h264dec: h264: 4:2:2 intra decoding

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

23 files changed, 623 insertions, 135 deletions

diff --git a/libavcodec/arm/h264dsp_init_arm.c b/libavcodec/arm/h264dsp_init_arm.c
index e9146405c2..e2c226557e 100644
--- a/libavcodec/arm/h264dsp_init_arm.c
+++ b/libavcodec/arm/h264dsp_init_arm.c
@@ -92,7 +92,7 @@ void ff_h264_idct8_add4_neon(uint8_t *dst, const int *block_offset,
                              DCTELEM *block, int stride,
                              const uint8_t nnzc[6*8]);
 
-static void ff_h264dsp_init_neon(H264DSPContext *c, const int bit_depth)
+static void ff_h264dsp_init_neon(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
 {
     if (bit_depth == 8) {
     c->h264_v_loop_filter_luma   = ff_h264_v_loop_filter_luma_neon;
@@ -122,14 +122,15 @@ static void ff_h264dsp_init_neon(H264DSPContext *c, const int bit_depth)
     c->h264_idct_dc_add     = ff_h264_idct_dc_add_neon;
     c->h264_idct_add16      = ff_h264_idct_add16_neon;
     c->h264_idct_add16intra = ff_h264_idct_add16intra_neon;
-    c->h264_idct_add8       = ff_h264_idct_add8_neon;
+    if (chroma_format_idc == 1)
+        c->h264_idct_add8   = ff_h264_idct_add8_neon;
     c->h264_idct8_add       = ff_h264_idct8_add_neon;
     c->h264_idct8_dc_add    = ff_h264_idct8_dc_add_neon;
     c->h264_idct8_add4      = ff_h264_idct8_add4_neon;
     }
 }
 
-void ff_h264dsp_init_arm(H264DSPContext *c, const int bit_depth)
+void ff_h264dsp_init_arm(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
 {
-    if (HAVE_NEON) ff_h264dsp_init_neon(c, bit_depth);
+    if (HAVE_NEON) ff_h264dsp_init_neon(c, bit_depth, chroma_format_idc);
 }
diff --git a/libavcodec/arm/h264pred_init_arm.c b/libavcodec/arm/h264pred_init_arm.c
index b1d4f005e8..62f891e0fb 100644
--- a/libavcodec/arm/h264pred_init_arm.c
+++ b/libavcodec/arm/h264pred_init_arm.c
@@ -42,7 +42,7 @@ void ff_pred8x8_0lt_dc_neon(uint8_t *src, int stride);
 void ff_pred8x8_l00_dc_neon(uint8_t *src, int stride);
 void ff_pred8x8_0l0_dc_neon(uint8_t *src, int stride);
 
-static void ff_h264_pred_init_neon(H264PredContext *h, int codec_id, const int bit_depth)
+static void ff_h264_pred_init_neon(H264PredContext *h, int codec_id, const int bit_depth, const int chroma_format_idc)
 {
     const int high_depth = bit_depth > 8;
 
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index e8b36aef90..07ef196185 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -63,8 +63,10 @@ void ff_h264_idct_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int strid
 void ff_h264_idct_add16_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
 void ff_h264_idct_add16intra_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
 void ff_h264_idct8_add4_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
+void ff_h264_idct_add8_422_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
 void ff_h264_idct_add8_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
 void ff_h264_luma_dc_dequant_idct_ ## depth ## _c(DCTELEM *output, DCTELEM *input, int qmul);\
+void ff_h264_chroma422_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);\
 void ff_h264_chroma_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);
 
 H264_IDCT( 8)
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index bfc73cb86d..29426f74b6 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -942,7 +942,7 @@ static void clone_tables(H264Context *dst, H264Context *src, int i){
     dst->list_counts              = src->list_counts;
 
     dst->s.obmc_scratchpad = NULL;
-    ff_h264_pred_init(&dst->hpc, src->s.codec_id, src->sps.bit_depth_luma);
+    ff_h264_pred_init(&dst->hpc, src->s.codec_id, src->sps.bit_depth_luma, src->sps.chroma_format_idc);
 }
 
 /**
@@ -970,8 +970,8 @@ static av_cold void common_init(H264Context *h){
     s->height = s->avctx->height;
     s->codec_id= s->avctx->codec->id;
 
-    ff_h264dsp_init(&h->h264dsp, 8);
-    ff_h264_pred_init(&h->hpc, s->codec_id, 8);
+    ff_h264dsp_init(&h->h264dsp, 8, 1);
+    ff_h264_pred_init(&h->hpc, s->codec_id, 8, 1);
 
     h->dequant_coeff_pps= -1;
     s->unrestricted_mv=1;
@@ -1428,11 +1428,13 @@ static void decode_postinit(H264Context *h, int setup_finished){
         ff_thread_finish_setup(s->avctx);
 }
 
-static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int chroma444, int simple){
+static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int simple){
     MpegEncContext * const s = &h->s;
     uint8_t *top_border;
     int top_idx = 1;
     const int pixel_shift = h->pixel_shift;
+    int chroma444 = CHROMA444;
+    int chroma422 = CHROMA422;
 
     src_y  -=   linesize;
     src_cb -= uvlinesize;
@@ -1456,6 +1458,14 @@ static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, ui
                             AV_COPY128(top_border+16, src_cb + 15*uvlinesize);
                             AV_COPY128(top_border+32, src_cr + 15*uvlinesize);
                         }
+                    } else if(chroma422){
+                        if (pixel_shift) {
+                            AV_COPY128(top_border+32, src_cb + 15*uvlinesize);
+                            AV_COPY128(top_border+48, src_cr + 15*uvlinesize);
+                        } else {
+                            AV_COPY64(top_border+16, src_cb +  15*uvlinesize);
+                            AV_COPY64(top_border+24, src_cr +  15*uvlinesize);
+                        }
                     } else {
                         if (pixel_shift) {
                             AV_COPY128(top_border+32, src_cb+7*uvlinesize);
@@ -1491,6 +1501,14 @@ static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, ui
                 AV_COPY128(top_border+16, src_cb + 16*linesize);
                 AV_COPY128(top_border+32, src_cr + 16*linesize);
             }
+        } else if(chroma422) {
+            if (pixel_shift) {
+                AV_COPY128(top_border+32, src_cb+16*uvlinesize);
+                AV_COPY128(top_border+48, src_cr+16*uvlinesize);
+            } else {
+                AV_COPY64(top_border+16, src_cb+16*uvlinesize);
+                AV_COPY64(top_border+24, src_cr+16*uvlinesize);
+            }
         } else {
             if (pixel_shift) {
                 AV_COPY128(top_border+32, src_cb+8*uvlinesize);
@@ -1769,10 +1787,11 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i
     /* is_h264 should always be true if SVQ3 is disabled. */
     const int is_h264 = !CONFIG_SVQ3_DECODER || simple || s->codec_id == CODEC_ID_H264;
     void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride);
+    const int block_h = 16>>s->chroma_y_shift;
 
     dest_y  = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize  ) * 16;
-    dest_cb = s->current_picture.f.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) *  8;
-    dest_cr = s->current_picture.f.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) *  8;
+    dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift)*8 + mb_y * s->uvlinesize * block_h;
+    dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift)*8 + mb_y * s->uvlinesize * block_h;
 
     s->dsp.prefetch(dest_y + (s->mb_x&3)*4*s->linesize + (64 << pixel_shift), s->linesize, 4);
     s->dsp.prefetch(dest_cb + (s->mb_x&7)*s->uvlinesize + (64 << pixel_shift), dest_cr - dest_cb, 2);
@@ -1785,8 +1804,8 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i
         block_offset = &h->block_offset[48];
         if(mb_y&1){ //FIXME move out of this function?
             dest_y -= s->linesize*15;
-            dest_cb-= s->uvlinesize*7;
-            dest_cr-= s->uvlinesize*7;
+            dest_cb-= s->uvlinesize*(block_h-1);
+            dest_cr-= s->uvlinesize*(block_h-1);
         }
         if(FRAME_MBAFF) {
             int list;
@@ -1833,12 +1852,12 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i
                         }
                     }
                 } else {
-                for (i = 0; i < 8; i++) {
+                for (i = 0; i < block_h; i++) {
                     uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize);
                     for (j = 0; j < 8; j++)
                         tmp_cb[j] = get_bits(&gb, bit_depth);
                 }
-                for (i = 0; i < 8; i++) {
+                for (i = 0; i < block_h; i++) {
                     uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize);
                     for (j = 0; j < 8; j++)
                         tmp_cr[j] = get_bits(&gb, bit_depth);
@@ -1856,7 +1875,7 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i
                         memset(dest_cr+ i*uvlinesize, 1 << (bit_depth - 1), 8);
                     }
                 } else {
-                for (i=0; i<8; i++) {
+                for (i=0; i<block_h; i++) {
                     memcpy(dest_cb+ i*uvlinesize, h->mb + 128 + i*4,  8);
                     memcpy(dest_cr+ i*uvlinesize, h->mb + 160 + i*4,  8);
                 }
@@ -1904,10 +1923,18 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i
                 }
             }else{
                 if(is_h264){
+                    int qp[2];
+                    if (CHROMA422) {
+                        qp[0] = h->chroma_qp[0]+3;
+                        qp[1] = h->chroma_qp[1]+3;
+                    } else {
+                        qp[0] = h->chroma_qp[0];
+                        qp[1] = h->chroma_qp[1];
+                    }
                     if(h->non_zero_count_cache[ scan8[CHROMA_DC_BLOCK_INDEX+0] ])
-                        h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*1 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]);
+                        h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*1 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][qp[0]][0]);
                     if(h->non_zero_count_cache[ scan8[CHROMA_DC_BLOCK_INDEX+1] ])
-                        h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*2 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]);
+                        h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*2 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][qp[1]][0]);
                     h->h264dsp.h264_idct_add8(dest, block_offset,
                                               h->mb, uvlinesize,
                                               h->non_zero_count_cache);
@@ -2545,11 +2572,13 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
 
     h->b_stride=  s->mb_width*4;
 
