| commit | 0ebaffccb294d90184ad78367de66b6307de3ac0 | |
| author | liuhongt <hongtao.liu@intel.com> | |
| Fri, 22 Mar 2024 06:40:00 +0000 (22 14:40 +0800) | ||
| committer | liuhongt <hongtao.liu@intel.com> | |
| Tue, 21 May 2024 05:21:52 +0000 (21 13:21 +0800) | ||
| tree | 3049011ba0d59a0fad60c72707aaaa88b4f0d388 | treesnapshot (tar.gz zip) |
| parent | 3bb8cdbd60cdb4dab45b97235dc045d65555b0a1 | commitdiff |
Use pblendw instead of pand to clear upper 16 bits.
For vec_pack_truncv8si/v4si w/o AVX512,
(const_vector:v4si (const_int 0xffff) x4) is used as mask to clear
upper 16 bits, but vpblendw with zero_vector can also be used, and
zero vector is cheaper than (const_vector:v4si (const_int 0xffff) x4).
gcc/ChangeLog:
PR target/114427
* config/i386/i386-expand.cc (expand_vec_perm_even_odd_pack):
Use pblendw instead of pand to clear upper bits.
gcc/testsuite/ChangeLog:
* gcc.target/i386/pr114427.c: New test.
For vec_pack_truncv8si/v4si w/o AVX512,
(const_vector:v4si (const_int 0xffff) x4) is used as mask to clear
upper 16 bits, but vpblendw with zero_vector can also be used, and
zero vector is cheaper than (const_vector:v4si (const_int 0xffff) x4).
gcc/ChangeLog:
PR target/114427
* config/i386/i386-expand.cc (expand_vec_perm_even_odd_pack):
Use pblendw instead of pand to clear upper bits.
gcc/testsuite/ChangeLog:
* gcc.target/i386/pr114427.c: New test.
| gcc/config/i386/i386-expand.cc | diffblobblamehistory | |
| gcc/testsuite/gcc.target/i386/pr114427.c | [new file with mode: 0644] | blob |