Richard Guenther [Wed, 25 Aug 2010 10:47:50 +0000 (10:47 +0000)]
re PR middle-end/45393 (ICE in redirect_eh_edge_1, at tree-eh.c:2135)
2010-08-25 Richard Guenther <rguenther@suse.de>
PR tree-optimization/45393
* tree-ssa-math-opts.c (execute_cse_sincos_1): Properly transfer
and clean EH info. Avoid SSA update.
(execute_cse_sincos): Cleanup the CFG if it has changed.
Jakub Jelinek [Wed, 25 Aug 2010 09:36:35 +0000 (11:36 +0200)]
re PR tree-optimization/45059 (error: invalid operands in binary operation, leads to ICE: verify_stmts failed)
PR tree-optimization/45059
* tree-ssa-reassoc.c (eliminate_redundant_comparison): Strip useless
type conversions from newop{1,2}. Assert t is a comparison and
newop{1,2} after the stripping are gimple vals.
Tejas Belagod [Wed, 25 Aug 2010 07:23:26 +0000 (08:23 +0100)]
vmull / vmovl support for Neon.
For Tejas Belagod
2010-08-25 Tejas Belagod <tejas.belagod@arm.com>
* lib/target-supports.exp (check_effective_target_vect_unpack):
Set vect_unpack supported flag to true for neon.
* config/arm/iterators.md (VU, SE, V_widen_l): New.
(V_unpack, US): New.
* config/arm/neon.md (vec_unpack<US>_hi_<mode>): Expansion for
vmovl.
(vec_unpack<US>_lo_<mode>): Likewise.
(neon_vec_unpack<US>_hi_<mode>): Instruction pattern for vmovl.
(neon_vec_unpack<US>_lo_<mode>): Likewise.
(vec_widen_<US>mult_lo_<mode>): Expansion for vmull.
(vec_widen_<US>mult_hi_<mode>): Likewise.
(neon_vec_<US>mult_lo_<mode>"): Instruction pattern for vmull.
(neon_vec_<US>mult_hi_<mode>"): Likewise.
(neon_unpack<US>_<mode>): Widening move intermediate step for
vectorizing without -mvectorize-with-neon-quad.
(neon_vec_<US>mult_<mode>): Widening multiply intermediate step
for vectorizing without -mvectorize-with-neon-quad.
* config/arm/predicates.md (vect_par_constant_high): Check for
high-half lanes of a vector.
(vect_par_constant_low): Check for low-half lanes of a vector.
Sebastian Pop [Tue, 24 Aug 2010 23:35:56 +0000 (23:35 +0000)]
Do not check whether memory references accessed in every iteration trap.
This patch relaxes the checks from gimple_could_trap_p in order to
allow the flag_loop_if_convert_stores to if-convert more loops
in which it is possible to prove that:
- the accesses to an array in a loop do not trap (more than the
original non-if-converted loop). This is true when the memory
accesses are executed at every iteration of the if-converted loop.
- the writes to memory occur on arrays that are not const qualified.
This is true when there exists at least one unconditional write to
the array in the analyzed program. In this patch this analysis is
limited to the loop to be if-converted.
* gimple.c (gimple_could_trap_p_1): Not static anymore.
Pass an extra bool parameter include_mem.
(gimple_could_trap_p): Adjust call to gimple_could_trap_p_1.
(gimple_assign_rhs_could_trap_p): Same.
* gimple.h (gimple_could_trap_p_1): Declared.
* tree-data-ref.h (same_data_refs_base_objects): New.
(same_data_refs): New.
* tree-if-conv.c (memrefs_read_or_written_unconditionally): New.
(write_memrefs_written_at_least_once): New.
(ifcvt_memrefs_wont_trap): New.
(operations_could_trap): New.
(ifcvt_could_trap_p): New.
(if_convertible_gimple_assign_stmt_p): Call ifcvt_could_trap_p.
Gets a vector of data refs.
(if_convertible_stmt_p): Same.
(if_convertible_loop_p_1): New.
(if_convertible_loop_p): Call if_convertible_loop_p_1.
Sebastian Pop [Tue, 24 Aug 2010 23:35:48 +0000 (23:35 +0000)]
Add flag -ftree-loop-if-convert-stores.
This patch adds a flag that controls the replacement of the memory
writes that are in predicated basic blocks with a full write:
for (...)
if (cond)
A[i] = foo
is replaced with:
for (...)
A[i] = cond ? foo : A[i]
In order to do this, we have to call gimple_could_trap_p instead of
gimple_assign_rhs_could_trap_p, as we have to also check that the LHS
of assign stmts does not trap.