+    s->chroma_y_shift = h->sps.chroma_format_idc <= 1; // 400 uses yuv420p
+
     s->width = 16*s->mb_width - (2>>CHROMA444)*FFMIN(h->sps.crop_right, (8<<CHROMA444)-1);
     if(h->sps.frame_mbs_only_flag)
-        s->height= 16*s->mb_height - (2>>CHROMA444)*FFMIN(h->sps.crop_bottom, (8<<CHROMA444)-1);
+        s->height= 16*s->mb_height - (1<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1);
     else
-        s->height= 16*s->mb_height - (4>>CHROMA444)*FFMIN(h->sps.crop_bottom, (8<<CHROMA444)-1);
+        s->height= 16*s->mb_height - (2<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1);
 
     if (s->context_initialized
         && (   s->width != s->avctx->width || s->height != s->avctx->height
@@ -2594,14 +2623,24 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
 
         switch (h->sps.bit_depth_luma) {
             case 9 :
-                s->avctx->pix_fmt = CHROMA444 ? PIX_FMT_YUV444P9 : PIX_FMT_YUV420P9;
+                if (CHROMA444)
+                    s->avctx->pix_fmt = PIX_FMT_YUV444P9;
+                else
+                    s->avctx->pix_fmt = PIX_FMT_YUV420P9;
                 break;
             case 10 :
-                s->avctx->pix_fmt = CHROMA444 ? PIX_FMT_YUV444P10 : PIX_FMT_YUV420P10;
+                if (CHROMA444)
+                    s->avctx->pix_fmt = PIX_FMT_YUV444P10;
+                else if (CHROMA422)
+                    s->avctx->pix_fmt = PIX_FMT_YUV422P10;
+                else
+                    s->avctx->pix_fmt = PIX_FMT_YUV420P10;
                 break;
             default:
                 if (CHROMA444){
                     s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ444P : PIX_FMT_YUV444P;
+                }else if (CHROMA422) {
+                    s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ422P : PIX_FMT_YUV422P;
                 }else{
                     s->avctx->pix_fmt = s->avctx->get_format(s->avctx,
                                                              s->avctx->codec->pix_fmts ?
@@ -3263,6 +3302,7 @@ static void loop_filter(H264Context *h, int start_x, int end_x){
     const int end_mb_y= s->mb_y + FRAME_MBAFF;
     const int old_slice_type= h->slice_type;
     const int pixel_shift = h->pixel_shift;
+    const int block_h = 16>>s->chroma_y_shift;
 
     if(h->deblocking_filter) {
         for(mb_x= start_x; mb_x<end_x; mb_x++){
@@ -3279,8 +3319,8 @@ static void loop_filter(H264Context *h, int start_x, int end_x){
                 s->mb_x= mb_x;
                 s->mb_y= mb_y;
                 dest_y  = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize  ) * 16;
-                dest_cb = s->current_picture.f.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444);
-                dest_cr = s->current_picture.f.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444);
+                dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift)*(8<<CHROMA444) + mb_y * s->uvlinesize * block_h;
+                dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift)*(8<<CHROMA444) + mb_y * s->uvlinesize * block_h;
                     //FIXME simplify above
 
                 if (MB_FIELD) {
@@ -3288,14 +3328,14 @@ static void loop_filter(H264Context *h, int start_x, int end_x){
                     uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2;
                     if(mb_y&1){ //FIXME move out of this function?
                         dest_y -= s->linesize*15;
-                        dest_cb-= s->uvlinesize*((8 << CHROMA444)-1);
-                        dest_cr-= s->uvlinesize*((8 << CHROMA444)-1);
+                        dest_cb-= s->uvlinesize*(block_h-1);
+                        dest_cr-= s->uvlinesize*(block_h-1);
                     }
                 } else {
                     linesize   = h->mb_linesize   = s->linesize;
                     uvlinesize = h->mb_uvlinesize = s->uvlinesize;
                 }
-                backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, CHROMA444, 0);
+                backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0);
                 if(fill_filter_caches(h, mb_type))
                     continue;
                 h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mb_xy]);
@@ -3731,13 +3771,15 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){
             if(avctx->has_b_frames < 2)
                 avctx->has_b_frames= !s->low_delay;
 
-            if (avctx->bits_per_raw_sample != h->sps.bit_depth_luma) {
+            if (avctx->bits_per_raw_sample != h->sps.bit_depth_luma ||
+                h->cur_chroma_format_idc != h->sps.chroma_format_idc) {
                 if (h->sps.bit_depth_luma >= 8 && h->sps.bit_depth_luma <= 10) {
                     avctx->bits_per_raw_sample = h->sps.bit_depth_luma;
+                    h->cur_chroma_format_idc = h->sps.chroma_format_idc;
                     h->pixel_shift = h->sps.bit_depth_luma > 8;
 
-                    ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma);
-                    ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma);
+                    ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma, h->sps.chroma_format_idc);
+                    ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma, h->sps.chroma_format_idc);
                     s->dsp.dct_bits = h->sps.bit_depth_luma > 8 ? 32 : 16;
                     dsputil_init(&s->dsp, s->avctx);
                 } else {
diff --git a/libavcodec/h264.h b/libavcodec/h264.h
index d448fc3330..2809e3253e 100644
--- a/libavcodec/h264.h
+++ b/libavcodec/h264.h
@@ -39,13 +39,6 @@
 #define interlaced_dct interlaced_dct_is_a_bad_name
 #define mb_intra mb_intra_is_not_initialized_see_mb_type
 
-#define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
-#define COEFF_TOKEN_VLC_BITS           8
-#define TOTAL_ZEROS_VLC_BITS           9
-#define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
-#define RUN_VLC_BITS                   3
-#define RUN7_VLC_BITS                  6
-
 #define MAX_SPS_COUNT 32
 #define MAX_PPS_COUNT 256
 
@@ -92,6 +85,7 @@
 #define CABAC h->pps.cabac
 #endif
 
+#define CHROMA422 (h->sps.chroma_format_idc == 2)
 #define CHROMA444 (h->sps.chroma_format_idc == 3)
 
 #define EXTENDED_SAR          255
@@ -582,6 +576,8 @@ typedef struct H264Context{
     // Timestamp stuff
     int sei_buffering_period_present;  ///< Buffering period SEI flag
     int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
+
+    int cur_chroma_format_idc;
 }H264Context;
 
 
@@ -809,7 +805,7 @@ static av_always_inline void write_back_non_zero_count(H264Context *h){
     AV_COPY32(&nnz[32], &nnz_cache[4+8*11]);
     AV_COPY32(&nnz[36], &nnz_cache[4+8*12]);
 
-    if(CHROMA444){
+    if(!h->s.chroma_y_shift){
         AV_COPY32(&nnz[24], &nnz_cache[4+8* 8]);
         AV_COPY32(&nnz[28], &nnz_cache[4+8* 9]);
         AV_COPY32(&nnz[40], &nnz_cache[4+8*13]);
diff --git a/libavcodec/h264_cabac.c b/libavcodec/h264_cabac.c
index 6bbf87f0ba..55cafc7bbc 100644
--- a/libavcodec/h264_cabac.c
+++ b/libavcodec/h264_cabac.c
@@ -1587,6 +1587,7 @@ static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCT
         9, 9,10,10, 8,11,12,11, 9, 9,10,10, 8,13,13, 9,
         9,10,10, 8,13,13, 9, 9,10,10,14,14,14,14,14 }
     };
+    static const uint8_t sig_coeff_offset_dc[7] = { 0, 0, 1, 1, 2, 2, 2 };
     /* node ctx: 0..3: abslevel1 (with abslevelgt1 == 0).
      * 4..7: abslevelgt1 + 3 (and abslevel1 doesn't matter).
      * map node ctx => cabac ctx for level=1 */
@@ -1651,12 +1652,20 @@ static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCT
         coeff_count= decode_significance_8x8_x86(CC, significant_coeff_ctx_base, index,
                                                  last_coeff_ctx_base, sig_off);
     } else {
-        coeff_count= decode_significance_x86(CC, max_coeff, significant_coeff_ctx_base, index,
-                                             last_coeff_ctx_base-significant_coeff_ctx_base);
+        if (is_dc && max_coeff == 8) { // dc 422
+            DECODE_SIGNIFICANCE(7, sig_coeff_offset_dc[last], sig_coeff_offset_dc[last]);
+        } else {
+            coeff_count= decode_significance_x86(CC, max_coeff, significant_coeff_ctx_base, index,
+                                                 last_coeff_ctx_base-significant_coeff_ctx_base);
+        }
 #else
         DECODE_SIGNIFICANCE( 63, sig_off[last], last_coeff_flag_offset_8x8[last] );
     } else {
-        DECODE_SIGNIFICANCE( max_coeff - 1, last, last );
+        if (is_dc && max_coeff == 8) { // dc 422
+            DECODE_SIGNIFICANCE(7, sig_coeff_offset_dc[last], sig_coeff_offset_dc[last]);
+        } else {
+            DECODE_SIGNIFICANCE(max_coeff - 1, last, last);
+        }
 #endif
     }
     assert(coeff_count > 0);
@@ -1692,6 +1701,8 @@ static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCT
             } \
         } else { \
             int coeff_abs = 2; \
+            if (is_dc && max_coeff == 8) \
+                node_ctx = FFMIN(node_ctx, 6); \
             ctx = coeff_abs_levelgt1_ctx[node_ctx] + abs_level_m1_ctx_base; \
             node_ctx = coeff_abs_level_transition[1][node_ctx]; \
 \
@@ -2315,22 +2326,31 @@ decode_intra_mb:
             decode_cabac_luma_residual(h, scan, scan8x8, pixel_shift, mb_type, cbp, 1);
             decode_cabac_luma_residual(h, scan, scan8x8, pixel_shift, mb_type, cbp, 2);
         } else {
+            const int num_c8x8 = h->sps.chroma_format_idc;
+
             if( cbp&0x30 ){
                 int c;
                 for( c = 0; c < 2; c++ ) {
                     //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-DC\n",c );
-                    decode_cabac_residual_dc(h, h->mb + ((256 + 16*16*c) << pixel_shift), 3, CHROMA_DC_BLOCK_INDEX+c, chroma_dc_scan, 4);
+                    decode_cabac_residual_dc(h, h->mb + ((256 + 16*16*c) << pixel_shift), 3,
+                                             CHROMA_DC_BLOCK_INDEX+c,
+                                             CHROMA422 ? chroma422_dc_scan : chroma_dc_scan,
+                                             4*num_c8x8);
                 }
             }
 
