Merge remote-tracking branch 'qatar/master'

* qatar/master:
  id3v2: fix doxy comment - 'machine byte order' makes no sense on char arrays
  VC1: restore mistakenly removed code
  twinvq: check output buffer size before decoding
  twinvq: return an error when the packet size is too small
  lavf: export some forgotten symbols with non-av prefixes.
  swscale: update altivec yuv2planeX asm to new per-plane API.
  swscale: make yuv2yuvX_10_sse2/avx 8/9/16-bits aware.
  yuv2planeX10 SIMD
  swscale: decide whether to use yuv2plane1/X on a per-plane basis.
  swscale: reintroduce full precision in 16-bit output.
  Split up yuv2yuvX functions
  Split out yuv2yuv1 luma and chroma in order to make them generic DSP functions
  lavc: replace references to deprecated AVCodecContext.error_recognition to use AVCodecContext.err_recognition
  lavc: translate non-flag-based er options into flag-based ef options at codec open
  add -err_filter AVOptions to access flag-based error recognition
  h264_weight: initialize "height" function argument properly.
  presets: spelling error in libvpx 1080p50_60
  avplay: fix fullscreen behaviour with SDL 1.2.14 on Mac OS X

Conflicts:
	ffplay.c
	libavformat/libavformat.v
	libswscale/swscale.c
	libswscale/x86/swscale_template.c
	tests/ref/lavfi/pixfmts_scale
Merged-by: Michael Niedermayer <michaelni@gmx.at>

libavcodec/aacdec.c

@@ -596,7 +596,7 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
             int ret = set_default_channel_config(avctx, new_che_pos, ac->m4ac.chan_config);
             if (!ret)
                 output_configure(ac, ac->che_pos, new_che_pos, ac->m4ac.chan_config, OC_GLOBAL_HDR);
-            else if (avctx->error_recognition >= FF_ER_EXPLODE)
+            else if (avctx->err_recognition & AV_EF_EXPLODE)
                 return AVERROR_INVALIDDATA;
         }
     }

libavcodec/ac3dec.c

@@ -1359,7 +1359,7 @@ static int ac3_decode_frame(AVCodecContext * avctx, void *data, int *data_size,
         if (s->frame_size > buf_size) {
             av_log(avctx, AV_LOG_ERROR, "incomplete frame\n");
             err = AAC_AC3_PARSE_ERROR_FRAME_SIZE;
-        } else if (avctx->error_recognition >= FF_ER_CAREFUL) {
+        } else if (avctx->err_recognition & AV_EF_CRCCHECK) {
             /* check for crc mismatch */
             if (av_crc(av_crc_get_table(AV_CRC_16_ANSI), 0, &buf[2], s->frame_size-2)) {
                 av_log(avctx, AV_LOG_ERROR, "frame CRC mismatch\n");

libavcodec/alsdec.c

@@ -393,7 +393,7 @@ static av_cold int read_specific_config(ALSDecContext *ctx)
         if (get_bits_left(&gb) < 32)
             return -1;
 
-        if (avctx->error_recognition >= FF_ER_CAREFUL) {
+        if (avctx->err_recognition & AV_EF_CRCCHECK) {
             ctx->crc_table = av_crc_get_table(AV_CRC_32_IEEE_LE);
             ctx->crc       = 0xFFFFFFFF;
             ctx->crc_org   = ~get_bits_long(&gb, 32);
@@ -1476,7 +1476,7 @@ static int decode_frame(AVCodecContext *avctx,
     }
 
     // update CRC
-    if (sconf->crc_enabled && avctx->error_recognition >= FF_ER_CAREFUL) {
+    if (sconf->crc_enabled && (avctx->err_recognition & AV_EF_CRCCHECK)) {
         int swap = HAVE_BIGENDIAN != sconf->msb_first;
 
         if (ctx->avctx->bits_per_raw_sample == 24) {
@@ -1710,7 +1710,7 @@ static av_cold int decode_init(AVCodecContext *avctx)
 
     // allocate crc buffer
     if (HAVE_BIGENDIAN != sconf->msb_first && sconf->crc_enabled &&
-        avctx->error_recognition >= FF_ER_CAREFUL) {
+        (avctx->err_recognition & AV_EF_CRCCHECK)) {
         ctx->crc_buffer = av_malloc(sizeof(*ctx->crc_buffer) *
                                     ctx->cur_frame_length *
                                     avctx->channels *

libavcodec/h261dec.c

@@ -136,7 +136,7 @@ static int h261_decode_gob_header(H261Context *h){
 
     if(s->qscale==0) {
         av_log(s->avctx, AV_LOG_ERROR, "qscale has forbidden 0 value\n");
-        if (s->avctx->error_recognition >= FF_ER_COMPLIANT)
+        if (s->avctx->err_recognition & AV_EF_BITSTREAM)
             return -1;
     }
 

libavcodec/h263dec.c

@@ -732,7 +732,7 @@ intrax8_decoded:
 av_log(avctx, AV_LOG_DEBUG, "%"PRId64"\n", rdtsc()-time);
 #endif
 
-    return (ret && avctx->error_recognition >= FF_ER_EXPLODE)?ret:get_consumed_bytes(s, buf_size);
+    return (ret && (avctx->err_recognition & AV_EF_EXPLODE))?ret:get_consumed_bytes(s, buf_size);
 }
 
 AVCodec ff_h263_decoder = {

libavcodec/h264.c

@@ -2893,7 +2893,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
             ff_thread_report_progress((AVFrame*)s->current_picture_ptr, INT_MAX, 1);
             ff_generate_sliding_window_mmcos(h);
             if (ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index) < 0 &&
-                s->avctx->error_recognition >= FF_ER_EXPLODE)
+                (s->avctx->err_recognition & AV_EF_EXPLODE))
                 return AVERROR_INVALIDDATA;
             /* Error concealment: if a ref is missing, copy the previous ref in its place.
              * FIXME: avoiding a memcpy would be nice, but ref handling makes many assumptions
@@ -3072,7 +3072,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
     }
 
     if(h->nal_ref_idc && ff_h264_decode_ref_pic_marking(h0, &s->gb) < 0 &&
-       s->avctx->error_recognition >= FF_ER_EXPLODE)
+       (s->avctx->err_recognition & AV_EF_EXPLODE))
         return AVERROR_INVALIDDATA;
 
     if(FRAME_MBAFF){

libavcodec/h264_refs.c

@@ -654,7 +654,7 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
 
     print_short_term(h);
     print_long_term(h);
-    return h->s.avctx->error_recognition >= FF_ER_EXPLODE ? err : 0;
+    return (h->s.avctx->err_recognition & AV_EF_EXPLODE) ? err : 0;
 }
 
 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb){

libavcodec/mjpegbdec.c

@@ -82,7 +82,7 @@ read_header:
         init_get_bits(&s->gb, buf_ptr+dqt_offs, (buf_end - (buf_ptr+dqt_offs))*8);
         s->start_code = DQT;
         if (ff_mjpeg_decode_dqt(s) < 0 &&
-            avctx->error_recognition >= FF_ER_EXPLODE)
+            (avctx->err_recognition & AV_EF_EXPLODE))
           return AVERROR_INVALIDDATA;
     }
 
