Merge remote-tracking branch 'newdev/master'

* newdev/master: (33 commits)
  Fix an infinite loop when RoQ encoded generated a frame with a size greater than the maximum valid size.
  Add kbdwin.o to AC3 decoder
  Detect byte-swapped AC-3 and support decoding it directly.
  cosmetics: indentation
  Always copy input data for AC3 decoder.
  ac3enc: make sym_quant() branch-free
  cosmetics: indentation
  Add a CPU flag for the Atom processor.
  id3v2: skip broken tags with invalid size
  id3v2: don't explicitly skip padding
  Make sure kbhit() is in conio.h
  fate: update wmv8-drm reference
  vc1: make P-frame deblock filter bit-exact.
  configure: Add the -D parameter to the dlltool command
  amr: Set the AVFMT_GENERIC_INDEX flag
  amr: Set the pkt->pos field properly to the start of the packet
  amr: Set the codec->bit_rate field based on the last packet
  rtsp: Specify unicast for TCP interleaved streams, too
  Set the correct target for mingw64 dlltool
  applehttp: Change the variable for stream position in seconds into int64_t
  ...

Conflicts:
	ffmpeg.c
	ffplay.c
	libavcodec/ac3dec.c
	libavformat/avio.h
	libavformat/id3v2.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

configure

@@ -1040,7 +1040,6 @@ HAVE_LIST="
     bswap
     closesocket
     cmov
-    conio_h
     dcbzl
     dev_bktr_ioctl_bt848_h
     dev_bktr_ioctl_meteor_h
@@ -1070,6 +1069,7 @@ HAVE_LIST="
     inet_aton
     inline_asm
     isatty
+    kbhit
     ldbrx
     libdc1394_1
     libdc1394_2
@@ -1118,7 +1118,6 @@ HAVE_LIST="
     sys_soundcard_h
     sys_videoio_h
     ten_operands
-    termios_h
     threads
     truncf
     vfp_args
@@ -2423,7 +2422,7 @@ case $target_os in
         LIBTARGET=i386
         if enabled x86_64; then
             enable malloc_aligned
-            LIBTARGET=x64
+            LIBTARGET="i386:x86-64"
         elif enabled arm; then
             LIBTARGET=arm
         fi
@@ -2433,7 +2432,7 @@ case $target_os in
         SLIBSUF=".dll"
         SLIBNAME_WITH_VERSION='$(SLIBPREF)$(FULLNAME)-$(LIBVERSION)$(SLIBSUF)'
         SLIBNAME_WITH_MAJOR='$(SLIBPREF)$(FULLNAME)-$(LIBMAJOR)$(SLIBSUF)'
-        SLIB_EXTRA_CMD=-'$(DLLTOOL) -m $(LIBTARGET) -d $$(@:$(SLIBSUF)=.def) -l $(SUBDIR)$(SLIBNAME_WITH_MAJOR:$(SLIBSUF)=.lib)'
+        SLIB_EXTRA_CMD=-'$(DLLTOOL) -m $(LIBTARGET) -d $$(@:$(SLIBSUF)=.def) -l $(SUBDIR)$(SLIBNAME_WITH_MAJOR:$(SLIBSUF)=.lib) -D $(SLIBNAME_WITH_MAJOR)'
         SLIB_INSTALL_EXTRA_CMD='-install -m 644 $(SUBDIR)$(SLIBNAME_WITH_MAJOR:$(SLIBSUF)=.lib) "$(SHLIBDIR)/$(SLIBNAME:$(SLIBSUF)=.lib)"; \
             install -m 644 $(SUBDIR)$(SLIBNAME_WITH_MAJOR:$(SLIBSUF)=.lib) "$(SHLIBDIR)/$(SLIBNAME_WITH_MAJOR:$(SLIBSUF)=.lib)"; \
             install -d "$(LIBDIR)"; \
@@ -2794,6 +2793,7 @@ check_func  ${malloc_prefix}posix_memalign      && enable posix_memalign
 check_func  setrlimit
 check_func  strerror_r
 check_func  strtok_r
+check_func_headers conio.h kbhit
 check_func_headers io.h setmode
 check_func_headers lzo/lzo1x.h lzo1x_999_compress
 check_lib2 "windows.h psapi.h" GetProcessMemoryInfo -lpsapi
@@ -2801,7 +2801,6 @@ check_func_headers windows.h GetProcessTimes
 check_func_headers windows.h MapViewOfFile
 check_func_headers windows.h VirtualAlloc
 
-check_header conio.h
 check_header dlfcn.h
 check_header dxva2api.h
 check_header libcrystalhd/libcrystalhd_if.h
@@ -2810,7 +2809,6 @@ check_header poll.h
 check_header sys/mman.h
 check_header sys/resource.h
 check_header sys/select.h
-check_header termios.h
 check_header vdpau/vdpau.h
 check_header vdpau/vdpau_x11.h
 check_header X11/extensions/XvMClib.h

ffmpeg.c

@@ -69,12 +69,7 @@
 #include <sys/select.h>
 #endif
 
-#if HAVE_TERMIOS_H
-#include <fcntl.h>
-#include <sys/ioctl.h>
-#include <sys/time.h>
-#include <termios.h>
-#elif HAVE_CONIO_H
+#if HAVE_KBHIT
 #include <conio.h>
 #endif
 #include <time.h>
@@ -343,12 +338,6 @@ typedef struct AVInputFile {
     int nb_streams;       /* nb streams we are aware of */
 } AVInputFile;
 
-#if HAVE_TERMIOS_H
-
-/* init terminal so that we can grab keys */
-static struct termios oldtty;
-#endif
-
 #if CONFIG_AVFILTER
 
 static int configure_filters(AVInputStream *ist, AVOutputStream *ost)
@@ -436,9 +425,6 @@ static int configure_filters(AVInputStream *ist, AVOutputStream *ost)
 static void term_exit(void)
 {
     av_log(NULL, AV_LOG_QUIET, "");
-#if HAVE_TERMIOS_H
-    tcsetattr (0, TCSANOW, &oldtty);
-#endif
 }
 
 static volatile int received_sigterm = 0;
@@ -452,26 +438,6 @@ sigterm_handler(int sig)
 
 static void term_init(void)
 {
-#if HAVE_TERMIOS_H
-    struct termios tty;
-
-    tcgetattr (0, &tty);
-    oldtty = tty;
-    atexit(term_exit);
-
-    tty.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP
-                          |INLCR|IGNCR|ICRNL|IXON);
-    tty.c_oflag |= OPOST;
-    tty.c_lflag &= ~(ECHO|ECHONL|ICANON|IEXTEN);
-    tty.c_cflag &= ~(CSIZE|PARENB);
-    tty.c_cflag |= CS8;
-    tty.c_cc[VMIN] = 1;
-    tty.c_cc[VTIME] = 0;
-
-    tcsetattr (0, TCSANOW, &tty);
-    signal(SIGQUIT, sigterm_handler); /* Quit (POSIX).  */
-#endif
-
     signal(SIGINT , sigterm_handler); /* Interrupt (ANSI).  */
     signal(SIGTERM, sigterm_handler); /* Termination (ANSI).  */
 #ifdef SIGXCPU
@@ -482,25 +448,7 @@ static void term_init(void)
 /* read a key without blocking */
 static int read_key(void)
 {
-#if HAVE_TERMIOS_H
-    int n = 1;
-    unsigned char ch;
-    struct timeval tv;
-    fd_set rfds;
-
-    FD_ZERO(&rfds);
-    FD_SET(0, &rfds);
-    tv.tv_sec = 0;
-    tv.tv_usec = 0;
-    n = select(1, &rfds, NULL, NULL, &tv);
-    if (n > 0) {
-        n = read(0, &ch, 1);
-        if (n == 1)
-            return ch;
-
-        return n;
-    }
-#elif HAVE_CONIO_H
+#if HAVE_KBHIT
     if(kbhit())
         return(getch());
 #endif
@@ -2511,7 +2459,11 @@ static int transcode(AVFormatContext **output_files,
 
     if (!using_stdin) {
         if(verbose >= 0)
+#if HAVE_KBHIT
             fprintf(stderr, "Press [q] to stop encoding\n");
+#else
+            fprintf(stderr, "Press ctrl-c to stop encoding\n");
+#endif
         url_set_interrupt_cb(decode_interrupt_cb);
     }
     term_init();

libavcodec/Makefile

@@ -57,7 +57,7 @@ OBJS-$(CONFIG_AAC_ENCODER)             += aacenc.o aaccoder.o    \
                                           psymodel.o iirfilter.o \
                                           mpeg4audio.o kbdwin.o
 OBJS-$(CONFIG_AASC_DECODER)            += aasc.o msrledec.o
-OBJS-$(CONFIG_AC3_DECODER)             += ac3dec.o ac3dec_data.o ac3.o
+OBJS-$(CONFIG_AC3_DECODER)             += ac3dec.o ac3dec_data.o ac3.o kbdwin.o
 OBJS-$(CONFIG_AC3_ENCODER)             += ac3enc_float.o ac3tab.o ac3.o kbdwin.o
 OBJS-$(CONFIG_AC3_FIXED_ENCODER)       += ac3enc_fixed.o ac3tab.o ac3.o
 OBJS-$(CONFIG_ALAC_DECODER)            += alac.o

libavcodec/ac3dec.c

@@ -208,6 +208,11 @@ static av_cold int ac3_decode_init(AVCodecContext *avctx)
     }
     s->downmixed = 1;
 
