September 2023 - ffmpeg-cvslog

avfft: wrap lavu/tx instead of ff_mdct
by Lynne 02 Sep '23

02 Sep '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Thu Nov 10 11:26:33 2022 +0100| [517e4fcca6b81e73536d4ef5edfd22ad5886e783] | committer: Lynne avfft: wrap lavu/tx instead of ff_mdct > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=517e4fcca6b81e735… --- libavcodec/avfft.c | 43 ++++++++++++++++++++++++++++++------------- 1 file changed, 30 insertions(+), 13 deletions(-) diff --git a/libavcodec/avfft.c b/libavcodec/avfft.c index e4b19af272..107b510828 100644 --- a/libavcodec/avfft.c +++ b/libavcodec/avfft.c @@ -75,43 +75,60 @@ av_cold void av_fft_end(FFTContext *s) } } -#if CONFIG_MDCT - FFTContext *av_mdct_init(int nbits, int inverse, double scale) { - FFTContext *s = av_malloc(sizeof(*s)); + int ret; + float scale_f = scale; + AVTXWrapper *s = av_malloc(sizeof(*s)); + if (!s) + return NULL; - if (s && ff_mdct_init(s, nbits, inverse, scale)) - av_freep(&s); + ret = av_tx_init(&s->ctx, &s->fn, AV_TX_FLOAT_MDCT, inverse, 1 << (nbits - 1), &scale_f, 0); + if (ret < 0) { + av_free(s); + return NULL; + } - return s; + if (inverse) { + ret = av_tx_init(&s->ctx2, &s->fn2, AV_TX_FLOAT_MDCT, inverse, 1 << (nbits - 1), + &scale_f, AV_TX_FULL_IMDCT); + if (ret < 0) { + av_tx_uninit(&s->ctx); + av_free(s); + return NULL; + } + } + + return (FFTContext *)s; } void av_imdct_calc(FFTContext *s, FFTSample *output, const FFTSample *input) { - s->imdct_calc(s, output, input); + AVTXWrapper *w = (AVTXWrapper *)s; + w->fn2(w->ctx2, output, (void *)input, sizeof(float)); } void av_imdct_half(FFTContext *s, FFTSample *output, const FFTSample *input) { - s->imdct_half(s, output, input); + AVTXWrapper *w = (AVTXWrapper *)s; + w->fn(w->ctx, output, (void *)input, sizeof(float)); } void av_mdct_calc(FFTContext *s, FFTSample *output, const FFTSample *input) { - s->mdct_calc(s, output, input); + AVTXWrapper *w = (AVTXWrapper *)s; + w->fn(w->ctx, output, (void *)input, sizeof(float)); } av_cold void av_mdct_end(FFTContext *s) { if (s) { - ff_mdct_end(s); - av_free(s); + AVTXWrapper *w = (AVTXWrapper *)s; + av_tx_uninit(&w->ctx); + av_free(w); } } -#endif /* CONFIG_MDCT */ - #if CONFIG_RDFT RDFTContext *av_rdft_init(int nbits, enum RDFTransformType trans)

1 0

avfft: wrap lavu/tx instead of ff_fft
by Lynne 02 Sep '23

02 Sep '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Thu Nov 10 11:23:38 2022 +0100| [dfcd4bbf5118b9fa601ba36e3c5984db61478d5f] | committer: Lynne avfft: wrap lavu/tx instead of ff_fft > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=dfcd4bbf5118b9fa6… --- libavcodec/avfft.c | 40 +++++++++++++++++++++++++++++++--------- 1 file changed, 31 insertions(+), 9 deletions(-) diff --git a/libavcodec/avfft.c b/libavcodec/avfft.c index 2200f37708..e4b19af272 100644 --- a/libavcodec/avfft.c +++ b/libavcodec/avfft.c @@ -18,38 +18,60 @@ #include "libavutil/attributes.h" #include "libavutil/mem.h" +#include "libavutil/tx.h" #include "avfft.h" #include "fft.h" #include "rdft.h" #include "dct.h" +typedef struct AVTXWrapper { + AVTXContext *ctx; + av_tx_fn fn; + + AVTXContext *ctx2; + av_tx_fn fn2; + + ptrdiff_t stride; +} AVTXWrapper; + /* FFT */ FFTContext *av_fft_init(int nbits, int inverse) { - FFTContext *s = av_mallocz(sizeof(*s)); - - if (s && ff_fft_init(s, nbits, inverse)) - av_freep(&s); + int ret; + float scale = 1.0f; + AVTXWrapper *s = av_malloc(sizeof(*s)); + if (!s) + return NULL; + + ret = av_tx_init(&s->ctx, &s->fn, AV_TX_FLOAT_FFT, inverse, 1 << nbits, + &scale, AV_TX_INPLACE); + if (ret < 0) { + av_free(s); + return NULL; + } - return s; + return (FFTContext *)s; } void av_fft_permute(FFTContext *s, FFTComplex *z) { - s->fft_permute(s, z); + /* Empty */ } void av_fft_calc(FFTContext *s, FFTComplex *z) { - s->fft_calc(s, z); + AVTXWrapper *w = (AVTXWrapper *)s; + w->fn(w->ctx, z, (void *)z, sizeof(AVComplexFloat)); } av_cold void av_fft_end(FFTContext *s) { if (s) { - ff_fft_end(s); - av_free(s); + AVTXWrapper *w = (AVTXWrapper *)s; + av_tx_uninit(&w->ctx); + av_tx_uninit(&w->ctx2); + av_free(w); } }

1 0

ffplay: port to lavu/tx
by Lynne 02 Sep '23

02 Sep '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Sat Feb 18 13:14:31 2023 +0100| [4acd08be6c4f39736179a3d90fd56b508e42ff6d] | committer: Lynne ffplay: port to lavu/tx > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=4acd08be6c4f39736… --- fftools/ffplay.c | 42 +++++++++++++++++++++++++++--------------- 1 file changed, 27 insertions(+), 15 deletions(-) diff --git a/fftools/ffplay.c b/fftools/ffplay.c index 5212ad053e..006da7ab57 100644 --- a/fftools/ffplay.c +++ b/fftools/ffplay.c @@ -47,7 +47,7 @@ #include "libavdevice/avdevice.h" #include "libswscale/swscale.h" #include "libavutil/opt.h" -#include "libavcodec/avfft.h" +#include "libavutil/tx.h" #include "libswresample/swresample.h" #include "libavfilter/avfilter.h" @@ -262,9 +262,11 @@ typedef struct VideoState { int16_t sample_array[SAMPLE_ARRAY_SIZE]; int sample_array_index; int last_i_start; - RDFTContext *rdft; + AVTXContext *rdft; + av_tx_fn rdft_fn; int rdft_bits; - FFTSample *rdft_data; + float *real_data; + AVComplexFloat *rdft_data; int xpos; double last_vis_time; SDL_Texture *vis_texture; @@ -1120,6 +1122,7 @@ static void video_audio_display(VideoState *s) fill_rectangle(s->xleft, y, s->width, 1); } } else { + int err = 0; if (realloc_texture(&s->vis_texture, SDL_PIXELFORMAT_ARGB8888, s->width, s->height, SDL_BLENDMODE_NONE, 1) < 0) return; @@ -1127,31 +1130,39 @@ static void video_audio_display(VideoState *s) s->xpos = 0; nb_display_channels= FFMIN(nb_display_channels, 2); if (rdft_bits != s->rdft_bits) { - av_rdft_end(s->rdft); - av_free(s->rdft_data); - s->rdft = av_rdft_init(rdft_bits, DFT_R2C); + const float rdft_scale = 1.0; + av_tx_uninit(&s->rdft); + av_freep(&s->real_data); + av_freep(&s->rdft_data); s->rdft_bits = rdft_bits; - s->rdft_data = av_malloc_array(nb_freq, 4 *sizeof(*s->rdft_data)); + s->real_data = av_malloc_array(nb_freq, 4 *sizeof(*s->real_data)); + s->rdft_data = av_malloc_array(nb_freq + 1, 2 *sizeof(*s->rdft_data)); + err = av_tx_init(&s->rdft, &s->rdft_fn, AV_TX_FLOAT_RDFT, + 0, 1 << rdft_bits, &rdft_scale, 0); } - if (!s->rdft || !s->rdft_data){ + if (err < 0 || !s->rdft_data) { av_log(NULL, AV_LOG_ERROR, "Failed to allocate buffers for RDFT, switching to waves display\n"); s->show_mode = SHOW_MODE_WAVES; } else { - FFTSample *data[2]; + float *data_in[2]; + AVComplexFloat *data[2]; SDL_Rect rect = {.x = s->xpos, .y = 0, .w = 1, .h = s->height}; uint32_t *pixels; int pitch; for (ch = 0; ch < nb_display_channels; ch++) { - data[ch] = s->rdft_data + 2 * nb_freq * ch; + data_in[ch] = s->real_data + 2 * nb_freq * ch; + data[ch] = s->rdft_data + nb_freq * ch; i = i_start + ch; for (x = 0; x < 2 * nb_freq; x++) { double w = (x-nb_freq) * (1.0 / nb_freq); - data[ch][x] = s->sample_array[i] * (1.0 - w * w); + data_in[ch][x] = s->sample_array[i] * (1.0 - w * w); i += channels; if (i >= SAMPLE_ARRAY_SIZE) i -= SAMPLE_ARRAY_SIZE; } - av_rdft_calc(s->rdft, data[ch]); + s->rdft_fn(s->rdft, data[ch], data_in[ch], sizeof(float)); + data[ch][0].im = data[ch][nb_freq].re; + data[ch][nb_freq].re = 0; } /* Least efficient way to do this, we should of course * directly access it but it is more than fast enough. */ @@ -1160,8 +1171,8 @@ static void video_audio_display(VideoState *s) pixels += pitch * s->height; for (y = 0; y < s->height; y++) { double w = 1 / sqrt(nb_freq); - int a = sqrt(w * sqrt(data[0][2 * y + 0] * data[0][2 * y + 0] + data[0][2 * y + 1] * data[0][2 * y + 1])); - int b = (nb_display_channels == 2 ) ? sqrt(w * hypot(data[1][2 * y + 0], data[1][2 * y + 1])) + int a = sqrt(w * sqrt(data[0][y].re * data[0][y].re + data[0][y].im * data[0][y].im)); + int b = (nb_display_channels == 2 ) ? sqrt(w * hypot(data[1][y].re, data[1][y].im)) : a; a = FFMIN(a, 255); b = FFMIN(b, 255); @@ -1197,7 +1208,8 @@ static void stream_component_close(VideoState *is, int stream_index) is->audio_buf = NULL; if (is->rdft) { - av_rdft_end(is->rdft); + av_tx_uninit(&is->rdft); + av_freep(&is->real_data); av_freep(&is->rdft_data); is->rdft = NULL; is->rdft_bits = 0;

1 0

wmavoice: convert DCT-I/DST-I to lavu/tx
by Lynne 02 Sep '23

02 Sep '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Fri Aug 4 21:16:30 2023 +0200| [a810126501e1ef0992d765720ff0d2629c5d1616] | committer: Lynne wmavoice: convert DCT-I/DST-I to lavu/tx This is the very last user of any lavc transform code. This also *corrects* wmavoice decoding, as the previous DCT/DST transforms were incorrect, bringing it closer to Microsoft's own wmavoice decoder. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a810126501e1ef099… --- libavcodec/wmavoice.c | 29 +++++++++++++++++------------ tests/fate/wma.mak | 9 ++++++--- 2 files changed, 23 insertions(+), 15 deletions(-) diff --git a/libavcodec/wmavoice.c b/libavcodec/wmavoice.c index 5ae92e2dbc..915315cb8a 100644 --- a/libavcodec/wmavoice.c +++ b/libavcodec/wmavoice.c @@ -42,8 +42,6 @@ #include "acelp_vectors.h" #include "acelp_filters.h" #include "lsp.h" -#include "dct.h" -#include "rdft.h" #include "sinewin.h" #define MAX_BLOCKS 8 ///< maximum number of blocks per frame @@ -266,8 +264,8 @@ typedef struct WMAVoiceContext { */ AVTXContext *rdft, *irdft; ///< contexts for FFT-calculation in the av_tx_fn rdft_fn, irdft_fn; ///< postfilter (for denoise filter) - DCTContext dct, dst; ///< contexts for phase shift (in Hilbert - ///< transform, part of postfilter) + AVTXContext *dct, *dst; ///< contexts for phase shift (in Hilbert + av_tx_fn dct_fn, dst_fn; ///< transform, part of postfilter) float sin[511], cos[511]; ///< 8-bit cosine/sine windows over [-pi,pi] ///< range float postfilter_agc; ///< gain control memory, used in @@ -391,10 +389,6 @@ static av_cold int wmavoice_decode_init(AVCodecContext *ctx) if (s->do_apf) { float scale = 1.0f; - if ((ret = ff_dct_init (&s->dct, 6, DCT_I)) < 0 || - (ret = ff_dct_init (&s->dst, 6, DST_I)) < 0) - return ret; - ret = av_tx_init(&s->rdft, &s->rdft_fn, AV_TX_FLOAT_RDFT, 0, 1 << 7, &scale, 0); if (ret < 0) return ret; @@ -403,6 +397,16 @@ static av_cold int wmavoice_decode_init(AVCodecContext *ctx) if (ret < 0) return ret; + scale = 1.0 / (1 << 6); + ret = av_tx_init(&s->dct, &s->dct_fn, AV_TX_FLOAT_DCT_I, 0, 1 << 6, &scale, 0); + if (ret < 0) + return ret; + + scale = 1.0 / (1 << 6); + ret = av_tx_init(&s->dst, &s->dst_fn, AV_TX_FLOAT_DST_I, 0, 1 << 6, &scale, 0); + if (ret < 0) + return ret; + ff_sine_window_init(s->cos, 256); memcpy(&s->sin[255], s->cos, 256 * sizeof(s->cos[0])); for (n = 0; n < 255; n++) { @@ -612,6 +616,7 @@ static void calc_input_response(WMAVoiceContext *s, float *lpcs_src, float irange, angle_mul, gain_mul, range, sq; LOCAL_ALIGNED_32(float, coeffs, [0x82]); LOCAL_ALIGNED_32(float, lpcs, [0x82]); + LOCAL_ALIGNED_32(float, lpcs_dct, [0x82]); int n, idx; memcpy(coeffs, coeffs_dst, 0x82*sizeof(float)); @@ -662,8 +667,8 @@ static void calc_input_response(WMAVoiceContext *s, float *lpcs_src, * is a sine input) by doing a phase shift (in theory, H(sin())=cos()). * Hilbert_Transform(RDFT(x)) = Laplace_Transform(x), which calculates the * "moment" of the LPCs in this filter. */ - s->dct.dct_calc(&s->dct, lpcs); - s->dst.dct_calc(&s->dst, lpcs); + s->dct_fn(s->dct, lpcs_dct, lpcs, sizeof(float)); + s->dst_fn(s->dst, lpcs, lpcs_dct, sizeof(float)); /* Split out the coefficient indexes into phase/magnitude pairs */ idx = 255 + av_clip(lpcs[64], -255, 255); @@ -2003,8 +2008,8 @@ static av_cold int wmavoice_decode_end(AVCodecContext *ctx) if (s->do_apf) { av_tx_uninit(&s->rdft); av_tx_uninit(&s->irdft); - ff_dct_end(&s->dct); - ff_dct_end(&s->dst); + av_tx_uninit(&s->dct); + av_tx_uninit(&s->dst); } return 0; diff --git a/tests/fate/wma.mak b/tests/fate/wma.mak index c13874ebfc..308dced9d6 100644 --- a/tests/fate/wma.mak +++ b/tests/fate/wma.mak @@ -20,17 +20,20 @@ fate-wmapro: $(FATE_WMAPRO-yes) FATE_WMAVOICE-$(call DEMDEC, ASF, WMAVOICE) += fate-wmavoice-7k fate-wmavoice-7k: CMD = pcm -i $(TARGET_SAMPLES)/wmavoice/streaming_CBR-7K.wma -fate-wmavoice-7k: REF = $(SAMPLES)/wmavoice/streaming_CBR-7K.pcm +fate-wmavoice-7k: REF = $(SAMPLES)/wmavoice/streaming_CBR-7K_ref.pcm +fate-wmavoice-7k: CMP_TARGET = 1368.61 fate-wmavoice-7k: FUZZ = 3 FATE_WMAVOICE-$(call DEMDEC, ASF, WMAVOICE) += fate-wmavoice-11k fate-wmavoice-11k: CMD = pcm -i $(TARGET_SAMPLES)/wmavoice/streaming_CBR-11K.wma -fate-wmavoice-11k: REF = $(SAMPLES)/wmavoice/streaming_CBR-11K.pcm +fate-wmavoice-11k: REF = $(SAMPLES)/wmavoice/streaming_CBR-11K_ref.pcm +fate-wmavoice-11k: CMP_TARGET = 965.24 fate-wmavoice-11k: FUZZ = 3 FATE_WMAVOICE-$(call DEMDEC, ASF, WMAVOICE) += fate-wmavoice-19k fate-wmavoice-19k: CMD = pcm -i $(TARGET_SAMPLES)/wmavoice/streaming_CBR-19K.wma -fate-wmavoice-19k: REF = $(SAMPLES)/wmavoice/streaming_CBR-19K.pcm +fate-wmavoice-19k: REF = $(SAMPLES)/wmavoice/streaming_CBR-19K_ref.pcm +fate-wmavoice-19k: CMP_TARGET = 689.33 fate-wmavoice-19k: FUZZ = 3 $(FATE_WMAVOICE-yes): CMP = stddev

