diff options
-rwxr-xr-x | configure | 2 | ||||
-rw-r--r-- | libavcodec/aacdec.c | 43 | ||||
-rw-r--r-- | libavcodec/avcodec.h | 27 | ||||
-rw-r--r-- | libavcodec/h264.c | 8 | ||||
-rw-r--r-- | libavcodec/h264_cabac.c | 2 | ||||
-rw-r--r-- | libavcodec/lpc.c | 5 | ||||
-rw-r--r-- | libavcodec/nellymoserdec.c | 2 | ||||
-rw-r--r-- | libavcodec/utils.c | 44 | ||||
-rw-r--r-- | libavcodec/version.h | 4 | ||||
-rw-r--r-- | libavcodec/w32pthreads.h | 16 | ||||
-rw-r--r-- | libavcodec/wmalosslessdec.c | 5 | ||||
-rw-r--r-- | libavcodec/x86/h264_i386.h | 4 | ||||
-rw-r--r-- | libavformat/asfdec.c | 2 | ||||
-rw-r--r-- | libavformat/flacdec.c | 5 | ||||
-rw-r--r-- | libavformat/flv.h | 1 | ||||
-rw-r--r-- | libavformat/flvdec.c | 76 | ||||
-rw-r--r-- | libavformat/idroqdec.c | 27 | ||||
-rw-r--r-- | libavformat/internal.h | 8 | ||||
-rw-r--r-- | libavformat/network.c | 5 | ||||
-rw-r--r-- | libavformat/utils.c | 43 | ||||
-rw-r--r-- | libavutil/aes.c | 2 | ||||
-rw-r--r-- | libpostproc/postprocess_altivec_template.c | 365 | ||||
-rw-r--r-- | libpostproc/postprocess_template.c | 3 |
23 files changed, 450 insertions, 249 deletions
@@ -3311,6 +3311,7 @@ elif enabled gcc; then check_cflags -fno-tree-vectorize check_cflags -Werror=implicit-function-declaration check_cflags -Werror=missing-prototypes + check_cflags -Werror=declaration-after-statement elif enabled llvm_gcc; then check_cflags -mllvm -stack-alignment=16 elif enabled clang; then @@ -3406,6 +3407,7 @@ echo "postprocessing support ${postproc-no}" echo "new filter support ${avfilter-no}" echo "network support ${network-no}" echo "threading support ${thread_type-no}" +echo "safe bitstream reader ${safe_bitstream_reader-no}" echo "SDL support ${sdl-no}" echo "Sun medialib support ${mlib-no}" echo "libdxva2 enabled ${dxva2-no}" diff --git a/libavcodec/aacdec.c b/libavcodec/aacdec.c index 3183ff2..c9ca0a6 100644 --- a/libavcodec/aacdec.c +++ b/libavcodec/aacdec.c @@ -722,16 +722,13 @@ static void decode_ltp(AACContext *ac, LongTermPrediction *ltp, /** * Decode Individual Channel Stream info; reference: table 4.6. - * - * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information. */ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, - GetBitContext *gb, int common_window) + GetBitContext *gb) { if (get_bits1(gb)) { av_log(ac->avctx, AV_LOG_ERROR, "Reserved bit set.\n"); - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + return AVERROR_INVALIDDATA; } ics->window_sequence[1] = ics->window_sequence[0]; ics->window_sequence[0] = get_bits(gb, 2); @@ -766,13 +763,11 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, if (ics->predictor_present) { if (ac->m4ac.object_type == AOT_AAC_MAIN) { if (decode_prediction(ac, ics, gb)) { - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + return AVERROR_INVALIDDATA; } } else if (ac->m4ac.object_type == AOT_AAC_LC) { av_log(ac->avctx, AV_LOG_ERROR, "Prediction is not allowed in AAC-LC.\n"); - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + return AVERROR_INVALIDDATA; } else { if ((ics->ltp.present = get_bits(gb, 1))) decode_ltp(ac, &ics->ltp, gb, ics->max_sfb); @@ -784,8 +779,7 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics, av_log(ac->avctx, AV_LOG_ERROR, "Number of scalefactor bands in group (%d) exceeds limit (%d).\n", ics->max_sfb, ics->num_swb); - memset(ics, 0, sizeof(IndividualChannelStream)); - return -1; + return AVERROR_INVALIDDATA; } return 0; @@ -1390,8 +1384,8 @@ static int decode_ics(AACContext *ac, SingleChannelElement *sce, global_gain = get_bits(gb, 8); if (!common_window && !scale_flag) { - if (decode_ics_info(ac, ics, gb, 0) < 0) - return -1; + if (decode_ics_info(ac, ics, gb) < 0) + return AVERROR_INVALIDDATA; } if (decode_band_types(ac, sce->band_type, sce->band_type_run_end, gb, ics) < 0) @@ -1507,8 +1501,8 @@ static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe) common_window = get_bits1(gb); if (common_window) { - if (decode_ics_info(ac, &cpe->ch[0].ics, gb, 1)) - return -1; + if (decode_ics_info(ac, &cpe->ch[0].ics, gb)) + return AVERROR_INVALIDDATA; i = cpe->ch[1].ics.use_kb_window[0]; cpe->ch[1].ics = cpe->ch[0].ics; cpe->ch[1].ics.use_kb_window[1] = i; @@ -2282,12 +2276,31 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data, static int aac_decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, AVPacket *avpkt) { + AACContext *ac = avctx->priv_data; const uint8_t *buf = avpkt->data; int buf_size = avpkt->size; GetBitContext gb; int buf_consumed; int buf_offset; int err; + int new_extradata_size; + const uint8_t *new_extradata = av_packet_get_side_data(avpkt, + AV_PKT_DATA_NEW_EXTRADATA, + &new_extradata_size); + + if (new_extradata) { + av_free(avctx->extradata); + avctx->extradata = av_mallocz(new_extradata_size + + FF_INPUT_BUFFER_PADDING_SIZE); + if (!avctx->extradata) + return AVERROR(ENOMEM); + avctx->extradata_size = new_extradata_size; + memcpy(avctx->extradata, new_extradata, new_extradata_size); + if (decode_audio_specific_config(ac, ac->avctx, &ac->m4ac, + avctx->extradata, + avctx->extradata_size*8, 1) < 0) + return AVERROR_INVALIDDATA; + } init_get_bits(&gb, buf, buf_size * 8); diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h index 0b9405f..90719dc 100644 --- a/libavcodec/avcodec.h +++ b/libavcodec/avcodec.h @@ -792,6 +792,10 @@ typedef struct RcOverride{ */ #define CODEC_CAP_SLICE_THREADS 0x2000 /** + * Codec supports changed parameters at any point. + */ +#define CODEC_CAP_PARAM_CHANGE 0x4000 +/** * Codec is lossless. */ #define CODEC_CAP_LOSSLESS 0x80000000 @@ -877,6 +881,8 @@ typedef struct AVPanScan{ enum AVPacketSideDataType { AV_PKT_DATA_PALETTE, + AV_PKT_DATA_NEW_EXTRADATA, + AV_PKT_DATA_PARAM_CHANGE, }; typedef struct AVPacket { @@ -946,6 +952,27 @@ typedef struct AVPacket { #define AV_PKT_FLAG_CORRUPT 0x0002 ///< The packet content is corrupted /** + * An AV_PKT_DATA_PARAM_CHANGE side data packet is laid out as follows: + * u32le param_flags + * if (param_flags & AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_COUNT) + * s32le channel_count + * if (param_flags & AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_LAYOUT) + * u64le channel_layout + * if (param_flags & AV_SIDE_DATA_PARAM_CHANGE_SAMPLE_RATE) + * s32le sample_rate + * if (param_flags & AV_SIDE_DATA_PARAM_CHANGE_DIMENSIONS) + * s32le width + * s32le height + */ + +enum AVSideDataParamChangeFlags { + AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_COUNT = 0x0001, + AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_LAYOUT = 0x0002, + AV_SIDE_DATA_PARAM_CHANGE_SAMPLE_RATE = 0x0004, + AV_SIDE_DATA_PARAM_CHANGE_DIMENSIONS = 0x0008, +}; + +/** * Audio Video Frame. * New fields can be added to the end of AVFRAME with minor version * bumps. Similarly fields that are marked as to be only accessed by diff --git a/libavcodec/h264.c b/libavcodec/h264.c index 652a4e9..9978188 100644 --- a/libavcodec/h264.c +++ b/libavcodec/h264.c @@ -4195,13 +4195,13 @@ int main(void){ s= show_bits(&gb, 24); - START_TIMER + {START_TIMER j= get_ue_golomb(&gb); if(j != i){ printf("mismatch! at %d (%d should be %d) bits:%6X\n", i, j, i, s); // return -1; } - STOP_TIMER("get_ue_golomb"); + STOP_TIMER("get_ue_golomb");} } @@ -4220,13 +4220,13 @@ int main(void){ s= show_bits(&gb, 24); - START_TIMER + {START_TIMER j= get_se_golomb(&gb); if(j != i - COUNT/2){ printf("mismatch! at %d (%d should be %d) bits:%6X\n", i, j, i, s); // return -1; } - STOP_TIMER("get_se_golomb"); + STOP_TIMER("get_se_golomb");} } printf("Testing RBSP\n"); diff --git a/libavcodec/h264_cabac.c b/libavcodec/h264_cabac.c index 0b33e30..5482294 100644 --- a/libavcodec/h264_cabac.c +++ b/libavcodec/h264_cabac.c @@ -1657,7 +1657,7 @@ decode_cabac_residual_internal(H264Context *h, DCTELEM *block, index[coeff_count++] = last;\ } const uint8_t *sig_off = significant_coeff_flag_offset_8x8[MB_FIELD]; -#if ARCH_X86 && HAVE_7REGS && HAVE_EBX_AVAILABLE && !defined(BROKEN_RELOCATIONS) +#if ARCH_X86 && HAVE_EBX_AVAILABLE && !defined(BROKEN_RELOCATIONS) coeff_count= decode_significance_8x8_x86(CC, significant_coeff_ctx_base, index, last_coeff_ctx_base, sig_off); } else { diff --git a/libavcodec/lpc.c b/libavcodec/lpc.c index c272088..3401226 100644 --- a/libavcodec/lpc.c +++ b/libavcodec/lpc.c @@ -35,8 +35,9 @@ static void lpc_apply_welch_window_c(const int32_t *data, int len, double w; double c; - assert(!(len&1)); //the optimization in r11881 does not support odd len - //if someone wants odd len extend the change in r11881 + /* The optimization in commit fa4ed8c does not support odd len. + * If someone wants odd len extend that change. */ + assert(!(len & 1)); n2 = (len >> 1); c = 2.0 / (len - 1.0); diff --git a/libavcodec/nellymoserdec.c b/libavcodec/nellymoserdec.c index 3386b82..c2e7cda 100644 --- a/libavcodec/nellymoserdec.c +++ b/libavcodec/nellymoserdec.c @@ -226,7 +226,7 @@ AVCodec ff_nellymoser_decoder = { .init = decode_init, .close = decode_end, .decode = decode_tag, - .capabilities = CODEC_CAP_DR1, + .capabilities = CODEC_CAP_DR1 | CODEC_CAP_PARAM_CHANGE, .long_name = NULL_IF_CONFIG_SMALL("Nellymoser Asao"), .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, diff --git a/libavcodec/utils.c b/libavcodec/utils.c index 03e367f..7d30f76 100644 --- a/libavcodec/utils.c +++ b/libavcodec/utils.c @@ -41,6 +41,7 @@ #include "thread.h" #include "audioconvert.h" #include "internal.h" +#include "bytestream.h" #include <stdlib.h> #include <stdarg.h> #include <limits.h> @@ -1028,6 +1029,47 @@ int attribute_align_arg avcodec_decode_audio3(AVCodecContext *avctx, int16_t *sa } #endif +static void apply_param_change(AVCodecContext *avctx, AVPacket *avpkt) +{ + int size = 0; + const uint8_t *data; + uint32_t flags; + + if (!(avctx->codec->capabilities & CODEC_CAP_PARAM_CHANGE)) + return; + + data = av_packet_get_side_data(avpkt, AV_PKT_DATA_PARAM_CHANGE, &size); + if (!data || size < 4) + return; + flags = bytestream_get_le32(&data); + size -= 4; + if (size < 4) /* Required for any of the changes */ + return; + if (flags & AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_COUNT) { + avctx->channels = bytestream_get_le32(&data); + size -= 4; + } + if (flags & AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_LAYOUT) { + if (size < 8) + return; + avctx->channel_layout = bytestream_get_le64(&data); + size -= 8; + } + if (size < 4) + return; + if (flags & AV_SIDE_DATA_PARAM_CHANGE_SAMPLE_RATE) { + avctx->sample_rate = bytestream_get_le32(&data); + size -= 4; + } + if (flags & AV_SIDE_DATA_PARAM_CHANGE_DIMENSIONS) { + if (size < 8) + return; + avctx->width = bytestream_get_le32(&data); + avctx->height = bytestream_get_le32(&data); + size -= 8; + } +} + int attribute_align_arg avcodec_decode_audio4(AVCodecContext *avctx, AVFrame *frame, int *got_frame_ptr, @@ -1044,6 +1086,8 @@ int attribute_align_arg avcodec_decode_audio4(AVCodecContext *avctx, if ((avctx->codec->capabilities & CODEC_CAP_DELAY) || avpkt->size) { av_packet_split_side_data(avpkt); + apply_param_change(avctx, avpkt); + avctx->pkt = avpkt; ret = avctx->codec->decode(avctx, frame, got_frame_ptr, avpkt); if (ret >= 0 && *got_frame_ptr) { diff --git a/libavcodec/version.h b/libavcodec/version.h index 360068a..5d4807c 100644 --- a/libavcodec/version.h +++ b/libavcodec/version.h @@ -21,8 +21,8 @@ #define AVCODEC_VERSION_H #define LIBAVCODEC_VERSION_MAJOR 53 -#define LIBAVCODEC_VERSION_MINOR 46 -#define LIBAVCODEC_VERSION_MICRO 1 +#define LIBAVCODEC_VERSION_MINOR 47 +#define LIBAVCODEC_VERSION_MICRO 0 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ LIBAVCODEC_VERSION_MINOR, \ diff --git a/libavcodec/w32pthreads.h b/libavcodec/w32pthreads.h index 2d1470c..3cdbc2c 100644 --- a/libavcodec/w32pthreads.h +++ b/libavcodec/w32pthreads.h @@ -139,7 +139,7 @@ static void pthread_cond_init(pthread_cond_t *cond, const void *unused_attr) win32_cond->semaphore = CreateSemaphore(NULL, 0, 0x7fffffff, NULL); if (!win32_cond->semaphore) return; - win32_cond->waiters_done = CreateEvent(NULL, FALSE, FALSE, NULL); + win32_cond->waiters_done = CreateEvent(NULL, TRUE, FALSE, NULL); if (!win32_cond->waiters_done) return; @@ -204,11 +204,10 @@ static void pthread_cond_wait(pthread_cond_t *cond, pthread_mutex_t *mutex) /* non native condition variables */ pthread_mutex_lock(&win32_cond->mtx_broadcast); - pthread_mutex_unlock(&win32_cond->mtx_broadcast); - pthread_mutex_lock(&win32_cond->mtx_waiter_count); win32_cond->waiter_count++; pthread_mutex_unlock(&win32_cond->mtx_waiter_count); + pthread_mutex_unlock(&win32_cond->mtx_broadcast); // unlock