[FFmpeg-cvslog] checkasm/sw_scale: hscale does not requires cpuflag test.
ffmpeg | branch: master | Alan Kelly | Fri
Jul 15 17:01:31 2022 +0200| [da0a37bab7434ef485146ce8575c7948db1fe3e2] |
committer: Anton Khirnov
checkasm/sw_scale: hscale does not requires cpuflag test.
This is done in ff_shuffle_filter_coefficients.
Signed-off-by: Anton Khirnov
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=da0a37bab7434ef485146ce8575c7948db1fe3e2
---
tests/checkasm/sw_scale.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/tests/checkasm/sw_scale.c b/tests/checkasm/sw_scale.c
index 9c07dd0421..86d266fb3e 100644
--- a/tests/checkasm/sw_scale.c
+++ b/tests/checkasm/sw_scale.c
@@ -278,8 +278,6 @@ static void check_hscale(void)
const uint8_t *src, const int16_t *filter,
const int32_t *filterPos, int filterSize);
-int cpu_flags = av_get_cpu_flags();
-
ctx = sws_alloc_context();
if (sws_init_context(ctx, NULL, NULL) < 0)
fail();
@@ -328,8 +326,7 @@ static void check_hscale(void)
ctx->dstW = ctx->chrDstW = input_sizes[dstWi];
ff_sws_init_scale(ctx);
memcpy(filterAvx2, filter, sizeof(uint16_t) * (SRC_PIXELS *
MAX_FILTER_WIDTH + MAX_FILTER_WIDTH));
-if ((cpu_flags & AV_CPU_FLAG_AVX2) && !(cpu_flags &
AV_CPU_FLAG_SLOW_GATHER))
-ff_shuffle_filter_coefficients(ctx, filterPosAvx, width,
filterAvx2, ctx->dstW);
+ff_shuffle_filter_coefficients(ctx, filterPosAvx, width,
filterAvx2, ctx->dstW);
if (check_func(ctx->hcScale, "hscale_%d_to_%d__fs_%d_dstW_%d",
ctx->srcBpc, ctx->dstBpc + 1, width, ctx->dstW)) {
memset(dst0, 0, SRC_PIXELS * sizeof(dst0[0]));
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libswscale: Enable hscale_avx2 for all input sizes.
ffmpeg | branch: master | Alan Kelly | Fri
Jul 15 16:59:43 2022 +0200| [a38293e4448c9389e604af9858984361a5677a20] |
committer: Anton Khirnov
libswscale: Enable hscale_avx2 for all input sizes.
ff_shuffle_filter_coefficients shuffles the tail as required.
Signed-off-by: Anton Khirnov
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a38293e4448c9389e604af9858984361a5677a20
---
libswscale/utils.c| 19 ---
libswscale/x86/swscale.c | 6 ++
tests/checkasm/sw_scale.c | 2 +-
3 files changed, 19 insertions(+), 8 deletions(-)
diff --git a/libswscale/utils.c b/libswscale/utils.c
index 34503e57f4..baa1791ebe 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -268,8 +268,7 @@ int ff_shuffle_filter_coefficients(SwsContext *c, int
*filterPos,
#if ARCH_X86_64
int i, j, k;
int cpu_flags = av_get_cpu_flags();
-// avx2 hscale filter processes 16 pixel blocks.
-if (!filter || dstW % 16 != 0)
+if (!filter)
return 0;
if (EXTERNAL_AVX2_FAST(cpu_flags) && !(cpu_flags &
AV_CPU_FLAG_SLOW_GATHER)) {
if ((c->srcBpc == 8) && (c->dstBpc <= 14)) {
@@ -281,9 +280,11 @@ int ff_shuffle_filter_coefficients(SwsContext *c, int
*filterPos,
}
// Do not swap filterPos for pixels which won't be processed by
// the main loop.
- for (i = 0; i + 8 <= dstW; i += 8) {
+ for (i = 0; i + 16 <= dstW; i += 16) {
FFSWAP(int, filterPos[i + 2], filterPos[i + 4]);
FFSWAP(int, filterPos[i + 3], filterPos[i + 5]);
+ FFSWAP(int, filterPos[i + 10], filterPos[i + 12]);
+ FFSWAP(int, filterPos[i + 11], filterPos[i + 13]);
}
if (filterSize > 4) {
// 16 pixels are processed at a time.
@@ -297,6 +298,18 @@ int ff_shuffle_filter_coefficients(SwsContext *c, int
*filterPos,
}
}
}
+ // 4 pixels are processed at a time in the tail.
+ for (; i < dstW; i += 4) {
+ // 4 filter coeffs are processed at a time.
+ int rem = dstW - i >= 4 ? 4 : dstW - i;
+ for (k = 0; k + 4 <= filterSize; k += 4) {
+ for (j = 0; j < rem; ++j) {
+ int from = (i + j) * filterSize + k;
+ int to = i * filterSize + j * 4 + k * 4;
+ memcpy([to], [from], 4 *
sizeof(int16_t));
+ }
+ }
+ }
}
av_free(filterCopy);
}
diff --git a/libswscale/x86/swscale.c b/libswscale/x86/swscale.c
index 89ef9f5d2b..ec1ca0e01c 100644
--- a/libswscale/x86/swscale.c
+++ b/libswscale/x86/swscale.c
@@ -625,10 +625,8 @@ switch(c->dstBpc){ \
if (EXTERNAL_AVX2_FAST(cpu_flags) && !(cpu_flags &
AV_CPU_FLAG_SLOW_GATHER)) {
if ((c->srcBpc == 8) && (c->dstBpc <= 14)) {
-if (c->chrDstW % 16 == 0)
-ASSIGN_AVX2_SCALE_FUNC(c->hcScale, c->hChrFilterSize);
-if (c->dstW % 16 == 0)
-ASSIGN_AVX2_SCALE_FUNC(c->hyScale, c->hLumFilterSize);
+ASSIGN_AVX2_SCALE_FUNC(c->hcScale, c->hChrFilterSize);
+ASSIGN_AVX2_SCALE_FUNC(c->hyScale, c->hLumFilterSize);
}
}
diff --git a/tests/checkasm/sw_scale.c b/tests/checkasm/sw_scale.c
index cbe4460a99..9c07dd0421 100644
--- a/tests/checkasm/sw_scale.c
+++ b/tests/checkasm/sw_scale.c
@@ -329,7 +329,7 @@ static void check_hscale(void)
ff_sws_init_scale(ctx);
memcpy(filterAvx2, filter, sizeof(uint16_t) * (SRC_PIXELS *
MAX_FILTER_WIDTH + MAX_FILTER_WIDTH));
if ((cpu_flags & AV_CPU_FLAG_AVX2) && !(cpu_flags &
AV_CPU_FLAG_SLOW_GATHER))
-ff_shuffle_filter_coefficients(ctx, filterPosAvx, width,
filterAvx2, SRC_PIXELS);
+ff_shuffle_filter_coefficients(ctx, filterPosAvx, width,
filterAvx2, ctx->dstW);
if (check_func(ctx->hcScale, "hscale_%d_to_%d__fs_%d_dstW_%d",
ctx->srcBpc, ctx->dstBpc + 1, width, ctx->dstW)) {
memset(dst0, 0, SRC_PIXELS * sizeof(dst0[0]));
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] sws: allow avx2 hscale to process inputs of any size.
