From patchwork Mon May 8 22:00:23 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Paul B Mahol X-Patchwork-Id: 3623 Delivered-To: ffmpegpatchwork@gmail.com Received: by 10.103.3.129 with SMTP id 123csp1048777vsd; Mon, 8 May 2017 15:00:46 -0700 (PDT) X-Received: by 10.223.157.29 with SMTP id k29mr43449804wre.156.1494280846449; Mon, 08 May 2017 15:00:46 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1494280846; cv=none; d=google.com; s=arc-20160816; b=zKRwIgwDmWOD7Dk7IEVjDb3IhwujBdWwVm7yaCCLQJUeP+Yc9+IpC3NUzuo+VqmXId 1hHkdMS4mp6sieYyUlAkNynw4LOaPlF/1L/h5KqraVNALmevPIUmITouWmpt5FMhJyLO Ik3sfa8lmxRV9niMzeUuu/UQEFKlSZ+R0Md7WPCFv51hTAzgcnfMKBD0D704ImPoFQQ1 3kpl0w8melO0xuJsDGqDrlZx6pyWdXp27fBU29L+AxxYTL9BouQw5aUHvs9+BlfIZSW0 gCInJ7cQTs1ezm86YN+yr/fzTQaQoB9djBbLuPy9gxWmabw0GEvgN0dFR7fVlePPRlFx 2chw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:mime-version:reply-to :list-subscribe:list-help:list-post:list-archive:list-unsubscribe :list-id:precedence:subject:references:in-reply-to:message-id:date :to:from:dkim-signature:delivered-to:arc-authentication-results; bh=8xxI8D/CsgMCyDDkkFH6+VG907RMqY24TxXYkAMCR14=; b=EFmsnXoBpbZSw5hwNTgH+wtMKLWe3xcJj8B5d3svNdRNlQ5PH3qGG4ZJFQf5/NeeUR ZG4EjEMYJ+RVg3NPP4gazbCk6qwfoTyxRTxCvY1Ffu9fHSSJ4f3UnWgKHegCVZXUmmUP FN/UAnR4F9aB8tzyWOUFS/8ow2G/ckdbqG3vXm/Jt4FpLdj/FgB8dIIteQ0UJKohTNFn wCWCqyXE7wSZvs9isZOJ2ZUAlFJ119kNHT9fpKx5xLoUtlCASexoqgZr9n0ykXX+WSlW ZvxfgWtetQSOIbiA+k1MITsIe002gtTSK2AcvMkejZ6yVgmoJ0jYICfAKwL3Od3iaiMG K3Yg== ARC-Authentication-Results: i=1; mx.google.com; dkim=neutral (body hash did not verify) header.i=@gmail.com; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=gmail.com Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id g48si15815003wrg.239.2017.05.08.15.00.45; Mon, 08 May 2017 15:00:46 -0700 (PDT) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; dkim=neutral (body hash did not verify) header.i=@gmail.com; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=gmail.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 4953668838F; Tue, 9 May 2017 01:00:36 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail-wr0-f175.google.com (mail-wr0-f175.google.com [209.85.128.175]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id EF439680570 for ; Tue, 9 May 2017 01:00:29 +0300 (EEST) Received: by mail-wr0-f175.google.com with SMTP id l50so55355422wrc.3 for ; Mon, 08 May 2017 15:00:36 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=from:to:subject:date:message-id:in-reply-to:references; bh=pVDFF/nxmAA+Fhpq/scxYBOTq00LB8QS5tvHwC+/RXs=; b=croH7QQDCr7i/2cw/Ng+DnMucbtFXqN8uHyIdsos/sFBN+g+n8HgPAKS+ldtL9buzX 6/htKpRiXdCfYochCEeabZvOT6a+65xS8K+KWhDzamnMD9sZNFXyLAppDXIK6ML6xCuB 7o+vKbcexPIZzKJLiTVJ0+cNQ9Nb4krgktUsedl+Lx4QgJ7P/8IUeaxrR0KjQ4QMkOdd mh21sy8KO+KP16XSLiUi2DB1OTV/sQNY9m3ur7bpFiFc3TG3RajzuFDnQ9eRMuBWmxUA JxGqztn4vCWtSo6F3tpPPs8GUIFFeA+4U5JdLEbpSkGyB8feasx4AEspnEc8gh6VJGnB sHOw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:subject:date:message-id:in-reply-to :references; bh=pVDFF/nxmAA+Fhpq/scxYBOTq00LB8QS5tvHwC+/RXs=; b=UaCPYTiB442TWoKRR76fnmpsomuHA0bgM01BEvdg69SWDJmK/LctgFSc1y2fjTx9EJ Mszyg6ZSL3TSi17AffLuBRlOIRCUXbQoVNEs2eTCpBKTYTKlaNeH2cN9bgpuYK7+rhsx 9uzBWuHgkz+JlAMuGiR1SQi+g9F5KvpAx4fPmy9uOZd3C8nBh38TUa62KEG8T1J1bFto Onr5+Jsv4hm+0okvwIMJwWuXinTPfiWgkz37XCs/cNMmUkwbTRFn0oEM7UVJVpvW84ul 5DX2BdbRsbLLllcGj4jE0y07R1zCtklZKmAy//Vu6wRtRktwUlmAhDUvQLfCvUBMUfJ5 uA+A== X-Gm-Message-State: AN3rC/6fC6dTpWpZPdj0YS//NMmLTktmIHz22oSM58C50j7YcQ/c0aDL rVhHSQylzsNyL5NZ X-Received: by 10.223.151.6 with SMTP id r6mr41777925wrb.189.1494280835342; Mon, 08 May 2017 15:00:35 -0700 (PDT) Received: from localhost.localdomain ([94.250.174.60]) by smtp.gmail.com with ESMTPSA id u76sm16011867wrb.27.2017.05.08.15.00.33 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Mon, 08 May 2017 15:00:34 -0700 (PDT) From: Paul B Mahol To: ffmpeg-devel@ffmpeg.org Date: Tue, 9 May 2017 00:00:23 +0200 Message-Id: <20170508220023.1262-1-onemda@gmail.com> X-Mailer: git-send-email 2.9.3 In-Reply-To: <20170507182200.14728-1-onemda@gmail.com> References: <20170507182200.14728-1-onemda@gmail.com> Subject: [FFmpeg-devel] [PATCH] avfilter: add arbitrary audio FIR filter X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches MIME-Version: 1.0 Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" Signed-off-by: Paul B Mahol --- configure | 2 + doc/filters.texi | 23 ++ libavfilter/Makefile | 1 + libavfilter/af_afir.c | 535 +++++++++++++++++++++++++++++++++++++++++ libavfilter/af_afir.h | 82 +++++++ libavfilter/allfilters.c | 1 + libavfilter/x86/Makefile | 2 + libavfilter/x86/af_afir.asm | 53 ++++ libavfilter/x86/af_afir_init.c | 35 +++ 9 files changed, 734 insertions(+) create mode 100644 libavfilter/af_afir.c create mode 100644 libavfilter/af_afir.h create mode 100644 libavfilter/x86/af_afir.asm create mode 100644 libavfilter/x86/af_afir_init.c diff --git a/configure b/configure index 2e1786a..a46c375 100755 --- a/configure +++ b/configure @@ -3081,6 +3081,8 @@ unix_protocol_select="network" # filters afftfilt_filter_deps="avcodec" afftfilt_filter_select="fft" +afir_filter_deps="avcodec" +afir_filter_select="fft" amovie_filter_deps="avcodec avformat" aresample_filter_deps="swresample" ass_filter_deps="libass" diff --git a/doc/filters.texi b/doc/filters.texi index f431274..0efce9a 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -878,6 +878,29 @@ afftfilt="1-clip((b/nb)*b,0,1)" @end example @end itemize +@section afir + +Apply an Arbitary Frequency Impulse Response filter. + +This filter uses second stream as FIR coefficients. +If second stream holds single channel, it will be used +for all input channels in first stream, otherwise +number of channels in second stream must be same as +number of channels in first stream. + +It accepts the following parameters: + +@table @option +@item dry +Set dry gain. This sets input gain. + +@item wet +Set wet gain. This sets final output gain. + +@item length +Set Impulse Response filter length. Default is 1, which means whole IR is processed. +@end table + @anchor{aformat} @section aformat diff --git a/libavfilter/Makefile b/libavfilter/Makefile