ppc: More consistent arch initialization
[ffmpeg.git] / libavcodec / ppc / h264dsp.c
1 /*
2  * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>
3  *
4  * This file is part of Libav.
5  *
6  * Libav 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.
10  *
11  * Libav 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.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with Libav; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20
21 #include "config.h"
22 #include "libavutil/attributes.h"
23 #include "libavutil/cpu.h"
24 #include "libavutil/intreadwrite.h"
25 #include "libavutil/ppc/types_altivec.h"
26 #include "libavutil/ppc/util_altivec.h"
27 #include "libavcodec/h264data.h"
28 #include "libavcodec/h264dsp.h"
29
30 #if HAVE_ALTIVEC
31
32 /****************************************************************************
33  * IDCT transform:
34  ****************************************************************************/
35
36 #define VEC_1D_DCT(vb0,vb1,vb2,vb3,va0,va1,va2,va3)               \
37     /* 1st stage */                                               \
38     vz0 = vec_add(vb0,vb2);       /* temp[0] = Y[0] + Y[2] */     \
39     vz1 = vec_sub(vb0,vb2);       /* temp[1] = Y[0] - Y[2] */     \
40     vz2 = vec_sra(vb1,vec_splat_u16(1));                          \
41     vz2 = vec_sub(vz2,vb3);       /* temp[2] = Y[1].1/2 - Y[3] */ \
42     vz3 = vec_sra(vb3,vec_splat_u16(1));                          \
43     vz3 = vec_add(vb1,vz3);       /* temp[3] = Y[1] + Y[3].1/2 */ \
44     /* 2nd stage: output */                                       \
45     va0 = vec_add(vz0,vz3);       /* x[0] = temp[0] + temp[3] */  \
46     va1 = vec_add(vz1,vz2);       /* x[1] = temp[1] + temp[2] */  \
47     va2 = vec_sub(vz1,vz2);       /* x[2] = temp[1] - temp[2] */  \
48     va3 = vec_sub(vz0,vz3)        /* x[3] = temp[0] - temp[3] */
49
50 #define VEC_TRANSPOSE_4(a0,a1,a2,a3,b0,b1,b2,b3) \
51     b0 = vec_mergeh( a0, a0 ); \
52     b1 = vec_mergeh( a1, a0 ); \
53     b2 = vec_mergeh( a2, a0 ); \
54     b3 = vec_mergeh( a3, a0 ); \
55     a0 = vec_mergeh( b0, b2 ); \
56     a1 = vec_mergel( b0, b2 ); \
57     a2 = vec_mergeh( b1, b3 ); \
58     a3 = vec_mergel( b1, b3 ); \
59     b0 = vec_mergeh( a0, a2 ); \
60     b1 = vec_mergel( a0, a2 ); \
61     b2 = vec_mergeh( a1, a3 ); \
62     b3 = vec_mergel( a1, a3 )
63
64 #define VEC_LOAD_U8_ADD_S16_STORE_U8(va)                      \
65     vdst_orig = vec_ld(0, dst);                               \
66     vdst = vec_perm(vdst_orig, zero_u8v, vdst_mask);          \
67     vdst_ss = (vec_s16) vec_mergeh(zero_u8v, vdst);         \
68     va = vec_add(va, vdst_ss);                                \
69     va_u8 = vec_packsu(va, zero_s16v);                        \
70     va_u32 = vec_splat((vec_u32)va_u8, 0);                  \
71     vec_ste(va_u32, element, (uint32_t*)dst);
72
73 static void ff_h264_idct_add_altivec(uint8_t *dst, int16_t *block, int stride)
74 {
75     vec_s16 va0, va1, va2, va3;
76     vec_s16 vz0, vz1, vz2, vz3;
77     vec_s16 vtmp0, vtmp1, vtmp2, vtmp3;
78     vec_u8 va_u8;
79     vec_u32 va_u32;
80     vec_s16 vdst_ss;
81     const vec_u16 v6us = vec_splat_u16(6);
82     vec_u8 vdst, vdst_orig;
83     vec_u8 vdst_mask = vec_lvsl(0, dst);
84     int element = ((unsigned long)dst & 0xf) >> 2;
85     LOAD_ZERO;
86
87     block[0] += 32;  /* add 32 as a DC-level for rounding */
88
89     vtmp0 = vec_ld(0,block);
