Fake-Merge remote-tracking branch 'ffmpeg-mt/master'

libavcodec/avcodec.h

@@ -725,10 +725,12 @@ typedef struct RcOverride{
  * Codec should fill in channel configuration and samplerate instead of container
  */
 #define CODEC_CAP_CHANNEL_CONF     0x0400
+
 /**
  * Codec is able to deal with negative linesizes
  */
 #define CODEC_CAP_NEG_LINESIZES    0x0800
+
 /**
  * Codec supports frame-level multithreading.
  */
@@ -2983,7 +2985,9 @@ typedef struct AVCodec {
     const enum AVSampleFormat *sample_fmts; ///< array of supported sample formats, or NULL if unknown, array is terminated by -1
     const int64_t *channel_layouts;         ///< array of support channel layouts, or NULL if unknown. array is terminated by 0
     uint8_t max_lowres;                     ///< maximum value for lowres supported by the decoder
+
     AVClass *priv_class;                    ///< AVClass for the private context
+
     const AVProfile *profiles;              ///< array of recognized profiles, or NULL if unknown, array is terminated by {FF_PROFILE_UNKNOWN}
 
     /**

libavcodec/dsputil.c

@@ -298,7 +298,7 @@ static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
 
 /* draw the edges of width 'w' of an image of size width, height */
 //FIXME check that this is ok for mpeg4 interlaced
-static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
+static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w, int sides)
 {
     uint8_t *ptr, *last_line;
     int i;
@@ -306,8 +306,8 @@ static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
     last_line = buf + (height - 1) * wrap;
     for(i=0;i<w;i++) {
         /* top and bottom */
-        memcpy(buf - (i + 1) * wrap, buf, width);
-        memcpy(last_line + (i + 1) * wrap, last_line, width);
+        if (sides&EDGE_TOP)    memcpy(buf - (i + 1) * wrap, buf, width);
+        if (sides&EDGE_BOTTOM) memcpy(last_line + (i + 1) * wrap, last_line, width);
     }
     /* left and right */
     ptr = buf;
@@ -318,10 +318,15 @@ static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
     }
     /* corners */
     for(i=0;i<w;i++) {
-        memset(buf - (i + 1) * wrap - w, buf[0], w); /* top left */
-        memset(buf - (i + 1) * wrap + width, buf[width-1], w); /* top right */
-        memset(last_line + (i + 1) * wrap - w, last_line[0], w); /* top left */
-        memset(last_line + (i + 1) * wrap + width, last_line[width-1], w); /* top right */
+        if (sides&EDGE_TOP) {
+            memset(buf - (i + 1) * wrap - w, buf[0], w); /* top left */
+            memset(buf - (i + 1) * wrap + width, buf[width-1], w); /* top right */
+        }
+
+        if (sides&EDGE_BOTTOM) {
+            memset(last_line + (i + 1) * wrap - w, last_line[0], w); /* top left */
+            memset(last_line + (i + 1) * wrap + width, last_line[width-1], w); /* top right */
+        }
     }
 }
 

libavcodec/dsputil.h

@@ -492,8 +492,10 @@ typedef struct DSPContext {
 #define BASIS_SHIFT 16
 #define RECON_SHIFT 6
 
-    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
+    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int sides);
 #define EDGE_WIDTH 16
+#define EDGE_TOP    1
+#define EDGE_BOTTOM 2
 
     void (*prefetch)(void *mem, int stride, int h);
 

libavcodec/h263dec.c

@@ -35,6 +35,7 @@
 #include "mpeg4video_parser.h"
 #include "msmpeg4.h"
 #include "vdpau_internal.h"
+#include "thread.h"
 #include "flv.h"
 #include "mpeg4video.h"
 
@@ -235,6 +236,7 @@ static int decode_slice(MpegEncContext *s){
                     if(++s->mb_x >= s->mb_width){
                         s->mb_x=0;
                         ff_draw_horiz_band(s, s->mb_y*mb_size, mb_size);
+                        MPV_report_decode_progress(s);
                         s->mb_y++;
                     }
                     return 0;
@@ -255,6 +257,7 @@ static int decode_slice(MpegEncContext *s){
         }
 
         ff_draw_horiz_band(s, s->mb_y*mb_size, mb_size);
+        MPV_report_decode_progress(s);
 
         s->mb_x= 0;
     }
@@ -639,6 +642,8 @@ retry:
     if(MPV_frame_start(s, avctx) < 0)
         return -1;
 
+    if (!s->divx_packed) ff_thread_finish_setup(avctx);
+
     if (CONFIG_MPEG4_VDPAU_DECODER && (s->avctx->codec->capabilities & CODEC_CAP_HWACCEL_VDPAU)) {
         ff_vdpau_mpeg4_decode_picture(s, s->gb.buffer, s->gb.buffer_end - s->gb.buffer);
         goto frame_end;

libavcodec/h264.h

@@ -392,9 +392,9 @@ typedef struct H264Context{
     /**
      * num_ref_idx_l0/1_active_minus1 + 1
      */
+    uint8_t *list_counts;            ///< Array of list_count per MB specifying the slice type
     unsigned int ref_count[2];   ///< counts frames or fields, depending on current mb mode
     unsigned int list_count;
-    uint8_t *list_counts;            ///< Array of list_count per MB specifying the slice type
     Picture ref_list[2][48];         /**< 0..15: frame refs, 16..47: mbaff field refs.
                                           Reordered version of default_ref_list
                                           according to picture reordering in slice header */
@@ -504,7 +504,9 @@ typedef struct H264Context{
     Picture *long_ref[32];
     Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
     Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
+    Picture *next_output_pic;
     int outputed_poc;
+    int next_outputed_poc;
 
     /**
      * memory management control operations buffer.

libavcodec/h264_direct.c

@@ -31,6 +31,7 @@
 #include "mpegvideo.h"
 #include "h264.h"
 #include "rectangle.h"
+#include "thread.h"
 
 //#undef NDEBUG
 #include <assert.h>
@@ -126,7 +127,7 @@ void ff_h264_direct_ref_list_init(H264Context * const h){
         h->col_parity= (FFABS(col_poc[0] - cur_poc) >= FFABS(col_poc[1] - cur_poc));
         ref1sidx=sidx= h->col_parity;
     }else if(!(s->picture_structure & h->ref_list[1][0].reference) && !h->ref_list[1][0].mbaff){ // FL -> FL & differ parity
-        h->col_fieldoff= s->mb_stride*(2*(h->ref_list[1][0].reference) - 3);
+        h->col_fieldoff= 2*(h->ref_list[1][0].reference) - 3;
     }
 
     if(cur->pict_type != FF_B_TYPE || h->direct_spatial_mv_pred)
@@ -140,11 +141,27 @@ void ff_h264_direct_ref_list_init(H264Context * const h){
     }
 }
 