index 0f99086..de5f992 100644 --- a/libavfilter/Makefile +++ b/libavfilter/Makefile @@ -37,6 +37,7 @@ OBJS-$(CONFIG_AEMPHASIS_FILTER) += af_aemphasis.o OBJS-$(CONFIG_AEVAL_FILTER) += aeval.o OBJS-$(CONFIG_AFADE_FILTER) += af_afade.o OBJS-$(CONFIG_AFFTFILT_FILTER) += af_afftfilt.o window_func.o +OBJS-$(CONFIG_AFIR_FILTER) += af_afir.o OBJS-$(CONFIG_AFORMAT_FILTER) += af_aformat.o OBJS-$(CONFIG_AGATE_FILTER) += af_agate.o OBJS-$(CONFIG_AINTERLEAVE_FILTER) += f_interleave.o diff --git a/libavfilter/af_afir.c b/libavfilter/af_afir.c new file mode 100644 index 0000000..eb59d53 --- /dev/null +++ b/libavfilter/af_afir.c @@ -0,0 +1,535 @@ +/* + * Copyright (c) 2017 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 + */ + +/** + * @file + * An arbitrary audio FIR filter + */ + +#include "libavutil/audio_fifo.h" +#include "libavutil/common.h" +#include "libavutil/float_dsp.h" +#include "libavutil/opt.h" +#include "libavcodec/avfft.h" + +#include "audio.h" +#include "avfilter.h" +#include "formats.h" +#include "internal.h" +#include "af_afir.h" + +static void fcmul_add_c(float *sum, const float *t, const float *c, int len) +{ + int n; + + for (n = 0; n < len; n++) { + const float cre = c[2 * n ]; + const float cim = c[2 * n + 1]; + const float tre = t[2 * n ]; + const float tim = t[2 * n + 1]; + + sum[2 * n ] += tre * cre - tim * cim; + sum[2 * n + 1] += tre * cim + tim * cre; + } + + sum[2 * n] += t[2 * n] * c[2 * n]; +} + +static int fir_channel(AVFilterContext *ctx, void *arg, int ch, int nb_jobs) +{ + AudioFIRContext *s = ctx->priv; + const float *src = (const float *)s->in[0]->extended_data[ch]; + int index1 = (s->index + 1) % 3; + int index2 = (s->index + 2) % 3; + float *sum = s->sum[ch]; + AVFrame *out = arg; + float *block; + float *dst; + int n, i, j; + + memset(sum, 0, sizeof(*sum) * s->fft_length); + block = s->block[ch] + s->part_index * s->block_size; + memset(block, 0, sizeof(*block) * s->fft_length); + + s->fdsp->vector_fmul_scalar(block + s->part_size, src, s->dry_gain, s->nb_samples); + emms_c(); + + av_rdft_calc(s->rdft[ch], block); + block[2 * s->part_size] = block[1]; + block[1] = 0; + + j = s->part_index; + + for (i = 0; i < s->nb_partitions; i++) { + const int coffset = i * s->coeff_size; + const FFTComplex *coeff = s->coeff[ch * !s->one2many] + coffset; + + block = s->block[ch] + j * s->block_size; + s->fcmul_add(sum, block, (const float *)coeff, s->part_size); + + if (j == 0) + j = s->nb_partitions; + j--; + } + + sum[1] = sum[2 * s->part_size]; + av_rdft_calc(s->irdft[ch], sum); + + dst = (float *)s->buffer->extended_data[ch] + index1 * s->part_size; + for (n = 0; n < s->part_size; n++) { + dst[n] += sum[n]; + } + + dst = (float *)s->buffer->extended_data[ch] + index2 * s->part_size; + + memcpy(dst, sum + s->part_size, s->part_size * sizeof(*dst)); + + dst = (float *)s->buffer->extended_data[ch] + s->index * s->part_size; + + if (out) { + float *ptr = (float *)out->extended_data[ch]; + s->fdsp->vector_fmul_scalar(ptr, dst, s->gain * s->wet_gain, out->nb_samples); + emms_c(); + } + + return 