@@ -116,7 +116,7 @@ read_header:
         s->mjpb_skiptosod = (sod_offs - sos_offs - show_bits(&s->gb, 16));
         s->start_code = SOS;
         if (ff_mjpeg_decode_sos(s, NULL, NULL) < 0 &&
-            avctx->error_recognition >= FF_ER_EXPLODE)
+            (avctx->err_recognition & AV_EF_EXPLODE))
           return AVERROR_INVALIDDATA;
     }
 

libavcodec/mjpegdec.c

@@ -1522,7 +1522,7 @@ eoi_parser:
                         break;
                     }
                     if (ff_mjpeg_decode_sos(s, NULL, NULL) < 0 &&
-                        avctx->error_recognition >= FF_ER_EXPLODE)
+                        (avctx->err_recognition & AV_EF_EXPLODE))
                       return AVERROR_INVALIDDATA;
                     break;
                 case DRI:

libavcodec/mpeg12.c

@@ -1378,7 +1378,7 @@ static int mpeg1_decode_picture(AVCodecContext *avctx,
     if (s->pict_type == AV_PICTURE_TYPE_P || s->pict_type == AV_PICTURE_TYPE_B) {
         s->full_pel[0] = get_bits1(&s->gb);
         f_code = get_bits(&s->gb, 3);
-        if (f_code == 0 && avctx->error_recognition >= FF_ER_COMPLIANT)
+        if (f_code == 0 && (avctx->err_recognition & AV_EF_BITSTREAM))
             return -1;
         s->mpeg_f_code[0][0] = f_code;
         s->mpeg_f_code[0][1] = f_code;
@@ -1386,7 +1386,7 @@ static int mpeg1_decode_picture(AVCodecContext *avctx,
     if (s->pict_type == AV_PICTURE_TYPE_B) {
         s->full_pel[1] = get_bits1(&s->gb);
         f_code = get_bits(&s->gb, 3);
-        if (f_code == 0 && avctx->error_recognition >= FF_ER_COMPLIANT)
+        if (f_code == 0 && (avctx->err_recognition & AV_EF_BITSTREAM))
             return -1;
         s->mpeg_f_code[1][0] = f_code;
         s->mpeg_f_code[1][1] = f_code;
@@ -1819,7 +1819,7 @@ static int mpeg_decode_slice(Mpeg1Context *s1, int mb_y,
                              && s->progressive_frame == 0 /* vbv_delay == 0xBBB || 0xE10*/;
 
                 if (left < 0 || (left && show_bits(&s->gb, FFMIN(left, 23)) && !is_d10)
-                    || (avctx->error_recognition >= FF_ER_AGGRESSIVE && left > 8)) {
+                    || ((avctx->err_recognition & AV_EF_BUFFER) && left > 8)) {
                     av_log(avctx, AV_LOG_ERROR, "end mismatch left=%d %0X\n", left, show_bits(&s->gb, FFMIN(left, 23)));
                     return -1;
                 } else
@@ -1911,7 +1911,7 @@ static int slice_decode_thread(AVCodecContext *c, void *arg)
 //av_log(c, AV_LOG_DEBUG, "ret:%d resync:%d/%d mb:%d/%d ts:%d/%d ec:%d\n",
 //ret, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, s->start_mb_y, s->end_mb_y, s->error_count);
         if (ret < 0) {
-            if (c->error_recognition >= FF_ER_EXPLODE)
+            if (c->err_recognition & AV_EF_EXPLODE)
                 return ret;
             if (s->resync_mb_x >= 0 && s->resync_mb_y >= 0)
                 ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, AC_ERROR | DC_ERROR | MV_ERROR);
@@ -1999,7 +1999,7 @@ static int mpeg1_decode_sequence(AVCodecContext *avctx,
     s->aspect_ratio_info = get_bits(&s->gb, 4);
     if (s->aspect_ratio_info == 0) {
         av_log(avctx, AV_LOG_ERROR, "aspect ratio has forbidden 0 value\n");
-        if (avctx->error_recognition >= FF_ER_COMPLIANT)
+        if (avctx->err_recognition & AV_EF_BITSTREAM)
             return -1;
     }
     s->frame_rate_index = get_bits(&s->gb, 4);
@@ -2287,7 +2287,7 @@ static int mpeg_decode_frame(AVCodecContext *avctx,
 
     if (avctx->extradata && !avctx->frame_number) {
         int ret = decode_chunks(avctx, picture, data_size, avctx->extradata, avctx->extradata_size);
-        if (ret < 0 && avctx->error_recognition >= FF_ER_EXPLODE)
+        if (ret < 0 && (avctx->err_recognition & AV_EF_EXPLODE))
             return ret;
     }
 
@@ -2347,7 +2347,7 @@ static int decode_chunks(AVCodecContext *avctx,
                     s->sync=1;
             } else {
                 av_log(avctx, AV_LOG_ERROR, "ignoring SEQ_START_CODE after %X\n", last_code);
-                if (avctx->error_recognition >= FF_ER_EXPLODE)
+                if (avctx->err_recognition & AV_EF_EXPLODE)
                     return AVERROR_INVALIDDATA;
             }
             break;
@@ -2381,7 +2381,7 @@ static int decode_chunks(AVCodecContext *avctx,
                 last_code = PICTURE_START_CODE;
             } else {
                 av_log(avctx, AV_LOG_ERROR, "ignoring pic after %X\n", last_code);
-                if (avctx->error_recognition >= FF_ER_EXPLODE)
+                if (avctx->err_recognition & AV_EF_EXPLODE)
                     return AVERROR_INVALIDDATA;
             }
             break;
@@ -2394,7 +2394,7 @@ static int decode_chunks(AVCodecContext *avctx,
                 mpeg_decode_sequence_extension(s);
                 } else {
                     av_log(avctx, AV_LOG_ERROR, "ignoring seq ext after %X\n", last_code);
-                    if (avctx->error_recognition >= FF_ER_EXPLODE)
+                    if (avctx->err_recognition & AV_EF_EXPLODE)
                         return AVERROR_INVALIDDATA;
                 }
                 break;
@@ -2412,7 +2412,7 @@ static int decode_chunks(AVCodecContext *avctx,
                     mpeg_decode_picture_coding_extension(s);
                 } else {
                     av_log(avctx, AV_LOG_ERROR, "ignoring pic cod ext after %X\n", last_code);
-                    if (avctx->error_recognition >= FF_ER_EXPLODE)
+                    if (avctx->err_recognition & AV_EF_EXPLODE)
                         return AVERROR_INVALIDDATA;
                 }
                 break;
@@ -2428,7 +2428,7 @@ static int decode_chunks(AVCodecContext *avctx,
                 s->sync=1;
             } else {
                 av_log(avctx, AV_LOG_ERROR, "ignoring GOP_START_CODE after %X\n", last_code);
-                if (avctx->error_recognition >= FF_ER_EXPLODE)
+                if (avctx->err_recognition & AV_EF_EXPLODE)
                     return AVERROR_INVALIDDATA;
             }
             break;
@@ -2475,7 +2475,7 @@ static int decode_chunks(AVCodecContext *avctx,
 
                 if (!s2->pict_type) {
                     av_log(avctx, AV_LOG_ERROR, "Missing picture start code\n");
-                    if (avctx->error_recognition >= FF_ER_EXPLODE)
+                    if (avctx->err_recognition & AV_EF_EXPLODE)
                         return AVERROR_INVALIDDATA;
                     break;
                 }
@@ -2516,7 +2516,7 @@ static int decode_chunks(AVCodecContext *avctx,
                     emms_c();
 
                     if (ret < 0) {
-                        if (avctx->error_recognition >= FF_ER_EXPLODE)
+                        if (avctx->err_recognition & AV_EF_EXPLODE)
                             return ret;
                         if (s2->resync_mb_x >= 0 && s2->resync_mb_y >= 0)
                             ff_er_add_slice(s2, s2->resync_mb_x, s2->resync_mb_y, s2->mb_x, s2->mb_y, AC_ERROR | DC_ERROR | MV_ERROR);

libavcodec/mpegaudiodec.c

@@ -79,7 +79,7 @@ typedef struct MPADecodeContext {
 #endif
     int adu_mode; ///< 0 for standard mp3, 1 for adu formatted mp3
     int dither_state;
-    int error_recognition;
+    int err_recognition;
     AVCodecContext* avctx;
     MPADSPContext mpadsp;
 } MPADecodeContext;
@@ -280,7 +280,7 @@ static av_cold int decode_init(AVCodecContext * avctx)
     ff_mpadsp_init(&s->mpadsp);
 