+    /* allocate context input buffer */
+        s->input_buffer = av_mallocz(AC3_FRAME_BUFFER_SIZE + FF_INPUT_BUFFER_PADDING_SIZE);
+        if (!s->input_buffer)
+            return AVERROR(ENOMEM);
+
     avctx->sample_fmt = AV_SAMPLE_FMT_S16;
     return 0;
 }
@@ -1306,26 +1311,17 @@ static int ac3_decode_frame(AVCodecContext * avctx, void *data, int *data_size,
     int blk, ch, err;
     const uint8_t *channel_map;
     const float *output[AC3_MAX_CHANNELS];
-    // if it seems to be byte-swapped AC-3 (aka DNET)
-    int is_swapped = buf_size >= 2 && AV_RB16(buf) == 0x770B;
-
-    /* initialize the GetBitContext with the start of valid AC-3 Frame */
-    if (is_swapped || avctx->error_recognition >= FF_ER_CAREFUL) {
-        /* allocate context input buffer */
-        if (!s->input_buffer)
-            s->input_buffer = av_mallocz(AC3_FRAME_BUFFER_SIZE + FF_INPUT_BUFFER_PADDING_SIZE);
-        if (!s->input_buffer)
-            return AVERROR(ENOMEM);
 
-        /* copy input buffer to decoder context to avoid reading past the end
-           of the buffer, which can be caused by a damaged input stream. */
-        if (is_swapped) {
-            int cnt = FFMIN(buf_size, AC3_FRAME_BUFFER_SIZE) >> 1;
-            s->dsp.bswap16_buf((uint16_t *)s->input_buffer, (const uint16_t *)buf, cnt);
-        } else
+    /* copy input buffer to decoder context to avoid reading past the end
+       of the buffer, which can be caused by a damaged input stream. */
+    if (buf_size >= 2 && AV_RB16(buf) == 0x770B) {
+        // seems to be byte-swapped AC-3
+        int cnt = FFMIN(buf_size, AC3_FRAME_BUFFER_SIZE) >> 1;
+        s->dsp.bswap16_buf((uint16_t *)s->input_buffer, (const uint16_t *)buf, cnt);
+    } else
         memcpy(s->input_buffer, buf, FFMIN(buf_size, AC3_FRAME_BUFFER_SIZE));
-        buf = s->input_buffer;
-    }
+    buf = s->input_buffer;
+    /* initialize the GetBitContext with the start of valid AC-3 Frame */
     init_get_bits(&s->gbc, buf, buf_size * 8);
 
     /* parse the syncinfo */

libavcodec/ac3enc.c

@@ -1096,17 +1096,7 @@ static int compute_bit_allocation(AC3EncodeContext *s)
  */
 static inline int sym_quant(int c, int e, int levels)
 {
-    int v;
-
-    if (c >= 0) {
-        v = (levels * (c << e)) >> 24;
-        v = (v + 1) >> 1;
-        v = (levels >> 1) + v;
-    } else {
-        v = (levels * ((-c) << e)) >> 24;
-        v = (v + 1) >> 1;
-        v = (levels >> 1) - v;
-    }
+    int v = ((((levels * c) >> (24 - e)) + 1) >> 1) + (levels >> 1);
     av_assert2(v >= 0 && v < levels);
     return v;
 }

libavcodec/dct.h

@@ -47,4 +47,6 @@ struct DCTContext {
 int  ff_dct_init(DCTContext *s, int nbits, enum DCTTransformType type);
 void ff_dct_end (DCTContext *s);
 
+void ff_dct_init_mmx(DCTContext *s);
+
 #endif

libavcodec/fft.h

@@ -99,7 +99,6 @@ int ff_fft_init(FFTContext *s, int nbits, int inverse);
 void ff_fft_init_altivec(FFTContext *s);
 void ff_fft_init_mmx(FFTContext *s);
 void ff_fft_init_arm(FFTContext *s);
-void ff_dct_init_mmx(DCTContext *s);
 
 void ff_fft_end(FFTContext *s);
 

libavcodec/ffv1.c

@@ -792,7 +792,7 @@ static av_cold int init_slice_contexts(FFV1Context *f){
         fs->slice_x     = sxs;
         fs->slice_y     = sys;
 
-        fs->sample_buffer = av_malloc(6 * (fs->width+6) * sizeof(*fs->sample_buffer));
+        fs->sample_buffer = av_malloc(9 * (fs->width+6) * sizeof(*fs->sample_buffer));
         if (!fs->sample_buffer)
             return AVERROR(ENOMEM);
     }

libavcodec/roqvideoenc.c

@@ -898,9 +898,20 @@ static void roq_encode_video(RoqContext *enc)
     for (i=0; i<enc->width*enc->height/64; i++)
         gather_data_for_cel(tempData->cel_evals + i, enc, tempData);
 
-    /* Quake 3 can't handle chunks bigger than 65536 bytes */
-    if (tempData->mainChunkSize/8 > 65536) {
-        enc->lambda *= .8;
+    /* Quake 3 can't handle chunks bigger than 65535 bytes */
+    if (tempData->mainChunkSize/8 > 65535) {
+        av_log(enc->avctx, AV_LOG_ERROR,
+               "Warning, generated a frame too big (%d > 65535), "
+               "try using a smaller qscale value.\n",
+               tempData->mainChunkSize/8);
+        enc->lambda *= 1.5;
+        tempData->mainChunkSize = 0;
+        memset(tempData->used_option, 0, sizeof(tempData->used_option));
+        memset(tempData->codebooks.usedCB4, 0,
+               sizeof(tempData->codebooks.usedCB4));
+        memset(tempData->codebooks.usedCB2, 0,
+               sizeof(tempData->codebooks.usedCB2));
+
         goto retry_encode;
     }
 

libavcodec/vc1.c

@@ -863,6 +863,7 @@ int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)
         }
     }
     if(v->panscanflag) {
+        av_log_missing_feature(v->s.avctx, "Pan-scan", 0);
         //...
     }
     v->rnd = get_bits1(gb);

libavcodec/vc1.h

@@ -236,7 +236,7 @@ typedef struct VC1Context{
     //@}
     int ttfrm;            ///< Transform type info present at frame level
     uint8_t ttmbf;        ///< Transform type flag
-    uint8_t ttblk4x4;     ///< Value of ttblk which indicates a 4x4 transform
+    int *ttblk_base, *ttblk; ///< Transform type at the block level
     int codingset;        ///< index of current table set from 11.8 to use for luma block decoding
     int codingset2;       ///< index of current table set from 11.8 to use for chroma block decoding
     int pqindex;          ///< raw pqindex used in coding set selection
@@ -311,6 +311,8 @@ typedef struct VC1Context{
     int x8_type;
 
     uint32_t *cbp_base, *cbp;
+    uint8_t *is_intra_base, *is_intra;
+    int16_t (*luma_mv_base)[2], (*luma_mv)[2];
     uint8_t bfraction_lut_index;///< Index for BFRACTION value (see Table 40, reproduced into ff_vc1_bfraction_lut[])
     uint8_t broken_link;        ///< Broken link flag (BROKEN_LINK syntax element)
     uint8_t closed_entry;       ///< Closed entry point flag (CLOSED_ENTRY syntax element)

libavcodec/vc1dec.c

@@ -209,6 +209,8 @@ static void vc1_mc_1mv(VC1Context *v, int dir)
     }
     uvmx = (mx + ((mx & 3) == 3)) >> 1;
     uvmy = (my + ((my & 3) == 3)) >> 1;
+    v->luma_mv[s->mb_x][0] = uvmx;
+    v->luma_mv[s->mb_x][1] = uvmy;
     if(v->fastuvmc) {
         uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
         uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
@@ -477,6 +479,7 @@ static void vc1_mc_4mv_chroma(VC1Context *v)
     } else {
         s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
         s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
+        v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;
         return; //no need to do MC for inter blocks
     }
 
@@ -484,6 +487,8 @@ static void vc1_mc_4mv_chroma(VC1Context *v)
     s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
     uvmx = (tx + ((tx&3) == 3)) >> 1;
     uvmy = (ty + ((ty&3) == 3)) >> 1;
+    v->luma_mv[s->mb_x][0] = uvmx;
+    v->luma_mv[s->mb_x][1] = uvmy;
     if(v->fastuvmc) {
         uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
         uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
@@ -652,8 +657,9 @@ static void vc1_mc_4mv_chroma(VC1Context *v)
 
 /** Predict and set motion vector
  */
-static inline void vc1_pred_mv(MpegEncContext *s, int n, int dmv_x, int dmv_y, int mv1, int r_x, int r_y, uint8_t* is_intra)
+static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y, int mv1, int r_x, int r_y, uint8_t* is_intra)
 {
+    MpegEncContext *s = &v->s;
     int xy, wrap, off = 0;
     int16_t *A, *B, *C;
     int px, py;
@@ -678,6 +684,7 @@ static inline void vc1_pred_mv(MpegEncContext *s, int n, int dmv_x, int dmv_y, i
             s->current_picture.motion_val[0][xy + wrap][1] = 0;
             s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
             s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
+            v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;
             s->current_picture.motion_val[1][xy + 1][0] = 0;
             s->current_picture.motion_val[1][xy + 1][1] = 0;
             s->current_picture.motion_val[1][xy + wrap][0] = 0;
@@ -1953,7 +1960,7 @@ static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int c
 /** Decode P block
  */
 static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,
-                              uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left)
+                              uint8_t *dst, int linesize, int skip_block, int *ttmb_out)
 {
     MpegEncContext *s = &v->s;
     GetBitContext *gb = &s->gb;
@@ -2011,10 +2018,6 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
             else{
                 v->vc1dsp.vc1_inv_trans_8x8_add(dst, linesize, block);
             }
-            if(apply_filter && cbp_top  & 0xC)
-                v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
-            if(apply_filter && cbp_left & 0xA)
-                v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);
         }
         break;
     case TT_4X4:
@@ -2038,10 +2041,6 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
                     v->vc1dsp.vc1_inv_trans_4x4_dc(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
                 else
                     v->vc1dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
-                if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))
-                    v->vc1dsp.vc1_v_loop_filter4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, v->pq);
-                if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))
-                    v->vc1dsp.vc1_h_loop_filter4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, v->pq);
             }
         }
         break;
@@ -2066,10 +2065,6 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
                     v->vc1dsp.vc1_inv_trans_8x4_dc(dst + j*4*linesize, linesize, block + off);
                 else
                     v->vc1dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
-                if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))
-                    v->vc1dsp.vc1_v_loop_filter8(dst + j*4*linesize, linesize, v->pq);
-                if(apply_filter && cbp_left & (2 << j))
-                    v->vc1dsp.vc1_h_loop_filter4(dst + j*4*linesize, linesize, v->pq);
             }
         }
         break;
@@ -2094,14 +2089,12 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
                     v->vc1dsp.vc1_inv_trans_4x8_dc(dst + j*4, linesize, block + off);
                 else
                     v->vc1dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
-                if(apply_filter && cbp_top & (2 << j))
-                    v->vc1dsp.vc1_v_loop_filter4(dst + j*4, linesize, v->pq);
-                if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))
-                    v->vc1dsp.vc1_h_loop_filter8(dst + j*4, linesize, v->pq);
             }
         }
         break;
     }
+    if (ttmb_out)
+        *ttmb_out |= ttblk << (n * 4);
     return pat;
 }
 
