Merge remote-tracking branch 'qatar/master'

* qatar/master:
  swscale: remove misplaced comment.
  ffmpeg: fix streaming to ffserver.
  swscale: split out RGB48 output functions from yuv2packed[12X]_c().
  build: move vpath directives to main Makefile
  swscale: fix JPEG-range YUV scaling artifacts.
  build: move ALLFFLIBS to a more logical place
  ARM: factor some repetitive code into macros
  Fix SVQ3 after adding 4:4:4 H.264 support
  H.264: fix CODEC_FLAG_GRAY
  4:4:4 H.264 decoding support
  ac3enc: fix allocation of floating point samples.

Conflicts:
	ffmpeg.c
	libavcodec/dsputil_template.c
	libavcodec/h264.c
	libavcodec/mpegvideo.c
	libavcodec/snow.c
	libswscale/swscale.c
	libswscale/swscale_internal.h

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

Makefile

@@ -2,6 +2,11 @@ include config.mak
 
 SRC_DIR = $(SRC_PATH_BARE)
 
+vpath %.c   $(SRC_DIR)
+vpath %.h   $(SRC_DIR)
+vpath %.S   $(SRC_DIR)
+vpath %.asm $(SRC_DIR)
+vpath %.v   $(SRC_DIR)
 vpath %.texi $(SRC_PATH_BARE)
 
 PROGS-$(CONFIG_FFMPEG)   += ffmpeg
@@ -24,6 +29,8 @@ ALLPROGS    = $(BASENAMES:%=%$(EXESUF))
 ALLPROGS_G  = $(BASENAMES:%=%_g$(EXESUF))
 ALLMANPAGES = $(BASENAMES:%=%.1)
 
+ALLFFLIBS = avcodec avdevice avfilter avformat avutil postproc swscale
+
 FFLIBS-$(CONFIG_AVDEVICE) += avdevice
 FFLIBS-$(CONFIG_AVFILTER) += avfilter
 FFLIBS-$(CONFIG_AVFORMAT) += avformat

common.mak

@@ -6,11 +6,6 @@
 all: all-yes
 
 ifndef SUBDIR
-vpath %.c   $(SRC_DIR)
-vpath %.h   $(SRC_DIR)
-vpath %.S   $(SRC_DIR)
-vpath %.asm $(SRC_DIR)
-vpath %.v   $(SRC_DIR)
 
 ifndef V
 Q      = @
@@ -25,8 +20,6 @@ $(foreach VAR,$(SILENT),$(eval override $(VAR) = @$($(VAR))))
 $(eval INSTALL = @$(call ECHO,INSTALL,$$(^:$(SRC_DIR)/%=%)); $(INSTALL))
 endif
 
-ALLFFLIBS = avcodec avdevice avfilter avformat avutil postproc swscale
-
 IFLAGS   := -I. -I$(SRC_PATH)
 CPPFLAGS := $(IFLAGS) $(CPPFLAGS)
 CFLAGS   += $(ECFLAGS)

libavcodec/ac3enc.c

@@ -2215,15 +2215,9 @@ static av_cold int allocate_buffers(AVCodecContext *avctx)
     AC3EncodeContext *s = avctx->priv_data;
     int channels = s->channels + 1; /* includes coupling channel */
 
-    FF_ALLOC_OR_GOTO(avctx, s->windowed_samples, AC3_WINDOW_SIZE *
-                     sizeof(*s->windowed_samples), alloc_fail);
-    FF_ALLOC_OR_GOTO(avctx, s->planar_samples, s->channels * sizeof(*s->planar_samples),
-                     alloc_fail);
-    for (ch = 0; ch < s->channels; ch++) {
-        FF_ALLOCZ_OR_GOTO(avctx, s->planar_samples[ch],
-                          (AC3_FRAME_SIZE+AC3_BLOCK_SIZE) * sizeof(**s->planar_samples),
-                          alloc_fail);
-    }
+    if (s->allocate_sample_buffers(s))
+        goto alloc_fail;
+
     FF_ALLOC_OR_GOTO(avctx, s->bap_buffer,  AC3_MAX_BLOCKS * channels *
                      AC3_MAX_COEFS * sizeof(*s->bap_buffer),  alloc_fail);
     FF_ALLOC_OR_GOTO(avctx, s->bap1_buffer, AC3_MAX_BLOCKS * channels *
@@ -2323,6 +2317,8 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx)
     AC3EncodeContext *s = avctx->priv_data;
     int ret, frame_size_58;
 
+    s->avctx = avctx;
+
     s->eac3 = avctx->codec_id == CODEC_ID_EAC3;
 
     avctx->frame_size = AC3_FRAME_SIZE;
@@ -2355,6 +2351,7 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx)
         s->apply_window                 = ff_ac3_fixed_apply_window;
         s->normalize_samples            = ff_ac3_fixed_normalize_samples;
         s->scale_coefficients           = ff_ac3_fixed_scale_coefficients;
+        s->allocate_sample_buffers      = ff_ac3_fixed_allocate_sample_buffers;
         s->deinterleave_input_samples   = ff_ac3_fixed_deinterleave_input_samples;
         s->apply_mdct                   = ff_ac3_fixed_apply_mdct;
         s->apply_channel_coupling       = ff_ac3_fixed_apply_channel_coupling;
@@ -2364,6 +2361,7 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx)
         s->mdct_init                    = ff_ac3_float_mdct_init;
         s->apply_window                 = ff_ac3_float_apply_window;
         s->scale_coefficients           = ff_ac3_float_scale_coefficients;
+        s->allocate_sample_buffers      = ff_ac3_float_allocate_sample_buffers;
         s->deinterleave_input_samples   = ff_ac3_float_deinterleave_input_samples;
         s->apply_mdct                   = ff_ac3_float_apply_mdct;
         s->apply_channel_coupling       = ff_ac3_float_apply_channel_coupling;

libavcodec/ac3enc.h

@@ -135,6 +135,7 @@ typedef struct AC3Block {
 typedef struct AC3EncodeContext {
     AVClass *av_class;                      ///< AVClass used for AVOption
     AC3EncOptions options;                  ///< encoding options
+    AVCodecContext *avctx;                  ///< parent AVCodecContext
     PutBitContext pb;                       ///< bitstream writer context
     DSPContext dsp;
     AC3DSPContext ac3dsp;                   ///< AC-3 optimized functions
@@ -230,6 +231,7 @@ typedef struct AC3EncodeContext {
     void (*scale_coefficients)(struct AC3EncodeContext *s);
 
     /* fixed vs. float templated function pointers */
+    int  (*allocate_sample_buffers)(struct AC3EncodeContext *s);
     void (*deinterleave_input_samples)(struct AC3EncodeContext *s,
                                        const SampleType *samples);
     void (*apply_mdct)(struct AC3EncodeContext *s);
@@ -276,6 +278,9 @@ void ff_ac3_float_scale_coefficients(AC3EncodeContext *s);
 
 /* prototypes for functions in ac3enc_template.c */
 
+int ff_ac3_fixed_allocate_sample_buffers(AC3EncodeContext *s);
+int ff_ac3_float_allocate_sample_buffers(AC3EncodeContext *s);
+
 void ff_ac3_fixed_deinterleave_input_samples(AC3EncodeContext *s,
                                              const SampleType *samples);
 void ff_ac3_float_deinterleave_input_samples(AC3EncodeContext *s,

libavcodec/ac3enc_template.c

@@ -31,6 +31,26 @@
 #include "ac3enc.h"
 
 
+int AC3_NAME(allocate_sample_buffers)(AC3EncodeContext *s)
+{
+    int ch;
+
+    FF_ALLOC_OR_GOTO(s->avctx, s->windowed_samples, AC3_WINDOW_SIZE *
+                     sizeof(*s->windowed_samples), alloc_fail);
+    FF_ALLOC_OR_GOTO(s->avctx, s->planar_samples, s->channels * sizeof(*s->planar_samples),
+                     alloc_fail);
+    for (ch = 0; ch < s->channels; ch++) {
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->planar_samples[ch],
+                          (AC3_FRAME_SIZE+AC3_BLOCK_SIZE) * sizeof(**s->planar_samples),
+                          alloc_fail);
+    }
+
+    return 0;
+alloc_fail:
+    return AVERROR(ENOMEM);
+}
+
+
 /**
  * Deinterleave input samples.
  * Channels are reordered from Libav's default order to AC-3 order.

libavcodec/arm/h264dsp_init_arm.c

@@ -122,7 +122,8 @@ 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;
+    //FIXME: reenable when asm is updated.
+    //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;

libavcodec/arm/mpegvideo_armv5te_s.S

@@ -35,6 +35,21 @@
  *
  * Inner loop should take 6 cycles per element on arm926ej-s (Nokia 770)
  */
+
+.macro  dequant_t       dst, src, mul, add, tmp
+        rsbs            \tmp, ip, \src, asr #16
+        addgt           \tmp, \add, #0
+        rsblt           \tmp, \add, #0
+        smlatbne        \dst, \src, \mul, \tmp
+.endm
+
+.macro  dequant_b       dst, src, mul, add, tmp
+        rsbs            \tmp, ip, \src, lsl #16
+        addgt           \tmp, \add, #0
+        rsblt           \tmp, \add, #0
+        smlabbne        \dst, \src, \mul, \tmp
+.endm
+
 function ff_dct_unquantize_h263_armv5te, export=1
         push            {r4-r9,lr}
         mov             ip, #0
@@ -44,50 +59,20 @@ function ff_dct_unquantize_h263_armv5te, export=1
 1:
         ldrd            r6, [r0, #8]
 
-        rsbs            r9, ip, r4, asr #16
-        addgt           r9, r2, #0
-        rsblt           r9, r2, #0
-        smlatbne        r9, r4, r1, r9
-
-        rsbs            lr, ip, r5, asr #16
-        addgt           lr, r2, #0
-        rsblt           lr, r2, #0
-        smlatbne        lr, r5, r1, lr
-
-        rsbs            r8, ip, r4, asl #16
-        addgt           r8, r2, #0
-        rsblt           r8, r2, #0
-        smlabbne        r4, r4, r1, r8
-
-        rsbs            r8, ip, r5, asl #16
-        addgt           r8, r2, #0
-        rsblt           r8, r2, #0
-        smlabbne        r5, r5, r1, r8
+        dequant_t       r9, r4, r1, r2, r9
+        dequant_t       lr, r5, r1, r2, lr
+        dequant_b       r4, r4, r1, r2, r8
+        dequant_b       r5, r5, r1, r2, r8
 
         strh            r4, [r0], #2
         strh            r9, [r0], #2
         strh            r5, [r0], #2
         strh            lr, [r0], #2
 
-        rsbs            r9, ip, r6, asr #16
-        addgt           r9, r2, #0
-        rsblt           r9, r2, #0
-        smlatbne        r9, r6, r1, r9
-
-        rsbs            lr, ip, r7, asr #16
-        addgt           lr, r2, #0
-        rsblt           lr, r2, #0
-        smlatbne        lr, r7, r1, lr
-
-        rsbs            r8, ip, r6, asl #16
-        addgt           r8, r2, #0
-        rsblt           r8, r2, #0
-        smlabbne        r6, r6, r1, r8
-
-        rsbs            r8, ip, r7, asl #16
-        addgt           r8, r2, #0
-        rsblt           r8, r2, #0
-        smlabbne        r7, r7, r1, r8
+        dequant_t       r9, r6, r1, r2, r9
+        dequant_t       lr, r7, r1, r2, lr
+        dequant_b       r6, r6, r1, r2, r8
+        dequant_b       r7, r7, r1, r2, r8
 
         strh            r6, [r0], #2
         strh            r9, [r0], #2

libavcodec/arm/simple_idct_armv5te.S

@@ -333,6 +333,20 @@ function idct_col_armv5te
         ldr    pc, [sp], #4
 endfunc
 
+.macro  clip   dst, src:vararg
+        movs   \dst, \src
+        movmi  \dst, #0
+        cmp    \dst, #255
+        movgt  \dst, #255
+.endm
+
+.macro  aclip  dst, src:vararg
+        adds   \dst, \src
+        movmi  \dst, #0
+        cmp    \dst, #255
+        movgt  \dst, #255
+.endm
+
 function idct_col_put_armv5te
         str    lr, [sp, #-4]!
 
