Mercurial > libavcodec.hg
view gsmdec.c @ 12334:435319d67bd8 libavcodec
Use word-writing instead of dword-writing (with two cached but otherwise
unchanged bytes) in the horizontal simple loopfilter. This makes the filter
quite a bit faster in itself (~30 cycles less on Core1), probably mostly
because we don't need a complex 4x4 transpose, but only a simple byte
interleave. Also allows using pextrw on SSE4, which speeds up even more
(e.g. 25% faster on Core i7).
author | rbultje |
---|---|
date | Sat, 31 Jul 2010 23:13:15 +0000 |
parents | 81871058b27c |
children |
line wrap: on
line source
/* * gsm 06.10 decoder * Copyright (c) 2010 Reimar Döffinger <Reimar.Doeffinger@gmx.de> * * 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 * GSM decoder */ #define ALT_BITSTREAM_READER_LE #include "avcodec.h" #include "get_bits.h" // input and output sizes in byte #define GSM_BLOCK_SIZE 33 #define GSM_MS_BLOCK_SIZE 65 #define GSM_FRAME_SIZE 160 typedef struct { // Contains first 120 elements from the previous frame // (used by long_term_synth according to the "lag"), // then in the following 160 elements the current // frame is constructed. int16_t ref_buf[280]; int v[9]; int lar[2][8]; int lar_idx; int msr; } GSMContext; static av_cold int gsm_init(AVCodecContext *avctx) { avctx->channels = 1; if (!avctx->sample_rate) avctx->sample_rate = 8000; avctx->sample_fmt = SAMPLE_FMT_S16; switch (avctx->codec_id) { case CODEC_ID_GSM: avctx->frame_size = GSM_FRAME_SIZE; avctx->block_align = GSM_BLOCK_SIZE; break; case CODEC_ID_GSM_MS: avctx->frame_size = 2 * GSM_FRAME_SIZE; avctx->block_align = GSM_MS_BLOCK_SIZE; } return 0; } static const int16_t dequant_tab[64][8] = { { -28, -20, -12, -4, 4, 12, 20, 28}, { -56, -40, -24, -8, 8, 24, 40, 56}, { -84, -60, -36, -12, 12, 36, 60, 84}, { -112, -80, -48, -16, 16, 48, 80, 112}, { -140, -100, -60, -20, 20, 60, 100, 140}, { -168, -120, -72, -24, 24, 72, 120, 168}, { -196, -140, -84, -28, 28, 84, 140, 196}, { -224, -160, -96, -32, 32, 96, 160, 224}, { -252, -180, -108, -36, 36, 108, 180, 252}, { -280, -200, -120, -40, 40, 120, 200, 280}, { -308, -220, -132, -44, 44, 132, 220, 308}, { -336, -240, -144, -48, 48, 144, 240, 336}, { -364, -260, -156, -52, 52, 156, 260, 364}, { -392, -280, -168, -56, 56, 168, 280, 392}, { -420, -300, -180, -60, 60, 180, 300, 420}, { -448, -320, -192, -64, 64, 192, 320, 448}, { -504, -360, -216, -72, 72, 216, 360, 504}, { -560, -400, -240, -80, 80, 240, 400, 560}, { -616, -440, -264, -88, 88, 264, 440, 616}, { -672, -480, -288, -96, 96, 288, 480, 672}, { -728, -520, -312, -104, 104, 312, 520, 728}, { -784, -560, -336, -112, 112, 336, 560, 784}, { -840, -600, -360, -120, 120, 360, 600, 840}, { -896, -640, -384, -128, 128, 384, 640, 896}, { -1008, -720, -432, -144, 144, 432, 720, 1008}, { -1120, -800, -480, -160, 160, 480, 800, 1120}, { -1232, -880, -528, -176, 176, 528, 880, 1232}, { -1344, -960, -576, -192, 192, 576, 960, 1344}, { -1456, -1040, -624, -208, 208, 624, 1040, 1456}, { -1568, -1120, -672, -224, 224, 672, 1120, 1568}, { -1680, -1200, -720, -240, 240, 720, 1200, 1680}, { -1792, -1280, -768, -256, 256, 768, 1280, 1792}, { -2016, -1440, -864, -288, 