@@ -2110,6 +2103,155 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
 static const int size_table  [6] = { 0, 2, 3, 4,  5,  8 };
 static const int offset_table[6] = { 0, 1, 3, 7, 15, 31 };
 
+static av_always_inline void vc1_apply_p_v_loop_filter(VC1Context *v, int block_num)
+{
+    MpegEncContext *s = &v->s;
+    int mb_cbp         = v->cbp[s->mb_x - s->mb_stride],
+        block_cbp      = mb_cbp      >> (block_num * 4), bottom_cbp,
+        mb_is_intra    = v->is_intra[s->mb_x - s->mb_stride],
+        block_is_intra = mb_is_intra >> (block_num * 4), bottom_is_intra;
+    int idx, linesize = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;
+    uint8_t *dst;
+
+    if(block_num > 3) {
+        dst      = s->dest[block_num - 3];
+    } else {
+        dst      = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 8) * linesize;
+    }
+    if (s->mb_y != s->mb_height || block_num < 2) {
+        int16_t (*mv)[2];
+        int mv_stride;
+
+        if(block_num > 3) {
+            bottom_cbp      = v->cbp[s->mb_x]      >> (block_num * 4);
+            bottom_is_intra = v->is_intra[s->mb_x] >> (block_num * 4);
+            mv              = &v->luma_mv[s->mb_x - s->mb_stride];
+            mv_stride       = s->mb_stride;
+        } else {
+            bottom_cbp      = (block_num < 2) ? (mb_cbp               >> ((block_num + 2) * 4)) :
+                                                (v->cbp[s->mb_x]      >> ((block_num - 2) * 4));
+            bottom_is_intra = (block_num < 2) ? (mb_is_intra          >> ((block_num + 2) * 4)) :
+                                                (v->is_intra[s->mb_x] >> ((block_num - 2) * 4));
+            mv_stride       = s->b8_stride;
+            mv              = &s->current_picture.motion_val[0][s->block_index[block_num] - 2 * mv_stride];
+        }
+
+        if (bottom_is_intra & 1 || block_is_intra & 1 ||
+            mv[0][0] != mv[mv_stride][0] || mv[0][1] != mv[mv_stride][1]) {
+            v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
+        } else {
+            idx = ((bottom_cbp >> 2) | block_cbp) & 3;
+            if(idx == 3) {
+                v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
+            } else if (idx) {
+                if (idx == 1)
+                    v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);
+                else
+                    v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);
+            }
+        }
+    }
+
+    dst -= 4 * linesize;
+    ttblk = (v->ttblk[s->mb_x - s->mb_stride] >> (block_num * 4)) & 0xf;
+    if (ttblk == TT_4X4 || ttblk == TT_8X4) {
+        idx = (block_cbp | (block_cbp >> 2)) & 3;
+        if (idx == 3) {
+            v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
+        } else if (idx) {
+            if (idx == 1)
+                v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);
+            else
+                v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);
+        }
+    }
+}
+
+static av_always_inline void vc1_apply_p_h_loop_filter(VC1Context *v, int block_num)
+{
+    MpegEncContext *s = &v->s;
+    int mb_cbp         = v->cbp[s->mb_x - 1 - s->mb_stride],
+        block_cbp      = mb_cbp      >> (block_num * 4), right_cbp,
+        mb_is_intra    = v->is_intra[s->mb_x - 1 - s->mb_stride],
+        block_is_intra = mb_is_intra >> (block_num * 4), right_is_intra;
+    int idx, linesize = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;
+    uint8_t *dst;
+
+    if (block_num > 3) {
+        dst = s->dest[block_num - 3] - 8 * linesize;
+    } else {
+        dst = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 16) * linesize - 8;
+    }
+
+    if (s->mb_x != s->mb_width || !(block_num & 5)) {
+        int16_t (*mv)[2];
+
+        if(block_num > 3) {
+            right_cbp      = v->cbp[s->mb_x - s->mb_stride] >> (block_num * 4);
+            right_is_intra = v->is_intra[s->mb_x - s->mb_stride] >> (block_num * 4);
+            mv             = &v->luma_mv[s->mb_x - s->mb_stride - 1];
+        }else{
+            right_cbp      = (block_num & 1) ? (v->cbp[s->mb_x - s->mb_stride]      >> ((block_num - 1) * 4)) :
+                                               (mb_cbp                              >> ((block_num + 1) * 4));
+            right_is_intra = (block_num & 1) ? (v->is_intra[s->mb_x - s->mb_stride] >> ((block_num - 1) * 4)) :
+                                               (mb_is_intra                         >> ((block_num + 1) * 4));
+            mv             = &s->current_picture.motion_val[0][s->block_index[block_num] - s->b8_stride * 2 - 2];
+        }
+        if (block_is_intra & 1 || right_is_intra & 1 || mv[0][0] != mv[1][0] || mv[0][1] != mv[1][1]) {
+            v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);
+        } else {
+            idx = ((right_cbp >> 1) | block_cbp) & 5; // FIXME check
+            if (idx == 5) {
+                v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);
+            } else if (idx) {
+                if (idx == 1)
+                    v->vc1dsp.vc1_h_loop_filter4(dst+4*linesize, linesize, v->pq);
+                else
+                    v->vc1dsp.vc1_h_loop_filter4(dst,            linesize, v->pq);
+            }
+        }
+    }
+
+    dst -= 4;
+    ttblk = (v->ttblk[s->mb_x - s->mb_stride - 1] >> (block_num * 4)) & 0xf;
+    if (ttblk == TT_4X4 || ttblk == TT_4X8) {
+        idx = (block_cbp | (block_cbp >> 1)) & 5;
+        if (idx == 5) {
+            v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);
+        } else if (idx) {
+            if (idx == 1)
+                v->vc1dsp.vc1_h_loop_filter4(dst + linesize*4, linesize, v->pq);
+            else
+                v->vc1dsp.vc1_h_loop_filter4(dst,              linesize, v->pq);
+        }
+    }
+}
+
+static void vc1_apply_p_loop_filter(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    int i;
+
+    for (i = 0; i < 6; i++) {
+        vc1_apply_p_v_loop_filter(v, i);
+    }
+
+    /* V always preceedes H, therefore we run H one MB before V;
+     * at the end of a row, we catch up to complete the row */
+    if (s->mb_x) {
+        for (i = 0; i < 6; i++) {
+            vc1_apply_p_h_loop_filter(v, i);
+        }
+        if (s->mb_x == s->mb_width - 1) {
+            s->mb_x++;
+            ff_update_block_index(s);
+            for (i = 0; i < 6; i++) {
+                vc1_apply_p_h_loop_filter(v, i);
+            }
+        }
+    }
+}
+
 /** Decode one P-frame MB (in Simple/Main profile)
  */
 static int vc1_decode_p_mb(VC1Context *v)
@@ -2129,8 +2271,7 @@ static int vc1_decode_p_mb(VC1Context *v)
     int first_block = 1;
     int dst_idx, off;
     int skipped, fourmv;
-    int block_cbp = 0, pat;
-    int apply_loop_filter;
+    int block_cbp = 0, pat, block_tt = 0, block_intra = 0;
 
     mquant = v->pq; /* Loosy initialization */
 
@@ -2143,7 +2284,6 @@ static int vc1_decode_p_mb(VC1Context *v)
     else
         skipped = v->s.mbskip_table[mb_pos];
 
-    apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);
     if (!fourmv) /* 1MV mode */
     {
         if (!skipped)
@@ -2157,7 +2297,7 @@ static int vc1_decode_p_mb(VC1Context *v)
                 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
             }
             s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
-            vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
+            vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
 