* common.opt (ftree-loop-if-convert-stores): New flag.
* doc/invoke.texi (ftree-loop-if-convert-stores): Documented.
* tree-if-conv.c (ifc_temp_var): Pass an extra parameter GSI. Insert
the created statement before GSI.
(if_convertible_phi_p): Allow virtual phi nodes when
flag_loop_if_convert_stores is set.
(if_convertible_gimple_assign_stmt_p): Allow memory reads and writes
Do not handle types that do not match is_gimple_reg_type.
Remove loop and bb parameters. Call gimple_could_trap_p instead of
when flag_loop_if_convert_stores is set, as LHS can contain
memory refs.
(if_convertible_stmt_p): Remove loop and bb parameters. Update calls
to if_convertible_gimple_assign_stmt_p.
(if_convertible_loop_p): Update call to if_convertible_stmt_p.
(replace_phi_with_cond_gimple_assign_stmt): Renamed
predicate_scalar_phi. Do not handle virtual phi nodes.
(ifconvert_phi_nodes): Renamed predicate_all_scalar_phis.
Call predicate_scalar_phi.
(insert_gimplified_predicates): Insert the gimplified predicate of a BB
just after the labels for flag_loop_if_convert_stores, otherwise
insert the predicate in the end of the BB.
(predicate_mem_writes): New.
(combine_blocks): Call predicate_all_scalar_phis. When
flag_loop_if_convert_stores is set, call predicate_mem_writes.
(tree_if_conversion): Call mark_sym_for_renaming when
flag_loop_if_convert_stores is set.
(main_tree_if_conversion): Return TODO_update_ssa_only_virtuals when
flag_loop_if_convert_stores is set.
Richard Guenther [Tue, 24 Aug 2010 13:25:25 +0000 (13:25 +0000)]
re PR middle-end/45379 (~10% slowdown on test_fpu at revision 163278)
2010-08-24 Richard Guenther <rguenther@suse.de>
PR middle-end/45379
* tree-ssa-address.c (create_mem_ref_raw): Drop to MEM_REF
if addr->index is NULL or zero.
* tree-ssa-alias.c (indirect_refs_may_alias_p): Handle
TARGET_MEM_REF more properly.
(indirect_ref_may_alias_decl_p): Likewise.
* emit-rtl.c (set_mem_attributes_minus_bitpos): Keep TARGET_MEM_REFs.
* alias.c (ao_ref_from_mem): Handle TARGET_MEM_REF more
properly.
Changpeng Fang [Mon, 23 Aug 2010 17:45:08 +0000 (17:45 +0000)]
pr45260 Don't generate prefetch if the address of base could not be taken.
* tree-flow.h (may_be_nonaddressable_p): New definition. Make the
existing static function global.
*tree-ssa-loop-ivopts.c (may_be_nonaddressable_p): This function
is changed to global.
*tree-ssa-loop-prefetch.c (gather_memory_references_ref): Call
may_be_nonaddressable_p on base, and don't collect this reference
if the address of the base could not be taken.
Janus Weil [Mon, 23 Aug 2010 12:26:42 +0000 (14:26 +0200)]
re PR fortran/45366 (Problem with procedure pointer dummy in PURE function)
2010-08-23 Janus Weil <janus@gcc.gnu.org>
PR fortran/45366
* resolve.c (resolve_procedure_interface): New function split off from
'resolve_symbol'.
(resolve_formal_arglist): Call it here ...
(resolve_symbol): ... and here.
* configure.ac (AC_C_BIGENDIAN, AC_TYPE_UINTPTR_T): New tests.
(ssize_t): Check via AC_TYPE_SSIZE_T instead of AC_CHECK_TYPE.
(ptrdiff_t): Check via AC_CHECK_TYPE.
* config.in, configure: Rebuild.
* system.h: Include stdint.h, if available.
* lex.c (WORDS_BIGENDIAN): Provide default.
(acc_char_mask_misalign, acc_char_replicate, acc_char_cmp,
acc_char_index, search_line_acc_char, repl_chars, search_line_mmx,
search_line_sse2, search_line_sse42, init_vectorized_lexer,
search_line_fast): New.
(_cpp_clean_line): Use search_line_fast. Restructure the fast
loop to make it clear when we're leaving the loop. Stay in the
fast loop for non-trigraph '?'.
Co-Authored-By: Andi Kleen <ak@linux.intel.com> Co-Authored-By: David S. Miller <davem@davemloft.net>
From-SVN: r163446
Jan Hubicka [Sat, 21 Aug 2010 12:36:19 +0000 (14:36 +0200)]
tree-pass.h (pass_ipa_cdtor_merge): New function.
