2 * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "libavutil/mem.h"
22 #include "libavutil/ppc/types_altivec.h"
23 #include "libavutil/ppc/util_altivec.h"
25 /* this code assume that stride % 16 == 0 */
27 #define CHROMA_MC8_ALTIVEC_CORE(BIAS1, BIAS2) \
28 vsrc2ssH = (vec_s16)VEC_MERGEH(zero_u8v,(vec_u8)vsrc2uc);\
29 vsrc3ssH = (vec_s16)VEC_MERGEH(zero_u8v,(vec_u8)vsrc3uc);\
31 psum = vec_mladd(vA, vsrc0ssH, BIAS1);\
32 psum = vec_mladd(vB, vsrc1ssH, psum);\
33 psum = vec_mladd(vC, vsrc2ssH, psum);\
34 psum = vec_mladd(vD, vsrc3ssH, psum);\
36 psum = vec_sr(psum, v6us);\
38 vdst = vec_ld(0, dst);\
39 ppsum = (vec_u8)vec_pack(psum, psum);\
40 vfdst = vec_perm(vdst, ppsum, fperm);\
42 OP_U8_ALTIVEC(fsum, vfdst, vdst);\
44 vec_st(fsum, 0, dst);\
52 #define CHROMA_MC8_ALTIVEC_CORE_SIMPLE \
54 vsrc0ssH = (vec_s16)VEC_MERGEH(zero_u8v,(vec_u8)vsrc0uc);\
55 vsrc1ssH = (vec_s16)VEC_MERGEH(zero_u8v,(vec_u8)vsrc1uc);\
57 psum = vec_mladd(vA, vsrc0ssH, v32ss);\
58 psum = vec_mladd(vE, vsrc1ssH, psum);\
59 psum = vec_sr(psum, v6us);\
61 vdst = vec_ld(0, dst);\
62 ppsum = (vec_u8)vec_pack(psum, psum);\
63 vfdst = vec_perm(vdst, ppsum, fperm);\
65 OP_U8_ALTIVEC(fsum, vfdst, vdst);\
67 vec_st(fsum, 0, dst);\
73 #define add28(a) vec_add(v28ss, a)
76 #define GET_VSRC1(vs0, off, b, perm0, s){ \
77 vec_u8 vsrcCuc, vsrcDuc; \
78 vsrcCuc = vec_ld(off, s); \
80 vsrcDuc = vec_ld(off + b, s); \
84 vs0 = vec_perm(vsrcCuc, vsrcDuc, perm0); \
86 #define GET_VSRC(vs0, vs1, off, b, perm0, perm1, s){ \
87 vec_u8 vsrcCuc, vsrcDuc; \
88 vsrcCuc = vec_ld(off, s); \
90 vsrcDuc = vec_ld(off + b, s); \
94 vs0 = vec_perm(vsrcCuc, vsrcDuc, perm0); \
95 if (reallyBadAlign){ \
98 vs1 = vec_perm(vsrcCuc, vsrcDuc, perm1); \
103 #define GET_VSRC1(vs0, off, b, perm0, s){ \
104 vs0 = vec_vsx_ld(off, s); \
106 #define GET_VSRC(vs0, vs1, off, b, perm0, perm1, s){ \
107 vs0 = vec_vsx_ld(off, s); \
108 vs1 = vec_vsx_ld(off + 1, s); \
110 #endif /* HAVE_BIGENDIAN */
112 #ifdef PREFIX_h264_chroma_mc8_altivec
113 static void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src,
114 ptrdiff_t stride, int h,
117 DECLARE_ALIGNED(16, signed int, ABCD)[4] =
118 {((8 - x) * (8 - y)),
125 const vec_s32 vABCD = vec_ld(0, ABCD);
126 const vec_s16 vA = VEC_SPLAT16(vABCD, 1);
127 const vec_s16 vB = VEC_SPLAT16(vABCD, 3);
128 const vec_s16 vC = VEC_SPLAT16(vABCD, 5);
129 const vec_s16 vD = VEC_SPLAT16(vABCD, 7);
130 const vec_s16 v32ss = vec_sl(vec_splat_s16(1),vec_splat_u16(5));
131 const vec_u16 v6us = vec_splat_u16(6);
133 vec_u8 vsrcperm0, vsrcperm1;
134 vec_u8 vsrc0uc, vsrc1uc;
135 vec_s16 vsrc0ssH, vsrc1ssH;
136 vec_u8 vsrc2uc, vsrc3uc;
137 vec_s16 vsrc2ssH, vsrc3ssH, psum;
138 vec_u8 vdst, ppsum, vfdst, fsum;
140 register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1;
141 register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0;