     avctx->sample_fmt= OUT_FMT;
-    s->error_recognition= avctx->error_recognition;
+    s->err_recognition = avctx->err_recognition;
 
     if (!init && !avctx->parse_only) {
         int offset;
@@ -1104,7 +1104,7 @@ static int huffman_decode(MPADecodeContext *s, GranuleDef *g,
                 s_index -= 4;
                 skip_bits_long(&s->gb, last_pos - pos);
                 av_log(s->avctx, AV_LOG_INFO, "overread, skip %d enddists: %d %d\n", last_pos - pos, end_pos-pos, end_pos2-pos);
-                if(s->error_recognition >= FF_ER_COMPLIANT)
+                if(s->err_recognition & AV_EF_BITSTREAM)
                     s_index=0;
                 break;
             }
@@ -1134,10 +1134,10 @@ static int huffman_decode(MPADecodeContext *s, GranuleDef *g,
     /* skip extension bits */
     bits_left = end_pos2 - get_bits_count(&s->gb);
 //av_log(NULL, AV_LOG_ERROR, "left:%d buf:%p\n", bits_left, s->in_gb.buffer);
-    if (bits_left < 0 && s->error_recognition >= FF_ER_COMPLIANT) {
+    if (bits_left < 0 && (s->err_recognition & AV_EF_BITSTREAM)) {
         av_log(s->avctx, AV_LOG_ERROR, "bits_left=%d\n", bits_left);
         s_index=0;
-    }else if(bits_left > 0 && s->error_recognition >= FF_ER_AGGRESSIVE){
+    }else if(bits_left > 0 && (s->err_recognition & AV_EF_BUFFER)){
         av_log(s->avctx, AV_LOG_ERROR, "bits_left=%d\n", bits_left);
         s_index=0;
     }

libavcodec/mxpegdec.c

@@ -275,11 +275,11 @@ static int mxpeg_decode_frame(AVCodecContext *avctx,
                     }
 
                     ret = ff_mjpeg_decode_sos(jpg, s->mxm_bitmask, reference_ptr);
-                    if (ret < 0 && avctx->error_recognition >= FF_ER_EXPLODE)
+                    if (ret < 0 && (avctx->err_recognition & AV_EF_EXPLODE))
                         return ret;
                 } else {
                     ret = ff_mjpeg_decode_sos(jpg, NULL, NULL);
-                    if (ret < 0 && avctx->error_recognition >= FF_ER_EXPLODE)
+                    if (ret < 0 && (avctx->err_recognition & AV_EF_EXPLODE))
                         return ret;
                 }
 

libavcodec/options.c

@@ -201,14 +201,19 @@ static const AVOption options[]={
 {"unofficial", "allow unofficial extensions", 0, AV_OPT_TYPE_CONST, {.dbl = FF_COMPLIANCE_UNOFFICIAL }, INT_MIN, INT_MAX, V|D|E, "strict"},
 {"experimental", "allow non standardized experimental things", 0, AV_OPT_TYPE_CONST, {.dbl = FF_COMPLIANCE_EXPERIMENTAL }, INT_MIN, INT_MAX, V|D|E, "strict"},
 {"b_qoffset", "qp offset between P and B frames", OFFSET(b_quant_offset), AV_OPT_TYPE_FLOAT, {.dbl = 1.25 }, -FLT_MAX, FLT_MAX, V|E},
+#if FF_API_ER
 {"er", "set error detection aggressivity", OFFSET(error_recognition), AV_OPT_TYPE_INT, {.dbl = FF_ER_CAREFUL }, INT_MIN, INT_MAX, A|V|D, "er"},
 {"careful", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = FF_ER_CAREFUL }, INT_MIN, INT_MAX, V|D, "er"},
 {"compliant", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = FF_ER_COMPLIANT }, INT_MIN, INT_MAX, V|D, "er"},
 {"aggressive", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = FF_ER_AGGRESSIVE }, INT_MIN, INT_MAX, V|D, "er"},
-#if FF_API_ER
 {"very_aggressive", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = FF_ER_VERY_AGGRESSIVE }, INT_MIN, INT_MAX, V|D, "er"},
-#endif /* FF_API_ER */
 {"explode", "abort decoding on error recognition", 0, AV_OPT_TYPE_CONST, {.dbl = FF_ER_EXPLODE }, INT_MIN, INT_MAX, V|D, "er"},
+#endif /* FF_API_ER */
+{"err_filter", "set error detection filter flags", OFFSET(err_recognition), AV_OPT_TYPE_FLAGS, {.dbl = AV_EF_CRCCHECK }, INT_MIN, INT_MAX, A|V|D, "err_filter"},
+{"crccheck", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = AV_EF_CRCCHECK }, INT_MIN, INT_MAX, V|D, "err_filter"},
+{"bitstream", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = AV_EF_BITSTREAM }, INT_MIN, INT_MAX, V|D, "err_filter"},
+{"buffer", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = AV_EF_BUFFER }, INT_MIN, INT_MAX, V|D, "err_filter"},
+{"explode", "abort decoding on minor error recognition", 0, AV_OPT_TYPE_CONST, {.dbl = AV_EF_EXPLODE }, INT_MIN, INT_MAX, V|D, "err_filter"},
 {"has_b_frames", NULL, OFFSET(has_b_frames), AV_OPT_TYPE_INT, {.dbl = DEFAULT }, INT_MIN, INT_MAX},
 {"block_align", NULL, OFFSET(block_align), AV_OPT_TYPE_INT, {.dbl = DEFAULT }, INT_MIN, INT_MAX},
 {"parse_only", NULL, OFFSET(parse_only), AV_OPT_TYPE_INT, {.dbl = DEFAULT }, INT_MIN, INT_MAX},

libavcodec/twinvq.c

@@ -822,7 +822,7 @@ static int twin_decode_frame(AVCodecContext * avctx, void *data,
     const ModeTab *mtab = tctx->mtab;
     float *out = data;
     enum FrameType ftype;
-    int window_type;
+    int window_type, out_size;
     static const enum FrameType wtype_to_ftype_table[] = {
         FT_LONG,   FT_LONG, FT_SHORT, FT_LONG,
         FT_MEDIUM, FT_LONG, FT_LONG,  FT_MEDIUM, FT_MEDIUM
@@ -831,8 +831,14 @@ static int twin_decode_frame(AVCodecContext * avctx, void *data,
     if (buf_size*8 < avctx->bit_rate*mtab->size/avctx->sample_rate + 8) {
         av_log(avctx, AV_LOG_ERROR,
                "Frame too small (%d bytes). Truncated file?\n", buf_size);
-        *data_size = 0;
-        return buf_size;
+        return AVERROR(EINVAL);
+    }
+
+    out_size = mtab->size * avctx->channels *
+               av_get_bytes_per_sample(avctx->sample_fmt);
+    if (*data_size < out_size) {
+        av_log(avctx, AV_LOG_ERROR, "output buffer is too small\n");
+        return AVERROR(EINVAL);
     }
 
     init_get_bits(&gb, buf, buf_size * 8);
@@ -857,7 +863,7 @@ static int twin_decode_frame(AVCodecContext * avctx, void *data,
         return buf_size;
     }
 
