[FFmpeg-cvslog] libswscale/aarch64: add another hscale specialization
ffmpeg | branch: master | Swinney, Jonathan | Sat Aug 13 20:48:30 2022 +| [75ffca7eef557bcc714d924048a6e184b39fa470] | committer: Martin Storsjö libswscale/aarch64: add another hscale specialization This specialization handles the case where filtersize is 4 mod 8, e.g. 12, 20, etc. Aarch64 was previously using the c function for this case. This implementation speeds up that case significantly. hscale_8_to_15__fs_12_dstW_512_c: 6234.1 hscale_8_to_15__fs_12_dstW_512_neon: 1505.6 Signed-off-by: Jonathan Swinney Signed-off-by: Martin Storsjö > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=75ffca7eef557bcc714d924048a6e184b39fa470 --- libswscale/aarch64/hscale.S | 107 +++ libswscale/aarch64/swscale.c | 18 2 files changed, 117 insertions(+), 8 deletions(-) diff --git a/libswscale/aarch64/hscale.S b/libswscale/aarch64/hscale.S index b7b21b7a0f..a16d3dca42 100644 --- a/libswscale/aarch64/hscale.S +++ b/libswscale/aarch64/hscale.S @@ -91,6 +91,113 @@ function ff_hscale8to15_X8_neon, export=1 ret endfunc +function ff_hscale8to15_X4_neon, export=1 +// x0 SwsContext *c (not used) +// x1 int16_t *dst +// w2 int dstW +// x3 const uint8_t *src +// x4 const int16_t *filter +// x5 const int32_t *filterPos +// w6 int filterSize + +// This function for filter sizes that are 4 mod 8. In other words, anything that's 0 mod 4 but not +// 0 mod 8. It also assumes that dstW is 0 mod 4. + +lsl w7, w6, #1 // w7 = filterSize * 2 +1: +ldp w8, w9, [x5] // filterPos[idx + 0], [idx + 1] +ldp w10, w11, [x5, #8] // filterPos[idx + 2], [idx + 3] + +moviv16.2d, #0 // initialize accumulator for idx + 0 +moviv17.2d, #0 // initialize accumulator for idx + 1 +moviv18.2d, #0 // initialize accumulator for idx + 2 +moviv19.2d, #0 // initialize accumulator for idx + 3 + +mov x12, x4 // filter pointer for idx + 0 +add x13, x4, x7 // filter pointer for idx + 1 +add x8, x3, w8, uxtw// srcp + filterPos[idx + 0] +add x9, x3, w9, uxtw// srcp + filterPos[idx + 1] + +add x14, x13, x7// filter pointer for idx + 2 +add x10, x3, w10, uxtw // srcp + filterPos[idx + 2] +add x11, x3, w11, uxtw // srcp + filterPos[idx + 3] + +mov w0, w6 // copy filterSize to a temp register, w0 +add x5, x5, #16 // advance the filterPos pointer +add x15, x14, x7// filter pointer for idx + 3 +mov x16, xzr// temp register for offsetting filter pointers + +2: +// This section loops over 8-wide chunks of filter size +ldr d4, [x8], #8// load 8 bytes from srcp for idx + 0 +ldr q0, [x12, x16] // load 8 values, 16 bytes from filter for idx + 0 + +ldr d5, [x9], #8// load 8 bytes from srcp for idx + 1 +ldr q1, [x13, x16] // load 8 values, 16 bytes from filter for idx + 1 + +uxtlv4.8h, v4.8b// unsigned extend long for idx + 0 +uxtlv5.8h, v5.8b// unsigned extend long for idx + 1 + +ldr d6, [x10], #8 // load 8 bytes from srcp for idx + 2 +ldr q2, [x14, x16] // load 8 values, 16 bytes from filter for idx + 2 + +smlal v16.4s, v0.4h, v4.4h// val += src[srcPos + j + 0..3] * filter[fs * i + j + 0..3], idx + 0 +smlal v17.4s, v1.4h, v5.4h// val += src[srcPos + j + 0..3] * filter[fs * i + j + 0..3], idx + 1 + +ldr d7, [x11], #8 // load 8 bytes from srcp for idx + 3 +ldr q3, [x15, x16] // load 8 values, 16 bytes from filter for idx + 3 + +sub w0, w0, #8 // decrement the remaining filterSize counter +smlal2 v16.4s, v0.8h, v4.8h// val += src[srcPos + j + 4..7] * filter[fs * i + j + 4..7], idx + 0 +smlal2 v17.4s, v1.8h, v5.8h// val += src[srcPos + j + 4..7] * filter[fs * i + j + 4..7], idx + 1 +uxtlv6.8h, v6.8b// unsigned extend long for idx + 2 +uxtlv7.8h, v7.8b
[FFmpeg-cvslog] fftools/ffmpeg: store a separate copy of input codec parameters
ffmpeg | branch: master | Anton Khirnov | Sat Aug 13
17:03:39 2022 +0200| [ab3147383006f1a31978efce23c6bad38a754e05] | committer:
Anton Khirnov
fftools/ffmpeg: store a separate copy of input codec parameters
Use it instead of AVStream.codecpar in the main thread. While
AVStream.codecpar is documented to only be updated when the stream is
added or avformat_find_stream_info(), it is actually updated during
demuxing. Accessing it from a different thread then constitutes a race.
