h264: Add add_pixels4/8() to h264dsp, and remove add_pixels4 from dsputil

These functions are mostly H264-specific (the only other user I can
spot is bink), and this allows us to special-case some functionality
for H264. Also remove the 16-bit-coeff with >8bpp versions (unused)
and merge the duplicate 32-bit-coeff for >8bpp (identical).

Signed-off-by: Martin Storsjö <martin@martin.st>

8 files changed, 120 insertions, 53 deletions

diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index 30c315eab8..62fa0acbde 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -403,6 +403,26 @@ static void put_signed_pixels_clamped_c(const int16_t *block,
     }
 }
 
+static void add_pixels8_c(uint8_t *restrict pixels,
+                          int16_t *block,
+                          int line_size)
+{
+    int i;
+
+    for(i=0;i<8;i++) {
+        pixels[0] += block[0];
+        pixels[1] += block[1];
+        pixels[2] += block[2];
+        pixels[3] += block[3];
+        pixels[4] += block[4];
+        pixels[5] += block[5];
+        pixels[6] += block[6];
+        pixels[7] += block[7];
+        pixels += line_size;
+        block += 8;
+    }
+}
+
 static void add_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels,
                                  int line_size)
 {
@@ -2678,6 +2698,8 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->shrink[2]= ff_shrink44;
     c->shrink[3]= ff_shrink88;
 
+    c->add_pixels8 = add_pixels8_c;
+
 #define hpel_funcs(prefix, idx, num) \
     c->prefix ## _pixels_tab idx [0] = prefix ## _pixels ## num ## _8_c; \
     c->prefix ## _pixels_tab idx [1] = prefix ## _pixels ## num ## _x2_8_c; \
@@ -2706,8 +2728,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->draw_edges                    = FUNCC(draw_edges            , depth);\
     c->clear_block                   = FUNCC(clear_block  ## dct   , depth);\
     c->clear_blocks                  = FUNCC(clear_blocks ## dct   , depth);\
-    c->add_pixels8                   = FUNCC(add_pixels8  ## dct   , depth);\
-    c->add_pixels4                   = FUNCC(add_pixels4  ## dct   , depth);\
 