+static void await_reference_mb_row(H264Context * const h, Picture *ref, int mb_y)
+{
+    int ref_field = ref->reference - 1;
+    int ref_field_picture = ref->field_picture;
+    int ref_height = 16*h->s.mb_height >> ref_field_picture;
+
+    if(!HAVE_PTHREADS || !(h->s.avctx->active_thread_type&FF_THREAD_FRAME))
+        return;
+
+    //FIXME it can be safe to access mb stuff
+    //even if pixels aren't deblocked yet
+
+    ff_thread_await_progress((AVFrame*)ref, FFMIN(16*mb_y >> ref_field_picture, ref_height-1),
+                             ref_field_picture && ref_field);
+}
+
 static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
     MpegEncContext * const s = &h->s;
     int b8_stride = 2;
     int b4_stride = h->b_stride;
-    int mb_xy = h->mb_xy;
+    int mb_xy = h->mb_xy, mb_y = s->mb_y;
     int mb_type_col[2];
     const int16_t (*l1mv0)[2], (*l1mv1)[2];
     const int8_t *l1ref0, *l1ref1;
@@ -157,6 +174,8 @@ static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
 
     assert(h->ref_list[1][0].reference&3);
 
+    await_reference_mb_row(h, &h->ref_list[1][0], s->mb_y + !!IS_INTERLACED(*mb_type));
+
 #define MB_TYPE_16x16_OR_INTRA (MB_TYPE_16x16|MB_TYPE_INTRA4x4|MB_TYPE_INTRA16x16|MB_TYPE_INTRA_PCM)
 
 
@@ -217,14 +236,17 @@ static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
 
     if(IS_INTERLACED(h->ref_list[1][0].mb_type[mb_xy])){ // AFL/AFR/FR/FL -> AFL/FL
         if(!IS_INTERLACED(*mb_type)){                    //     AFR/FR    -> AFL/FL
+            mb_y = (s->mb_y&~1) + h->col_parity;
             mb_xy= s->mb_x + ((s->mb_y&~1) + h->col_parity)*s->mb_stride;
             b8_stride = 0;
         }else{
-            mb_xy += h->col_fieldoff; // non zero for FL -> FL & differ parity
+            mb_y  += h->col_fieldoff;
+            mb_xy += s->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
         }
         goto single_col;
     }else{                                               // AFL/AFR/FR/FL -> AFR/FR
         if(IS_INTERLACED(*mb_type)){                     // AFL       /FL -> AFR/FR
+            mb_y = s->mb_y&~1;
             mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
             mb_type_col[0] = h->ref_list[1][0].mb_type[mb_xy];
             mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy + s->mb_stride];
@@ -260,6 +282,8 @@ single_col:
         }
     }
 
+    await_reference_mb_row(h, &h->ref_list[1][0], mb_y);
+
     l1mv0  = &h->ref_list[1][0].motion_val[0][h->mb2b_xy [mb_xy]];
     l1mv1  = &h->ref_list[1][0].motion_val[1][h->mb2b_xy [mb_xy]];
     l1ref0 = &h->ref_list[1][0].ref_index [0][4*mb_xy];
@@ -384,7 +408,7 @@ static void pred_temp_direct_motion(H264Context * const h, int *mb_type){
     MpegEncContext * const s = &h->s;
     int b8_stride = 2;
     int b4_stride = h->b_stride;
-    int mb_xy = h->mb_xy;
+    int mb_xy = h->mb_xy, mb_y = s->mb_y;
     int mb_type_col[2];
     const int16_t (*l1mv0)[2], (*l1mv1)[2];
     const int8_t *l1ref0, *l1ref1;
@@ -394,16 +418,21 @@ static void pred_temp_direct_motion(H264Context * const h, int *mb_type){
 
     assert(h->ref_list[1][0].reference&3);
 
+    await_reference_mb_row(h, &h->ref_list[1][0], s->mb_y + !!IS_INTERLACED(*mb_type));
+
     if(IS_INTERLACED(h->ref_list[1][0].mb_type[mb_xy])){ // AFL/AFR/FR/FL -> AFL/FL
         if(!IS_INTERLACED(*mb_type)){                    //     AFR/FR    -> AFL/FL
+            mb_y = (s->mb_y&~1) + h->col_parity;
             mb_xy= s->mb_x + ((s->mb_y&~1) + h->col_parity)*s->mb_stride;
             b8_stride = 0;
         }else{
-            mb_xy += h->col_fieldoff; // non zero for FL -> FL & differ parity
+            mb_y  += h->col_fieldoff;
+            mb_xy += s->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
         }
         goto single_col;
     }else{                                               // AFL/AFR/FR/FL -> AFR/FR
         if(IS_INTERLACED(*mb_type)){                     // AFL       /FL -> AFR/FR
+            mb_y = s->mb_y&~1;
             mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
             mb_type_col[0] = h->ref_list[1][0].mb_type[mb_xy];
             mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy + s->mb_stride];
@@ -440,6 +469,8 @@ single_col:
         }
     }
 
+    await_reference_mb_row(h, &h->ref_list[1][0], mb_y);
+
     l1mv0  = &h->ref_list[1][0].motion_val[0][h->mb2b_xy [mb_xy]];
     l1mv1  = &h->ref_list[1][0].motion_val[1][h->mb2b_xy [mb_xy]];
     l1ref0 = &h->ref_list[1][0].ref_index [0][4*mb_xy];

libavcodec/mdec.c

@@ -31,6 +31,7 @@
 #include "dsputil.h"
 #include "mpegvideo.h"
 #include "mpeg12.h"
+#include "thread.h"
 
 typedef struct MDECContext{
     AVCodecContext *avctx;
@@ -162,10 +163,10 @@ static int decode_frame(AVCodecContext *avctx,
     int i;
 
     if(p->data[0])
-        avctx->release_buffer(avctx, p);
+        ff_thread_release_buffer(avctx, p);
 
     p->reference= 0;
-    if(avctx->get_buffer(avctx, p) < 0){
+    if(ff_thread_get_buffer(avctx, p) < 0){
         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return -1;
     }
@@ -238,6 +239,18 @@ static av_cold int decode_init(AVCodecContext *avctx){
     return 0;
 }
 
+static av_cold int decode_init_thread_copy(AVCodecContext *avctx){
+    MDECContext * const a = avctx->priv_data;
+    AVFrame *p = (AVFrame*)&a->picture;
+
+    avctx->coded_frame= p;
+    a->avctx= avctx;
+
+    p->qscale_table= av_mallocz( p->qstride * a->mb_height);
+
+    return 0;
+}
+
 static av_cold int decode_end(AVCodecContext *avctx){
     MDECContext * const a = avctx->priv_data;
 
@@ -259,7 +272,8 @@ AVCodec ff_mdec_decoder = {
     NULL,
     decode_end,
     decode_frame,
-    CODEC_CAP_DR1,
+    CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
     .long_name= NULL_IF_CONFIG_SMALL("Sony PlayStation MDEC (Motion DECoder)"),
+    .init_thread_copy= ONLY_IF_THREADS_ENABLED(decode_init_thread_copy)
 };
 

libavcodec/mimic.c

@@ -27,6 +27,7 @@
 #include "get_bits.h"
 #include "bytestream.h"
 #include "dsputil.h"
+#include "thread.h"
 