             /* FIXME Set DC val for inter block ? */
             if (s->mb_intra && !mb_has_coeffs)
@@ -2211,38 +2351,10 @@ static int vc1_decode_p_mb(VC1Context *v)
                         if(v->a_avail)
                             v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
                     }
-                    if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
-                        int left_cbp, top_cbp;
-                        if(i & 4){
-                            left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);
-                            top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
-                        }else{
-                            left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));
-                            top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
-                        }
-                        if(left_cbp & 0xC)
-                            v->vc1dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
-                        if(top_cbp  & 0xA)
-                            v->vc1dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
-                    }
                     block_cbp |= 0xF << (i << 2);
+                    block_intra |= 1 << i;
                 } else if(val) {
-                    int left_cbp = 0, top_cbp = 0, filter = 0;
-                    if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
-                        filter = 1;
-                        if(i & 4){
-                            left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);
-                            top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
-                        }else{
-                            left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));
-                            top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
-                        }
-                        if(left_cbp & 0xC)
-                            v->vc1dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
-                        if(top_cbp  & 0xA)
-                            v->vc1dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
-                    }
-                    pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);
+                    pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);
                     block_cbp |= pat << (i << 2);
                     if(!v->ttmbf && ttmb < 8) ttmb = -1;
                     first_block = 0;
@@ -2258,9 +2370,8 @@ static int vc1_decode_p_mb(VC1Context *v)
             }
             s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;
             s->current_picture.qscale_table[mb_pos] = 0;
-            vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
+            vc1_pred_mv(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
             vc1_mc_1mv(v, 0);
-            return 0;
         }
     } //1MV mode
     else //4MV mode
@@ -2284,7 +2395,7 @@ static int vc1_decode_p_mb(VC1Context *v)
                     if(val) {
                         GET_MVDATA(dmv_x, dmv_y);
                     }
-                    vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
+                    vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
                     if(!s->mb_intra) vc1_mc_4mv_luma(v, i);
                     intra_count += s->mb_intra;
                     is_intra[i] = s->mb_intra;
@@ -2299,8 +2410,9 @@ static int vc1_decode_p_mb(VC1Context *v)
                 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];
             }
             // if there are no coded blocks then don't do anything more
-            if(!intra_count && !coded_inter) return 0;
             dst_idx = 0;
+            if(!intra_count && !coded_inter)
+                goto end;
             GET_MQUANT();
             s->current_picture.qscale_table[mb_pos] = mquant;
             /* test if block is intra and has pred */
@@ -2344,44 +2456,15 @@ static int vc1_decode_p_mb(VC1Context *v)
                         if(v->a_avail)
                             v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
                     }
-                    if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
-                        int left_cbp, top_cbp;
-                        if(i & 4){
-                            left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);
-                            top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
-                        }else{
-                            left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));
-                            top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
-                        }
-                        if(left_cbp & 0xC)
-                            v->vc1dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
-                        if(top_cbp  & 0xA)
-                            v->vc1dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
-                    }
                     block_cbp |= 0xF << (i << 2);
+                    block_intra |= 1 << i;
                 } else if(is_coded[i]) {
-                    int left_cbp = 0, top_cbp = 0, filter = 0;
-                    if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
-                        filter = 1;
-                        if(i & 4){
-                            left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);
-                            top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
-                        }else{
-                            left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));
-                            top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
-                        }
-                        if(left_cbp & 0xC)
-                            v->vc1dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
-                        if(top_cbp  & 0xA)
-                            v->vc1dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
-                    }
-                    pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);
+                    pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);
                     block_cbp |= pat << (i << 2);
                     if(!v->ttmbf && ttmb < 8) ttmb = -1;
                     first_block = 0;
                 }
             }
-            return 0;
         }
         else //Skipped MB
         {
@@ -2393,18 +2476,19 @@ static int vc1_decode_p_mb(VC1Context *v)
             }
             for (i=0; i<4; i++)
             {
-                vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
+                vc1_pred_mv(v, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
                 vc1_mc_4mv_luma(v, i);
             }
             vc1_mc_4mv_chroma(v);
             s->current_picture.qscale_table[mb_pos] = 0;
-            return 0;
         }
     }
+end:
     v->cbp[s->mb_x] = block_cbp;
+    v->ttblk[s->mb_x] = block_tt;
+    v->is_intra[s->mb_x] = block_intra;
 
-    /* Should never happen */
-    return -1;
+    return 0;
 }
 
 /** Decode one B-frame MB (in Main profile)
@@ -2546,7 +2630,7 @@ static void vc1_decode_b_mb(VC1Context *v)
                        i & 4 ? s->uvlinesize : s->linesize,
                        s->block[i]);
         } else if(val) {
-            vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), 0, 0, 0);
+            vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), NULL);
             if(!v->ttmbf && ttmb < 8) ttmb = -1;
             first_block = 0;
         }
@@ -2686,7 +2770,7 @@ static void vc1_decode_i_blocks(VC1Context *v)
 
 /** Decode blocks of I-frame for advanced profile
  */
-static void vc1_decode_i_blocks_adv(VC1Context *v)
+static void vc1_decode_i_blocks_adv(VC1Context *v, int mby_start, int mby_end)
 {
     int k;
     MpegEncContext *s = &v->s;
@@ -2728,8 +2812,15 @@ static void vc1_decode_i_blocks_adv(VC1Context *v)
     s->mb_x = s->mb_y = 0;
     s->mb_intra = 1;
     s->first_slice_line = 1;
+    s->mb_y = mby_start;
+    if (mby_start) {
+        s->mb_x = 0;
+        ff_init_block_index(s);
+        memset(&s->coded_block[s->block_index[0]-s->b8_stride], 0,
+               s->b8_stride * sizeof(*s->coded_block));
+    }
     idct8x8_fn = v->vc1dsp.vc1_inv_trans_8x8_put_signed[0];
-    for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
+    for(; s->mb_y < mby_end; s->mb_y++) {
         s->mb_x = 0;
         ff_init_block_index(s);
         for(;s->mb_x < s->mb_width; s->mb_x++) {
@@ -2815,7 +2906,7 @@ static void vc1_decode_i_blocks_adv(VC1Context *v)
             if(v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq);
 
             if(get_bits_count(&s->gb) > v->bits) {
-                ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
+                ff_er_add_slice(s, 0, mby_start, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
                 return;
             }
@@ -2828,12 +2919,13 @@ static void vc1_decode_i_blocks_adv(VC1Context *v)
     }
     if (v->s.loop_filter)
         ff_draw_horiz_band(s, (s->mb_height-1)*16, 16);
-    ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
+    ff_er_add_slice(s, 0, mby_start, s->mb_width - 1, mby_end - 1, (AC_END|DC_END|MV_END));
 }
 
-static void vc1_decode_p_blocks(VC1Context *v)
+static void vc1_decode_p_blocks(VC1Context *v, int mby_start, int mby_end)
 {
     MpegEncContext *s = &v->s;
+    int apply_loop_filter;
 
     /* select codingmode used for VLC tables selection */
     switch(v->c_ac_table_index){
@@ -2860,29 +2952,45 @@ static void vc1_decode_p_blocks(VC1Context *v)
         break;
     }
 
+    apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);
     s->first_slice_line = 1;
     memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);
-    for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
+    for(s->mb_y = mby_start; s->mb_y < mby_end; s->mb_y++) {
         s->mb_x = 0;
         ff_init_block_index(s);
         for(; s->mb_x < s->mb_width; s->mb_x++) {
             ff_update_block_index(s);
 
             vc1_decode_p_mb(v);
+            if (s->mb_y != mby_start && apply_loop_filter)
+                vc1_apply_p_loop_filter(v);
             if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
-                ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
+                ff_er_add_slice(s, 0, mby_start, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits,s->mb_x,s->mb_y);
                 return;
             }
         }
         memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride);
-        ff_draw_horiz_band(s, s->mb_y * 16, 16);
+        memmove(v->ttblk_base, v->ttblk, sizeof(v->ttblk_base[0])*s->mb_stride);
+        memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0])*s->mb_stride);
+        memmove(v->luma_mv_base, v->luma_mv, sizeof(v->luma_mv_base[0])*s->mb_stride);
+        if (s->mb_y != mby_start) ff_draw_horiz_band(s, (s->mb_y-1) * 16, 16);
         s->first_slice_line = 0;
     }
-    ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
+    if (apply_loop_filter) {
+        s->mb_x = 0;
+        ff_init_block_index(s);
+        for (; s->mb_x < s->mb_width; s->mb_x++) {
+            ff_update_block_index(s);
+            vc1_apply_p_loop_filter(v);
+        }
+    }
+    if (mby_end >= mby_start)
+        ff_draw_horiz_band(s, (mby_end-1) * 16, 16);
+    ff_er_add_slice(s, 0, mby_start, s->mb_width - 1, mby_end - 1, (AC_END|DC_END|MV_END));
 }
 
-static void vc1_decode_b_blocks(VC1Context *v)
+static void vc1_decode_b_blocks(VC1Context *v, int mby_start, int mby_end)
 {
     MpegEncContext *s = &v->s;
 
@@ -2912,7 +3020,7 @@ static void vc1_decode_b_blocks(VC1Context *v)
     }
 
     s->first_slice_line = 1;
-    for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
+    for(s->mb_y = mby_start; s->mb_y < mby_end; s->mb_y++) {
         s->mb_x = 0;
         ff_init_block_index(s);
         for(; s->mb_x < s->mb_width; s->mb_x++) {
@@ -2920,7 +3028,7 @@ static void vc1_decode_b_blocks(VC1Context *v)
 
             vc1_decode_b_mb(v);
             if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
-                ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
+                ff_er_add_slice(s, 0, mby_start, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits,s->mb_x,s->mb_y);
                 return;
             }
@@ -2934,7 +3042,7 @@ static void vc1_decode_b_blocks(VC1Context *v)
     }
     if (v->s.loop_filter)
         ff_draw_horiz_band(s, (s->mb_height-1)*16, 16);
-    ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
+    ff_er_add_slice(s, 0, mby_start, s->mb_width - 1, mby_end - 1, (AC_END|DC_END|MV_END));
 }
 
 static void vc1_decode_skip_blocks(VC1Context *v)
@@ -2956,18 +3064,17 @@ static void vc1_decode_skip_blocks(VC1Context *v)
     s->pict_type = FF_P_TYPE;
 }
 
