July 2023 - ffmpeg-cvslog - lists.ffmpeg.org

avfilter/avf_showcwt: improve analysis
by Paul B Mahol 22 Jul '23

22 Jul '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Fri Jul 21 23:11:11 2023 +0200| [e6168e43a2eec4757cc8d7f981f9d29e1f12dbed] | committer: Paul B Mahol avfilter/avf_showcwt: improve analysis Make inverse FFT step always power of 2 in size. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e6168e43a2eec4757… --- libavfilter/avf_showcwt.c | 78 +++++++++++++++++++++++++---------------------- 1 file changed, 42 insertions(+), 36 deletions(-) diff --git a/libavfilter/avf_showcwt.c b/libavfilter/avf_showcwt.c index ee0d697e1e..ef6798e9cf 100644 --- a/libavfilter/avf_showcwt.c +++ b/libavfilter/avf_showcwt.c @@ -69,14 +69,10 @@ typedef struct ShowCWTContext { char *rate_str; AVRational auto_frame_rate; AVRational frame_rate; - AVTXContext **fft; - AVTXContext **ifft; - av_tx_fn tx_fn; - av_tx_fn itx_fn; - int fft_in_size; - int fft_out_size; - int ifft_in_size; - int ifft_out_size; + AVTXContext **fft, **ifft; + av_tx_fn tx_fn, itx_fn; + int fft_in_size, fft_out_size; + int ifft_in_size, ifft_out_size; int pos; int64_t in_pts; int64_t old_pts; @@ -84,8 +80,7 @@ typedef struct ShowCWTContext { float *frequency_band; AVFrame *kernel; unsigned *index; - int *kernel_start; - int *kernel_stop; + int *kernel_start, *kernel_stop; AVFrame *cache; AVFrame *outpicref; AVFrame *fft_in; @@ -105,19 +100,14 @@ typedef struct ShowCWTContext { int slide; int new_frame; int direction; - int hop_size; - int hop_index; - int ihop_size; - int ihop_index; - int input_padding_size; - int input_sample_count; - int output_padding_size; - int output_sample_count; + int hop_size, ihop_size; + int hop_index, ihop_index; + int input_padding_size, output_padding_size; + int input_sample_count, output_sample_count; int frequency_band_count; float logarithmic_basis; int frequency_scale; - float minimum_frequency; - float maximum_frequency; + float minimum_frequency, maximum_frequency; float deviation; float bar_ratio; int bar_size; @@ -585,9 +575,9 @@ static int run_channel_cwt(AVFilterContext *ctx, void *arg, int jobnr, int nb_jo ShowCWTContext *s = ctx->priv; const int ch = *(int *)arg; AVComplexFloat *dst = (AVComplexFloat *)s->fft_out->extended_data[ch]; - const int output_sample_count = s->output_sample_count; + const int output_padding_size = s->output_padding_size; const int ihop_size = s->ihop_size; - const int ioffset = (s->output_padding_size - ihop_size) >> 1; + const int ioffset = (output_padding_size - ihop_size) >> 1; const int count = s->frequency_band_count; const int start = (count * jobnr) / nb_jobs; const int end = (count * (jobnr+1)) / nb_jobs; @@ -611,7 +601,7 @@ static int run_channel_cwt(AVFilterContext *ctx, void *arg, int jobnr, int nb_jo s->fdsp->vector_fmul((float *)dstx, (const float *)srcx, (const float *)kernelx, FFALIGN(kernel_range * 2, 16)); - memset(isrc, 0, sizeof(*isrc) * output_sample_count); + memset(isrc, 0, sizeof(*isrc) * output_padding_size); for (int i = 0; i < kernel_range; i++) { const unsigned n = index[i + kernel_start]; @@ -627,9 +617,10 @@ static int run_channel_cwt(AVFilterContext *ctx, void *arg, int jobnr, int nb_jo return 0; } -static void compute_kernel(AVFilterContext *ctx) +static int compute_kernel(AVFilterContext *ctx) { ShowCWTContext *s = ctx->priv; + const float correction = s->input_padding_size / (float)s->input_sample_count; const int size = s->input_sample_count; const float scale_factor = 1.f/(float)size; const int output_sample_count = s->output_sample_count; @@ -637,17 +628,18 @@ static void compute_kernel(AVFilterContext *ctx) int *kernel_start = s->kernel_start; int *kernel_stop = s->kernel_stop; unsigned *index = s->index; + int kernel_min = INT_MAX; + int kernel_max = 0; for (int y = 0; y < fsize; y++) { AVComplexFloat *kernel = (AVComplexFloat *)s->kernel->extended_data[y]; - float frequency = s->frequency_band[y*2]; + float frequency = s->frequency_band[y*2] * correction; float deviation = 1.f / (s->frequency_band[y*2+1] * - output_sample_count); + output_sample_count * correction); for (int n = 0; n < size; n++) { float ff, f = fabsf(n-frequency); - f = size - fabsf(f - size); ff = expf(-f*f*deviation) * scale_factor; kernel[n].re = ff; kernel[n].im = ff; @@ -655,21 +647,33 @@ static void compute_kernel(AVFilterContext *ctx) for (int n = 0; n < size; n++) { if (kernel[n].re != 0.f) { + if (kernel[n].re > FLT_MIN) + av_log(ctx, AV_LOG_DEBUG, "out of range kernel\n"); kernel_start[y] = n; + kernel_min = FFMIN(kernel_start[y], kernel_min); break; } } for (int n = 0; n < size; n++) { if (kernel[size - n - 1].re != 0.f) { + if (kernel[size - n - 1].re > FLT_MIN) + av_log(ctx, AV_LOG_DEBUG, "out of range kernel\n"); kernel_stop[y] = size - n - 1; + kernel_max = FFMAX(kernel_stop[y], kernel_max); break; } } } for (int n = 0; n < size; n++) - index[n] = n % output_sample_count; + index[n] = n % s->output_padding_size; + + av_log(ctx, AV_LOG_DEBUG, "kernel_min: %d\n", kernel_min); + av_log(ctx, AV_LOG_DEBUG, "kernel_max: %d\n", kernel_max); + av_log(ctx, AV_LOG_DEBUG, "kernel_range: %d\n", kernel_max - kernel_min); + + return kernel_max - kernel_min; } static int config_output(AVFilterLink *outlink) @@ -708,9 +712,12 @@ static int config_output(AVFilterLink *outlink) s->nb_consumed_samples = 65536; s->input_sample_count = s->nb_consumed_samples; - s->hop_size = s->nb_consumed_samples >> 1; - s->input_padding_size = 65536; - s->output_padding_size = FFMAX(16, av_rescale(s->input_padding_size, s->pps, inlink->sample_rate)); + s->input_padding_size = 1 << (32 - ff_clz(s->input_sample_count)); + s->output_sample_count = FFMAX(1, av_rescale(s->input_sample_count, s->pps, inlink->sample_rate)); + s->output_padding_size = 1 << (32 - ff_clz(s->output_sample_count)); + + s->hop_size = s->input_padding_size >> 1; + s->ihop_size = s->output_padding_size >> 1; outlink->w = s->w; outlink->h = s->h; @@ -719,11 +726,8 @@ static int config_output(AVFilterLink *outlink) s->fft_in_size = FFALIGN(s->input_padding_size, av_cpu_max_align()); s->fft_out_size = FFALIGN(s->input_padding_size, av_cpu_max_align()); - s->output_sample_count = s->output_padding_size; - s->ifft_in_size = FFALIGN(s->output_padding_size, av_cpu_max_align()); s->ifft_out_size = FFALIGN(s->output_padding_size, av_cpu_max_align()); - s->ihop_size = s->output_padding_size >> 1; s->fft = av_calloc(s->nb_threads, sizeof(*s->fft)); if (!s->fft) @@ -821,7 +825,7 @@ static int config_output(AVFilterLink *outlink) s->outpicref->color_range = AVCOL_RANGE_JPEG; - factor = s->nb_consumed_samples / (float)inlink->sample_rate; + factor = s->input_sample_count / (float)inlink->sample_rate; minimum_frequency *= factor; maximum_frequency *= factor; @@ -861,7 +865,9 @@ static int config_output(AVFilterLink *outlink) minimum_frequency, s->frequency_scale, s->deviation); av_log(ctx, AV_LOG_DEBUG, "input_sample_count: %d\n", s->input_sample_count); + av_log(ctx, AV_LOG_DEBUG, "input_padding_size: %d\n", s->input_padding_size); av_log(ctx, AV_LOG_DEBUG, "output_sample_count: %d\n", s->output_sample_count); + av_log(ctx, AV_LOG_DEBUG, "output_padding_size: %d\n", s->output_padding_size); switch (s->direction) { case DIRECTION_LR: @@ -1042,7 +1048,7 @@ static int output_frame(AVFilterContext *ctx) if (s->slide != SLIDE_FRAME || s->new_frame == 1) { int64_t pts_offset = s->new_frame ? 0LL : av_rescale(s->ihop_index, s->hop_size, s->ihop_size); - pts_offset = av_rescale_q(pts_offset - 16384LL, av_make_q(1, inlink->sample_rate), inlink->time_base); + pts_offset = av_rescale_q(pts_offset - s->input_sample_count/2, av_make_q(1, inlink->sample_rate), inlink->time_base); s->outpicref->pts = av_rescale_q(s->in_pts + pts_offset, inlink->time_base, outlink->time_base); s->outpicref->duration = 1; }

1 0

avformat/id3v2: Free buffer in decode_str()
by Michael Niedermayer 22 Jul '23

22 Jul '23

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Fri Jun 30 01:57:03 2023 +0200| [c41ab871c703bc41a4753c784adff1c5fe3795ca] | committer: Michael Niedermayer avformat/id3v2: Free buffer in decode_str() Fixes: memleak Fixes: 60058/clusterfuzz-testcase-minimized-ffmpeg_dem_IFF_fuzzer-5665259244093440 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c41ab871c703bc41a… --- libavformat/id3v2.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/libavformat/id3v2.c b/libavformat/id3v2.c index 38c86a8e79..69193933e0 100644 --- a/libavformat/id3v2.c +++ b/libavformat/id3v2.c @@ -309,8 +309,10 @@ static int decode_str(AVFormatContext *s, AVIOContext *pb, int encoding, avio_w8(dynbuf, 0); dynsize = avio_close_dyn_buf(dynbuf, dst); - if (dynsize <= 0) + if (dynsize <= 0) { + av_freep(dst); return AVERROR(ENOMEM); + } *maxread = left; return 0;

