1 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4 (movsi_aarch64): Likewise.
5 (load_pairsi): Likewise.
6 (load_pairdi): Likewise.
7 (store_pairsi): Likewise.
8 (store_pairdi): Likewise.
9 (load_pairsf): Likewise.
10 (load_pairdf): Likewise.
11 (store_pairsf): Likewise.
12 (store_pairdf): Likewise.
13 (zero_extend): Likewise.
14 (trunc): Swap alternatives.
15 (fcvt_target): Add '?' to prefer w over r.
17 2018-05-14 Jakub Jelinek <jakub@redhat.com>
20 * config/i386/i386.md: Disallow non-commutative arithmetics in
21 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
22 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
23 in the peephole2 before it.
25 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
27 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
28 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
29 (ix86_handle_option): Handle -mcldemote.
30 * config.gcc: New header.
31 * config/i386/cldemoteintrin.h: New file.
32 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
33 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
35 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
36 OPTION_MASK_ISA_CLDEMOTE.
37 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
38 (ix86_valid_target_attribute_inner_p): Ditto.
39 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
40 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
41 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
42 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
43 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
45 * config/i386/i386.opt: Add -mcldemote.
46 * config/i386/x86intrin.h: New header.
47 * doc/invoke.texi: Add -mcldemote.
49 2018-05-14 Richard Biener <rguenther@suse.de>
51 * doc/match-and-simplify.texi: Adjust :s documentation.
53 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
55 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
57 (REORDER_45): Likewise.
59 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
62 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
63 * vec.c (qsort_chk): Use gcc_qsort.
64 * Makefile.in (OBJS-libcommon): Add sort.o.
65 (build/sort.o): New target. Use it...
66 (BUILD_RTL): ... here, and...
67 (build/gencfn-macros): ... here, and...
68 (build/genmatch): ... here.
70 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
71 Chung-Ju Wu <jasonwucj@gmail.com>
73 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
74 * config/nds32/nds32-graywolf.md: New file.
75 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
76 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
78 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
79 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
80 * config/nds32/nds32.md (pipeline_model): Add graywolf.
81 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
82 * config/nds32/pipelines.md: Include n15 settings.
84 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
85 Chung-Ju Wu <jasonwucj@gmail.com>
87 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
88 * config/nds32/nds32-n13.md: New file.
89 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
90 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
92 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
93 * config/nds32/nds32.md (pipeline_model): Add n13.
94 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
95 * config/nds32/pipelines.md: Include n13 settings.
97 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
98 Chung-Ju Wu <jasonwucj@gmail.com>
100 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
101 * config/nds32/nds32-n10.md: New file.
102 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
103 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
105 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
106 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
107 * config/nds32/nds32.md (pipeline_model): Add n10.
108 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
109 * config/nds32/pipelines.md: Include n10 settings.
111 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
112 Kito Cheng <kito.cheng@gmail.com>
113 Chung-Ju Wu <jasonwucj@gmail.com>
115 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
116 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
117 Add enum values for DSP extension instructions.
118 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
120 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
121 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
123 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
124 * config/nds32/nds32-dspext.md: New file for DSP implementation.
125 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
126 * config/nds32/nds32-intrinsic.md: Likewise.
127 * config/nds32/nds32_intrinsic.h: Likewise.
128 * config/nds32/nds32-md-auxiliary.c: Likewise.
129 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
130 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
131 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
132 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
133 * config/nds32/nds32-protos.h: New declarations for DSP extension.
134 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
135 TYPE_DMAC in switch statement.
136 * config/nds32/nds32.c: New checking and implementation for DSP
137 extension instructions.
138 * config/nds32/nds32.h: Likewise.
139 * config/nds32/nds32.md: Likewise.
140 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
141 * config/nds32/predicates.md: Implement new predicates for DSP
144 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
146 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
147 Reformat alternatives and attributes so it is easier to identify
148 which constraints/attributes go with which instruction.
149 (mov<mode>_hardfloat32, FMOVE64): Likewise.
150 (mov<mode>_softfloat32, FMOVE64): Likewise.
151 (mov<mode>_hardfloat64, FMOVE64): Likewise.
152 (mov<mode>_softfloat64, FMOVE64): Likewise.
154 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
156 * doc/extend.texi (PowerPC Built-in Functions): Rename this
158 (Basic PowerPC Built-in Functions): The new name of the
159 subsection previously known as "PowerPC Built-in Functions".
160 (Basic PowerPC Built-in Functions Available on all Configurations):
162 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
163 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
164 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
165 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
167 2018-05-11 Martin Jambor <mjambor@suse.cz>
170 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
173 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
176 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
178 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
180 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
181 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
182 (ix86_handle_option): Handle -mwaitpkg.
183 * config.gcc: New header.
184 * config/i386/cpuid.h (bit_WAITPKG): New bit.
185 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
186 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
188 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
189 OPTION_MASK_ISA_WAITPKG.
190 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
191 (ix86_option_override_internal): Add PTA_WAITPKG.
192 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
193 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
195 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
196 __builtin_ia32_umwait and __builtin_ia32_tpause.
197 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
198 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
199 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
200 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
201 UNSPECV_TPAUSE): New.
202 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
203 * config/i386/i386.opt: Add -mwaitpkg.
204 * config/i386/waitpkgintrin.h: New file.
205 * config/i386/x86intrin.h: New header.
206 * doc/invoke.texi: Add -mwaitpkg.
208 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
211 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
213 (cortex-m0): Use armv6s-m isa.
214 (cortex-m0plus): Likewise.
215 (cortex-m1): Likewise.
216 (cortex-m0.small-multiply): Likewise.
217 (cortex-m0plus.small-multiply): Likewise.
218 (cortex-m1.small-multiply): Likewise.
220 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
221 Jakub Jelinek <jakub@redhat.com>
223 PR tree-optimization/85692
224 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
225 source permute as well.
227 2018-05-11 Martin Liska <mliska@suse.cz>
230 * doc/extend.texi: Document LLVM style format for no_sanitize
233 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
235 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
236 mode_supports_vsx_dform_quad to mode_supports_dq_form.
237 (mode_supports_vsx_dform_quad): Likewise.
238 (mode_supports_vmx_dform): Move these functions to be next to the
239 other mode_supports functions.
240 (mode_supports_dq_form): Likewise.
241 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
242 mode_supports_dq_form.
243 (reg_offset_addressing_ok_p): Likewise.
244 (offsettable_ok_by_alignment): Likewise.
245 (rs6000_legitimate_offset_address_p): Likewise.
246 (legitimate_lo_sum_address_p): Likewise.
247 (rs6000_legitimize_address): Likewise.
248 (rs6000_legitimize_reload_address): Likewise.
249 (rs6000_secondary_reload_inner): Likewise.
250 (rs6000_preferred_reload_class): Likewise.
251 (rs6000_output_move_128bit): Likewise.
253 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
255 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
256 Generate SImode target register for null target.
257 <case IX86_BUILTIN_XGETBV>: Ditto.
258 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
259 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
261 2018-05-10 Carl Love <cel@us.ibm.com>
263 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
264 dcbtt and dcbtstt if operands[2] is 0.
266 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
269 * config/i386/sse.md (usadv64qi): New expander.
271 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
273 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
274 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
275 -maltivec=be support.
276 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
277 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
278 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
279 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
280 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
281 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
282 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
283 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
284 altivec_vsumsws): Adjust.
285 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
286 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
287 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
289 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
290 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
291 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
292 (altivec_lve<VI_char>x): Delete expand.
293 (*altivec_lve<VI_char>x_internal): Rename to...
294 (altivec_lve<VI_char>x): ... this.
295 (altivec_lvxl_<mode>): Delete expand.
296 (*altivec_lvxl_<mode>_internal): Rename to ...
297 (altivec_lvxl_<mode>): ... this.
298 (altivec_stvxl_<mode>): Delete expand.
299 (*altivec_stvxl_<mode>_internal): Rename to ...
300 (altivec_stvxl_<mode>): ... this.
301 (altivec_stve<VI_char>x): Delete expand.
302 (*altivec_stve<VI_char>x_internal): Rename to ...
303 (altivec_stve<VI_char>x): ... this.
304 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
305 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
306 reduc_plus_scal_<mode>): Adjust.
307 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
309 (rs6000_cpu_cpp_builtins): Adjust.
310 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
311 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
312 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
313 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
314 -maltivec=be support.
315 (rs6000_split_vec_extract_var): Adjust.
316 (rs6000_split_v4si_init): Adjust.
317 (swap_selector_for_mode): Delete.
318 (altivec_expand_lvx_be, altivec_expand_stvx_be,
319 altivec_expand_stvex_be): Delete.
320 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
321 -maltivec=be support.
322 (rs6000_gimple_fold_builtin): Ditto.
323 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
325 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
326 (TARGET_DIRECT_MOVE_64BIT): Adjust.
327 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
328 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
329 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
330 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
331 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
332 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
333 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
334 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
335 anonymous split): Adjust.
336 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
337 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
339 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
341 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
342 when --with-gxx-include-dir is also specified.
343 * configure: Regenerate.
345 2018-05-09 Jim Wilson <jimw@sifive.com>
348 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
349 * config/riscv/t-withmultilib: New.
350 * config/riscv/withmultilib.h: New.
351 * doc/install.texi: Document RISC-V --with-multilib-list support.
353 2018-05-09 Richard Biener <rguenther@suse.de>
355 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
357 (vect_bb_vectorization_profitable_p): Adjust. Compute
358 actual scalar cost using the cost vector and the add_stmt_cost
361 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
363 PR rtl-optimization/85645
364 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
365 in the REG_CFA_REGISTER note for LR, don't leave it empty.
367 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
369 PR rtl-optimization/85645
370 * shrink-wrap.c (spread_components): Return a boolean saying if
371 anything was changed.
372 (try_shrink_wrapping_separate): Iterate spread_components until
373 nothing changes anymore.
375 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
377 PR rtl-optimization/85645
378 * regrename.c (build_def_use): Also kill the chains that include the
379 destination of a REG_CFA_REGISTER note.
381 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
383 PR rtl-optimization/85645
384 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
385 insn that has a REG_CFA_REGISTER note.
387 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
389 * cfgexpand.c (expand_clobber): New function.
390 (expand_gimple_stmt_1): Use it.
391 * tree-vect-stmts.c (vect_clobber_variable): New function,
393 (vectorizable_simd_clone_call): ...here.
394 (vectorizable_store): Emit a clobber either side of an
395 IFN_STORE_LANES sequence.
396 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
398 2018-05-09 Tom de Vries <tom@codesourcery.com>
401 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
402 (define_insn "trap_if_false"): Add exit after trap.
404 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
406 PR rtl-optimization/85638
407 * bb-reorder.c: Include common/common-target.h.
408 (create_forwarder_block): New function extracted from...
409 (fix_up_crossing_landing_pad): ...here. Rename into...
410 (dw2_fix_up_crossing_landing_pad): ...this.
411 (sjlj_fix_up_crossing_landing_pad): New function.
412 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
413 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
414 from both partitions and exit the loop after one iteration.
416 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
419 * doc/extend.texi (PowerPC Built-in Functions): Rename this
421 (Basic PowerPC Built-in Functions): The new name of the
422 subsection previously known as "PowerPC Built-in Functions".
423 (Basic PowerPC Built-in Functions Available on all Configurations):
425 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
427 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
429 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
431 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
434 2018-05-08 Jim Wilson <jimw@sifive.com>
436 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
437 (LD_EMUL_SUFFIX): New.
440 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
442 * doc/extend.texi (PowerPC Built-in Functions): Rename this
444 (Basic PowerPC Built-in Functions): The new name of the
445 subsection previously known as "PowerPC Built-in Functions".
446 (Basic PowerPC Built-in Functions Available on all Configurations):
448 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
450 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
452 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
454 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
457 2018-05-08 Jakub Jelinek <jakub@redhat.com>
460 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
461 after cmpelim optimization.
463 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
465 * config.gcc: Support "goldmont".
466 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
467 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
469 * config/i386/i386.c (m_GOLDMONT): Define.
470 (processor_target_table): Add "goldmont".
471 (PTA_GOLDMONT): Define.
472 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
473 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
474 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
475 (fold_builtin_cpu): Add "goldmont".
476 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
477 (ix86_option_override_internal): Add "goldmont".
478 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
479 (processor_type): Add PROCESSOR_GOLDMONT.
480 * config/i386/i386.md: Add CPU "glm".
481 * config/i386/glm.md: New file.
482 * config/i386/x86-tune.def: Add m_GOLDMONT.
483 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
485 2018-05-08 Jakub Jelinek <jakub@redhat.com>
488 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
490 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
491 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
492 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
495 * config/i386/i386.c (ix86_fold_builtin): Handle
496 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
499 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
500 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
502 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
505 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
506 (check_arch): Likewise.
507 (check_fpu): Return the result rather than printing it.
508 (end arch): Fix operator precedence.
510 (END): Print the result from check_fpu.
512 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
513 Alan Hayward <alan.hayward@arm.com>
514 David Sherwood <david.sherwood@arm.com>
516 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
517 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
518 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
519 (*fcmuo<mode>_and): New patterns.
521 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
523 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
524 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
525 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
526 (cmp_op, sve_imm_con): New code attributes.
527 (SVE_COND_INT_CMP, imm_con): Delete.
528 (cmp_op): Remove above unspecs from int attribute.
529 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
531 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
532 comparison-specific unspecs.
533 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
534 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
535 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
536 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
537 (*vec_fcm<cmp_op><mode>): Rename to...
538 (*fcm<cmp_op><mode>): ...this and adjust likewise.
539 (*vec_fcmuo<mode>): Rename to...
540 (*fcmuo<mode>): ...this and adjust likewise.
541 (*pred_fcm<cmp_op><mode>): New pattern.
542 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
543 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
545 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
547 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
548 (aarch64_emit_sve_predicated_cond): New function.
549 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
550 (aarch64_emit_unspec_cond_or): Replace with...
551 (aarch64_emit_sve_or_conds): ...this new function. Use
552 aarch64_emit_sve_ptrue_op for the individual comparisons and
553 aarch64_emit_binop to OR them together.
554 (aarch64_emit_inverted_unspec_cond): Replace with...
555 (aarch64_emit_sve_inverted_cond): ...this new function. Use
556 aarch64_emit_sve_ptrue_op for the comparison and
557 aarch64_emit_unop to invert the result.
558 (aarch64_expand_sve_vec_cmp_float): Update after the above
559 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
561 2018-05-07 Nathan Sidwell <nathan@acm.org>
563 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
564 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
565 (Backwards Compatibility): Likewise.
567 2018-05-07 Luis Machado <luis.machado@linaro.org>
571 2018-05-07 Luis Machado <luis.machado@linaro.org>
573 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
574 <prefetch_dynamic_strides>: New const bool field.
575 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
576 prefetch_dynamic_strides.
577 (exynosm1_prefetch_tune): Likewise.
578 (thunderxt88_prefetch_tune): Likewise.
579 (thunderx_prefetch_tune): Likewise.
580 (thunderx2t99_prefetch_tune): Likewise.
581 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
582 (aarch64_override_options_internal): Update to set
583 PARAM_PREFETCH_DYNAMIC_STRIDES.
584 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
585 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
586 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
587 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
588 prefetch-dynamic-strides setting.
590 2018-05-07 Luis Machado <luis.machado@linaro.org>
592 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
593 <minimum_stride>: New const int field.
594 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
595 minimum_stride field.
596 (exynosm1_prefetch_tune): Likewise.
597 (thunderxt88_prefetch_tune): Likewise.
598 (thunderx_prefetch_tune): Likewise.
599 (thunderx2t99_prefetch_tune): Likewise.
600 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
601 (aarch64_override_options_internal): Update to set
602 PARAM_PREFETCH_MINIMUM_STRIDE.
603 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
604 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
605 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
606 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
607 stride is constant and is below the minimum stride threshold.
609 2018-05-07 Luis Machado <luis.machado@linaro.org>
611 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
614 2018-05-07 Luis Machado <luis.machado@linaro.org>
616 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
617 <prefetch_dynamic_strides>: New const bool field.
618 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
619 prefetch_dynamic_strides.
620 (exynosm1_prefetch_tune): Likewise.
621 (thunderxt88_prefetch_tune): Likewise.
622 (thunderx_prefetch_tune): Likewise.
623 (thunderx2t99_prefetch_tune): Likewise.
624 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
625 (aarch64_override_options_internal): Update to set
626 PARAM_PREFETCH_DYNAMIC_STRIDES.
627 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
628 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
629 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
630 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
631 prefetch-dynamic-strides setting.
633 2018-05-07 Luis Machado <luis.machado@linaro.org>
635 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
636 <minimum_stride>: New const int field.
637 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
638 minimum_stride field.
639 (exynosm1_prefetch_tune): Likewise.
640 (thunderxt88_prefetch_tune): Likewise.
641 (thunderx_prefetch_tune): Likewise.
642 (thunderx2t99_prefetch_tune): Likewise.
643 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
644 (aarch64_override_options_internal): Update to set
645 PARAM_PREFETCH_MINIMUM_STRIDE.
646 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
647 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
648 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
649 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
650 stride is constant and is below the minimum stride threshold.
652 2018-05-06 Jakub Jelinek <jakub@redhat.com>
655 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
656 the type is addressable. Don't force op into register if it has
659 2018-05-05 Roland McGrath <mcgrathr@google.com>
662 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
663 any section for which we don't know a specific type it should have,
664 regardless of name. Previously this was done only for the exact
665 names ".init_array", ".fini_array", and ".preinit_array".
666 (default_elf_asm_named_section): Add comment about
667 relationship with default_section_type_flags and SECTION_NOTYPE.
668 (get_section): Don't consider it a type conflict if one side has
669 SECTION_NOTYPE and the other doesn't, as long as neither has the
670 SECTION_BSS et al used in the default_section_type_flags logic.
672 2018-05-05 Tom de Vries <tom@codesourcery.com>
675 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
676 (workaround_barsyncs): New function.
677 (nvptx_reorg): Use workaround_barsyncs.
678 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
679 (define_expand "nvptx_membar_cta"): New define_expand.
680 (define_insn "*nvptx_membar_cta"): New insn.
682 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
684 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
685 To improve optimization opportunities.
686 * builtin-types.def: The new needed builtin types for the above.
688 2018-05-04 Richard Biener <rguenther@suse.de>
690 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
691 * gimple-ssa-store-merging.c
692 (imm_store_chain_info::output_merged_store): Remove redundant create,
693 release split_store vector contents on failure.
694 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
695 scalar stmt vector on cache hit.
697 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
699 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
701 * config.gcc (powerpc-xilinx-eabi*): Remove.
702 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
704 (fusion_addis_mem_combo_load): Ditto.
705 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
707 (rs6000_cpu_cpp_builtins): Ditto.
708 * config/rs6000/rs6000-linux.c
709 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
710 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
711 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
713 (rs6000_setup_reg_addr_masks): Ditto.
714 (rs6000_init_hard_regno_mode_ok): Ditto.
715 (rs6000_option_override_internal): Ditto.
716 (legitimate_lo_sum_address_p): Ditto.
717 (rs6000_legitimize_address): Ditto.
718 (rs6000_legitimize_reload_address): Ditto.
719 (rs6000_legitimate_address_p): Ditto.
720 (abi_v4_pass_in_fpr): Ditto.
721 (setup_incoming_varargs): Ditto.
722 (rs6000_gimplify_va_arg): Ditto.
723 (rs6000_split_multireg_move): Ditto.
724 (rs6000_savres_strategy): Ditto.
725 (rs6000_emit_prologue_components): Ditto.
726 (rs6000_emit_epilogue_components): Ditto.
727 (rs6000_emit_prologue): Ditto.
728 (rs6000_emit_epilogue): Ditto.
729 (rs6000_elf_file_end): Ditto.
730 (rs6000_function_value): Ditto.
731 (rs6000_libcall_value): Ditto.
732 * config/rs6000/rs6000.h: Ditto.
733 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
734 (TARGET_MINMAX): ... this. New.
735 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
736 * config/rs6000/rs6000.md: Remove Xilinx FP support.
737 (*movsi_internal1_single): Delete.
738 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
739 mfpu=, mxilinx-fpu): Delete.
740 * config/rs6000/singlefp.h: Delete.
741 * config/rs6000/sysv4.h: Remove Xilinx FP support.
742 * config/rs6000/t-rs6000: Ditto.
743 * config/rs6000/t-xilinx: Delete.
744 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
745 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
746 (VStype_simple): Delete.
747 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
748 * config/rs6000/xfpu.h: Delete.
749 * config/rs6000/xfpu.md: Delete.
750 * config/rs6000/xilinx.h: Delete.
751 * config/rs6000/xilinx.opt: Delete.
752 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
753 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
755 2018-05-04 Tom de Vries <tom@codesourcery.com>
758 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
761 2018-05-04 Richard Biener <rguenther@suse.de>
764 * tree-complex.c (update_complex_assignment): We are always in SSA form.
765 (expand_complex_div_wide): Likewise.
766 (expand_complex_operations_1): Likewise.
767 (expand_complex_libcall): Preserve EH info of the original stmt.
768 (tree_lower_complex): Handle removed blocks.
769 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
770 on complex multiplication and division libcall builtins.
772 2018-05-04 Richard Biener <rguenther@suse.de>
775 * fold-const.c (negate_expr_p): Restrict negation of operand
776 zero of a division to when we know that can happen without
778 (fold_negate_expr_1): Likewise.
780 2018-05-04 Jakub Jelinek <jakub@redhat.com>
783 * real.h (real_nextafter): Declare.
784 * real.c (real_nextafter): New function.
785 * fold-const-call.c (fold_const_nextafter): New function.
786 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
788 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
789 even when arg1_mode is different from arg0_mode.
791 2018-05-03 Nathan Sidwell <nathan@acm.org>
793 * doc/extend.texi (Deprecated Features): Remove
795 (Backwards Compatibility): Likewise.
796 * doc/invoke.texi (C++ Language Options): Likewise.
797 (C++ Dialect Options): Likewise.
799 2018-05-03 Jakub Jelinek <jakub@redhat.com>
802 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
803 _mm512_mask_mullox_epi64): New intrinsics.
805 2018-05-03 Tom de Vries <tom@codesourcery.com>
808 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
809 dump files): Add offload-tree.
811 2018-05-03 Richard Biener <rguenther@suse.de>
813 PR tree-optimization/85615
814 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
815 to loops not nested in BBs loop father to avoid creating multi-entry
818 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
820 PR tree-optimization/70291
821 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
822 arguments. Change return type to tree. Emit libcall as a new
823 statement rather than replacing existing one when inplace_p is true.
824 (expand_complex_multiplication_components): New function.
825 (expand_complex_multiplication): Expand floating-point complex
826 multiplication using the above.
827 (expand_complex_division): Rename inner_type parameter to type.
828 Update expand_complex_libcall call-site.
829 (expand_complex_operations_1): Update expand_complex_multiplication
830 and expand_complex_division call-sites.
832 2018-05-02 Jakub Jelinek <jakub@redhat.com>
835 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
836 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
837 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
838 the highest significant bit of the shift count mask is clear. In
839 check whether and[sq]i3 is needed verify that all significant bits
840 of the shift count other than the highest are set.
842 2018-05-02 Tom de Vries <tom@codesourcery.com>
845 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
846 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
847 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
848 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
849 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
850 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
851 __builtin_goacc_parlevel_size.
853 2018-05-02 Richard Biener <rguenther@suse.de>
855 PR tree-optimization/85597
856 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
857 do not use split vect_get_vec_defs call but call vect_get_slp_defs
860 2018-05-02 Tom de Vries <tom@codesourcery.com>
863 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
864 dump files): Add ltrans-tree.
866 2018-05-02 Tom de Vries <tom@codesourcery.com>
869 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
870 dump files): Add wpa-ipa.
872 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
874 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
875 powerpc*-*-linux*paired* target.
876 * config/rs6000/750cl.h: Delete.
877 * config/rs6000/paired.h: Delete.
878 * config/rs6000/paired.md: Delete.
879 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
881 * config/rs6000/rs6000-builtin.def: Remove paired float support.
882 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
883 comment. Remove paired float support.
884 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
885 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
887 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
888 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
890 * config/rs6000/rs6000.c: Remove paired float support.
891 (paired_expand_vector_init, paired_expand_vector_move,
892 paired_emit_vector_compare, paired_emit_vector_cond_expr,
893 (paired_expand_lv_builtin, paired_expand_stv_builtin,
894 paired_expand_builtin, paired_expand_predicate_builtin,
895 paired_init_builtins): Delete.
896 * config/rs6000/rs6000.h: Remove paired float support.
897 * config/rs6000/rs6000.md: Remove paired float support.
898 (move_from_CR_ov_bit): Delete.
899 * config/rs6000/rs6000.opt (mpaired): Delete.
900 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
901 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
903 2018-05-02 Richard Biener <rguenther@suse.de>
906 * gimplify.c (gimplify_save_expr): When in SSA form allow
907 SAVE_EXPRs to compute to SSA vars.
909 2018-05-02 Jakub Jelinek <jakub@redhat.com>
912 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
913 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
914 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
915 clobber operands[2], instead use a new pseudo. Formatting fixes.
917 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
919 PR tree-optimization/85586
920 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
921 exit early for statements in the same group if the accesses are
924 2018-05-02 Tom de Vries <tom@codesourcery.com>
927 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
930 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
932 PR tree-optimization/85143
933 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
935 2018-05-01 Tom de Vries <tom@codesourcery.com>
938 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
939 not found" error message.
941 2018-05-01 Tom de Vries <tom@codesourcery.com>
944 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
945 * vec.c (test_ordered_remove_if): New function.
946 (vec_c_tests): Call test_ordered_remove_if.
947 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
948 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
949 * tree-vect-patterns.c (vect_pattern_recog_1): Use
950 VEC_ORDERED_REMOVE_IF.
952 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
954 PR tree-optimization/82665
955 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
956 pointer subtraction where arguments come from a memchr call.
958 2018-05-01 Jakub Jelinek <jakub@redhat.com>
960 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
961 --push-state --as-needed and --pop-state instead of --as-needed and
962 --no-as-needed if ld supports it.
963 * configure: Regenerated.
966 * doc/install.texi2html: Replace _002d with - and _002a with * in
967 generated html files using sed.
969 2018-04-30 David Malcolm <dmalcolm@redhat.com>
972 * gcc-rich-location.c (blank_line_before_p): New function.
973 (use_new_line): New function.
974 (gcc_rich_location::add_fixit_insert_formatted): New function.
975 * gcc-rich-location.h
976 (gcc_rich_location::add_fixit_insert_formatted): New function.
978 2018-04-30 David Malcolm <dmalcolm@redhat.com>
980 * selftest.c (assert_streq): Rename "expected" and "actual" to
981 "val1" and "val2". Extend NULL-handling to cover both inputs
982 symmetrically, while still requiring both to be non-NULL for a pass.
983 * selftest.h (assert_streq): Rename "expected" and "actual" to
985 (ASSERT_EQ): Likewise.
986 (ASSERT_EQ_AT): Likewise.
987 (ASSERT_KNOWN_EQ): Likewise.
988 (ASSERT_KNOWN_EQ_AT): Likewise.
989 (ASSERT_NE): Likewise.
990 (ASSERT_MAYBE_NE): Likewise.
991 (ASSERT_MAYBE_NE_AT): Likewise.
992 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
993 the assertion to pass.
994 (ASSERT_STREQ_AT): Likewise.
996 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
998 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
999 interaction with -pie.
1001 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1003 * selftest.h: Fix alphabetization of per-source-file selftest
1006 2018-04-30 Jason Merrill <jason@redhat.com>
1008 PR c++/61982 - dead stores to destroyed objects.
1009 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1012 2018-04-30 Jason Merrill <jason@redhat.com>
1014 * tree.c (build_clobber): New.
1015 * tree.h: Declare it.
1016 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1018 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1020 * diagnostic-show-locus.c (layout::layout): Update for
1021 location_get_source_line returning a char_span.
1022 (struct char_span): Move to input.h.
1023 (struct correction): Update for fields in char_span becoming
1025 (struct source_line): Update for location_get_source_line
1026 returning a char_span.
1027 (layout::print_line): Likewise.
1028 * edit-context.c (edited_file::print_content): Likewise.
1029 (edited_file::print_diff_hunk): Likewise.
1030 (edited_file::print_run_of_changed_lines): Likewise.
1031 (edited_file::get_num_lines): Likewise.
1032 (edited_line::edited_line): Likewise.
1033 * final.c (asm_show_source): Likewise.
1034 * input.c (location_get_source_line): Convert return type
1035 from const char * to char_span, losing the final "line_len"
1037 (dump_location_info): Update for the above.
1038 (get_substring_ranges_for_loc): Likewise. Use a char_span
1039 when handling the literal within the line.
1040 (test_reading_source_line): Update for location_get_source_line
1041 returning a char_span.
1042 * input.h (class char_span): Move here from
1043 diagnostic-show-locus.c, converting from a struct to a class.
1044 Make data members private.
1045 (char_span::operator bool): New.
1046 (char_span::length): New.
1047 (char_span::get_buffer): New.
1048 (char_span::operator[]): New.
1049 (char_span::subspan): Make const.
1050 (char_span::xstrdup): New.
1051 (location_get_source_line): Convert return type from const char *
1052 to char_span, losing the final "line_size" param.
1054 2018-04-30 Jan Hubicka <jh@suse.cz>
1056 * lto-wrapper.c (ltrans_priorities): New static var.
1057 (cmp_priority): New.
1058 (run_gcc): Read priorities and if doing parallel build order
1059 the Makefile by them.
1061 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1063 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1065 2018-04-30 Richard Biener <rguenther@suse.de>
1067 * tree-cfg.c (verify_address): Remove base argument, add
1068 flag whether to check TREE_ADDRESSABLE and do that.
1069 (verify_expr): Remove.
1070 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1071 (verify_gimple_assign_single): Likewise.
1072 (verify_gimple_switch): Likewise.
1073 (verify_expr_location_1): Dereference tp once. Add (disabled)
1074 piece from verify_expr.
1075 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1077 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1079 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1081 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1083 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1084 (small_data_pattern): Likewise.
1085 (arc_rewrite_small_data): Likewise.
1086 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1087 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1088 (get_symbol_alignment): New function.
1089 (legitimate_small_data_address_p): Likewise.
1090 (legitimate_scaled_address): Update, call
1091 legitimate_small_data_address_p.
1092 (output_sdata): New static variable.
1093 (arc_print_operand): Update how we handle small data operands.
1094 (arc_print_operand_address): Likewise.
1095 (arc_legitimate_address_p): Update, use
1096 legitimate_small_data_address_p.
1097 (arc_rewrite_small_data_p): Remove.
1098 (arc_rewrite_small_data_1): Likewise.
1099 (arc_rewrite_small_data): Likewise.
1100 (small_data_pattern): Likewise.
1101 (compact_sda_memory_operand): Update to use
1102 legitimate_small_data_address_p and get_symbol_alignment.
1103 (prepare_move_operands): Don't rewite sdata pattern.
1104 (prepare_extend_operands): Remove.
1105 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1107 (zero_extendqisi2): Likewise.
1108 (zero_extendhisi2): Likewise.
1109 (extendqihi2): Likewise.
1110 (extendqisi2): Likewise.
1111 (extendhisi2): Likewise.
1115 * config/arc/constraints.md (Usd): Change it to memory constraint.
1117 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1119 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1120 as source of std instructions.
1121 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1122 allow 6-bit constants as source for store instructions.
1123 (movdi_insn): Update instruction pattern to allow 6-bit constants
1124 as source for store instructions.
1126 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1128 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1130 2018-04-30 Nathan Sidwell <nathan@acm.org>
1131 Sandra Loosemore <sandra@codesourcery.com>
1133 * dumpfile.c (dump_open): Allow '-' for stdout.
1134 * doc/invoke.texi (Developer Options): Document dump filename
1135 determination early. Document stdin/stdout selection.
1137 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1139 Microblaze Target: PIC data text relative
1141 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1142 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1144 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1145 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1146 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1147 New addressing mode for data-text relative position indepenedent code.
1148 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1149 'ADDRESS_SYMBOLIC_TXT_REL'.
1150 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1151 (microblaze_legitimate_pic_operand): Exclude function calls from
1152 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1153 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1155 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1156 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1157 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1158 for 'address + offset'.
1159 (microblaze_expand_prologue): Add new function prologue call for
1161 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1162 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1163 table in case of TARGET_PIC_DATA_TEXT_REL.
1164 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1165 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1166 Add new macros 'UNSPEC_TEXT',
1167 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1168 + exclude function calls from 'UNSPEC_PLT' in case of data text
1170 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1171 new target hook for generating address diff vector tables in case of
1173 * doc/tm.texi : Regenerate.
1174 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1175 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1176 of addr diff vector generation.
1177 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1178 target hook definition.
1179 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1180 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1181 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1182 option for data text relative.
1184 2018-04-30 Richard Biener <rguenther@suse.de>
1186 * tree-chrec.h (evolution_function_is_constant_p): Remove
1188 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1190 2018-04-30 Richard Biener <rguenther@suse.de>
1193 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1195 2018-04-30 Richard Biener <rguenther@suse.de>
1197 PR tree-optimization/28364
1198 PR tree-optimization/85275
1199 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1200 copying first exit test.
1202 2018-04-28 Mark Wielaard <mark@klomp.org>
1204 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1206 (dwarf_AT): Handle DW_AT_addr_base.
1207 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1209 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1212 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1213 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1214 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1215 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1217 2018-04-28 Richard Biener <rguenther@suse.de>
1219 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1220 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1221 to reflect use. Only add interesting stmts.
1223 2018-04-27 Martin Jambor <mjambor@suse.cz>
1226 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1227 the jump function allows for passing through aggregate values.
1229 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1231 * input.h (in_system_header_at): Convert from macro to inline
1233 (from_macro_expansion_at): Likewise.
1234 (from_macro_definition_at): Likewise.
1236 2018-04-27 Jeff Law <law@redhat.com>
1238 * config.gcc: Mark tile* targets as deprecated/obsolete.
1240 2018-04-27 Richard Biener <rguenther@suse.de>
1242 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1245 2018-04-27 Richard Biener <rguenther@suse.de>
1247 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1249 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1251 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1252 with Yd constraint. Set "preferred_for_speed" attribute from
1253 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1255 (*movdi_internal): Ditto.
1256 (movti_interunit splitters): Remove
1257 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1258 (movdi_interunit splitters): Ditto.
1259 * config/i386/constraints.md (Ye): Remove.
1260 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1262 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1265 * config/aarch64/constraints.md (Usg): Limit to 31.
1268 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1270 PR tree-optimization/85529
1271 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1272 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1273 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1274 zero extension or masking of the MSB bit.
1275 (optimize_range_tests): Add FIRST_BB argument, pass it through
1276 to optimize_range_tests_var_bound.
1277 (maybe_optimize_range_tests, reassociate_bb): Adjust
1278 optimize_range_tests callers.
1280 2018-04-26 Richard Biener <rguenther@suse.de>
1281 Jakub Jelinek <jakub@redhat.com>
1283 * cgraph.h (symbol_table): Just declare debug method here.
1284 * symtab.c (symbol_table::debug): Define.
1286 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1288 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1290 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1292 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1293 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1294 (*movdi_internal): Substitute Yi and Yj constraint with x
1295 and Ym and Yn constraint with y constraint. Update "isa"
1296 attribute and set "preferred_for_speed" attribute from
1297 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1298 (*movsi_internal): Ditto.
1299 (*movdf_internal): Ditto.
1300 (*movsf_internal): Ditto.
1301 (*zero_extendsidi2): Ditto.
1302 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1303 (sse2_loadld): Ditto.
1304 (*vec_extract<ssevecmodelower>_0): Ditto.
1305 (*vec_extractv4si_0_zext_sse4): Ditto.
1306 (vec_concatv2di): Ditto.
1307 (*vec_dup<mode>): Ditto.
1308 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1309 * config/i386/constraints.md (Yi): Remove.
1314 2018-04-26 Nathan Sidwell <nathan@acm.org>
1316 * dumpfile.c (dump_open): New.
1317 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1318 (dump_finish): Detect stdio/stderr by value not name.
1320 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1322 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1324 2018-04-26 Tom de Vries <tom@codesourcery.com>
1327 * config/nvptx/nvptx.c (verify_neutering_jumps)
1328 (verify_neutering_labels): New function
1329 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1331 2018-04-26 Tom de Vries <tom@codesourcery.com>
1334 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1335 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1338 2018-04-26 Richard Biener <rguenther@suse.de>
1339 Tom de Vries <tom@codesourcery.com>
1342 * lto-streamer-out.c (output_function): Fixup loops if required to match
1343 discovery done in the reader.
1345 2018-04-26 Richard Biener <rguenther@suse.de>
1347 PR tree-optimization/85116
1348 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1349 have a loop exit from the single latch predecessor. Remove
1350 case of header with just condition.
1351 (ch_base::copy_headers): Exclude infinite loops from any
1353 (pass_ch::execute): Record exits.
1355 2018-04-26 Richard Biener <rguenther@suse.de>
1357 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1358 prologue cost vector and pass it to vect_get_load_cost.
1359 (vect_get_peeling_costs_all_drs): Likewise.
1360 (vect_peeling_hash_get_lowest_cost): Likewise.
1361 (vect_enhance_data_refs_alignment): Likewise.
1363 2018-04-26 Richard Biener <rguenther@suse.de>
1366 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1367 checking of integer<->pointer conversions.
1368 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1369 sign-/zero-extending pointer types.
1370 (expand_omp_for_static_chunk): Likewise.
1372 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1373 Jean Lee <xiaoyur347@gmail.com>
1375 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1376 (TARGET_ASAN_SHADOW_OFFSET): Define.
1377 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1378 true for -fsanitize=address.
1380 2018-04-25 Mark Wielaard <mark@klomp.org>
1382 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1385 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1387 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1388 than "alu", remove explicit "memory" and "imm_disp" attributes.
1389 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1392 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1393 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1394 gen_lowpart_no_emit.
1396 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1399 * config/i386/i386.c (ix86_expand_builtin): Change memory
1400 operand to XI, extend p0 to Pmode.
1401 * config/i386/i386.md: Change unspec volatile and operand
1402 1 mode to XI, change operand 0 mode to P.
1404 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1406 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1407 GET_MODE_MASK before any checking.
1408 (nds32_can_use_bset_p): Likewise.
1409 (nds32_can_use_btgl_p): Likewise.
1411 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1413 * config/nds32/nds32-doubleword.md: New define_split pattern for
1414 illegal register number.
1416 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1418 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1420 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1422 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1424 2018-04-25 Richard Biener <rguenther@suse.de>
1426 * lto-streamer.h (LTO_major_version): Bump to 8.
1428 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1430 * BASE-VER: Set to 9.0.0.
1432 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1434 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1435 in __abskf2 and __powikf2.
1437 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1440 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1441 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1442 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1443 Use the above on operand 2. Reindent.
1444 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1446 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1449 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1451 * config/i386/i386.opt (mcet): Removed.
1452 * doc/install.texi: Remove -mcet documentation.
1453 * doc/invoke.texi: Likewise.
1455 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1458 * doc/install.texi: Remove -mcet from bootstrap-cet.
1460 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1463 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1464 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1468 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1469 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1470 containing a CONST_VECTOR.
1472 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1474 * doc/install.texi: Update newlib dependency for nvptx.
1476 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1479 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1480 instead of INTVAL when shifting x left.
1482 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1484 PR tree-optimization/85478
1485 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1486 vect_grouped_store_supported for single element vectors.
1488 2018-04-24 Richard Biener <rguenther@suse.de>
1491 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1492 load cost increase to the case of non-constant step.
1494 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1497 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1498 destination if any_malformed_asm.
1500 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1503 * expr.c (store_field): In the bitfield case, if the value comes from
1504 a function call and is returned in registers by means of a PARALLEL,
1505 do not change the mode of the temporary unless BLKmode and VOIDmode.
1507 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1509 PR rtl-optimization/85423
1510 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1511 dependencies to debug insns when the previous insn is non-debug.
1513 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1515 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1516 enums into a single definition.
1517 (fls): Fix predicates and printing.
1520 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1522 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1523 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1524 and short u6 immediate.
1525 (check_if_valid_sleep_operand): Remove.
1526 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1528 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1530 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1531 flag_always_save_lp condition.
1532 * config/nds32/nds32.opt (malways-save-lp): New option.
1534 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1536 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1537 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1538 * config/nds32/nds32.h
1539 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1540 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1542 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1544 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1545 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1547 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1548 Chung-Ju Wu <jasonwucj@gmail.com>
1550 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1552 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1553 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1555 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1557 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1559 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1561 * config/nds32/nds32-protos.h (nds32_data_alignment,
1562 nds32_local_alignment): Declare.
1563 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1564 nds32_local_alignment): New functions.
1565 (TARGET_CONSTANT_ALIGNMENT): Define.
1566 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1568 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1570 * config/nds32/nds32.c
1571 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1572 (TARGET_MODES_TIEABLE_P): Likewise.
1574 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1576 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1579 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1580 Chung-Ju Wu <jasonwucj@gmail.com>
1582 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1583 for unaligned access.
1584 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1586 * config/nds32/nds32-intrinsic.md: Likewise.
1587 * config/nds32/nds32_intrinsic.h: Likewise.
1588 * config/nds32/nds32.h (nds32_builtins): Likewise.
1589 * config/nds32/nds32.opt (munaligned-access): New option.
1590 * config/nds32/nds32.c (nds32_asm_file_start): Display
1591 flag_unaligned_access status.
1593 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1595 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1596 -mno-relax is present.
1597 * config/riscv/linux.h (LINK_SPEC): Ditto.
1599 2018-04-20 Martin Sebor <msebor@redhat.com>
1602 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1604 (gimple_fold_builtin_stxcpy_chk): Same.
1605 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1607 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1610 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1611 __powikf2 when long double is IEEE 128-bit.
1613 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1615 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1616 step to make sure stack always aligned.
1618 2018-04-20 Carl Love <cel@us.ibm.com>
1621 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1622 size check for arg0.
