github-mirrors/FFmpeg
1
0
Fork 0
mirror of https://mirror.skon.top/https://github.com/FFmpeg/FFmpeg synced 2026-04-20 21:00:41 +08:00
Code Issues Packages Projects Releases Wiki Activity

avcodec/x86/huffyuvencdsp: Remove MMX sub_hfyu_median_pred_int16

Browse Source 
Superseded by SSE2 and AVX2.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This commit is contained in:
Andreas Rheinhardt
2026-02-26 02:44:37 +01:00
parent 154bcd1054
commit d345e902d2
4 changed files with 2 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
libavcodec/huffyuvenc.c
View File

@@ -39,7 +39,6 @@
#include "huffyuvencdsp.h"
#include "lossless_videoencdsp.h"
#include "put_bits.h"
#include "libavutil/emms.h"
#include "libavutil/mem.h"
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
@@ -940,7 +939,6 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
} else {
av_log(avctx, AV_LOG_ERROR, "Format not supported!\n");
}
emms_c();
size += (put_bits_count(&s->pb) + 31) / 8;
put_bits(&s->pb, 16, 0);

40
libavcodec/x86/huffyuvencdsp.asm
View File

@@ -55,46 +55,6 @@ INIT_YMM avx2
DIFF_INT16
%endif
INIT_MMX mmxext
cglobal sub_hfyu_median_pred_int16, 7,7,0, dst, src1, src2, mask, w, left, left_top
add wd, wd
movd mm7, maskd
SPLATW mm7, mm7
movq mm0, [src1q]
movq mm2, [src2q]
psllq mm0, 16
psllq mm2, 16
movd mm6, [left_topq]
por mm0, mm6
movd mm6, [leftq]
por mm2, mm6
xor maskq, maskq
.loop:
movq mm1, [src1q + maskq]
movq mm3, [src2q + maskq]
movq mm4, mm2
psubw mm2, mm0
paddw mm2, mm1
pand mm2, mm7
movq mm5, mm4
pmaxsw mm4, mm1
pminsw mm1, mm5
pminsw mm4, mm2
pmaxsw mm4, mm1
psubw mm3, mm4
pand mm3, mm7
movq [dstq + maskq], mm3
add maskq, 8
movq mm0, [src1q + maskq - 2]
movq mm2, [src2q + maskq - 2]
cmp maskq, wq
jb .loop
movzx maskd, word [src1q + wq - 2]
mov [left_topq], maskd
movzx maskd, word [src2q + wq - 2]
mov [leftq], maskd
RET
%macro SUB_HFYU_MEDIAN_PRED_INT16 1 ; u,s for pmaxuw vs pmaxsw
cglobal sub_hfyu_median_pred_int16, 7,7,6, dst, src1, src2, mask, w, left, left_top
movd xm5, maskd

6
libavcodec/x86/huffyuvencdsp_init.c
View File

@@ -31,8 +31,6 @@ void ff_diff_int16_sse2(uint16_t *dst, const uint16_t *src1, const uint16_t *src
unsigned mask, int w);
void ff_diff_int16_avx2(uint16_t *dst, const uint16_t *src1, const uint16_t *src2,
unsigned mask, int w);
void ff_sub_hfyu_median_pred_int16_mmxext(uint16_t *dst, const uint16_t *src1, const uint16_t *src2,
unsigned mask, int w, int *left, int *left_top);
void ff_sub_hfyu_median_pred_int16_sse2(uint16_t *dst, const uint16_t *src1, const uint16_t *src2,
unsigned mask, int w, int *left, int *left_top);
void ff_sub_hfyu_median_pred_int16_avx2(uint16_t *dst, const uint16_t *src1, const uint16_t *src2,
@@ -42,10 +40,6 @@ av_cold void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c, int bpp, int wid
{
av_unused int cpu_flags = av_get_cpu_flags();
if (EXTERNAL_MMXEXT(cpu_flags) && bpp < 16) {
c->sub_hfyu_median_pred_int16 = ff_sub_hfyu_median_pred_int16_mmxext;
}
if (EXTERNAL_SSE2(cpu_flags)) {
c->diff_int16 = ff_diff_int16_sse2;
if (bpp < 16 && width >= 8)

4
tests/checkasm/huffyuvencdsp.c
View File

@@ -42,8 +42,8 @@ static void check_sub_hfyu_median_pred_int16(const char *aligned, unsigned width
static const int bpps[] = { 9, 16, };
HuffYUVEncDSPContext c;
declare_func_emms(AV_CPU_FLAG_MMXEXT, void, uint16_t *dst, const uint16_t *src1,
const uint16_t *src2, unsigned mask, int w, int *left, int *left_top);
declare_func(void, uint16_t *dst, const uint16_t *src1,
const uint16_t *src2, unsigned mask, int w, int *left, int *left_top);
for (size_t i = 0; i < FF_ARRAY_ELEMS(bpps); ++i) {
const int bpp = bpps[i];