ffmpeg | branch: master | Alan Kelly | Tue Apr 26 10:00:02 2022 +0200| [a6724285fd45111436dd5242eab2c489182aa5c2] | committer: Anton Khirnov sws: allow avx2 hscale to process inputs of any size. The main loop processes blocks of 16 pixels. The tail processes blocks of size 4. Signed-off-by: Anton Khirnov > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a6724285fd45111436dd5242eab2c489182aa5c2 --- libswscale/x86/scale_avx2.asm | 44 ++- 1 file changed, 43 insertions(+), 1 deletion(-) diff --git a/libswscale/x86/scale_avx2.asm b/libswscale/x86/scale_avx2.asm index 20acdbd633..37095e596a 100644 --- a/libswscale/x86/scale_avx2.asm +++ b/libswscale/x86/scale_avx2.asm @@ -53,6 +53,9 @@ cglobal hscale8to15_%1, 7, 9, 16, pos0, dst, w, srcmem, filter, fltpos, fltsize, mova m14, [four] shr fltsized, 2 %endif +cmp wq, 0x10 +jl .tail_loop +sub wq, 0x10 .loop: movu m1, [fltposq] movu m2, [fltposq+32] @@ -101,7 +104,46 @@ cglobal hscale8to15_%1, 7, 9, 16, pos0, dst, w, srcmem, filter, fltpos, fltsize, add fltposq, 0x40 add countq, 0x10 cmp countq, wq -jl .loop +jle .loop + +add wq, 0x10 +cmp countq, wq +jge .end + +.tail_loop: +movu xm1, [fltposq] +%ifidn %1, X4 +pxor xm9, xm9 +pxor xm10, xm10 +xor innerq, innerq +.tail_innerloop: +%endif +vpcmpeqd xm13, xm13 +vpgatherdd xm3,[srcmemq + xm1], xm13 +vpunpcklbw xm5, xm3, xm0 +vpunpckhbw xm6, xm3, xm0 +vpmaddwd xm5, xm5, [filterq] +vpmaddwd xm6, xm6, [filterq + 0x10] +add filterq, 0x20 +%ifidn %1, X4 +paddd xm9, xm5 +paddd xm10, xm6 +paddd xm1, xm14 +add innerq, 1 +cmp innerq, fltsizeq +jl .tail_innerloop +vphaddd xm5, xm9, xm10 +%else +vphaddd xm5, xm5, xm6 +%endif +vpsrad xm5, 7 +vpackssdw xm5, xm5, xm5 +vmovq [dstq + countq * 2], xm5 +add fltposq, 0x10 +add countq, 0x4 +cmp countq, wq +jl .tail_loop +.end: REP_RET %endmacro ___ ffmpeg-cvslog mailing list ffmpeg-cvslog@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog To unsubscribe, visit link above, or email ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] sws: Replace call to yuv2yuvX_mmx by yuv2yuvX_mmxext
ffmpeg | branch: master | Alan Kelly | Wed
Aug 17 11:20:39 2022 +0200| [51a34e8525fea2bbc29b42831d7a17f34e8518d3] |
committer: Andreas Rheinhardt
sws: Replace call to yuv2yuvX_mmx by yuv2yuvX_mmxext
Signed-off-by: Andreas Rheinhardt
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=51a34e8525fea2bbc29b42831d7a17f34e8518d3
---
libswscale/x86/swscale.c | 7 ++-
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/libswscale/x86/swscale.c b/libswscale/x86/swscale.c
index 32d441245d..89ef9f5d2b 100644
--- a/libswscale/x86/swscale.c
+++ b/libswscale/x86/swscale.c
@@ -205,20 +205,17 @@ static void yuv2yuvX_ ##opt(const int16_t *filter, int
filterSize, \
int remainder = (dstW % step); \
int pixelsProcessed = dstW - remainder; \
if(((uintptr_t)dest) & 15){ \
-yuv2yuvX_mmx(filter, filterSize, src, dest, dstW, dither, offset); \
+yuv2yuvX_mmxext(filter, filterSize, src, dest, dstW, dither, offset); \
return; \
} \
if(pixelsProcessed > 0) \
ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset,
pixelsProcessed + offset, dither, offset); \
if(remainder > 0){ \
- ff_yuv2yuvX_mmx(filter, filterSize - 1, pixelsProcessed, dest - offset,
pixelsProcessed + remainder + offset, dither, offset); \
+ ff_yuv2yuvX_mmxext(filter, filterSize - 1, pixelsProcessed, dest -
offset, pixelsProcessed + remainder + offset, dither, offset); \
} \
return; \
}
-#if HAVE_MMX_EXTERNAL
-YUV2YUVX_FUNC_MMX(mmx, 16)
-#endif
#if HAVE_MMXEXT_EXTERNAL
YUV2YUVX_FUNC_MMX(mmxext, 16)
#endif
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libswscale: Re-factor ff_shuffle_filter_coefficients.
ffmpeg | branch: master | Alan Kelly | Thu Feb 17
11:03:52 2022 +0100| [e534d98af3bfdc2c926b15301404e2d85524a048] | committer:
Michael Niedermayer
libswscale: Re-factor ff_shuffle_filter_coefficients.
Make the code more readable and follow the style guide.
Signed-off-by: Michael Niedermayer
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e534d98af3bfdc2c926b15301404e2d85524a048
---
libswscale/utils.c | 66 +-
1 file changed, 36 insertions(+), 30 deletions(-)
diff --git a/libswscale/utils.c b/libswscale/utils.c
index 344c87dfdf..7c8e1bbdde 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -278,42 +278,48 @@ static const FormatEntry format_entries[] = {
[AV_PIX_FMT_P416LE] = { 1, 1 },
};
-int ff_shuffle_filter_coefficients(SwsContext *c, int *filterPos, int
filterSize, int16_t *filter, int dstW){
+int ff_shuffle_filter_coefficients(SwsContext *c, int *filterPos,
+ int filterSize, int16_t *filter,
+ int dstW)
+{
#if ARCH_X86_64
-int i, j, k, l;
+int i, j, k;
int cpu_flags = av_get_cpu_flags();
+// avx2 hscale filter processes 16 pixel blocks.
+if (!filter || dstW % 16 != 0)
+return 0;
if (EXTERNAL_AVX2_FAST(cpu_flags) && !(cpu_flags &
AV_CPU_FLAG_SLOW_GATHER)) {
-if ((c->srcBpc == 8) && (c->dstBpc <= 14)){
-if (dstW % 16 == 0){
-if (filter != NULL){
-for (i = 0; i < dstW; i += 8){
-FFSWAP(int, filterPos[i + 2], filterPos[i+4]);
-FFSWAP(int, filterPos[i + 3], filterPos[i+5]);
-}
-if (filterSize > 4){
-int16_t *tmp2 = av_malloc(dstW * filterSize * 2);
-if (!tmp2)
-return AVERROR(ENOMEM);
-memcpy(tmp2, filter, dstW * filterSize * 2);
-for (i = 0; i < dstW; i += 16){//pixel
-for (k = 0; k < filterSize / 4; ++k){//fcoeff
-for (j = 0; j < 16; ++j){//inner pixel
-for (l = 0; l < 4; ++l){//coeff
-int from = i * filterSize + j *
filterSize + k * 4 + l;
-int to = (i) * filterSize + j * 4 + l
+ k * 64;
-filter[to] = tmp2[from];
-}
-}
-}
-}
-av_free(tmp2);
-}
-}
-}
+if ((c->srcBpc == 8) && (c->dstBpc <= 14)) {
+ int16_t *filterCopy = NULL;
+ if (filterSize > 4) {
+ if (!FF_ALLOC_TYPED_ARRAY(filterCopy, dstW * filterSize))
+ return AVERROR(ENOMEM);
+ memcpy(filterCopy, filter, dstW * filterSize * sizeof(int16_t));
+ }
+ // Do not swap filterPos for pixels which won't be processed by
+ // the main loop.