 #define MIMIC_HEADER_SIZE   20
 
@@ -51,6 +52,10 @@ typedef struct {
     ScanTable       scantable;
     DSPContext      dsp;
     VLC             vlc;
+
+    /* Kept in the context so multithreading can have a constant to read from */
+    int             next_cur_index;
+    int             next_prev_index;
 } MimicContext;
 
 static const uint32_t huffcodes[] = {
@@ -121,6 +126,23 @@ static av_cold int mimic_decode_init(AVCodecContext *avctx)
     return 0;
 }
 
+static int mimic_decode_update_thread_context(AVCodecContext *avctx, const AVCodecContext *avctx_from)
+{
+    MimicContext *dst = avctx->priv_data, *src = avctx_from->priv_data;
+
+    if (avctx == avctx_from) return 0;
+
+    dst->cur_index  = src->next_cur_index;
+    dst->prev_index = src->next_prev_index;
+
+    memcpy(dst->buf_ptrs, src->buf_ptrs, sizeof(src->buf_ptrs));
+    memcpy(dst->flipped_ptrs, src->flipped_ptrs, sizeof(src->flipped_ptrs));
+
+    memset(&dst->buf_ptrs[dst->cur_index], 0, sizeof(AVFrame));
+
+    return 0;
+}
+
 static const int8_t vlcdec_lookup[9][64] = {
     {    0, },
     {   -1,   1, },
@@ -205,7 +227,7 @@ static int vlc_decode_block(MimicContext *ctx, int num_coeffs, int qscale)
 static int decode(MimicContext *ctx, int quality, int num_coeffs,
                   int is_iframe)
 {
-    int y, x, plane;
+    int y, x, plane, cur_row = 0;
 
     for(plane = 0; plane < 3; plane++) {
         const int is_chroma = !!plane;
@@ -236,6 +258,7 @@ static int decode(MimicContext *ctx, int quality, int num_coeffs,
                         int index = (ctx->cur_index+backref)&15;
                         uint8_t *p = ctx->flipped_ptrs[index].data[0];
 
+                        ff_thread_await_progress(&ctx->buf_ptrs[index], cur_row, 0);
                         if(p) {
                             p += src -
                                 ctx->flipped_ptrs[ctx->prev_index].data[plane];
@@ -246,6 +269,7 @@ static int decode(MimicContext *ctx, int quality, int num_coeffs,
                         }
                     }
                 } else {
+                    ff_thread_await_progress(&ctx->buf_ptrs[ctx->prev_index], cur_row, 0);
                     ctx->dsp.put_pixels_tab[1][0](dst, src, stride, 8);
                 }
                 src += 8;
@@ -253,6 +277,8 @@ static int decode(MimicContext *ctx, int quality, int num_coeffs,
             }
             src += (stride - ctx->num_hblocks[plane])<<3;
             dst += (stride - ctx->num_hblocks[plane])<<3;
+
+            ff_thread_report_progress(&ctx->buf_ptrs[ctx->cur_index], cur_row++, 0);
         }
     }
 
@@ -326,14 +352,20 @@ static int mimic_decode_frame(AVCodecContext *avctx, void *data,
     }
 
     ctx->buf_ptrs[ctx->cur_index].reference = 1;
-    if(avctx->get_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index])) {
+    ctx->buf_ptrs[ctx->cur_index].pict_type = is_pframe ? FF_P_TYPE:FF_I_TYPE;
+    if(ff_thread_get_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index])) {
         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return -1;
     }
 
+    ctx->next_prev_index = ctx->cur_index;
+    ctx->next_cur_index  = (ctx->cur_index - 1) & 15;
+
     prepare_avpic(ctx, &ctx->flipped_ptrs[ctx->cur_index],
                   (AVPicture*) &ctx->buf_ptrs[ctx->cur_index]);
 
+    ff_thread_finish_setup(avctx);
+
     av_fast_malloc(&ctx->swap_buf, &ctx->swap_buf_size,
                                  swap_buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
     if(!ctx->swap_buf)
@@ -345,21 +377,23 @@ static int mimic_decode_frame(AVCodecContext *avctx, void *data,
     init_get_bits(&ctx->gb, ctx->swap_buf, swap_buf_size << 3);
 
     if(!decode(ctx, quality, num_coeffs, !is_pframe)) {
-        avctx->release_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index]);
-        return -1;
+        if (avctx->active_thread_type&FF_THREAD_FRAME)
+            ff_thread_report_progress(&ctx->buf_ptrs[ctx->cur_index], INT_MAX, 0);
+        else {
+            ff_thread_release_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index]);
+            return -1;
+        }
     }
 
-    ctx->buf_ptrs[ctx->cur_index].pict_type = is_pframe ? FF_P_TYPE:FF_I_TYPE;
     *(AVFrame*)data = ctx->buf_ptrs[ctx->cur_index];
     *data_size = sizeof(AVFrame);
 
-    ctx->prev_index = ctx->cur_index;
-    ctx->cur_index--;
-    ctx->cur_index &= 15;
+    ctx->prev_index = ctx->next_prev_index;
+    ctx->cur_index  = ctx->next_cur_index;
 
     /* Only release frames that aren't used for backreferences anymore */
     if(ctx->buf_ptrs[ctx->cur_index].data[0])
-        avctx->release_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index]);
+        ff_thread_release_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index]);
 
     return buf_size;
 }
@@ -370,9 +404,12 @@ static av_cold int mimic_decode_end(AVCodecContext *avctx)
     int i;
 
     av_free(ctx->swap_buf);
+
+    if(avctx->is_copy) return 0;
+
     for(i = 0; i < 16; i++)
         if(ctx->buf_ptrs[i].data[0])
-            avctx->release_buffer(avctx, &ctx->buf_ptrs[i]);
+            ff_thread_release_buffer(avctx, &ctx->buf_ptrs[i]);
     free_vlc(&ctx->vlc);
 
     return 0;
@@ -387,6 +424,7 @@ AVCodec ff_mimic_decoder = {
     NULL,
     mimic_decode_end,
     mimic_decode_frame,
-    CODEC_CAP_DR1,
+    CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
     .long_name = NULL_IF_CONFIG_SMALL("Mimic"),
+    .update_thread_context = ONLY_IF_THREADS_ENABLED(mimic_decode_update_thread_context)
 };

libavcodec/mpeg12.c

@@ -37,6 +37,7 @@
 #include "bytestream.h"
 #include "vdpau_internal.h"
 #include "xvmc_internal.h"
+#include "thread.h"
 
 //#undef NDEBUG
 //#include <assert.h>
@@ -1179,6 +1180,27 @@ static av_cold int mpeg_decode_init(AVCodecContext *avctx)
     return 0;
 }
 