@@ -341,27 +355,15 @@ function idct_col_put_armv5te
         ldmfd  sp!, {a3, a4}
         ldr    lr, [sp, #32]
         add    a2, a3, v1
-        movs   a2, a2, asr #20
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        clip   a2, a2, asr #20
         add    ip, a4, v2
-        movs   ip, ip, asr #20
-        movmi  ip, #0
-        cmp    ip, #255
-        movgt  ip, #255
+        clip   ip, ip, asr #20
         orr    a2, a2, ip, lsl #8
         sub    a3, a3, v1
-        movs   a3, a3, asr #20
-        movmi  a3, #0
-        cmp    a3, #255
-        movgt  a3, #255
+        clip   a3, a3, asr #20
         sub    a4, a4, v2
-        movs   a4, a4, asr #20
-        movmi  a4, #0
-        cmp    a4, #255
+        clip   a4, a4, asr #20
         ldr    v1, [sp, #28]
-        movgt  a4, #255
         strh   a2, [v1]
         add    a2, v1, #2
         str    a2, [sp, #28]
@@ -371,79 +373,43 @@ function idct_col_put_armv5te
         strh   a2, [v2, v1]!
 
         sub    a2, a3, v3
-        movs   a2, a2, asr #20
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        clip   a2, a2, asr #20
         sub    ip, a4, v4
-        movs   ip, ip, asr #20
-        movmi  ip, #0
-        cmp    ip, #255
-        movgt  ip, #255
+        clip   ip, ip, asr #20
         orr    a2, a2, ip, lsl #8
         strh   a2, [v1, lr]!
         add    a3, a3, v3
-        movs   a2, a3, asr #20
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        clip   a2, a3, asr #20
         add    a4, a4, v4
-        movs   a4, a4, asr #20
-        movmi  a4, #0
-        cmp    a4, #255
-        movgt  a4, #255
+        clip   a4, a4, asr #20
         orr    a2, a2, a4, lsl #8
         ldmfd  sp!, {a3, a4}
         strh   a2, [v2, -lr]!
 
         add    a2, a3, v5
-        movs   a2, a2, asr #20
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        clip   a2, a2, asr #20
         add    ip, a4, v6
-        movs   ip, ip, asr #20
-        movmi  ip, #0
-        cmp    ip, #255
-        movgt  ip, #255
+        clip   ip, ip, asr #20
         orr    a2, a2, ip, lsl #8
         strh   a2, [v1, lr]!
         sub    a3, a3, v5
-        movs   a2, a3, asr #20
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        clip   a2, a3, asr #20
         sub    a4, a4, v6
-        movs   a4, a4, asr #20
-        movmi  a4, #0
-        cmp    a4, #255
-        movgt  a4, #255
+        clip   a4, a4, asr #20
         orr    a2, a2, a4, lsl #8
         ldmfd  sp!, {a3, a4}
         strh   a2, [v2, -lr]!
 
         add    a2, a3, v7
-        movs   a2, a2, asr #20
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        clip   a2, a2, asr #20
         add    ip, a4, fp
-        movs   ip, ip, asr #20
-        movmi  ip, #0
-        cmp    ip, #255
-        movgt  ip, #255
+        clip   ip, ip, asr #20
         orr    a2, a2, ip, lsl #8
         strh   a2, [v1, lr]
         sub    a3, a3, v7
-        movs   a2, a3, asr #20
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        clip   a2, a3, asr #20
         sub    a4, a4, fp
-        movs   a4, a4, asr #20
-        movmi  a4, #0
-        cmp    a4, #255
-        movgt  a4, #255
+        clip   a4, a4, asr #20
         orr    a2, a2, a4, lsl #8
         strh   a2, [v2, -lr]
 
@@ -460,36 +426,22 @@ function idct_col_add_armv5te
         ldmfd  sp!, {a3, a4}
         ldrh   ip, [lr]
         add    a2, a3, v1
-        mov    a2, a2, asr #20
         sub    a3, a3, v1
         and    v1, ip, #255
-        adds   a2, a2, v1
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        aclip  a2, v1, a2, asr #20
         add    v1, a4, v2
         mov    v1, v1, asr #20
-        adds   v1, v1, ip, lsr #8
-        movmi  v1, #0
-        cmp    v1, #255
-        movgt  v1, #255
+        aclip  v1, v1, ip, lsr #8
         orr    a2, a2, v1, lsl #8
         ldr    v1, [sp, #32]
         sub    a4, a4, v2
         rsb    v2, v1, v1, lsl #3
         ldrh   ip, [v2, lr]!
         strh   a2, [lr]
-        mov    a3, a3, asr #20
         and    a2, ip, #255
-        adds   a3, a3, a2
-        movmi  a3, #0
-        cmp    a3, #255
-        movgt  a3, #255
+        aclip  a3, a2, a3, asr #20
         mov    a4, a4, asr #20
-        adds   a4, a4, ip, lsr #8
-        movmi  a4, #0
-        cmp    a4, #255
-        movgt  a4, #255
+        aclip  a4, a4, ip, lsr #8
         add    a2, lr, #2
         str    a2, [sp, #28]
         orr    a2, a3, a4, lsl #8
@@ -498,102 +450,60 @@ function idct_col_add_armv5te
         ldmfd  sp!, {a3, a4}
         ldrh   ip, [lr, v1]!
         sub    a2, a3, v3
-        mov    a2, a2, asr #20
         add    a3, a3, v3
         and    v3, ip, #255
-        adds   a2, a2, v3
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        aclip  a2, v3, a2, asr #20
         sub    v3, a4, v4
         mov    v3, v3, asr #20
-        adds   v3, v3, ip, lsr #8
-        movmi  v3, #0
-        cmp    v3, #255
-        movgt  v3, #255
+        aclip  v3, v3, ip, lsr #8
         orr    a2, a2, v3, lsl #8
         add    a4, a4, v4
         ldrh   ip, [v2, -v1]!
         strh   a2, [lr]
-        mov    a3, a3, asr #20
         and    a2, ip, #255
-        adds   a3, a3, a2
-        movmi  a3, #0
-        cmp    a3, #255
-        movgt  a3, #255
+        aclip  a3, a2, a3, asr #20
         mov    a4, a4, asr #20
-        adds   a4, a4, ip, lsr #8
-        movmi  a4, #0
-        cmp    a4, #255
-        movgt  a4, #255
+        aclip  a4, a4, ip, lsr #8
         orr    a2, a3, a4, lsl #8
         strh   a2, [v2]
 
         ldmfd  sp!, {a3, a4}
         ldrh   ip, [lr, v1]!
         add    a2, a3, v5
-        mov    a2, a2, asr #20
         sub    a3, a3, v5
         and    v3, ip, #255
-        adds   a2, a2, v3
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        aclip  a2, v3, a2, asr #20
         add    v3, a4, v6
         mov    v3, v3, asr #20
-        adds   v3, v3, ip, lsr #8
-        movmi  v3, #0
-        cmp    v3, #255
-        movgt  v3, #255
+        aclip  v3, v3, ip, lsr #8
         orr    a2, a2, v3, lsl #8
         sub    a4, a4, v6
         ldrh   ip, [v2, -v1]!
         strh   a2, [lr]
-        mov    a3, a3, asr #20
         and    a2, ip, #255
-        adds   a3, a3, a2
-        movmi  a3, #0
-        cmp    a3, #255
-        movgt  a3, #255
+        aclip  a3, a2, a3, asr #20
         mov    a4, a4, asr #20
-        adds   a4, a4, ip, lsr #8
-        movmi  a4, #0
-        cmp    a4, #255
-        movgt  a4, #255
+        aclip  a4, a4, ip, lsr #8
         orr    a2, a3, a4, lsl #8
         strh   a2, [v2]
 
         ldmfd  sp!, {a3, a4}
         ldrh   ip, [lr, v1]!
         add    a2, a3, v7
-        mov    a2, a2, asr #20
         sub    a3, a3, v7
         and    v3, ip, #255
-        adds   a2, a2, v3
-        movmi  a2, #0
-        cmp    a2, #255
-        movgt  a2, #255
+        aclip  a2, v3, a2, asr #20
         add    v3, a4, fp
         mov    v3, v3, asr #20
-        adds   v3, v3, ip, lsr #8
-        movmi  v3, #0
-        cmp    v3, #255
-        movgt  v3, #255
+        aclip  v3, v3, ip, lsr #8
         orr    a2, a2, v3, lsl #8
         sub    a4, a4, fp
         ldrh   ip, [v2, -v1]!
         strh   a2, [lr]
-        mov    a3, a3, asr #20
         and    a2, ip, #255
-        adds   a3, a3, a2
-        movmi  a3, #0
-        cmp    a3, #255
-        movgt  a3, #255
+        aclip  a3, a2, a3, asr #20
         mov    a4, a4, asr #20
-        adds   a4, a4, ip, lsr #8
-        movmi  a4, #0
-        cmp    a4, #255
-        movgt  a4, #255
+        aclip  a4, a4, ip, lsr #8
         orr    a2, a3, a4, lsl #8
         strh   a2, [v2]
 

libavcodec/dsputil.h

@@ -505,7 +505,7 @@ typedef struct DSPContext {
 #define BASIS_SHIFT 16
 #define RECON_SHIFT 6
 
-    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int sides);
+    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
 #define EDGE_WIDTH 16
 #define EDGE_TOP    1
 #define EDGE_BOTTOM 2

libavcodec/dsputil_template.c

@@ -79,7 +79,7 @@ static inline void FUNC(copy_block16)(uint8_t *dst, const uint8_t *src, int dstS
 
 /* draw the edges of width 'w' of an image of size width, height */
 //FIXME check that this is ok for mpeg4 interlaced
-static void FUNCC(draw_edges)(uint8_t *p_buf, int p_wrap, int width, int height, int w, int sides)
+static void FUNCC(draw_edges)(uint8_t *p_buf, int p_wrap, int width, int height, int w, int h, int sides)
 {
     pixel *buf = (pixel*)p_buf;
     int wrap = p_wrap / sizeof(pixel);
@@ -106,10 +106,10 @@ static void FUNCC(draw_edges)(uint8_t *p_buf, int p_wrap, int width, int height,
     buf -= w;
     last_line = buf + (height - 1) * wrap;
     if (sides & EDGE_TOP)
-        for(i = 0; i < w; i++)
+        for(i = 0; i < h; i++)
             memcpy(buf - (i + 1) * wrap, buf, (width + w + w) * sizeof(pixel)); // top
     if (sides & EDGE_BOTTOM)
-        for (i = 0; i < w; i++)
+        for (i = 0; i < h; i++)
             memcpy(last_line + (i + 1) * wrap, last_line, (width + w + w) * sizeof(pixel)); // bottom
 }
 

libavcodec/h264.h

@@ -39,9 +39,6 @@
 #define interlaced_dct interlaced_dct_is_a_bad_name
 #define mb_intra mb_intra_is_not_initialized_see_mb_type
 
-#define LUMA_DC_BLOCK_INDEX   24
-#define CHROMA_DC_BLOCK_INDEX 25
-
 #define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
 #define COEFF_TOKEN_VLC_BITS           8
 #define TOTAL_ZEROS_VLC_BITS           9
@@ -60,8 +57,6 @@
  * of progressive decoding by about 2%. */
 #define ALLOW_INTERLACE
 
-#define ALLOW_NOCHROMA
-
 #define FMO 0
 
 /**
@@ -85,16 +80,12 @@
 #endif
 #define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
 
-#ifdef ALLOW_NOCHROMA
-#define CHROMA h->sps.chroma_format_idc
-#else
-#define CHROMA 1
-#endif
-
 #ifndef CABAC
 #define CABAC h->pps.cabac
 #endif
 
+#define CHROMA444 (h->sps.chroma_format_idc == 3)
+
 #define EXTENDED_SAR          255
 
 #define MB_TYPE_REF0       MB_TYPE_ACPRED //dirty but it fits in 16 bit
@@ -198,7 +189,7 @@ typedef struct SPS{
     int num_reorder_frames;
     int scaling_matrix_present;
     uint8_t scaling_matrix4[6][16];
-    uint8_t scaling_matrix8[2][64];
+    uint8_t scaling_matrix8[6][64];
     int nal_hrd_parameters_present_flag;
     int vcl_hrd_parameters_present_flag;
     int pic_struct_present_flag;
@@ -233,7 +224,7 @@ typedef struct PPS{
     int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
     int transform_8x8_mode;     ///< transform_8x8_mode_flag
     uint8_t scaling_matrix4[6][16];
-    uint8_t scaling_matrix8[2][64];
+    uint8_t scaling_matrix8[6][64];
     uint8_t chroma_qp_table[2][64];  ///< pre-scaled (with chroma_qp_index_offset) version of qp_table
     int chroma_qp_diff;
 }PPS;
@@ -298,21 +289,15 @@ typedef struct H264Context{
     unsigned int top_samples_available;
     unsigned int topright_samples_available;
     unsigned int left_samples_available;
-    uint8_t (*top_borders[2])[(16+2*8)*2];
+    uint8_t (*top_borders[2])[(16*3)*2];
 
     /**
      * non zero coeff count cache.
      * is 64 if not available.
      */
-    DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[6*8];
+    DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15*8];
 
-    /*
-    .UU.YYYY
-    .UU.YYYY
-    .vv.YYYY
-    .VV.YYYY
-    */
-    uint8_t (*non_zero_count)[32];
+    uint8_t (*non_zero_count)[48];
 
     /**
      * Motion vector cache.
@@ -336,7 +321,7 @@ typedef struct H264Context{
      * block_offset[ 0..23] for frame macroblocks
      * block_offset[24..47] for field macroblocks
      */
-    int block_offset[2*(16+8)];
+    int block_offset[2*(16*3)];
 
     uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
     uint32_t *mb2br_xy;
@@ -356,9 +341,9 @@ typedef struct H264Context{
     PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
 
     uint32_t dequant4_buffer[6][QP_MAX_NUM+1][16]; //FIXME should these be moved down?
-    uint32_t dequant8_buffer[2][QP_MAX_NUM+1][64];
+    uint32_t dequant8_buffer[6][QP_MAX_NUM+1][64];
     uint32_t (*dequant4_coeff[6])[16];
-    uint32_t (*dequant8_coeff[2])[64];
+    uint32_t (*dequant8_coeff[6])[64];
 
     int slice_num;
     uint16_t *slice_table;     ///< slice_table_base + 2*mb_stride + 1
@@ -408,15 +393,15 @@ typedef struct H264Context{
     GetBitContext *intra_gb_ptr;
     GetBitContext *inter_gb_ptr;
 
-    DECLARE_ALIGNED(16, DCTELEM, mb)[16*24*2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
-    DECLARE_ALIGNED(16, DCTELEM, mb_luma_dc)[16*2];
+    DECLARE_ALIGNED(16, DCTELEM, mb)[16*48*2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
+    DECLARE_ALIGNED(16, DCTELEM, mb_luma_dc)[3][16*2];
     DCTELEM mb_padding[256*2];        ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either check that i is not too large or ensure that there is some unused stuff after mb
 