+ for (i = 0; i + 8 <= dstW; i += 8) {
+ FFSWAP(int, filterPos[i + 2], filterPos[i + 4]);
+ FFSWAP(int, filterPos[i + 3], filterPos[i + 5]);
+ }
+ if (filterSize > 4) {
+ // 16 pixels are processed at a time.
+ for (i = 0; i + 16 <= dstW; i += 16) {
+ // 4 filter coeffs are processed at a time.
+ for (k = 0; k + 4 <= filterSize; k += 4) {
+ for (j = 0; j < 16; ++j) {
+ int from = (i + j) * filterSize + k;
+ int to = i * filterSize + j * 4 + k * 16;
+ memcpy([to], [from], 4 *
sizeof(int16_t));
+ }
+ }
+ }
+ }
+ av_free(filterCopy);
}
}
-return 0;
#endif
+return 0;
}
int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libswscale: Check and propagate memory allocation errors from ff_shuffle_filter_coefficients.
ffmpeg | branch: master | Alan Kelly | Thu Feb 17
11:03:21 2022 +0100| [f1a5414c97a594d6f1d011860753794681ec56c5] | committer:
Michael Niedermayer
libswscale: Check and propagate memory allocation errors from
ff_shuffle_filter_coefficients.
Signed-off-by: Michael Niedermayer
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f1a5414c97a594d6f1d011860753794681ec56c5
---
libswscale/swscale_internal.h | 2 +-
libswscale/utils.c| 11 ---
2 files changed, 9 insertions(+), 4 deletions(-)
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index 3a78d95ba6..26d28d42e6 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -1144,5 +1144,5 @@ void ff_sws_slice_worker(void *priv, int jobnr, int
threadnr,
#define MAX_LINES_AHEAD 4
//shuffle filter and filterPos for hyScale and hcScale filters in avx2
-void ff_shuffle_filter_coefficients(SwsContext *c, int* filterPos, int
filterSize, int16_t *filter, int dstW);
+int ff_shuffle_filter_coefficients(SwsContext *c, int* filterPos, int
filterSize, int16_t *filter, int dstW);
#endif /* SWSCALE_SWSCALE_INTERNAL_H */
diff --git a/libswscale/utils.c b/libswscale/utils.c
index c5ea8853d5..344c87dfdf 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -278,7 +278,7 @@ static const FormatEntry format_entries[] = {
[AV_PIX_FMT_P416LE] = { 1, 1 },
};
-void ff_shuffle_filter_coefficients(SwsContext *c, int *filterPos, int
filterSize, int16_t *filter, int dstW){
+int ff_shuffle_filter_coefficients(SwsContext *c, int *filterPos, int
filterSize, int16_t *filter, int dstW){
#if ARCH_X86_64
int i, j, k, l;
int cpu_flags = av_get_cpu_flags();
@@ -292,6 +292,8 @@ void ff_shuffle_filter_coefficients(SwsContext *c, int
*filterPos, int filterSiz
}
if (filterSize > 4){
int16_t *tmp2 = av_malloc(dstW * filterSize * 2);
+if (!tmp2)
+return AVERROR(ENOMEM);
memcpy(tmp2, filter, dstW * filterSize * 2);
for (i = 0; i < dstW; i += 16){//pixel
for (k = 0; k < filterSize / 4; ++k){//fcoeff
@@ -310,6 +312,7 @@ void ff_shuffle_filter_coefficients(SwsContext *c, int
*filterPos, int filterSiz
}
}
}
+return 0;
#endif
}
@@ -1836,7 +1839,8 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter
*srcFilter,
get_local_pos(c, 0, 0, 0),
get_local_pos(c, 0, 0, 0))) < 0)
goto fail;
-ff_shuffle_filter_coefficients(c, c->hLumFilterPos,
c->hLumFilterSize, c->hLumFilter, dstW);
+if (ff_shuffle_filter_coefficients(c, c->hLumFilterPos,
c->hLumFilterSize, c->hLumFilter, dstW) < 0)
+goto nomem;
if ((ret = initFilter(>hChrFilter, >hChrFilterPos,
>hChrFilterSize, c->chrXInc,
c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
@@ -1846,7 +1850,8 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter
*srcFilter,
get_local_pos(c, c->chrSrcHSubSample,
c->src_h_chr_pos, 0),
get_local_pos(c, c->chrDstHSubSample,
c->dst_h_chr_pos, 0))) < 0)
goto fail;
-ff_shuffle_filter_coefficients(c, c->hChrFilterPos,
c->hChrFilterSize, c->hChrFilter, c->chrDstW);
+if (ff_shuffle_filter_coefficients(c, c->hChrFilterPos,
c->hChrFilterSize, c->hChrFilter, c->chrDstW) < 0)
+goto nomem;
}
} // initialize horizontal stuff
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libavutil/cpu: Add AV_CPU_FLAG_SLOW_GATHER.
ffmpeg | branch: master | Alan Kelly | Tue
Dec 21 20:56:41 2021 +0100| [ffbab99f2c22be06ef3c564fd38320d40e48a2b5] |
committer: James Almer
libavutil/cpu: Add AV_CPU_FLAG_SLOW_GATHER.
This flag is set on Haswell and earlier and all AMD cpus.
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ffbab99f2c22be06ef3c564fd38320d40e48a2b5
---
doc/APIchanges | 3 +++
libavutil/cpu.h | 1 +
libavutil/version.h | 4 ++--
libavutil/x86/cpu.c | 15 ++-
4 files changed, 20 insertions(+), 3 deletions(-)
diff --git a/doc/APIchanges b/doc/APIchanges
index 93fc45ced4..ac75040274 100644
--- a/doc/APIchanges
+++ b/doc/APIchanges
@@ -14,6 +14,9 @@ libavutil: 2021-04-27
API changes, most recent first:
+2021-12-21 - xx - lavu 57.12.100 - cpu.h
+ Add AV_CPU_FLAG_SLOW_GATHER.
+
2021-12-20 - xx - lavu 57.11.101 - display.h
Modified the documentation of av_display_rotation_set()
to match its longstanding actual behaviour of treating
diff --git a/libavutil/cpu.h b/libavutil/cpu.h
index ae443eccad..ce9bf14bf7 100644
--- a/libavutil/cpu.h
+++ b/libavutil/cpu.h
@@ -54,6 +54,7 @@
#define AV_CPU_FLAG_BMI10x2 ///< Bit Manipulation Instruction Set 1
#define AV_CPU_FLAG_BMI20x4 ///< Bit Manipulation Instruction Set 2
#define AV_CPU_FLAG_AVX512 0x10 ///< AVX-512 functions: requires OS
support even if YMM/ZMM registers aren't used
+#define AV_CPU_FLAG_SLOW_GATHER 0x200 ///< CPU has slow gathers.
