dsputil: template get_pixels() for different bit depths

Signed-off-by: Mans Rullgard <mans@mansr.com>

9 files changed, 39 insertions, 30 deletions

diff --git a/libavcodec/alpha/dsputil_alpha.c b/libavcodec/alpha/dsputil_alpha.c
index 039608b..3efcf09 100644
--- a/libavcodec/alpha/dsputil_alpha.c
+++ b/libavcodec/alpha/dsputil_alpha.c
@@ -321,7 +321,8 @@ void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx)
         c->put_pixels_clamped = put_pixels_clamped_mvi_asm;
         c->add_pixels_clamped = add_pixels_clamped_mvi_asm;
 
-        c->get_pixels       = get_pixels_mvi;
+        if (!high_bit_depth)
+            c->get_pixels   = get_pixels_mvi;
         c->diff_pixels      = diff_pixels_mvi;
         c->sad[0]           = pix_abs16x16_mvi_asm;
         c->sad[1]           = pix_abs8x8_mvi;
diff --git a/libavcodec/arm/dsputil_init_armv6.c b/libavcodec/arm/dsputil_init_armv6.c
index d442415..07124c2 100644
--- a/libavcodec/arm/dsputil_init_armv6.c
+++ b/libavcodec/arm/dsputil_init_armv6.c
@@ -106,8 +106,9 @@ void av_cold ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx)
     c->avg_pixels_tab[1][0] = ff_avg_pixels8_armv6;
     }
 
+    if (!high_bit_depth)
+        c->get_pixels = ff_get_pixels_armv6;
     c->add_pixels_clamped = ff_add_pixels_clamped_armv6;
-    c->get_pixels = ff_get_pixels_armv6;
     c->diff_pixels = ff_diff_pixels_armv6;
 
     c->pix_abs[0][0] = ff_pix_abs16_armv6;
diff --git a/libavcodec/bfin/dsputil_bfin.c b/libavcodec/bfin/dsputil_bfin.c
index 691c060..f88ae66 100644
--- a/libavcodec/bfin/dsputil_bfin.c
+++ b/libavcodec/bfin/dsputil_bfin.c
@@ -199,12 +199,12 @@ void dsputil_init_bfin( DSPContext* c, AVCodecContext *avctx )
 {
     const int high_bit_depth = avctx->codec_id == CODEC_ID_H264 && avctx->bits_per_raw_sample > 8;
 
-    c->get_pixels         = ff_bfin_get_pixels;
     c->diff_pixels        = ff_bfin_diff_pixels;
     c->put_pixels_clamped = ff_bfin_put_pixels_clamped;
     c->add_pixels_clamped = ff_bfin_add_pixels_clamped;
 
     if (!high_bit_depth)
+    c->get_pixels         = ff_bfin_get_pixels;
     c->clear_blocks       = bfin_clear_blocks;
     c->pix_sum            = ff_bfin_pix_sum;
     c->pix_norm1          = ff_bfin_pix_norm1;
diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index b181640..09e58f4 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -307,25 +307,6 @@ static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
     return s;
 }
 
-static void get_pixels_c(DCTELEM *restrict block, const uint8_t *pixels, int line_size)
-{
-    int i;
-
-    /* read the pixels */
-    for(i=0;i<8;i++) {
-        block[0] = pixels[0];
-        block[1] = pixels[1];
-        block[2] = pixels[2];
-        block[3] = pixels[3];
-        block[4] = pixels[4];
-        block[5] = pixels[5];
-        block[6] = pixels[6];
-        block[7] = pixels[7];
-        pixels += line_size;
-        block += 8;
-    }
-}
-
 static void diff_pixels_c(DCTELEM *restrict block, const uint8_t *s1,
                           const uint8_t *s2, int stride){
     int i;
@@ -2927,7 +2908,6 @@ av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)
         }
     }
 
-    c->get_pixels = get_pixels_c;
     c->diff_pixels = diff_pixels_c;
     c->put_pixels_clamped = ff_put_pixels_clamped_c;
     c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_c;
@@ -3160,6 +3140,7 @@ av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)
 
 
 #define BIT_DEPTH_FUNCS(depth, dct)\
+    c->get_pixels                    = FUNCC(get_pixels   ## dct   , depth);\
     c->draw_edges                    = FUNCC(draw_edges            , depth);\
     c->emulated_edge_mc              = FUNC (ff_emulated_edge_mc   , depth);\
     c->clear_block                   = FUNCC(clear_block  ## dct   , depth);\
diff --git a/libavcodec/dsputil_template.c b/libavcodec/dsputil_template.c
index 5863275..72ed6bf 100644
--- a/libavcodec/dsputil_template.c
+++ b/libavcodec/dsputil_template.c
@@ -193,6 +193,29 @@ void FUNC(ff_emulated_edge_mc)(uint8_t *buf, const uint8_t *src, int linesize, i
 }
 
