2 * Copyright (C) 2009 David Conrad
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
25 #include "libavutil/attributes.h"
26 #include "libavutil/cpu.h"
27 #include "libavutil/ppc/cpu.h"
28 #include "libavutil/ppc/util_altivec.h"
30 #include "libavcodec/vp3dsp.h"
34 static const vec_s16 constants =
35 {0, 64277, 60547, 54491, 46341, 36410, 25080, 12785};
37 static const vec_u8 interleave_high =
38 {0, 1, 16, 17, 4, 5, 20, 21, 8, 9, 24, 25, 12, 13, 28, 29};
40 static const vec_u8 interleave_high =
41 {2, 3, 18, 19, 6, 7, 22, 23, 10, 11, 26, 27, 14, 15, 30, 31};
45 vec_s16 A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;\
46 vec_s16 Ed, Gd, Add, Bdd, Fd, Hd;\
47 vec_s16 eight = vec_splat_s16(8);\
48 vec_u16 four = vec_splat_u16(4);\
50 vec_s16 C1 = vec_splat(constants, 1);\
51 vec_s16 C2 = vec_splat(constants, 2);\
52 vec_s16 C3 = vec_splat(constants, 3);\
53 vec_s16 C4 = vec_splat(constants, 4);\
54 vec_s16 C5 = vec_splat(constants, 5);\
55 vec_s16 C6 = vec_splat(constants, 6);\
56 vec_s16 C7 = vec_splat(constants, 7);\
58 vec_s16 b0 = vec_ld(0x00, block);\
59 vec_s16 b1 = vec_ld(0x10, block);\
60 vec_s16 b2 = vec_ld(0x20, block);\
61 vec_s16 b3 = vec_ld(0x30, block);\
62 vec_s16 b4 = vec_ld(0x40, block);\
63 vec_s16 b5 = vec_ld(0x50, block);\
64 vec_s16 b6 = vec_ld(0x60, block);\
65 vec_s16 b7 = vec_ld(0x70, block);
67 // these functions do (a*C)>>16
68 // things are tricky because a is signed, but C unsigned.
69 // M15 is used if C fits in 15 bit unsigned (C6,C7)
70 // M16 is used if C requires 16 bits unsigned
71 static inline vec_s16 M15(vec_s16 a, vec_s16 C)
73 return (vec_s16)vec_perm(vec_mule(a,C), vec_mulo(a,C), interleave_high);
75 static inline vec_s16 M16(vec_s16 a, vec_s16 C)
77 return vec_add(a, M15(a, C));
80 #define IDCT_1D(ADD, SHIFT)\
81 A = vec_add(M16(b1, C1), M15(b7, C7));\
82 B = vec_sub(M15(b1, C7), M16(b7, C1));\
83 C = vec_add(M16(b3, C3), M16(b5, C5));\
84 D = vec_sub(M16(b5, C3), M16(b3, C5));\
86 Ad = M16(vec_sub(A, C), C4);\
87 Bd = M16(vec_sub(B, D), C4);\
92 E = ADD(M16(vec_add(b0, b4), C4));\
93 F = ADD(M16(vec_sub(b0, b4), C4));\
95 G = vec_add(M16(b2, C2), M15(b6, C6));\
96 H = vec_sub(M15(b2, C6), M16(b6, C2));\
101 Add = vec_add(F, Ad);\
102 Bdd = vec_sub(Bd, H);\
104 Fd = vec_sub(F, Ad);\
105 Hd = vec_add(Bd, H);\
107 b0 = SHIFT(vec_add(Gd, Cd));\
108 b7 = SHIFT(vec_sub(Gd, Cd));\
110 b1 = SHIFT(vec_add(Add, Hd));\
111 b2 = SHIFT(vec_sub(Add, Hd));\
113 b3 = SHIFT(vec_add(Ed, Dd));\
114 b4 = SHIFT(vec_sub(Ed, Dd));\
116 b5 = SHIFT(vec_add(Fd, Bdd));\
117 b6 = SHIFT(vec_sub(Fd, Bdd));
120 #define ADD8(a) vec_add(a, eight)
121 #define SHIFT4(a) vec_sra(a, four)
123 static void vp3_idct_put_altivec(uint8_t *dst, ptrdiff_t stride, int16_t block[64])
128 // pixels are signed; so add 128*16 in addition to the normal 8
129 vec_s16 v2048 = vec_sl(vec_splat_s16(1), vec_splat_u16(11));
130 eight = vec_add(eight, v2048);
133 TRANSPOSE8(b0, b1, b2, b3, b4, b5, b6, b7);
134 IDCT_1D(ADD8, SHIFT4)
137 t = vec_packsu(a, a);\
138 vec_ste((vec_u32)t, 0, (unsigned int *)dst);\
139 vec_ste((vec_u32)t, 4, (unsigned int *)dst);
141 PUT(b0) dst += stride;
142 PUT(b1) dst += stride;
143 PUT(b2) dst += stride;
144 PUT(b3) dst += stride;
145 PUT(b4) dst += stride;
146 PUT(b5) dst += stride;
147 PUT(b6) dst += stride;
149 memset(block, 0, sizeof(*block) * 64);
152 static void vp3_idct_add_altivec(uint8_t *dst, ptrdiff_t stride, int16_t block[64])
157 vec_u8 vdst_mask = vec_mergeh(vec_splat_u8(-1), vec_lvsl(0, dst));
162 TRANSPOSE8(b0, b1, b2, b3, b4, b5, b6, b7);
163 IDCT_1D(ADD8, SHIFT4)
167 vdst = vec_ld(0, dst);\
168 vdst_16 = (vec_s16)vec_perm(vdst, zero_u8v, vdst_mask);
171 vdst = vec_vsx_ld(0,dst);\
172 vdst_16 = (vec_s16)vec_mergeh(vdst, zero_u8v);
177 vdst_16 = vec_adds(a, vdst_16);\
178 t = vec_packsu(vdst_16, vdst_16);\
179 vec_ste((vec_u32)t, 0, (unsigned int *)dst);\
180 vec_ste((vec_u32)t, 4, (unsigned int *)dst);
182 ADD(b0) dst += stride;
183 ADD(b1) dst += stride;
184 ADD(b2) dst += stride;
185 ADD(b3) dst += stride;
186 ADD(b4) dst += stride;
187 ADD(b5) dst += stride;
188 ADD(b6) dst += stride;
190 memset(block, 0, sizeof(*block) * 64);
193 #endif /* HAVE_ALTIVEC */
195 av_cold void ff_vp3dsp_init_ppc(VP3DSPContext *c, int flags)
198 if (!PPC_ALTIVEC(av_get_cpu_flags()))
201 c->idct_put = vp3_idct_put_altivec;
202 c->idct_add = vp3_idct_add_altivec;