#define AV_CPU_FLAG_ALTIVEC 0x0001 ///< standard
#define AV_CPU_FLAG_VSX 0x0002 ///< ISA 2.06
diff --git a/libavutil/version.h b/libavutil/version.h
index 678401fcf5..668f9206fe 100644
--- a/libavutil/version.h
+++ b/libavutil/version.h
@@ -79,8 +79,8 @@
*/
#define LIBAVUTIL_VERSION_MAJOR 57
-#define LIBAVUTIL_VERSION_MINOR 11
-#define LIBAVUTIL_VERSION_MICRO 101
+#define LIBAVUTIL_VERSION_MINOR 12
+#define LIBAVUTIL_VERSION_MICRO 100
#define LIBAVUTIL_VERSION_INT AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \
LIBAVUTIL_VERSION_MINOR, \
diff --git a/libavutil/x86/cpu.c b/libavutil/x86/cpu.c
index bcd41a50a2..441b4695d5 100644
--- a/libavutil/x86/cpu.c
+++ b/libavutil/x86/cpu.c
@@ -146,8 +146,21 @@ int ff_get_cpu_flags_x86(void)
if (max_std_level >= 7) {
cpuid(7, eax, ebx, ecx, edx);
#if HAVE_AVX2
-if ((rval & AV_CPU_FLAG_AVX) && (ebx & 0x0020))
+if ((rval & AV_CPU_FLAG_AVX) && (ebx & 0x0020)) {
rval |= AV_CPU_FLAG_AVX2;
+cpuid(1, eax, ebx, ecx, std_caps);
+family = ((eax >> 8) & 0xf) + ((eax >> 20) & 0xff);
+model = ((eax >> 4) & 0xf) + ((eax >> 12) & 0xf0);
+/* Haswell has slow gather */
+if (!strncmp(vendor.c, "GenuineIntel", 12))
+if (family == 6 && model < 70)
+rval |= AV_CPU_FLAG_SLOW_GATHER;
+/* Zen 3 and earlier have slow gather */
+if (!strncmp(vendor.c, "AuthenticAMD", 12))
+if (family <= 0x19)
+rval |= AV_CPU_FLAG_SLOW_GATHER;
+}
+
#if HAVE_AVX512 /* F, CD, BW, DQ, VL */
if ((xcr0_lo & 0xe0) == 0xe0) { /* OPMASK/ZMM state */
if ((rval & AV_CPU_FLAG_AVX2) && (ebx & 0xd003) == 0xd003)
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libswscale: Test AV_CPU_FLAG_SLOW_GATHER for hscale functions.
ffmpeg | branch: master | Alan Kelly | Mon
Dec 20 15:45:45 2021 +0100| [eebe406c808e6061ee76e93a616537b5369dbf40] |
committer: James Almer
libswscale: Test AV_CPU_FLAG_SLOW_GATHER for hscale functions.
This is instead of EXTERNAL_AVX2_FAST so that the avx2 hscale functions
are only used where they are faster.
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=eebe406c808e6061ee76e93a616537b5369dbf40
---
libswscale/utils.c| 2 +-
libswscale/x86/swscale.c | 2 +-
tests/checkasm/sw_scale.c | 2 +-
3 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/libswscale/utils.c b/libswscale/utils.c
index d4a72d3ce1..7158384f0b 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -282,7 +282,7 @@ void ff_shuffle_filter_coefficients(SwsContext *c, int
*filterPos, int filterSiz
#if ARCH_X86_64
int i, j, k, l;
int cpu_flags = av_get_cpu_flags();
-if (EXTERNAL_AVX2_FAST(cpu_flags)){
+if (EXTERNAL_AVX2_FAST(cpu_flags) && !(cpu_flags &
AV_CPU_FLAG_SLOW_GATHER)) {
if ((c->srcBpc == 8) && (c->dstBpc <= 14)){
if (dstW % 16 == 0){
if (filter != NULL){
diff --git a/libswscale/x86/swscale.c b/libswscale/x86/swscale.c
index c49a05c37b..ffc7691c12 100644
--- a/libswscale/x86/swscale.c
+++ b/libswscale/x86/swscale.c
@@ -578,7 +578,7 @@ switch(c->dstBpc){ \
break; \
}
-if (EXTERNAL_AVX2_FAST(cpu_flags)) {
+if (EXTERNAL_AVX2_FAST(cpu_flags) && !(cpu_flags &
AV_CPU_FLAG_SLOW_GATHER)) {
if ((c->srcBpc == 8) && (c->dstBpc <= 14)) {
if (c->chrDstW % 16 == 0)
ASSIGN_AVX2_SCALE_FUNC(c->hcScale, c->hChrFilterSize);
diff --git a/tests/checkasm/sw_scale.c b/tests/checkasm/sw_scale.c
index f4912e6c2c..3c0a083b42 100644
--- a/tests/checkasm/sw_scale.c
+++ b/tests/checkasm/sw_scale.c
@@ -217,7 +217,7 @@ static void check_hscale(void)
}
ff_sws_init_scale(ctx);
memcpy(filterAvx2, filter, sizeof(uint16_t) * (SRC_PIXELS *
MAX_FILTER_WIDTH + MAX_FILTER_WIDTH));
-if (cpu_flags & AV_CPU_FLAG_AVX2)
+if ((cpu_flags & AV_CPU_FLAG_AVX2) && !(cpu_flags &
AV_CPU_FLAG_SLOW_GATHER))
ff_shuffle_filter_coefficients(ctx, filterPosAvx, width,
filterAvx2, SRC_PIXELS);
if (check_func(ctx->hcScale, "hscale_%d_to_%d_width%d",
ctx->srcBpc, ctx->dstBpc + 1, width)) {
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] x86/scale_avx2: Change asm indent from 2 to 4 spaces.
ffmpeg | branch: master | Alan Kelly | Thu Dec 16 17:27:10 2021 +0100| [9092e58c4469f5488f305fdb85e34e61bba7b04a] | committer: James Almer x86/scale_avx2: Change asm indent from 2 to 4 spaces. Signed-off-by: James Almer > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9092e58c4469f5488f305fdb85e34e61bba7b04a --- libswscale/x86/scale_avx2.asm | 96 +-- 1 file changed, 48 insertions(+), 48 deletions(-) diff --git a/libswscale/x86/scale_avx2.asm b/libswscale/x86/scale_avx2.asm index 2cd7e968d3..eb472db12f 100644 --- a/libswscale/x86/scale_avx2.asm +++ b/libswscale/x86/scale_avx2.asm @@ -45,63 +45,63 @@ SECTION .text %macro SCALE_FUNC 1 cglobal hscale8to15_%1, 7, 9, 16, pos0, dst, w, srcmem, filter, fltpos, fltsize, count, inner - pxor m0, m0 - mova m15, [swizzle] - mov countq, $0 - movsxd wq, wd +pxor m0, m0 +mova m15, [swizzle] +mov countq, $0 +movsxd wq, wd %ifidn %1, X4 - mova m14, [four] - shr fltsized, 2 +mova m14, [four] +shr fltsized, 2 %endif .loop: - movu m1, [fltposq] - movu m2, [fltposq+32] +movu m1, [fltposq] +movu m2, [fltposq+32] %ifidn %1, X4 - pxor m9, m9 - pxor m10, m10 - pxor m11, m11 - pxor m12, m12 - mov innerq, $0 +pxor m9, m9 +pxor m10, m10 +pxor m11, m11 +pxor m12, m12 +mov innerq, $0 .innerloop: %endif - vpcmpeqd m13, m13 - vpgatherdd m3,[srcmemq + m1], m13 - vpcmpeqd m13, m13 - vpgatherdd m4,[srcmemq + m2], m13 - vpunpcklbw m5, m3, m0 - vpunpckhbw m6, m3, m0 - vpunpcklbw m7, m4, m0 - vpunpckhbw m8, m4, m0 - vpmaddwd m5, m5, [filterq] - vpmaddwd m6, m6, [filterq + 32] - vpmaddwd m7, m7, [filterq + 64] - vpmaddwd m8, m8, [filterq + 96] - add filterq, $80 +vpcmpeqd m13, m13 +vpgatherdd m3,[srcmemq + m1], m13 +vpcmpeqd m13, m13 +vpgatherdd m4,[srcmemq + m2], m13 +vpunpcklbw m5, m3, m0 +vpunpckhbw m6, m3, m0 +vpunpcklbw m7, m4, m0 +vpunpckhbw