1 0

wmavoice: convert RDFT to lavu/tx
by Lynne 02 Sep '23

02 Sep '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Fri Aug 4 20:20:10 2023 +0200| [d895d3c8c79e301f9d6f3aab0cc754ac2b7d78fb] | committer: Lynne wmavoice: convert RDFT to lavu/tx > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d895d3c8c79e301f9… --- libavcodec/wmavoice.c | 75 ++++++++++++++++++++++++++++++--------------------- 1 file changed, 45 insertions(+), 30 deletions(-) diff --git a/libavcodec/wmavoice.c b/libavcodec/wmavoice.c index 44fda0e2d6..5ae92e2dbc 100644 --- a/libavcodec/wmavoice.c +++ b/libavcodec/wmavoice.c @@ -31,6 +31,7 @@ #include "libavutil/float_dsp.h" #include "libavutil/mem_internal.h" #include "libavutil/thread.h" +#include "libavutil/tx.h" #include "avcodec.h" #include "codec_internal.h" #include "decode.h" @@ -263,8 +264,8 @@ typedef struct WMAVoiceContext { * smoothing and so on, and context variables for FFT/iFFT. * @{ */ - RDFTContext rdft, irdft; ///< contexts for FFT-calculation in the - ///< postfilter (for denoise filter) + AVTXContext *rdft, *irdft; ///< contexts for FFT-calculation in the + av_tx_fn rdft_fn, irdft_fn; ///< postfilter (for denoise filter) DCTContext dct, dst; ///< contexts for phase shift (in Hilbert ///< transform, part of postfilter) float sin[511], cos[511]; ///< 8-bit cosine/sine windows over [-pi,pi] @@ -277,9 +278,9 @@ typedef struct WMAVoiceContext { ///< by postfilter float denoise_filter_cache[MAX_FRAMESIZE]; int denoise_filter_cache_size; ///< samples in #denoise_filter_cache - DECLARE_ALIGNED(32, float, tilted_lpcs_pf)[0x80]; + DECLARE_ALIGNED(32, float, tilted_lpcs_pf)[0x82]; ///< aligned buffer for LPC tilting - DECLARE_ALIGNED(32, float, denoise_coeffs_pf)[0x80]; + DECLARE_ALIGNED(32, float, denoise_coeffs_pf)[0x82]; ///< aligned buffer for denoise coefficients DECLARE_ALIGNED(32, float, synth_filter_out_buf)[0x80 + MAX_LSPS_ALIGN16]; ///< aligned buffer for postfilter speech @@ -388,12 +389,20 @@ static av_cold int wmavoice_decode_init(AVCodecContext *ctx) s->spillover_bitsize = 3 + av_ceil_log2(ctx->block_align); s->do_apf = flags & 0x1; if (s->do_apf) { - if ((ret = ff_rdft_init(&s->rdft, 7, DFT_R2C)) < 0 || - (ret = ff_rdft_init(&s->irdft, 7, IDFT_C2R)) < 0 || - (ret = ff_dct_init (&s->dct, 6, DCT_I)) < 0 || + float scale = 1.0f; + + if ((ret = ff_dct_init (&s->dct, 6, DCT_I)) < 0 || (ret = ff_dct_init (&s->dst, 6, DST_I)) < 0) return ret; + ret = av_tx_init(&s->rdft, &s->rdft_fn, AV_TX_FLOAT_RDFT, 0, 1 << 7, &scale, 0); + if (ret < 0) + return ret; + + ret = av_tx_init(&s->irdft, &s->irdft_fn, AV_TX_FLOAT_RDFT, 1, 1 << 7, &scale, 0); + if (ret < 0) + return ret; + ff_sine_window_init(s->cos, 256); memcpy(&s->sin[255], s->cos, 256 * sizeof(s->cos[0])); for (n = 0; n < 255; n++) { @@ -596,20 +605,24 @@ static float tilt_factor(const float *lpcs, int n_lpcs) /** * Derive denoise filter coefficients (in real domain) from the LPCs. */ -static void calc_input_response(WMAVoiceContext *s, float *lpcs, - int fcb_type, float *coeffs, int remainder) +static void calc_input_response(WMAVoiceContext *s, float *lpcs_src, + int fcb_type, float *coeffs_dst, int remainder) { float last_coeff, min = 15.0, max = -15.0; float irange, angle_mul, gain_mul, range, sq; + LOCAL_ALIGNED_32(float, coeffs, [0x82]); + LOCAL_ALIGNED_32(float, lpcs, [0x82]); int n, idx; + memcpy(coeffs, coeffs_dst, 0x82*sizeof(float)); + /* Create frequency power spectrum of speech input (i.e. RDFT of LPCs) */ - s->rdft.rdft_calc(&s->rdft, lpcs); + s->rdft_fn(s->rdft, lpcs, lpcs_src, sizeof(float)); #define log_range(var, assign) do { \ float tmp = log10f(assign); var = tmp; \ max = FFMAX(max, tmp); min = FFMIN(min, tmp); \ } while (0) - log_range(last_coeff, lpcs[1] * lpcs[1]); + log_range(last_coeff, lpcs[64] * lpcs[64]); for (n = 1; n < 64; n++) log_range(lpcs[n], lpcs[n * 2] * lpcs[n * 2] + lpcs[n * 2 + 1] * lpcs[n * 2 + 1]); @@ -668,25 +681,25 @@ static void calc_input_response(WMAVoiceContext *s, float *lpcs, coeffs[n * 2 + 1] = coeffs[n] * s->sin[idx]; coeffs[n * 2] = coeffs[n] * s->cos[idx]; } - coeffs[1] = last_coeff; + coeffs[64] = last_coeff; /* move into real domain */ - s->irdft.rdft_calc(&s->irdft, coeffs); + s->irdft_fn(s->irdft, coeffs_dst, coeffs, sizeof(AVComplexFloat)); /* tilt correction and normalize scale */ - memset(&coeffs[remainder], 0, sizeof(coeffs[0]) * (128 - remainder)); + memset(&coeffs_dst[remainder], 0, sizeof(coeffs_dst[0]) * (128 - remainder)); if (s->denoise_tilt_corr) { float tilt_mem = 0; - coeffs[remainder - 1] = 0; + coeffs_dst[remainder - 1] = 0; ff_tilt_compensation(&tilt_mem, - -1.8 * tilt_factor(coeffs, remainder - 1), - coeffs, remainder); + -1.8 * tilt_factor(coeffs_dst, remainder - 1), + coeffs_dst, remainder); } - sq = (1.0 / 64.0) * sqrtf(1 / avpriv_scalarproduct_float_c(coeffs, coeffs, + sq = (1.0 / 64.0) * sqrtf(1 / avpriv_scalarproduct_float_c(coeffs_dst, coeffs_dst, remainder)); for (n = 0; n < remainder; n++) - coeffs[n] *= sq; + coeffs_dst[n] *= sq; } /** @@ -722,6 +735,8 @@ static void wiener_denoise(WMAVoiceContext *s, int fcb_type, int remainder, lim, n; if (fcb_type != FCB_TYPE_SILENCE) { + LOCAL_ALIGNED_32(float, coeffs_f, [0x82]); + LOCAL_ALIGNED_32(float, synth_f, [0x82]); float *tilted_lpcs = s->tilted_lpcs_pf, *coeffs = s->denoise_coeffs_pf, tilt_mem = 0; @@ -742,16 +757,16 @@ static void wiener_denoise(WMAVoiceContext *s, int fcb_type, /* apply coefficients (in frequency spectrum domain), i.e. complex * number multiplication */ memset(&synth_pf[size], 0, sizeof(synth_pf[0]) * (128 - size)); - s->rdft.rdft_calc(&s->rdft, synth_pf); - s->rdft.rdft_calc(&s->rdft, coeffs); - synth_pf[0] *= coeffs[0]; - synth_pf[1] *= coeffs[1]; - for (n = 1; n < 64; n++) { - float v1 = synth_pf[n * 2], v2 = synth_pf[n * 2 + 1]; - synth_pf[n * 2] = v1 * coeffs[n * 2] - v2 * coeffs[n * 2 + 1]; - synth_pf[n * 2 + 1] = v2 * coeffs[n * 2] + v1 * coeffs[n * 2 + 1]; + s->rdft_fn(s->rdft, synth_f, synth_pf, sizeof(float)); + s->rdft_fn(s->rdft, coeffs_f, coeffs, sizeof(float)); + synth_f[0] *= coeffs_f[0]; + synth_f[1] *= coeffs_f[1]; + for (n = 1; n <= 64; n++) { + float v1 = synth_f[n * 2], v2 = synth_f[n * 2 + 1]; + synth_f[n * 2] = v1 * coeffs_f[n * 2] - v2 * coeffs_f[n * 2 + 1]; + synth_f[n * 2 + 1] = v2 * coeffs_f[n * 2] + v1 * coeffs_f[n * 2 + 1]; } - s->irdft.rdft_calc(&s->irdft, synth_pf); + s->irdft_fn(s->irdft, synth_pf, synth_f, sizeof(AVComplexFloat)); } /* merge filter output with the history of previous runs */ @@ -1986,8 +2001,8 @@ static av_cold int wmavoice_decode_end(AVCodecContext *ctx) WMAVoiceContext *s = ctx->priv_data; if (s->do_apf) { - ff_rdft_end(&s->rdft); - ff_rdft_end(&s->irdft); + av_tx_uninit(&s->rdft); + av_tx_uninit(&s->irdft); ff_dct_end(&s->dct); ff_dct_end(&s->dst); }