     switch (avctx->bits_per_raw_sample) {
     case 9:
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index d0ba95ceeb..e69ac32e6e 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -153,7 +153,6 @@ typedef struct DSPContext {
     void (*put_signed_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
     void (*add_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
     void (*add_pixels8)(uint8_t *pixels, int16_t *block, int line_size);
-    void (*add_pixels4)(uint8_t *pixels, int16_t *block, int line_size);
     int (*sum_abs_dctelem)(int16_t *block/*align 16*/);
     /**
      * translational global motion compensation.
diff --git a/libavcodec/dsputil_template.c b/libavcodec/dsputil_template.c
index 5244e20b84..ec7f9305ec 100644
--- a/libavcodec/dsputil_template.c
+++ b/libavcodec/dsputil_template.c
@@ -89,48 +89,6 @@ static void FUNCC(get_pixels ## suffix)(int16_t *restrict _block,       \
     }                                                                   \
 }                                                                       \
                                                                         \
-static void FUNCC(add_pixels8 ## suffix)(uint8_t *restrict _pixels,     \
-                                         int16_t *_block,               \
-                                         int line_size)                 \
-{                                                                       \
-    int i;                                                              \
-    pixel *restrict pixels = (pixel *restrict)_pixels;                  \
-    dctcoef *block = (dctcoef*)_block;                                  \
-    line_size /= sizeof(pixel);                                         \
-                                                                        \
-    for(i=0;i<8;i++) {                                                  \
-        pixels[0] += block[0];                                          \
-        pixels[1] += block[1];                                          \
-        pixels[2] += block[2];                                          \
-        pixels[3] += block[3];                                          \
-        pixels[4] += block[4];                                          \
-        pixels[5] += block[5];                                          \
-        pixels[6] += block[6];                                          \
-        pixels[7] += block[7];                                          \
-        pixels += line_size;                                            \
-        block += 8;                                                     \
-    }                                                                   \
-}                                                                       \
-                                                                        \
-static void FUNCC(add_pixels4 ## suffix)(uint8_t *restrict _pixels,     \
-                                         int16_t *_block,               \
-                                         int line_size)                 \
-{                                                                       \
-    int i;                                                              \
-    pixel *restrict pixels = (pixel *restrict)_pixels;                  \
-    dctcoef *block = (dctcoef*)_block;                                  \
-    line_size /= sizeof(pixel);                                         \
-                                                                        \
-    for(i=0;i<4;i++) {                                                  \
-        pixels[0] += block[0];                                          \
-        pixels[1] += block[1];                                          \
-        pixels[2] += block[2];                                          \
-        pixels[3] += block[3];                                          \
-        pixels += line_size;                                            \
-        block += 4;                                                     \
-    }                                                                   \
-}                                                                       \
-                                                                        \
 static void FUNCC(clear_block ## suffix)(int16_t *block)                \
 {                                                                       \
     memset(block, 0, sizeof(dctcoef)*64);                               \
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 79e4b0baf6..37cac01af3 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -2116,7 +2116,7 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
         if (IS_8x8DCT(mb_type)) {
             if (transform_bypass) {
                 idct_dc_add  =
-                idct_add     = h->dsp.add_pixels8;
+                idct_add     = h->h264dsp.h264_add_pixels8;
             } else {
                 idct_dc_add = h->h264dsp.h264_idct8_dc_add;
                 idct_add    = h->h264dsp.h264_idct8_add;
@@ -2141,7 +2141,7 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
         } else {
             if (transform_bypass) {
                 idct_dc_add  =
-                    idct_add = h->dsp.add_pixels4;
+                idct_add     = h->h264dsp.h264_add_pixels4;
             } else {
                 idct_dc_add = h->h264dsp.h264_idct_dc_add;
                 idct_add    = h->h264dsp.h264_idct_add;
@@ -2238,9 +2238,9 @@ static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, int mb_type,
                         for (i = 0; i < 16; i++)
                             if (h->non_zero_count_cache[scan8[i + p * 16]] ||
                                 dctcoef_get(h->mb, pixel_shift, i * 16 + p * 256))
-                                h->dsp.add_pixels4(dest_y + block_offset[i],
-                                                   h->mb + (i * 16 + p * 256 << pixel_shift),
-                                                   linesize);
+                                h->h264dsp.h264_add_pixels4(dest_y + block_offset[i],
+                                                            h->mb + (i * 16 + p * 256 << pixel_shift),
+                                                            linesize);
                     }
                 } else {
                     h->h264dsp.h264_idct_add16intra(dest_y, block_offset,
@@ -2251,8 +2251,8 @@ static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, int mb_type,
             } else if (h->cbp & 15) {
                 if (transform_bypass) {
                     const int di = IS_8x8DCT(mb_type) ? 4 : 1;
-                    idct_add = IS_8x8DCT(mb_type) ? h->dsp.add_pixels8
-                                                  : h->dsp.add_pixels4;
+                    idct_add = IS_8x8DCT(mb_type) ? h->h264dsp.h264_add_pixels8
+                                                  : h->h264dsp.h264_add_pixels4;
                     for (i = 0; i < 16; i += di)
                         if (h->non_zero_count_cache[scan8[i + p * 16]])
                             idct_add(dest_y + block_offset[i],
diff --git a/libavcodec/h264_mb_template.c b/libavcodec/h264_mb_template.c
index 7ff7edd2cd..fdefed4304 100644
--- a/libavcodec/h264_mb_template.c
+++ b/libavcodec/h264_mb_template.c
@@ -207,7 +207,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
                                                             h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
                                                             uvlinesize);
                 } else {
-                    idct_add = h->dsp.add_pixels4;
+                    idct_add = h->h264dsp.h264_add_pixels4;
                     for (j = 1; j < 3; j++) {
                         for (i = j * 16; i < j * 16 + 4; i++)
                             if (h->non_zero_count_cache[scan8[i]] ||
diff --git a/libavcodec/h264addpx_template.c b/libavcodec/h264addpx_template.c
new file mode 100644
index 0000000000..d1babc54e5
--- /dev/null
+++ b/libavcodec/h264addpx_template.c
@@ -0,0 +1,68 @@
+/*
+ * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
+ * Copyright (c) 2003-2011 Michael Niedermayer <michaelni@gmx.at>
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * H.264 / AVC / MPEG4 part10 DSP functions.
+ * @author Michael Niedermayer <michaelni@gmx.at>
+ */
+
+#include "bit_depth_template.c"
+
+static void FUNCC(ff_h264_add_pixels4)(uint8_t *_dst, int16_t *_src, int stride)
+{
+    int i;
+    pixel *dst = (pixel *) _dst;
+    dctcoef *src = (dctcoef *) _src;
+    stride /= sizeof(pixel);
+
+    for (i = 0; i < 4; i++) {
+        dst[0] += src[0];
+        dst[1] += src[1];
+        dst[2] += src[2];
+        dst[3] += src[3];
+
+        dst += stride;
+        src += 4;
+    }
+}
+
+static void FUNCC(ff_h264_add_pixels8)(uint8_t *_dst, int16_t *_src, int stride)
+{
+    int i;
+    pixel *dst = (pixel *) _dst;
+    dctcoef *src = (dctcoef *) _src;
+    stride /= sizeof(pixel);
+
+    for (i = 0; i < 8; i++) {
+        dst[0] += src[0];
+        dst[1] += src[1];
+        dst[2] += src[2];
+        dst[3] += src[3];
+        dst[4] += src[4];
+        dst[5] += src[5];
+        dst[6] += src[6];
+        dst[7] += src[7];
+
+        dst += stride;
+        src += 8;
+    }
+}
diff --git a/libavcodec/h264dsp.c b/libavcodec/h264dsp.c
index d29c685008..b739758e24 100644
--- a/libavcodec/h264dsp.c
+++ b/libavcodec/h264dsp.c
@@ -43,11 +43,29 @@
 #include "h264dsp_template.c"
 #undef BIT_DEPTH
 