1624 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1625 Tom de Vries <tom@codesourcery.com>
1628 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1629 Emit insns for calls too.
1630 (nvptx_find_par): Always look for worker-level predecessor insn.
1631 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1633 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1634 (nvptx_process_pars): Propagate frames for calls.
1636 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1639 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1641 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1642 (ix86_handle_option): Don't handle OPT_mibt.
1643 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1645 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1647 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1648 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1649 (ix86_target_macros): Define __CET__ with flag_cf_protection
1650 for -fcf-protection.
1651 * config/i386/i386.c (isa2_opts): Remove -mibt.
1652 * config/i386/i386.h (TARGET_IBT): Removed.
1653 (TARGET_IBT_P): Likewise.
1654 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1655 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1656 * config/i386/i386.opt (mcet): Update help message.
1659 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1660 -mcet as an alias for -mshstk.
1662 2018-04-20 Richard Biener <rguenther@suse.de>
1665 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1666 complexity by forcing a single use of the multiply operand.
1668 2018-04-20 Martin Jambor <mjambor@suse.cz>
1671 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1672 recursion dependency to only apply to non-clones.
1674 2018-04-20 Martin Jambor <mjambor@suse.cz>
1677 * ipa-cp.c (create_specialized_node): Check that clones of
1678 self-recursive edges exist during IPA-CP.
1680 2018-04-19 Toon Moene <toon@moene.org>
1682 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1685 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1687 PR tree-optimization/85467
1688 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1689 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1690 VECTOR_CST element to type.
1692 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1695 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1696 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1697 (builtin_longjmp): Likewise.
1698 (save_stack_nonlocal): New pattern.
1699 (restore_stack_nonlocal): Likewise.
1701 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1704 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1705 Replace ASM_OUTPUT_LABEL with fprintf.
1707 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1710 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1711 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1712 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1713 define __IBT__ and __SHSTK__ for -fcf-protection.
1714 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1716 (ix86_trampoline_init): Likewise.
1717 (x86_output_mi_thunk): Likewise.
1718 (ix86_notrack_prefixed_insn_p): Likewise.
1719 (ix86_option_override_internal): Don't disallow -fcf-protection.
1720 * config/i386/i386.md (rdssp<mode>): Also enable for
1722 (incssp<mode>): Likewise.
1723 (nop_endbr): Likewise.
1724 * config/i386/i386.opt (mcet): Change help message to built-in
1728 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1729 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1730 enable CET built-in functions.
1732 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1734 * common/config/i386/i386-common.c
1735 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1736 OPTION_MASK_ISA_MOVDIRI_UNSET,
1737 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1738 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1739 * config.gcc (movdirintrin.h): New header.
1740 * config/i386/cpuid.h (bit_MOVDIRI,
1741 bit_MOVDIR64B): New bits.
1742 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1744 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1745 (VOID, PVOID, PCVOID)): New function types.
1746 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1747 __builtin_ia32_directstoreu_u64,
1748 __builtin_ia32_movdir64b): New builtins.
1749 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1750 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1752 (ix86_valid_target_attribute_inner_p): Ditto.
1753 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1754 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1755 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1756 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1757 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1758 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1759 (movdiri<mode>, movdir64b_<mode>): New.
1760 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1761 * config/i386/immintrin.h: Include movdirintrin.h.
1762 * config/i386/movdirintrin.h: New file.
1763 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1765 2018-04-19 Richard Biener <rguenther@suse.de>
1768 * cfg.c (clear_bb_flags): When loop state says we have
1769 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1771 2018-04-19 Richard Biener <rguenther@suse.de>
1773 PR tree-optimization/84737
1774 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1775 copying restrict info.
1776 (vect_setup_realignment): Use it.
1777 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1778 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1779 the first DR to all generated stores.
1780 (vectorizable_load): Likewise for loads.
1782 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1784 PR tree-optimization/85446
1785 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1786 the integral and pointer types to have the same precision.
1788 * doc/install.texi: Document --disable-cet being the default and
1791 2018-04-18 Martin Liska <mliska@suse.cz>
1793 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1796 2018-04-18 Martin Liska <mliska@suse.cz>
1799 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1802 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1803 arguments if they are comparable.
1805 2018-04-18 Martin Liska <mliska@suse.cz>
1808 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1811 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1814 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1817 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1818 ENDBR after calling __morestack.
1820 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1823 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1824 by using gcc_base_ver to generate a gcc_driver_version, and use
1825 it when generating GCC_DRIVER_NAME.
1826 * configure: Regenerate.
1828 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1831 * config.gcc: Obsolete powerpc*-*-*spe*.
1833 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1836 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1837 (stabstr_D): Change type of unum from unsigned int to
1838 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1841 2018-04-17 Jim Wilson <jimw@sifive.com>
1844 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1845 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1846 Set arg_pointer_offset after using pretend_args_size.
1848 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1850 PR rtl-optimization/85431
1851 * dse.c (record_store): Ignore zero width stores.
1854 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1855 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1856 __builtin_stack_restore rather than after it.
1857 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1858 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1859 argument instead of virtual_dynamic_stack_rtx.
1861 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1863 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1865 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1866 Add note to error message to explain internal mapping of overloaded
1867 built-in function name to non-overloaded built-in function name.
1868 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1871 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1874 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1875 where the inputs overlap with the output.
1877 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1879 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1880 (=v, v) alternative and explicit "memory" attribute.
1881 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1882 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1884 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1885 "sselog1" type instead of "sselog".
1886 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1887 "sselog". Remove explicit "memory" attribute.
1888 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1889 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1891 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1892 "sselog1" type instead of "sselog". Remove explicit "memory"
1894 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1895 use "sselog1" type instead of "sselog". Remove explicit "memory"
1897 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1898 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1900 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1901 "sselog1" type instead of "sselog". Remove explicit "memory"
1903 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1904 use "sselog1" type instead of "sselog". Remove explicit "memory"
1908 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1911 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1914 2018-04-17 Martin Jambor <mjambor@suse.cz>
1917 * ipa-cp.c (create_specialized_node): Call
1918 expand_all_artificial_thunks if necessary.
1920 2018-04-17 Martin Liska <mliska@suse.cz>
1923 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1924 in message, remote space in between '_G' and '('.
1926 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1929 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1930 avx512f_vmcmp<mode>3<round_saeonly_name>,
1931 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1932 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1933 avx512f_rndscale<mode><round_saeonly_name>,
1934 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1935 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1936 Use %<iptr>2 instead of %2 for -masm=intel.
1937 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1938 avx512f_vcvttss2usi<round_saeonly_name>,
1939 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1941 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1942 avx512f_vcvttsd2usi<round_saeonly_name>,
1943 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1944 Use %q1 instead of %1 for -masm=intel.
1945 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1946 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1947 of %3 for -masm=intel.
1948 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1950 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1952 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1954 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1955 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1957 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1959 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1960 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1961 %g1 and one with %0 and %1.
1962 (avx512er_vmrcp28<mode><round_saeonly_name>,
1963 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1965 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1966 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1967 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1968 of %0 and %{%4%} for -masm=intel.
1969 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1970 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1971 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1972 order of %0 and %{%5%}%{z%} for -masm=intel.
1974 2018-04-17 Jan Hubicka <jh@suse.cz>
1977 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1979 2018-04-17 Martin Liska <mliska@suse.cz>
1982 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1983 for target_clone error message. Make default implementation
1984 clone to be a local declaration.
1985 (separate_attrs): Add new argument and check for an empty
1987 (expand_target_clones): Handle it.
1988 (ipa_target_clone): Make redirection just for target_clones
1991 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1992 Tom de Vries <tom@codesourcery.com>
1995 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1996 tiled basic blocks without omp continue statements.
1998 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2001 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2002 vec_extract expression as having side effects to make sure it gets
2005 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2008 * config/i386/i386.c (get_builtin_code_for_version): Check
2011 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2014 * gcc/config.gcc: Support "skylake".
2015 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2017 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2018 (processor_target_table): Add "skylake".
2019 (ix86_option_override_internal): Add "skylake".
2020 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2021 PROCESSOR_CANNONLAKE.
2022 (get_builtin_code_for_version): Fix priority for
2023 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2024 PROCESSOR_SKYLAKE-AVX512.
2025 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2026 (processor_type): Add PROCESSOR_SKYLAKE.
2028 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2029 Jason Merrill <jason@redhat.com>
2032 * convert.c (convert_to_integer_1): Use direct recursion for
2033 enumeral types and types with a precision less than the number
2034 of bits in their mode.
2036 2018-04-16 Julia Koval <julia.koval@intel.com>
2039 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2040 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2042 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2045 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2046 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2047 and -mno-direct-move.
2049 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2052 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2053 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2054 Ensure negative shifts result in {0}.
2056 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2058 PR rtl-optimization/79916
2059 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2060 regs (if any) to define how to gnerate SD moves when LRA is in
2063 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2065 PR rtl-optimization/85393
2066 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2067 * except.c (expand_dw2_landing_pad_for_region): Make static.
2068 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2069 a label and unconditional jump to old_bb, rather than
2070 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2073 PR rtl-optimization/85376
2074 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2075 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2076 instead of a specific value.
2078 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2079 Bin Cheng <bin.cheng@arm.com>
2081 PR tree-optimization/82965
2082 PR tree-optimization/83991
2083 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2084 by_profile_only parameter.
2085 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2086 information if the loop was predicted to iterate too many times.
2087 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2089 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2092 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2095 2018-04-13 Martin Liska <mliska@suse.cz>
2096 Jakub Jelinek <jakub@redhat.com>
2099 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2100 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2101 * builtins.c (expand_builtin_memory_copy_args): Use
2102 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2103 handle dest_addr == pc_rtx.
2105 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2108 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2109 asked to not generate direct moves.
2110 (fix_trunc<mode>si2_stfiwx): Similar.
2111 (fix_trunc<mode>si2_internal): Similar.
2113 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2116 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2117 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2118 lookup if dest in some wider mode is known to be const0_rtx and
2119 if so, record permanent equivalence for it to be ZERO_EXTEND of
2120 the narrower mode destination.
2122 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2124 * lto-streamer-out.c (output_function): Revert 259346.
2125 * omp-expand.c (expand_oacc_for): Likewise.
2127 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2129 PR rtl-optimization/85354
2130 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2131 * sel-sched.c (sel_global_init): ... here.
2133 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2136 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2137 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2138 mode for PE-COFF targets.
2139 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2140 (i386_pe_asm_lto_end): Likewise.
2141 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2142 (TARGET_ASM_LTO_END): Likewise.
2143 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2144 (i386_pe_asm_lto_start): New function.
2145 (i386_pe_asm_lto_end): Likewise.
2147 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2148 Richard Biener <rguenther@suse.de>
2151 * lto-streamer-out.c (output_function): Fix CFG loop state before
2153 * omp-expand.c (expand_oacc_for): Handle calls to internal
2154 functions like regular functions.
2156 2018-04-12 Richard Biener <rguenther@suse.de>
2159 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2160 for the early LTO debug to properly generate references to it
2161 during DIE emission. Do not re-use that for the skeleton for
2163 (dwarf2out_early_finish): Likewise.
2165 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2168 * config/i386/sse.md
2169 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2170 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2171 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2172 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2173 and output is a reg, avoid creating invalid lowpart subreg, but
2174 instead split into a 512-bit move. Don't split if not AVX512VL,
2175 input is xmm16+ reg and output is a mem.
2176 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2177 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2178 xmm16+ reg and output is a mem.
2180 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2182 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2183 also for flag_dwarf2_cfi_asm.
2185 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2187 PR rtl-optimization/85342
2188 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2189 a bool scalar var inside of the loop instead. Don't try to update
2190 recog_data.operand after failed apply_change_group.
2192 2018-04-12 Tom de Vries <tom@codesourcery.com>
2195 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2196 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2197 array with flexible array member as array without given dimension.
2198 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2199 argument for undefined param to true.
2201 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2204 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2205 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2206 from PowerPC section.
2207 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2208 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2209 help text that is too long.
2210 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2211 help text that is too long.
2212 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2213 help text that is too long.
2215 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2217 * config/alpha/alpha.md (stack_probe_internal): Rename
2218 from "probe_stack". Update all callers.
2220 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2222 PR rtl-optimization/84566
2223 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2224 sched_macro_fuse_insns.
2226 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2229 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2230 (compute_block_dependences): ... from here.
2232 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2234 PR tree-optimization/85331
2235 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2236 from int to HOST_WIDE_INT.
2238 2018-04-11 Martin Jambor <mjambor@suse.cz>
2241 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2242 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2243 not the same as the source val.
2244 (cgraph_edge_brings_value_p): New parameter.
2245 (gather_edges_for_value): Pass destination value to
2246 cgraph_edge_brings_value_p.
2247 (perhaps_add_new_callers): Likewise.
2248 (get_info_about_necessary_edges): Likewise and exclude values brought
2249 only by self-recursive edges.
2250 (create_specialized_node): Redirect only clones of self-calling edges.
2251 (+self_recursive_pass_through_p): New function.
2252 (find_more_scalar_values_for_callers_subset): Use it.
2253 (find_aggregate_values_for_callers_subset): Likewise.
2254 (known_aggs_to_agg_replacement_list): Removed.
2255 (decide_whether_version_node): Re-calculate known constants for all
2256 remaining context clones.
2258 2018-04-11 Richard Biener <rguenther@suse.de>
2261 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2262 from early DWARF output.
2263 (dwarf2out_early_finish): Output line info unconditionally into
2264 early DWARF and add reference to it.
2266 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2269 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2270 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2271 other than V2DFmode using iptr mode attribute.
2272 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2274 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2276 PR rtl-optimization/84659
2277 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2279 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2282 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2284 (output_loc_list): Pass address of a dummy size variable even in the
2285 locview handling loop.
2286 (index_location_lists): Add comment on why skip_loc_list_entry can't
2289 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2292 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2295 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2298 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2299 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2300 and -mstring-compare-inline-limit.
2302 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2305 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2306 for stack clash protection in a register whenever we need it to be in
2309 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2311 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2312 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2314 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2317 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2319 (mlong-double-): Ditto.
2320 * config/rs6000/sysv4.opt (msdata=): Ditto.
2321 (mtls-size=): Ditto.
2323 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2325 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2326 erroneous entries for
2327 "vector int vec_ldl (int, long int *)", and
2328 "vector unsigned int vec_ldl (int, unsigned long int *)".
2329 Add comments and entries for
2330 "vector bool char vec_ldl (int, bool char *)",
2331 "vector bool short vec_ldl (int, bool short *)",
2332 "vector bool int vec_ldl (int, bool int *)",
2333 "vector bool long long vec_ldl (int, bool long long *)",
2334 "vector pixel vec_ldl (int, pixel *)",
2335 "vector long long vec_ldl (int, long long *)",
2336 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2337 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2338 type tree bool_long_long_type_node and correct definition of
2339 bool_V2DI_type_node to make reference to this new type tree.
2340 (rs6000_mangle_type): Replace erroneous reference to
2341 bool_long_type_node with bool_long_long_type_node.
2342 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2343 comments to emphasize sign distinctions for char and int types and
2344 replace RS6000_BTI_bool_long constant with
2345 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2346 use of RS6000_BTI_pixel.
2347 (bool_long_type_node): Remove this macro definition.
2348 (bool_long_long_type_node): New macro definition
2350 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2352 PR rtl-optimization/85300
2353 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2354 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2355 simplify_unary_operation fails.
2357 2018-04-10 Martin Liska <mliska@suse.cz>
2359 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2360 cgraph_edge and ipa_ref.
2362 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2366 * config/i386/sse.md
2367 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2368 computation of the VEC_MERGE selector from mask.
2369 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2370 Fix decoding of the VEC_MERGE selector into mask.
2372 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2374 PR tree-optimization/85286
2375 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2377 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2379 * final.c (final_1): Set insn_last_address as well as
2380 insn_current_address.
2382 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2385 * explow.c (emit_stack_probe): Call validize_mem on memory location
2386 before passing it to gen_probe_stack. Create address operand and
2387 legitimize it for the probe_stack_address case.
2389 2018-04-09 Jan Hubicka <jh@suse.cz>
2392 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2393 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2394 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2396 (free_lang_data): Rebuild type inheritance graph.
2398 2018-04-09 Martin Sebor <msebor@redhat.com>
2400 * invoke.texi (-finline-small-functions): Mention other optimization
2402 (-findirect-inlining, -fpartial-inlining): Same.
2403 (-finline-functions-called-once): Same.
2404 (-freorder-blocks-and-partition): Same.
2406 2018-04-09 Jan Hubicka <jh@suse.cz>
2409 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2410 jumps; choose last target that matches the criteria (i.e.
2411 no partition changes for non-crossing jumps).
2412 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2413 support for redirecting crossing jumps to non-crossing.
2415 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2417 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2418 also for naked functions.
2420 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2422 * config/arc/arc.md (add_shift): New pattern.
2423 (add_shift2): Likewise.
2424 (sub_shift): Likewise.
2425 (sub_shift_cmp0_noout): Likewise.
2426 (compare_si_ashiftsi): Likewise.
2427 (xbfu_cmp0_noout): New combine pattern.
2428 (xbfu_cmp0"): Likewise.
2429 (movsi_set_cc_insn): Place the predicable variant first.
2430 (commutative_binary_cmp0_noout): Remove clobber.
2431 (commutative_binary_cmp0): New pattern.
2432 (noncommutative_binary_cmp0): Likewise.
2433 (noncommutative_binary_cmp0_noout): Likewise.
2434 (noncommutative_binary_comparison_result_used): Removed.
2435 (rsub_cmp0): New pattern.
2436 (rsub_cmp0_noout): Likewise.
2437 (extzvsi): Changed, keep only meaningful variants.
2438 (SQH, SEZ): New iterators.
2439 (SQH_postfix): New mode attribute.
2440 (SEZ_prefix): New code attribute.
2441 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2442 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2443 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2445 (noncommutative_operator): Check the availability of barrel
2448 2018-04-09 Richard Biener <rguenther@suse.de>
2450 PR tree-optimization/85284
2451 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2452 Only use the niter constraining form of simple_iv when the exit
2455 2018-04-09 Tom de Vries <tom@codesourcery.com>
2458 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2459 (define_expand "*memory_barrier"): New define_expand.
2460 (define_insn "memory_barrier"): New insn.
2462 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2464 PR rtl-optimization/80463
2465 PR rtl-optimization/83972
2466 PR rtl-optimization/83480
2468 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2469 correct producer for the insn.
2470 (tidy_control_flow): Fixup seqnos in case of debug insns.
2472 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2474 PR rtl-optimization/83913
2476 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2477 different sched-times when merging exprs.
2479 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2481 PR rtl-optimization/83962
2483 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2484 tidy_fallthru_edge and tidy_control_flow.
2486 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2488 PR rtl-optimization/83530
2490 * sel-sched.c (force_next_insn): New global variable.
2491 (remove_insn_for_debug): When force_next_insn is true, also leave only
2492 next insn in the ready list.
2493 (sel_sched_region): When the region wasn't scheduled, make another pass
2494 over it with force_next_insn set to 1.
2496 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2498 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2500 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2501 for interrupt control.
2502 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2503 functions for interrupt control.
2504 * config/nds32/nds32-intrinsic.md: Likewise.
2505 * config/nds32/nds32_intrinsic.h: Likewise.
2506 * config/nds32/nds32.h (nds32_builtins): Likewise.
2508 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2510 * config/nds32/nds32.c (nds32_init_machine_status,
2511 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2512 strict_aligned_p field.
2513 (nds32_expand_to_rtl_hook): New function.
2514 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2515 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2517 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2518 Chung-Ju Wu <jasonwucj@gmail.com>
2520 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2521 * config/nds32/nds32-n7.md: New file.
2522 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2523 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2525 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2526 * config/nds32/nds32.md (pipeline_model): Add n7.
2527 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2528 * config/nds32/pipelines.md: Include n7 settings.
2530 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2531 Chung-Ju Wu <jasonwucj@gmail.com>
2533 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2534 * config/nds32/nds32-e8.md: New file.
2535 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2536 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2538 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2539 * config/nds32/nds32.md (pipeline_model): Add e8.
2540 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2541 * config/nds32/pipelines.md: Include e8 settings.
2543 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2544 Chung-Ju Wu <jasonwucj@gmail.com>
2546 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2547 * config/nds32/nds32-n8.md: New file.
2548 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2549 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2551 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2552 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2553 * config/nds32/nds32.md (pipeline_model): Add n8.
2554 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2555 * config/nds32/pipelines.md: Include n8 settings.
2557 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2558 Chung-Ju Wu <jasonwucj@gmail.com>
2560 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2561 * config/nds32/nds32-n9-2r1w.md: New file.
2562 * config/nds32/nds32-n9-3r2w.md: New file.
2563 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2564 nds32_register_ports): New or modify for cpu n9.
2565 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2567 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2568 * config/nds32/nds32-utils.c: New file.
2569 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2570 TARGET_MUL_SLOW): Define.
2571 * config/nds32/nds32.md (pipeline_model): New attribute.
2572 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2573 New options that support cpu n9.
2574 * config/nds32/pipelines.md: Include n9 settings.
2575 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2577 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2579 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2580 information if necessary.
2581 (output_cond_branch_compare_zero): Likewise.
2582 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2583 (nds32_target_alignment): Refine for alignment.
2584 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2585 (FUNCTION_BOUNDARY): Modify.
2586 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2588 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2590 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2592 * config/nds32/constants.md (unspec_volatile_element): Add values for
2593 TLB operation and data prefetch.
2594 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2595 functions for TLB operation and data prefetch.
2596 * config/nds32/nds32-intrinsic.md: Likewise.
2597 * config/nds32/nds32_intrinsic.h: Likewise.
2598 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2599 (nds32_print_operand): Likewise.
2600 * config/nds32/nds32.h (nds32_builtins): Likewise.
2602 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2603 Andrew Pinski <pinsika@gcc.gnu.org>
2606 * match.pd: Use constant_boolean_node of correct type instead of
2607 boolean_true_node or boolean_false_node for simplifying
2608 pointer comparisons to zero.
2610 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2612 PR tree-optimization/80021
2613 * tree.c (verify_type_variant): Make error call in verify_variant_match
2614 translatable and remove final full stop.
2616 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2618 * config/nds32/constants.md (unspec_volatile_element): Add
2619 UNSPEC_VOLATILE_EH_RETURN.
2620 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2621 nds32_output_stack_pop): Support dwarf exception handling process.
2622 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2623 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2624 exception handling process.
2625 (nds32_compute_stack_frame): Likewise.
2626 (nds32_return_addr_rtx): Likewise.
2627 (nds32_initial_elimination_offset): Likewise.
2628 (nds32_expand_prologue): Likewise.
2629 (nds32_expand_epilogue): Likewise.
2630 (nds32_dynamic_chain_address): New function.
2631 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2633 (DYNAMIC_CHAIN_ADDRESS): Define.
2634 (EH_RETURN_DATA_REGNO): Define.
2635 (EH_RETURN_STACKADJ_RTX): Define.
2636 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2637 patterns for dwarf exception handling.
2639 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2641 * config/nds32/nds32.h: Clean up obsolete macros.
2643 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2645 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2646 Add enum values for particular instructions.
2647 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2648 particular intrinsic functions.
2649 * config/nds32/nds32-intrinsic.md: Likewise.
2650 * config/nds32/nds32_intrinsic.h: Likewise.
2651 * config/nds32/nds32.h (nds32_builtins): Likewise.
2652 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2653 (btst, ave): New patterns for particular instructions.
2655 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2657 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2658 Add enum values for atomic load/store and memory sync.
2659 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2661 * config/nds32/nds32-intrinsic.md: Likewise.
2662 * config/nds32/nds32_intrinsic.h: Likewise.
2663 * config/nds32/nds32.h (nds32_builtins): Likewise.
2665 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2667 PR tree-optimization/85257
2668 * fold-const.c (native_encode_vector): If not all elts could fit
2669 and off is -1, return 0 rather than offset.
2670 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2671 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2672 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2673 adjust buffer in native_interpret_expr call.
2675 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2677 * config/nds32/constants.md (unspec_volatile_element): Add cache
2678 control enum values.
2679 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2680 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2681 * config/nds32/nds32.c (nds32_cctl_names): New.
2682 (nds32_print_operand): Handle cache control register names.
2683 * config/nds32/nds32.h (nds32_builtins): New enum values.
2684 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2686 * config/nds32/nds32.md (type): Add mmu.
2687 * config/nds32/pipelines.md (simple_insn): Add mmu.
2689 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2691 * config/nds32/nds32.md (type): Remove call.
2692 * config/nds32/pipelines.md (simple_insn): Likewise.
2694 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2696 * config/nds32/constants.md (unspec_volatile_element): Add
2697 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2698 UNSPEC_VOLATILE_FMFCFG.
2699 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2700 description for fmfcfg and fmfcsr.
2701 (bdesc_1arg): Add fmtcsr.
2702 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2703 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2704 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2705 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2706 unspec_fmfcfg): New patterns.
2707 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2708 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2709 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2710 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2711 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2712 __nds32__fmfcfg): Define.
2714 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2716 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2717 intrinsic register names.
2718 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2719 intrinsic register enum values and macros.
2721 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2723 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2724 for load/store addressing form.
2725 (nds32_print_operand_address): Likewise.
2727 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2730 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2731 based on LABEL_REF. Remove useless assertion.
2732 (pic_address_needs_scratch): Fix formatting.
2733 (sparc_legitimize_pic_address): Minor tweaks.
2734 (sparc_delegitimize_address): Adjust assertion accordingly.
2735 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2736 into symbolic_operand.
2737 (movsi_high_pic_label_ref): Likewise.
2738 (movsi_lo_sum_pic_label_ref): Likewise.
2739 (movdi_pic_label_ref): Likewise.
2740 (movdi_high_pic_label_ref): Likewise.
2741 (movdi_lo_sum_pic_label_ref): Likewise.
2743 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2745 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2746 custom LIB_SPEC setup.
2748 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2749 Kito Cheng <kito.cheng@gmail.com>
2751 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2752 * config/riscv/freebsd.h: New.
2754 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2756 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2757 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2760 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2761 Kito Cheng <kito.cheng@gmail.com>
2763 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2764 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2765 * config/nds32/nds32-protos.h (nds32_output_call,
2766 nds32_output_return): Declare.
2767 * config/nds32/nds32.md: Refine all the call and return patterns.
2769 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2772 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2773 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2775 PR rtl-optimization/84872
2776 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2777 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2780 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2782 * expr.c (copy_blkmode_to_reg): Revert 254862.
2783 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2785 2018-04-06 Richard Biener <rguenther@suse.de>
2788 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2789 after seeing a component reference with an adjacent field. Treat
2790 refs to arrays at struct end of external decls similar to
2791 refs to unconstrained commons.
2793 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2796 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2797 look through SAVE_EXPRs with non-side-effects argument. Adjust
2799 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2802 2018-04-06 Richard Biener <rguenther@suse.de>
2805 * alias.c (find_base_term): New wrapper around find_base_term
2806 unwinding CSELIB_VAL_PTR changes.
2807 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2810 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2812 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2814 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2815 constant definitions.
2816 ("nop"): lr 0,0 -> nopr r0
2817 ("nop_lr0", "nop_lr1"): New insn definitions.
2819 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2821 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2822 NDS32_V3PUSH_AVAILABLE_P macro.
2824 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2825 Chung-Ju Wu <jasonwucj@gmail.com>
2827 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2828 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2829 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2830 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2831 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2832 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2833 * config/nds32/constraints.md: New constraints and checking for hard
2834 float configuration.
2835 * config/nds32/iterators.md: New mode iterator and attribute for hard
2836 float configuration.
2837 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2839 * config/nds32/nds32-fpu.md: New file.
2840 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2841 deal with hard float code generation.
2842 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2844 (abi_type, float_reg_number): New enum type.
2845 * config/nds32/nds32-predicates.c: New predicates for hard float.
2846 * config/nds32/nds32-protos.h: Declare functions for hard float.
2847 * config/nds32/nds32.c: Implementation for hard float configuration.
2848 * config/nds32/nds32.h: Definitions for hard float configuration.
2849 * config/nds32/nds32.md: Include hard float machine description and
2850 modify patterns for hard float configuration.
2851 * config/nds32/nds32.opt: New options for hard float configuration.
2852 * config/nds32/predicates.md: New predicates for hard float
2855 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2857 * common/config/nds32/nds32-common.c
2858 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2860 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2863 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2864 CONSTRUCTOR_ELT (ctor, ...)->value.
2866 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2869 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2871 2018-04-05 Tom de Vries <tom@codesourcery.com>
2874 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2877 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2878 Kito Cheng <kito.cheng@gmail.com>
2880 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2881 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2882 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2883 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2885 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2886 Kito Cheng <kito.cheng@gmail.com>
2888 * config/nds32/constraints.md (Ufe): New memory constraint.
2889 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2890 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2891 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2893 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2894 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2896 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2898 * config/nds32/nds32.md: Use optimize_size in the condition for
2899 alu-shift instructions.
2901 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2903 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2905 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2907 * config/nds32/nds32.md (negsi2): Refine pattern.
2909 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2910 Chung-Ju Wu <jasonwucj@gmail.com>
2912 * config/nds32/iterators.md (shift_rotate): New code iterator.
2913 (shift): New code attribute.
2914 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2915 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2916 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2917 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2918 bit-wise operations.
2919 (andsi3, *andsi3): Ditto.
2920 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2921 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2922 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2923 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2924 nds32_ior_operand, nds32_xor_operand): New predicates.
2926 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2928 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2929 (addsi3, subsi3): ... this.
2931 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2933 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2935 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2937 * config/nds32/nds32.md: Adjust indention.
2939 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2941 * config/nds32/nds32.md (feature): New attribute.
2943 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2945 * config/nds32/nds32.md (subtype): New attribute.
2947 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2950 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2951 expansion to perform a bitwise AND of the argument followed by a
2952 boolean negation of the result.
2954 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2956 PR rtl-optimization/84878
2957 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2958 the basic block. Assert the use reference is not artificial and that
2959 it has an associated insn.
2961 2018-04-04 Michael Matz <matz@suse.de>
2963 * builtins.c (compute_objsize): Pass correct operand
2964 to array_at_struct_end_p.
2966 2018-04-04 Richard Biener <rguenther@suse.de>
2969 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2970 from contexts for DINFO_LEVEL_TERSE and below.
2972 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2974 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2975 resiter_operand condition.
2976 * config/nds32/nds32.md (*move<mode>): Ditto.
2978 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2979 Monk Chiang <sh.chiang04@gmail.com>
2981 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2983 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2985 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2987 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2988 Kito Cheng <kito.cheng@gmail.com>
2990 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2991 nds32_cond_code_str, output_cond_branch,
2992 output_cond_branch_compare_zero, nds32_expand_cbranch,
2993 nds32_expand_cstore, nds32_expand_movcc,
2994 nds32_output_cbranchsi4_equality_zero,
2995 nds32_output_cbranchsi4_equality_reg,
2996 nds32_output_cbranchsi4_equality_reg_or_const_int,
2997 nds32_output_cbranchsi4_greater_less_zero: New functions.
2998 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2999 nds32_expand_cstore, nds32_expand_movcc,
3000 nds32_output_cbranchsi4_equality_zero,
3001 nds32_output_cbranchsi4_equality_reg,
3002 nds32_output_cbranchsi4_equality_reg_or_const_int,
3003 nds32_output_cbranchsi4_greater_less_zero): Declare.
3004 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3005 nds32_rimm11s_operand): New predicates.
3006 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3007 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3010 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3012 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3013 * config/nds32/nds32.md: Ditto.
3014 * config/nds32/pipelines.md: Ditto.
3016 2018-04-04 Richard Biener <rguenther@suse.de>
3018 PR tree-optimization/85168
3019 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3020 propagating abnormals.
3022 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3024 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3026 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3027 Kito Cheng <kito.cheng@gmail.com>
3029 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3030 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3031 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3032 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3033 * config/nds32/nds32.md (sibcall_internal): New.
3034 (sibcall_register): Remove.
3035 (sibcall_immediate): Remove.
3036 (sibcall_value_internal): New.
3037 (sibcall_value_register): Remove.
3038 (sibcall_value_immediate): Remove.
3039 * config/nds32/predicates.md (nds32_general_register_operand): New.
3040 (nds32_call_address_operand): New.
3042 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3044 PR rtl-optimization/85167
3045 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3046 bb_defs if *split_p, instead preinitialize it to NULL.
3048 PR tree-optimization/85156
3049 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3050 evaluating the argument multiple times.
3052 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3054 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3056 (_mm_cvtpd_ps): Likewise.
3057 (_mm_cvttpd_epi32): Likewise.
3058 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3059 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3060 vector, pixel, and bool following altivec.h include.
3062 2018-04-03 Martin Sebor <msebor@redhat.com>
3064 * doc/extend.texi (Common Function Attributes): Clarify.
3065 (const attribute): Likewise.
3066 (pure attribute): Likewise.
3068 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3071 * config/i386/i386.c (ix86_expand_vector_set): Use
3072 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3074 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3076 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3077 instructions when changing rounding bits to preserve precision bits
3078 in the x87 control word.
3080 2018-04-03 Martin Liska <mliska@suse.cz>
3082 PR tree-optimization/82491
3083 * rtl.h (strip_offset_and_add): Replace += suboffset with
3084 poly_uint64 () + suboffset.
3086 2018-03-29 Martin Liska <mliska@suse.cz>
3087 Martin Jambor <mjambor@suse.cz>
3090 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3091 param_type is not an integral or pointer type.
3093 2018-04-03 Richard Biener <rguenther@suse.de>
3095 * sese.h (recompute_all_dominators): Remove.
3097 2018-04-02 Martin Sebor <msebor@redhat.com>
3099 * doc/invoke.texi (-Wrestrict): Fix typos.
3101 2018-04-02 Jim Wilson <jimw@sifive.com>
3103 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3104 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3105 (<optab>di3, <optab>si3_extend): Likewise.
3106 (<optab>si3_mask, <optab>si3_mask_1): New.
3107 (<optab>di3_mask, <optab>di3_mask_1): New.
3108 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3109 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3110 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3112 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3114 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3117 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3119 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3120 (nds32_canonicalize_comparison): New function.
3122 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3123 Kito Cheng <kito.cheng@gmail.com>
3124 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3126 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3127 * config/nds32/constants.md (unspec_volatile_element): Add
3128 UNSPEC_VOLATILE_RELAX_GROUP.
3129 * config/nds32/nds32-relax-opt.c: New file.
3130 * config/nds32/nds32-predicates.c
3131 (nds32_symbol_load_store_p): New function.
3132 * config/nds32/nds32-protos.h
3133 (nds32_symbol_load_store_p): Declare function.
3134 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3135 * config/nds32/nds32.c
3136 (nds32_register_pass): New function to register pass.
3137 (nds32_register_passes): New function to register passes.
3138 * config/nds32/nds32.md (relax_group): New pattern.
3139 * config/nds32/nds32.opt (mrelax-hint): New option.
3140 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3142 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3144 * config/nds32/t-nds32: Modify files dependency.
3146 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3148 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3149 (PROFILE_HOOK): Define its implementation.
3151 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3153 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3154 type and 32-bit size.
3156 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3159 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3160 (V_128_256): New mode iterator.
3161 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3162 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3163 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3165 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3166 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3168 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3171 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3172 NaN inputs correctly.
3174 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3177 * config/rs6000/vsx.md (??r): New mode attribute.
3178 (*vsx_mov<mode>_64bit): Use it.
3179 (*vsx_mov<mode>_32bit): Likewise.
3181 2018-03-30 Martin Sebor <msebor@redhat.com>
3183 PR tree-optimization/84818
3184 * builtins.c (check_access): Use warning_n.
3186 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3189 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3191 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3194 2018-03-30 Julia Koval <julia.koval@intel.com>
3197 * x86-tune.def (movx, partial_reg_dependency): Enable for
3200 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3203 * lra-constraints.c (process_alt_operands): Move setting
3204 this_alternative_matches below.
3206 2018-03-29 Martin Liska <mliska@suse.cz>
3209 * doc/invoke.texi: Document how LTO works with debug info.
3210 Describe auto-load support of binutils. Mention 'x86-64'
3211 as valid option value of -march option.
3213 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3215 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3218 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3219 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3220 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3223 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3226 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3227 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3228 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3229 for RS6000_BTM_POWERPC64.
3230 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3231 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3232 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3237 2018-03-28 Carl Love <cel@us.ibm.com>
3240 2017-09-27 Carl Love <cel@us.ibm.com>
3242 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3243 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3244 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3247 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3249 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3250 instead of __vector bool.
3251 (_mm_max_pu8): Likewise.
3252 (_mm_min_pi16): Likewise.
3254 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3257 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3258 (DIVWEUO): Likewise.
3260 (DIVDEUO): Likewise.
3261 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3262 DIVWEUO and DIVDEUO.
3263 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3264 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3265 (div_extend): Likewise.
3266 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3268 (__builtin_divweuo): Likewise.
3269 (__builtin_divdeo): Likewise.
3270 (__builtin_divdeuo): Likewise.
3272 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3275 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3276 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3278 PR tree-optimization/82004
3279 * gimple-match-head.c (optimize_pow_to_exp): New function.
3280 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3281 Don't fold to exp if optimize_pow_to_exp is false.
3283 2018-03-28 Martin Liska <mliska@suse.cz>
3286 * calls.c (initialize_argument_information): Fix trailing space.
3287 * common.opt: Fix typo and provide better explanation for
3288 -fsanitize-coverage option.
3289 * config/i386/i386.opt: Fix typo.
3291 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3292 Martin Liska <mliska@suse.cz>
3295 * gimplify.c (asan_poison_variable): Don't do the check for
3296 gimplify_omp_ctxp here.
3297 (gimplify_decl_expr): Do it here.
3298 (gimplify_target_expr): Likewise.
3300 2018-03-28 Martin Liska <mliska@suse.cz>
3303 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3304 chkp_type_bounds_count if MPX is not enabled.
3306 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3308 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3310 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3313 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3314 function to create the function decl for complex long double
3315 multiply and divide for -mabi=ieeelongdouble.
3316 (init_float128_ieee): Call it.
3318 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3321 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3322 -fcf-protection=branch -mibt.
3323 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3325 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3328 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3330 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3333 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3334 extern array declarations.
3336 2018-03-27 Richard Biener <rguenther@suse.de>
3339 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3340 explicit single_use checks.
3342 2018-03-27 Richard Biener <rguenther@suse.de>
3344 PR tree-optimization/85082
3345 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3348 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3350 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3351 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3352 Turn on fasynchronous-unwind-tables and funwind-tables.
3354 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3357 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3358 (*bmi_blsr_<mode>_ccz): Ditto.
3360 2018-03-26 Tom de Vries <tom@codesourcery.com>
3362 PR tree-optimization/85063
3363 * omp-general.c (offloading_function_p): New function. Factor out
3365 * omp-offload.c (pass_omp_target_link::gate): ... here.
3366 * omp-general.h (offloading_function_p): Declare.
3367 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3368 with attribute omp declare target for offloading functions.
3370 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3372 PR tree-optimization/84005
3373 * tree-data-ref.h (get_base_for_alignment): Declare.
3374 * tree-data-ref.c (get_base_for_alignment_1): New function.
3375 (get_base_for_alignment): Likewise.
3376 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3377 get_base_for_alignment to find a suitable base object, instead
3378 of always using drb->base_address.
3380 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3383 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3384 known size by default.
3386 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3389 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3390 and non BLKmode operands.
3392 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3395 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3396 Clean up attributes.
3398 2018-03-23 Richard Biener <rguenther@suse.de>
3401 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3402 we are going to emit early debug for LTO.
3404 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3407 * function.c (match_asm_constraints_1): Don't optimize if input
3408 doesn't satisfy general_operand predicate for output's mode.
3411 * alias.c (write_dependence_p): Don't require for x_canonicalized
3412 non-VOIDmode if x has VOIDmode.
3415 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3416 just don't try to optimize it rather than assert it never happens.
3418 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3420 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3421 macro expansions for definition of ST_INTERNAL_<mode> and
3422 LD_INTERNAL_<mode> builtins.
3423 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3425 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3427 (altivec_expand_st_builtin): Likewise.
3428 (altivec_expand_builtin): Remove calls to deleted functions.
3429 (rs6000_address_for_altivec): Delete this function.
3430 * config/rs6000/vector.md: Remove expands for
3431 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3433 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3436 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3437 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3438 re-computing once computed.
3439 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3440 (arm_init_machine_status): Initialize
3441 machine->static_chain_stack_bytes.
3443 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3446 * doc/extend.texi: Add four new prototypes for vec_ld.
3447 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3448 definitions for more logical presentation.
3449 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3450 entries for V1TI variants of __builtin_altivec_ld builtin.
3451 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3452 handling of V1TI variant of LVX icode pattern.
3453 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3454 (rs6000_gimple_fold_builtin): Likewise.
3455 (altivec_init_builtins): Add code to define
3456 __builtin_altivec_lvx_v1ti function.
3458 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3461 * function.c (match_asm_constraints_1): Don't do the optimization
3462 if input isn't a REG, SUBREG, MEM or constant.
3464 2018-03-22 Tom de Vries <tom@codesourcery.com>
3466 PR tree-optimization/84956
3467 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3468 bb_has_abnormal_pred.
3470 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3473 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3474 DECL_INITIAL (decl) to decl at the end.
3475 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3478 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3480 * doc/extend.texi (__builtin_tgmath): Document when complex
3481 integer types are treated as _Complex _Float64.
3483 2018-03-21 Tom de Vries <tom@codesourcery.com>
3485 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3487 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3489 PR tree-optimization/84960
3490 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3491 if it is ENTRY block, move them into single succ of ENTRY in that case.
3493 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3495 PR tree-optimization/84811
3496 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3497 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3498 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3499 (known_subrange_p): Update accordingly. Cast each value involved
3500 in the size comparison, rather than casting the result of the
3503 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3505 PR tree-optimization/84982
3506 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3507 by flipping the least significant bit rather than all bits from
3508 bitpos to bitpos + bitsize - 1.