Ideally, some mechanism should eventually be provided for signalling
parameter updates to the user. Then the demuxing thread could pick up
the changes and propagate them to the decoder.
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ab3147383006f1a31978efce23c6bad38a754e05
---
fftools/ffmpeg.c | 39 ---
fftools/ffmpeg.h | 6 ++
fftools/ffmpeg_opt.c | 6 +-
3 files changed, 31 insertions(+), 20 deletions(-)
diff --git a/fftools/ffmpeg.c b/fftools/ffmpeg.c
index 8eb7759392..ef7177fc33 100644
--- a/fftools/ffmpeg.c
+++ b/fftools/ffmpeg.c
@@ -608,6 +608,7 @@ static void ffmpeg_cleanup(int ret)
av_freep(&ist->dts_buffer);
avcodec_free_context(&ist->dec_ctx);
+avcodec_parameters_free(&ist->par);
av_freep(&input_streams[i]);
}
@@ -1492,7 +1493,7 @@ static void print_final_stats(int64_t total_size)
for (j = 0; j < f->nb_streams; j++) {
InputStream *ist = input_streams[f->ist_index + j];
-enum AVMediaType type = ist->st->codecpar->codec_type;
+enum AVMediaType type = ist->par->codec_type;
total_size+= ist->data_size;
total_packets += ist->nb_packets;
@@ -1809,7 +1810,7 @@ static void flush_encoders(void)
for (x = 0; x < fg->nb_inputs; x++) {
InputFilter *ifilter = fg->inputs[x];
if (ifilter->format < 0 &&
-ifilter_parameters_from_codecpar(ifilter,
ifilter->ist->st->codecpar) < 0) {
+ifilter_parameters_from_codecpar(ifilter,
ifilter->ist->par) < 0) {
av_log(NULL, AV_LOG_ERROR, "Error copying paramerets
from input stream\n");
exit_program(1);
}
@@ -1912,11 +1913,11 @@ static void do_streamcopy(InputStream *ist,
OutputStream *ost, const AVPacket *p
if (pkt->dts == AV_NOPTS_VALUE) {
opkt->dts = av_rescale_q(ist->dts, AV_TIME_BASE_Q, ost->mux_timebase);
} else if (ost->st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
-int duration = av_get_audio_frame_duration2(ist->st->codecpar,
pkt->size);
+int duration = av_get_audio_frame_duration2(ist->par, pkt->size);
if(!duration)
-duration = ist->st->codecpar->frame_size;
+duration = ist->par->frame_size;
opkt->dts = av_rescale_delta(ist->st->time_base, pkt->dts,
-(AVRational){1,
ist->st->codecpar->sample_rate}, duration,
+(AVRational){1, ist->par->sample_rate},
duration,
&ist->filter_in_rescale_delta_last,
ost->mux_timebase);
/* dts will be set immediately afterwards to what pts is now */
opkt->pts = opkt->dts - ost_tb_start_time;
@@ -1976,7 +1977,7 @@ static int ifilter_send_frame(InputFilter *ifilter,
AVFrame *frame, int keep_ref
/* determine if the parameters for this input changed */
need_reinit = ifilter->format != frame->format;
-switch (ifilter->ist->st->codecpar->codec_type) {
+switch (ifilter->ist->par->codec_type) {
case AVMEDIA_TYPE_AUDIO:
need_reinit |= ifilter->sample_rate!= frame->sample_rate ||
av_channel_layout_compare(&ifilter->ch_layout,
&frame->ch_layout);
@@ -2056,7 +2057,7 @@ static int ifilter_send_eof(InputFilter *ifilter, int64_t
pts)
} else {
// the filtergraph was never configured
if (ifilter->format < 0) {
-ret = ifilter_parameters_from_codecpar(ifilter,
ifilter->ist->st->codecpar);
+ret = ifilter_parameters_from_codecpar(ifilter, ifilter->ist->par);
if (ret < 0)
return ret;
}
@@ -2212,9 +2213,9 @@ static int decode_video(InputStream *ist, AVPacket *pkt,
int *got_output, int64_
// The following line may be required in some cases where there is no
parser
// or the parser does not has_b_frames correctly
-if (ist->st->codecpar->video_delay < ist->dec_ctx->has_b_frames) {
+if (ist->par->video_delay < ist->dec_ctx->has_b_frames) {
if (ist->dec_ctx->codec_id == AV_CODEC_ID_H264) {
-ist->st->codecpar->video_delay = ist->dec_ctx->has_b_frames;
+ist->par->video_delay = ist->dec_ctx->has_b_frames;
} else
av_log(ist->dec_ctx, AV_LOG_WARNING,
"video_delay is larger in
[FFmpeg-cvslog] doc/APIchanges: add missing rgbaf16 pixfmt entry
ffmpeg | branch: master | Timo Rothenpieler | Tue Aug
16 12:31:03 2022 +0200| [317f5252c09d6deee7025907eea2dfe44935c891] | committer:
Timo Rothenpieler
doc/APIchanges: add missing rgbaf16 pixfmt entry
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=317f5252c09d6deee7025907eea2dfe44935c891
---
doc/APIchanges | 3 +++
1 file changed, 3 insertions(+)
diff --git a/doc/APIchanges b/doc/APIchanges
index a196bc40d4..b3ba07ee7c 100644
--- a/doc/APIchanges
+++ b/doc/APIchanges
@@ -14,6 +14,9 @@ libavutil: 2021-04-27
API changes, most recent first:
+2022-08-07 - e95b08a7dd - lavu 57.33.101 - pixfmt.h
+ Add AV_PIX_FMT_RGBAF16{BE,LE} pixel formats.