m8, m4, m0 +vpmaddwd m5, m5, [filterq] +vpmaddwd m6, m6, [filterq + 32] +vpmaddwd m7, m7, [filterq + 64] +vpmaddwd m8, m8, [filterq + 96] +add filterq, $80 %ifidn %1, X4 - paddd m9, m5 - paddd m10, m6 - paddd m11, m7 - paddd m12, m8 - paddd m1, m14 - paddd m2, m14 - add innerq, $1 - cmp innerq, fltsizeq - jl .innerloop - vphaddd m5, m9, m10 - vphaddd m6, m11, m12 +paddd m9, m5 +paddd m10, m6 +paddd m11, m7 +paddd m12, m8 +paddd m1, m14 +paddd m2, m14 +add innerq, $1 +cmp innerq, fltsizeq +jl .innerloop +vphaddd m5, m9, m10 +vphaddd m6, m11, m12 %else - vphaddd m5, m5, m6 - vphaddd m6, m7, m8 +vphaddd m5, m5, m6 +vphaddd m6, m7, m8 %endif - vpsrad m5, 7 - vpsrad m6, 7 - vpackssdw m5, m5, m6 - vpermd m5, m15, m5 - vmovdqu [dstq + countq * 2], m5 - add fltposq, $40 - add countq, $10 - cmp countq, wq - jl .loop +vpsrad m5, 7 +vpsrad m6, 7 +vpackssdw m5, m5, m6 +vpermd m5, m15, m5 +vmovdqu [dstq + countq * 2], m5 +add fltposq, $40 +add countq, $10 +cmp countq, wq +jl .loop REP_RET %endmacro ___ ffmpeg-cvslog mailing list ffmpeg-cvslog@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog To unsubscribe, visit link above, or email ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] x86/swscale: fix minor coding style issues
ffmpeg | branch: master | Alan Kelly | Thu Dec 16
17:05:48 2021 +0100| [86663963e6419a127cf52a03758855f4f8f8689f] | committer:
James Almer
x86/swscale: fix minor coding style issues
Signed-off-by: James Almer
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=86663963e6419a127cf52a03758855f4f8f8689f
---
libswscale/x86/swscale.c | 14 +++---
tests/checkasm/sw_scale.c | 3 +--
2 files changed, 8 insertions(+), 9 deletions(-)
diff --git a/libswscale/x86/swscale.c b/libswscale/x86/swscale.c
index 164b06d6ba..c49a05c37b 100644
--- a/libswscale/x86/swscale.c
+++ b/libswscale/x86/swscale.c
@@ -578,13 +578,13 @@ switch(c->dstBpc){ \
break; \
}
-if (EXTERNAL_AVX2_FAST(cpu_flags)){
- if ((c->srcBpc == 8) && (c->dstBpc <= 14)){
-if(c->chrDstW % 16 == 0)
- ASSIGN_AVX2_SCALE_FUNC(c->hcScale, c->hChrFilterSize);
-if(c->dstW % 16 == 0)
- ASSIGN_AVX2_SCALE_FUNC(c->hyScale, c->hLumFilterSize);
- }
+if (EXTERNAL_AVX2_FAST(cpu_flags)) {
+if ((c->srcBpc == 8) && (c->dstBpc <= 14)) {
+if (c->chrDstW % 16 == 0)
+ASSIGN_AVX2_SCALE_FUNC(c->hcScale, c->hChrFilterSize);
+if (c->dstW % 16 == 0)
+ASSIGN_AVX2_SCALE_FUNC(c->hyScale, c->hLumFilterSize);
+}
}
if (EXTERNAL_AVX2_FAST(cpu_flags)) {
diff --git a/tests/checkasm/sw_scale.c b/tests/checkasm/sw_scale.c
index 011cb46428..f4912e6c2c 100644
--- a/tests/checkasm/sw_scale.c
+++ b/tests/checkasm/sw_scale.c
@@ -217,9 +217,8 @@ static void check_hscale(void)
}
ff_sws_init_scale(ctx);
memcpy(filterAvx2, filter, sizeof(uint16_t) * (SRC_PIXELS *
MAX_FILTER_WIDTH + MAX_FILTER_WIDTH));
-if (cpu_flags & AV_CPU_FLAG_AVX2){
+if (cpu_flags & AV_CPU_FLAG_AVX2)
ff_shuffle_filter_coefficients(ctx, filterPosAvx, width,
filterAvx2, SRC_PIXELS);
-}
if (check_func(ctx->hcScale, "hscale_%d_to_%d_width%d",
ctx->srcBpc, ctx->dstBpc + 1, width)) {
memset(dst0, 0, SRC_PIXELS * sizeof(dst0[0]));
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libswscale: Adds ff_hscale8to15_4_avx2 and ff_hscale8to15_X4_avx2 for all filter sizes.
ffmpeg | branch: master | Alan Kelly | Wed Dec 15
10:35:02 2021 +0100| [f900a19fa94b1a55b660ec2e5c13419d59754bc0] | committer:
James Almer
libswscale: Adds ff_hscale8to15_4_avx2 and ff_hscale8to15_X4_avx2 for all
filter sizes.
Fixes so that fate under 64 bit Windows passes.
These functions replace all ff_hscale8to15_*_ssse3 when avx2 is available.
Signed-off-by: James Almer
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f900a19fa94b1a55b660ec2e5c13419d59754bc0
---
libswscale/swscale_internal.h | 2 +
libswscale/utils.c| 37 ++
libswscale/x86/Makefile | 1 +
libswscale/x86/scale_avx2.asm | 112 ++
libswscale/x86/swscale.c | 19 +++
tests/checkasm/sw_scale.c | 20 ++--
6 files changed, 186 insertions(+), 5 deletions(-)
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index 708facba67..64aa0b9804 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -1105,4 +1105,6 @@ void ff_sws_slice_worker(void *priv, int jobnr, int
threadnr,
//number of extra lines to process
#define MAX_LINES_AHEAD 4
+//shuffle filter and filterPos for hyScale and hcScale filters in avx2
+void ff_shuffle_filter_coefficients(SwsContext *c, int* filterPos, int
filterSize, int16_t *filter, int dstW);
#endif /* SWSCALE_SWSCALE_INTERNAL_H */
diff --git a/libswscale/utils.c b/libswscale/utils.c
index ae92ac9fbc..d4a72d3ce1 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -278,6 +278,41 @@ static const FormatEntry format_entries[] = {
[AV_PIX_FMT_P416LE] = { 1, 0 },
};
+void ff_shuffle_filter_coefficients(SwsContext *c, int *filterPos, int
filterSize, int16_t *filter, int dstW){
+#if ARCH_X86_64
+int i, j, k, l;
+int cpu_flags = av_get_cpu_flags();
+if (EXTERNAL_AVX2_FAST(cpu_flags)){
+if ((c->srcBpc == 8) && (c->dstBpc <= 14)){
+if (dstW % 16 == 0){
+if (filter != NULL){
+for (i = 0; i < dstW; i += 8){
+FFSWAP(int, filterPos[i + 2], filterPos[i+4]);
+FFSWAP(int, filterPos[i + 3], filterPos[i+5]);
+}
+if (filterSize > 4){
+int16_t *tmp2 = av_malloc(dstW * filterSize * 2);
+memcpy(tmp2, filter, dstW * filterSize * 2);
+for (i = 0; i < dstW; i += 16){//pixel
+for (k = 0; k < filterSize / 4; ++k){//fcoeff
+for (j = 0; j < 16; ++j){//inner pixel
+for (l = 0; l < 4; ++l){//coeff
+int from = i * filterSize + j *
filterSize + k * 4 + l;
+int to = (i) * filterSize + j * 4 + l
+ k * 64;
+filter[to] = tmp2[from];
+}
+}
+}
+}
+av_free(tmp2);
+}
+}
+}
+}
+}
+#endif
+}
+
int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
{
return (unsigned)pix_fmt < FF_ARRAY_ELEMS(format_entries) ?