-static void vc1_decode_blocks(VC1Context *v)
+static void vc1_decode_blocks(VC1Context *v, int mby_start, int mby_end)
 {
 
     v->s.esc3_level_length = 0;
     if(v->x8_type){
         ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );
     }else{
-
         switch(v->s.pict_type) {
         case FF_I_TYPE:
             if(v->profile == PROFILE_ADVANCED)
-                vc1_decode_i_blocks_adv(v);
+                vc1_decode_i_blocks_adv(v, mby_start, mby_end);
             else
                 vc1_decode_i_blocks(v);
             break;
@@ -2975,16 +3082,16 @@ static void vc1_decode_blocks(VC1Context *v)
             if(v->p_frame_skipped)
                 vc1_decode_skip_blocks(v);
             else
-                vc1_decode_p_blocks(v);
+                vc1_decode_p_blocks(v, mby_start, mby_end);
             break;
         case FF_B_TYPE:
             if(v->bi_type){
                 if(v->profile == PROFILE_ADVANCED)
-                    vc1_decode_i_blocks_adv(v);
+                    vc1_decode_i_blocks_adv(v, mby_start, mby_end);
                 else
                     vc1_decode_i_blocks(v);
             }else
-                vc1_decode_b_blocks(v);
+                vc1_decode_b_blocks(v, mby_start, mby_end);
             break;
         }
     }
@@ -3116,6 +3223,12 @@ static av_cold int vc1_decode_init(AVCodecContext *avctx)
 
     v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);
     v->cbp = v->cbp_base + s->mb_stride;
+    v->ttblk_base = av_malloc(sizeof(v->ttblk_base[0]) * 2 * s->mb_stride);
+    v->ttblk = v->ttblk_base + s->mb_stride;
+    v->is_intra_base = av_malloc(sizeof(v->is_intra_base[0]) * 2 * s->mb_stride);
+    v->is_intra = v->is_intra_base + s->mb_stride;
+    v->luma_mv_base = av_malloc(sizeof(v->luma_mv_base[0]) * 2 * s->mb_stride);
+    v->luma_mv = v->luma_mv_base + s->mb_stride;
 
     /* allocate block type info in that way so it could be used with s->block_index[] */
     v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);
@@ -3145,12 +3258,17 @@ static int vc1_decode_frame(AVCodecContext *avctx,
                             AVPacket *avpkt)
 {
     const uint8_t *buf = avpkt->data;
-    int buf_size = avpkt->size;
+    int buf_size = avpkt->size, n_slices = 0, i;
     VC1Context *v = avctx->priv_data;
     MpegEncContext *s = &v->s;
     AVFrame *pict = data;
     uint8_t *buf2 = NULL;
     const uint8_t *buf_start = buf;
+    struct {
+        uint8_t *buf;
+        GetBitContext gb;
+        int mby_start;
+    } *slices = NULL;
 
     /* no supplementary picture */
     if (buf_size == 0 || (buf_size == 4 && AV_RB32(buf) == VC1_CODE_ENDOFSEQ)) {
@@ -3205,10 +3323,20 @@ static int vc1_decode_frame(AVCodecContext *avctx,
                     init_get_bits(&s->gb, buf2, buf_size2*8);
                     vc1_decode_entry_point(avctx, v, &s->gb);
                     break;
-                case VC1_CODE_SLICE:
-                    av_log(avctx, AV_LOG_ERROR, "Sliced decoding is not implemented (yet)\n");
-                    av_free(buf2);
-                    return -1;
+                case VC1_CODE_SLICE: {
+                    int buf_size3;
+                    slices = av_realloc(slices, sizeof(*slices) * (n_slices+1));
+                    if (!slices) goto err;
+                    slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
+                    if (!slices[n_slices].buf) goto err;
+                    buf_size3 = vc1_unescape_buffer(start + 4, size,
+                                                    slices[n_slices].buf);
+                    init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,
+                                  buf_size3 << 3);
+                    slices[n_slices].mby_start = get_bits(&slices[n_slices].gb, 9);
+                    n_slices++;
+                    break;
+                }
                 }
             }
         }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */
@@ -3217,15 +3345,14 @@ static int vc1_decode_frame(AVCodecContext *avctx,
             divider = find_next_marker(buf, buf + buf_size);
             if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){
                 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");
-                av_free(buf2);
-                return -1;
+                goto err;
             }
 
             buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);
             // TODO
             if(!v->warn_interlaced++)
                 av_log(v->s.avctx, AV_LOG_ERROR, "Interlaced WVC1 support is not implemented\n");
-            av_free(buf2);return -1;
+            goto err;
         }else{
             buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);
         }
@@ -3235,19 +3362,16 @@ static int vc1_decode_frame(AVCodecContext *avctx,
     // do parse frame header
     if(v->profile < PROFILE_ADVANCED) {
         if(vc1_parse_frame_header(v, &s->gb) == -1) {
-            av_free(buf2);
-            return -1;
+            goto err;
         }
     } else {
         if(vc1_parse_frame_header_adv(v, &s->gb) == -1) {
-            av_free(buf2);
-            return -1;
+            goto err;
         }
     }
 
     if(v->res_sprite && (s->pict_type!=FF_I_TYPE)){
-        av_free(buf2);
-        return -1;
+        goto err;
     }
 
     // for hurry_up==5
@@ -3256,33 +3380,29 @@ static int vc1_decode_frame(AVCodecContext *avctx,
 
     /* skip B-frames if we don't have reference frames */
     if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){
-        av_free(buf2);
-        return -1;//buf_size;
+        goto err;
     }
     /* skip b frames if we are in a hurry */
     if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;
     if(   (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
        || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
        ||  avctx->skip_frame >= AVDISCARD_ALL) {
-        av_free(buf2);
-        return buf_size;
+        goto end;
     }
     /* skip everything if we are in a hurry>=5 */
     if(avctx->hurry_up>=5) {
-        av_free(buf2);
-        return -1;//buf_size;
+        goto err;
     }
 
     if(s->next_p_frame_damaged){
         if(s->pict_type==FF_B_TYPE)
-            return buf_size;
+            goto end;
         else
             s->next_p_frame_damaged=0;
     }
 
     if(MPV_frame_start(s, avctx) < 0) {
-        av_free(buf2);
-        return -1;
+        goto err;
     }
 
     s->me.qpel_put= s->dsp.put_qpel_pixels_tab;
@@ -3293,17 +3413,23 @@ static int vc1_decode_frame(AVCodecContext *avctx,
         ff_vdpau_vc1_decode_picture(s, buf_start, (buf + buf_size) - buf_start);
     else if (avctx->hwaccel) {
         if (avctx->hwaccel->start_frame(avctx, buf, buf_size) < 0)
-            return -1;
+            goto err;
         if (avctx->hwaccel->decode_slice(avctx, buf_start, (buf + buf_size) - buf_start) < 0)
-            return -1;
+            goto err;
         if (avctx->hwaccel->end_frame(avctx) < 0)
-            return -1;
+            goto err;
     } else {
         ff_er_frame_start(s);
 
         v->bits = buf_size * 8;
-        vc1_decode_blocks(v);
-//av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);
+        for (i = 0; i <= n_slices; i++) {
+            if (i && get_bits1(&s->gb))
+                vc1_parse_frame_header_adv(v, &s->gb);
+            vc1_decode_blocks(v, i == 0 ? 0 : FFMAX(0, slices[i-1].mby_start),
+                i == n_slices ? s->mb_height : FFMIN(s->mb_height, slices[i].mby_start));
+            if (i != n_slices) s->gb = slices[i].gb;
+        }
+//av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), s->gb.size_in_bits);
 //  if(get_bits_count(&s->gb) > buf_size * 8)
 //      return -1;
         ff_er_frame_end(s);
@@ -3324,8 +3450,19 @@ assert(s->current_picture.pict_type == s->pict_type);
         ff_print_debug_info(s, pict);
     }
 
+end:
     av_free(buf2);
+    for (i = 0; i < n_slices; i++)
+        av_free(slices[i].buf);
+    av_free(slices);
     return buf_size;
+
+err:
+    av_free(buf2);
+    for (i = 0; i < n_slices; i++)
+        av_free(slices[i].buf);
+    av_free(slices);
+    return -1;
 }
 
 
@@ -3345,6 +3482,9 @@ static av_cold int vc1_decode_end(AVCodecContext *avctx)
     av_freep(&v->over_flags_plane);
     av_freep(&v->mb_type_base);
     av_freep(&v->cbp_base);
+    av_freep(&v->ttblk_base);
+    av_freep(&v->is_intra_base); // FIXME use v->mb_type[]
+    av_freep(&v->luma_mv_base);
     ff_intrax8_common_end(&v->x8);
     return 0;
 }

libavformat/amr.c

@@ -121,6 +121,7 @@ static int amr_read_packet(AVFormatContext *s,
 {
     AVCodecContext *enc = s->streams[0]->codec;
     int read, size = 0, toc, mode;
+    int64_t pos = avio_tell(s->pb);
 
     if (url_feof(s->pb))
     {
@@ -153,8 +154,11 @@ static int amr_read_packet(AVFormatContext *s,
         return AVERROR(EIO);
     }
 