0; +} + +static int fir_frame(AudioFIRContext *s, AVFilterLink *outlink) +{ + AVFilterContext *ctx = outlink->src; + AVFrame *out = NULL; + int ret; + + s->nb_samples = FFMIN(s->part_size, av_audio_fifo_size(s->fifo[0])); + + if (!s->want_skip) { + out = ff_get_audio_buffer(outlink, s->nb_samples); + if (!out) + return AVERROR(ENOMEM); + } + + s->in[0] = ff_get_audio_buffer(ctx->inputs[0], s->nb_samples); + if (!s->in[0]) { + av_frame_free(&out); + return AVERROR(ENOMEM); + } + + av_audio_fifo_peek(s->fifo[0], (void **)s->in[0]->extended_data, s->nb_samples); + + ctx->internal->execute(ctx, fir_channel, out, NULL, outlink->channels); + + s->part_index = (s->part_index + 1) % s->nb_partitions; + + av_audio_fifo_drain(s->fifo[0], s->nb_samples); + + if (!s->want_skip) { + out->pts = s->pts; + if (s->pts != AV_NOPTS_VALUE) + s->pts += av_rescale_q(out->nb_samples, (AVRational){1, outlink->sample_rate}, outlink->time_base); + } + + s->index++; + if (s->index == 3) + s->index = 0; + + av_frame_free(&s->in[0]); + + if (s->want_skip == 1) { + s->want_skip = 0; + ret = 0; + } else { + ret = ff_filter_frame(outlink, out); + } + + return ret; +} + +static int convert_coeffs(AVFilterContext *ctx) +{ + AudioFIRContext *s = ctx->priv; + int i, ch, n, N; + float power = 0; + + s->nb_taps = av_audio_fifo_size(s->fifo[1]); + if (s->nb_taps <= 0) + return AVERROR(EINVAL); + + for (n = 4; (1 << n) < s->nb_taps; n++); + N = FFMIN(n, 16); + s->ir_length = 1 << n; + s->fft_length = (1 << (N + 1)) + 1; + s->part_size = 1 << (N - 1); + s->block_size = FFALIGN(s->fft_length, 16); + s->coeff_size = FFALIGN(s->part_size + 1, 16); + s->nb_partitions = (s->nb_taps + s->part_size - 1) / s->part_size; + s->nb_coeffs = s->ir_length + s->nb_partitions; + + for (ch = 0; ch < ctx->inputs[0]->channels; ch++) { + s->sum[ch] = av_calloc(s->fft_length, sizeof(**s->sum)); + if (!s->sum[ch]) + return AVERROR(ENOMEM); + } + + for (ch = 0; ch < ctx->inputs[1]->channels; ch++) { + s->coeff[ch] = av_calloc(s->nb_partitions * s->coeff_size, sizeof(**s->coeff)); + if (!s->coeff[ch]) + return AVERROR(ENOMEM); + } + + for (ch = 0; ch < ctx->inputs[0]->channels; ch++) { + s->block[ch] = av_calloc(s->nb_partitions * s->block_size, sizeof(**s->block)); + if (!s->block[ch]) + return AVERROR(ENOMEM); + } + + for (ch = 0; ch < ctx->inputs[0]->channels; ch++) { + s->rdft[ch] = av_rdft_init(N, DFT_R2C); + s->irdft[ch] = av_rdft_init(N, IDFT_C2R); + if (!s->rdft[ch] || !s->irdft[ch]) + return AVERROR(ENOMEM); + } + + s->in[1] = ff_get_audio_buffer(ctx->inputs[1], s->nb_taps); + if (!s->in[1]) + return AVERROR(ENOMEM); + + s->buffer = ff_get_audio_buffer(ctx->inputs[0], s->part_size * 3); + if (!s->buffer) + return AVERROR(ENOMEM); + + av_audio_fifo_read(s->fifo[1], (void **)s->in[1]->extended_data, s->nb_taps); + + for (ch = 0; ch < ctx->inputs[1]->channels; ch++) { + float *time = (float *)s->in[1]->extended_data[!s->one2many * ch]; + float *block = s->block[ch]; + FFTComplex *coeff = s->coeff[ch]; + + for (i = FFMAX(1, s->length * s->nb_taps); i < s->nb_taps; i++) + time[i] = 