+
2022-08-xx - xx - lavu 57.33.100 - hwcontext_qsv.h
Add loader field to AVQSVDeviceContext
___
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: updated tests for sw_scale
ffmpeg | branch: master | Swinney, Jonathan | Sat Aug 13
20:55:55 2022 +| [4dcd191a507b76003bc8a1d0d845d99e3ba9fabf] | committer:
Martin Storsjö
checkasm: updated tests for sw_scale
Change the reference to exactly match the C reference in swscale,
instead of exactly matching the x86 SIMD implementations (which
differs slightly). Test with and without SWS_ACCURATE_RND - if this
flag isn't set, the output must match the C reference exactly,
otherwise it is allowed to be off by 2.
Mark a couple x86 functions as unavailable when SWS_ACCURATE_RND
is set - apparently this discrepancy hasn't been noticed in other
exact tests before.
Add a test for yuv2plane1.
Signed-off-by: Jonathan Swinney
Signed-off-by: Martin Storsjö
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=4dcd191a507b76003bc8a1d0d845d99e3ba9fabf
---
libswscale/x86/swscale.c | 8 +-
tests/checkasm/sw_scale.c | 188 --
2 files changed, 154 insertions(+), 42 deletions(-)
diff --git a/libswscale/x86/swscale.c b/libswscale/x86/swscale.c
index 628f12137c..32d441245d 100644
--- a/libswscale/x86/swscale.c
+++ b/libswscale/x86/swscale.c
@@ -534,7 +534,8 @@ switch(c->dstBpc){ \
ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse2, sse2);
ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse2, ,
HAVE_ALIGNED_STACK || ARCH_X86_64);
-ASSIGN_VSCALE_FUNC(c->yuv2plane1, sse2);
+if (!(c->flags & SWS_ACCURATE_RND))
+ASSIGN_VSCALE_FUNC(c->yuv2plane1, sse2);
switch (c->srcFormat) {
case AV_PIX_FMT_YA8:
@@ -583,14 +584,15 @@ switch(c->dstBpc){ \
ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse4,
if (!isBE(c->dstFormat)) c->yuv2planeX =
ff_yuv2planeX_16_sse4,
HAVE_ALIGNED_STACK || ARCH_X86_64);
-if (c->dstBpc == 16 && !isBE(c->dstFormat))
+if (c->dstBpc == 16 && !isBE(c->dstFormat) && !(c->flags &
SWS_ACCURATE_RND))
c->yuv2plane1 = ff_yuv2plane1_16_sse4;
}
if (EXTERNAL_AVX(cpu_flags)) {
ASSIGN_VSCALEX_FUNC(c->yuv2planeX, avx, ,
HAVE_ALIGNED_STACK || ARCH_X86_64);
-ASSIGN_VSCALE_FUNC(c->yuv2plane1, avx);
+if (!(c->flags & SWS_ACCURATE_RND))
+ASSIGN_VSCALE_FUNC(c->yuv2plane1, avx);
switch (c->srcFormat) {
case AV_PIX_FMT_YUYV422:
diff --git a/tests/checkasm/sw_scale.c b/tests/checkasm/sw_scale.c
index b643a47c30..859993db6f 100644
--- a/tests/checkasm/sw_scale.c
+++ b/tests/checkasm/sw_scale.c
@@ -35,40 +35,140 @@
AV_WN32(buf + j, rnd()); \
} while (0)
-// This reference function is the same approximate algorithm employed by the
-// SIMD functions
-static void ref_function(const int16_t *filter, int filterSize,
- const int16_t **src, uint8_t
*dest, int dstW,
- const uint8_t *dither, int
offset)
+static void yuv2planeX_8_ref(const int16_t *filter, int filterSize,
+ const int16_t **src, uint8_t *dest, int dstW,
+ const uint8_t *dither, int offset)
{
-int i, d;
-d = ((filterSize - 1) * 8 + dither[0]) >> 4;
-for ( i = 0; i < dstW; i++) {
-int16_t val = d;
+// This corresponds to the yuv2planeX_8_c function
+int i;
+for (i = 0; i < dstW; i++) {
+int val = dither[(i + offset) & 7] << 12;
int j;
-union {
-int val;
-int16_t v[2];
-} t;
-for (j = 0; j < filterSize; j++){
-t.val = (int)src[j][i + offset] * (int)filter[j];
-val += t.v[1];
+for (j = 0; j < filterSize; j++)
+val += src[j][i] * filter[j];
+
+dest[i]= av_clip_uint8(val >> 19);
+}
+}
+