+    /* Both AMR formats have 50 frames per second */
+    s->streams[0]->codec->bit_rate = size*8*50;
+
     pkt->stream_index = 0;
-    pkt->pos= avio_tell(s->pb);
+    pkt->pos = pos;
     pkt->data[0]=toc;
     pkt->duration= enc->codec_id == CODEC_ID_AMR_NB ? 160 : 320;
     read = avio_read(s->pb, pkt->data+1, size-1);
@@ -177,6 +181,7 @@ AVInputFormat ff_amr_demuxer = {
     amr_read_header,
     amr_read_packet,
     NULL,
+    .flags = AVFMT_GENERIC_INDEX,
 };
 #endif
 

libavformat/applehttp.c

@@ -496,7 +496,8 @@ static int applehttp_read_seek(AVFormatContext *s, int stream_index,
                                int64_t timestamp, int flags)
 {
     AppleHTTPContext *c = s->priv_data;
-    int pos = 0, i;
+    int64_t pos = 0;
+    int i;
     struct variant *var = c->variants[0];
 
     if ((flags & AVSEEK_FLAG_BYTE) || !c->finished)

libavformat/asfdec.c

@@ -25,6 +25,7 @@
 #include "libavutil/avstring.h"
 #include "libavcodec/mpegaudio.h"
 #include "avformat.h"
+#include "avio_internal.h"
 #include "riff.h"
 #include "asf.h"
 #include "asfcrypt.h"
@@ -1241,7 +1242,7 @@ static int asf_read_seek(AVFormatContext *s, int stream_index, int64_t pts, int
 
     /* Try using the protocol's read_seek if available */
     if(s->pb) {
-        int ret = av_url_read_fseek(s->pb, stream_index, pts, flags);
+        int ret = ffio_read_seek(s->pb, stream_index, pts, flags);
         if(ret >= 0)
             asf_reset_header(s);
         if (ret != AVERROR(ENOSYS))

libavformat/avio.h

@@ -415,6 +415,9 @@ attribute_deprecated void         put_tag(AVIOContext *s, const char *tag);
  * @}
  */
 
+attribute_deprecated int     av_url_read_fpause(AVIOContext *h,    int pause);
+attribute_deprecated int64_t av_url_read_fseek( AVIOContext *h,    int stream_index,
+                                                int64_t timestamp, int flags);
 
 /**
  * @defgroup old_url_f_funcs Old url_f* functions
@@ -512,10 +515,6 @@ int64_t avio_size(AVIOContext *s);
  */
 int url_feof(AVIOContext *s);
 
-int av_url_read_fpause(AVIOContext *h, int pause);
-int64_t av_url_read_fseek(AVIOContext *h, int stream_index,
-                          int64_t timestamp, int flags);
-
 /** @warning currently size is limited */
 #ifdef __GNUC__
 int avio_printf(AVIOContext *s, const char *fmt, ...) __attribute__ ((__format__ (__printf__, 2, 3)));
@@ -622,9 +621,10 @@ int url_resetbuf(AVIOContext *s, int flags);
 int avio_open(AVIOContext **s, const char *url, int flags);
 
 int avio_close(AVIOContext *s);
-URLContext *url_fileno(AVIOContext *s);
 
 #if FF_API_OLD_AVIO
+attribute_deprecated URLContext *url_fileno(AVIOContext *s);
+
 /**
  * @deprecated use AVIOContext.max_packet_size directly.
  */

libavformat/avio_internal.h

@@ -66,4 +66,9 @@ uint64_t ffio_read_varlen(AVIOContext *bc);
 /** @warning must be called before any I/O */
 int ffio_set_buf_size(AVIOContext *s, int buf_size);
 
+int     ffio_read_pause(AVIOContext *h,    int pause);
+int64_t ffio_read_seek( AVIOContext *h,    int stream_index,
+                        int64_t timestamp, int flags);
+
+
 #endif // AVFORMAT_AVIO_INTERNAL_H

libavformat/aviobuf.c

@@ -406,6 +406,15 @@ void put_flush_packet(AVIOContext *s)
 {
     avio_flush(s);
 }
+int av_url_read_fpause(AVIOContext *s, int pause)
+{
+    return ffio_read_pause(s, pause);
+}
+int64_t av_url_read_fseek(AVIOContext *s, int stream_index,
+                         int64_t timestamp, int flags)
+{
+    return ffio_read_seek(s, stream_index, timestamp, flags);
+}
 #endif
 
 int avio_put_str(AVIOContext *s, const char *str)
@@ -932,10 +941,12 @@ int avio_close(AVIOContext *s)
     return url_close(h);
 }
 
+#if FF_API_OLD_AVIO
 URLContext *url_fileno(AVIOContext *s)
 {
     return s->opaque;
 }
+#endif
 
 int avio_printf(AVIOContext *s, const char *fmt, ...)
 {
@@ -978,15 +989,15 @@ int url_fget_max_packet_size(AVIOContext *s)
 }
 #endif
 
-int av_url_read_fpause(AVIOContext *s, int pause)
+int ffio_read_pause(AVIOContext *s, int pause)
 {
     if (!s->read_pause)
         return AVERROR(ENOSYS);
     return s->read_pause(s->opaque, pause);
 }
 
-int64_t av_url_read_fseek(AVIOContext *s, int stream_index,
-                          int64_t timestamp, int flags)
+int64_t ffio_read_seek(AVIOContext *s, int stream_index,
+                       int64_t timestamp, int flags)
 {
     URLContext *h = s->opaque;
     int64_t ret;

libavformat/flvdec.c

@@ -28,6 +28,7 @@
 #include "libavcodec/bytestream.h"
 #include "libavcodec/mpeg4audio.h"
 #include "avformat.h"
+#include "avio_internal.h"
 #include "flv.h"
 
 typedef struct {
@@ -461,7 +462,7 @@ leave:
 static int flv_read_seek(AVFormatContext *s, int stream_index,
     int64_t ts, int flags)
 {
-    return av_url_read_fseek(s->pb, stream_index, ts, flags);
+    return ffio_read_seek(s->pb, stream_index, ts, flags);
 }
 
 #if 0 /* don't know enough to implement this */
@@ -482,7 +483,7 @@ static int flv_read_seek2(AVFormatContext *s, int stream_index,
             ts = av_rescale_rnd(ts, 1000, AV_TIME_BASE,
                 flags & AVSEEK_FLAG_BACKWARD ? AV_ROUND_DOWN : AV_ROUND_UP);
         }
-        ret = av_url_read_fseek(s->pb, stream_index, ts, flags);
+        ret = ffio_read_seek(s->pb, stream_index, ts, flags);
     }
 
     if (ret == AVERROR(ENOSYS))

libavformat/id3v2.c

@@ -187,9 +187,9 @@ static void ff_id3v2_parse(AVFormatContext *s, int len, uint8_t version, uint8_t
     int isv34, unsync;
     unsigned tlen;
     char tag[5];
-    int64_t next;
+    int64_t next, end = avio_tell(s->pb) + len;
     int taghdrlen;
-    const char *reason;
+    const char *reason = NULL;
     AVIOContext pb;
     unsigned char *buffer = NULL;
     int buffer_size = 0;
@@ -282,20 +282,15 @@ static void ff_id3v2_parse(AVFormatContext *s, int len, uint8_t version, uint8_t
         avio_seek(s->pb, next, SEEK_SET);
     }
 
-    if (len > 0) {
-        /* Skip padding */
-        avio_skip(s->pb, len);
-    }
     if (version == 4 && flags & 0x10) /* Footer preset, always 10 bytes, skip over it */
-        avio_skip(s->pb, 10);
-
-    av_free(buffer);
-    return;
+        end += 10;
 
   error:
-    av_log(s, AV_LOG_INFO, "ID3v2.%d tag skipped, cannot handle %s\n", version, reason);
-    avio_seek(s->pb, len, SEEK_CUR);
+    if (reason)
+        av_log(s, AV_LOG_INFO, "ID3v2.%d tag skipped, cannot handle %s\n", version, reason);
+    avio_seek(s->pb, end, SEEK_SET);
     av_free(buffer);
+    return;
 }
 
 void ff_id3v2_read(AVFormatContext *s, const char *magic)

libavformat/mp3dec.c

@@ -90,7 +90,7 @@ static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
         return -1;
 
     /* Check for Xing / Info tag */
-    avio_seek(s->pb, xing_offtbl[c.lsf == 1][c.nb_channels == 1], SEEK_CUR);
+    avio_skip(s->pb, xing_offtbl[c.lsf == 1][c.nb_channels == 1]);
     v = avio_rb32(s->pb);
     if(v == MKBETAG('X', 'i', 'n', 'g') || v == MKBETAG('I', 'n', 'f', 'o')) {
         v = avio_rb32(s->pb);

libavformat/rmenc.c

@@ -436,7 +436,7 @@ static int rm_write_trailer(AVFormatContext *s)
 
     if (!url_is_streamed(s->pb)) {
         /* end of file: finish to write header */
-        index_pos = avio_seek(pb, 0, SEEK_CUR);
+        index_pos = avio_tell(pb);
         data_size = index_pos - rm->data_pos;
 
         /* FIXME: write index */

libavformat/rtsp.c

@@ -1157,7 +1157,7 @@ int ff_rtsp_make_setup_request(AVFormatContext *s, const char *host, int port,
                 continue;
             snprintf(transport, sizeof(transport) - 1,
                      "%s/TCP;", trans_pref);
-            if (rt->server_type == RTSP_SERVER_WMS)
+            if (rt->transport != RTSP_TRANSPORT_RDT)
                 av_strlcat(transport, "unicast;", sizeof(transport));
             av_strlcatf(transport, sizeof(transport),
                         "interleaved=%d-%d",

libavformat/utils.c

@@ -2554,7 +2554,7 @@ int av_read_play(AVFormatContext *s)
     if (s->iformat->read_play)
         return s->iformat->read_play(s);
     if (s->pb)
-        return av_url_read_fpause(s->pb, 0);
+        return ffio_read_pause(s->pb, 0);
     return AVERROR(ENOSYS);
 }
 
