Avoid AVX instructions in ScaleRowUp2_Linear_SSSE3

The "vpackuswb   %%xmm2,%%xmm0,%%xmm0" and "vmovdqu     %%xmm0,(%1)"
instructions in ScaleRowUp2_Linear_SSSE3() are AVX instructions. They
cause an illegal instruction exception on CPUs that do not support AVX.

Bug: libyuv:927
Bug: chromium:1312551
Change-Id: I87b2aaf041e7d185e7e8fb07172d4f37482e9d08
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3585881
Reviewed-by: Mirko Bonadei <mbonadei@chromium.org>
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Commit-Queue: Wan-Teh Chang <wtc@google.com>
This commit is contained in:
Wan-Teh Chang 2022-04-14 15:19:40 -07:00 committed by libyuv LUCI CQ
parent 15e6d46448
commit 18f9110516
3 changed files with 4 additions and 4 deletions

View File

@ -1,6 +1,6 @@
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 1818
Version: 1819
License: BSD
License File: LICENSE

View File

@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 1818
#define LIBYUV_VERSION 1819
#endif // INCLUDE_LIBYUV_VERSION_H_

View File

@ -1283,8 +1283,8 @@ void ScaleRowUp2_Linear_SSSE3(const uint8_t* src_ptr,
"paddw %%xmm4,%%xmm2 \n" // 3*near+far+2 (hi)
"psrlw $2,%%xmm0 \n" // 3/4*near+1/4*far (lo)
"psrlw $2,%%xmm2 \n" // 3/4*near+1/4*far (hi)
"vpackuswb %%xmm2,%%xmm0,%%xmm0 \n"
"vmovdqu %%xmm0,(%1) \n"
"packuswb %%xmm2,%%xmm0 \n"
"movdqu %%xmm0,(%1) \n"
"lea 0x8(%0),%0 \n"
"lea 0x10(%1),%1 \n" // 8 sample to 16 sample