3510 2018-03-21 Nathan Sidwell <nathan@acm.org>
3512 * doc/extend.texi (Deprecated Features): Remove mention of
3513 long-deleted deprecations.
3515 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3518 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3519 * configure: Regenerate.
3521 2018-03-21 Tom de Vries <tom@codesourcery.com>
3523 PR tree-optimization/83126
3524 * tree-parloops.c (num_phis): New function.
3525 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3527 2018-03-21 Nathan Sidwell <nathan@acm.org>
3529 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3530 mention anon-struct/union members and trailing attributes.
3532 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3534 PR tree-optimization/84969
3535 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3536 builtin memset partitions if they set different rhs values.
3538 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3540 PR rtl-optimization/84989
3541 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3542 VEC_DUPLICATE with scalar result mode.
3544 2018-03-21 Martin Liska <mliska@suse.cz>
3547 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3548 not intended return statement.
3550 2018-03-21 Martin Liska <mliska@suse.cz>
3553 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3554 (chkp_find_bound_slots_1): Limit number of iterations.
3556 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3559 * Minor grammar fixes for x86 options.
3561 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3564 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3565 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3567 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3570 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3571 (altivec_lvx_<mode>_1op): Likewise.
3572 (altivec_stvx_<mode>_2op): Likewise.
3573 (altivec_stvx_<mode>_1op): Likewise.
3574 (altivec_lvx_<VM2:mode>): New define_expand.
3575 (altivec_stvx_<VM2:mode>): Likewise.
3576 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3577 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3578 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3579 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3580 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3581 (rs6000_gen_lvx): Likewise.
3582 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3583 (altivec_expand_stv_builtin): Likewise.
3584 (altivec_expand_builtin): Likewise.
3585 * config/rs6000/vector.md: Likewise.
3587 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3590 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3593 2018-03-20 Richard Biener <rguenther@suse.de>
3596 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3597 sign-conversions as zero, fall back to standard scalar_stmt
3600 2018-03-20 Martin Liska <mliska@suse.cz>
3603 * predict.c (rebuild_frequencies): Handle case when we have
3604 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3606 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3609 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3610 flag_section_anchors.
3611 * varasm.c (use_blocks_for_decl_p): Remove hack for
3612 dw2_force_const_mem.
3615 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3617 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3618 be created, use lowpart_subreg of operands[0] rather than operands[0]
3620 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3621 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3622 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3623 and n constraint instead of aarch64_shift_imm_di and Usd.
3624 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3625 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3627 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3630 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3631 to favor GPR over NEON registers.
3632 (<shift>di3_neon): Likewise.
3634 2018-03-20 Tom de Vries <tom@codesourcery.com>
3637 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3638 (nvptx_process_pars): Emit bar.sync asap and alap.
3640 2018-03-20 Tom de Vries <tom@codesourcery.com>
3643 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3644 seen_label if seen_label is already set.
3646 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3649 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3650 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3651 Use 1U instead of 1. Formatting fixes.
3654 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3655 instead of TREE_TYPE (s1) for the return value.
3657 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3659 PR tree-optimization/84946
3660 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3661 bitsize + bitsize in poly_uint64 rather than poly_int64.
3664 * dwarf2asm.c: Include fold-const.c.
3665 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3666 of decl rather than decl itself.
3668 PR rtl-optimization/84643
3669 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3671 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3674 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3675 calling assemble_variable.
3677 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3680 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3681 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3683 2018-03-19 Jim Wilson <jimw@sifive.com>
3686 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3687 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3688 (riscv_first_stack_step): Likewise.
3689 (riscv_option_override): Use STACK_BOUNDARY instead of
3691 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3693 (BIGGEST_ALIGNMENT): Set to 128.
3694 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3695 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3698 2018-03-19 Richard Biener <rguenther@suse.de>
3700 PR tree-optimization/84933
3701 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3702 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3704 2018-03-19 Richard Biener <rguenther@suse.de>
3706 PR tree-optimization/84859
3707 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3708 (cond_if_else_store_replacement): Perform sinking operation on
3709 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3710 Generalize what a BB with a single eligible store is.
3712 2018-03-19 Richard Biener <rguenther@suse.de>
3714 PR tree-optimization/84929
3715 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3716 chrec_is_positive against non-chrec arg.
3718 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3721 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3723 2018-03-18 Martin Liska <mliska@suse.cz>
3725 PR rtl-optimization/84635
3726 * regrename.c (build_def_use): Use matches_mode only when
3729 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3731 PR tree-optimization/84913
3732 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3733 vectorize chains of COND_EXPRs.
3735 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3737 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3739 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3741 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3743 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3745 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3747 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3748 Kito Cheng <kito.cheng@gmail.com>
3750 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3751 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3752 (nds32_adjust_reg_alloc_order): New function.
3753 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3755 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3757 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3758 nds32_print_operand, nds32_print_operand_address): Use
3759 HOST_WIDE_INT_PRINT_DEC instead.
3761 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3763 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3765 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3768 * config/i386/i386.c (initial_ix86_tune_features,
3769 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3771 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3772 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3773 rather than 1u << ix86_tune. Formatting fix.
3774 (ix86_option_override_internal): Change ix86_arch_mask from
3775 unsigned int to unsigned HOST_WIDE_INT, initialize to
3776 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3777 (ix86_function_specific_restore): Likewise.
3779 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3782 * postreload.c (reload_combine_recognize_pattern): Perform
3783 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3784 truncate_int_for_mode the result for the destination's mode.
3787 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3788 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3790 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3793 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3796 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3797 regno_allocno_class_array and sorted_pseudos.
3798 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3799 insns where regno is used.
3801 2018-03-16 Martin Liska <mliska@suse.cz>
3804 * multiple_target.c (create_dispatcher_calls): Redirect
3805 reference in the symbol table.
3807 2018-03-16 Martin Liska <mliska@suse.cz>
3810 * multiple_target.c (create_dispatcher_calls): Redirect also
3813 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3817 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3818 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3819 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3821 2018-03-16 Julia Koval <julia.koval@intel.com>
3823 * doc/invoke.texi (Skylake Server): Add CLWB.
3824 Cannonlake): Remove CLWB.
3826 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3828 PR tree-optimization/84841
3829 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3831 (FLOAT_ONE_CONST_TYPE): Define.
3832 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3833 (sort_by_operand_rank): Put entries with higher constant_type last
3834 rather than first to match comments.
3836 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3838 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3841 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3844 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3845 check and use address of target always.
3847 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3850 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3851 (indirect_thunk_bnd_needed): Likewise.
3852 (indirect_thunks_used): Likewise.
3853 (indirect_thunks_bnd_used): Likewise.
3854 (indirect_return_needed): New.
3855 (indirect_return_bnd_needed): Likewise.
3856 (output_indirect_thunk_function): Add a bool argument for
3858 (output_indirect_thunk_function): Don't generate alias for
3859 function return thunk.
3860 (ix86_code_end): Call output_indirect_thunk_function to generate
3861 function return thunks.
3862 (ix86_output_function_return): Set indirect_return_bnd_needed
3863 and indirect_return_needed instead of indirect_thunk_bnd_needed
3864 and indirect_thunk_needed.
3866 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3868 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3869 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3870 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3872 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3873 Paul Hua <paul.hua.gm@gmail.com>
3876 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3878 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3880 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3881 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3884 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3887 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3888 instead of GET_MODE_SIZE when comparing Units.
3890 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3893 * varasm.c (hash_section): Return an unchangeble hash value
3894 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3895 Return !aarch64_can_use_per_function_literal_pools_p ().
3897 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3900 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3901 rather than address of cmode as last argument to prepare_cmp_insn.
3903 2018-03-15 Julia Koval <julia.koval@intel.com>
3905 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3906 F_AVX512VNNI, F_AVX512BITALG): New.
3908 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3911 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3912 insn for floating-point loads and stores.
3914 2018-03-14 Carl Love <cel@us.ibm.com>
3916 * config/rs6000/rs6000-c.c: Add macro definitions for
3917 ALTIVEC_BUILTIN_VEC_PERMXOR.
3918 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3919 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3920 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3921 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3923 * config/doc/extend.texi: Add prototypes for vec_permxor.
3925 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3928 * diagnostic-show-locus.c (class layout_point): Convert m_line
3929 from int to linenum_type.
3930 (line_span::comparator): Use linenum "compare" function when
3931 comparing line numbers.
3932 (test_line_span): New function.
3933 (layout_range::contains_point): Convert param "row" from int to
3935 (layout_range::intersects_line_p): Likewise.
3936 (layout::will_show_line_p): Likewise.
3937 (layout::print_source_line): Likewise.
3938 (layout::should_print_annotation_line_p): Likewise.
3939 (layout::print_annotation_line): Likewise.
3940 (layout::print_leading_fixits): Likewise.
3941 (layout::annotation_line_showed_range_p): Likewise.
3942 (struct line_corrections): Likewise for field m_row.
3943 (line_corrections::line_corrections): Likewise for param "row".
3944 (layout::print_trailing_fixits): Likewise.
3945 (layout::get_state_at_point): Likewise.
3946 (layout::get_x_bound_for_row): Likewise.
3947 (layout::print_line): Likewise.
3948 (diagnostic_show_locus): Likewise for locals "last_line" and
3950 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3951 * input.c (selftest::test_linenum_comparisons): New function.
3952 (selftest::input_c_tests): Call it.
3953 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3954 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3955 * selftest.h (ASSERT_GT): New macro.
3956 (ASSERT_GT_AT): New macro.
3957 (ASSERT_LT): New macro.
3958 (ASSERT_LT_AT): New macro.
3960 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3962 PR rtl-optimization/84780
3963 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3965 2018-03-14 Martin Liska <mliska@suse.cz>
3967 * tree.c (record_node_allocation_statistics): Use
3968 get_stats_node_kind.
3969 (get_stats_node_kind): New function extracted from
3970 record_node_allocation_statistics.
3971 (free_node): Use get_stats_node_kind.
3973 2018-03-14 Richard Biener <rguenther@suse.de>
3975 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3976 that the value-set of ANTIC_IN doesn't grow.
3979 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3981 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3982 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3984 2018-03-14 Julia Koval <julia.koval@intel.com>
3986 * config.gcc (icelake-client, icelake-server): New.
3988 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3989 (initial_ix86_arch_features): Ditto.
3990 (PTA_SKYLAKE): Add SGX.
3991 (PTA_ICELAKE): Remove.
3992 (PTA_ICELAKE_CLIENT): New.
3993 (PTA_ICELAKE_SERVER): New.
3994 (ix86_option_override_internal): Split up icelake on icelake client and
3996 (get_builtin_code_for_version): Ditto.
3997 (fold_builtin_cpu): Ditto.
3998 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3999 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4000 * config/i386/i386.h (processor_type): Ditto.
4001 * doc/invoke.texi: Ditto.
4003 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4006 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4007 INTEGER_CST offset, add it together with bitpos / 8 and
4008 sign extend based on POINTER_SIZE.
4012 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4015 * config/i386/constraints.md (Yc): New register constraint.
4016 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4017 Use Yc constraint for alternative 2 of operand 0. Remove
4018 preferred_for_speed attribute.
4020 2018-03-14 Richard Biener <rguenther@suse.de>
4022 PR tree-optimization/84830
4023 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4024 with the old one to avoid oscillations.
4026 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4029 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4031 (assign_by_spills): Return a flag of reload assignment failure.
4032 Do not process the reload assignment failures. Do not spill other
4033 reload pseudos if they has the same reg class. Update n if
4035 (lra_assign): Add a return arg. Set up from the result of
4036 assign_by_spills call.
4037 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4038 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4039 usage_insns if it is not NULL.
4040 (spill_hard_reg_in_range): New function.
4041 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4042 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4043 function prototypes.
4044 (lra_assign): Change prototype.
4045 * lra.c (lra): Add code to deal with fails by splitting hard reg
4048 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4050 * config/riscv/riscv.opt (mrelax): New option.
4051 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4052 "norelax" when riscv_mrelax is disabled.
4053 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4055 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4058 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4059 reassociation for int modes.
4061 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4063 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4064 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4066 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4067 * config/aarch64/aarch64-sve.md
4068 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4069 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4070 (*extend<mode><Vwide>2): Rename to...
4071 (aarch64_sve_extend<mode><Vwide>2): ...this.
4072 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4073 renaming the old pattern to...
4074 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4076 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4077 define_expand, renaming the old pattern to...
4078 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4079 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4080 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4081 account when deciding which SVE instruction the optab should use.
4082 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4084 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4086 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4087 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4088 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4089 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4090 (tlsdesc_small_<mode>): Turn a define_expand and use
4091 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4092 (tlsdesc_small_advsimd_<mode>): ...this.
4093 (tlsdesc_small_sve_<mode>): New pattern.
4095 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4097 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4098 (UNSPEC_UMUL_HIGHPART): New constants.
4099 (MUL_HIGHPART): New int iteraor.
4100 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4101 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4103 (*<su>mul<mode>3_highpart): New define_insn.
4105 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4108 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4111 2018-03-13 Martin Liska <mliska@suse.cz>
4114 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4116 (sem_item_optimizer::~sem_item_optimizer): Release it.
4117 (sem_item_optimizer::merge_classes): Register variable aliases.
4118 (sem_item_optimizer::fixup_pt_set): New function.
4119 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4120 * ipa-icf.h: Declare new variables and functions.
4122 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4125 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4126 integer_pow2p@2 and test integer_pow2p in condition.
4127 (A < 0 ? C : 0): Similarly for @1.
4130 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4131 characters starting at p contain '\0' character, don't look beyond
4135 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4136 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4139 * reg-stack.c (change_stack): Change update_end var from int to
4140 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4141 also call set_block_for_insn on the newly added insns and rescan.
4144 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4145 on the last operand.
4148 * tree.c (stabilize_reference_1): Return save_expr (e) for
4149 STATEMENT_LIST even if it doesn't have side-effects.
4151 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4153 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4155 2018-03-12 Renlin Li <renlin.li@arm.com>
4157 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4158 aarch64_output_scalar_simd_mov_immediate.
4160 2018-03-12 Martin Sebor <msebor@redhat.com>
4162 PR tree-optimization/83456
4163 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4164 for perfectly overlapping calls to memcpy.
4165 (gimple_fold_builtin_memory_chk): Same.
4166 (gimple_fold_builtin_strcpy): Handle no-warning.
4167 (gimple_fold_builtin_stxcpy_chk): Same.
4168 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4170 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4172 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4173 parameter. Use it for SFmode.
4174 (rs6000_function_arg_advance_1): Adjust.
4175 (rs6000_function_arg): Adjust.
4176 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4178 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4180 PR rtl-optimization/84169
4181 PR rtl-optimization/84780
4182 * combine.c (can_combine_p): Check for a 2-insn combination whether
4183 the destination register is used between the two insns, too.
4185 2018-03-12 Richard Biener <rguenther@suse.de>
4187 PR tree-optimization/84803
4188 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4189 for refs DR analysis didn't process.
4191 2018-03-12 Richard Biener <rguenther@suse.de>
4193 PR tree-optimization/84777
4194 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4195 force-vectorize loops ignore whether we are optimizing for size.
4197 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4199 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4200 (TARGET_MD_ASM_ADJUST): Define.
4202 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4203 Kito Cheng <kito.cheng@gmail.com>
4204 Chung-Ju Wu <jasonwucj@gmail.com>
4206 * config/nds32/nds32.c (nds32_compute_stack_frame,
4207 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4208 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4209 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4210 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4211 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4212 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4213 * config/nds32/nds32.md (prologue, epilogue): Use macro
4214 NDS32_V3PUSH_AVAILABLE_P to do checking.
4216 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4219 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4220 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4221 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4222 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4223 addition of most attributes on !orig_type_die or the attribute not
4224 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4226 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4227 Chung-Ju Wu <jasonwucj@gmail.com>
4229 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4231 * config/nds32/nds32.opt (mvh): New option.
4233 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4234 Chung-Ju Wu <jasonwucj@gmail.com>
4236 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4238 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4239 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4242 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4243 Chung-Ju Wu <jasonwucj@gmail.com>
4245 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4247 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4248 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4250 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4251 Kito Cheng <kito.cheng@gmail.com>
4252 Chung-Ju Wu <jasonwucj@gmail.com>
4254 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4255 UNSPEC_FFMISM and UNSPEC_FLMISM.
4256 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4257 for ffb, ffmism and flmism.
4258 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4259 (unspec_ffmism): Ditto.
4260 (unspec_flmism): Ditto.
4261 (nds32_expand_builtin_impl): Check if string extension is available.
4262 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4263 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4265 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4268 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4271 * lra-assigns.c (assign_by_spills): Return a flag of reload
4272 assignment failure. Do not process the reload assignment
4273 failures. Do not spill other reload pseudos if they has the same
4275 (lra_assign): Add a return arg. Set up from the result of
4276 assign_by_spills call.
4277 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4278 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4279 usage_insns if it is not NULL.
4280 (spill_hard_reg_in_range): New function.
4281 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4282 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4283 function prototypes.
4284 (lra_assign): Change prototype.
4285 * lra.c (lra): Add code to deal with fails by splitting hard reg
4288 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4291 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4293 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4296 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4297 (dw_val_node): Add val_symbolic_view.
4298 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4299 (symview_upper_bound): New.
4300 (new_line_info_table): Initialize symviews_since_reset.
4301 (dwarf2out_source_line): Count symviews_since_reset and set
4302 symview_upper_bound.
4303 (dw_val_equal_p): Handle symview.
4304 (add_AT_symview): New.
4305 (print_dw_val): Handle symview.
4306 (attr_checksum, attr_checksum_ordered): Likewise.
4307 (same_dw_val_p, size_of_die): Likewise.
4308 (value_format, output_die): Likewise.
4309 (add_high_low_attributes): Use add_AT_symview for entry_view.
4310 (dwarf2out_finish): Reset symview_upper_bound, clear
4313 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4316 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4317 Add strict argument and use it.
4318 (rs6000_split_multireg_move): Update for new strict argument.
4319 (mem_operand_gpr): Disallow all non-offsettable addresses.
4320 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4322 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4325 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4326 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4327 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4330 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4331 decl, use remap_type if we want to use the type.
4333 2018-03-09 Martin Sebor <msebor@redhat.com>
4335 PR tree-optimization/84526
4336 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4338 (builtin_access::generic_overlap): Be prepared to handle non-array
4341 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4343 PR rtl-optimization/84682
4344 * lra-constraints.c (process_address_1): Check is_address flag
4345 for address constraints.
4346 (process_alt_operands): Likewise.
4347 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4348 preprocess_constraints.
4349 * recog.h (preprocess_constraints): Add oploc parameter.
4351 * recog.c (preprocess_constraints): Test address_operand for
4352 CT_ADDRESS constraints.
4354 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4357 * lra-assigns.c (assign_by_spills): Return a flag of reload
4358 assignment failure. Do not process the reload assignment
4359 failures. Do not spill other reload pseudos if they has the same
4361 (lra_assign): Add a return arg. Set up from the result of
4362 assign_by_spills call.
4363 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4364 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4365 usage_insns if it is not NULL.
4366 (spill_hard_reg_in_range): New function.
4367 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4368 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4369 function prototypes.
4370 (lra_assign): Change prototype.
4371 * lra.c (lra): Add code to deal with fails by splitting hard reg
4374 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4377 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4378 Accept complain bool parameter. Only emit errors if it is true.
4379 (arm_parse_cpu_option_name): Likewise.
4380 (arm_target_thumb_only): Adjust callers of the above.
4381 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4382 prototype to take a default true bool parameter.
4383 (arm_parse_arch_option_name): Likewise.
4385 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4386 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4390 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4391 * configure: Regenerate.
4392 * configure.ac ("linker --version-script option"): New.
4393 ("linker soname option"): New.
4395 2018-03-09 Richard Biener <rguenther@suse.de>
4397 PR tree-optimization/84775
4398 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4399 immediate uses of predicate stmts and mark them modified.
4402 PR tree-optimization/84178
4403 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4405 (version_loop_for_if_conversion): Delay update_ssa call.
4406 (tree_if_conversion): Delay update_ssa until after predicate
4409 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4412 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4413 when the function accesses prior frames.
4415 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4418 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4419 gen_llsym, otherwise call maybe_gen_llsym.
4422 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4423 has ',' character inside of it.
4425 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4428 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4429 as clobbering CC_REGNUM.
4431 2018-03-08 Richard Biener <rguenther@suse.de>
4434 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4435 (follow_copies_to_constant): Do not follow SSA names registered
4438 2018-03-08 Richard Biener <rguenther@suse.de>
4440 PR tree-optimization/84178
4441 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4443 (version_loop_for_if_conversion): Delay update_ssa call.
4444 (tree_if_conversion): Delay update_ssa until after predicate
4447 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4449 PR tree-optimization/84178
4450 * tree-if-conv.c (release_bb_predicate): Remove the
4451 the assertion that the stmts have NULL use_ops.
4452 Discard the statements, asserting that they haven't
4453 yet been added to a BB.
4455 2018-03-08 Richard Biener <rguenther@suse.de>
4457 PR tree-optimization/84746
4458 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4459 (phi_translate): Pass in destination ANTIC_OUT set.
4460 (phi_translate_1): Likewise. For a simplified result lookup
4461 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4462 (phi_translate_set): Adjust.
4463 (do_pre_regular_insertion): Likewise.
4464 (do_pre_partial_partial_insertion): Likewise.
4466 2018-03-08 Martin Liska <mliska@suse.cz>
4468 PR gcov-profile/84735
4469 * doc/gcov.texi: Document usage of profile files.
4470 * gcov-io.h: Document changes in the format.
4472 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4476 * dwarf2out.c (struct dw_line_info_table): Update comments for
4478 (FORCE_RESET_NEXT_VIEW): New.
4479 (FORCE_RESETTING_VIEW_P): New.
4480 (RESETTING_VIEW_P): Check for -1 too.
4481 (ZERO_VIEW_P): Likewise.
4482 (new_line_info_table): Force-reset next view.
4483 (dwarf2out_begin_function): Likewise.
4484 (dwarf2out_source_line): Simplify zero_view_p initialization.
4485 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4486 view directly. Omit view when omitting .loc at line 0.
4488 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4490 PR tree-optimization/84740
4491 * tree-switch-conversion.c (process_switch): Call build_constructors
4492 only if info.phi_count is non-zero.
4494 PR tree-optimization/84739
4495 * tree-tailcall.c (find_tail_calls): Check call arguments against
4496 DECL_ARGUMENTS (current_function_decl) rather than
4497 DECL_ARGUMENTS (func) when checking for tail recursion.
4499 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4501 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4502 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4503 Volker Reichelt's entry and add entries for people that perform
4504 GCC fuzzy testing and report numerous bugs.
4506 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4509 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4510 readonly data in sdata, if that is disabled.
4511 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4512 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4513 -mreadonly-in-sdata option.
4515 2018-03-07 Martin Sebor <msebor@redhat.com>
4517 PR tree-optimization/84468
4518 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4519 basic block when looking for nul assignment.
4521 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4524 * except.h (output_function_exception_table): Adjust prototype.
4525 * except.c (output_function_exception_table): Remove FNNAME parameter
4526 and add SECTION parameter. Ouput one part of the table at a time.
4527 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4528 the first part of the exception table and emit unwind directives.
4529 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4530 (i386_pe_seh_cold_init): Likewise.
4531 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4532 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4533 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4534 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4535 * config/i386/winnt.c: Include except.h.
4536 (struct seh_frame_state): Add reg_offset, after_prologue and
4537 in_cold_section fields.
4538 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4539 (i386_pe_seh_cold_init): New function.
4540 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4541 to seh->in_cold_section.
4542 (seh_emit_push): Record the offset of the push.
4543 (seh_emit_save): Record the offet of the save.
4544 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4545 Test seh->after_prologue to disregard the epilogue.
4546 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4547 (i386_pe_end_cold_function): New function.
4549 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4552 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4553 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4556 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4558 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4559 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4562 * multiple_target.c: Include tree-inline.h and intl.h.
4563 (expand_target_clones): Diagnose and fail if node->definition and
4564 !tree_versionable_function_p (node->decl).
4566 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4568 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4570 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4571 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4572 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4574 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4577 * combine.c (try_combine): Use reg_or_subregno instead of handling
4578 just paradoxical SUBREGs and REGs.
4580 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4582 * config/arc/arc.c (arc_finalize_pic): Remove function.
4583 (arc_must_save_register): We use single base PIC register, remove
4584 checks to save/restore the PIC register.
4585 (arc_expand_prologue): Likewise.
4586 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4588 (arc_verify_short): Likewise.
4589 (arc_attr_type): Likewise.
4590 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4591 (walk_stores): Likewise.
4592 (arc_address_cost): Make it static.
4593 (arc_verify_short): Likewise.
4594 (branch_dest): Likewise.
4595 (arc_attr_type): Likewise.
4596 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4597 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4598 (arc_final_prescan_insn): Remove inserting the nops due to
4599 hardware hazards. It is done in reorg step.
4600 (insn_length_variant_t): Remove.
4601 (insn_length_parameters_t): Likewise.
4602 (arc_insn_length_parameters): Likewise.
4603 (arc_get_insn_variants): Likewise.
4604 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4606 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4609 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4612 PR tree-optimization/84687
4613 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4615 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4617 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4619 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4620 Rename to ppc_speculation_barrier.
4621 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4622 __builtin_ppc_speculation_barrier.
4624 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4627 * combine.c (combine_simplify_rtx): Don't try to simplify if
4628 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4631 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4633 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4634 to 32 bytes when compiling for POWER9.
4636 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4639 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4640 regparm >= 3 with no arg reg available also for calls with
4641 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4644 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4646 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4648 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4651 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4653 2018-03-05 Richard Biener <rguenther@suse.de>
4655 PR tree-optimization/84486
4656 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4657 When inserting a __builtin_assume_aligned call set the LHS
4658 SSA name alignment info accordingly.
4660 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4662 PR tree-optimization/84114
4663 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4664 Avoid reassociation of FLOAT_MODE addition.
4666 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4668 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4669 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4670 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4671 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4672 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4673 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4674 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4676 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4677 __builtin_ia32_wbinvd): New builtins.
4678 (SPECIAL_ARGS2): New.
4679 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4680 (SPECIAL_ARGS2): New.
4681 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4682 (ix86_valid_target_attribute_inner_p): Ditto.
4683 (ix86_init_mmx_sse_builtins): Add special_args2.
4684 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4685 TARGET_WBNOINVD_P): New.
4686 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4687 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4688 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4689 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4690 * config/i386/pconfigintrin.h: New file.
4691 * config/i386/wbnoinvdintrin.h: Ditto.
4692 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4693 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4695 2018-03-05 Richard Biener <rguenther@suse.de>
4697 PR tree-optimization/84670
4698 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4700 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4701 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4702 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4703 doesn't grow if all successors (recursively) were visited at least
4706 2018-03-05 Richard Biener <rguenther@suse.de>
4708 PR tree-optimization/84650
4709 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4710 if executed in the loop pipeline.
4712 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4714 * doc/configfiles.texi (Configuration Files): Move info about
4715 conditionalizing $target-protos.h to...
4716 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4717 differs from $target-protos.h.
4719 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4720 Chung-Ju Wu <jasonwucj@gmail.com>
4722 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4723 * config/nds32/nds32-multiple.md (setmemsi): Define.
4724 * config/nds32/nds32-memory-manipulation.c
4725 (nds32_gen_dup_4_byte_to_word_value): New.
4726 (emit_setmem_word_loop): New.
4727 (emit_setmem_byte_loop): New.
4728 (nds32_expand_setmem_loop): New.
4729 (nds32_expand_setmem_loop_v3m): New.
4730 (nds32_expand_setmem_unroll): New.
4731 (nds32_expand_setmem): New.
4733 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4734 Chung-Ju Wu <jasonwucj@gmail.com>
4736 * config/nds32/nds32-memory-manipulation.c
4737 (nds32_emit_load_store): New.
4738 (nds32_emit_post_inc_load_store): New.
4739 (nds32_emit_mem_move): New.
4740 (nds32_emit_mem_move_block): New.
4741 (nds32_expand_movmemsi_loop_unknown_size): New.
4742 (nds32_expand_movmemsi_loop_known_size): New.
4743 (nds32_expand_movmemsi_loop): New.
4744 (nds32_expand_movmemsi_unroll): New.
4745 (nds32_expand_movmemqi): Rename ...
4746 (nds32_expand_movmemsi): ... to this.
4747 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4748 (movmemsi): ... to this.
4749 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4750 (nds32_expand_movmemsi): ... to this.
4752 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4753 Monk Chiang <sh.chiang04@gmail.com>
4754 Chung-Ju Wu <jasonwucj@gmail.com>
4756 * config/nds32/nds32-protos.h
4757 (nds32_expand_load_multiple): New arguments.
4758 (nds32_expand_store_multiple): Ditto.
4759 (nds32_valid_multiple_load_store): Rename ...
4760 (nds32_valid_multiple_load_store_p): ... to this.
4761 * config/nds32/nds32-memory-manipulation.c
4762 (nds32_expand_load_multiple): Refine implementation.
4763 (nds32_expand_store_multiple): Ditto.
4764 * config/nds32/nds32-multiple.md
4765 (load_multiple): Update nds32_expand_load_multiple interface.
4766 (store_multiple): Update nds32_expand_store_multiple interface.
4767 * config/nds32/nds32-predicates.c
4768 (nds32_valid_multiple_load_store): Rename ...
4769 (nds32_valid_multiple_load_store_p): ... to this and refine
4771 * config/nds32/predicates.md
4772 (nds32_load_multiple_and_update_address_operation): New predicate.
4773 (nds32_store_multiple_and_update_address_operation): New predicate.
4775 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4776 Chung-Ju Wu <jasonwucj@gmail.com>
4778 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4779 (combo): New attribute.
4780 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4782 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4784 * config/nds32/nds32.opt: Change -mcmodel= default value.
4786 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4787 Monk Chiang <sh.chiang04@gmail.com>
4788 Chung-Ju Wu <jasonwucj@gmail.com>
4790 * config/nds32/constants.md (unspec_element): New enum.
4791 * config/nds32/constraints.md (Umw): New constraint.
4792 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4793 * config/nds32/nds32-intrinsic.md: Likewise.
4794 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4795 (nds32_valid_smw_lwm_base_p): New.
4796 (nds32_output_smw_single_word): New.
4797 (nds32_output_lmw_single_word): New.
4798 (nds32_expand_unaligned_load): New.
4799 (nds32_expand_unaligned_store): New.
4800 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4801 (nds32_output_smw_single_word): Declare.
4802 (nds32_output_lmw_single_word): Declare.
4803 (nds32_expand_unaligned_load): Declare.
4804 (nds32_expand_unaligned_store): Declare.
4805 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4806 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4807 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4808 NDS32_BUILTIN_UASTORE_DW.
4809 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4812 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4813 Kito Cheng <kito.cheng@gmail.com>
4814 Chung-Ju Wu <jasonwucj@gmail.com>
4816 * config/nds32/nds32-intrinsic.c
4817 (nds32_expand_builtin_null_ftype_reg): Delete.
4818 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4819 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4820 (nds32_read_argument): New.
4821 (nds32_legitimize_target): Ditto.
4822 (nds32_legitimize_argument): Ditto.
4823 (nds32_check_constant_argument): Ditto.
4824 (nds32_expand_unop_builtin): Ditto.
4825 (nds32_expand_unopimm_builtin): Ditto.
4826 (nds32_expand_binop_builtin): Ditto.
4827 (nds32_builtin_decl_impl): Ditto.
4828 (builtin_description): Ditto.
4829 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4830 (nds32_init_builtins_impl): Ditto.
4831 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4832 (nds32_builtin_decl): New.
4833 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4834 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4836 2018-03-02 Jeff Law <law@redhat.com>
4838 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4839 (redundant_insn, fill_simple_delay_slots): Likewise.
4840 (fill_slots_from_thread): Likewise.
4841 * resource.c (mark_referenced_resources): Likewise.
4842 (mark_set_resources, find_dead_or_set_registers): Likewise.
4844 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4846 * substring-locations.h (format_warning_va): Formatting fix for
4848 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4850 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4851 * substring-locations.c: Include intl.h.
4852 (format_warning_va): Turned into small wrapper around
4853 format_warning_n_va, renamed to ...
4854 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4855 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4857 (format_warning_at_substring_n): New function.
4858 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4859 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4860 format_warning_at_substring with just a shorter name instead of
4861 const function pointer.
4862 (fmtwarn_n): New function.
4863 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4864 appropriate, get rid of all the fmtstr temporaries, move conditionals
4865 with G_() wrapped string literals directly into fmtwarn arguments,
4866 cast dir.len to (int), formatting fixes.
4868 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4870 * doc/invoke.texi: Remove "Cilk Plus" references.
4872 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4873 Richard Biener <rguenther@suse.de>
4876 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4877 for error or warning attributes if CALL_FROM_THUNK_P is set.
4880 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4883 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4884 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4887 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4888 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4890 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4891 instead of -1U in last predictors element's probability member.
4893 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4896 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4897 arguments if they are comparable.
4899 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4901 PR tree-optimization/84634
4902 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4903 masks and masked_loop_p with a single loop_masks, making sure it's
4904 null for bb vectorization.
4906 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4908 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4909 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4910 loop->force_vectorize to check whether there is no alias.
4912 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4915 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4917 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4919 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4920 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4921 instead of a loop around prev_real_insn.
4922 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4926 * config/i386/i386.c (ix86_print_operand): Use conditional
4927 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4930 2018-03-02 Richard Biener <rguenther@suse.de>
4932 PR tree-optimization/84427
4933 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4934 (bitmap_set_subtract_values): Rewrite to handle multiple
4937 (prune_clobbered_mems): Likewise.
4938 (phi_translate): Take edge instead of pred/phiblock.
4939 (phi_translate_1): Likewise.
4940 (phi_translate_set): Likewise. Insert all translated
4941 exprs for a value into the set, keeping possibly multiple
4942 expressions per value.
4943 (compute_antic_aux): Adjust for phi_translate changes.
4944 When intersecting union the expressions and prune those
4945 not in the final value set, keeping possibly multiple
4946 expressions per value. Do not use value-insertion
4947 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4948 all expressions. Add verification that the value-sets
4949 only shrink during iteration.
4950 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4951 (do_pre_regular_insertion): Likewise.
4952 (do_pre_partial_partial_insertion): Likewise.
4954 2018-03-02 Richard Biener <rguenther@suse.de>
4957 * config/darwin.c (saved_debug_info_level): New static global.
4958 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4959 (darwin_asm_lto_end): Restore debug info generation settings.
4961 2018-03-01 Martin Liska <mliska@suse.cz>
4964 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4967 2018-03-01 Richard Biener <rguenther@suse.de>
4970 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4971 type annotation with LTO when debug was disabled at compile-time.
4973 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4975 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4977 (mips_final_postscan_insn): Likewise.
4979 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4981 PR rtl-optimization/84528
4982 * alias.c (init_alias_target): Add commentary.
4983 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4984 a unique base value if the frame pointer is not eliminated
4985 to the stack pointer.
4987 2018-03-01 Tom de Vries <tom@codesourcery.com>
4989 PR rtl-optimization/83327
4990 * lra-int.h (hard_regs_spilled_into): Declare.
4991 * lra.c (hard_regs_spilled_into): Define.
4992 (init_reg_info): Init hard_regs_spilled_into.
4993 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4994 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4995 (process_bb_lives): Handle hard_regs_spilled_into.
4996 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4997 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4999 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5001 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5002 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5003 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5004 * config/rs6000/aix72.h: New file.
5006 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5008 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5009 instead of warning_at with conditional singular and plural messages
5013 * stor-layout.c (update_alignment_for_field): For
5014 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5015 && !DECL_PACKED (field), do the alignment update, just use
5016 only desired_align instead of MAX (type_align, desired_align)
5018 (place_field): Don't do known_align < desired_align handling
5019 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5020 is non-NULL, instead do it after rli->prev_field handling and
5021 only if not within a bitfield word. For DECL_PACKED (field)
5022 use type_align of BITS_PER_UNIT.
5024 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5026 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5027 superfluous parentheses and trailing spaces.
5029 2018-02-28 Richard Biener <rguenther@suse.de>
5031 PR tree-optimization/84584
5032 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5033 SCoPs with fake exit edge.
5035 2018-02-28 Martin Liska <mliska@suse.cz>
5038 * timevar.c (timer::print): Fix format to properly print 100%
5041 2018-02-28 Richard Biener <rguenther@suse.de>
5044 * genmatch.c (capture_info::walk_match): Do not mark
5045 captured expressions without operands as expr_p given
5046 they act more like predicates and should be subject to
5047 "lost tail" side-effect preserving.
5049 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5051 PR rtl-optimization/81611
5052 * auto-inc-dec.c (attempt_change): Move dead note from
5053 mem_insn if it's the next use of regno
5054 (find_address): Take address use of reg holding
5055 non-incremented value. Add parm to limit search to the named
5057 (merge_in_block): Attempt to use a mem insn that is the next
5058 use of the original regno.
5060 2018-02-27 Martin Sebor <msebor@redhat.com>
5063 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5064 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5066 2018-02-27 Martin Sebor <msebor@redhat.com>
5068 PR translation/84207
5069 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5070 n argument to unsigned HOST_WIDE_INT.
5071 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5072 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5073 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5074 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5076 2018-02-27 Richard Biener <rguenther@suse.de>
5078 PR tree-optimization/84512
5079 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5080 Do not use the estimate returned from record_stmt_cost for
5081 the scalar iteration cost but sum properly using add_stmt_cost.
5083 2018-02-27 Richard Biener <rguenther@suse.de>
5085 PR tree-optimization/84466
5086 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5087 Adjust last change to less strictly validate use operands.
5089 2018-02-27 Martin Liska <mliska@suse.cz>
5091 PR gcov-profile/84548
5092 * gcov.c (process_file): Allow partial overlap and consider it
5093 also as group functions.
5094 (output_lines): Properly calculate range of lines for a group.
5096 2018-02-27 Martin Liska <mliska@suse.cz>
5098 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5099 'ggc' suffixes. Change first column width.
5100 (timer::print): Fix formatting of the column.
5102 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5104 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5105 preserve inline entry blocks for the sake of debug inline
5106 entry point markers alone.
5107 (remove_unused_locals): Suggest in comments a better place to
5108 force the preservation of inline entry blocks that are
5109 otherwise unused, but do not preserve them.
5111 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5113 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5115 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5118 * config/i386/constraints.md (Bs): Replace
5119 ix86_indirect_branch_register with
5120 TARGET_INDIRECT_BRANCH_REGISTER.
5122 * config/i386/i386.md (indirect_jump): Likewise.
5123 (tablejump): Likewise.
5124 (*sibcall_memory): Likewise.
5125 (*sibcall_value_memory): Likewise.
5126 Peepholes of indirect call and jump via memory: Likewise.
5127 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5128 (*sibcall_value_GOT_32): Likewise.
5129 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5130 (GOT_memory_operand): Likewise.
5131 (call_insn_operand): Likewise.
5132 (sibcall_insn_operand): Likewise.
5133 (GOT32_symbol_operand): Likewise.
5134 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5136 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5138 PR rtl-optimization/83496
5139 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5140 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5141 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5142 redundant insn, if any.
5143 (relax_delay_slots): Likewise.
5144 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5146 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5148 PR tree-optimization/83965
5149 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5150 that grouped statements are part of a reduction chain. Return
5151 true if the statement is not marked as a reduction itself but
5153 (vect_recog_dot_prod_pattern): Don't check whether the statement
5154 is part of a group here.
5155 (vect_recog_sad_pattern): Likewise.
5156 (vect_recog_widen_sum_pattern): Likewise.
5158 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5161 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5163 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5166 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5168 (ix86_output_indirect_function_return): New prototype.
5169 (ix86_split_simple_return_pop_internal): Likewise.
5170 * config/i386/i386.c (indirect_return_via_cx): New.
5171 (indirect_return_via_cx_bnd): Likewise.
5172 (indirect_thunk_name): Handle return va CX_REG.
5173 (output_indirect_thunk_function): Create alias for
5174 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5175 (ix86_output_indirect_jmp): Remove the bool argument.
5176 (ix86_output_indirect_function_return): New function.
5177 (ix86_split_simple_return_pop_internal): Likewise.
5178 * config/i386/i386.md (*indirect_jump): Don't pass false
5179 to ix86_output_indirect_jmp.
5180 (*tablejump_1): Likewise.
5181 (simple_return_pop_internal): Change it to define_insn_and_split.
5182 Call ix86_split_simple_return_pop_internal to split it for
5184 (simple_return_indirect_internal): Call
5185 ix86_output_indirect_function_return instead of
5186 ix86_output_indirect_jmp.
5188 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5191 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5192 memset and value initialization afterwards.
5194 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5196 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5198 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5201 * common/config/aarch64/aarch64-common.c
5202 (aarch_option_optimization_table[]): Switch
5203 off fomit-frame-pointer
5205 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5206 Chung-Ju Wu <jasonwucj@gmail.com>
5208 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5210 (store_multiple): Ditto.
5212 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5214 * config.gcc: Add --with-cpu support for nds32 target.
5215 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5216 * config/nds32/nds32.opt: Add -mcpu= option.
5218 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5220 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5221 isel=yes): Warn for these deprecated options.
5223 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5225 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5226 ISA_2_5_MASKS_EMBEDDED.
5228 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5230 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5231 p->max as pointers rather than using iterative_hash_expr.
5233 2018-02-23 Carl Love <cel@us.ibm.com>
5235 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5236 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5238 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5239 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5240 P8V_BUILTIN_VEC_VUNSIGNED2.
5242 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5245 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5246 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5248 * lra-constraints.c (curr_insn_transform): Set up
5249 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5251 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5252 * lra-eliminations.c (spill_pseudos): Ditto.
5253 (process_insn_for_elimination): Ditto.
5254 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5255 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5256 LRA_NON_CLOBBERED_ALT.
5258 2018-02-22 Martin Sebor <msebor@redhat.com>
5260 PR tree-optimization/84480
5261 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5262 to maybe_diag_stxncpy_trunc. Call it.