@@ -1801,6 +1836,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter
*srcFilter,
get_local_pos(c, 0, 0, 0),
get_local_pos(c, 0, 0, 0))) < 0)
goto fail;
+ff_shuffle_filter_coefficients(c, c->hLumFilterPos,
c->hLumFilterSize, c->hLumFilter, dstW);
if ((ret = initFilter(>hChrFilter, >hChrFilterPos,
>hChrFilterSize, c->chrXInc,
c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
@@ -1810,6 +1846,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter
*srcFilter,
get_local_pos(c, c->chrSrcHSubSample,
c->src_h_chr_pos, 0),
get_local_pos(c, c->chrDstHSubSample,
c->dst_h_chr_pos, 0))) < 0)
goto fail;
+ff_shuffle_filter_coefficients(c, c->hChrFilterPos,
c->hChrFilterSize, c->hChrFilter, c->chrDstW);
}
} // initialize horizontal stuff
diff --git a/libswscale/x86/Makefile b/libswscale/x86/Makefile
index bfe383364e..68391494be 100644
--- a/libswscale/x86/Makefile
+++ b/libswscale/x86/Makefile
@@ -11,6 +11,7 @@ OBJS-$(CONFIG_XMM_CLOBBER_TEST) += x86/w64xmmtest.o
X86ASM-OBJS += x86/input.o \
x86/output.o \
x86/scale.o \
+ x86/scale_avx2.o \
[FFmpeg-cvslog] libswscale/x86/swscale: Only call ff_yuv2yuvX functions if the input size is > 0
ffmpeg | branch: release/4.4 | Alan Kelly | Thu Apr 1
12:00:16 2021 +0200| [95aacf30e3803c57d91ff62975b375e394f61d49] | committer:
Michael Niedermayer
libswscale/x86/swscale: Only call ff_yuv2yuvX functions if the input size is > 0
Signed-off-by: Michael Niedermayer
(cherry picked from commit dc57762cb43619f91fd2a5d95510fa3b14cfeaaf)
Signed-off-by: Michael Niedermayer
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=95aacf30e3803c57d91ff62975b375e394f61d49
---
libswscale/x86/swscale.c | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/libswscale/x86/swscale.c b/libswscale/x86/swscale.c
index cc9e8b0155..0848a31461 100644
--- a/libswscale/x86/swscale.c
+++ b/libswscale/x86/swscale.c
@@ -197,7 +197,8 @@ static void yuv2yuvX_ ##opt(const int16_t *filter, int
filterSize, \
const int16_t **src, uint8_t *dest, int dstW, \
const uint8_t *dither, int offset) \
{ \
-ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset, dstW +
offset, dither, offset); \
+if(dstW > 0) \
+ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset, dstW +
offset, dither, offset); \
return; \
}
@@ -215,7 +216,8 @@ static void yuv2yuvX_ ##opt(const int16_t *filter, int
filterSize, \
yuv2yuvX_mmx(filter, filterSize, src, dest, dstW, dither, offset); \
return; \
} \
-ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset,
pixelsProcessed + offset, dither, offset); \
+if(pixelsProcessed > 0) \
+ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset,
pixelsProcessed + offset, dither, offset); \
if(remainder > 0){ \
ff_yuv2yuvX_mmx(filter, filterSize - 1, pixelsProcessed, dest - offset,
pixelsProcessed + remainder + offset, dither, offset); \
} \
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libswscale/x86/yuv2yuvX: Removes unrolling for mmx and mmxext
ffmpeg | branch: release/4.4 | Alan Kelly | Thu Apr 1 12:00:15 2021 +0200| [4aeedf4c2a8f35be667d5dd40c84bd27730ef1db] | committer: Michael Niedermayer libswscale/x86/yuv2yuvX: Removes unrolling for mmx and mmxext Signed-off-by: Michael Niedermayer (cherry picked from commit 3ce8d092448827842c451807f03010ad5129fd8f) Signed-off-by: Michael Niedermayer > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=4aeedf4c2a8f35be667d5dd40c84bd27730ef1db --- libswscale/x86/yuv2yuvX.asm | 14 +- 1 file changed, 13 insertions(+), 1 deletion(-) diff --git a/libswscale/x86/yuv2yuvX.asm b/libswscale/x86/yuv2yuvX.asm index 521880dabe..b6294cb919 100644 --- a/libswscale/x86/yuv2yuvX.asm +++ b/libswscale/x86/yuv2yuvX.asm @@ -37,8 +37,10 @@ SECTION .text cglobal yuv2yuvX, 7, 7, 8, filter, filterSize, src, dest, dstW, dither, offset %if notcpuflag(sse3) %define movr mova +%define unroll 1 %else %define movr movdqu +%define unroll 2 %endif movsxdifnidn dstWq, dstWd movsxdifnidn offsetq, offsetd @@ -70,8 +72,10 @@ cglobal yuv2yuvX, 7, 7, 8, filter, filterSize, src, dest, dstW, dither, offset .outerloop: mova m4, m7 mova m3, m7 +%if cpuflag(sse3) mova m6, m7 mova m1, m7 +%endif .loop: %if cpuflag(avx2) vpbroadcastq m0, [filterSizeq + 8] @@ -84,28 +88,36 @@ cglobal yuv2yuvX, 7, 7, 8, filter, filterSize, src, dest, dstW, dither, offset pmulhw m5, m0, [srcq + offsetq * 2 + mmsize] paddwm3, m3, m2 paddwm4, m4, m5 +%if cpuflag(sse3) pmulhw m2, m0, [srcq + offsetq * 2 + 2 * mmsize] pmulhw m5, m0, [srcq + offsetq * 2 + 3 * mmsize] paddwm6, m6, m2 paddwm1, m1, m5 +%endif add filterSizeq, $10 mov srcq, [filterSizeq] test srcq, srcq jnz .loop psrawm3, m3, 3 psrawm4, m4, 3 +%if cpuflag(sse3) psrawm6, m6, 3 psrawm1, m1, 3 +%endif packuswb m3, m3, m4 +%if cpuflag(sse3) packuswb m6, m6, m1 +%endif mov srcq, [filterq] %if cpuflag(avx2) vpermq m3, m3, 216 vpermq m6, m6, 216 %endif movr [destq + offsetq], m3 +%if cpuflag(sse3) movr [destq + offsetq + mmsize], m6 -add offsetq, mmsize * 2 +%endif +add offsetq, mmsize * unroll mov filterSizeq, filterq cmp offsetq, dstWq jb .outerloop ___ ffmpeg-cvslog mailing list ffmpeg-cvslog@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog To unsubscribe, visit link above, or email ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] tests/checkasm/sw_scale: adds additional tests sizes for yux2yuvX
ffmpeg | branch: release/4.4 | Alan Kelly | Thu Apr 1
12:00:17 2021 +0200| [6bc2058d00b119d265c9970eac213d2922d15129] | committer:
Michael Niedermayer
tests/checkasm/sw_scale: adds additional tests sizes for yux2yuvX
Signed-off-by: Michael Niedermayer
(cherry picked from commit e1484bc455dff500f8b35b58d434924bca0e03d6)
Signed-off-by: Michael Niedermayer
