1/*
2 * VC-1 and WMV3 decoder - DSP functions AltiVec-optimized
3 * Copyright (c) 2006 Konstantin Shishkov
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22#include "config.h"
23#include "libavutil/attributes.h"
24#include "libavutil/cpu.h"
25#include "libavutil/ppc/cpu.h"
26#include "libavutil/ppc/types_altivec.h"
27#include "libavutil/ppc/util_altivec.h"
28#include "libavcodec/vc1dsp.h"
29
30#if HAVE_ALTIVEC
31
32// main steps of 8x8 transform
33#define STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_rnd) \
34do { \
35    t0 = vec_sl(vec_add(s0, s4), vec_2); \
36    t0 = vec_add(vec_sl(t0, vec_1), t0); \
37    t0 = vec_add(t0, vec_rnd); \
38    t1 = vec_sl(vec_sub(s0, s4), vec_2); \
39    t1 = vec_add(vec_sl(t1, vec_1), t1); \
40    t1 = vec_add(t1, vec_rnd); \
41    t2 = vec_add(vec_sl(s6, vec_2), vec_sl(s6, vec_1)); \
42    t2 = vec_add(t2, vec_sl(s2, vec_4)); \
43    t3 = vec_add(vec_sl(s2, vec_2), vec_sl(s2, vec_1)); \
44    t3 = vec_sub(t3, vec_sl(s6, vec_4)); \
45    t4 = vec_add(t0, t2); \
46    t5 = vec_add(t1, t3); \
47    t6 = vec_sub(t1, t3); \
48    t7 = vec_sub(t0, t2); \
49\
50    t0 = vec_sl(vec_add(s1, s3), vec_4); \
51    t0 = vec_add(t0, vec_sl(s5, vec_3)); \
52    t0 = vec_add(t0, vec_sl(s7, vec_2)); \
53    t0 = vec_add(t0, vec_sub(s5, s3)); \
54\
55    t1 = vec_sl(vec_sub(s1, s5), vec_4); \
56    t1 = vec_sub(t1, vec_sl(s7, vec_3)); \
57    t1 = vec_sub(t1, vec_sl(s3, vec_2)); \
58    t1 = vec_sub(t1, vec_add(s1, s7)); \
59\
60    t2 = vec_sl(vec_sub(s7, s3), vec_4); \
61    t2 = vec_add(t2, vec_sl(s1, vec_3)); \
62    t2 = vec_add(t2, vec_sl(s5, vec_2)); \
63    t2 = vec_add(t2, vec_sub(s1, s7)); \
64\
65    t3 = vec_sl(vec_sub(s5, s7), vec_4); \
66    t3 = vec_sub(t3, vec_sl(s3, vec_3)); \
67    t3 = vec_add(t3, vec_sl(s1, vec_2)); \
68    t3 = vec_sub(t3, vec_add(s3, s5)); \
69\
70    s0 = vec_add(t4, t0); \
71    s1 = vec_add(t5, t1); \
72    s2 = vec_add(t6, t2); \
73    s3 = vec_add(t7, t3); \
74    s4 = vec_sub(t7, t3); \
75    s5 = vec_sub(t6, t2); \
76    s6 = vec_sub(t5, t1); \
77    s7 = vec_sub(t4, t0); \
78}while(0)
79
80#define SHIFT_HOR8(s0, s1, s2, s3, s4, s5, s6, s7) \
81do { \
82    s0 = vec_sra(s0, vec_3); \
83    s1 = vec_sra(s1, vec_3); \
84    s2 = vec_sra(s2, vec_3); \
85    s3 = vec_sra(s3, vec_3); \
86    s4 = vec_sra(s4, vec_3); \
87    s5 = vec_sra(s5, vec_3); \
88    s6 = vec_sra(s6, vec_3); \
89    s7 = vec_sra(s7, vec_3); \
90}while(0)
91
92#define SHIFT_VERT8(s0, s1, s2, s3, s4, s5, s6, s7) \
93do { \
94    s0 = vec_sra(s0, vec_7); \
95    s1 = vec_sra(s1, vec_7); \
96    s2 = vec_sra(s2, vec_7); \
97    s3 = vec_sra(s3, vec_7); \
98    s4 = vec_sra(vec_add(s4, vec_1s), vec_7); \
99    s5 = vec_sra(vec_add(s5, vec_1s), vec_7); \
100    s6 = vec_sra(vec_add(s6, vec_1s), vec_7); \
101    s7 = vec_sra(vec_add(s7, vec_1s), vec_7); \
102}while(0)
103
104/* main steps of 4x4 transform */
105#define STEP4(s0, s1, s2, s3, vec_rnd) \
106do { \
107    t1 = vec_add(vec_sl(s0, vec_4), s0); \
108    t1 = vec_add(t1, vec_rnd); \
109    t2 = vec_add(vec_sl(s2, vec_4), s2); \
110    t0 = vec_add(t1, t2); \