0; + + for (i = 0; i < s->nb_partitions; i++) { + const float scale = 1.f / s->part_size; + const int toffset = i * s->part_size; + const int coffset = i * s->coeff_size; + const int boffset = s->part_size; + const int remaining = s->nb_taps - (i * s->part_size); + const int size = remaining >= s->part_size ? s->part_size : remaining; + + memset(block, 0, sizeof(*block) * s->fft_length); + for (n = 0; n < size; n++) { + power += time[n + toffset] * time[n + toffset]; + block[n + boffset] = time[n + toffset]; + } + + av_rdft_calc(s->rdft[0], block); + + coeff[coffset].re = block[0] * scale; + coeff[coffset].im = 0; + for (n = 1; n < s->part_size; n++) { + coeff[coffset + n].re = block[2 * n] * scale; + coeff[coffset + n].im = block[2 * n + 1] * scale; + } + coeff[coffset + s->part_size].re = block[1] * scale; + coeff[coffset + s->part_size].im = 0; + } + } + + av_frame_free(&s->in[1]); + s->gain = 1.f / sqrtf(power); + av_log(ctx, AV_LOG_DEBUG, "nb_taps: %d\n", s->nb_taps); + av_log(ctx, AV_LOG_DEBUG, "nb_partitions: %d\n", s->nb_partitions); + av_log(ctx, AV_LOG_DEBUG, "partition size: %d\n", s->part_size); + av_log(ctx, AV_LOG_DEBUG, "ir_length: %d\n", s->ir_length); + + s->have_coeffs = 1; + + return 0; +} + +static int read_ir(AVFilterLink *link, AVFrame *frame) +{ + AVFilterContext *ctx = link->dst; + AudioFIRContext *s = ctx->priv; + int nb_taps, max_nb_taps; + + av_audio_fifo_write(s->fifo[1], (void **)frame->extended_data, + frame->nb_samples); + av_frame_free(&frame); + + nb_taps = av_audio_fifo_size(s->fifo[1]); + max_nb_taps = MAX_IR_DURATION * ctx->outputs[0]->sample_rate; + if (nb_taps > max_nb_taps) { + av_log(ctx, AV_LOG_ERROR, "Too big number of coefficients: %d > %d.\n", nb_taps, max_nb_taps); + return AVERROR(EINVAL); + } + + return 0; +} + +static int filter_frame(AVFilterLink *link, AVFrame *frame) +{ + AVFilterContext *ctx = link->dst; + AudioFIRContext *s = ctx->priv; + AVFilterLink *outlink = ctx->outputs[0]; + int ret = 0; + + av_audio_fifo_write(s->fifo[0], (void **)frame->extended_data, + frame->nb_samples); + if (s->pts == AV_NOPTS_VALUE) + s->pts = frame->pts; + + av_frame_free(&frame); + + if (!s->have_coeffs && s->eof_coeffs) { + ret = convert_coeffs(ctx); + if (ret < 0) + return ret; + } + + if (s->have_coeffs) { + while (av_audio_fifo_size(s->fifo[0]) >= s->part_size) { + ret = fir_frame(s, outlink); + if (ret < 0) + break; + } + } + return ret; +} + +static int request_frame(AVFilterLink *outlink) +{ + AVFilterContext *ctx = outlink->src; + AudioFIRContext *s = ctx->priv; + int ret; + + if (!s->eof_coeffs) { + ret = ff_request_frame(ctx->inputs[1]); + if (ret == AVERROR_EOF) { + s->eof_coeffs = 1; + ret = 0; + } + return ret; + } + ret = ff_request_frame(ctx->inputs[0]); + if (ret == AVERROR_EOF && s->have_coeffs) { + if (s->need_padding) { + AVFrame *silence = ff_get_audio_buffer(outlink, s->part_size); + + if (!silence) + return AVERROR(ENOMEM); + av_audio_fifo_write(s->fifo[0], (void **)silence->extended_data, + silence->nb_samples); + av_frame_free(&silence); + s->need_padding = 