     /**
      * Cabac
      */
     CABACContext cabac;
-    uint8_t      cabac_state[460];
+    uint8_t      cabac_state[1024];
 
     /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
     uint16_t     *cbp_table;
@@ -721,27 +706,43 @@ o-o o-o
 */
 
 /* Scan8 organization:
- *   0 1 2 3 4 5 6 7
- * 0   u u y y y y y
- * 1 u U U y Y Y Y Y
- * 2 u U U y Y Y Y Y
- * 3   v v y Y Y Y Y
- * 4 v V V y Y Y Y Y
- * 5 v V V   DYDUDV
+ *    0 1 2 3 4 5 6 7
+ * 0  DY    y y y y y
+ * 1        y Y Y Y Y
+ * 2        y Y Y Y Y
+ * 3        y Y Y Y Y
+ * 4        y Y Y Y Y
+ * 5  DU    u u u u u
+ * 6        u U U U U
+ * 7        u U U U U
+ * 8        u U U U U
+ * 9        u U U U U
+ * 10 DV    v v v v v
+ * 11       v V V V V
+ * 12       v V V V V
+ * 13       v V V V V
+ * 14       v V V V V
  * DY/DU/DV are for luma/chroma DC.
  */
 
+#define LUMA_DC_BLOCK_INDEX   48
+#define CHROMA_DC_BLOCK_INDEX 49
+
 //This table must be here because scan8[constant] must be known at compiletime
-static const uint8_t scan8[16 + 2*4 + 3]={
- 4+1*8, 5+1*8, 4+2*8, 5+2*8,
- 6+1*8, 7+1*8, 6+2*8, 7+2*8,
- 4+3*8, 5+3*8, 4+4*8, 5+4*8,
- 6+3*8, 7+3*8, 6+4*8, 7+4*8,
- 1+1*8, 2+1*8,
- 1+2*8, 2+2*8,
- 1+4*8, 2+4*8,
- 1+5*8, 2+5*8,
- 4+5*8, 5+5*8, 6+5*8
+static const uint8_t scan8[16*3 + 3]={
+ 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8,
+ 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8,
+ 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8,
+ 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8,
+ 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8,
+ 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8,
+ 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8,
+ 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8,
+ 4+11*8, 5+11*8, 4+12*8, 5+12*8,
+ 6+11*8, 7+11*8, 6+12*8, 7+12*8,
+ 4+13*8, 5+13*8, 4+14*8, 5+14*8,
+ 6+13*8, 7+13*8, 6+14*8, 7+14*8,
+ 0+ 0*8, 0+ 5*8, 0+10*8
 };
 
 static av_always_inline uint32_t pack16to32(int a, int b){
@@ -773,11 +774,11 @@ static void fill_decode_neighbors(H264Context *h, int mb_type){
     MpegEncContext * const s = &h->s;
     const int mb_xy= h->mb_xy;
     int topleft_xy, top_xy, topright_xy, left_xy[2];
-    static const uint8_t left_block_options[4][16]={
-        {0,1,2,3,7,10,8,11,7+0*8, 7+1*8, 7+2*8, 7+3*8, 2+0*8, 2+3*8, 2+1*8, 2+2*8},
-        {2,2,3,3,8,11,8,11,7+2*8, 7+2*8, 7+3*8, 7+3*8, 2+1*8, 2+2*8, 2+1*8, 2+2*8},
-        {0,0,1,1,7,10,7,10,7+0*8, 7+0*8, 7+1*8, 7+1*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8},
-        {0,2,0,2,7,10,7,10,7+0*8, 7+2*8, 7+0*8, 7+2*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8}
+    static const uint8_t left_block_options[4][32]={
+        {0,1,2,3,7,10,8,11,3+0*4, 3+1*4, 3+2*4, 3+3*4, 1+4*4, 1+8*4, 1+5*4, 1+9*4},
+        {2,2,3,3,8,11,8,11,3+2*4, 3+2*4, 3+3*4, 3+3*4, 1+5*4, 1+9*4, 1+5*4, 1+9*4},
+        {0,0,1,1,7,10,7,10,3+0*4, 3+0*4, 3+1*4, 3+1*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4},
+        {0,2,0,2,7,10,7,10,3+0*4, 3+2*4, 3+0*4, 3+2*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4}
     };
 
     h->topleft_partition= -1;
@@ -947,32 +948,41 @@ static void fill_decode_caches(H264Context *h, int mb_type){
 */
 //FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec)
     if(top_type){
-        AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][4+3*8]);
-            h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][1+1*8];
-            h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][2+1*8];
-
-            h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][1+2*8];
-            h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][2+2*8];
-    }else {
-            h->non_zero_count_cache[1+8*0]=
-            h->non_zero_count_cache[2+8*0]=
-
-            h->non_zero_count_cache[1+8*3]=
-            h->non_zero_count_cache[2+8*3]=
-            AV_WN32A(&h->non_zero_count_cache[4+8*0], CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040);
+        AV_COPY32(&h->non_zero_count_cache[4+8* 0], &h->non_zero_count[top_xy][4*3]);
+        if(CHROMA444){
+            AV_COPY32(&h->non_zero_count_cache[4+8* 5], &h->non_zero_count[top_xy][4* 7]);
+            AV_COPY32(&h->non_zero_count_cache[4+8*10], &h->non_zero_count[top_xy][4*11]);
+        }else{
+            AV_COPY32(&h->non_zero_count_cache[4+8* 5], &h->non_zero_count[top_xy][4* 5]);
+            AV_COPY32(&h->non_zero_count_cache[4+8*10], &h->non_zero_count[top_xy][4* 9]);
+        }
+    }else{
+        uint32_t top_empty = CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040;
+        AV_WN32A(&h->non_zero_count_cache[4+8* 0], top_empty);
+        AV_WN32A(&h->non_zero_count_cache[4+8* 5], top_empty);
+        AV_WN32A(&h->non_zero_count_cache[4+8*10], top_empty);
     }
 
     for (i=0; i<2; i++) {
         if(left_type[i]){
-            h->non_zero_count_cache[3+8*1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]];
-            h->non_zero_count_cache[3+8*2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]];
-                h->non_zero_count_cache[0+8*1 +   8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]];
-                h->non_zero_count_cache[0+8*4 +   8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]];
+            h->non_zero_count_cache[3+8* 1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]];
+            h->non_zero_count_cache[3+8* 2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]];
+            if(CHROMA444){
+                h->non_zero_count_cache[3+8* 6 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]+4*4];
+                h->non_zero_count_cache[3+8* 7 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]+4*4];
+                h->non_zero_count_cache[3+8*11 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]+8*4];
+                h->non_zero_count_cache[3+8*12 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]+8*4];
+            }else{
+                h->non_zero_count_cache[3+8* 6 +   8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]];
+                h->non_zero_count_cache[3+8*11 +   8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]];
+            }
         }else{
-                h->non_zero_count_cache[3+8*1 + 2*8*i]=
-                h->non_zero_count_cache[3+8*2 + 2*8*i]=
-                h->non_zero_count_cache[0+8*1 +   8*i]=
-                h->non_zero_count_cache[0+8*4 +   8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64;
+            h->non_zero_count_cache[3+8* 1 + 2*8*i]=
+            h->non_zero_count_cache[3+8* 2 + 2*8*i]=
+            h->non_zero_count_cache[3+8* 6 + 2*8*i]=
+            h->non_zero_count_cache[3+8* 7 + 2*8*i]=
+            h->non_zero_count_cache[3+8*11 + 2*8*i]=
+            h->non_zero_count_cache[3+8*12 + 2*8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64;
         }
     }
 
@@ -981,15 +991,15 @@ static void fill_decode_caches(H264Context *h, int mb_type){
         if(top_type) {
             h->top_cbp = h->cbp_table[top_xy];
         } else {
-            h->top_cbp = IS_INTRA(mb_type) ? 0x1CF : 0x00F;
+            h->top_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F;
         }
         // left_cbp
         if (left_type[0]) {
-            h->left_cbp = (h->cbp_table[left_xy[0]] & 0x1f0)
+            h->left_cbp =   (h->cbp_table[left_xy[0]] & 0x7F0)
                         |  ((h->cbp_table[left_xy[0]]>>(left_block[0]&(~1)))&2)
                         | (((h->cbp_table[left_xy[1]]>>(left_block[2]&(~1)))&2) << 2);
         } else {
-            h->left_cbp = IS_INTRA(mb_type) ? 0x1CF : 0x00F;
+            h->left_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F;
         }
     }
     }
@@ -1190,11 +1200,21 @@ static inline int pred_intra_mode(H264Context *h, int n){
 static inline void write_back_non_zero_count(H264Context *h){
     const int mb_xy= h->mb_xy;
 
-    AV_COPY64(&h->non_zero_count[mb_xy][ 0], &h->non_zero_count_cache[0+8*1]);
-    AV_COPY64(&h->non_zero_count[mb_xy][ 8], &h->non_zero_count_cache[0+8*2]);
-    AV_COPY32(&h->non_zero_count[mb_xy][16], &h->non_zero_count_cache[0+8*5]);
-    AV_COPY32(&h->non_zero_count[mb_xy][20], &h->non_zero_count_cache[4+8*3]);
-    AV_COPY64(&h->non_zero_count[mb_xy][24], &h->non_zero_count_cache[0+8*4]);
+    AV_COPY32(&h->non_zero_count[mb_xy][ 0], &h->non_zero_count_cache[4+8* 1]);
+    AV_COPY32(&h->non_zero_count[mb_xy][ 4], &h->non_zero_count_cache[4+8* 2]);
+    AV_COPY32(&h->non_zero_count[mb_xy][ 8], &h->non_zero_count_cache[4+8* 3]);
+    AV_COPY32(&h->non_zero_count[mb_xy][12], &h->non_zero_count_cache[4+8* 4]);
+    AV_COPY32(&h->non_zero_count[mb_xy][16], &h->non_zero_count_cache[4+8* 6]);
+    AV_COPY32(&h->non_zero_count[mb_xy][20], &h->non_zero_count_cache[4+8* 7]);
+    AV_COPY32(&h->non_zero_count[mb_xy][32], &h->non_zero_count_cache[4+8*11]);
+    AV_COPY32(&h->non_zero_count[mb_xy][36], &h->non_zero_count_cache[4+8*12]);
+
+    if(CHROMA444){
+        AV_COPY32(&h->non_zero_count[mb_xy][24], &h->non_zero_count_cache[4+8* 8]);
+        AV_COPY32(&h->non_zero_count[mb_xy][28], &h->non_zero_count_cache[4+8* 9]);
+        AV_COPY32(&h->non_zero_count[mb_xy][40], &h->non_zero_count_cache[4+8*13]);
+        AV_COPY32(&h->non_zero_count[mb_xy][44], &h->non_zero_count_cache[4+8*14]);
+    }
 }
 
 static inline void write_back_motion(H264Context *h, int mb_type){
@@ -1267,8 +1287,7 @@ static void av_unused decode_mb_skip(H264Context *h){
     const int mb_xy= h->mb_xy;
     int mb_type=0;
 
-    memset(h->non_zero_count[mb_xy], 0, 32);
-    memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui
+    memset(h->non_zero_count[mb_xy], 0, 48);
 
     if(MB_FIELD)
         mb_type|= MB_TYPE_INTERLACED;

libavcodec/h264_cavlc.c

@@ -371,12 +371,12 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in
 
     //FIXME put trailing_onex into the context
 
-    if(n >= CHROMA_DC_BLOCK_INDEX){
+    if(max_coeff <= 8){
         coeff_token= get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1);
         total_coeff= coeff_token>>2;
     }else{
-        if(n == LUMA_DC_BLOCK_INDEX){
-            total_coeff= pred_non_zero_count(h, 0);
+        if(n >= LUMA_DC_BLOCK_INDEX){
+            total_coeff= pred_non_zero_count(h, (n - LUMA_DC_BLOCK_INDEX)*16);
             coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
             total_coeff= coeff_token>>2;
         }else{
@@ -482,7 +482,8 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in
     if(total_coeff == max_coeff)
         zeros_left=0;
     else{
-        if(n >= CHROMA_DC_BLOCK_INDEX)
+        /* 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
             zeros_left= get_vlc2(gb, (total_zeros_vlc-1)[ total_coeff ].table, TOTAL_ZEROS_VLC_BITS, 1);
@@ -536,12 +537,80 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in
     return 0;
 }
 
