ffmpeg sync

CVS patchset: 6437
CVS date: 2004/04/25 18:57:04

8 files changed, 973 insertions, 82 deletions

diff --git a/src/libffmpeg/libavcodec/ppc/Makefile.am b/src/libffmpeg/libavcodec/ppc/Makefile.am
index fbd734c29..50b9d802e 100644
--- a/src/libffmpeg/libavcodec/ppc/Makefile.am
+++ b/src/libffmpeg/libavcodec/ppc/Makefile.am
@@ -11,6 +11,7 @@ noinst_LTLIBRARIES = libavcodec_ppc.la
 
 libavcodec_ppc_src =  dsputil_altivec.c \
 		      dsputil_ppc.c \
+		      fdct_altivec.c \
 		      fft_altivec.c \
 		      idct_altivec.c \
 		      gmc_altivec.c \
diff --git a/src/libffmpeg/libavcodec/ppc/dsputil_altivec.c b/src/libffmpeg/libavcodec/ppc/dsputil_altivec.c
index 633cae68b..1bc6fb009 100644
--- a/src/libffmpeg/libavcodec/ppc/dsputil_altivec.c
+++ b/src/libffmpeg/libavcodec/ppc/dsputil_altivec.c
@@ -1,7 +1,7 @@
 /*
  * Copyright (c) 2002 Brian Foley
  * Copyright (c) 2002 Dieter Shirley
- * Copyright (c) 2003 Romain Dolbeau <romain@dolbeau.org>
+ * Copyright (c) 2003-2004 Romain Dolbeau <romain@dolbeau.org>
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
@@ -1302,6 +1302,357 @@ POWERPC_PERF_STOP_COUNT(altivec_put_no_rnd_pixels16_xy2_num, 1);
 #endif /* ALTIVEC_USE_REFERENCE_C_CODE */
 }
 