1 0

avcodec/rka: Fix integer overflow in decode_filter()
by Michael Niedermayer 22 Jul '23

22 Jul '23

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Fri Jun 30 01:24:20 2023 +0200| [f9b29451e422e672595e95eefd03879a4398a1f6] | committer: Michael Niedermayer avcodec/rka: Fix integer overflow in decode_filter() Fixes: signed integer overflow: 2147443649 + 65535 cannot be represented in type 'int' Fixes: 60054/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_RKA_fuzzer-5095674572832768 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f9b29451e422e6725… --- libavcodec/rka.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/rka.c b/libavcodec/rka.c index d56f4faee4..7646d776be 100644 --- a/libavcodec/rka.c +++ b/libavcodec/rka.c @@ -745,7 +745,7 @@ static int decode_filter(RKAContext *s, ChContext *ctx, ACoder *ac, int off, uns } ctx->buf1[off] = sum - ctx->buf0[off + -1]; ctx->buf0[off] = sum; - m += FFABS(ctx->buf1[off]); + m += (unsigned)FFABS(ctx->buf1[off]); } } if (ctx->cmode2 != 0) {

1 0

avformat/rka: bps < 8 is invalid
by Michael Niedermayer 22 Jul '23

22 Jul '23

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Apr 9 16:01:35 2023 +0200| [167b4f56f187edafd94fa2e706897ac3bb450511] | committer: Michael Niedermayer avformat/rka: bps < 8 is invalid Fixes: division by zero Fixes: 57828/clusterfuzz-testcase-minimized-ffmpeg_dem_RKA_fuzzer-6571818338353152 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=167b4f56f187edafd… --- libavformat/rka.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavformat/rka.c b/libavformat/rka.c index 39e5b3bce1..36e25ade01 100644 --- a/libavformat/rka.c +++ b/libavformat/rka.c @@ -72,7 +72,7 @@ static int rka_read_header(AVFormatContext *s) if (channels == 0) return AVERROR_INVALIDDATA; bps = par->extradata[13]; - if (bps == 0) + if (bps < 8) return AVERROR_INVALIDDATA; size_offset = avio_rl32(s->pb); framepos = avio_tell(s->pb);

1 0

avcodec: Rename ff_kbd_window_init() as it will be needed from outside libavcodec
by Michael Niedermayer 22 Jul '23

22 Jul '23

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Fri Jun 16 19:22:49 2023 +0200| [fd5aa93a37b3fa21195c6d7b22ef655124020e09] | committer: Michael Niedermayer avcodec: Rename ff_kbd_window_init() as it will be needed from outside libavcodec Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=fd5aa93a37b3fa211… --- libavcodec/aacdec_template.c | 8 ++++---- libavcodec/aactab.c | 4 ++-- libavcodec/ac3dec.c | 2 +- libavcodec/ac3enc_fixed.c | 2 +- libavcodec/ac3enc_float.c | 2 +- libavcodec/dolby_e.c | 4 ++-- libavcodec/kbdwin.c | 4 ++-- libavcodec/kbdwin.h | 6 +++--- 8 files changed, 16 insertions(+), 16 deletions(-) diff --git a/libavcodec/aacdec_template.c b/libavcodec/aacdec_template.c index 444dc4fa9d..2385ea58b0 100644 --- a/libavcodec/aacdec_template.c +++ b/libavcodec/aacdec_template.c @@ -1159,8 +1159,8 @@ static av_cold void aac_static_table_init(void) 352); // window initialization - AAC_RENAME(ff_kbd_window_init)(AAC_RENAME(aac_kbd_long_960), 4.0, 960); - AAC_RENAME(ff_kbd_window_init)(AAC_RENAME(aac_kbd_short_120), 6.0, 120); + AAC_RENAME(avpriv_kbd_window_init)(AAC_RENAME(aac_kbd_long_960), 4.0, 960); + AAC_RENAME(avpriv_kbd_window_init)(AAC_RENAME(aac_kbd_short_120), 6.0, 120); #if !USE_FIXED AAC_RENAME(ff_sine_window_init)(AAC_RENAME(sine_960), 960); @@ -1168,8 +1168,8 @@ static av_cold void aac_static_table_init(void) AAC_RENAME(ff_init_ff_sine_windows)(9); ff_aac_float_common_init(); #else - AAC_RENAME(ff_kbd_window_init)(AAC_RENAME2(aac_kbd_long_1024), 4.0, 1024); - AAC_RENAME(ff_kbd_window_init)(AAC_RENAME2(aac_kbd_short_128), 6.0, 128); + AAC_RENAME(avpriv_kbd_window_init)(AAC_RENAME2(aac_kbd_long_1024), 4.0, 1024); + AAC_RENAME(avpriv_kbd_window_init)(AAC_RENAME2(aac_kbd_short_128), 6.0, 128); init_sine_windows_fixed(); #endif diff --git a/libavcodec/aactab.c b/libavcodec/aactab.c index 0f4941d5df..d0006eac35 100644 --- a/libavcodec/aactab.c +++ b/libavcodec/aactab.c @@ -48,8 +48,8 @@ DECLARE_ALIGNED(32, float, ff_aac_kbd_short_128)[128]; static av_cold void aac_float_common_init(void) { - ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024); - ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128); + avpriv_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024); + avpriv_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128); ff_init_ff_sine_windows(10); ff_init_ff_sine_windows(7); } diff --git a/libavcodec/ac3dec.c b/libavcodec/ac3dec.c index fc0cbeb493..8c63f015f4 100644 --- a/libavcodec/ac3dec.c +++ b/libavcodec/ac3dec.c @@ -228,7 +228,7 @@ static av_cold int ac3_decode_init(AVCodecContext *avctx) if ((ret = av_tx_init(&s->tx_256, &s->tx_fn_256, IMDCT_TYPE, 1, 256, &scale, 0))) return ret; - AC3_RENAME(ff_kbd_window_init)(s->window, 5.0, 256); + AC3_RENAME(avpriv_kbd_window_init)(s->window, 5.0, 256); ff_bswapdsp_init(&s->bdsp); #if (USE_FIXED) diff --git a/libavcodec/ac3enc_fixed.c b/libavcodec/ac3enc_fixed.c index 88dfd66b91..079a43dc39 100644 --- a/libavcodec/ac3enc_fixed.c +++ b/libavcodec/ac3enc_fixed.c @@ -82,7 +82,7 @@ static av_cold int ac3_fixed_mdct_init(AC3EncodeContext *s) if (!iwin) return AVERROR(ENOMEM); - ff_kbd_window_init(fwin, 5.0, AC3_BLOCK_SIZE); + avpriv_kbd_window_init(fwin, 5.0, AC3_BLOCK_SIZE); for (int i = 0; i < AC3_BLOCK_SIZE; i++) iwin[i] = lrintf(fwin[i] * (1 << 22)); diff --git a/libavcodec/ac3enc_float.c b/libavcodec/ac3enc_float.c index ae351a535e..9664adbf63 100644 --- a/libavcodec/ac3enc_float.c +++ b/libavcodec/ac3enc_float.c @@ -92,7 +92,7 @@ static av_cold int ac3_float_mdct_init(AC3EncodeContext *s) return AVERROR(ENOMEM); } - ff_kbd_window_init(window, 5.0, AC3_BLOCK_SIZE); + avpriv_kbd_window_init(window, 5.0, AC3_BLOCK_SIZE); s->mdct_window = window; return av_tx_init(&s->tx, &s->tx_fn, AV_TX_FLOAT_MDCT, 0, diff --git a/libavcodec/dolby_e.c b/libavcodec/dolby_e.c index 921c33f3ba..b8dac0fa3f 100644 --- a/libavcodec/dolby_e.c +++ b/libavcodec/dolby_e.c @@ -1200,7 +1200,7 @@ static av_cold void init_tables(void) gain_tab[i] = exp2f((i - 960) / 64.0f); // short 1 - ff_kbd_window_init(window, 3.0f, 128); + avpriv_kbd_window_init(window, 3.0f, 128); for (i = 0; i < 128; i++) window[128 + i] = window[127 - i]; @@ -1227,7 +1227,7 @@ static av_cold void init_tables(void) window[1088 + i] = 1.0f; // long - ff_kbd_window_init(window + 1408, 3.0f, 256); + avpriv_kbd_window_init(window + 1408, 3.0f, 256); for (i = 0; i < 640; i++) window[1664 + i] = 1.0f; for (i = 0; i < 256; i++) diff --git a/libavcodec/kbdwin.c b/libavcodec/kbdwin.c index 82755874d4..eacdb46774 100644 --- a/libavcodec/kbdwin.c +++ b/libavcodec/kbdwin.c @@ -56,12 +56,12 @@ av_cold static int kbd_window_init(float *float_window, int *int_window, float a return 0; } -av_cold int ff_kbd_window_init(float *window, float alpha, int n) +av_cold int avpriv_kbd_window_init(float *window, float alpha, int n) { return kbd_window_init(window, NULL, alpha, n); } -av_cold int ff_kbd_window_init_fixed(int32_t *window, float alpha, int n) +av_cold int avpriv_kbd_window_init_fixed(int32_t *window, float alpha, int n) { return kbd_window_init(NULL, window, alpha, n); } diff --git a/libavcodec/kbdwin.h b/libavcodec/kbdwin.h index 452fc46596..0cb2073c5f 100644 --- a/libavcodec/kbdwin.h +++ b/libavcodec/kbdwin.h @@ -22,7 +22,7 @@ #include <stdint.h> /** - * Maximum window size for ff_kbd_window_init. + * Maximum window size for avpriv_kbd_window_init. */ #define FF_KBD_WINDOW_MAX 1024 @@ -34,7 +34,7 @@ * * @return if n is larger than FF_KBD_WINDOW_MAX then AVERROR(ENOMEM) is possible */ -int ff_kbd_window_init(float *window, float alpha, int n); -int ff_kbd_window_init_fixed(int32_t *window, float alpha, int n); +int avpriv_kbd_window_init(float *window, float alpha, int n); +int avpriv_kbd_window_init_fixed(int32_t *window, float alpha, int n); #endif /* AVCODEC_KBDWIN_H */