the external mutex pthread_mutex_unlock(mutex); @@ -216,7 +215,7 @@ static void pthread_cond_wait(pthread_cond_t *cond, pthread_mutex_t *mutex) pthread_mutex_lock(&win32_cond->mtx_waiter_count); win32_cond->waiter_count--; - last_waiter = !win32_cond->waiter_count && win32_cond->is_broadcast; + last_waiter = !win32_cond->waiter_count || !win32_cond->is_broadcast; pthread_mutex_unlock(&win32_cond->mtx_waiter_count); if (last_waiter) @@ -235,13 +234,20 @@ static void pthread_cond_signal(pthread_cond_t *cond) return; } + pthread_mutex_lock(&win32_cond->mtx_broadcast); + /* non-native condition variables */ pthread_mutex_lock(&win32_cond->mtx_waiter_count); have_waiter = win32_cond->waiter_count; pthread_mutex_unlock(&win32_cond->mtx_waiter_count); - if (have_waiter) + if (have_waiter) { ReleaseSemaphore(win32_cond->semaphore, 1, NULL); + WaitForSingleObject(win32_cond->waiters_done, INFINITE); + ResetEvent(win32_cond->waiters_done); + } + + pthread_mutex_unlock(&win32_cond->mtx_broadcast); } static void w32thread_init(void) diff --git a/libavcodec/wmalosslessdec.c b/libavcodec/wmalosslessdec.c index 2b32bcc..0af01de 100644 --- a/libavcodec/wmalosslessdec.c +++ b/libavcodec/wmalosslessdec.c @@ -1338,6 +1338,7 @@ static int decode_packet(AVCodecContext *avctx, *data_size = 0; if (s->packet_done || s->packet_loss) { + int seekable_frame_in_packet, spliced_packet; s->packet_done = 0; /** sanity check for the buffer length */ @@ -1351,8 +1352,8 @@ static int decode_packet(AVCodecContext *avctx, /** parse packet header */ init_get_bits(gb, buf, s->buf_bit_size); packet_sequence_number = get_bits(gb, 4); - int seekable_frame_in_packet = get_bits1(gb); - int spliced_packet = get_bits1(gb); + seekable_frame_in_packet = get_bits1(gb); + spliced_packet = get_bits1(gb); /** get number of bits that need to be added to the previous frame */ num_bits_prev_frame = get_bits(gb, s->log2_frame_size); diff --git a/libavcodec/x86/h264_i386.h b/libavcodec/x86/h264_i386.h index 089762b..0bc6a61 100644 --- a/libavcodec/x86/h264_i386.h +++ b/libavcodec/x86/h264_i386.h @@ -36,7 +36,7 @@ //FIXME use some macros to avoid duplicating get_cabac (cannot be done yet //as that would make optimization work hard) -#if HAVE_6REGS && !defined(BROKEN_RELOCATIONS) +#if HAVE_EBX_AVAILABLE && !defined(BROKEN_RELOCATIONS) static int decode_significance_x86(CABACContext *c, int max_coeff, uint8_t *significant_coeff_ctx_base, int *index, x86_reg last_off){ @@ -144,6 +144,6 @@ static int decode_significance_8x8_x86(CABACContext *c, ); return coeff_count; } -#endif /* HAVE_6REGS && !defined(BROKEN_RELOCATIONS) */ +#endif /* HAVE_EBX_AVAILABLE && !defined(BROKEN_RELOCATIONS) */ #endif /* AVCODEC_X86_H264_I386_H */ diff --git a/libavformat/asfdec.c b/libavformat/asfdec.c index 0bb210b..5af112d 100644 --- a/libavformat/asfdec.c +++ b/libavformat/asfdec.c @@ -987,7 +987,7 @@ static int ff_asf_parse_packet(AVFormatContext *s, AVIOContext *pb, AVPacket *pk asf_st->packet_pos= asf->packet_pos; if (asf_st->pkt.data && asf_st->palette_changed) { uint8_t *pal; - pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, + pal = av_packet_new_side_data(&asf_st->pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE); if (!pal) { av_log(s, AV_LOG_ERROR, "Cannot append palette to packet\n"); diff --git a/libavformat/flacdec.c b/libavformat/flacdec.c index 163d2fa..070f9af 100644 --- a/libavformat/flacdec.c +++ b/libavformat/flacdec.c @@ -102,7 +102,7 @@ static int flac_read_header(AVFormatContext *s, uint8_t isrc[13]; uint64_t start; const uint8_t *offset; - int i, j, chapters, track, ti; + int i, chapters, track, ti; if (metadata_size < 431) return AVERROR_INVALIDDATA; offset = buffer + 395; @@ -119,8 +119,7 @@ static int flac_read_header(AVFormatContext *s, offset += 14; ti = bytestream_get_byte(&offset); if (ti <= 0) return AVERROR_INVALIDDATA; - for (j = 0; j < ti; j++) - offset += 12; + offset += ti * 12; avpriv_new_chapter(s, track, st->time_base, start, AV_NOPTS_VALUE, isrc); } } else { diff --git a/libavformat/flv.h b/libavformat/flv.h index 2165357..ef8eaab 100644 --- a/libavformat/flv.h +++ b/libavformat/flv.h @@ -66,6 +66,7 @@ enum { FLV_STREAM_TYPE_VIDEO, FLV_STREAM_TYPE_AUDIO, FLV_STREAM_TYPE_DATA, + FLV_STREAM_TYPE_NB, }; enum { diff --git a/libavformat/flvdec.c b/libavformat/flvdec.c index 3494529..765eee5 100644 --- a/libavformat/flvdec.c +++ b/libavformat/flvdec.c @@ -37,6 +37,10 @@ typedef struct { int wrong_dts; ///< wrong dts due to negative cts + uint8_t *new_extradata[FLV_STREAM_TYPE_NB]; + int new_extradata_size[FLV_STREAM_TYPE_NB]; + int last_sample_rate; + int last_channels; } FLVContext; static int flv_probe(AVProbeData *p) @@ -50,8 +54,7 @@ static int flv_probe(AVProbeData *p) return 0; } -static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream, int flv_codecid) { - AVCodecContext *acodec = astream->codec; +static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream, AVCodecContext *acodec, int flv_codecid) { switch(flv_codecid) { //no distinction between S16 and S8 PCM codec flags case FLV_CODECID_PCM: @@ -411,6 +414,15 @@ static int flv_read_header(AVFormatContext *s, return 0; } +static int flv_read_close(AVFormatContext *s) +{ + int i; + FLVContext *flv = s->priv_data; + for(i=0; i<FLV_STREAM_TYPE_NB; i++) + av_freep(&flv->new_extradata[i]); + return 0; +} + static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size) { av_free(st->codec->extradata); @@ -422,6 +434,18 @@ static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size) return 0; } +static int flv_queue_extradata(FLVContext *flv, AVIOContext *pb, int stream, + int size) +{ + av_free(flv->new_extradata[stream]); + flv->new_extradata[stream] = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE); + if (!flv->new_extradata[stream]) + return AVERROR(ENOMEM); + flv->new_extradata_size[stream] = size; + avio_read(pb, flv->new_extradata[stream], size); + return 0; +} + static int flv_read_packet(AVFormatContext *s, AVPacket *pkt) { FLVContext *flv = s->priv_data; @@ -429,6 +453,7 @@ static int