diff options
| author | Jan Beulich <jbeulich@suse.com> | 2023-06-15 10:52:35 +0200 |
|---|---|---|
| committer | Jan Beulich <jbeulich@suse.com> | 2023-06-15 10:52:35 +0200 |
| commit | a4df0ce78d6f1beee5d2ef53e2a1a9f3fc377100 (patch) | |
| tree | 711185aaecfb7a945b180aca89d3d5b7aebc2dbf /gcc | |
| parent | a5e06412e783ad42a39f0b66797654b3f8f2056f (diff) | |
| download | gcc-a4df0ce78d6f1beee5d2ef53e2a1a9f3fc377100.zip gcc-a4df0ce78d6f1beee5d2ef53e2a1a9f3fc377100.tar.gz gcc-a4df0ce78d6f1beee5d2ef53e2a1a9f3fc377100.tar.bz2 | |
x86/AVX512: use VMOVDDUP for broadcast to V2DF
Like is already the case for the AVX/AVX2 form, VMOVDDUP - acting on
double precision floating values - is more appropriate to use here, and
it can also result in shorter insn encodings when source is memory or
%xmm0...%xmm7, and no masking is applied (in allowing a 2-byte VEX
prefix then instead of a 3-byte one).
gcc/
* config/i386/sse.md (<avx512>_vec_dup<mode><mask_name>): Use
vmovddup.
Diffstat (limited to 'gcc')
| -rw-r--r-- | gcc/config/i386/sse.md | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index a0824ef..70d7410 100644 --- a/gcc/config/i386/sse.md +++ b/gcc/config/i386/sse.md @@ -25922,9 +25922,9 @@ "TARGET_AVX512F" { /* There is no DF broadcast (in AVX-512*) to 128b register. - Mimic it with integer variant. */ + Mimic it with vmovddup, just like vec_dupv2df<mask_name> does. */ if (<MODE>mode == V2DFmode) - return "vpbroadcastq\t{%1, %0<mask_operand2>|%0<mask_operand2>, %q1}"; + return "vmovddup\t{%1, %0<mask_operand2>|%0<mask_operand2>, %q1}"; return "v<sseintprefix>broadcast<bcstscalarsuff>\t{%1, %0<mask_operand2>|%0<mask_operand2>, %<iptr>1}"; } |