+static int mpeg_decode_update_thread_context(AVCodecContext *avctx, const AVCodecContext *avctx_from)
+{
+    Mpeg1Context *ctx = avctx->priv_data, *ctx_from = avctx_from->priv_data;
+    MpegEncContext *s = &ctx->mpeg_enc_ctx, *s1 = &ctx_from->mpeg_enc_ctx;
+    int err;
+
+    if(avctx == avctx_from || !ctx_from->mpeg_enc_ctx_allocated || !s1->context_initialized)
+        return 0;
+
+    err = ff_mpeg_update_thread_context(avctx, avctx_from);
+    if(err) return err;
+
+    if(!ctx->mpeg_enc_ctx_allocated)
+        memcpy(s + 1, s1 + 1, sizeof(Mpeg1Context) - sizeof(MpegEncContext));
+
+    if(!(s->pict_type == FF_B_TYPE || s->low_delay))
+        s->picture_number++;
+
+    return 0;
+}
+
 static void quant_matrix_rebuild(uint16_t *matrix, const uint8_t *old_perm,
                                      const uint8_t *new_perm){
     uint16_t temp_matrix[64];
@@ -1595,6 +1617,9 @@ static int mpeg_field_start(MpegEncContext *s, const uint8_t *buf, int buf_size)
         }
 
         *s->current_picture_ptr->pan_scan= s1->pan_scan;
+
+        if (HAVE_PTHREADS && avctx->active_thread_type&FF_THREAD_FRAME)
+            ff_thread_finish_setup(avctx);
     }else{ //second field
             int i;
 
@@ -1769,6 +1794,7 @@ static int mpeg_decode_slice(Mpeg1Context *s1, int mb_y,
             const int mb_size= 16>>s->avctx->lowres;
 
             ff_draw_horiz_band(s, mb_size*(s->mb_y>>field_pic), mb_size);
+            MPV_report_decode_progress(s);
 
             s->mb_x = 0;
             s->mb_y += 1<<field_pic;
@@ -1924,7 +1950,8 @@ static int slice_end(AVCodecContext *avctx, AVFrame *pict)
             *pict= *(AVFrame*)s->current_picture_ptr;
             ff_print_debug_info(s, pict);
         } else {
-            s->picture_number++;
+            if (avctx->active_thread_type&FF_THREAD_FRAME)
+                s->picture_number++;
             /* latency of 1 frame for I- and P-frames */
             /* XXX: use another variable than picture_number */
             if (s->last_picture_ptr != NULL) {
@@ -2260,7 +2287,7 @@ static int decode_chunks(AVCodecContext *avctx,
         buf_ptr = ff_find_start_code(buf_ptr,buf_end, &start_code);
         if (start_code > 0x1ff){
             if(s2->pict_type != FF_B_TYPE || avctx->skip_frame <= AVDISCARD_DEFAULT){
-                if(avctx->thread_count > 1){
+                if(HAVE_THREADS && avctx->active_thread_type&FF_THREAD_SLICE){
                     int i;
 
                     avctx->execute(avctx, slice_decode_thread,  &s2->thread_context[0], NULL, s->slice_count, sizeof(void*));
@@ -2422,7 +2449,7 @@ static int decode_chunks(AVCodecContext *avctx,
                     break;
                 }
 
-                if(avctx->thread_count > 1){
+                if(HAVE_THREADS && avctx->active_thread_type&FF_THREAD_SLICE){
                     int threshold= (s2->mb_height*s->slice_count + avctx->thread_count/2) / avctx->thread_count;
                     if(threshold <= mb_y){
                         MpegEncContext *thread_context= s2->thread_context[s->slice_count];
@@ -2496,6 +2523,7 @@ AVCodec ff_mpeg1video_decoder = {
     .flush= flush,
     .max_lowres= 3,
     .long_name= NULL_IF_CONFIG_SMALL("MPEG-1 video"),
+    .update_thread_context= ONLY_IF_THREADS_ENABLED(mpeg_decode_update_thread_context)
 };
 
 AVCodec ff_mpeg2video_decoder = {
@@ -2532,7 +2560,7 @@ AVCodec ff_mpegvideo_decoder = {
 
 #if CONFIG_MPEG_XVMC_DECODER
 static av_cold int mpeg_mc_decode_init(AVCodecContext *avctx){
-    if( avctx->thread_count > 1)
+    if( avctx->active_thread_type & FF_THREAD_SLICE )
         return -1;
     if( !(avctx->slice_flags & SLICE_FLAG_CODED_ORDER) )
         return -1;

libavcodec/mpeg4videodec.c

@@ -23,6 +23,7 @@
 #include "mpegvideo.h"
 #include "mpeg4video.h"
 #include "h263.h"
+#include "thread.h"
 
 // The defines below define the number of bits that are read at once for
 // reading vlc values. Changing these may improve speed and data cache needs
@@ -373,7 +374,13 @@ int mpeg4_decode_video_packet_header(MpegEncContext *s)
         return -1;
     }
     if(s->pict_type == FF_B_TYPE){
-        while(s->next_picture.mbskip_table[ s->mb_index2xy[ mb_num ] ]) mb_num++;
+        int mb_x = 0, mb_y = 0;
+
+        while(s->next_picture.mbskip_table[ s->mb_index2xy[ mb_num ] ]) {
+            if (!mb_x) ff_thread_await_progress((AVFrame*)s->next_picture_ptr, mb_y++, 0);
+            mb_num++;
+            if (++mb_x == s->mb_width) mb_x = 0;
+        }
         if(mb_num >= s->mb_num) return -1; // slice contains just skipped MBs which where already decoded
     }
 
@@ -1303,6 +1310,8 @@ static int mpeg4_decode_mb(MpegEncContext *s,
                 s->last_mv[i][1][0]=
                 s->last_mv[i][1][1]= 0;
             }
+
+            ff_thread_await_progress((AVFrame*)s->next_picture_ptr, s->mb_y, 0);
         }
 
         /* if we skipped it in the future P Frame than skip it now too */
@@ -1482,6 +1491,12 @@ end:
     if(s->codec_id==CODEC_ID_MPEG4){
         if(mpeg4_is_resync(s)){
             const int delta= s->mb_x + 1 == s->mb_width ? 2 : 1;
+
+            if(s->pict_type==FF_B_TYPE){
+                ff_thread_await_progress((AVFrame*)s->next_picture_ptr,
+                                        (s->mb_x + delta >= s->mb_width) ? FFMIN(s->mb_y+1, s->mb_height-1) : s->mb_y, 0);
+            }
+
             if(s->pict_type==FF_B_TYPE && s->next_picture.mbskip_table[xy + delta])
                 return SLICE_OK;
             return SLICE_END;
@@ -2239,11 +2254,12 @@ AVCodec ff_mpeg4_decoder = {
     NULL,
     ff_h263_decode_end,
     ff_h263_decode_frame,
-    CODEC_CAP_DRAW_HORIZ_BAND | CODEC_CAP_DR1 | CODEC_CAP_TRUNCATED | CODEC_CAP_DELAY,
+    CODEC_CAP_DRAW_HORIZ_BAND | CODEC_CAP_DR1 | CODEC_CAP_TRUNCATED | CODEC_CAP_DELAY | CODEC_CAP_FRAME_THREADS,
     .flush= ff_mpeg_flush,
     .max_lowres= 3,
     .long_name= NULL_IF_CONFIG_SMALL("MPEG-4 part 2"),
     .pix_fmts= ff_hwaccel_pixfmt_list_420,
+    .update_thread_context= ONLY_IF_THREADS_ENABLED(ff_mpeg_update_thread_context)
 };
 
