Compiling reduced testcase g++.target/i386/part-vect-vcondhf.C results in ICE since r14-4964-g7eed861e8ca3f5. $ cat part-vect-vcondhf.C typedef _Float16 __attribute__((__vector_size__(8))) __v4hf; __v4hf cf, df; __attribute__((__vector_size__(4 * sizeof(short)))) short csf_ne_d; __v4hf csf_ne() { return csf_ne_d ? cf : df; } $ g++ part-vect-vcondhf.C -mxop part-vect-vcondhf.C: In function ‘__v4hf csf_ne()’: part-vect-vcondhf.C:4:46: error: unrecognizable insn: 4 | __v4hf csf_ne() { return csf_ne_d ? cf : df; } | ^ (insn 10 9 13 2 (set (reg:V4HF 102 [ _6 ]) (if_then_else:V4HF (subreg:V4HF (reg:V4HI 99 [ _2 ]) 0) (reg:V4HF 100 [ cf.1_3 ]) (reg:V4HF 101 [ df.2_4 ]))) "part-vect-vcondhf.C":4:42 -1 (nil)) during RTL pass: vregs part-vect-vcondhf.C:4:46: internal compiler error: in extract_insn, at recog.cc:2812 0x8da3b4 _fatal_insn(char const*, rtx_def const*, char const*, int, char const*) /home/mjires/git/GCC/master/gcc/rtl-error.cc:108 0x8da3d0 _fatal_insn_not_found(rtx_def const*, char const*, int, char const*) /home/mjires/git/GCC/master/gcc/rtl-error.cc:116 0x8d88f1 extract_insn(rtx_insn*) /home/mjires/git/GCC/master/gcc/recog.cc:2812 0xf77cc0 instantiate_virtual_regs_in_insn /home/mjires/git/GCC/master/gcc/function.cc:1611 0xf77cc0 instantiate_virtual_regs /home/mjires/git/GCC/master/gcc/function.cc:1994 0xf77cc0 execute /home/mjires/git/GCC/master/gcc/function.cc:2041 Please submit a full bug report, with preprocessed source (by using -freport-bug). Please include the complete backtrace with any bug report. See <https://gcc.gnu.org/bugs/> for instructions. $ g++ -v Using built-in specs. COLLECT_GCC=g++ COLLECT_LTO_WRAPPER=/home/mjires/built/master/libexec/gcc/x86_64-pc-linux-gnu/14.0.0/lto-wrapper Target: x86_64-pc-linux-gnu Configured with: /home/mjires/git/GCC/master/configure --prefix=/home/mjires/built/master --disable-bootstrap --enable-checking --enable-languages=c,c++,fortran,lto : (reconfigured) /home/mjires/git/GCC/master/configure --prefix=/home/mjires/built/master --disable-bootstrap --enable-checking --enable-languages=c,c++,fortran,lto Thread model: posix Supported LTO compression algorithms: zlib zstd gcc version 14.0.0 20231207 (experimental) (GCC)
Confirmed.
Mine.
The master branch has been updated by hongtao Liu <liuhongt@gcc.gnu.org>: https://gcc.gnu.org/g:2c2df123de432356d28ee7e8a9627dc9e80a399d commit r14-6393-g2c2df123de432356d28ee7e8a9627dc9e80a399d Author: liuhongt <hongtao.liu@intel.com> Date: Fri Dec 8 11:19:52 2023 +0800 Support vpcmov for V4HF/V4BF/V2HF/V2BF under TARGET_XOP. gcc/ChangeLog: PR target/112904 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn. gcc/testsuite/ChangeLog: * g++.target/i386/pr112904.C: New test.
Fixed in GCC14.
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>: https://gcc.gnu.org/g:585dc19ef051959d688e6ea89b41998600480ecb commit r14-6530-g585dc19ef051959d688e6ea89b41998600480ecb Author: Jakub Jelinek <jakub@redhat.com> Date: Thu Dec 14 08:01:04 2023 +0100 testsuite: Fix up pr112904.C test [PR112904] On Fri, Dec 08, 2023 at 03:12:00PM +0800, liuhongt wrote: > * g++.target/i386/pr112904.C: New test. The new test FAILs on i686-linux and even on x86_64-linux I think it doesn't actually test what was reported, unless one performs testing with -march= for some XOP enabled CPU or -mxop. The following patch fixes that, tested on x86_64-linux with make check-g++ RUNTESTFLAGS='--target_board=unix\{-m32,-m32/-mno-sse/-mno-mmx,-m64\} i386.exp=pr112904.C' 2023-12-14 Jakub Jelinek <jakub@redhat.com> PR target/112904 * g++.target/i386/pr112904.C: Add dg-do compile, dg-options -mxop and for ia32 also dg-additional-options -mmmx.