90     vtmp1 = vec_sld(vtmp0, vtmp0, 8);
91     vtmp2 = vec_ld(16,block);
92     vtmp3 = vec_sld(vtmp2, vtmp2, 8);
93     memset(block, 0, 16 * sizeof(int16_t));
94
95     VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
96     VEC_TRANSPOSE_4(va0,va1,va2,va3,vtmp0,vtmp1,vtmp2,vtmp3);
97     VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
98
99     va0 = vec_sra(va0,v6us);
100     va1 = vec_sra(va1,v6us);
101     va2 = vec_sra(va2,v6us);
102     va3 = vec_sra(va3,v6us);
103
104     VEC_LOAD_U8_ADD_S16_STORE_U8(va0);
105     dst += stride;
106     VEC_LOAD_U8_ADD_S16_STORE_U8(va1);
107     dst += stride;
108     VEC_LOAD_U8_ADD_S16_STORE_U8(va2);
109     dst += stride;
110     VEC_LOAD_U8_ADD_S16_STORE_U8(va3);
111 }
112
113 #define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7,  d0, d1, d2, d3, d4, d5, d6, d7) {\
114     /*        a0  = SRC(0) + SRC(4); */ \
115     vec_s16 a0v = vec_add(s0, s4);    \
116     /*        a2  = SRC(0) - SRC(4); */ \
117     vec_s16 a2v = vec_sub(s0, s4);    \
118     /*        a4  =           (SRC(2)>>1) - SRC(6); */ \
119     vec_s16 a4v = vec_sub(vec_sra(s2, onev), s6);    \
120     /*        a6  =           (SRC(6)>>1) + SRC(2); */ \
121     vec_s16 a6v = vec_add(vec_sra(s6, onev), s2);    \
122     /*        b0  =         a0 + a6; */ \
123     vec_s16 b0v = vec_add(a0v, a6v);  \
124     /*        b2  =         a2 + a4; */ \
125     vec_s16 b2v = vec_add(a2v, a4v);  \
126     /*        b4  =         a2 - a4; */ \
127     vec_s16 b4v = vec_sub(a2v, a4v);  \
128     /*        b6  =         a0 - a6; */ \
129     vec_s16 b6v = vec_sub(a0v, a6v);  \
130     /* a1 =  SRC(5) - SRC(3) - SRC(7) - (SRC(7)>>1); */ \
131     /*        a1 =             (SRC(5)-SRC(3)) -  (SRC(7)  +  (SRC(7)>>1)); */ \
132     vec_s16 a1v = vec_sub( vec_sub(s5, s3), vec_add(s7, vec_sra(s7, onev)) ); \
133     /* a3 =  SRC(7) + SRC(1) - SRC(3) - (SRC(3)>>1); */ \
134     /*        a3 =             (SRC(7)+SRC(1)) -  (SRC(3)  +  (SRC(3)>>1)); */ \
135     vec_s16 a3v = vec_sub( vec_add(s7, s1), vec_add(s3, vec_sra(s3, onev)) );\
136     /* a5 =  SRC(7) - SRC(1) + SRC(5) + (SRC(5)>>1); */ \
137     /*        a5 =             (SRC(7)-SRC(1)) +   SRC(5) +   (SRC(5)>>1); */ \
138     vec_s16 a5v = vec_add( vec_sub(s7, s1), vec_add(s5, vec_sra(s5, onev)) );\
139     /*        a7 =                SRC(5)+SRC(3) +  SRC(1) +   (SRC(1)>>1); */ \
140     vec_s16 a7v = vec_add( vec_add(s5, s3), vec_add(s1, vec_sra(s1, onev)) );\
141     /*        b1 =                  (a7>>2)  +  a1; */ \
142     vec_s16 b1v = vec_add( vec_sra(a7v, twov), a1v); \
143     /*        b3 =          a3 +        (a5>>2); */ \
144     vec_s16 b3v = vec_add(a3v, vec_sra(a5v, twov)); \
145     /*        b5 =                  (a3>>2)  -   a5; */ \
146     vec_s16 b5v = vec_sub( vec_sra(a3v, twov), a5v); \
147     /*        b7 =           a7 -        (a1>>2); */ \
148     vec_s16 b7v = vec_sub( a7v, vec_sra(a1v, twov)); \
149     /* DST(0,    b0 + b7); */ \
150     d0 = vec_add(b0v, b7v); \
151     /* DST(1,    b2 + b5); */ \
152     d1 = vec_add(b2v, b5v); \
153     /* DST(2,    b4 + b3); */ \
154     d2 = vec_add(b4v, b3v); \
155     /* DST(3,    b6 + b1); */ \
156     d3 = vec_add(b6v, b1v); \
157     /* DST(4,    b6 - b1); */ \
158     d4 = vec_sub(b6v, b1v); \
159     /* DST(5,    b4 - b3); */ \
160     d5 = vec_sub(b4v, b3v); \
161     /* DST(6,    b2 - b5); */ \