 

libavcodec/mpegvideo.h

@@ -123,6 +123,7 @@ typedef struct Picture{
     int ref_poc[2][2][16];      ///< h264 POCs of the frames used as reference (FIXME need per slice)
     int ref_count[2][2];        ///< number of entries in ref_poc              (FIXME need per slice)
     int mbaff;                  ///< h264 1 -> MBAFF frame 0-> not MBAFF
+    int field_picture;          ///< whether or not the picture was encoded in seperate fields
 
     int mb_var_sum;             ///< sum of MB variance for current frame
     int mc_mb_var_sum;          ///< motion compensated MB variance for current frame
@@ -292,6 +293,8 @@ typedef struct MpegEncContext {
     Picture *last_picture_ptr;     ///< pointer to the previous picture.
     Picture *next_picture_ptr;     ///< pointer to the next picture (for bidir pred)
     Picture *current_picture_ptr;  ///< pointer to the current picture
+    int picture_count;             ///< number of allocated pictures (MAX_PICTURE_COUNT * avctx->thread_count)
+    int picture_range_start, picture_range_end; ///< the part of picture that this context can allocate in
     uint8_t *visualization_buffer[3]; //< temporary buffer vor MV visualization
     int last_dc[3];                ///< last DC values for MPEG1
     int16_t *dc_val_base;
@@ -681,6 +684,7 @@ typedef struct MpegEncContext {
     void (*denoise_dct)(struct MpegEncContext *s, DCTELEM *block);
 } MpegEncContext;
 
+#define REBASE_PICTURE(pic, new_ctx, old_ctx) (pic ? &new_ctx->picture[pic - old_ctx->picture] : NULL)
 
 void MPV_decode_defaults(MpegEncContext *s);
 int MPV_common_init(MpegEncContext *s);
@@ -706,6 +710,9 @@ void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
 int ff_find_unused_picture(MpegEncContext *s, int shared);
 void ff_denoise_dct(MpegEncContext *s, DCTELEM *block);
 void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
+int MPV_lowest_referenced_row(MpegEncContext *s, int dir);
+void MPV_report_decode_progress(MpegEncContext *s);
+int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
 const uint8_t *ff_find_start_code(const uint8_t *p, const uint8_t *end, uint32_t *state);
 void ff_set_qscale(MpegEncContext * s, int qscale);
 

libavcodec/mpegvideo_enc.c

@@ -36,6 +36,7 @@
 #include "mjpegenc.h"
 #include "msmpeg4.h"
 #include "faandct.h"
+#include "thread.h"
 #include "aandcttab.h"
 #include "flv.h"
 #include "mpeg4video.h"
@@ -1238,9 +1239,9 @@ int MPV_encode_picture(AVCodecContext *avctx,
 {
     MpegEncContext *s = avctx->priv_data;
     AVFrame *pic_arg = data;
-    int i, stuffing_count;
+    int i, stuffing_count, context_count = avctx->active_thread_type&FF_THREAD_SLICE ? avctx->thread_count : 1;
 
-    for(i=0; i<avctx->thread_count; i++){
+    for(i=0; i<context_count; i++){
         int start_y= s->thread_context[i]->start_mb_y;
         int   end_y= s->thread_context[i]->  end_mb_y;
         int h= s->mb_height;
@@ -1304,7 +1305,7 @@ vbv_retry:
                     s->last_non_b_time= s->time - s->pp_time;
                 }
 //                av_log(NULL, AV_LOG_ERROR, "R:%d ", s->next_lambda);
-                for(i=0; i<avctx->thread_count; i++){
+                for(i=0; i<context_count; i++){
                     PutBitContext *pb= &s->thread_context[i]->pb;
                     init_put_bits(pb, pb->buf, pb->buf_end - pb->buf);
                 }
@@ -2771,6 +2772,7 @@ static int encode_picture(MpegEncContext *s, int picture_number)
 {
     int i;
     int bits;
+    int context_count = s->avctx->active_thread_type&FF_THREAD_SLICE ? s->avctx->thread_count : 1;
 
     s->picture_number = picture_number;
 
@@ -2810,7 +2812,7 @@ static int encode_picture(MpegEncContext *s, int picture_number)
     }
 
     s->mb_intra=0; //for the rate distortion & bit compare functions
-    for(i=1; i<s->avctx->thread_count; i++){
+    for(i=1; i<context_count; i++){
         ff_update_duplicate_context(s->thread_context[i], s);
     }
 
@@ -2823,11 +2825,11 @@ static int encode_picture(MpegEncContext *s, int picture_number)
         s->lambda2= (s->lambda2* (int64_t)s->avctx->me_penalty_compensation + 128)>>8;
         if(s->pict_type != FF_B_TYPE && s->avctx->me_threshold==0){
             if((s->avctx->pre_me && s->last_non_b_pict_type==FF_I_TYPE) || s->avctx->pre_me==2){
-                s->avctx->execute(s->avctx, pre_estimate_motion_thread, &s->thread_context[0], NULL, s->avctx->thread_count, sizeof(void*));
+                s->avctx->execute(s->avctx, pre_estimate_motion_thread, &s->thread_context[0], NULL, context_count, sizeof(void*));
             }
         }
 
-        s->avctx->execute(s->avctx, estimate_motion_thread, &s->thread_context[0], NULL, s->avctx->thread_count, sizeof(void*));
+        s->avctx->execute(s->avctx, estimate_motion_thread, &s->thread_context[0], NULL, context_count, sizeof(void*));
     }else /* if(s->pict_type == FF_I_TYPE) */{
         /* I-Frame */
         for(i=0; i<s->mb_stride*s->mb_height; i++)
@@ -2835,10 +2837,10 @@ static int encode_picture(MpegEncContext *s, int picture_number)
 
         if(!s->fixed_qscale){
             /* finding spatial complexity for I-frame rate control */
-            s->avctx->execute(s->avctx, mb_var_thread, &s->thread_context[0], NULL, s->avctx->thread_count, sizeof(void*));
+            s->avctx->execute(s->avctx, mb_var_thread, &s->thread_context[0], NULL, context_count, sizeof(void*));
         }
     }
-    for(i=1; i<s->avctx->thread_count; i++){
+    for(i=1; i<context_count; i++){
         merge_context_after_me(s, s->thread_context[i]);
     }
     s->current_picture.mc_mb_var_sum= s->current_picture_ptr->mc_mb_var_sum= s->me.mc_mb_var_sum_temp;
@@ -2974,11 +2976,11 @@ static int encode_picture(MpegEncContext *s, int picture_number)
     bits= put_bits_count(&s->pb);
     s->header_bits= bits - s->last_bits;
 