5263 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5264 from gimple_fold_builtin_strcpy. Print inlining stack.
5265 (handle_builtin_stxncpy): Print inlining stack.
5266 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5268 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5271 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5272 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5273 and -fcheck-pointer-bounds are used together.
5274 (indirect_thunk_prefix): New enum.
5275 (indirect_thunk_need_prefix): New function.
5276 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5277 "_nt" instead of "_bnd" for NOTRACK prefix.
5278 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5279 (output_indirect_thunk_function): Likewise.
5281 (ix86_code_end): Update output_indirect_thunk_function calls.
5282 (ix86_output_indirect_branch_via_reg): Replace
5283 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5284 (ix86_output_indirect_branch_via_push): Likewise.
5285 (ix86_output_function_return): Likewise.
5286 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5287 incompatible with -fcf-protection=branch and
5288 -fcheck-pointer-bounds.
5290 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5293 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5294 Change gcc_assert call to output_operand_lossage.
5296 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5298 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5300 2018-02-22 DJ Delorie <dj@redhat.com>
5301 Sebastian Perta <sebastian.perta@renesas.com>
5302 Oleg Endo <olegendo@gcc.gnu.org>
5304 * config/rx/rx.c (rx_rtx_costs): New function.
5305 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5307 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5309 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5311 2018-02-22 Martin Liska <mliska@suse.cz>
5314 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5315 Add "native" as a possible value.
5317 2018-02-22 Martin Liska <mliska@suse.cz>
5320 * config/i386/i386.c (ix86_option_override_internal):
5321 Add "native" as a possible value for -march and -mtune.
5323 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5326 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5327 to all type variants.
5329 PR tree-optimization/84503
5330 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5331 width as info->bitpos + info->bitsize - start.
5332 (merged_store_group::merge_overlapping): Simplify width computation.
5333 (check_no_overlap): New function.
5334 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5335 start + width and last_order of the group, fail if check_no_overlap
5337 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5338 to group if check_no_overlap fails.
5340 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5342 * config/rs6000/altivec.md: Delete contraint arguments to
5343 define_expand, define_split, and define_peephole2, and in
5344 define_insn_and_split if always unused.
5345 * config/rs6000/darwin.md: Ditto.
5346 * config/rs6000/dfp.md: Ditto.
5347 * config/rs6000/rs6000.md: Ditto.
5348 * config/rs6000/sync.md: Ditto.
5349 * config/rs6000/vector.md: Ditto.
5350 * config/rs6000/vsx.md: Ditto.
5352 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5354 * config/rs6000/altivec.md: Write output control strings as braced
5355 blocks instead of double-quoted strings.
5356 * config/rs6000/darwin.md: Ditto.
5357 * config/rs6000/rs6000.md: Ditto.
5358 * config/rs6000/vector.md: Ditto.
5359 * config/rs6000/vsx.md: Ditto.
5361 2018-02-21 Jason Merrill <jason@redhat.com>
5363 PR c++/84314 - ICE with templates and fastcall attribute.
5364 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5366 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5368 * ipa-cp.c (determine_versionability): Fix comment typos.
5370 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5373 * ipa-cp.c (determine_versionability): Do not version functions caling
5376 2018-02-21 Martin Liska <mliska@suse.cz>
5379 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5380 Add "native" as a possible value.
5381 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5382 the macro when native cpu detection is available.
5384 2018-02-21 Martin Liska <mliska@suse.cz>
5387 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5388 Add "native" as a possible value.
5389 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5390 when native cpu detection is available.
5392 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5393 Martin Sebor <msebor@redhat.com>
5395 PR tree-optimization/84478
5396 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5398 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5399 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5400 support which is conservatively correct, for 2 only stay conservative
5401 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5402 argument to the 2 argument get_range_strlen, adjust 6 arg
5403 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5405 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5406 (gimple_fold_builtin_strlen): Pass true as last argument to
5409 2018-02-20 Martin Sebor <msebor@redhat.com>
5412 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5413 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5414 (builtin_memref::builtin_memref): Factor out parts into
5415 set_base_and_offset and call it.
5417 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5420 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5421 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5422 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5423 search at the associated MODE_INT.
5425 2018-02-20 Jeff Law <law@redhat.com>
5428 PR tree-optimization/81592
5430 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5431 for range data rather than using global data.
5432 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5433 range data rather than using global data.
5434 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5435 pass it to children as needed.
5436 (struct directive::fmtresult): Similarly.
5437 (struct directive::set_width): Similarly.
5438 (struct directive::set_precision): Similarly.
5439 (format_integer, format_directive, parse_directive): Similarly.
5440 (format_none): Accept unnamed vr_values parameter.
5441 (format_percent, format_floating, format_character): Similarly.
5442 (format_string, format_plain): Similarly.
5443 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5444 the EVRP range analyzer for range data rather than using global data.
5445 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5446 gimple-ssa-evrp-analyze.h
5447 (class sprintf_dom_walker): Add after_dom_children member function.
5448 Add evrp_range_analyzer member.
5449 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5450 range analyzer as needed.
5451 (sprintf_dom_walker::after_dom_children): New member function.
5452 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5454 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5455 (evrp_range_analyzer::pop_to_marker): Likewise.
5457 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5459 PR tree-optimization/84419
5460 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5461 with the required type if its current type is compatible but
5464 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5467 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5468 after vectorization.
5470 2018-02-20 Martin Liska <mliska@suse.cz>
5473 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5474 possible values if we don't have a hint.
5476 2018-02-20 Martin Liska <mliska@suse.cz>
5480 * final.c (shorten_branches): Build align_tab array with one
5482 * opts.c (finish_options): Add alignment option limit check.
5483 (MAX_CODE_ALIGN): Likewise.
5484 (MAX_CODE_ALIGN_VALUE): Likewise.
5485 * doc/invoke.texi: Document maximum allowed option value for
5486 all -falign-* options.
5488 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5491 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5492 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5493 * var-tracking.c (emit_note_insn_var_location): Remove all references
5494 to NOTE_INSN_CALL_ARG_LOCATION.
5495 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5496 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5497 Use copy_rtx_if_shared.
5498 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5499 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5500 (dwarf2out_var_location): Remove handling of
5501 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5503 * final.c (final_scan_insn): Remove all references to
5504 NOTE_INSN_CALL_ARG_LOCATION.
5505 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5506 before dumping final insns.
5507 * except.c (emit_note_eh_region_end): Remove all references to
5508 NOTE_INSN_CALL_ARG_LOCATION.
5509 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5510 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5511 * config/arc/arc.c (hwloop_optimize): Likewise.
5512 * config/arm/arm.c (create_fix_barrier): Likewise.
5513 * config/s390/s390.c (s390_chunkify_start): Likewise.
5514 * config/sh/sh.c (find_barrier): Likewise.
5515 * config/i386/i386.c (rest_of_insert_endbranch,
5516 ix86_seh_fixup_eh_fallthru): Likewise.
5517 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5518 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5519 * config/frv/frv.c (frv_function_prologue): Likewise.
5520 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5522 (note_outside_basic_block_p): Remove all references to
5523 NOTE_INSN_CALL_ARG_LOCATION.
5524 * gengtype.c (adjust_field_rtx_def): Likewise.
5525 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5527 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5528 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5531 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5534 PR tree-optimization/84452
5535 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5536 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5539 2018-02-19 Martin Liska <mliska@suse.cz>
5542 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5544 2018-02-19 Martin Liska <mliska@suse.cz>
5545 Richard Sandiford <richard.sandiford@linaro.org>
5547 PR tree-optimization/82491
5548 * gimple-fold.c (get_base_constructor): Make earlier bail out
5551 2018-02-19 Carl Love <cel@us.ibm.com>
5553 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5554 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5556 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5557 P8V_BUILTIN_VEC_NEG.
5559 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5561 * config/rl78/rl78.md (movdf): New define expand.
5563 2018-02-19 Martin Liska <mliska@suse.cz>
5566 * doc/invoke.texi: Fix typo.
5567 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5569 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5571 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5572 handle rs6000_single_float and rs6000_double_float specially for
5575 2018-02-16 Jeff Law <law@redhat.com>
5577 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5580 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5583 * ipa-inline.c (inline_small_functions): Fix a typo.
5585 2018-02-16 Nathan Sidwell <nathan@acm.org>
5587 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5589 2018-02-16 Carl Love <cel@us.ibm.com>
5591 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5592 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5593 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5594 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5595 expansion to P8V_BUILTIN_VEC_FLOAT2.
5597 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5599 PR rtl-optimization/70023
5600 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5601 src_regno into account.
5603 2018-02-16 Carl Love <cel@us.ibm.com>
5605 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5606 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5607 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5608 * config/rs6000/rs6000.c: Remove case statements for
5609 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5610 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5611 and P9V_BUILTIN_VEC_VINSERT4B.
5612 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5613 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5614 * config/rs6000/vsx.md:
5615 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5618 2018-02-16 Carl Love <cel@us.ibm.com>
5620 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5622 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5624 * config/rs6000/rs6000-c.c: Add the definitions for
5625 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5626 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5627 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5628 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5629 definition for insert4b and define insn *insert3b_internal.
5630 * doc/extend.texi: Add documentation for vec_extract4b.
5632 2018-02-16 Nathan Sidwell <nathan@acm.org>
5634 * doc/extend.texi (Backwards Compatibility): Mention friend
5635 injection. Note for-scope is deprecated.
5636 * doc/invoke.texi (-ffriend-injection): Deprecate.
5638 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5640 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5641 that moved to I2, also allow destinations that are a paradoxical
5642 subreg (instead of a normal reg).
5644 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5647 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5650 2018-02-16 Richard Biener <rguenther@suse.de>
5652 PR tree-optimization/84037
5653 PR tree-optimization/84016
5655 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5656 Adjust vec_construct for the fact we need additional higher latency
5657 128bit inserts for AVX256 and AVX512 vector builds.
5658 (ix86_add_stmt_cost): Scale vector construction cost for
5661 2018-02-16 Richard Biener <rguenther@suse.de>
5663 PR tree-optimization/84417
5664 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5665 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5666 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5668 2018-02-16 Martin Liska <mliska@suse.cz>
5671 * internal-fn.def (ASAN_CHECK): Set proper flags.
5672 (ASAN_MARK): Likewise.
5674 2018-02-16 Julia Koval <julia.koval@intel.com>
5676 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5677 from PTA_CANNONLAKE.
5679 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5682 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5683 Use ++iter rather than iter++ for std::list iterators.
5684 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5685 defer deleting them until all nodes in the forest are processed. Do
5686 free even leaf nodes. Change to_process into auto_vec.
5689 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5690 * vec.h (vec_default_construct): Use memset instead of placement new
5691 if BROKEN_VALUE_INITIALIZATION is defined.
5692 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5693 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5696 PR rtl-optimization/83723
5697 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5698 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5699 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5701 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5703 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5705 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5707 PR rtl-optimization/81443
5708 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5709 from inner REGs to paradoxical SUBREGs.
5711 2018-02-16 Richard Biener <rguenther@suse.de>
5713 PR tree-optimization/84399
5714 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5715 For operands we can analyze at their definition make sure we can
5716 analyze them at each use as well.
5718 2018-02-16 Richard Biener <rguenther@suse.de>
5720 PR tree-optimization/84190
5721 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5722 volatile accesses if the decl isn't volatile.
5724 2018-02-15 Jason Merrill <jason@redhat.com>
5726 PR c++/84314 - ICE with templates and fastcall attribute.
5727 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5728 TYPE_CANONICAL on an existing type.
5730 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5732 PR tree-optimization/84383
5733 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5734 dstoff nor call operand_equal_p if dstbase is NULL.
5736 PR tree-optimization/84334
5737 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5738 also a CONSTANT_CLASS_P, punt.
5740 2018-02-14 Jim Wilson <jimw@sifive.com>
5742 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5743 first SMALL_OPERAND check. New local min_second_step. Move assert
5744 to where locals are set. Add TARGET_RVC support.
5745 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5747 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5749 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5751 2018-02-14 Martin Sebor <msebor@redhat.com>
5753 PR tree-optimization/83698
5754 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5755 arrays constrain the offset range to their bounds.
5756 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5757 (builtin_access::overlap): Avoid setting the size of overlap if it's
5759 (maybe_diag_overlap): Also consider arrays when deciding what values
5760 of offsets to include in diagnostics.
5762 2018-02-14 Martin Sebor <msebor@redhat.com>
5765 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5766 that correspond to the kind of a declaration.
5768 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5771 * config/pa/pa.md: Load address of PIC label using the linkage table
5772 if the label is nonlocal.
5774 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5776 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5777 warning message if user requests -maltivec=be.
5778 * doc/invoke.texi: Document deprecation of -maltivec=be.
5780 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5783 * config/rs6000/rs6000-c.c: Update definitions for
5784 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5785 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5787 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5790 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5791 add _get_ssp intrinsics. Remove argument from
5792 __builtin_ia32_rdssp[d|q].
5793 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5794 * config/i386/i386-builtin.def: Remove argument from
5795 __builtin_ia32_rdssp[d|q].
5796 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5797 ix86_expand_special_args_builtin for _rdssp[d|q].
5798 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5799 Clear register before usage.
5800 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5801 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5803 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5805 PR tree-optimization/84357
5806 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5807 operand 1 of an ARRAY_REF too.
5809 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5812 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5813 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5815 (set_of_reg): New struct.
5816 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5817 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5818 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5820 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5821 Split into bitclr, bitset, bitinvert patterns if appropriate.
5822 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5823 use rx_fuse_in_memory_bitop.
5824 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5825 to named insn, correct maximum insn length.
5827 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5830 * machmode.def: Define a complex mode for PARTIAL_INT.
5831 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5833 * doc/rtl.texi: Document CSPImode.
5834 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5836 (msp430_hard_regno_nregs_with_padding): Likewise.
5838 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5841 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5843 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5845 PR rtl-optimization/84169
5846 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5847 we generated a parallel as new i3 and we split that to new i2 and i3
5848 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5849 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5850 those to i2, not i1. Partially rewrite this scan code.
5852 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5855 * stor-layout.c (place_field): For variable length fields, adjust
5856 offset_align afterwards not just based on the field's alignment,
5857 but also on the size.
5860 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5861 of exps and logs in the use_exp2 case.
5863 2018-02-13 Jeff Law <law@redhat.com>
5865 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5868 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5871 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5875 * common.opt (gas-loc-support, gas-locview-support): New.
5876 (ginline-points, ginternal-reset-location-views): New.
5877 * doc/invoke.texi: Document them. Use @itemx where intended.
5878 (gvariable-location-views): Adjust.
5879 * target.def (reset_location_view): New.
5880 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5881 (TARGET_RESET_LOCATION_VIEW): New.
5882 * doc/tm.texi: Rebuilt.
5883 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5884 (dwarf2out_default_as_locview_support): New.
5885 (output_asm_line_debug_info): Use option variables.
5886 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5887 (output_loc_list): Likewise.
5888 (add_high_low_attributes): Check option variables.
5889 Don't output entry view attribute in strict mode.
5890 (gen_inlined_subroutine_die): Check option variables.
5891 (dwarf2out_inline_entry): Likewise.
5892 (init_sections_and_labels): Likewise.
5893 (dwarf2out_early_finish): Likewise.
5894 (maybe_reset_location_view): New, from...
5895 (dwarf2out_var_location): ... here. Call it.
5896 * debug.h (dwarf2out_default_as_loc_support): Declare.
5897 (dwarf2out_default_as_locview_support): Declare.
5898 * hooks.c (hook_int_rtx_insn_0): New.
5899 * hooks.h (hook_int_rtx_insn_0): Declare.
5900 * toplev.c (process_options): Take -gas-loc-support and
5901 -gas-locview-support from dwarf2out. Enable
5902 -gvariable-location-views by default only with locview
5903 assembler support. Enable -ginternal-reset-location-views by
5904 default only if the target defines the corresponding hook.
5905 Enable -ginline-points by default if location views are
5906 enabled; force it disabled if statement frontiers are
5908 * tree-inline.c (expand_call_inline): Check option variables.
5909 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5911 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5913 PR tree-optimization/84321
5914 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5915 handling. Also check whether the anti-range contains any values
5916 that satisfy the mask; switch to a VR_RANGE if not.
5918 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5921 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5923 2018-02-13 Martin Jambor <mjambor@suse.cz>
5926 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5927 of call statements, also set location of a load to a temporary.
5929 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5931 * config/rl78/rl78.c (add_vector_labels): New function.
5932 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5933 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5934 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5935 which checks that no arguments are passed.
5936 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5937 * doc/extend.texi: Documentation for the new attribute.
5939 2018-02-13 Andreas Schwab <schwab@suse.de>
5941 * config/riscv/linux.h (CPP_SPEC): Define.
5943 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5946 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5947 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5948 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5949 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5950 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5951 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5952 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5953 temporarily for AES and PCLMUL builtins.
5955 PR tree-optimization/84339
5956 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5957 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5961 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5962 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5963 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5965 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5967 * omp-simd-clone.h: New file.
5968 * omp-simd-clone.c: Include omp-simd-clone.h.
5969 (expand_simd_clones): No longer static.
5970 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5971 cgraph.h and omp-simd-clone.h.
5972 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5973 (vect_recog_widen_shift_pattern): Formatting fix.
5974 (vect_pattern_recog_1): Don't check optab for calls.
5977 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5978 operands[2] into a REG before using gen_lowpart on it.
5980 2018-02-12 Jeff Law <law@redhat.com>
5983 * config/sh/sh.c (find_barrier): Consider a sibling call
5986 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5987 successfully back substituting a reg.
5989 2018-02-12 Richard Biener <rguenther@suse.de>
5991 PR tree-optimization/84037
5992 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5993 parameter, move visited init to caller.
5994 (vect_slp_analyze_operations): Separate cost from validity
5995 check, initialize visited once for all instances.
5996 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5998 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6000 (vect_model_promotion_demotion_cost): Likewise.
6001 (vectorizable_bswap): Guard cost modeling with !slp_node
6002 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6004 (vectorizable_call): Likewise.
6005 (vectorizable_conversion): Likewise.
6006 (vectorizable_assignment): Likewise.
6007 (vectorizable_shift): Likewise.
6008 (vectorizable_operation): Likewise.
6009 (vectorizable_store): Likewise.
6010 (vectorizable_load): Likewise.
6011 (vectorizable_condition): Likewise.
6012 (vectorizable_comparison): Likewise.
6014 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6017 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6018 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6019 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6021 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6024 * params.def (inline-min-speedup): Increase from 8 to 15.
6025 (max-inline-insns-auto): Decrease from 40 to 30.
6026 * ipa-split.c (consider_split): Add some buffer for function to
6027 be considered inlining candidate.
6028 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6031 2018-02-12 Richard Biener <rguenther@suse.de>
6033 PR tree-optimization/84037
6034 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6035 matched stmts if we cannot swap the non-matched ones.
6037 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6039 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6040 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6041 _mm_maskz_scalef_round_ss): New intrinsics.
6042 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6043 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6044 __builtin_ia32_scalefss_round): Remove.
6045 (__builtin_ia32_scalefsd_mask_round,
6046 __builtin_ia32_scalefss_mask_round): New intrinsics.
6047 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6048 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6049 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6050 "<round_constraint>")): Changed to ...
6051 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6052 "<round_scalar_constraint>")): ... this.
6053 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6054 %0, %1, %2<round_op3>}"): Changed to ...
6055 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6056 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6057 %2<round_scalar_mask_op3>}"): ... this.
6058 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6060 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6062 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6063 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6064 (_mm_maskz_sqrt_round_ss): New intrinsics.
6065 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6066 (__builtin_ia32_sqrtsd_mask_round)
6067 (__builtin_ia32_sqrtss_mask_round): New builtins.
6068 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6069 (__builtin_ia32_sqrtss_round): Remove.
6070 (__builtin_ia32_sqrtsd_mask_round)
6071 (__builtin_ia32_sqrtss_mask_round): New builtins.
6072 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6073 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6074 ((match_operand:VF_128 1 "vector_operand"
6075 "xBm,<round_constraint>")): Changed to ...
6076 ((match_operand:VF_128 1 "vector_operand"
6077 "xBm,<round_scalar_constraint>")): ... this.
6078 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6079 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6080 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6081 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6082 %<iptr>1<round_scalar_mask_op3>}): ... this.
6083 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6084 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6086 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6089 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6090 Cast vec_cmpeq result to correct type.
6091 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6092 Cast vec_cmpgt result to correct type.
6094 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6096 * final.c (final_scan_insn_1): Renamed from...
6097 (final_scan_insn): ... this. New wrapper, to recover
6098 seen from the outermost call in recursive ones.
6099 * config/sparc/sparc.c (output_return): Drop seen from call.
6100 (output_sibcall): Likewise.
6101 * config/visium/visium.c (output_branch): Likewise.
6103 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6105 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6108 2018-02-10 Alan Modra <amodra@gmail.com>
6111 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6112 Specify LR as an input.
6114 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6117 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6118 remove_member_access_dummy_vars): New functions.
6119 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6120 lower_omp_1, execute_lower_omp): Use them.
6122 PR rtl-optimization/84308
6123 * shrink-wrap.c (spread_components): Release todo vector.
6125 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6127 PR rtl-optimization/57193
6128 * ira-color.c (struct allocno_color_data): Add member
6129 conflict_allocno_hard_prefs.
6130 (update_conflict_allocno_hard_prefs): New.
6131 (bucket_allocno_compare_func): Add a preference based on
6132 conflict_allocno_hard_prefs.
6133 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6134 (color_allocnos): Remove a dead code. Initiate
6135 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6137 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6140 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6141 constraint from =wa to wa. Avoid a subreg on the output operand,
6142 instead use a pseudo and subreg it in a move.
6143 (p9_xxbrd_<mode>): Changed to ...
6144 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6145 (p9_xxbrd_v2df): New expander.
6146 (p9_xxbrw_<mode>): Changed to ...
6147 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6148 (p9_xxbrw_v4sf): New expander.
6150 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6152 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6154 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6157 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6158 multiply in 32-bit mode.
6159 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6160 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6163 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6165 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6166 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6167 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6168 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6170 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6173 * dwarf2out.c (is_trivial_indirect_ref): New function.
6174 (dwarf2out_late_global_decl): Do not generate a location
6175 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6176 and that are not defined in the current unit.
6178 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6180 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6181 instead of a libcall for UNORDERED.
6183 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6186 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6187 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6189 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6192 * config/s390/s390.c (s390_set_current_function): Invoke
6193 s390_indirect_branch_settings also if fndecl didn't change.
6195 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6197 * config/rs6000/rs6000.md (blockage): Set length to zero.
6199 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6201 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6203 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6206 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6207 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6211 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6212 PARALLEL incoming that failed vt_get_decl_and_offset check.
6215 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6216 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6218 (get_variable_section): For decls in named .bss* sections pass true as
6219 second argument to bss_initializer_p.
6221 2018-02-09 Marek Polacek <polacek@redhat.com>
6222 Jakub Jelinek <jakub@redhat.com>
6225 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6226 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6227 Sync some changes from cxx_fold_indirect_ref.
6229 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6231 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6233 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6234 (BLOCK_INLINE_ENTRY_LABEL): New.
6235 (dwarf2out_var_location): Disregard inline entry markers.
6236 (inline_entry_data): New struct.
6237 (inline_entry_data_hasher): New hashtable type.
6238 (inline_entry_data_hasher::hash): New.
6239 (inline_entry_data_hasher::equal): New.
6240 (inline_entry_data_table): New variable.
6241 (add_high_low_attributes): Add DW_AT_entry_pc and
6242 DW_AT_GNU_entry_view attributes if a pending entry is found
6243 in inline_entry_data_table. Add old entry_pc attribute only
6244 if debug nonbinding markers are disabled.
6245 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6246 markers are enabled.
6247 (block_within_block_p, dwarf2out_inline_entry): New.
6248 (dwarf2out_finish): Check that no entries remained in
6249 inline_entry_data_table.
6250 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6251 (final_scan_insn, notice_source_line): Likewise.
6252 (rest_of_clean_state): Skip inline entry markers.
6253 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6255 * gimple.c (gimple_build_debug_inline_entry): New.
6256 * gimple.h (enum gimple_debug_subcode): Add
6257 GIMPLE_DEBUG_INLINE_ENTRY.
6258 (gimple_build_debug_inline_entry): Declare.
6259 (gimple_debug_inline_entry_p): New.
6260 (gimple_debug_nonbind_marker_p): Adjust.
6261 * insn-notes.def (INLINE_ENTRY): New.
6262 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6263 inline entry marker notes.
6264 (print_insn): Likewise.
6265 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6266 (INSN_DEBUG_MARKER_KIND): Likewise.
6267 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6268 * tree-inline.c (expand_call_inline): Build and insert
6269 debug_inline_entry stmt.
6270 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6271 inline entry blocks early, if nonbind markers are enabled.
6272 (dump_scope_block): Dump fragment info.
6273 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6274 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6275 (gimple_build_debug_inline_entry): New.
6276 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6277 Enable/disable inline entry points too.
6278 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6279 (DEBUG_INSN): Describe inline entry markers.
6281 * common.opt (gvariable-location-views): New.
6282 (gvariable-location-views=incompat5): New.
6283 * config.in: Rebuilt.
6284 * configure: Rebuilt.
6285 * configure.ac: Test assembler for view support.
6286 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6287 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6288 * dwarf2out.c (var_loc_view): New typedef.
6289 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6290 (dwarf2out_locviews_in_attribute): New.
6291 (dwarf2out_locviews_in_loclist): New.
6292 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6293 (enum dw_line_info_opcode): Add LI_adv_address.
6294 (struct dw_line_info_table): Add view.
6295 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6296 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6297 (zero_view_p): New variable.
6298 (ZERO_VIEW_P): New macro.
6299 (output_asm_line_debug_info): New.
6300 (struct var_loc_node): Add view.
6301 (add_AT_view_list, AT_loc_list): New.
6302 (add_var_loc_to_decl): Add view param. Test it against last.
6303 (new_loc_list): Add view params. Record them.
6304 (AT_loc_list_ptr): Handle loc and view lists.
6305 (view_list_to_loc_list_val_node): New.
6306 (print_dw_val): Handle dw_val_class_view_list.
6307 (size_of_die): Likewise.
6308 (value_format): Likewise.
6309 (loc_list_has_views): New.
6310 (gen_llsym): Set vl_symbol too.
6311 (maybe_gen_llsym, skip_loc_list_entry): New.
6312 (dwarf2out_maybe_output_loclist_view_pair): New.
6313 (output_loc_list): Output view list or entries too.
6314 (output_view_list_offset): New.
6315 (output_die): Handle dw_val_class_view_list.
6316 (output_dwarf_version): New.
6317 (output_compilation_unit_header): Use it.
6318 (output_skeleton_debug_sections): Likewise.
6319 (output_rnglists, output_line_info): Likewise.
6320 (output_pubnames, output_aranges): Update version comments.
6321 (output_one_line_info_table): Output view numbers in asm comments.
6322 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6323 Call maybe_gen_llsym.
6324 (loc_list_from_tree_1): Adjust.
6325 (add_AT_location_description): Create view list attribute if
6326 needed, check it's absent otherwise.
6327 (convert_cfa_to_fb_loc_list): Adjust.
6328 (maybe_emit_file): Call output_asm_line_debug_info for test.
6329 (dwarf2out_var_location): Reset views as needed. Precompute
6330 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6331 attribute. Set view.
6332 (new_line_info_table): Reset next view.
6333 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6334 (dwarf2out_source_line): Likewise. Output view resets and labels to
6335 the assembler, or select appropriate line info opcodes.
6336 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6337 (optimize_string_length): Catch it. Adjust.
6338 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6339 dw_val_class_view_list, and remove it if no longer needed.
6340 (hash_loc_list): Hash view numbers.
6341 (loc_list_hasher::equal): Compare them.
6342 (optimize_location_lists): Check whether a view list symbol is
6343 needed, and whether the locview attribute is present, and
6344 whether they match. Remove the locview attribute if no longer
6346 (index_location_lists): Call skip_loc_list_entry for test.
6347 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6348 Use output_dwarf_version.
6349 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6350 (struct dw_val_node): Add val_view_list.
6351 * final.c (SEEN_NEXT_VIEW): New.
6352 (set_next_view_needed): New.
6353 (clear_next_view_needed): New.
6354 (maybe_output_next_view): New.
6355 (final_start_function): Rename to...
6356 (final_start_function_1): ... this. Take pointer to FIRST,
6357 add SEEN parameter. Emit param bindings in the initial view.
6358 (final_start_function): Reintroduce SEEN-less interface.
6359 (final): Rename to...
6360 (final_1): ... this. Take SEEN parameter. Output final pending
6361 next view at the end.
6362 (final): Reintroduce seen-less interface.
6363 (final_scan_insn): Output pending next view before switching
6364 sections or ending a block. Mark the next view as needed when
6365 outputting variable locations. Notify debug backend of section
6366 changes, and of location view changes.
6367 (rest_of_handle_final): Adjust.
6368 * toplev.c (process_options): Autodetect value for debug variable
6369 location views option. Warn on incompat5 without -gdwarf-5.
6370 * doc/invoke.texi (gvariable-location-views): New.
6371 (gvariable-location-views=incompat5): New.
6372 (gno-variable-location-views): New.
6374 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6376 PR tree-optimization/84136
6377 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6378 that the result of find_edge is non-NULL.
6380 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6383 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6384 storing integer register in SImode. Fix cost of 256 and 512
6385 byte aligned SSE register store.
6387 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6389 * config/i386/i386.c (ix86_multiplication_cost): Fix
6390 multiplication cost for TARGET_AVX512DQ.
6392 2018-02-08 Marek Polacek <polacek@redhat.com>
6394 PR tree-optimization/84238
6395 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6398 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6400 PR tree-optimization/84265
6401 * tree-vect-stmts.c (vectorizable_store): Don't treat
6402 VMAT_CONTIGUOUS accesses as grouped.
6403 (vectorizable_load): Likewise.
6405 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6407 PR tree-optimization/81635
6408 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6409 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6410 (test_round_for_mask): New functions.
6411 (wide_int_cc_tests): Call test_round_for_mask.
6412 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6413 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6414 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6415 range returned by get_range_info.
6417 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6420 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6421 * symtab.c: Include builtins.h
6422 (symtab_node::output_to_lto_symbol_table_p): Move here
6423 from lto-streamer-out.c:output_symbol_p.
6424 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6425 (output_symbol_p): Move all logic to symtab.c
6426 (produce_symtab): Update.
6428 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6430 * config/s390/s390-opts.h (enum indirect_branch): Define.
6431 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6432 (s390_indirect_branch_via_thunk)
6433 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6434 (enum s390_indirect_branch_type): Define.
6435 * config/s390/s390.c (struct s390_frame_layout, struct
6436 machine_function): Remove.
6437 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6438 (indirect_branch_table_label_no, indirect_branch_table_name):
6440 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6441 (enum s390_indirect_branch_option): Define.
6442 (s390_return_addr_from_memory): New function.
6443 (s390_handle_string_attribute): New function.
6444 (s390_attribute_table): Add new attribute handler.
6445 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6446 (s390_indirect_branch_via_thunk): New function.
6447 (s390_indirect_branch_via_inline_thunk): New function.
6448 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6449 sibling call optimization for non z10 compiles.
6450 (s390_emit_call): Force indirect branch target to be a single
6451 register. Add r1 clobber for non-z10 compiles.
6452 (s390_emit_epilogue): Emit return jump via return_use expander.
6453 (s390_reorg): Handle JUMP_INSNs as execute targets.
6454 (s390_option_override_internal): Perform validity checks for the
6455 new command line options.
6456 (s390_indirect_branch_attrvalue): New function.
6457 (s390_indirect_branch_settings): New function.
6458 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6459 (s390_output_indirect_thunk_function): New function.
6460 (s390_code_end): Implement target hook.
6461 (s390_case_values_threshold): Implement target hook.
6462 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6464 * config/s390/s390.h (struct s390_frame_layout)
6465 (struct machine_function): Move here from s390.c.
6466 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6467 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6468 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6469 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6470 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6471 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6472 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6473 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6474 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6475 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6476 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6477 (mnemonic attribute): Add values which aren't recognized
6479 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6480 pattern for branch conversion. Fix mnemonic attribute.
6481 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6482 indirect branch via thunk if requested.
6483 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6484 ("*indirect_jump"): Disable for branch conversion using out of
6486 ("indirect_jump_via_thunk<mode>_z10")
6487 ("indirect_jump_via_thunk<mode>")
6488 ("indirect_jump_via_inlinethunk<mode>_z10")
6489 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6490 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6491 ("casesi_jump_via_inlinethunk<mode>_z10")
6492 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6493 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6494 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6495 ("*indirect2_jump"): Disable for branch conversion.
6496 ("casesi_jump"): Turn into expander and expand patterns for branch
6498 ("return_use"): New expander.
6499 ("*return"): Emit return via thunk and rename it to ...
6500 ("*return<mode>"): ... this one.
6501 * config/s390/s390.opt: Add new options and and enum for the
6504 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6506 * lra-constraints.c (match_reload): Unconditionally use
6507 gen_lowpart_SUBREG, rather than selecting between that
6508 and equivalent gen_rtx_SUBREG code.
6510 2018-02-08 Richard Biener <rguenther@suse.de>
6512 PR tree-optimization/84233
6513 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6514 changed flag instead of boguously re-using phi_inserted.
6516 2018-02-08 Martin Jambor <mjambor@suse.cz>
6518 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6519 static local variables.
6521 2018-02-08 Richard Biener <rguenther@suse.de>
6523 PR tree-optimization/84278
6524 * tree-vect-stmts.c (vectorizable_store): When looking for
6525 smaller vector types to perform grouped strided loads/stores
6526 make sure the mode is supported by the target.
6527 (vectorizable_load): Likewise.
6529 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6531 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6532 Increase LDP/STP opportunities by adding adjacent callee-saves.
6534 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6536 PR rtl-optimization/84068
6537 PR rtl-optimization/83459
6538 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6540 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6542 PR tree-optimization/84224
6543 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6544 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6547 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6550 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6551 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6552 count, remove one USE.
6554 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6556 * doc/install.texi (Configuration): Document the
6557 --with-long-double-format={ibm,ieee} PowerPC configuration
6561 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6562 Convert from define_expand to be define_insn_and_split. Rework
6563 float/double/_Float128 conversions to QI/HI/SImode to work with
6564 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6565 conversions to QI/HImode types did a store and then a load to
6566 truncate the value. For conversions to VSX registers, don't split
6567 the insn, instead emit the code directly. Use the code iterator
6568 any_fix to combine signed and unsigned conversions.
6569 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6570 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6571 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6572 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6573 (fix_<mode>di2_hw): Likewise.
6574 (fixuns_<mode>di2_hw): Likewise.
6575 (fix_<mode>si2_hw): Likewise.
6576 (fixuns_<mode>si2_hw): Likewise.
6577 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6578 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6579 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6580 fix<uns>_trunc<SFDF:mode>si2_p8.
6581 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6583 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6584 (fix<uns>_<mode>_mem): Likewise.
6585 (fctiw<u>z_<mode>_mem): Likewise.
6586 (fix<uns>_<mode>_mem): Likewise.
6587 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6588 the register allocator from doing a direct move to the GPRs to do
6589 a store, and instead use the ISA 3.0 store byte/half-word from
6590 vector register instruction. For IEEE 128-bit floating point,
6591 also optimize stores of 32-bit ints.
6592 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6594 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6596 * genextract.c (push_pathstr_operand): New function to support
6598 (walk_rtx): Call push_pathstr_operand.
6599 (print_path): Support [a-zA-Z].
6601 2018-02-07 Richard Biener <rguenther@suse.de>
6603 PR tree-optimization/84037
6604 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6605 (cse_and_gimplify_to_preheader): Declare.
6606 (vect_get_place_in_interleaving_chain): Likewise.
6607 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6609 (_loop_vec_info::~_loop_vec_info): Delete it.
6610 (cse_and_gimplify_to_preheader): New function.
6611 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6612 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6613 (vectorizable_load): Likewise. For grouped stores always base
6614 the IV on the first element.
6615 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6616 condition before gimplifying.
6618 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6620 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6621 *DIV_EXPR and *MOD_EXPR.
6623 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6626 * config/i386/i386.c (ix86_option_override_internal): Mask out
6627 the CF_SET bit when checking -fcf-protection.
6629 2018-02-07 Tom de Vries <tom@codesourcery.com>
6632 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6635 2018-02-07 Richard Biener <rguenther@suse.de>
6637 PR tree-optimization/84204
6638 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6641 PR tree-optimization/84205
6642 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6643 special-case isl_ast_op_zdiv_r.
6645 PR tree-optimization/84223
6646 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6647 Only add conditions from within the region.
6648 (gather_bbs::after_dom_children): Adjust.
6650 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6653 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6654 * config/avr/avr.md: Only post-reload split REG-REG moves if
6655 either register is GENERAL_REG_P.
6657 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6659 PR tree-optimization/84235
6660 * tree-ssa-scopedtables.c
6661 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6662 if the subtraction is performed in floating point type where NaNs are
6663 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6664 build 1. Formatting fix.
6666 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6669 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6670 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6671 and skip it regardless of bb boundaries. Use CALL_P macro,
6672 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6673 unnecessarily, formatting fix.
6675 2018-02-06 Michael Collison <michael.collison@arm.com>
6677 * config/arm/thumb2.md:
6678 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6679 (*thumb_mov_notscc): Ditto.
6681 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6684 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6685 unsigned_fix, not "s".
6687 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6689 * configure.ac (gcc_fn_eh_frame_ro): New function.
6690 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6691 correct .eh_frame permissions.
6692 * configure: Regenerate.
6694 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6696 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6699 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6701 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6702 Display warning message for -mno-speculate-indirect-jumps.
6704 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6706 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6708 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6710 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6712 PR tree-optimization/84225
6713 * tree-eh.c (find_trapping_overflow): Only call
6714 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6716 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6719 * config/i386/i386.c: Reimplement the check of possible options
6720 -mibt/-mshstk conbination. Change error messages.
6721 * doc/invoke.texi: Fix a typo: remove extra '='.
6723 2018-02-06 Marek Polacek <polacek@redhat.com>
6725 PR tree-optimization/84228
6726 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6728 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6731 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6732 emitted arch directives.
6733 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6734 __ARM_FEATURE_COPROC before changing architectures.
6736 2018-02-06 Richard Biener <rguenther@suse.de>
6738 * config/i386/i386.c (print_reg): Fix typo.
6739 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6741 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6743 * configure: Regenerate.
6745 2018-02-05 Martin Sebor <msebor@redhat.com>
6747 PR tree-optimization/83369
6748 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6751 2018-02-05 Martin Liska <mliska@suse.cz>
6753 * doc/invoke.texi: Cherry-pick upstream r323995.
6755 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6757 * ira.c (ira_init_register_move_cost): Adjust comment.
6759 2018-02-05 Martin Liska <mliska@suse.cz>
6761 PR gcov-profile/84137
6762 * doc/gcov.texi: Fix typo in documentation.
6764 2018-02-05 Martin Liska <mliska@suse.cz>
6766 PR gcov-profile/83879
6767 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6768 using dlopen functionality.
6770 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6772 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6773 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6774 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6775 _mm_maskz_range_round_ss): New intrinsics.
6776 (__builtin_ia32_rangesd128_round)
6777 (__builtin_ia32_rangess128_round): Remove.
6778 (__builtin_ia32_rangesd128_mask_round,
6779 __builtin_ia32_rangess128_mask_round): New builtins.
6780 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6781 __builtin_ia32_rangess128_round): Remove.
6782 (__builtin_ia32_rangesd128_mask_round,
6783 __builtin_ia32_rangess128_mask_round): New builtins.
6784 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6785 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6786 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6787 "<round_saeonly_constraint>")): Changed to ...
6788 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6789 "<round_saeonly_scalar_constraint>")): ... this.
6790 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6791 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6792 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6793 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6794 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6796 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6798 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6800 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6801 Remove all values except native, 8540 and 8548.
6803 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6805 * config/i386/i386.c (ix86_output_function_return): Pass
6806 INVALID_REGNUM, instead of -1, as invalid register number to
6807 indirect_thunk_name and output_indirect_thunk.
6809 2018-02-02 Julia Koval <julia.koval@intel.com>
6811 * config.gcc: Add -march=icelake.
6812 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6813 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6814 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6815 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6816 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6817 (processor_target_table): Add icelake.
6818 (ix86_option_override_internal): Handle new PTAs.
6819 (get_builtin_code_for_version): Handle icelake.
6820 (M_INTEL_COREI7_ICELAKE): New.
6821 (fold_builtin_cpu): Handle icelake.
6822 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6823 * doc/invoke.texi: Add -march=icelake.
6825 2018-02-02 Julia Koval <julia.koval@intel.com>
6827 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6828 to wide_int_bitmask.
6829 * wide-int-bitmask.h: New.
6831 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6834 * config/i386/i386.md: Replace Pmode with word_mode in
6835 builtin_setjmp_setup and builtin_longjmp to support x32.
6837 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6841 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6843 (rs6000_supported_cpu_names): New static variable.
6844 (linux_cpu_translation_table): Likewise.
6845 (elf_platform) <cpu>: Define new static variable and use it.
6846 Translate kernel AT_PLATFORM name to canonical name if needed.
6847 Error if platform name is unknown.
6849 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6852 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6854 2018-02-01 Jeff Law <law@redhat.com>
6857 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6858 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6859 the scratch if RELEASE_VIA_POP is false.
6860 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6861 If we have to save a temporary register, decrement SIZE appropriately.
6862 Pass new arguments to release_scratch_register_on_entry.
6863 (ix86_adjust_stack_and_probe): Likewise.
6864 (ix86_emit_probe_stack_range): Pass new arguments to
6865 release_scratch_register_on_entry.
6867 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6869 PR rtl-optimization/84157
6870 * combine.c (change_zero_ext): Use REG_P predicate in
6871 front of HARD_REGISTER_P predicate.