flv_read_packet(AVFormatContext *s, AVPacket *pkt) int stream_type=-1; int64_t next, pos; int64_t dts, pts = AV_NOPTS_VALUE; + int sample_rate, channels; AVStream *st = NULL; for(;;avio_skip(s->pb, 4)){ /* pkt size is repeated at end. skip it */ @@ -518,13 +543,24 @@ static int flv_read_packet(AVFormatContext *s, AVPacket *pkt) } if(stream_type == FLV_STREAM_TYPE_AUDIO){ + int bits_per_coded_sample; + channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1; + sample_rate = (44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3); + bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8; if(!st->codec->channels || !st->codec->sample_rate || !st->codec->bits_per_coded_sample) { - st->codec->channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1; - st->codec->sample_rate = (44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3); - st->codec->bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8; + st->codec->channels = channels; + st->codec->sample_rate = sample_rate; + st->codec->bits_per_coded_sample = bits_per_coded_sample; } if(!st->codec->codec_id){ - flv_set_audio_codec(s, st, flags & FLV_AUDIO_CODECID_MASK); + flv_set_audio_codec(s, st, st->codec, flags & FLV_AUDIO_CODECID_MASK); + flv->last_sample_rate = st->codec->sample_rate; + flv->last_channels = st->codec->channels; + } else { + AVCodecContext ctx; + ctx.sample_rate = sample_rate; + flv_set_audio_codec(s, st, &ctx, flags & FLV_AUDIO_CODECID_MASK); + sample_rate = ctx.sample_rate; } } else if(stream_type == FLV_STREAM_TYPE_VIDEO) { size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK); @@ -545,8 +581,13 @@ static int flv_read_packet(AVFormatContext *s, AVPacket *pkt) if (flv->wrong_dts) dts = AV_NOPTS_VALUE; } - - if (type == 0 && !st->codec->extradata) { + if (type == 0 && (!st->codec->extradata || st->codec->codec_id != CODEC_ID_H264)) { + if (st->codec->extradata) { + if ((ret = flv_queue_extradata(flv, s->pb, stream_type, size)) < 0) + return ret; + ret = AVERROR(EAGAIN); + goto leave; + } if ((ret = flv_get_extradata(s, st, size)) < 0) return ret; if (st->codec->codec_id == CODEC_ID_AAC) { @@ -583,8 +624,22 @@ static int flv_read_packet(AVFormatContext *s, AVPacket *pkt) pkt->dts = dts; pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts; pkt->stream_index = st->index; - if(st->codec->codec_id == CODEC_ID_NELLYMOSER) - av_packet_new_side_data(pkt, 'F', 1)[0]= flags; + if (flv->new_extradata[stream_type]) { + uint8_t *side = av_packet_new_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA, + flv->new_extradata_size[stream_type]); + if (side) { + memcpy(side, flv->new_extradata[stream_type], + flv->new_extradata_size[stream_type]); + av_freep(&flv->new_extradata[stream_type]); + flv->new_extradata_size[stream_type] = 0; + } + } + if (stream_type == FLV_STREAM_TYPE_AUDIO && (sample_rate != flv->last_sample_rate || + channels != flv->last_channels)) { + flv->last_sample_rate = sample_rate; + flv->last_channels = channels; + ff_add_param_change(pkt, channels, 0, sample_rate, 0, 0); + } if ( stream_type == FLV_STREAM_TYPE_AUDIO || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY) || @@ -640,6 +695,7 @@ AVInputFormat ff_flv_demuxer = { #if 0 .read_seek2 = flv_read_seek2, #endif + .read_close = flv_read_close, .extensions = "flv", .value = CODEC_ID_FLV1, }; diff --git a/libavformat/idroqdec.c b/libavformat/idroqdec.c index 9ac32eb..fffee9d 100644 --- a/libavformat/idroqdec.c +++ b/libavformat/idroqdec.c @@ -45,6 +45,7 @@ typedef struct RoqDemuxContext { + int frame_rate; int width; int height; int audio_channels; @@ -71,29 +72,21 @@ static int roq_read_header(AVFormatContext *s, { RoqDemuxContext *roq = s->priv_data; AVIOContext *pb = s->pb; - int framerate; - AVStream *st; unsigned char preamble[RoQ_CHUNK_PREAMBLE_SIZE]; /* get the main header */ if (avio_read(pb, preamble, RoQ_CHUNK_PREAMBLE_SIZE) != RoQ_CHUNK_PREAMBLE_SIZE) return AVERROR(EIO); - framerate = AV_RL16(&preamble[6]); + roq->frame_rate = AV_RL16(&preamble[6]); /* init private context parameters */ roq->width = roq->height = roq->audio_channels = roq->video_pts = roq->audio_frame_count = 0; roq->audio_stream_index = -1; + roq->video_stream_index = -1; - st = avformat_new_stream(s, NULL); - if (!st) - return AVERROR(ENOMEM); - avpriv_set_pts_info(st, 63, 1, framerate); - roq->video_stream_index = st->index; - st->codec->codec_type = AVMEDIA_TYPE_VIDEO; - st->codec->codec_id = CODEC_ID_ROQ; - st->codec->codec_tag = 0; /* no fourcc */ + s->ctx_flags |= AVFMTCTX_NOHEADER; return 0; } @@ -131,8 +124,16 @@ static int roq_read_packet(AVFormatContext *s, switch (chunk_type) { case RoQ_INFO: - if (!roq->width || !roq->height) { - AVStream *st = s->streams[roq->video_stream_index]; + if (roq->video_stream_index == -1) { + AVStream *st = avformat_new_stream(s, NULL); + if (!st) + return AVERROR(ENOMEM); + avpriv_set_pts_info(st, 63, 1, roq->frame_rate); + roq->video_stream_index = st->index; + st->codec->codec_type = AVMEDIA_TYPE_VIDEO; + st->codec->codec_id = CODEC_ID_ROQ; + st->codec->codec_tag = 0; /* no fourcc */ + if (avio_read(pb, preamble, RoQ_CHUNK_PREAMBLE_SIZE) != RoQ_CHUNK_PREAMBLE_SIZE) return AVERROR(EIO); st->codec->width = roq->width = AV_RL16(preamble); diff --git a/libavformat/internal.h b/libavformat/internal.h index c3be0df..c8376f7 100644 --- a/libavformat/internal.h +++ b/libavformat/internal.h @@ -298,4 +298,12 @@ int64_t ff_gen_search(AVFormatContext *s, int stream_index, void avpriv_set_pts_info(AVStream *s, int pts_wrap_bits, unsigned int pts_num, unsigned int pts_den); +/** + * Add side data to a packet for changing parameters to the given values. + * Parameters set to 0 aren't included in the change. + */ +int ff_add_param_change(AVPacket *pkt, int32_t channels, + uint64_t channel_layout, int32_t sample_rate, + int32_t width, int32_t height); + #endif /* AVFORMAT_INTERNAL_H */ diff --git a/libavformat/network.c b/libavformat/network.c index e27c64d..9beaaaf 100644 --- a/libavformat/network.c +++ b/libavformat/network.c @@ -124,13 +124,16 @@ int ff_network_inited_globally; int ff_network_init(void) { +#if HAVE_WINSOCK2_H + WSADATA wsaData; +#endif + if (!ff_network_inited_globally) av_log(NULL, AV_LOG_WARNING, "Using network protocols without global " "network initialization. Please use " "avformat_network_init(), this will " "become mandatory later.