-    for(i=1; i<s->avctx->thread_count; i++){
+    for(i=1; i<context_count; i++){
         update_duplicate_context_after_me(s->thread_context[i], s);
     }
-    s->avctx->execute(s->avctx, encode_thread, &s->thread_context[0], NULL, s->avctx->thread_count, sizeof(void*));
-    for(i=1; i<s->avctx->thread_count; i++){
+    s->avctx->execute(s->avctx, encode_thread, &s->thread_context[0], NULL, context_count, sizeof(void*));
+    for(i=1; i<context_count; i++){
         merge_context_after_encode(s, s->thread_context[i]);
     }
     emms_c();

libavcodec/snow.c

@@ -1976,9 +1976,9 @@ static int frame_start(SnowContext *s){
    int h= s->avctx->height;
 
     if(s->current_picture.data[0]){
-        s->dsp.draw_edges(s->current_picture.data[0], s->current_picture.linesize[0], w   , h   , EDGE_WIDTH  );
-        s->dsp.draw_edges(s->current_picture.data[1], s->current_picture.linesize[1], w>>1, h>>1, EDGE_WIDTH/2);
-        s->dsp.draw_edges(s->current_picture.data[2], s->current_picture.linesize[2], w>>1, h>>1, EDGE_WIDTH/2);
+        s->dsp.draw_edges(s->current_picture.data[0], s->current_picture.linesize[0], w   , h   , EDGE_WIDTH  , EDGE_TOP|EDGE_BOTTOM);
+        s->dsp.draw_edges(s->current_picture.data[1], s->current_picture.linesize[1], w>>1, h>>1, EDGE_WIDTH/2, EDGE_TOP|EDGE_BOTTOM);
+        s->dsp.draw_edges(s->current_picture.data[2], s->current_picture.linesize[2], w>>1, h>>1, EDGE_WIDTH/2, EDGE_TOP|EDGE_BOTTOM);
     }
 
     release_buffer(s->avctx);

libavcodec/utils.c

@@ -366,6 +366,7 @@ void avcodec_default_release_buffer(AVCodecContext *s, AVFrame *pic){
     assert(pic->type==FF_BUFFER_TYPE_INTERNAL);
     assert(s->internal_buffer_count);
 
+    if(s->internal_buffer){
     buf = NULL; /* avoids warning */
     for(i=0; i<s->internal_buffer_count; i++){ //just 3-5 checks so is not worth to optimize
         buf= &((InternalBuffer*)s->internal_buffer)[i];
@@ -377,6 +378,7 @@ void avcodec_default_release_buffer(AVCodecContext *s, AVFrame *pic){
     last = &((InternalBuffer*)s->internal_buffer)[s->internal_buffer_count];
 
     FFSWAP(InternalBuffer, *buf, *last);
+    }
 
     for(i=0; i<4; i++){
         pic->data[i]=NULL;
@@ -572,6 +574,7 @@ int attribute_align_arg avcodec_open(AVCodecContext *avctx, AVCodec *codec)
             goto free_and_end;
         }
     }
+
     ret=0;
 end:
     entangled_thread_counter--;
@@ -1298,9 +1301,9 @@ int av_lockmgr_register(int (*cb)(void **mutex, enum AVLockOp op))
 unsigned int ff_toupper4(unsigned int x)
 {
     return     toupper( x     &0xFF)
-            + (toupper((x>>8 )&0xFF)<<8 )
-            + (toupper((x>>16)&0xFF)<<16)
-            + (toupper((x>>24)&0xFF)<<24);
+    + (toupper((x>>8 )&0xFF)<<8 )
+    + (toupper((x>>16)&0xFF)<<16)
+    + (toupper((x>>24)&0xFF)<<24);
 }
 
 #if !HAVE_PTHREADS

libavcodec/vp8.c

@@ -471,7 +471,7 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y)
                                   mb - 1 /* left */,
                                   mb + 1 /* top-left */ };
     enum { CNT_ZERO, CNT_NEAREST, CNT_NEAR, CNT_SPLITMV };
-    enum { EDGE_TOP, EDGE_LEFT, EDGE_TOPLEFT };
+    enum { VP8_EDGE_TOP, VP8_EDGE_LEFT, VP8_EDGE_TOPLEFT };
     int idx = CNT_ZERO;
     int cur_sign_bias = s->sign_bias[mb->ref_frame];
     int8_t *sign_bias = s->sign_bias;
@@ -512,7 +512,7 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y)
         mb->mode = VP8_MVMODE_MV;
 
         /* If we have three distinct MVs, merge first and last if they're the same */
-        if (cnt[CNT_SPLITMV] && AV_RN32A(&near_mv[1+EDGE_TOP]) == AV_RN32A(&near_mv[1+EDGE_TOPLEFT]))
+        if (cnt[CNT_SPLITMV] && AV_RN32A(&near_mv[1+VP8_EDGE_TOP]) == AV_RN32A(&near_mv[1+VP8_EDGE_TOPLEFT]))
             cnt[CNT_NEAREST] += 1;
 
         /* Swap near and nearest if necessary */
@@ -526,9 +526,9 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y)
 
                 /* Choose the best mv out of 0,0 and the nearest mv */
                 clamp_mv(s, &mb->mv, &near_mv[CNT_ZERO + (cnt[CNT_NEAREST] >= cnt[CNT_ZERO])]);
-                cnt[CNT_SPLITMV] = ((mb_edge[EDGE_LEFT]->mode    == VP8_MVMODE_SPLIT) +
-                                    (mb_edge[EDGE_TOP]->mode     == VP8_MVMODE_SPLIT)) * 2 +
-                                    (mb_edge[EDGE_TOPLEFT]->mode == VP8_MVMODE_SPLIT);
+                cnt[CNT_SPLITMV] = ((mb_edge[VP8_EDGE_LEFT]->mode    == VP8_MVMODE_SPLIT) +
+                                    (mb_edge[VP8_EDGE_TOP]->mode     == VP8_MVMODE_SPLIT)) * 2 +
+                                    (mb_edge[VP8_EDGE_TOPLEFT]->mode == VP8_MVMODE_SPLIT);
 
                 if (vp56_rac_get_prob_branchy(c, vp8_mode_contexts[cnt[CNT_SPLITMV]][3])) {
                     mb->mode = VP8_MVMODE_SPLIT;

libavcodec/x86/dsputil_mmx.c

@@ -783,7 +783,7 @@ static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale){
 
 /* draw the edges of width 'w' of an image of size width, height
    this mmx version can only handle w==8 || w==16 */
-static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
+static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w, int sides)
 {
     uint8_t *ptr, *last_line;
     int i;
@@ -838,34 +838,39 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
 