+static av_always_inline int decode_luma_residual(H264Context *h, GetBitContext *gb, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p){
+    int i4x4, i8x8;
+    MpegEncContext * const s = &h->s;
+    int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1];
+    if(IS_INTRA16x16(mb_type)){
+        AV_ZERO128(h->mb_luma_dc[p]+0);
+        AV_ZERO128(h->mb_luma_dc[p]+8);
+        AV_ZERO128(h->mb_luma_dc[p]+16);
+        AV_ZERO128(h->mb_luma_dc[p]+24);
+        if( decode_residual(h, h->intra_gb_ptr, h->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX+p, scan, NULL, 16) < 0){
+            return -1; //FIXME continue if partitioned and other return -1 too
+        }
+
+        assert((cbp&15) == 0 || (cbp&15) == 15);
+
+        if(cbp&15){
+            for(i8x8=0; i8x8<4; i8x8++){
+                for(i4x4=0; i4x4<4; i4x4++){
+                    const int index= i4x4 + 4*i8x8 + p*16;
+                    if( decode_residual(h, h->intra_gb_ptr, h->mb + (16*index << pixel_shift),
+                        index, scan + 1, h->dequant4_coeff[p][qscale], 15) < 0 ){
+                        return -1;
+                    }
+                }
+            }
+            return 0xf;
+        }else{
+            fill_rectangle(&h->non_zero_count_cache[scan8[p*16]], 4, 4, 8, 0, 1);
+            return 0;
+        }
+    }else{
+        int cqm = (IS_INTRA( mb_type ) ? 0:3)+p;
+        /* For CAVLC 4:4:4, we need to keep track of the luma 8x8 CBP for deblocking nnz purposes. */
+        int new_cbp = 0;
+        for(i8x8=0; i8x8<4; i8x8++){
+            if(cbp & (1<<i8x8)){
+                if(IS_8x8DCT(mb_type)){
+                    DCTELEM *buf = &h->mb[64*i8x8+256*p << pixel_shift];
+                    uint8_t *nnz;
+                    for(i4x4=0; i4x4<4; i4x4++){
+                        const int index= i4x4 + 4*i8x8 + p*16;
+                        if( decode_residual(h, gb, buf, index, scan8x8+16*i4x4,
+                                            h->dequant8_coeff[cqm][qscale], 16) < 0 )
+                            return -1;
+                    }
+                    nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ];
+                    nnz[0] += nnz[1] + nnz[8] + nnz[9];
+                    new_cbp |= !!nnz[0] << i8x8;
+                }else{
+                    for(i4x4=0; i4x4<4; i4x4++){
+                        const int index= i4x4 + 4*i8x8 + p*16;
+                        if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index,
+                                            scan, h->dequant4_coeff[cqm][qscale], 16) < 0 ){
+                            return -1;
+                        }
+                        new_cbp |= h->non_zero_count_cache[ scan8[index] ] << i8x8;
+                    }
+                }
+            }else{
+                uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ];
+                nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;
+            }
+        }
+        return new_cbp;
+    }
+}
+
 int ff_h264_decode_mb_cavlc(H264Context *h){
     MpegEncContext * const s = &h->s;
     int mb_xy;
     int partition_count;
     unsigned int mb_type, cbp;
     int dct8x8_allowed= h->pps.transform_8x8_mode;
+    int decode_chroma = h->sps.chroma_format_idc == 1 || h->sps.chroma_format_idc == 2;
     const int pixel_shift = h->pixel_shift;
 
     mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride;
@@ -608,19 +677,21 @@ decode_intra_mb:
 
     if(IS_INTRA_PCM(mb_type)){
         unsigned int x;
+        static const uint16_t mb_sizes[4] = {256,384,512,768};
+        const int mb_size = mb_sizes[h->sps.chroma_format_idc]*h->sps.bit_depth_luma >> 3;
 
         // We assume these blocks are very rare so we do not optimize it.
         align_get_bits(&s->gb);
 
         // The pixels are stored in the same order as levels in h->mb array.
-        for(x=0; x < (CHROMA ? 384 : 256)*h->sps.bit_depth_luma/8; x++){
+        for(x=0; x < mb_size; x++){
             ((uint8_t*)h->mb)[x]= get_bits(&s->gb, 8);
         }
 
         // In deblocking, the quantizer is 0
         s->current_picture.qscale_table[mb_xy]= 0;
         // All coeffs are present
-        memset(h->non_zero_count[mb_xy], 16, 32);
+        memset(h->non_zero_count[mb_xy], 16, 48);
 
         s->current_picture.mb_type[mb_xy]= mb_type;
         return 0;
@@ -668,7 +739,7 @@ decode_intra_mb:
             if(h->intra16x16_pred_mode < 0)
                 return -1;
         }
-        if(CHROMA){
+        if(decode_chroma){
             pred_mode= ff_h264_check_intra_pred_mode(h, get_ue_golomb_31(&s->gb));
             if(pred_mode < 0)
                 return -1;
@@ -896,15 +967,19 @@ decode_intra_mb:
 
     if(!IS_INTRA16x16(mb_type)){
         cbp= get_ue_golomb(&s->gb);
-        if(cbp > 47){
-            av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y);
-            return -1;
-        }
 
-        if(CHROMA){
+        if(decode_chroma){
+            if(cbp > 47){
+                av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y);
+                return -1;
+            }
             if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp[cbp];
             else                     cbp= golomb_to_inter_cbp   [cbp];
         }else{
+            if(cbp > 15){
+                av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y);
+                return -1;
+            }
             if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp_gray[cbp];
             else                     cbp= golomb_to_inter_cbp_gray[cbp];
         }
@@ -918,8 +993,9 @@ decode_intra_mb:
     s->current_picture.mb_type[mb_xy]= mb_type;
 
     if(cbp || IS_INTRA16x16(mb_type)){
-        int i8x8, i4x4, chroma_idx;
+        int i4x4, chroma_idx;
         int dquant;
+        int ret;
         GetBitContext *gb= IS_INTRA(mb_type) ? h->intra_gb_ptr : h->inter_gb_ptr;
         const uint8_t *scan, *scan8x8;
         const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8);
@@ -947,85 +1023,45 @@ decode_intra_mb:
 
         h->chroma_qp[0]= get_chroma_qp(h, 0, s->qscale);
         h->chroma_qp[1]= get_chroma_qp(h, 1, s->qscale);
-        if(IS_INTRA16x16(mb_type)){
-            AV_ZERO128(h->mb_luma_dc+0);
-            AV_ZERO128(h->mb_luma_dc+8);
-            AV_ZERO128(h->mb_luma_dc+16);
-            AV_ZERO128(h->mb_luma_dc+24);
-            if( decode_residual(h, h->intra_gb_ptr, h->mb_luma_dc, LUMA_DC_BLOCK_INDEX, scan, h->dequant4_coeff[0][s->qscale], 16) < 0){
-                return -1; //FIXME continue if partitioned and other return -1 too
-            }
 
-            assert((cbp&15) == 0 || (cbp&15) == 15);
+        if( (ret = decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 0)) < 0 ){
+            return -1;
+        }
+        h->cbp_table[mb_xy] |= ret << 12;
+        if(CHROMA444){
+            if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 1) < 0 ){
+                return -1;
+            }
+            if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 2) < 0 ){
+                return -1;
+            }
+        } else {
+            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){
+                        return -1;
+                    }
+            }
 
-            if(cbp&15){
-                for(i8x8=0; i8x8<4; i8x8++){
+            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= i4x4 + 4*i8x8;
-                        if( decode_residual(h, h->intra_gb_ptr, h->mb + (16*index << pixel_shift), index, scan + 1, h->dequant4_coeff[0][s->qscale], 15) < 0 ){
+                        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;
                         }
                     }
                 }
             }else{
-                fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1);
+                fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1);
+                fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1);
             }
-        }else{
-            for(i8x8=0; i8x8<4; i8x8++){
-                if(cbp & (1<<i8x8)){
-                    if(IS_8x8DCT(mb_type)){
-                        DCTELEM *buf = &h->mb[64*i8x8 << pixel_shift];
-                        uint8_t *nnz;
-                        for(i4x4=0; i4x4<4; i4x4++){
-                            if( decode_residual(h, gb, buf, i4x4+4*i8x8, scan8x8+16*i4x4,
-                                                h->dequant8_coeff[IS_INTRA( mb_type ) ? 0:1][s->qscale], 16) <0 )
-                                return -1;
-                        }
-                        nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ];
-                        nnz[0] += nnz[1] + nnz[8] + nnz[9];
-                    }else{
-                        for(i4x4=0; i4x4<4; i4x4++){
-                            const int index= i4x4 + 4*i8x8;
-
-                            if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan, h->dequant4_coeff[IS_INTRA( mb_type ) ? 0:3][s->qscale], 16) <0 ){
-                                return -1;
-                            }
-                        }
-                    }
-                }else{
-                    uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ];
-                    nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;
-                }
-            }
-        }
-
-        if(cbp&0x30){
-            for(chroma_idx=0; chroma_idx<2; chroma_idx++)
-                if( decode_residual(h, gb, h->mb + ((256 + 16*4*chroma_idx) << pixel_shift), CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma_dc_scan, NULL, 4) < 0){
-                    return -1;
-                }
-        }
-
-        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 + 4*chroma_idx + i4x4;
-                    if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){
-                        return -1;
-                    }
-                }
-            }
-        }else{
-            uint8_t * const nnz= &h->non_zero_count_cache[0];
-            nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
-            nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
         }
     }else{
-        uint8_t * const nnz= &h->non_zero_count_cache[0];
-        fill_rectangle(&nnz[scan8[0]], 4, 4, 8, 0, 1);
-        nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
-        nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
+        fill_rectangle(&h->non_zero_count_cache[scan8[ 0]], 4, 4, 8, 0, 1);
+        fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1);
+        fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1);
     }
     s->current_picture.qscale_table[mb_xy]= s->qscale;
     write_back_non_zero_count(h);

libavcodec/h264_loopfilter.c

@@ -217,10 +217,11 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
     int mb_xy;
     int mb_type, left_type;
     int qp, qp0, qp1, qpc, qpc0, qpc1, qp_thresh;
+    int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
 
     mb_xy = h->mb_xy;
 
-    if(!h->top_type || !h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff) {
+    if(!h->top_type || !h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff || CHROMA444) {
         ff_h264_filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);
         return;
     }
@@ -262,16 +263,18 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
             filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, h);
             filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, h);
         }
-        if(left_type){
-            filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, h);
-            filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, h);
+        if(chroma){
+            if(left_type){
+                filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, h);
+                filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, h);
+            }
+            filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, h);
+            filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, h);
+            filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, h);
+            filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, h);
+            filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, h);
+            filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, h);
         }
-        filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, h);
-        filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, h);
-        filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, h);
-        filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, h);
-        filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, h);
-        filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, h);
         return;
     } else {
         LOCAL_ALIGNED_8(int16_t, bS, [2], [4][4]);
@@ -298,7 +301,7 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
 #define FILTER(hv,dir,edge)\
         if(AV_RN64A(bS[dir][edge])) {                                   \
             filter_mb_edge##hv( &img_y[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qp : qp##dir, h );\
-            if(!(edge&1)) {\
+            if(chroma && !(edge&1)) {\
                 filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h );\
                 filter_mb_edgec##hv( &img_cr[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h );\
             }\
@@ -353,9 +356,10 @@ 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 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 chroma, int chroma444, int dir) {
     MpegEncContext * const s = &h->s;
     int edge;
+    int chroma_qp_avg[2];
     const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy;
     const int mbm_type = dir == 0 ? h->left_type[0] : h->top_type;
 
@@ -394,7 +398,7 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
                         bS[2]= 1+((h->cbp_table[mbn_xy] & 8)||h->non_zero_count_cache[scan8[0]+2]);
                         bS[3]= 1+((h->cbp_table[mbn_xy] & 8)||h->non_zero_count_cache[scan8[0]+3]);
                     }else{
-                    const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 4+3*8;
+                    const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 3*4;
                     int i;
                     for( i = 0; i < 4; i++ ) {
                         bS[i] = 1 + !!(h->non_zero_count_cache[scan8[0]+i] | mbn_nnz[i]);
@@ -407,10 +411,17 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
                 tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize);
                 { int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
                 filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, h );
-                filter_mb_edgech( &img_cb[j*uvlinesize], tmp_uvlinesize, bS,
-                                ( h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1, h);
-                filter_mb_edgech( &img_cr[j*uvlinesize], tmp_uvlinesize, bS,
-                                ( h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1, h);
+                chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
+                chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
+                if (chroma) {
+                    if (chroma444) {
+                        filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], h);
+                        filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h);
+                    } else {
+                        filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], h);
+                        filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h);
+                    }
+                }
             }
         }else{
             DECLARE_ALIGNED(8, int16_t, bS)[4];
@@ -465,23 +476,29 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
                 //tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], s->current_picture.qscale_table[mbn_xy]);
                 tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
                 //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
+                chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1;
+                chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1;
                 if( dir == 0 ) {
                     filter_mb_edgev( &img_y[0], linesize, bS, qp, h );
-                    {
-                        int qp= ( h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1;
-                        filter_mb_edgecv( &img_cb[0], uvlinesize, bS, qp, h);
-                        if(h->pps.chroma_qp_diff)
-                            qp= ( h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1;
-                        filter_mb_edgecv( &img_cr[0], uvlinesize, bS, qp, h);
+                    if (chroma) {
+                        if (chroma444) {
+                            filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h);
+                            filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h);
+                        } else {
+                            filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h);
+                            filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h);
+                        }
                     }
                 } else {
                     filter_mb_edgeh( &img_y[0], linesize, bS, qp, h );
-                    {
-                        int qp= ( h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1;
-                        filter_mb_edgech( &img_cb[0], uvlinesize, bS, qp, h);
-                        if(h->pps.chroma_qp_diff)
-                            qp= ( h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1;
-                        filter_mb_edgech( &img_cr[0], uvlinesize, bS, qp, h);
+                    if (chroma) {
+                        if (chroma444) {
+                            filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h);
+                            filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h);
+                        } else {
+                            filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h);
+                            filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h);
+                        }
                     }
                 }
             }
@@ -545,15 +562,25 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
         //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
         if( dir == 0 ) {
             filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, h );
-            if( (edge&1) == 0 ) {
-                filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h);
-                filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h);
+            if (chroma) {
+                if (chroma444) {
+                    filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h);
+                    filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h);
+                } else if( (edge&1) == 0 ) {
+                    filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h);
+                    filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h);
+                }
             }
         } else {
             filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, h );
-            if( (edge&1) == 0 ) {
-                filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h);
-                filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h);
+            if (chroma) {
+                if (chroma444) {
+                    filter_mb_edgeh ( &img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h);
+                    filter_mb_edgeh ( &img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h);
+                } else if( (edge&1) == 0 ) {
+                    filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h);
+                    filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h);
+                }
             }
         }
     }
@@ -566,6 +593,7 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
     const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4;
     int first_vertical_edge_done = 0;
     av_unused int dir;
+    int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
 