* tree-pass.h (pass_ipa_cdtor_merge): New function.
* cgraphunit.c (static_ctors, static_dtors): Move to ipa.c; make
heap allocated.
(record_cdtor_fn): Move to ipa.c; do not test for
have_ctors_dtors.
(build_cdtor): Move to ipa.c; add code avoiding construction
when target have ctors/dtors and there is only one ctor/dtor at given
priority.
(compare_ctor, compare_dtor): Move to ipa.c; use DECL_UID to stabilize sort;
reverse order of constructors.
(cgraph_build_cdtor_fns):Move to ipa.c; rename to build_cdtor_fns.
(cgraph_finalize_function): Do not call record_cdtor_fn.
(cgraph_finalize_compilation_unit): Do not call cgraph_build_cdtor_fns.
(cgraph_build_static_cdtor): Move to ipa.c.
* ipa.c: Include target.h and tree-iterator.h.
(cgraph_build_static_cdtor, static_ctors, static_dtors,
record_cdtor_fn, build_cdtor, compare_ctor, compare_dtor,
build_cdtor_fns, ipa_cdtor_merge, gate_ipa_cdtor_merge,
pass_ipa_cdtor_merge): New.
* passes.c (init_optimization_passes): Enqueue pass_ipa_cdtor_merge.
* ipa-prop.c (update_indirect_edges_after_inlining): Avoid out of bounds access.
Tobias Burnus [Sat, 21 Aug 2010 10:12:53 +0000 (12:12 +0200)]
re PR fortran/36158 (Transformational function BESSEL_YN(n1,n2,x) and BESSEL_JN missing)
2010-08-21 Tobias Burnus <burnus@net-b.de>
PR fortran/36158
PR fortran/33197
* intrinsic.c (add_sym): Init value attribute.
(set_attr_value): New function.
(add_functions) Use it and add JN/YN resolvers.
* symbol.c (gfc_copy_formal_args_intr): Copy value attr.
* intrinsic.h (gfc_resolve_bessel_n2): New prototype.
* gfortran.h (gfc_intrinsic_arg): Add value attribute.
* iresolve.c (gfc_resolve_bessel_n2): New function.
* trans-intrinsic.c (gfc_get_symbol_for_expr): Create
formal arg list.
(gfc_conv_intrinsic_function,gfc_is_intrinsic_libcall):
Add GFC_ISYM_JN2/GFC_ISYM_YN2 as case value.
* simplify.c (): For YN set to -INF if previous values
was -INF.
* trans-expr.c (gfc_conv_procedure_call): Don't crash
if sym->as is NULL.
* iresolve.c (gfc_resolve_extends_type_of): Set the
type of the dummy argument to the one of the actual.
pex-common.c (pex_read_err): Set stderr_pipe to -1 if a corresponding stream has been opened.
* pex-common.c (pex_read_err): Set stderr_pipe to -1 if a
corresponding stream has been opened.
(pex_free): Close pipe file descriptors corresponding to child's
stdout and stderr before waiting.
Olivier Hainque [Fri, 20 Aug 2010 22:08:41 +0000 (22:08 +0000)]
tree.h (alias_diag_flags): New enum.
* tree.h (alias_diag_flags): New enum.
(alias_pair): Add an 'emitted_diags' field.
* varasm.c (finish_aliases_1): Honor and update
* p->emitted_diags.
(assemble_alias): Initialize emitted_diags of new pairs.
Jakub Jelinek [Fri, 20 Aug 2010 18:07:12 +0000 (20:07 +0200)]
re PR rtl-optimization/45353 (ICE: RTL check: expected elt 3 type 'B', have '0' (rtx barrier) in sel_bb_head, at sel-sched-ir.c:4329 with -fselective-scheduling and __builtin_unreachable())
PR rtl-optimization/45353
* sel-sched-ir.c (sel_bb_head): Return NULL even if next_nonnote_insn
after bb_note is a BARRIER.
Jan Hubicka [Fri, 20 Aug 2010 13:29:36 +0000 (15:29 +0200)]
lto.c (has_analyzed_clone_p): New function
* lto.c (has_analyzed_clone_p): New function
(lto_materialize_function): Use callgraph to determine if
body is needed.
(materialize_cgraph): Remove DECL_IS_BUILTIN check.
Nathan Sidwell [Fri, 20 Aug 2010 12:59:40 +0000 (12:59 +0000)]
missed from rest of commit
missed from rest of commit
* gimplify.c (gimplify_modify_expr): When assigning to volatiles,
copy the src value and return a copy.