-    *data_size = mtab->size*avctx->channels*4;
+    *data_size = out_size;
 
     return buf_size;
 }

libavcodec/utils.c

@@ -610,6 +610,16 @@ int attribute_align_arg avcodec_open2(AVCodecContext *avctx, AVCodec *codec, AVD
         goto free_and_end;
     }
     avctx->frame_number = 0;
+#if FF_API_ER
+
+    av_log(avctx, AV_LOG_DEBUG, "err{or,}_recognition separate: %d; %d\n",
+           avctx->error_recognition, avctx->err_recognition);
+    /* FF_ER_CAREFUL (==1) implies AV_EF_CRCCHECK (== 1<<1 - 1),
+       FF_ER_COMPLIANT (==2) implies AV_EF_{CRCCHECK,BITSTREAM} (== 1<<2 - 1), et cetera} */
+    avctx->err_recognition |= (1<<(avctx->error_recognition-(avctx->error_recognition>=FF_ER_VERY_AGGRESSIVE))) - 1;
+    av_log(avctx, AV_LOG_DEBUG, "err{or,}_recognition combined: %d; %d\n",
+           avctx->error_recognition, avctx->err_recognition);
+#endif
 
     if (!HAVE_THREADS)
         av_log(avctx, AV_LOG_WARNING, "Warning: not compiled with thread support, using thread emulation\n");

libavcodec/vc1dec.c

@@ -930,6 +930,8 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir)
     if (!v->field_mode || (v->field_mode && !v->numref)) {
         valid_count = get_chroma_mv(mvx, mvy, intra, 0, &tx, &ty);
         if (!valid_count) {
+            s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;
+            s->current_picture.f.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 intra blocks
         }
@@ -941,6 +943,8 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir)
         if (dominant)
             chroma_ref_type = !v->cur_field_type;
     }
+    s->current_picture.f.motion_val[1][s->block_index[0]][0] = tx;
+    s->current_picture.f.motion_val[1][s->block_index[0]][1] = ty;
     uvmx = (tx + ((tx & 3) == 3)) >> 1;
     uvmy = (ty + ((ty & 3) == 3)) >> 1;
 

libavcodec/x86/h264_weight_10bit.asm

@@ -44,6 +44,7 @@ SECTION .text
     PROLOGUE 0,6,8
     movifnidn  r0, r0mp
     movifnidn r1d, r1m
+    movifnidn r2d, r2m
     movifnidn r4d, r4m
     movifnidn r5d, r5m
 %endmacro

libavformat/id3v2.h

@@ -62,7 +62,7 @@ typedef struct ID3v2ExtraMetaGEOB {
 /**
  * Detect ID3v2 Header.
  * @param buf   must be ID3v2_HEADER_SIZE byte long
- * @param magic magic bytes to identify the header, machine byte order.
+ * @param magic magic bytes to identify the header.
  * If in doubt, use ID3v2_DEFAULT_MAGIC.
  */
 int ff_id3v2_match(const uint8_t *buf, const char *magic);

libavformat/libavformat.v

@@ -23,5 +23,10 @@ LIBAVFORMAT_$MAJOR {
                 ff_timefilter_new;
                 ff_timefilter_update;
                 ff_timefilter_reset;
+                get_*;
+                put_*;
+                udp_set_remote_url;
+                udp_get_local_port;
+                init_checksum;
         local: *;
 };

libswscale/ppc/swscale_altivec.c

@@ -94,34 +94,29 @@ altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW)
 
 //FIXME remove the usage of scratch buffers.
 static void
-yuv2yuvX_altivec_real(SwsContext *c,
-                      const int16_t *lumFilter, const int16_t **lumSrc,
-                      int lumFilterSize, const int16_t *chrFilter,
-                      const int16_t **chrUSrc, const int16_t **chrVSrc,
-                      int chrFilterSize, const int16_t **alpSrc,
-                      uint8_t *dest[4], int dstW, int chrDstW)
+yuv2planeX_altivec(const int16_t *filter, int filterSize,
+                   const int16_t **src, uint8_t *dest, int dstW,
+                   const uint8_t *dither, int offset)
 {
-    uint8_t *yDest = dest[0], *uDest = dest[1], *vDest = dest[2];
-    const uint8_t *lumDither = c->lumDither8, *chrDither = c->chrDither8;
     register int i, j;
     {
         DECLARE_ALIGNED(16, int, val)[dstW];
 
         for (i=0; i<dstW; i++)
-            val[i] = lumDither[i & 7] << 12;
+            val[i] = dither[(i + offset) & 7] << 12;
 
-        for (j = 0; j < lumFilterSize; j++) {
-            vector signed short l1, vLumFilter = vec_ld(j << 1, lumFilter);
-            vector unsigned char perm, perm0 = vec_lvsl(j << 1, lumFilter);
+        for (j = 0; j < filterSize; j++) {
+            vector signed short l1, vLumFilter = vec_ld(j << 1, filter);
+            vector unsigned char perm, perm0 = vec_lvsl(j << 1, filter);
             vLumFilter = vec_perm(vLumFilter, vLumFilter, perm0);
             vLumFilter = vec_splat(vLumFilter, 0); // lumFilter[j] is loaded 8 times in vLumFilter
 
-            perm = vec_lvsl(0, lumSrc[j]);
-            l1 = vec_ld(0, lumSrc[j]);
+            perm = vec_lvsl(0, src[j]);
+            l1 = vec_ld(0, src[j]);
 
             for (i = 0; i < (dstW - 7); i+=8) {
                 int offset = i << 2;
-                vector signed short l2 = vec_ld((i << 1) + 16, lumSrc[j]);
+                vector signed short l2 = vec_ld((i << 1) + 16, src[j]);
 
                 vector signed int v1 = vec_ld(offset, val);
                 vector signed int v2 = vec_ld(offset + 16, val);
@@ -143,73 +138,10 @@ yuv2yuvX_altivec_real(SwsContext *c,
                 l1 = l2;
             }
             for ( ; i < dstW; i++) {
-                val[i] += lumSrc[j][i] * lumFilter[j];
+                val[i] += src[j][i] * filter[j];
             }
         }
-        altivec_packIntArrayToCharArray(val, yDest, dstW);
-    }
-    if (uDest != 0) {
-        DECLARE_ALIGNED(16, int, u)[chrDstW];
-        DECLARE_ALIGNED(16, int, v)[chrDstW];
-
-        for (i=0; i<chrDstW; i++) {
-            u[i] = chrDither[i & 7] << 12;
-            v[i] = chrDither[(i + 3) & 7] << 12;
-        }
-
-        for (j = 0; j < chrFilterSize; j++) {
-            vector signed short l1, l1_V, vChrFilter = vec_ld(j << 1, chrFilter);
-            vector unsigned char perm, perm0 = vec_lvsl(j << 1, chrFilter);
-            vChrFilter = vec_perm(vChrFilter, vChrFilter, perm0);
-            vChrFilter = vec_splat(vChrFilter, 0); // chrFilter[j] is loaded 8 times in vChrFilter
-
-            perm = vec_lvsl(0, chrUSrc[j]);
-            l1 = vec_ld(0, chrUSrc[j]);
-            l1_V = vec_ld(0, chrVSrc[j]);
-
-            for (i = 0; i < (chrDstW - 7); i+=8) {
-                int offset = i << 2;
-                vector signed short l2 = vec_ld((i << 1) + 16, chrUSrc[j]);
-                vector signed short l2_V = vec_ld((i << 1) + 16, chrVSrc[j]);
-
-                vector signed int v1 = vec_ld(offset, u);
-                vector signed int v2 = vec_ld(offset + 16, u);
-                vector signed int v1_V = vec_ld(offset, v);
-                vector signed int v2_V = vec_ld(offset + 16, v);
-
-                vector signed short ls = vec_perm(l1, l2, perm); // chrUSrc[j][i] ... chrUSrc[j][i+7]
-                vector signed short ls_V = vec_perm(l1_V, l2_V, perm); // chrVSrc[j][i] ... chrVSrc[j][i]
-
-                vector signed int i1 = vec_mule(vChrFilter, ls);
-                vector signed int i2 = vec_mulo(vChrFilter, ls);
-                vector signed int i1_V = vec_mule(vChrFilter, ls_V);
-                vector signed int i2_V = vec_mulo(vChrFilter, ls_V);
-
-                vector signed int vf1 = vec_mergeh(i1, i2);
-                vector signed int vf2 = vec_mergel(i1, i2); // chrUSrc[j][i] * chrFilter[j] ... chrUSrc[j][i+7] * chrFilter[j]
-                vector signed int vf1_V = vec_mergeh(i1_V, i2_V);
-                vector signed int vf2_V = vec_mergel(i1_V, i2_V); // chrVSrc[j][i] * chrFilter[j] ... chrVSrc[j][i+7] * chrFilter[j]
-
-                vector signed int vo1 = vec_add(v1, vf1);
-                vector signed int vo2 = vec_add(v2, vf2);
-                vector signed int vo1_V = vec_add(v1_V, vf1_V);
-                vector signed int vo2_V = vec_add(v2_V, vf2_V);
-
-                vec_st(vo1, offset, u);
-                vec_st(vo2, offset + 16, u);
-                vec_st(vo1_V, offset, v);
-                vec_st(vo2_V, offset + 16, v);
-
-                l1 = l2;
-                l1_V = l2_V;
-            }
-            for ( ; i < chrDstW; i++) {
-                u[i] += chrUSrc[j][i] * chrFilter[j];
-                v[i] += chrVSrc[j][i] * chrFilter[j];
-            }
-        }
-        altivec_packIntArrayToCharArray(u, uDest, chrDstW);
-        altivec_packIntArrayToCharArray(v, vDest, chrDstW);
+        altivec_packIntArrayToCharArray(val, dest, dstW);
     }
 }
 
