-Index: /trunk/common/ppc/quant.c\r
-===================================================================\r
---- /trunk/common/ppc/quant.c (revision 601)\r
-+++ /trunk/common/ppc/quant.c (revision 621)\r
-@@ -18,8 +18,4 @@\r
- * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.\r
- *****************************************************************************/\r
--\r
--#ifdef HAVE_ALTIVEC_H\r
--#include <altivec.h>\r
--#endif\r
- \r
- #include "common/common.h"\r
-@@ -54,29 +50,29 @@\r
- temp2v = vec_xor(temp2v, mskB); \\r
- temp1v = vec_adds(temp1v, vec_and(mskA, one)); \\r
--vec_st(temp1v, (dct0), dct); \\r
-+vec_st(temp1v, (dct0), (int16_t*)dct); \\r
- temp2v = vec_adds(temp2v, vec_and(mskB, one)); \\r
--vec_st(temp2v, (dct1), dct);\r
-+vec_st(temp2v, (dct1), (int16_t*)dct);\r
- \r
- void x264_quant_4x4_altivec( int16_t dct[4][4], int quant_mf[4][4], int const i_qbits, int const f ) {\r
- vector bool short mskA;\r
-- vec_s32_t i_qbitsv;\r
-+ vec_u32_t i_qbitsv;\r
- vec_u16_t coefvA;\r
- vec_u32_t multEvenvA, multOddvA;\r
-- vec_u32_t mfvA;\r
-+ vec_u16_t mfvA;\r
- vec_s16_t zerov, one;\r
-- vec_s32_t fV;\r
-+ vec_u32_t fV;\r
- \r
- vector bool short mskB;\r
- vec_u16_t coefvB;\r
- vec_u32_t multEvenvB, multOddvB;\r
-- vec_u32_t mfvB;\r
-+ vec_u16_t mfvB;\r
- \r
- vec_s16_t temp1v, temp2v;\r
- \r
-- vect_sint_u qbits_u;\r
-+ vect_int_u qbits_u;\r
- qbits_u.s[0]=i_qbits;\r
- i_qbitsv = vec_splat(qbits_u.v, 0);\r
- \r
-- vect_sint_u f_u;\r
-+ vect_int_u f_u;\r
- f_u.s[0]=f;\r
- \r
-@@ -114,16 +110,16 @@\r
- temp2v = vec_xor(temp2v, mskB); \\r
- temp1v = vec_add(temp1v, vec_and(mskA, one)); \\r
--vec_st(temp1v, (dct0), dct); \\r
-+vec_st(temp1v, (dct0), (int16_t*)dct); \\r
- temp2v = vec_add(temp2v, vec_and(mskB, one)); \\r
--vec_st(temp2v, (dct1), dct);\r
-+vec_st(temp2v, (dct1), (int16_t*)dct);\r
- \r
- \r
- void x264_quant_4x4_dc_altivec( int16_t dct[4][4], int i_quant_mf, int const i_qbits, int const f ) {\r
- vector bool short mskA;\r
-- vec_s32_t i_qbitsv;\r
-+ vec_u32_t i_qbitsv;\r
- vec_u16_t coefvA;\r
- vec_u32_t multEvenvA, multOddvA;\r
- vec_s16_t zerov, one;\r
-- vec_s32_t fV;\r
-+ vec_u32_t fV;\r
- \r
- vector bool short mskB;\r
-@@ -133,15 +129,14 @@\r
- vec_s16_t temp1v, temp2v;\r
- \r
-- vec_u32_t mfv;\r
-- vect_int_u mf_u;\r
-+ vec_u16_t mfv;\r
-+ vect_ushort_u mf_u;\r
- mf_u.s[0]=i_quant_mf;\r
- mfv = vec_splat( mf_u.v, 0 );\r
-- mfv = vec_packs( mfv, mfv);\r
- \r
-- vect_sint_u qbits_u;\r
-+ vect_int_u qbits_u;\r
- qbits_u.s[0]=i_qbits;\r
- i_qbitsv = vec_splat(qbits_u.v, 0);\r
- \r
-- vect_sint_u f_u;\r
-+ vect_int_u f_u;\r
- f_u.s[0]=f;\r
- fV = vec_splat(f_u.v, 0);\r
-@@ -156,13 +151,15 @@\r
- void x264_quant_8x8_altivec( int16_t dct[8][8], int quant_mf[8][8], int const i_qbits, int const f ) {\r
- vector bool short mskA;\r
-- vec_s32_t i_qbitsv;\r
-+ vec_u32_t i_qbitsv;\r
- vec_u16_t coefvA;\r
-- vec_s32_t multEvenvA, multOddvA, mfvA;\r
-+ vec_u32_t multEvenvA, multOddvA;\r
-+ vec_u16_t mfvA;\r
- vec_s16_t zerov, one;\r
-- vec_s32_t fV;\r
-+ vec_u32_t fV;\r
- \r
- vector bool short mskB;\r
- vec_u16_t coefvB;\r
-- vec_u32_t multEvenvB, multOddvB, mfvB;\r
