Mercurial > libavcodec.hg

comparison ppc/int_altivec.c @ 8494:1615d6b75ada libavcodec

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Cleanup _t types in libavcodec/ppc
author lu_zero
date Sat, 27 Dec 2008 11:21:28 +0000
parents a8a79f5385f6
children e9d9d946f213
comparison
equal deleted inserted replaced
8493:469f3e5bcf13 8494:1615d6b75ada
77 } 77 }
78 78
79 static void add_int16_altivec(int16_t * v1, int16_t * v2, int order) 79 static void add_int16_altivec(int16_t * v1, int16_t * v2, int order)
80 { 80 {
81 int i; 81 int i;
82 register vec_s16_t vec, *pv; 82 register vec_s16 vec, *pv;
83 83
84 for(i = 0; i < order; i += 8){ 84 for(i = 0; i < order; i += 8){
85 pv = (vec_s16_t*)v2; 85 pv = (vec_s16*)v2;
86 vec = vec_perm(pv[0], pv[1], vec_lvsl(0, v2)); 86 vec = vec_perm(pv[0], pv[1], vec_lvsl(0, v2));
87 vec_st(vec_add(vec_ld(0, v1), vec), 0, v1); 87 vec_st(vec_add(vec_ld(0, v1), vec), 0, v1);
88 v1 += 8; 88 v1 += 8;
89 v2 += 8; 89 v2 += 8;
90 } 90 }
91 } 91 }
92 92
93 static void sub_int16_altivec(int16_t * v1, int16_t * v2, int order) 93 static void sub_int16_altivec(int16_t * v1, int16_t * v2, int order)
94 { 94 {
95 int i; 95 int i;
96 register vec_s16_t vec, *pv; 96 register vec_s16 vec, *pv;
97 97
98 for(i = 0; i < order; i += 8){ 98 for(i = 0; i < order; i += 8){
99 pv = (vec_s16_t*)v2; 99 pv = (vec_s16*)v2;
100 vec = vec_perm(pv[0], pv[1], vec_lvsl(0, v2)); 100 vec = vec_perm(pv[0], pv[1], vec_lvsl(0, v2));
101 vec_st(vec_sub(vec_ld(0, v1), vec), 0, v1); 101 vec_st(vec_sub(vec_ld(0, v1), vec), 0, v1);
102 v1 += 8; 102 v1 += 8;
103 v2 += 8; 103 v2 += 8;
104 } 104 }
106 106
107 static int32_t scalarproduct_int16_altivec(int16_t * v1, int16_t * v2, int order, const int shift) 107 static int32_t scalarproduct_int16_altivec(int16_t * v1, int16_t * v2, int order, const int shift)
108 { 108 {
109 int i; 109 int i;
110 LOAD_ZERO; 110 LOAD_ZERO;
111 register vec_s16_t vec1, *pv; 111 register vec_s16 vec1, *pv;
112 register vec_s32_t res = vec_splat_s32(0), t; 112 register vec_s32 res = vec_splat_s32(0), t;
113 register vec_u32_t shifts; 113 register vec_u32 shifts;
114 DECLARE_ALIGNED_16(int32_t, ires); 114 DECLARE_ALIGNED_16(int32_t, ires);
115 115
116 shifts = zero_u32v; 116 shifts = zero_u32v;
117 if(shift & 0x10) shifts = vec_add(shifts, vec_sl(vec_splat_u32(0x08), vec_splat_u32(0x1))); 117 if(shift & 0x10) shifts = vec_add(shifts, vec_sl(vec_splat_u32(0x08), vec_splat_u32(0x1)));
118 if(shift & 0x08) shifts = vec_add(shifts, vec_splat_u32(0x08)); 118 if(shift & 0x08) shifts = vec_add(shifts, vec_splat_u32(0x08));
119 if(shift & 0x04) shifts = vec_add(shifts, vec_splat_u32(0x04)); 119 if(shift & 0x04) shifts = vec_add(shifts, vec_splat_u32(0x04));
120 if(shift & 0x02) shifts = vec_add(shifts, vec_splat_u32(0x02)); 120 if(shift & 0x02) shifts = vec_add(shifts, vec_splat_u32(0x02));
121 if(shift & 0x01) shifts = vec_add(shifts, vec_splat_u32(0x01)); 121 if(shift & 0x01) shifts = vec_add(shifts, vec_splat_u32(0x01));
122 122
123 for(i = 0; i < order; i += 8){ 123 for(i = 0; i < order; i += 8){
124 pv = (vec_s16_t*)v1; 124 pv = (vec_s16*)v1;
125 vec1 = vec_perm(pv[0], pv[1], vec_lvsl(0, v1)); 125 vec1 = vec_perm(pv[0], pv[1], vec_lvsl(0, v1));
126 t = vec_msum(vec1, vec_ld(0, v2), zero_s32v); 126 t = vec_msum(vec1, vec_ld(0, v2), zero_s32v);
127 t = vec_sr(t, shifts); 127 t = vec_sr(t, shifts);
128 res = vec_sums(t, res); 128 res = vec_sums(t, res);
129 v1 += 8; 129 v1 += 8;