Use H264 MMX chroma functions to accelerate RV40 decoding.

Patch by Mathieu Velten (matmaul A gmail)

Originally committed as revision 16419 to svn://svn.ffmpeg.org/ffmpeg/trunk

4 files changed, 104 insertions, 21 deletions

diff --git a/libavcodec/x86/dsputil_h264_template_mmx.c b/libavcodec/x86/dsputil_h264_template_mmx.c
index 0bf8732e35..43f4393098 100644
--- a/libavcodec/x86/dsputil_h264_template_mmx.c
+++ b/libavcodec/x86/dsputil_h264_template_mmx.c
@@ -25,9 +25,8 @@
  * H264_CHROMA_OP must be defined to empty for put and pavgb/pavgusb for avg
  * H264_CHROMA_MC8_MV0 must be defined to a (put|avg)_pixels8 function
  */
-static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y, int rnd)
+static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y, const uint64_t *rnd_reg)
 {
-    const uint64_t *rnd_reg;
     DECLARE_ALIGNED_8(uint64_t, AA);
     DECLARE_ALIGNED_8(uint64_t, DD);
     int i;
@@ -45,17 +44,15 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
         /* 1 dimensional filter only */
         const int dxy = x ? 1 : stride;
 
-        rnd_reg = rnd ? &ff_pw_4 : &ff_pw_3;
-
         __asm__ volatile(
             "movd %0, %%mm5\n\t"
             "movq %1, %%mm4\n\t"
-            "movq %2, %%mm6\n\t"         /* mm6 = rnd */
+            "movq %2, %%mm6\n\t"         /* mm6 = rnd >> 3 */
             "punpcklwd %%mm5, %%mm5\n\t"
             "punpckldq %%mm5, %%mm5\n\t" /* mm5 = B = x */
             "pxor %%mm7, %%mm7\n\t"
             "psubw %%mm5, %%mm4\n\t"     /* mm4 = A = 8-x */
-            :: "rm"(x+y), "m"(ff_pw_8), "m"(*rnd_reg));
+            :: "rm"(x+y), "m"(ff_pw_8), "m"(*(rnd_reg+1)));
 
         for(i=0; i<h; i++) {
             __asm__ volatile(
@@ -78,7 +75,7 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
                 "pmullw %%mm5, %%mm2\n\t"
                 "pmullw %%mm5, %%mm3\n\t"
 
-                /* dst[0..7] = (A * src[0..7] + B * src[1..8] + 4) >> 3 */
+                /* dst[0..7] = (A * src[0..7] + B * src[1..8] + (rnd >> 3)) >> 3 */
                 "paddw %%mm6, %%mm0\n\t"
                 "paddw %%mm6, %%mm1\n\t"
                 "paddw %%mm2, %%mm0\n\t"
@@ -97,7 +94,6 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
     }
 
     /* general case, bilinear */
-    rnd_reg = rnd ? &ff_pw_32.a : &ff_pw_28.a;
     __asm__ volatile("movd %2, %%mm4\n\t"
                  "movd %3, %%mm6\n\t"
                  "punpcklwd %%mm4, %%mm4\n\t"
@@ -172,7 +168,7 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
             : : "m" (src[0]), "m" (src[1]), "m" (DD));
 
         __asm__ volatile(
-            /* dst[0..7] = ([mm2,mm3] + 32) >> 6 */
+            /* dst[0..7] = ([mm2,mm3] + rnd) >> 6 */
             "paddw %1, %%mm2\n\t"
             "paddw %1, %%mm3\n\t"
             "psrlw $6, %%mm2\n\t"
@@ -185,7 +181,7 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
     }
 }
 
-static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 4*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 4*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y, const uint64_t *rnd_reg)
 {
     __asm__ volatile(
         "pxor   %%mm7, %%mm7        \n\t"
@@ -249,7 +245,7 @@ static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 4*/, uint8_t *src/*align 1*
         "sub $2, %2                 \n\t"
         "jnz 1b                     \n\t"
         : "+r"(dst), "+r"(src), "+r"(h)
-        : "r"((x86_reg)stride), "m"(ff_pw_32), "m"(x), "m"(y)
+        : "r"((x86_reg)stride), "m"(*rnd_reg), "m"(x), "m"(y)
     );
 }
 
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index c6588ed02d..f76a4fa33a 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -1733,6 +1733,7 @@ PREFETCH(prefetch_3dnow, prefetch)
 #undef PREFETCH
 
 #include "h264dsp_mmx.c"
+#include "rv40dsp_mmx.c"
 
 /* CAVS specific */
 void ff_cavsdsp_init_mmx2(DSPContext* c, AVCodecContext *avctx);
@@ -2638,6 +2639,9 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
         c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_mmx;
         c->put_no_rnd_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_mmx_nornd;
 
