vc1dsp: introduce cases for 8x8 and 16x16

This allows further unrolling the DSP implementation where possible.

x86 and ARM DSP modified by simply moving the multiple calls from vc1dec
to the DSP code. Decoding improvements should only occurs because of the
compiler actually able to unroll more.

Decoding time: ~8.80s -> 8.64s (ie around 2%)

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

1 files changed, 108 insertions, 33 deletions

diff --git a/libavcodec/vc1dsp.c b/libavcodec/vc1dsp.c
index c37b956..acc3aea 100644
--- a/libavcodec/vc1dsp.c
+++ b/libavcodec/vc1dsp.c
@@ -643,6 +643,64 @@ static av_always_inline void OPNAME ## vc1_mspel_mc(uint8_t *dst,             \
         src += stride;                                                        \
     }                                                                         \
 }\
+static av_always_inline void OPNAME ## vc1_mspel_mc_16(uint8_t *dst,          \
+                                                       const uint8_t *src,    \
+                                                       ptrdiff_t stride,      \
+                                                       int hmode,             \
+                                                       int vmode,             \
+                                                       int rnd)               \
+{                                                                             \
+    int i, j;                                                                 \
+                                                                              \
+    if (vmode) { /* Horizontal filter to apply */                             \
+        int r;                                                                \
+                                                                              \
+        if (hmode) { /* Vertical filter to apply, output to tmp */            \
+            static const int shift_value[] = { 0, 5, 1, 5 };                  \
+            int shift = (shift_value[hmode] + shift_value[vmode]) >> 1;       \
+            int16_t tmp[19 * 16], *tptr = tmp;                                \
+                                                                              \
+            r = (1 << (shift - 1)) + rnd - 1;                                 \
+                                                                              \
+            src -= 1;                                                         \
+            for (j = 0; j < 16; j++) {                                        \
+                for (i = 0; i < 19; i++)                                      \
+                    tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode) + r) >> shift; \
+                src  += stride;                                               \
+                tptr += 19;                                                   \
+            }                                                                 \
+                                                                              \
+            r    = 64 - rnd;                                                  \
+            tptr = tmp + 1;                                                   \
+            for (j = 0; j < 16; j++) {                                        \
+                for (i = 0; i < 16; i++)                                      \
+                    OP(dst[i], (vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode) + r) >> 7); \
+                dst  += stride;                                               \
+                tptr += 19;                                                   \
+            }                                                                 \
+                                                                              \
+            return;                                                           \
+        } else { /* No horizontal filter, output 8 lines to dst */            \
+            r = 1 - rnd;                                                      \
+                                                                              \
+            for (j = 0; j < 16; j++) {                                        \
+                for (i = 0; i < 16; i++)                                      \
+                    OP(dst[i], vc1_mspel_filter(src + i, stride, vmode, r));  \
+                src += stride;                                                \
+                dst += stride;                                                \
+            }                                                                 \
+            return;                                                           \
+        }                                                                     \
+    }                                                                         \
+                                                                              \
+    /* Horizontal mode with no vertical mode */                               \
+    for (j = 0; j < 16; j++) {                                                \
+        for (i = 0; i < 16; i++)                                              \
+            OP(dst[i], vc1_mspel_filter(src + i, 1, hmode, rnd));             \
+        dst += stride;                                                        \
+        src += stride;                                                        \
+    }                                                                         \
+}\
 static void OPNAME ## pixels8x8_c(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int rnd){\
     int i;\
     for(i=0; i<8; i++){\
@@ -651,6 +709,17 @@ static void OPNAME ## pixels8x8_c(uint8_t *block, const uint8_t *pixels, ptrdiff
         pixels+=line_size;\
         block +=line_size;\
     }\
+}\
+static void OPNAME ## pixels16x16_c(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int rnd){\
+    int i;\
+    for(i=0; i<16; i++){\
+        OP4(*(uint32_t*)(block   ), AV_RN32(pixels   ));\
+        OP4(*(uint32_t*)(block+ 4), AV_RN32(pixels+ 4));\
+        OP4(*(uint32_t*)(block+ 8), AV_RN32(pixels+ 8));\
+        OP4(*(uint32_t*)(block+12), AV_RN32(pixels+12));\
+        pixels+=line_size;\
+        block +=line_size;\
+    }\
 }
 
 #define op_put(a, b) a = av_clip_uint8(b)
@@ -675,6 +744,18 @@ static void avg_vc1_mspel_mc ## a ## b ## _c(uint8_t *dst,                    \
                                              ptrdiff_t stride, int rnd)       \
 {                                                                             \
     avg_vc1_mspel_mc(dst, src, stride, a, b, rnd);                            \
+}                                                                             \
+static void put_vc1_mspel_mc ## a ## b ## _16_c(uint8_t *dst,                 \
+                                                const uint8_t *src,           \
+                                                ptrdiff_t stride, int rnd)    \
+{                                                                             \
+    put_vc1_mspel_mc_16(dst, src, stride, a, b, rnd);                         \
+}                                                                             \
+static void avg_vc1_mspel_mc ## a ## b ## _16_c(uint8_t *dst,                 \
+                                                const uint8_t *src,           \
+                                                ptrdiff_t stride, int rnd)    \
+{                                                                             \
+    avg_vc1_mspel_mc_16(dst, src, stride, a, b, rnd);                         \
 }
 