142 vsrcperm0 = vec_lvsl(0, src);
143 vsrcperm1 = vec_lvsl(1, src);
146 if (((unsigned long)dst) % 16 == 0) {
147 fperm = (vec_u8){0x10, 0x11, 0x12, 0x13,
148 0x14, 0x15, 0x16, 0x17,
149 0x08, 0x09, 0x0A, 0x0B,
150 0x0C, 0x0D, 0x0E, 0x0F};
152 fperm = (vec_u8){0x00, 0x01, 0x02, 0x03,
153 0x04, 0x05, 0x06, 0x07,
154 0x18, 0x19, 0x1A, 0x1B,
155 0x1C, 0x1D, 0x1E, 0x1F};
158 GET_VSRC(vsrc0uc, vsrc1uc, 0, 16, vsrcperm0, vsrcperm1, src);
160 vsrc0ssH = (vec_s16)VEC_MERGEH(zero_u8v,(vec_u8)vsrc0uc);
161 vsrc1ssH = (vec_s16)VEC_MERGEH(zero_u8v,(vec_u8)vsrc1uc);
164 for (i = 0 ; i < h ; i++) {
165 GET_VSRC(vsrc2uc, vsrc3uc, stride, 16, vsrcperm0, vsrcperm1, src);
166 CHROMA_MC8_ALTIVEC_CORE(v32ss, noop);
169 const vec_s16 vE = vec_add(vB, vC);
170 if (ABCD[2]) { // x == 0 B == 0
171 for (i = 0 ; i < h ; i++) {
172 GET_VSRC1(vsrc1uc, stride, 15, vsrcperm0, src);
173 CHROMA_MC8_ALTIVEC_CORE_SIMPLE;
176 } else { // y == 0 C == 0
177 for (i = 0 ; i < h ; i++) {
178 GET_VSRC(vsrc0uc, vsrc1uc, 0, 15, vsrcperm0, vsrcperm1, src);
179 CHROMA_MC8_ALTIVEC_CORE_SIMPLE;
186 /* this code assume that stride % 16 == 0 */
187 #ifdef PREFIX_no_rnd_vc1_chroma_mc8_altivec
188 static void PREFIX_no_rnd_vc1_chroma_mc8_altivec(uint8_t *dst, uint8_t *src,
189 ptrdiff_t stride, int h,
192 DECLARE_ALIGNED(16, signed int, ABCD)[4] =
193 {((8 - x) * (8 - y)),
200 const vec_s32 vABCD = vec_ld(0, ABCD);
201 const vec_s16 vA = VEC_SPLAT16(vABCD, 1);
202 const vec_s16 vB = VEC_SPLAT16(vABCD, 3);
203 const vec_s16 vC = VEC_SPLAT16(vABCD, 5);
204 const vec_s16 vD = VEC_SPLAT16(vABCD, 7);
205 const vec_s16 v28ss = vec_sub(vec_sl(vec_splat_s16(1),vec_splat_u16(5)),vec_splat_s16(4));
206 const vec_u16 v6us = vec_splat_u16(6);
208 vec_u8 vsrcperm0, vsrcperm1;
209 vec_u8 vsrc0uc, vsrc1uc;
210 vec_s16 vsrc0ssH, vsrc1ssH;
211 vec_u8 vsrc2uc, vsrc3uc;
212 vec_s16 vsrc2ssH, vsrc3ssH, psum;
213 vec_u8 vdst, ppsum, vfdst, fsum;
215 register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1;
216 register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0;
217 vsrcperm0 = vec_lvsl(0, src);
218 vsrcperm1 = vec_lvsl(1, src);
221 if (((unsigned long)dst) % 16 == 0) {
222 fperm = (vec_u8){0x10, 0x11, 0x12, 0x13,
223 0x14, 0x15, 0x16, 0x17,
224 0x08, 0x09, 0x0A, 0x0B,
225 0x0C, 0x0D, 0x0E, 0x0F};
227 fperm = (vec_u8){0x00, 0x01, 0x02, 0x03,
228 0x04, 0x05, 0x06, 0x07,
229 0x18, 0x19, 0x1A, 0x1B,
230 0x1C, 0x1D, 0x1E, 0x1F};
233 GET_VSRC(vsrc0uc, vsrc1uc, 0, 16, vsrcperm0, vsrcperm1, src);
235 vsrc0ssH = (vec_s16)VEC_MERGEH(zero_u8v, (vec_u8)vsrc0uc);
236 vsrc1ssH = (vec_s16)VEC_MERGEH(zero_u8v, (vec_u8)vsrc1uc);
238 for (i = 0 ; i < h ; i++) {
239 GET_VSRC(vsrc2uc, vsrc3uc, stride, 16, vsrcperm0, vsrcperm1, src);
240 CHROMA_MC8_ALTIVEC_CORE(vec_splat_s16(0), add28);
247 #undef CHROMA_MC8_ALTIVEC_CORE