H.264 idct functions that include the chroma, inter luma and intra16 luma loops

thus avoiding the calling overhead.
New functions are not yet used.

Originally committed as revision 16206 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/dsputil.c

@@ -4281,6 +4281,10 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
         c->h264_idct8_add= ff_h264_idct8_add_c;
         c->h264_idct_dc_add= ff_h264_idct_dc_add_c;
         c->h264_idct8_dc_add= ff_h264_idct8_dc_add_c;
+        c->h264_idct_add16     = ff_h264_idct_add16_c;
+        c->h264_idct8_add4     = ff_h264_idct8_add4_c;
+        c->h264_idct_add8      = ff_h264_idct_add8_c;
+        c->h264_idct_add16intra= ff_h264_idct_add16intra_c;
     }
 
     c->get_pixels = get_pixels_c;

libavcodec/dsputil.h

@@ -60,6 +60,10 @@ void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
 void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
 void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block);
 void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block);
+void ff_h264_idct_add16_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add16intra_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct8_add4_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add8_c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
 
 void ff_vector_fmul_add_add_c(float *dst, const float *src0, const float *src1,
                               const float *src2, int src3, int blocksize, int step);
@@ -441,6 +445,10 @@ typedef struct DSPContext {
     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_dct)(DCTELEM block[4][4]);
+    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]);
 
     /* snow wavelet */
     void (*vertical_compose97i)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);

libavcodec/h264idct.c

@@ -165,3 +165,55 @@ void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride){
         dst += stride;
     }
 }
+
+//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,
+};
+
+void ff_h264_idct_add16_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i++){
+        int nnz = nnzc[ scan8[i] ];
+        if(nnz){
+            if(nnz==1 && block[i*16]) ff_h264_idct_dc_add_c(dst + block_offset[i], block + i*16, stride);
+            else                      idct_internal        (dst + block_offset[i], block + i*16, stride, 4, 6, 1);
+        }
+    }
+}
+
+void ff_h264_idct_add16intra_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i++){
+        if(nnzc[ scan8[i] ]) idct_internal        (dst + block_offset[i], block + i*16, stride, 4, 6, 1);
+        else if(block[i*16]) ff_h264_idct_dc_add_c(dst + block_offset[i], block + i*16, stride);
+    }
+}
+
+void ff_h264_idct8_add4_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i+=4){
+        int nnz = nnzc[ scan8[i] ];
+        if(nnz){
+            if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_c(dst + block_offset[i], block + i*16, stride);
+            else                      ff_h264_idct8_add_c   (dst + block_offset[i], block + i*16, stride);
+        }
+    }
+}
+
+void ff_h264_idct_add8_c(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_c   (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
+        else if(block[i*16])
+            ff_h264_idct_dc_add_c(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
+    }
+}

libavcodec/i386/dsputil_mmx.c

@@ -2629,8 +2629,11 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
         c->h264_idct_add= ff_h264_idct_add_mmx;
         c->h264_idct8_dc_add=
         c->h264_idct8_add= ff_h264_idct8_add_mmx;
-        if (mm_flags & FF_MM_SSE2)
-            c->h264_idct8_add= ff_h264_idct8_add_sse2;
+
+        c->h264_idct_add16     = ff_h264_idct_add16_mmx;
+        c->h264_idct8_add4     = ff_h264_idct8_add4_mmx;
+        c->h264_idct_add8      = ff_h264_idct_add8_mmx;
+        c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx;
 
         if (mm_flags & FF_MM_MMXEXT) {
             c->prefetch = prefetch_mmx2;
@@ -2651,6 +2654,10 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
 
             c->h264_idct_dc_add= ff_h264_idct_dc_add_mmx2;
             c->h264_idct8_dc_add= ff_h264_idct8_dc_add_mmx2;
+            c->h264_idct_add16     = ff_h264_idct_add16_mmx2;
+            c->h264_idct8_add4     = ff_h264_idct8_add4_mmx2;
+            c->h264_idct_add8      = ff_h264_idct_add8_mmx2;
+            c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx2;
 
             if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
                 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmx2;
@@ -2807,6 +2814,9 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             H264_QPEL_FUNCS(0, 0, sse2);
         }
         if(mm_flags & FF_MM_SSE2){
+            c->h264_idct8_add = ff_h264_idct8_add_sse2;
+            c->h264_idct8_add4= ff_h264_idct8_add4_sse2;
+
             H264_QPEL_FUNCS(0, 1, sse2);
             H264_QPEL_FUNCS(0, 2, sse2);
             H264_QPEL_FUNCS(0, 3, sse2);

libavcodec/i386/h264dsp_mmx.c

@@ -376,6 +376,101 @@ static void ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
     }
 }
 
+//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 void ff_h264_idct_add16_mmx(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i++){
+        if(nnzc[ scan8[i] ])
+            ff_h264_idct_add_mmx(dst + block_offset[i], block + i*16, stride);
+    }
+}
+
+static void ff_h264_idct8_add4_mmx(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i+=4){
+        if(nnzc[ scan8[i] ])
+            ff_h264_idct8_add_mmx(dst + block_offset[i], block + i*16, stride);
+    }
+}
+
+
+static void ff_h264_idct_add16_mmx2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i++){
+        int nnz = nnzc[ scan8[i] ];
+        if(nnz){
+            if(nnz==1 && block[i*16]) ff_h264_idct_dc_add_mmx2(dst + block_offset[i], block + i*16, stride);
+            else                      ff_h264_idct_add_mmx    (dst + block_offset[i], block + i*16, stride);
+        }
+    }
+}
+
+static void ff_h264_idct_add16intra_mmx(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i++){
+        if(nnzc[ scan8[i] ] || block[i*16])
+            ff_h264_idct_add_mmx(dst + block_offset[i], block + i*16, stride);
+    }
+}
+
+static void ff_h264_idct_add16intra_mmx2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i++){
+        if(nnzc[ scan8[i] ]) ff_h264_idct_add_mmx    (dst + block_offset[i], block + i*16, stride);
+        else if(block[i*16]) ff_h264_idct_dc_add_mmx2(dst + block_offset[i], block + i*16, stride);
+    }
+}
+
+static void ff_h264_idct8_add4_mmx2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i+=4){
+        int nnz = nnzc[ scan8[i] ];
+        if(nnz){
+            if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_mmx2(dst + block_offset[i], block + i*16, stride);
+            else                      ff_h264_idct8_add_mmx    (dst + block_offset[i], block + i*16, stride);
+        }
+    }
+}
+
+static void ff_h264_idct8_add4_sse2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
+    int i;
+    for(i=0; i<16; i+=4){
+        int nnz = nnzc[ scan8[i] ];
+        if(nnz){
+            if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_mmx2(dst + block_offset[i], block + i*16, stride);
+            else                      ff_h264_idct8_add_sse2   (dst + block_offset[i], block + i*16, stride);
+        }
+    }
+}
+
+static void ff_h264_idct_add8_mmx(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] ] || block[i*16])
+            ff_h264_idct_add_mmx    (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
+    }
+}
+
+static void ff_h264_idct_add8_mmx2(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_mmx    (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
+        else if(block[i*16])
+            ff_h264_idct_dc_add_mmx2(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
+    }
+}
 
 /***********************************/
 /* deblocking */