+static int cmp_off_by_n(const uint8_t *ref, const uint8_t *test, size_t n, int
accuracy)
+{
+for (size_t i = 0; i < n; i++) {
+if (abs(ref[i] - test[i]) > accuracy)
+return 1;
+}
+return 0;
+}
+
+static void print_data(uint8_t *p, size_t len, size_t offset)
+{
+size_t i = 0;
+for (; i < len; i++) {
+if (i % 8 == 0) {
+printf("0x%04zx: ", i+offset);
+}
+printf("0x%02x ", (uint32_t) p[i]);
+if (i % 8 == 7) {
+printf("\n");
}
-dest[i]= av_clip_uint8(val>>3);
}
+if (i % 8 != 0) {
+printf("\n");
+}
+}
+
+static size_t show_differences(uint8_t *a, uint8_t *b, size_t len)
+{
+for (size_t i = 0; i < len; i++) {
+if (a[i] != b[i]) {
+size_t offset_of_mismatch = i;
+size_t offset;
+if (i >= 8) i-=8;
+offset = i & (~7);
+printf("test a:\n");
+print_data(&a[offset], 32, offset);
+printf("\ntest b:\n");
+print_data(&b[offset], 32, offset);
+
[FFmpeg-cvslog] swscale/aarch64: vscale optimization
ffmpeg | branch: master | Swinney, Jonathan | Sat Aug 13
20:56:02 2022 +| [3e708722a2d06b8c7290d359ccaea9cd9885e701] | committer:
Martin Storsjö
swscale/aarch64: vscale optimization
Use scalar times vector multiply accumlate instructions instead of
vector times vector to remove the need for replicating load instructions
which are slightly slower.
On AWS c7g (Graviton 3, Neoverse V1) instances:
yuv2yuvX_8_0_512_accurate_neon: 1144.8 987.4
yuv2yuvX_16_0_512_accurate_neon: 2080.5 1869.4
Signed-off-by: Jonathan Swinney
Signed-off-by: Martin Storsjö
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3e708722a2d06b8c7290d359ccaea9cd9885e701
---
libswscale/aarch64/output.S | 11 +--
1 file changed, 5 insertions(+), 6 deletions(-)
diff --git a/libswscale/aarch64/output.S b/libswscale/aarch64/output.S
index af71de6050..991750cf31 100644
--- a/libswscale/aarch64/output.S
+++ b/libswscale/aarch64/output.S
@@ -34,16 +34,15 @@ function ff_yuv2planeX_8_neon, export=1
mov x9, x2 // srcp= src
mov x10, x0 // filterp = filter
3: ldp x11, x12, [x9], #16 // get 2 pointers:
src[j] and src[j+1]
+ldr s7, [x10], #4 // read 2x16-bit
coeff X and Y at filter[j] and filter[j+1]
add x11, x11, x7, lsl #1// &src[j ][i]
add x12, x12, x7, lsl #1// &src[j+1][i]
ld1 {v5.8H}, [x11] // read 8x16-bit @
src[j ][i + {0..7}]: A,B,C,D,E,F,G,H
ld1 {v6.8H}, [x12] // read 8x16-bit @
src[j+1][i + {0..7}]: I,J,K,L,M,N,O,P
-ld1r{v7.8H}, [x10], #2 // read 1x16-bit
coeff X at filter[j ] and duplicate across lanes
-ld1r{v16.8H}, [x10], #2 // read 1x16-bit
coeff Y at filter[j+1] and duplicate across lanes
-smlal v3.4S, v5.4H, v7.4H // val0 +=
{A,B,C,D} * X
-smlal2 v4.4S, v5.8H, v7.8H // val1 +=
{E,F,G,H} * X
-smlal v3.4S, v6.4H, v16.4H// val0 +=
{I,J,K,L} * Y
-smlal2 v4.4S, v6.8H, v16.8H// val1 +=
{M,N,O,P} * Y
+smlal v3.4S, v5.4H, v7.H[0] // val0 +=
{A,B,C,D} * X
+smlal2 v4.4S, v5.8H, v7.H[0] // val1 +=
{E,F,G,H} * X
+smlal v3.4S, v6.4H, v7.H[1] // val0 +=
{I,J,K,L} * Y
+smlal2 v4.4S, v6.8H, v7.H[1] // val1 +=
{M,N,O,P} * Y
subsw8, w8, #2 // tmpfilterSize
-= 2
b.gt3b // loop until
filterSize consumed
___
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/aarch64: add vscale specializations
ffmpeg | branch: master | Swinney, Jonathan | Sat Aug 13
20:56:06 2022 +| [0d7caa5b09b680aa2c8ae677a8d0fdc134b7e658] | committer:
Martin Storsjö
swscale/aarch64: add vscale specializations
This commit adds new code paths for vscale when filterSize is 2, 4, or
8. By using specialized code with unrolling to match the filterSize we
can improve performance.
