Factorize some duplicated code from CAVS and H.264 into a common file.

patch by Christophe Gisquet, christophe.gisquet free fr

Originally committed as revision 11504 to svn://svn.ffmpeg.org/ffmpeg/trunk

4 files changed, 57 insertions, 73 deletions

diff --git a/libavcodec/i386/cavsdsp_mmx.c b/libavcodec/i386/cavsdsp_mmx.c
index 4a23d0bc60..141382fb07 100644
--- a/libavcodec/i386/cavsdsp_mmx.c
+++ b/libavcodec/i386/cavsdsp_mmx.c
@@ -32,22 +32,6 @@
  *
  ****************************************************************************/
 
-#define SUMSUB_BA( a, b ) \
-    "paddw "#b", "#a" \n\t"\
-    "paddw "#b", "#b" \n\t"\
-    "psubw "#a", "#b" \n\t"
-
-#define SBUTTERFLY(a,b,t,n)\
-    "movq " #a ", " #t "              \n\t" /* abcd */\
-    "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
-    "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */
-
-#define TRANSPOSE4(a,b,c,d,t)\
-    SBUTTERFLY(a,b,t,wd) /* a=aebf t=cgdh */\
-    SBUTTERFLY(c,d,b,wd) /* c=imjn b=kolp */\
-    SBUTTERFLY(a,c,d,dq) /* a=aeim d=bfjn */\
-    SBUTTERFLY(t,b,c,dq) /* t=cgko c=dhlp */
-
 static inline void cavs_idct8_1d(int16_t *block, uint64_t bias)
 {
     asm volatile(
diff --git a/libavcodec/i386/dsputil_mmx.c b/libavcodec/i386/dsputil_mmx.c
index 8328c28114..9b787291c8 100644
--- a/libavcodec/i386/dsputil_mmx.c
+++ b/libavcodec/i386/dsputil_mmx.c
@@ -203,17 +203,6 @@ DECLARE_ALIGNED_16(const double, ff_pd_2[2]) = { 2.0, 2.0 };
 #undef DEF
 #undef PAVGB
 
-#define SBUTTERFLY(a,b,t,n,m)\
-    "mov" #m " " #a ", " #t "         \n\t" /* abcd */\
-    "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
-    "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */\
-
-#define TRANSPOSE4(a,b,c,d,t)\
-    SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\
-    SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\
-    SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\
-    SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */
-
 /***********************************/
 /* standard MMX */
 
@@ -1545,46 +1534,6 @@ static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *src1, uint8_t
 #define DIFF_PIXELS_4x8(p1,p2,stride,temp) DIFF_PIXELS_8(d, q,   %%mm,  p1, p2, stride, temp)
 #define DIFF_PIXELS_8x8(p1,p2,stride,temp) DIFF_PIXELS_8(q, dqa, %%xmm, p1, p2, stride, temp)
 
-#ifdef ARCH_X86_64
-// permutes 01234567 -> 05736421
-#define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
-    SBUTTERFLY(a,b,%%xmm8,wd,dqa)\
-    SBUTTERFLY(c,d,b,wd,dqa)\
-    SBUTTERFLY(e,f,d,wd,dqa)\
-    SBUTTERFLY(g,h,f,wd,dqa)\
-    SBUTTERFLY(a,c,h,dq,dqa)\
-    SBUTTERFLY(%%xmm8,b,c,dq,dqa)\
-    SBUTTERFLY(e,g,b,dq,dqa)\
-    SBUTTERFLY(d,f,g,dq,dqa)\
-    SBUTTERFLY(a,e,f,qdq,dqa)\
-    SBUTTERFLY(%%xmm8,d,e,qdq,dqa)\
-    SBUTTERFLY(h,b,d,qdq,dqa)\
-    SBUTTERFLY(c,g,b,qdq,dqa)\
-    "movdqa %%xmm8, "#g"              \n\t"
-#else
-#define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
-    "movdqa "#h", "#t"                \n\t"\
-    SBUTTERFLY(a,b,h,wd,dqa)\
-    "movdqa "#h", 16"#t"              \n\t"\
-    "movdqa "#t", "#h"                \n\t"\
-    SBUTTERFLY(c,d,b,wd,dqa)\
-    SBUTTERFLY(e,f,d,wd,dqa)\
-    SBUTTERFLY(g,h,f,wd,dqa)\
-    SBUTTERFLY(a,c,h,dq,dqa)\
-    "movdqa "#h", "#t"                \n\t"\
-    "movdqa 16"#t", "#h"              \n\t"\
-    SBUTTERFLY(h,b,c,dq,dqa)\
-    SBUTTERFLY(e,g,b,dq,dqa)\
-    SBUTTERFLY(d,f,g,dq,dqa)\
-    SBUTTERFLY(a,e,f,qdq,dqa)\
-    SBUTTERFLY(h,d,e,qdq,dqa)\
-    "movdqa "#h", 16"#t"              \n\t"\
-    "movdqa "#t", "#h"                \n\t"\
-    SBUTTERFLY(h,b,d,qdq,dqa)\
-    SBUTTERFLY(c,g,b,qdq,dqa)\
-    "movdqa 16"#t", "#g"              \n\t"
-#endif
-
 #define LBUTTERFLY2(a1,b1,a2,b2)\
     "paddw " #b1 ", " #a1 "           \n\t"\
     "paddw " #b2 ", " #a2 "           \n\t"\
diff --git a/libavcodec/i386/dsputil_mmx.h b/libavcodec/i386/dsputil_mmx.h
index fc5d01e70b..a1571fc1e3 100644
--- a/libavcodec/i386/dsputil_mmx.h
+++ b/libavcodec/i386/dsputil_mmx.h
@@ -52,4 +52,61 @@ extern const uint64_t ff_pb_FC;
 extern const double ff_pd_1[2];
 extern const double ff_pd_2[2];
 