0; + } + + while (av_audio_fifo_size(s->fifo[0]) > 0) { + ret = fir_frame(s, outlink); + if (ret < 0) + return ret; + } + ret = AVERROR_EOF; + } + return ret; +} + +static int query_formats(AVFilterContext *ctx) +{ + AVFilterFormats *formats; + AVFilterChannelLayouts *layouts; + static const enum AVSampleFormat sample_fmts[] = { + AV_SAMPLE_FMT_FLTP, + AV_SAMPLE_FMT_NONE + }; + int ret, i; + + layouts = ff_all_channel_counts(); + if ((ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0) + return ret; + + for (i = 0; i < 2; i++) { + layouts = ff_all_channel_counts(); + if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0) + return ret; + } + + formats = ff_make_format_list(sample_fmts); + if ((ret = ff_set_common_formats(ctx, formats)) < 0) + return ret; + + formats = ff_all_samplerates(); + return ff_set_common_samplerates(ctx, formats); +} + +static int config_output(AVFilterLink *outlink) +{ + AVFilterContext *ctx = outlink->src; + AudioFIRContext *s = ctx->priv; + + if (ctx->inputs[0]->channels != ctx->inputs[1]->channels && + ctx->inputs[1]->channels != 1) { + av_log(ctx, AV_LOG_ERROR, + "Second input must have same number of channels as first input or " + "exactly 1 channel.\n"); + return AVERROR(EINVAL); + } + + s->one2many = ctx->inputs[1]->channels == 1; + outlink->sample_rate = ctx->inputs[0]->sample_rate; + outlink->time_base = ctx->inputs[0]->time_base; + outlink->channel_layout = ctx->inputs[0]->channel_layout; + outlink->channels = ctx->inputs[0]->channels; + + s->fifo[0] = av_audio_fifo_alloc(ctx->inputs[0]->format, ctx->inputs[0]->channels, 1024); + s->fifo[1] = av_audio_fifo_alloc(ctx->inputs[1]->format, ctx->inputs[1]->channels, 1024); + if (!s->fifo[0] || !s->fifo[1]) + return AVERROR(ENOMEM); + + s->sum = av_calloc(outlink->channels, sizeof(*s->sum)); + s->coeff = av_calloc(ctx->inputs[1]->channels, sizeof(*s->coeff)); + s->block = av_calloc(ctx->inputs[0]->channels, sizeof(*s->block)); + s->rdft = av_calloc(outlink->channels, sizeof(*s->rdft)); + s->irdft = av_calloc(outlink->channels, sizeof(*s->irdft)); + if (!s->sum || !s->coeff || !s->block || !s->rdft || !s->irdft) + return AVERROR(ENOMEM); + + s->nb_channels = outlink->channels; + s->nb_coef_channels = ctx->inputs[1]->channels; + s->want_skip = 1; + s->need_padding = 1; + s->pts = AV_NOPTS_VALUE; + + return 0; +} + +static av_cold void uninit(AVFilterContext *ctx) +{ + AudioFIRContext *s = ctx->priv; + int ch; + + if (s->sum) { + for (ch = 0; ch < s->nb_channels; ch++) { + av_freep(&s->sum[ch]); + } + } + av_freep(&s->sum); + + if (s->coeff) { + for (ch = 0; ch < s->nb_coef_channels; ch++) { + av_freep(&s->coeff[ch]); + } + } + av_freep(&s->coeff); + + if (s->block) { + for (ch = 0; ch < s->nb_channels; ch++) { + av_freep(&s->block[ch]); + } + } + av_freep(&s->block); + + if (s->rdft) { + for (ch = 0; ch < s->nb_channels; ch++) { + av_rdft_end(s->rdft[ch]); + } + } + av_freep(&s->rdft); + + if (s->irdft) { + for (ch = 0; ch < s->nb_channels; ch++) { + av_rdft_end(s->irdft[ch]); + } + } + av_freep(&s->irdft); + + av_frame_free(&s->in[0]); + av_frame_free(&s->in[1]); + av_frame_free(&s->buffer); + + av_audio_fifo_free(s->fifo[0]); + av_audio_fifo_free(s->fifo[1]); + + av_freep(&s->fdsp); +} + +static av_cold int init(AVFilterContext *ctx) +{ + AudioFIRContext *s = ctx->priv; + + s->fcmul_add = fcmul_add_c; + + s->fdsp = avpriv_float_dsp_alloc(0); + if (!s->fdsp) + return AVERROR(ENOMEM); + + if (ARCH_X86) + ff_afir_init_x86(s); + + return 0; +} + +static const AVFilterPad afir_inputs[] = { + { + .name = "main", + .type = AVMEDIA_TYPE_AUDIO, + .filter_frame = filter_frame, + },{ + .name = "ir", + .type = AVMEDIA_TYPE_AUDIO, + .filter_frame = read_ir, + }, + { NULL } +}; + +static const AVFilterPad afir_outputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_AUDIO, + .config_props = config_output, + .request_frame = request_frame, + }, + { NULL } +}; + +#define AF AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM +#define OFFSET(x) offsetof(AudioFIRContext, x) + +static const AVOption afir_options[] = { + { "dry", "set dry gain", OFFSET(dry_gain), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 1, AF }, + { "wet", "set wet gain", OFFSET(wet_gain), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 1, AF }, + { "length", "set IR length", OFFSET(length), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 1, AF }, + { NULL } +}; + +AVFILTER_DEFINE_CLASS(afir); + +AVFilter ff_af_afir = { + .name = "afir", + .description = NULL_IF_CONFIG_SMALL("Apply Finite Impulse Response filter with supplied coefficients in 2nd stream."), + .priv_size = sizeof(AudioFIRContext), + .priv_class = &afir_class, + .query_formats = query_formats, + .init = init, + .uninit = uninit, + .inputs = afir_inputs, + .outputs = afir_outputs, + .flags = AVFILTER_FLAG_SLICE_THREADS, +}; diff --git a/libavfilter/af_afir.h b/libavfilter/af_afir.h new file mode 100644 index 0000000..5379199 --- /dev/null +++ b/libavfilter/af_afir.h @@ -0,0 +1,82 @@ +/* + * Copyright (c) 2017 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 + */ + +#ifndef AVFILTER_AFIR_H +#define AVFILTER_AFIR_H + +#include "libavutil/audio_fifo.h" +#include "libavutil/common.h" +#include "libavutil/float_dsp.h" +#include "libavutil/opt.h" +#include "libavcodec/avfft.h" + +#include "audio.h" +#include "avfilter.h" +#include "formats.h" +#include "internal.h" + +#define MAX_IR_DURATION 30 + +typedef struct AudioFIRContext { + const AVClass *class; + + float wet_gain; + float dry_gain; + float length; + + float gain; + + int eof_coeffs; + int have_coeffs; + int nb_coeffs; + int nb_taps; + int part_size; + int part_index; + int coeff_size; + int block_size; + int nb_partitions; + int nb_channels; + int ir_length; + int fft_length; + int nb_coef_channels; + int one2many; + int nb_samples; + int want_skip; + int need_padding; + + RDFTContext **rdft, **irdft; + float **sum; + float **block; + FFTComplex **coeff; + + AVAudioFifo *fifo[2]; + AVFrame *in[2]; + AVFrame *buffer; + int64_t pts; + int index; + + AVFloatDSPContext *fdsp; + void (*fcmul_add)(float *sum, const float *t, const float *c, + int