+int hadamard8_diff8x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h){
+POWERPC_PERF_DECLARE(altivec_hadamard8_diff8x8_num, 1);
+  int sum;
+POWERPC_PERF_START_COUNT(altivec_hadamard8_diff8x8_num, 1);
+  register const_vector unsigned char vzero = (const_vector unsigned char)vec_splat_u8(0);
+  register vector signed short temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
+#ifdef CONFIG_DARWIN
+  {
+    register const_vector signed short vprod1 = (const_vector signed short)( 1,-1, 1,-1, 1,-1, 1,-1);
+    register const_vector signed short vprod2 = (const_vector signed short)( 1, 1,-1,-1, 1, 1,-1,-1);
+    register const_vector signed short vprod3 = (const_vector signed short)( 1, 1, 1, 1,-1,-1,-1,-1);
+    register const_vector unsigned char perm1 = (const_vector unsigned char)
+      (0x02, 0x03, 0x00, 0x01,
+       0x06, 0x07, 0x04, 0x05,
+       0x0A, 0x0B, 0x08, 0x09,
+       0x0E, 0x0F, 0x0C, 0x0D);
+    register const_vector unsigned char perm2 = (const_vector unsigned char)
+      (0x04, 0x05, 0x06, 0x07,
+       0x00, 0x01, 0x02, 0x03,
+       0x0C, 0x0D, 0x0E, 0x0F,
+       0x08, 0x09, 0x0A, 0x0B);
+    register const_vector unsigned char perm3 = (const_vector unsigned char)
+      (0x08, 0x09, 0x0A, 0x0B,
+       0x0C, 0x0D, 0x0E, 0x0F,
+       0x00, 0x01, 0x02, 0x03,
+       0x04, 0x05, 0x06, 0x07);
+#else
+    register const_vector signed short vprod1 = (const_vector signed short){ 1,-1, 1,-1, 1,-1, 1,-1};
+    register const_vector signed short vprod2 = (const_vector signed short){ 1, 1,-1,-1, 1, 1,-1,-1};
+    register const_vector signed short vprod3 = (const_vector signed short){ 1, 1, 1, 1,-1,-1,-1,-1};
+    register const_vector unsigned char perm1 = (const_vector unsigned char)
+      {0x02, 0x03, 0x00, 0x01,
+       0x06, 0x07, 0x04, 0x05,
+       0x0A, 0x0B, 0x08, 0x09,
+       0x0E, 0x0F, 0x0C, 0x0D};
+    register const_vector unsigned char perm2 = (const_vector unsigned char)
+      {0x04, 0x05, 0x06, 0x07,
+       0x00, 0x01, 0x02, 0x03,
+       0x0C, 0x0D, 0x0E, 0x0F,
+       0x08, 0x09, 0x0A, 0x0B};
+    register const_vector unsigned char perm3 = (const_vector unsigned char)
+      {0x08, 0x09, 0x0A, 0x0B,
+       0x0C, 0x0D, 0x0E, 0x0F,
+       0x00, 0x01, 0x02, 0x03,
+       0x04, 0x05, 0x06, 0x07};
+#endif
+
+#define ONEITERBUTTERFLY(i, res)					\
+    {									\
+      register vector unsigned char src1, src2, srcO;		       	\
+      register vector unsigned char dst1, dst2, dstO;		       	\
+      src1 = vec_ld(stride * i, src);					\
+      if ((((stride * i) + (unsigned long)src) & 0x0000000F) > 8)	\
+	src2 = vec_ld((stride * i) + 16, src);				\
+      srcO = vec_perm(src1, src2, vec_lvsl(stride * i, src));		\
+      dst1 = vec_ld(stride * i, dst);					\
+      if ((((stride * i) + (unsigned long)dst) & 0x0000000F) > 8)	\
+	dst2 = vec_ld((stride * i) + 16, dst);				\
+      dstO = vec_perm(dst1, dst2, vec_lvsl(stride * i, dst));		\
+      /* promote the unsigned chars to signed shorts */			\
+      /* we're in the 8x8 function, we only care for the first 8 */	\
+      register vector signed short srcV =			       	\
+	(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
+      register vector signed short dstV =			       	\
+	(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
+      /* substractions inside the first butterfly */			\
+      register vector signed short but0 = vec_sub(srcV, dstV);	       	\
+      register vector signed short op1 = vec_perm(but0, but0, perm1);  	\
+      register vector signed short but1 = vec_mladd(but0, vprod1, op1);	\
+      register vector signed short op2 = vec_perm(but1, but1, perm2);  	\
+      register vector signed short but2 = vec_mladd(but1, vprod2, op2);	\
+      register vector signed short op3 = vec_perm(but2, but2, perm3);  	\
+      res = vec_mladd(but2, vprod3, op3);				\
+    }
+    ONEITERBUTTERFLY(0, temp0);
+    ONEITERBUTTERFLY(1, temp1);
+    ONEITERBUTTERFLY(2, temp2);
+    ONEITERBUTTERFLY(3, temp3);
+    ONEITERBUTTERFLY(4, temp4);
+    ONEITERBUTTERFLY(5, temp5);
+    ONEITERBUTTERFLY(6, temp6);
+    ONEITERBUTTERFLY(7, temp7);
+  }
+#undef ONEITERBUTTERFLY
+  {
+    register vector signed int vsum;
+    register vector signed short line0 = vec_add(temp0, temp1);
+    register vector signed short line1 = vec_sub(temp0, temp1);
+    register vector signed short line2 = vec_add(temp2, temp3);
+    register vector signed short line3 = vec_sub(temp2, temp3);
+    register vector signed short line4 = vec_add(temp4, temp5);
+    register vector signed short line5 = vec_sub(temp4, temp5);
+    register vector signed short line6 = vec_add(temp6, temp7);
+    register vector signed short line7 = vec_sub(temp6, temp7);
+    
+    register vector signed short line0B = vec_add(line0, line2);
+    register vector signed short line2B = vec_sub(line0, line2);
+    register vector signed short line1B = vec_add(line1, line3);
+    register vector signed short line3B = vec_sub(line1, line3);
+    register vector signed short line4B = vec_add(line4, line6);
+    register vector signed short line6B = vec_sub(line4, line6);
+    register vector signed short line5B = vec_add(line5, line7);
+    register vector signed short line7B = vec_sub(line5, line7);
+    
+    register vector signed short line0C = vec_add(line0B, line4B);
+    register vector signed short line4C = vec_sub(line0B, line4B);
+    register vector signed short line1C = vec_add(line1B, line5B);
+    register vector signed short line5C = vec_sub(line1B, line5B);
+    register vector signed short line2C = vec_add(line2B, line6B);
+    register vector signed short line6C = vec_sub(line2B, line6B);
+    register vector signed short line3C = vec_add(line3B, line7B);
+    register vector signed short line7C = vec_sub(line3B, line7B);
+    
+    vsum = vec_sum4s(vec_abs(line0C), vec_splat_s32(0));
+    vsum = vec_sum4s(vec_abs(line1C), vsum);
+    vsum = vec_sum4s(vec_abs(line2C), vsum);
+    vsum = vec_sum4s(vec_abs(line3C), vsum);
+    vsum = vec_sum4s(vec_abs(line4C), vsum);
+    vsum = vec_sum4s(vec_abs(line5C), vsum);
+    vsum = vec_sum4s(vec_abs(line6C), vsum);
+    vsum = vec_sum4s(vec_abs(line7C), vsum);
+    vsum = vec_sums(vsum, (vector signed int)vzero);
+    vsum = vec_splat(vsum, 3);
+    vec_ste(vsum, 0, &sum);
+  }
+POWERPC_PERF_STOP_COUNT(altivec_hadamard8_diff8x8_num, 1);
+  return sum;
+}
+
+/*
+  16x8 works with 16 elements ; it allows to avoid replicating
+  loads, and give the compiler more rooms for scheduling.
+  It's only used from inside hadamard8_diff16_altivec.
+  
+  Unfortunately, it seems gcc-3.3 is a bit dumb, and
+  the compiled code has a LOT of spill code, it seems
+  gcc (unlike xlc) cannot keep everything in registers
+  by itself. The following code include hand-made
+  registers allocation. It's not clean, but on
+  a 7450 the resulting code is much faster (best case
+  fall from 700+ cycles to 550).
+  
+  xlc doesn't add spill code, but it doesn't know how to
+  schedule for the 7450, and its code isn't much faster than
+  gcc-3.3 on the 7450 (but uses 25% less instructions...)
+  
+  On the 970, the hand-made RA is still a win (arount 690
+  vs. around 780), but xlc goes to around 660 on the
+  regular C code...
+*/
+
+static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h) {
+  int sum;
+  register vector signed short
+    temp0 asm ("v0"),
+    temp1 asm ("v1"),
+    temp2 asm ("v2"),
+    temp3 asm ("v3"),
+    temp4 asm ("v4"),
+    temp5 asm ("v5"),
+    temp6 asm ("v6"),
+    temp7 asm ("v7");
+  register vector signed short
+    temp0S asm ("v8"),
+    temp1S asm ("v9"),
+    temp2S asm ("v10"),
+    temp3S asm ("v11"),
+    temp4S asm ("v12"),
+    temp5S asm ("v13"),
+    temp6S asm ("v14"),
+    temp7S asm ("v15");
+  register const_vector unsigned char vzero asm ("v31")= (const_vector unsigned char)vec_splat_u8(0);
+  {
+#ifdef CONFIG_DARWIN
+    register const_vector signed short vprod1 asm ("v16")= (const_vector signed short)( 1,-1, 1,-1, 1,-1, 1,-1);
+    register const_vector signed short vprod2 asm ("v17")= (const_vector signed short)( 1, 1,-1,-1, 1, 1,-1,-1);
+    register const_vector signed short vprod3 asm ("v18")= (const_vector signed short)( 1, 1, 1, 1,-1,-1,-1,-1);
+    register const_vector unsigned char perm1 asm ("v19")= (const_vector unsigned char)
+      (0x02, 0x03, 0x00, 0x01,
+       0x06, 0x07, 0x04, 0x05,
+       0x0A, 0x0B, 0x08, 0x09,
+       0x0E, 0x0F, 0x0C, 0x0D);
+    register const_vector unsigned char perm2 asm ("v20")= (const_vector unsigned char)
+      (0x04, 0x05, 0x06, 0x07,
+       0x00, 0x01, 0x02, 0x03,
+       0x0C, 0x0D, 0x0E, 0x0F,
+       0x08, 0x09, 0x0A, 0x0B);
+    register const_vector unsigned char perm3 asm ("v21")= (const_vector unsigned char)
+      (0x08, 0x09, 0x0A, 0x0B,
+       0x0C, 0x0D, 0x0E, 0x0F,
+       0x00, 0x01, 0x02, 0x03,
+       0x04, 0x05, 0x06, 0x07);
+#else
+    register const_vector signed short vprod1 = (const_vector signed short){ 1,-1, 1,-1, 1,-1, 1,-1};
+    register const_vector signed short vprod2 = (const_vector signed short){ 1, 1,-1,-1, 1, 1,-1,-1};
+    register const_vector signed short vprod3 = (const_vector signed short){ 1, 1, 1, 1,-1,-1,-1,-1};