+        c->put_rv40_chroma_pixels_tab[0]= put_rv40_chroma_mc8_mmx;
+        c->put_rv40_chroma_pixels_tab[1]= put_rv40_chroma_mc4_mmx;
+
         c->h264_idct_dc_add=
         c->h264_idct_add= ff_h264_idct_add_mmx;
         c->h264_idct8_dc_add=
@@ -2723,6 +2727,9 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmx2);
             SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, mmx2);
 
+            c->avg_rv40_chroma_pixels_tab[0]= avg_rv40_chroma_mc8_mmx2;
+            c->avg_rv40_chroma_pixels_tab[1]= avg_rv40_chroma_mc4_mmx2;
+
             c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_mmx2_rnd;
             c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_mmx2;
             c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_mmx2;
@@ -2808,6 +2815,9 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_3dnow_rnd;
             c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_3dnow;
 
+            c->avg_rv40_chroma_pixels_tab[0]= avg_rv40_chroma_mc8_3dnow;
+            c->avg_rv40_chroma_pixels_tab[1]= avg_rv40_chroma_mc4_3dnow;
+
             if (ENABLE_CAVS_DECODER)
                 ff_cavsdsp_init_3dnow(c, avctx);
         }
diff --git a/libavcodec/x86/h264dsp_mmx.c b/libavcodec/x86/h264dsp_mmx.c
index 511aeb4c82..8eeb65779b 100644
--- a/libavcodec/x86/h264dsp_mmx.c
+++ b/libavcodec/x86/h264dsp_mmx.c
@@ -2084,22 +2084,30 @@ H264_MC_816(H264_MC_H, ssse3)
 H264_MC_816(H264_MC_HV, ssse3)
 #endif
 
+/* rnd interleaved with rnd div 8, use p+1 to access rnd div 8 */
+DECLARE_ALIGNED_8(static const uint64_t, h264_rnd_reg[4]) = {
+    0x0020002000200020ULL, 0x0004000400040004ULL, 0x001C001C001C001CULL, 0x0003000300030003ULL
+};
 
 #define H264_CHROMA_OP(S,D)
 #define H264_CHROMA_OP4(S,D,T)
-#define H264_CHROMA_MC8_TMPL put_h264_chroma_mc8_mmx
-#define H264_CHROMA_MC4_TMPL put_h264_chroma_mc4_mmx
+#define H264_CHROMA_MC8_TMPL put_h264_chroma_generic_mc8_mmx
+#define H264_CHROMA_MC4_TMPL put_h264_chroma_generic_mc4_mmx
 #define H264_CHROMA_MC2_TMPL put_h264_chroma_mc2_mmx2
 #define H264_CHROMA_MC8_MV0 put_pixels8_mmx
 #include "dsputil_h264_template_mmx.c"
 
 static void put_h264_chroma_mc8_mmx_rnd(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
 {
-    put_h264_chroma_mc8_mmx(dst, src, stride, h, x, y, 1);
+    put_h264_chroma_generic_mc8_mmx(dst, src, stride, h, x, y, h264_rnd_reg);
 }
 static void put_h264_chroma_mc8_mmx_nornd(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
 {
-    put_h264_chroma_mc8_mmx(dst, src, stride, h, x, y, 0);
+    put_h264_chroma_generic_mc8_mmx(dst, src, stride, h, x, y, h264_rnd_reg+2);
+}
+static void put_h264_chroma_mc4_mmx(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    put_h264_chroma_generic_mc4_mmx(dst, src, stride, h, x, y, h264_rnd_reg);
 }
 
