| blob | 3f1752d4cebba7712c6cf4f48481c09a81b3aa7a |
1 Index: /common/ppc/quant.c
2 ===================================================================
3 --- /common/ppc/quant.c (revision 601)
4 +++ /common/ppc/quant.c (revision 621)
5 @@ -18,8 +18,4 @@
6 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
7 *****************************************************************************/
8 -
9 -#ifdef HAVE_ALTIVEC_H
10 -#include <altivec.h>
11 -#endif
13 #include "common/common.h"
14 @@ -54,29 +50,29 @@
15 temp2v = vec_xor(temp2v, mskB); \
16 temp1v = vec_adds(temp1v, vec_and(mskA, one)); \
17 -vec_st(temp1v, (dct0), dct); \
18 +vec_st(temp1v, (dct0), (int16_t*)dct); \
19 temp2v = vec_adds(temp2v, vec_and(mskB, one)); \
20 -vec_st(temp2v, (dct1), dct);
21 +vec_st(temp2v, (dct1), (int16_t*)dct);
23 void x264_quant_4x4_altivec( int16_t dct[4][4], int quant_mf[4][4], int const i_qbits, int const f ) {
24 vector bool short mskA;
25 - vec_s32_t i_qbitsv;
26 + vec_u32_t i_qbitsv;
27 vec_u16_t coefvA;
28 vec_u32_t multEvenvA, multOddvA;
29 - vec_u32_t mfvA;
30 + vec_u16_t mfvA;
31 vec_s16_t zerov, one;
32 - vec_s32_t fV;
33 + vec_u32_t fV;
35 vector bool short mskB;
36 vec_u16_t coefvB;
37 vec_u32_t multEvenvB, multOddvB;
38 - vec_u32_t mfvB;
39 + vec_u16_t mfvB;
41 vec_s16_t temp1v, temp2v;
43 - vect_sint_u qbits_u;
44 + vect_int_u qbits_u;
45 qbits_u.s[0]=i_qbits;
46 i_qbitsv = vec_splat(qbits_u.v, 0);
48 - vect_sint_u f_u;
49 + vect_int_u f_u;
50 f_u.s[0]=f;
52 @@ -114,16 +110,16 @@
53 temp2v = vec_xor(temp2v, mskB); \
54 temp1v = vec_add(temp1v, vec_and(mskA, one)); \
55 -vec_st(temp1v, (dct0), dct); \
56 +vec_st(temp1v, (dct0), (int16_t*)dct); \
57 temp2v = vec_add(temp2v, vec_and(mskB, one)); \
58 -vec_st(temp2v, (dct1), dct);
59 +vec_st(temp2v, (dct1), (int16_t*)dct);
62 void x264_quant_4x4_dc_altivec( int16_t dct[4][4], int i_quant_mf, int const i_qbits, int const f ) {
63 vector bool short mskA;
64 - vec_s32_t i_qbitsv;
65 + vec_u32_t i_qbitsv;
66 vec_u16_t coefvA;
67 vec_u32_t multEvenvA, multOddvA;
68 vec_s16_t zerov, one;
69 - vec_s32_t fV;
70 + vec_u32_t fV;
72 vector bool short mskB;
73 @@ -133,15 +129,14 @@
74 vec_s16_t temp1v, temp2v;
76 - vec_u32_t mfv;
77 - vect_int_u mf_u;
78 + vec_u16_t mfv;
79 + vect_ushort_u mf_u;
80 mf_u.s[0]=i_quant_mf;
81 mfv = vec_splat( mf_u.v, 0 );
82 - mfv = vec_packs( mfv, mfv);
84 - vect_sint_u qbits_u;
85 + vect_int_u qbits_u;
86 qbits_u.s[0]=i_qbits;
87 i_qbitsv = vec_splat(qbits_u.v, 0);
89 - vect_sint_u f_u;
90 + vect_int_u f_u;
91 f_u.s[0]=f;
92 fV = vec_splat(f_u.v, 0);
93 @@ -156,13 +151,15 @@
94 void x264_quant_8x8_altivec( int16_t dct[8][8], int quant_mf[8][8], int const i_qbits, int const f ) {
95 vector bool short mskA;
96 - vec_s32_t i_qbitsv;
97 + vec_u32_t i_qbitsv;
98 vec_u16_t coefvA;
99 - vec_s32_t multEvenvA, multOddvA, mfvA;
100 + vec_u32_t multEvenvA, multOddvA;
101 + vec_u16_t mfvA;
102 vec_s16_t zerov, one;
103 - vec_s32_t fV;
104 + vec_u32_t fV;
106 vector bool short mskB;
107 vec_u16_t coefvB;
108 - vec_u32_t multEvenvB, multOddvB, mfvB;
109 + vec_u32_t multEvenvB, multOddvB;
110 + vec_u16_t mfvB;
112 vec_s16_t temp1v, temp2v;
