diff options
| author | Noah Goldstein <goldstein.w.n@gmail.com> | 2021-10-23 01:26:47 -0400 |
|---|---|---|
| committer | Noah Goldstein <goldstein.w.n@gmail.com> | 2021-10-23 13:02:42 -0500 |
| commit | bad852b61b79503fcb3c5fc379c70f768df3e1fb (patch) | |
| tree | 3b37966225e1c354d49314058c513bed713883c3 /sysdeps/x86_64/multiarch | |
| parent | d8e7d0638153d27c5982619c881223bd791e844e (diff) | |
| download | glibc-bad852b61b79503fcb3c5fc379c70f768df3e1fb.zip glibc-bad852b61b79503fcb3c5fc379c70f768df3e1fb.tar.gz glibc-bad852b61b79503fcb3c5fc379c70f768df3e1fb.tar.bz2 | |
x86: Replace sse2 instructions with avx in memcmp-evex-movbe.S
This commit replaces two usages of SSE2 'movups' with AVX 'vmovdqu'.
it could potentially be dangerous to use SSE2 if this function is ever
called without using 'vzeroupper' beforehand. While compilers appear
to use 'vzeroupper' before function calls if AVX2 has been used, using
SSE2 here is more brittle. Since it is not absolutely necessary it
should be avoided.
It costs 2-extra bytes but the extra bytes should only eat into
alignment padding.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
Diffstat (limited to 'sysdeps/x86_64/multiarch')
| -rw-r--r-- | sysdeps/x86_64/multiarch/memcmp-evex-movbe.S | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S b/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S index 2761b54..640f6757 100644 --- a/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S +++ b/sysdeps/x86_64/multiarch/memcmp-evex-movbe.S @@ -561,13 +561,13 @@ L(between_16_31): /* From 16 to 31 bytes. No branch when size == 16. */ /* Use movups to save code size. */ - movups (%rsi), %xmm2 + vmovdqu (%rsi), %xmm2 VPCMP $4, (%rdi), %xmm2, %k1 kmovd %k1, %eax testl %eax, %eax jnz L(return_vec_0_lv) /* Use overlapping loads to avoid branches. */ - movups -16(%rsi, %rdx, CHAR_SIZE), %xmm2 + vmovdqu -16(%rsi, %rdx, CHAR_SIZE), %xmm2 VPCMP $4, -16(%rdi, %rdx, CHAR_SIZE), %xmm2, %k1 addl $(CHAR_PER_VEC - (16 / CHAR_SIZE)), %edx kmovd %k1, %eax |