111    t1 = vec_sub(t1, t2); \
112    t3 = vec_sl(vec_sub(s3, s1), vec_1); \
113    t3 = vec_add(t3, vec_sl(t3, vec_2)); \
114    t2 = vec_add(t3, vec_sl(s1, vec_5)); \
115    t3 = vec_add(t3, vec_sl(s3, vec_3)); \
116    t3 = vec_add(t3, vec_sl(s3, vec_2)); \
117    s0 = vec_add(t0, t2); \
118    s1 = vec_sub(t1, t3); \
119    s2 = vec_add(t1, t3); \
120    s3 = vec_sub(t0, t2); \
121}while (0)
122
123#define SHIFT_HOR4(s0, s1, s2, s3) \
124    s0 = vec_sra(s0, vec_3); \
125    s1 = vec_sra(s1, vec_3); \
126    s2 = vec_sra(s2, vec_3); \
127    s3 = vec_sra(s3, vec_3);
128
129#define SHIFT_VERT4(s0, s1, s2, s3) \
130    s0 = vec_sra(s0, vec_7); \
131    s1 = vec_sra(s1, vec_7); \
132    s2 = vec_sra(s2, vec_7); \
133    s3 = vec_sra(s3, vec_7);
134
135/** Do inverse transform on 8x8 block
136*/
137static void vc1_inv_trans_8x8_altivec(int16_t block[64])
138{
139    vector signed short src0, src1, src2, src3, src4, src5, src6, src7;
140    vector signed int s0, s1, s2, s3, s4, s5, s6, s7;
141    vector signed int s8, s9, sA, sB, sC, sD, sE, sF;
142    vector signed int t0, t1, t2, t3, t4, t5, t6, t7;
143    const vector signed int vec_64 = vec_sl(vec_splat_s32(4), vec_splat_u32(4));
144    const vector unsigned int vec_7 = vec_splat_u32(7);
145    const vector unsigned int vec_4 = vec_splat_u32(4);
146    const vector  signed int vec_4s = vec_splat_s32(4);
147    const vector unsigned int vec_3 = vec_splat_u32(3);
148    const vector unsigned int vec_2 = vec_splat_u32(2);
149    const vector  signed int vec_1s = vec_splat_s32(1);
150    const vector unsigned int vec_1 = vec_splat_u32(1);
151
152    src0 = vec_ld(  0, block);
153    src1 = vec_ld( 16, block);
154    src2 = vec_ld( 32, block);
155    src3 = vec_ld( 48, block);
156    src4 = vec_ld( 64, block);
157    src5 = vec_ld( 80, block);
158    src6 = vec_ld( 96, block);
159    src7 = vec_ld(112, block);
160
161    s0 = vec_unpackl(src0);
162    s1 = vec_unpackl(src1);
163    s2 = vec_unpackl(src2);
164    s3 = vec_unpackl(src3);
165    s4 = vec_unpackl(src4);
166    s5 = vec_unpackl(src5);
167    s6 = vec_unpackl(src6);
168    s7 = vec_unpackl(src7);
169    s8 = vec_unpackh(src0);
170    s9 = vec_unpackh(src1);
171    sA = vec_unpackh(src2);
172    sB = vec_unpackh(src3);
173    sC = vec_unpackh(src4);
174    sD = vec_unpackh(src5);
175    sE = vec_unpackh(src6);
176    sF = vec_unpackh(src7);
177    STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_4s);
178    SHIFT_HOR8(s0, s1, s2, s3, s4, s5, s6, s7);
179    STEP8(s8, s9, sA, sB, sC, sD, sE, sF, vec_4s);
180    SHIFT_HOR8(s8, s9, sA, sB, sC, sD, sE, sF);
181    src0 = vec_pack(s8, s0);
182    src1 = vec_pack(s9, s1);
183    src2 = vec_pack(sA, s2);
184    src3 = vec_pack(sB, s3);
185    src4 = vec_pack(sC, s4);
186    src5 = vec_pack(sD, s5);
187    src6 = vec_pack(sE, s6);
188    src7 = vec_pack(sF, s7);
189    TRANSPOSE8(src0, src1, src2, src3, src4, src5, src6, src7);
190
191    s0 = vec_unpackl(src0);
192    s1 = vec_unpackl(src1);
193    s2 = vec_unpackl(src2);
194    s3 = vec_unpackl(src3);
195    s4 = vec_unpackl(src4);
