H.264: Add x86 assembly for 10-bit H.264 predict functions

Mainly ported from 8-bit H.264 predict.

Some code ported from x264. LGPL ok by author.
Signed-off-by: NRonald S. Bultje <rsbultje@gmail.com>

libavcodec/x86/dsputil_mmx.c

@@ -42,7 +42,7 @@ DECLARE_ALIGNED(8,  const uint64_t, ff_wtwo) = 0x0002000200020002ULL;
 DECLARE_ALIGNED(16, const uint64_t, ff_pdw_80000000)[2] =
 {0x8000000080000000ULL, 0x8000000080000000ULL};
 
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_1  ) = 0x0001000100010001ULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_1  ) = {0x0001000100010001ULL, 0x0001000100010001ULL};
 DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_2  ) = {0x0002000200020002ULL, 0x0002000200020002ULL};
 DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_3  ) = {0x0003000300030003ULL, 0x0003000300030003ULL};
 DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_4  ) = {0x0004000400040004ULL, 0x0004000400040004ULL};

libavcodec/x86/h264_intrapred_init.c

@@ -43,9 +43,41 @@ PRED4x4(horizontal_down, 10, avx)
 #define PRED8x8(TYPE, DEPTH, OPT) \
 void ff_pred8x8_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, int stride);
 
+PRED8x8(dc, 10, mmxext)
+PRED8x8(dc, 10, sse2)
+PRED8x8(top_dc, 10, mmxext)
+PRED8x8(top_dc, 10, sse2)
 PRED8x8(vertical, 10, sse2)
 PRED8x8(horizontal, 10, sse2)
 
+#define PRED8x8L(TYPE, DEPTH, OPT)\
+void ff_pred8x8l_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, int has_topleft, int has_topright, int stride);
+
+PRED8x8L(dc, 10, sse2)
+PRED8x8L(dc, 10, ssse3)
+PRED8x8L(top_dc, 10, sse2)
+PRED8x8L(top_dc, 10, ssse3)
+PRED8x8L(vertical, 10, sse2)
+PRED8x8L(vertical, 10, ssse3)
+PRED8x8L(horizontal, 10, sse2)
+PRED8x8L(horizontal, 10, ssse3)
+PRED8x8L(down_left, 10, sse2)
+PRED8x8L(down_left, 10, ssse3)
+PRED8x8L(down_right, 10, sse2)
+PRED8x8L(down_right, 10, ssse3)
+PRED8x8L(vertical_right, 10, sse2)
+PRED8x8L(vertical_right, 10, ssse3)
+PRED8x8L(horizontal_up, 10, sse2)
+PRED8x8L(horizontal_up, 10, ssse3)
+
+#define PRED16x16(TYPE, DEPTH, OPT)\
+void ff_pred16x16_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, int stride);
+
+PRED16x16(vertical, 10, mmxext)
+PRED16x16(vertical, 10, sse2)
+PRED16x16(horizontal, 10, mmxext)
+PRED16x16(horizontal, 10, sse2)
+
 void ff_pred16x16_vertical_mmx     (uint8_t *src, int stride);
 void ff_pred16x16_vertical_sse     (uint8_t *src, int stride);
 void ff_pred16x16_horizontal_mmx   (uint8_t *src, int stride);
@@ -253,6 +285,12 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
         if (mm_flags & AV_CPU_FLAG_MMX2) {
             h->pred4x4[DC_PRED             ] = ff_pred4x4_dc_10_mmxext;
             h->pred4x4[HOR_UP_PRED         ] = ff_pred4x4_horizontal_up_10_mmxext;
+
+            h->pred8x8[DC_PRED8x8          ] = ff_pred8x8_dc_10_mmxext;
+            h->pred8x8[TOP_DC_PRED8x8      ] = ff_pred8x8_top_dc_10_mmxext;
+
+            h->pred16x16[VERT_PRED8x8      ] = ff_pred16x16_vertical_10_mmxext;
+            h->pred16x16[HOR_PRED8x8       ] = ff_pred16x16_horizontal_10_mmxext;
         }
         if (mm_flags & AV_CPU_FLAG_SSE2) {
             h->pred4x4[DIAG_DOWN_LEFT_PRED ] = ff_pred4x4_down_left_10_sse2;
@@ -261,13 +299,33 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
             h->pred4x4[VERT_RIGHT_PRED     ] = ff_pred4x4_vertical_right_10_sse2;
             h->pred4x4[HOR_DOWN_PRED       ] = ff_pred4x4_horizontal_down_10_sse2;
 