On AWS c7g (Graviton 3, Neoverse V1) instances:
before after
yuv2yuvX_2_0_512_accurate_neon: 558.8268.9
yuv2yuvX_4_0_512_accurate_neon: 637.5434.9
yuv2yuvX_8_0_512_accurate_neon: 1144.8 806.2
yuv2yuvX_16_0_512_accurate_neon: 2080.5 1853.7
Signed-off-by: Jonathan Swinney
Signed-off-by: Martin Storsjö
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=0d7caa5b09b680aa2c8ae677a8d0fdc134b7e658
---
libswscale/aarch64/output.S | 177 +++
libswscale/aarch64/swscale.c | 13
2 files changed, 190 insertions(+)
diff --git a/libswscale/aarch64/output.S b/libswscale/aarch64/output.S
index 991750cf31..b8a2818c9b 100644
--- a/libswscale/aarch64/output.S
+++ b/libswscale/aarch64/output.S
@@ -21,13 +21,33 @@
#include "libavutil/aarch64/asm.S"
function ff_yuv2planeX_8_neon, export=1
+// x0 - const int16_t *filter,
+// x1 - int filterSize,
+// x2 - const int16_t **src,
+// x3 - uint8_t *dest,
+// w4 - int dstW,
+// x5 - const uint8_t *dither,
+// w6 - int offset
+
ld1 {v0.8B}, [x5] // load 8x8-bit
dither
+and w6, w6, #7
cbz w6, 1f // check if
offsetting present
ext v0.8B, v0.8B, v0.8B, #3 // honor
offsetting which can be 0 or 3 only
1: uxtlv0.8H, v0.8B// extend dither
to 16-bit
ushll v1.4S, v0.4H, #12 // extend dither
to 32-bit with left shift by 12 (part 1)
ushll2 v2.4S, v0.8H, #12 // extend dither
to 32-bit with left shift by 12 (part 2)
+cmp w1, #8 // if filterSize
== 8, branch to specialized version
+b.eq6f
+cmp w1, #4 // if filterSize
== 4, branch to specialized version
+b.eq8f
+cmp w1, #2 // if filterSize
== 2, branch to specialized version
+b.eq10f
+
+// The filter size does not match of the of specialized implementations. It is
either even or odd. If it is even
+// then use the first section below.
mov x7, #0 // i = 0
+tbnzw1, #0, 4f // if filterSize %
2 != 0 branch to specialized version
+// fs % 2 == 0
2: mov v3.16B, v1.16B // initialize
accumulator part 1 with dithering value
mov v4.16B, v2.16B // initialize
accumulator part 2 with dithering value
mov w8, w1 // tmpfilterSize =
filterSize
@@ -54,4 +74,161 @@ function ff_yuv2planeX_8_neon, export=1
add x7, x7, #8 // i += 8
b.gt2b // loop until
width consumed
ret
+
+// If filter size is odd (most likely == 1), then use this section.
+// fs % 2 != 0
+4: mov v3.16B, v1.16B // initialize
accumulator part 1 with dithering value
+mov v4.16B, v2.16B // initialize
accumulator part 2 with dithering value
+mov w8, w1 // tmpfilterSize =
filterSize
+mov x9, x2 // srcp= src
+mov x10, x0 // filterp = filter
+5: ldr x11, [x9], #8 // get 1 pointer:
src[j]
+ldr h6, [x10], #2 // read 1 16 bit
coeff X at filter[j]
+add x11, x11, x7, lsl #1// &src[j ][i]
+ld1 {v5.8H}, [x11] // read 8x16-bit @
src[j ][i + {0..7}]: A,B,C,D,E,F,G,H
+smlal v3.4S, v5.4H, v6.H[0] // val0 +=
{A,B,C,D} * X
+smlal2 v4.4S, v5.8H, v6.H[0] // val1 +=
{E,F,G,H} * X
+subsw8, w8, #1 // tmpfilterSize
-= 2
+b.gt5b // loop until
filterSize consumed
+
+sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)
+sqshrun2v3.8h, v4.4s, #16 // clip16(val1>>16)
+uqshrn v3.8b, v3.8h, #3// clip8(v
[FFmpeg-cvslog] checkasm/hevc_add_res: add 12bit test
ffmpeg | branch: master | J. Dekker | Thu Jun 23 20:04:05
2022 +0200| [ea6ecb12aa9ebfbc985f71938a6cccf5046ca826] | committer: J. Dekker
checkasm/hevc_add_res: add 12bit test
Also fix the bug where in every other byte only the lower 2 bits were
used in the 8bit test.