1 0

lavu/tx: add DCT-I and DST-I transforms
by Lynne 02 Sep '23

02 Sep '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Thu Aug 3 18:23:02 2023 +0200| [ef8fd7bc3c479ec00b3e1f490d44f9c93894d210] | committer: Lynne lavu/tx: add DCT-I and DST-I transforms These are true, actual DCT-I and DST-I transforms, unlike the libavcodec versions, which are plainly not. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ef8fd7bc3c479ec00… --- libavutil/tx.h | 24 +++++++++++ libavutil/tx_template.c | 103 ++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 127 insertions(+) diff --git a/libavutil/tx.h b/libavutil/tx.h index d178e8ee9d..4696988cae 100644 --- a/libavutil/tx.h +++ b/libavutil/tx.h @@ -105,6 +105,30 @@ enum AVTXType { AV_TX_DOUBLE_DCT = 10, AV_TX_INT32_DCT = 11, + /** + * Discrete Cosine Transform I + * + * The forward transform is a DCT-I. + * The inverse transform is a DCT-I multiplied by 2/(N + 1). + * + * The input array is always overwritten. + */ + AV_TX_FLOAT_DCT_I = 12, + AV_TX_DOUBLE_DCT_I = 13, + AV_TX_INT32_DCT_I = 14, + + /** + * Discrete Sine Transform I + * + * The forward transform is a DST-I. + * The inverse transform is a DST-I multiplied by 2/(N + 1). + * + * The input array is always overwritten. + */ + AV_TX_FLOAT_DST_I = 15, + AV_TX_DOUBLE_DST_I = 16, + AV_TX_INT32_DST_I = 17, + /* Not part of the API, do not use */ AV_TX_NB, }; diff --git a/libavutil/tx_template.c b/libavutil/tx_template.c index c56dcf0826..c026cb40c4 100644 --- a/libavutil/tx_template.c +++ b/libavutil/tx_template.c @@ -2004,6 +2004,107 @@ static const FFTXCodelet TX_NAME(ff_tx_dctIII_def) = { .prio = FF_TX_PRIO_BASE, }; +static av_cold int TX_NAME(ff_tx_dcstI_init)(AVTXContext *s, + const FFTXCodelet *cd, + uint64_t flags, + FFTXCodeletOptions *opts, + int len, int inv, + const void *scale) +{ + int ret; + SCALE_TYPE rsc = *((SCALE_TYPE *)scale); + + if (inv) { + len *= 2; + s->len *= 2; + rsc *= 0.5; + } + + /* We want a half-complex RDFT */ + flags |= cd->type == TX_TYPE(DCT_I) ? AV_TX_REAL_TO_REAL : + AV_TX_REAL_TO_IMAGINARY; + + if ((ret = ff_tx_init_subtx(s, TX_TYPE(RDFT), flags, NULL, + (len - 1 + 2*(cd->type == TX_TYPE(DST_I)))*2, + 0, &rsc))) + return ret; + + s->tmp = av_mallocz((len + 1)*2*sizeof(TXSample)); + if (!s->tmp) + return AVERROR(ENOMEM); + + return 0; +} + +static void TX_NAME(ff_tx_dctI)(AVTXContext *s, void *_dst, + void *_src, ptrdiff_t stride) +{ + TXSample *dst = _dst; + TXSample *src = _src; + const int len = s->len - 1; + TXSample *tmp = (TXSample *)s->tmp; + + stride /= sizeof(TXSample); + + for (int i = 0; i < len; i++) + tmp[i] = tmp[2*len - i] = src[i * stride]; + + tmp[len] = src[len * stride]; /* Middle */ + + s->fn[0](&s->sub[0], dst, tmp, sizeof(TXSample)); +} + +static void TX_NAME(ff_tx_dstI)(AVTXContext *s, void *_dst, + void *_src, ptrdiff_t stride) +{ + TXSample *dst = _dst; + TXSample *src = _src; + const int len = s->len + 1; + TXSample *tmp = (void *)s->tmp; + + stride /= sizeof(TXSample); + + tmp[0] = 0; + + for (int i = 1; i < len; i++) { + TXSample a = src[(i - 1) * stride]; + tmp[i] = -a; + tmp[2*len - i] = a; + } + + tmp[len] = 0; /* i == n, Nyquist */ + + s->fn[0](&s->sub[0], dst, tmp, sizeof(float)); +} + +static const FFTXCodelet TX_NAME(ff_tx_dctI_def) = { + .name = TX_NAME_STR("dctI"), + .function = TX_NAME(ff_tx_dctI), + .type = TX_TYPE(DCT_I), + .flags = AV_TX_UNALIGNED | AV_TX_INPLACE | FF_TX_OUT_OF_PLACE, + .factors = { 2, TX_FACTOR_ANY }, + .nb_factors = 2, + .min_len = 2, + .max_len = TX_LEN_UNLIMITED, + .init = TX_NAME(ff_tx_dcstI_init), + .cpu_flags = FF_TX_CPU_FLAGS_ALL, + .prio = FF_TX_PRIO_BASE, +}; + +static const FFTXCodelet TX_NAME(ff_tx_dstI_def) = { + .name = TX_NAME_STR("dstI"), + .function = TX_NAME(ff_tx_dstI), + .type = TX_TYPE(DST_I), + .flags = AV_TX_UNALIGNED | AV_TX_INPLACE | FF_TX_OUT_OF_PLACE, + .factors = { 2, TX_FACTOR_ANY }, + .nb_factors = 2, + .min_len = 2, + .max_len = TX_LEN_UNLIMITED, + .init = TX_NAME(ff_tx_dcstI_init), + .cpu_flags = FF_TX_CPU_FLAGS_ALL, + .prio = FF_TX_PRIO_BASE, +}; + int TX_TAB(ff_tx_mdct_gen_exp)(AVTXContext *s, int *pre_tab) { int off = 0; @@ -2101,6 +2202,8 @@ const FFTXCodelet * const TX_NAME(ff_tx_codelet_list)[] = { &TX_NAME(ff_tx_rdft_c2r_def), &TX_NAME(ff_tx_dctII_def), &TX_NAME(ff_tx_dctIII_def), + &TX_NAME(ff_tx_dctI_def), + &TX_NAME(ff_tx_dstI_def), NULL, };

1 0

lavu/tx: add real to real and real to imaginary RDFT transforms
by Lynne 02 Sep '23