             if( cbp&0x20 ) {
-                int c, i;
+                int c, i, i8x8;
                 for( c = 0; c < 2; c++ ) {
+                    DCTELEM *mb = h->mb + (16*(16 + 16*c) << pixel_shift);
                     qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[c]];
-                    for( i = 0; i < 4; i++ ) {
-                        const int index = 16 + 16 * c + i;
-                        //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16 );
-                        decode_cabac_residual_nondc(h, h->mb + (16*index << pixel_shift), 4, index, scan + 1, qmul, 15);
+                    for (i8x8 = 0; i8x8 < num_c8x8; i8x8++) {
+                        for (i = 0; i < 4; i++) {
+                            const int index = 16 + 16 * c + 8*i8x8 + i;
+                            //av_log(s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16);
+                            decode_cabac_residual_nondc(h, mb, 4, index, scan + 1, qmul, 15);
+                            mb += 16<<pixel_shift;
+                        }
                     }
                 }
             } else {
diff --git a/libavcodec/h264_cavlc.c b/libavcodec/h264_cavlc.c
index 0cd147fd7a..cc38370047 100644
--- a/libavcodec/h264_cavlc.c
+++ b/libavcodec/h264_cavlc.c
@@ -62,6 +62,30 @@ static const uint8_t chroma_dc_coeff_token_bits[4*5]={
  2, 3, 2, 0,
 };
 
+static const uint8_t chroma422_dc_coeff_token_len[4*9]={
+  1,  0,  0,  0,
+  7,  2,  0,  0,
+  7,  7,  3,  0,
+  9,  7,  7,  5,
+  9,  9,  7,  6,
+ 10, 10,  9,  7,
+ 11, 11, 10,  7,
+ 12, 12, 11, 10,
+ 13, 12, 12, 11,
+};
+
+static const uint8_t chroma422_dc_coeff_token_bits[4*9]={
+  1,   0,  0, 0,
+ 15,   1,  0, 0,
+ 14,  13,  1, 0,
+  7,  12, 11, 1,
+  6,   5, 10, 1,
+  7,   6,  4, 9,
+  7,   6,  5, 8,
+  7,   6,  5, 4,
+  7,   5,  4, 4,
+};
+
 static const uint8_t coeff_token_len[4][4*17]={
 {
      1, 0, 0, 0,
@@ -172,6 +196,26 @@ static const uint8_t chroma_dc_total_zeros_bits[3][4]= {
     { 1, 0, 0, 0,},
 };
 
+static const uint8_t chroma422_dc_total_zeros_len[7][8]= {
+    { 1, 3, 3, 4, 4, 4, 5, 5 },
+    { 3, 2, 3, 3, 3, 3, 3 },
+    { 3, 3, 2, 2, 3, 3 },
+    { 3, 2, 2, 2, 3 },
+    { 2, 2, 2, 2 },
+    { 2, 2, 1 },
+    { 1, 1 },
+};
+
+static const uint8_t chroma422_dc_total_zeros_bits[7][8]= {
+    { 1, 2, 3, 2, 3, 1, 1, 0 },
+    { 0, 1, 1, 4, 5, 6, 7 },
+    { 0, 1, 1, 2, 6, 7 },
+    { 6, 0, 1, 2, 7 },
+    { 0, 1, 2, 3 },
+    { 0, 1, 1 },
+    { 0, 1 },
+};
+
 static const uint8_t run_len[7][16]={
     {1,1},
     {1,2,2},
@@ -200,6 +244,10 @@ static VLC chroma_dc_coeff_token_vlc;
 static VLC_TYPE chroma_dc_coeff_token_vlc_table[256][2];
 static const int chroma_dc_coeff_token_vlc_table_size = 256;
 
+static VLC chroma422_dc_coeff_token_vlc;
+static VLC_TYPE chroma422_dc_coeff_token_vlc_table[8192][2];
+static const int chroma422_dc_coeff_token_vlc_table_size = 8192;
+
 static VLC total_zeros_vlc[15];
 static VLC_TYPE total_zeros_vlc_tables[15][512][2];
 static const int total_zeros_vlc_tables_size = 512;
@@ -208,6 +256,10 @@ static VLC chroma_dc_total_zeros_vlc[3];
 static VLC_TYPE chroma_dc_total_zeros_vlc_tables[3][8][2];
 static const int chroma_dc_total_zeros_vlc_tables_size = 8;
 
+static VLC chroma422_dc_total_zeros_vlc[7];
+static VLC_TYPE chroma422_dc_total_zeros_vlc_tables[7][32][2];
+static const int chroma422_dc_total_zeros_vlc_tables_size = 32;
+
 static VLC run_vlc[6];
 static VLC_TYPE run_vlc_tables[6][8][2];
 static const int run_vlc_tables_size = 8;
@@ -219,6 +271,14 @@ static const int run7_vlc_table_size = 96;
 #define LEVEL_TAB_BITS 8
 static int8_t cavlc_level_tab[7][1<<LEVEL_TAB_BITS][2];
 
+#define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
+#define CHROMA422_DC_COEFF_TOKEN_VLC_BITS 13
+#define COEFF_TOKEN_VLC_BITS           8
+#define TOTAL_ZEROS_VLC_BITS           9
+#define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
+#define CHROMA422_DC_TOTAL_ZEROS_VLC_BITS 5
+#define RUN_VLC_BITS                   3
+#define RUN7_VLC_BITS                  6
 
 /**
  * gets the predicted number of non-zero coefficients.
@@ -277,6 +337,13 @@ av_cold void ff_h264_decode_init_vlc(void){
                  &chroma_dc_coeff_token_bits[0], 1, 1,
                  INIT_VLC_USE_NEW_STATIC);
 
+        chroma422_dc_coeff_token_vlc.table = chroma422_dc_coeff_token_vlc_table;
+        chroma422_dc_coeff_token_vlc.table_allocated = chroma422_dc_coeff_token_vlc_table_size;
+        init_vlc(&chroma422_dc_coeff_token_vlc, CHROMA422_DC_COEFF_TOKEN_VLC_BITS, 4*9,
+                 &chroma422_dc_coeff_token_len [0], 1, 1,
+                 &chroma422_dc_coeff_token_bits[0], 1, 1,
+                 INIT_VLC_USE_NEW_STATIC);
+
         offset = 0;
         for(i=0; i<4; i++){
             coeff_token_vlc[i].table = coeff_token_vlc_tables+offset;
@@ -303,6 +370,17 @@ av_cold void ff_h264_decode_init_vlc(void){
                      &chroma_dc_total_zeros_bits[i][0], 1, 1,
                      INIT_VLC_USE_NEW_STATIC);
         }
+
+        for(i=0; i<7; i++){
+            chroma422_dc_total_zeros_vlc[i].table = chroma422_dc_total_zeros_vlc_tables[i];
+            chroma422_dc_total_zeros_vlc[i].table_allocated = chroma422_dc_total_zeros_vlc_tables_size;
+            init_vlc(&chroma422_dc_total_zeros_vlc[i],
+                     CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 8,
+                     &chroma422_dc_total_zeros_len [i][0], 1, 1,
+                     &chroma422_dc_total_zeros_bits[i][0], 1, 1,
+                     INIT_VLC_USE_NEW_STATIC);
+        }
+
         for(i=0; i<15; i++){
             total_zeros_vlc[i].table = total_zeros_vlc_tables[i];
             total_zeros_vlc[i].table_allocated = total_zeros_vlc_tables_size;
@@ -372,7 +450,10 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in
     //FIXME put trailing_onex into the context
 
     if(max_coeff <= 8){
-        coeff_token= get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1);
+        if (max_coeff == 4)
+            coeff_token = get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1);
+        else
+            coeff_token = get_vlc2(gb, chroma422_dc_coeff_token_vlc.table, CHROMA422_DC_COEFF_TOKEN_VLC_BITS, 1);
         total_coeff= coeff_token>>2;
     }else{
         if(n >= LUMA_DC_BLOCK_INDEX){
@@ -482,11 +563,16 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in
     if(total_coeff == max_coeff)
         zeros_left=0;
     else{
-        /* FIXME: we don't actually support 4:2:2 yet. */
-        if(max_coeff <= 8)
-            zeros_left= get_vlc2(gb, (chroma_dc_total_zeros_vlc-1)[ total_coeff ].table, CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1);
-        else
+        if (max_coeff <= 8) {
+            if (max_coeff == 4)
+                zeros_left = get_vlc2(gb, (chroma_dc_total_zeros_vlc-1)[total_coeff].table,
+                                      CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1);
+            else
+                zeros_left = get_vlc2(gb, (chroma422_dc_total_zeros_vlc-1)[total_coeff].table,
+                                      CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 1);
+        } else {
             zeros_left= get_vlc2(gb, (total_zeros_vlc-1)[ total_coeff ].table, TOTAL_ZEROS_VLC_BITS, 1);
+        }
     }
 