     if (FRAME_MBAFF
             // and current and left pair do not have the same interlaced type
@@ -589,11 +617,11 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
         } else {
             static const uint8_t offset[2][2][8]={
                 {
-                    {7+8*0, 7+8*0, 7+8*0, 7+8*0, 7+8*1, 7+8*1, 7+8*1, 7+8*1},
-                    {7+8*2, 7+8*2, 7+8*2, 7+8*2, 7+8*3, 7+8*3, 7+8*3, 7+8*3},
+                    {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1},
+                    {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3},
                 },{
-                    {7+8*0, 7+8*1, 7+8*2, 7+8*3, 7+8*0, 7+8*1, 7+8*2, 7+8*3},
-                    {7+8*0, 7+8*1, 7+8*2, 7+8*3, 7+8*0, 7+8*1, 7+8*2, 7+8*3},
+                    {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
+                    {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
                 }
             };
             const uint8_t *off= offset[MB_FIELD][mb_y&1];
@@ -634,25 +662,29 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
         if(MB_FIELD){
             filter_mb_mbaff_edgev ( h, img_y                ,   linesize, bS  , 1, qp [0] );
             filter_mb_mbaff_edgev ( h, img_y  + 8*  linesize,   linesize, bS+4, 1, qp [1] );
-            filter_mb_mbaff_edgecv( h, img_cb,                uvlinesize, bS  , 1, bqp[0] );
-            filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1] );
-            filter_mb_mbaff_edgecv( h, img_cr,                uvlinesize, bS  , 1, rqp[0] );
-            filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1] );
+            if (chroma){
+                filter_mb_mbaff_edgecv( h, img_cb,                uvlinesize, bS  , 1, bqp[0] );
+                filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1] );
+                filter_mb_mbaff_edgecv( h, img_cr,                uvlinesize, bS  , 1, rqp[0] );
+                filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1] );
+            }
         }else{
             filter_mb_mbaff_edgev ( h, img_y              , 2*  linesize, bS  , 2, qp [0] );
             filter_mb_mbaff_edgev ( h, img_y  +   linesize, 2*  linesize, bS+1, 2, qp [1] );
-            filter_mb_mbaff_edgecv( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0] );
-            filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1] );
-            filter_mb_mbaff_edgecv( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0] );
-            filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1] );
+            if (chroma){
+                filter_mb_mbaff_edgecv( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0] );
+                filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1] );
+                filter_mb_mbaff_edgecv( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0] );
+                filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1] );
+            }
         }
     }
 
 #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, 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, chroma, CHROMA444, 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, 0);
-    filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, 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, 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, chroma, CHROMA444, 1);
 #endif
 }

libavcodec/h264_ps.c

@@ -269,7 +269,7 @@ static void decode_scaling_matrices(H264Context *h, SPS *sps, PPS *pps, int is_s
         fallback_sps ? sps->scaling_matrix4[0] : default_scaling4[0],
         fallback_sps ? sps->scaling_matrix4[3] : default_scaling4[1],
         fallback_sps ? sps->scaling_matrix8[0] : default_scaling8[0],
-        fallback_sps ? sps->scaling_matrix8[1] : default_scaling8[1]
+        fallback_sps ? sps->scaling_matrix8[3] : default_scaling8[1]
     };
     if(get_bits1(&s->gb)){
         sps->scaling_matrix_present |= is_sps;
@@ -281,7 +281,15 @@ static void decode_scaling_matrices(H264Context *h, SPS *sps, PPS *pps, int is_s
         decode_scaling_list(h,scaling_matrix4[5],16,default_scaling4[1],scaling_matrix4[4]); // Inter, Cb
         if(is_sps || pps->transform_8x8_mode){
             decode_scaling_list(h,scaling_matrix8[0],64,default_scaling8[0],fallback[2]);  // Intra, Y
-            decode_scaling_list(h,scaling_matrix8[1],64,default_scaling8[1],fallback[3]);  // Inter, Y
+            if(h->sps.chroma_format_idc == 3){
+                decode_scaling_list(h,scaling_matrix8[1],64,default_scaling8[0],scaling_matrix8[0]);  // Intra, Cr
+                decode_scaling_list(h,scaling_matrix8[2],64,default_scaling8[0],scaling_matrix8[1]);  // Intra, Cb
+            }
+            decode_scaling_list(h,scaling_matrix8[3],64,default_scaling8[1],fallback[3]);  // Inter, Y
+            if(h->sps.chroma_format_idc == 3){
+                decode_scaling_list(h,scaling_matrix8[4],64,default_scaling8[1],scaling_matrix8[3]);  // Inter, Cr
+                decode_scaling_list(h,scaling_matrix8[5],64,default_scaling8[1],scaling_matrix8[4]);  // Inter, Cb
+            }
         }
     }
 }
@@ -395,7 +403,7 @@ 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 >= 8 || sps->crop_bottom >= 8){
+        if(sps->crop_right >= (8<<CHROMA444) || sps->crop_bottom >= (8<<CHROMA444)){
             av_log(h->s.avctx, AV_LOG_ERROR, "brainfart cropping not supported, this could look slightly wrong ...\n");
         }
     }else{

libavcodec/h264dsp.h

@@ -66,10 +66,10 @@ typedef struct H264DSPContext{
     void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
     void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
 
-    void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
-    void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
-    void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
-    void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
+    void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[15*8]);
+    void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[15*8]);
+    void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[15*8]);
+    void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[15*8]);
     void (*h264_luma_dc_dequant_idct)(DCTELEM *output, DCTELEM *input/*align 16*/, int qmul);
     void (*h264_chroma_dc_dequant_idct)(DCTELEM *block, int qmul);
 }H264DSPContext;

libavcodec/h264idct_template.c

@@ -30,15 +30,19 @@
 #ifndef AVCODEC_H264IDCT_INTERNAL_H
 #define AVCODEC_H264IDCT_INTERNAL_H
 //FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split
-static const uint8_t scan8[16 + 2*4]={
- 4+1*8, 5+1*8, 4+2*8, 5+2*8,
- 6+1*8, 7+1*8, 6+2*8, 7+2*8,
- 4+3*8, 5+3*8, 4+4*8, 5+4*8,
- 6+3*8, 7+3*8, 6+4*8, 7+4*8,
- 1+1*8, 2+1*8,
- 1+2*8, 2+2*8,
- 1+4*8, 2+4*8,
- 1+5*8, 2+5*8,
+static const uint8_t scan8[16*3]={
+ 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8,
+ 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8,
+ 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8,
+ 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8,
+ 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8,
+ 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8,
+ 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8,
+ 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8,
+ 4+11*8, 5+11*8, 4+12*8, 5+12*8,
+ 6+11*8, 7+11*8, 6+12*8, 7+12*8,
+ 4+13*8, 5+13*8, 4+14*8, 5+14*8,
+ 6+13*8, 7+13*8, 6+14*8, 7+14*8
 };
 #endif
 
@@ -190,7 +194,7 @@ void FUNCC(ff_h264_idct8_dc_add)(uint8_t *p_dst, DCTELEM *block, int stride){
     }
 }
 
-void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
     int i;
     for(i=0; i<16; i++){
         int nnz = nnzc[ scan8[i] ];
@@ -201,7 +205,7 @@ void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *b
     }
 }
 
-void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
     int i;
     for(i=0; i<16; i++){
         if(nnzc[ scan8[i] ])             FUNCC(idct_internal      )(dst + block_offset[i], block + i*16*sizeof(pixel), stride, 4, 6, 1);
@@ -209,7 +213,7 @@ void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTEL
     }
 }
 
-void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
     int i;
     for(i=0; i<16; i+=4){
         int nnz = nnzc[ scan8[i] ];
@@ -220,13 +224,15 @@ void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *b
     }
 }
 
-void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
-    int i;
-    for(i=16; i<16+8; i++){
-        if(nnzc[ scan8[i] ])
-            FUNCC(ff_h264_idct_add   )(dest[(i&4)>>2] + block_offset[i], block + i*16*sizeof(pixel), stride);
-        else if(((dctcoef*)block)[i*16])
-            FUNCC(ff_h264_idct_dc_add)(dest[(i&4)>>2] + block_offset[i], block + i*16*sizeof(pixel), stride);
+void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
+    int i, j;
+    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);
+        }
     }
 }
 /**

libavcodec/mpegvideo.c

@@ -1180,12 +1180,17 @@ void MPV_frame_end(MpegEncContext *s)
        && s->current_picture.reference
        && !s->intra_only
        && !(s->flags&CODEC_FLAG_EMU_EDGE)) {
-        int edges = EDGE_BOTTOM | EDGE_TOP, h = s->v_edge_pos;
-
-            s->dsp.draw_edges(s->current_picture.data[0], s->linesize  , s->h_edge_pos   , h   , EDGE_WIDTH  , edges);
-            s->dsp.draw_edges(s->current_picture.data[1], s->uvlinesize, s->h_edge_pos>>1, h>>1, EDGE_WIDTH/2, edges);
-            s->dsp.draw_edges(s->current_picture.data[2], s->uvlinesize, s->h_edge_pos>>1, h>>1, EDGE_WIDTH/2, edges);
-
+            int hshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_w;
+            int vshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_h;
+            s->dsp.draw_edges(s->current_picture.data[0], s->linesize  ,
+                              s->h_edge_pos             , s->v_edge_pos,
+                              EDGE_WIDTH        , EDGE_WIDTH        , EDGE_TOP | EDGE_BOTTOM);
+            s->dsp.draw_edges(s->current_picture.data[1], s->uvlinesize,
+                              s->h_edge_pos>>hshift, s->v_edge_pos>>vshift,
+                              EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, EDGE_TOP | EDGE_BOTTOM);
+            s->dsp.draw_edges(s->current_picture.data[2], s->uvlinesize,
+                              s->h_edge_pos>>hshift, s->v_edge_pos>>vshift,
+                              EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, EDGE_TOP | EDGE_BOTTOM);
     }
 
     emms_c();
@@ -2289,14 +2294,19 @@ void ff_draw_horiz_band(MpegEncContext *s, int y, int h){
        && !s->intra_only
        && !(s->flags&CODEC_FLAG_EMU_EDGE)) {
         int sides = 0, edge_h;
+        int hshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_w;
+        int vshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_h;
         if (y==0) sides |= EDGE_TOP;
         if (y + h >= s->v_edge_pos) sides |= EDGE_BOTTOM;
 
         edge_h= FFMIN(h, s->v_edge_pos - y);
 
-        s->dsp.draw_edges(s->current_picture_ptr->data[0] +  y    *s->linesize  , s->linesize  , s->h_edge_pos   , edge_h   , EDGE_WIDTH  , sides);
-        s->dsp.draw_edges(s->current_picture_ptr->data[1] + (y>>1)*s->uvlinesize, s->uvlinesize, s->h_edge_pos>>1, edge_h>>1, EDGE_WIDTH/2, sides);
-        s->dsp.draw_edges(s->current_picture_ptr->data[2] + (y>>1)*s->uvlinesize, s->uvlinesize, s->h_edge_pos>>1, edge_h>>1, EDGE_WIDTH/2, sides);
+        s->dsp.draw_edges(s->current_picture_ptr->data[0] +  y         *s->linesize  , s->linesize,
+                          s->h_edge_pos        , edge_h        , EDGE_WIDTH        , EDGE_WIDTH        , sides);
+        s->dsp.draw_edges(s->current_picture_ptr->data[1] + (y>>vshift)*s->uvlinesize, s->uvlinesize,
+                          s->h_edge_pos>>hshift, edge_h>>hshift, EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, sides);
+        s->dsp.draw_edges(s->current_picture_ptr->data[2] + (y>>vshift)*s->uvlinesize, s->uvlinesize,
+                          s->h_edge_pos>>hshift, edge_h>>hshift, EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, sides);
     }
 
     h= FFMIN(h, s->avctx->height - y);

libavcodec/ppc/h264_altivec.c

@@ -527,7 +527,7 @@ static void ff_h264_idct8_dc_add_altivec(uint8_t *dst, DCTELEM *block, int strid
     h264_idct_dc_add_internal(dst, block, stride, 8);
 }
 
-static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
     int i;
     for(i=0; i<16; i++){
         int nnz = nnzc[ scan8[i] ];
@@ -538,7 +538,7 @@ static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, DC
     }
 }
 
-static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
     int i;
     for(i=0; i<16; i++){
         if(nnzc[ scan8[i] ]) ff_h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
@@ -546,7 +546,7 @@ static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offse
     }
 }
 
-static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
     int i;
     for(i=0; i<16; i+=4){
         int nnz = nnzc[ scan8[i] ];
@@ -557,13 +557,15 @@ static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, DC
     }
 }
 