Signed-off-by: J. Dekker
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ea6ecb12aa9ebfbc985f71938a6cccf5046ca826
---
tests/checkasm/hevc_add_res.c | 15 ---
1 file changed, 8 insertions(+), 7 deletions(-)
diff --git a/tests/checkasm/hevc_add_res.c b/tests/checkasm/hevc_add_res.c
index 0c896adaca..f17d121939 100644
--- a/tests/checkasm/hevc_add_res.c
+++ b/tests/checkasm/hevc_add_res.c
@@ -36,14 +36,14 @@
} \
} while (0)
-#define randomize_buffers2(buf, size) \
+#define randomize_buffers2(buf, size, mask) \
do { \
int j;\
for (j = 0; j < size; j++)\
-AV_WN16A(buf + j * 2, rnd() & 0x3FF); \
+AV_WN16A(buf + j * 2, rnd() & mask); \
} while (0)
-static void compare_add_res(int size, ptrdiff_t stride, int overflow_test)
+static void compare_add_res(int size, ptrdiff_t stride, int overflow_test, int
mask)
{
LOCAL_ALIGNED_32(int16_t, res0, [32 * 32]);
LOCAL_ALIGNED_32(int16_t, res1, [32 * 32]);
@@ -53,7 +53,7 @@ static void compare_add_res(int size, ptrdiff_t stride, int
overflow_test)
declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *dst, int16_t *res,
ptrdiff_t stride);
randomize_buffers(res0, size);
-randomize_buffers2(dst0, size);
+randomize_buffers2(dst0, size, mask);
if (overflow_test)
res0[0] = 0x8000;
memcpy(res1, res0, sizeof(*res0) * size);
@@ -69,6 +69,7 @@ static void compare_add_res(int size, ptrdiff_t stride, int
overflow_test)
static void check_add_res(HEVCDSPContext h, int bit_depth)
{
int i;
+int mask = bit_depth == 8 ? 0x : bit_depth == 10 ? 0x03FF : 0x07FF;
for (i = 2; i <= 5; i++) {
int block_size = 1 << i;
@@ -76,9 +77,9 @@ static void check_add_res(HEVCDSPContext h, int bit_depth)
ptrdiff_t stride = block_size << (bit_depth > 8);
if (check_func(h.add_residual[i - 2], "hevc_add_res_%dx%d_%d",
block_size, block_size, bit_depth)) {
-compare_add_res(size, stride, 0);
+compare_add_res(size, stride, 0, mask);
// overflow test for res = -32768
-compare_add_res(size, stride, 1);
+compare_add_res(size, stride, 1, mask);
}
}
}
@@ -87,7 +88,7 @@ void checkasm_check_hevc_add_res(void)
{
int bit_depth;
-for (bit_depth = 8; bit_depth <= 10; bit_depth++) {
+for (bit_depth = 8; bit_depth <= 12; bit_depth++) {
HEVCDSPContext h;
ff_hevc_dsp_init(&h, bit_depth);
___
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] lavc/aarch64: reformat add_res funcs
ffmpeg | branch: master | J. Dekker | Thu Jun 23 20:04:06
2022 +0200| [aa9eabb7a5283fd90b3274ac4b6ba0d16e42] | committer: J. Dekker
lavc/aarch64: reformat add_res funcs
Signed-off-by: J. Dekker
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=aa9eabb7a5283fd90b3274ac4b6ba0d16e42
---
libavcodec/aarch64/hevcdsp_idct_neon.S | 216 -
1 file changed, 108 insertions(+), 108 deletions(-)
diff --git a/libavcodec/aarch64/hevcdsp_idct_neon.S
b/libavcodec/aarch64/hevcdsp_idct_neon.S
index 0869431294..484eea8437 100644
--- a/libavcodec/aarch64/hevcdsp_idct_neon.S
+++ b/libavcodec/aarch64/hevcdsp_idct_neon.S
@@ -27,21 +27,21 @@
#include "libavutil/aarch64/asm.S"
const trans, align=4
-.short 64, 83, 64, 36
-.short 89, 75, 50, 18
-.short 90, 87, 80, 70
-.short 57, 43, 25, 9
-.short 90, 90, 88, 85
-.short 82, 78, 73, 67
-.short 61, 54, 46, 38
-.short 31, 22, 13, 4
+.short 64, 83, 64, 36
+.short 89, 75, 50, 18
+.short 90, 87, 80, 70
+.short 57, 43, 25, 9
+.short 90, 90, 88, 85
+.short 82, 78, 73, 67
+.short 61, 54, 46, 38
+.short 31, 22, 13, 4
endconst
.macro clip10 in1, in2, c1, c2
-smax\in1, \in1, \c1
-smax\in2, \in2, \c1
-smin\in1, \in1, \c2
-smin\in2, \in2, \c2
+smax\in1, \in1, \c1
+smax\in2, \in2, \c1
+smin\in1, \in1, \c2
+smin\in2, \in2, \c2
.endm
function ff_hevc_add_residual_4x4_8_neon, export=1