 #define STORE_BLOCK(type) \
@@ -993,7 +1079,7 @@ decode_intra_mb:
     s->current_picture.f.mb_type[mb_xy] = mb_type;
 
     if(cbp || IS_INTRA16x16(mb_type)){
-        int i4x4, chroma_idx;
+        int i4x4, i8x8, chroma_idx;
         int dquant;
         int ret;
         GetBitContext *gb= IS_INTRA(mb_type) ? h->intra_gb_ptr : h->inter_gb_ptr;
@@ -1036,9 +1122,14 @@ decode_intra_mb:
                 return -1;
             }
         } else {
+            const int num_c8x8 = h->sps.chroma_format_idc;
+
             if(cbp&0x30){
                 for(chroma_idx=0; chroma_idx<2; chroma_idx++)
-                    if( decode_residual(h, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift), CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma_dc_scan, NULL, 4) < 0){
+                    if (decode_residual(h, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift),
+                                        CHROMA_DC_BLOCK_INDEX+chroma_idx,
+                                        CHROMA422 ? chroma422_dc_scan : chroma_dc_scan,
+                                        NULL, 4*num_c8x8) < 0) {
                         return -1;
                     }
             }
@@ -1046,10 +1137,13 @@ decode_intra_mb:
             if(cbp&0x20){
                 for(chroma_idx=0; chroma_idx<2; chroma_idx++){
                     const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]];
-                    for(i4x4=0; i4x4<4; i4x4++){
-                        const int index= 16 + 16*chroma_idx + i4x4;
-                        if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){
-                            return -1;
+                    DCTELEM *mb = h->mb + (16*(16 + 16*chroma_idx) << pixel_shift);
+                    for (i8x8=0; i8x8<num_c8x8; i8x8++) {
+                        for (i4x4=0; i4x4<4; i4x4++) {
+                            const int index= 16 + 16*chroma_idx + 8*i8x8 + i4x4;
+                            if (decode_residual(h, gb, mb, index, scan + 1, qmul, 15) < 0)
+                                return -1;
+                            mb += 16<<pixel_shift;
                         }
                     }
                 }
diff --git a/libavcodec/h264_loopfilter.c b/libavcodec/h264_loopfilter.c
index 0b7806994f..67399395a1 100644
--- a/libavcodec/h264_loopfilter.c
+++ b/libavcodec/h264_loopfilter.c
@@ -212,6 +212,7 @@ static void av_always_inline h264_filter_mb_fast_internal( H264Context *h, int m
     MpegEncContext * const s = &h->s;
     int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
     int chroma444 = CHROMA444;
+    int chroma422 = CHROMA422;
 
     int mb_xy = h->mb_xy;
     int left_type= h->left_type[LTOP];
@@ -289,6 +290,23 @@ static void av_always_inline h264_filter_mb_fast_internal( H264Context *h, int m
                     filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
                     filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
                 }
+            }else if(chroma422){
+                if(left_type){
+                    filter_mb_edgecv(&img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
+                    filter_mb_edgecv(&img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
+                }
+                filter_mb_edgecv(&img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
+                filter_mb_edgecv(&img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
+                if(top_type){
+                    filter_mb_edgech(&img_cb[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
+                    filter_mb_edgech(&img_cr[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
+                }
+                filter_mb_edgech(&img_cb[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
+                filter_mb_edgech(&img_cr[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
+                filter_mb_edgech(&img_cb[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
+                filter_mb_edgech(&img_cr[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
+                filter_mb_edgech(&img_cb[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
+                filter_mb_edgech(&img_cr[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
             }else{
                 if(left_type){
                     filter_mb_edgecv( &img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
@@ -411,10 +429,12 @@ static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){
     return v;
 }
 
-static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int a, int b, int chroma, int chroma444, int dir) {
+static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int a, int b, int chroma, int dir) {
     MpegEncContext * const s = &h->s;
     int edge;
     int chroma_qp_avg[2];
+    int chroma444 = CHROMA444;
+    int chroma422 = CHROMA422;
     const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy;
     const int mbm_type = dir == 0 ? h->left_type[LTOP] : h->top_type;
 
@@ -564,8 +584,9 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
     for( edge = 1; edge < edges; edge++ ) {
         DECLARE_ALIGNED(8, int16_t, bS)[4];
         int qp;
+        const int deblock_edge = !IS_8x8DCT(mb_type & (edge<<24)); // (edge&1) && IS_8x8DCT(mb_type)
 
-        if( IS_8x8DCT(mb_type & (edge<<24)) ) // (edge&1) && IS_8x8DCT(mb_type)
+        if (!deblock_edge && (!chroma422 || dir == 0))
             continue;
 
         if( IS_INTRA(mb_type)) {
@@ -627,14 +648,23 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
                 }
             }
         } else {
-            filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0 );
-            if (chroma) {
-                if (chroma444) {
-                    filter_mb_edgeh ( &img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
-                    filter_mb_edgeh ( &img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
-                } else if( (edge&1) == 0 ) {
-                    filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
-                    filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
+            if (chroma422) {
+                if (deblock_edge)
+                    filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0);
+                if (chroma) {
+                    filter_mb_edgech(&img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
+                    filter_mb_edgech(&img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
+                }
+            } else {
+                filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0);
+                if (chroma) {
+                    if (chroma444) {
+                        filter_mb_edgeh (&img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
+                        filter_mb_edgeh (&img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
+                    } else if ((edge&1) == 0) {
+                        filter_mb_edgech(&img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
+                        filter_mb_edgech(&img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
+                    }
                 }
             }
         }
@@ -726,6 +756,11 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
                     filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
                     filter_mb_mbaff_edgev ( h, img_cr,                uvlinesize, bS  , 1, rqp[0], a, b, 1 );
                     filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
+                } else if (CHROMA422) {
+                    filter_mb_mbaff_edgecv(h, img_cb,                uvlinesize, bS  , 1, bqp[0], a, b, 1);
+                    filter_mb_mbaff_edgecv(h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1);
+                    filter_mb_mbaff_edgecv(h, img_cr,                uvlinesize, bS  , 1, rqp[0], a, b, 1);
+                    filter_mb_mbaff_edgecv(h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1);
                 }else{
                     filter_mb_mbaff_edgecv( h, img_cb,                uvlinesize, bS  , 1, bqp[0], a, b, 1 );
                     filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
@@ -754,9 +789,9 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
 
 #if CONFIG_SMALL
     for( dir = 0; dir < 2; dir++ )
-        filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, a, b, chroma, CHROMA444, dir);
+        filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, a, b, chroma, dir);
 #else
-    filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, a, b, chroma, CHROMA444, 0);
-    filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0,                        a, b, chroma, CHROMA444, 1);
+    filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, a, b, chroma, 0);
+    filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0,                        a, b, chroma, 1);
 #endif
 }
diff --git a/libavcodec/h264_mvpred.h b/libavcodec/h264_mvpred.h
index d2677eaa6c..b46d0985a8 100644
--- a/libavcodec/h264_mvpred.h
+++ b/libavcodec/h264_mvpred.h
@@ -510,7 +510,7 @@ static void fill_decode_caches(H264Context *h, int mb_type){
     if(top_type){
         nnz = h->non_zero_count[top_xy];
         AV_COPY32(&nnz_cache[4+8* 0], &nnz[4*3]);
-        if(CHROMA444){
+        if(!s->chroma_y_shift){
             AV_COPY32(&nnz_cache[4+8* 5], &nnz[4* 7]);
             AV_COPY32(&nnz_cache[4+8*10], &nnz[4*11]);
         }else{
@@ -534,6 +534,11 @@ static void fill_decode_caches(H264Context *h, int mb_type){
                 nnz_cache[3+8* 7 + 2*8*i]= nnz[left_block[8+1+2*i]+4*4];
                 nnz_cache[3+8*11 + 2*8*i]= nnz[left_block[8+0+2*i]+8*4];
                 nnz_cache[3+8*12 + 2*8*i]= nnz[left_block[8+1+2*i]+8*4];
+            }else if(CHROMA422) {
+                nnz_cache[3+8* 6 + 2*8*i]= nnz[left_block[8+0+2*i]-2+4*4];
+                nnz_cache[3+8* 7 + 2*8*i]= nnz[left_block[8+1+2*i]-2+4*4];
+                nnz_cache[3+8*11 + 2*8*i]= nnz[left_block[8+0+2*i]-2+8*4];
+                nnz_cache[3+8*12 + 2*8*i]= nnz[left_block[8+1+2*i]-2+8*4];
             }else{
                 nnz_cache[3+8* 6 +   8*i]= nnz[left_block[8+4+2*i]];
                 nnz_cache[3+8*11 +   8*i]= nnz[left_block[8+5+2*i]];
diff --git a/libavcodec/h264_ps.c b/libavcodec/h264_ps.c
index 61fb12ce0c..3c7b04abc6 100644
--- a/libavcodec/h264_ps.c
+++ b/libavcodec/h264_ps.c
@@ -396,7 +396,8 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){
 #endif
     sps->crop= get_bits1(&s->gb);
     if(sps->crop){
-        int crop_limit = sps->chroma_format_idc == 3 ? 16 : 8;
+        int crop_vertical_limit = sps->chroma_format_idc & 2 ? 16 : 8;
+        int crop_horizontal_limit = sps->chroma_format_idc == 3 ? 16 : 8;
         sps->crop_left  = get_ue_golomb(&s->gb);
         sps->crop_right = get_ue_golomb(&s->gb);
         sps->crop_top   = get_ue_golomb(&s->gb);
@@ -404,7 +405,7 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){
         if(sps->crop_left || sps->crop_top){
             av_log(h->s.avctx, AV_LOG_ERROR, "insane cropping not completely supported, this could look slightly wrong ...\n");
         }
-        if(sps->crop_right >= crop_limit || sps->crop_bottom >= crop_limit){
+        if(sps->crop_right >= crop_horizontal_limit || sps->crop_bottom >= crop_vertical_limit){
             av_log(h->s.avctx, AV_LOG_ERROR, "brainfart cropping not supported, this could look slightly wrong ...\n");
         }
     }else{
diff --git a/libavcodec/h264data.h b/libavcodec/h264data.h
index c459523f71..a5ed069e94 100644
--- a/libavcodec/h264data.h
+++ b/libavcodec/h264data.h
@@ -80,7 +80,14 @@ static const uint8_t luma_dc_field_scan[16]={
 
 static const uint8_t chroma_dc_scan[4]={
  (0+0*2)*16, (1+0*2)*16,
- (0+1*2)*16, (1+1*2)*16,  //FIXME
+ (0+1*2)*16, (1+1*2)*16,
+};
+
+static const uint8_t chroma422_dc_scan[8]={
+ (0+0*2)*16, (0+1*2)*16,
+ (1+0*2)*16, (0+2*2)*16,
+ (0+3*2)*16, (1+1*2)*16,
+ (1+2*2)*16, (1+3*2)*16,
 };
 