+    register const_vector unsigned char perm1 = (const_vector unsigned char)
+      {0x02, 0x03, 0x00, 0x01,
+       0x06, 0x07, 0x04, 0x05,
+       0x0A, 0x0B, 0x08, 0x09,
+       0x0E, 0x0F, 0x0C, 0x0D};
+    register const_vector unsigned char perm2 = (const_vector unsigned char)
+      {0x04, 0x05, 0x06, 0x07,
+       0x00, 0x01, 0x02, 0x03,
+       0x0C, 0x0D, 0x0E, 0x0F,
+       0x08, 0x09, 0x0A, 0x0B};
+    register const_vector unsigned char perm3 = (const_vector unsigned char)
+      {0x08, 0x09, 0x0A, 0x0B,
+       0x0C, 0x0D, 0x0E, 0x0F,
+       0x00, 0x01, 0x02, 0x03,
+       0x04, 0x05, 0x06, 0x07};
+#endif
+#define ONEITERBUTTERFLY(i, res1, res2)					\
+    {									\
+      register vector unsigned char src1 asm ("v22"), src2 asm ("v23"); \
+      register vector unsigned char dst1 asm ("v24"), dst2 asm ("v25"); \
+      src1 = vec_ld(stride * i, src);					\
+      src2 = vec_ld((stride * i) + 16, src);				\
+      register vector unsigned char srcO asm ("v22") = vec_perm(src1, src2, vec_lvsl(stride * i, src)); \
+      dst1 = vec_ld(stride * i, dst);					\
+      dst2 = vec_ld((stride * i) + 16, dst);				\
+      register vector unsigned char dstO asm ("v23") = vec_perm(dst1, dst2, vec_lvsl(stride * i, dst)); \
+      /* promote the unsigned chars to signed shorts */			\
+      register vector signed short srcV asm ("v24") =                   \
+	(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
+      register vector signed short dstV asm ("v25") =                   \
+	(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
+      register vector signed short srcW asm ("v26") =                   \
+	(vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)srcO); \
+      register vector signed short dstW asm ("v27") =                   \
+	(vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)dstO); \
+      /* substractions inside the first butterfly */			\
+      register vector signed short but0 asm ("v28") = vec_sub(srcV, dstV); \
+      register vector signed short but0S asm ("v29") = vec_sub(srcW, dstW); \
+      register vector signed short op1 asm ("v30") = vec_perm(but0, but0, perm1); \
+      register vector signed short but1 asm ("v22") = vec_mladd(but0, vprod1, op1); \
+      register vector signed short op1S asm ("v23") = vec_perm(but0S, but0S, perm1); \
+      register vector signed short but1S asm ("v24") = vec_mladd(but0S, vprod1, op1S); \
+      register vector signed short op2 asm ("v25") = vec_perm(but1, but1, perm2); \
+      register vector signed short but2 asm ("v26") = vec_mladd(but1, vprod2, op2); \
+      register vector signed short op2S asm ("v27") = vec_perm(but1S, but1S, perm2); \
+      register vector signed short but2S asm ("v28") = vec_mladd(but1S, vprod2, op2S); \
+      register vector signed short op3 asm ("v29") = vec_perm(but2, but2, perm3); \
+      res1 = vec_mladd(but2, vprod3, op3);				\
+      register vector signed short op3S asm ("v30") = vec_perm(but2S, but2S, perm3); \
+      res2 = vec_mladd(but2S, vprod3, op3S);				\
+    }
+    ONEITERBUTTERFLY(0, temp0, temp0S);
+    ONEITERBUTTERFLY(1, temp1, temp1S);
+    ONEITERBUTTERFLY(2, temp2, temp2S);
+    ONEITERBUTTERFLY(3, temp3, temp3S);
+    ONEITERBUTTERFLY(4, temp4, temp4S);
+    ONEITERBUTTERFLY(5, temp5, temp5S);
+    ONEITERBUTTERFLY(6, temp6, temp6S);
+    ONEITERBUTTERFLY(7, temp7, temp7S);
+  }
+#undef ONEITERBUTTERFLY
+  {
+    register vector signed int vsum;
+    register vector signed short line0 = vec_add(temp0, temp1);
+    register vector signed short line1 = vec_sub(temp0, temp1);
+    register vector signed short line2 = vec_add(temp2, temp3);
+    register vector signed short line3 = vec_sub(temp2, temp3);
+    register vector signed short line4 = vec_add(temp4, temp5);
+    register vector signed short line5 = vec_sub(temp4, temp5);
+    register vector signed short line6 = vec_add(temp6, temp7);
+    register vector signed short line7 = vec_sub(temp6, temp7);
+      
+    register vector signed short line0B = vec_add(line0, line2);
+    register vector signed short line2B = vec_sub(line0, line2);
+    register vector signed short line1B = vec_add(line1, line3);
+    register vector signed short line3B = vec_sub(line1, line3);
+    register vector signed short line4B = vec_add(line4, line6);
+    register vector signed short line6B = vec_sub(line4, line6);
+    register vector signed short line5B = vec_add(line5, line7);
+    register vector signed short line7B = vec_sub(line5, line7);
+      
+    register vector signed short line0C = vec_add(line0B, line4B);
+    register vector signed short line4C = vec_sub(line0B, line4B);
+    register vector signed short line1C = vec_add(line1B, line5B);
+    register vector signed short line5C = vec_sub(line1B, line5B);
+    register vector signed short line2C = vec_add(line2B, line6B);
+    register vector signed short line6C = vec_sub(line2B, line6B);
+    register vector signed short line3C = vec_add(line3B, line7B);
+    register vector signed short line7C = vec_sub(line3B, line7B);
+      
+    vsum = vec_sum4s(vec_abs(line0C), vec_splat_s32(0));
+    vsum = vec_sum4s(vec_abs(line1C), vsum);
+    vsum = vec_sum4s(vec_abs(line2C), vsum);
+    vsum = vec_sum4s(vec_abs(line3C), vsum);
+    vsum = vec_sum4s(vec_abs(line4C), vsum);
+    vsum = vec_sum4s(vec_abs(line5C), vsum);
+    vsum = vec_sum4s(vec_abs(line6C), vsum);
+    vsum = vec_sum4s(vec_abs(line7C), vsum);
+
+    register vector signed short line0S = vec_add(temp0S, temp1S);
+    register vector signed short line1S = vec_sub(temp0S, temp1S);
+    register vector signed short line2S = vec_add(temp2S, temp3S);
+    register vector signed short line3S = vec_sub(temp2S, temp3S);
+    register vector signed short line4S = vec_add(temp4S, temp5S);
+    register vector signed short line5S = vec_sub(temp4S, temp5S);
+    register vector signed short line6S = vec_add(temp6S, temp7S);
+    register vector signed short line7S = vec_sub(temp6S, temp7S);
+
+    register vector signed short line0BS = vec_add(line0S, line2S);
+    register vector signed short line2BS = vec_sub(line0S, line2S);
+    register vector signed short line1BS = vec_add(line1S, line3S);
+    register vector signed short line3BS = vec_sub(line1S, line3S);
+    register vector signed short line4BS = vec_add(line4S, line6S);
+    register vector signed short line6BS = vec_sub(line4S, line6S);
+    register vector signed short line5BS = vec_add(line5S, line7S);
+    register vector signed short line7BS = vec_sub(line5S, line7S);
+
+    register vector signed short line0CS = vec_add(line0BS, line4BS);
+    register vector signed short line4CS = vec_sub(line0BS, line4BS);
+    register vector signed short line1CS = vec_add(line1BS, line5BS);
+    register vector signed short line5CS = vec_sub(line1BS, line5BS);
+    register vector signed short line2CS = vec_add(line2BS, line6BS);
+    register vector signed short line6CS = vec_sub(line2BS, line6BS);
+    register vector signed short line3CS = vec_add(line3BS, line7BS);
+    register vector signed short line7CS = vec_sub(line3BS, line7BS);
+
+    vsum = vec_sum4s(vec_abs(line0CS), vsum);
+    vsum = vec_sum4s(vec_abs(line1CS), vsum);
+    vsum = vec_sum4s(vec_abs(line2CS), vsum);
+    vsum = vec_sum4s(vec_abs(line3CS), vsum);
+    vsum = vec_sum4s(vec_abs(line4CS), vsum);
+    vsum = vec_sum4s(vec_abs(line5CS), vsum);
+    vsum = vec_sum4s(vec_abs(line6CS), vsum);
+    vsum = vec_sum4s(vec_abs(line7CS), vsum);
+    vsum = vec_sums(vsum, (vector signed int)vzero);
+    vsum = vec_splat(vsum, 3);
+    vec_ste(vsum, 0, &sum);
+  }
+  return sum;
+}
+
+int hadamard8_diff16_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h){
+POWERPC_PERF_DECLARE(altivec_hadamard8_diff16_num, 1);
+  int score;
+POWERPC_PERF_START_COUNT(altivec_hadamard8_diff16_num, 1);
+  score = hadamard8_diff16x8_altivec(s, dst, src, stride, 8);
+  if (h==16) {
+    dst += 8*stride;
+    src += 8*stride;
+    score += hadamard8_diff16x8_altivec(s, dst, src, stride, 8);
+  }
+POWERPC_PERF_STOP_COUNT(altivec_hadamard8_diff16_num, 1);
+  return score;
+}
+
 int has_altivec(void)
 {
 #ifdef CONFIG_DARWIN
diff --git a/src/libffmpeg/libavcodec/ppc/dsputil_altivec.h b/src/libffmpeg/libavcodec/ppc/dsputil_altivec.h
index 93448a1ad..e2729ab22 100644
--- a/src/libffmpeg/libavcodec/ppc/dsputil_altivec.h
+++ b/src/libffmpeg/libavcodec/ppc/dsputil_altivec.h
@@ -1,6 +1,7 @@
 /*
  * Copyright (c) 2002 Brian Foley
  * Copyright (c) 2002 Dieter Shirley
+ * Copyright (c) 2003-2004 Romain Dolbeau <romain@dolbeau.org>
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
@@ -45,6 +46,8 @@ extern void put_pixels8_xy2_altivec(uint8_t *block, const uint8_t *pixels, int l
 extern void put_no_rnd_pixels8_xy2_altivec(uint8_t *block, const uint8_t *pixels, int line_size, int h);
 extern void put_pixels16_xy2_altivec(uint8_t * block, const uint8_t * pixels, int line_size, int h);
 extern void put_no_rnd_pixels16_xy2_altivec(uint8_t * block, const uint8_t * pixels, int line_size, int h);
+extern int hadamard8_diff8x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h);
+extern int hadamard8_diff16_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h);
 