-+ vec_u32_t multEvenvB, multOddvB;\r
-+ vec_u16_t mfvB;\r
- \r
- vec_s16_t temp1v, temp2v;\r
-@@ -172,5 +169,5 @@\r
- i_qbitsv = vec_splat(qbits_u.v, 0);\r
- \r
-- vect_sint_u f_u;\r
-+ vect_int_u f_u;\r
- f_u.s[0]=f;\r
- fV = vec_splat(f_u.v, 0);\r
-Index: /trunk/common/ppc/dct.c\r
-===================================================================\r
---- /trunk/common/ppc/dct.c (revision 604)\r
-+++ /trunk/common/ppc/dct.c (revision 621)\r
-@@ -61,6 +61,6 @@\r
- VEC_DCT( dct0v, dct1v, dct2v, dct3v, tmp0v, tmp1v, tmp2v, tmp3v );\r
- \r
-- vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, dct);\r
-- vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, dct);\r
-+ vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, (int16_t*)dct);\r
-+ vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, (int16_t*)dct);\r
- }\r
- \r
-@@ -95,12 +95,12 @@\r
- VEC_DCT( dct4v, dct5v, dct6v, dct7v, tmp4v, tmp5v, tmp6v, tmp7v );\r
- \r
-- vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, dct);\r
-- vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, dct);\r
-- vec_st(vec_perm(tmp4v, tmp5v, permHighv), 32, dct);\r
-- vec_st(vec_perm(tmp6v, tmp7v, permHighv), 48, dct);\r
-- vec_st(vec_perm(tmp0v, tmp1v, permLowv), 64, dct);\r
-- vec_st(vec_perm(tmp2v, tmp3v, permLowv), 80, dct);\r
-- vec_st(vec_perm(tmp4v, tmp5v, permLowv), 96, dct);\r
-- vec_st(vec_perm(tmp6v, tmp7v, permLowv), 112, dct);\r
-+ vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, (int16_t*)dct);\r
-+ vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, (int16_t*)dct);\r
-+ vec_st(vec_perm(tmp4v, tmp5v, permHighv), 32, (int16_t*)dct);\r
-+ vec_st(vec_perm(tmp6v, tmp7v, permHighv), 48, (int16_t*)dct);\r
-+ vec_st(vec_perm(tmp0v, tmp1v, permLowv), 64, (int16_t*)dct);\r
-+ vec_st(vec_perm(tmp2v, tmp3v, permLowv), 80, (int16_t*)dct);\r
-+ vec_st(vec_perm(tmp4v, tmp5v, permLowv), 96, (int16_t*)dct);\r
-+ vec_st(vec_perm(tmp6v, tmp7v, permLowv), 112, (int16_t*)dct);\r
- }\r
- \r
-@@ -312,6 +312,6 @@\r
- void x264_add8x8_idct8_altivec( uint8_t *dst, int16_t dct[8][8] )\r
- {\r
-- vec_s16_t onev = vec_splat_s16(1);\r
-- vec_s16_t twov = vec_splat_s16(2);\r
-+ vec_u16_t onev = vec_splat_s16(1);\r
-+ vec_u16_t twov = vec_splat_s16(2);\r
- \r
- dct[0][0] += 32; // rounding for the >>6 at the end\r
-@@ -342,5 +342,5 @@\r
- vec_u8_t perm_ldv = vec_lvsl(0, dst);\r
- vec_u8_t perm_stv = vec_lvsr(8, dst);\r
-- vec_s16_t sixv = vec_splat_s16(6);\r
-+ vec_u16_t sixv = vec_splat_s16(6);\r
- const vec_u8_t sel = (vec_u8_t) CV(0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1);\r
- LOAD_ZERO;\r
-Index: /trunk/common/ppc/quant.h\r
-===================================================================\r
---- /trunk/common/ppc/quant.h (revision 601)\r
-+++ /trunk/common/ppc/quant.h (revision 621)\r
-@@ -19,4 +19,8 @@\r
- *****************************************************************************/\r
- \r
-+#ifdef SYS_LINUX\r
-+#include <altivec.h>\r
-+#endif\r
-+\r
- #ifndef _PPC_QUANT_H\r
- #define _PPC_QUANT_H 1\r
-@@ -28,8 +32,7 @@\r
- \r
- typedef union {\r
-- signed int s[4];\r
-- vector signed int v;\r
