MrBesen/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

avcodec/x86/videodsp: Remove obsolete MMX, 3dnow, SSE functions

Browse Source 
x64 always has MMX, MMXEXT, SSE and SSE2 and this means
that some functions for MMX, MMXEXT, SSE and 3dnow are always
overridden by other functions (unless one e.g. explicitly
disables SSE2). So given that the only systems which 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 2022-06-12 02:13:39 +02:00
parent 2718a3be1f
commit 19abc4c0a9
2 changed files with 3 additions and 106 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
libavcodec/x86/videodsp.asm
View File

@ -45,7 +45,6 @@ SECTION .text
jnz .%1_y_loop
%endmacro
%macro vvar_fn 0
; .----. <- zero
; | | <- top is copied from first line in body of source
; |----| <- start_y
@ -53,6 +52,7 @@ SECTION .text
; |----| <- end_y
; | | <- bottom is copied from last line in body of source
; '----' <- bh
INIT_XMM sse
%if ARCH_X86_64
cglobal emu_edge_vvar, 7, 8, 1, dst, dst_stride, src, src_stride, \
start_y, end_y, bh, w
@ -81,15 +81,6 @@ cglobal emu_edge_vvar, 1, 6, 1, dst, src, start_y, end_y, bh, w
V_COPY_ROW bottom, bhq ; v_copy_row(bottom, bh)
.end: ; }
RET
%endmacro
%if ARCH_X86_32
INIT_MMX mmx
vvar_fn
%endif
INIT_XMM sse
vvar_fn
%macro hvar_fn 0
cglobal emu_edge_hvar, 5, 6, 1, dst, dst_stride, start_x, n_words, h, w
@ -105,11 +96,7 @@ cglobal emu_edge_hvar, 5, 6, 1, dst, dst_stride, start_x, n_words, h, w
imul wd, 0x01010101 ; w *= 0x01010101
movd m0, wd
mov wq, n_wordsq ; initialize w
%if cpuflag(sse2)
pshufd m0, m0, q0000 ; splat
%else ; mmx
punpckldq m0, m0 ; splat
%endif ; mmx/sse
%endif ; avx2
.x_loop: ; do {
movu [dstq+wq*2], m0 ; write($reg, $mmsize)
@ -123,11 +110,6 @@ cglobal emu_edge_hvar, 5, 6, 1, dst, dst_stride, start_x, n_words, h, w
RET
%endmacro
%if ARCH_X86_32
INIT_MMX mmx
hvar_fn
%endif
INIT_XMM sse2
hvar_fn
@ -338,9 +320,6 @@ cglobal emu_edge_vfix %+ %%n, 1, 5, 1, dst, src, start_y, end_y, bh
INIT_MMX mmx
VERTICAL_EXTEND 1, 15
%if ARCH_X86_32
VERTICAL_EXTEND 16, 22
%endif
INIT_XMM sse
VERTICAL_EXTEND 16, 22
@ -438,9 +417,6 @@ cglobal emu_edge_hfix %+ %%n, 4, 5, 1, dst, dst_stride, start_x, bh, val
INIT_MMX mmx
H_EXTEND 2, 14
%if ARCH_X86_32
H_EXTEND 16, 22
%endif
INIT_XMM sse2
H_EXTEND 16, 22
@ -450,19 +426,11 @@ INIT_XMM avx2
H_EXTEND 8, 22
%endif
%macro PREFETCH_FN 1
INIT_MMX mmxext
cglobal prefetch, 3, 3, 0, buf, stride, h
.loop:
%1 [bufq]
prefetcht0 [bufq]
add bufq, strideq
dec hd
jg .loop
REP_RET
%endmacro
INIT_MMX mmxext
PREFETCH_FN prefetcht0
%if ARCH_X86_32
INIT_MMX 3dnow
PREFETCH_FN prefetch
%endif