@@ -2563,7 +2563,7 @@ int av_read_pause(AVFormatContext *s)
     if (s->iformat->read_pause)
         return s->iformat->read_pause(s);
     if (s->pb)
-        return av_url_read_fpause(s->pb, 1);
+        return ffio_read_pause(s->pb, 1);
     return AVERROR(ENOSYS);
 }
 

libavformat/wtv.c

@@ -701,10 +701,10 @@ static AVStream * parse_media_type(AVFormatContext *s, AVStream *st, int sid,
             return NULL;
         if (!ff_guidcmp(formattype, format_videoinfo2)) {
             int consumed = parse_videoinfoheader2(s, st);
-            avio_seek(pb, FFMAX(size - consumed, 0), SEEK_CUR);
+            avio_skip(pb, FFMAX(size - consumed, 0));
         } else if (!ff_guidcmp(formattype, format_mpeg2_video)) {
             int consumed = parse_videoinfoheader2(s, st);
-            avio_seek(pb, FFMAX(size - consumed, 0), SEEK_CUR);
+            avio_skip(pb, FFMAX(size - consumed, 0));
         } else {
             if (ff_guidcmp(formattype, format_none))
                 av_log(s, AV_LOG_WARNING, "unknown formattype:"PRI_GUID"\n", ARG_GUID(formattype));

libavutil/cpu.c

@@ -58,6 +58,7 @@ int main(void)
            cpu_flags & AV_CPU_FLAG_SSE3     ? "SSE3 "       : "",
            cpu_flags & AV_CPU_FLAG_SSE3SLOW ? "SSE3(slow) " : "",
            cpu_flags & AV_CPU_FLAG_SSSE3    ? "SSSE3 "      : "",
+           cpu_flags & AV_CPU_FLAG_ATOM     ? "Atom "       : "",
            cpu_flags & AV_CPU_FLAG_SSE4     ? "SSE4.1 "     : "",
            cpu_flags & AV_CPU_FLAG_SSE42    ? "SSE4.2 "     : "",
            cpu_flags & AV_CPU_FLAG_AVX      ? "AVX "        : "",

libavutil/cpu.h

@@ -34,6 +34,7 @@
 #define AV_CPU_FLAG_SSE3         0x0040 ///< Prescott SSE3 functions
 #define AV_CPU_FLAG_SSE3SLOW 0x20000000 ///< SSE3 supported, but usually not faster
 #define AV_CPU_FLAG_SSSE3        0x0080 ///< Conroe SSSE3 functions
+#define AV_CPU_FLAG_ATOM     0x10000000 ///< Atom processor, some SSSE3 instructions are slower
 #define AV_CPU_FLAG_SSE4         0x0100 ///< Penryn SSE4.1 functions
 #define AV_CPU_FLAG_SSE42        0x0200 ///< Nehalem SSE4.2 functions
 #define AV_CPU_FLAG_AVX          0x4000 ///< AVX functions: requires OS support even if YMM registers aren't used

libavutil/x86/cpu.c

@@ -135,16 +135,24 @@ int ff_get_cpu_flags_x86(void)
         }
     }
 
-    if (!strncmp(vendor.c, "GenuineIntel", 12) &&
-        family == 6 && (model == 9 || model == 13 || model == 14)) {
-        /* 6/9 (pentium-m "banias"), 6/13 (pentium-m "dothan"), and 6/14 (core1 "yonah")
-         * theoretically support sse2, but it's usually slower than mmx,
-         * so let's just pretend they don't. AV_CPU_FLAG_SSE2 is disabled and
-         * AV_CPU_FLAG_SSE2SLOW is enabled so that SSE2 is not used unless
-         * explicitly enabled by checking AV_CPU_FLAG_SSE2SLOW. The same
-         * situation applies for AV_CPU_FLAG_SSE3 and AV_CPU_FLAG_SSE3SLOW. */
-        if (rval & AV_CPU_FLAG_SSE2) rval ^= AV_CPU_FLAG_SSE2SLOW|AV_CPU_FLAG_SSE2;
-        if (rval & AV_CPU_FLAG_SSE3) rval ^= AV_CPU_FLAG_SSE3SLOW|AV_CPU_FLAG_SSE3;
+    if (!strncmp(vendor.c, "GenuineIntel", 12)) {
+        if (family == 6 && (model == 9 || model == 13 || model == 14)) {
+            /* 6/9 (pentium-m "banias"), 6/13 (pentium-m "dothan"), and 6/14 (core1 "yonah")
+            * theoretically support sse2, but it's usually slower than mmx,
+            * so let's just pretend they don't. AV_CPU_FLAG_SSE2 is disabled and
+            * AV_CPU_FLAG_SSE2SLOW is enabled so that SSE2 is not used unless
+            * explicitly enabled by checking AV_CPU_FLAG_SSE2SLOW. The same
+            * situation applies for AV_CPU_FLAG_SSE3 and AV_CPU_FLAG_SSE3SLOW. */
+            if (rval & AV_CPU_FLAG_SSE2) rval ^= AV_CPU_FLAG_SSE2SLOW|AV_CPU_FLAG_SSE2;
+            if (rval & AV_CPU_FLAG_SSE3) rval ^= AV_CPU_FLAG_SSE3SLOW|AV_CPU_FLAG_SSE3;
+        }
+        /* The Atom processor has SSSE3 support, which is useful in many cases,
+         * but sometimes the SSSE3 version is slower than the SSE2 equivalent
+         * on the Atom, but is generally faster on other processors supporting
+         * SSSE3. This flag allows for selectively disabling certain SSSE3
+         * functions on the Atom. */
+        if (family == 6 && model == 28)
+            rval |= AV_CPU_FLAG_ATOM;
     }
 
     return rval;

tests/ref/fate/vc1

@@ -2,14 +2,14 @@
 0, 3600, 38016, 0xf4715db5
 0, 7200, 38016, 0xf4715db5
 0, 10800, 38016, 0xf46af0e1
-0, 14400, 38016, 0x96992cf1
-0, 18000, 38016, 0xbaadd874
-0, 21600, 38016, 0x751f4328
-0, 25200, 38016, 0x751f4328
-0, 28800, 38016, 0xf7294772
-0, 32400, 38016, 0xf7294772
-0, 36000, 38016, 0xf1d12133
-0, 39600, 38016, 0xf1d12133
-0, 43200, 38016, 0xf1d12133
-0, 46800, 38016, 0xf1d12133
-0, 50400, 38016, 0xf1d12133
+0, 14400, 38016, 0x9c1c2cf1
+0, 18000, 38016, 0xff12d87f
+0, 21600, 38016, 0x7408432b
+0, 25200, 38016, 0x7408432b
+0, 28800, 38016, 0x8d11479a
+0, 32400, 38016, 0x8d11479a
+0, 36000, 38016, 0xc4a121ab
+0, 39600, 38016, 0xc4a121ab
+0, 43200, 38016, 0xc4a121ab
+0, 46800, 38016, 0xc4a121ab
+0, 50400, 38016, 0xc4a121ab