 extern void gmc1_altivec(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y16, int rounder);
 
diff --git a/src/libffmpeg/libavcodec/ppc/dsputil_ppc.c b/src/libffmpeg/libavcodec/ppc/dsputil_ppc.c
index b8372e51e..b70de7328 100644
--- a/src/libffmpeg/libavcodec/ppc/dsputil_ppc.c
+++ b/src/libffmpeg/libavcodec/ppc/dsputil_ppc.c
@@ -1,6 +1,7 @@
 /*
  * Copyright (c) 2002 Brian Foley
  * Copyright (c) 2002 Dieter Shirley
+ * Copyright (c) 2003-2004 Romain Dolbeau <romain@dolbeau.org>
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
@@ -59,6 +60,8 @@ static unsigned char* perfname[] = {
   "put_no_rnd_pixels8_xy2_altivec",
   "put_pixels16_xy2_altivec",
   "put_no_rnd_pixels16_xy2_altivec",
+  "hadamard8_diff8x8_altivec",
+  "hadamard8_diff16_altivec",
   "clear_blocks_dcbz32_ppc",
   "clear_blocks_dcbz128_ppc"
 };
@@ -262,7 +265,7 @@ void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx)
         c->add_bytes= add_bytes_altivec;
 #endif /* 0 */
         c->put_pixels_tab[0][0] = put_pixels16_altivec;
-        /* the tow functions do the same thing, so use the same code */
+        /* the two functions do the same thing, so use the same code */
         c->put_no_rnd_pixels_tab[0][0] = put_pixels16_altivec;
         c->avg_pixels_tab[0][0] = avg_pixels16_altivec;
 // next one disabled as it's untested.
@@ -276,6 +279,9 @@ void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx)
         
 	c->gmc1 = gmc1_altivec;
 
+	c->hadamard8_diff[0] = hadamard8_diff16_altivec;
+	c->hadamard8_diff[1] = hadamard8_diff8x8_altivec;
+
 #ifdef CONFIG_ENCODERS
 	if (avctx->dct_algo == FF_DCT_AUTO ||
 	    avctx->dct_algo == FF_DCT_ALTIVEC)
diff --git a/src/libffmpeg/libavcodec/ppc/dsputil_ppc.h b/src/libffmpeg/libavcodec/ppc/dsputil_ppc.h
index d672edfcb..8b34c6b45 100644
--- a/src/libffmpeg/libavcodec/ppc/dsputil_ppc.h
+++ b/src/libffmpeg/libavcodec/ppc/dsputil_ppc.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003 Romain Dolbeau <romain@dolbeau.org>
+ * Copyright (c) 2003-2004 Romain Dolbeau <romain@dolbeau.org>
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
@@ -50,6 +50,8 @@ enum powerpc_perf_index {
   altivec_put_no_rnd_pixels8_xy2_num,
   altivec_put_pixels16_xy2_num,
   altivec_put_no_rnd_pixels16_xy2_num,
+  altivec_hadamard8_diff8x8_num,
+  altivec_hadamard8_diff16_num,
   powerpc_clear_blocks_dcbz32,
   powerpc_clear_blocks_dcbz128,
   powerpc_perf_total
@@ -63,6 +65,8 @@ enum powerpc_data_index {
 };
 extern unsigned long long perfdata[POWERPC_NUM_PMC_ENABLED][powerpc_perf_total][powerpc_data_total];
 