6873 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6875 * config/avr/avr.c (avr_option_override): Move disabling of
6876 -fdelete-null-pointer-checks to...
6877 * common/config/avr/avr-common.c (avr_option_optimization_table):
6880 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6882 PR tree-optimization/81635
6883 * tree-data-ref.c (split_constant_offset_1): For types that
6884 wrap on overflow, try to use range info to prove that wrapping
6887 2018-02-01 Renlin Li <renlin.li@arm.com>
6890 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6892 (aarch64_register_move_cost): Likewise.
6893 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6895 (REG_CLASS_NAMES): Likewise.
6896 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6897 TAILCALL_ADDR_REGS. Remove IP registers.
6898 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6900 2018-02-01 Richard Biener <rguenther@suse.de>
6902 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6903 for specifying RPO order and allow NULL for that.
6904 * domwalk.c (dom_walker::dom_walker): Likewise.
6905 (dom_walker::walk): Handle NULL RPO order.
6906 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6908 (rewrite_update_dom_walker): Likewise.
6909 (mark_def_dom_walker): Likewise.
6911 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6913 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6914 (aarch64_maybe_expand_sve_subreg_move): Declare.
6915 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6916 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6918 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6919 that are semantically a reverse operation.
6920 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6921 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6922 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6924 (aarch64_can_change_mode_class): For big-endian, forbid changes
6925 between two SVE modes if they have different element sizes.
6927 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6929 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6930 the TImode handling for big-endian targets.
6932 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6934 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6935 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6937 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6938 Remove BSWAP handing for big-endian targets and use the form of
6939 LD1RQ appropariate for the mode.
6941 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6943 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6944 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6947 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6950 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6951 check for operands that need to go through aarch64_sve_reload_be.
6953 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6955 PR tree-optimization/81661
6956 PR tree-optimization/84117
6957 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6958 * tree-eh.c: Include gimplify.h.
6959 (find_trapping_overflow, replace_trapping_overflow,
6960 rewrite_to_non_trapping_overflow): New functions.
6961 * tree-vect-loop.c: Include tree-eh.h.
6962 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6963 * tree-data-ref.c: Include tree-eh.h.
6964 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6966 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6968 PR rtl-optimization/84123
6969 * combine.c (change_zero_ext): Check if hard register satisfies
6970 can_change_dest_mode before calling gen_lowpart_SUBREG.
6972 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6975 * ira.c (ira_init_register_move_cost): Remove assert.
6977 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6979 PR rtl-optimization/84071
6980 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6981 * doc/tm.texi: Regenerate.
6983 2018-01-31 Richard Biener <rguenther@suse.de>
6985 PR tree-optimization/84132
6986 * tree-data-ref.c (analyze_miv_subscript): Properly
6987 check whether evolution_function_is_affine_multivariate_p
6988 before calling gcd_of_steps_may_divide_p.
6990 2018-01-31 Julia Koval <julia.koval@intel.com>
6993 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6994 * config/i386/i386.md (rdpid_rex64) New.
6995 (rdpid): Make 32bit only.
6997 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7000 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7001 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7003 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7006 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7008 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7010 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7012 PR rtl-optimization/84071
7013 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7014 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7016 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7018 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7019 (arc_attribute_table): Add 'aux' attribute.
7020 (arc_in_small_data_p): Consider aux like variables.
7021 (arc_is_aux_reg_p): New function.
7022 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7023 (arc_get_aux_arg): New function.
7024 (prepare_move_operands): Handle aux-register access.
7025 (arc_handle_aux_attribute): New function.
7026 * doc/extend.texi (ARC Variable attributes): Add subsection.
7028 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7030 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7031 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7032 (arc_attribute_table): Add 'uncached' attribute.
7033 (arc_print_operand): Print '.di' flag for uncached memory
7035 (arc_in_small_data_p): Do not consider for small data the uncached
7037 (arc_is_uncached_mem_p): New function.
7038 * config/arc/predicates.md (compact_store_memory_operand): Check
7039 for uncached memory accesses.
7040 (nonvol_nonimm_operand): Likewise.
7041 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7043 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7046 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7047 falign-loops=): Add Optimization flag.
7049 2018-01-30 Jeff Law <law@redhat.com>
7052 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7053 INT_REGISTERS_SAVED. Check it prior to calling
7054 get_scratch_register_on_entry.
7055 (ix86_adjust_stack_and_probe): Similarly.
7056 (ix86_emit_probe_stack_range): Similarly.
7057 (ix86_expand_prologue): Corresponding changes.
7059 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7062 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7063 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7065 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7068 * lra-constraints.c (curr_insn_transform): Process AND in the
7071 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7073 PR rtl-optimization/83986
7074 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7075 dependence against last_pending_memory_flush in addition to
7078 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7080 PR tree-optimization/81611
7081 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7084 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7087 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7090 2018-01-30 Richard Biener <rguenther@suse.de>
7091 Jakub Jelinek <jakub@redhat.com>
7093 PR tree-optimization/84111
7094 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7095 inner loops added during recursion, as they don't have up-to-date
7098 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7101 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7102 (can_inline_edge_by_limits_p): ... here.
7103 (can_early_inline_edge_p, check_callers,
7104 update_caller_keys, update_callee_keys, recursive_inlining,
7105 add_new_edges_to_heap, speculation_useful_p,
7106 inline_small_functions,
7107 inline_small_functions, flatten_function,
7108 inline_to_all_callers_1): Update.
7110 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7112 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7115 2018-01-30 Richard Biener <rguenther@suse.de>
7117 PR tree-optimization/83008
7118 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7119 invariant and constant vector uses in stmts when they need
7122 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7125 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7126 * configure: Regenerate.
7128 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7130 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7132 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7133 Use gen_rtx_REG rather than gen_lowpart.
7135 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7137 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7138 rather than 0 when creating partial subregs.
7140 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7142 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7145 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7148 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7149 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7150 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7151 flags. This restores the settings used before the 2017-07-24.
7152 Turning off pre increment/decrement/modify allows IVOPTS to
7153 optimize DF/SF loops where the index is an int.
7155 2018-01-29 Richard Biener <rguenther@suse.de>
7156 Kelvin Nilsen <kelvin@gcc.gnu.org>
7159 * tree-vect-stmts.c (vectorizable_call): Don't call
7160 targetm.vectorize_builtin_md_vectorized_function if callee is
7163 2018-01-22 Carl Love <cel@us.ibm.com>
7165 * doc/extend.tex: Fix typo in second arg in
7166 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7168 2018-01-29 Richard Biener <rguenther@suse.de>
7170 PR tree-optimization/84086
7171 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7172 (flush_ssaname_freelist): When SSA names were released reset
7173 the SCEV hash table.
7175 2018-01-29 Richard Biener <rguenther@suse.de>
7177 PR tree-optimization/84057
7178 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7179 removed paths when removing edges.
7181 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7183 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7184 -mfunction-return=@var{choice}.
7186 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7189 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7191 (layout::print_source_line): Likewise.
7192 (test_get_line_width_without_trailing_whitespace): Add test cases.
7194 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7197 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7200 2018-01-26 Jim Wilson <jimw@sifive.com>
7202 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7204 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7207 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7209 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7210 and CMP + SUB-immediate -> SUBS.
7212 2018-01-26 Martin Sebor <msebor@redhat.com>
7214 PR tree-optimization/83896
7215 * tree-ssa-strlen.c (get_string_len): Rename...
7216 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7217 Avoid assuming length is constant.
7218 (handle_char_store): Use HOST_WIDE_INT for string length.
7220 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7223 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7224 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7226 2018-01-26 Richard Biener <rguenther@suse.de>
7228 PR rtl-optimization/84003
7229 * dse.c (record_store): Only record redundant stores when
7230 the earlier store aliases at least all accesses the later one does.
7232 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7234 PR rtl-optimization/83985
7235 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7236 REG_CFA_RESTORE insns.
7237 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7238 REG_CFA_RESTORE insns here.
7241 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7242 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7244 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7246 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7247 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7248 (arc_init): Likewise.
7249 (arc_override_options): Likewise.
7250 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7252 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7254 * config/arc/arc.h (TARGET_DBNZ): Define.
7255 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7256 properly set the tune attribute.
7257 (dbnz): Use TARGET_DBNZ guard.
7258 * config/arc/arc.opt (mtune): Add core3 option.
7260 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7262 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7263 recognize new pic like addresses.
7264 (arc_delegitimize_address): Clean up.
7266 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7268 * config/arc/arc-arches.def: Option mrf16 valid for all
7270 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7271 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7272 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7273 * config/arc/arc-tables.opt: Regenerate.
7274 * config/arc/arc.c (arc_conditional_register_usage): Handle
7275 reduced register file case.
7276 (arc_file_start): Set must have build attributes.
7277 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7279 * config/arc/arc.opt (mrf16): Add new option.
7280 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7281 * config/arc/genmultilib.awk: Handle new mrf16 option.
7282 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7283 * config/arc/t-multilib: Regenerate.
7284 * doc/invoke.texi (ARC Options): Document mrf16 option.
7286 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7288 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7289 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7290 (arc_attribute_table): Add 'secure_call' attribute.
7291 (arc_print_operand): Print secure call operand.
7292 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7294 (arc_is_secure_call_p): New function. * config/arc/arc.md
7295 (call_i): Add support for sjli instruction.
7296 (call_value_i): Likewise.
7297 * config/arc/constraints.md (Csc): New constraint.
7299 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7300 John Eric Martin <John.Martin@emmicro-us.com>
7302 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7303 * config/arc/arc.c (_arc_jli_section): New struct.
7304 (arc_jli_section): New type.
7305 (rc_jli_sections): New static variable.
7306 (arc_handle_jli_attribute): New function.
7307 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7308 (arc_file_end): New function.
7309 (TARGET_ASM_FILE_END): Define.
7310 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7311 (arc_add_jli_section): New function.
7312 (jli_call_scan): Likewise.
7313 (arc_reorg): Call jli_call_scan.
7314 (arc_output_addsi): Remove 'S' from printing asm operand.
7315 (arc_is_jli_call_p): New function.
7316 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7318 (movhi_insn): Likewise.
7319 (movsi_insn): Likewise.
7320 (movsi_set_cc_insn): Likewise.
7321 (loadqi_update): Likewise.
7322 (load_zeroextendqisi_update): Likewise.
7323 (load_signextendqisi_update): Likewise.
7324 (loadhi_update): Likewise.
7325 (load_zeroextendhisi_update): Likewise.
7326 (load_signextendhisi_update): Likewise.
7327 (loadsi_update): Likewise.
7328 (loadsf_update): Likewise.
7329 (movsicc_insn): Likewise.
7330 (bset_insn): Likewise.
7331 (bxor_insn): Likewise.
7332 (bclr_insn): Likewise.
7333 (bmsk_insn): Likewise.
7334 (bicsi3_insn): Likewise.
7335 (cmpsi_cc_c_insn): Likewise.
7336 (movsi_ne): Likewise.
7337 (movsi_cond_exec): Likewise.
7338 (clrsbsi2): Likewise.
7349 (call_i): Remove 'S' asm letter, add jli instruction.
7350 (call_value_i): Likewise.
7351 * config/arc/arc.op (mjli-always): New option.
7352 * config/arc/constraints.md (Cji): New constraint.
7353 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7355 (subsf3_fpx): Likewise.
7356 (mulsf3_fpx): Likewise.
7357 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7359 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7361 * doc/invoke.texi (ARC): Document mjli-always option.
7363 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7365 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7366 avoid addition with 0 and use incw and decw where possible.
7368 2018-01-26 Richard Biener <rguenther@suse.de>
7370 PR tree-optimization/81082
7371 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7372 association if it requires casting to unsigned.
7373 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7374 from fold_plusminus_mult_expr to catch important cases late when
7375 range info is available.
7377 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7379 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7380 * configure.ac (hidden_linkonce): New test.
7381 * configure: Regenerate.
7382 * config.in: Regenerate.
7384 2018-01-26 Julia Koval <julia.koval@intel.com>
7386 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7387 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7388 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7389 _mm_mask_bitshuffle_epi64_mask): Fix type.
7390 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7391 USI_FTYPE_V4DI_V4DI_USI): Remove.
7392 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7393 __builtin_ia32_vpshufbitqmb256_mask,
7394 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7395 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7396 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7398 2018-01-26 Alan Modra <amodra@gmail.com>
7401 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7402 UNSPEC_VBPERMQ. Sort other unspecs.
7404 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7406 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7408 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7411 * predict.c (drop_profile): Do not push/pop cfun; update also
7413 (handle_missing_profiles): Fix logic looking for zero profiles.
7415 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7418 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7419 on functions with #pragma omp declare simd or functions with simd
7421 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7422 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7423 Remove trailing \n from warning_at calls.
7425 2018-01-25 Tom de Vries <tom@codesourcery.com>
7428 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7429 for neutered workers.
7431 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7434 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7435 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7437 2018-01-24 Jeff Law <law@redhat.com>
7440 * i386.c (get_probe_interval): Move to earlier point.
7441 (ix86_compute_frame_layout): If -fstack-clash-protection and
7442 the frame is larger than the probe interval, then use pushes
7443 to save registers rather than reg->mem moves.
7444 (ix86_expand_prologue): Remove conditional for int_registers_saved
7447 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7450 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7451 min/max for never referenced object.
7453 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7456 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7458 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7459 attributes from DECL_ATTRIBUTES (decl) without affecting
7460 DECL_ATTRIBUTES (current_function_decl).
7461 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7462 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7464 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7466 PR tree-optimization/83979
7467 * fold-const.c (fold_comparison): Use constant_boolean_node
7468 instead of boolean_{true,false}_node.
7470 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7472 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7475 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7477 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7478 Simplify the clause that sets the length attribute.
7479 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7480 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7481 clause that sets the length attribute.
7482 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7484 2018-01-24 Tom de Vries <tom@codesourcery.com>
7487 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7488 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7489 Add strict parameter.
7490 (prevent_branch_around_nothing): Insert dummy insn between branch to
7491 label and label with no ptx insn inbetween.
7492 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7494 2018-01-24 Tom de Vries <tom@codesourcery.com>
7497 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7498 for neutered threads in warp.
7499 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7501 2018-01-24 Richard Biener <rguenther@suse.de>
7503 PR tree-optimization/83176
7504 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7507 2018-01-24 Richard Biener <rguenther@suse.de>
7509 PR tree-optimization/82819
7510 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7511 code generating pluses that are no-ops in the target precision.
7513 2018-01-24 Richard Biener <rguenther@suse.de>
7516 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7518 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7520 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7521 to merge probabilities.
7522 * predict.c (probably_never_executed): Also mark as cold functions
7523 with global 0 profile and guessed local profile.
7524 * profile-count.c (profile_probability::combine_with_count): New
7526 * profile-count.h (profile_probability::operator*,
7527 profile_probability::operator*=, profile_probability::operator/,
7528 profile_probability::operator/=): Reduce precision to adjusted
7529 and set value to guessed on contradictory divisions.
7530 (profile_probability::combine_with_freq): Remove.
7531 (profile_probability::combine_wiht_count): Declare.
7532 (profile_count::force_nonzero):: Set to adjusted.
7533 (profile_count::probability_in):: Set quality to adjusted.
7534 * tree-ssa-tail-merge.c (replace_block_by): Use
7537 2018-01-23 Andrew Waterman <andrew@sifive.com>
7538 Jim Wilson <jimw@sifive.com>
7540 * config/riscv/riscv.c (riscv_stack_boundary): New.
7541 (riscv_option_override): Set riscv_stack_boundary. Handle
7542 riscv_preferred_stack_boundary_arg.
7543 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7544 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7545 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7546 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7547 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7548 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7550 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7553 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7554 of struct ix86_frame.
7555 (ix86_expand_epilogue): Likewise. Add a local variable for
7556 the reg_save_offset field in struct ix86_frame.
7558 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7560 PR tree-optimization/82604
7561 * tree-loop-distribution.c (enum partition_kind): New enum item
7562 PKIND_PARTIAL_MEMSET.
7563 (partition_builtin_p): Support above new enum item.
7564 (generate_code_for_partition): Ditto.
7565 (compute_access_range): Differentiate cases that equality can be
7566 proven at all loops, the innermost loops or no loops.
7567 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7568 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7569 (finalize_partitions, distribute_loop): Don't fuse partition of
7570 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7571 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7574 2018-01-23 Martin Liska <mliska@suse.cz>
7576 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7577 order to ignore the predictor.
7578 (PRED_POLYMORPHIC_CALL): Likewise.
7579 (PRED_RECURSIVE_CALL): Likewise.
7581 2018-01-23 Martin Liska <mliska@suse.cz>
7583 * tree-profile.c (tree_profiling): Print function header to
7584 aware reader which function we are working on.
7585 * value-prof.c (gimple_find_values_to_profile): Do not print
7586 not interesting value histograms.
7588 2018-01-23 Martin Liska <mliska@suse.cz>
7590 * profile-count.h (enum profile_quality): Add
7591 profile_uninitialized as the first value. Do not number values
7592 as they are zero based.
7593 (profile_count::verify): Update sanity check.
7594 (profile_probability::verify): Likewise.
7596 2018-01-23 Nathan Sidwell <nathan@acm.org>
7598 * doc/invoke.texi (ffor-scope): Deprecate.
7600 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7602 PR tree-optimization/83510
7603 * domwalk.c (set_all_edges_as_executable): New function.
7604 (dom_walker::dom_walker): Convert bool param
7605 "skip_unreachable_blocks" to enum reachability. Move setup of
7606 edge flags to set_all_edges_as_executable and only do it when
7607 reachability is REACHABLE_BLOCKS.
7608 * domwalk.h (enum dom_walker::reachability): New enum.
7609 (dom_walker::dom_walker): Convert bool param
7610 "skip_unreachable_blocks" to enum reachability.
7611 (set_all_edges_as_executable): New decl.
7612 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7613 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7615 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7616 but converting true to REACHABLE_BLOCKS.
7617 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7619 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7620 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7621 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7623 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7624 if check_all_array_refs will be called.
7626 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7628 * tree.c (selftest::test_location_wrappers): Add more test
7631 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7633 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7634 (selftest::test_bit_in_range): Likewise.
7636 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7639 * doc/sourcebuild.texi (vect_float): Say that the selector
7640 only describes the situation when -funsafe-math-optimizations is on.
7641 (vect_float_strict): Document.
7643 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7645 PR tree-optimization/83965
7646 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7647 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7648 instead of checking only for a reduction.
7649 (vect_recog_widen_sum_pattern): Likewise.
7651 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7653 * predict.c (probably_never_executed): Only use precise profile info.
7654 (compute_function_frequency): Skip after inlining hack since we now
7655 have quality checking.
7657 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7659 * profile-count.h (profile_probability::very_unlikely,
7660 profile_probability::unlikely, profile_probability::even): Set
7661 precision to guessed.
7663 2018-01-23 Richard Biener <rguenther@suse.de>
7665 PR tree-optimization/83963
7666 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7667 Properly terminate dominator walk when crossing the exit edge not
7668 when visiting its source block.
7670 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7673 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7674 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7676 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7678 PR tree-optimization/83957
7679 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7680 semicolon after for body surrounded by braces.
7682 PR tree-optimization/83081
7683 * profile-count.h (profile_probability::split): New method.
7684 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7685 Use profile_probability::split.
7686 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7687 when splitting a single conditional jump into 2.
7689 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7691 PR tree-optimization/69452
7692 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7695 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7697 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7698 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7699 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7701 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7703 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7704 * config/rl78/rl78.md (movdi): New define_expand.
7705 * config/rl78/rl78.c (rl78_split_movdi): New function.
7707 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7710 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7712 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7713 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7714 128-bit to produce an UNSPEC move to get the double word with the
7715 signbit and then a shift directly to do signbit.
7716 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7717 implementation with a new version that just does either a direct
7718 move or a regular move. Move memory interface to separate insns.
7719 Move insns so they are next to the expander.
7720 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7721 with signbit move. Split big and little endian case.
7722 (signbit<mode>2_dm_mem_le): Likewise.
7723 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7724 (signbit<mode>2_dm2): Likewise.
7726 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7728 * config/rl78/rl78.md (anddi3): New define_expand.
7730 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7732 * config/rl78/rl78.md (umindi3): New define_expand.
7734 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7736 * config/rl78/rl78.md (smindi3): New define_expand.
7738 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7740 * config/rl78/rl78.md (smaxdi3): New define_expand.
7742 2018-01-22 Carl Love <cel@us.ibm.com>
7744 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7745 LVX_V1TI): Add macro expansion.
7746 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7747 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7748 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7749 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7750 Change check to determine if the instruction is a byte reversing
7751 entry. Fix typo in comment.
7752 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7753 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7754 Add def_builtin calls for new builtins.
7755 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7756 Add define_insn expansion.
7758 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7760 * config/rl78/rl78.md (umaxdi3): New define_expand.
7762 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7764 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7765 for non-QImode registers.
7767 2018-01-22 Richard Biener <rguenther@suse.de>
7769 PR tree-optimization/83963
7770 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7771 including the loop exit block.
7772 (scop_detection::merge_sese): Likewise.
7773 (scop_detection::add_scop): Do it here instead.
7775 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7777 * doc/sourcebuild.texi (arm_softfloat): Document.
7779 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7782 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7783 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7784 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7786 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7787 David Edelsohn <dje.gcc@gmail.com>
7790 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7791 Change "crset eq" to "crset 2".
7792 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7793 (*call_indirect_aix<mode>_nospec): Likewise.
7794 (*call_value_indirect_aix<mode>_nospec): Likewise.
7795 (*call_indirect_elfv2<mode>_nospec): Likewise.
7796 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7797 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7798 change assembly output from . to $.
7799 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7800 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7801 (*tablejump<mode>_internal1_nospec): Likewise.
7803 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7806 * config/sh/sh_optimize_sett_clrt.cc:
7807 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7809 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7811 PR tree-optimization/83940
7812 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7813 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7814 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7815 use it to pass back the definition type.
7816 (vect_check_store_rhs): Likewise rhs_dt_out.
7817 (vect_build_gather_load_calls): Add a mask_dt argument and use
7818 it instead of a call to vect_is_simple_use.
7819 (vectorizable_store): Update calls to vect_check_load_store_mask
7820 and vect_check_store_rhs. Use the dt returned by the latter instead
7821 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7822 instead of calls to vect_is_simple_use. Pass the scalar rather
7823 than the vector operand to vect_is_simple_use when handling
7824 second and subsequent copies of an rhs value.
7825 (vectorizable_load): Update calls to vect_check_load_store_mask
7826 and vect_build_gather_load_calls. Use the cached mask_dt and
7827 gs_info.offset_dt instead of calls to vect_is_simple_use.
7829 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7832 * tree-emutls.c: Include gimplify.h.
7833 (lower_emutls_2): New function.
7834 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7835 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7836 it before further processing.
7839 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7840 UINTVAL (trueop1) instead of INTVAL (op1).
7842 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7846 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7847 INCOMING_FRAME_SP_OFFSET if not defined.
7848 (scan_trace): Add ENTRY argument. If true and
7849 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7850 emit a note to adjust the CFA offset.
7851 (create_cfi_notes): Adjust scan_trace callers.
7852 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7853 INCOMING_FRAME_SP_OFFSET in the CIE.
7854 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7855 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7857 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7858 * doc/tm.texi: Regenerated.
7860 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7862 PR rtl-optimization/83147
7863 * lra-constraints.c (remove_inheritance_pseudos): Use
7864 lra_substitute_pseudo_within_insn.
7866 2018-01-19 Tom de Vries <tom@codesourcery.com>
7867 Cesar Philippidis <cesar@codesourcery.com>
7870 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7872 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7875 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7876 spaces for function labels.
7878 2018-01-19 Martin Liska <mliska@suse.cz>
7880 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7881 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7882 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7883 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7884 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7885 (PRED_CONST_RETURN): Change from 69 to 65.
7886 (PRED_NULL_RETURN): Change from 91 to 71.
7887 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7888 (PRED_LOOP_GUARD): Change from 66 to 73.
7890 2018-01-19 Martin Liska <mliska@suse.cz>
7892 * predict.c (predict_insn_def): Add new assert.
7893 (struct branch_predictor): Change type to signed integer.
7894 (test_prediction_value_range): Amend test to cover
7896 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7897 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7898 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7899 (PRED_LOOP_IV_COMPARE): Likewise.
7900 * predict.h (PROB_UNINITIALIZED): Define new constant.
7902 2018-01-19 Martin Liska <mliska@suse.cz>
7904 * predict.c (dump_prediction): Add new format for
7905 analyze_brprob.py script which is enabled with -details
7907 * profile-count.h (precise_p): New function.
7909 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7911 PR tree-optimization/83922
7912 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7913 there are no statements that need masking.
7914 (vect_active_double_reduction_p): New function.
7915 (vect_analyze_loop_operations): Use it when handling phis that
7916 are not in the loop header.
7918 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7920 PR tree-optimization/83914
7921 * tree-vect-loop.c (vectorizable_induction): Don't convert
7922 init_expr or apply the peeling adjustment for inductions
7923 that are nested within the vectorized loop.
7925 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7927 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7930 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7934 * function.h (gimplify_parameters): Add gimple_seq * argument.
7935 * function.c: Include gimple.h and options.h.
7936 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7937 for the added local temporaries if needed.
7938 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7939 if there are any parameter cleanups, wrap whole body into a
7940 try/finally with the cleanups.
7942 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7945 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7946 Use GET_MODE_CLASS for scalar floating point.
7948 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7952 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7953 Fix call of call_cgraph_insertion_hooks.
7955 2018-01-18 Martin Sebor <msebor@redhat.com>
7957 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7959 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7962 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7965 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7968 * common.opt: (-ffile-prefix-map): New option.
7969 * opts.c (common_handle_option): Defer it.
7970 * opts-global.c (handle_common_deferred_options): Handle it.
7971 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7972 * file-prefix-map.h: New file.
7973 (remap_debug_filename, add_debug_prefix_map): ...here.
7974 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7975 * final.c (debug_prefix_map, add_debug_prefix_map
7976 remap_debug_filename): Move to...
7977 * file-prefix-map.c: New file.
7978 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7979 generalize, get rid of alloca(), use strrchr() instead of strchr().
7980 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7981 Implement in terms of add_prefix_map().
7982 (remap_macro_filename, remap_debug_filename): Implement in term of
7984 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7985 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7986 * dbxout.c: Include file-prefix-map.h.
7987 * varasm.c: Likewise.
7988 * vmsdbgout.c: Likewise.
7989 * xcoffout.c: Likewise.
7990 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7991 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7992 * doc/invoke.texi (-ffile-prefix-map): Document.
7993 (-fdebug-prefix-map): Update description.
7995 2018-01-18 Martin Liska <mliska@suse.cz>
7997 * config/i386/i386.c (indirect_thunk_name): Document that also
7999 (output_indirect_thunk): Document why both instructions
8000 (pause and lfence) are generated.
8002 2018-01-18 Richard Biener <rguenther@suse.de>
8004 PR tree-optimization/83887
8005 * graphite-scop-detection.c
8006 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8007 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8008 (scop_detection::merge_sese): Re-implement with a flood-fill
8009 algorithm that properly finds a SESE region if it exists.
8011 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8014 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8015 pointer_diff optimizations use view_convert instead of convert.
8017 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8019 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8020 Generate different code for -mno-speculate-indirect-jumps.
8021 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8022 (*call_indirect_aix<mode>): Disable for
8023 -mno-speculate-indirect-jumps.
8024 (*call_indirect_aix<mode>_nospec): New define_insn.
8025 (*call_value_indirect_aix<mode>): Disable for
8026 -mno-speculate-indirect-jumps.
8027 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8028 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8029 -mno-speculate-indirect-jumps.
8030 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8032 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8034 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8035 long double type, set the flags for noting the default long double
8036 type, even if we don't pass or return a long double type.
8038 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8041 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8044 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8046 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8047 support for merge[hl].
8048 (fold_mergehl_helper): New helper function.
8049 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8050 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8051 (altivec_vmrglw_direct): Add xxmrglw insn.
8053 2018-01-17 Andrew Waterman <andrew@sifive.com>
8055 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8056 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8058 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8061 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8062 call the lto_location_cache before reading the
8063 DECL_SOURCE_LOCATION of the types.
8065 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8066 Richard Sandiford <richard.sandiford@linaro.org>
8068 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8069 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8070 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8071 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8072 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8074 * config/aarch64/constraints.md (aarch64_movti_operand):
8076 * config/aarch64/predicates.md (Uti): Add new constraint.
8078 2018-01-17 Carl Love <cel@us.ibm.com>
8080 * config/rs6000/vsx.md (define_expand xl_len_r,
8081 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8082 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8084 (define_expand, define_insn): Move the shift left from the
8085 define_insn to the define_expand for lxvl and stxvl instructions.
8086 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8087 and XL_LEN_R definitions to PURE.
8089 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8091 * config/i386/i386.c (indirect_thunk_name): Declare regno
8092 as unsigned int. Compare regno with INVALID_REGNUM.
8093 (output_indirect_thunk): Ditto.
8094 (output_indirect_thunk_function): Ditto.
8095 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8096 in the call to output_indirect_thunk_function.
8098 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8101 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8102 rather than the size of inner_type to determine the stack slot size
8103 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8105 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8108 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8111 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8113 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8114 endian Linux systems to optionally enable multilibs for selecting
8115 the long double type if the user configured an explicit type.
8116 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8117 have no long double multilibs if not defined.
8118 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8119 warn if the user used -mabi={ieee,ibm}longdouble and we built
8120 multilibs for long double.
8121 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8122 appropriate multilib option.
8123 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8125 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8126 for building long double multilibs.
8127 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8129 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8131 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8134 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8136 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8139 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8142 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8145 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8147 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8148 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8150 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8152 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8153 different rtl trees depending on TARGET_64BIT.
8154 (rs6000_gen_lvx): Likewise.
8156 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8158 * config/visium/visium.md (nop): Tweak comment.
8159 (hazard_nop): Likewise.
8161 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8163 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8164 -mspeculate-indirect-jumps.
8165 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8166 for -mno-speculate-indirect-jumps.
8167 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8168 (*call_value_indirect_elfv2<mode>): Disable for
8169 -mno-speculate-indirect-jumps.
8170 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8171 (indirect_jump): Emit different RTL for
8172 -mno-speculate-indirect-jumps.
8173 (*indirect_jump<mode>): Disable for
8174 -mno-speculate-indirect-jumps.
8175 (*indirect_jump<mode>_nospec): New define_insn.
8176 (tablejump): Emit different RTL for
8177 -mno-speculate-indirect-jumps.
8178 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8179 (tablejumpsi_nospec): New define_expand.
8180 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8181 (tablejumpdi_nospec): New define_expand.
8182 (*tablejump<mode>_internal1): Disable for
8183 -mno-speculate-indirect-jumps.
8184 (*tablejump<mode>_internal1_nospec): New define_insn.
8185 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8188 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8190 * caller-save.c (insert_save): Drop unnecessary parameter. All
8193 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8194 Richard Biener <rguenth@suse.de>
8197 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8198 return early, inline manually is_gimple_sizepos. Make sure if we
8199 call gimplify_expr we don't end up with a gimple constant.
8200 * tree.c (variably_modified_type_p): Don't return true for
8201 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8202 * gimplify.h (is_gimple_sizepos): Remove.
8204 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8206 PR tree-optimization/83857
8207 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8208 vectorizable_live_operation for pure SLP statements.
8209 (vectorizable_live_operation): Handle PHIs.
8211 2018-01-16 Richard Biener <rguenther@suse.de>
8213 PR tree-optimization/83867
8214 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8215 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8217 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8220 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8221 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8222 If off is not INTEGER_CST, issue a may not be aligned warning
8223 rather than isn't aligned. Use isn%'t rather than isn't.
8224 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8226 <case MULT_EXPR>: Improve the case when bottom and one of the
8227 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8228 operand, in that case check if the other operand is multiple of
8229 bottom divided by the INTEGER_CST operand.
8231 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8234 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8235 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8236 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8237 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8238 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8239 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8240 (pa_function_arg_size): New function.
8242 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8244 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8245 in a separate statement.
8247 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8249 PR tree-optimization/83847
8250 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8251 group gathers and scatters.
8253 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8255 PR rtl-optimization/86620
8256 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8258 PR rtl-optimization/83213
8259 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8260 to last if both are JUMP_INSNs.
8262 PR tree-optimization/83843
8263 * gimple-ssa-store-merging.c
8264 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8265 store_immediate_info for bswap/nop orig_stores.
8267 2018-01-15 Andrew Waterman <andrew@sifive.com>
8269 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8271 <UDIV>: Increase cost if !TARGET_DIV.
8273 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8275 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8276 (define_attr "cr_logical_3op"): New.
8277 (cceq_ior_compare): Adjust.
8278 (cceq_ior_compare_complement): Adjust.
8279 (*cceq_rev_compare): Adjust.
8280 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8281 (is_cracked_insn): Adjust.
8282 (insn_must_be_first_in_group): Adjust.
8283 * config/rs6000/40x.md: Adjust.
8284 * config/rs6000/440.md: Adjust.
8285 * config/rs6000/476.md: Adjust.
8286 * config/rs6000/601.md: Adjust.
8287 * config/rs6000/603.md: Adjust.
8288 * config/rs6000/6xx.md: Adjust.
8289 * config/rs6000/7450.md: Adjust.
8290 * config/rs6000/7xx.md: Adjust.
8291 * config/rs6000/8540.md: Adjust.
8292 * config/rs6000/cell.md: Adjust.
8293 * config/rs6000/e300c2c3.md: Adjust.
8294 * config/rs6000/e500mc.md: Adjust.
8295 * config/rs6000/e500mc64.md: Adjust.
8296 * config/rs6000/e5500.md: Adjust.
8297 * config/rs6000/e6500.md: Adjust.
8298 * config/rs6000/mpc.md: Adjust.
8299 * config/rs6000/power4.md: Adjust.
8300 * config/rs6000/power5.md: Adjust.
8301 * config/rs6000/power6.md: Adjust.
8302 * config/rs6000/power7.md: Adjust.
8303 * config/rs6000/power8.md: Adjust.
8304 * config/rs6000/power9.md: Adjust.
8305 * config/rs6000/rs64.md: Adjust.
8306 * config/rs6000/titan.md: Adjust.
8308 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8310 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8311 ix86_indirect_branch_register logic.
8313 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8315 * config/i386/constraints.md (Bs): Update
8316 ix86_indirect_branch_register check. Don't check
8317 ix86_indirect_branch_register with GOT_memory_operand.
8319 * config/i386/predicates.md (GOT_memory_operand): Don't check
8320 ix86_indirect_branch_register here.
8321 (GOT32_symbol_operand): Likewise.
8323 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8325 * config/i386/predicates.md (constant_call_address_operand):
8326 Rewrite ix86_indirect_branch_register logic.
8327 (sibcall_insn_operand): Likewise.
8329 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8331 * config/i386/constraints.md (Bs): Replace
8332 ix86_indirect_branch_thunk_register with
8333 ix86_indirect_branch_register.
8335 * config/i386/i386.md (indirect_jump): Likewise.
8336 (tablejump): Likewise.
8337 (*sibcall_memory): Likewise.
8338 (*sibcall_value_memory): Likewise.
8339 Peepholes of indirect call and jump via memory: Likewise.
8340 * config/i386/i386.opt: Likewise.
8341 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8342 (GOT_memory_operand): Likewise.
8343 (call_insn_operand): Likewise.
8344 (sibcall_insn_operand): Likewise.
8345 (GOT32_symbol_operand): Likewise.
8347 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8350 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8351 type rather than type addr's type points to.
8352 (expand_omp_atomic_mutex): Likewise.
8353 (expand_omp_atomic): Likewise.
8355 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8358 * config/i386/i386.c (output_indirect_thunk_function): Use
8359 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8360 for __x86_return_thunk.
8362 2018-01-15 Richard Biener <rguenther@suse.de>
8365 * expmed.c (extract_bit_field_1): Fix typo.
8367 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8370 * config/arm/iterators.md (VF): New mode iterator.
8371 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8372 Remove integer-related logic from pattern.
8373 (neon_vabd<mode>_3): Likewise.
8375 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8378 * common.opt (fstrict-overflow): No longer an alias.
8379 (fwrapv-pointer): New option.
8380 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8381 also for pointer types based on flag_wrapv_pointer.
8382 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8383 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8384 opts->x_flag_wrapv got set.
8385 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8386 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8387 POINTER_TYPE_OVERFLOW_UNDEFINED.
8388 * match.pd: Likewise in address comparison pattern.
8389 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8391 2018-01-15 Richard Biener <rguenther@suse.de>
8394 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8395 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8396 Reset type names to their identifier if their TYPE_DECL doesn't
8397 have linkage (and thus is used for ODR and devirt).
8398 (save_debug_info_for_decl): Remove.
8399 (save_debug_info_for_type): Likewise.
8400 (add_tree_to_fld_list): Adjust.
8401 * tree-pretty-print.c (dump_generic_node): Make dumping of
8402 type names more robust.
8404 2018-01-15 Richard Biener <rguenther@suse.de>
8406 * BASE-VER: Bump to 8.0.1.
8408 2018-01-14 Martin Sebor <msebor@redhat.com>
8411 * builtins.c (check_access): Avoid warning when the no-warning bit
8414 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8416 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8417 * ira-color (allocno_hard_regs_compare): Likewise.
8419 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8422 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8423 Use .pushsection/.popsection.
8425 2018-01-14 Martin Sebor <msebor@redhat.com>
8428 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8430 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8432 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8433 entry from extra_headers.
8434 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8435 extra_headers, make the list bitwise identical to the i?86-*-* one.
8437 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8439 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8440 -mcmodel=large with -mindirect-branch=thunk,
8441 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8442 -mfunction-return=thunk-extern.
8443 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8444 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8445 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8447 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8449 * config/i386/i386.c (print_reg): Print the name of the full
8450 integer register without '%'.
8451 (ix86_print_operand): Handle 'V'.
8452 * doc/extend.texi: Document 'V' modifier.
8454 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8456 * config/i386/constraints.md (Bs): Disallow memory operand for
8457 -mindirect-branch-register.
8459 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8460 (GOT_memory_operand): Likewise.
8461 (call_insn_operand): Likewise.
8462 (sibcall_insn_operand): Likewise.
8463 (GOT32_symbol_operand): Likewise.
8464 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8465 for -mindirect-branch-register.
8466 (tablejump): Likewise.
8467 (*sibcall_memory): Likewise.
8468 (*sibcall_value_memory): Likewise.
8469 Disallow peepholes of indirect call and jump via memory for
8470 -mindirect-branch-register.
8471 (*call_pop): Replace m with Bw.
8472 (*call_value_pop): Likewise.
8473 (*sibcall_pop_memory): Replace m with Bs.
8474 * config/i386/i386.opt (mindirect-branch-register): New option.
8475 * doc/invoke.texi: Document -mindirect-branch-register option.
8477 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8479 * config/i386/i386-protos.h (ix86_output_function_return): New.
8480 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8481 set function_return_type.
8482 (indirect_thunk_name): Add ret_p to indicate thunk for function
8484 (output_indirect_thunk_function): Pass false to
8485 indirect_thunk_name.
8486 (ix86_output_indirect_branch_via_reg): Likewise.
8487 (ix86_output_indirect_branch_via_push): Likewise.
8488 (output_indirect_thunk_function): Create alias for function
8489 return thunk if regno < 0.
8490 (ix86_output_function_return): New function.
8491 (ix86_handle_fndecl_attribute): Handle function_return.
8492 (ix86_attribute_table): Add function_return.
8493 * config/i386/i386.h (machine_function): Add
8494 function_return_type.
8495 * config/i386/i386.md (simple_return_internal): Use
8496 ix86_output_function_return.
8497 (simple_return_internal_long): Likewise.
8498 * config/i386/i386.opt (mfunction-return=): New option.
8499 (indirect_branch): Mention -mfunction-return=.
8500 * doc/extend.texi: Document function_return function attribute.
8501 * doc/invoke.texi: Document -mfunction-return= option.
8503 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8505 * config/i386/i386-opts.h (indirect_branch): New.
8506 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8507 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8508 with local indirect jump when converting indirect call and jump.
8509 (ix86_set_indirect_branch_type): New.
8510 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8511 (indirectlabelno): New.
8512 (indirect_thunk_needed): Likewise.
8513 (indirect_thunk_bnd_needed): Likewise.
8514 (indirect_thunks_used): Likewise.
8515 (indirect_thunks_bnd_used): Likewise.
8516 (INDIRECT_LABEL): Likewise.
8517 (indirect_thunk_name): Likewise.
8518 (output_indirect_thunk): Likewise.
8519 (output_indirect_thunk_function): Likewise.
8520 (ix86_output_indirect_branch_via_reg): Likewise.
8521 (ix86_output_indirect_branch_via_push): Likewise.
8522 (ix86_output_indirect_branch): Likewise.
8523 (ix86_output_indirect_jmp): Likewise.
8524 (ix86_code_end): Call output_indirect_thunk_function if needed.
8525 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8527 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8528 (ix86_attribute_table): Add indirect_branch.
8529 * config/i386/i386.h (machine_function): Add indirect_branch_type
8530 and has_local_indirect_jump.
8531 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8533 (tablejump): Likewise.
8534 (*indirect_jump): Use ix86_output_indirect_jmp.
8535 (*tablejump_1): Likewise.
8536 (simple_return_indirect_internal): Likewise.
8537 * config/i386/i386.opt (mindirect-branch=): New option.
8538 (indirect_branch): New.
8541 (thunk-inline): Likewise.
8542 (thunk-extern): Likewise.
8543 * doc/extend.texi: Document indirect_branch function attribute.
8544 * doc/invoke.texi: Document -mindirect-branch= option.
8546 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8549 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8551 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8553 * ipa-inline.c (want_inline_small_function_p): Return false if
8554 inlining has already failed with CIF_FINAL_ERROR.
8555 (update_caller_keys): Call want_inline_small_function_p before
8557 (update_callee_keys): Likewise.
8559 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8561 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8563 (rs6000_quadword_masked_address_p): Likewise.