288, 864, 1440, 2016}, { -2240, -1600, -960, -320, 320, 960, 1600, 2240}, { -2464, -1760, -1056, -352, 352, 1056, 1760, 2464}, { -2688, -1920, -1152, -384, 384, 1152, 1920, 2688}, { -2912, -2080, -1248, -416, 416, 1248, 2080, 2912}, { -3136, -2240, -1344, -448, 448, 1344, 2240, 3136}, { -3360, -2400, -1440, -480, 480, 1440, 2400, 3360}, { -3584, -2560, -1536, -512, 512, 1536, 2560, 3584}, { -4032, -2880, -1728, -576, 576, 1728, 2880, 4032}, { -4480, -3200, -1920, -640, 640, 1920, 3200, 4480}, { -4928, -3520, -2112, -704, 704, 2112, 3520, 4928}, { -5376, -3840, -2304, -768, 768, 2304, 3840, 5376}, { -5824, -4160, -2496, -832, 832, 2496, 4160, 5824}, { -6272, -4480, -2688, -896, 896, 2688, 4480, 6272}, { -6720, -4800, -2880, -960, 960, 2880, 4800, 6720}, { -7168, -5120, -3072, -1024, 1024, 3072, 5120, 7168}, { -8063, -5759, -3456, -1152, 1152, 3456, 5760, 8064}, { -8959, -6399, -3840, -1280, 1280, 3840, 6400, 8960}, { -9855, -7039, -4224, -1408, 1408, 4224, 7040, 9856}, {-10751, -7679, -4608, -1536, 1536, 4608, 7680, 10752}, {-11647, -8319, -4992, -1664, 1664, 4992, 8320, 11648}, {-12543, -8959, -5376, -1792, 1792, 5376, 8960, 12544}, {-13439, -9599, -5760, -1920, 1920, 5760, 9600, 13440}, {-14335, -10239, -6144, -2048, 2048, 6144, 10240, 14336}, {-16127, -11519, -6912, -2304, 2304, 6912, 11519, 16127}, {-17919, -12799, -7680, -2560, 2560, 7680, 12799, 17919}, {-19711, -14079, -8448, -2816, 2816, 8448, 14079, 19711}, {-21503, -15359, -9216, -3072, 3072, 9216, 15359, 21503}, {-23295, -16639, -9984, -3328, 3328, 9984, 16639, 23295}, {-25087, -17919, -10752, -3584, 3584, 10752, 17919, 25087}, {-26879, -19199, -11520, -3840, 3840, 11520, 19199, 26879}, {-28671, -20479, -12288, -4096, 4096, 12288, 20479, 28671} }; static void apcm_dequant_add(GetBitContext *gb, int16_t *dst) { int i; int maxidx = get_bits(gb, 6); const int16_t *tab = dequant_tab[maxidx]; for (i = 0; i < 13; i++) dst[3*i] += tab[get_bits(gb, 3)]; } static inline int gsm_mult(int a, int b) { return (a * b + (1 << 14)) >> 15; } static const uint16_t long_term_gain_tab[4] = { 3277, 11469, 21299, 32767 }; static void long_term_synth(int16_t *dst, int lag, int gain_idx) { int i; const int16_t *src = dst - lag; uint16_t gain = long_term_gain_tab[gain_idx]; for (i = 0; i < 40; i++) dst[i] = gsm_mult(gain, src[i]); } static inline int decode_log_area(int coded, int factor, int offset) { coded <<= 10; coded -= offset; return gsm_mult(coded, factor) << 1; } static av_noinline int get_rrp(int filtered) { int abs = FFABS(filtered); if (abs < 11059) abs <<= 1; else if (abs < 20070) abs += 11059; else abs = (abs >> 2) + 26112; return filtered < 0 ? -abs : abs; } static int filter_value(int in, int rrp[8], int v[9]) { int i; for (i = 7; i >= 0; i--) { in -= gsm_mult(rrp[i], v[i]); v[i + 1] = v[i] + gsm_mult(rrp[i], in); } v[0] = in; return in; } static void short_term_synth(GSMContext *ctx, int16_t *dst, const int16_t *src) { int i; int rrp[8]; int *lar = ctx->lar[ctx->lar_idx]; int *lar_prev = ctx->lar[ctx->lar_idx ^ 1]; for (i = 0; i < 8; i++) rrp[i] = get_rrp((lar_prev[i] >> 2) + (lar_prev[i] >> 1) + (lar[i] >> 2)); for (i = 0; i < 