+#ifndef POWERPC_MODE_64BITS
+#define POWERP_PMC_DATATYPE unsigned long
 #define POWERPC_GET_PMC1(a) asm volatile("mfspr %0, 937" : "=r" (a))
 #define POWERPC_GET_PMC2(a) asm volatile("mfspr %0, 938" : "=r" (a))
 #if (POWERPC_NUM_PMC_ENABLED > 2)
@@ -79,7 +83,30 @@ extern unsigned long long perfdata[POWERPC_NUM_PMC_ENABLED][powerpc_perf_total][
 #define POWERPC_GET_PMC5(a) do {} while (0)
 #define POWERPC_GET_PMC6(a) do {} while (0)
 #endif
-#define POWERPC_PERF_DECLARE(a, cond) unsigned long pmc_start[POWERPC_NUM_PMC_ENABLED], pmc_stop[POWERPC_NUM_PMC_ENABLED], pmc_loop_index;
+#else /* POWERPC_MODE_64BITS */
+#define POWERP_PMC_DATATYPE unsigned long long
+#define POWERPC_GET_PMC1(a) asm volatile("mfspr %0, 771" : "=r" (a))
+#define POWERPC_GET_PMC2(a) asm volatile("mfspr %0, 772" : "=r" (a))
+#if (POWERPC_NUM_PMC_ENABLED > 2)
+#define POWERPC_GET_PMC3(a) asm volatile("mfspr %0, 773" : "=r" (a))
+#define POWERPC_GET_PMC4(a) asm volatile("mfspr %0, 774" : "=r" (a))
+#else
+#define POWERPC_GET_PMC3(a) do {} while (0)
+#define POWERPC_GET_PMC4(a) do {} while (0)
+#endif
+#if (POWERPC_NUM_PMC_ENABLED > 4)
+#define POWERPC_GET_PMC5(a) asm volatile("mfspr %0, 775" : "=r" (a))
+#define POWERPC_GET_PMC6(a) asm volatile("mfspr %0, 776" : "=r" (a))
+#else
+#define POWERPC_GET_PMC5(a) do {} while (0)
+#define POWERPC_GET_PMC6(a) do {} while (0)
+#endif
+#endif /* POWERPC_MODE_64BITS */
+#define POWERPC_PERF_DECLARE(a, cond)				\
+  POWERP_PMC_DATATYPE						\
+    pmc_start[POWERPC_NUM_PMC_ENABLED],				\
+    pmc_stop[POWERPC_NUM_PMC_ENABLED],				\
+    pmc_loop_index;
 #define POWERPC_PERF_START_COUNT(a, cond) do { \
   POWERPC_GET_PMC6(pmc_start[5]); \
   POWERPC_GET_PMC5(pmc_start[4]); \
@@ -101,9 +128,9 @@ extern unsigned long long perfdata[POWERPC_NUM_PMC_ENABLED][powerpc_perf_total][
         pmc_loop_index < POWERPC_NUM_PMC_ENABLED; \
         pmc_loop_index++)         \
     {                             \
-      if (pmc_stop[pmc_loop_index] >= pmc_start[pmc_loop_index]) \
-      {                           \
-        unsigned long diff =      \
+      if (pmc_stop[pmc_loop_index] >= pmc_start[pmc_loop_index])  \
+	{							  \
+        POWERP_PMC_DATATYPE diff =				  \
           pmc_stop[pmc_loop_index] - pmc_start[pmc_loop_index];   \
         if (diff < perfdata[pmc_loop_index][a][powerpc_data_min]) \
           perfdata[pmc_loop_index][a][powerpc_data_min] = diff;   \
diff --git a/src/libffmpeg/libavcodec/ppc/fdct_altivec.c b/src/libffmpeg/libavcodec/ppc/fdct_altivec.c
new file mode 100644
index 000000000..99df5ced3
--- /dev/null
+++ b/src/libffmpeg/libavcodec/ppc/fdct_altivec.c
@@ -0,0 +1,498 @@
+/* ffmpeg/libavcodec/ppc/fdct_altivec.c, this file is part of the
+ * AltiVec optimized library for the FFMPEG Multimedia System
+ * Copyright (C) 2003  James Klicman <james@klicman.org>
+ *
+ * This library 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.
+ *
+ * This library 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 this library; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+
+#include "../common.h"
+#include "../dsputil.h"
+#include "dsputil_altivec.h"
+#include "gcc_fixes.h"
+
+
+#define vs16(v) ((vector signed short)(v))
+#define vs32(v) ((vector signed int)(v))
+#define vu8(v)  ((vector unsigned char)(v))
+#define vu16(v) ((vector unsigned short)(v))
+#define vu32(v) ((vector unsigned int)(v))
+
+
+#define C1     0.98078525066375732421875000 /* cos(1*PI/16) */
+#define C2     0.92387950420379638671875000 /* cos(2*PI/16) */
+#define C3     0.83146959543228149414062500 /* cos(3*PI/16) */
+#define C4     0.70710676908493041992187500 /* cos(4*PI/16) */
+#define C5     0.55557024478912353515625000 /* cos(5*PI/16) */
+#define C6     0.38268342614173889160156250 /* cos(6*PI/16) */
+#define C7     0.19509032368659973144531250 /* cos(7*PI/16) */
+#define SQRT_2 1.41421353816986083984375000 /* sqrt(2)      */
+
+
+#define W0 -(2 * C2)
+#define W1 (2 * C6)
+#define W2 (SQRT_2 * C6)
+#define W3 (SQRT_2 * C3)
+#define W4 (SQRT_2 * (-C1 + C3 + C5 - C7))
+#define W5 (SQRT_2 * ( C1 + C3 - C5 + C7))
+#define W6 (SQRT_2 * ( C1 + C3 + C5 - C7))
+#define W7 (SQRT_2 * ( C1 + C3 - C5 - C7))
+#define W8 (SQRT_2 * ( C7 - C3))
+#define W9 (SQRT_2 * (-C1 - C3))
+#define WA (SQRT_2 * (-C3 - C5))
+#define WB (SQRT_2 * ( C5 - C3))
+
+
+static vector float fdctconsts[3] = {
+    (vector float)AVV( W0, W1, W2, W3 ),
+    (vector float)AVV( W4, W5, W6, W7 ),
+    (vector float)AVV( W8, W9, WA, WB )
+};
+
+#define LD_W0 vec_splat(cnsts0, 0)
+#define LD_W1 vec_splat(cnsts0, 1)
+#define LD_W2 vec_splat(cnsts0, 2)
+#define LD_W3 vec_splat(cnsts0, 3)
+#define LD_W4 vec_splat(cnsts1, 0)
+#define LD_W5 vec_splat(cnsts1, 1)
+#define LD_W6 vec_splat(cnsts1, 2)
+#define LD_W7 vec_splat(cnsts1, 3)
+#define LD_W8 vec_splat(cnsts2, 0)
+#define LD_W9 vec_splat(cnsts2, 1)
+#define LD_WA vec_splat(cnsts2, 2)
+#define LD_WB vec_splat(cnsts2, 3)
+
+
+#define FDCTROW(b0,b1,b2,b3,b4,b5,b6,b7) /* {{{ */                  \
+    x0 = vec_add(b0, b7);               /* x0 = b0 + b7; */         \
+    x7 = vec_sub(b0, b7);               /* x7 = b0 - b7; */         \
+    x1 = vec_add(b1, b6);               /* x1 = b1 + b6; */         \
+    x6 = vec_sub(b1, b6);               /* x6 = b1 - b6; */         \
+    x2 = vec_add(b2, b5);               /* x2 = b2 + b5; */         \
+    x5 = vec_sub(b2, b5);               /* x5 = b2 - b5; */         \
+    x3 = vec_add(b3, b4);               /* x3 = b3 + b4; */         \
+    x4 = vec_sub(b3, b4);               /* x4 = b3 - b4; */         \
+                                                                    \
+    b7 = vec_add(x0, x3);               /* b7 = x0 + x3; */         \
+    b1 = vec_add(x1, x2);               /* b1 = x1 + x2; */         \
+    b0 = vec_add(b7, b1);               /* b0 = b7 + b1; */         \
+    b4 = vec_sub(b7, b1);               /* b4 = b7 - b1; */         \
+                                                                    \
+    b2 = vec_sub(x0, x3);               /* b2 = x0 - x3; */         \
+    b6 = vec_sub(x1, x2);               /* b6 = x1 - x2; */         \
+    b5 = vec_add(b6, b2);               /* b5 = b6 + b2; */         \
+    cnst = LD_W2;                                                   \
+    b5 = vec_madd(cnst, b5, mzero);     /* b5 = b5 * W2; */         \
+    cnst = LD_W1;                                                   \
+    b2 = vec_madd(cnst, b2, b5);        /* b2 = b5 + b2 * W1; */    \
+    cnst = LD_W0;                                                   \
+    b6 = vec_madd(cnst, b6, b5);        /* b6 = b5 + b6 * W0; */    \
+                                                                    \
+    x0 = vec_add(x4, x7);               /* x0 = x4 + x7; */         \
+    x1 = vec_add(x5, x6);               /* x1 = x5 + x6; */         \
+    x2 = vec_add(x4, x6);               /* x2 = x4 + x6; */         \
+    x3 = vec_add(x5, x7);               /* x3 = x5 + x7; */         \
+    x8 = vec_add(x2, x3);               /* x8 = x2 + x3; */         \
+    cnst = LD_W3;                                                   \
+    x8 = vec_madd(cnst, x8, mzero);     /* x8 = x8 * W3; */         \