02 Sep '23

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Thu Aug 3 18:21:23 2023 +0200| [11e22730e1eb7ec6b5953c16b3b4df58be14e2de] | committer: Lynne lavu/tx: add real to real and real to imaginary RDFT transforms These are in-place transforms, required for DCT-I and DST-I. Templated as the mod2 variant requires minor modifications, and is required specifically for DCT-I/DST-I. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=11e22730e1eb7ec6b… --- doc/APIchanges | 3 + libavutil/tx.c | 18 ++++- libavutil/tx.h | 10 +++ libavutil/tx_template.c | 175 ++++++++++++++++++++++++++++++++++++++---------- libavutil/version.h | 2 +- 5 files changed, 167 insertions(+), 41 deletions(-) diff --git a/doc/APIchanges b/doc/APIchanges index ad1efe708d..db3242b667 100644 --- a/doc/APIchanges +++ b/doc/APIchanges @@ -2,6 +2,9 @@ The last version increases of all libraries were on 2023-02-09 API changes, most recent first: +2023-07-xx - xxxxxxxxxx - lavu 58.18.100 - tx.h + Add AV_TX_REAL_TO_REAL and AV_TX_REAL_TO_IMAGINARY + 2023-08-18 - xxxxxxxxxx - lavu 58.17.100 - channel_layout.h All AV_CHANNEL_LAYOUT_* macros are now compatible with C++ 17 and older. diff --git a/libavutil/tx.c b/libavutil/tx.c index e25abf998f..e9826e6107 100644 --- a/libavutil/tx.c +++ b/libavutil/tx.c @@ -437,7 +437,9 @@ int ff_tx_decompose_length(int dst[TX_MAX_DECOMPOSITIONS], enum AVTXType type, /* Check direction for non-orthogonal codelets */ if (((cd->flags & FF_TX_FORWARD_ONLY) && inv) || - ((cd->flags & (FF_TX_INVERSE_ONLY | AV_TX_FULL_IMDCT)) && !inv)) + ((cd->flags & (FF_TX_INVERSE_ONLY | AV_TX_FULL_IMDCT)) && !inv) || + ((cd->flags & (FF_TX_FORWARD_ONLY | AV_TX_REAL_TO_REAL)) && inv) || + ((cd->flags & (FF_TX_FORWARD_ONLY | AV_TX_REAL_TO_IMAGINARY)) && inv)) continue; /* Check if the CPU supports the required ISA */ @@ -560,6 +562,10 @@ static void print_flags(AVBPrint *bp, uint64_t f) av_bprintf(bp, "%spreshuf", prev > 1 ? sep : ""); if ((f & AV_TX_FULL_IMDCT) && ++prev) av_bprintf(bp, "%simdct_full", prev > 1 ? sep : ""); + if ((f & AV_TX_REAL_TO_REAL) && ++prev) + av_bprintf(bp, "%sreal_to_real", prev > 1 ? sep : ""); + if ((f & AV_TX_REAL_TO_IMAGINARY) && ++prev) + av_bprintf(bp, "%sreal_to_imaginary", prev > 1 ? sep : ""); if ((f & FF_TX_ASM_CALL) && ++prev) av_bprintf(bp, "%sasm_call", prev > 1 ? sep : ""); av_bprintf(bp, "]"); @@ -717,7 +723,11 @@ av_cold int ff_tx_init_subtx(AVTXContext *s, enum AVTXType type, uint64_t req_flags = flags; /* Flags the codelet may require to be present */ - uint64_t inv_req_mask = AV_TX_FULL_IMDCT | FF_TX_PRESHUFFLE | FF_TX_ASM_CALL; + uint64_t inv_req_mask = AV_TX_FULL_IMDCT | + AV_TX_REAL_TO_REAL | + AV_TX_REAL_TO_IMAGINARY | + FF_TX_PRESHUFFLE | + FF_TX_ASM_CALL; /* Unaligned codelets are compatible with the aligned flag */ if (req_flags & FF_TX_ALIGNED) @@ -742,7 +752,9 @@ av_cold int ff_tx_init_subtx(AVTXContext *s, enum AVTXType type, /* Check direction for non-orthogonal codelets */ if (((cd->flags & FF_TX_FORWARD_ONLY) && inv) || - ((cd->flags & (FF_TX_INVERSE_ONLY | AV_TX_FULL_IMDCT)) && !inv)) + ((cd->flags & (FF_TX_INVERSE_ONLY | AV_TX_FULL_IMDCT)) && !inv) || + ((cd->flags & (FF_TX_FORWARD_ONLY | AV_TX_REAL_TO_REAL)) && inv) || + ((cd->flags & (FF_TX_FORWARD_ONLY | AV_TX_REAL_TO_IMAGINARY)) && inv)) continue; /* Check if the requested flags match from both sides */ diff --git a/libavutil/tx.h b/libavutil/tx.h index 064edbc097..d178e8ee9d 100644 --- a/libavutil/tx.h +++ b/libavutil/tx.h @@ -149,6 +149,16 @@ enum AVTXFlags { * Ignored for all transforms but inverse MDCTs. */ AV_TX_FULL_IMDCT = 1ULL << 2, + + /** + * Perform a real to half-complex RDFT. + * Only the real, or imaginary coefficients will + * be output, depending on the flag used. Only available for forward RDFTs. + * Output array must have enough space to hold N complex values + * (regular size for a real to complex transform). + */ + AV_TX_REAL_TO_REAL = 1ULL << 3, + AV_TX_REAL_TO_IMAGINARY = 1ULL << 4, }; /** diff --git a/libavutil/tx_template.c b/libavutil/tx_template.c index c4ec9502e0..c56dcf0826 100644 --- a/libavutil/tx_template.c +++ b/libavutil/tx_template.c @@ -1613,14 +1613,17 @@ static av_cold int TX_NAME(ff_tx_rdft_init)(AVTXContext *s, int ret; double f, m; TXSample *tab; + int len4 = FFALIGN(len, 4) / 4; s->scale_d = *((SCALE_TYPE *)scale); s->scale_f = s->scale_d; + flags &= ~(AV_TX_REAL_TO_REAL | AV_TX_REAL_TO_IMAGINARY); + if ((ret = ff_tx_init_subtx(s, TX_TYPE(FFT), flags, NULL, len >> 1, inv, scale))) return ret; - if (!(s->exp = av_mallocz((8 + (len >> 2) - 1)*sizeof(*s->exp)))) + if (!(s->exp = av_mallocz((8 + 2*len4)*sizeof(*s->exp)))) return AVERROR(ENOMEM); tab = (TXSample *)s->exp; @@ -1639,17 +1642,20 @@ static av_cold int TX_NAME(ff_tx_rdft_init)(AVTXContext *s, *tab++ = RESCALE( (0.5 - inv) * m); *tab++ = RESCALE(-(0.5 - inv) * m); - for (int i = 0; i < len >> 2; i++) + for (int i = 0; i < len4; i++) *tab++ = RESCALE(cos(i*f)); - for (int i = len >> 2; i >= 0; i--) - *tab++ = RESCALE(cos(i*f) * (inv ? +1.0 : -1.0)); + + tab = ((TXSample *)s->exp) + len4 + 8; + + for (int i = 0; i < len4; i++) + *tab++ = RESCALE(cos(((float)len/4.0 - (float)i + 0)*f) * (inv ? +1.0 : -1.0)); return 0; } -#define DECL_RDFT(name, inv) \ -static void TX_NAME(ff_tx_rdft_ ##name)(AVTXContext *s, void *_dst, \ - void *_src, ptrdiff_t stride) \ +#define DECL_RDFT(n, inv) \ +static void TX_NAME(ff_tx_rdft_ ##n)(AVTXContext *s, void *_dst, \ + void *_src, ptrdiff_t stride) \ { \ const int len2 = s->len >> 1; \ const int len4 = s->len >> 2; \ @@ -1698,40 +1704,131 @@ static void TX_NAME(ff_tx_rdft_ ##name)(AVTXContext *s, void *_dst, \ data[len2].re = data[0].im; \ data[ 0].im = data[len2].im = 0; \ } \ -} +} \ + \ +static const FFTXCodelet TX_NAME(ff_tx_rdft_ ##n## _def) = { \ + .name = TX_NAME_STR("rdft_" #n), \ + .function = TX_NAME(ff_tx_rdft_ ##n), \ + .type = TX_TYPE(RDFT), \ + .flags = AV_TX_UNALIGNED | AV_TX_INPLACE | FF_TX_OUT_OF_PLACE | \ + (inv ? FF_TX_INVERSE_ONLY : FF_TX_FORWARD_ONLY), \ + .factors = { 4, TX_FACTOR_ANY }, \ + .nb_factors = 2, \ + .min_len = 4, \ + .max_len = TX_LEN_UNLIMITED, \ + .init = TX_NAME(ff_tx_rdft_init), \ + .cpu_flags = FF_TX_CPU_FLAGS_ALL, \ + .prio = FF_TX_PRIO_BASE, \ +}; -DECL_RDFT(r2c, 0) -DECL_RDFT(c2r, 1) +DECL_RDFT(r2c, 0) +DECL_RDFT(c2r, 1) -static const FFTXCodelet TX_NAME(ff_tx_rdft_r2c_def) = { - .name = TX_NAME_STR("rdft_r2c"), - .function = TX_NAME(ff_tx_rdft_r2c), - .type = TX_TYPE(RDFT), - .flags = AV_TX_UNALIGNED | AV_TX_INPLACE | - FF_TX_OUT_OF_PLACE | FF_TX_FORWARD_ONLY, - .factors = { 2, TX_FACTOR_ANY }, - .nb_factors = 2, - .min_len = 2, - .max_len = TX_LEN_UNLIMITED, - .init = TX_NAME(ff_tx_rdft_init), - .cpu_flags = FF_TX_CPU_FLAGS_ALL, - .prio = FF_TX_PRIO_BASE, +#define DECL_RDFT_HALF(n, mode, mod2) \ +static void TX_NAME(ff_tx_rdft_ ##n)(AVTXContext *s, void *_dst, \ + void *_src, ptrdiff_t stride) \ +{ \ + const int len = s->len; \ + const int len2 = len >> 1; \ + const int len4 = len >> 2; \ + const int aligned_len4 = FFALIGN(len, 4)/4; \ + const TXSample *fact = (void *)s->exp; \ + const TXSample *tcos = fact + 8; \ + const TXSample *tsin = tcos + aligned_len4; \ + TXComplex *data = _dst; \ + TXSample *out = _dst; /* Half-complex is forward-only */ \ + TXSample tmp_dc; \ + av_unused TXSample tmp_mid; \ + TXSample tmp[4]; \ + TXComplex sf, sl; \ + \ + s->fn[0](&s->sub[0], _dst, _src, sizeof(TXComplex)); \ + \ + tmp_dc = data[0].re; \ + data[ 0].re = tmp_dc + data[0].im; \ + tmp_dc = tmp_dc - data[0].im; \ + \ + data[ 0].re = MULT(fact[0], data[ 0].re); \ + tmp_dc = MULT(fact[1], tmp_dc); \ + data[len4].re = MULT(fact[2], data[len4].re); \ + \ + if (!mod2) { \ + data[len4].im = MULT(fact[3], data[len4].im); \ + } else { \ + sf = data[len4]; \ + sl = data[len4 + 1]; \ + if (mode == AV_TX_REAL_TO_REAL) \ + tmp[0] = MULT(fact[4], (sf.re + sl.re)); \ + else \ + tmp[0] = MULT(fact[5], (sf.im - sl.im)); \ + tmp[1] = MULT(fact[6], (sf.im + sl.im)); \ + tmp[2] = MULT(fact[7], (sf.re - sl.re)); \ + \ + if (mode == AV_TX_REAL_TO_REAL) { \ + tmp[3] = tmp[1]*tcos[len4] - tmp[2]*tsin[len4]; \ + tmp_mid = (tmp[0] - tmp[3]); \ + } else { \ + tmp[3] = tmp[1]*tsin[len4] + tmp[2]*tcos[len4]; \ + tmp_mid = (tmp[0] + tmp[3]); \ + } \ + } \ + \ + /* NOTE: unrolling this breaks non-mod8 lengths */ \ + for (int i = 1; i <= len4; i++) { \ + TXSample tmp[4]; \ + TXComplex sf = data[i]; \ + TXComplex sl = data[len2 - i]; \ + \ + if (mode == AV_TX_REAL_TO_REAL) \ + tmp[0] = MULT(fact[4], (sf.re + sl.re)); \ + else \ + tmp[0] = MULT(fact[5], (sf.im - sl.im)); \ + \ + tmp[1] = MULT(fact[6], (sf.im + sl.im)); \ + tmp[2] = MULT(fact[7], (sf.re - sl.re)); \ + \ + if (mode == AV_TX_REAL_TO_REAL) { \ + tmp[3] = tmp[1]*tcos[i] - tmp[2]*tsin[i]; \ + out[i] = (tmp[0] + tmp[3]); \ + out[len - i] = (tmp[0] - tmp[3]); \ + } else { \ + tmp[3] = tmp[1]*tsin[i] + tmp[2]*tcos[i]; \ + out[i - 1] = (tmp[3] - tmp[0]); \ + out[len - i - 1] = (tmp[0] + tmp[3]); \ + } \ + } \ + \ + for (int i = 1; i < (len4 + (mode == AV_TX_REAL_TO_IMAGINARY)); i++) \ + out[len2 - i] = out[len - i]; \ + \ + if (mode == AV_TX_REAL_TO_REAL) { \ + out[len2] = tmp_dc; \ + if (mod2) \ + out[len4 + 1] = tmp_mid; \ + } else if (mod2) { \ + out[len4] = tmp_mid; \ + } \ +} \ + \ +static const FFTXCodelet TX_NAME(ff_tx_rdft_ ##n## _def) = { \ + .name = TX_NAME_STR("rdft_" #n), \ + .function = TX_NAME(ff_tx_rdft_ ##n), \ + .type = TX_TYPE(RDFT), \ + .flags = AV_TX_UNALIGNED | AV_TX_INPLACE | mode | \ + FF_TX_OUT_OF_PLACE | FF_TX_FORWARD_ONLY, \ + .factors = { 2 + 2*(!mod2), TX_FACTOR_ANY }, \ + .nb_factors = 2, \ + .min_len = 2 + 2*(!mod2), \ + .max_len = TX_LEN_UNLIMITED, \ + .init = TX_NAME(ff_tx_rdft_init), \ + .cpu_flags = FF_TX_CPU_FLAGS_ALL, \ + .prio = FF_TX_PRIO_BASE, \ }; -static const FFTXCodelet TX_NAME(ff_tx_rdft_c2r_def) = { - .name = TX_NAME_STR("rdft_c2r"), - .function = TX_NAME(ff_tx_rdft_c2r), - .type = TX_TYPE(RDFT), - .flags = AV_TX_UNALIGNED | AV_TX_INPLACE | - FF_TX_OUT_OF_PLACE | FF_TX_INVERSE_ONLY, - .factors = { 2, TX_FACTOR_ANY }, - .nb_factors = 2, - .min_len = 2, - .max_len = TX_LEN_UNLIMITED, - .init = TX_NAME(ff_tx_rdft_init), - .cpu_flags = FF_TX_CPU_FLAGS_ALL, - .prio = FF_TX_PRIO_BASE, -}; +DECL_RDFT_HALF(r2r, AV_TX_REAL_TO_REAL, 0) +DECL_RDFT_HALF(r2r_mod2, AV_TX_REAL_TO_REAL, 1) +DECL_RDFT_HALF(r2i, AV_TX_REAL_TO_IMAGINARY, 0) +DECL_RDFT_HALF(r2i_mod2, AV_TX_REAL_TO_IMAGINARY, 1) static av_cold int TX_NAME(ff_tx_dct_init)(AVTXContext *s, const FFTXCodelet *cd, @@ -1997,6 +2094,10 @@ const FFTXCodelet * const TX_NAME(ff_tx_codelet_list)[] = { &TX_NAME(ff_tx_mdct_naive_inv_def), &TX_NAME(ff_tx_mdct_inv_full_def), &TX_NAME(ff_tx_rdft_r2c_def), + &TX_NAME(ff_tx_rdft_r2r_def), + &TX_NAME(ff_tx_rdft_r2r_mod2_def), + &TX_NAME(ff_tx_rdft_r2i_def), + &TX_NAME(ff_tx_rdft_r2i_mod2_def), &TX_NAME(ff_tx_rdft_c2r_def), &TX_NAME(ff_tx_dctII_def), &TX_NAME(ff_tx_dctIII_def), diff --git a/libavutil/version.h b/libavutil/version.h index bc43baca9f..6d27f91cce 100644 --- a/libavutil/version.h +++ b/libavutil/version.h @@ -79,7 +79,7 @@ */ #define LIBAVUTIL_VERSION_MAJOR 58 -#define LIBAVUTIL_VERSION_MINOR 17 +#define LIBAVUTIL_VERSION_MINOR 18 #define LIBAVUTIL_VERSION_MICRO 100 #define LIBAVUTIL_VERSION_INT AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \

