https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102327
--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> --- The master branch has been updated by hongtao Liu <liuho...@gcc.gnu.org>: https://gcc.gnu.org/g:7fc4d6000a0d6c847c4c07373fbd4a17e54c678f commit r12-3547-g7fc4d6000a0d6c847c4c07373fbd4a17e54c678f Author: liuhongt <hongtao....@intel.com> Date: Wed Sep 15 15:19:28 2021 +0800 Optimize for V{8,16,32}HFmode vec_set/extract/init. gcc/ChangeLog: PR target/102327 * config/i386/i386-expand.c (ix86_expand_vector_init_interleave): Use puncklwd to pack 2 HFmodes. (ix86_expand_vector_set): Use blendw instead of pinsrw. * config/i386/i386.c (ix86_can_change_mode_class): Adjust for AVX512FP16 which supports 16bit vector load. * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>): Rename to .. (avx512bw_interleave_high<mode><mask_name>): .. this, and extend to V32HFmode. (avx2_interleave_highv16hi<mask_name>): Rename to .. (avx2_interleave_high<mode><mask_name>): .. this, and extend to V16HFmode. (vec_interleave_highv8hi<mask_name>): Rename to .. (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode. (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>): Rename to .. (<mask_codefor>avx512bw_interleave_low<mode><mask_name>): this, and extend to V32HFmode. (avx2_interleave_lowv16hi<mask_name>): Rename to .. (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode. (vec_interleave_lowv8hi<mask_name>): Rename to .. (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode. (sse4_1_pblendw): Rename to .. (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode. (avx2_pblendph): New define_expand. (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use sseintmodesuffix instead of ssemodesuffix. (blendsuf): New mode attr. gcc/testsuite/ChangeLog: * gcc.target/i386/pr102327-1.c: New test. * gcc.target/i386/pr102327-2.c: New test. * gcc.target/i386/avx512fp16-1c.c: Adjust testcase.