-static void ff_h264_idct_add8_altivec(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
-    int i;
-    for(i=16; i<16+8; i++){
-        if(nnzc[ scan8[i] ])
-            ff_h264_idct_add_altivec(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
-        else if(block[i*16])
-            h264_idct_dc_add_altivec(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
+static void ff_h264_idct_add8_altivec(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
+    int i, j;
+    for (j = 1; j < 3; j++) {
+        for(i = j * 16; i < j * 16 + 4; i++){
+            if(nnzc[ scan8[i] ])
+                ff_h264_idct_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
+            else if(block[i*16])
+                h264_idct_dc_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
+        }
     }
 }
 

libavcodec/snow.c

@@ -1976,9 +1976,15 @@ static int frame_start(SnowContext *s){
    int h= s->avctx->height;
 
     if(s->current_picture.data[0]){
-        s->dsp.draw_edges(s->current_picture.data[0], s->current_picture.linesize[0], w   , h   , EDGE_WIDTH  , EDGE_TOP|EDGE_BOTTOM);
-        s->dsp.draw_edges(s->current_picture.data[1], s->current_picture.linesize[1], w>>1, h>>1, EDGE_WIDTH/2, EDGE_TOP|EDGE_BOTTOM);
-        s->dsp.draw_edges(s->current_picture.data[2], s->current_picture.linesize[2], w>>1, h>>1, EDGE_WIDTH/2, EDGE_TOP|EDGE_BOTTOM);
+        s->dsp.draw_edges(s->current_picture.data[0],
+                          s->current_picture.linesize[0], w   , h   ,
+                          EDGE_WIDTH  , EDGE_WIDTH  , EDGE_TOP | EDGE_BOTTOM);
+        s->dsp.draw_edges(s->current_picture.data[1],
+                          s->current_picture.linesize[1], w>>1, h>>1,
+                          EDGE_WIDTH/2, EDGE_WIDTH/2, EDGE_TOP | EDGE_BOTTOM);
+        s->dsp.draw_edges(s->current_picture.data[2],
+                          s->current_picture.linesize[2], w>>1, h>>1,
+                          EDGE_WIDTH/2, EDGE_WIDTH/2, EDGE_TOP | EDGE_BOTTOM);
     }
 
     release_buffer(s->avctx);

libavcodec/svq3.c

@@ -635,8 +635,9 @@ static int svq3_decode_mb(SVQ3Context *svq3, unsigned int mb_type)
         memset(h->intra4x4_pred_mode+h->mb2br_xy[mb_xy], DC_PRED, 8);
     }
     if (!IS_SKIP(mb_type) || s->pict_type == AV_PICTURE_TYPE_B) {
-        memset(h->non_zero_count_cache + 8, 0, 4*9*sizeof(uint8_t));
-        s->dsp.clear_blocks(h->mb);
+        memset(h->non_zero_count_cache + 8, 0, 14*8*sizeof(uint8_t));
+        s->dsp.clear_blocks(h->mb+  0);
+        s->dsp.clear_blocks(h->mb+384);
     }
 
     if (!IS_INTRA16x16(mb_type) && (!IS_SKIP(mb_type) || s->pict_type == AV_PICTURE_TYPE_B)) {
@@ -656,8 +657,8 @@ static int svq3_decode_mb(SVQ3Context *svq3, unsigned int mb_type)
         }
     }
     if (IS_INTRA16x16(mb_type)) {
-        AV_ZERO128(h->mb_luma_dc+0);
-        AV_ZERO128(h->mb_luma_dc+8);
+        AV_ZERO128(h->mb_luma_dc[0]+0);
+        AV_ZERO128(h->mb_luma_dc[0]+8);
         if (svq3_decode_block(&s->gb, h->mb_luma_dc, 0, 1)){
             av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding intra luma dc\n");
             return -1;
@@ -683,20 +684,23 @@ static int svq3_decode_mb(SVQ3Context *svq3, unsigned int mb_type)
         }
 
         if ((cbp & 0x30)) {
-            for (i = 0; i < 2; ++i) {
-              if (svq3_decode_block(&s->gb, &h->mb[16*(16 + 4*i)], 0, 3)){
+            for (i = 1; i < 3; ++i) {
+              if (svq3_decode_block(&s->gb, &h->mb[16*16*i], 0, 3)){
                 av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding chroma dc block\n");
                 return -1;
               }
             }
 
             if ((cbp & 0x20)) {
-                for (i = 0; i < 8; i++) {
-                    h->non_zero_count_cache[ scan8[16+i] ] = 1;
-
-                    if (svq3_decode_block(&s->gb, &h->mb[16*(16 + i)], 1, 1)){
-                        av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding chroma ac block\n");
-                        return -1;
+                for (i = 1; i < 3; i++) {
+                    for (j = 0; j < 4; j++) {
+                        k = 16*i + j;
+                        h->non_zero_count_cache[ scan8[k] ] = 1;
+
+                        if (svq3_decode_block(&s->gb, &h->mb[16*k], 1, 1)){
+                            av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding chroma ac block\n");
+                            return -1;
+                        }
                     }
                 }
             }

libavcodec/x86/dsputil_mmx.c

@@ -762,7 +762,7 @@ static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale){
 
 /* draw the edges of width 'w' of an image of size width, height
    this mmx version can only handle w==8 || w==16 */
-static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w, int sides)
+static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides)
 {
     uint8_t *ptr, *last_line;
     int i;
@@ -817,7 +817,7 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w,
 
     /* top and bottom (and hopefully also the corners) */
     if (sides&EDGE_TOP) {
-        for(i = 0; i < w; i += 4) {
+        for(i = 0; i < h; i += 4) {
             ptr= buf - (i + 1) * wrap - w;
             __asm__ volatile(
                     "1:                             \n\t"

libavcodec/x86/h264_i386.h

@@ -36,7 +36,7 @@
 #if ARCH_X86 && HAVE_7REGS && HAVE_EBX_AVAILABLE && !defined(BROKEN_RELOCATIONS)
 static int decode_significance_x86(CABACContext *c, int max_coeff,
                                    uint8_t *significant_coeff_ctx_base,
-                                   int *index){
+                                   int *index, x86_reg last_off){
     void *end= significant_coeff_ctx_base + max_coeff - 1;
     int minusstart= -(int)significant_coeff_ctx_base;
     int minusindex= 4-(int)index;
@@ -52,10 +52,12 @@ static int decode_significance_x86(CABACContext *c, int max_coeff,
 
         "test $1, %%edx                         \n\t"
         " jz 3f                                 \n\t"
+        "add  %7, %1                            \n\t"
 
-        BRANCHLESS_GET_CABAC("%%edx", "%3", "61(%1)", "%%ebx",
+        BRANCHLESS_GET_CABAC("%%edx", "%3", "(%1)", "%%ebx",
                              "%%bx", "%%esi", "%%eax", "%%al")
 
+        "sub  %7, %1                            \n\t"
         "mov  %2, %%"REG_a"                     \n\t"
         "movl %4, %%ecx                         \n\t"
         "add  %1, %%"REG_c"                     \n\t"
@@ -82,7 +84,7 @@ static int decode_significance_x86(CABACContext *c, int max_coeff,
         "movl %%esi, "RANGE    "(%3)            \n\t"
         "movl %%ebx, "LOW      "(%3)            \n\t"
         :"=&a"(coeff_count), "+r"(significant_coeff_ctx_base), "+m"(index)
-        :"r"(c), "m"(minusstart), "m"(end), "m"(minusindex)
+        :"r"(c), "m"(minusstart), "m"(end), "m"(minusindex), "m"(last_off)
         : "%"REG_c, "%ebx", "%edx", "%esi", "memory"
     );
     return coeff_count;
@@ -90,7 +92,7 @@ static int decode_significance_x86(CABACContext *c, int max_coeff,
 
 static int decode_significance_8x8_x86(CABACContext *c,
                                        uint8_t *significant_coeff_ctx_base,
-                                       int *index, const uint8_t *sig_off){
+                                       int *index, x86_reg last_off, const uint8_t *sig_off){
     int minusindex= 4-(int)index;
     int coeff_count;
     x86_reg last=0;
@@ -114,8 +116,9 @@ static int decode_significance_8x8_x86(CABACContext *c,
 
         "movzbl "MANGLE(last_coeff_flag_offset_8x8)"(%%edi), %%edi\n\t"
         "add %5, %%"REG_D"                      \n\t"
+        "add %7, %%"REG_D"                      \n\t"
 
-        BRANCHLESS_GET_CABAC("%%edx", "%3", "15(%%"REG_D")", "%%ebx",
+        BRANCHLESS_GET_CABAC("%%edx", "%3", "(%%"REG_D")", "%%ebx",
                              "%%bx", "%%esi", "%%eax", "%%al")
 
         "mov %2, %%"REG_a"                      \n\t"
@@ -142,7 +145,7 @@ static int decode_significance_8x8_x86(CABACContext *c,
         "movl %%esi, "RANGE    "(%3)            \n\t"
         "movl %%ebx, "LOW      "(%3)            \n\t"
         :"=&a"(coeff_count),"+m"(last), "+m"(index)
-        :"r"(c), "m"(minusindex), "m"(significant_coeff_ctx_base), "m"(sig_off)
+        :"r"(c), "m"(minusindex), "m"(significant_coeff_ctx_base), "m"(sig_off), "m"(last_off)
         : "%"REG_c, "%ebx", "%edx", "%esi", "%"REG_D, "memory"
     );
     return coeff_count;

libavcodec/x86/h264_idct.asm

@@ -32,14 +32,18 @@
 SECTION_RODATA
 
 ; FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split
-scan8_mem: db 4+1*8, 5+1*8, 4+2*8, 5+2*8
-           db 6+1*8, 7+1*8, 6+2*8, 7+2*8
-           db 4+3*8, 5+3*8, 4+4*8, 5+4*8
-           db 6+3*8, 7+3*8, 6+4*8, 7+4*8
-           db 1+1*8, 2+1*8
-           db 1+2*8, 2+2*8
-           db 1+4*8, 2+4*8
-           db 1+5*8, 2+5*8
+scan8_mem: db  4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8
+           db  6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8
+           db  4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8
+           db  6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8
+           db  4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8
+           db  6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8
+           db  4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8
+           db  6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8
+           db  4+11*8, 5+11*8, 4+12*8, 5+12*8
+           db  6+11*8, 7+11*8, 6+12*8, 7+12*8
+           db  4+13*8, 5+13*8, 4+14*8, 5+14*8
+           db  6+13*8, 7+13*8, 6+14*8, 7+14*8
 %ifdef PIC
 %define scan8 r11
 %else
@@ -617,6 +621,8 @@ cglobal h264_idct_add8_8_mmx, 5, 7, 0
     mov         r10, r0
 %endif
     call         h264_idct_add8_mmx_plane
+    mov          r5, 32
+    add          r2, 384
 %ifdef ARCH_X86_64
     add         r10, gprsize
 %else
@@ -678,6 +684,8 @@ cglobal h264_idct_add8_8_mmx2, 5, 7, 0
     lea         r11, [scan8_mem]
 %endif
     call h264_idct_add8_mmx2_plane
+    mov          r5, 32
+    add          r2, 384
 %ifdef ARCH_X86_64
     add         r10, gprsize
 %else
@@ -810,12 +818,12 @@ cglobal h264_idct_add16intra_8_sse2, 5, 7, 8
     test        r0, r0
     jz .try%1dc
 %ifdef ARCH_X86_64
-    mov        r0d, dword [r1+%1*8+64]
+    mov        r0d, dword [r1+(%1&1)*8+64*(1+(%1>>1))]
     add         r0, [r10]
 %else
     mov         r0, r0m
     mov         r0, [r0]
-    add         r0, dword [r1+%1*8+64]
+    add         r0, dword [r1+(%1&1)*8+64*(1+(%1>>1))]
 %endif
     call        x264_add8x4_idct_sse2
     jmp .cycle%1end
@@ -824,16 +832,18 @@ cglobal h264_idct_add16intra_8_sse2, 5, 7, 8
     or         r0w, word [r2+32]
     jz .cycle%1end
 %ifdef ARCH_X86_64
-    mov        r0d, dword [r1+%1*8+64]
+    mov        r0d, dword [r1+(%1&1)*8+64*(1+(%1>>1))]
     add         r0, [r10]
 %else
     mov         r0, r0m
     mov         r0, [r0]
-    add         r0, dword [r1+%1*8+64]
+    add         r0, dword [r1+(%1&1)*8+64*(1+(%1>>1))]
 %endif
     call        h264_idct_dc_add8_mmx2
 .cycle%1end
-%if %1 < 3
+%if %1 == 1
+    add         r2, 384+64
+%elif %1 < 3
     add         r2, 64
 %endif
 %endmacro
@@ -845,15 +855,15 @@ cglobal h264_idct_add8_8_sse2, 5, 7, 8
 %ifdef ARCH_X86_64
     mov         r10, r0
 %endif
-    add8_sse2_cycle 0, 0x09
-    add8_sse2_cycle 1, 0x11
+    add8_sse2_cycle 0, 0x34
+    add8_sse2_cycle 1, 0x3c
 %ifdef ARCH_X86_64
     add         r10, gprsize
 %else
     add        r0mp, gprsize
 %endif
-    add8_sse2_cycle 2, 0x21
-    add8_sse2_cycle 3, 0x29
+    add8_sse2_cycle 2, 0x5c
+    add8_sse2_cycle 3, 0x64
     RET
 