@@ -405,7 +337,7 @@ void ff_sws_init_swScale_altivec(SwsContext *c)
     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) &&
         dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21 &&
         !c->alpPixBuf) {
-        c->yuv2yuvX     = yuv2yuvX_altivec_real;
+        c->yuv2planeX = yuv2planeX_altivec;
     }
 
     /* The following list of supported dstFormat values should

libswscale/swscale_internal.h

@@ -61,56 +61,58 @@ typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t* src[],
                        int srcStride[], int srcSliceY, int srcSliceH,
                        uint8_t* dst[], int dstStride[]);
 
+
 /**
- * Write one line of horizontally scaled Y/U/V/A to planar output
+ * Write one line of horizontally scaled data to planar output
  * without any additional vertical scaling (or point-scaling).
  *
- * @param c       SWS scaling context
- * @param lumSrc  scaled luma (Y) source data, 15bit for 8-10bit output,
- *                19-bit for 16bit output (in int32_t)
- * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
- *                19-bit for 16bit output (in int32_t)
- * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
+ * @param src     scaled source data, 15bit for 8-10bit output,
  *                19-bit for 16bit output (in int32_t)
- * @param alpSrc  scaled alpha (A) source data, 15bit for 8-10bit output,
+ * @param dest    pointer to the output plane. For >8bit
+ *                output, this is in uint16_t
+ * @param dstW    width of destination in pixels
+ * @param dither  ordered dither array of type int16_t and size 8
+ * @param offset  Dither offset
+ */
+typedef void (*yuv2planar1_fn) (const int16_t *src, uint8_t *dest, int dstW,
+                                const uint8_t *dither, int offset);
+
+/**
+ * Write one line of horizontally scaled data to planar output
+ * with multi-point vertical scaling between input pixels.
+ *
+ * @param filter        vertical luma/alpha scaling coefficients, 12bit [0,4096]
+ * @param src           scaled luma (Y) or alpha (A) source data, 15bit for 8-10bit output,
  *                      19-bit for 16bit output (in int32_t)
- * @param dest    pointer to the 4 output planes (Y/U/V/A). For >8bit
+ * @param filterSize    number of vertical input lines to scale
+ * @param dest          pointer to output plane. For >8bit
  *                      output, this is in uint16_t
- * @param dstW    width of dest[0], dest[3], lumSrc and alpSrc in pixels
- * @param chrDstW width of dest[1], dest[2], chrUSrc and chrVSrc
+ * @param dstW          width of destination pixels
+ * @param offset        Dither offset
  */
-typedef void (*yuv2planar1_fn) (struct SwsContext *c,
-                                const int16_t *lumSrc, const int16_t *chrUSrc,
-                                const int16_t *chrVSrc, const int16_t *alpSrc,
-                                uint8_t *dest[4], int dstW, int chrDstW);
+typedef void (*yuv2planarX_fn) (const int16_t *filter, int filterSize,
+                                const int16_t **src, uint8_t *dest, int dstW,
+                                const uint8_t *dither, int offset);
+
 /**
- * Write one line of horizontally scaled Y/U/V/A to planar output
+ * Write one line of horizontally scaled chroma to interleaved output
  * with multi-point vertical scaling between input pixels.
  *
  * @param c             SWS scaling context
- * @param lumFilter     vertical luma/alpha scaling coefficients, 12bit [0,4096]
- * @param lumSrc        scaled luma (Y) source data, 15bit for 8-10bit output,
- *                      19-bit for 16bit output (in int32_t)
- * @param lumFilterSize number of vertical luma/alpha input lines to scale
  * @param chrFilter     vertical chroma scaling coefficients, 12bit [0,4096]
  * @param chrUSrc       scaled chroma (U) source data, 15bit for 8-10bit output,
  *                      19-bit for 16bit output (in int32_t)
  * @param chrVSrc       scaled chroma (V) source data, 15bit for 8-10bit output,
  *                      19-bit for 16bit output (in int32_t)
  * @param chrFilterSize number of vertical chroma input lines to scale
- * @param alpSrc        scaled alpha (A) source data, 15bit for 8-10bit output,
- *                      19-bit for 16bit output (in int32_t)
- * @param dest          pointer to the 4 output planes (Y/U/V/A). For >8bit
+ * @param dest          pointer to the output plane. For >8bit
  *                      output, this is in uint16_t
- * @param dstW          width of dest[0], dest[3], lumSrc and alpSrc in pixels
- * @param chrDstW       width of dest[1], dest[2], chrUSrc and chrVSrc
+ * @param dstW          width of chroma planes
  */
-typedef void (*yuv2planarX_fn) (struct SwsContext *c, const int16_t *lumFilter,
-                                const int16_t **lumSrc, int lumFilterSize,
-                                const int16_t *chrFilter, const int16_t **chrUSrc,
-                                const int16_t **chrVSrc,  int chrFilterSize,
-                                const int16_t **alpSrc, uint8_t *dest[4],
-                                int dstW, int chrDstW);
+typedef void (*yuv2interleavedX_fn) (struct SwsContext *c, const int16_t *chrFilter, int chrFilterSize,
+                                     const int16_t **chrUSrc, const int16_t **chrVSrc,
+                                     uint8_t *dest, int dstW);
+
 /**
  * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
  * output without any additional vertical scaling (or point-scaling). Note
@@ -412,8 +414,9 @@ typedef struct SwsContext {
 #endif
 
     /* function pointers for swScale() */
-    yuv2planar1_fn yuv2yuv1;
-    yuv2planarX_fn yuv2yuvX;
+    yuv2planar1_fn yuv2plane1;
+    yuv2planarX_fn yuv2planeX;
+    yuv2interleavedX_fn yuv2nv12cX;
     yuv2packed1_fn yuv2packed1;
     yuv2packed2_fn yuv2packed2;
     yuv2packedX_fn yuv2packedX;

libswscale/x86/scale.asm

 ;******************************************************************************
-;* x86-optimized horizontal line scaling functions
+;* x86-optimized horizontal/vertical line scaling functions
 ;* Copyright (c) 2011 Ronald S. Bultje <rsbultje@gmail.com>
+;*                    Kieran Kunhya <kieran@kunhya.com>
 ;*
 ;* This file is part of Libav.
 ;*
@@ -28,6 +29,11 @@ max_19bit_int: times 4 dd 0x7ffff
 max_19bit_flt: times 4 dd 524287.0
 minshort:      times 8 dw 0x8000
 unicoeff:      times 4 dd 0x20000000
+yuv2yuvX_16_start:  times 4 dd 0x4000 - 0x40000000
+yuv2yuvX_10_start:  times 4 dd 0x10000
+yuv2yuvX_9_start:   times 4 dd 0x20000
+yuv2yuvX_10_upper:  times 8 dw 0x3ff
+yuv2yuvX_9_upper:   times 8 dw 0x1ff
 