     for(i=0;i<w;i+=4) {
         /* top and bottom (and hopefully also the corners) */
-        ptr= buf - (i + 1) * wrap - w;
-        __asm__ volatile(
-                "1:                             \n\t"
-                "movq (%1, %0), %%mm0           \n\t"
-                "movq %%mm0, (%0)               \n\t"
-                "movq %%mm0, (%0, %2)           \n\t"
-                "movq %%mm0, (%0, %2, 2)        \n\t"
-                "movq %%mm0, (%0, %3)           \n\t"
-                "add $8, %0                     \n\t"
-                "cmp %4, %0                     \n\t"
-                " jb 1b                         \n\t"
-                : "+r" (ptr)
-                : "r" ((x86_reg)buf - (x86_reg)ptr - w), "r" ((x86_reg)-wrap), "r" ((x86_reg)-wrap*3), "r" (ptr+width+2*w)
-        );
-        ptr= last_line + (i + 1) * wrap - w;
-        __asm__ volatile(
-                "1:                             \n\t"
-                "movq (%1, %0), %%mm0           \n\t"
-                "movq %%mm0, (%0)               \n\t"
-                "movq %%mm0, (%0, %2)           \n\t"
-                "movq %%mm0, (%0, %2, 2)        \n\t"
-                "movq %%mm0, (%0, %3)           \n\t"
-                "add $8, %0                     \n\t"
-                "cmp %4, %0                     \n\t"
-                " jb 1b                         \n\t"
-                : "+r" (ptr)
-                : "r" ((x86_reg)last_line - (x86_reg)ptr - w), "r" ((x86_reg)wrap), "r" ((x86_reg)wrap*3), "r" (ptr+width+2*w)
-        );
+        if (sides&EDGE_TOP) {
+            ptr= buf - (i + 1) * wrap - w;
+            __asm__ volatile(
+                    "1:                             \n\t"
+                    "movq (%1, %0), %%mm0           \n\t"
+                    "movq %%mm0, (%0)               \n\t"
+                    "movq %%mm0, (%0, %2)           \n\t"
+                    "movq %%mm0, (%0, %2, 2)        \n\t"
+                    "movq %%mm0, (%0, %3)           \n\t"
+                    "add $8, %0                     \n\t"
+                    "cmp %4, %0                     \n\t"
+                    " jb 1b                         \n\t"
+                    : "+r" (ptr)
+                    : "r" ((x86_reg)buf - (x86_reg)ptr - w), "r" ((x86_reg)-wrap), "r" ((x86_reg)-wrap*3), "r" (ptr+width+2*w)
+            );
+        }
+
+        if (sides&EDGE_BOTTOM) {
+            ptr= last_line + (i + 1) * wrap - w;
+            __asm__ volatile(
+                    "1:                             \n\t"
+                    "movq (%1, %0), %%mm0           \n\t"
+                    "movq %%mm0, (%0)               \n\t"
+                    "movq %%mm0, (%0, %2)           \n\t"
+                    "movq %%mm0, (%0, %2, 2)        \n\t"
+                    "movq %%mm0, (%0, %3)           \n\t"
+                    "add $8, %0                     \n\t"
+                    "cmp %4, %0                     \n\t"
+                    " jb 1b                         \n\t"
+                    : "+r" (ptr)
+                    : "r" ((x86_reg)last_line - (x86_reg)ptr - w), "r" ((x86_reg)wrap), "r" ((x86_reg)wrap*3), "r" (ptr+width+2*w)
+            );
+        }
     }
 }
 

libavutil/internal.h

@@ -181,7 +181,6 @@
 #   define NULL_IF_CONFIG_SMALL(x) x
 #endif
 
-
 /**
  * Define a function with only the non-default version specified.
  *

mt-work/email.sh

+#!/bin/sh -v
+
+# args [where to put patches] [smtp server] [destination]
+
+git format-patch -o "$1" --inline --subject-prefix=soc --thread origin
+git send-email --no-chain-reply-to --smtp-server $2 --to $3 --dry-run "$1"

mt-work/mplayer.diff

+diff --git a/libmpcodecs/vd_ffmpeg.c b/libmpcodecs/vd_ffmpeg.c
+index 7c68a20..135e6b1 100644
+--- a/libmpcodecs/vd_ffmpeg.c
++++ b/libmpcodecs/vd_ffmpeg.c
+@@ -280,7 +280,7 @@ static int init(sh_video_t *sh){
+         return 0;
+     }
+ 
+-    if(vd_use_slices && (lavc_codec->capabilities&CODEC_CAP_DRAW_HORIZ_BAND) && !do_vis_debug)
++    if(vd_use_slices && (lavc_codec->capabilities&CODEC_CAP_DRAW_HORIZ_BAND) && !do_vis_debug && lavc_param_threads <= 1)
+         ctx->do_slices=1;
+ 
+     if(lavc_codec->capabilities&CODEC_CAP_DR1 && !do_vis_debug && lavc_codec->id != CODEC_ID_H264 && lavc_codec->id != CODEC_ID_INTERPLAY_VIDEO && lavc_codec->id != CODEC_ID_ROQ && lavc_codec->id != CODEC_ID_VP8 && lavc_codec->id != CODEC_ID_LAGARITH)

mt-work/raw.sh

+#!/bin/bash
+
+fn=`basename "$1"`
+for th in 1 4; do
+    time ./ffmpeg_g -threads $th -skip_loop_filter all -vsync 0 -y -t 30 -i "$1" -an -f rawvideo "raw/n-$fn-$th.yuv"
+done
+
+#for th in 1 4; do
+#    time ./ffmpeg_g -threads $th -vsync 0 -y -t 30 -i "$1" -an -f rawvideo "raw/$fn-$th.yuv"
+#done

mt-work/test.sh

+#!/bin/bash
+
+fn=`basename "$1"`
+for th in 1 2 3 4; do
+    time ./ffmpeg_g -threads $th -flags2 +fast  -vsync 0 -y -t 30 -i "$1" -an -f framecrc "crc/$fn-$th.txt" >/dev/null 2>&1
+done
+
+./ffmpeg_g -threads 1 -y -t 10 -i "$1" -an -f framecrc "crc/$fn-1-vsync.txt" >/dev/null 2>&1
+./ffmpeg_g -threads 3 -y -t 10 -i "$1" -an -f framecrc "crc/$fn-3-vsync.txt" >/dev/null 2>&1
+
+md5 "crc/$fn-"[1234].txt
+echo
+md5 "crc/$fn-"*vsync.txt

mt-work/todo.txt

+Todo
+
+-- For other people
+- Multithread vp8 or vc1.
+- Multithread an intra codec like mjpeg.
+- Fix mpeg1 (see below).
+- Try the first three items under Optimization.
+- Fix h264 (see below).
+- Try mpeg4 (see below).
+
+-- Bug fixes
+
+General critical:
+- Error resilience has to run before ff_report_frame_progress()
+is called. Otherwise there will be race conditions. (This might already
+work.) In general testing error paths should be done more.
+
+h264:
+- Files split at the wrong NAL unit don't (and can't)
+be decoded with threads (e.g. TS split so PPS is after
+the frame, PAFF with two fields in a packet). Scan the
+packet at the start of decode and don't finish setup
+until all PPS/SPS have been encountered.
+
+mpeg4:
+- Packed B-frames need to be explicitly split up
+when frame threading is on. It's not very fast
+without this.
+- The buffer age optimization is disabled due to
+the way buffers are allocated across threads. The
+branch 'fix_buffer_age' has an attempt to fix it
+which breaks ffplay.
+- Support interlaced.
+
+mpeg1/2:
+- Seeking always prints "first frame not a keyframe"
+with threads on. Currently disabled for this reason.
+
+-- Prove correct
+
+- decode_update_progress() in h264.c
+h264_race_checking branch has some work on h264,
+but not that function. It might be worth putting
+the branch under #ifdef DEBUG in mainline, but
+the code would have to be cleaner.
+- MPV_lowest_referenced_row() and co in mpegvideo.c
+- Same in vp3.
+
+-- Optimization
+
+- EMU_EDGE is always set for h264 PAFF+MT
+because draw_edges() writes into the other field's
+thread's pixels.
+- Check update_thread_context() functions and make
+sure they only copy what they need to.
+- Try some more optimization of the "ref < 48; ref++"
+loop in h264.c await_references(), try turning the list0/list1 check
+above into a loop without being slower.
+- Support frame+slice threading at the same time
+by assigning slice_count threads for frame threads
+to use with execute(). This is simpler but unbalanced
+if only one frame thread uses any.
+
+-- Features
+
+- Support streams with width/height changing. This
+requires flushing all current frames (and buffering
+the input in the meantime), closing the codec and
+reopening it. Or don't support it.
+- Support encoding. Might need more threading primitives
+for good ratecontrol; would be nice for audio and libavfilter too.
+- Async decoding part 1: instead of trying to
+start every thread at the beginning, return a picture
+if the earliest thread is already done, but don't wait
+for it. Not sure what effect this would have.
+- Part 2: have an API that doesn't wait for the decoding
+thread, only returns EAGAIN if it's not ready. What will
+it do with the next input packet if it returns that?
+- Have an API that returns finished pictures but doesn't
+require sending new ones. Maybe allow NULL avpkt when
+not at the end of the stream.
+
+-- Samples
+
+http://astrange.ithinksw.net/ffmpeg/mt-samples/
+
+See yuvcmp.c in this directory to compare decoded samples.
+
+For debugging, try commenting out ff_thread_finish_setup calls so
+that only one thread runs at once, and then binary search+
+scatter printfs to look for differences in codec contexts.