 // zigzag_scan8x8_cavlc[i] = zigzag_scan8x8[(i/4) + 16*(i%4)]
diff --git a/libavcodec/h264dsp.c b/libavcodec/h264dsp.c
index 96a38ff77d..cf0067b8f0 100644
--- a/libavcodec/h264dsp.c
+++ b/libavcodec/h264dsp.c
@@ -41,7 +41,7 @@
 #include "h264dsp_template.c"
 #undef BIT_DEPTH
 
-void ff_h264dsp_init(H264DSPContext *c, const int bit_depth)
+void ff_h264dsp_init(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
 {
 #undef FUNC
 #define FUNC(a, depth) a ## _ ## depth ## _c
@@ -53,10 +53,16 @@ void ff_h264dsp_init(H264DSPContext *c, const int bit_depth)
     c->h264_idct8_dc_add= FUNC(ff_h264_idct8_dc_add, depth);\
     c->h264_idct_add16     = FUNC(ff_h264_idct_add16, depth);\
     c->h264_idct8_add4     = FUNC(ff_h264_idct8_add4, depth);\
-    c->h264_idct_add8      = FUNC(ff_h264_idct_add8, depth);\
+    if (chroma_format_idc == 1)\
+        c->h264_idct_add8  = FUNC(ff_h264_idct_add8, depth);\
+    else\
+        c->h264_idct_add8  = FUNC(ff_h264_idct_add8_422, depth);\
     c->h264_idct_add16intra= FUNC(ff_h264_idct_add16intra, depth);\
     c->h264_luma_dc_dequant_idct= FUNC(ff_h264_luma_dc_dequant_idct, depth);\
-    c->h264_chroma_dc_dequant_idct= FUNC(ff_h264_chroma_dc_dequant_idct, depth);\
+    if (chroma_format_idc == 1)\
+        c->h264_chroma_dc_dequant_idct= FUNC(ff_h264_chroma_dc_dequant_idct, depth);\
+    else\
+        c->h264_chroma_dc_dequant_idct= FUNC(ff_h264_chroma422_dc_dequant_idct, depth);\
 \
     c->weight_h264_pixels_tab[0]= FUNC(weight_h264_pixels16x16, depth);\
     c->weight_h264_pixels_tab[1]= FUNC(weight_h264_pixels16x8, depth);\
@@ -86,11 +92,23 @@ void ff_h264dsp_init(H264DSPContext *c, const int bit_depth)
     c->h264_h_loop_filter_luma_intra= FUNC(h264_h_loop_filter_luma_intra, depth);\
     c->h264_h_loop_filter_luma_mbaff_intra= FUNC(h264_h_loop_filter_luma_mbaff_intra, depth);\
     c->h264_v_loop_filter_chroma= FUNC(h264_v_loop_filter_chroma, depth);\
-    c->h264_h_loop_filter_chroma= FUNC(h264_h_loop_filter_chroma, depth);\
-    c->h264_h_loop_filter_chroma_mbaff= FUNC(h264_h_loop_filter_chroma_mbaff, depth);\
+    if (chroma_format_idc == 1)\
+        c->h264_h_loop_filter_chroma= FUNC(h264_h_loop_filter_chroma, depth);\
+    else\
+        c->h264_h_loop_filter_chroma= FUNC(h264_h_loop_filter_chroma422, depth);\
+    if (chroma_format_idc == 1)\
+        c->h264_h_loop_filter_chroma_mbaff= FUNC(h264_h_loop_filter_chroma_mbaff, depth);\
+    else\
+        c->h264_h_loop_filter_chroma_mbaff= FUNC(h264_h_loop_filter_chroma422_mbaff, depth);\
     c->h264_v_loop_filter_chroma_intra= FUNC(h264_v_loop_filter_chroma_intra, depth);\
-    c->h264_h_loop_filter_chroma_intra= FUNC(h264_h_loop_filter_chroma_intra, depth);\
-    c->h264_h_loop_filter_chroma_mbaff_intra= FUNC(h264_h_loop_filter_chroma_mbaff_intra, depth);\
+    if (chroma_format_idc == 1)\
+        c->h264_h_loop_filter_chroma_intra= FUNC(h264_h_loop_filter_chroma_intra, depth);\
+    else\
+        c->h264_h_loop_filter_chroma_intra= FUNC(h264_h_loop_filter_chroma422_intra, depth);\
+    if (chroma_format_idc == 1)\
+        c->h264_h_loop_filter_chroma_mbaff_intra= FUNC(h264_h_loop_filter_chroma_mbaff_intra, depth);\
+    else\
+        c->h264_h_loop_filter_chroma_mbaff_intra= FUNC(h264_h_loop_filter_chroma422_mbaff_intra, depth);\
     c->h264_loop_filter_strength= NULL;
 
     switch (bit_depth) {
@@ -105,7 +123,7 @@ void ff_h264dsp_init(H264DSPContext *c, const int bit_depth)
         break;
     }
 
-    if (ARCH_ARM) ff_h264dsp_init_arm(c, bit_depth);
-    if (HAVE_ALTIVEC) ff_h264dsp_init_ppc(c, bit_depth);
-    if (HAVE_MMX) ff_h264dsp_init_x86(c, bit_depth);
+    if (ARCH_ARM) ff_h264dsp_init_arm(c, bit_depth, chroma_format_idc);
+    if (HAVE_ALTIVEC) ff_h264dsp_init_ppc(c, bit_depth, chroma_format_idc);
+    if (HAVE_MMX) ff_h264dsp_init_x86(c, bit_depth, chroma_format_idc);
 }
diff --git a/libavcodec/h264dsp.h b/libavcodec/h264dsp.h
index b61967adfc..c79ab0a625 100644
--- a/libavcodec/h264dsp.h
+++ b/libavcodec/h264dsp.h
@@ -74,9 +74,9 @@ typedef struct H264DSPContext{
     void (*h264_chroma_dc_dequant_idct)(DCTELEM *block, int qmul);
 }H264DSPContext;
 
-void ff_h264dsp_init(H264DSPContext *c, const int bit_depth);
-void ff_h264dsp_init_arm(H264DSPContext *c, const int bit_depth);
-void ff_h264dsp_init_ppc(H264DSPContext *c, const int bit_depth);
-void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth);
+void ff_h264dsp_init(H264DSPContext *c, const int bit_depth, const int chroma_format_idc);
+void ff_h264dsp_init_arm(H264DSPContext *c, const int bit_depth, const int chroma_format_idc);
+void ff_h264dsp_init_ppc(H264DSPContext *c, const int bit_depth, const int chroma_format_idc);
+void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chroma_format_idc);
 
 #endif /* AVCODEC_H264DSP_H */
diff --git a/libavcodec/h264dsp_template.c b/libavcodec/h264dsp_template.c
index 906d99f739..3023541634 100644
--- a/libavcodec/h264dsp_template.c
+++ b/libavcodec/h264dsp_template.c
@@ -275,6 +275,14 @@ static void FUNCC(h264_h_loop_filter_chroma_mbaff)(uint8_t *pix, int stride, int
 {
     FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 1, alpha, beta, tc0);
 }
+static void FUNCC(h264_h_loop_filter_chroma422)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
+{
+    FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 4, alpha, beta, tc0);
+}
+static void FUNCC(h264_h_loop_filter_chroma422_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
+{
+    FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
+}
 
 static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma_intra)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta)
 {
@@ -312,3 +320,11 @@ static void FUNCC(h264_h_loop_filter_chroma_mbaff_intra)(uint8_t *pix, int strid
 {
     FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 1, alpha, beta);
 }
+static void FUNCC(h264_h_loop_filter_chroma422_intra)(uint8_t *pix, int stride, int alpha, int beta)
+{
+    FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 4, alpha, beta);
+}
+static void FUNCC(h264_h_loop_filter_chroma422_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
+{
+    FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
+}
diff --git a/libavcodec/h264idct_template.c b/libavcodec/h264idct_template.c
index 3b386695a2..64bc70dd47 100644
--- a/libavcodec/h264idct_template.c
+++ b/libavcodec/h264idct_template.c
@@ -224,6 +224,39 @@ void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *
         }
     }
 }
+
+void FUNCC(ff_h264_idct_add8_422)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
+    int i, j;
+
+#if 0
+    av_log(NULL, AV_LOG_INFO, "idct\n");
+    int32_t *b = block;
+    for (int i = 0; i < 256; i++) {
+        av_log(NULL, AV_LOG_INFO, "%5d ", b[i+256]);
+        if (!((i+1) % 16))
+            av_log(NULL, AV_LOG_INFO, "\n");
+    }
+#endif
+
+    for(j=1; j<3; j++){
+        for(i=j*16; i<j*16+4; i++){
+            if(nnzc[ scan8[i] ])
+                FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
+            else if(((dctcoef*)block)[i*16])
+                FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
+        }
+    }
+
+    for(j=1; j<3; j++){
+        for(i=j*16+4; i<j*16+8; i++){
+            if(nnzc[ scan8[i+4] ])
+                FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
+            else if(((dctcoef*)block)[i*16])
+                FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
+        }
+    }
+}
+
 /**
  * IDCT transforms the 16 dc values and dequantizes them.
  * @param qmul quantization parameter
@@ -263,6 +296,42 @@ void FUNCC(ff_h264_luma_dc_dequant_idct)(DCTELEM *p_output, DCTELEM *p_input, in
 #undef stride
 }
 