 ;void ff_h264_luma_dc_dequant_idct_mmx(DCTELEM *output, DCTELEM *input, int qmul)

libavcodec/x86/h264_idct_10bit.asm

@@ -29,14 +29,18 @@ SECTION_RODATA
 
 pw_pixel_max: times 8 dw ((1 << 10)-1)
 pd_32:        times 4 dd 32
-scan8_mem: db 4+1*8, 5+1*8, 4+2*8, 5+2*8
-           db 6+1*8, 7+1*8, 6+2*8, 7+2*8
-           db 4+3*8, 5+3*8, 4+4*8, 5+4*8
-           db 6+3*8, 7+3*8, 6+4*8, 7+4*8
-           db 1+1*8, 2+1*8
-           db 1+2*8, 2+2*8
-           db 1+4*8, 2+4*8
-           db 1+5*8, 2+5*8
+scan8_mem: db  4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8
+           db  6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8
+           db  4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8
+           db  6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8
+           db  4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8
+           db  6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8
+           db  4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8
+           db  6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8
+           db  4+11*8, 5+11*8, 4+12*8, 5+12*8
+           db  6+11*8, 7+11*8, 6+12*8, 7+12*8
+           db  4+13*8, 5+13*8, 4+14*8, 5+14*8
+           db  6+13*8, 7+13*8, 6+14*8, 7+14*8
 
 %ifdef PIC
 %define scan8 r11
@@ -306,7 +310,7 @@ INIT_AVX
 IDCT_ADD16INTRA_10 avx
 %endif
 
-%assign last_block 24
+%assign last_block 36
 ;-----------------------------------------------------------------------------
 ; h264_idct_add8(pixel **dst, const int *block_offset, dctcoef *block, int stride, const uint8_t nnzc[6*8])
 ;-----------------------------------------------------------------------------
@@ -317,21 +321,22 @@ cglobal h264_idct_add8_10_%1,5,7
 %endif
     add      r2, 1024
     mov      r0, [r0]
-    ADD16_OP_INTRA %1, 16, 1+1*8
-    ADD16_OP_INTRA %1, 18, 1+2*8
+    ADD16_OP_INTRA %1, 16, 4+ 6*8
+    ADD16_OP_INTRA %1, 18, 4+ 7*8
+    add      r2, 1024-128*2
 %ifdef ARCH_X86_64
     mov      r0, [r10+gprsize]
 %else
     mov      r0, r0m
     mov      r0, [r0+gprsize]
 %endif
-    ADD16_OP_INTRA %1, 20, 1+4*8
-    ADD16_OP_INTRA %1, 22, 1+5*8
+    ADD16_OP_INTRA %1, 32, 4+11*8
+    ADD16_OP_INTRA %1, 34, 4+12*8
     REP_RET
     AC %1, 16
     AC %1, 18
-    AC %1, 20
-    AC %1, 22
+    AC %1, 32
+    AC %1, 34
 
 %endmacro ; IDCT_ADD8
 

libswscale/swscale.c

@@ -506,6 +506,13 @@ static void yuv2nv12X_c(SwsContext *c, const int16_t *lumFilter,
         }
 }
 
+#define output_pixel(pos, val) \
+        if (target == PIX_FMT_GRAY16BE) { \
+            AV_WB16(pos, val); \
+        } else { \
+            AV_WL16(pos, val); \
+        }
+
 static av_always_inline void
 yuv2gray16_X_c_template(SwsContext *c, const int16_t *lumFilter,
                         const int16_t **lumSrc, int lumFilterSize,
@@ -516,12 +523,6 @@ yuv2gray16_X_c_template(SwsContext *c, const int16_t *lumFilter,
 {
     int i;
 
-#define output_pixel(pos, val) \
-        if (target == PIX_FMT_GRAY16BE) { \
-            AV_WB16(pos, val); \
-        } else { \
-            AV_WL16(pos, val); \
-        }
     for (i = 0; i < (dstW >> 1); i++) {
         int j;
         int Y1 = 1 << 18;
@@ -583,10 +584,11 @@ yuv2gray16_1_c_template(SwsContext *c, const uint16_t *buf0,
         output_pixel(&dest[2 * i2 + 0], Y1);
         output_pixel(&dest[2 * i2 + 2], Y2);
     }
-#undef output_pixel
 }
 
-#define YUV2PACKEDWRAPPER(name, ext, fmt) \
+#undef output_pixel
+
+#define YUV2PACKEDWRAPPER(name, base, ext, fmt) \
 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
                         const int16_t **lumSrc, int lumFilterSize, \
                         const int16_t *chrFilter, const int16_t **chrUSrc, \
@@ -594,7 +596,7 @@ static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
                         const int16_t **alpSrc, uint8_t *dest, int dstW, \
                         int y) \
 { \
-    name ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
+    name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
                           chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
                           alpSrc, dest, dstW, y, fmt); \
 } \
@@ -606,7 +608,7 @@ static void name ## ext ## _2_c(SwsContext *c, const uint16_t *buf0, \
                         const uint16_t *abuf1, uint8_t *dest, int dstW, \
                         int yalpha, int uvalpha, int y) \
 { \
-    name ## _2_c_template(c, buf0, buf1, ubuf0, ubuf1, \
+    name ## base ## _2_c_template(c, buf0, buf1, ubuf0, ubuf1, \
                           vbuf0, vbuf1, abuf0, abuf1, \
                           dest, dstW, yalpha, uvalpha, y, fmt); \
 } \
@@ -618,13 +620,20 @@ static void name ## ext ## _1_c(SwsContext *c, const uint16_t *buf0, \
                         int uvalpha, enum PixelFormat dstFormat, \
                         int flags, int y) \
 { \
-    name ## _1_c_template(c, buf0, ubuf0, ubuf1, vbuf0, \
+    name ## base ## _1_c_template(c, buf0, ubuf0, ubuf1, vbuf0, \
                           vbuf1, abuf0, dest, dstW, uvalpha, \
                           dstFormat, flags, y, fmt); \
 }
 
-YUV2PACKEDWRAPPER(yuv2gray16, LE, PIX_FMT_GRAY16LE);
-YUV2PACKEDWRAPPER(yuv2gray16, BE, PIX_FMT_GRAY16BE);
+YUV2PACKEDWRAPPER(yuv2gray16,, LE, PIX_FMT_GRAY16LE);
+YUV2PACKEDWRAPPER(yuv2gray16,, BE, PIX_FMT_GRAY16BE);
+
+#define output_pixel(pos, acc) \
+    if (target == PIX_FMT_MONOBLACK) { \
+        pos = acc; \
+    } else { \
+        pos = ~acc; \
+    }
 
 static av_always_inline void
 yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter,
@@ -639,12 +648,6 @@ yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter,
     int i;
     int acc = 0;
 
-#define output_pixel(pos, acc) \
-    if (target == PIX_FMT_MONOBLACK) { \
-        pos = acc; \
-    } else { \
-        pos = ~acc; \
-    }
     for (i = 0; i < dstW - 1; i += 2) {
         int j;
         int Y1 = 1 << 18;
@@ -718,21 +721,12 @@ yuv2mono_1_c_template(SwsContext *c, const uint16_t *buf0,
         acc += acc + g[(buf0[i + 7] >> 7) + d128[7]];
         output_pixel(*dest++, acc);
     }
-#undef output_pixel
 }
 
-YUV2PACKEDWRAPPER(yuv2mono, white, PIX_FMT_MONOWHITE);
-YUV2PACKEDWRAPPER(yuv2mono, black, PIX_FMT_MONOBLACK);
+#undef output_pixel
 
-static av_always_inline void
-yuv2422_X_c_template(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, int dstW,
-                     int y, enum PixelFormat target)
-{
-    int i;
+YUV2PACKEDWRAPPER(yuv2mono,, white, PIX_FMT_MONOWHITE);
+YUV2PACKEDWRAPPER(yuv2mono,, black, PIX_FMT_MONOBLACK);
 
 #define output_pixels(pos, Y1, U, Y2, V) \
     if (target == PIX_FMT_YUYV422) { \
@@ -747,6 +741,16 @@ yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter,
         dest[pos + 3] = Y2; \
     }
 
+static av_always_inline void
+yuv2422_X_c_template(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, int dstW,
+                     int y, enum PixelFormat target)
+{
+    int i;
+
     for (i = 0; i < (dstW >> 1); i++) {
         int j;
         int Y1 = 1 << 18;
@@ -828,11 +832,156 @@ yuv2422_1_c_template(SwsContext *c, const uint16_t *buf0,
             output_pixels(i * 4, Y1, U, Y2, V);
         }
     }
+}
+
 #undef output_pixels
+
+YUV2PACKEDWRAPPER(yuv2, 422, yuyv422, PIX_FMT_YUYV422);
+YUV2PACKEDWRAPPER(yuv2, 422, uyvy422, PIX_FMT_UYVY422);
+
+#define r_b ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? r : b)
+#define b_r ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? b : r)
+
+static av_always_inline void
+yuv2rgb48_X_c_template(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, int dstW,
+                       int y, enum PixelFormat target)
+{
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int j;
+        int Y1 = 1 << 18;
+        int Y2 = 1 << 18;
+        int U  = 1 << 18;
+        int V  = 1 << 18;
+        const uint8_t *r, *g, *b;
+
+        for (j = 0; j < lumFilterSize; j++) {
+            Y1 += lumSrc[j][i * 2]     * lumFilter[j];
+            Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
+        }
+        for (j = 0; j < chrFilterSize; j++) {
+            U += chrUSrc[j][i] * chrFilter[j];
+            V += chrVSrc[j][i] * chrFilter[j];
+        }
+        Y1 >>= 19;
+        Y2 >>= 19;
+        U  >>= 19;
+        V  >>= 19;
+        if ((Y1 | Y2 | U | V) & 0x100) {
+            Y1 = av_clip_uint8(Y1);
+            Y2 = av_clip_uint8(Y2);
+            U  = av_clip_uint8(U);
+            V  = av_clip_uint8(V);
+        }
+
+        /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/
+        r = (const uint8_t *) c->table_rV[V];
+        g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]);
+        b = (const uint8_t *) c->table_bU[U];
+
+        dest[ 0] = dest[ 1] = r_b[Y1];
+        dest[ 2] = dest[ 3] =   g[Y1];
+        dest[ 4] = dest[ 5] = b_r[Y1];
+        dest[ 6] = dest[ 7] = r_b[Y2];
+        dest[ 8] = dest[ 9] =   g[Y2];
+        dest[10] = dest[11] = b_r[Y2];
+        dest += 12;
+    }
+}
+
+static av_always_inline void
+yuv2rgb48_2_c_template(SwsContext *c, const uint16_t *buf0,
+                       const uint16_t *buf1, const uint16_t *ubuf0,
+                       const uint16_t *ubuf1, const uint16_t *vbuf0,
+                       const uint16_t *vbuf1, const uint16_t *abuf0,
+                       const uint16_t *abuf1, uint8_t *dest, int dstW,
+                       int yalpha, int uvalpha, int y,
+                       enum PixelFormat target)
+{
+    int  yalpha1 = 4095 - yalpha;
+    int uvalpha1 = 4095 - uvalpha;
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int Y1 = (buf0[i * 2]     * yalpha1  + buf1[i * 2]     * yalpha)  >> 19;
+        int Y2 = (buf0[i * 2 + 1] * yalpha1  + buf1[i * 2 + 1] * yalpha)  >> 19;
+        int U  = (ubuf0[i]        * uvalpha1 + ubuf1[i]        * uvalpha) >> 19;
+        int V  = (vbuf0[i]        * uvalpha1 + vbuf1[i]        * uvalpha) >> 19;
+        const uint8_t *r = (const uint8_t *) c->table_rV[V],
+                      *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]),
+                      *b = (const uint8_t *) c->table_bU[U];
+
+        dest[ 0] = dest[ 1] = r_b[Y1];
+        dest[ 2] = dest[ 3] =   g[Y1];
+        dest[ 4] = dest[ 5] = b_r[Y1];
+        dest[ 6] = dest[ 7] = r_b[Y2];
+        dest[ 8] = dest[ 9] =   g[Y2];
+        dest[10] = dest[11] = b_r[Y2];
+        dest += 12;
+    }
+}
+
+static av_always_inline void
+yuv2rgb48_1_c_template(SwsContext *c, const uint16_t *buf0,
+                       const uint16_t *ubuf0, const uint16_t *ubuf1,
+                       const uint16_t *vbuf0, const uint16_t *vbuf1,
+                       const uint16_t *abuf0, uint8_t *dest, int dstW,
+                       int uvalpha, enum PixelFormat dstFormat,
+                       int flags, int y, enum PixelFormat target)
+{
+    int i;
+
+    if (uvalpha < 2048) {
+        for (i = 0; i < (dstW >> 1); i++) {
+            int Y1 = buf0[i * 2]     >> 7;
+            int Y2 = buf0[i * 2 + 1] >> 7;
+            int U  = ubuf1[i]        >> 7;
+            int V  = vbuf1[i]        >> 7;
+            const uint8_t *r = (const uint8_t *) c->table_rV[V],
+                          *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]),
+                          *b = (const uint8_t *) c->table_bU[U];
+
+            dest[ 0] = dest[ 1] = r_b[Y1];
+            dest[ 2] = dest[ 3] =   g[Y1];
+            dest[ 4] = dest[ 5] = b_r[Y1];
+            dest[ 6] = dest[ 7] = r_b[Y2];
+            dest[ 8] = dest[ 9] =   g[Y2];
+            dest[10] = dest[11] = b_r[Y2];
+            dest += 12;
+        }
+    } else {
+        for (i = 0; i < (dstW >> 1); i++) {
+            int Y1 =  buf0[i * 2]          >> 7;
+            int Y2 =  buf0[i * 2 + 1]      >> 7;
+            int U  = (ubuf0[i] + ubuf1[i]) >> 8;
+            int V  = (vbuf0[i] + vbuf1[i]) >> 8;
+            const uint8_t *r = (const uint8_t *) c->table_rV[V],
+                          *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]),
+                          *b = (const uint8_t *) c->table_bU[U];
+
+            dest[ 0] = dest[ 1] = r_b[Y1];
+            dest[ 2] = dest[ 3] =   g[Y1];
+            dest[ 4] = dest[ 5] = b_r[Y1];
+            dest[ 6] = dest[ 7] = r_b[Y2];
+            dest[ 8] = dest[ 9] =   g[Y2];
+            dest[10] = dest[11] = b_r[Y2];
+            dest += 12;
+        }
+    }
 }
 