--} vect_sint_u;\r
--\r
-+ unsigned short s[8];\r
-+ vector unsigned short v;\r
-+} vect_ushort_u;\r
- \r
- void x264_quant_4x4_altivec( int16_t dct[4][4], int quant_mf[4][4], int const i_qbits, int const f );\r
+Index: /common/ppc/quant.c
+===================================================================
+--- /common/ppc/quant.c (revision 601)
++++ /common/ppc/quant.c (revision 621)
+@@ -18,8 +18,4 @@
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
+ *****************************************************************************/
+-
+-#ifdef HAVE_ALTIVEC_H
+-#include <altivec.h>
+-#endif
+
+ #include "common/common.h"
+@@ -54,29 +50,29 @@
+ temp2v = vec_xor(temp2v, mskB); \
+ temp1v = vec_adds(temp1v, vec_and(mskA, one)); \
+-vec_st(temp1v, (dct0), dct); \
++vec_st(temp1v, (dct0), (int16_t*)dct); \
+ temp2v = vec_adds(temp2v, vec_and(mskB, one)); \
+-vec_st(temp2v, (dct1), dct);
++vec_st(temp2v, (dct1), (int16_t*)dct);
+
+ void x264_quant_4x4_altivec( int16_t dct[4][4], int quant_mf[4][4], int const i_qbits, int const f ) {
+ vector bool short mskA;
+- vec_s32_t i_qbitsv;
++ vec_u32_t i_qbitsv;
+ vec_u16_t coefvA;
+ vec_u32_t multEvenvA, multOddvA;
+- vec_u32_t mfvA;
++ vec_u16_t mfvA;
+ vec_s16_t zerov, one;
+- vec_s32_t fV;
++ vec_u32_t fV;
+
+ vector bool short mskB;
+ vec_u16_t coefvB;
+ vec_u32_t multEvenvB, multOddvB;
+- vec_u32_t mfvB;
++ vec_u16_t mfvB;
+
+ vec_s16_t temp1v, temp2v;
+
+- vect_sint_u qbits_u;
++ vect_int_u qbits_u;
+ qbits_u.s[0]=i_qbits;
+ i_qbitsv = vec_splat(qbits_u.v, 0);
+
+- vect_sint_u f_u;
++ vect_int_u f_u;
+ f_u.s[0]=f;
+
+@@ -114,16 +110,16 @@
+ temp2v = vec_xor(temp2v, mskB); \
+ temp1v = vec_add(temp1v, vec_and(mskA, one)); \
+-vec_st(temp1v, (dct0), dct); \
++vec_st(temp1v, (dct0), (int16_t*)dct); \
+ temp2v = vec_add(temp2v, vec_and(mskB, one)); \
+-vec_st(temp2v, (dct1), dct);
++vec_st(temp2v, (dct1), (int16_t*)dct);
+
+
+ void x264_quant_4x4_dc_altivec( int16_t dct[4][4], int i_quant_mf, int const i_qbits, int const f ) {
+ vector bool short mskA;
+- vec_s32_t i_qbitsv;
++ vec_u32_t i_qbitsv;
+ vec_u16_t coefvA;
+ vec_u32_t multEvenvA, multOddvA;
+ vec_s16_t zerov, one;
+- vec_s32_t fV;
++ vec_u32_t fV;
+
+ vector bool short mskB;
+@@ -133,15 +129,14 @@
+ vec_s16_t temp1v, temp2v;
+
+- vec_u32_t mfv;
+- vect_int_u mf_u;
++ vec_u16_t mfv;
++ vect_ushort_u mf_u;
+ mf_u.s[0]=i_quant_mf;
+ mfv = vec_splat( mf_u.v, 0 );
+- mfv = vec_packs( mfv, mfv);
+
+- vect_sint_u qbits_u;
++ vect_int_u qbits_u;
+ qbits_u.s[0]=i_qbits;
+ i_qbitsv = vec_splat(qbits_u.v, 0);
+
+- vect_sint_u f_u;
++ vect_int_u f_u;
+ f_u.s[0]=f;
+ fV = vec_splat(f_u.v, 0);
+@@ -156,13 +151,15 @@
+ void x264_quant_8x8_altivec( int16_t dct[8][8], int quant_mf[8][8], int const i_qbits, int const f ) {
+ vector bool short mskA;
+- vec_s32_t i_qbitsv;
++ vec_u32_t i_qbitsv;
+ vec_u16_t coefvA;
+- vec_s32_t multEvenvA, multOddvA, mfvA;
++ vec_u32_t multEvenvA, multOddvA;
++ vec_u16_t mfvA;
+ vec_s16_t zerov, one;
+- vec_s32_t fV;
++ vec_u32_t fV;
+
+ vector bool short mskB;
+ vec_u16_t coefvB;
+- vec_u32_t multEvenvB, multOddvB, mfvB;
++ vec_u32_t multEvenvB, multOddvB;