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=6bc2058d00b119d265c9970eac213d2922d15129
---
tests/checkasm/sw_scale.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/tests/checkasm/sw_scale.c b/tests/checkasm/sw_scale.c
index a10118704b..3ac0f9082f 100644
--- a/tests/checkasm/sw_scale.c
+++ b/tests/checkasm/sw_scale.c
@@ -68,8 +68,8 @@ static void check_yuv2yuvX(void)
#define FILTER_SIZES 4
static const int filter_sizes[FILTER_SIZES] = {1, 4, 8, 16};
#define LARGEST_INPUT_SIZE 512
-#define INPUT_SIZES 4
-static const int input_sizes[INPUT_SIZES] = {128, 144, 256, 512};
+#define INPUT_SIZES 6
+static const int input_sizes[INPUT_SIZES] = {8, 24, 128, 144, 256, 512};
declare_func_emms(AV_CPU_FLAG_MMX, void, const int16_t *filter,
int filterSize, const int16_t **src, uint8_t *dest,
@@ -107,7 +107,7 @@ static void check_yuv2yuvX(void)
for(j = 0; j < 4; ++j)
vFilterData[i].coeff[j + 4] = filter_coeff[i];
}
-if (check_func(ctx->yuv2planeX, "yuv2yuvX_%d_%d",
filter_sizes[fsi], osi)){
+if (check_func(ctx->yuv2planeX, "yuv2yuvX_%d_%d_%d",
filter_sizes[fsi], osi, dstW)){
memset(dst0, 0, LARGEST_INPUT_SIZE * sizeof(dst0[0]));
memset(dst1, 0, LARGEST_INPUT_SIZE * sizeof(dst1[0]));
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libswscale/x86/swscale: Only call ff_yuv2yuvX functions if the input size is > 0
ffmpeg | branch: master | Alan Kelly | Thu Apr 1
12:00:16 2021 +0200| [dc57762cb43619f91fd2a5d95510fa3b14cfeaaf] | committer:
Michael Niedermayer
libswscale/x86/swscale: Only call ff_yuv2yuvX functions if the input size is > 0
Signed-off-by: Michael Niedermayer
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=dc57762cb43619f91fd2a5d95510fa3b14cfeaaf
---
libswscale/x86/swscale.c | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/libswscale/x86/swscale.c b/libswscale/x86/swscale.c
index cc9e8b0155..0848a31461 100644
--- a/libswscale/x86/swscale.c
+++ b/libswscale/x86/swscale.c
@@ -197,7 +197,8 @@ static void yuv2yuvX_ ##opt(const int16_t *filter, int
filterSize, \
const int16_t **src, uint8_t *dest, int dstW, \
const uint8_t *dither, int offset) \
{ \
-ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset, dstW +
offset, dither, offset); \
+if(dstW > 0) \
+ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset, dstW +
offset, dither, offset); \
return; \
}
@@ -215,7 +216,8 @@ static void yuv2yuvX_ ##opt(const int16_t *filter, int
filterSize, \
yuv2yuvX_mmx(filter, filterSize, src, dest, dstW, dither, offset); \
return; \
} \
-ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset,
pixelsProcessed + offset, dither, offset); \
+if(pixelsProcessed > 0) \
+ff_yuv2yuvX_ ##opt(filter, filterSize - 1, 0, dest - offset,
pixelsProcessed + offset, dither, offset); \
if(remainder > 0){ \
ff_yuv2yuvX_mmx(filter, filterSize - 1, pixelsProcessed, dest - offset,
pixelsProcessed + remainder + offset, dither, offset); \
} \
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] libswscale/x86/yuv2yuvX: Removes unrolling for mmx and mmxext
ffmpeg | branch: master | Alan Kelly | Thu Apr 1 12:00:15 2021 +0200| [3ce8d092448827842c451807f03010ad5129fd8f] | committer: Michael Niedermayer libswscale/x86/yuv2yuvX: Removes unrolling for mmx and mmxext Signed-off-by: Michael Niedermayer > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3ce8d092448827842c451807f03010ad5129fd8f --- libswscale/x86/yuv2yuvX.asm | 14 +- 1 file changed, 13 insertions(+), 1 deletion(-) diff --git a/libswscale/x86/yuv2yuvX.asm b/libswscale/x86/yuv2yuvX.asm index 521880dabe..b6294cb919 100644 --- a/libswscale/x86/yuv2yuvX.asm +++ b/libswscale/x86/yuv2yuvX.asm @@ -37,8 +37,10 @@ SECTION .text cglobal yuv2yuvX, 7, 7, 8, filter, filterSize, src, dest, dstW, dither, offset %if notcpuflag(sse3) %define movr mova +%define unroll 1 %else %define movr movdqu +%define unroll 2 %endif movsxdifnidn dstWq, dstWd movsxdifnidn offsetq, offsetd @@ -70,8 +72,10 @@ cglobal yuv2yuvX, 7, 7, 8, filter, filterSize, src, dest, dstW, dither, offset .outerloop: mova m4, m7 mova m3, m7 +%if cpuflag(sse3) mova m6, m7 mova m1, m7 +%endif .loop: %if cpuflag(avx2) vpbroadcastq m0, [filterSizeq + 8] @@ -84,28 +88,36 @@ cglobal yuv2yuvX, 7, 7, 8, filter, filterSize, src, dest, dstW, dither, offset pmulhw m5, m0, [srcq + offsetq * 2 + mmsize] paddwm3, m3, m2 paddwm4, m4, m5 +%if cpuflag(sse3) pmulhw m2, m0, [srcq + offsetq * 2 + 2 * mmsize] pmulhw m5, m0, [srcq + offsetq * 2 + 3 * mmsize] paddwm6, m6, m2 paddwm1, m1, m5 +%endif add filterSizeq, $10 mov srcq, [filterSizeq] test srcq, srcq jnz .loop psrawm3, m3, 3 psrawm4, m4, 3 +%if cpuflag(sse3) psrawm6, m6, 3 psrawm1, m1, 3 +%endif packuswb m3, m3, m4 +%if cpuflag(sse3) packuswb m6, m6, m1 +%endif mov srcq, [filterq] %if cpuflag(avx2) vpermq m3, m3, 216 vpermq m6, m6, 216 %endif movr [destq + offsetq], m3 +%if cpuflag(sse3) movr [destq + offsetq + mmsize], m6 -add offsetq, mmsize * 2 +%endif +add offsetq, mmsize * unroll mov filterSizeq, filterq cmp offsetq, dstWq jb .outerloop ___ ffmpeg-cvslog mailing list ffmpeg-cvslog@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog To unsubscribe, visit link above, or email ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] tests/checkasm/sw_scale: adds additional tests sizes for yux2yuvX
ffmpeg | branch: master | Alan Kelly | Thu Apr 1
12:00:17 2021 +0200| [e1484bc455dff500f8b35b58d434924bca0e03d6] | committer:
Michael Niedermayer
tests/checkasm/sw_scale: adds additional tests sizes for yux2yuvX
Signed-off-by: Michael Niedermayer
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e1484bc455dff500f8b35b58d434924bca0e03d6
---
tests/checkasm/sw_scale.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/tests/checkasm/sw_scale.c b/tests/checkasm/sw_scale.c