 SECTION .text
 
@@ -429,3 +435,233 @@ INIT_XMM
 SCALE_FUNCS2 sse2,  6, 7, 8
 SCALE_FUNCS2 ssse3, 6, 6, 8
 SCALE_FUNCS2 sse4,  6, 6, 8
+
+;-----------------------------------------------------------------------------
+; vertical line scaling
+;
+; void yuv2plane1_<output_size>_<opt>(const int16_t *src, uint8_t *dst, int dstW,
+;                                     const uint8_t *dither, int offset)
+; and
+; void yuv2planeX_<output_size>_<opt>(const int16_t *filter, int filterSize,
+;                                     const int16_t **src, uint8_t *dst, int dstW,
+;                                     const uint8_t *dither, int offset)
+;
+; Scale one or $filterSize lines of source data to generate one line of output
+; data. The input is 15-bit in int16_t if $output_size is [8,10] and 19-bit in
+; int32_t if $output_size is 16. $filter is 12-bits. $filterSize is a multiple
+; of 2. $offset is either 0 or 3. $dither holds 8 values.
+;-----------------------------------------------------------------------------
+
+%macro yuv2planeX_fn 4
+
+%ifdef ARCH_X86_32
+%define cntr_reg r1
+%define movsx mov
+%else
+%define cntr_reg r11
+%define movsx movsxd
+%endif
+
+cglobal yuv2planeX_%2_%1, %4, 7, %3
+%if %2 == 8 || %2 == 9 || %2 == 10
+    pxor            m6,  m6
+%endif ; %2 == 8/9/10
+
+%if %2 == 8
+%ifdef ARCH_X86_32
+%assign pad 0x2c - (stack_offset & 15)
+    SUB             rsp, pad
+%define m_dith m7
+%else ; x86-64
+%define m_dith m9
+%endif ; x86-32
+
+    ; create registers holding dither
+    movq        m_dith, [r5]             ; dither
+    test            r6d, r6d
+    jz              .no_rot
+%if mmsize == 16
+    punpcklqdq  m_dith,  m_dith
+%endif ; mmsize == 16
+    PALIGNR     m_dith,  m_dith,  3,  m0
+.no_rot:
+%if mmsize == 16
+    punpcklbw   m_dith,  m6
+%ifdef ARCH_X86_64
+    punpcklwd       m8,  m_dith,  m6
+    pslld           m8,  12
+%else ; x86-32
+    punpcklwd       m5,  m_dith,  m6
+    pslld           m5,  12
+%endif ; x86-32/64
+    punpckhwd   m_dith,  m6
+    pslld       m_dith,  12
+%ifdef ARCH_X86_32
+    mova      [rsp+ 0],  m5
+    mova      [rsp+16],  m_dith
+%endif
+%else ; mmsize == 8
+    punpcklbw       m5,  m_dith,  m6
+    punpckhbw   m_dith,  m6
+    punpcklwd       m4,  m5,  m6
+    punpckhwd       m5,  m6
+    punpcklwd       m3,  m_dith,  m6
+    punpckhwd   m_dith,  m6
+    pslld           m4,  12
+    pslld           m5,  12
+    pslld           m3,  12
+    pslld       m_dith,  12
+    mova      [rsp+ 0],  m4
+    mova      [rsp+ 8],  m5
+    mova      [rsp+16],  m3
+    mova      [rsp+24],  m_dith
+%endif ; mmsize == 8/16
+%endif ; %2 == 8
+
+    xor             r5,  r5
+
+.pixelloop
+%assign %%i 0
+    ; the rep here is for the 8bit output mmx case, where dither covers
+    ; 8 pixels but we can only handle 2 pixels per register, and thus 4
+    ; pixels per iteration. In order to not have to keep track of where
+    ; we are w.r.t. dithering, we unroll the mmx/8bit loop x2.
+%if %2 == 8
+%rep 16/mmsize
+%endif ; %2 == 8
+
+%if %2 == 8
+%ifdef ARCH_X86_32
+    mova            m2, [rsp+mmsize*(0+%%i)]
+    mova            m1, [rsp+mmsize*(1+%%i)]
+%else ; x86-64
+    mova            m2,  m8
+    mova            m1,  m_dith
+%endif ; x86-32/64
+%else ; %2 == 9/10/16
+    mova            m1, [yuv2yuvX_%2_start]
+    mova            m2,  m1
+%endif ; %2 == 8/9/10/16
+    movsx     cntr_reg,  r1m
+.filterloop_ %+ %%i
+    ; input pixels
+    mov             r6, [r2+gprsize*cntr_reg-2*gprsize]
+%if %2 == 16
+    mova            m3, [r6+r5*4]
+    mova            m5, [r6+r5*4+mmsize]
+%else ; %2 == 8/9/10
+    mova            m3, [r6+r5*2]
+%endif ; %2 == 8/9/10/16
+    mov             r6, [r2+gprsize*cntr_reg-gprsize]
+%if %2 == 16
+    mova            m4, [r6+r5*4]
+    mova            m6, [r6+r5*4+mmsize]
+%else ; %2 == 8/9/10
+    mova            m4, [r6+r5*2]
+%endif ; %2 == 8/9/10/16
+
+    ; coefficients
+    movd            m0, [r0+2*cntr_reg-4]; coeff[0], coeff[1]
+%if %2 == 16
+    pshuflw         m7,  m0,  0          ; coeff[0]
+    pshuflw         m0,  m0,  0x55       ; coeff[1]
+    pmovsxwd        m7,  m7              ; word -> dword
+    pmovsxwd        m0,  m0              ; word -> dword
+
+    pmulld          m3,  m7
+    pmulld          m5,  m7
+    pmulld          m4,  m0
+    pmulld          m6,  m0
+
+    paddd           m2,  m3
+    paddd           m1,  m5
+    paddd           m2,  m4
+    paddd           m1,  m6
+%else ; %2 == 10/9/8
+    punpcklwd       m5,  m3,  m4
+    punpckhwd       m3,  m4
+    SPLATD          m0,  m0
+
+    pmaddwd         m5,  m0
+    pmaddwd         m3,  m0
+
+    paddd           m2,  m5
+    paddd           m1,  m3
+%endif ; %2 == 8/9/10/16
+
+    sub       cntr_reg,  2
+    jg .filterloop_ %+ %%i
+
+%if %2 == 16
+    psrad           m2,  31 - %2
+    psrad           m1,  31 - %2
+%else ; %2 == 10/9/8
+    psrad           m2,  27 - %2
+    psrad           m1,  27 - %2
+%endif ; %2 == 8/9/10/16
+
+%if %2 == 8
+    packssdw        m2,  m1
+    packuswb        m2,  m2
+    movh     [r3+r5*1],  m2
+%else ; %2 == 9/10/16
+%if %2 == 16
+    packssdw        m2,  m1
+    paddw           m2, [minshort]
+%else ; %2 == 9/10
+%ifidn %1, sse4
+    packusdw        m2,  m1
+%elifidn %1, avx
+    packusdw        m2,  m1
+%else ; mmx2/sse2
+    packssdw        m2,  m1
+    pmaxsw          m2,  m6
+%endif ; mmx2/sse2/sse4/avx
+    pminsw          m2, [yuv2yuvX_%2_upper]
+%endif ; %2 == 9/10/16
+    mova     [r3+r5*2],  m2
+%endif ; %2 == 8/9/10/16
+
+    add             r5,  mmsize/2
+    sub             r4d, mmsize/2
+%if %2 == 8
+%assign %%i %%i+2
+%endrep
+%endif ; %2 == 8
+    jg .pixelloop
+
+%if %2 == 8
+%ifdef ARCH_X86_32
+    ADD             rsp, pad
+    RET
+%else ; x86-64
+    REP_RET
+%endif ; x86-32/64
+%else ; %2 == 9/10/16
+    REP_RET
+%endif ; %2 == 8/9/10/16
+%endmacro
+
+%define PALIGNR PALIGNR_MMX
+%ifdef ARCH_X86_32
+INIT_MMX
+yuv2planeX_fn mmx,   8,  0, 7
+yuv2planeX_fn mmx2,  9,  0, 5
+yuv2planeX_fn mmx2, 10,  0, 5
+%endif
+
+INIT_XMM
+yuv2planeX_fn sse2,  8, 10, 7
+yuv2planeX_fn sse2,  9,  7, 5
+yuv2planeX_fn sse2, 10,  7, 5
+
+%define PALIGNR PALIGNR_SSSE3
+yuv2planeX_fn sse4,  8, 10, 7
+yuv2planeX_fn sse4,  9,  7, 5
+yuv2planeX_fn sse4, 10,  7, 5
+yuv2planeX_fn sse4, 16,  8, 5
+
+INIT_AVX
+yuv2planeX_fn avx,   8, 10, 7
+yuv2planeX_fn avx,   9,  7, 5
+yuv2planeX_fn avx,  10,  7, 5