8564 (quad_aligned_load_p): Likewise.
8565 (quad_aligned_store_p): Likewise.
8566 (const_load_sequence_p): Add comment to describe the outer-most loop.
8567 (mimic_memory_attributes_and_flags): New function.
8568 (rs6000_gen_stvx): Likewise.
8569 (replace_swapped_aligned_store): Likewise.
8570 (rs6000_gen_lvx): Likewise.
8571 (replace_swapped_aligned_load): Likewise.
8572 (replace_swapped_load_constant): Capitalize argument name in
8573 comment describing this function.
8574 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8575 and stores that access quad-word aligned addresses and replace
8576 with stvx or lvx instructions when appropriate.
8577 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8578 New function prototype.
8579 (rs6000_quadword_masked_address_p): Likewise.
8580 (rs6000_gen_lvx): Likewise.
8581 (rs6000_gen_stvx): Likewise.
8582 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8583 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8584 when memory address is aligned.
8585 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8586 this split to select lvx instruction when memory address is aligned.
8587 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8588 instruction when memory address is aligned.
8589 (*vsx_le_perm_load_v16qi): Likewise.
8590 (four unnamed splitters): Modify to select the stvx instruction
8591 when memory is aligned.
8593 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8595 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8596 which appears in the queue multiple times.
8598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8599 Alan Hayward <alan.hayward@arm.com>
8600 David Sherwood <david.sherwood@arm.com>
8602 * tree-vectorizer.h (vec_lower_bound): New structure.
8603 (_loop_vec_info): Add check_nonzero and lower_bounds.
8604 (LOOP_VINFO_CHECK_NONZERO): New macro.
8605 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8606 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8607 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8608 fields. Make seg_len the distance travelled, not including the
8610 (dr_direction_indicator): Declare.
8611 (dr_zero_step_indicator): Likewise.
8612 (dr_known_forward_stride_p): Likewise.
8613 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8615 (runtime_alias_check_p): Allow runtime alias checks with
8617 (operator ==): Compare access_size and align.
8618 (prune_runtime_alias_test_list): Rework for new distinction between
8619 the access_size and seg_len.
8620 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8622 (get_segment_min_max): New function.
8623 (create_intersect_range_checks): Use it.
8624 (dr_step_indicator): New function.
8625 (dr_direction_indicator): Likewise.
8626 (dr_zero_step_indicator): Likewise.
8627 (dr_known_forward_stride_p): Likewise.
8628 * tree-loop-distribution.c (data_ref_segment_size): Return
8629 DR_STEP * (niters - 1).
8630 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8632 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8633 (vect_preserves_scalar_order_p): New function, split out from...
8634 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8635 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8636 (vect_vfa_access_size): New function.
8637 (vect_vfa_align): Likewise.
8638 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8639 (dump_lower_bound): New function.
8640 (vect_check_lower_bound): Likewise.
8641 (vect_small_gap_p): Likewise.
8642 (vectorizable_with_step_bound_p): Likewise.
8643 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8644 depencies if the vectorization factor is 1. Convert the checks
8645 for nonzero steps into checks on the bounds of DR_STEP. Try using
8646 a bunds check for variable steps if the minimum required step is
8647 relatively small. Update calls to the dr_with_seg_len
8648 constructor and to vect_compile_time_alias.
8649 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8651 (vect_loop_versioning): Call it.
8652 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8654 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8657 Alan Hayward <alan.hayward@arm.com>
8658 David Sherwood <david.sherwood@arm.com>
8660 * doc/sourcebuild.texi (vect_scatter_store): Document.
8661 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8663 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8665 * genopinit.c (main): Add supports_vec_scatter_store and
8666 supports_vec_scatter_store_cached to target_optabs.
8667 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8668 IFN_MASK_SCATTER_STORE.
8669 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8671 * internal-fn.h (internal_store_fn_p): Declare.
8672 (internal_fn_stored_value_index): Likewise.
8673 * internal-fn.c (scatter_store_direct): New macro.
8674 (expand_scatter_store_optab_fn): New function.
8675 (direct_scatter_store_optab_supported_p): New macro.
8676 (internal_store_fn_p): New function.
8677 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8678 IFN_MASK_SCATTER_STORE.
8679 (internal_fn_mask_index): Likewise.
8680 (internal_fn_stored_value_index): New function.
8681 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8683 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8684 * optabs-query.c (supports_vec_scatter_store_p): New function.
8685 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8686 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8687 true for scatter stores.
8688 (vect_gather_scatter_fn_p): Handle scatter stores too.
8689 (vect_check_gather_scatter): Consider using scatter stores if
8690 supports_vec_scatter_store_p.
8691 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8693 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8694 internal_fn_stored_value_index.
8695 (check_load_store_masking): Handle scatter stores too.
8696 (vect_get_store_rhs): Make public.
8697 (vectorizable_call): Use internal_store_fn_p.
8698 (vectorizable_store): Handle scatter store internal functions.
8699 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8700 when deciding whether the end of the group has been reached.
8701 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8702 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8703 (mask_scatter_store<mode>): New insns.
8705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8706 Alan Hayward <alan.hayward@arm.com>
8707 David Sherwood <david.sherwood@arm.com>
8709 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8710 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8711 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8713 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8714 Use vect_truncate_gather_scatter_offset if we can't treat the
8715 operation as a normal gather load or scatter store.
8716 (get_group_load_store_type): Take the gather_scatter_info
8717 as argument. Try using a gather load or scatter store for
8718 single-element groups.
8719 (get_load_store_type): Update calls to get_group_load_store_type
8720 and vect_use_strided_gather_scatters_p.
8722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8723 Alan Hayward <alan.hayward@arm.com>
8724 David Sherwood <david.sherwood@arm.com>
8726 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8727 optional tree argument.
8728 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8730 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8731 but continue to use the current value as a fallback.
8732 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8733 to compare the updates.
8734 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8735 (get_load_store_type): Use it when handling a strided access.
8736 (vect_get_strided_load_store_ops): New function.
8737 (vect_get_data_ptr_increment): Likewise.
8738 (vectorizable_load): Handle strided gather loads. Always pass
8739 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8742 Alan Hayward <alan.hayward@arm.com>
8743 David Sherwood <david.sherwood@arm.com>
8745 * doc/md.texi (gather_load@var{m}): Document.
8746 (mask_gather_load@var{m}): Likewise.
8747 * genopinit.c (main): Add supports_vec_gather_load and
8748 supports_vec_gather_load_cached to target_optabs.
8749 * optabs-tree.c (init_tree_optimization_optabs): Use
8750 ggc_cleared_alloc to allocate target_optabs.
8751 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8752 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8754 * internal-fn.h (internal_load_fn_p): Declare.
8755 (internal_gather_scatter_fn_p): Likewise.
8756 (internal_fn_mask_index): Likewise.
8757 (internal_gather_scatter_fn_supported_p): Likewise.
8758 * internal-fn.c (gather_load_direct): New macro.
8759 (expand_gather_load_optab_fn): New function.
8760 (direct_gather_load_optab_supported_p): New macro.
8761 (direct_internal_fn_optab): New function.
8762 (internal_load_fn_p): Likewise.
8763 (internal_gather_scatter_fn_p): Likewise.
8764 (internal_fn_mask_index): Likewise.
8765 (internal_gather_scatter_fn_supported_p): Likewise.
8766 * optabs-query.c (supports_at_least_one_mode_p): New function.
8767 (supports_vec_gather_load_p): Likewise.
8768 * optabs-query.h (supports_vec_gather_load_p): Declare.
8769 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8770 and memory_type field.
8771 (NUM_PATTERNS): Bump to 15.
8772 * tree-vect-data-refs.c: Include internal-fn.h.
8773 (vect_gather_scatter_fn_p): New function.
8774 (vect_describe_gather_scatter_call): Likewise.
8775 (vect_check_gather_scatter): Try using internal functions for
8776 gather loads. Recognize existing calls to a gather load function.
8777 (vect_analyze_data_refs): Consider using gather loads if
8778 supports_vec_gather_load_p.
8779 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8780 (vect_get_gather_scatter_offset_type): Likewise.
8781 (vect_convert_mask_for_vectype): Likewise.
8782 (vect_add_conversion_to_patterm): Likewise.
8783 (vect_try_gather_scatter_pattern): Likewise.
8784 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8785 (vect_vect_recog_func_ptrs): Add it.
8786 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8787 internal_fn_mask_index and internal_gather_scatter_fn_p.
8788 (check_load_store_masking): Take the gather_scatter_info as an
8789 argument and handle gather loads.
8790 (vect_get_gather_scatter_ops): New function.
8791 (vectorizable_call): Check internal_load_fn_p.
8792 (vectorizable_load): Likewise. Handle gather load internal
8794 (vectorizable_store): Update call to check_load_store_masking.
8795 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8796 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8797 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8798 (aarch64_gather_scale_operand_d): New predicates.
8799 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8800 (mask_gather_load<mode>): New insns.
8802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8803 Alan Hayward <alan.hayward@arm.com>
8804 David Sherwood <david.sherwood@arm.com>
8806 * optabs.def (fold_left_plus_optab): New optab.
8807 * doc/md.texi (fold_left_plus_@var{m}): Document.
8808 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8809 * internal-fn.c (fold_left_direct): Define.
8810 (expand_fold_left_optab_fn): Likewise.
8811 (direct_fold_left_optab_supported_p): Likewise.
8812 * fold-const-call.c (fold_const_fold_left): New function.
8813 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8814 * tree-parloops.c (valid_reduction_p): New function.
8815 (gather_scalar_reductions): Use it.
8816 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8817 (vect_finish_replace_stmt): Declare.
8818 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8819 (needs_fold_left_reduction_p): New function, split out from...
8820 (vect_is_simple_reduction): ...here. Accept reductions that
8821 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8822 (vect_force_simple_reduction): Also store the reduction type in
8823 the assignment's STMT_VINFO_REDUC_TYPE.
8824 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8825 (merge_with_identity): New function.
8826 (vect_expand_fold_left): Likewise.
8827 (vectorize_fold_left_reduction): Likewise.
8828 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8829 scalar phi in place for it. Check for target support and reject
8830 cases that would reassociate the operation. Defer the transform
8831 phase to vectorize_fold_left_reduction.
8832 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8833 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8834 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8838 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8839 call to ifc_temp_var.
8841 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8842 Alan Hayward <alan.hayward@arm.com>
8843 David Sherwood <david.sherwood@arm.com>
8845 * target.def (legitimize_address_displacement): Take the original
8846 offset as a poly_int.
8847 * targhooks.h (default_legitimize_address_displacement): Update
8849 * targhooks.c (default_legitimize_address_displacement): Likewise.
8850 * doc/tm.texi: Regenerate.
8851 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8852 as an argument, moving assert of ad->disp == ad->disp_term to...
8853 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8854 Try calling targetm.legitimize_address_displacement before expanding
8855 the address rather than afterwards, and adjust for the new interface.
8856 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8857 Match the new hook interface. Handle SVE addresses.
8858 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8863 * Makefile.in (OBJS): Add early-remat.o.
8864 * target.def (select_early_remat_modes): New hook.
8865 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8866 * doc/tm.texi: Regenerate.
8867 * targhooks.h (default_select_early_remat_modes): Declare.
8868 * targhooks.c (default_select_early_remat_modes): New function.
8869 * timevar.def (TV_EARLY_REMAT): New timevar.
8870 * passes.def (pass_early_remat): New pass.
8871 * tree-pass.h (make_pass_early_remat): Declare.
8872 * early-remat.c: New file.
8873 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8875 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8878 Alan Hayward <alan.hayward@arm.com>
8879 David Sherwood <david.sherwood@arm.com>
8881 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8882 vfm1 with a bound_epilog parameter.
8883 (vect_do_peeling): Update calls accordingly, and move the prologue
8884 call earlier in the function. Treat the base bound_epilog as 0 for
8885 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8886 this base when peeling for gaps.
8887 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8888 with fully-masked loops.
8889 (vect_estimate_min_profitable_iters): Handle the single peeled
8890 iteration in that case.
8892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8893 Alan Hayward <alan.hayward@arm.com>
8894 David Sherwood <david.sherwood@arm.com>
8896 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8897 single-element interleaving even if the size is not a power of 2.
8898 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8899 accesses for single-element interleaving if the group size is
8902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8903 Alan Hayward <alan.hayward@arm.com>
8904 David Sherwood <david.sherwood@arm.com>
8906 * doc/md.texi (fold_extract_last_@var{m}): Document.
8907 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8908 * optabs.def (fold_extract_last_optab): New optab.
8909 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8910 * internal-fn.c (fold_extract_direct): New macro.
8911 (expand_fold_extract_optab_fn): Likewise.
8912 (direct_fold_extract_optab_supported_p): Likewise.
8913 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8914 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8915 EXTRACT_LAST_REDUCTION.
8916 (get_initial_def_for_reduction): Do not create an initial vector
8917 for EXTRACT_LAST_REDUCTION reductions.
8918 (vectorizable_reduction): Leave the scalar phi in place for
8919 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8920 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8921 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8922 transform phase to vectorizable_condition.
8923 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8925 (vect_finish_stmt_generation): ...here.
8926 (vect_finish_replace_stmt): New function.
8927 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8928 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8930 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8933 Alan Hayward <alan.hayward@arm.com>
8934 David Sherwood <david.sherwood@arm.com>
8936 * doc/md.texi (extract_last_@var{m}): Document.
8937 * optabs.def (extract_last_optab): New optab.
8938 * internal-fn.def (EXTRACT_LAST): New internal function.
8939 * internal-fn.c (cond_unary_direct): New macro.
8940 (expand_cond_unary_optab_fn): Likewise.
8941 (direct_cond_unary_optab_supported_p): Likewise.
8942 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8943 loops using EXTRACT_LAST.
8944 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8945 (extract_last_<mode>): ...this optab.
8946 (vec_extract<mode><Vel>): Update accordingly.
8948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8949 Alan Hayward <alan.hayward@arm.com>
8950 David Sherwood <david.sherwood@arm.com>
8952 * target.def (empty_mask_is_expensive): New hook.
8953 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8954 * doc/tm.texi: Regenerate.
8955 * targhooks.h (default_empty_mask_is_expensive): Declare.
8956 * targhooks.c (default_empty_mask_is_expensive): New function.
8957 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8958 if the target says that empty masks are expensive.
8959 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8961 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8964 Alan Hayward <alan.hayward@arm.com>
8965 David Sherwood <david.sherwood@arm.com>
8967 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8968 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8969 (vect_use_loop_mask_for_alignment_p): New function.
8970 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8971 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8972 niters_skip argument. Make sure that the first niters_skip elements
8973 of the first iteration are inactive.
8974 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8975 Update call to vect_set_loop_masks_directly.
8976 (get_misalign_in_elems): New function, split out from...
8977 (vect_gen_prolog_loop_niters): ...here.
8978 (vect_update_init_of_dr): Take a code argument that specifies whether
8979 the adjustment should be added or subtracted.
8980 (vect_update_init_of_drs): Likewise.
8981 (vect_prepare_for_masked_peels): New function.
8982 (vect_do_peeling): Skip prologue peeling if we're using a mask
8983 instead. Update call to vect_update_inits_of_drs.
8984 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8986 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8987 alignment. Do not include the number of peeled iterations in
8988 the minimum threshold in that case.
8989 (vectorizable_induction): Adjust the start value down by
8990 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8991 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8992 Take the number of skipped iterations into account when calculating
8994 * tree-vect-stmts.c (vect_gen_while_not): New function.
8996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8997 Alan Hayward <alan.hayward@arm.com>
8998 David Sherwood <david.sherwood@arm.com>
9000 * doc/sourcebuild.texi (vect_fully_masked): Document.
9001 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9003 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9005 (vect_analyze_loop_2): ...here. Don't check the vectorization
9006 factor against the number of loop iterations if the loop is
9009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9010 Alan Hayward <alan.hayward@arm.com>
9011 David Sherwood <david.sherwood@arm.com>
9013 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9014 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9015 (dump_groups): Update accordingly.
9016 (iv_use::mem_type): New member variable.
9017 (address_p): New function.
9018 (record_use): Add a mem_type argument and initialize the new
9020 (record_group_use): Add a mem_type argument. Use address_p.
9021 Remove obsolete null checks of base_object. Update call to record_use.
9022 (find_interesting_uses_op): Update call to record_group_use.
9023 (find_interesting_uses_cond): Likewise.
9024 (find_interesting_uses_address): Likewise.
9025 (get_mem_type_for_internal_fn): New function.
9026 (find_address_like_use): Likewise.
9027 (find_interesting_uses_stmt): Try find_address_like_use before
9028 calling find_interesting_uses_op.
9029 (addr_offset_valid_p): Use the iv mem_type field as the type
9030 of the addressed memory.
9031 (add_autoinc_candidates): Likewise.
9032 (get_address_cost): Likewise.
9033 (split_small_address_groups_p): Use address_p.
9034 (split_address_groups): Likewise.
9035 (add_iv_candidate_for_use): Likewise.
9036 (autoinc_possible_for_pair): Likewise.
9037 (rewrite_groups): Likewise.
9038 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9039 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9040 (get_alias_ptr_type_for_ptr_address): New function.
9041 (rewrite_use_address): Rewrite address uses in calls that were
9042 identified by find_address_like_use.
9044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9045 Alan Hayward <alan.hayward@arm.com>
9046 David Sherwood <david.sherwood@arm.com>
9048 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9050 * gimple-expr.h (is_gimple_addressable: Likewise.
9051 * gimple-expr.c (is_gimple_address): Likewise.
9052 * internal-fn.c (expand_call_mem_ref): New function.
9053 (expand_mask_load_optab_fn): Use it.
9054 (expand_mask_store_optab_fn): Likewise.
9056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9057 Alan Hayward <alan.hayward@arm.com>
9058 David Sherwood <david.sherwood@arm.com>
9060 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9061 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9062 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9063 (cond_umax@var{mode}): Document.
9064 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9065 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9066 (cond_umin_optab, cond_umax_optab): New optabs.
9067 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9068 (COND_IOR, COND_XOR): New internal functions.
9069 * internal-fn.h (get_conditional_internal_fn): Declare.
9070 * internal-fn.c (cond_binary_direct): New macro.
9071 (expand_cond_binary_optab_fn): Likewise.
9072 (direct_cond_binary_optab_supported_p): Likewise.
9073 (get_conditional_internal_fn): New function.
9074 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9075 Cope with reduction statements that are vectorized as calls rather
9077 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9078 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9079 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9080 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9081 (UNSPEC_COND_EOR): New unspecs.
9082 (optab): Add mappings for them.
9083 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9084 (sve_int_op, sve_fp_op): New int attributes.
9086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9087 Alan Hayward <alan.hayward@arm.com>
9088 David Sherwood <david.sherwood@arm.com>
9090 * optabs.def (while_ult_optab): New optab.
9091 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9092 * internal-fn.def (WHILE_ULT): New internal function.
9093 * internal-fn.h (direct_internal_fn_supported_p): New override
9094 that takes two types as argument.
9095 * internal-fn.c (while_direct): New macro.
9096 (expand_while_optab_fn): New function.
9097 (convert_optab_supported_p): Likewise.
9098 (direct_while_optab_supported_p): New macro.
9099 * wide-int.h (wi::udiv_ceil): New function.
9100 * tree-vectorizer.h (rgroup_masks): New structure.
9101 (vec_loop_masks): New typedef.
9102 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9104 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9105 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9106 (vect_max_vf): New function.
9107 (slpeel_make_loop_iterate_ntimes): Delete.
9108 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9109 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9110 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9111 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9112 internal-fn.h, stor-layout.h and optabs-query.h.
9113 (vect_set_loop_mask): New function.
9114 (add_preheader_seq): Likewise.
9115 (add_header_seq): Likewise.
9116 (interleave_supported_p): Likewise.
9117 (vect_maybe_permute_loop_masks): Likewise.
9118 (vect_set_loop_masks_directly): Likewise.
9119 (vect_set_loop_condition_masked): Likewise.
9120 (vect_set_loop_condition_unmasked): New function, split out from
9121 slpeel_make_loop_iterate_ntimes.
9122 (slpeel_make_loop_iterate_ntimes): Rename to..
9123 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9124 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9125 (vect_do_peeling): Update call accordingly.
9126 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9128 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9129 mask_compare_type, can_fully_mask_p and fully_masked_p.
9130 (release_vec_loop_masks): New function.
9131 (_loop_vec_info): Use it to free the loop masks.
9132 (can_produce_all_loop_masks_p): New function.
9133 (vect_get_max_nscalars_per_iter): Likewise.
9134 (vect_verify_full_masking): Likewise.
9135 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9136 retries, and free the mask rgroups before retrying. Check loop-wide
9137 reasons for disallowing fully-masked loops. Make the final decision
9138 about whether use a fully-masked loop or not.
9139 (vect_estimate_min_profitable_iters): Do not assume that peeling
9140 for the number of iterations will be needed for fully-masked loops.
9141 (vectorizable_reduction): Disable fully-masked loops.
9142 (vectorizable_live_operation): Likewise.
9143 (vect_halve_mask_nunits): New function.
9144 (vect_double_mask_nunits): Likewise.
9145 (vect_record_loop_mask): Likewise.
9146 (vect_get_loop_mask): Likewise.
9147 (vect_transform_loop): Handle the case in which the final loop
9148 iteration might handle a partial vector. Call vect_set_loop_condition
9149 instead of slpeel_make_loop_iterate_ntimes.
9150 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9151 (check_load_store_masking): New function.
9152 (prepare_load_store_mask): Likewise.
9153 (vectorizable_store): Handle fully-masked loops.
9154 (vectorizable_load): Likewise.
9155 (supportable_widening_operation): Use vect_halve_mask_nunits for
9157 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9158 (vect_gen_while): New function.
9159 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9160 (aarch64_uqdec<mode>): New insn.
9162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9163 Alan Hayward <alan.hayward@arm.com>
9164 David Sherwood <david.sherwood@arm.com>
9166 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9167 (reduc_xor_scal_optab): New optabs.
9168 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9169 (reduc_xor_scal_@var{m}): Document.
9170 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9171 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9173 * fold-const-call.c (fold_const_call): Handle them.
9174 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9175 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9176 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9177 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9178 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9179 (UNSPEC_XORV): New unspecs.
9180 (optab): Add entries for them.
9181 (BITWISEV): New int iterator.
9182 (bit_reduc_op): New int attributes.
9184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9185 Alan Hayward <alan.hayward@arm.com>
9186 David Sherwood <david.sherwood@arm.com>
9188 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9189 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9190 * optabs.def (vec_shl_insert_optab): New optab.
9191 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9192 (duplicate_and_interleave): Likewise.
9193 * tree-vect-loop.c: Include internal-fn.h.
9194 (neutral_op_for_slp_reduction): New function, split out from
9195 get_initial_defs_for_reduction.
9196 (get_initial_def_for_reduction): Handle option 2 for variable-length
9197 vectors by loading the neutral value into a vector and then shifting
9198 the initial value into element 0.
9199 (get_initial_defs_for_reduction): Replace the code argument with
9200 the neutral value calculated by neutral_op_for_slp_reduction.
9201 Use gimple_build_vector for constant-length vectors.
9202 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9203 but the first group_size elements have a neutral value.
9204 Use duplicate_and_interleave otherwise.
9205 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9206 Update call to get_initial_defs_for_reduction. Handle SLP
9207 reductions for variable-length vectors by creating one vector
9208 result for each scalar result, with the elements associated
9209 with other scalar results stubbed out with the neutral value.
9210 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9211 Require IFN_VEC_SHL_INSERT for double reductions on
9212 variable-length vectors, or SLP reductions that have
9213 a neutral value. Require can_duplicate_and_interleave_p
9214 support for variable-length unchained SLP reductions if there
9215 is no neutral value, such as for MIN/MAX reductions. Also require
9216 the number of vector elements to be a multiple of the number of
9217 SLP statements when doing variable-length unchained SLP reductions.
9218 Update call to vect_create_epilog_for_reduction.
9219 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9220 and remove initial values.
9221 (duplicate_and_interleave): Make public.
9222 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9223 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9226 Alan Hayward <alan.hayward@arm.com>
9227 David Sherwood <david.sherwood@arm.com>
9229 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9230 (can_duplicate_and_interleave_p): New function.
9231 (vect_get_and_check_slp_defs): Take the vector of statements
9232 rather than just the current one. Remove excess parentheses.
9233 Restriction rejectinon of vect_constant_def and vect_external_def
9234 for variable-length vectors to boolean types, or types for which
9235 can_duplicate_and_interleave_p is false.
9236 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9237 (duplicate_and_interleave): New function.
9238 (vect_get_constant_vectors): Use gimple_build_vector for
9239 constant-length vectors and suitable variable-length constant
9240 vectors. Use duplicate_and_interleave for other variable-length
9241 vectors. Don't defer the update when inserting new statements.
9243 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9244 Alan Hayward <alan.hayward@arm.com>
9245 David Sherwood <david.sherwood@arm.com>
9247 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9248 min_profitable_iters doesn't go negative.
9250 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9251 Alan Hayward <alan.hayward@arm.com>
9252 David Sherwood <david.sherwood@arm.com>
9254 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9255 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9256 * optabs.def (vec_mask_load_lanes_optab): New optab.
9257 (vec_mask_store_lanes_optab): Likewise.
9258 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9259 (MASK_STORE_LANES): Likewise.
9260 * internal-fn.c (mask_load_lanes_direct): New macro.
9261 (mask_store_lanes_direct): Likewise.
9262 (expand_mask_load_optab_fn): Handle masked operations.
9263 (expand_mask_load_lanes_optab_fn): New macro.
9264 (expand_mask_store_optab_fn): Handle masked operations.
9265 (expand_mask_store_lanes_optab_fn): New macro.
9266 (direct_mask_load_lanes_optab_supported_p): Likewise.
9267 (direct_mask_store_lanes_optab_supported_p): Likewise.
9268 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9270 (vect_load_lanes_supported): Likewise.
9271 * tree-vect-data-refs.c (strip_conversion): New function.
9272 (can_group_stmts_p): Likewise.
9273 (vect_analyze_data_ref_accesses): Use it instead of checking
9274 for a pair of assignments.
9275 (vect_store_lanes_supported): Take a masked_p parameter.
9276 (vect_load_lanes_supported): Likewise.
9277 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9278 vect_store_lanes_supported and vect_load_lanes_supported.
9279 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9280 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9281 parameter. Don't allow gaps for masked accesses.
9282 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9283 and vect_load_lanes_supported.
9284 (get_load_store_type): Take a masked_p parameter and update
9285 call to get_group_load_store_type.
9286 (vectorizable_store): Update call to get_load_store_type.
9287 Handle IFN_MASK_STORE_LANES.
9288 (vectorizable_load): Update call to get_load_store_type.
9289 Handle IFN_MASK_LOAD_LANES.
9291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9292 Alan Hayward <alan.hayward@arm.com>
9293 David Sherwood <david.sherwood@arm.com>
9295 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9297 * config/aarch64/aarch64-protos.h
9298 (aarch64_sve_struct_memory_operand_p): Declare.
9299 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9300 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9301 (VPRED, vpred): Handle SVE structure modes.
9302 * config/aarch64/constraints.md (Utx): New constraint.
9303 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9304 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9305 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9306 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9307 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9308 structure modes. Split into pieces after RA.
9309 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9310 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9312 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9313 SVE structure modes.
9314 (aarch64_classify_address): Likewise.
9315 (sizetochar): Move earlier in file.
9316 (aarch64_print_operand): Handle SVE register lists.
9317 (aarch64_array_mode): New function.
9318 (aarch64_sve_struct_memory_operand_p): Likewise.
9319 (TARGET_ARRAY_MODE): Redefine.
9321 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9322 Alan Hayward <alan.hayward@arm.com>
9323 David Sherwood <david.sherwood@arm.com>
9325 * target.def (array_mode): New target hook.
9326 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9327 * doc/tm.texi: Regenerate.
9328 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9329 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9330 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9332 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9336 Alan Hayward <alan.hayward@arm.com>
9337 David Sherwood <david.sherwood@arm.com>
9339 * fold-const.c (fold_binary_loc): Check the argument types
9340 rather than the result type when testing for a vector operation.
9342 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9344 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9345 * doc/tm.texi: Regenerate.
9347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9348 Alan Hayward <alan.hayward@arm.com>
9349 David Sherwood <david.sherwood@arm.com>
9351 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9352 (sve): Document new AArch64 extension.
9353 * doc/md.texi (w): Extend the description of the AArch64
9354 constraint to include SVE vectors.
9355 (Upl, Upa): Document new AArch64 predicate constraints.
9356 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9358 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9359 (msve-vector-bits=): New option.
9360 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9361 SVE when these are disabled.
9362 (sve): New extension.
9363 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9364 modes. Adjust their number of units based on aarch64_sve_vg.
9365 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9366 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9367 aarch64_addr_query_type.
9368 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9369 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9370 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9371 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9372 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9373 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9374 (aarch64_simd_imm_zero_p): Delete.
9375 (aarch64_check_zero_based_sve_index_immediate): Declare.
9376 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9377 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9378 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9379 (aarch64_sve_float_mul_immediate_p): Likewise.
9380 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9382 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9383 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9384 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9385 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9386 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9387 (aarch64_regmode_natural_size): Likewise.
9388 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9389 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9391 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9392 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9393 for VG and the SVE predicate registers.
9394 (V_ALIASES): Add a "z"-prefixed alias.
9395 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9396 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9397 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9398 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9399 (REG_CLASS_NAMES): Add entries for them.
9400 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9401 and the predicate registers.
9402 (aarch64_sve_vg): Declare.
9403 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9404 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9405 (REGMODE_NATURAL_SIZE): Define.
9406 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9408 * config/aarch64/aarch64.c: Include cfgrtl.h.
9409 (simd_immediate_info): Add a constructor for series vectors,
9410 and an associated step field.
9411 (aarch64_sve_vg): New variable.
9412 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9413 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9414 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9415 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9416 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9417 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9418 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9419 (aarch64_get_mask_mode): New functions.
9420 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9421 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9422 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9423 predicate modes and predicate registers. Explicitly restrict
9424 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9425 to store a vector mode if it is recognized by
9426 aarch64_classify_vector_mode.
9427 (aarch64_regmode_natural_size): New function.
9428 (aarch64_hard_regno_caller_save_mode): Return the original mode
9430 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9431 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9432 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9433 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9435 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9436 does not overlap dest if the function is frame-related. Handle
9438 (aarch64_split_add_offset): New function.
9439 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9440 them aarch64_add_offset.
9441 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9442 and update call to aarch64_sub_sp.
9443 (aarch64_add_cfa_expression): New function.
9444 (aarch64_expand_prologue): Pass extra temporary registers to the
9445 functions above. Handle the case in which we need to emit new
9446 DW_CFA_expressions for registers that were originally saved
9447 relative to the stack pointer, but now have to be expressed
9448 relative to the frame pointer.
9449 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9451 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9452 IP0 and IP1 values for SVE frames.
9453 (aarch64_expand_vec_series): New function.
9454 (aarch64_expand_sve_widened_duplicate): Likewise.
9455 (aarch64_expand_sve_const_vector): Likewise.
9456 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9457 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9458 into the register, rather than emitting a SET directly.
9459 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9460 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9461 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9462 (offset_9bit_signed_scaled_p): New functions.
9463 (aarch64_replicate_bitmask_imm): New function.
9464 (aarch64_bitmask_imm): Use it.
9465 (aarch64_cannot_force_const_mem): Reject expressions involving
9466 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9467 (aarch64_classify_index): Handle SVE indices, by requiring
9468 a plain register index with a scale that matches the element size.
9469 (aarch64_classify_address): Handle SVE addresses. Assert that
9470 the mode of the address is VOIDmode or an integer mode.
9471 Update call to aarch64_classify_symbol.
9472 (aarch64_classify_symbolic_expression): Update call to
9473 aarch64_classify_symbol.
9474 (aarch64_const_vec_all_in_range_p): New function.
9475 (aarch64_print_vector_float_operand): Likewise.
9476 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9477 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9478 and the FP immediates 1.0 and 0.5.
9479 (aarch64_print_address_internal): Handle SVE addresses.
9480 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9481 (aarch64_regno_regclass): Handle predicate registers.
9482 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9484 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9485 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9486 (aarch64_convert_sve_vector_bits): New function.
9487 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9488 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9490 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9491 Handle SVE vector and predicate modes. Accept VL-based constants
9492 that need only one temporary register, and VL offsets that require
9493 no temporary registers.
9494 (aarch64_conditional_register_usage): Mark the predicate registers
9495 as fixed if SVE isn't available.
9496 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9497 Return true for SVE vector and predicate modes.
9498 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9499 rather than an unsigned int. Handle SVE modes.
9500 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9502 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9504 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9505 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9506 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9507 (aarch64_sve_float_mul_immediate_p): New functions.
9508 (aarch64_sve_valid_immediate): New function.
9509 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9510 Explicitly reject structure modes. Check for INDEX constants.
9511 Handle PTRUE and PFALSE constants.
9512 (aarch64_check_zero_based_sve_index_immediate): New function.
9513 (aarch64_simd_imm_zero_p): Delete.
9514 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9515 vector modes. Accept constants in the range of CNT[BHWD].
9516 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9517 ask for an Advanced SIMD mode.
9518 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9519 (aarch64_simd_vector_alignment): Handle SVE predicates.
9520 (aarch64_vectorize_preferred_vector_alignment): New function.
9521 (aarch64_simd_vector_alignment_reachable): Use it instead of
9523 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9524 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9526 (MAX_VECT_LEN): Delete.
9527 (expand_vec_perm_d): Add a vec_flags field.
9528 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9529 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9530 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9532 (aarch64_evpc_rev): Rename to...
9533 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9534 (aarch64_evpc_rev_global): New function.
9535 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9536 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9538 (aarch64_evpc_sve_tbl): New function.
9539 (aarch64_expand_vec_perm_const_1): Update after rename of
9540 aarch64_evpc_rev. Handle SVE permutes too, trying
9541 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9542 than aarch64_evpc_tbl.
9543 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9544 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9545 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9546 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9547 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9548 (aarch64_expand_sve_vcond): New functions.
9549 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9550 of aarch64_vector_mode_p.
9551 (aarch64_dwarf_poly_indeterminate_value): New function.
9552 (aarch64_compute_pressure_classes): Likewise.
9553 (aarch64_can_change_mode_class): Likewise.
9554 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9555 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9556 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9557 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9558 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9559 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9560 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9561 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9563 (Dn, Dl, Dr): Accept const as well as const_vector.
9564 (Dz): Likewise. Compare against CONST0_RTX.
9565 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9566 of "vector" where appropriate.
9567 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9568 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9569 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9570 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9571 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9572 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9573 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9574 (v_int_equiv): Extend to SVE modes.
9575 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9577 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9578 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9579 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9580 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9581 (SVE_COND_FP_CMP): New int iterators.
9582 (perm_hilo): Handle the new unpack unspecs.
9583 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9585 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9586 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9587 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9588 (aarch64_equality_operator, aarch64_constant_vector_operand)
9589 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9590 (aarch64_sve_nonimmediate_operand): Likewise.
9591 (aarch64_sve_general_operand): Likewise.
9592 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9593 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9594 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9595 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9596 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9597 (aarch64_sve_float_arith_immediate): Likewise.
9598 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9599 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9600 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9601 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9602 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9603 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9604 (aarch64_sve_float_arith_operand): Likewise.
9605 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9606 (aarch64_sve_float_mul_operand): Likewise.
9607 (aarch64_sve_vec_perm_operand): Likewise.
9608 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9609 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9610 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9611 as well as const_vector.
9612 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9613 in file. Use CONST0_RTX and CONSTM1_RTX.
9614 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9615 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9616 Use aarch64_simd_imm_zero.
9617 * config/aarch64/aarch64-sve.md: New file.
9618 * config/aarch64/aarch64.md: Include it.
9619 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9620 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9621 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9622 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9623 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9624 (sve): New attribute.
9625 (enabled): Disable instructions with the sve attribute unless
9627 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9628 aarch64_expand_mov_immediate.
9629 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9630 CNT[BHSD] immediates.
9631 (movti): Split CONST_POLY_INT moves into two halves.
9632 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9633 Split additions that need a temporary here if the destination
9634 is the stack pointer.
9635 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9636 (*add<mode>3_poly_1): New instruction.
9637 (set_clobber_cc): New expander.
9639 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9641 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9642 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9643 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9644 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9645 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9646 Change innermode from fixed_mode_size to machine_mode.
9647 (simplify_subreg): Update call accordingly. Handle a constant-sized
9648 subreg of a variable-length CONST_VECTOR.
9650 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9651 Alan Hayward <alan.hayward@arm.com>
9652 David Sherwood <david.sherwood@arm.com>
9654 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9655 (add_offset_to_base): New function, split out from...
9656 (create_mem_ref): ...here. When handling a scale other than 1,
9657 check first whether the address is valid without the offset.
9658 Add it into the base if so, leaving the index and scale as-is.
9660 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9663 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9664 fold_for_warn before checking if arg2 is INTEGER_CST.
9666 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9668 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9669 (store_multiple_operation): Delete.
9670 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9671 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9672 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9673 guarded by TARGET_STRING.
9674 (rs6000_output_load_multiple): Delete.
9675 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9676 OPTION_MASK_STRING / TARGET_STRING handling.
9677 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9678 (const rs6000_opt_masks) <"string">: Change mask to 0.
9679 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9680 (MASK_STRING): Delete.
9681 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9683 (load_multiple): Delete.
9690 (store_multiple): Delete.
9697 (movmemsi_8reg): Delete.
9698 (corresponding unnamed define_insn): Delete.
9699 (movmemsi_6reg): Delete.
9700 (corresponding unnamed define_insn): Delete.
9701 (movmemsi_4reg): Delete.
9702 (corresponding unnamed define_insn): Delete.
9703 (movmemsi_2reg): Delete.
9704 (corresponding unnamed define_insn): Delete.
9705 (movmemsi_1reg): Delete.
9706 (corresponding unnamed define_insn): Delete.
9707 * config/rs6000/rs6000.opt (mno-string): New.
9708 (mstring): Replace by deprecation warning stub.
9709 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9711 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9713 * regrename.c (regrename_do_replace): If replacing the same
9714 reg multiple times, try to reuse last created gen_raw_REG.
9717 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9718 main to workaround a bug in GDB.
9720 2018-01-12 Tom de Vries <tom@codesourcery.com>
9723 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9725 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9727 PR rtl-optimization/80481
9728 * ira-color.c (get_cap_member): New function.
9729 (allocnos_conflict_by_live_ranges_p): Use it.
9730 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9731 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9733 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9736 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9737 (*saddl_se_1): Ditto.
9739 (*ssubl_se_1): Ditto.
9741 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9743 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9744 rather than wi::to_widest for DR_INITs.
9745 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9746 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9747 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9749 (vect_analyze_group_access_1): Note that here.
9751 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9753 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9754 polynomial type sizes.
9756 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9758 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9759 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9760 (gimple_add_tmp_var): Likewise.
9762 2018-01-12 Martin Liska <mliska@suse.cz>
9764 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9765 (gimple_alloc_sizes): Likewise.
9766 (dump_gimple_statistics): Use PRIu64 in printf format.
9767 * gimple.h: Change uint64_t to int.
9769 2018-01-12 Martin Liska <mliska@suse.cz>
9771 * tree-core.h: Use uint64_t instead of int.
9772 * tree.c (tree_node_counts): Likewise.
9773 (tree_node_sizes): Likewise.
9774 (dump_tree_statistics): Use PRIu64 in printf format.
9776 2018-01-12 Martin Liska <mliska@suse.cz>
9778 * Makefile.in: As qsort_chk is implemented in vec.c, add
9779 vec.o to linkage of gencfn-macros.
9780 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9781 passing the info to record_node_allocation_statistics.
9782 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9784 * ggc-common.c (struct ggc_usage): Add operator== and use
9785 it in operator< and compare function.
9786 * mem-stats.h (struct mem_usage): Likewise.
9787 * vec.c (struct vec_usage): Remove operator< and compare
9788 function. Can be simply inherited.
9790 2018-01-12 Martin Jambor <mjambor@suse.cz>
9793 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9794 * tree-ssa-math-opts.c: Include domwalk.h.
9795 (convert_mult_to_fma_1): New function.
9796 (fma_transformation_info): New type.
9797 (fma_deferring_state): Likewise.
9798 (cancel_fma_deferring): New function.
9799 (result_of_phi): Likewise.
9800 (last_fma_candidate_feeds_initial_phi): Likewise.
9801 (convert_mult_to_fma): Added deferring logic, split actual
9802 transformation to convert_mult_to_fma_1.
9803 (math_opts_dom_walker): New type.
9804 (math_opts_dom_walker::after_dom_children): New method, body moved
9805 here from pass_optimize_widening_mul::execute, added deferring logic
9807 (pass_optimize_widening_mul::execute): Moved most of code to
9808 math_opts_dom_walker::after_dom_children.
9809 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9810 * config/i386/i386.c (ix86_option_override_internal): Added
9811 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9813 2018-01-12 Richard Biener <rguenther@suse.de>
9816 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9817 inline instance vars.
9819 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9822 * config/rx/rx.c (rx_is_restricted_memory_address):
9825 2018-01-12 Richard Biener <rguenther@suse.de>
9827 PR tree-optimization/80846
9828 * target.def (split_reduction): New target hook.
9829 * targhooks.c (default_split_reduction): New function.
9830 * targhooks.h (default_split_reduction): Declare.
9831 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9832 target requests first reduce vectors by combining low and high
9834 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9835 (get_vectype_for_scalar_type_and_size): Export.
9836 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9837 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9838 * doc/tm.texi: Regenerate.
9839 * config/i386/i386.c (ix86_split_reduction): Implement
9840 TARGET_VECTORIZE_SPLIT_REDUCTION.
9842 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9845 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9846 in PIC mode except for TARGET_VXWORKS_RTP.
9847 * config/sparc/sparc.c: Include cfgrtl.h.
9848 (TARGET_INIT_PIC_REG): Define.
9849 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9850 (sparc_pic_register_p): New predicate.