13; i++) dst[i] = filter_value(src[i], rrp, ctx->v); for (i = 0; i < 8; i++) rrp[i] = get_rrp((lar_prev[i] >> 1) + (lar [i] >> 1)); for (i = 13; i < 27; i++) dst[i] = filter_value(src[i], rrp, ctx->v); for (i = 0; i < 8; i++) rrp[i] = get_rrp((lar_prev[i] >> 2) + (lar [i] >> 1) + (lar[i] >> 2)); for (i = 27; i < 40; i++) dst[i] = filter_value(src[i], rrp, ctx->v); for (i = 0; i < 8; i++) rrp[i] = get_rrp(lar[i]); for (i = 40; i < 160; i++) dst[i] = filter_value(src[i], rrp, ctx->v); ctx->lar_idx ^= 1; } static int postprocess(int16_t *data, int msr) { int i; for (i = 0; i < 160; i++) { msr = av_clip_int16(data[i] + gsm_mult(msr, 28180)); data[i] = av_clip_int16(msr << 1) & ~7; } return msr; } static int gsm_decode_block(AVCodecContext *avctx, int16_t *samples, GetBitContext *gb) { GSMContext *ctx = avctx->priv_data; int i; int16_t *ref_dst = ctx->ref_buf + 120; int *lar = ctx->lar[ctx->lar_idx]; lar[0] = decode_log_area(get_bits(gb, 6), 13107, 1 << 15); lar[1] = decode_log_area(get_bits(gb, 6), 13107, 1 << 15); lar[2] = decode_log_area(get_bits(gb, 5), 13107, (1 << 14) + 2048*2); lar[3] = decode_log_area(get_bits(gb, 5), 13107, (1 << 14) - 2560*2); lar[4] = decode_log_area(get_bits(gb, 4), 19223, (1 << 13) + 94*2); lar[5] = decode_log_area(get_bits(gb, 4), 17476, (1 << 13) - 1792*2); lar[6] = decode_log_area(get_bits(gb, 3), 31454, (1 << 12) - 341*2); lar[7] = decode_log_area(get_bits(gb, 3), 29708, (1 << 12) - 1144*2); for (i = 0; i < 4; i++) { int lag = get_bits(gb, 7); int gain_idx = get_bits(gb, 2); int offset = get_bits(gb, 2); lag = av_clip(lag, 40, 120); long_term_synth(ref_dst, lag, gain_idx); apcm_dequant_add(gb, ref_dst + offset); ref_dst += 40; } memcpy(ctx->ref_buf, ctx->ref_buf + 160, 120 * sizeof(*ctx->ref_buf)); short_term_synth(ctx, samples, ctx->ref_buf + 120); // for optimal speed this could be merged with short_term_synth, // not done yet because it is a bit ugly ctx->msr = postprocess(samples, ctx->msr); return 0; } static int gsm_decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPacket *avpkt) { int res; GetBitContext gb; const uint8_t *buf = avpkt->data; int buf_size = avpkt->size; int16_t *samples = data; int frame_bytes = 2 * avctx->frame_size; if (*data_size < frame_bytes) return -1; *data_size = 0; if(buf_size < avctx->block_align) return AVERROR_INVALIDDATA; init_get_bits(&gb, buf, buf_size * 8); switch (avctx->codec_id) { case CODEC_ID_GSM: if (get_bits(&gb, 4) != 0xd) av_log(avctx, AV_LOG_WARNING, "Missing GSM magic!\n"); res = gsm_decode_block(avctx, samples, &gb); if (res < 0) return res; break; case CODEC_ID_GSM_MS: res = gsm_decode_block(avctx, samples, &gb); if (res < 0) return res; res = gsm_decode_block(avctx, samples + GSM_FRAME_SIZE, &gb); if (res < 0) return res; } *data_size = frame_bytes; return avctx->block_align; } AVCodec gsm_decoder = { "gsm", AVMEDIA_TYPE_AUDIO, CODEC_ID_GSM, sizeof(GSMContext), gsm_init, NULL, NULL, gsm_decode_frame, .long_name = NULL_IF_CONFIG_SMALL("GSM"), }; AVCodec gsm_ms_decoder = { "gsm_ms", AVMEDIA_TYPE_AUDIO, CODEC_ID_GSM_MS, sizeof(GSMContext), gsm_init, NULL, NULL, gsm_decode_frame, .long_name = NULL_IF_CONFIG_SMALL("GSM Microsoft variant"), };