1 0

avcodec/kbdwin: Support arbitrary sized windows
by Michael Niedermayer 22 Jul '23

22 Jul '23

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Wed Jun 14 17:22:55 2023 +0200| [cf00f60bab1f79213c274a6cd4357b32bd5c0101] | committer: Michael Niedermayer avcodec/kbdwin: Support arbitrary sized windows Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=cf00f60bab1f79213… --- libavcodec/kbdwin.c | 22 ++++++++++++++-------- libavcodec/kbdwin.h | 8 +++++--- 2 files changed, 19 insertions(+), 11 deletions(-) diff --git a/libavcodec/kbdwin.c b/libavcodec/kbdwin.c index 163f9d5251..82755874d4 100644 --- a/libavcodec/kbdwin.c +++ b/libavcodec/kbdwin.c @@ -19,20 +19,23 @@ #include "libavutil/avassert.h" #include "libavutil/mathematics.h" #include "libavutil/attributes.h" +#include "libavutil/mem.h" #include "kbdwin.h" -av_cold static void kbd_window_init(float *float_window, int *int_window, float alpha, int n) +av_cold static int kbd_window_init(float *float_window, int *int_window, float alpha, int n) { int i; double sum = 0.0, tmp; double scale = 0.0; - double temp[FF_KBD_WINDOW_MAX / 2 + 1]; + double temp_small[FF_KBD_WINDOW_MAX / 2 + 1]; + double *temp= n<=FF_KBD_WINDOW_MAX ? temp_small : av_malloc((n/2+1) * sizeof(*temp)); double alpha2 = 4 * (alpha * M_PI / n) * (alpha * M_PI / n); - av_assert0(n <= FF_KBD_WINDOW_MAX); + if (!temp) + return AVERROR(ENOMEM); for (i = 0; i <= n / 2; i++) { - tmp = i * (n - i) * alpha2; + tmp = alpha2 * i * (n - i); temp[i] = av_bessel_i0(sqrt(tmp)); scale += temp[i] * (1 + (i && i<n/2)); } @@ -48,14 +51,17 @@ av_cold static void kbd_window_init(float *float_window, int *int_window, float if (float_window) float_window[i] = sqrt(sum * scale); else int_window[i] = lrint(2147483647 * sqrt(sum * scale)); } + if (temp != temp_small) + av_free(temp); + return 0; } -av_cold void ff_kbd_window_init(float *window, float alpha, int n) +av_cold int ff_kbd_window_init(float *window, float alpha, int n) { - kbd_window_init(window, NULL, alpha, n); + return kbd_window_init(window, NULL, alpha, n); } -av_cold void ff_kbd_window_init_fixed(int32_t *window, float alpha, int n) +av_cold int ff_kbd_window_init_fixed(int32_t *window, float alpha, int n) { - kbd_window_init(NULL, window, alpha, n); + return kbd_window_init(NULL, window, alpha, n); } diff --git a/libavcodec/kbdwin.h b/libavcodec/kbdwin.h index 4185c4206f..452fc46596 100644 --- a/libavcodec/kbdwin.h +++ b/libavcodec/kbdwin.h @@ -30,9 +30,11 @@ * Generate a Kaiser-Bessel Derived Window. * @param window pointer to half window * @param alpha determines window shape - * @param n size of half window, max FF_KBD_WINDOW_MAX + * @param n size of half window + * + * @return if n is larger than FF_KBD_WINDOW_MAX then AVERROR(ENOMEM) is possible */ -void ff_kbd_window_init(float *window, float alpha, int n); -void ff_kbd_window_init_fixed(int32_t *window, float alpha, int n); +int ff_kbd_window_init(float *window, float alpha, int n); +int ff_kbd_window_init_fixed(int32_t *window, float alpha, int n); #endif /* AVCODEC_KBDWIN_H */

1 0

avcodec/pcm: allow Changing parameters
by Michael Niedermayer 22 Jul '23

22 Jul '23

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Mon Jun 5 15:56:58 2023 +0200| [94d44dbe212b3ecb67256c4edfc3d7c3c3ac4472] | committer: Michael Niedermayer avcodec/pcm: allow Changing parameters SDR needs this for switching between mono and stereo stations Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=94d44dbe212b3ecb6… --- libavcodec/pcm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/pcm.c b/libavcodec/pcm.c index 23955ba2dd..467ecb4fe0 100644 --- a/libavcodec/pcm.c +++ b/libavcodec/pcm.c @@ -578,7 +578,7 @@ const FFCodec ff_ ## name_ ## _decoder = { \ .priv_data_size = sizeof(PCMDecode), \ .init = pcm_decode_init, \ FF_CODEC_DECODE_CB(pcm_decode_frame), \ - .p.capabilities = AV_CODEC_CAP_DR1, \ + .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_PARAM_CHANGE, \ .p.sample_fmts = (const enum AVSampleFormat[]){ sample_fmt_, \ AV_SAMPLE_FMT_NONE }, \ }

1 0

avutil/tx_template: extend to 2M
by Michael Niedermayer 22 Jul '23

22 Jul '23

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Tue May 30 02:59:11 2023 +0200| [8f48a62151f4a299574959df9a48813303ef4edb] | committer: Michael Niedermayer avutil/tx_template: extend to 2M Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=8f48a62151f4a2995… --- libavutil/tx_template.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/libavutil/tx_template.c b/libavutil/tx_template.c index 983de75a47..c4ec9502e0 100644 --- a/libavutil/tx_template.c +++ b/libavutil/tx_template.c @@ -43,6 +43,10 @@ SR_TABLE(32768) \ SR_TABLE(65536) \ SR_TABLE(131072) \ + SR_TABLE(262144) \ + SR_TABLE(524288) \ + SR_TABLE(1048576) \ + SR_TABLE(2097152) \ #define SR_TABLE(len) \ TABLE_DEF(len, len/4 + 1); @@ -717,6 +721,10 @@ DECL_SR_CODELET(16384,8192,4096) DECL_SR_CODELET(32768,16384,8192) DECL_SR_CODELET(65536,32768,16384) DECL_SR_CODELET(131072,65536,32768) +DECL_SR_CODELET(262144,131072,65536) +DECL_SR_CODELET(524288,262144,131072) +DECL_SR_CODELET(1048576,524288,262144) +DECL_SR_CODELET(2097152,1048576,524288) static av_cold int TX_NAME(ff_tx_fft_init)(AVTXContext *s, const FFTXCodelet *cd, @@ -1947,6 +1955,10 @@ const FFTXCodelet * const TX_NAME(ff_tx_codelet_list)[] = { &TX_NAME(ff_tx_fft32768_ns_def), &TX_NAME(ff_tx_fft65536_ns_def), &TX_NAME(ff_tx_fft131072_ns_def), + &TX_NAME(ff_tx_fft262144_ns_def), + &TX_NAME(ff_tx_fft524288_ns_def), + &TX_NAME(ff_tx_fft1048576_ns_def), + &TX_NAME(ff_tx_fft2097152_ns_def), /* Prime factor codelets */ &TX_NAME(ff_tx_fft3_ns_def),

1 0

[ffmpeg-radio] branch master updated (db12164d0a -> bbdfeb4ca4)
by ffmpeg-git 22 Jul '23

22 Jul '23

This is an automated email from the git hooks/post-receive script. Git pushed a change to branch master in repository libavradio. from db12164d0a avradio/sdr: workaround inverted gain parameter on sdrplay new 6029450572 avradio/sdrdemux: Add Mittelwelle / Mediumwave / Mediumfrequency band new 4561bcd02e avradio/sdrdemux: Fix corner case in snap2band new ec21dd22b4 avradio/sdrdemux: Fix seeking to stations at the edge of the range new e06dd1e69c avradio/sdrdemux: snap2station() should look only at the vissible window new 4e424b2141 avradio/sdr: snap2station() documentation new 018370ac29 avradio/sdrdemux: do not hack bandwidth value in random places new 37717d5141 avradio/sdrdemux: snap2station consider more distant stations new e3eb650a52 tests: Add avradio/sdrdemux tests new 70d6058364 avradio/sdr: Remove code setting frequency from variable being 0 new 326fe6e456 avradio/sdrdemux: no need to set wanted_freq from sdrfile_initial_setup() new 195eacee23 avradio/sdr: More atomics, less Mutexes new a3f7ae9830 avradio/sdrdemux: + vs. - bug new 2b2f054d1e avradio/sdrdemux: snap2band doesnt change anything so use const new bbdfeb4ca4 avradio/sdrdemux: Fix wraparound with seeking The 14 revisions listed above as "new" are entirely new to this repository and will be described in separate emails. The revisions listed as "add" were already present in the repository and have only been added to this reference. Summary of changes: libavradio/sdr.h | 6 ++-- libavradio/sdrdemux.c | 96 +++++++++++++++++++++++++++++++++++-------------- libavradio/sdrinradio.c | 5 +-- tests/Makefile | 1 + tests/fate/sdr.mak | 11 ++++++ tests/ref/fate/sdr-am | 55 ++++++++++++++++++++++++++++ tests/ref/fate/sdr-fm | 25 +++++++++++++ 7 files changed, 167 insertions(+), 32 deletions(-) create mode 100644 tests/fate/sdr.mak create mode 100644 tests/ref/fate/sdr-am create mode 100644 tests/ref/fate/sdr-fm