+            h->pred8x8[DC_PRED8x8          ] = ff_pred8x8_dc_10_sse2;
+            h->pred8x8[TOP_DC_PRED8x8      ] = ff_pred8x8_top_dc_10_sse2;
             h->pred8x8[VERT_PRED8x8        ] = ff_pred8x8_vertical_10_sse2;
             h->pred8x8[HOR_PRED8x8         ] = ff_pred8x8_horizontal_10_sse2;
+
+            h->pred8x8l[VERT_PRED           ] = ff_pred8x8l_vertical_10_sse2;
+            h->pred8x8l[HOR_PRED            ] = ff_pred8x8l_horizontal_10_sse2;
+            h->pred8x8l[DC_PRED             ] = ff_pred8x8l_dc_10_sse2;
+            h->pred8x8l[TOP_DC_PRED         ] = ff_pred8x8l_top_dc_10_sse2;
+            h->pred8x8l[DIAG_DOWN_LEFT_PRED ] = ff_pred8x8l_down_left_10_sse2;
+            h->pred8x8l[DIAG_DOWN_RIGHT_PRED] = ff_pred8x8l_down_right_10_sse2;
+            h->pred8x8l[VERT_RIGHT_PRED     ] = ff_pred8x8l_vertical_right_10_sse2;
+            h->pred8x8l[HOR_UP_PRED         ] = ff_pred8x8l_horizontal_up_10_sse2;
+
+            h->pred16x16[VERT_PRED8x8      ] = ff_pred16x16_vertical_10_sse2;
+            h->pred16x16[HOR_PRED8x8       ] = ff_pred16x16_horizontal_10_sse2;
         }
         if (mm_flags & AV_CPU_FLAG_SSSE3) {
             h->pred4x4[DIAG_DOWN_RIGHT_PRED] = ff_pred4x4_down_right_10_ssse3;
             h->pred4x4[VERT_RIGHT_PRED     ] = ff_pred4x4_vertical_right_10_ssse3;
             h->pred4x4[HOR_DOWN_PRED       ] = ff_pred4x4_horizontal_down_10_ssse3;
+
+            h->pred8x8l[VERT_PRED           ] = ff_pred8x8l_vertical_10_ssse3;
+            h->pred8x8l[HOR_PRED            ] = ff_pred8x8l_horizontal_10_ssse3;
+            h->pred8x8l[DC_PRED             ] = ff_pred8x8l_dc_10_ssse3;
+            h->pred8x8l[TOP_DC_PRED         ] = ff_pred8x8l_top_dc_10_ssse3;
+            h->pred8x8l[DIAG_DOWN_LEFT_PRED ] = ff_pred8x8l_down_left_10_ssse3;
         }
 #if HAVE_AVX
         if (mm_flags & AV_CPU_FLAG_AVX) {

libavcodec/x86/x86util.asm

@@ -528,6 +528,14 @@
 %endif
 %endmacro
 
+%macro SPLATD 2-3 0
+%if mmsize == 16
+    pshufd %1, %2, (%3)*0x55
+%else
+    pshufw %1, %2, (%3)*0x11 + ((%3)+1)*0x44
+%endif
+%endmacro
+
 %macro CLIPW 3 ;(dst, min, max)
     pmaxsw %1, %2
     pminsw %1, %3