71
libavcodec/x86/videodsp_init.c
View File

@ -52,26 +52,6 @@ extern emu_edge_vfix_func ff_emu_edge_vfix12_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix13_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix14_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix15_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix16_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix17_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix18_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix19_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix20_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix21_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix22_mmx;
#if ARCH_X86_32
static emu_edge_vfix_func * const vfixtbl_mmx[22] = {
&ff_emu_edge_vfix1_mmx, &ff_emu_edge_vfix2_mmx, &ff_emu_edge_vfix3_mmx,
&ff_emu_edge_vfix4_mmx, &ff_emu_edge_vfix5_mmx, &ff_emu_edge_vfix6_mmx,
&ff_emu_edge_vfix7_mmx, &ff_emu_edge_vfix8_mmx, &ff_emu_edge_vfix9_mmx,
&ff_emu_edge_vfix10_mmx, &ff_emu_edge_vfix11_mmx, &ff_emu_edge_vfix12_mmx,
&ff_emu_edge_vfix13_mmx, &ff_emu_edge_vfix14_mmx, &ff_emu_edge_vfix15_mmx,
&ff_emu_edge_vfix16_mmx, &ff_emu_edge_vfix17_mmx, &ff_emu_edge_vfix18_mmx,
&ff_emu_edge_vfix19_mmx, &ff_emu_edge_vfix20_mmx, &ff_emu_edge_vfix21_mmx,
&ff_emu_edge_vfix22_mmx
};
#endif
extern emu_edge_vvar_func ff_emu_edge_vvar_mmx;
extern emu_edge_vfix_func ff_emu_edge_vfix16_sse;
extern emu_edge_vfix_func ff_emu_edge_vfix17_sse;
extern emu_edge_vfix_func ff_emu_edge_vfix18_sse;
@ -103,19 +83,6 @@ extern emu_edge_hfix_func ff_emu_edge_hfix8_mmx;
extern emu_edge_hfix_func ff_emu_edge_hfix10_mmx;
extern emu_edge_hfix_func ff_emu_edge_hfix12_mmx;
extern emu_edge_hfix_func ff_emu_edge_hfix14_mmx;
extern emu_edge_hfix_func ff_emu_edge_hfix16_mmx;
extern emu_edge_hfix_func ff_emu_edge_hfix18_mmx;
extern emu_edge_hfix_func ff_emu_edge_hfix20_mmx;
extern emu_edge_hfix_func ff_emu_edge_hfix22_mmx;
#if ARCH_X86_32
static emu_edge_hfix_func * const hfixtbl_mmx[11] = {
ff_emu_edge_hfix2_mmx, ff_emu_edge_hfix4_mmx, ff_emu_edge_hfix6_mmx,
ff_emu_edge_hfix8_mmx, ff_emu_edge_hfix10_mmx, ff_emu_edge_hfix12_mmx,
ff_emu_edge_hfix14_mmx, ff_emu_edge_hfix16_mmx, ff_emu_edge_hfix18_mmx,
ff_emu_edge_hfix20_mmx, ff_emu_edge_hfix22_mmx
};
#endif
extern emu_edge_hvar_func ff_emu_edge_hvar_mmx;
extern emu_edge_hfix_func ff_emu_edge_hfix16_sse2;
extern emu_edge_hfix_func ff_emu_edge_hfix18_sse2;
extern emu_edge_hfix_func ff_emu_edge_hfix20_sse2;
@ -221,30 +188,6 @@ static av_always_inline void emulated_edge_mc(uint8_t *dst, const uint8_t *src,
}
}
#if ARCH_X86_32
static av_noinline void emulated_edge_mc_mmx(uint8_t *buf, const uint8_t *src,
ptrdiff_t buf_stride,
ptrdiff_t src_stride,
int block_w, int block_h,
int src_x, int src_y, int w, int h)
{
emulated_edge_mc(buf, src, buf_stride, src_stride, block_w, block_h,
src_x, src_y, w, h, vfixtbl_mmx, &ff_emu_edge_vvar_mmx,
hfixtbl_mmx, &ff_emu_edge_hvar_mmx);
}
static av_noinline void emulated_edge_mc_sse(uint8_t *buf, const uint8_t *src,
ptrdiff_t buf_stride,
ptrdiff_t src_stride,
int block_w, int block_h,
int src_x, int src_y, int w, int h)
{
emulated_edge_mc(buf, src, buf_stride, src_stride, block_w, block_h,
src_x, src_y, w, h, vfixtbl_sse, &ff_emu_edge_vvar_sse,
hfixtbl_mmx, &ff_emu_edge_hvar_mmx);
}
#endif
static av_noinline void emulated_edge_mc_sse2(uint8_t *buf, const uint8_t *src,
ptrdiff_t buf_stride,
ptrdiff_t src_stride,
@ -273,29 +216,15 @@ static av_noinline void emulated_edge_mc_avx2(uint8_t *buf, const uint8_t *src,
#endif /* HAVE_X86ASM */
void ff_prefetch_mmxext(uint8_t *buf, ptrdiff_t stride, int h);
void ff_prefetch_3dnow(uint8_t *buf, ptrdiff_t stride, int h);
av_cold void ff_videodsp_init_x86(VideoDSPContext *ctx, int bpc)
{
#if HAVE_X86ASM
int cpu_flags = av_get_cpu_flags();
#if ARCH_X86_32
if (EXTERNAL_MMX(cpu_flags) && bpc <= 8) {
ctx->emulated_edge_mc = emulated_edge_mc_mmx;
}
if (EXTERNAL_AMD3DNOW(cpu_flags)) {
ctx->prefetch = ff_prefetch_3dnow;
}
#endif /* ARCH_X86_32 */
if (EXTERNAL_MMXEXT(cpu_flags)) {
ctx->prefetch = ff_prefetch_mmxext;
}
#if ARCH_X86_32
if (EXTERNAL_SSE(cpu_flags) && bpc <= 8) {
ctx->emulated_edge_mc = emulated_edge_mc_sse;
}
#endif /* ARCH_X86_32 */
if (EXTERNAL_SSE2(cpu_flags) && bpc <= 8) {
ctx->emulated_edge_mc = emulated_edge_mc_sse2;
}