From 42315dabce376fd7085e2a1bbab4d230d3d2ccd8 Mon Sep 17 00:00:00 2001 From: Kostya Shishkov Date: Tue, 15 Mar 2011 09:37:48 +0100 Subject: Chronomaster DFA decoder Signed-off-by: Anton Khirnov --- libavcodec/Makefile | 1 + libavcodec/allcodecs.c | 1 + libavcodec/avcodec.h | 1 + libavcodec/dfa.c | 395 +++++++++++++++++++++++++++++++++++++++++++++++++ libavcodec/version.h | 2 +- 5 files changed, 399 insertions(+), 1 deletion(-) create mode 100644 libavcodec/dfa.c (limited to 'libavcodec') diff --git a/libavcodec/Makefile b/libavcodec/Makefile index 85d9a34..4cb61ca 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -103,6 +103,7 @@ OBJS-$(CONFIG_COOK_DECODER) += cook.o OBJS-$(CONFIG_CSCD_DECODER) += cscd.o OBJS-$(CONFIG_CYUV_DECODER) += cyuv.o OBJS-$(CONFIG_DCA_DECODER) += dca.o synth_filter.o dcadsp.o +OBJS-$(CONFIG_DFA_DECODER) += dfa.o OBJS-$(CONFIG_DNXHD_DECODER) += dnxhddec.o dnxhddata.o OBJS-$(CONFIG_DNXHD_ENCODER) += dnxhdenc.o dnxhddata.o \ mpegvideo_enc.o motion_est.o \ diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index 8de6ad8..9aa29b7 100644 --- a/libavcodec/allcodecs.c +++ b/libavcodec/allcodecs.c @@ -88,6 +88,7 @@ void avcodec_register_all(void) REGISTER_DECODER (CLJR, cljr); REGISTER_DECODER (CSCD, cscd); REGISTER_DECODER (CYUV, cyuv); + REGISTER_DECODER (DFA, dfa); REGISTER_ENCDEC (DNXHD, dnxhd); REGISTER_DECODER (DPX, dpx); REGISTER_DECODER (DSICINVIDEO, dsicinvideo); diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h index 150c99d..4cd78a3 100644 --- a/libavcodec/avcodec.h +++ b/libavcodec/avcodec.h @@ -212,6 +212,7 @@ enum CodecID { CODEC_ID_LAGARITH, CODEC_ID_PRORES, CODEC_ID_JV, + CODEC_ID_DFA, /* various PCM "codecs" */ CODEC_ID_PCM_S16LE= 0x10000, diff --git a/libavcodec/dfa.c b/libavcodec/dfa.c new file mode 100644 index 0000000..1023197 --- /dev/null +++ b/libavcodec/dfa.c @@ -0,0 +1,395 @@ +/* + * Chronomaster DFA Video Decoder + * Copyright (c) 2011 Konstantin Shishkov + * based on work by Vladimir "VAG" Gneushev + * + * This file is part of Libav. + * + * Libav 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. + * + * Libav 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 Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "avcodec.h" +#include "libavutil/intreadwrite.h" +#include "bytestream.h" +#include "libavutil/lzo.h" // for av_memcpy_backptr + +typedef struct DfaContext { + AVFrame pic; + + uint32_t pal[256]; + uint8_t *frame_buf; +} DfaContext; + +static av_cold int dfa_decode_init(AVCodecContext *avctx) +{ + DfaContext *s = avctx->priv_data; + + avctx->pix_fmt = PIX_FMT_PAL8; + + s->frame_buf = av_mallocz(avctx->width * avctx->height + AV_LZO_OUTPUT_PADDING); + if (!s->frame_buf) + return AVERROR(ENOMEM); + + return 0; +} + +static int decode_copy(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end) +{ + const int size = width * height; + + if (src_end - src < size) + return -1; + bytestream_get_buffer(&src, frame, size); + return 0; +} + +static int decode_tsw1(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end) +{ + const uint8_t *frame_start = frame; + const uint8_t *frame_end = frame + width * height; + int mask = 0x10000, bitbuf = 0; + int v, offset, count, segments; + + segments = bytestream_get_le32(&src); + frame += bytestream_get_le32(&src); + if (frame < frame_start || frame > frame_end) + return -1; + while (segments--) { + if (mask == 0x10000) { + if (src >= src_end) + return -1; + bitbuf = bytestream_get_le16(&src); + mask = 1; + } + if (src_end - src < 2 || frame_end - frame < 2) + return -1; + if (bitbuf & mask) { + v = bytestream_get_le16(&src); + offset = (v & 0x1FFF) << 1; + count = ((v >> 13) + 2) << 1; + if (frame - offset < frame_start || frame_end - frame < count) + return -1; + av_memcpy_backptr(frame, offset, count); + frame += count; + } else { + *frame++ = *src++; + *frame++ = *src++; + } + mask <<= 1; + } + + return 