2 14

libavutil: Remove TOMI CPU
by Kieran Kunhya 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Kieran Kunhya <kierank(a)obe.tv> | Sat Dec 17 21:53:27 2022 +0000| [a4e616824b785df2a3598a47a306e2242bbe3a25] | committer: Kieran Kunhya libavutil: Remove TOMI CPU > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a4e616824b785df2a… --- configure | 1 - libavutil/intreadwrite.h | 2 - libavutil/tomi/intreadwrite.h | 150 ------------------------------------------ 3 files changed, 153 deletions(-) diff --git a/configure b/configure index d46e868a0c..eb2fe392b4 100755 --- a/configure +++ b/configure @@ -2089,7 +2089,6 @@ ARCH_LIST=" sparc64 tilegx tilepro - tomi x86 x86_32 x86_64 diff --git a/libavutil/intreadwrite.h b/libavutil/intreadwrite.h index 4c8413a536..c4679fdba4 100644 --- a/libavutil/intreadwrite.h +++ b/libavutil/intreadwrite.h @@ -72,8 +72,6 @@ typedef union { # include "mips/intreadwrite.h" #elif ARCH_PPC # include "ppc/intreadwrite.h" -#elif ARCH_TOMI -# include "tomi/intreadwrite.h" #elif ARCH_X86 # include "x86/intreadwrite.h" #endif diff --git a/libavutil/tomi/intreadwrite.h b/libavutil/tomi/intreadwrite.h deleted file mode 100644 index 7dec4158d3..0000000000 --- a/libavutil/tomi/intreadwrite.h +++ /dev/null @@ -1,150 +0,0 @@ -/* - * Copyright (c) 2010 Mans Rullgard <mans(a)mansr.com> - * - * This file is part of FFmpeg. - * - * FFmpeg is free software; you can redistribute it and/or - * modify it under the terms of the GNU Lesser General Public - * License as published by the Free Software Foundation; either - * version 2.1 of the License, or (at your option) any later version. - * - * FFmpeg is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * Lesser General Public License for more details. - * - * You should have received a copy of the GNU Lesser General Public - * License along with FFmpeg; if not, write to the Free Software - * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - */ - -#ifndef AVUTIL_TOMI_INTREADWRITE_H -#define AVUTIL_TOMI_INTREADWRITE_H - -#include <stdint.h> - -#include "config.h" -#include "libavutil/attributes.h" - -#define AV_RB16 AV_RB16 -static av_always_inline uint16_t AV_RB16(const void *p) -{ - uint16_t v; - __asm__ ("loadacc, (%1+) \n\t" - "rol8 \n\t" - "storeacc, %0 \n\t" - "loadacc, (%1+) \n\t" - "add, %0 \n\t" - : "=r"(v), "+a"(p)); - return v; -} - -#define AV_WB16 AV_WB16 -static av_always_inline void AV_WB16(void *p, uint16_t v) -{ - __asm__ volatile ("loadacc, %1 \n\t" - "lsr8 \n\t" - "storeacc, (%0+) \n\t" - "loadacc, %1 \n\t" - "storeacc, (%0+) \n\t" - : "+&a"(p) : "r"(v)); -} - -#define AV_RL16 AV_RL16 -static av_always_inline uint16_t AV_RL16(const void *p) -{ - uint16_t v; - __asm__ ("loadacc, (%1+) \n\t" - "storeacc, %0 \n\t" - "loadacc, (%1+) \n\t" - "rol8 \n\t" - "add, %0 \n\t" - : "=r"(v), "+a"(p)); - return v; -} - -#define AV_WL16 AV_WL16 -static av_always_inline void AV_WL16(void *p, uint16_t v) -{ - __asm__ volatile ("loadacc, %1 \n\t" - "storeacc, (%0+) \n\t" - "lsr8 \n\t" - "storeacc, (%0+) \n\t" - : "+&a"(p) : "r"(v)); -} - -#define AV_RB32 AV_RB32 -static av_always_inline uint32_t AV_RB32(const void *p) -{ - uint32_t v; - __asm__ ("loadacc, (%1+) \n\t" - "rol8 \n\t" - "rol8 \n\t" - "rol8 \n\t" - "storeacc, %0 \n\t" - "loadacc, (%1+) \n\t" - "rol8 \n\t" - "rol8 \n\t" - "add, %0 \n\t" - "loadacc, (%1+) \n\t" - "rol8 \n\t" - "add, %0 \n\t" - "loadacc, (%1+) \n\t" - "add, %0 \n\t" - : "=r"(v), "+a"(p)); - return v; -} - -#define AV_WB32 AV_WB32 -static av_always_inline void AV_WB32(void *p, uint32_t v) -{ - __asm__ volatile ("loadacc, #4 \n\t" - "add, %0 \n\t" - "loadacc, %1 \n\t" - "storeacc, (-%0) \n\t" - "lsr8 \n\t" - "storeacc, (-%0) \n\t" - "lsr8 \n\t" - "storeacc, (-%0) \n\t" - "lsr8 \n\t" - "storeacc, (-%0) \n\t" - : "+&a"(p) : "r"(v)); -} - -#define AV_RL32 AV_RL32 -static av_always_inline uint32_t AV_RL32(const void *p) -{ - uint32_t v; - __asm__ ("loadacc, (%1+) \n\t" - "storeacc, %0 \n\t" - "loadacc, (%1+) \n\t" - "rol8 \n\t" - "add, %0 \n\t" - "loadacc, (%1+) \n\t" - "rol8 \n\t" - "rol8 \n\t" - "add, %0 \n\t" - "loadacc, (%1+) \n\t" - "rol8 \n\t" - "rol8 \n\t" - "rol8 \n\t" - "add, %0 \n\t" - : "=r"(v), "+a"(p)); - return v; -} - -#define AV_WL32 AV_WL32 -static av_always_inline void AV_WL32(void *p, uint32_t v) -{ - __asm__ volatile ("loadacc, %1 \n\t" - "storeacc, (%0+) \n\t" - "lsr8 \n\t" - "storeacc, (%0+) \n\t" - "lsr8 \n\t" - "storeacc, (%0+) \n\t" - "lsr8 \n\t" - "storeacc, (%0+) \n\t" - : "+&a"(p) : "r"(v)); -} - -#endif /* AVUTIL_TOMI_INTREADWRITE_H */

1 0

swscale/rgb2rgb2: rework RISC-V V shuffle_bytes_{1230,3012}
by Rémi Denis-Courmont 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Rémi Denis-Courmont <remi(a)remlab.net> | Tue Jul 18 20:25:13 2023 +0300| [c2b38619c081658b303d1669155803d2aebf3f6a] | committer: Rémi Denis-Courmont swscale/rgb2rgb2: rework RISC-V V shuffle_bytes_{1230,3012} This avoids strided loads. Before: shuffle_bytes_1230_rvv_i32: 308.7 shuffle_bytes_3012_rvv_i32: 308.7 After: shuffle_bytes_1230_rvv_i32: 46.7 shuffle_bytes_3012_rvv_i32: 46.7 > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c2b38619c081658b3… --- libswscale/riscv/rgb2rgb_rvv.S | 30 ++++++++++++++++++++---------- 1 file changed, 20 insertions(+), 10 deletions(-) diff --git a/libswscale/riscv/rgb2rgb_rvv.S b/libswscale/riscv/rgb2rgb_rvv.S index 366f80c1c3..0ac3afff7c 100644 --- a/libswscale/riscv/rgb2rgb_rvv.S +++ b/libswscale/riscv/rgb2rgb_rvv.S @@ -49,19 +49,29 @@ func ff_shuffle_bytes_2103_rvv, zve32x endfunc func ff_shuffle_bytes_1230_rvv, zve32x - addi t1, a0, 2 - addi t2, a0, 3 - addi t3, a0, 0 - addi a0, a0, 1 - j 1b + li t1, 24 + li t2, 8 + j 3f endfunc func ff_shuffle_bytes_3012_rvv, zve32x - addi t1, a0, 0 - addi t2, a0, 1 - addi t3, a0, 2 - addi a0, a0, 3 - j 1b + li t1, 8 + li t2, 24 +3: + srai a2, a2, 2 +4: + vsetvli t0, a2, e32, m8, ta, ma + vle32.v v8, (a0) + sub a2, a2, t0 + vsll.vx v16, v8, t1 + sh2add a0, t0, a0 + vsrl.vx v8, v8, t2 + vor.vv v16, v16, v8 + vse32.v v16, (a1) + sh2add a1, t0, a1 + bnez a2, 4b + + ret endfunc func ff_shuffle_bytes_3210_rvv, zve32x

1 0

swscale/rgb2rgb2: rework RISC-V V shuffle_bytes_{0321,2103}
by Rémi Denis-Courmont 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Rémi Denis-Courmont <remi(a)remlab.net> | Tue Jul 18 21:39:59 2023 +0300| [15982554e694e3ccf8d72677e401ded86ca5605b] | committer: Rémi Denis-Courmont swscale/rgb2rgb2: rework RISC-V V shuffle_bytes_{0321,2103} This avoids strided loads. Before: shuffle_bytes_0321_rvv_i32: 307.7 shuffle_bytes_2103_rvv_i32: 308.7 After: shuffle_bytes_0321_rvv_i32: 59.7 shuffle_bytes_2103_rvv_i32: 61.5 > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=15982554e694e3ccf… --- libswscale/riscv/rgb2rgb_rvv.S | 45 +++++++++++++++++++----------------------- 1 file changed, 20 insertions(+), 25 deletions(-) diff --git a/libswscale/riscv/rgb2rgb_rvv.S b/libswscale/riscv/rgb2rgb_rvv.S index 8f30d760cb..366f80c1c3 100644 --- a/libswscale/riscv/rgb2rgb_rvv.S +++ b/libswscale/riscv/rgb2rgb_rvv.S @@ -21,38 +21,33 @@ #include "libavutil/riscv/asm.S" func ff_shuffle_bytes_0321_rvv, zve32x - addi t1, a0, 3 - addi t2, a0, 2 - addi t3, a0, 1 + li t1, 0x00ff00ff + j 1f +endfunc + +func ff_shuffle_bytes_2103_rvv, zve32x + li t1, ~0x00ff00ff 1: + not t2, t1 srai a2, a2, 2 - li t4, 4 2: - vsetvli t0, a2, e8, m1, ta, ma - sub a2, a2, t0 - vlse8.v v8, (a0), t4 - sh2add a0, t0, a0 - vlse8.v v9, (t1), t4 - sh2add t1, t0, t1 - vlse8.v v10, (t2), t4 - sh2add t2, t0, t2 - vlse8.v v11, (t3), t4 - sh2add t3, t0, t3 - vsseg4e8.v v8, (a1) - sh2add a1, t0, a1 - bnez a2, 2b + vsetvli t0, a2, e32, m8, ta, ma + vle32.v v8, (a0) + sub a2, a2, t0 + vand.vx v16, v8, t2 + sh2add a0, t0, a0 + vand.vx v8, v8, t1 + vsrl.vi v24, v16, 16 + vsll.vi v16, v16, 16 + vor.vv v8, v8, v24 + vor.vv v8, v16, v8 + vse32.v v8, (a1) + sh2add a1, t0, a1 + bnez a2, 2b ret endfunc -func ff_shuffle_bytes_2103_rvv, zve32x - addi t1, a0, 1 - addi t2, a0, 0 - addi t3, a0, 3 - addi a0, a0, 2 - j 1b -endfunc - func ff_shuffle_bytes_1230_rvv, zve32x addi t1, a0, 2 addi t2, a0, 3