@@ -50,13 +50,13 @@ function ff_hevc_add_residual_4x4_8_neon, export=1
ld1 {v2.s}[1], [x0], x2
ld1 {v2.s}[2], [x0], x2
ld1 {v2.s}[3], [x0], x2
-sub x0, x0, x2, lsl #2
-uxtl v6.8h, v2.8b
-uxtl2v7.8h, v2.16b
-sqaddv0.8h, v0.8h, v6.8h
-sqaddv1.8h, v1.8h, v7.8h
-sqxtun v0.8b, v0.8h
-sqxtun2 v0.16b, v1.8h
+sub x0, x0, x2, lsl #2
+uxtlv6.8h, v2.8b
+uxtl2 v7.8h, v2.16b
+sqadd v0.8h, v0.8h, v6.8h
+sqadd v1.8h, v1.8h, v7.8h
+sqxtun v0.8b, v0.8h
+sqxtun2 v0.16b, v1.8h
st1 {v0.s}[0], [x0], x2
st1 {v0.s}[1], [x0], x2
st1 {v0.s}[2], [x0], x2
@@ -70,63 +70,63 @@ function ff_hevc_add_residual_4x4_10_neon, export=1
ld1 {v2.d}[0], [x12], x2
ld1 {v2.d}[1], [x12], x2
ld1 {v3.d}[0], [x12], x2
-sqaddv0.8h, v0.8h, v2.8h
+sqadd v0.8h, v0.8h, v2.8h
ld1 {v3.d}[1], [x12], x2
-movi v4.8h, #0
-sqaddv1.8h, v1.8h, v3.8h
-mvni v5.8h, #0xFC, lsl #8 // movi #0x3FF
-clip10 v0.8h, v1.8h, v4.8h, v5.8h
-st1 {v0.d}[0], [x0], x2
-st1 {v0.d}[1], [x0], x2
-st1 {v1.d}[0], [x0], x2
-st1 {v1.d}[1], [x0], x2
+moviv4.8h, #0
+sqadd v1.8h, v1.8h, v3.8h
+mvniv5.8h, #0xFC, lsl #8 // movi #0x3FF
+clip10 v0.8h, v1.8h, v4.8h, v5.8h
+st1 {v0.d}[0], [x0], x2
+st1 {v0.d}[1], [x0], x2
+st1 {v1.d}[0], [x0], x2
+st1 {v1.d}[1], [x0], x2
ret
endfunc
function ff_hevc_add_residual_8x8_8_neon, export=1
-add x12, x0, x2
-add x2, x2, x2
-mov x3, #8
-1: subs x3, x3, #2
-ld1 {v2.d}[0], [x0]
-ld1 {v2.d}[1],[x12]
-uxtl v3.8h, v2.8b
+add x12, x0, x2
+add x2, x2, x2
+mov x3, #8
+1: subsx3, x3, #2
+ld1 {v2.d}[0], [x0]
+ld1 {v2.d}[1], [x12]
+uxtlv3.8h, v2.8b
ld1 {v0.8h-v1.8h}, [x1], #32
-uxtl2v2.8h, v2.16b
-sqaddv0.8h, v0.8h, v3.8h
-sqaddv1.8h, v1.8h, v2.8h
-sqxtun v0.8b, v0.8h
-sqxtun2 v0.16b, v1.8h
-st1 {v0.d}[0], [x0], x2
-st1 {v0.d}[1],[x12], x2
-bne 1b
+uxtl2 v2.8h, v2.16b
+sqadd v0.8h, v0.8h, v3.8h
+sqadd v1.8h, v1.8h, v2.8h
+sqxtun v0.8b, v0.8h
+sqxtun2 v0.16b, v1.8h
+s
[FFmpeg-cvslog] checkasm: Provide enough alignment in the new yuv2plane1 test
ffmpeg | branch: master | Martin Storsjö | Tue Aug 16 23:46:35 2022 +0300| [21c2c57ba5a8b426ad9c07902ec957dbbb20cf36] | committer: Martin Storsjö checkasm: Provide enough alignment in the new yuv2plane1 test This fixes the checkasm test in some setups on x86. Signed-off-by: Martin Storsjö > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=21c2c57ba5a8b426ad9c07902ec957dbbb20cf36 --- 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 859993db6f..cbe4460a99 100644 --- a/tests/checkasm/sw_scale.c +++ b/tests/checkasm/sw_scale.c @@ -114,9 +114,9 @@ static void check_yuv2yuv1(int accurate) const int16_t *src, uint8_t *dest, int dstW, const uint8_t *dither, int offset); -LOCAL_ALIGNED_8(int16_t, src_pixels, [LARGEST_INPUT_SIZE]); -LOCAL_ALIGNED_8(uint8_t, dst0, [LARGEST_INPUT_SIZE]); -LOCAL_ALIGNED_8(uint8_t, dst1, [LARGEST_INPUT_SIZE]); +LOCAL_ALIGNED_16(int16_t, src_pixels, [LARGEST_INPUT_SIZE]); +LOCAL_ALIGNED_16(uint8_t, dst0, [LARGEST_INPUT_SIZE]); +LOCAL_ALIGNED_16(uint8_t, dst1, [LARGEST_INPUT_SIZE]); LOCAL_ALIGNED_8(uint8_t, dither, [8]); randomize_buffers((uint8_t*)dither, 8); ___ 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: Don't hardcode the height to 8 in sad8_xy2_mmx
ffmpeg | branch: master | Martin Storsjö | Wed Jul 13
23:39:40 2022 +0300| [dc55e6357818e21e26afbcdfdf2dd7368ec0e8fd] | committer:
Martin Storsjö
x86: Don't hardcode the height to 8 in sad8_xy2_mmx
The height is hardcoded in some of the me_cmp functions, but not
in all of them. But in the case of all other functions, it's hardcoded
in the same place in SIMD functions as in the C reference functions,
while this one function differs from the behaviour of the C code.