1 0

avformat: add OSQ demuxer
by Paul B Mahol 01 Sep '23

01 Sep '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Tue Jun 27 19:51:54 2023 +0200| [fba454617555524900c99a4a236fb3a3266d081d] | committer: Paul B Mahol avformat: add OSQ demuxer > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=fba45461755552490… --- Changelog | 1 + libavformat/Makefile | 1 + libavformat/allformats.c | 1 + libavformat/osq.c | 118 +++++++++++++++++++++++++++++++++++++++++++++++ libavformat/version.h | 4 +- 5 files changed, 123 insertions(+), 2 deletions(-) diff --git a/Changelog b/Changelog index c010e86159..6c2622dc4e 100644 --- a/Changelog +++ b/Changelog @@ -30,6 +30,7 @@ version <next>: - support for the P_SKIP hinting to speed up libx264 encoding - Support HEVC,VP9,AV1 codec in enhanced flv format - apsnr and asisdr audio filters +- OSQ demuxer and decoder version 6.0: diff --git a/libavformat/Makefile b/libavformat/Makefile index f9944baab9..cc1b12360a 100644 --- a/libavformat/Makefile +++ b/libavformat/Makefile @@ -433,6 +433,7 @@ OBJS-$(CONFIG_OMA_DEMUXER) += omadec.o pcm.o oma.o OBJS-$(CONFIG_OMA_MUXER) += omaenc.o rawenc.o oma.o id3v2enc.o OBJS-$(CONFIG_OPUS_MUXER) += oggenc.o \ vorbiscomment.o +OBJS-$(CONFIG_OSQ_DEMUXER) += osq.o rawdec.o OBJS-$(CONFIG_PAF_DEMUXER) += paf.o OBJS-$(CONFIG_PCM_ALAW_DEMUXER) += pcmdec.o pcm.o OBJS-$(CONFIG_PCM_ALAW_MUXER) += pcmenc.o rawenc.o diff --git a/libavformat/allformats.c b/libavformat/allformats.c index 6324952bd2..f4210e4932 100644 --- a/libavformat/allformats.c +++ b/libavformat/allformats.c @@ -331,6 +331,7 @@ extern const FFOutputFormat ff_ogv_muxer; extern const AVInputFormat ff_oma_demuxer; extern const FFOutputFormat ff_oma_muxer; extern const FFOutputFormat ff_opus_muxer; +extern const AVInputFormat ff_osq_demuxer; extern const AVInputFormat ff_paf_demuxer; extern const AVInputFormat ff_pcm_alaw_demuxer; extern const FFOutputFormat ff_pcm_alaw_muxer; diff --git a/libavformat/osq.c b/libavformat/osq.c new file mode 100644 index 0000000000..c5a63d3398 --- /dev/null +++ b/libavformat/osq.c @@ -0,0 +1,118 @@ +/* + * OSQ demuxer + * Copyright (c) 2023 Paul B Mahol + * + * 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 + */ + +#include "libavutil/intreadwrite.h" +#include "avio_internal.h" +#include "avformat.h" +#include "demux.h" +#include "internal.h" +#include "rawdec.h" + +static int osq_probe(const AVProbeData *p) +{ + if (AV_RL32(p->buf) != MKTAG('O','S','Q',' ')) + return 0; + if (AV_RL32(p->buf + 4) != 48) + return 0; + if (AV_RL16(p->buf + 8) != 1) + return 0; + if (p->buf[10] == 0) + return 0; + if (p->buf[11] == 0) + return 0; + if (AV_RL32(p->buf + 12) == 0) + return 0; + if (AV_RL16(p->buf + 16) == 0) + return 0; + + return AVPROBE_SCORE_MAX; +} + +static int osq_read_header(AVFormatContext *s) +{ + uint32_t t, size; + AVStream *st; + int ret; + + t = avio_rl32(s->pb); + if (t != MKTAG('O','S','Q',' ')) + return AVERROR_INVALIDDATA; + + size = avio_rl32(s->pb); + if (size != 48) + return AVERROR_INVALIDDATA; + + st = avformat_new_stream(s, NULL); + if (!st) + return AVERROR(ENOMEM); + if ((ret = ff_get_extradata(s, st->codecpar, s->pb, size)) < 0) + return ret; + + t = avio_rl32(s->pb); + if (t != MKTAG('R','I','F','F')) + return AVERROR_INVALIDDATA; + avio_skip(s->pb, 8); + + t = avio_rl32(s->pb); + if (t != MKTAG('f','m','t',' ')) + return AVERROR_INVALIDDATA; + size = avio_rl32(s->pb); + avio_skip(s->pb, size); + + t = avio_rl32(s->pb); + size = avio_rl32(s->pb); + while (t != MKTAG('d','a','t','a')) { + avio_skip(s->pb, size); + + t = avio_rl32(s->pb); + size = avio_rl32(s->pb); + if (avio_feof(s->pb)) + return AVERROR_INVALIDDATA; + } + + st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO; + st->codecpar->codec_id = AV_CODEC_ID_OSQ; + st->codecpar->sample_rate = AV_RL32(st->codecpar->extradata + 4); + if (st->codecpar->sample_rate <= 0) + return AVERROR_INVALIDDATA; + st->codecpar->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC; + st->codecpar->ch_layout.nb_channels = st->codecpar->extradata[3]; + if (st->codecpar->ch_layout.nb_channels == 0) + return AVERROR_INVALIDDATA; + st->start_time = 0; + st->duration = AV_RL32(st->codecpar->extradata + 16); + avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate); + + return 0; +} + +const AVInputFormat ff_osq_demuxer = { + .name = "osq", + .long_name = NULL_IF_CONFIG_SMALL("raw OSQ"), + .read_probe = osq_probe, + .read_header = osq_read_header, + .read_packet = ff_raw_read_partial_packet, + .extensions = "osq", + .flags = AVFMT_NOBINSEARCH | AVFMT_NOGENSEARCH | AVFMT_NO_BYTE_SEEK | AVFMT_NOTIMESTAMPS, + .raw_codec_id = AV_CODEC_ID_OSQ, + .priv_data_size = sizeof(FFRawDemuxerContext), + .priv_class = &ff_raw_demuxer_class, +}; diff --git a/libavformat/version.h b/libavformat/version.h index 787ee8c90b..1055753772 100644 --- a/libavformat/version.h +++ b/libavformat/version.h @@ -31,8 +31,8 @@ #include "version_major.h" -#define LIBAVFORMAT_VERSION_MINOR 10 -#define LIBAVFORMAT_VERSION_MICRO 101 +#define LIBAVFORMAT_VERSION_MINOR 11 +#define LIBAVFORMAT_VERSION_MICRO 100 #define LIBAVFORMAT_VERSION_INT AV_VERSION_INT(LIBAVFORMAT_VERSION_MAJOR, \ LIBAVFORMAT_VERSION_MINOR, \