libswscale/x86/swscale_mmx.c

@@ -213,6 +213,23 @@ SCALE_FUNCS_SSE(sse2);
 SCALE_FUNCS_SSE(ssse3);
 SCALE_FUNCS_SSE(sse4);
 
+#define VSCALEX_FUNC(size, opt) \
+extern void ff_yuv2planeX_ ## size ## _ ## opt(const int16_t *filter, int filterSize, \
+                                               const int16_t **src, uint8_t *dest, int dstW, \
+                                               const uint8_t *dither, int offset)
+#define VSCALEX_FUNCS(opt1, opt2) \
+    VSCALEX_FUNC(8,  opt1); \
+    VSCALEX_FUNC(9,  opt2); \
+    VSCALEX_FUNC(10, opt2)
+
+#if ARCH_X86_32
+VSCALEX_FUNCS(mmx,  mmx2);
+#endif
+VSCALEX_FUNCS(sse2, sse2);
+VSCALEX_FUNCS(sse4, sse4);
+VSCALEX_FUNC(16, sse4);
+VSCALEX_FUNCS(avx,  avx);
+
 void ff_sws_init_swScale_mmx(SwsContext *c)
 {
     int cpu_flags = av_get_cpu_flags();
@@ -249,10 +266,18 @@ void ff_sws_init_swScale_mmx(SwsContext *c)
     case 8:  ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
     default: ASSIGN_SCALE_FUNC2(hscalefn, X, opt1, opt2); break; \
     }
+#define ASSIGN_VSCALEX_FUNC(vscalefn, opt1, opt2, opt2chk, do_16_case) \
+switch(c->dstBpc){ \
+    case 16:                                     do_16_case;                           break; \
+    case 10: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2planeX_10_ ## opt2; break; \
+    case 9:  if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2planeX_9_  ## opt2; break; \
+    default:                                     vscalefn = ff_yuv2planeX_8_  ## opt1; break; \
+    }
 #if ARCH_X86_32
     if (cpu_flags & AV_CPU_FLAG_MMX) {
         ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx);
         ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx);
+        ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMX2,);
     }
 #endif
 #define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
@@ -266,6 +291,7 @@ void ff_sws_init_swScale_mmx(SwsContext *c)
     if (cpu_flags & AV_CPU_FLAG_SSE2) {
         ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse2, sse2);
         ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse2, sse2);
+        ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse2, sse2, 1,);
     }
     if (cpu_flags & AV_CPU_FLAG_SSSE3) {
         ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, ssse3, ssse3);
@@ -275,6 +301,12 @@ void ff_sws_init_swScale_mmx(SwsContext *c)
         /* Xto15 don't need special sse4 functions */
         ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse4, ssse3);
         ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse4, ssse3);
+        ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse4, sse4, 1,
+                            if (!isBE(c->dstFormat)) c->yuv2planeX = ff_yuv2planeX_16_sse4);
+    }
+
+    if (cpu_flags & AV_CPU_FLAG_AVX) {
+        ASSIGN_VSCALEX_FUNC(c->yuv2planeX, avx, avx, 1,);
     }
 #endif
 }

tests/ref/lavfi/pixdesc

@@ -40,16 +40,16 @@ yuv420p9le          9ed4b1dfabc53fd9e586ff6c4c43af80
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
 yuv422p10be         bdc13b630fd668b34c6fe1aae28dfc71
 yuv422p10le         d0607c260a45c973e6639f4e449730ad
-yuv422p16be         4e9b3b3467aeebb6a528cee5966800ed
-yuv422p16le         f87c81bf16916b64d201359be0b4b6f4
+yuv422p16be         5499502e1c29534a158a1fe60e889f60
+yuv422p16le         e3d61fde6978591596bc36b914386623
 yuv422p9be          29b71579946940a8c00fa844c9dff507
 yuv422p9le          062b7f9cbb972bf36b5bdb1a7623701a
 yuv440p             5a064afe2b453bb52cdb3f176b1aa1cf
 yuv444p             0a98447b78fd476aa39686da6a74fa2e
 yuv444p10be         e65cbae7e4f1892c23defbc8e8052cf6
 yuv444p10le         767179dd82846cf00ee4c340c9c1ab74
-yuv444p16be         3ad639fff73e56f3b09dd20c335478d6
-yuv444p16le         8a7e66dc91ab7971fd24a9105ff2699b
+yuv444p16be         1c6ea2c2f5e539006112ceec3d4e7d90
+yuv444p16le         20f86bc2f68d2b3f1f2b48b97b2189f4
 yuv444p9be          6ab31f4c12b533ce318ecdff83cdd054
 yuv444p9le          f0606604a5c08becab6ba500124c4b7c
 yuva420p            a29884f3f3dfe1e00b961bc17bef3d47

tests/ref/lavfi/pixfmts_copy

@@ -40,16 +40,16 @@ yuv420p9le          9ed4b1dfabc53fd9e586ff6c4c43af80
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
 yuv422p10be         bdc13b630fd668b34c6fe1aae28dfc71
 yuv422p10le         d0607c260a45c973e6639f4e449730ad
-yuv422p16be         4e9b3b3467aeebb6a528cee5966800ed
-yuv422p16le         f87c81bf16916b64d201359be0b4b6f4
+yuv422p16be         5499502e1c29534a158a1fe60e889f60
+yuv422p16le         e3d61fde6978591596bc36b914386623
 yuv422p9be          29b71579946940a8c00fa844c9dff507
 yuv422p9le          062b7f9cbb972bf36b5bdb1a7623701a
 yuv440p             5a064afe2b453bb52cdb3f176b1aa1cf
 yuv444p             0a98447b78fd476aa39686da6a74fa2e
 yuv444p10be         e65cbae7e4f1892c23defbc8e8052cf6
 yuv444p10le         767179dd82846cf00ee4c340c9c1ab74
-yuv444p16be         3ad639fff73e56f3b09dd20c335478d6
-yuv444p16le         8a7e66dc91ab7971fd24a9105ff2699b
+yuv444p16be         1c6ea2c2f5e539006112ceec3d4e7d90
+yuv444p16le         20f86bc2f68d2b3f1f2b48b97b2189f4
 yuv444p9be          6ab31f4c12b533ce318ecdff83cdd054
 yuv444p9le          f0606604a5c08becab6ba500124c4b7c
 yuva420p            a29884f3f3dfe1e00b961bc17bef3d47