 PUT_VC1_MSPEL(1, 0)
@@ -878,6 +959,11 @@ static void sprite_v_double_twoscale_c(uint8_t *dst,
 }
 
 #endif /* CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER */
+#define FN_ASSIGN(X, Y) \
+    dsp->put_vc1_mspel_pixels_tab[1][X+4*Y] = put_vc1_mspel_mc##X##Y##_c; \
+    dsp->put_vc1_mspel_pixels_tab[0][X+4*Y] = put_vc1_mspel_mc##X##Y##_16_c; \
+    dsp->avg_vc1_mspel_pixels_tab[1][X+4*Y] = avg_vc1_mspel_mc##X##Y##_c; \
+    dsp->avg_vc1_mspel_pixels_tab[0][X+4*Y] = avg_vc1_mspel_mc##X##Y##_16_c
 
 av_cold void ff_vc1dsp_init(VC1DSPContext *dsp)
 {
@@ -902,39 +988,28 @@ av_cold void ff_vc1dsp_init(VC1DSPContext *dsp)
     dsp->vc1_v_loop_filter16  = vc1_v_loop_filter16_c;
     dsp->vc1_h_loop_filter16  = vc1_h_loop_filter16_c;
 
-    dsp->put_vc1_mspel_pixels_tab[0]  = put_pixels8x8_c;
-    dsp->put_vc1_mspel_pixels_tab[1]  = put_vc1_mspel_mc10_c;
-    dsp->put_vc1_mspel_pixels_tab[2]  = put_vc1_mspel_mc20_c;
-    dsp->put_vc1_mspel_pixels_tab[3]  = put_vc1_mspel_mc30_c;
-    dsp->put_vc1_mspel_pixels_tab[4]  = put_vc1_mspel_mc01_c;
-    dsp->put_vc1_mspel_pixels_tab[5]  = put_vc1_mspel_mc11_c;
-    dsp->put_vc1_mspel_pixels_tab[6]  = put_vc1_mspel_mc21_c;
-    dsp->put_vc1_mspel_pixels_tab[7]  = put_vc1_mspel_mc31_c;
-    dsp->put_vc1_mspel_pixels_tab[8]  = put_vc1_mspel_mc02_c;
-    dsp->put_vc1_mspel_pixels_tab[9]  = put_vc1_mspel_mc12_c;
-    dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c;
-    dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c;
-    dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c;
-    dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_c;
-    dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c;
-    dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c;
-
-    dsp->avg_vc1_mspel_pixels_tab[0]  = avg_pixels8x8_c;
-    dsp->avg_vc1_mspel_pixels_tab[1]  = avg_vc1_mspel_mc10_c;
-    dsp->avg_vc1_mspel_pixels_tab[2]  = avg_vc1_mspel_mc20_c;
-    dsp->avg_vc1_mspel_pixels_tab[3]  = avg_vc1_mspel_mc30_c;
-    dsp->avg_vc1_mspel_pixels_tab[4]  = avg_vc1_mspel_mc01_c;
-    dsp->avg_vc1_mspel_pixels_tab[5]  = avg_vc1_mspel_mc11_c;
-    dsp->avg_vc1_mspel_pixels_tab[6]  = avg_vc1_mspel_mc21_c;
-    dsp->avg_vc1_mspel_pixels_tab[7]  = avg_vc1_mspel_mc31_c;
-    dsp->avg_vc1_mspel_pixels_tab[8]  = avg_vc1_mspel_mc02_c;
-    dsp->avg_vc1_mspel_pixels_tab[9]  = avg_vc1_mspel_mc12_c;
-    dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_c;
-    dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_c;
-    dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_c;
-    dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_c;
-    dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_c;
-    dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_c;
+    dsp->put_vc1_mspel_pixels_tab[0][0] = put_pixels16x16_c;
+    dsp->avg_vc1_mspel_pixels_tab[0][0] = avg_pixels16x16_c;
+    dsp->put_vc1_mspel_pixels_tab[1][0] = put_pixels8x8_c;
+    dsp->avg_vc1_mspel_pixels_tab[1][0] = avg_pixels8x8_c;
+    FN_ASSIGN(0, 1);
+    FN_ASSIGN(0, 2);
+    FN_ASSIGN(0, 3);
+
+    FN_ASSIGN(1, 0);
+    FN_ASSIGN(1, 1);
+    FN_ASSIGN(1, 2);
+    FN_ASSIGN(1, 3);
+
+    FN_ASSIGN(2, 0);
+    FN_ASSIGN(2, 1);
+    FN_ASSIGN(2, 2);
+    FN_ASSIGN(2, 3);
+
+    FN_ASSIGN(3, 0);
+    FN_ASSIGN(3, 1);
+    FN_ASSIGN(3, 2);
+    FN_ASSIGN(3, 3);
 
     dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = put_no_rnd_vc1_chroma_mc8_c;
     dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = avg_no_rnd_vc1_chroma_mc8_c;