+void FUNCC(ff_h264_chroma422_dc_dequant_idct)(DCTELEM *p_block, int qmul){
+    const int stride= 16*2;
+    const int xStride= 16;
+    int i;
+    int temp[8];
+    static const uint8_t x_offset[2]={0, 16};
+    dctcoef *block = (dctcoef*)p_block;
+
+    for(i=0; i<4; i++){
+        temp[2*i+0] = block[stride*i + xStride*0] + block[stride*i + xStride*1];
+        temp[2*i+1] = block[stride*i + xStride*0] - block[stride*i + xStride*1];
+    }
+
+    for(i=0; i<2; i++){
+        const int offset= x_offset[i];
+        const int z0= temp[2*0+i] + temp[2*2+i];
+        const int z1= temp[2*0+i] - temp[2*2+i];
+        const int z2= temp[2*1+i] - temp[2*3+i];
+        const int z3= temp[2*1+i] + temp[2*3+i];
+
+        block[stride*0+offset]= ((z0 + z3)*qmul + 128) >> 8;
+        block[stride*1+offset]= ((z1 + z2)*qmul + 128) >> 8;
+        block[stride*2+offset]= ((z1 - z2)*qmul + 128) >> 8;
+        block[stride*3+offset]= ((z0 - z3)*qmul + 128) >> 8;
+    }
+
+#if 0
+    av_log(NULL, AV_LOG_INFO, "after chroma dc\n");
+    for (int i = 0; i < 256; i++) {
+        av_log(NULL, AV_LOG_INFO, "%5d ", block[i]);
+        if (!((i+1) % 16))
+            av_log(NULL, AV_LOG_INFO, "\n");
+    }
+#endif
+}
+
 void FUNCC(ff_h264_chroma_dc_dequant_idct)(DCTELEM *p_block, int qmul){
     const int stride= 16*2;
     const int xStride= 16;
diff --git a/libavcodec/h264pred.c b/libavcodec/h264pred.c
index f6533cf9ba..eb4166f1e3 100644
--- a/libavcodec/h264pred.c
+++ b/libavcodec/h264pred.c
@@ -363,7 +363,7 @@ static void pred8x8_tm_vp8_c(uint8_t *src, int stride){
 /**
  * Set the intra prediction function pointers.
  */
-void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth){
+void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth, const int chroma_format_idc){
 //    MpegEncContext * const s = &h->s;
 
 #undef FUNC
@@ -436,20 +436,39 @@ void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth){
     h->pred8x8l[TOP_DC_PRED         ]= FUNCC(pred8x8l_top_dc              , depth);\
     h->pred8x8l[DC_128_PRED         ]= FUNCC(pred8x8l_128_dc              , depth);\
 \
-    h->pred8x8[VERT_PRED8x8   ]= FUNCC(pred8x8_vertical                   , depth);\
-    h->pred8x8[HOR_PRED8x8    ]= FUNCC(pred8x8_horizontal                 , depth);\
+    if (chroma_format_idc == 1) {\
+        h->pred8x8[VERT_PRED8x8   ]= FUNCC(pred8x8_vertical               , depth);\
+        h->pred8x8[HOR_PRED8x8    ]= FUNCC(pred8x8_horizontal             , depth);\
+    } else {\
+        h->pred8x8[VERT_PRED8x8   ]= FUNCC(pred8x16_vertical              , depth);\
+        h->pred8x8[HOR_PRED8x8    ]= FUNCC(pred8x16_horizontal            , depth);\
+    }\
     if (codec_id != CODEC_ID_VP8) {\
-        h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x8_plane                    , depth);\
+        if (chroma_format_idc == 1) {\
+            h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x8_plane                , depth);\
+        } else {\
+            h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x16_plane               , depth);\
+        }\
     } else\
         h->pred8x8[PLANE_PRED8x8]= FUNCD(pred8x8_tm_vp8);\
     if(codec_id != CODEC_ID_RV40 && codec_id != CODEC_ID_VP8){\
-        h->pred8x8[DC_PRED8x8     ]= FUNCC(pred8x8_dc                     , depth);\
-        h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x8_left_dc                , depth);\
-        h->pred8x8[TOP_DC_PRED8x8 ]= FUNCC(pred8x8_top_dc                 , depth);\
-        h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l0t, depth);\
-        h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0lt, depth);\
-        h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l00, depth);\
-        h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0l0, depth);\
+        if (chroma_format_idc == 1) {\
+            h->pred8x8[DC_PRED8x8     ]= FUNCC(pred8x8_dc                     , depth);\
+            h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x8_left_dc                , depth);\
+            h->pred8x8[TOP_DC_PRED8x8 ]= FUNCC(pred8x8_top_dc                 , depth);\
+            h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l0t, depth);\
+            h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0lt, depth);\
+            h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l00, depth);\
+            h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0l0, depth);\
+        } else {\
+            h->pred8x8[DC_PRED8x8     ]= FUNCC(pred8x16_dc                    , depth);\
+            h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x16_left_dc               , depth);\
+            h->pred8x8[TOP_DC_PRED8x8 ]= FUNCC(pred8x16_top_dc                , depth);\
+            h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l0t, depth);\
+            h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0lt, depth);\
+            h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l00, depth);\
+            h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0l0, depth);\
+        }\
     }else{\
         h->pred8x8[DC_PRED8x8     ]= FUNCD(pred8x8_dc_rv40);\
         h->pred8x8[LEFT_DC_PRED8x8]= FUNCD(pred8x8_left_dc_rv40);\
@@ -459,7 +478,11 @@ void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth){
             h->pred8x8[DC_129_PRED8x8]= FUNCC(pred8x8_129_dc              , depth);\
         }\
     }\
-    h->pred8x8[DC_128_PRED8x8 ]= FUNCC(pred8x8_128_dc                     , depth);\
+    if (chroma_format_idc == 1) {\
+        h->pred8x8[DC_128_PRED8x8 ]= FUNCC(pred8x8_128_dc                 , depth);\
+    } else {\
+        h->pred8x8[DC_128_PRED8x8 ]= FUNCC(pred8x16_128_dc                , depth);\
+    }\
 \
     h->pred16x16[DC_PRED8x8     ]= FUNCC(pred16x16_dc                     , depth);\
     h->pred16x16[VERT_PRED8x8   ]= FUNCC(pred16x16_vertical               , depth);\
@@ -506,6 +529,6 @@ void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth){
             break;
     }
 
-    if (ARCH_ARM) ff_h264_pred_init_arm(h, codec_id, bit_depth);
-    if (HAVE_MMX) ff_h264_pred_init_x86(h, codec_id, bit_depth);
+    if (ARCH_ARM) ff_h264_pred_init_arm(h, codec_id, bit_depth, chroma_format_idc);
+    if (HAVE_MMX) ff_h264_pred_init_x86(h, codec_id, bit_depth, chroma_format_idc);
 }
diff --git a/libavcodec/h264pred.h b/libavcodec/h264pred.h
index daf123f968..599cdb228b 100644
--- a/libavcodec/h264pred.h
+++ b/libavcodec/h264pred.h
@@ -101,8 +101,8 @@ typedef struct H264PredContext{
     void (*pred16x16_add[3])(uint8_t *pix/*align 16*/, const int *block_offset, const DCTELEM *block/*align 16*/, int stride);
 }H264PredContext;
 
-void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth);
-void ff_h264_pred_init_arm(H264PredContext *h, int codec_id, const int bit_depth);
-void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth);
+void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth, const int chroma_format_idc);
+void ff_h264_pred_init_arm(H264PredContext *h, int codec_id, const int bit_depth, const int chroma_format_idc);
+void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth, const int chroma_format_idc);
 
 #endif /* AVCODEC_H264PRED_H */
diff --git a/libavcodec/h264pred_template.c b/libavcodec/h264pred_template.c
index 36f6d4e12f..eb5250501a 100644
--- a/libavcodec/h264pred_template.c
+++ b/libavcodec/h264pred_template.c
@@ -454,6 +454,19 @@ static void FUNCC(pred8x8_vertical)(uint8_t *_src, int _stride){
     }
 }
 
+static void FUNCC(pred8x16_vertical)(uint8_t *_src, int _stride){
+    int i;
+    pixel *src = (pixel*)_src;
+    int stride = _stride>>(sizeof(pixel)-1);
+    const pixel4 a= AV_RN4PA(((pixel4*)(src-stride))+0);
+    const pixel4 b= AV_RN4PA(((pixel4*)(src-stride))+1);
+
+    for(i=0; i<16; i++){
+        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
+        AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
+    }
+}
+
 static void FUNCC(pred8x8_horizontal)(uint8_t *_src, int stride){
     int i;
     pixel *src = (pixel*)_src;
@@ -466,6 +479,17 @@ static void FUNCC(pred8x8_horizontal)(uint8_t *_src, int stride){
     }
 }
 
+static void FUNCC(pred8x16_horizontal)(uint8_t *_src, int stride){
+    int i;
+    pixel *src = (pixel*)_src;
+    stride >>= sizeof(pixel)-1;
+    for(i=0; i<16; i++){
+        const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
+        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
+        AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
+    }
+}
+
 #define PRED8x8_X(n, v)\
 static void FUNCC(pred8x8_##n##_dc)(uint8_t *_src, int stride){\
     int i;\
@@ -482,6 +506,11 @@ PRED8x8_X(127, (1<<(BIT_DEPTH-1))-1);
 PRED8x8_X(128, (1<<(BIT_DEPTH-1))+0);
 PRED8x8_X(129, (1<<(BIT_DEPTH-1))+1);
 
+static void FUNCC(pred8x16_128_dc)(uint8_t *_src, int stride){
+    FUNCC(pred8x8_128_dc)(_src, stride);
+    FUNCC(pred8x8_128_dc)(_src+8*stride, stride);
+}
+
 static void FUNCC(pred8x8_left_dc)(uint8_t *_src, int stride){
     int i;
     int dc0, dc2;
@@ -507,6 +536,11 @@ static void FUNCC(pred8x8_left_dc)(uint8_t *_src, int stride){
     }
 }
 
+static void FUNCC(pred8x16_left_dc)(uint8_t *_src, int stride){
+    FUNCC(pred8x8_left_dc)(_src, stride);
+    FUNCC(pred8x8_left_dc)(_src+8*stride, stride);
+}
+
 static void FUNCC(pred8x8_top_dc)(uint8_t *_src, int stride){
     int i;
     int dc0, dc1;
@@ -532,6 +566,27 @@ static void FUNCC(pred8x8_top_dc)(uint8_t *_src, int stride){
     }
 }
 