tests/ref/lavfi/pixfmts_crop

@@ -33,8 +33,8 @@ yuv422p16be         167e4338811a7d272925a4c6417d60da
 yuv422p16le         3359395d5875d581fa1e975013d30114
 yuv440p             2472417d980e395ad6843cbb8b633b29
 yuv444p             1f151980486848c96bc5585ced99003e
-yuv444p16be         5d0c0ea66ab43c0c590d8c2a9256e43f
-yuv444p16le         3c0a747c1b64feb0ab8dfba92f92579a
+yuv444p16be         1ce8fcd4712d525af983e6179d6a4f9e
+yuv444p16le         5f1441e18345aadb3f881dac99c6c08a
 yuva420p            7536753dfbc7932560fb50c921369a0e
 yuvj420p            21f891093006d42d7683b0e1d773a657
 yuvj422p            9a43d474c407590ad8f213880586b45e

tests/ref/lavfi/pixfmts_hflip

@@ -29,12 +29,12 @@ yuv420p             2d5c80f9ba2ddd85b2aeda3564cc7d64
 yuv420p16be         1c4fa93d0744de3cdc6d34ab55db3fb4
 yuv420p16le         92c74f5759068c381e4a066fe7faf2e0
 yuv422p             6e728f4eb9eae287c224f396d84be6ea
-yuv422p16be         69cf0605496c321546899a8442ee64fb
-yuv422p16le         f0b443fea72f4b6f462859a73b159664
+yuv422p16be         a05d43cd62b790087bd37083174557de
+yuv422p16le         6954abebcbc62d81068d58d0c62bdd5b
 yuv440p             a99e2b57ed601f39852715c9d675d0d3
 yuv444p             947e47f7bb5fdccc659d19b7df2b6fc3
-yuv444p16be         bc7d53923cff1d7e98d24540845fb64b
-yuv444p16le         5df206a93f85ef8b77f5bdc81d9b0a0b
+yuv444p16be         58c012e5ab73b066ef3c2b6411a395f1
+yuv444p16le         32c12794e184042a59738ab2de608c8d
 yuva420p            d83ec0c01498189f179ec574918185f1
 yuvj420p            df3aaaec3bb157c3bde5f0365af30f4f
 yuvj422p            d113871528d510a192797af59df9c05c

tests/ref/lavfi/pixfmts_null

@@ -40,16 +40,16 @@ yuv420p9le          9ed4b1dfabc53fd9e586ff6c4c43af80
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
 yuv422p10be         bdc13b630fd668b34c6fe1aae28dfc71
 yuv422p10le         d0607c260a45c973e6639f4e449730ad
-yuv422p16be         4e9b3b3467aeebb6a528cee5966800ed
-yuv422p16le         f87c81bf16916b64d201359be0b4b6f4
+yuv422p16be         5499502e1c29534a158a1fe60e889f60
+yuv422p16le         e3d61fde6978591596bc36b914386623
 yuv422p9be          29b71579946940a8c00fa844c9dff507
 yuv422p9le          062b7f9cbb972bf36b5bdb1a7623701a
 yuv440p             5a064afe2b453bb52cdb3f176b1aa1cf
 yuv444p             0a98447b78fd476aa39686da6a74fa2e
 yuv444p10be         e65cbae7e4f1892c23defbc8e8052cf6
 yuv444p10le         767179dd82846cf00ee4c340c9c1ab74
-yuv444p16be         3ad639fff73e56f3b09dd20c335478d6
-yuv444p16le         8a7e66dc91ab7971fd24a9105ff2699b
+yuv444p16be         1c6ea2c2f5e539006112ceec3d4e7d90
+yuv444p16le         20f86bc2f68d2b3f1f2b48b97b2189f4
 yuv444p9be          6ab31f4c12b533ce318ecdff83cdd054
 yuv444p9le          f0606604a5c08becab6ba500124c4b7c
 yuva420p            a29884f3f3dfe1e00b961bc17bef3d47

tests/ref/lavfi/pixfmts_scale

@@ -33,23 +33,23 @@ yuv411p             1143e7c5cc28fe0922b051b17733bc4c
 yuv420p             fdad2d8df8985e3d17e73c71f713cb14
 yuv420p10be         6d335e75b553da590135cf8bb999610c
 yuv420p10le         d510ddbabefd03ef39ec943fcb51b709
-yuv420p16be         2a75942af24fbdc1fdfe189c6e7bf589
-yuv420p16le         c4264d92a7c273967a778f4f5daddbe3
+yuv420p16be         31988e9a5d6acacaa710f67bc1172f3a
+yuv420p16le         f5390ce399f88e0e4e2621ed7833b250
 yuv420p9be          ec4983b7a949c0472110a7a2c58e278a
 yuv420p9le          c136dce5913a722eee44ab72cff664b2
 yuv422p             918e37701ee7377d16a8a6c119c56a40
 yuv422p10be         cea7ca6b0e66d6f29539885896c88603
 yuv422p10le         a10c4a5837547716f13cd61918b145f9
-yuv422p16be         285993ee0c0f4f8e511ee46f93c5f38c
-yuv422p16le         61bfcee8e54465f760164f5a75d40b5e
+yuv422p16be         e7e34fe9264784763ab6cb406524c0f3
+yuv422p16le         c435b76b08204dda6908640fb5fd4621
 yuv422p9be          82494823944912f73cebc58ad2979bbd
 yuv422p9le          fc69c8a21f473916a4b4225636b97e06
 yuv440p             461503fdb9b90451020aa3b25ddf041c
 yuv444p             81b2eba962d12e8d64f003ac56f6faf2
 yuv444p10be         e9d3c8e744b8b0d8187ca092fa203fc9
 yuv444p10le         02f0a336e9da062a64df1ba487e102c5
-yuv444p16be         2677f3074d255f9dab625e9e2e092ca5
-yuv444p16le         65fa92521ef97088599ea83f9508cd5b
+yuv444p16be         0da9bed80f5542682ab286f3261cf24c
+yuv444p16le         a0c5d3c7bf3f181db503cf8e450d1335
 yuv444p9be          9ac2643ce7f7e5c4e17c8c9fd8494d4a
 yuv444p9le          896a1cc9cccca1ba410dd53942d33cc4
 yuva420p            8673a9131fb47de69788863f93a50eb7

tests/ref/lavfi/pixfmts_vflip

@@ -40,16 +40,16 @@ yuv420p9le          0f1e371a1374d3cba2205b70cc7cac90
 yuv422p             d7f5cb44d9b0210d66d6a8762640ab34
 yuv422p10be         588fe319b96513c32e21d3e32b45447f
 yuv422p10le         11b57f2bd9661024153f3973b9090cdb
-yuv422p16be         c092d083548c2a144c372a98c46875c7
-yuv422p16le         c071b9397a416d51cbe339345cbcba84
+yuv422p16be         9bd8f8c961822b586fa4cf992be54acc
+yuv422p16le         9c4a1239605c7952b736ac3130163f14
 yuv422p9be          7c6f1e140b3999ee7d923854e507752a
 yuv422p9le          51f10d79c07989060dd06e767e6d7d60
 yuv440p             876385e96165acf51271b20e5d85a416
 yuv444p             9c3c667d1613b72d15bc6d851c5eb8f7
 yuv444p10be         944a4997c4edb3a8dd0f0493cfd5a1fd
 yuv444p10le         2d0947ae89ecc6a501eee6832cb27e06
-yuv444p16be         6a954614fd2a8ae0df53e4fd76937af8
-yuv444p16le         65613965fb58cc4c3cd480a68b6540ea
+yuv444p16be         de2dedfc6f12073ffead113f86e07ecf
+yuv444p16le         8e83323cf102d6c823a03ae8a7b7e033
 yuv444p9be          6ac92b7dc9ab2fc59bee99204886899a
 yuv444p9le          85aef13a654953d3455d89770b0d74bd
 yuva420p            c705d1cf061d8c6580ac690b55f92276