+                                                                    \
+    cnst = LD_W8;                                                   \
+    x0 = vec_madd(cnst, x0, mzero);     /* x0 *= W8; */             \
+    cnst = LD_W9;                                                   \
+    x1 = vec_madd(cnst, x1, mzero);     /* x1 *= W9; */             \
+    cnst = LD_WA;                                                   \
+    x2 = vec_madd(cnst, x2, x8);        /* x2 = x2 * WA + x8; */    \
+    cnst = LD_WB;                                                   \
+    x3 = vec_madd(cnst, x3, x8);        /* x3 = x3 * WB + x8; */    \
+                                                                    \
+    cnst = LD_W4;                                                   \
+    b7 = vec_madd(cnst, x4, x0);        /* b7 = x4 * W4 + x0; */    \
+    cnst = LD_W5;                                                   \
+    b5 = vec_madd(cnst, x5, x1);        /* b5 = x5 * W5 + x1; */    \
+    cnst = LD_W6;                                                   \
+    b3 = vec_madd(cnst, x6, x1);        /* b3 = x6 * W6 + x1; */    \
+    cnst = LD_W7;                                                   \
+    b1 = vec_madd(cnst, x7, x0);        /* b1 = x7 * W7 + x0; */    \
+                                                                    \
+    b7 = vec_add(b7, x2);               /* b7 = b7 + x2; */         \
+    b5 = vec_add(b5, x3);               /* b5 = b5 + x3; */         \
+    b3 = vec_add(b3, x2);               /* b3 = b3 + x2; */         \
+    b1 = vec_add(b1, x3);               /* b1 = b1 + x3; */         \
+    /* }}} */
+
+#define FDCTCOL(b0,b1,b2,b3,b4,b5,b6,b7) /* {{{ */                  \
+    x0 = vec_add(b0, b7);               /* x0 = b0 + b7; */         \
+    x7 = vec_sub(b0, b7);               /* x7 = b0 - b7; */         \
+    x1 = vec_add(b1, b6);               /* x1 = b1 + b6; */         \
+    x6 = vec_sub(b1, b6);               /* x6 = b1 - b6; */         \
+    x2 = vec_add(b2, b5);               /* x2 = b2 + b5; */         \
+    x5 = vec_sub(b2, b5);               /* x5 = b2 - b5; */         \
+    x3 = vec_add(b3, b4);               /* x3 = b3 + b4; */         \
+    x4 = vec_sub(b3, b4);               /* x4 = b3 - b4; */         \
+                                                                    \
+    b7 = vec_add(x0, x3);               /* b7 = x0 + x3; */         \
+    b1 = vec_add(x1, x2);               /* b1 = x1 + x2; */         \
+    b0 = vec_add(b7, b1);               /* b0 = b7 + b1; */         \
+    b4 = vec_sub(b7, b1);               /* b4 = b7 - b1; */         \
+                                                                    \
+    b2 = vec_sub(x0, x3);               /* b2 = x0 - x3; */         \
+    b6 = vec_sub(x1, x2);               /* b6 = x1 - x2; */         \
+    b5 = vec_add(b6, b2);               /* b5 = b6 + b2; */         \
+    cnst = LD_W2;                                                   \
+    b5 = vec_madd(cnst, b5, mzero);     /* b5 = b5 * W2; */         \
+    cnst = LD_W1;                                                   \
+    b2 = vec_madd(cnst, b2, b5);        /* b2 = b5 + b2 * W1; */    \
+    cnst = LD_W0;                                                   \
+    b6 = vec_madd(cnst, b6, b5);        /* b6 = b5 + b6 * W0; */    \
+                                                                    \
+    x0 = vec_add(x4, x7);               /* x0 = x4 + x7; */         \
+    x1 = vec_add(x5, x6);               /* x1 = x5 + x6; */         \
+    x2 = vec_add(x4, x6);               /* x2 = x4 + x6; */         \
+    x3 = vec_add(x5, x7);               /* x3 = x5 + x7; */         \
+    x8 = vec_add(x2, x3);               /* x8 = x2 + x3; */         \
+    cnst = LD_W3;                                                   \
+    x8 = vec_madd(cnst, x8, mzero);     /* x8 = x8 * W3; */         \
+                                                                    \
+    cnst = LD_W8;                                                   \
+    x0 = vec_madd(cnst, x0, mzero);     /* x0 *= W8; */             \
+    cnst = LD_W9;                                                   \
+    x1 = vec_madd(cnst, x1, mzero);     /* x1 *= W9; */             \
+    cnst = LD_WA;                                                   \
+    x2 = vec_madd(cnst, x2, x8);        /* x2 = x2 * WA + x8; */    \
+    cnst = LD_WB;                                                   \
+    x3 = vec_madd(cnst, x3, x8);        /* x3 = x3 * WB + x8; */    \
+                                                                    \
+    cnst = LD_W4;                                                   \
+    b7 = vec_madd(cnst, x4, x0);        /* b7 = x4 * W4 + x0; */    \
+    cnst = LD_W5;                                                   \
+    b5 = vec_madd(cnst, x5, x1);        /* b5 = x5 * W5 + x1; */    \
+    cnst = LD_W6;                                                   \
+    b3 = vec_madd(cnst, x6, x1);        /* b3 = x6 * W6 + x1; */    \
+    cnst = LD_W7;                                                   \
+    b1 = vec_madd(cnst, x7, x0);        /* b1 = x7 * W7 + x0; */    \
+                                                                    \
+    b7 = vec_add(b7, x2);               /* b7 += x2; */             \
+    b5 = vec_add(b5, x3);               /* b5 += x3; */             \
+    b3 = vec_add(b3, x2);               /* b3 += x2; */             \
+    b1 = vec_add(b1, x3);               /* b1 += x3; */             \
+    /* }}} */
+
+
+
+/* two dimensional discrete cosine transform */
+
+void fdct_altivec(int16_t *block)
+{
+POWERPC_PERF_DECLARE(altivec_fdct, 1);
+#ifdef ALTIVEC_USE_REFERENCE_C_CODE
+POWERPC_PERF_START_COUNT(altivec_fdct, 1);
+    void ff_jpeg_fdct_islow(int16_t *block);
+    ff_jpeg_fdct_islow(block);
+POWERPC_PERF_STOP_COUNT(altivec_fdct, 1);
+#else /* ALTIVEC_USE_REFERENCE_C_CODE */
+    vector signed short *bp;
+    vector float *cp;
+    vector float b00, b10, b20, b30, b40, b50, b60, b70;
+    vector float b01, b11, b21, b31, b41, b51, b61, b71;
+    vector float mzero, cnst, cnsts0, cnsts1, cnsts2;
+    vector float x0, x1, x2, x3, x4, x5, x6, x7, x8;
+
+    POWERPC_PERF_START_COUNT(altivec_fdct, 1);
+
+
+    /* setup constants {{{ */
+    /* mzero = -0.0 */
+    vu32(mzero) = vec_splat_u32(-1);
+    vu32(mzero) = vec_sl(vu32(mzero), vu32(mzero));
+    cp = fdctconsts;
+    cnsts0 = vec_ld(0, cp); cp++;
+    cnsts1 = vec_ld(0, cp); cp++;
+    cnsts2 = vec_ld(0, cp);
+    /* }}} */
+
+
+    /* 8x8 matrix transpose (vector short[8]) {{{ */
+#define MERGE_S16(hl,a,b) vec_merge##hl(vs16(a), vs16(b))
+
+    bp = (vector signed short*)block;
+    vs16(b00) = vec_ld(0,    bp);
+    vs16(b40) = vec_ld(16*4, bp);
+    vs16(b01) = MERGE_S16(h, b00, b40);
+    vs16(b11) = MERGE_S16(l, b00, b40);
+    bp++;
+    vs16(b10) = vec_ld(0,    bp);
+    vs16(b50) = vec_ld(16*4, bp);
+    vs16(b21) = MERGE_S16(h, b10, b50);
+    vs16(b31) = MERGE_S16(l, b10, b50);
+    bp++;
+    vs16(b20) = vec_ld(0,    bp);
+    vs16(b60) = vec_ld(16*4, bp);
+    vs16(b41) = MERGE_S16(h, b20, b60);
+    vs16(b51) = MERGE_S16(l, b20, b60);
+    bp++;
+    vs16(b30) = vec_ld(0,    bp);