1 0

swscale: inline ff_shuffle_bytes_3210_rvv
by Rémi Denis-Courmont 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Rémi Denis-Courmont <remi(a)remlab.net> | Tue Jul 18 21:11:15 2023 +0300| [d3948e4db50666da05368d2bc84134fea2e20006] | committer: Rémi Denis-Courmont swscale: inline ff_shuffle_bytes_3210_rvv No functional changes. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d3948e4db50666da0… --- libswscale/riscv/rgb2rgb_rvv.S | 19 ++++++++++++++++++- 1 file changed, 18 insertions(+), 1 deletion(-) diff --git a/libswscale/riscv/rgb2rgb_rvv.S b/libswscale/riscv/rgb2rgb_rvv.S index bbdfdbebbc..8f30d760cb 100644 --- a/libswscale/riscv/rgb2rgb_rvv.S +++ b/libswscale/riscv/rgb2rgb_rvv.S @@ -74,7 +74,24 @@ func ff_shuffle_bytes_3210_rvv, zve32x addi t2, a0, 1 addi t3, a0, 0 addi a0, a0, 3 - j 1b + srai a2, a2, 2 + li t4, 4 +1: + vsetvli t0, a2, e8, m1, ta, ma + sub a2, a2, t0 + vlse8.v v8, (a0), t4 + sh2add a0, t0, a0 + vlse8.v v9, (t1), t4 + sh2add t1, t0, t1 + vlse8.v v10, (t2), t4 + sh2add t2, t0, t2 + vlse8.v v11, (t3), t4 + sh2add t3, t0, t3 + vsseg4e8.v v8, (a1) + sh2add a1, t0, a1 + bnez a2, 1b + + ret endfunc func ff_interleave_bytes_rvv, zve32x

1 0

vulkan_av1: remove an invalid debug print
by Lynne 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Fri Jul 21 19:34:00 2023 +0200| [e6a12a5214b8d14f1fe88c655548e411c1503ff9] | committer: Lynne vulkan_av1: remove an invalid debug print Leftover. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e6a12a5214b8d14f1… --- libavcodec/vulkan_av1.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/libavcodec/vulkan_av1.c b/libavcodec/vulkan_av1.c index e06457bbc3..d0f4019248 100644 --- a/libavcodec/vulkan_av1.c +++ b/libavcodec/vulkan_av1.c @@ -484,8 +484,6 @@ static int vk_av1_start_frame(AVCodecContext *avctx, ap->av1_frame_header.film_grain.ar_coeffs_cr_plus_128[24] = film_grain->ar_coeffs_cr_plus_128[24]; } - av_log(avctx, AV_LOG_DEBUG, "Created frame parameters"); - /* Workaround for a spec issue. */ ap->dec = dec;

1 0

lavu/vulkan: remove unused field from the execution pool structure
by Lynne 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Wed Jul 19 05:38:32 2023 +0200| [3e3d46309bfb4049c4bcea3d219be7bd299b6ab7] | committer: Lynne lavu/vulkan: remove unused field from the execution pool structure > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3e3d46309bfb4049c… --- libavutil/vulkan.h | 1 - 1 file changed, 1 deletion(-) diff --git a/libavutil/vulkan.h b/libavutil/vulkan.h index 7171fb3c42..20b81105dd 100644 --- a/libavutil/vulkan.h +++ b/libavutil/vulkan.h @@ -208,7 +208,6 @@ typedef struct FFVkExecContext { } FFVkExecContext; typedef struct FFVkExecPool { - FFVkQueueFamilyCtx *qf; FFVkExecContext *contexts; atomic_int_least64_t idx;

1 0

lavu/vulkan: remove threadsafe buffer index load and fix a signed overflow
by Lynne 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Wed Jul 19 04:54:37 2023 +0200| [97890c2b553aa2447c0ea1a3cea37f25f59a3449] | committer: Lynne lavu/vulkan: remove threadsafe buffer index load and fix a signed overflow It's not needed anymore. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=97890c2b553aa2447… --- libavutil/vulkan.c | 4 +--- libavutil/vulkan.h | 2 +- 2 files changed, 2 insertions(+), 4 deletions(-) diff --git a/libavutil/vulkan.c b/libavutil/vulkan.c index 26b9b6f1fb..48f5f4b5dc 100644 --- a/libavutil/vulkan.c +++ b/libavutil/vulkan.c @@ -282,8 +282,6 @@ int ff_vk_exec_pool_init(FFVulkanContext *s, FFVkQueueFamilyCtx *qf, VkCommandPoolCreateInfo cqueue_create; VkCommandBufferAllocateInfo cbuf_create; - atomic_init(&pool->idx, 0); - /* Create command pool */ cqueue_create = (VkCommandPoolCreateInfo) { .sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO, @@ -472,7 +470,7 @@ VkResult ff_vk_exec_get_query(FFVulkanContext *s, FFVkExecContext *e, FFVkExecContext *ff_vk_exec_get(FFVkExecPool *pool) { - int idx = atomic_fetch_add_explicit(&pool->idx, 1, memory_order_relaxed); + uint32_t idx = pool->idx++; idx %= pool->pool_size; return &pool->contexts[idx]; } diff --git a/libavutil/vulkan.h b/libavutil/vulkan.h index bbbc9374ae..7171fb3c42 100644 --- a/libavutil/vulkan.h +++ b/libavutil/vulkan.h @@ -151,7 +151,7 @@ typedef struct FFVulkanPipeline { } FFVulkanPipeline; typedef struct FFVkExecContext { - int idx; + uint32_t idx; const struct FFVkExecPool *parent; pthread_mutex_t lock;