* doc/extend.texi (Volatiles): Move from C++ to C and expand.
(C++ Volatiles): Adjust to describe C++ semantics only.
Nathan Sidwell [Fri, 20 Aug 2010 12:22:11 +0000 (12:22 +0000)]
gimplify.c (gimplify_modify_expr): When assigning to volatiles, copy the src value and return a copy.
gcc/
* gimplify.c (gimplify_modify_expr): When assigning to volatiles,
copy the src value and return a copy.
* doc/extend.texi (Volatiles): Move from C++ to C and expand.
(C++ Volatiles): Adjust to describe C++ semantics only.
Bernd Schmidt [Thu, 19 Aug 2010 17:31:57 +0000 (17:31 +0000)]
re PR tree-optimization/42172 (inefficient bit fields assignments)
PR target/42172
* combine.c (combine_validate_cost): New arg I0. All callers changed.
Take its cost into account if nonnull.
(insn_a_feeds_b): New static function.
(combine_instructions): Look for four-insn combinations.
(can_combine_p): New args PRED2, SUCC2. All callers changed. Take
them into account when computing all_adjacent and looking for other
uses.
(combinable_i3pat): New args I0DEST, I0_NOT_IN_SRC. All callers
changed. Treat them like I1DEST and I1_NOT_IN_SRC.
(try_combine): New arg I0. Handle four-insn combinations.
(distribute_notes): New arg ELIM_I0. All callers changed. Treat it
like ELIM_I1.
Daniel Kraft [Thu, 19 Aug 2010 16:02:30 +0000 (18:02 +0200)]
re PR fortran/29785 (Fortran 2003: POINTER Rank Remapping)
2010-08-19 Daniel Kraft <d@domob.eu>
PR fortran/29785
PR fortran/45016
* trans.h (struct gfc_se): New flag `byref_noassign'.
* trans-array.h (gfc_conv_shift_descriptor_lbound): New method.
(gfc_conv_array_extent_dim), (gfc_conv_descriptor_size): New methods.
* expr.c (gfc_check_pointer_assign): Allow bounds and rank remapping
and check for compile-time errors with those.
* trans-decl.c (trans_associate_var): Use new routine
`gfc_conv_shift_descriptor_lbound' instead of doing it manually.
* trans-array.c (gfc_conv_shift_descriptor_lbound): New method.
(gfc_conv_array_extent_dim), (gfc_conv_descriptor_size): New methods.
(gfc_array_init_size): Use new `gfc_conv_array_extent_dim'.
(gfc_conv_expr_descriptor): Handle new flag `byref_noassign'.
* trans-expr.c (gfc_trans_pointer_assignment): Handle bounds and
rank remapping for assignment.
2010-08-19 Daniel Kraft <d@domob.eu>
PR fortran/29785
PR fortran/45016
* gfortran.dg/pointer_assign_5.f90: Remove 'not implemented' error.
* gfortran.dg/pointer_remapping_1.f90: New test.
* gfortran.dg/pointer_remapping_2.f03: New test.
* gfortran.dg/pointer_remapping_3.f08: New test.
* gfortran.dg/pointer_remapping_4.f03: New test.
* gfortran.dg/pointer_remapping_5.f08: New test.
* gfortran.dg/pointer_remapping_6.f08: New test.
Uros Bizjak [Thu, 19 Aug 2010 15:55:01 +0000 (17:55 +0200)]
i386.md (*lea_1): Use P mode iterator.
2010-08-19 Uros Bizjak <ubizjak@gmail.com>
* config/i386/i386.md (*lea_1): Use P mode iterator.
(lea add splitter): Also handle DImode operands.
(DImode lea add splitter): Use x86_64_immediate_operand for operand 2
predicate. Do not use ix86_lea_for_add_ok.
(zext DImode lea add splitter): Use ix86_lea_for_add_ok.
(lea ashift splitter): Also handle DImode operands.
(DImode lea ashift splitter): Remove splitter.
2010-08-19 Uros Bizjak <ubizjak@gmail.com>
* config/i386/i386.md (*add<SWI:mode>3_cconly_overflow): Do not use
ix86_binary_operator_ok.
Ian Bolton [Thu, 19 Aug 2010 08:27:59 +0000 (08:27 +0000)]
For Ian Bolton <ian.bolton@arm.com>
2010-08-19 Ian Bolton <ian.bolton@arm.com>
PR target/45070
* gcc.c-torture/execute/pr45070.c: New.
* config/arm/arm.c (arm_output_epilogue): Ensure that return
value of size 1-3 is handled correctly.