++ vec_u16_t mfvB;
+
+ vec_s16_t temp1v, temp2v;
+@@ -172,5 +169,5 @@
+ i_qbitsv = vec_splat(qbits_u.v, 0);
+
+- vect_sint_u f_u;
++ vect_int_u f_u;
+ f_u.s[0]=f;
+ fV = vec_splat(f_u.v, 0);
+Index: /common/ppc/dct.c
+===================================================================
+--- /common/ppc/dct.c (revision 604)
++++ /common/ppc/dct.c (revision 621)
+@@ -61,6 +61,6 @@
+ VEC_DCT( dct0v, dct1v, dct2v, dct3v, tmp0v, tmp1v, tmp2v, tmp3v );
+
+- vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, dct);
+- vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, dct);
++ vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, (int16_t*)dct);
++ vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, (int16_t*)dct);
+ }
+
+@@ -95,12 +95,12 @@
+ VEC_DCT( dct4v, dct5v, dct6v, dct7v, tmp4v, tmp5v, tmp6v, tmp7v );
+
+- vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, dct);
+- vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, dct);
+- vec_st(vec_perm(tmp4v, tmp5v, permHighv), 32, dct);
+- vec_st(vec_perm(tmp6v, tmp7v, permHighv), 48, dct);
+- vec_st(vec_perm(tmp0v, tmp1v, permLowv), 64, dct);
+- vec_st(vec_perm(tmp2v, tmp3v, permLowv), 80, dct);
+- vec_st(vec_perm(tmp4v, tmp5v, permLowv), 96, dct);
+- vec_st(vec_perm(tmp6v, tmp7v, permLowv), 112, dct);
++ vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, (int16_t*)dct);
++ vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, (int16_t*)dct);
++ vec_st(vec_perm(tmp4v, tmp5v, permHighv), 32, (int16_t*)dct);
++ vec_st(vec_perm(tmp6v, tmp7v, permHighv), 48, (int16_t*)dct);
++ vec_st(vec_perm(tmp0v, tmp1v, permLowv), 64, (int16_t*)dct);
++ vec_st(vec_perm(tmp2v, tmp3v, permLowv), 80, (int16_t*)dct);
++ vec_st(vec_perm(tmp4v, tmp5v, permLowv), 96, (int16_t*)dct);
++ vec_st(vec_perm(tmp6v, tmp7v, permLowv), 112, (int16_t*)dct);
+ }
+
+@@ -312,6 +312,6 @@
+ void x264_add8x8_idct8_altivec( uint8_t *dst, int16_t dct[8][8] )
+ {
+- vec_s16_t onev = vec_splat_s16(1);
+- vec_s16_t twov = vec_splat_s16(2);
++ vec_u16_t onev = vec_splat_s16(1);
++ vec_u16_t twov = vec_splat_s16(2);
+
+ dct[0][0] += 32; // rounding for the >>6 at the end
+@@ -342,5 +342,5 @@
+ vec_u8_t perm_ldv = vec_lvsl(0, dst);
+ vec_u8_t perm_stv = vec_lvsr(8, dst);
+- vec_s16_t sixv = vec_splat_s16(6);
++ vec_u16_t sixv = vec_splat_s16(6);
+ const vec_u8_t sel = (vec_u8_t) CV(0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1);
+ LOAD_ZERO;
+Index: /common/ppc/quant.h
+===================================================================
+--- /common/ppc/quant.h (revision 601)
++++ /common/ppc/quant.h (revision 621)
+@@ -19,4 +19,8 @@
+ *****************************************************************************/
+
++#ifdef SYS_LINUX
++#include <altivec.h>
++#endif
++
+ #ifndef _PPC_QUANT_H
+ #define _PPC_QUANT_H 1
+@@ -28,8 +32,7 @@
+
+ typedef union {
+- signed int s[4];
+- vector signed int v;
+-} vect_sint_u;
+-
++ unsigned short s[8];
++ vector unsigned short v;
++} vect_ushort_u;
+
+ void x264_quant_4x4_altivec( int16_t dct[4][4], int quant_mf[4][4], int const i_qbits, int const f );
projects / handbrake / commitdiff