113 @@ -172,5 +169,5 @@
114 i_qbitsv = vec_splat(qbits_u.v, 0);
116 - vect_sint_u f_u;
117 + vect_int_u f_u;
118 f_u.s[0]=f;
119 fV = vec_splat(f_u.v, 0);
120 Index: /common/ppc/dct.c
121 ===================================================================
122 --- /common/ppc/dct.c (revision 604)
123 +++ /common/ppc/dct.c (revision 621)
124 @@ -61,6 +61,6 @@
125 VEC_DCT( dct0v, dct1v, dct2v, dct3v, tmp0v, tmp1v, tmp2v, tmp3v );
127 - vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, dct);
128 - vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, dct);
129 + vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, (int16_t*)dct);
130 + vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, (int16_t*)dct);
131 }
133 @@ -95,12 +95,12 @@
134 VEC_DCT( dct4v, dct5v, dct6v, dct7v, tmp4v, tmp5v, tmp6v, tmp7v );
136 - vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, dct);
137 - vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, dct);
138 - vec_st(vec_perm(tmp4v, tmp5v, permHighv), 32, dct);
139 - vec_st(vec_perm(tmp6v, tmp7v, permHighv), 48, dct);
140 - vec_st(vec_perm(tmp0v, tmp1v, permLowv), 64, dct);
141 - vec_st(vec_perm(tmp2v, tmp3v, permLowv), 80, dct);
142 - vec_st(vec_perm(tmp4v, tmp5v, permLowv), 96, dct);
143 - vec_st(vec_perm(tmp6v, tmp7v, permLowv), 112, dct);
144 + vec_st(vec_perm(tmp0v, tmp1v, permHighv), 0, (int16_t*)dct);
145 + vec_st(vec_perm(tmp2v, tmp3v, permHighv), 16, (int16_t*)dct);
146 + vec_st(vec_perm(tmp4v, tmp5v, permHighv), 32, (int16_t*)dct);
147 + vec_st(vec_perm(tmp6v, tmp7v, permHighv), 48, (int16_t*)dct);
148 + vec_st(vec_perm(tmp0v, tmp1v, permLowv), 64, (int16_t*)dct);
149 + vec_st(vec_perm(tmp2v, tmp3v, permLowv), 80, (int16_t*)dct);
150 + vec_st(vec_perm(tmp4v, tmp5v, permLowv), 96, (int16_t*)dct);
151 + vec_st(vec_perm(tmp6v, tmp7v, permLowv), 112, (int16_t*)dct);
152 }
154 @@ -312,6 +312,6 @@
155 void x264_add8x8_idct8_altivec( uint8_t *dst, int16_t dct[8][8] )
156 {
157 - vec_s16_t onev = vec_splat_s16(1);
158 - vec_s16_t twov = vec_splat_s16(2);
159 + vec_u16_t onev = vec_splat_s16(1);
160 + vec_u16_t twov = vec_splat_s16(2);
162 dct[0][0] += 32; // rounding for the >>6 at the end
163 @@ -342,5 +342,5 @@
164 vec_u8_t perm_ldv = vec_lvsl(0, dst);
165 vec_u8_t perm_stv = vec_lvsr(8, dst);
166 - vec_s16_t sixv = vec_splat_s16(6);
167 + vec_u16_t sixv = vec_splat_s16(6);
168 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);
169 LOAD_ZERO;
170 Index: /common/ppc/quant.h
171 ===================================================================
172 --- /common/ppc/quant.h (revision 601)
173 +++ /common/ppc/quant.h (revision 621)
174 @@ -19,4 +19,8 @@
175 *****************************************************************************/
177 +#ifdef SYS_LINUX
178 +#include <altivec.h>
179 +#endif
180 +
181 #ifndef _PPC_QUANT_H
182 #define _PPC_QUANT_H 1
183 @@ -28,8 +32,7 @@
185 typedef union {
186 - signed int s[4];
187 - vector signed int v;
188 -} vect_sint_u;
189 -
190 + unsigned short s[8];
191 + vector unsigned short v;
192 +} vect_ushort_u;
194 void x264_quant_4x4_altivec( int16_t dct[4][4], int quant_mf[4][4], int const i_qbits, int const f );