1 0

lavc/vulkan_decode: use a single execution pool per thread
by Lynne 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Wed Jul 19 05:39:07 2023 +0200| [c06ad641ec36ea0e9011be8d6182612c22f6196b] | committer: Lynne lavc/vulkan_decode: use a single execution pool per thread The spec says command buffer pools must be externally synchronized objects. This still lets us pool some, just not as much. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c06ad641ec36ea0e9… --- libavcodec/vulkan_decode.c | 86 ++++++++++++++++++++++++++++++++++------------ libavcodec/vulkan_decode.h | 3 +- 2 files changed, 66 insertions(+), 23 deletions(-) diff --git a/libavcodec/vulkan_decode.c b/libavcodec/vulkan_decode.c index 973c7ca548..f20733fb39 100644 --- a/libavcodec/vulkan_decode.c +++ b/libavcodec/vulkan_decode.c @@ -42,12 +42,53 @@ static const VkExtensionProperties *dec_ext[] = { #endif }; +static const VkVideoProfileInfoKHR *get_video_profile(FFVulkanDecodeShared *ctx, enum AVCodecID codec_id) +{ + const VkVideoProfileListInfoKHR *profile_list; + + VkStructureType profile_struct_type = + codec_id == AV_CODEC_ID_H264 ? VK_STRUCTURE_TYPE_VIDEO_DECODE_H264_PROFILE_INFO_KHR : + codec_id == AV_CODEC_ID_HEVC ? VK_STRUCTURE_TYPE_VIDEO_DECODE_H265_PROFILE_INFO_KHR : + codec_id == AV_CODEC_ID_AV1 ? VK_STRUCTURE_TYPE_VIDEO_DECODE_AV1_PROFILE_INFO_MESA : + 0; + + profile_list = ff_vk_find_struct(ctx->s.hwfc->create_pnext, + VK_STRUCTURE_TYPE_VIDEO_PROFILE_LIST_INFO_KHR); + if (!profile_list) + return NULL; + + for (int i = 0; i < profile_list->profileCount; i++) + if (ff_vk_find_struct(profile_list->pProfiles[i].pNext, profile_struct_type)) + return &profile_list->pProfiles[i]; + + return NULL; +} + int ff_vk_update_thread_context(AVCodecContext *dst, const AVCodecContext *src) { int err; FFVulkanDecodeContext *src_ctx = src->internal->hwaccel_priv_data; FFVulkanDecodeContext *dst_ctx = dst->internal->hwaccel_priv_data; + if (!dst_ctx->exec_pool.cmd_bufs) { + FFVulkanDecodeShared *ctx = (FFVulkanDecodeShared *)src_ctx->shared_ref->data; + + const VkVideoProfileInfoKHR *profile = get_video_profile(ctx, dst->codec_id); + if (!profile) { + av_log(dst, AV_LOG_ERROR, "Video profile missing from frames context!"); + return AVERROR(EINVAL); + } + + err = ff_vk_exec_pool_init(&ctx->s, &ctx->qf, + &dst_ctx->exec_pool, + src_ctx->exec_pool.pool_size, + src_ctx->exec_pool.nb_queries, + VK_QUERY_TYPE_RESULT_STATUS_ONLY_KHR, 0, + profile); + if (err < 0) + return err; + } + err = av_buffer_replace(&dst_ctx->shared_ref, src_ctx->shared_ref); if (err < 0) return err; @@ -271,7 +312,7 @@ void ff_vk_decode_flush(AVCodecContext *avctx) }; VkCommandBuffer cmd_buf; - FFVkExecContext *exec = ff_vk_exec_get(&ctx->exec_pool); + FFVkExecContext *exec = ff_vk_exec_get(&dec->exec_pool); ff_vk_exec_start(&ctx->s, exec); cmd_buf = exec->buf; @@ -317,7 +358,7 @@ int ff_vk_decode_frame(AVCodecContext *avctx, size_t data_size = FFALIGN(vp->slices_size, ctx->caps.minBitstreamBufferSizeAlignment); - FFVkExecContext *exec = ff_vk_exec_get(&ctx->exec_pool); + FFVkExecContext *exec = ff_vk_exec_get(&dec->exec_pool); /* The current decoding reference has to be bound as an inactive reference */ VkVideoReferenceSlotInfoKHR *cur_vk_ref; @@ -326,7 +367,7 @@ int ff_vk_decode_frame(AVCodecContext *avctx, cur_vk_ref[0].slotIndex = -1; decode_start.referenceSlotCount++; - if (ctx->exec_pool.nb_queries) { + if (dec->exec_pool.nb_queries) { int64_t prev_sub_res = 0; ff_vk_exec_wait(&ctx->s, exec); ret = ff_vk_exec_get_query(&ctx->s, exec, NULL, &prev_sub_res); @@ -495,14 +536,14 @@ int ff_vk_decode_frame(AVCodecContext *avctx, vk->CmdBeginVideoCodingKHR(cmd_buf, &decode_start); /* Start status query */ - if (ctx->exec_pool.nb_queries) - vk->CmdBeginQuery(cmd_buf, ctx->exec_pool.query_pool, exec->query_idx + 0, 0); + if (dec->exec_pool.nb_queries) + vk->CmdBeginQuery(cmd_buf, dec->exec_pool.query_pool, exec->query_idx + 0, 0); vk->CmdDecodeVideoKHR(cmd_buf, &vp->decode_info); /* End status query */ - if (ctx->exec_pool.nb_queries) - vk->CmdEndQuery(cmd_buf, ctx->exec_pool.query_pool, exec->query_idx + 0); + if (dec->exec_pool.nb_queries) + vk->CmdEndQuery(cmd_buf, dec->exec_pool.query_pool, exec->query_idx + 0); vk->CmdEndVideoCodingKHR(cmd_buf, &decode_end); @@ -555,9 +596,6 @@ static void free_common(void *opaque, uint8_t *data) FFVulkanContext *s = &ctx->s; FFVulkanFunctions *vk = &ctx->s.vkfn; - /* Wait on and free execution pool */ - ff_vk_exec_pool_free(s, &ctx->exec_pool); - /* Destroy layered view */ if (ctx->layered_view) vk->DestroyImageView(s->hwctx->act_dev, ctx->layered_view, s->hwctx->alloc); @@ -1029,6 +1067,11 @@ void ff_vk_decode_free_params(void *opaque, uint8_t *data) int ff_vk_decode_uninit(AVCodecContext *avctx) { FFVulkanDecodeContext *dec = avctx->internal->hwaccel_priv_data; + FFVulkanDecodeShared *ctx = (FFVulkanDecodeShared *)dec->shared_ref->data; + + /* Wait on and free execution pool */ + ff_vk_exec_pool_free(&ctx->s, &dec->exec_pool); + av_buffer_pool_uninit(&dec->tmp_pool); av_buffer_unref(&dec->session_params); av_buffer_unref(&dec->shared_ref); @@ -1044,8 +1087,7 @@ int ff_vk_decode_init(AVCodecContext *avctx) FFVulkanDecodeShared *ctx; FFVulkanContext *s; FFVulkanFunctions *vk; - FFVkQueueFamilyCtx qf_dec; - const VkVideoProfileListInfoKHR *profile_list; + const VkVideoProfileInfoKHR *profile; VkVideoDecodeH264SessionParametersCreateInfoKHR h264_params = { .sType = VK_STRUCTURE_TYPE_VIDEO_DECODE_H264_SESSION_PARAMETERS_CREATE_INFO_KHR, @@ -1089,10 +1131,9 @@ int ff_vk_decode_init(AVCodecContext *avctx) s->device = (AVHWDeviceContext *)s->frames->device_ref->data; s->hwctx = s->device->hwctx; - profile_list = ff_vk_find_struct(s->hwfc->create_pnext, - VK_STRUCTURE_TYPE_VIDEO_PROFILE_LIST_INFO_KHR); - if (!profile_list) { - av_log(avctx, AV_LOG_ERROR, "Profile list missing from frames context!"); + profile = get_video_profile(ctx, avctx->codec_id); + if (!profile) { + av_log(avctx, AV_LOG_ERROR, "Video profile missing from frames context!"); return AVERROR(EINVAL); } @@ -1101,7 +1142,7 @@ int ff_vk_decode_init(AVCodecContext *avctx) goto fail; /* Create queue context */ - qf = ff_vk_qf_init(s, &qf_dec, VK_QUEUE_VIDEO_DECODE_BIT_KHR); + qf = ff_vk_qf_init(s, &ctx->qf, VK_QUEUE_VIDEO_DECODE_BIT_KHR); /* Check for support */ if (!(s->video_props[qf].videoCodecOperations & @@ -1123,14 +1164,14 @@ int ff_vk_decode_init(AVCodecContext *avctx) session_create.pictureFormat = s->hwfc->format[0]; session_create.referencePictureFormat = session_create.pictureFormat; session_create.pStdHeaderVersion = dec_ext[avctx->codec_id]; - session_create.pVideoProfile = &profile_list->pProfiles[0]; + session_create.pVideoProfile = profile; - /* Create decode exec context. + /* Create decode exec context for this specific main thread. * 2 async contexts per thread was experimentally determined to be optimal * for a majority of streams. */ - err = ff_vk_exec_pool_init(s, &qf_dec, &ctx->exec_pool, 2*avctx->thread_count, + err = ff_vk_exec_pool_init(s, &ctx->qf, &dec->exec_pool, 2, nb_q, VK_QUERY_TYPE_RESULT_STATUS_ONLY_KHR, 0, - session_create.pVideoProfile); + profile); if (err < 0) goto fail; @@ -1168,7 +1209,8 @@ int ff_vk_decode_init(AVCodecContext *avctx) dpb_frames->height = s->frames->height; dpb_hwfc = dpb_frames->hwctx; - dpb_hwfc->create_pnext = (void *)profile_list; + dpb_hwfc->create_pnext = (void *)ff_vk_find_struct(ctx->s.hwfc->create_pnext, + VK_STRUCTURE_TYPE_VIDEO_PROFILE_LIST_INFO_KHR); dpb_hwfc->format[0] = s->hwfc->format[0]; dpb_hwfc->tiling = VK_IMAGE_TILING_OPTIMAL; dpb_hwfc->usage = VK_IMAGE_USAGE_VIDEO_DECODE_DPB_BIT_KHR | diff --git a/libavcodec/vulkan_decode.h b/libavcodec/vulkan_decode.h index 4e45cbde71..1b4e1cc712 100644 --- a/libavcodec/vulkan_decode.h +++ b/libavcodec/vulkan_decode.h @@ -37,7 +37,7 @@ typedef struct FFVulkanDecodeProfileData { typedef struct FFVulkanDecodeShared { FFVulkanContext s; FFVkVideoCommon common; - FFVkExecPool exec_pool; + FFVkQueueFamilyCtx qf; VkVideoCapabilitiesKHR caps; VkVideoDecodeCapabilitiesKHR dec_caps; @@ -56,6 +56,7 @@ typedef struct FFVulkanDecodeShared { typedef struct FFVulkanDecodeContext { AVBufferRef *shared_ref; AVBufferRef *session_params; + FFVkExecPool exec_pool; int dedicated_dpb; /* Oddity #1 - separate DPB images */ int layered_dpb; /* Madness #1 - layered DPB images */

1 0

hwcontext_vulkan: hide Linux-only header after 571756bf2fe2
by Jan Beich 21 Jul '23

21 Jul '23

ffmpeg | branch: master | Jan Beich <jbeich(a)FreeBSD.org> | Sat Jul 15 04:57:47 2023 +0200| [e6bd8b132372f64a3378c069cb9855d3d91bde09] | committer: Lynne hwcontext_vulkan: hide Linux-only header after 571756bf2fe2 major/minor are in <sys/types.h> on BSDs and <sys/mkdev.h> on Solaris-like. libavutil/hwcontext_vulkan.c:55:10: fatal error: 'sys/sysmacros.h' file not found #include <sys/sysmacros.h> ^~~~~~~~~~~~~~~~~ > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e6bd8b132372f64a3… --- libavutil/hwcontext_vulkan.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c index ca802cc86e..75314f1407 100644 --- a/libavutil/hwcontext_vulkan.c +++ b/libavutil/hwcontext_vulkan.c @@ -52,7 +52,9 @@ #if CONFIG_VAAPI #include <va/va_drmcommon.h> #endif +#ifdef __linux__ #include <sys/sysmacros.h> +#endif #include <sys/stat.h> #include <xf86drm.h> #include <drm_fourcc.h>

1 0

avfilter/avf_showcwt: fix index storing
by Paul B Mahol 20 Jul '23

20 Jul '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Fri Jul 21 00:28:01 2023 +0200| [74474a551b63f778776e94d094c5ea55c3639b9b] | committer: Paul B Mahol avfilter/avf_showcwt: fix index storing > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=74474a551b63f7787… --- libavfilter/avf_showcwt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavfilter/avf_showcwt.c b/libavfilter/avf_showcwt.c index 51a45ef636..ee0d697e1e 100644 --- a/libavfilter/avf_showcwt.c +++ b/libavfilter/avf_showcwt.c @@ -662,7 +662,7 @@ static void compute_kernel(AVFilterContext *ctx) for (int n = 0; n < size; n++) { if (kernel[size - n - 1].re != 0.f) { - kernel_stop[y] = size - n; + kernel_stop[y] = size - n - 1; break; } }