196    s5 = vec_unpackl(src5);
197    s6 = vec_unpackl(src6);
198    s7 = vec_unpackl(src7);
199    s8 = vec_unpackh(src0);
200    s9 = vec_unpackh(src1);
201    sA = vec_unpackh(src2);
202    sB = vec_unpackh(src3);
203    sC = vec_unpackh(src4);
204    sD = vec_unpackh(src5);
205    sE = vec_unpackh(src6);
206    sF = vec_unpackh(src7);
207    STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_64);
208    SHIFT_VERT8(s0, s1, s2, s3, s4, s5, s6, s7);
209    STEP8(s8, s9, sA, sB, sC, sD, sE, sF, vec_64);
210    SHIFT_VERT8(s8, s9, sA, sB, sC, sD, sE, sF);
211    src0 = vec_pack(s8, s0);
212    src1 = vec_pack(s9, s1);
213    src2 = vec_pack(sA, s2);
214    src3 = vec_pack(sB, s3);
215    src4 = vec_pack(sC, s4);
216    src5 = vec_pack(sD, s5);
217    src6 = vec_pack(sE, s6);
218    src7 = vec_pack(sF, s7);
219
220    vec_st(src0,  0, block);
221    vec_st(src1, 16, block);
222    vec_st(src2, 32, block);
223    vec_st(src3, 48, block);
224    vec_st(src4, 64, block);
225    vec_st(src5, 80, block);
226    vec_st(src6, 96, block);
227    vec_st(src7,112, block);
228}
229
230/** Do inverse transform on 8x4 part of block
231*/
232static void vc1_inv_trans_8x4_altivec(uint8_t *dest, int stride, int16_t *block)
233{
234    vector signed short src0, src1, src2, src3, src4, src5, src6, src7;
235    vector signed int s0, s1, s2, s3, s4, s5, s6, s7;
236    vector signed int s8, s9, sA, sB, sC, sD, sE, sF;
237    vector signed int t0, t1, t2, t3, t4, t5, t6, t7;
238    const vector signed int vec_64 = vec_sl(vec_splat_s32(4), vec_splat_u32(4));
239    const vector unsigned int vec_7 = vec_splat_u32(7);
240    const vector unsigned int vec_5 = vec_splat_u32(5);
241    const vector unsigned int vec_4 = vec_splat_u32(4);
242    const vector  signed int vec_4s = vec_splat_s32(4);
243    const vector unsigned int vec_3 = vec_splat_u32(3);
244    const vector unsigned int vec_2 = vec_splat_u32(2);
245    const vector unsigned int vec_1 = vec_splat_u32(1);
246    vector unsigned char tmp;
247    vector signed short tmp2, tmp3;
248    vector unsigned char perm0, perm1, p0, p1, p;
249
250    src0 = vec_ld(  0, block);
251    src1 = vec_ld( 16, block);
252    src2 = vec_ld( 32, block);
253    src3 = vec_ld( 48, block);
254    src4 = vec_ld( 64, block);
255    src5 = vec_ld( 80, block);
256    src6 = vec_ld( 96, block);
257    src7 = vec_ld(112, block);
258
259    TRANSPOSE8(src0, src1, src2, src3, src4, src5, src6, src7);
260    s0 = vec_unpackl(src0);
261    s1 = vec_unpackl(src1);
262    s2 = vec_unpackl(src2);
263    s3 = vec_unpackl(src3);
264    s4 = vec_unpackl(src4);
265    s5 = vec_unpackl(src5);
266    s6 = vec_unpackl(src6);
267    s7 = vec_unpackl(src7);
268    s8 = vec_unpackh(src0);
269    s9 = vec_unpackh(src1);
270    sA = vec_unpackh(src2);
271    sB = vec_unpackh(src3);
272    sC = vec_unpackh(src4);
273    sD = vec_unpackh(src5);
274    sE = vec_unpackh(src6);
275    sF = vec_unpackh(src7);
276    STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_4s);
277    SHIFT_HOR8(s0, s1, s2, s3, s4, s5, s6, s7);
278    STEP8(s8, s9, sA, sB, sC, sD, sE, sF, vec_4s);