index a10118704b..3ac0f9082f 100644
--- a/tests/checkasm/sw_scale.c
+++ b/tests/checkasm/sw_scale.c
@@ -68,8 +68,8 @@ static void check_yuv2yuvX(void)
#define FILTER_SIZES 4
static const int filter_sizes[FILTER_SIZES] = {1, 4, 8, 16};
#define LARGEST_INPUT_SIZE 512
-#define INPUT_SIZES 4
-static const int input_sizes[INPUT_SIZES] = {128, 144, 256, 512};
+#define INPUT_SIZES 6
+static const int input_sizes[INPUT_SIZES] = {8, 24, 128, 144, 256, 512};
declare_func_emms(AV_CPU_FLAG_MMX, void, const int16_t *filter,
int filterSize, const int16_t **src, uint8_t *dest,
@@ -107,7 +107,7 @@ static void check_yuv2yuvX(void)
for(j = 0; j < 4; ++j)
vFilterData[i].coeff[j + 4] = filter_coeff[i];
}
-if (check_func(ctx->yuv2planeX, "yuv2yuvX_%d_%d",
filter_sizes[fsi], osi)){
+if (check_func(ctx->yuv2planeX, "yuv2yuvX_%d_%d_%d",
filter_sizes[fsi], osi, dstW)){
memset(dst0, 0, LARGEST_INPUT_SIZE * sizeof(dst0[0]));
memset(dst1, 0, LARGEST_INPUT_SIZE * sizeof(dst1[0]));
___
ffmpeg-cvslog mailing list
ffmpeg-cvslog@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog
To unsubscribe, visit link above, or email
ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] checkasm/sw_scale: properly initialize src_pixer and filter_coeff buffers
ffmpeg | branch: master | Alan Kelly | Fri Feb 19 14:55:39 2021 +0100| [ee18edb13a9ae3041df961dd5003c2055b5cab35] | committer: James Almer checkasm/sw_scale: properly initialize src_pixer and filter_coeff buffers Fixes valgrind uninitialised value warnings. Signed-off-by: James Almer > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ee18edb13a9ae3041df961dd5003c2055b5cab35 --- tests/checkasm/sw_scale.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tests/checkasm/sw_scale.c b/tests/checkasm/sw_scale.c index 7504f8b45f..dee1af820c 100644 --- a/tests/checkasm/sw_scale.c +++ b/tests/checkasm/sw_scale.c @@ -86,8 +86,8 @@ static void check_yuv2yuvX(void) uint16_t coeff[8]; } *vFilterData; uint8_t d_val = rnd(); -randomize_buffers(filter_coeff, LARGEST_FILTER); -randomize_buffers(src_pixels, LARGEST_FILTER * LARGEST_INPUT_SIZE); +randomize_buffers((uint8_t*)src_pixels, LARGEST_FILTER * LARGEST_INPUT_SIZE * sizeof(int16_t)); +randomize_buffers((uint8_t*)filter_coeff, LARGEST_FILTER * sizeof(int16_t)); ctx = sws_alloc_context(); if (sws_init_context(ctx, NULL, NULL) < 0) fail(); ___ ffmpeg-cvslog mailing list ffmpeg-cvslog@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-cvslog To unsubscribe, visit link above, or email ffmpeg-cvslog-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-cvslog] swscale: move yuv2yuvX_sse3 to yasm, unrolls main loop
ffmpeg | branch: master | Alan Kelly | Thu
Jan 14 15:47:03 2021 +0100| [554c2bc7086f49ef5a6a989ad6bc4bc11807eb6f] |
committer: Paul B Mahol
swscale: move yuv2yuvX_sse3 to yasm, unrolls main loop
And other small optimizations for ~20% speedup.
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=554c2bc7086f49ef5a6a989ad6bc4bc11807eb6f
---
libswscale/x86/Makefile | 1 +
libswscale/x86/swscale.c | 130 +++-
libswscale/x86/swscale_template.c | 82 ---
libswscale/x86/yuv2yuvX.asm | 136 ++
tests/checkasm/sw_scale.c | 103 +
5 files changed, 294 insertions(+), 158 deletions(-)
diff --git a/libswscale/x86/Makefile b/libswscale/x86/Makefile
index 831d5359aa..bfe383364e 100644
--- a/libswscale/x86/Makefile
+++ b/libswscale/x86/Makefile
@@ -13,3 +13,4 @@ X86ASM-OBJS += x86/input.o
\
x86/scale.o \
x86/rgb_2_rgb.o \
x86/yuv_2_rgb.o \
+ x86/yuv2yuvX.o \
diff --git a/libswscale/x86/swscale.c b/libswscale/x86/swscale.c
index 15c0b22f20..3df193a067 100644
--- a/libswscale/x86/swscale.c
+++ b/libswscale/x86/swscale.c
@@ -63,6 +63,16 @@ DECLARE_ASM_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) =
0x8080808080808080ULL;
DECLARE_ASM_ALIGNED(8, const uint64_t, ff_w)=
0x0001000100010001ULL;
+#define YUV2YUVX_FUNC_DECL(opt) \
+static void yuv2yuvX_ ##opt(const int16_t *filter, int filterSize, const
int16_t **src, \
+ uint8_t *dest, int dstW, \
+ const uint8_t *dither, int offset); \
+
+YUV2YUVX_FUNC_DECL(mmx)
+YUV2YUVX_FUNC_DECL(mmxext)
+YUV2YUVX_FUNC_DECL(sse3)
+YUV2YUVX_FUNC_DECL(avx2)
+
//MMX versions
#if HAVE_MMX_INLINE
#undef RENAME
@@ -198,81 +208,44 @@ void ff_updateMMXDitherTables(SwsContext *c, int dstY)
}
#if HAVE_MMXEXT
-static void yuv2yuvX_sse3(const int16_t *filter, int filterSize,
- const int16_t **src, uint8_t *dest, int dstW,
- const uint8_t *dither, int offset)
-{
-if(((uintptr_t)dest) & 15){
-yuv2yuvX_mmxext(filter, filterSize, src, dest, dstW, dither, offset);
-return;
-}
-filterSize--;
-#define MAIN_FUNCTION \
-"pxor %%xmm0, %%xmm0 \n\t" \
-"punpcklbw %%xmm0, %%xmm3 \n\t" \
-"movd %4, %%xmm1 \n\t" \
-"punpcklwd %%xmm1, %%xmm1 \n\t" \
-"punpckldq %%xmm1, %%xmm1 \n\t" \
-"punpcklqdq %%xmm1, %%xmm1 \n\t" \
-"psllw $3, %%xmm1 \n\t" \
-"paddw %%xmm1, %%xmm3 \n\t" \
-"psraw $4, %%xmm3 \n\t" \
-"movdqa %%xmm3, %%xmm4 \n\t" \
-"movdqa %%xmm3, %%xmm7 \n\t" \
-"movl %3, %%ecx \n\t" \
-"mov %0, %%"FF_REG_d"\n\t"\
-"mov(%%"FF_REG_d"), %%"FF_REG_S" \n\t"\
-".p2align 4 \n\t" /* FIXME
Unroll? */\
-"1: \n\t"\
-"movddup 8(%%"FF_REG_d"), %%xmm0 \n\t" /*
filterCoeff */\
-"movdqa (%%"FF_REG_S", %%"FF_REG_c", 2), %%xmm2 \n\t" /*
srcData */\
-"movdqa16(%%"FF_REG_S", %%"FF_REG_c", 2), %%xmm5 \n\t" /*
srcData */\
-"add$16, %%"FF_REG_d"\n\t"\
-"mov(%%"FF_REG_d"), %%"FF_REG_S" \n\t"\
-"test %%"FF_REG_S", %%"FF_REG_S" \n\t"\
-"pmulhw %%xmm0, %%xmm2 \n\t"\
-"pmulhw %%xmm0, %%xmm5 \n\t"\
-"paddw%%xmm2, %%xmm3 \n\t"\
-"paddw%%xmm5, %%xmm4 \n\t"\
-" jnz1b \n\t"\
-"psraw $3, %%xmm3 \n\t"\
-"psraw $3, %%xmm4 \n\t"\
-"packuswb %%xmm4, %%xmm3 \n\t"\
-"movntdq %%xmm3, (%1, %%"FF_REG_c")