1 0

avfilter/avf_showcwt: move variables out of loop
by Paul B Mahol 20 Jul '23

20 Jul '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Thu Jul 20 22:30:42 2023 +0200| [86a1904f417cdbeac48f4def86cd51e9b29d1ed1] | committer: Paul B Mahol avfilter/avf_showcwt: move variables out of loop > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=86a1904f417cdbeac… --- libavfilter/avf_showcwt.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/libavfilter/avf_showcwt.c b/libavfilter/avf_showcwt.c index 578312cdd6..51a45ef636 100644 --- a/libavfilter/avf_showcwt.c +++ b/libavfilter/avf_showcwt.c @@ -634,12 +634,12 @@ static void compute_kernel(AVFilterContext *ctx) const float scale_factor = 1.f/(float)size; const int output_sample_count = s->output_sample_count; const int fsize = s->frequency_band_count; + int *kernel_start = s->kernel_start; + int *kernel_stop = s->kernel_stop; unsigned *index = s->index; for (int y = 0; y < fsize; y++) { AVComplexFloat *kernel = (AVComplexFloat *)s->kernel->extended_data[y]; - int *kernel_start = s->kernel_start; - int *kernel_stop = s->kernel_stop; float frequency = s->frequency_band[y*2]; float deviation = 1.f / (s->frequency_band[y*2+1] * output_sample_count);

1 0

avfilter/vf_pseudocolor: add four more presets
by Paul B Mahol 20 Jul '23

20 Jul '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Thu Jul 20 09:37:20 2023 +0200| [3dc93acc56258e25ae43176aafb85e2191695105] | committer: Paul B Mahol avfilter/vf_pseudocolor: add four more presets > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3dc93acc56258e25a… --- doc/filters.texi | 4 +++ libavfilter/vf_pseudocolor.c | 70 +++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 73 insertions(+), 1 deletion(-) diff --git a/doc/filters.texi b/doc/filters.texi index 1f1890b97d..9690332092 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -19711,6 +19711,10 @@ Available LUTs: @item spectral @item cool @item heat +@item fiery +@item blues +@item green +@item helix @end table @item opacity diff --git a/libavfilter/vf_pseudocolor.c b/libavfilter/vf_pseudocolor.c index 5b6cd5e70a..31532e2824 100644 --- a/libavfilter/vf_pseudocolor.c +++ b/libavfilter/vf_pseudocolor.c @@ -70,6 +70,10 @@ enum Curves { SPECTRAL, COOL, HEAT, + FIERY, + BLUES, + GREEN, + HELIX, NB_CURVES, }; @@ -91,6 +95,10 @@ enum Presets { PRESET_SPECTRAL, PRESET_COOL, PRESET_HEAT, + PRESET_FIERY, + PRESET_BLUES, + PRESET_GREEN, + PRESET_HELIX, NB_PRESETS, }; @@ -162,6 +170,26 @@ static double heatfunv(double x) return 0.25 * sin(2.0 * x * M_PI) + 0.5; } +static double fieryfunu(double x) +{ + return 0.75 - 0.25 * cos(2.0 * x * M_PI); +} + +static double fieryfunv(double x) +{ + return 0.25 + 0.25 * cos(2.0 * x * M_PI); +} + +static double helixfunu(double x) +{ + return 0.5 + 0.15 * sin(5.0 * x * M_PI + M_PI); +} + +static double helixfunv(double x) +{ + return 0.5 + 0.15 * cos(6.0 * x * M_PI + M_PI_2); +} + static const Curve curves[] = { [MAGMA] = {{ @@ -222,6 +250,38 @@ static const Curve curves[] = .offset = { 0., 0., 0 }, .yuv = 1, .fun = { heatfunu, limit, heatfunv }, }, + [FIERY] = {{ + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + }, + .offset = { 0., 0., 0 }, + .yuv = 1, + .fun = { fieryfunu, limit, fieryfunv }, }, + [BLUES] = {{ + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + }, + .offset = { 0., 0., 0 }, + .yuv = 1, + .fun = { fieryfunv, limit, fieryfunu }, }, + [GREEN] = {{ + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + }, + .offset = { 0., 0., 0 }, + .yuv = 1, + .fun = { fieryfunv, limit, fieryfunv }, }, + [HELIX] = {{ + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + { 0, 0, 0, 0, 0, 0, 1./256, 0 }, + }, + .offset = { 0., 0., 0 }, + .yuv = 1, + .fun = { helixfunu, limit, helixfunv }, }, }; static const Preset presets[] = @@ -242,7 +302,11 @@ static const Preset presets[] = [PRESET_SOLAR] = { 1, &full_range, &curves[SOLAR], NULL }, [PRESET_SPECTRAL]= { 1, &full_range, &curves[SPECTRAL],NULL }, [PRESET_COOL] = { 1, &full_range, &curves[COOL], NULL }, - [PRESET_HEAT]= { 1, &full_range, &curves[HEAT],NULL }, + [PRESET_HEAT] = { 1, &full_range, &curves[HEAT], NULL }, + [PRESET_FIERY] = { 1, &full_range, &curves[FIERY], NULL }, + [PRESET_BLUES] = { 1, &full_range, &curves[BLUES], NULL }, + [PRESET_GREEN] = { 1, &full_range, &curves[GREEN], NULL }, + [PRESET_HELIX] = { 1, &full_range, &curves[HELIX], NULL }, }; typedef struct PseudoColorContext { @@ -300,6 +364,10 @@ static const AVOption pseudocolor_options[] = { { "spectral", NULL, 0, AV_OPT_TYPE_CONST, {.i64=PRESET_SPECTRAL},.flags = FLAGS, "preset" }, { "cool", NULL, 0, AV_OPT_TYPE_CONST, {.i64=PRESET_COOL}, .flags = FLAGS, "preset" }, { "heat", NULL, 0, AV_OPT_TYPE_CONST, {.i64=PRESET_HEAT}, .flags = FLAGS, "preset" }, + { "fiery", NULL, 0, AV_OPT_TYPE_CONST, {.i64=PRESET_FIERY}, .flags = FLAGS, "preset" }, + { "blues", NULL, 0, AV_OPT_TYPE_CONST, {.i64=PRESET_BLUES}, .flags = FLAGS, "preset" }, + { "green", NULL, 0, AV_OPT_TYPE_CONST, {.i64=PRESET_GREEN}, .flags = FLAGS, "preset" }, + { "helix", NULL, 0, AV_OPT_TYPE_CONST, {.i64=PRESET_HELIX}, .flags = FLAGS, "preset" }, { "opacity", "set pseudocolor opacity",OFFSET(opacity), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 1, .flags = FLAGS }, { NULL } };

1 0

lavc/audiodsp: rework RISC-V V scalar product
by Rémi Denis-Courmont 20 Jul '23

20 Jul '23

ffmpeg | branch: master | Rémi Denis-Courmont <remi(a)remlab.net> | Mon Jul 17 20:46:06 2023 +0300| [44cac1def0fab47592882afe6695e39c20781839] | committer: Rémi Denis-Courmont lavc/audiodsp: rework RISC-V V scalar product Take vector reduction out of the loop and unroll. Before: audiodsp.scalarproduct_int16_c: 12321.0 audiodsp.scalarproduct_int16_rvv_i32: 4175.7 After: audiodsp.scalarproduct_int16_c: 12320.5 audiodsp.scalarproduct_int16_rvv_i32: 1230.2 > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=44cac1def0fab4759… --- libavcodec/riscv/audiodsp_rvv.S | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/libavcodec/riscv/audiodsp_rvv.S b/libavcodec/riscv/audiodsp_rvv.S index af1e07bef9..f7eba2114f 100644 --- a/libavcodec/riscv/audiodsp_rvv.S +++ b/libavcodec/riscv/audiodsp_rvv.S @@ -21,21 +21,22 @@ #include "libavutil/riscv/asm.S" func ff_scalarproduct_int16_rvv, zve32x - vsetivli zero, 1, e32, m1, ta, ma - vmv.s.x v8, zero + vsetvli t0, zero, e32, m8, ta, ma + vmv.v.x v8, zero + vmv.s.x v0, zero 1: - vsetvli t0, a2, e16, m1, ta, ma + vsetvli t0, a2, e16, m4, tu, ma vle16.v v16, (a0) sub a2, a2, t0 vle16.v v24, (a1) sh1add a0, t0, a0 - vwmul.vv v0, v16, v24 + vwmacc.vv v8, v16, v24 sh1add a1, t0, a1 - vsetvli zero, t0, e32, m2, ta, ma - vredsum.vs v8, v0, v8 bnez a2, 1b - vmv.x.s a0, v8 + vsetvli t0, zero, e32, m8, ta, ma + vredsum.vs v0, v8, v0 + vmv.x.s a0, v0 ret endfunc

