This naive hist[p[x]]++ loop suffers badly when there are large regions of
identical values in the image, because of store-to-load forwarding delay.
Splitting up the histogram into four "parallel" histograms and processing
them one at a time speeds things up significantly, about 40% on my end.
Since psadbw only exists for 8-bits, we have to emulate it for 16-bit
inputs. The simplest sequence is to use a normal subtraction, which is safe
as long as the inputs do not exceed 32767 - so limit this implementation
to 15-bit inputs and below.
For 16-bit inputs, we could in theory instead use a pminw / pmaxw to ensure
the resulting difference does not overflow, but this is slower, and also
breaks the subsequent use of pmaddwd, so I opted to skip 16-bit SIMD for
now.
scene_sad10_c: 114175.6 ( 1.00x)
scene_sad10_avx2: 9617.7 (11.87x)
scene_sad10_avx512: 5208.8 (21.92x)
scene_sad12_c: 114537.8 ( 1.00x)
scene_sad12_avx2: 9614.0 (11.91x)
scene_sad12_avx512: 5186.3 (22.08x)
scene_sad14_c: 114113.9 ( 1.00x)
scene_sad14_avx2: 9612.9 (11.87x)
scene_sad14_avx512: 5186.0 (22.00x)
scene_sad15_c: 114108.9 ( 1.00x)
scene_sad15_avx2: 9612.3 (11.87x)
scene_sad15_avx512: 5186.4 (22.00x)
scene_sad16_c: 114136.0 ( 1.00x)
Trivial to add, but a lot faster (on my machine).
scene_sad8_c: 114476.4 ( 1.00x)
scene_sad8_sse2: 8644.3 (13.24x)
scene_sad8_avx2: 4520.1 (25.33x)
scene_sad8_avx512: 3153.0 (36.31x)
the SSL_shutdown in tls_close need call the url_bio_bwrite
so we should keep udp still alive
Signed-off-by: Jack Lau <jacklau1222@qq.com>
Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
It's possible to call linker indirectly through driver like Clang. In
which cases linker args has to be prefixed with -Wl.
Signed-off-by: Kacper Michajłow <kasper93@gmail.com>
This aligns declared function types in checkasm with real definition.
Fixes FATE: checkasm-{sw_rgb,sw_scale,sw_yuv2rgb,sw_yuv2yuv}
Fixes: runtime error: call to function <func> through pointer to incorrect function type
Fixes: c1a0e65763
Signed-off-by: Kacper Michajłow <kasper93@gmail.com>
Skipping backwards (and even forwards) resets the EOF flag, and can thus
lead to infinite looping if the conditions are just right.
Fixes: Infinite loop
Fixes: 427538726/clusterfuzz-testcase-minimized-ffmpeg_dem_FLV_fuzzer-6582567304495104
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
Old Mingw-w64 releases provided by some distros seemingly don't have them, so
check for them and disable the dtls protocol if unavailable.
Signed-off-by: James Almer <jamrial@gmail.com>
If the image data is not at the start of the buffer allocation, such as
when the buffer has padding before the image data, this function maps too
much memory, since src_data + src_buf->size exceeds the buffer size.
Fix this by subtracting the difference between the buffer start and the
provided image data pointer from the size of the memory range to map.
An easy way to reproduce this issue is using the vf_pad filter, which
allocates image data buffers with a nonzero offset whenever padding is
requested before the start of the image data.
The issue is that:
float en[2];
...
tns->n_filt[w] = is8 ? 1 : order != TNS_MAX_ORDER ? 2 : 3;
for (g = 0; g < tns->n_filt[w]; g++) {
tns->direction[w][g] = slant != 2 ? slant : en[g] < en[!g];
When using the AAC Main profile, n_filt = 3, and slant is by
default 2 (normal long frames), g can go above 1.
en is the evolution of energy in the frequency domain for every
band at the given window. E.g. whether the energy is concentrated
at the top of each band, or the bottom.
For 2-pole filters, its straightforward.
For 3-pole filters, we need more than 2 measurements.
This commit properly implements support for 3-pole filters, by measuring
the band energy across three areas.
Do note that even xHE-AAC caps n_filt to 2, and only AAC Main allows
n_filt == 3.
Fixes https://trac.ffmpeg.org/ticket/11418
Add check for the return value of av_malloc_array() and av_calloc()
to avoid potential NULL pointer dereference.
Fixes: dcfd24b10c ("avcodec/alsdec: Implement floating point sample data decoding")
Signed-off-by: Jiasheng Jiang <jiashengjiangcool@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
This adds support for high bit depth formats, as well as formats with fewer
than 3 planes. The implementation for HBD is the same as for 8 bit formats,
just right shifted to 8 bits.
It's worth pointing out that this also works for HDR formats (and even DV),
because the underlying implementation is just trying to minimize the histogram
difference. If anything, using a HDR format will result in a *more* accurate
detection, because HDR formats tend to be more perceptually uniform.
Enables timeline editing options for overlay_cuda similar to what overlay allows
Example overlaying an image on a video between 30 to 60 seconds:
ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i sample-video.mp4 -i sample-image.jpg
-filter_complex "[1:v]hwupload_cuda[image],[0:v]scale_npp=format=yuv420p[video],[video][image]overlay_cuda=enable='between(t,30,60)'"
-c:v h264_nvenc -c:a copy -y overlay-output-gpu.mp4
Signed-off-by: Jorge Estrada <jestrada.list@gmail.com>
Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