+    vs16(b70) = vec_ld(16*4, bp);
+    vs16(b61) = MERGE_S16(h, b30, b70);
+    vs16(b71) = MERGE_S16(l, b30, b70);
+
+    vs16(x0) = MERGE_S16(h, b01, b41);
+    vs16(x1) = MERGE_S16(l, b01, b41);
+    vs16(x2) = MERGE_S16(h, b11, b51);
+    vs16(x3) = MERGE_S16(l, b11, b51);
+    vs16(x4) = MERGE_S16(h, b21, b61);
+    vs16(x5) = MERGE_S16(l, b21, b61);
+    vs16(x6) = MERGE_S16(h, b31, b71);
+    vs16(x7) = MERGE_S16(l, b31, b71);
+
+    vs16(b00) = MERGE_S16(h, x0, x4);
+    vs16(b10) = MERGE_S16(l, x0, x4);
+    vs16(b20) = MERGE_S16(h, x1, x5);
+    vs16(b30) = MERGE_S16(l, x1, x5);
+    vs16(b40) = MERGE_S16(h, x2, x6);
+    vs16(b50) = MERGE_S16(l, x2, x6);
+    vs16(b60) = MERGE_S16(h, x3, x7);
+    vs16(b70) = MERGE_S16(l, x3, x7);
+
+#undef MERGE_S16
+    /* }}} */
+
+
+/* Some of the initial calculations can be done as vector short before
+ * conversion to vector float.  The following code section takes advantage
+ * of this.
+ */
+#if 1
+    /* fdct rows {{{ */
+    vs16(x0) = vec_add(vs16(b00), vs16(b70));
+    vs16(x7) = vec_sub(vs16(b00), vs16(b70));
+    vs16(x1) = vec_add(vs16(b10), vs16(b60));
+    vs16(x6) = vec_sub(vs16(b10), vs16(b60));
+    vs16(x2) = vec_add(vs16(b20), vs16(b50));
+    vs16(x5) = vec_sub(vs16(b20), vs16(b50));
+    vs16(x3) = vec_add(vs16(b30), vs16(b40));
+    vs16(x4) = vec_sub(vs16(b30), vs16(b40));
+
+    vs16(b70) = vec_add(vs16(x0), vs16(x3));
+    vs16(b10) = vec_add(vs16(x1), vs16(x2));
+
+    vs16(b00) = vec_add(vs16(b70), vs16(b10));
+    vs16(b40) = vec_sub(vs16(b70), vs16(b10));
+
+#define CTF0(n) \
+    vs32(b##n##1) = vec_unpackl(vs16(b##n##0)); \
+    vs32(b##n##0) = vec_unpackh(vs16(b##n##0)); \
+    b##n##1 = vec_ctf(vs32(b##n##1), 0); \
+    b##n##0 = vec_ctf(vs32(b##n##0), 0);
+
+    CTF0(0);
+    CTF0(4);
+
+    vs16(b20) = vec_sub(vs16(x0), vs16(x3));
+    vs16(b60) = vec_sub(vs16(x1), vs16(x2));
+
+    CTF0(2);
+    CTF0(6);
+
+#undef CTF0
+
+    x0 = vec_add(b60, b20);
+    x1 = vec_add(b61, b21);
+
+    cnst = LD_W2;
+    x0 = vec_madd(cnst, x0, mzero);
+    x1 = vec_madd(cnst, x1, mzero);
+    cnst = LD_W1;
+    b20 = vec_madd(cnst, b20, x0);
+    b21 = vec_madd(cnst, b21, x1);
+    cnst = LD_W0;
+    b60 = vec_madd(cnst, b60, x0);
+    b61 = vec_madd(cnst, b61, x1);
+
+#define CTFX(x,b) \
+    vs32(b##0) = vec_unpackh(vs16(x)); \
+    vs32(b##1) = vec_unpackl(vs16(x)); \
+    b##0 = vec_ctf(vs32(b##0), 0); \
+    b##1 = vec_ctf(vs32(b##1), 0); \
+
+    CTFX(x4, b7);
+    CTFX(x5, b5);
+    CTFX(x6, b3);
+    CTFX(x7, b1);
+
+#undef CTFX
+
+
+    x0 = vec_add(b70, b10);
+    x1 = vec_add(b50, b30);
+    x2 = vec_add(b70, b30);
+    x3 = vec_add(b50, b10);
+    x8 = vec_add(x2, x3);
+    cnst = LD_W3;
+    x8 = vec_madd(cnst, x8, mzero);
+
+    cnst = LD_W8;
+    x0 = vec_madd(cnst, x0, mzero);
+    cnst = LD_W9;
+    x1 = vec_madd(cnst, x1, mzero);
+    cnst = LD_WA;
+    x2 = vec_madd(cnst, x2, x8);
+    cnst = LD_WB;
+    x3 = vec_madd(cnst, x3, x8);
+
+    cnst = LD_W4;
+    b70 = vec_madd(cnst, b70, x0);
+    cnst = LD_W5;
+    b50 = vec_madd(cnst, b50, x1);
+    cnst = LD_W6;
+    b30 = vec_madd(cnst, b30, x1);
+    cnst = LD_W7;
+    b10 = vec_madd(cnst, b10, x0);
+
+    b70 = vec_add(b70, x2);
+    b50 = vec_add(b50, x3);
+    b30 = vec_add(b30, x2);
+    b10 = vec_add(b10, x3);
+
+
+    x0 = vec_add(b71, b11);
+    x1 = vec_add(b51, b31);
+    x2 = vec_add(b71, b31);
+    x3 = vec_add(b51, b11);
+    x8 = vec_add(x2, x3);
+    cnst = LD_W3;
+    x8 = vec_madd(cnst, x8, mzero);
+
+    cnst = LD_W8;
+    x0 = vec_madd(cnst, x0, mzero);
+    cnst = LD_W9;
+    x1 = vec_madd(cnst, x1, mzero);
+    cnst = LD_WA;
+    x2 = vec_madd(cnst, x2, x8);
+    cnst = LD_WB;
+    x3 = vec_madd(cnst, x3, x8);
+
+    cnst = LD_W4;
+    b71 = vec_madd(cnst, b71, x0);
+    cnst = LD_W5;
+    b51 = vec_madd(cnst, b51, x1);
+    cnst = LD_W6;
+    b31 = vec_madd(cnst, b31, x1);
+    cnst = LD_W7;
+    b11 = vec_madd(cnst, b11, x0);
+
+    b71 = vec_add(b71, x2);
+    b51 = vec_add(b51, x3);
+    b31 = vec_add(b31, x2);
+    b11 = vec_add(b11, x3);
+    /* }}} */
+#else
+    /* convert to float {{{ */
+#define CTF(n) \
+    vs32(b##n##1) = vec_unpackl(vs16(b##n##0)); \
+    vs32(b##n##0) = vec_unpackh(vs16(b##n##0)); \
+    b##n##1 = vec_ctf(vs32(b##n##1), 0); \
+    b##n##0 = vec_ctf(vs32(b##n##0), 0); \
+
+    CTF(0);
+    CTF(1);
+    CTF(2);
+    CTF(3);
+    CTF(4);
+    CTF(5);
+    CTF(6);
+    CTF(7);
+
+#undef CTF
+    /* }}} */
+
+    FDCTROW(b00, b10, b20, b30, b40, b50, b60, b70);
+    FDCTROW(b01, b11, b21, b31, b41, b51, b61, b71);
+#endif
+
+
+    /* 8x8 matrix transpose (vector float[8][2]) {{{ */
+    x0 = vec_mergel(b00, b20);
+    x1 = vec_mergeh(b00, b20);
+    x2 = vec_mergel(b10, b30);
+    x3 = vec_mergeh(b10, b30);
+
+    b00 = vec_mergeh(x1, x3);
+    b10 = vec_mergel(x1, x3);
+    b20 = vec_mergeh(x0, x2);
+    b30 = vec_mergel(x0, x2);
+
+    x4 = vec_mergel(b41, b61);
+    x5 = vec_mergeh(b41, b61);
+    x6 = vec_mergel(b51, b71);
+    x7 = vec_mergeh(b51, b71);
+
+    b41 = vec_mergeh(x5, x7);
+    b51 = vec_mergel(x5, x7);
+    b61 = vec_mergeh(x4, x6);
+    b71 = vec_mergel(x4, x6);
+
+    x0 = vec_mergel(b01, b21);
+    x1 = vec_mergeh(b01, b21);
+    x2 = vec_mergel(b11, b31);
+    x3 = vec_mergeh(b11, b31);
+
+    x4 = vec_mergel(b40, b60);
+    x5 = vec_mergeh(b40, b60);
+    x6 = vec_mergel(b50, b70);
+    x7 = vec_mergeh(b50, b70);
+
+    b40 = vec_mergeh(x1, x3);
+    b50 = vec_mergel(x1, x3);
+    b60 = vec_mergeh(x0, x2);
+    b70 = vec_mergel(x0, x2);
+
+    b01 = vec_mergeh(x5, x7);
+    b11 = vec_mergel(x5, x7);
+    b21 = vec_mergeh(x4, x6);
+    b31 = vec_mergel(x4, x6);
+    /* }}} */
+
+
+    FDCTCOL(b00, b10, b20, b30, b40, b50, b60, b70);
+    FDCTCOL(b01, b11, b21, b31, b41, b51, b61, b71);
+
+
+    /* round, convert back to short {{{ */
+#define CTS(n) \
+    b##n##0 = vec_round(b##n##0); \
+    b##n##1 = vec_round(b##n##1); \
+    vs32(b##n##0) = vec_cts(b##n##0, 0); \
+    vs32(b##n##1) = vec_cts(b##n##1, 0); \
+    vs16(b##n##0) = vec_pack(vs32(b##n##0), vs32(b##n##1)); \
+    vec_st(vs16(b##n##0), 0, bp);
+
+    bp = (vector signed short*)block;
+    CTS(0); bp++;
+    CTS(1); bp++;
+    CTS(2); bp++;
+    CTS(3); bp++;
+    CTS(4); bp++;
+    CTS(5); bp++;
+    CTS(6); bp++;
+    CTS(7);
+
+#undef CTS
+    /* }}} */
+
+POWERPC_PERF_STOP_COUNT(altivec_fdct, 1);
+#endif /* ALTIVEC_USE_REFERENCE_C_CODE */
+}
+
+/* vim:set foldmethod=marker foldlevel=0: */
diff --git a/src/libffmpeg/libavcodec/ppc/mpegvideo_altivec.c b/src/libffmpeg/libavcodec/ppc/mpegvideo_altivec.c
index 51b387792..91e744af9 100644
--- a/src/libffmpeg/libavcodec/ppc/mpegvideo_altivec.c
+++ b/src/libffmpeg/libavcodec/ppc/mpegvideo_altivec.c
@@ -1,6 +1,9 @@
 /*
  * Copyright (c) 2002 Dieter Shirley
  *
+ * dct_unquantize_h263_altivec:
+ * Copyright (c) 2003 Romain Dolbeau <romain@dolbeau.org>
+ *
  * This library 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
diff --git a/src/libffmpeg/libavcodec/ppc/mpegvideo_ppc.c b/src/libffmpeg/libavcodec/ppc/mpegvideo_ppc.c
index ce4bf8a47..c8269eb9a 100644
--- a/src/libffmpeg/libavcodec/ppc/mpegvideo_ppc.c
+++ b/src/libffmpeg/libavcodec/ppc/mpegvideo_ppc.c
@@ -1,84 +1,86 @@
-/*