0; +} + +static int decode_dsw1(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end) +{ + const uint8_t *frame_start = frame; + const uint8_t *frame_end = frame + width * height; + int mask = 0x10000, bitbuf = 0; + int v, offset, count, segments; + + segments = bytestream_get_le16(&src); + while (segments--) { + if (mask == 0x10000) { + if (src >= src_end) + return -1; + bitbuf = bytestream_get_le16(&src); + mask = 1; + } + if (src_end - src < 2 || frame_end - frame < 2) + return -1; + if (bitbuf & mask) { + v = bytestream_get_le16(&src); + offset = (v & 0x1FFF) << 1; + count = ((v >> 13) + 2) << 1; + if (frame - offset < frame_start || frame_end - frame < count) + return -1; + // can't use av_memcpy_backptr() since it can overwrite following pixels + for (v = 0; v < count; v++) + frame[v] = frame[v - offset]; + frame += count; + } else if (bitbuf & (mask << 1)) { + frame += bytestream_get_le16(&src); + } else { + *frame++ = *src++; + *frame++ = *src++; + } + mask <<= 2; + } + + return 0; +} + +static int decode_dds1(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end) +{ + const uint8_t *frame_start = frame; + const uint8_t *frame_end = frame + width * height; + int mask = 0x10000, bitbuf = 0; + int i, v, offset, count, segments; + + segments = bytestream_get_le16(&src); + while (segments--) { + if (mask == 0x10000) { + if (src >= src_end) + return -1; + bitbuf = bytestream_get_le16(&src); + mask = 1; + } + if (src_end - src < 2 || frame_end - frame < 2) + return -1; + if (bitbuf & mask) { + v = bytestream_get_le16(&src); + offset = (v & 0x1FFF) << 2; + count = ((v >> 13) + 2) << 1; + if (frame - offset < frame_start || frame_end - frame < count*2 + width) + return -1; + for (i = 0; i < count; i++) { + frame[0] = frame[1] = + frame[width] = frame[width + 1] = frame[-offset]; + + frame += 2; + } + } else if (bitbuf & (mask << 1)) { + frame += bytestream_get_le16(&src) * 2; + } else { + frame[0] = frame[1] = + frame[width] = frame[width + 1] = *src++; + frame += 2; + frame[0] = frame[1] = + frame[width] = frame[width + 1] = *src++; + frame += 2; + } + mask <<= 2; + } + + return 0; +} + +static int decode_bdlt(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end) +{ + const uint8_t *frame_end = frame + width * height; + uint8_t *line_ptr; + int count, lines, segments; + + count = bytestream_get_le16(&src); + if (count >= height || width * count < 0) + return -1; + frame += width * count; + lines = bytestream_get_le16(&src); + if (frame + lines * width > frame_end || src >= src_end) + return -1; + + while (lines--) { + line_ptr = frame; + frame += width; + segments = *src++; + while (segments--) { + if (src_end - src < 3) + return -1; + line_ptr += *src++; + if (line_ptr >= frame) + return -1; + count = (int8_t)*src++; + if (count >= 0) { + if (line_ptr + count > frame || src_end - src < count) + return -1; + bytestream_get_buffer(&src, line_ptr, count); + } else { + count = -count; + if (line_ptr + count > frame || src >= src_end) + return -1; + memset(line_ptr, *src++, count); + } + line_ptr += count; + } + } + + return 0; +} + +static int decode_wdlt(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end) +{ + const uint8_t *frame_end = frame + width * height; + uint8_t *line_ptr; + int count, i, v, lines, segments; + + lines = bytestream_get_le16(&src); + if (frame + lines * width > frame_end || src >= src_end) + return -1; + + while (lines--) { + segments = bytestream_get_le16(&src); + while ((segments & 0xC000) == 0xC000) { + frame -= (int16_t)segments * width; + if (frame >= frame_end) + return -1; + segments = bytestream_get_le16(&src); + } + if (segments & 0x8000) { + frame[width - 1] = segments & 0xFF; + segments = bytestream_get_le16(&src); + } + line_ptr = frame; + frame += width; + while (segments--) { + if (src_end - src < 2) + return -1; + line_ptr += *src++; + if (line_ptr >= frame) + return -1; + count = (int8_t)*src++; + if (count >= 0) { + if (line_ptr + count*2 > frame || src_end - src < count*2) + return -1; + bytestream_get_buffer(&src, line_ptr, count*2); + line_ptr += count * 2; + } else { + count = -count; + if (line_ptr + count*2 > frame || src_end - src < 2) + return -1; + v = bytestream_get_le16(&src); + for (i = 0; i < count; i++) + bytestream_put_le16(&line_ptr, v); + } + } + } + + return 0; +} + +static int decode_unk6(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end) +{ + return -1; +} + +static int decode_blck(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end) +{ + memset(frame, 0, width * height); + return 0; +} + + +typedef int (*chunk_decoder)(uint8_t *frame, int width, int height, + const uint8_t *src, const uint8_t *src_end); + +static const chunk_decoder decoder[8] = { + decode_copy, decode_tsw1, decode_bdlt, decode_wdlt, + decode_unk6, decode_dsw1, decode_blck, decode_dds1, +}; + +static const char* chunk_name[8] = { + "COPY", "TSW1", "BDLT", "WDLT", "????", "DSW1", "BLCK", "DDS1" +}; + +static int dfa_decode_frame(AVCodecContext *avctx, + void *data, int *data_size, + AVPacket *avpkt) +{ + DfaContext *s = avctx->priv_data; + const uint8_t *buf = avpkt->data; + const uint8_t *buf_end = avpkt->data + avpkt->size; + const uint8_t *tmp_buf; + uint32_t chunk_type, chunk_size; + uint8_t *dst; + int ret; + int i, pal_elems; + + if (s->pic.data[0]) + avctx->release_buffer(avctx, &s->pic); + + if ((ret = avctx->get_buffer(avctx, &s->pic))) { + av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); + return ret; + } + + while (buf < buf_end) { + chunk_size = AV_RL32(buf + 4); + chunk_type = AV_RL32(buf + 8); + buf += 12; + if (buf_end - buf < chunk_size) { + av_log(avctx, AV_LOG_ERROR, "Chunk size is too big (%d bytes)\n", chunk_size); + return -1; + } + if (!chunk_type) + break; + if (chunk_type == 1) { + pal_elems = FFMIN(chunk_size / 3, 256); + tmp_buf = buf; + for (i = 0; i < pal_elems; i++) { + s->pal[i] = bytestream_get_be24(&tmp_buf) << 2; + s->pal[i] |= (s->pal[i] >> 6) & 0x333; + } + s->pic.palette_has_changed = 1; + } else if (chunk_type <= 9) { + if (decoder[chunk_type - 2](s->frame_buf, avctx->width, avctx->height, + buf, buf + chunk_size)) { + av_log(avctx, AV_LOG_ERROR, "Error decoding %s chunk\n", + chunk_name[chunk_type - 2]); + return -1; + } + } else { + av_log(avctx, AV_LOG_WARNING, "Ignoring unknown chunk type %d\n", + chunk_type); + } + buf += chunk_size; + } + + buf = s->frame_buf; + dst = s->pic.data[0]; + for (i = 0; i < avctx->height; i++) { + memcpy(dst, buf, avctx->width); + dst += s->pic.linesize[0]; + buf += avctx->width; + } + memcpy(s->pic.data[1], s->pal, sizeof(s->pal)); + + *data_size = sizeof(AVFrame); + *(AVFrame*)data = s->pic; + + return avpkt->size; +} + +static av_cold int dfa_decode_end(AVCodecContext *avctx) +{ + DfaContext *s = avctx->priv_data; + + if (s->pic.data[0]) + avctx->release_buffer(avctx, &s->pic); + + av_freep(&s->frame_buf); + + return 0; +} + +AVCodec ff_dfa_decoder = { + "dfa", + AVMEDIA_TYPE_VIDEO, + CODEC_ID_DFA, + sizeof(DfaContext), + dfa_decode_init, + NULL, + dfa_decode_end, + dfa_decode_frame, + CODEC_CAP_DR1, + .long_name = NULL_IF_CONFIG_SMALL("Chronomaster DFA"), +}; diff --git a/libavcodec/version.h b/libavcodec/version.h index 73a6f33..a1c8365 100644 --- a/libavcodec/version.h +++ b/libavcodec/version.h @@ -21,7 +21,7 @@ #define AVCODEC_VERSION_H #define LIBAVCODEC_VERSION_MAJOR 52 -#define LIBAVCODEC_VERSION_MINOR 115 +#define LIBAVCODEC_VERSION_MINOR 116 #define LIBAVCODEC_VERSION_MICRO 0 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ -- cgit v1.1 From c4549bd66a94bec1284dcca0da19e5038a82cbe3 Mon Sep 17 00:00:00 2001 From: Justin Ruggles Date: Mon, 28 Mar 2011 14:01:27 -0400 Subject: ac3enc: avoid memcpy() of exponents and baps in EXP_REUSE case by using exponent reference blocks. --- libavcodec/ac3enc.c | 34 +++++++++++++++++----------------- 1 file changed, 17 insertions(+), 17 deletions(-) (limited to 'libavcodec') diff --git a/libavcodec/ac3enc.c b/libavcodec/ac3enc.c index f7a366b..c641400 100644 --- a/libavcodec/ac3enc.c +++ b/libavcodec/ac3enc.c @@ -112,6 +112,7 @@ typedef struct AC3Block { uint8_t coeff_shift[AC3_MAX_CHANNELS]; ///< fixed-point coefficient shift values uint8_t new_rematrixing_strategy; ///< send new rematrixing flags in this block uint8_t rematrixing_flags[4]; ///< rematrixing flags + struct AC3Block *exp_ref_block[AC3_MAX_CHANNELS]; ///< reference blocks for EXP_REUSE } AC3Block; /** @@ -692,7 +693,7 @@ static void encode_exponents_blk_ch(uint8_t *exp, int nb_exps, int exp_strategy) static void encode_exponents(AC3EncodeContext *s) { int blk, blk1, ch; - uint8_t *exp, *exp1, *exp_strategy; + uint8_t *exp, *exp_strategy; int nb_coefs, num_reuse_blocks; for (ch = 0; ch < s->channels; ch++) { @@ -704,9 +705,13 @@ static void encode_exponents(AC3EncodeContext *s) while (blk < AC3_MAX_BLOCKS) { blk1 = blk + 1; - /* count the number of EXP_REUSE blocks after the current block */ - while (blk1 < AC3_MAX_BLOCKS && exp_strategy[blk1] == EXP_REUSE) + /* count the number of EXP_REUSE blocks after the current block + and set exponent reference block pointers */ + s->blocks[blk].exp_ref_block[ch] = &s->blocks[blk]; + while (blk1 < AC3_MAX_BLOCKS && exp_strategy[blk1] == EXP_REUSE) { + s->blocks[blk1].exp_ref_block[ch] = &s->blocks[blk]; blk1++; + } num_reuse_blocks = blk1 - blk - 1; /* for the EXP_REUSE case we select the min of the exponents */ @@ -714,15 +719,8 @@ static void encode_exponents(AC3EncodeContext *s) encode_exponents_blk_ch(exp, nb_coefs, exp_strategy[blk]); - /* copy encoded exponents for reuse case */ - exp1 = exp + AC3_MAX_COEFS; - while (blk < blk1-1) { - memcpy(exp1, exp, nb_coefs * sizeof(*exp)); - exp1 += AC3_MAX_COEFS; - blk++; - } + exp += AC3_MAX_COEFS * (num_reuse_blocks + 1); blk = blk1; - exp = exp1; } } } @@ -1035,7 +1033,7 @@ static int bit_alloc(AC3EncodeContext *s, int snr_offset) reset_block_bap(s); mantissa_bits = 0; for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { - AC3Block *block = &s->blocks[blk]; + AC3Block *block; // initialize grouped mantissa counts. these are set so that they are // padded to the next whole group size when bits are counted in // compute_mantissa_size_final @@ -1047,9 +1045,8 @@ static int bit_alloc(AC3EncodeContext *s, int snr_offset) blocks within a frame are the exponent values. We can take advantage of that by reusing the bit allocation pointers whenever we reuse exponents. */ - if (s->exp_strategy[ch][blk] == EXP_REUSE) { - memcpy(block->bap[ch], s->blocks[blk-1].bap[ch], AC3_MAX_COEFS); - } else { + block = s->blocks[blk].exp_ref_block[ch]; + if (s->exp_strategy[ch][blk] != EXP_REUSE) { ff_ac3_bit_alloc_calc_bap(block->mask[ch], block->psd[ch], 0, s->nb_coefs[ch], snr_offset, s->bit_alloc.floor, ff_ac3_bap_tab, @@ -1352,12 +1349,14 @@ static void quantize_mantissas(AC3EncodeContext *s) for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { AC3Block *block = &s->blocks[blk]; + AC3Block *ref_block; s->mant1_cnt = s->mant2_cnt = s->mant4_cnt = 0; s->qmant1_ptr = s->qmant2_ptr = s->qmant4_ptr = NULL; for (ch = 0; ch < s->channels; ch++) { + ref_block = block->exp_ref_block[ch]; quantize_mantissas_blk_ch(s, block->fixed_coef[ch], - block->exp[ch], block->bap[ch], + ref_block->exp[ch], ref_block->bap[ch], block->qmant[ch], s->nb_coefs[ch]); } } @@ -1516,9 +1515,10 @@ static void output_audio_block(AC3EncodeContext *s, int blk) /* mantissas */ for (ch = 0; ch < s->channels; ch++) { int b, q; + AC3Block *ref_block = block->exp_ref_block[ch]; for (i = 0; i < s->nb_coefs[ch]; i++) { q = block->qmant[ch][i]; - b = block->bap[ch][i]; + b = ref_block->bap[ch][i]; switch (b) { case 0: break; case 1: if (q != 128) put_bits(&s->pb, 5, q); break; -- cgit v1.1