1 0

lavu/float_dsp: rework RISC-V V scalar product
by Rémi Denis-Courmont 20 Jul '23

20 Jul '23

ffmpeg | branch: master | Rémi Denis-Courmont <remi(a)remlab.net> | Mon Jul 17 20:11:08 2023 +0300| [29b9d616c212f20bc0d3e24928919a14efc047f8] | committer: Rémi Denis-Courmont lavu/float_dsp: rework RISC-V V scalar product 1) Take the reductive sum out of the loop, leaving a regular vector addition in the loop. 2) Merge the addition and the multiplication. 3) Unroll. Before: scalarproduct_float_rvv_f32: 832.5 After: scalarproduct_float_rvv_f32: 275.2 > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=29b9d616c212f20bc… --- libavutil/riscv/float_dsp_rvv.S | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) diff --git a/libavutil/riscv/float_dsp_rvv.S b/libavutil/riscv/float_dsp_rvv.S index fa8f9dc212..7cfc890bc2 100644 --- a/libavutil/riscv/float_dsp_rvv.S +++ b/libavutil/riscv/float_dsp_rvv.S @@ -166,20 +166,22 @@ endfunc // a0 = (a0).(a1) [0..a2-1] func ff_scalarproduct_float_rvv, zve32f - vsetivli zero, 1, e32, m1, ta, ma - vmv.s.x v8, zero + vsetvli t0, zero, e32, m8, ta, ma + vmv.v.x v8, zero + vmv.s.x v0, zero 1: - vsetvli t0, a2, e32, m1, ta, ma + vsetvli t0, a2, e32, m8, tu, ma vle32.v v16, (a0) sub a2, a2, t0 vle32.v v24, (a1) sh2add a0, t0, a0 - vfmul.vv v16, v16, v24 + vfmacc.vv v8, v16, v24 sh2add a1, t0, a1 - vfredusum.vs v8, v16, v8 bnez a2, 1b - vfmv.f.s fa0, v8 + vsetvli t0, zero, e32, m8, ta, ma + vfredusum.vs v0, v8, v0 + vfmv.f.s fa0, v0 NOHWF fmv.x.w a0, fa0 ret endfunc

1 0

lavu/float_dsp: unroll RISC-V V loops
by Rémi Denis-Courmont 20 Jul '23

20 Jul '23

ffmpeg | branch: master | Rémi Denis-Courmont <remi(a)remlab.net> | Mon Jul 17 19:33:09 2023 +0300| [b710f881cefde728f869689193885b82af6b90cd] | committer: Rémi Denis-Courmont lavu/float_dsp: unroll RISC-V V loops butterflies_float_c: 1057.0 butterflies_float_rvv_f32: 351.0 (before) butterflies_float_rvv_f32: 329.5 (after) vector_dmac_scalar_c: 819.0 vector_dmac_scalar_rvv_f64: 670.5 (before) vector_dmac_scalar_rvv_f64: 431.0 (after) vector_dmul_c: 800.2 vector_dmul_rvv_f64: 541.5 (before) vector_dmul_rvv_f64: 426.0 (after) vector_dmul_scalar_c: 545.7 vector_dmul_scalar_rvv_f64: 670.7 (before) vector_dmul_scalar_rvv_f64: 324.7 (after) vector_fmac_scalar_c: 804.5 vector_fmac_scalar_rvv_f32: 412.7 (before) vector_fmac_scalar_rvv_f32: 214.5 (after) vector_fmul_c: 811.2 vector_fmul_rvv_f32: 285.7 (before) vector_fmul_rvv_f32: 214.2 (after) vector_fmul_add_c: 1313.0 vector_fmul_add_rvv_f32: 349.0 (before) vector_fmul_add_rvv_f32: 290.2 (after) vector_fmul_reverse_c: 815.7 vector_fmul_reverse_rvv_f32: 529.2 (before) vector_fmul_reverse_rvv_f32: 515.7 (after) vector_fmul_scalar_c: 546.0 vector_fmul_scalar_rvv_f32: 350.2 (before) vector_fmul_scalar_rvv_f32: 169.5 (after) > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=b710f881cefde728f… --- libavutil/riscv/float_dsp_rvv.S | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/libavutil/riscv/float_dsp_rvv.S b/libavutil/riscv/float_dsp_rvv.S index 77961b7387..fa8f9dc212 100644 --- a/libavutil/riscv/float_dsp_rvv.S +++ b/libavutil/riscv/float_dsp_rvv.S @@ -23,7 +23,7 @@ // (a0) = (a1) * (a2) [0..a3-1] func ff_vector_fmul_rvv, zve32f 1: - vsetvli t0, a3, e32, m1, ta, ma + vsetvli t0, a3, e32, m8, ta, ma vle32.v v16, (a1) sub a3, a3, t0 vle32.v v24, (a2) @@ -42,7 +42,7 @@ func ff_vector_fmac_scalar_rvv, zve32f NOHWF fmv.w.x fa0, a2 NOHWF mv a2, a3 1: - vsetvli t0, a2, e32, m1, ta, ma + vsetvli t0, a2, e32, m8, ta, ma slli t1, t0, 2 vle32.v v24, (a1) sub a2, a2, t0 @@ -61,7 +61,7 @@ func ff_vector_fmul_scalar_rvv, zve32f NOHWF fmv.w.x fa0, a2 NOHWF mv a2, a3 1: - vsetvli t0, a2, e32, m1, ta, ma + vsetvli t0, a2, e32, m8, ta, ma vle32.v v16, (a1) sub a2, a2, t0 vfmul.vf v16, v16, fa0 @@ -82,7 +82,7 @@ func ff_vector_fmul_window_rvv, zve32f sh2add t3, t1, a3 li t1, -4 // byte stride 1: - vsetvli t2, a4, e32, m1, ta, ma + vsetvli t2, a4, e32, m4, ta, ma vle32.v v16, (a1) slli t4, t2, 2 vlse32.v v20, (a2), t1 @@ -109,7 +109,7 @@ endfunc // (a0) = (a1) * (a2) + (a3) [0..a4-1] func ff_vector_fmul_add_rvv, zve32f 1: - vsetvli t0, a4, e32, m1, ta, ma + vsetvli t0, a4, e32, m8, ta, ma vle32.v v8, (a1) sub a4, a4, t0 vle32.v v16, (a2) @@ -131,7 +131,7 @@ func ff_vector_fmul_reverse_rvv, zve32f li t2, -4 // byte stride addi a2, a2, -4 1: - vsetvli t0, a3, e32, m1, ta, ma + vsetvli t0, a3, e32, m8, ta, ma slli t1, t0, 2 vle32.v v16, (a1) sub a3, a3, t0 @@ -149,7 +149,7 @@ endfunc // (a0) = (a0) + (a1), (a1) = (a0) - (a1) [0..a2-1] func ff_butterflies_float_rvv, zve32f 1: - vsetvli t0, a2, e32, m1, ta, ma + vsetvli t0, a2, e32, m8, ta, ma vle32.v v16, (a0) sub a2, a2, t0 vle32.v v24, (a1) @@ -187,7 +187,7 @@ endfunc // (a0) = (a1) * (a2) [0..a3-1] func ff_vector_dmul_rvv, zve64d 1: - vsetvli t0, a3, e64, m1, ta, ma + vsetvli t0, a3, e64, m8, ta, ma vle64.v v16, (a1) sub a3, a3, t0 vle64.v v24, (a2) @@ -206,7 +206,7 @@ func ff_vector_dmac_scalar_rvv, zve64d NOHWD fmv.d.x fa0, a2 NOHWD mv a2, a3 1: - vsetvli t0, a2, e64, m1, ta, ma + vsetvli t0, a2, e64, m8, ta, ma vle64.v v24, (a1) sub a2, a2, t0 vle64.v v16, (a0) @@ -224,7 +224,7 @@ func ff_vector_dmul_scalar_rvv, zve64d NOHWD fmv.d.x fa0, a2 NOHWD mv a2, a3 1: - vsetvli t0, a2, e64, m1, ta, ma + vsetvli t0, a2, e64, m8, ta, ma vle64.v v16, (a1) sub a2, a2, t0 vfmul.vf v16, v16, fa0

1 0

avcodec/x86/mathops: clip constants used with shift instructions within inline assembly
by Rémi Denis-Courmont 20 Jul '23

20 Jul '23

ffmpeg | branch: release/6.0 | Rémi Denis-Courmont <remi(a)remlab.net> | Sun Jul 16 18:18:02 2023 +0300| [cc703cf60759d9798f440a9417e4efa2fcbe2747] | committer: James Almer avcodec/x86/mathops: clip constants used with shift instructions within inline assembly Fixes assembling with binutil as >= 2.41 Signed-off-by: James Almer <jamrial(a)gmail.com> (cherry picked from commit effadce6c756247ea8bae32dc13bb3e6f464f0eb) > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=cc703cf60759d9798… --- libavcodec/x86/mathops.h | 26 +++++++++++++++++++++++--- 1 file changed, 23 insertions(+), 3 deletions(-) diff --git a/libavcodec/x86/mathops.h b/libavcodec/x86/mathops.h index 6298f5ed19..ca7e2dffc1 100644 --- a/libavcodec/x86/mathops.h +++ b/libavcodec/x86/mathops.h @@ -35,12 +35,20 @@ static av_always_inline av_const int MULL(int a, int b, unsigned shift) { int rt, dummy; + if (__builtin_constant_p(shift)) __asm__ ( "imull %3 \n\t" "shrdl %4, %%edx, %%eax \n\t" :"=a"(rt), "=d"(dummy) - :"a"(a), "rm"(b), "ci"((uint8_t)shift) + :"a"(a), "rm"(b), "i"(shift & 0x1F) ); + else + __asm__ ( + "imull %3 \n\t" + "shrdl %4, %%edx, %%eax \n\t" + :"=a"(rt), "=d"(dummy) + :"a"(a), "rm"(b), "c"((uint8_t)shift) + ); return rt; } @@ -113,19 +121,31 @@ __asm__ volatile(\ // avoid +32 for shift optimization (gcc should do that ...) #define NEG_SSR32 NEG_SSR32 static inline int32_t NEG_SSR32( int32_t a, int8_t s){ + if (__builtin_constant_p(s)) __asm__ ("sarl %1, %0\n\t" : "+r" (a) - : "ic" ((uint8_t)(-s)) + : "i" (-s & 0x1F) ); + else + __asm__ ("sarl %1, %0\n\t" + : "+r" (a) + : "c" ((uint8_t)(-s)) + ); return a; } #define NEG_USR32 NEG_USR32 static inline uint32_t NEG_USR32(uint32_t a, int8_t s){ + if (__builtin_constant_p(s)) __asm__ ("shrl %1, %0\n\t" : "+r" (a) - : "ic" ((uint8_t)(-s)) + : "i" (-s & 0x1F) ); + else + __asm__ ("shrl %1, %0\n\t" + : "+r" (a) + : "c" ((uint8_t)(-s)) + ); return a; }

1 0