+static void FUNCC(pred8x16_top_dc)(uint8_t *_src, int stride){
+    int i;
+    int dc0, dc1;
+    pixel4 dc0splat, dc1splat;
+    pixel *src = (pixel*)_src;
+    stride >>= sizeof(pixel)-1;
+
+    dc0=dc1=0;
+    for(i=0;i<4; i++){
+        dc0+= src[i-stride];
+        dc1+= src[4+i-stride];
+    }
+    dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
+    dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
+
+    for(i=0; i<16; i++){
+        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
+        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
+    }
+}
+
 static void FUNCC(pred8x8_dc)(uint8_t *_src, int stride){
     int i;
     int dc0, dc1, dc2;
@@ -560,6 +615,48 @@ static void FUNCC(pred8x8_dc)(uint8_t *_src, int stride){
     }
 }
 
+static void FUNCC(pred8x16_dc)(uint8_t *_src, int stride){
+    int i;
+    int dc0, dc1, dc2, dc3, dc4;
+    pixel4 dc0splat, dc1splat, dc2splat, dc3splat, dc4splat, dc5splat, dc6splat, dc7splat;
+    pixel *src = (pixel*)_src;
+    stride >>= sizeof(pixel)-1;
+
+    dc0=dc1=dc2=dc3=dc4=0;
+    for(i=0;i<4; i++){
+        dc0+= src[-1+i*stride] + src[i-stride];
+        dc1+= src[4+i-stride];
+        dc2+= src[-1+(i+4)*stride];
+        dc3+= src[-1+(i+8)*stride];
+        dc4+= src[-1+(i+12)*stride];
+    }
+    dc0splat = PIXEL_SPLAT_X4((dc0 + 4)>>3);
+    dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
+    dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
+    dc3splat = PIXEL_SPLAT_X4((dc1 + dc2 + 4)>>3);
+    dc4splat = PIXEL_SPLAT_X4((dc3 + 2)>>2);
+    dc5splat = PIXEL_SPLAT_X4((dc1 + dc3 + 4)>>3);
+    dc6splat = PIXEL_SPLAT_X4((dc4 + 2)>>2);
+    dc7splat = PIXEL_SPLAT_X4((dc1 + dc4 + 4)>>3);
+
+    for(i=0; i<4; i++){
+        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
+        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
+    }
+    for(i=4; i<8; i++){
+        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
+        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc3splat);
+    }
+    for(i=8; i<12; i++){
+        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc4splat);
+        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc5splat);
+    }
+    for(i=12; i<16; i++){
+        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc6splat);
+        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc7splat);
+    }
+}
+
 //the following 4 function should not be optimized!
 static void FUNC(pred8x8_mad_cow_dc_l0t)(uint8_t *src, int stride){
     FUNCC(pred8x8_top_dc)(src, stride);
@@ -618,6 +715,47 @@ static void FUNCC(pred8x8_plane)(uint8_t *_src, int _stride){
   }
 }
 
+static void FUNCC(pred8x16_plane)(uint8_t *_src, int _stride){
+  int j, k;
+  int a;
+  INIT_CLIP
+  pixel *src = (pixel*)_src;
+  int stride = _stride>>(sizeof(pixel)-1);
+  const pixel * const src0 = src +3-stride;
+  const pixel *       src1 = src +8*stride-1;
+  const pixel *       src2 = src1-2*stride;    // == src+6*stride-1;
+  int H = src0[1] - src0[-1];
+  int V = src1[0] - src2[ 0];
+
+  for (k = 2; k <= 4; ++k) {
+      src1 += stride; src2 -= stride;
+      H += k*(src0[k] - src0[-k]);
+      V += k*(src1[0] - src2[ 0]);
+  }
+  for (; k <= 8; ++k) {
+      src1 += stride; src2 -= stride;
+      V += k*(src1[0] - src2[0]);
+  }
+
+  H = (17*H+16) >> 5;
+  V = (5*V+32) >> 6;
+
+  a = 16*(src1[0] + src2[8] + 1) - 7*V - 3*H;
+  for(j=16; j>0; --j) {
+    int b = a;
+    a += V;
+    src[0] = CLIP((b    ) >> 5);
+    src[1] = CLIP((b+  H) >> 5);
+    src[2] = CLIP((b+2*H) >> 5);
+    src[3] = CLIP((b+3*H) >> 5);
+    src[4] = CLIP((b+4*H) >> 5);
+    src[5] = CLIP((b+5*H) >> 5);
+    src[6] = CLIP((b+6*H) >> 5);
+    src[7] = CLIP((b+7*H) >> 5);
+    src += stride;
+  }
+}
+
 #define SRC(x,y) src[(x)+(y)*stride]
 #define PL(y) \
     const int l##y = (SRC(-1,y-1) + 2*SRC(-1,y) + SRC(-1,y+1) + 2) >> 2;
diff --git a/libavcodec/ppc/h264_altivec.c b/libavcodec/ppc/h264_altivec.c
index 223971bd1a..00fb0a73d9 100644
--- a/libavcodec/ppc/h264_altivec.c
+++ b/libavcodec/ppc/h264_altivec.c
@@ -999,12 +999,13 @@ void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {
     }
 }
 
-void ff_h264dsp_init_ppc(H264DSPContext *c, const int bit_depth)
+void ff_h264dsp_init_ppc(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
 {
     if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) {
     if (bit_depth == 8) {
         c->h264_idct_add = ff_h264_idct_add_altivec;
-        c->h264_idct_add8 = ff_h264_idct_add8_altivec;
+        if (chroma_format_idc == 1)
+            c->h264_idct_add8 = ff_h264_idct_add8_altivec;
         c->h264_idct_add16 = ff_h264_idct_add16_altivec;
         c->h264_idct_add16intra = ff_h264_idct_add16intra_altivec;
         c->h264_idct_dc_add= h264_idct_dc_add_altivec;
diff --git a/libavcodec/rv34.c b/libavcodec/rv34.c
index 324cf717c2..d1df3461cc 100644
--- a/libavcodec/rv34.c
+++ b/libavcodec/rv34.c
@@ -1412,7 +1412,7 @@ av_cold int ff_rv34_decode_init(AVCodecContext *avctx)
     if (MPV_common_init(s) < 0)
         return -1;
 
-    ff_h264_pred_init(&r->h, CODEC_ID_RV40, 8);
+    ff_h264_pred_init(&r->h, CODEC_ID_RV40, 8, 1);
 
 #if CONFIG_RV30_DECODER
     if (avctx->codec_id == CODEC_ID_RV30)
diff --git a/libavcodec/vp8.c b/libavcodec/vp8.c
index fb3f7f733a..7cbf8b1434 100644
--- a/libavcodec/vp8.c
+++ b/libavcodec/vp8.c
@@ -1721,7 +1721,7 @@ static av_cold int vp8_decode_init(AVCodecContext *avctx)
     avctx->pix_fmt = PIX_FMT_YUV420P;
 
     dsputil_init(&s->dsp, avctx);
-    ff_h264_pred_init(&s->hpc, CODEC_ID_VP8, 8);
+    ff_h264_pred_init(&s->hpc, CODEC_ID_VP8, 8, 1);
     ff_vp8dsp_init(&s->vp8dsp);
 
     return 0;
diff --git a/libavcodec/x86/h264_intrapred_init.c b/libavcodec/x86/h264_intrapred_init.c
index 7220dd75c1..cdf7d55d96 100644
--- a/libavcodec/x86/h264_intrapred_init.c
+++ b/libavcodec/x86/h264_intrapred_init.c
@@ -167,7 +167,7 @@ void ff_pred4x4_tm_vp8_mmxext      (uint8_t *src, const uint8_t *topright, int s
 void ff_pred4x4_tm_vp8_ssse3       (uint8_t *src, const uint8_t *topright, int stride);
 void ff_pred4x4_vertical_vp8_mmxext(uint8_t *src, const uint8_t *topright, int stride);
 
-void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth)
+void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth, const int chroma_format_idc)
 {
     int mm_flags = av_get_cpu_flags();
 