+/* in/out: mma=mma+mmb, mmb=mmb-mma */
+#define SUMSUB_BA( a, b ) \
+    "paddw "#b", "#a" \n\t"\
+    "paddw "#b", "#b" \n\t"\
+    "psubw "#a", "#b" \n\t"
+
+#define SBUTTERFLY(a,b,t,n,m)\
+    "mov" #m " " #a ", " #t "         \n\t" /* abcd */\
+    "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
+    "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */\
+
+#define TRANSPOSE4(a,b,c,d,t)\
+    SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\
+    SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\
+    SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\
+    SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */
+
+#ifdef ARCH_X86_64
+// permutes 01234567 -> 05736421
+#define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
+    SBUTTERFLY(a,b,%%xmm8,wd,dqa)\
+    SBUTTERFLY(c,d,b,wd,dqa)\
+    SBUTTERFLY(e,f,d,wd,dqa)\
+    SBUTTERFLY(g,h,f,wd,dqa)\
+    SBUTTERFLY(a,c,h,dq,dqa)\
+    SBUTTERFLY(%%xmm8,b,c,dq,dqa)\
+    SBUTTERFLY(e,g,b,dq,dqa)\
+    SBUTTERFLY(d,f,g,dq,dqa)\
+    SBUTTERFLY(a,e,f,qdq,dqa)\
+    SBUTTERFLY(%%xmm8,d,e,qdq,dqa)\
+    SBUTTERFLY(h,b,d,qdq,dqa)\
+    SBUTTERFLY(c,g,b,qdq,dqa)\
+    "movdqa %%xmm8, "#g"              \n\t"
+#else
+#define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
+    "movdqa "#h", "#t"                \n\t"\
+    SBUTTERFLY(a,b,h,wd,dqa)\
+    "movdqa "#h", 16"#t"              \n\t"\
+    "movdqa "#t", "#h"                \n\t"\
+    SBUTTERFLY(c,d,b,wd,dqa)\
+    SBUTTERFLY(e,f,d,wd,dqa)\
+    SBUTTERFLY(g,h,f,wd,dqa)\
+    SBUTTERFLY(a,c,h,dq,dqa)\
+    "movdqa "#h", "#t"                \n\t"\
+    "movdqa 16"#t", "#h"              \n\t"\
+    SBUTTERFLY(h,b,c,dq,dqa)\
+    SBUTTERFLY(e,g,b,dq,dqa)\
+    SBUTTERFLY(d,f,g,dq,dqa)\
+    SBUTTERFLY(a,e,f,qdq,dqa)\
+    SBUTTERFLY(h,d,e,qdq,dqa)\
+    "movdqa "#h", 16"#t"              \n\t"\
+    "movdqa "#t", "#h"                \n\t"\
+    SBUTTERFLY(h,b,d,qdq,dqa)\
+    SBUTTERFLY(c,g,b,qdq,dqa)\
+    "movdqa 16"#t", "#g"              \n\t"
+#endif
+
 #endif /* FFMPEG_DSPUTIL_MMX_H */
diff --git a/libavcodec/i386/h264dsp_mmx.c b/libavcodec/i386/h264dsp_mmx.c
index 14511db1e7..1ecb0cbef1 100644
--- a/libavcodec/i386/h264dsp_mmx.c
+++ b/libavcodec/i386/h264dsp_mmx.c
@@ -23,12 +23,6 @@
 /***********************************/
 /* IDCT */
 
-/* in/out: mma=mma+mmb, mmb=mmb-mma */
-#define SUMSUB_BA( a, b ) \
-    "paddw "#b", "#a" \n\t"\
-    "paddw "#b", "#b" \n\t"\
-    "psubw "#a", "#b" \n\t"
-
 #define SUMSUB_BADC( a, b, c, d ) \
     "paddw "#b", "#a" \n\t"\
     "paddw "#d", "#c" \n\t"\