len); +} AudioFIRContext; + +void ff_afir_init_x86(AudioFIRContext *s); + +#endif /* AVFILTER_AFIR_H */ diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c index 8fb87eb..555c442 100644 --- a/libavfilter/allfilters.c +++ b/libavfilter/allfilters.c @@ -50,6 +50,7 @@ static void register_all(void) REGISTER_FILTER(AEVAL, aeval, af); REGISTER_FILTER(AFADE, afade, af); REGISTER_FILTER(AFFTFILT, afftfilt, af); + REGISTER_FILTER(AFIR, afir, af); REGISTER_FILTER(AFORMAT, aformat, af); REGISTER_FILTER(AGATE, agate, af); REGISTER_FILTER(AINTERLEAVE, ainterleave, af); diff --git a/libavfilter/x86/Makefile b/libavfilter/x86/Makefile index b6195f8..135e75f 100644 --- a/libavfilter/x86/Makefile +++ b/libavfilter/x86/Makefile @@ -1,3 +1,4 @@ +OBJS-$(CONFIG_AFIR_FILTER) += x86/af_afir_init.o OBJS-$(CONFIG_BLEND_FILTER) += x86/vf_blend_init.o OBJS-$(CONFIG_BWDIF_FILTER) += x86/vf_bwdif_init.o OBJS-$(CONFIG_COLORSPACE_FILTER) += x86/colorspacedsp_init.o @@ -23,6 +24,7 @@ OBJS-$(CONFIG_VOLUME_FILTER) += x86/af_volume_init.o OBJS-$(CONFIG_W3FDIF_FILTER) += x86/vf_w3fdif_init.o OBJS-$(CONFIG_YADIF_FILTER) += x86/vf_yadif_init.o +YASM-OBJS-$(CONFIG_AFIR_FILTER) += x86/af_afir.o YASM-OBJS-$(CONFIG_BLEND_FILTER) += x86/vf_blend.o YASM-OBJS-$(CONFIG_BWDIF_FILTER) += x86/vf_bwdif.o YASM-OBJS-$(CONFIG_COLORSPACE_FILTER) += x86/colorspacedsp.o diff --git a/libavfilter/x86/af_afir.asm b/libavfilter/x86/af_afir.asm new file mode 100644 index 0000000..b425055 --- /dev/null +++ b/libavfilter/x86/af_afir.asm @@ -0,0 +1,53 @@ +;***************************************************************************** +;* x86-optimized functions for afir filter +;* Copyright (c) 2017 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/x86/x86util.asm" + +SECTION_RODATA 32 + +SECTION .text + +;------------------------------------------------------------------------------ +; void ff_fcmul_add(float *sum, const float *t, const float *c, int len) +;------------------------------------------------------------------------------ + +INIT_XMM sse3 +cglobal fcmul_add, 4,4,3, sum, t, c, len + shl lend, 3 + add lend, mmsize + add tq, lenq + add cq, lenq + add sumq, lenq + neg lenq +ALIGN 16 +.loop: + movsldup m0, [tq + lenq] + movaps m1, [cq + lenq] + mulps m0, m1 + shufps m1, m1, 0xb1 + movshdup m2, [tq + lenq] + mulps m2, m1 + addsubps m0, m2; + addps m0, [sumq + lenq] + movaps [sumq + lenq], m0 + add lenq, mmsize + jl .loop + REP_RET diff --git a/libavfilter/x86/af_afir_init.c b/libavfilter/x86/af_afir_init.c new file mode 100644 index 0000000..1cd5290 --- /dev/null +++ b/libavfilter/x86/af_afir_init.c @@ -0,0 +1,35 @@ +/* + * 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 "config.h" +#include "libavutil/attributes.h" +#include "libavutil/cpu.h" +#include "libavutil/x86/cpu.h" +#include "libavfilter/af_afir.h" + +void ff_fcmul_add_sse3(float *sum, const float *t, const float *c, + int len); + +av_cold void ff_afir_init_x86(AudioFIRContext *s) +{ + int cpu_flags = av_get_cpu_flags(); + + if (EXTERNAL_SSE3(cpu_flags)) { + s->fcmul_add = ff_fcmul_add_sse3; + } +}