9851 (sparc_legitimate_address_p): Use it.
9852 (sparc_legitimize_pic_address): Likewise.
9853 (sparc_delegitimize_address): Likewise.
9854 (sparc_mode_dependent_address_p): Likewise.
9855 (gen_load_pcrel_sym): Remove 4th parameter.
9856 (load_got_register): Adjust call to above. Remove obsolete stuff.
9857 (sparc_expand_prologue): Do not call load_got_register here.
9858 (sparc_flat_expand_prologue): Likewise.
9859 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9860 (sparc_use_pseudo_pic_reg): New function.
9861 (sparc_init_pic_reg): Likewise.
9862 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9863 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9865 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9867 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9868 Add item for branch_cost.
9870 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9872 PR rtl-optimization/83565
9873 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9874 not extend the result to a larger mode for rotate operations.
9875 (num_sign_bit_copies1): Likewise.
9877 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9880 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9882 Use values-Xc.o for -pedantic.
9883 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9885 2018-01-12 Martin Liska <mliska@suse.cz>
9888 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9890 (possible_polymorphic_call_targets): Use it.
9891 (ipa_devirt): Likewise.
9893 2018-01-12 Martin Liska <mliska@suse.cz>
9895 * profile-count.h (enum profile_quality): Use 0 as invalid
9896 enum value of profile_quality.
9898 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9900 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9901 -mext-string options.
9903 2018-01-12 Richard Biener <rguenther@suse.de>
9905 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9906 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9907 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9909 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9911 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9913 * configure.ac (--with-long-double-format): Add support for the
9914 configuration option to change the default long double format on
9916 * config.gcc (powerpc*-linux*-*): Likewise.
9917 * configure: Regenerate.
9918 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9919 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9920 used without modification.
9922 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9924 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9925 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9926 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9927 MISC_BUILTIN_SPEC_BARRIER.
9928 (rs6000_init_builtins): Likewise.
9929 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9931 (speculation_barrier): New define_insn.
9932 * doc/extend.texi: Document __builtin_speculation_barrier.
9934 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9937 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9938 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9939 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9941 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9942 integral modes instead of "ss" and "sd".
9943 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9944 vectors with 32-bit and 64-bit elements.
9945 (vecdupssescalarmodesuffix): New mode attribute.
9946 (vec_dup<mode>): Use it.
9948 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9951 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9952 frame if argument is passed on stack.
9954 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9957 * ree.c (combine_reaching_defs): Optimize also
9958 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9959 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9961 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9964 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9966 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9969 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9970 after they are computed.
9972 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9974 PR tree-optimization/83695
9975 * gimple-loop-linterchange.cc
9976 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9977 reset cached scev information after interchange.
9978 (pass_linterchange::execute): Remove call to scev_reset_htab.
9980 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9982 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9983 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9984 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9985 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9986 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9987 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9988 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9989 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9990 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9991 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9992 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9993 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9994 (V_lane_reg): Likewise.
9995 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9997 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9998 (vfmal_lane_low<mode>_intrinsic,
9999 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10000 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10001 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10002 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10003 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10004 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10006 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10008 * config/arm/arm-cpus.in (fp16fml): New feature.
10009 (ALL_SIMD): Add fp16fml.
10010 (armv8.2-a): Add fp16fml as an option.
10011 (armv8.3-a): Likewise.
10012 (armv8.4-a): Add fp16fml as part of fp16.
10013 * config/arm/arm.h (TARGET_FP16FML): Define.
10014 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10016 * config/arm/arm-modes.def (V2HF): Define.
10017 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10018 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10019 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10020 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10021 vfmsl_low, vfmsl_high): New set of builtins.
10022 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10023 (vfml_op): New code attribute.
10024 (VFMLHALVES): New int iterator.
10025 (VFML, VFMLSEL): New mode attributes.
10026 (V_reg): Define mapping for V2HF.
10027 (V_hi, V_lo): New mode attributes.
10028 (VF_constraint): Likewise.
10029 (vfml_half, vfml_half_selector): New int attributes.
10030 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10032 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10033 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10035 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10036 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10037 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10038 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10040 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10041 Document new effective target and option set.
10043 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10045 * config/arm/arm-cpus.in (armv8_4): New feature.
10046 (ARMv8_4a): New fgroup.
10047 (armv8.4-a): New arch.
10048 * config/arm/arm-tables.opt: Regenerate.
10049 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10050 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10051 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10052 Add matching rules for -march=armv8.4-a and extensions.
10053 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10055 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10058 * config/rx/rx.md (BW): New mode attribute.
10059 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10061 2018-01-11 Richard Biener <rguenther@suse.de>
10063 PR tree-optimization/83435
10064 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10065 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10066 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10068 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10069 Alan Hayward <alan.hayward@arm.com>
10070 David Sherwood <david.sherwood@arm.com>
10072 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10074 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10075 (aarch64_print_address_internal): Use it to check for a zero offset.
10077 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10078 Alan Hayward <alan.hayward@arm.com>
10079 David Sherwood <david.sherwood@arm.com>
10081 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10082 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10083 Return a poly_int64 rather than a HOST_WIDE_INT.
10084 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10085 rather than a HOST_WIDE_INT.
10086 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10087 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10088 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10089 final_offset from HOST_WIDE_INT to poly_int64.
10090 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10091 to_constant when getting the number of units in an Advanced SIMD
10093 (aarch64_builtin_vectorized_function): Check for a constant number
10095 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10097 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10098 attribute instead of GET_MODE_NUNITS.
10099 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10100 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10101 GET_MODE_SIZE for fixed-size registers.
10102 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10103 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10104 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10105 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10106 (aarch64_print_operand, aarch64_print_address_internal)
10107 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10108 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10109 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10110 Handle polynomial GET_MODE_SIZE.
10111 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10112 wider than SImode without modification.
10113 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10114 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10115 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10116 passing and returning SVE modes.
10117 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10118 rather than GEN_INT.
10119 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10120 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10121 (aarch64_allocate_and_probe_stack_space): Likewise.
10122 (aarch64_layout_frame): Cope with polynomial offsets.
10123 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10124 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10125 polynomial offsets.
10126 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10127 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10128 poly_int64 rather than a HOST_WIDE_INT.
10129 (aarch64_get_separate_components, aarch64_process_components)
10130 (aarch64_expand_prologue, aarch64_expand_epilogue)
10131 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10132 (aarch64_anchor_offset): New function, split out from...
10133 (aarch64_legitimize_address): ...here.
10134 (aarch64_builtin_vectorization_cost): Handle polynomial
10135 TYPE_VECTOR_SUBPARTS.
10136 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10138 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10139 number of elements from the PARALLEL rather than the mode.
10140 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10141 rather than GET_MODE_BITSIZE.
10142 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10143 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10144 (aarch64_expand_vec_perm_const_1): Handle polynomial
10145 d->perm.length () and d->perm elements.
10146 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10147 Apply to_constant to d->perm elements.
10148 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10149 polynomial CONST_VECTOR_NUNITS.
10150 (aarch64_move_pointer): Take amount as a poly_int64 rather
10152 (aarch64_progress_pointer): Avoid temporary variable.
10153 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10154 the mode attribute instead of GET_MODE.
10156 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10157 Alan Hayward <alan.hayward@arm.com>
10158 David Sherwood <david.sherwood@arm.com>
10160 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10161 x exists before using it.
10162 (aarch64_add_constant_internal): Rename to...
10163 (aarch64_add_offset_1): ...this. Replace regnum with separate
10164 src and dest rtxes. Handle the case in which they're different,
10165 including when the offset is zero. Replace scratchreg with an rtx.
10166 Use 2 additions if there is no spare register into which we can
10167 move a 16-bit constant.
10168 (aarch64_add_constant): Delete.
10169 (aarch64_add_offset): Replace reg with separate src and dest
10170 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10171 Use aarch64_add_offset_1.
10172 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10173 an rtx rather than an int. Take the delta as a poly_int64
10174 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10175 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10176 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10177 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10178 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10179 and aarch64_add_sp.
10180 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10181 aarch64_add_constant.
10183 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10185 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10186 Use scalar_float_mode.
10188 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10190 * config/aarch64/aarch64-simd.md
10191 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10192 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10193 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10194 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10195 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10196 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10197 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10198 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10199 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10200 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10202 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10205 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10206 targ_options->x_arm_arch_string is non NULL.
10208 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10210 * config/aarch64/aarch64.h
10211 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10213 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10216 * expmed.c (emit_store_flag_force): Swap if const op0
10217 and change VOIDmode to mode of op0.
10219 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10221 PR rtl-optimization/83761
10222 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10223 than bytes to mode_for_size.
10225 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10227 PR middle-end/83189
10228 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10229 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10232 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10234 PR middle-end/83575
10235 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10236 when in layout mode.
10237 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10238 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10241 2018-01-10 Michael Collison <michael.collison@arm.com>
10243 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10244 * config/aarch64/aarch64-option-extension.def: Add
10245 AARCH64_OPT_EXTENSION of 'fp16fml'.
10246 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10247 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10248 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10249 * config/aarch64/constraints.md (Ui7): New constraint.
10250 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10251 (VFMLA_SEL_W): Ditto.
10254 (VFMLA16_LOW): New int iterator.
10255 (VFMLA16_HIGH): Ditto.
10256 (UNSPEC_FMLAL): New unspec.
10257 (UNSPEC_FMLSL): Ditto.
10258 (UNSPEC_FMLAL2): Ditto.
10259 (UNSPEC_FMLSL2): Ditto.
10260 (f16mac): New code attribute.
10261 * config/aarch64/aarch64-simd-builtins.def
10262 (aarch64_fmlal_lowv2sf): Ditto.
10263 (aarch64_fmlsl_lowv2sf): Ditto.
10264 (aarch64_fmlalq_lowv4sf): Ditto.
10265 (aarch64_fmlslq_lowv4sf): Ditto.
10266 (aarch64_fmlal_highv2sf): Ditto.
10267 (aarch64_fmlsl_highv2sf): Ditto.
10268 (aarch64_fmlalq_highv4sf): Ditto.
10269 (aarch64_fmlslq_highv4sf): Ditto.
10270 (aarch64_fmlal_lane_lowv2sf): Ditto.
10271 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10272 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10273 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10274 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10275 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10276 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10277 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10278 (aarch64_fmlal_lane_highv2sf): Ditto.
10279 (aarch64_fmlsl_lane_highv2sf): Ditto.
10280 (aarch64_fmlal_laneq_highv2sf): Ditto.
10281 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10282 (aarch64_fmlalq_lane_highv4sf): Ditto.
10283 (aarch64_fmlsl_lane_highv4sf): Ditto.
10284 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10285 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10286 * config/aarch64/aarch64-simd.md:
10287 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10288 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10289 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10290 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10291 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10292 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10293 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10294 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10295 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10296 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10297 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10298 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10299 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10300 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10301 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10302 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10303 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10304 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10305 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10306 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10307 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10308 (vfmlsl_low_u32): Ditto.
10309 (vfmlalq_low_u32): Ditto.
10310 (vfmlslq_low_u32): Ditto.
10311 (vfmlal_high_u32): Ditto.
10312 (vfmlsl_high_u32): Ditto.
10313 (vfmlalq_high_u32): Ditto.
10314 (vfmlslq_high_u32): Ditto.
10315 (vfmlal_lane_low_u32): Ditto.
10316 (vfmlsl_lane_low_u32): Ditto.
10317 (vfmlal_laneq_low_u32): Ditto.
10318 (vfmlsl_laneq_low_u32): Ditto.
10319 (vfmlalq_lane_low_u32): Ditto.
10320 (vfmlslq_lane_low_u32): Ditto.
10321 (vfmlalq_laneq_low_u32): Ditto.
10322 (vfmlslq_laneq_low_u32): Ditto.
10323 (vfmlal_lane_high_u32): Ditto.
10324 (vfmlsl_lane_high_u32): Ditto.
10325 (vfmlal_laneq_high_u32): Ditto.
10326 (vfmlsl_laneq_high_u32): Ditto.
10327 (vfmlalq_lane_high_u32): Ditto.
10328 (vfmlslq_lane_high_u32): Ditto.
10329 (vfmlalq_laneq_high_u32): Ditto.
10330 (vfmlslq_laneq_high_u32): Ditto.
10331 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10332 (AARCH64_FL_FOR_ARCH8_4): New.
10333 (AARCH64_ISA_F16FML): New ISA flag.
10334 (TARGET_F16FML): New feature flag for fp16fml.
10335 (doc/invoke.texi): Document new fp16fml option.
10337 2018-01-10 Michael Collison <michael.collison@arm.com>
10339 * config/aarch64/aarch64-builtins.c:
10340 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10341 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10342 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10343 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10344 (AARCH64_ISA_SHA3): New ISA flag.
10345 (TARGET_SHA3): New feature flag for sha3.
10346 * config/aarch64/iterators.md (sha512_op): New int attribute.
10347 (CRYPTO_SHA512): New int iterator.
10348 (UNSPEC_SHA512H): New unspec.
10349 (UNSPEC_SHA512H2): Ditto.
10350 (UNSPEC_SHA512SU0): Ditto.
10351 (UNSPEC_SHA512SU1): Ditto.
10352 * config/aarch64/aarch64-simd-builtins.def
10353 (aarch64_crypto_sha512hqv2di): New builtin.
10354 (aarch64_crypto_sha512h2qv2di): Ditto.
10355 (aarch64_crypto_sha512su0qv2di): Ditto.
10356 (aarch64_crypto_sha512su1qv2di): Ditto.
10357 (aarch64_eor3qv8hi): Ditto.
10358 (aarch64_rax1qv2di): Ditto.
10359 (aarch64_xarqv2di): Ditto.
10360 (aarch64_bcaxqv8hi): Ditto.
10361 * config/aarch64/aarch64-simd.md:
10362 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10363 (aarch64_crypto_sha512su0qv2di): Ditto.
10364 (aarch64_crypto_sha512su1qv2di): Ditto.
10365 (aarch64_eor3qv8hi): Ditto.
10366 (aarch64_rax1qv2di): Ditto.
10367 (aarch64_xarqv2di): Ditto.
10368 (aarch64_bcaxqv8hi): Ditto.
10369 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10370 (vsha512h2q_u64): Ditto.
10371 (vsha512su0q_u64): Ditto.
10372 (vsha512su1q_u64): Ditto.
10373 (veor3q_u16): Ditto.
10374 (vrax1q_u64): Ditto.
10375 (vxarq_u64): Ditto.
10376 (vbcaxq_u16): Ditto.
10377 * config/arm/types.md (crypto_sha512): New type attribute.
10378 (crypto_sha3): Ditto.
10379 (doc/invoke.texi): Document new sha3 option.
10381 2018-01-10 Michael Collison <michael.collison@arm.com>
10383 * config/aarch64/aarch64-builtins.c:
10384 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10385 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10386 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10387 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10388 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10389 (AARCH64_ISA_SM4): New ISA flag.
10390 (TARGET_SM4): New feature flag for sm4.
10391 * config/aarch64/aarch64-simd-builtins.def
10392 (aarch64_sm3ss1qv4si): Ditto.
10393 (aarch64_sm3tt1aq4si): Ditto.
10394 (aarch64_sm3tt1bq4si): Ditto.
10395 (aarch64_sm3tt2aq4si): Ditto.
10396 (aarch64_sm3tt2bq4si): Ditto.
10397 (aarch64_sm3partw1qv4si): Ditto.
10398 (aarch64_sm3partw2qv4si): Ditto.
10399 (aarch64_sm4eqv4si): Ditto.
10400 (aarch64_sm4ekeyqv4si): Ditto.
10401 * config/aarch64/aarch64-simd.md:
10402 (aarch64_sm3ss1qv4si): Ditto.
10403 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10404 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10405 (aarch64_sm4eqv4si): Ditto.
10406 (aarch64_sm4ekeyqv4si): Ditto.
10407 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10408 (sm3part_op): Ditto.
10409 (CRYPTO_SM3TT): Ditto.
10410 (CRYPTO_SM3PART): Ditto.
10411 (UNSPEC_SM3SS1): New unspec.
10412 (UNSPEC_SM3TT1A): Ditto.
10413 (UNSPEC_SM3TT1B): Ditto.
10414 (UNSPEC_SM3TT2A): Ditto.
10415 (UNSPEC_SM3TT2B): Ditto.
10416 (UNSPEC_SM3PARTW1): Ditto.
10417 (UNSPEC_SM3PARTW2): Ditto.
10418 (UNSPEC_SM4E): Ditto.
10419 (UNSPEC_SM4EKEY): Ditto.
10420 * config/aarch64/constraints.md (Ui2): New constraint.
10421 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10422 * config/arm/types.md (crypto_sm3): New type attribute.
10423 (crypto_sm4): Ditto.
10424 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10425 (vsm3tt1aq_u32): Ditto.
10426 (vsm3tt1bq_u32): Ditto.
10427 (vsm3tt2aq_u32): Ditto.
10428 (vsm3tt2bq_u32): Ditto.
10429 (vsm3partw1q_u32): Ditto.
10430 (vsm3partw2q_u32): Ditto.
10431 (vsm4eq_u32): Ditto.
10432 (vsm4ekeyq_u32): Ditto.
10433 (doc/invoke.texi): Document new sm4 option.
10435 2018-01-10 Michael Collison <michael.collison@arm.com>
10437 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10438 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10439 (AARCH64_FL_FOR_ARCH8_4): New.
10440 (AARCH64_FL_V8_4): New flag.
10441 (doc/invoke.texi): Document new armv8.4-a option.
10443 2018-01-10 Michael Collison <michael.collison@arm.com>
10445 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10446 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10447 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10448 * config/aarch64/aarch64-option-extension.def: Add
10449 AARCH64_OPT_EXTENSION of 'sha2'.
10450 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10451 (crypto): Disable sha2 and aes if crypto disabled.
10452 (crypto): Enable aes and sha2 if enabled.
10453 (simd): Disable sha2 and aes if simd disabled.
10454 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10456 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10457 (TARGET_SHA2): New feature flag for sha2.
10458 (TARGET_AES): New feature flag for aes.
10459 * config/aarch64/aarch64-simd.md:
10460 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10461 conditional on TARGET_AES.
10462 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10463 (aarch64_crypto_sha1hsi): Make pattern conditional
10465 (aarch64_crypto_sha1hv4si): Ditto.
10466 (aarch64_be_crypto_sha1hv4si): Ditto.
10467 (aarch64_crypto_sha1su1v4si): Ditto.
10468 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10469 (aarch64_crypto_sha1su0v4si): Ditto.
10470 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10471 (aarch64_crypto_sha256su0v4si): Ditto.
10472 (aarch64_crypto_sha256su1v4si): Ditto.
10473 (doc/invoke.texi): Document new aes and sha2 options.
10475 2018-01-10 Martin Sebor <msebor@redhat.com>
10477 PR tree-optimization/83781
10478 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10481 2018-01-11 Martin Sebor <msebor@gmail.com>
10482 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10484 PR tree-optimization/83501
10485 PR tree-optimization/81703
10487 * tree-ssa-strlen.c (get_string_cst): Rename...
10488 (get_string_len): ...to this. Handle global constants.
10489 (handle_char_store): Adjust.
10491 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10492 Jim Wilson <jimw@sifive.com>
10494 * config/riscv/riscv-protos.h (riscv_output_return): New.
10495 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10496 (riscv_attribute_table, riscv_output_return),
10497 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10498 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10499 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10501 (riscv_expand_prologue): Add early return for naked function.
10502 (riscv_expand_epilogue): Likewise.
10503 (riscv_function_ok_for_sibcall): Return false for naked function.
10504 (riscv_set_current_function): New.
10505 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10506 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10507 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10508 * doc/extend.texi (RISC-V Function Attributes): New.
10510 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10512 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10513 check for 128-bit long double before checking TCmode.
10514 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10515 128-bit long doubles before checking TFmode or TCmode.
10516 (FLOAT128_IBM_P): Likewise.
10518 2018-01-10 Martin Sebor <msebor@redhat.com>
10520 PR tree-optimization/83671
10521 * builtins.c (c_strlen): Unconditionally return zero for the empty
10523 Use -Warray-bounds for warnings.
10524 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10525 for non-constant array indices with COMPONENT_REF, arrays of
10526 arrays, and pointers to arrays.
10527 (gimple_fold_builtin_strlen): Determine and set length range for
10528 non-constant character arrays.
10530 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10532 PR middle-end/81897
10533 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10536 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10538 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10540 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10543 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10544 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10545 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10546 indexed_or_indirect_operand predicate.
10547 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10548 (*vsx_le_perm_load_v8hi): Likewise.
10549 (*vsx_le_perm_load_v16qi): Likewise.
10550 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10551 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10552 (*vsx_le_perm_store_v8hi): Likewise.
10553 (*vsx_le_perm_store_v16qi): Likewise.
10554 (eight unnamed splitters): Likewise.
10556 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10558 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10559 * config/rs6000/emmintrin.h: Likewise.
10560 * config/rs6000/mmintrin.h: Likewise.
10561 * config/rs6000/xmmintrin.h: Likewise.
10563 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10566 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10568 * tree.c (tree_nop_conversion): Return true for location wrapper
10570 (maybe_wrap_with_location): New function.
10571 (selftest::check_strip_nops): New function.
10572 (selftest::test_location_wrappers): New function.
10573 (selftest::tree_c_tests): Call it.
10574 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10575 (maybe_wrap_with_location): New decl.
10576 (EXPR_LOCATION_WRAPPER_P): New macro.
10577 (location_wrapper_p): New inline function.
10578 (tree_strip_any_location_wrapper): New inline function.
10580 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10583 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10584 stack_realign_offset for the largest alignment of stack slot
10586 (ix86_find_max_used_stack_alignment): New function.
10587 (ix86_finalize_stack_frame_flags): Use it. Set
10588 max_used_stack_alignment if we don't realign stack.
10589 * config/i386/i386.h (machine_function): Add
10590 max_used_stack_alignment.
10592 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10594 * config/arm/arm.opt (-mbranch-cost): New option.
10595 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10598 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10601 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10602 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10604 2018-01-10 Richard Biener <rguenther@suse.de>
10607 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10608 early out so it also covers the case where we have a non-NULL
10611 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10613 PR tree-optimization/83753
10614 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10615 for non-strided grouped accesses if the number of elements is 1.
10617 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10620 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10621 * i386.h (TARGET_USE_GATHER): Define.
10622 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10624 2018-01-10 Martin Liska <mliska@suse.cz>
10627 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10628 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10630 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10631 CLEANUP_NO_PARTITIONING is not set.
10633 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10635 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10636 for vectors, as a partial revert of r254296.
10637 * rtl.h (const_vec_p): Delete.
10638 (const_vec_duplicate_p): Don't test for vector CONSTs.
10639 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10640 * expmed.c (make_tree): Likewise.
10643 * common.md (E, F): Use CONSTANT_P instead of checking for
10645 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10646 checking for CONST_VECTOR.
10648 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10650 PR middle-end/83575
10651 * predict.c (force_edge_cold): Handle in more sane way edges
10652 with no prediction.
10654 2018-01-09 Carl Love <cel@us.ibm.com>
10656 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10658 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10659 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10660 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10661 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10662 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10663 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10664 * config/rs6000/rs6000-protos.h: Add extern defition for
10665 rs6000_generate_float2_double_code.
10666 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10668 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10669 (float2_v2df): Add define_expand.
10671 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10674 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10675 op_mode in the force_to_mode call.
10677 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10679 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10680 instead of checking each element individually.
10681 (aarch64_evpc_uzp): Likewise.
10682 (aarch64_evpc_zip): Likewise.
10683 (aarch64_evpc_ext): Likewise.
10684 (aarch64_evpc_rev): Likewise.
10685 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10686 instead of checking each element individually. Return true without
10688 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10689 whether all selected elements come from the same input, instead of
10690 checking each element individually. Remove calls to gen_rtx_REG,
10691 start_sequence and end_sequence and instead assert that no rtl is
10694 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10696 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10697 order of HIGH and CONST checks.
10699 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10701 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10702 if the destination isn't an SSA_NAME.
10704 2018-01-09 Richard Biener <rguenther@suse.de>
10706 PR tree-optimization/83668
10707 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10709 (canonicalize_loop_form): ... here, renamed from ...
10710 (canonicalize_loop_closed_ssa_form): ... this and amended to
10711 swap successor edges for loop exit blocks to make us use
10712 the RPO order we need for initial schedule generation.
10714 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10716 PR tree-optimization/64811
10717 * match.pd: When optimizing comparisons with Inf, avoid
10718 introducing or losing exceptions from comparisons with NaN.
10720 2018-01-09 Martin Liska <mliska@suse.cz>
10723 * asan.c (shadow_mem_size): Add gcc_assert.
10725 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10727 Don't save registers in main().
10730 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10731 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10732 * config/avr/avr.c (avr_set_current_function): Don't error if
10733 naked, OS_task or OS_main are specified at the same time.
10734 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10736 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10738 * common/config/avr/avr-common.c (avr_option_optimization_table):
10739 Switch on -mmain-is-OS_task for optimizing compilations.
10741 2018-01-09 Richard Biener <rguenther@suse.de>
10743 PR tree-optimization/83572
10744 * graphite.c: Include cfganal.h.
10745 (graphite_transform_loops): Connect infinite loops to exit
10746 and remove fake edges at the end.
10748 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10750 * ipa-inline.c (edge_badness): Revert accidental checkin.
10752 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10755 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10756 symbols; not inline clones.
10758 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10761 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10762 hard registers. Formatting fixes.
10764 PR preprocessor/83722
10765 * gcc.c (try_generate_repro): Pass
10766 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10767 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10770 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10771 Kito Cheng <kito.cheng@gmail.com>
10773 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10774 (riscv_leaf_function_p): Delete.
10775 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10777 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10779 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10781 (do_ifelse): New function.
10782 (do_isel): New function.
10783 (do_sub3): New function.
10784 (do_add3): New function.
10785 (do_load_mask_compare): New function.
10786 (do_overlap_load_compare): New function.
10787 (expand_compare_loop): New function.
10788 (expand_block_compare): Call expand_compare_loop() when appropriate.
10789 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10790 option description.
10791 (-mblock-compare-inline-loop-limit): New option.
10793 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10796 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10797 Reverse order of second and third operands in first alternative.
10798 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10799 of first and second elements in UNSPEC_VPERMR vector.
10800 (altivec_expand_vec_perm_le): Likewise.
10802 2018-01-08 Jeff Law <law@redhat.com>
10804 PR rtl-optimizatin/81308
10805 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10806 (process_switch): If group_case_labels makes a change, then set
10808 (pass_convert_switch::execute): If a switch is converted, then
10809 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10811 PR rtl-optimization/81308
10812 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10815 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10817 PR target/83663 - Revert r255946
10818 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10819 generation for cases where splatting a value is not useful.
10820 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10821 across a vec_duplicate and a paradoxical subreg forming a vector
10822 mode to a vec_concat.
10824 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10826 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10827 -march=armv8.3-a variants.
10828 * config/arm/t-multilib: Likewise.
10829 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10831 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10833 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10835 (cceq_ior_compare_complement): Give it a name so I can use it, and
10836 change boolean_or_operator predicate to boolean_operator so it can
10837 be used to generate a crand.
10838 (eqne): New code iterator.
10839 (bd/bd_neg): New code_attrs.
10840 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10841 a single define_insn.
10842 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10843 decrement (bdnzt/bdnzf/bdzt/bdzf).
10844 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10845 with the new names of the branch decrement patterns, and added the
10846 names of the branch decrement conditional patterns.
10848 2018-01-08 Richard Biener <rguenther@suse.de>
10850 PR tree-optimization/83563
10851 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10854 2018-01-08 Richard Biener <rguenther@suse.de>
10856 PR middle-end/83713
10857 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10859 2018-01-08 Richard Biener <rguenther@suse.de>
10861 PR tree-optimization/83685
10862 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10863 references to abnormals.
10865 2018-01-08 Richard Biener <rguenther@suse.de>
10868 * dwarf2out.c (output_indirect_strings): Handle empty
10869 skeleton_debug_str_hash.
10870 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10872 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10874 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10875 (emit_store_direct): Likewise.
10876 (arc_trampoline_adjust_address): Likewise.
10877 (arc_asm_trampoline_template): New function.
10878 (arc_initialize_trampoline): Use asm_trampoline_template.
10879 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10880 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10881 * config/arc/arc.md (flush_icache): Delete pattern.
10883 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10885 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10886 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10889 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10892 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10893 by not USED_FOR_TARGET.
10894 (make_pass_resolve_sw_modes): Likewise.
10896 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10898 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10901 2018-01-08 Richard Biener <rguenther@suse.de>
10903 PR middle-end/83580
10904 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10906 2018-01-08 Richard Biener <rguenther@suse.de>
10908 PR middle-end/83517
10909 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10911 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10913 PR middle-end/81897
10914 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10915 basic blocks with a small number of successors.
10916 (convert_control_dep_chain_into_preds): Improve handling of
10918 (dump_predicates): Split apart into...
10919 (dump_pred_chain): ...here...
10920 (dump_pred_info): ...and here.
10921 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10922 (can_chain_union_be_invalidated_p): Improve check for invalidation
10924 (uninit_uses_cannot_happen): Avoid unnecessary if
10925 convert_control_dep_chain_into_preds yielded nothing.
10927 2018-01-06 Martin Sebor <msebor@redhat.com>
10929 PR tree-optimization/83640
10930 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10931 subtracting negative offset from size.
10932 (builtin_access::overlap): Adjust offset bounds of the access to fall
10933 within the size of the object if possible.
10935 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10937 PR rtl-optimization/83699
10938 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10939 extract_bit_field_as_subreg to cases in which the extracted
10940 value is also a vector.
10942 * lra-constraints.c (process_alt_operands): Test for the equivalence
10943 substitutions when detecting a possible reload cycle.
10945 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10948 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10949 by default if flag_selective_schedling{,2}. Formatting fixes.
10951 PR rtl-optimization/83682
10952 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10953 if it has non-VECTOR_MODE element mode.
10954 (vec_duplicate_p): Likewise.
10956 PR middle-end/83694
10957 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10958 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10960 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10963 * config/i386/i386-builtin.def
10964 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10965 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10966 Require also OPTION_MASK_ISA_AVX512F in addition to
10967 OPTION_MASK_ISA_GFNI.
10968 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10969 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10970 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10971 to OPTION_MASK_ISA_GFNI.
10972 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10973 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10974 OPTION_MASK_ISA_AVX512BW.
10975 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10976 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10977 addition to OPTION_MASK_ISA_GFNI.
10978 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10979 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10980 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10981 to OPTION_MASK_ISA_GFNI.
10982 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10983 a requirement for all ISAs rather than any of them with a few
10985 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10987 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10988 bitmasks to be enabled with 3 exceptions, instead of requiring any
10989 enabled ISA with lots of exceptions.
10990 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10991 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10992 Change avx512bw in isa attribute to avx512f.
10993 * config/i386/sgxintrin.h: Add license boilerplate.
10994 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10995 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10996 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10997 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10999 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11000 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11001 temporarily sse2 rather than sse if not enabled already.
11004 * config/i386/sse.md (VI248_VLBW): Rename to ...
11005 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11006 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11007 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11008 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11009 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11010 mode iterator instead of VI248_VLBW.
11012 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11014 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11015 (record_modified): Skip clobbers; add debug output.
11016 (param_change_prob): Use sreal frequencies.
11018 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11020 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11021 punt for user-aligned variables.
11023 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11025 * tree-chrec.c (chrec_contains_symbols): Return true for
11028 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11031 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11032 of (x|y) == x for BICS pattern.
11034 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11036 PR tree-optimization/83605
11037 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11038 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11041 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11043 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11044 * config/epiphany/rtems.h: New file.
11046 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11047 Uros Bizjak <ubizjak@gmail.com>
11050 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11051 QIreg_operand instead of register_operand predicate.
11052 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11053 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11054 comments instead of -fmitigate[-_]rop.
11056 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11059 * cgraphunit.c (symbol_table::compile): Switch to text_section
11060 before calling assembly_start debug hook.
11061 * run-rtl-passes.c (run_rtl_passes): Likewise.
11064 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11066 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11067 range_int_cst_p rather than !symbolic_range_p before calling
11068 extract_range_from_multiplicative_op_1.
11070 2018-01-04 Jeff Law <law@redhat.com>
11072 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11073 redundant test in assertion.
11075 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11077 * doc/rtl.texi: Document machine_mode wrapper classes.
11079 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11081 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11082 using tree_to_uhwi.
11084 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11086 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11087 the VEC_PERM_EXPR fold to fail.
11089 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11092 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11093 to switched_sections.
11095 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11098 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11099 test for d.testing.
11101 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11104 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11105 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11107 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11110 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11111 is BLKmode and bitpos not zero or mode change is needed.
11113 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11116 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11119 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11122 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11123 instead of MULT rtx. Update all corresponding splitters.
11124 (*saddl_se): Ditto.
11125 (*ssub<modesuffix>): Ditto.
11126 (*ssubl_se): Ditto.
11127 (*cmp_sadd_di): Update split patterns.
11128 (*cmp_sadd_si): Ditto.
11129 (*cmp_sadd_sidi): Ditto.
11130 (*cmp_ssub_di): Ditto.
11131 (*cmp_ssub_si): Ditto.
11132 (*cmp_ssub_sidi): Ditto.
11133 * config/alpha/predicates.md (const23_operand): New predicate.
11134 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11135 Look for ASHIFT, not MULT inner operand.
11136 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11138 2018-01-04 Martin Liska <mliska@suse.cz>
11140 PR gcov-profile/83669
11141 * gcov.c (output_intermediate_file): Add version to intermediate
11143 * doc/gcov.texi: Document new field 'version' in intermediate
11144 file format. Fix location of '-k' option of gcov command.
11146 2018-01-04 Martin Liska <mliska@suse.cz>
11149 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11151 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11153 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11155 2018-01-03 Martin Sebor <msebor@redhat.com>
11157 PR tree-optimization/83655
11158 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11159 checking calls with invalid arguments.
11161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11163 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11164 (vectorizable_mask_load_store): Delete.
11165 (vectorizable_call): Return false for masked loads and stores.
11166 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11167 instead of gimple_assign_rhs1.
11168 (vectorizable_load): Handle IFN_MASK_LOAD.
11169 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11173 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11175 (vectorizable_mask_load_store): ...here.
11176 (vectorizable_load): ...and here.
11178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11180 * tree-vect-stmts.c (vect_build_all_ones_mask)
11181 (vect_build_zero_merge_argument): New functions, split out from...
11182 (vectorizable_load): ...here.
11184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11186 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11188 (vectorizable_mask_load_store): ...here.
11189 (vectorizable_store): ...and here.
11191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11193 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11195 (vectorizable_mask_load_store): ...here.
11197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11199 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11200 (vect_model_store_cost): Take a vec_load_store_type instead of a
11202 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11203 (vect_model_store_cost): Take a vec_load_store_type instead of a
11205 (vectorizable_mask_load_store): Update accordingly.
11206 (vectorizable_store): Likewise.
11207 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11211 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11212 IFN_MASK_LOAD calls here rather than...
11213 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11216 Alan Hayward <alan.hayward@arm.com>
11217 David Sherwood <david.sherwood@arm.com>
11219 * expmed.c (extract_bit_field_1): For vector extracts,
11220 fall back to extract_bit_field_as_subreg if vec_extract
11223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11224 Alan Hayward <alan.hayward@arm.com>
11225 David Sherwood <david.sherwood@arm.com>
11227 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11228 they are variable or constant sized.
11229 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11230 slots for constant-sized data.
11232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11233 Alan Hayward <alan.hayward@arm.com>
11234 David Sherwood <david.sherwood@arm.com>
11236 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11237 handling COND_EXPRs with boolean comparisons, try to find a better
11238 basis for the mask type than the boolean itself.
11240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11242 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11243 is calculated and how it can be overridden.
11244 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11245 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11247 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11251 Alan Hayward <alan.hayward@arm.com>
11252 David Sherwood <david.sherwood@arm.com>
11254 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11255 Remove the mode argument.
11256 (aarch64_simd_valid_immediate): Remove the mode and inverse
11258 * config/aarch64/iterators.md (bitsize): New iterator.
11259 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11260 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11261 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11262 aarch64_simd_valid_immediate.
11263 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11264 (aarch64_reg_or_bic_imm): Likewise.
11265 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11266 with an insn_type enum and msl with a modifier_type enum.
11267 Replace element_width with a scalar_mode. Change the shift
11268 to unsigned int. Add constructors for scalar_float_mode and
11269 scalar_int_mode elements.
11270 (aarch64_vect_float_const_representable_p): Delete.
11271 (aarch64_can_const_movi_rtx_p)
11272 (aarch64_simd_scalar_immediate_valid_for_move)
11273 (aarch64_simd_make_constant): Update call to
11274 aarch64_simd_valid_immediate.
11275 (aarch64_advsimd_valid_immediate_hs): New function.
11276 (aarch64_advsimd_valid_immediate): Likewise.
11277 (aarch64_simd_valid_immediate): Remove mode and inverse
11278 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11279 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11280 and aarch64_float_const_representable_p on the result.
11281 (aarch64_output_simd_mov_immediate): Remove mode argument.
11282 Update call to aarch64_simd_valid_immediate and use of
11283 simd_immediate_info.
11284 (aarch64_output_scalar_simd_mov_immediate): Update call
11287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11288 Alan Hayward <alan.hayward@arm.com>
11289 David Sherwood <david.sherwood@arm.com>
11291 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11292 (mode_nunits): Likewise CONST_MODE_NUNITS.
11293 * machmode.def (ADJUST_NUNITS): Document.
11294 * genmodes.c (mode_data::need_nunits_adj): New field.
11295 (blank_mode): Update accordingly.
11296 (adj_nunits): New variable.
11297 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11299 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11300 listed in adj_nunits.
11301 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11302 listed in adj_nunits. Don't emit case statements for such modes.
11303 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11304 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11305 nothing if adj_nunits is nonnull.
11306 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11307 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11308 (emit_mode_fbit): Update use of print_maybe_const_decl.
11309 (emit_move_size): Likewise. Treat the array as non-const
11311 (emit_mode_adjustments): Handle adj_nunits.
11313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11315 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11316 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11317 (VECTOR_MODES): Use it.
11318 (make_vector_modes): Take the prefix as an argument.
11320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11321 Alan Hayward <alan.hayward@arm.com>
11322 David Sherwood <david.sherwood@arm.com>
11324 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11325 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11326 for MODE_VECTOR_BOOL.
11327 * machmode.def (VECTOR_BOOL_MODE): Document.
11328 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11329 (make_vector_bool_mode): New function.
11330 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11332 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11333 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11335 * stor-layout.c (int_mode_for_mode): Likewise.
11336 * tree.c (build_vector_type_for_mode): Likewise.
11337 * varasm.c (output_constant_pool_2): Likewise.
11338 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11339 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11340 for MODE_VECTOR_BOOL.
11341 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11342 of mode class checks.
11343 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11344 instead of a list of mode class checks.
11345 (expand_vector_scalar_condition): Likewise.
11346 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11349 Alan Hayward <alan.hayward@arm.com>
11350 David Sherwood <david.sherwood@arm.com>
11352 * machmode.h (mode_size): Change from unsigned short to
11354 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11355 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11356 or if measurement_type is not polynomial.
11357 (fixed_size_mode::includes_p): Check for constant-sized modes.
11358 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11359 return a poly_uint16 rather than an unsigned short.
11360 (emit_mode_size): Change the type of mode_size from unsigned short
11361 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11362 (emit_mode_adjustments): Cope with polynomial vector sizes.
11363 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11365 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11367 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11368 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11369 * caller-save.c (setup_save_areas): Likewise.
11370 (replace_reg_with_saved_mem): Likewise.
11371 * calls.c (emit_library_call_value_1): Likewise.
11372 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11373 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11374 (gen_lowpart_for_combine): Likewise.
11375 * convert.c (convert_to_integer_1): Likewise.
11376 * cse.c (equiv_constant, cse_insn): Likewise.
11377 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11378 (cselib_subst_to_values): Likewise.
11379 * dce.c (word_dce_process_block): Likewise.
11380 * df-problems.c (df_word_lr_mark_ref): Likewise.
11381 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11382 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11383 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11384 (rtl_for_decl_location): Likewise.
11385 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11386 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11387 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11388 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11389 (expand_expr_real_1): Likewise.
11390 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11391 (pad_below): Likewise.
11392 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11393 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11394 * ira.c (get_subreg_tracking_sizes): Likewise.
11395 * ira-build.c (ira_create_allocno_objects): Likewise.
11396 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11397 (ira_sort_regnos_for_alter_reg): Likewise.
11398 * ira-costs.c (record_operand_costs): Likewise.
11399 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11400 (resolve_simple_move): Likewise.
11401 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11402 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11403 (lra_constraints): Likewise.
11404 (CONST_POOL_OK_P): Reject variable-sized modes.
11405 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11406 (add_pseudo_to_slot, lra_spill): Likewise.
11407 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11408 * optabs-query.c (get_best_extraction_insn): Likewise.
11409 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11410 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11411 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11412 * recog.c (offsettable_address_addr_space_p): Likewise.
11413 * regcprop.c (maybe_mode_change): Likewise.
11414 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11415 * regrename.c (build_def_use): Likewise.
11416 * regstat.c (dump_reg_info): Likewise.
11417 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11418 (find_reloads, find_reloads_subreg_address): Likewise.
11419 * reload1.c (eliminate_regs_1): Likewise.
11420 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11421 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11422 (simplify_binary_operation_1, simplify_subreg): Likewise.
11423 * targhooks.c (default_function_arg_padding): Likewise.
11424 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11425 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11426 (verify_gimple_assign_ternary): Likewise.
11427 * tree-inline.c (estimate_move_cost): Likewise.
11428 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11429 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11430 (get_address_cost_ainc): Likewise.
11431 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11432 (vect_supportable_dr_alignment): Likewise.
11433 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11434 (vectorizable_reduction): Likewise.
11435 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11436 (vectorizable_operation, vectorizable_load): Likewise.