@@ -176,14 +176,17 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
         if (mm_flags & AV_CPU_FLAG_MMX) {
             h->pred16x16[VERT_PRED8x8         ] = ff_pred16x16_vertical_mmx;
             h->pred16x16[HOR_PRED8x8          ] = ff_pred16x16_horizontal_mmx;
-            h->pred8x8  [VERT_PRED8x8         ] = ff_pred8x8_vertical_mmx;
-            h->pred8x8  [HOR_PRED8x8          ] = ff_pred8x8_horizontal_mmx;
+            if (chroma_format_idc == 1) {
+                h->pred8x8  [VERT_PRED8x8     ] = ff_pred8x8_vertical_mmx;
+                h->pred8x8  [HOR_PRED8x8      ] = ff_pred8x8_horizontal_mmx;
+            }
             if (codec_id == CODEC_ID_VP8) {
                 h->pred16x16[PLANE_PRED8x8    ] = ff_pred16x16_tm_vp8_mmx;
                 h->pred8x8  [PLANE_PRED8x8    ] = ff_pred8x8_tm_vp8_mmx;
                 h->pred4x4  [TM_VP8_PRED      ] = ff_pred4x4_tm_vp8_mmx;
             } else {
-                h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_plane_mmx;
+                if (chroma_format_idc == 1)
+                    h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_plane_mmx;
                 if (codec_id == CODEC_ID_SVQ3) {
                     h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_plane_svq3_mmx;
                 } else if (codec_id == CODEC_ID_RV40) {
@@ -197,7 +200,8 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
         if (mm_flags & AV_CPU_FLAG_MMX2) {
             h->pred16x16[HOR_PRED8x8            ] = ff_pred16x16_horizontal_mmxext;
             h->pred16x16[DC_PRED8x8             ] = ff_pred16x16_dc_mmxext;
-            h->pred8x8  [HOR_PRED8x8            ] = ff_pred8x8_horizontal_mmxext;
+            if (chroma_format_idc == 1)
+                h->pred8x8[HOR_PRED8x8          ] = ff_pred8x8_horizontal_mmxext;
             h->pred8x8l [TOP_DC_PRED            ] = ff_pred8x8l_top_dc_mmxext;
             h->pred8x8l [DC_PRED                ] = ff_pred8x8l_dc_mmxext;
             h->pred8x8l [HOR_PRED               ] = ff_pred8x8l_horizontal_mmxext;
@@ -221,8 +225,10 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
                 h->pred4x4  [HOR_UP_PRED        ] = ff_pred4x4_horizontal_up_mmxext;
             }
             if (codec_id == CODEC_ID_SVQ3 || codec_id == CODEC_ID_H264) {
-                h->pred8x8  [TOP_DC_PRED8x8     ] = ff_pred8x8_top_dc_mmxext;
-                h->pred8x8  [DC_PRED8x8         ] = ff_pred8x8_dc_mmxext;
+                if (chroma_format_idc == 1) {
+                    h->pred8x8[TOP_DC_PRED8x8   ] = ff_pred8x8_top_dc_mmxext;
+                    h->pred8x8[DC_PRED8x8       ] = ff_pred8x8_dc_mmxext;
+                }
             }
             if (codec_id == CODEC_ID_VP8) {
                 h->pred16x16[PLANE_PRED8x8      ] = ff_pred16x16_tm_vp8_mmxext;
@@ -231,7 +237,8 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
                 h->pred4x4  [TM_VP8_PRED        ] = ff_pred4x4_tm_vp8_mmxext;
                 h->pred4x4  [VERT_PRED          ] = ff_pred4x4_vertical_vp8_mmxext;
             } else {
-                h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_plane_mmx2;
+                if (chroma_format_idc == 1)
+                    h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_plane_mmx2;
                 if (codec_id == CODEC_ID_SVQ3) {
                     h->pred16x16[PLANE_PRED8x8  ] = ff_pred16x16_plane_svq3_mmx2;
                 } else if (codec_id == CODEC_ID_RV40) {
@@ -257,7 +264,8 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
                 h->pred16x16[PLANE_PRED8x8    ] = ff_pred16x16_tm_vp8_sse2;
                 h->pred8x8  [PLANE_PRED8x8    ] = ff_pred8x8_tm_vp8_sse2;
             } else {
-                h->pred8x8  [PLANE_PRED8x8    ] = ff_pred8x8_plane_sse2;
+                if (chroma_format_idc == 1)
+                    h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_plane_sse2;
                 if (codec_id == CODEC_ID_SVQ3) {
                     h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_plane_svq3_sse2;
                 } else if (codec_id == CODEC_ID_RV40) {
@@ -271,7 +279,8 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
         if (mm_flags & AV_CPU_FLAG_SSSE3) {
             h->pred16x16[HOR_PRED8x8          ] = ff_pred16x16_horizontal_ssse3;
             h->pred16x16[DC_PRED8x8           ] = ff_pred16x16_dc_ssse3;
-            h->pred8x8  [HOR_PRED8x8          ] = ff_pred8x8_horizontal_ssse3;
+            if (chroma_format_idc == 1)
+                h->pred8x8  [HOR_PRED8x8      ] = ff_pred8x8_horizontal_ssse3;
             h->pred8x8l [TOP_DC_PRED          ] = ff_pred8x8l_top_dc_ssse3;
             h->pred8x8l [DC_PRED              ] = ff_pred8x8l_dc_ssse3;
             h->pred8x8l [HOR_PRED             ] = ff_pred8x8l_horizontal_ssse3;
@@ -286,7 +295,8 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
                 h->pred8x8  [PLANE_PRED8x8    ] = ff_pred8x8_tm_vp8_ssse3;
                 h->pred4x4  [TM_VP8_PRED      ] = ff_pred4x4_tm_vp8_ssse3;
             } else {
-                h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_plane_ssse3;
+                if (chroma_format_idc == 1)
+                    h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_plane_ssse3;
                 if (codec_id == CODEC_ID_SVQ3) {
                     h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_plane_svq3_ssse3;
                 } else if (codec_id == CODEC_ID_RV40) {
@@ -301,7 +311,8 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
             h->pred4x4[DC_PRED             ] = ff_pred4x4_dc_10_mmxext;
             h->pred4x4[HOR_UP_PRED         ] = ff_pred4x4_horizontal_up_10_mmxext;
 
-            h->pred8x8[DC_PRED8x8          ] = ff_pred8x8_dc_10_mmxext;
+            if (chroma_format_idc == 1)
+                h->pred8x8[DC_PRED8x8      ] = ff_pred8x8_dc_10_mmxext;
 
             h->pred8x8l[DC_128_PRED        ] = ff_pred8x8l_128_dc_10_mmxext;
 
@@ -319,11 +330,13 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
             h->pred4x4[VERT_RIGHT_PRED     ] = ff_pred4x4_vertical_right_10_sse2;
             h->pred4x4[HOR_DOWN_PRED       ] = ff_pred4x4_horizontal_down_10_sse2;
 
-            h->pred8x8[DC_PRED8x8          ] = ff_pred8x8_dc_10_sse2;
-            h->pred8x8[TOP_DC_PRED8x8      ] = ff_pred8x8_top_dc_10_sse2;
-            h->pred8x8[PLANE_PRED8x8       ] = ff_pred8x8_plane_10_sse2;
-            h->pred8x8[VERT_PRED8x8        ] = ff_pred8x8_vertical_10_sse2;
-            h->pred8x8[HOR_PRED8x8         ] = ff_pred8x8_horizontal_10_sse2;
+            if (chroma_format_idc == 1) {
+                h->pred8x8[DC_PRED8x8      ] = ff_pred8x8_dc_10_sse2;
+                h->pred8x8[TOP_DC_PRED8x8  ] = ff_pred8x8_top_dc_10_sse2;
+                h->pred8x8[PLANE_PRED8x8   ] = ff_pred8x8_plane_10_sse2;
+                h->pred8x8[VERT_PRED8x8    ] = ff_pred8x8_vertical_10_sse2;
+                h->pred8x8[HOR_PRED8x8     ] = ff_pred8x8_horizontal_10_sse2;
+            }
 
             h->pred8x8l[VERT_PRED           ] = ff_pred8x8l_vertical_10_sse2;
             h->pred8x8l[HOR_PRED            ] = ff_pred8x8l_horizontal_10_sse2;
diff --git a/libavcodec/x86/h264dsp_mmx.c b/libavcodec/x86/h264dsp_mmx.c
index 68e543681f..71beb262c9 100644
--- a/libavcodec/x86/h264dsp_mmx.c
+++ b/libavcodec/x86/h264dsp_mmx.c
@@ -350,7 +350,7 @@ H264_BIWEIGHT_10_SSE( 4,  8, 10)
 H264_BIWEIGHT_10_SSE( 4,  4, 10)
 H264_BIWEIGHT_10_SSE( 4,  2, 10)
 
-void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
+void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
 {
     int mm_flags = av_get_cpu_flags();
 
@@ -368,7 +368,8 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
 
         c->h264_idct_add16          = ff_h264_idct_add16_8_mmx;
         c->h264_idct8_add4          = ff_h264_idct8_add4_8_mmx;
-        c->h264_idct_add8           = ff_h264_idct_add8_8_mmx;
+        if (chroma_format_idc == 1)
+            c->h264_idct_add8       = ff_h264_idct_add8_8_mmx;
         c->h264_idct_add16intra     = ff_h264_idct_add16intra_8_mmx;
         c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_mmx;
 
@@ -377,13 +378,16 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
             c->h264_idct8_dc_add   = ff_h264_idct8_dc_add_8_mmx2;
             c->h264_idct_add16     = ff_h264_idct_add16_8_mmx2;
             c->h264_idct8_add4     = ff_h264_idct8_add4_8_mmx2;
-            c->h264_idct_add8      = ff_h264_idct_add8_8_mmx2;
+            if (chroma_format_idc == 1)
+                c->h264_idct_add8  = ff_h264_idct_add8_8_mmx2;
             c->h264_idct_add16intra= ff_h264_idct_add16intra_8_mmx2;
 
             c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_8_mmxext;
-            c->h264_h_loop_filter_chroma= ff_deblock_h_chroma_8_mmxext;
             c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_8_mmxext;
-            c->h264_h_loop_filter_chroma_intra= ff_deblock_h_chroma_intra_8_mmxext;
+            if (chroma_format_idc == 1) {
+                c->h264_h_loop_filter_chroma= ff_deblock_h_chroma_8_mmxext;
+                c->h264_h_loop_filter_chroma_intra= ff_deblock_h_chroma_intra_8_mmxext;
+            }
 #if ARCH_X86_32
             c->h264_v_loop_filter_luma= ff_deblock_v_luma_8_mmxext;
             c->h264_h_loop_filter_luma= ff_deblock_h_luma_8_mmxext;
@@ -413,7 +417,8 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
 
                 c->h264_idct_add16          = ff_h264_idct_add16_8_sse2;
                 c->h264_idct8_add4          = ff_h264_idct8_add4_8_sse2;
-                c->h264_idct_add8           = ff_h264_idct_add8_8_sse2;
+                if (chroma_format_idc == 1)
+                    c->h264_idct_add8       = ff_h264_idct_add8_8_sse2;
                 c->h264_idct_add16intra     = ff_h264_idct_add16intra_8_sse2;
                 c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_sse2;
 
@@ -472,7 +477,8 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
                 c->h264_idct8_dc_add   = ff_h264_idct8_dc_add_10_sse2;
 
                 c->h264_idct_add16     = ff_h264_idct_add16_10_sse2;
-                c->h264_idct_add8      = ff_h264_idct_add8_10_sse2;
+                if (chroma_format_idc == 1)
+                    c->h264_idct_add8  = ff_h264_idct_add8_10_sse2;
                 c->h264_idct_add16intra= ff_h264_idct_add16intra_10_sse2;
 #if HAVE_ALIGNED_STACK
                 c->h264_idct8_add      = ff_h264_idct8_add_10_sse2;
@@ -532,7 +538,8 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
                 c->h264_idct8_dc_add   = ff_h264_idct8_dc_add_10_avx;
 
                 c->h264_idct_add16     = ff_h264_idct_add16_10_avx;
-                c->h264_idct_add8      = ff_h264_idct_add8_10_avx;
+                if (chroma_format_idc == 1)
+                    c->h264_idct_add8  = ff_h264_idct_add8_10_avx;
                 c->h264_idct_add16intra= ff_h264_idct_add16intra_10_avx;
 #if HAVE_ALIGNED_STACK
                 c->h264_idct8_add      = ff_h264_idct8_add_10_avx;