\n"); #if HAVE_WINSOCK2_H - WSADATA wsaData; if (WSAStartup(MAKEWORD(1,1), &wsaData)) return 0; #endif diff --git a/libavformat/utils.c b/libavformat/utils.c index edcae2f..65b1fe3 100644 --- a/libavformat/utils.c +++ b/libavformat/utils.c @@ -26,6 +26,7 @@ #include "internal.h" #include "libavcodec/internal.h" #include "libavcodec/raw.h" +#include "libavcodec/bytestream.h" #include "libavutil/opt.h" #include "libavutil/dict.h" #include "libavutil/pixdesc.h" @@ -4290,3 +4291,45 @@ int avformat_network_deinit(void) #endif return 0; } + +int ff_add_param_change(AVPacket *pkt, int32_t channels, + uint64_t channel_layout, int32_t sample_rate, + int32_t width, int32_t height) +{ + uint32_t flags = 0; + int size = 4; + uint8_t *data; + if (!pkt) + return AVERROR(EINVAL); + if (channels) { + size += 4; + flags |= AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_COUNT; + } + if (channel_layout) { + size += 8; + flags |= AV_SIDE_DATA_PARAM_CHANGE_CHANNEL_LAYOUT; + } + if (sample_rate) { + size += 4; + flags |= AV_SIDE_DATA_PARAM_CHANGE_SAMPLE_RATE; + } + if (width || height) { + size += 8; + flags |= AV_SIDE_DATA_PARAM_CHANGE_DIMENSIONS; + } + data = av_packet_new_side_data(pkt, AV_PKT_DATA_PARAM_CHANGE, size); + if (!data) + return AVERROR(ENOMEM); + bytestream_put_le32(&data, flags); + if (channels) + bytestream_put_le32(&data, channels); + if (channel_layout) + bytestream_put_le64(&data, channel_layout); + if (sample_rate) + bytestream_put_le32(&data, sample_rate); + if (width || height) { + bytestream_put_le32(&data, width); + bytestream_put_le32(&data, height); + } + return 0; +} diff --git a/libavutil/aes.c b/libavutil/aes.c index 46b290c..7950902 100644 --- a/libavutil/aes.c +++ b/libavutil/aes.c @@ -33,7 +33,7 @@ typedef union { typedef struct AVAES { // Note: round_key[16] is accessed in the init code, but this only - // overwrites state, which does not matter (see also r7471). + // overwrites state, which does not matter (see also commit ba554c0). av_aes_block round_key[15]; av_aes_block state[2]; int rounds; diff --git a/libpostproc/postprocess_altivec_template.c b/libpostproc/postprocess_altivec_template.c index c53ebc0..3a37562 100644 --- a/libpostproc/postprocess_altivec_template.c +++ b/libpostproc/postprocess_altivec_template.c @@ -530,6 +530,39 @@ static inline void doVertDefFilter_altivec(uint8_t src[], int stride, PPContext } static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { + const vector signed int vsint32_8 = vec_splat_s32(8); + const vector unsigned int vuint32_4 = vec_splat_u32(4); + const vector signed char neg1 = vec_splat_s8(-1); + + const vector unsigned char permA1 = (vector unsigned char) + {0x00, 0x01, 0x02, 0x10, 0x11, 0x12, 0x1F, 0x1F, + 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F}; + const vector unsigned char permA2 = (vector unsigned char) + {0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x10, 0x11, + 0x12, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F}; + const vector unsigned char permA1inc = (vector unsigned char) + {0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; + const vector unsigned char permA2inc = (vector unsigned char) + {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x01, + 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; + const vector unsigned char magic = (vector unsigned char) + {0x01, 0x02, 0x01, 0x02, 0x04, 0x02, 0x01, 0x02, + 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; + const vector unsigned char extractPerm = (vector unsigned char) + {0x10, 0x10, 0x10, 0x01, 0x10, 0x10, 0x10, 0x01, + 0x10, 0x10, 0x10, 0x01, 0x10, 0x10, 0x10, 0x01}; + const vector unsigned char extractPermInc = (vector unsigned char) + {0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, + 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01}; + const vector unsigned char identity = vec_lvsl(0,(unsigned char *)0); + const vector unsigned char tenRight = (vector unsigned char) + {0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; + const vector unsigned char eightLeft = (vector unsigned char) + {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x08}; + /* this code makes no assumption on src or stride. One could remove the recomputation of the perm @@ -539,11 +572,9 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { src & stride :-( */ uint8_t *srcCopy = src; - DECLARE_ALIGNED(16, uint8_t, dt)[16]; + DECLARE_ALIGNED(16, uint8_t, dt)[16] = { deringThreshold }; const vector signed int zero = vec_splat_s32(0); - vector unsigned char v_dt; - dt[0] = deringThreshold; - v_dt = vec_splat(vec_ld(0, dt), 0); + vector unsigned char v_dt = vec_splat(vec_ld(0, dt), 0); #define LOAD_LINE(i) \ const vector unsigned char perm##i = \ @@ -565,6 +596,11 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { #undef LOAD_LINE vector unsigned char v_avg; + DECLARE_ALIGNED(16, signed int, S)[8]; + DECLARE_ALIGNED(16, int, tQP2)[4] = { c->QP/2 + 1 }; + vector signed int vQP2 = vec_ld(0, tQP2); + vQP2 = vec_splat(vQP2, 0); + { const vector unsigned char trunc_perm = (vector unsigned char) {0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, @@ -575,21 +611,22 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { const vector unsigned char trunc_src78 = vec_perm(src7, src8, trunc_perm); #define EXTRACT(op) do { \ - const vector unsigned char s##op##_1 = vec_##op(trunc_src12, trunc_src34); \ - const vector unsigned char s##op##_2 = vec_##op(trunc_src56, trunc_src78); \ - const vector unsigned char s##op##_6 = vec_##op(s##op##_1, s##op##_2); \ - const vector unsigned char s##op##_8h = vec_mergeh(s##op##_6, s##op##_6); \ - const vector unsigned char s##op##_8l = vec_mergel(s##op##_6, s##op##_6); \ - const vector unsigned char s##op##_9 = vec_##op(s##op##_8h, s##op##_8l); \ - const vector unsigned char s##op##_9h = vec_mergeh(s##op##_9, s##op##_9); \ - const vector unsigned char s##op##_9l = vec_mergel(s##op##_9, s##op##_9); \ - const vector unsigned char s##op##_10 = vec_##op(s##op##_9h, s##op##_9l); \ - const vector unsigned char s##op##_10h = vec_mergeh(s##op##_10, s##op##_10); \ - const vector unsigned char s##op##_10l = vec_mergel(s##op##_10, s##op##_10); \ - const vector unsigned char s##op##_11 = vec_##op(s##op##_10h, s##op##_10l); \ - const vector unsigned char s##op##_11h = vec_mergeh(s##op##_11, s##op##_11); \ - const vector unsigned char s##op##_11l = vec_mergel(s##op##_11, s##op##_11); \ - v_##op = vec_##op(s##op##_11h, s##op##_11l); } while (0) + const vector unsigned char s_1 = vec_##op(trunc_src12, trunc_src34); \ + const vector unsigned char s_2 = vec_##op(trunc_src56, trunc_src78); \ + const vector unsigned char s_6 = vec_##op(s_1, s_2); \ + const vector unsigned char s_8h = vec_mergeh(s_6, s_6); \ + const vector unsigned char s_8l = vec_mergel(s_6, s_6); \ + const vector unsigned char s_9 = vec_##op(s_8h, s_8l); \ + const vector unsigned char s_9h = vec_mergeh(s_9, s_9); \ + const vector unsigned char s_9l = vec_mergel(s_9, s_9); \ + const vector unsigned char s_10 = vec_##op(s_9h, s_9l); \ + const vector unsigned char s_10h = vec_mergeh(s_10, s_10); \ + const vector unsigned char s_10l = vec_mergel(s_10, s_10); \ + const vector unsigned char s_11 = vec_##op(s_10h, s_10l); \ + const vector unsigned char s_11h = vec_mergeh(s_11, s_11); \ + const vector unsigned char s_11l = vec_mergel(s_11, s_11); \ + v_##op = vec_##op(s_11h, s_11l); \ +} while (0) vector unsigned char v_min; vector unsigned char v_max; @@ -603,7 +640,6 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { v_avg = vec_avg(v_min, v_max); } - DECLARE_ALIGNED(16, signed int, S)[8]; { const vector unsigned short mask1 = (vector unsigned short) {0x0001, 0x0002, 0x0004, 0x0008, @@ -615,22 +651,27 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { const vector unsigned int vuint32_16 = vec_sl(vec_splat_u32(1), vec_splat_u32(4)); const vector unsigned int vuint32_1 = vec_splat_u32(1); + vector signed int sumA2; + vector signed int sumB2; + vector signed int sum0, sum1, sum2, sum3, sum4; + vector signed int sum5, sum6, sum7, sum8, sum9; + #define COMPARE(i) \ - vector signed int sum##i; \ do { \ - const vector unsigned char cmp##i = \ + const vector unsigned char cmp = \ (vector unsigned char)vec_cmpgt(src##i, v_avg); \ - const vector unsigned short cmpHi##i = \ - (vector unsigned short)vec_mergeh(cmp##i, cmp##i); \ - const vector unsigned short cmpLi##i = \ - (vector unsigned short)vec_mergel(cmp##i, cmp##i); \ - const vector signed short cmpHf##i = \ - (vector signed short)vec_and(cmpHi##i, mask1); \ - const vector signed short cmpLf##i = \ - (vector signed short)vec_and(cmpLi##i, mask2); \ - const vector signed int sump##i = vec_sum4s(cmpHf##i, zero); \ - const vector signed int sumq##i = vec_sum4s(cmpLf##i, sump##i); \ - sum##i = vec_sums(sumq##i, zero); } while (0) + const vector unsigned short cmpHi = \ + (vector unsigned short)vec_mergeh(cmp, cmp); \ + const vector unsigned short cmpLi = \ + (vector unsigned short)vec_mergel(cmp, cmp); \ + const vector signed short cmpHf = \ + (vector signed short)vec_and(cmpHi, mask1); \ + const vector signed short cmpLf = \ + (vector signed short)vec_and(cmpLi, mask2); \ + const vector signed int sump = vec_sum4s(cmpHf, zero); \ + const vector signed int sumq = vec_sum4s(cmpLf, sump); \ + sum##i = vec_sums(sumq, zero); \ + } while (0) COMPARE(0); COMPARE(1); @@ -644,8 +685,6 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { COMPARE(9); #undef COMPARE - vector signed int sumA2; - vector signed int sumB2; { const vector signed int sump02 = vec_mergel(sum0, sum2); const vector signed int sump13 = vec_mergel(sum1, sum3); @@ -699,86 +738,43 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { /* I'm not sure the following is actually faster than straight, unvectorized C code :-( */ - DECLARE_ALIGNED(16, int, tQP2)[4]; - tQP2[0]= c->QP/2 + 1; - vector signed int vQP2 = vec_ld(0, tQP2); - vQP2 = vec_splat(vQP2, 0); - const vector signed int vsint32_8 = vec_splat_s32(8); - const vector unsigned int vuint32_4 = vec_splat_u32(4); - - const vector unsigned char permA1 = (vector unsigned char) - {0x00, 0x01, 0x02, 0x10, 0x11, 0x12, 0x1F, 0x1F, - 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F}; - const vector unsigned char permA2 = (vector unsigned char) - {0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x10, 0x11, - 0x12, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F}; - const vector unsigned char permA1inc = (vector unsigned char) - {0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x00, 0x00, - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; - const vector unsigned char permA2inc = (vector unsigned char) - {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x01, - 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; - const vector unsigned char magic = (vector unsigned char) - {0x01, 0x02, 0x01, 0x02, 0x04, 0x02, 0x01, 0x02, - 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; - const vector unsigned char extractPerm = (vector unsigned char) - {0x10, 0x10, 0x10, 0x01, 0x10, 0x10, 0x10, 0x01, - 0x10, 0x10, 0x10, 0x01, 0x10, 0x10, 0x10, 0x01}; - const vector unsigned char extractPermInc = (vector unsigned char) - {0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, - 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01}; - const vector unsigned char identity = vec_lvsl(0,(unsigned char *)0); - const vector unsigned char tenRight = (vector unsigned char) - {0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}; - const vector unsigned char eightLeft = (vector unsigned char) - {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x08}; - - -#define F_INIT(i) \ - vector unsigned char tenRightM##i = tenRight; \ - vector unsigned char permA1M##i = permA1; \ - vector unsigned char permA2M##i = permA2; \ - vector unsigned char extractPermM##i = extractPerm +#define F_INIT() \ + vector unsigned char tenRightM = tenRight; \ + vector unsigned char permA1M = permA1; \ + vector unsigned char permA2M = permA2; \ + vector unsigned char extractPermM = extractPerm #define F2(i, j, k, l) \ if (S[i] & (1 << (l+1))) { \ - const vector unsigned char a_##j##_A##l = \ - vec_perm(src##i, src##j, permA1M##i); \ - const vector unsigned char a_##j##_B##l = \ - vec_perm(a_##j##_A##l, src##k, permA2M##i); \ - const vector signed int a_##j##_sump##l = \ - (vector signed int)vec_msum(a_##j##_B##l, magic, \ - (vector unsigned int)zero); \ - vector signed int F_##j##_##l = \ - vec_sr(vec_sums(a_##j##_sump##l, vsint32_8), vuint32_4); \ - F_##j##_##l = vec_splat(F_##j##_##l, 3); \ - const vector signed int p_##j##_##l = \ - (vector signed int)vec_perm(src##j, \ - (vector unsigned char)zero, \ - extractPermM##i); \ - const vector signed int sum_##j##_##l = vec_add( p_##j##_##l, vQP2);\ - const vector signed int diff_##j##_##l = vec_sub( p_##j##_##l, vQP2);\ - vector signed int newpm_##j##_##l; \ - if (vec_all_lt(sum_##j##_##l, F_##j##_##l)) \ - newpm_##j##_##l = sum_##j##_##l; \ - else if (vec_all_gt(diff_##j##_##l, F_##j##_##l)) \ - newpm_##j##_##l = diff_##j##_##l; \ - else newpm_##j##_##l = F_##j##_##l; \ - const vector unsigned char newpm2_##j##_##l = \ - vec_splat((vector unsigned char)newpm_##j##_##l, 15); \ - const vector unsigned char mask##j##l = vec_add(identity, \ - tenRightM##i); \ - src##j = vec_perm(src##j, newpm2_##j##_##l, mask##j##l); \ + const vector unsigned char a_A = vec_perm(src##i, src##j, permA1M); \ + const vector unsigned char a_B = vec_perm(a_A, src##k, permA2M); \ + const vector signed int a_sump = \ + (vector signed int)vec_msum(a_B, magic, (vector unsigned int)zero);\ + vector signed int F = vec_sr(vec_sums(a_sump, vsint32_8), vuint32_4); \ + const vector signed int p = \ + (vector signed int)vec_perm(src##j, (vector unsigned char)zero, \ + extractPermM); \ + const vector signed int sum = vec_add(p, vQP2); \ + const vector signed int diff = vec_sub(p, vQP2); \ + vector signed int newpm; \ + vector unsigned char newpm2, mask; \ + F = vec_splat(F, 3); \ + if (vec_all_lt(sum, F)) \ + newpm = sum; \ + else if (vec_all_gt(diff, F)) \ + newpm = diff; \ + else newpm = F; \ + newpm2 = vec_splat((vector unsigned char)newpm, 15); \ + mask = vec_add(identity, tenRightM); \ + src##j = vec_perm(src##j, newpm2, mask); \ } \ - permA1M##i = vec_add(permA1M##i, permA1inc); \ - permA2M##i = vec_add(permA2M##i, permA2inc); \ - tenRightM##i = vec_sro(tenRightM##i, eightLeft); \ - extractPermM##i = vec_add(extractPermM##i, extractPermInc) + permA1M = vec_add(permA1M, permA1inc); \ + permA2M = vec_add(permA2M, permA2inc); \ + tenRightM = vec_sro(tenRightM, eightLeft); \ + extractPermM = vec_add(extractPermM, extractPermInc) -#define ITER(i, j, k) \ - F_INIT(i); \ +#define ITER(i, j, k) do { \ + F_INIT(); \ F2(i, j, k, 0); \ F2(i, j, k, 1); \ F2(i, j, k, 2); \ @@ -786,7 +782,8 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { F2(i, j, k, 4); \ F2(i, j, k, 5); \ F2(i, j, k, 6); \ - F2(i, j, k, 7) + F2(i, j, k, 7); \ +} while (0) ITER(0, 1, 2); ITER(1, 2, 3); @@ -797,19 +794,18 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { ITER(6, 7, 8); ITER(7, 8, 9); - const vector signed char neg1 = vec_splat_s8(-1); - -#define STORE_LINE(i) \ - const vector unsigned char permST##i = \ +#define STORE_LINE(i) do { \ + const vector unsigned char permST = \ vec_lvsr(i * stride, srcCopy); \ - const vector unsigned char maskST##i = \ + const vector unsigned char maskST = \ vec_perm((vector unsigned char)zero, \ - (vector unsigned char)neg1, permST##i);\ - src##i = vec_perm(src##i ,src##i, permST##i); \ - sA##i= vec_sel(sA##i, src##i, maskST##i); \ - sB##i= vec_sel(src##i, sB##i, maskST##i); \ + (vector unsigned char)neg1, permST); \ + src##i = vec_perm(src##i ,src##i, permST); \ + sA##i= vec_sel(sA##i, src##i, maskST); \ + sB##i= vec_sel(src##i, sB##i, maskST); \ vec_st(sA##i, i * stride, srcCopy); \ - vec_st(sB##i, i * stride + 16, srcCopy) + vec_st(sB##i, i * stride + 16, srcCopy); \ +} while (0) STORE_LINE(1); STORE_LINE(2); @@ -832,16 +828,16 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride, uint8_t *tempBlurred, uint32_t *tempBlurredPast, int *maxNoise) { + const vector signed char neg1 = vec_splat_s8(-1); + const vector unsigned char permHH = (const vector unsigned char){0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F}; + const vector signed int zero = vec_splat_s32(0); const vector signed short vsint16_1 = vec_splat_s16(1); vector signed int v_dp = zero; vector signed int v_sysdp = zero; int d, sysd, i; - tempBlurredPast[127]= maxNoise[0]; - tempBlurredPast[128]= maxNoise[1]; - tempBlurredPast[129]= maxNoise[2]; - #define LOAD_LINE(src, i) \ register int j##src##i = i * stride; \ vector unsigned char perm##src##i = vec_lvsl(j##src##i, src); \ @@ -872,11 +868,12 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride, LOAD_LINE(tempBlurred, 7); #undef LOAD_LINE -#define ACCUMULATE_DIFFS(i) \ - vector signed short v_d##i = vec_sub(v_tempBlurredAss##i, \ - v_srcAss##i); \ - v_dp = vec_msums(v_d##i, v_d##i, v_dp); \ - v_sysdp = vec_msums(v_d##i, vsint16_1, v_sysdp) +#define ACCUMULATE_DIFFS(i) do { \ + vector signed short v_d = vec_sub(v_tempBlurredAss##i, \ + v_srcAss##i); \ + v_dp = vec_msums(v_d, v_d, v_dp); \ + v_sysdp = vec_msums(v_d, vsint16_1, v_sysdp); \ + } while (0) ACCUMULATE_DIFFS(0); ACCUMULATE_DIFFS(1); @@ -888,6 +885,10 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride, ACCUMULATE_DIFFS(7); #undef ACCUMULATE_DIFFS + tempBlurredPast[127]= maxNoise[0]; + tempBlurredPast[128]= maxNoise[1]; + tempBlurredPast[129]= maxNoise[2]; + v_dp = vec_sums(v_dp, zero); v_sysdp = vec_sums(v_sysdp, zero); @@ -938,13 +939,12 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride, const vector signed short vsint16_4 = vec_splat_s16(4); const vector unsigned short vuint16_3 = vec_splat_u16(3); -#define OP(i) \ - const vector signed short v_temp##i = \ - vec_mladd(v_tempBlurredAss##i, \ - vsint16_7, v_srcAss##i); \ - const vector signed short v_temp2##i = \ - vec_add(v_temp##i, vsint16_4); \ - v_tempBlurredAss##i = vec_sr(v_temp2##i, vuint16_3) +#define OP(i) do { \ + const vector signed short v_temp = \ + vec_mladd(v_tempBlurredAss##i, vsint16_7, v_srcAss##i); \ + const vector signed short v_temp2 = vec_add(v_temp, vsint16_4); \ + v_tempBlurredAss##i = vec_sr(v_temp2, vuint16_3); \ + } while (0) OP(0); OP(1); @@ -959,13 +959,13 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride, const vector signed short vsint16_3 = vec_splat_s16(3); const vector signed short vsint16_2 = vec_splat_s16(2); -#define OP(i) \ - const vector signed short v_temp##i = \ - vec_mladd(v_tempBlurredAss##i, \ - vsint16_3, v_srcAss##i); \ - const vector signed short v_temp2##i = \ - vec_add(v_temp##i, vsint16_2); \ - v_tempBlurredAss##i = vec_sr(v_temp2##i, (vector unsigned short)vsint16_2) +#define OP(i) do { \ + const vector signed short v_temp = \ + vec_mladd(v_tempBlurredAss##i, vsint16_3, v_srcAss##i); \ + const vector signed short v_temp2 = vec_add(v_temp, vsint16_2); \ + v_tempBlurredAss##i = \ + vec_sr(v_temp2, (vector unsigned short)vsint16_2); \ + } while (0) OP(0); OP(1); @@ -979,27 +979,19 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride, } } - const vector signed char neg1 = vec_splat_s8(-1); - const vector unsigned char permHH = (const vector unsigned char){0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F}; - -#define PACK_AND_STORE(src, i) \ - const vector unsigned char perms##src##i = \ - vec_lvsr(i * stride, src); \ - const vector unsigned char vf##src##i = \ - vec_packsu(v_tempBlurredAss##i, (vector signed short)zero); \ - const vector unsigned char vg##src##i = \ - vec_perm(vf##src##i, v_##src##A##i, permHH); \ - const vector unsigned char mask##src##i = \ - vec_perm((vector unsigned char)zero, (vector unsigned char)neg1, perms##src##i); \ - const vector unsigned char vg2##src##i = \ - vec_perm(vg##src##i, vg##src##i, perms##src##i); \ - const vector unsigned char svA##src##i = \ - vec_sel(v_##src##A1##i, vg2##src##i, mask##src##i); \ - const vector unsigned char svB##src##i = \ - vec_sel(vg2##src##i, v_##src##A2##i, mask##src##i); \ - vec_st(svA##src##i, i * stride, src); \ - vec_st(svB##src##i, i * stride + 16, src) +#define PACK_AND_STORE(src, i) do { \ + const vector unsigned char perms = vec_lvsr(i * stride, src); \ + const vector unsigned char vf = \ + vec_packsu(v_tempBlurredAss##1, (vector signed short)zero); \ + const vector unsigned char vg = vec_perm(vf, v_##src##A##i, permHH); \ + const vector unsigned char mask = \ + vec_perm((vector unsigned char)zero, (vector unsigned char)neg1, perms); \ + const vector unsigned char vg2 = vec_perm(vg, vg, perms); \ + const vector unsigned char svA = vec_sel(v_##src##A1##i, vg2, mask); \ + const vector unsigned char svB = vec_sel(vg2, v_##src##A2##i, mask); \ + vec_st(svA, i * stride, src); \ + vec_st(svB, i * stride + 16, src); \ +} while (0) PACK_AND_STORE(src, 0); PACK_AND_STORE(src, 1); @@ -1127,6 +1119,7 @@ static inline void transpose_16x8_char_toPackedAlign_altivec(unsigned char* dst, static inline void transpose_8x16_char_fromPackedAlign_altivec(unsigned char* dst, unsigned char* src, int stride) { const vector unsigned char zero = vec_splat_u8(0); + const vector signed char neg1 = vec_splat_s8(-1); #define LOAD_DOUBLE_LINE(i, j) \ vector unsigned char src##i = vec_ld(i * 16, src); \ @@ -1187,26 +1180,28 @@ static inline void transpose_8x16_char_fromPackedAlign_altivec(unsigned char* ds temp7 = vec_mergel(tempD, tempL); - const vector signed char neg1 = vec_splat_s8(-1); -#define STORE_DOUBLE_LINE(i, j) \ - vector unsigned char dstA##i = vec_ld(i * stride, dst); \ - vector unsigned char dstB##i = vec_ld(i * stride + 16, dst); \ - vector unsigned char dstA##j = vec_ld(j * stride, dst); \ - vector unsigned char dstB##j = vec_ld(j * stride+ 16, dst); \ - vector unsigned char align##i = vec_lvsr(i * stride, dst); \ - vector unsigned char align##j = vec_lvsr(j * stride, dst); \ - vector unsigned char mask##i = vec_perm(zero, (vector unsigned char)neg1, align##i); \ - vector unsigned char mask##j = vec_perm(zero, (vector unsigned char)neg1, align##j); \ - vector unsigned char dstR##i = vec_perm(temp##i, temp##i, align##i);\ - vector unsigned char dstR##j = vec_perm(temp##j, temp##j, align##j);\ - vector unsigned char dstAF##i = vec_sel(dstA##i, dstR##i, mask##i); \ - vector unsigned char dstBF##i = vec_sel(dstR##i, dstB##i, mask##i); \ - vector unsigned char dstAF##j = vec_sel(dstA##j, dstR##j, mask##j); \ - vector unsigned char dstBF##j = vec_sel(dstR##j, dstB##j, mask##j); \ - vec_st(dstAF##i, i * stride, dst); \ - vec_st(dstBF##i, i * stride + 16, dst); \ - vec_st(dstAF##j, j * stride, dst); \ - vec_st(dstBF##j, j * stride + 16, dst) +#define STORE_DOUBLE_LINE(i, j) do { \ + vector unsigned char dstAi = vec_ld(i * stride, dst); \ + vector unsigned char dstBi = vec_ld(i * stride + 16, dst); \ + vector unsigned char dstAj = vec_ld(j * stride, dst); \ + vector unsigned char dstBj = vec_ld(j * stride+ 16, dst); \ + vector unsigned char aligni = vec_lvsr(i * stride, dst); \ + vector unsigned char alignj = vec_lvsr(j * stride, dst); \ + vector unsigned char maski = \ + vec_perm(zero, (vector unsigned char)neg1, aligni); \ + vector unsigned char maskj = \ + vec_perm(zero, (vector unsigned char)neg1, alignj); \ + vector unsigned char dstRi = vec_perm(temp##i, temp##i, aligni); \ + vector unsigned char dstRj = vec_perm(temp##j, temp##j, alignj); \ + vector unsigned char dstAFi = vec_sel(dstAi, dstRi, maski); \ + vector unsigned char dstBFi = vec_sel(dstRi, dstBi, maski); \ + vector unsigned char dstAFj = vec_sel(dstAj, dstRj, maskj); \ + vector unsigned char dstBFj = vec_sel(dstRj, dstBj, maskj); \ + vec_st(dstAFi, i * stride, dst); \ + vec_st(dstBFi, i * stride + 16, dst); \ + vec_st(dstAFj, j * stride, dst); \ + vec_st(dstBFj, j * stride + 16, dst); \ +} while (0) STORE_DOUBLE_LINE(0,1); STORE_DOUBLE_LINE(2,3); diff --git a/libpostproc/postprocess_template.c b/libpostproc/postprocess_template.c index 934855b..04dd36c 100644 --- a/libpostproc/postprocess_template.c +++ b/libpostproc/postprocess_template.c @@ -3515,9 +3515,10 @@ static void RENAME(postProcess)(const uint8_t src[], int srcStride, uint8_t dst[ else if(mode & H_DEBLOCK){ #if HAVE_ALTIVEC DECLARE_ALIGNED(16, unsigned char, tempBlock)[272]; + int t; transpose_16x8_char_toPackedAlign_altivec(tempBlock, dstBlock - (4 + 1), stride); - const int t=vertClassify_altivec(tempBlock-48, 16, &c); + t = vertClassify_altivec(tempBlock-48, 16, &c); if(t==1) { doVertLowPass_altivec(tempBlock-48, 16, &c); transpose_8x16_char_fromPackedAlign_altivec(dstBlock - (4 + 1), tempBlock, stride); |