(Before 542765ce3eccbca587d54262a512cbdb1407230d, there were a
couple other sad8_*_mmx functions with similar hardcoded height.)
Signed-off-by: Martin Storsjö
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=dc55e6357818e21e26afbcdfdf2dd7368ec0e8fd
---
libavcodec/x86/me_cmp_init.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/libavcodec/x86/me_cmp_init.c b/libavcodec/x86/me_cmp_init.c
index f140381c98..bc1051c27e 100644
--- a/libavcodec/x86/me_cmp_init.c
+++ b/libavcodec/x86/me_cmp_init.c
@@ -202,13 +202,12 @@ static inline int sum_mmx(void)
static int sad8_xy2_ ## suf(MpegEncContext *v, const uint8_t *blk2, \
const uint8_t *blk1, ptrdiff_t stride, int h) \
{ \
-av_assert2(h == 8); \
__asm__ volatile ( \
"pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
::);\
\
-sad8_4_ ## suf(blk1, blk2, stride, 8); \
+sad8_4_ ## suf(blk1, blk2, stride, h); \
\
return sum_ ## suf(); \
} \
___
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] tools: Make sure to create the tools directory before building decode_simple.o
ffmpeg | branch: master | Martin Storsjö | Mon Aug 8 11:33:26 2022 +0300| [1eaa575cf11c054b0b724480f3070fc908faf8ef] | committer: Martin Storsjö tools: Make sure to create the tools directory before building decode_simple.o This directory dependency is normally added implicitly by rules in ffbuild/common.mak; for tools it's created by a rule for TOOLOBJS. TOOLOBJS is populated implicitly from TOOLS, and decode_simple.o doesn't end up there because it's an odd occurrance of a lone object file in the tools subdirectory, not belonging to any other tool. Signed-off-by: Martin Storsjö > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1eaa575cf11c054b0b724480f3070fc908faf8ef --- tools/Makefile | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/Makefile b/tools/Makefile index f4d1327b9f..4afa23342d 100644 --- a/tools/Makefile +++ b/tools/Makefile @@ -20,6 +20,8 @@ tools/target_io_dem_fuzzer.o: tools/target_dem_fuzzer.c tools/venc_data_dump$(EXESUF): tools/decode_simple.o tools/scale_slice_test$(EXESUF): tools/decode_simple.o +tools/decode_simple.o: | tools + OUTDIRS += tools clean:: ___ 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: motion: Test different h parameters
ffmpeg | branch: master | Martin Storsjö | Tue Jul 12
23:58:01 2022 +0300| [d69d12a5b9236b9d2f1fd247ea452f84cdd1aaf9] | committer:
Martin Storsjö
checkasm: motion: Test different h parameters
Previously, the checkasm test always passed h=8, so no other cases
were tested.
Out of the me_cmp functions, in practice, some functions are hardcoded
to always assume a 8x8 block (ignoring the h parameter), while others
do use the parameter. For those with hardcoded height, both the
reference C function and the assembly implementations ignore the
parameter similarly.
The documentation for the functions indicate that heights between
w/2 and 2*w, within the range of 4 to 16, should be supported. This
patch just tests random heights in that range, without knowing what
width the current function actually uses.
Signed-off-by: Martin Storsjö
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d69d12a5b9236b9d2f1fd247ea452f84cdd1aaf9
---
tests/checkasm/motion.c | 10 ++
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/tests/checkasm/motion.c b/tests/checkasm/motion.c
index 631a9ed29f..87b20d1c10 100644
--- a/tests/checkasm/motion.c
+++ b/tests/checkasm/motion.c
@@ -45,7 +45,7 @@ static void test_motion(const char *name, me_cmp_func
test_func)
/* motion estimation can look up to 17 bytes ahead */
static const int look_ahead = 17;
-int i, x, y, d1, d2;
+int i, x, y, h, d1, d2;
uint8_t *ptr;
LOCAL_ALIGNED_16(uint8_t, img1, [WIDTH * HEIGHT]);
@@ -68,14 +68,16 @@ static void test_motion(const char *name, me_cmp_func
test_func)
for (i = 0; i < ITERATIONS; i++) {
x = rnd() % (WIDTH - look_ahead);
y = rnd() % (HEIGHT - look_ahead);
+// Pick a random h between 4 and 16; pick an even value.
+h = 4 + ((rnd() % (16 + 1 - 4)) & ~1);
ptr = img2 + y * WIDTH + x;
-d2 = call_ref(NULL, img1, ptr, WIDTH, 8);
-d1 = call_new(NULL, img1, ptr, WIDTH, 8);
+d2 = call_ref(NULL, img1, ptr, WIDTH, h);
+d1 = call_new(NULL, img1, ptr, WIDTH, h);
if (d1 != d2) {
fail();
-printf("func: %s, x=%d y=%d, error: asm=%d c=%d\n", name, x,
y, d1, d2);
+printf("func: %s, x=%d y=%d h=%d, error: asm=%d c=%d\n", name,
x, y, h, d1, d2);
break;
}
}
___
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".