 #undef H264_CHROMA_OP
@@ -2112,14 +2120,18 @@ static void put_h264_chroma_mc8_mmx_nornd(uint8_t *dst/*align 8*/, uint8_t *src/
 #define H264_CHROMA_OP(S,D) "pavgb " #S ", " #D " \n\t"
 #define H264_CHROMA_OP4(S,D,T) "movd  " #S ", " #T " \n\t"\
                                "pavgb " #T ", " #D " \n\t"
-#define H264_CHROMA_MC8_TMPL avg_h264_chroma_mc8_mmx2
-#define H264_CHROMA_MC4_TMPL avg_h264_chroma_mc4_mmx2
+#define H264_CHROMA_MC8_TMPL avg_h264_chroma_generic_mc8_mmx2
+#define H264_CHROMA_MC4_TMPL avg_h264_chroma_generic_mc4_mmx2
 #define H264_CHROMA_MC2_TMPL avg_h264_chroma_mc2_mmx2
 #define H264_CHROMA_MC8_MV0 avg_pixels8_mmx2
 #include "dsputil_h264_template_mmx.c"
 static void avg_h264_chroma_mc8_mmx2_rnd(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
 {
-    avg_h264_chroma_mc8_mmx2(dst, src, stride, h, x, y, 1);
+    avg_h264_chroma_generic_mc8_mmx2(dst, src, stride, h, x, y, h264_rnd_reg);
+}
+static void avg_h264_chroma_mc4_mmx2(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    avg_h264_chroma_generic_mc4_mmx2(dst, src, stride, h, x, y, h264_rnd_reg);
 }
 #undef H264_CHROMA_OP
 #undef H264_CHROMA_OP4
@@ -2131,13 +2143,17 @@ static void avg_h264_chroma_mc8_mmx2_rnd(uint8_t *dst/*align 8*/, uint8_t *src/*
 #define H264_CHROMA_OP(S,D) "pavgusb " #S ", " #D " \n\t"
 #define H264_CHROMA_OP4(S,D,T) "movd " #S ", " #T " \n\t"\
                                "pavgusb " #T ", " #D " \n\t"
-#define H264_CHROMA_MC8_TMPL avg_h264_chroma_mc8_3dnow
-#define H264_CHROMA_MC4_TMPL avg_h264_chroma_mc4_3dnow
+#define H264_CHROMA_MC8_TMPL avg_h264_chroma_generic_mc8_3dnow
+#define H264_CHROMA_MC4_TMPL avg_h264_chroma_generic_mc4_3dnow
 #define H264_CHROMA_MC8_MV0 avg_pixels8_3dnow
 #include "dsputil_h264_template_mmx.c"
 static void avg_h264_chroma_mc8_3dnow_rnd(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
 {
-    avg_h264_chroma_mc8_3dnow(dst, src, stride, h, x, y, 1);
+    avg_h264_chroma_generic_mc8_3dnow(dst, src, stride, h, x, y, h264_rnd_reg);
+}
+static void avg_h264_chroma_mc4_3dnow(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    avg_h264_chroma_generic_mc4_3dnow(dst, src, stride, h, x, y, h264_rnd_reg);
 }
 #undef H264_CHROMA_OP
 #undef H264_CHROMA_OP4
diff --git a/libavcodec/x86/rv40dsp_mmx.c b/libavcodec/x86/rv40dsp_mmx.c
new file mode 100644
index 0000000000..47461c68aa
--- /dev/null
+++ b/libavcodec/x86/rv40dsp_mmx.c
@@ -0,0 +1,61 @@
+/*
+ * Copyright (c) 2008 Konstantin Shishkov, Mathieu Velten
+ *
+ * MMX-optimized DSP functions for RV40, based on H.264 optimizations by
+ * Michael Niedermayer and Loren Merritt
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "dsputil_mmx.h"
+
+/* bias interleaved with bias div 8, use p+1 to access bias div 8 */
+DECLARE_ALIGNED_8(static const uint64_t, rv40_bias_reg[4][8]) = {
+    { 0x0000000000000000ULL, 0x0000000000000000ULL, 0x0010001000100010ULL, 0x0002000200020002ULL,
+      0x0020002000200020ULL, 0x0004000400040004ULL, 0x0010001000100010ULL, 0x0002000200020002ULL },
+    { 0x0020002000200020ULL, 0x0004000400040004ULL, 0x001C001C001C001CULL, 0x0003000300030003ULL,
+      0x0020002000200020ULL, 0x0004000400040004ULL, 0x001C001C001C001CULL, 0x0003000300030003ULL },
+    { 0x0000000000000000ULL, 0x0000000000000000ULL, 0x0020002000200020ULL, 0x0004000400040004ULL,
+      0x0010001000100010ULL, 0x0002000200020002ULL, 0x0020002000200020ULL, 0x0004000400040004ULL },
+    { 0x0020002000200020ULL, 0x0004000400040004ULL, 0x001C001C001C001CULL, 0x0003000300030003ULL,
+      0x0020002000200020ULL, 0x0004000400040004ULL, 0x001C001C001C001CULL, 0x0003000300030003ULL }
+};
+
+static void put_rv40_chroma_mc8_mmx(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    put_h264_chroma_generic_mc8_mmx(dst, src, stride, h, x, y, &rv40_bias_reg[y>>1][x&(~1)]);
+}
+static void put_rv40_chroma_mc4_mmx(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    put_h264_chroma_generic_mc4_mmx(dst, src, stride, h, x, y, &rv40_bias_reg[y>>1][x&(~1)]);
+}
+static void avg_rv40_chroma_mc8_mmx2(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    avg_h264_chroma_generic_mc8_mmx2(dst, src, stride, h, x, y, &rv40_bias_reg[y>>1][x&(~1)]);
+}
+static void avg_rv40_chroma_mc4_mmx2(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    avg_h264_chroma_generic_mc4_mmx2(dst, src, stride, h, x, y, &rv40_bias_reg[y>>1][x&(~1)]);
+}
+static void avg_rv40_chroma_mc8_3dnow(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    avg_h264_chroma_generic_mc8_3dnow(dst, src, stride, h, x, y, &rv40_bias_reg[y>>1][x&(~1)]);
+}
+static void avg_rv40_chroma_mc4_3dnow(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
+{
+    avg_h264_chroma_generic_mc4_3dnow(dst, src, stride, h, x, y, &rv40_bias_reg[y>>1][x&(~1)]);
+}