 #define DCTELEM_FUNCS(dctcoef, suffix)                                  \
+static void FUNCC(get_pixels ## suffix)(DCTELEM *restrict _block,       \
+                                        const uint8_t *_pixels,         \
+                                        int line_size)                  \
+{                                                                       \
+    const pixel *pixels = (const pixel *) _pixels;                      \
+    dctcoef *restrict block = (dctcoef *) _block;                       \
+    int i;                                                              \
+                                                                        \
+    /* read the pixels */                                               \
+    for(i=0;i<8;i++) {                                                  \
+        block[0] = pixels[0];                                           \
+        block[1] = pixels[1];                                           \
+        block[2] = pixels[2];                                           \
+        block[3] = pixels[3];                                           \
+        block[4] = pixels[4];                                           \
+        block[5] = pixels[5];                                           \
+        block[6] = pixels[6];                                           \
+        block[7] = pixels[7];                                           \
+        pixels += line_size / sizeof(pixel);                            \
+        block += 8;                                                     \
+    }                                                                   \
+}                                                                       \
+                                                                        \
 static void FUNCC(add_pixels8 ## suffix)(uint8_t *restrict _pixels,     \
                                          DCTELEM *_block,               \
                                          int line_size)                 \
diff --git a/libavcodec/mlib/dsputil_mlib.c b/libavcodec/mlib/dsputil_mlib.c
index c0f2c03..b915ffe 100644
--- a/libavcodec/mlib/dsputil_mlib.c
+++ b/libavcodec/mlib/dsputil_mlib.c
@@ -423,11 +423,12 @@ void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx)
 {
     const int high_bit_depth = avctx->codec_id == CODEC_ID_H264 && avctx->bits_per_raw_sample > 8;
 
-    c->get_pixels  = get_pixels_mlib;
     c->diff_pixels = diff_pixels_mlib;
     c->add_pixels_clamped = add_pixels_clamped_mlib;
 
     if (!high_bit_depth) {
+    c->get_pixels  = get_pixels_mlib;
+
     c->put_pixels_tab[0][0] = put_pixels16_mlib;
     c->put_pixels_tab[0][1] = put_pixels16_x2_mlib;
     c->put_pixels_tab[0][2] = put_pixels16_y2_mlib;
diff --git a/libavcodec/ppc/dsputil_altivec.c b/libavcodec/ppc/dsputil_altivec.c
index 7bdd570..f052672 100644
--- a/libavcodec/ppc/dsputil_altivec.c
+++ b/libavcodec/ppc/dsputil_altivec.c
@@ -1387,11 +1387,10 @@ void dsputil_init_altivec(DSPContext* c, AVCodecContext *avctx)
     c->sse[0]= sse16_altivec;
     c->pix_sum = pix_sum_altivec;
     c->diff_pixels = diff_pixels_altivec;
-    c->get_pixels = get_pixels_altivec;
-    if (!high_bit_depth)
-    c->clear_block = clear_block_altivec;
     c->add_bytes= add_bytes_altivec;
     if (!high_bit_depth) {
+    c->get_pixels = get_pixels_altivec;
+    c->clear_block = clear_block_altivec;
     c->put_pixels_tab[0][0] = put_pixels16_altivec;
     /* the two functions do the same thing, so use the same code */
     c->put_no_rnd_pixels_tab[0][0] = put_pixels16_altivec;
diff --git a/libavcodec/ps2/dsputil_mmi.c b/libavcodec/ps2/dsputil_mmi.c
index 5857096..94a5a8c 100644
--- a/libavcodec/ps2/dsputil_mmi.c
+++ b/libavcodec/ps2/dsputil_mmi.c
@@ -152,9 +152,9 @@ void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx)
 
     c->put_pixels_tab[0][0] = put_pixels16_mmi;
     c->put_no_rnd_pixels_tab[0][0] = put_pixels16_mmi;
-    }
 
     c->get_pixels = get_pixels_mmi;
+    }
 
     if (avctx->bits_per_raw_sample <= 8 &&
         (idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_PS2)) {
diff --git a/libavcodec/x86/dsputilenc_mmx.c b/libavcodec/x86/dsputilenc_mmx.c
index ea03e92..7362234 100644
--- a/libavcodec/x86/dsputilenc_mmx.c
+++ b/libavcodec/x86/dsputilenc_mmx.c
@@ -1098,6 +1098,7 @@ static int ssd_int8_vs_int16_mmx(const int8_t *pix1, const int16_t *pix2, int si
 void dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
 {
     int mm_flags = av_get_cpu_flags();
+    int bit_depth = avctx->bits_per_raw_sample;
 
     if (mm_flags & AV_CPU_FLAG_MMX) {
         const int dct_algo = avctx->dct_algo;
@@ -1112,7 +1113,8 @@ void dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
             }
         }
 
-        c->get_pixels = get_pixels_mmx;
+        if (bit_depth <= 8)
+            c->get_pixels = get_pixels_mmx;
         c->diff_pixels = diff_pixels_mmx;
         c->pix_sum = pix_sum16_mmx;
 
@@ -1159,7 +1161,8 @@ void dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
         }
 
         if(mm_flags & AV_CPU_FLAG_SSE2){
-            c->get_pixels = get_pixels_sse2;
+            if (bit_depth <= 8)
+                c->get_pixels = get_pixels_sse2;
             c->sum_abs_dctelem= sum_abs_dctelem_sse2;
 #if HAVE_YASM && HAVE_ALIGNED_STACK
             c->hadamard8_diff[0]= ff_hadamard8_diff16_sse2;