RepoMirrors/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2025-01-01 12:22:09 +00:00
Code Issues Projects Releases Wiki Activity

avcodec/x86/mpegvideoencdsp: Remove obsolete MMX(EXT) functions

Browse Source 
x64 always has MMX, MMXEXT, SSE and SSE2 and this means
that some functions for MMX, MMXEXT and 3dnow are always
overridden by other functions (unless one e.g. explicitly
disables SSE2) for x64. So given that the only systems that
benefit from these functions are truely ancient 32bit x86s
they are removed.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This commit is contained in:
Andreas Rheinhardt 2022-06-11 04:09:24 +02:00
parent 542765ce3e
commit 55d8618a47
2 changed files with 2 additions and 61 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
libavcodec/x86/mpegvideoencdsp.asm
View File

@ -23,21 +23,15 @@
%include "libavutil/x86/x86util.asm"
SECTION_RODATA
cextern pw_1
SECTION .text
; int ff_pix_sum16_mmx(uint8_t *pix, int line_size)
; int ff_pix_sum16(uint8_t *pix, int line_size)
; %1 = number of loops
; %2 = number of GPRs used
%macro PIX_SUM16 3
cglobal pix_sum16, 2, %2, 6
movsxdifnidn r1, r1d
mov r2, %1
%if mmsize == 16
lea r3, [r1*3]
%endif
%if notcpuflag(xop)
pxor m5, m5
%endif
@ -50,56 +44,27 @@ cglobal pix_sum16, 2, %2, 6
vphaddubq m3, [r0+r3]
%else
mova m0, [r0]
%if mmsize == 8
mova m1, [r0+8]
%if cpuflag(mmxext)
mova m2, [r0+r1]
mova m3, [r0+r1+8]
%endif
%else ; sse2
mova m1, [r0+r1]
mova m2, [r0+r1*2]
mova m3, [r0+r3]
%endif
%if cpuflag(mmxext)
psadbw m0, m5
psadbw m1, m5
psadbw m2, m5
psadbw m3, m5
%else ; mmx
punpckhbw m2, m0, m5
punpcklbw m0, m5
punpckhbw m3, m1, m5
punpcklbw m1, m5
%endif ; cpuflag(mmxext)
%endif ; cpuflag(xop)
paddw m1, m0
paddw m3, m2
paddw m3, m1
paddw m4, m3
%if cpuflag(mmxext)
lea r0, [r0+r1*%3]
%else
add r0, r1
%endif
dec r2
jne .loop
%if mmsize == 16
pshufd m0, m4, q0032
paddd m4, m0
%elif notcpuflag(mmxext)
HADDW m4, m5
%endif
movd eax, m4
RET
%endmacro
%if ARCH_X86_32
INIT_MMX mmx
PIX_SUM16 16, 3, 0
INIT_MMX mmxext
PIX_SUM16 8, 4, 2
%endif
INIT_XMM sse2
PIX_SUM16 4, 4, 4
%if HAVE_XOP_EXTERNAL
@ -107,7 +72,7 @@ INIT_XMM xop
PIX_SUM16 4, 4, 4
%endif
; int ff_pix_norm1_mmx(uint8_t *pix, int line_size)
; int ff_pix_norm1(uint8_t *pix, int line_size)
; %1 = number of xmm registers used
; %2 = number of loops
%macro PIX_NORM1 2
@ -118,11 +83,7 @@ cglobal pix_norm1, 2, 3, %1
pxor m5, m5
.loop:
mova m2, [r0+0]
%if mmsize == 8
mova m3, [r0+8]
%else
mova m3, [r0+r1]
%endif
punpckhbw m1, m2, m0
punpcklbw m2, m0
punpckhbw m4, m3, m0
@ -135,11 +96,7 @@ cglobal pix_norm1, 2, 3, %1
paddd m4, m3
paddd m5, m2
paddd m5, m4
%if mmsize == 8
add r0, r1
%else
lea r0, [r0+r1*2]
%endif
dec r2
jne .loop
HADDD m5, m1
@ -147,8 +104,6 @@ cglobal pix_norm1, 2, 3, %1
RET
%endmacro
INIT_MMX mmx
PIX_NORM1 0, 16
INIT_XMM sse2
PIX_NORM1 6, 8

14
libavcodec/x86/mpegvideoencdsp_init.c
View File

@ -23,11 +23,8 @@
#include "libavcodec/avcodec.h"
#include "libavcodec/mpegvideoencdsp.h"
int ff_pix_sum16_mmx(uint8_t *pix, int line_size);
int ff_pix_sum16_mmxext(uint8_t *pix, int line_size);
int ff_pix_sum16_sse2(uint8_t *pix, int line_size);
int ff_pix_sum16_xop(uint8_t *pix, int line_size);
int ff_pix_norm1_mmx(uint8_t *pix, int line_size);
int ff_pix_norm1_sse2(uint8_t *pix, int line_size);
#if HAVE_INLINE_ASM
@ -219,17 +216,6 @@ av_cold void ff_mpegvideoencdsp_init_x86(MpegvideoEncDSPContext *c,
{
int cpu_flags = av_get_cpu_flags();
#if ARCH_X86_32
if (EXTERNAL_MMX(cpu_flags)) {
c->pix_sum = ff_pix_sum16_mmx;
c->pix_norm1 = ff_pix_norm1_mmx;
}
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->pix_sum = ff_pix_sum16_mmxext;
}
#endif
if (EXTERNAL_SSE2(cpu_flags)) {
c->pix_sum = ff_pix_sum16_sse2;
c->pix_norm1 = ff_pix_norm1_sse2;