mt-work/valgrind-check.sh

+#!/bin/bash
+
+valgrind --leak-check=full ./ffmpeg_g -threads 3 -vsync 0 -y -t 30 -i "$1" -an -f framecrc /dev/null
\ No newline at end of file

mt-work/yuvcmp.c

+/*
+ * originally by Andreas Öman (andoma)
+ * some changes by Alexander Strange
+ */
+
+#include <string.h>
+#include <stdlib.h>
+#include <inttypes.h>
+#include <stdio.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+
+
+int
+main(int argc, char **argv)
+{
+    int fd[2];
+    int print_pixels = 0;
+    int dump_blocks = 0;
+
+    int width;
+    int height;
+    int to_skip = 0;
+
+    if (argc < 6) {
+        fprintf(stderr, "%s [YUV file 1] [YUV file 2] width height pixelcmp|blockdump (# to skip)\n", argv[0]);
+        return 1;
+    }
+
+    width  = atoi(argv[3]);
+    height = atoi(argv[4]);
+    if (argc > 6)
+        to_skip = atoi(argv[6]);
+
+    uint8_t *Y[2], *C[2][2];
+    int i, v, c, p;
+    int lsiz = width * height;
+    int csiz = width * height / 4;
+    int x, y;
+    int cwidth = width / 2;
+    int fr = to_skip;
+    int mb;
+    char *mberrors;
+    int mb_x, mb_y;
+    uint8_t *a;
+    uint8_t *b;
+    int die = 0;
+
+    print_pixels = strstr(argv[5], "pixelcmp") ? 1 : 0;
+    dump_blocks  = strstr(argv[5], "blockdump") ? 1 : 0;
+
+    for(i = 0; i < 2; i++) {
+        Y[i] = malloc(lsiz);
+        C[0][i] = malloc(csiz);
+        C[1][i] = malloc(csiz);
+
+        fd[i] = open(argv[1 + i], O_RDONLY);
+        if(fd[i] == -1) {
+            perror("open");
+            exit(1);
+        }
+        fcntl(fd[i], F_NOCACHE, 1);
+
+        if (to_skip)
+            lseek(fd[i], to_skip * (lsiz + 2*csiz), SEEK_SET);
+    }
+
+    mb_x = width / 16;
+    mb_y = height / 16;
+
+    mberrors = malloc(mb_x * mb_y);
+
+    while(!die) {
+        memset(mberrors, 0, mb_x * mb_y);
+
+        printf("Loading frame %d\n", ++fr);
+
+        for(i = 0; i < 2; i++) {
+            v = read(fd[i], Y[i], lsiz);
+            if(v != lsiz) {
+                fprintf(stderr, "Unable to read Y from file %d, exiting\n", i + 1);
+                return 1;
+            }
+        }
+
+
+        for(c = 0; c < lsiz; c++) {
+            if(Y[0][c] != Y[1][c]) {
+                x = c % width;
+                y = c / width;
+
+                mb = x / 16 + (y / 16) * mb_x;
+
+                if(print_pixels)
+                    printf("Luma diff 0x%02x != 0x%02x at pixel (%4d,%-4d) mb(%d,%d) #%d\n",
+                           Y[0][c],
+                           Y[1][c],
+                           x, y,
+                           x / 16,
+                           y / 16,
+                           mb);
+
+                mberrors[mb] |= 1;
+            }
+        }
+
+        /* Chroma planes */
+
+        for(p = 0; p < 2; p++) {
+
+            for(i = 0; i < 2; i++) {
+                v = read(fd[i], C[p][i], csiz);
+                if(v != csiz) {
+                    fprintf(stderr, "Unable to read %c from file %d, exiting\n",
+                            "UV"[p], i + 1);
+                    return 1;
+                }
+            }
+
+            for(c = 0; c < csiz; c++) {
+                if(C[p][0][c] != C[p][1][c]) {
+                    x = c % cwidth;
+                    y = c / cwidth;
+
+                    mb = x / 8 + (y / 8) * mb_x;
+
+                    mberrors[mb] |= 2 << p;
+
+                    if(print_pixels)
+
+                        printf("c%c diff 0x%02x != 0x%02x at pixel (%4d,%-4d) "
+                               "mb(%3d,%-3d) #%d\n",
+                               p ? 'r' : 'b',
+                               C[p][0][c],
+                               C[p][1][c],
+
+                               x, y,
+                               x / 8,
+                               y / 8,
+                               x / 8 + y / 8 * cwidth / 8);
+                }
+            }
+        }
+
+        for(i = 0; i < mb_x * mb_y; i++) {
+            x = i % mb_x;
+            y = i / mb_x;
+
+            if(mberrors[i]) {
+                die = 1;
+
+                printf("MB (%3d,%-3d) %4d %d %c%c%c damaged\n",
+                       x, y, i, mberrors[i],
+                       mberrors[i] & 1 ? 'Y' : ' ',
+                       mberrors[i] & 2 ? 'U' : ' ',
+                       mberrors[i] & 4 ? 'V' : ' ');
+
+                if(dump_blocks) {
+                    a = Y[0] + x * 16 + y * 16 * width;
+                    b = Y[1] + x * 16 + y * 16 * width;
+
+                    for(y = 0; y < 16; y++) {
+                        printf("%c ", "TB"[y&1]);
+                        for(x = 0; x < 16; x++)
+                            printf("%02x%c", a[x + y * width],
+                                   a[x + y * width] != b[x + y * width] ? '<' : ' ');
+
+                        printf("| ");
+                        for(x = 0; x < 16; x++)
+                            printf("%02x%c", b[x + y * width],
+                                   a[x + y * width] != b[x + y * width] ? '<' : ' ');
+
+                        printf("\n");
+                    }
+                }
+            }
+        }
+    }
+
+    return 0;
+}