1 0

avcodec: add OSQ audio decoder
by Paul B Mahol 01 Sep '23

01 Sep '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Tue Jun 27 19:54:25 2023 +0200| [7ef9d31071021c05e6b792af3f25b7b9ceaa9258] | committer: Paul B Mahol avcodec: add OSQ audio decoder > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=7ef9d31071021c05e… --- libavcodec/Makefile | 1 + libavcodec/allcodecs.c | 1 + libavcodec/codec_desc.c | 7 + libavcodec/codec_id.h | 1 + libavcodec/osq.c | 475 ++++++++++++++++++++++++++++++++++++++++++++++++ libavcodec/version.h | 2 +- 6 files changed, 486 insertions(+), 1 deletion(-) diff --git a/libavcodec/Makefile b/libavcodec/Makefile index f961d0abd6..08fd151619 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -579,6 +579,7 @@ OBJS-$(CONFIG_OPUS_DECODER) += opusdec.o opusdec_celt.o opus_celt.o \ opusdsp.o opus_parse.o opus_rc.o OBJS-$(CONFIG_OPUS_ENCODER) += opusenc.o opusenc_psy.o opus_celt.o \ opus_pvq.o opus_rc.o opustab.o +OBJS-$(CONFIG_OSQ_DECODER) += osq.o OBJS-$(CONFIG_PAF_AUDIO_DECODER) += pafaudio.o OBJS-$(CONFIG_PAF_VIDEO_DECODER) += pafvideo.o OBJS-$(CONFIG_PAM_DECODER) += pnmdec.o pnm.o diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index 8775d15a4f..6e95ca5636 100644 --- a/libavcodec/allcodecs.c +++ b/libavcodec/allcodecs.c @@ -516,6 +516,7 @@ extern const FFCodec ff_nellymoser_decoder; extern const FFCodec ff_on2avc_decoder; extern const FFCodec ff_opus_encoder; extern const FFCodec ff_opus_decoder; +extern const FFCodec ff_osq_decoder; extern const FFCodec ff_paf_audio_decoder; extern const FFCodec ff_qcelp_decoder; extern const FFCodec ff_qdm2_decoder; diff --git a/libavcodec/codec_desc.c b/libavcodec/codec_desc.c index 4406dd8318..f556bb94d5 100644 --- a/libavcodec/codec_desc.c +++ b/libavcodec/codec_desc.c @@ -3413,6 +3413,13 @@ static const AVCodecDescriptor codec_descriptors[] = { .long_name = NULL_IF_CONFIG_SMALL("AC-4"), .props = AV_CODEC_PROP_LOSSY, }, + { + .id = AV_CODEC_ID_OSQ, + .type = AVMEDIA_TYPE_AUDIO, + .name = "osq", + .long_name = NULL_IF_CONFIG_SMALL("OSQ (Original Sound Quality)"), + .props = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSLESS, + }, /* subtitle codecs */ { diff --git a/libavcodec/codec_id.h b/libavcodec/codec_id.h index a5a0cb8525..29b410b8d3 100644 --- a/libavcodec/codec_id.h +++ b/libavcodec/codec_id.h @@ -543,6 +543,7 @@ enum AVCodecID { AV_CODEC_ID_WAVARC, AV_CODEC_ID_RKA, AV_CODEC_ID_AC4, + AV_CODEC_ID_OSQ, /* subtitle codecs */ AV_CODEC_ID_FIRST_SUBTITLE = 0x17000, ///< A dummy ID pointing at the start of subtitle codecs. diff --git a/libavcodec/osq.c b/libavcodec/osq.c new file mode 100644 index 0000000000..47587cd812 --- /dev/null +++ b/libavcodec/osq.c @@ -0,0 +1,475 @@ +/* + * OSQ audio decoder + * Copyright (c) 2023 Paul B Mahol + * + * 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 + */ + +#include "libavutil/internal.h" +#include "libavutil/intreadwrite.h" +#include "avcodec.h" +#include "codec_internal.h" +#include "decode.h" +#include "internal.h" +#define BITSTREAM_READER_LE +#include "get_bits.h" +#include "unary.h" + +#define OFFSET 5 + +typedef struct OSQChannel { + unsigned prediction; + unsigned coding_mode; + unsigned residue_parameter; + unsigned residue_bits; + unsigned history[3]; + unsigned pos, count; + double sum; + int32_t prev; +} OSQChannel; + +typedef struct OSQContext { + GetBitContext gb; + OSQChannel ch[2]; + + uint8_t *bitstream; + size_t max_framesize; + size_t bitstream_size; + + int decorrelate; + int frame_samples; + int64_t nb_samples; + + int32_t *decode_buffer[2]; + + AVPacket *pkt; + int pkt_offset; +} OSQContext; + +static av_cold int osq_close(AVCodecContext *avctx) +{ + OSQContext *s = avctx->priv_data; + + av_freep(&s->bitstream); + s->bitstream_size = 0; + + for (int ch = 0; ch < FF_ARRAY_ELEMS(s->decode_buffer); ch++) + av_freep(&s->decode_buffer[ch]); + + return 0; +} + +static av_cold int osq_init(AVCodecContext *avctx) +{ + OSQContext *s = avctx->priv_data; + + if (avctx->extradata_size < 48) + return AVERROR(EINVAL); + + if (avctx->extradata[0] != 1) { + av_log(avctx, AV_LOG_ERROR, "Unsupported version.\n"); + return AVERROR_INVALIDDATA; + } + + avctx->sample_rate = AV_RL32(avctx->extradata + 4); + if (avctx->sample_rate < 1) + return AVERROR_INVALIDDATA; + + av_channel_layout_uninit(&avctx->ch_layout); + avctx->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC; + avctx->ch_layout.nb_channels = avctx->extradata[3]; + if (avctx->ch_layout.nb_channels < 1) + return AVERROR_INVALIDDATA; + if (avctx->ch_layout.nb_channels > FF_ARRAY_ELEMS(s->decode_buffer)) + return AVERROR_INVALIDDATA; + + switch (avctx->extradata[2]) { + case 8: avctx->sample_fmt = AV_SAMPLE_FMT_U8P; break; + case 16: avctx->sample_fmt = AV_SAMPLE_FMT_S16P; break; + case 20: + case 24: + case 28: + case 32: avctx->sample_fmt = AV_SAMPLE_FMT_S32P; break; + default: return AVERROR_INVALIDDATA; + } + + s->nb_samples = AV_RL64(avctx->extradata + 16); + s->frame_samples = AV_RL16(avctx->extradata + 8); + s->max_framesize = (s->frame_samples * 16 + 1024) * avctx->ch_layout.nb_channels; + + s->bitstream = av_calloc(s->max_framesize + AV_INPUT_BUFFER_PADDING_SIZE, sizeof(*s->bitstream)); + if (!s->bitstream) + return AVERROR(ENOMEM); + + for (int ch = 0; ch < avctx->ch_layout.nb_channels; ch++) { + s->decode_buffer[ch] = av_calloc(s->frame_samples + OFFSET, + sizeof(*s->decode_buffer[ch])); + if (!s->decode_buffer[ch]) + return AVERROR(ENOMEM); + } + + s->pkt = avctx->internal->in_pkt; + + return 0; +} + +static void reset_stats(OSQChannel *cb) +{ + memset(cb->history, 0, sizeof(cb->history)); + cb->pos = cb->count = cb->sum = 0; +} + +static void update_stats(OSQChannel *cb, int val) +{ + cb->sum += FFABS(val) - cb->history[cb->pos]; + cb->history[cb->pos] = FFABS(val); + cb->pos++; + cb->count++; + if (cb->pos >= FF_ARRAY_ELEMS(cb->history)) + cb->pos = 0; +} + +static int update_residue_parameter(OSQChannel *cb) +{ + double sum, x; + int rice_k; + + sum = cb->sum; + x = sum / cb->count; + rice_k = av_ceil_log2(x); + if (rice_k >= 30) { + rice_k = floor(sum / 1.4426952 + 0.5); + if (rice_k < 1) + rice_k = 1; + } + + return rice_k; +} + +static uint32_t get_urice(GetBitContext *gb, int k) +{ + uint32_t z, x, b; + + x = get_unary(gb, 1, 512); + b = get_bits_long(gb, k); + z = b | x << k; + + return z; +} + +static int32_t get_srice(GetBitContext *gb, int x) +{ + int32_t y = get_urice(gb, x); + return get_bits1(gb) ? -y : y; +} + +static int osq_channel_parameters(AVCodecContext *avctx, int ch) +{ + OSQContext *s = avctx->priv_data; + OSQChannel *cb = &s->ch[ch]; + GetBitContext *gb = &s->gb; + + cb->prev = 0; + cb->prediction = get_urice(gb, 5); + cb->coding_mode = get_urice(gb, 3); + if (cb->prediction >= 15) + return AVERROR_INVALIDDATA; + if (cb->coding_mode > 0 && cb->coding_mode < 3) { + cb->residue_parameter = get_urice(gb, 4); + if (!cb->residue_parameter || cb->residue_parameter >= 31) + return AVERROR_INVALIDDATA; + } else if (cb->coding_mode == 3) { + cb->residue_bits = get_urice(gb, 4); + if (!cb->residue_bits || cb->residue_bits >= 31) + return AVERROR_INVALIDDATA; + } else if (cb->coding_mode) { + return AVERROR_INVALIDDATA; + } + + if (cb->coding_mode == 2) + reset_stats(cb); + + return 0; +} + +#define A (-1) +#define B (-2) +#define C (-3) +#define D (-4) +#define E (-5) +#define P2 ((dst[A] + dst[A]) - dst[B]) +#define P3 ((dst[A] - dst[B]) * 3 + dst[C]) + +static int do_decode(AVCodecContext *avctx, AVFrame *frame, int decorrelate, int downsample) +{ + OSQContext *s = avctx->priv_data; + const int nb_channels = avctx->ch_layout.nb_channels; + const int nb_samples = frame->nb_samples; + GetBitContext *gb = &s->gb; + + for (int n = 0; n < nb_samples; n++) { + for (int ch = 0; ch < nb_channels; ch++) { + OSQChannel *cb = &s->ch[ch]; + int32_t *dst = s->decode_buffer[ch] + OFFSET; + int32_t p, prev = cb->prev; + + if (nb_channels == 2 && ch == 1 && decorrelate != s->decorrelate) { + if (!decorrelate) { + s->decode_buffer[1][OFFSET+A] += s->decode_buffer[0][OFFSET+B]; + s->decode_buffer[1][OFFSET+B] += s->decode_buffer[0][OFFSET+C]; + s->decode_buffer[1][OFFSET+C] += s->decode_buffer[0][OFFSET+D]; + s->decode_buffer[1][OFFSET+D] += s->decode_buffer[0][OFFSET+E]; + } else { + s->decode_buffer[1][OFFSET+A] -= s->decode_buffer[0][OFFSET+B]; + s->decode_buffer[1][OFFSET+B] -= s->decode_buffer[0][OFFSET+C]; + s->decode_buffer[1][OFFSET+C] -= s->decode_buffer[0][OFFSET+D]; + s->decode_buffer[1][OFFSET+D] -= s->decode_buffer[0][OFFSET+E]; + } + s->decorrelate = decorrelate; + } + + if (!cb->coding_mode) { + dst[n] = 0; + } else if (cb->coding_mode == 3) { + dst[n] = get_sbits_long(gb, cb->residue_bits); + } else { + dst[n] = get_srice(gb, cb->residue_parameter); + } + + if (get_bits_left(gb) < 0) { + av_log(avctx, AV_LOG_ERROR, "overread!\n"); + return AVERROR_INVALIDDATA; + } + + p = prev / 2; + prev = dst[n]; + + switch (cb->prediction) { + case 0: + break; + case 1: + dst[n] += dst[A]; + break; + case 2: + dst[n] += dst[A] + p; + break; + case 3: + dst[n] += P2; + break; + case 4: + dst[n] += P2 + p; + break; + case 5: + dst[n] += P3; + break; + case 6: + dst[n] += P3 + p; + break; + case 7: + dst[n] += (P2 + P3) / 2 + p; + break; + case 8: + dst[n] += (P2 + P3) / 2; + break; + case 9: + dst[n] += (P2 * 2 + P3) / 3 + p; + break; + case 10: + dst[n] += (P2 + P3 * 2) / 3 + p; + break; + case 11: + dst[n] += (dst[A] + dst[B]) / 2; + break; + case 12: + dst[n] += dst[B]; + break; + case 13: + dst[n] += (dst[D] + dst[B]) / 2; + break; + case 14: + dst[n] += (P2 + dst[A]) / 2 + p; + break; + default: + return AVERROR_INVALIDDATA; + } + + cb->prev = prev; + + if (downsample) + dst[n] *= 256; + + dst[E] = dst[D]; + dst[D] = dst[C]; + dst[C] = dst[B]; + dst[B] = dst[A]; + dst[A] = dst[n]; + + if (cb->coding_mode == 2) { + update_stats(cb, dst[n]); + cb->residue_parameter = update_residue_parameter(cb); + } + + if (nb_channels == 2 && ch == 1) { + if (decorrelate) + dst[n] += s->decode_buffer[0][OFFSET+n]; + } + + if (downsample) + dst[A] /= 256; + } + } + + return 0; +} + +static int osq_decode_block(AVCodecContext *avctx, AVFrame *frame) +{ + const int nb_channels = avctx->ch_layout.nb_channels; + OSQContext *s = avctx->priv_data; + int ret, decorrelate, downsample; + GetBitContext *gb = &s->gb; + + skip_bits1(gb); + decorrelate = get_bits1(gb); + downsample = get_bits1(gb); + + for (int ch = 0; ch < nb_channels; ch++) { + if ((ret = osq_channel_parameters(avctx, ch)) < 0) { + av_log(avctx, AV_LOG_ERROR, "invalid channel parameters\n"); + return ret; + } + } + + if ((ret = do_decode(avctx, frame, decorrelate, downsample)) < 0) + return ret; + + align_get_bits(gb); + + switch (avctx->sample_fmt) { + case AV_SAMPLE_FMT_U8P: + for (int ch = 0; ch < nb_channels; ch++) { + uint8_t *dst = (uint8_t *)frame->extended_data[ch]; + int32_t *src = s->decode_buffer[ch] + OFFSET; + + for (int n = 0; n < frame->nb_samples; n++) + dst[n] = av_clip_uint8(src[n] + 0x80); + } + break; + case AV_SAMPLE_FMT_S16P: + for (int ch = 0; ch < nb_channels; ch++) { + int16_t *dst = (int16_t *)frame->extended_data[ch]; + int32_t *src = s->decode_buffer[ch] + OFFSET; + + for (int n = 0; n < frame->nb_samples; n++) + dst[n] = (int16_t)src[n]; + } + break; + case AV_SAMPLE_FMT_S32P: + for (int ch = 0; ch < nb_channels; ch++) { + int32_t *dst = (int32_t *)frame->extended_data[ch]; + int32_t *src = s->decode_buffer[ch] + OFFSET; + + for (int n = 0; n < frame->nb_samples; n++) + dst[n] = src[n]; + } + break; + default: + return AVERROR_BUG; + } + + return 0; +} + +static int osq_receive_frame(AVCodecContext *avctx, AVFrame *frame) +{ + OSQContext *s = avctx->priv_data; + GetBitContext *gb = &s->gb; + int ret, n; + + while (s->bitstream_size < s->max_framesize) { + int size; + + if (!s->pkt->data) { + ret = ff_decode_get_packet(avctx, s->pkt); + if (ret == AVERROR_EOF && s->bitstream_size > 0) + break; + if (ret < 0) + return ret; + } + + size = FFMIN(s->pkt->size - s->pkt_offset, s->max_framesize - s->bitstream_size); + memcpy(s->bitstream + s->bitstream_size, s->pkt->data + s->pkt_offset, size); + s->bitstream_size += size; + s->pkt_offset += size; + + if (s->pkt_offset == s->pkt->size) { + av_packet_unref(s->pkt); + s->pkt_offset = 0; + } + } + + frame->nb_samples = FFMIN(s->frame_samples, s->nb_samples); + if (frame->nb_samples <= 0) + return AVERROR_EOF; + + if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) + goto fail; + + if ((ret = init_get_bits8(gb, s->bitstream, s->bitstream_size)) < 0) + goto fail; + + if ((ret = osq_decode_block(avctx, frame)) < 0) + goto fail; + + s->nb_samples -= frame->nb_samples; + + n = get_bits_count(gb) / 8; + if (n > s->bitstream_size) { + ret = AVERROR_INVALIDDATA; + goto fail; + } + + memmove(s->bitstream, &s->bitstream[n], s->bitstream_size - n); + s->bitstream_size -= n; + + return 0; + +fail: + s->bitstream_size = 0; + s->pkt_offset = 0; + av_packet_unref(s->pkt); + + return ret; +} + +const FFCodec ff_osq_decoder = { + .p.name = "osq", + CODEC_LONG_NAME("OSQ (Original Sound Quality)"), + .p.type = AVMEDIA_TYPE_AUDIO, + .p.id = AV_CODEC_ID_OSQ, + .priv_data_size = sizeof(OSQContext), + .init = osq_init, + FF_CODEC_RECEIVE_FRAME_CB(osq_receive_frame), + .close = osq_close, + .p.capabilities = AV_CODEC_CAP_CHANNEL_CONF | + AV_CODEC_CAP_DR1, + .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, + .p.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P, + AV_SAMPLE_FMT_S16P, + AV_SAMPLE_FMT_S32P, + AV_SAMPLE_FMT_NONE }, +}; diff --git a/libavcodec/version.h b/libavcodec/version.h index 728ab8839d..e0fe2eb7b8 100644 --- a/libavcodec/version.h +++ b/libavcodec/version.h @@ -29,7 +29,7 @@ #include "version_major.h" -#define LIBAVCODEC_VERSION_MINOR 23 +#define LIBAVCODEC_VERSION_MINOR 24 #define LIBAVCODEC_VERSION_MICRO 100 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \

1 0

fftools/ffmpeg_mux: replace monotonous with monotonic
by Leo Izen 01 Sep '23

01 Sep '23

ffmpeg | branch: master | Leo Izen <leo.izen(a)gmail.com> | Tue Aug 29 16:54:54 2023 -0400| [60be62d29388fdc681540dff387e0304e610bf39] | committer: Leo Izen fftools/ffmpeg_mux: replace monotonous with monotonic The word "monotonous" means "spoken in a monotone" which is not what we mean here. We mean "monotonic" i.e. nondecreasing. Signed-off-by: Leo Izen <leo.izen(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=60be62d29388fdc68… --- fftools/ffmpeg_mux.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fftools/ffmpeg_mux.c b/fftools/ffmpeg_mux.c index e74b5f09f5..7a924dba6c 100644 --- a/fftools/ffmpeg_mux.c +++ b/fftools/ffmpeg_mux.c @@ -115,7 +115,7 @@ static int write_packet(Muxer *mux, OutputStream *ost, AVPacket *pkt) int loglevel = max - pkt->dts > 2 || ost->type == AVMEDIA_TYPE_VIDEO ? AV_LOG_WARNING : AV_LOG_DEBUG; if (exit_on_error) loglevel = AV_LOG_ERROR; - av_log(s, loglevel, "Non-monotonous DTS in output stream " + av_log(s, loglevel, "Non-monotonic DTS in output stream " "%d:%d; previous: %"PRId64", current: %"PRId64"; ", ost->file_index, ost->st->index, ms->last_mux_dts, pkt->dts); if (exit_on_error) {

1 0