11437 * tree.c (build_same_sized_truth_vector_type): Likewise.
11438 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11439 * var-tracking.c (emit_note_insn_var_location): Likewise.
11440 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11441 (ADDR_VEC_ALIGN): Likewise.
11443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11444 Alan Hayward <alan.hayward@arm.com>
11445 David Sherwood <david.sherwood@arm.com>
11447 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11449 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11450 or if measurement_type is polynomial.
11451 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11452 * combine.c (make_extraction): Likewise.
11453 * dse.c (find_shift_sequence): Likewise.
11454 * dwarf2out.c (mem_loc_descriptor): Likewise.
11455 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11456 (extract_bit_field, extract_low_bits): Likewise.
11457 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11458 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11459 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11460 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11461 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11462 * reload.c (find_reloads): Likewise.
11463 * reload1.c (alter_reg): Likewise.
11464 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11465 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11466 * tree-if-conv.c (predicate_mem_writes): Likewise.
11467 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11468 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11469 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11470 * valtrack.c (dead_debug_insert_temp): Likewise.
11471 * varasm.c (mergeable_constant_section): Likewise.
11472 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11475 Alan Hayward <alan.hayward@arm.com>
11476 David Sherwood <david.sherwood@arm.com>
11478 * expr.c (expand_assignment): Cope with polynomial mode sizes
11479 when assigning to a CONCAT.
11481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11482 Alan Hayward <alan.hayward@arm.com>
11483 David Sherwood <david.sherwood@arm.com>
11485 * machmode.h (mode_precision): Change from unsigned short to
11487 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11489 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11490 or if measurement_type is not polynomial.
11491 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11492 in which the mode is already known to be a scalar_int_mode.
11493 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11494 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11496 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11497 for GET_MODE_PRECISION.
11498 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11499 for GET_MODE_PRECISION.
11500 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11502 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11503 (expand_field_assignment, make_extraction): Likewise.
11504 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11505 (get_last_value): Likewise.
11506 * convert.c (convert_to_integer_1): Likewise.
11507 * cse.c (cse_insn): Likewise.
11508 * expr.c (expand_expr_real_1): Likewise.
11509 * lra-constraints.c (simplify_operand_subreg): Likewise.
11510 * optabs-query.c (can_atomic_load_p): Likewise.
11511 * optabs.c (expand_atomic_load): Likewise.
11512 (expand_atomic_store): Likewise.
11513 * ree.c (combine_reaching_defs): Likewise.
11514 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11515 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11516 * tree.h (type_has_mode_precision_p): Likewise.
11517 * ubsan.c (instrument_si_overflow): Likewise.
11519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11520 Alan Hayward <alan.hayward@arm.com>
11521 David Sherwood <david.sherwood@arm.com>
11523 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11524 polynomial numbers of units.
11525 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11526 (valid_vector_subparts_p): New function.
11527 (build_vector_type): Remove temporary shim and take the number
11528 of units as a poly_uint64 rather than an int.
11529 (build_opaque_vector_type): Take the number of units as a
11530 poly_uint64 rather than an int.
11531 * tree.c (build_vector_from_ctor): Handle polynomial
11532 TYPE_VECTOR_SUBPARTS.
11533 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11534 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11535 (build_vector_from_val): If the number of units is variable,
11536 use build_vec_duplicate_cst for constant operands and
11537 VEC_DUPLICATE_EXPR otherwise.
11538 (make_vector_type): Remove temporary is_constant ().
11539 (build_vector_type, build_opaque_vector_type): Take the number of
11540 units as a poly_uint64 rather than an int.
11541 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11543 * cfgexpand.c (expand_debug_expr): Likewise.
11544 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11545 (store_constructor, expand_expr_real_1): Likewise.
11546 (const_scalar_mask_from_tree): Likewise.
11547 * fold-const-call.c (fold_const_reduction): Likewise.
11548 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11549 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11550 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11551 (fold_relational_const): Likewise.
11552 (native_interpret_vector): Likewise. Change the size from an
11553 int to an unsigned int.
11554 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11555 TYPE_VECTOR_SUBPARTS.
11556 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11557 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11558 duplicating a non-constant operand into a variable-length vector.
11559 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11560 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11561 * ipa-icf.c (sem_variable::equals): Likewise.
11562 * match.pd: Likewise.
11563 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11564 * print-tree.c (print_node): Likewise.
11565 * stor-layout.c (layout_type): Likewise.
11566 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11567 * tree-cfg.c (verify_gimple_comparison): Likewise.
11568 (verify_gimple_assign_binary): Likewise.
11569 (verify_gimple_assign_ternary): Likewise.
11570 (verify_gimple_assign_single): Likewise.
11571 * tree-pretty-print.c (dump_generic_node): Likewise.
11572 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11573 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11574 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11575 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11576 (vect_shift_permute_load_chain): Likewise.
11577 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11578 (expand_vector_condition, optimize_vector_constructor): Likewise.
11579 (lower_vec_perm, get_compute_type): Likewise.
11580 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11581 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11582 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11583 (vect_recog_mask_conversion_pattern): Likewise.
11584 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11585 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11586 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11587 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11588 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11589 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11590 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11591 (supportable_widening_operation): Likewise.
11592 (supportable_narrowing_operation): Likewise.
11593 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11595 * varasm.c (output_constant): Likewise.
11597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11598 Alan Hayward <alan.hayward@arm.com>
11599 David Sherwood <david.sherwood@arm.com>
11601 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11602 so that both the length == 3 and length != 3 cases set up their
11603 own permute vectors. Add comments explaining why we know the
11604 number of elements is constant.
11605 (vect_permute_load_chain): Likewise.
11607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11608 Alan Hayward <alan.hayward@arm.com>
11609 David Sherwood <david.sherwood@arm.com>
11611 * machmode.h (mode_nunits): Change from unsigned char to
11613 (ONLY_FIXED_SIZE_MODES): New macro.
11614 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11615 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11616 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11618 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11619 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11620 or if measurement_type is not polynomial.
11621 * genmodes.c (ZERO_COEFFS): New macro.
11622 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11624 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11625 Use ZERO_COEFFS when emitting initializers.
11626 * data-streamer.h (bp_pack_poly_value): New function.
11627 (bp_unpack_poly_value): Likewise.
11628 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11629 for GET_MODE_NUNITS.
11630 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11631 for GET_MODE_NUNITS.
11632 * tree.c (make_vector_type): Remove temporary shim and make
11633 the real function take the number of units as a poly_uint64
11634 rather than an int.
11635 (build_vector_type_for_mode): Handle polynomial nunits.
11636 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11637 * emit-rtl.c (const_vec_series_p_1): Likewise.
11638 (gen_rtx_CONST_VECTOR): Likewise.
11639 * fold-const.c (test_vec_duplicate_folding): Likewise.
11640 * genrecog.c (validate_pattern): Likewise.
11641 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11642 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11643 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11644 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11645 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11646 * rtlanal.c (subreg_get_info): Likewise.
11647 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11648 (vect_grouped_load_supported): Likewise.
11649 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11650 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11651 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11652 (simplify_const_unary_operation, simplify_binary_operation_1)
11653 (simplify_const_binary_operation, simplify_ternary_operation)
11654 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11655 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11656 instead of CONST_VECTOR_NUNITS.
11657 * varasm.c (output_constant_pool_2): Likewise.
11658 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11659 explicit-encoded elements in the XVEC for variable-length vectors.
11661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11663 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11666 Alan Hayward <alan.hayward@arm.com>
11667 David Sherwood <david.sherwood@arm.com>
11669 * coretypes.h (fixed_size_mode): Declare.
11670 (fixed_size_mode_pod): New typedef.
11671 * builtins.h (target_builtins::x_apply_args_mode)
11672 (target_builtins::x_apply_result_mode): Change type to
11673 fixed_size_mode_pod.
11674 * builtins.c (apply_args_size, apply_result_size, result_vector)
11675 (expand_builtin_apply_args_1, expand_builtin_apply)
11676 (expand_builtin_return): Update accordingly.
11678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11680 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11681 * cselib.c (cselib_hash_rtx): Likewise.
11682 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11683 CONST_VECTOR encoding.
11685 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11686 Jeff Law <law@redhat.com>
11689 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11690 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11691 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11692 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11695 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11696 explicitly probe *sp in a noreturn function if there were any callee
11697 register saves or frame pointer is needed.
11699 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11702 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11703 BLKmode for ternary, binary or unary expressions.
11706 * var-tracking.c (delete_vta_debug_insn): New inline function.
11707 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11708 insns from get_insns () to NULL instead of each bb separately.
11709 Use delete_vta_debug_insn. No longer static.
11710 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11711 delete_vta_debug_insns callers.
11712 * rtl.h (delete_vta_debug_insns): Declare.
11713 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11714 instead of variable_tracking_main.
11716 2018-01-03 Martin Sebor <msebor@redhat.com>
11718 PR tree-optimization/83603
11719 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11720 arguments past the endof the argument list in functions declared
11721 without a prototype.
11722 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11723 Avoid checking when arguments are null.
11725 2018-01-03 Martin Sebor <msebor@redhat.com>
11728 * doc/extend.texi (attribute const): Fix a typo.
11729 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11730 issuing -Wsuggest-attribute for void functions.
11732 2018-01-03 Martin Sebor <msebor@redhat.com>
11734 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11735 offset_int::from instead of wide_int::to_shwi.
11736 (maybe_diag_overlap): Remove assertion.
11737 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11738 * gimple-ssa-sprintf.c (format_directive): Same.
11739 (parse_directive): Same.
11740 (sprintf_dom_walker::compute_format_length): Same.
11741 (try_substitute_return_value): Same.
11743 2018-01-03 Jeff Law <law@redhat.com>
11745 PR middle-end/83654
11746 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11747 non-constant residual for zero at runtime and avoid probing in
11748 that case. Reorganize code for trailing problem to mirror handling
11751 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11753 PR tree-optimization/83501
11754 * tree-ssa-strlen.c (get_string_cst): New.
11755 (handle_char_store): Call get_string_cst.
11757 2018-01-03 Martin Liska <mliska@suse.cz>
11759 PR tree-optimization/83593
11760 * tree-ssa-strlen.c: Include tree-cfg.h.
11761 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11762 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11763 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11765 (strlen_dom_walker::before_dom_children): Call
11766 gimple_purge_dead_eh_edges. Dump tranformation with details
11768 (strlen_dom_walker::before_dom_children): Update call by adding
11769 new argument cleanup_eh.
11770 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11772 2018-01-03 Martin Liska <mliska@suse.cz>
11775 * cif-code.def (VARIADIC_THUNK): New enum value.
11776 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11779 2018-01-03 Jan Beulich <jbeulich@suse.com>
11781 * sse.md (mov<mode>_internal): Tighten condition for when to use
11782 vmovdqu<ssescalarsize> for TI and OI modes.
11784 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11786 Update copyright years.
11788 2018-01-03 Martin Liska <mliska@suse.cz>
11791 * ipa-visibility.c (function_and_variable_visibility): Skip
11792 functions with noipa attribure.
11794 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11796 * gcc.c (process_command): Update copyright notice dates.
11797 * gcov-dump.c (print_version): Ditto.
11798 * gcov.c (print_version): Ditto.
11799 * gcov-tool.c (print_version): Ditto.
11800 * gengtype.c (create_file): Ditto.
11801 * doc/cpp.texi: Bump @copying's copyright year.
11802 * doc/cppinternals.texi: Ditto.
11803 * doc/gcc.texi: Ditto.
11804 * doc/gccint.texi: Ditto.
11805 * doc/gcov.texi: Ditto.
11806 * doc/install.texi: Ditto.
11807 * doc/invoke.texi: Ditto.
11809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11811 * vector-builder.h (vector_builder::m_full_nelts): Change from
11812 unsigned int to poly_uint64.
11813 (vector_builder::full_nelts): Update prototype accordingly.
11814 (vector_builder::new_vector): Likewise.
11815 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11816 (vector_builder::operator ==): Likewise.
11817 (vector_builder::finalize): Likewise.
11818 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11819 Take the number of elements as a poly_uint64 rather than an
11821 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11822 from unsigned int to poly_uint64.
11823 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11824 (vec_perm_indices::new_vector): Likewise.
11825 (vec_perm_indices::length): Likewise.
11826 (vec_perm_indices::nelts_per_input): Likewise.
11827 (vec_perm_indices::input_nelts): Likewise.
11828 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11829 number of elements per input as a poly_uint64 rather than an
11830 unsigned int. Use the original encoding for variable-length
11831 vectors, rather than clamping each individual element.
11832 For the second and subsequent elements in each pattern,
11833 clamp the step and base before clamping their sum.
11834 (vec_perm_indices::series_p): Handle polynomial element counts.
11835 (vec_perm_indices::all_in_range_p): Likewise.
11836 (vec_perm_indices_to_tree): Likewise.
11837 (vec_perm_indices_to_rtx): Likewise.
11838 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11839 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11840 (tree_vector_builder::new_binary_operation): Handle polynomial
11841 element counts. Return false if we need to know the number
11842 of elements at compile time.
11843 * fold-const.c (fold_vec_perm): Punt if the number of elements
11844 isn't known at compile time.
11846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11848 * vec-perm-indices.h (vec_perm_builder): Change element type
11849 from HOST_WIDE_INT to poly_int64.
11850 (vec_perm_indices::element_type): Update accordingly.
11851 (vec_perm_indices::clamp): Handle polynomial element_types.
11852 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11853 (vec_perm_indices::all_in_range_p): Likewise.
11854 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11856 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11857 polynomial vec_perm_indices element types.
11858 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11859 * fold-const.c (fold_vec_perm): Likewise.
11860 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11861 * tree-vect-generic.c (lower_vec_perm): Likewise.
11862 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11863 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11864 element type to HOST_WIDE_INT.
11866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11867 Alan Hayward <alan.hayward@arm.com>
11868 David Sherwood <david.sherwood@arm.com>
11870 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11871 rather than an int. Use plus_constant.
11872 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11873 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11876 Alan Hayward <alan.hayward@arm.com>
11877 David Sherwood <david.sherwood@arm.com>
11879 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11880 a HOST_WIDE_INT to a poly_int64.
11882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11883 Alan Hayward <alan.hayward@arm.com>
11884 David Sherwood <david.sherwood@arm.com>
11886 * calls.c (load_register_parameters): Cope with polynomial
11887 mode sizes. Require a constant size for BLKmode parameters
11888 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11889 forces a parameter to be padded at the lsb end in order to
11890 fill a complete number of words, require the parameter size
11891 to be ordered wrt UNITS_PER_WORD.
11893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11894 Alan Hayward <alan.hayward@arm.com>
11895 David Sherwood <david.sherwood@arm.com>
11897 * reload1.c (spill_stack_slot_width): Change element type
11898 from unsigned int to poly_uint64_pod.
11899 (alter_reg): Treat mode sizes as polynomial.
11901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11902 Alan Hayward <alan.hayward@arm.com>
11903 David Sherwood <david.sherwood@arm.com>
11905 * reload.c (complex_word_subreg_p): New function.
11906 (reload_inner_reg_of_subreg, push_reload): Use it.
11908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11909 Alan Hayward <alan.hayward@arm.com>
11910 David Sherwood <david.sherwood@arm.com>
11912 * lra-constraints.c (process_alt_operands): Reject matched
11913 operands whose sizes aren't ordered.
11914 (match_reload): Refer to this check here.
11916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11917 Alan Hayward <alan.hayward@arm.com>
11918 David Sherwood <david.sherwood@arm.com>
11920 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11921 that the mode size is in the set {1, 2, 4, 8, 16}.
11923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11924 Alan Hayward <alan.hayward@arm.com>
11925 David Sherwood <david.sherwood@arm.com>
11927 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11928 Use plus_constant instead of gen_rtx_PLUS.
11930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11931 Alan Hayward <alan.hayward@arm.com>
11932 David Sherwood <david.sherwood@arm.com>
11934 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11935 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11936 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11937 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11938 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11939 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11940 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11941 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11942 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11943 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11945 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11946 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11947 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11948 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11949 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11950 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11951 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11952 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11953 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11954 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11956 * expr.c (emit_move_resolve_push): Treat the input and result
11957 of PUSH_ROUNDING as a poly_int64.
11958 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11959 (emit_push_insn): Likewise.
11960 * lra-eliminations.c (mark_not_eliminable): Likewise.
11961 * recog.c (push_operand): Likewise.
11962 * reload1.c (elimination_effects): Likewise.
11963 * rtlanal.c (nonzero_bits1): Likewise.
11964 * calls.c (store_one_arg): Likewise. Require the padding to be
11965 known at compile time.
11967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11968 Alan Hayward <alan.hayward@arm.com>
11969 David Sherwood <david.sherwood@arm.com>
11971 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11972 Use plus_constant instead of gen_rtx_PLUS.
11974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11975 Alan Hayward <alan.hayward@arm.com>
11976 David Sherwood <david.sherwood@arm.com>
11978 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11979 rather than an int.
11981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11982 Alan Hayward <alan.hayward@arm.com>
11983 David Sherwood <david.sherwood@arm.com>
11985 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11986 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11987 via stack temporaries. Treat the mode size as polynomial too.
11989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11990 Alan Hayward <alan.hayward@arm.com>
11991 David Sherwood <david.sherwood@arm.com>
11993 * expr.c (expand_expr_real_2): When handling conversions involving
11994 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11995 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11996 as a poly_uint64 too.
11998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11999 Alan Hayward <alan.hayward@arm.com>
12000 David Sherwood <david.sherwood@arm.com>
12002 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12005 Alan Hayward <alan.hayward@arm.com>
12006 David Sherwood <david.sherwood@arm.com>
12008 * combine.c (can_change_dest_mode): Handle polynomial
12009 REGMODE_NATURAL_SIZE.
12010 * expmed.c (store_bit_field_1): Likewise.
12011 * expr.c (store_constructor): Likewise.
12012 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12013 and polynomial REGMODE_NATURAL_SIZE.
12014 (gen_lowpart_common): Likewise.
12015 * reginfo.c (record_subregs_of_mode): Likewise.
12016 * rtlanal.c (read_modify_subreg_p): Likewise.
12018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12019 Alan Hayward <alan.hayward@arm.com>
12020 David Sherwood <david.sherwood@arm.com>
12022 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12023 numbers of elements.
12025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12026 Alan Hayward <alan.hayward@arm.com>
12027 David Sherwood <david.sherwood@arm.com>
12029 * match.pd: Cope with polynomial numbers of vector elements.
12031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12032 Alan Hayward <alan.hayward@arm.com>
12033 David Sherwood <david.sherwood@arm.com>
12035 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12036 in a POINTER_PLUS_EXPR.
12038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12039 Alan Hayward <alan.hayward@arm.com>
12040 David Sherwood <david.sherwood@arm.com>
12042 * omp-simd-clone.c (simd_clone_subparts): New function.
12043 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12044 (ipa_simd_modify_function_body): Likewise.
12046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12047 Alan Hayward <alan.hayward@arm.com>
12048 David Sherwood <david.sherwood@arm.com>
12050 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12051 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12052 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12053 (expand_vector_condition, vector_element): Likewise.
12054 (subparts_gt): New function.
12055 (get_compute_type): Use subparts_gt.
12056 (count_type_subparts): Delete.
12057 (expand_vector_operations_1): Use subparts_gt instead of
12058 count_type_subparts.
12060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12061 Alan Hayward <alan.hayward@arm.com>
12062 David Sherwood <david.sherwood@arm.com>
12064 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12065 (vect_compile_time_alias): ...this new function. Do the calculation
12066 on poly_ints rather than trees.
12067 (vect_prune_runtime_alias_test_list): Update call accordingly.
12069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12070 Alan Hayward <alan.hayward@arm.com>
12071 David Sherwood <david.sherwood@arm.com>
12073 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12075 (vect_schedule_slp_instance): Likewise.
12077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12078 Alan Hayward <alan.hayward@arm.com>
12079 David Sherwood <david.sherwood@arm.com>
12081 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12082 constant and extern definitions for variable-length vectors.
12083 (vect_get_constant_vectors): Note that the number of units
12084 is known to be constant.
12086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12087 Alan Hayward <alan.hayward@arm.com>
12088 David Sherwood <david.sherwood@arm.com>
12090 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12091 of units as polynomial. Choose between WIDE and NARROW based
12094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12095 Alan Hayward <alan.hayward@arm.com>
12096 David Sherwood <david.sherwood@arm.com>
12098 * tree-vect-stmts.c (simd_clone_subparts): New function.
12099 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12102 Alan Hayward <alan.hayward@arm.com>
12103 David Sherwood <david.sherwood@arm.com>
12105 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12106 vectors as polynomial. Use build_index_vector for
12107 IFN_GOMP_SIMD_LANE.
12109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12110 Alan Hayward <alan.hayward@arm.com>
12111 David Sherwood <david.sherwood@arm.com>
12113 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12114 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12115 for variable-length vectors.
12116 (vectorizable_mask_load_store): Treat the number of units as
12117 polynomial, asserting that it is constant if the condition has
12118 already been enforced.
12119 (vectorizable_store, vectorizable_load): Likewise.
12121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12122 Alan Hayward <alan.hayward@arm.com>
12123 David Sherwood <david.sherwood@arm.com>
12125 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12126 of units as polynomial. Punt if we can't tell at compile time
12127 which vector contains the final result.
12129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12130 Alan Hayward <alan.hayward@arm.com>
12131 David Sherwood <david.sherwood@arm.com>
12133 * tree-vect-loop.c (vectorizable_induction): Treat the number
12134 of units as polynomial. Punt on SLP inductions. Use an integer
12135 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12136 cast of such a series for variable-length floating-point
12139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12140 Alan Hayward <alan.hayward@arm.com>
12141 David Sherwood <david.sherwood@arm.com>
12143 * tree.h (build_index_vector): Declare.
12144 * tree.c (build_index_vector): New function.
12145 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12146 of units as polynomial, forcibly converting it to a constant if
12147 vectorizable_reduction has already enforced the condition.
12148 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12149 to create a {1,2,3,...} vector.
12150 (vectorizable_reduction): Treat the number of units as polynomial.
12151 Choose vectype_in based on the largest scalar element size rather
12152 than the smallest number of units. Enforce the restrictions
12155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12156 Alan Hayward <alan.hayward@arm.com>
12157 David Sherwood <david.sherwood@arm.com>
12159 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12160 number of units as polynomial.
12162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12163 Alan Hayward <alan.hayward@arm.com>
12164 David Sherwood <david.sherwood@arm.com>
12166 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12167 * target.def (autovectorize_vector_sizes): Return the vector sizes
12168 by pointer, using vector_sizes rather than a bitmask.
12169 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12170 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12171 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12173 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12174 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12175 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12176 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12177 * omp-general.c (omp_max_vf): Likewise.
12178 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12179 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12180 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12181 * tree-vect-slp.c (vect_slp_bb): Likewise.
12182 * doc/tm.texi: Regenerate.
12183 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12185 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12186 the vector size as a poly_uint64 rather than an unsigned int.
12187 (current_vector_size): Change from an unsigned int to a poly_uint64.
12188 (get_vectype_for_scalar_type): Update accordingly.
12189 * tree.h (build_truth_vector_type): Take the size and number of
12190 units as a poly_uint64 rather than an unsigned int.
12191 (build_vector_type): Add a temporary overload that takes
12192 the number of units as a poly_uint64 rather than an unsigned int.
12193 * tree.c (make_vector_type): Likewise.
12194 (build_truth_vector_type): Take the number of units as a poly_uint64
12195 rather than an unsigned int.
12197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12198 Alan Hayward <alan.hayward@arm.com>
12199 David Sherwood <david.sherwood@arm.com>
12201 * target.def (get_mask_mode): Take the number of units and length
12202 as poly_uint64s rather than unsigned ints.
12203 * targhooks.h (default_get_mask_mode): Update accordingly.
12204 * targhooks.c (default_get_mask_mode): Likewise.
12205 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12206 * doc/tm.texi: Regenerate.
12208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12209 Alan Hayward <alan.hayward@arm.com>
12210 David Sherwood <david.sherwood@arm.com>
12212 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12213 * omp-general.c (omp_max_vf): Likewise.
12214 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12215 (expand_omp_simd): Handle polynomial safelen.
12216 * omp-low.c (omplow_simd_context): Add a default constructor.
12217 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12218 (lower_rec_simd_input_clauses): Update accordingly.
12219 (lower_rec_input_clauses): Likewise.
12221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12222 Alan Hayward <alan.hayward@arm.com>
12223 David Sherwood <david.sherwood@arm.com>
12225 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12226 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12227 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12228 (vect_analyze_slp_cost): Likewise.
12229 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12230 (vect_model_load_cost): Likewise.
12232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12233 Alan Hayward <alan.hayward@arm.com>
12234 David Sherwood <david.sherwood@arm.com>
12236 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12237 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12238 from an unsigned int * to a poly_uint64_pod *.
12239 (calculate_unrolling_factor): New function.
12240 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12243 Alan Hayward <alan.hayward@arm.com>
12244 David Sherwood <david.sherwood@arm.com>
12246 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12247 from an unsigned int to a poly_uint64.
12248 (_loop_vec_info::slp_unrolling_factor): Likewise.
12249 (_loop_vec_info::vectorization_factor): Change from an int
12251 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12252 (vect_get_num_vectors): New function.
12253 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12254 (vect_get_num_copies): Use vect_get_num_vectors.
12255 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12256 to an unsigned int *.
12257 (vect_analyze_data_refs): Change min_vf from an int * to a
12259 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12260 than an unsigned HOST_WIDE_INT.
12261 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12262 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12263 to an unsigned int *.
12264 (vect_analyze_data_ref_dependences): Likewise.
12265 (vect_compute_data_ref_alignment): Handle polynomial vf.
12266 (vect_enhance_data_refs_alignment): Likewise.
12267 (vect_prune_runtime_alias_test_list): Likewise.
12268 (vect_shift_permute_load_chain): Likewise.
12269 (vect_supportable_dr_alignment): Likewise.
12270 (dependence_distance_ge_vf): Take the vectorization factor as a
12271 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12272 (vect_analyze_data_refs): Change min_vf from an int * to a
12274 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12275 vfm1 as a poly_uint64 rather than an int. Make the same change
12276 for the returned bound_scalar.
12277 (vect_gen_vector_loop_niters): Handle polynomial vf.
12278 (vect_do_peeling): Likewise. Update call to
12279 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12280 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12282 * tree-vect-loop.c (vect_determine_vectorization_factor)
12283 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12284 (vect_get_known_peeling_cost): Likewise.
12285 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12286 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12287 (vect_transform_loop): Likewise. Use the lowest possible VF when
12288 updating the upper bounds of the loop.
12289 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12290 rather than an int.
12291 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12292 polynomial unroll factors.
12293 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12294 (vect_make_slp_decision): Likewise.
12295 (vect_supported_load_permutation_p): Likewise, and polynomial
12297 (vect_analyze_slp_cost): Handle polynomial vf.
12298 (vect_slp_analyze_node_operations): Likewise.
12299 (vect_slp_analyze_bb_1): Likewise.
12300 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12301 than an unsigned HOST_WIDE_INT.
12302 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12303 (vectorizable_load): Handle polynomial vf.
12304 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12306 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12309 Alan Hayward <alan.hayward@arm.com>
12310 David Sherwood <david.sherwood@arm.com>
12312 * match.pd: Handle bit operations involving three constants
12313 and try to fold one pair.
12315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12317 * tree-vect-loop-manip.c: Include gimple-fold.h.
12318 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12319 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12320 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12321 Add a path that uses a step of VF instead of 1, but disable it
12323 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12324 and niters_no_overflow parameters. Update calls to
12325 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12326 Create a new SSA name if the latter choses to use a ste other
12327 than zero, and return it via niters_vector_mult_vf_var.
12328 * tree-vect-loop.c (vect_transform_loop): Update calls to
12329 vect_do_peeling, vect_gen_vector_loop_niters and
12330 slpeel_make_loop_iterate_ntimes.
12331 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12332 (vect_gen_vector_loop_niters): Update declarations after above changes.
12334 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12336 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12337 128-bit round to integer instructions.
12338 (ceil<mode>2): Likewise.
12339 (btrunc<mode>2): Likewise.
12340 (round<mode>2): Likewise.
12342 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12344 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12345 unaligned VSX load/store on P8/P9.
12346 (expand_block_clear): Allow the use of unaligned VSX
12347 load/store on P8/P9.
12349 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12351 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12353 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12354 swap associated with both a load and a store.
12356 2018-01-02 Andrew Waterman <andrew@sifive.com>
12358 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12359 * config/riscv/riscv.md (clear_cache): Use it.
12361 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12363 * web.c: Remove out-of-date comment.
12365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12367 * expr.c (fixup_args_size_notes): Check that any existing
12368 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12369 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12370 (emit_single_push_insn): ...here.
12372 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12374 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12375 (const_vector_encoded_nelts): New function.
12376 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12377 (const_vector_int_elt, const_vector_elt): Declare.
12378 * emit-rtl.c (const_vector_int_elt_1): New function.
12379 (const_vector_elt): Likewise.
12380 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12381 of CONST_VECTOR_ELT.
12383 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12385 * expr.c: Include rtx-vector-builder.h.
12386 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12387 directly on the tree encoding.
12388 (const_vector_from_tree): Likewise.
12389 * optabs.c: Include rtx-vector-builder.h.
12390 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12391 sequence of "u" values.
12392 * vec-perm-indices.c: Include rtx-vector-builder.h.
12393 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12394 directly on the vec_perm_indices encoding.
12396 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12398 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12399 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12400 * rtx-vector-builder.h: New file.
12401 * rtx-vector-builder.c: Likewise.
12402 * rtl.h (rtx_def::u2): Add a const_vector field.
12403 (CONST_VECTOR_NPATTERNS): New macro.
12404 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12405 (CONST_VECTOR_DUPLICATE_P): Likewise.
12406 (CONST_VECTOR_STEPPED_P): Likewise.
12407 (CONST_VECTOR_ENCODED_ELT): Likewise.
12408 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12409 (unwrap_const_vec_duplicate): Likewise.
12410 (const_vec_series_p): Check for a non-duplicated vector encoding.
12411 Say that the function only returns true for integer vectors.
12412 * emit-rtl.c: Include rtx-vector-builder.h.
12413 (gen_const_vec_duplicate_1): Delete.
12414 (gen_const_vector): Call gen_const_vec_duplicate instead of
12415 gen_const_vec_duplicate_1.
12416 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12417 (gen_const_vec_duplicate): Use rtx_vector_builder.
12418 (gen_const_vec_series): Likewise.
12419 (gen_rtx_CONST_VECTOR): Likewise.
12420 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12421 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12422 Build a new vector rather than modifying a CONST_VECTOR in-place.
12423 (handle_special_swappables): Update call accordingly.
12424 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12425 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12426 Build a new vector rather than modifying a CONST_VECTOR in-place.
12427 (handle_special_swappables): Update call accordingly.
12429 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12431 * simplify-rtx.c (simplify_const_binary_operation): Use
12432 CONST_VECTOR_ELT instead of XVECEXP.
12434 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12436 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12437 the selector elements to be different from the data elements
12438 if the selector is a VECTOR_CST.
12439 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12440 ssizetype for the selector.
12442 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12444 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12445 before testing each element individually.
12446 * tree-vect-generic.c (lower_vec_perm): Likewise.
12448 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12450 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12451 * selftest-run-tests.c (selftest::run_tests): Call it.
12452 * vector-builder.h (vector_builder::operator ==): New function.
12453 (vector_builder::operator !=): Likewise.
12454 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12455 (vec_perm_indices::all_from_input_p): New function.
12456 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12457 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12458 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12459 instead of reading the VECTOR_CST directly. Detect whether both
12460 vector inputs are the same before constructing the vec_perm_indices,
12461 and update the number of inputs argument accordingly. Use the
12462 utility functions added above. Only construct sel2 if we need to.
12464 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12466 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12467 the broadcast of the low byte.
12468 (expand_mult_highpart): Use an explicit encoding for the permutes.
12469 * optabs-query.c (can_mult_highpart_p): Likewise.
12470 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12471 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12472 (vectorizable_bswap): Likewise.
12473 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12474 explicit encoding for the power-of-2 permutes.
12475 (vect_permute_store_chain): Likewise.
12476 (vect_grouped_load_supported): Likewise.
12477 (vect_permute_load_chain): Likewise.
12479 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12481 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12482 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12483 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12484 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12485 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12486 (vect_gen_perm_mask_any): Likewise.
12488 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12490 * int-vector-builder.h: New file.
12491 * vec-perm-indices.h: Include int-vector-builder.h.
12492 (vec_perm_indices): Redefine as an int_vector_builder.
12493 (auto_vec_perm_indices): Delete.
12494 (vec_perm_builder): Redefine as a stand-alone class.
12495 (vec_perm_indices::vec_perm_indices): New function.
12496 (vec_perm_indices::clamp): Likewise.
12497 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12498 (vec_perm_indices::new_vector): New function.
12499 (vec_perm_indices::new_expanded_vector): Update for new
12500 vec_perm_indices class.
12501 (vec_perm_indices::rotate_inputs): New function.
12502 (vec_perm_indices::all_in_range_p): Operate directly on the
12503 encoded form, without computing elided elements.
12504 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12505 encoding. Update for new vec_perm_indices class.
12506 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12507 the given vec_perm_builder.
12508 (expand_vec_perm_var): Update vec_perm_builder constructor.
12509 (expand_mult_highpart): Use vec_perm_builder instead of
12510 auto_vec_perm_indices.
12511 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12512 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12513 or double series encoding as appropriate.
12514 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12515 vec_perm_indices instead of auto_vec_perm_indices.
12516 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12517 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12518 (vect_permute_store_chain): Likewise.
12519 (vect_grouped_load_supported): Likewise.
12520 (vect_permute_load_chain): Likewise.
12521 (vect_shift_permute_load_chain): Likewise.
12522 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12523 (vect_transform_slp_perm_load): Likewise.
12524 (vect_schedule_slp_instance): Likewise.
12525 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12526 (vectorizable_mask_load_store): Likewise.
12527 (vectorizable_bswap): Likewise.
12528 (vectorizable_store): Likewise.
12529 (vectorizable_load): Likewise.
12530 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12531 vec_perm_indices instead of auto_vec_perm_indices. Use
12532 tree_to_vec_perm_builder to read the vector from a tree.
12533 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12534 vec_perm_builder instead of a vec_perm_indices.
12535 (have_whole_vector_shift): Use vec_perm_builder and
12536 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12537 truncation to calc_vec_perm_mask_for_shift.
12538 (vect_create_epilog_for_reduction): Likewise.
12539 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12540 from auto_vec_perm_indices to vec_perm_indices.
12541 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12542 instead of changing individual elements.
12543 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12544 the vector in d.perm.
12545 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12546 from auto_vec_perm_indices to vec_perm_indices.
12547 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12548 instead of changing individual elements.
12549 (arm_vectorize_vec_perm_const): Use new_vector to install
12550 the vector in d.perm.
12551 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12552 Update vec_perm_builder constructor.
12553 (rs6000_expand_interleave): Likewise.
12554 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12555 (rs6000_expand_interleave): Likewise.
12557 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12559 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12560 to qimode could truncate the indices.
12561 * optabs.c (expand_vec_perm_var): Likewise.
12563 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12565 * Makefile.in (OBJS): Add vec-perm-indices.o.
12566 * vec-perm-indices.h: New file.
12567 * vec-perm-indices.c: Likewise.
12568 * target.h (vec_perm_indices): Replace with a forward class
12570 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12571 * optabs.h: Include vec-perm-indices.h.
12572 (expand_vec_perm): Delete.
12573 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12574 (expand_vec_perm_const): Declare.
12575 * target.def (vec_perm_const_ok): Replace with...
12576 (vec_perm_const): ...this new hook.
12577 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12578 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12579 * doc/tm.texi: Regenerate.
12580 * optabs.def (vec_perm_const): Delete.
12581 * doc/md.texi (vec_perm_const): Likewise.
12582 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12583 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12584 expand_vec_perm for constant permutation vectors. Assert that
12585 the mode of variable permutation vectors is the integer equivalent
12586 of the mode that is being permuted.
12587 * optabs-query.h (selector_fits_mode_p): Declare.
12588 * optabs-query.c: Include vec-perm-indices.h.
12589 (selector_fits_mode_p): New function.
12590 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12591 is defined, instead of checking whether the vec_perm_const_optab
12592 exists. Use targetm.vectorize.vec_perm_const instead of
12593 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12594 fit in the vector mode before using a variable permute.
12595 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12596 vec_perm_indices instead of an rtx.
12597 (expand_vec_perm): Replace with...
12598 (expand_vec_perm_const): ...this new function. Take the selector
12599 as a vec_perm_indices rather than an rtx. Also take the mode of
12600 the selector. Update call to shift_amt_for_vec_perm_mask.
12601 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12602 Use vec_perm_indices::new_expanded_vector to expand the original
12603 selector into bytes. Check whether the indices fit in the vector
12604 mode before using a variable permute.
12605 (expand_vec_perm_var): Make global.
12606 (expand_mult_highpart): Use expand_vec_perm_const.
12607 * fold-const.c: Includes vec-perm-indices.h.
12608 * tree-ssa-forwprop.c: Likewise.
12609 * tree-vect-data-refs.c: Likewise.
12610 * tree-vect-generic.c: Likewise.
12611 * tree-vect-loop.c: Likewise.
12612 * tree-vect-slp.c: Likewise.
12613 * tree-vect-stmts.c: Likewise.
12614 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12616 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12617 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12618 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12619 (aarch64_vectorize_vec_perm_const): ...this new function.
12620 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12621 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12622 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12623 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12624 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12625 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12626 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12628 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12629 check for NEON modes.
12630 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12631 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12632 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12633 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12635 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12636 the old VEC_PERM_CONST conditions.
12637 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12638 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12639 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12640 (ia64_vectorize_vec_perm_const_ok): Merge into...
12641 (ia64_vectorize_vec_perm_const): ...this new function.
12642 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12643 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12644 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12645 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12646 * config/mips/mips.c (mips_expand_vec_perm_const)
12647 (mips_vectorize_vec_perm_const_ok): Merge into...
12648 (mips_vectorize_vec_perm_const): ...this new function.
12649 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12650 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12651 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12652 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12653 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12654 (rs6000_expand_vec_perm_const): Delete.
12655 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12657 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12658 (altivec_expand_vec_perm_const_le): Take each operand individually.
12659 Operate on constant selectors rather than rtxes.
12660 (altivec_expand_vec_perm_const): Likewise. Update call to
12661 altivec_expand_vec_perm_const_le.
12662 (rs6000_expand_vec_perm_const): Delete.
12663 (rs6000_vectorize_vec_perm_const_ok): Delete.
12664 (rs6000_vectorize_vec_perm_const): New function.
12665 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12666 an element count and rtx array.
12667 (rs6000_expand_extract_even): Update call accordingly.
12668 (rs6000_expand_interleave): Likewise.
12669 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12670 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12671 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12672 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12673 (rs6000_expand_vec_perm_const): Delete.
12674 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12675 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12676 (altivec_expand_vec_perm_const_le): Take each operand individually.
12677 Operate on constant selectors rather than rtxes.
12678 (altivec_expand_vec_perm_const): Likewise. Update call to
12679 altivec_expand_vec_perm_const_le.
12680 (rs6000_expand_vec_perm_const): Delete.
12681 (rs6000_vectorize_vec_perm_const_ok): Delete.
12682 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12683 reference to the SPE evmerge intructions.
12684 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12685 an element count and rtx array.
12686 (rs6000_expand_extract_even): Update call accordingly.
12687 (rs6000_expand_interleave): Likewise.
12688 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12689 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12691 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12693 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12695 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12696 vector mode and that that mode matches the mode of the data
12698 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12699 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12700 directly using expand_vec_perm_1 when forcing selectors into
12702 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12704 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12706 * optabs-query.h (can_vec_perm_p): Delete.
12707 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12708 * optabs-query.c (can_vec_perm_p): Split into...
12709 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12710 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12711 particular selector is valid.
12712 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12713 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12714 (vect_grouped_load_supported): Likewise.
12715 (vect_shift_permute_load_chain): Likewise.
12716 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12717 (vect_transform_slp_perm_load): Likewise.
12718 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12719 (vectorizable_bswap): Likewise.
12720 (vect_gen_perm_mask_checked): Likewise.
12721 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12722 implementations of variable permutation vectors into account
12723 when deciding which selector to use.
12724 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12725 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12726 with a false third argument.
12727 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12728 to test whether the constant selector is valid and can_vec_perm_var_p
12729 to test whether a variable selector is valid.
12731 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12733 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12734 * optabs-query.c (can_vec_perm_p): Likewise.
12735 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12736 instead of vec_perm_indices.
12737 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12738 (vect_gen_perm_mask_checked): Likewise,
12739 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12740 (vect_gen_perm_mask_checked): Likewise,
12742 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12744 * optabs-query.h (qimode_for_vec_perm): Declare.
12745 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12746 (qimode_for_vec_perm): ...this new function.
12747 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12749 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12751 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12752 does not have a conditional at the top.
12754 2018-01-02 Richard Biener <rguenther@suse.de>
12756 * ipa-inline.c (big_speedup_p): Fix expression.
12758 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12761 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12764 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12768 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12769 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12770 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12771 cond_taken_branch_cost 3->4.
12773 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12775 PR tree-optimization/83581
12776 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12777 TODO_cleanup_cfg if any changes have been made.
12779 PR middle-end/83608
12780 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12781 convert_modes if target mode has the right side, but different mode
12784 PR middle-end/83609
12785 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12786 last argument when extracting from CONCAT. If either from_real or
12787 from_imag is NULL, use expansion through memory. If result is not
12788 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12789 the parts directly to inner mode, if even that fails, use expansion
12792 PR middle-end/83623
12793 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12794 check for bswap in mode rather than HImode and use that in expand_unop
12797 Copyright (C) 2018 Free Software Foundation, Inc.
12799 Copying and distribution of this file, with or without modification,
12800 are permitted in any medium without royalty provided the copyright
12801 notice and this notice are preserved.