- * Copyright (c) 2002 Dieter Shirley

- *

- * This library 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 of the License, or (at your option) any later version.

- *

- * This library 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 this library; if not, write to the Free Software

- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

- */

- 

-#include "../dsputil.h"

-#include "../mpegvideo.h"

-#include <time.h>

-

-#ifdef HAVE_ALTIVEC

-#include "dsputil_altivec.h"

-#endif

-

-extern int dct_quantize_altivec(MpegEncContext *s,  

-        DCTELEM *block, int n,

-        int qscale, int *overflow);

+/*
+ * Copyright (c) 2002 Dieter Shirley
+ *
+ * This library 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 of the License, or (at your option) any later version.
+ *
+ * This library 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 this library; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+ 
+#include "../dsputil.h"
+#include "../mpegvideo.h"
+#include <time.h>
+
+#ifdef HAVE_ALTIVEC
+#include "dsputil_altivec.h"
+#endif
+
+extern int dct_quantize_altivec(MpegEncContext *s,  
+        DCTELEM *block, int n,
+        int qscale, int *overflow);
 extern void dct_unquantize_h263_altivec(MpegEncContext *s,
                                         DCTELEM *block, int n, int qscale);
-

-extern void idct_put_altivec(uint8_t *dest, int line_size, int16_t *block);

-extern void idct_add_altivec(uint8_t *dest, int line_size, int16_t *block);

-

-

-void MPV_common_init_ppc(MpegEncContext *s)

-{

+
+extern void idct_put_altivec(uint8_t *dest, int line_size, int16_t *block);
+extern void idct_add_altivec(uint8_t *dest, int line_size, int16_t *block);
+
+
+void MPV_common_init_ppc(MpegEncContext *s)
+{
 #ifdef HAVE_ALTIVEC
-    if (has_altivec())

-    {

-        if ((s->avctx->idct_algo == FF_IDCT_AUTO) ||

-                (s->avctx->idct_algo == FF_IDCT_ALTIVEC))

-        {

-            s->dsp.idct_put = idct_put_altivec;

-            s->dsp.idct_add = idct_add_altivec;

+    if (has_altivec())
+    {
+        if ((s->avctx->idct_algo == FF_IDCT_AUTO) ||
+                (s->avctx->idct_algo == FF_IDCT_ALTIVEC))
+        {
+            s->dsp.idct_put = idct_put_altivec;
+            s->dsp.idct_add = idct_add_altivec;
 #ifndef ALTIVEC_USE_REFERENCE_C_CODE
-            s->dsp.idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;

+            s->dsp.idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
 #else /* ALTIVEC_USE_REFERENCE_C_CODE */
             s->dsp.idct_permutation_type = FF_NO_IDCT_PERM;
 #endif /* ALTIVEC_USE_REFERENCE_C_CODE */
-        }

-

-        // Test to make sure that the dct required alignments are met.

-        if ((((long)(s->q_intra_matrix) & 0x0f) != 0) ||

-                (((long)(s->q_inter_matrix) & 0x0f) != 0))

-        {

-            av_log(s->avctx, AV_LOG_INFO, "Internal Error: q-matrix blocks must be 16-byte aligned "

-                    "to use Altivec DCT. Reverting to non-altivec version.\n");

-            return;

-        }

-

-        if (((long)(s->intra_scantable.inverse) & 0x0f) != 0)

-        {

-            av_log(s->avctx, AV_LOG_INFO, "Internal Error: scan table blocks must be 16-byte aligned "

-                    "to use Altivec DCT. Reverting to non-altivec version.\n");

-            return;

-        }

-

-

-        if ((s->avctx->dct_algo == FF_DCT_AUTO) ||

-                (s->avctx->dct_algo == FF_DCT_ALTIVEC))

-        {

-            s->dct_quantize = dct_quantize_altivec;

+        }
+
+        // Test to make sure that the dct required alignments are met.
+        if ((((long)(s->q_intra_matrix) & 0x0f) != 0) ||
+                (((long)(s->q_inter_matrix) & 0x0f) != 0))
+        {
+            av_log(s->avctx, AV_LOG_INFO, "Internal Error: q-matrix blocks must be 16-byte aligned "
+                    "to use Altivec DCT. Reverting to non-altivec version.\n");
+            return;
+        }
+
+        if (((long)(s->intra_scantable.inverse) & 0x0f) != 0)
+        {
+            av_log(s->avctx, AV_LOG_INFO, "Internal Error: scan table blocks must be 16-byte aligned "
+                    "to use Altivec DCT. Reverting to non-altivec version.\n");
+            return;
+        }
+
+
+        if ((s->avctx->dct_algo == FF_DCT_AUTO) ||
+                (s->avctx->dct_algo == FF_DCT_ALTIVEC))
+        {
+#if 0 /* seems to cause trouble under some circumstances */
+            s->dct_quantize = dct_quantize_altivec;
+#endif
             s->dct_unquantize_h263_intra = dct_unquantize_h263_altivec;
             s->dct_unquantize_h263_inter = dct_unquantize_h263_altivec;
-        }

-    } else

-#endif

-    {

-        /* Non-AltiVec PPC optimisations here */

-    }

-}

-

+        }
+    } else
+#endif
+    {
+        /* Non-AltiVec PPC optimisations here */
+    }
+}
+