279    SHIFT_HOR8(s8, s9, sA, sB, sC, sD, sE, sF);
280    src0 = vec_pack(s8, s0);
281    src1 = vec_pack(s9, s1);
282    src2 = vec_pack(sA, s2);
283    src3 = vec_pack(sB, s3);
284    src4 = vec_pack(sC, s4);
285    src5 = vec_pack(sD, s5);
286    src6 = vec_pack(sE, s6);
287    src7 = vec_pack(sF, s7);
288    TRANSPOSE8(src0, src1, src2, src3, src4, src5, src6, src7);
289
290    s0 = vec_unpackh(src0);
291    s1 = vec_unpackh(src1);
292    s2 = vec_unpackh(src2);
293    s3 = vec_unpackh(src3);
294    s8 = vec_unpackl(src0);
295    s9 = vec_unpackl(src1);
296    sA = vec_unpackl(src2);
297    sB = vec_unpackl(src3);
298    STEP4(s0, s1, s2, s3, vec_64);
299    SHIFT_VERT4(s0, s1, s2, s3);
300    STEP4(s8, s9, sA, sB, vec_64);
301    SHIFT_VERT4(s8, s9, sA, sB);
302    src0 = vec_pack(s0, s8);
303    src1 = vec_pack(s1, s9);
304    src2 = vec_pack(s2, sA);
305    src3 = vec_pack(s3, sB);
306
307    p0 = vec_lvsl (0, dest);
308    p1 = vec_lvsl (stride, dest);
309    p = vec_splat_u8 (-1);
310    perm0 = vec_mergeh (p, p0);
311    perm1 = vec_mergeh (p, p1);
312
313#define ADD(dest,src,perm)                                              \
314    /* *(uint64_t *)&tmp = *(uint64_t *)dest; */                        \
315    tmp = vec_ld (0, dest);                                             \
316    tmp2 = (vector signed short)vec_perm (tmp, vec_splat_u8(0), perm);  \
317    tmp3 = vec_adds (tmp2, src);                                        \
318    tmp = vec_packsu (tmp3, tmp3);                                      \
319    vec_ste ((vector unsigned int)tmp, 0, (unsigned int *)dest);        \
320    vec_ste ((vector unsigned int)tmp, 4, (unsigned int *)dest);
321
322    ADD (dest, src0, perm0)      dest += stride;
323    ADD (dest, src1, perm1)      dest += stride;
324    ADD (dest, src2, perm0)      dest += stride;
325    ADD (dest, src3, perm1)
326}
327
328#define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
329#define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
330
331#define OP_U8_ALTIVEC                          PUT_OP_U8_ALTIVEC
332#define PREFIX_no_rnd_vc1_chroma_mc8_altivec   put_no_rnd_vc1_chroma_mc8_altivec
333#include "h264chroma_template.c"
334#undef OP_U8_ALTIVEC
335#undef PREFIX_no_rnd_vc1_chroma_mc8_altivec
336
337#define OP_U8_ALTIVEC                          AVG_OP_U8_ALTIVEC
338#define PREFIX_no_rnd_vc1_chroma_mc8_altivec   avg_no_rnd_vc1_chroma_mc8_altivec
339#include "h264chroma_template.c"
340#undef OP_U8_ALTIVEC
341#undef PREFIX_no_rnd_vc1_chroma_mc8_altivec
342
343#endif /* HAVE_ALTIVEC */
344
345av_cold void ff_vc1dsp_init_ppc(VC1DSPContext *dsp)
346{
347#if HAVE_ALTIVEC
348    if (!PPC_ALTIVEC(av_get_cpu_flags()))
349        return;
350
351    dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_altivec;
352    dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_altivec;
353    dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = put_no_rnd_vc1_chroma_mc8_altivec;
354    dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = avg_no_rnd_vc1_chroma_mc8_altivec;
355#endif /* HAVE_ALTIVEC */
356}
357