tests/ref/fate/wmv8-drm

@@ -34,129 +34,129 @@
 0, 123750, 84480, 0xabf7c25d
 0, 127500, 84480, 0x912600ee
 0, 131250, 84480, 0x7ee7c70b
-0, 135000, 84480, 0xe212b0d1
-0, 138750, 84480, 0xd4fa6c16
-0, 142500, 84480, 0xd10fa126
-0, 146250, 84480, 0xda91d3aa
-0, 150000, 84480, 0xd90be940
-0, 153750, 84480, 0x908a009b
-0, 157500, 84480, 0xdd26d6e9
-0, 161250, 84480, 0xfeda8de9
-0, 165000, 84480, 0x96d66505
-0, 168750, 84480, 0xf43b4b1b
-0, 172500, 84480, 0xd44122c2
-0, 176250, 84480, 0xc65da7a9
-0, 180000, 84480, 0xbf4178f2
-0, 183750, 84480, 0x96be846a
-0, 187500, 84480, 0x73e6459e
-0, 191250, 84480, 0x70086917
-0, 195000, 84480, 0x9b12571d
-0, 198750, 84480, 0xe4ce7bce
-0, 202500, 84480, 0x845672e5
-0, 206250, 84480, 0x1c17e189
-0, 210000, 84480, 0x12a877d0
-0, 213750, 84480, 0xca62e8ed
-0, 217500, 84480, 0x87beb28f
-0, 221250, 84480, 0x1ba915a4
-0, 225000, 84480, 0x159fc9c4
-0, 228750, 84480, 0xcec3e3ef
-0, 232500, 84480, 0x96a42f48
-0, 236250, 84480, 0xcd17decf
-0, 240000, 84480, 0x4cf95d6c
-0, 243750, 84480, 0xa258400e
-0, 247500, 84480, 0xb9c566d8
-0, 251250, 84480, 0xc39f6dab
-0, 255000, 84480, 0xd08a0880
-0, 258750, 84480, 0x41e3a70f
-0, 262500, 84480, 0x0ded83d9
-0, 266250, 84480, 0x74a915ea
-0, 270000, 84480, 0x00c2c849
-0, 273750, 84480, 0x709a9b07
-0, 277500, 84480, 0xfb276daa
-0, 281250, 84480, 0x89ca744b
-0, 285000, 84480, 0x457ba569
-0, 288750, 84480, 0xc37c8d57
-0, 292500, 84480, 0x1890a97e
-0, 296250, 84480, 0xe455b8d0
-0, 300000, 84480, 0x1f6bdbce
-0, 303750, 84480, 0x9df90ed1
-0, 307500, 84480, 0xff80328c
-0, 311250, 84480, 0x0e115940
-0, 315000, 84480, 0x09ddc11d
-0, 318750, 84480, 0xeef241e9
-0, 322500, 84480, 0xbe3578b4
-0, 326250, 84480, 0x2a83f0f3
-0, 330000, 84480, 0xb41d01c2
-0, 333750, 84480, 0x630efd97
-0, 337500, 84480, 0xd592140d
-0, 341250, 84480, 0x7110f627
-0, 345000, 84480, 0x2dbaa590
-0, 348750, 84480, 0x13d80be5
-0, 352500, 84480, 0xb8f3740d
-0, 356250, 84480, 0xd9a6fcef
-0, 360000, 84480, 0x56bfa4e3
-0, 363750, 84480, 0x49364550
-0, 367500, 84480, 0x0a2f1690
-0, 371250, 84480, 0xb87fccbf
-0, 375000, 84480, 0xb2aa69f7
-0, 378750, 84480, 0x176bf378
-0, 382500, 84480, 0xe7a6de45
-0, 386250, 84480, 0x8e3900a6
-0, 390000, 84480, 0x24a72099
-0, 393750, 84480, 0x5394df95
-0, 397500, 84480, 0xd648c92e
-0, 401250, 84480, 0xab95fb62
-0, 405000, 84480, 0xe3239cb0
-0, 408750, 84480, 0xb4f93467
-0, 412500, 84480, 0x94d0d01d
-0, 416250, 84480, 0x6316a25e
-0, 420000, 84480, 0x6316a25e
-0, 423750, 84480, 0x6316a25e
-0, 427500, 84480, 0x6316a25e
-0, 431250, 84480, 0x6316a25e
-0, 435000, 84480, 0x6316a25e
-0, 438750, 84480, 0x6316a25e
+0, 135000, 84480, 0x09c5b0d1
+0, 138750, 84480, 0x6dbe6c0c
+0, 142500, 84480, 0x0fe0a120
+0, 146250, 84480, 0x2352d3a2
+0, 150000, 84480, 0xb22ce92e
+0, 153750, 84480, 0x31db0099
+0, 157500, 84480, 0xad2dd73a
+0, 161250, 84480, 0xb9af8e20
+0, 165000, 84480, 0x7b956549
+0, 168750, 84480, 0x3f774b87
+0, 172500, 84480, 0x824a23a3
+0, 176250, 84480, 0x4469a8d8
+0, 180000, 84480, 0xc80c7a0a
+0, 183750, 84480, 0xcf958549
+0, 187500, 84480, 0x449746e3
+0, 191250, 84480, 0xbac66a82
+0, 195000, 84480, 0x99e85855
+0, 198750, 84480, 0xa4a17d17
+0, 202500, 84480, 0xe29c7587
+0, 206250, 84480, 0x551de592
+0, 210000, 84480, 0xe0877bce
+0, 213750, 84480, 0x9660eb35
+0, 217500, 84480, 0x0a34b644
+0, 221250, 84480, 0x352919f0
+0, 225000, 84480, 0xef56ce27
+0, 228750, 84480, 0x030fe862
+0, 232500, 84480, 0x2eba33e2
+0, 236250, 84480, 0x242de401
+0, 240000, 84480, 0xbadd61ca
+0, 243750, 84480, 0x2060465b
+0, 247500, 84480, 0x256e6965
+0, 251250, 84480, 0x243b7084
+0, 255000, 84480, 0x8b3c0b47
+0, 258750, 84480, 0xc174a9af
+0, 262500, 84480, 0xb6d48686
+0, 266250, 84480, 0xa3dd1871
+0, 270000, 84480, 0x04cdcaf7
+0, 273750, 84480, 0x55f89c94
+0, 277500, 84480, 0xda657032
+0, 281250, 84480, 0x38ba7698
+0, 285000, 84480, 0x4d03a7f2
+0, 288750, 84480, 0x115d9035
+0, 292500, 84480, 0x24c6acc6
+0, 296250, 84480, 0xdd2bbcae
+0, 300000, 84480, 0xb4fee0b9
+0, 303750, 84480, 0xc51c14e0
+0, 307500, 84480, 0xfb7737de
+0, 311250, 84480, 0x38675fb0
+0, 315000, 84480, 0x4752c710
+0, 318750, 84480, 0xfeb7491b
+0, 322500, 84480, 0xaa248122
+0, 326250, 84480, 0x9a4af87c
+0, 330000, 84480, 0xedcf09df
+0, 333750, 84480, 0x563a05df
+0, 337500, 84480, 0x0dde1e03
+0, 341250, 84480, 0xd8f0ff65
+0, 345000, 84480, 0xbeb9ae1a
+0, 348750, 84480, 0x416d1468
+0, 352500, 84480, 0x66c87d4c
+0, 356250, 84480, 0xa67c0774
+0, 360000, 84480, 0xd8f8aec1
+0, 363750, 84480, 0xadfa502b
+0, 367500, 84480, 0x50bf20e4
+0, 371250, 84480, 0xbcb3d8cc
+0, 375000, 84480, 0xa54677d7
+0, 378750, 84480, 0x3566042d
+0, 382500, 84480, 0x4c9eed57
+0, 386250, 84480, 0xc3b90e58
+0, 390000, 84480, 0x3c042bfa
+0, 393750, 84480, 0x19f8e890
+0, 397500, 84480, 0xd3dacfb9
+0, 401250, 84480, 0x2365fc6f
+0, 405000, 84480, 0xa2c19d00
+0, 408750, 84480, 0xce94336f
+0, 412500, 84480, 0xfa9bcf14
+0, 416250, 84480, 0x24d6a243
+0, 420000, 84480, 0x24d6a243
+0, 423750, 84480, 0x24d6a243
+0, 427500, 84480, 0x24d6a243
+0, 431250, 84480, 0x24d6a243
+0, 435000, 84480, 0x24d6a243
+0, 438750, 84480, 0x24d6a243
 0, 442500, 84480, 0xae1c8854
-0, 446250, 84480, 0x30fe68bf
-0, 450000, 84480, 0xfd1435c8
-0, 453750, 84480, 0xddf57fab
-0, 457500, 84480, 0xc3553a51
-0, 461250, 84480, 0xd9ce7ae8
-0, 465000, 84480, 0x671765cb
-0, 468750, 84480, 0x78336eab
-0, 472500, 84480, 0xb66b659c
-0, 476250, 84480, 0x0e1f68bb
-0, 480000, 84480, 0x8f4669dc
-0, 483750, 84480, 0x60c47360
-0, 487500, 84480, 0x93037246
-0, 491250, 84480, 0xdebe620d
-0, 495000, 84480, 0x7555161e
-0, 498750, 84480, 0xcbaf4311
-0, 502500, 84480, 0x8e3783a0
-0, 506250, 84480, 0x3888008f
-0, 510000, 84480, 0x8a4fa114
-0, 513750, 84480, 0xac0bf10b
-0, 517500, 84480, 0xb485ff7f
-0, 521250, 84480, 0xeaec2133
-0, 525000, 84480, 0x105827cd
-0, 528750, 84480, 0x55ff4b2d
-0, 532500, 84480, 0x78c64a49
-0, 536250, 84480, 0x3897731d
-0, 540000, 84480, 0x9ca891aa
-0, 543750, 84480, 0x24a6ab0a
-0, 547500, 84480, 0x88fdc6fb
-0, 551250, 84480, 0x05afea61
-0, 555000, 84480, 0xe703e2cf
-0, 558750, 84480, 0x9eb0e64a
-0, 562500, 84480, 0x92a7f0ab
-0, 566250, 84480, 0xec2bfbfb
-0, 570000, 84480, 0x7db600ad
-0, 573750, 84480, 0x4abd6393
-0, 577500, 84480, 0x54ce06b5
-0, 581250, 84480, 0x7cb6f150
-0, 585000, 84480, 0x099d4aeb
-0, 588750, 84480, 0x0bcfdc31
-0, 592500, 84480, 0xe4e72d1f
-0, 596250, 84480, 0xbce22331
-0, 600000, 84480, 0x020545d7
-0, 603750, 84480, 0x71869e48
+0, 446250, 84480, 0xbb8968bf
+0, 450000, 84480, 0x6f923623
+0, 453750, 84480, 0x22e98029
+0, 457500, 84480, 0x8ac33af3
+0, 461250, 84480, 0x05947b6e
+0, 465000, 84480, 0xfc35661a
+0, 468750, 84480, 0x0e6b6e47
+0, 472500, 84480, 0x82c764bb
+0, 476250, 84480, 0x57a36833
+0, 480000, 84480, 0xc8dd690a
+0, 483750, 84480, 0x02c47232
+0, 487500, 84480, 0x6645715d
+0, 491250, 84480, 0xc64860f7
+0, 495000, 84480, 0x4f5614b3
+0, 498750, 84480, 0xa70842ca
+0, 502500, 84480, 0x379d8458
+0, 506250, 84480, 0xa14701cf
+0, 510000, 84480, 0xad1aa2b2
+0, 513750, 84480, 0xee28f320
+0, 517500, 84480, 0x505801e9
+0, 521250, 84480, 0x7947233b
+0, 525000, 84480, 0x3ce72a9d
+0, 528750, 84480, 0xa6834e64
+0, 532500, 84480, 0xfebf4d70
+0, 536250, 84480, 0x4a0775e2
+0, 540000, 84480, 0x9d7e945b
+0, 543750, 84480, 0xaa9eadd9
+0, 547500, 84480, 0xaa85c9b1
+0, 551250, 84480, 0xa005edaf
+0, 555000, 84480, 0x7fc4e5cc
+0, 558750, 84480, 0xb0f6e8d1
+0, 562500, 84480, 0x9ef9f330
+0, 566250, 84480, 0xbe14ff1f
+0, 570000, 84480, 0xd494048c
+0, 573750, 84480, 0x046166a7
+0, 577500, 84480, 0x052a09b2
+0, 581250, 84480, 0x71fff4ab
+0, 585000, 84480, 0xb9684e41
+0, 588750, 84480, 0x1ddce068
+0, 592500, 84480, 0xb9de300e
+0, 596250, 84480, 0x13962590
+0, 600000, 84480, 0xde79482f
+0, 603750, 84480, 0x7d1ca064