+#define BIT_DEPTH 8
+#include "h264addpx_template.c"
+#undef BIT_DEPTH
+
+#define BIT_DEPTH 16
+#include "h264addpx_template.c"
+#undef BIT_DEPTH
+
 void ff_h264dsp_init(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
 {
 #undef FUNC
 #define FUNC(a, depth) a ## _ ## depth ## _c
 
+#define ADDPX_DSP(depth) \
+    c->h264_add_pixels4 = FUNC(ff_h264_add_pixels4, depth);\
+    c->h264_add_pixels8 = FUNC(ff_h264_add_pixels8, depth)
+
+    if (bit_depth > 8 && bit_depth <= 16) {
+        ADDPX_DSP(16);
+    } else {
+        ADDPX_DSP(8);
+    }
+
 #define H264_DSP(depth) \
     c->h264_idct_add= FUNC(ff_h264_idct_add, depth);\
     c->h264_idct8_add= FUNC(ff_h264_idct8_add, depth);\
diff --git a/libavcodec/h264dsp.h b/libavcodec/h264dsp.h
index c6bcd9e214..757caf8162 100644
--- a/libavcodec/h264dsp.h
+++ b/libavcodec/h264dsp.h
@@ -101,6 +101,10 @@ typedef struct H264DSPContext {
     void (*h264_luma_dc_dequant_idct)(int16_t *output,
                                       int16_t *input /*align 16*/, int qmul);
     void (*h264_chroma_dc_dequant_idct)(int16_t *block, int qmul);
+
+    /* bypass-transform */
+    void (*h264_add_pixels8)(uint8_t *dst, int16_t *block, int stride);
+    void (*h264_add_pixels4)(uint8_t *dst, int16_t *block, int stride);
 } H264DSPContext;
 
 void ff_h264dsp_init(H264DSPContext *c, const int bit_depth,