diff options
| author | Jakub Jelinek <jakub@gcc.gnu.org> | 2014-01-04 10:57:36 +0100 |
|---|---|---|
| committer | Jakub Jelinek <jakub@gcc.gnu.org> | 2014-01-04 10:57:36 +0100 |
| commit | dad5ed2eb152e85ae2ebb3582d57d895488011a4 (patch) | |
| tree | 0825b32b591ed13bbe28597519cb1b2c610c122a /gcc/tree-vect-loop-manip.c | |
| parent | e8ed37508957f989a10b0d66dc0910ced0b2716b (diff) | |
| download | gcc-dad5ed2eb152e85ae2ebb3582d57d895488011a4.zip gcc-dad5ed2eb152e85ae2ebb3582d57d895488011a4.tar.gz gcc-dad5ed2eb152e85ae2ebb3582d57d895488011a4.tar.bz2 | |
sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d} or vmovdqu* for misaligned_operand.
* config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
or vmovdqu* for misaligned_operand.
(<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
<sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
* config/i386/i386.c (ix86_expand_special_args_builtin): Set
aligned_mem for AVX512F masked aligned load and store builtins and for
non-temporal moves.
* gcc.target/i386/avx512f-vmovdqu32-1.c: Allow vmovdqu64 instead of
vmovdqu32.
From-SVN: r206332
Diffstat (limited to 'gcc/tree-vect-loop-manip.c')
0 files changed, 0 insertions, 0 deletions