-YUV2PACKEDWRAPPER(yuv2422, yuyv, PIX_FMT_YUYV422);
-YUV2PACKEDWRAPPER(yuv2422, uyvy, PIX_FMT_UYVY422);
+#undef r_b
+#undef b_r
+
+YUV2PACKEDWRAPPER(yuv2, rgb48, rgb48be, PIX_FMT_RGB48BE);
+//YUV2PACKEDWRAPPER(yuv2, rgb48, rgb48le, PIX_FMT_RGB48LE);
+YUV2PACKEDWRAPPER(yuv2, rgb48, bgr48be, PIX_FMT_BGR48BE);
+//YUV2PACKEDWRAPPER(yuv2, rgb48, bgr48le, PIX_FMT_BGR48LE);
 
 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
     for (i=0; i<(dstW>>1); i++) {\
@@ -974,36 +1123,6 @@ YUV2PACKEDWRAPPER(yuv2422, uyvy, PIX_FMT_UYVY422);
 
 #define YSCALE_YUV_2_ANYRGB_C(func)\
     switch(c->dstFormat) {\
-    case PIX_FMT_RGB48BE:\
-    case PIX_FMT_RGB48LE:\
-        func(uint8_t,0)\
-            ((uint8_t*)dest)[ 0]= r[Y1];\
-            ((uint8_t*)dest)[ 1]= r[Y1];\
-            ((uint8_t*)dest)[ 2]= g[Y1];\
-            ((uint8_t*)dest)[ 3]= g[Y1];\
-            ((uint8_t*)dest)[ 4]= b[Y1];\
-            ((uint8_t*)dest)[ 5]= b[Y1];\
-            ((uint8_t*)dest)[ 6]= r[Y2];\
-            ((uint8_t*)dest)[ 7]= r[Y2];\
-            ((uint8_t*)dest)[ 8]= g[Y2];\
-            ((uint8_t*)dest)[ 9]= g[Y2];\
-            ((uint8_t*)dest)[10]= b[Y2];\
-            ((uint8_t*)dest)[11]= b[Y2];\
-            dest+=12;\
-        }\
-        break;\
-    case PIX_FMT_BGR48BE:\
-    case PIX_FMT_BGR48LE:\
-        func(uint8_t,0)\
-            ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\
-            ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\
-            ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\
-            ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\
-            ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\
-            ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\
-            dest+=12;\
-        }\
-        break;\
     case PIX_FMT_RGBA:\
     case PIX_FMT_BGRA:\
         if (CONFIG_SMALL) {\
@@ -1294,19 +1413,21 @@ static av_always_inline void fillPlane(uint8_t* plane, int stride,
     }
 }
 
+#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos))
+
+#define r ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? b_r : r_b)
+#define b ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? r_b : b_r)
+
 static av_always_inline void
 rgb48ToY_c_template(int16_t *dst, const uint16_t *src, int width,
                     enum PixelFormat origin)
 {
     int i;
     for (i = 0; i < width; i++) {
-#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos))
-        int a = input_pixel(&src[i*3+0]);
-        int g = input_pixel(&src[i*3+1]);
-        int c = input_pixel(&src[i*3+2]);
+        int r_b = input_pixel(&src[i*3+0]);
+        int   g = input_pixel(&src[i*3+1]);
+        int b_r = input_pixel(&src[i*3+2]);
 
-#define r ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? c : a)
-#define b ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? a : c)
         dst[i] = (RY*r + GY*g + BY*b + (32<<(RGB2YUV_SHIFT-1+8)) + (1<<(RGB2YUV_SHIFT-7+8))) >> (RGB2YUV_SHIFT-6+8);
     }
 }
@@ -1319,9 +1440,9 @@ rgb48ToUV_c_template(int16_t *dstU, int16_t *dstV,
     int i;
     assert(src1==src2);
     for (i = 0; i < width; i++) {
-        int a = input_pixel(&src1[3*i + 0]);
-        int g = input_pixel(&src1[3*i + 1]);
-        int c = input_pixel(&src1[3*i + 2]);
+        int r_b = input_pixel(&src1[i*3+0]);
+        int   g = input_pixel(&src1[i*3+1]);
+        int b_r = input_pixel(&src1[i*3+2]);
 
         dstU[i] = (RU*r + GU*g + BU*b + (256<<(RGB2YUV_SHIFT-1+8)) + (1<<(RGB2YUV_SHIFT-7+8))) >> (RGB2YUV_SHIFT-6+8);
         dstV[i] = (RV*r + GV*g + BV*b + (256<<(RGB2YUV_SHIFT-1+8)) + (1<<(RGB2YUV_SHIFT-7+8))) >> (RGB2YUV_SHIFT-6+8);
@@ -1336,17 +1457,18 @@ rgb48ToUV_half_c_template(int16_t *dstU, int16_t *dstV,
     int i;
     assert(src1==src2);
     for (i = 0; i < width; i++) {
-        int a = (input_pixel(&src1[6*i + 0])) + (input_pixel(&src1[6*i + 3]));
-        int g = (input_pixel(&src1[6*i + 1])) + (input_pixel(&src1[6*i + 4]));
-        int c = (input_pixel(&src1[6*i + 2])) + (input_pixel(&src1[6*i + 5]));
+        int r_b = (input_pixel(&src1[6*i + 0])) + (input_pixel(&src1[6*i + 3]));
+        int   g = (input_pixel(&src1[6*i + 1])) + (input_pixel(&src1[6*i + 4]));
+        int b_r = (input_pixel(&src1[6*i + 2])) + (input_pixel(&src1[6*i + 5]));
 
         dstU[i]= (RU*r + GU*g + BU*b + (256U<<(RGB2YUV_SHIFT+8)) + (1<<(RGB2YUV_SHIFT-6+8))) >> (RGB2YUV_SHIFT-5+8);
         dstV[i]= (RV*r + GV*g + BV*b + (256U<<(RGB2YUV_SHIFT+8)) + (1<<(RGB2YUV_SHIFT-6+8))) >> (RGB2YUV_SHIFT-5+8);
     }
+}
+
 #undef r
 #undef b
 #undef input_pixel
-}
 
 #define rgb48funcs(pattern, BE_LE, origin) \
 static void pattern ## 48 ## BE_LE ## ToY_c(uint8_t *dst, const uint8_t *src, \
@@ -1374,6 +1496,10 @@ rgb48funcs(rgb, BE, PIX_FMT_RGB48BE);
 rgb48funcs(bgr, LE, PIX_FMT_BGR48LE);
 rgb48funcs(bgr, BE, PIX_FMT_BGR48BE);
 
+#define input_pixel(i) ((origin == PIX_FMT_RGBA || origin == PIX_FMT_BGRA || \
+                         origin == PIX_FMT_ARGB || origin == PIX_FMT_ABGR) ? AV_RN32A(&src[(i)*4]) : \
+                        (isBE(origin) ? AV_RB16(&src[(i)*2]) : AV_RL16(&src[(i)*2])))
+
 static av_always_inline void
 rgb16_32ToY_c_template(int16_t *dst, const uint8_t *src,
                        int width, enum PixelFormat origin,
@@ -1386,9 +1512,6 @@ rgb16_32ToY_c_template(int16_t *dst, const uint8_t *src,
     int i;
 
     for (i = 0; i < width; i++) {
-#define input_pixel(i) ((origin == PIX_FMT_RGBA || origin == PIX_FMT_BGRA || \
-                         origin == PIX_FMT_ARGB || origin == PIX_FMT_ABGR) ? AV_RN32A(&src[(i)*4]) : \
-                        (isBE(origin) ? AV_RB16(&src[(i)*2]) : AV_RL16(&src[(i)*2])))
         int px = input_pixel(i) >> shp;
         int b = (px & maskb) >> shb;
         int g = (px & maskg) >> shg;
@@ -1454,9 +1577,10 @@ rgb16_32ToUV_half_c_template(int16_t *dstU, int16_t *dstV,
         dstU[i] = (ru * r + gu * g + bu * b + (unsigned)rnd) >> ((S)-6+1);
         dstV[i] = (rv * r + gv * g + bv * b + (unsigned)rnd) >> ((S)-6+1);
     }
-#undef input_pixel
 }
 
+#undef input_pixel
+
 #define rgb16_32_wrapper(fmt, name, shr, shg, shb, shp, maskr, \
                          maskg, maskb, rsh, gsh, bsh, S) \
 static void name ## ToY_c(uint8_t *dst, const uint8_t *src, \
@@ -1637,6 +1761,22 @@ static av_always_inline void nvXXtoUV_c(uint8_t *dst1, uint8_t *dst2,
     }
 }
 
+static void nv12ToUV_c(uint8_t *dstU, uint8_t *dstV,
+                       const uint8_t *src1, const uint8_t *src2,
+                       int width, uint32_t *unused)
+{
+    nvXXtoUV_c(dstU, dstV, src1, width);
+}
+
+static void nv21ToUV_c(uint8_t *dstU, uint8_t *dstV,
+                       const uint8_t *src1, const uint8_t *src2,
+                       int width, uint32_t *unused)
+{
+    nvXXtoUV_c(dstV, dstU, src1, width);
+}
+
+#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos))
+
 // FIXME Maybe dither instead.
 static av_always_inline void
 yuv9_OR_10ToUV_c_template(uint8_t *dstU, uint8_t *dstV,
@@ -1647,7 +1787,6 @@ yuv9_OR_10ToUV_c_template(uint8_t *dstU, uint8_t *dstV,
     const uint16_t *srcU = (const uint16_t *) _srcU;
     const uint16_t *srcV = (const uint16_t *) _srcV;
 
-#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos))
     for (i = 0; i < width; i++) {
         dstU[i] = input_pixel(&srcU[i]) >> (depth - 8);
         dstV[i] = input_pixel(&srcV[i]) >> (depth - 8);
@@ -1663,9 +1802,10 @@ yuv9_or_10ToY_c_template(uint8_t *dstY, const uint8_t *_srcY,
 
     for (i = 0; i < width; i++)
         dstY[i] = input_pixel(&srcY[i]) >> (depth - 8);
-#undef input_pixel
 }
 
+#undef input_pixel
+
 #define YUV_NBPS(depth, BE_LE, origin) \
 static void BE_LE ## depth ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
                                      const uint8_t *srcU, const uint8_t *srcV, \
@@ -1684,20 +1824,6 @@ YUV_NBPS( 9, BE, PIX_FMT_YUV420P9BE);
 YUV_NBPS(10, LE, PIX_FMT_YUV420P10LE);
 YUV_NBPS(10, BE, PIX_FMT_YUV420P10BE);
 
-static inline void nv12ToUV_c(uint8_t *dstU, uint8_t *dstV,
-                              const uint8_t *src1, const uint8_t *src2,
-                              int width, uint32_t *unused)
-{
-    nvXXtoUV_c(dstU, dstV, src1, width);
-}
-
-static inline void nv21ToUV_c(uint8_t *dstU, uint8_t *dstV,
-                              const uint8_t *src1, const uint8_t *src2,
-                              int width, uint32_t *unused)
-{
-    nvXXtoUV_c(dstV, dstU, src1, width);
-}
-
 static void bgr24ToY_c(int16_t *dst, const uint8_t *src,
                        int width, uint32_t *unused)
 {
@@ -2003,14 +2129,34 @@ find_c_packed_planar_out_funcs(SwsContext *c,
             *yuv2packedX = yuv2monoblack_X_c;
             break;
         case PIX_FMT_YUYV422:
-            *yuv2packed1 = yuv2422yuyv_1_c;
-            *yuv2packed2 = yuv2422yuyv_2_c;
-            *yuv2packedX = yuv2422yuyv_X_c;
+            *yuv2packed1 = yuv2yuyv422_1_c;
+            *yuv2packed2 = yuv2yuyv422_2_c;
+            *yuv2packedX = yuv2yuyv422_X_c;
             break;
         case PIX_FMT_UYVY422:
-            *yuv2packed1 = yuv2422uyvy_1_c;
-            *yuv2packed2 = yuv2422uyvy_2_c;
-            *yuv2packedX = yuv2422uyvy_X_c;
+            *yuv2packed1 = yuv2uyvy422_1_c;
+            *yuv2packed2 = yuv2uyvy422_2_c;
+            *yuv2packedX = yuv2uyvy422_X_c;
+            break;
+        case PIX_FMT_RGB48LE:
+            //*yuv2packed1 = yuv2rgb48le_1_c;
+            //*yuv2packed2 = yuv2rgb48le_2_c;
+            //*yuv2packedX = yuv2rgb48le_X_c;
+            //break;
+        case PIX_FMT_RGB48BE:
+            *yuv2packed1 = yuv2rgb48be_1_c;
+            *yuv2packed2 = yuv2rgb48be_2_c;
+            *yuv2packedX = yuv2rgb48be_X_c;
+            break;
+        case PIX_FMT_BGR48LE:
+            //*yuv2packed1 = yuv2bgr48le_1_c;
+            //*yuv2packed2 = yuv2bgr48le_2_c;
+            //*yuv2packedX = yuv2bgr48le_X_c;
+            //break;
+        case PIX_FMT_BGR48BE:
+            *yuv2packed1 = yuv2bgr48be_1_c;
+            *yuv2packed2 = yuv2bgr48be_2_c;
+            *yuv2packedX = yuv2bgr48be_X_c;
             break;
         default:
             *yuv2packed1 = yuv2packed1_c;