162     d6 = vec_sub(b2v, b5v); \
163     /* DST(7,    b0 - b7); */ \
164     d7 = vec_sub(b0v, b7v); \
165 }
166
167 #define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel) { \
168     /* unaligned load */                                       \
169     vec_u8 hv = vec_ld( 0, dest );                           \
170     vec_u8 lv = vec_ld( 7, dest );                           \
171     vec_u8 dstv   = vec_perm( hv, lv, (vec_u8)perm_ldv );  \
172     vec_s16 idct_sh6 = vec_sra(idctv, sixv);                 \
173     vec_u16 dst16 = (vec_u16)vec_mergeh(zero_u8v, dstv);   \
174     vec_s16 idstsum = vec_adds(idct_sh6, (vec_s16)dst16);  \
175     vec_u8 idstsum8 = vec_packsu(zero_s16v, idstsum);        \
176     vec_u8 edgehv;                                           \
177     /* unaligned store */                                      \
178     vec_u8 bodyv  = vec_perm( idstsum8, idstsum8, perm_stv );\
179     vec_u8 edgelv = vec_perm( sel, zero_u8v, perm_stv );     \
180     lv    = vec_sel( lv, bodyv, edgelv );                      \
181     vec_st( lv, 7, dest );                                     \
182     hv    = vec_ld( 0, dest );                                 \
183     edgehv = vec_perm( zero_u8v, sel, perm_stv );              \
184     hv    = vec_sel( hv, bodyv, edgehv );                      \
185     vec_st( hv, 0, dest );                                     \
186  }
187
188 static void ff_h264_idct8_add_altivec( uint8_t *dst, int16_t *dct, int stride ) {
189     vec_s16 s0, s1, s2, s3, s4, s5, s6, s7;
190     vec_s16 d0, d1, d2, d3, d4, d5, d6, d7;
191     vec_s16 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7;
192
193     vec_u8 perm_ldv = vec_lvsl(0, dst);
194     vec_u8 perm_stv = vec_lvsr(8, dst);
195
196     const vec_u16 onev = vec_splat_u16(1);
197     const vec_u16 twov = vec_splat_u16(2);
198     const vec_u16 sixv = vec_splat_u16(6);
199
200     const vec_u8 sel = (vec_u8) {0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1};
201     LOAD_ZERO;
202
203     dct[0] += 32; // rounding for the >>6 at the end
204
205     s0 = vec_ld(0x00, (int16_t*)dct);
206     s1 = vec_ld(0x10, (int16_t*)dct);
207     s2 = vec_ld(0x20, (int16_t*)dct);
208     s3 = vec_ld(0x30, (int16_t*)dct);
209     s4 = vec_ld(0x40, (int16_t*)dct);
210     s5 = vec_ld(0x50, (int16_t*)dct);
211     s6 = vec_ld(0x60, (int16_t*)dct);
212     s7 = vec_ld(0x70, (int16_t*)dct);
213     memset(dct, 0, 64 * sizeof(int16_t));
214
215     IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7,
216                      d0, d1, d2, d3, d4, d5, d6, d7);
217
218     TRANSPOSE8( d0,  d1,  d2,  d3,  d4,  d5,  d6, d7 );
219
220     IDCT8_1D_ALTIVEC(d0,  d1,  d2,  d3,  d4,  d5,  d6, d7,
221                      idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7);
222
223     ALTIVEC_STORE_SUM_CLIP(&dst[0*stride], idct0, perm_ldv, perm_stv, sel);
224     ALTIVEC_STORE_SUM_CLIP(&dst[1*stride], idct1, perm_ldv, perm_stv, sel);
225     ALTIVEC_STORE_SUM_CLIP(&dst[2*stride], idct2, perm_ldv, perm_stv, sel);
226     ALTIVEC_STORE_SUM_CLIP(&dst[3*stride], idct3, perm_ldv, perm_stv, sel);
227     ALTIVEC_STORE_SUM_CLIP(&dst[4*stride], idct4, perm_ldv, perm_stv, sel);
228     ALTIVEC_STORE_SUM_CLIP(&dst[5*stride], idct5, perm_ldv, perm_stv, sel);
229     ALTIVEC_STORE_SUM_CLIP(&dst[6*stride], idct6, perm_ldv, perm_stv, sel);
230     ALTIVEC_STORE_SUM_CLIP(&dst[7*stride], idct7, perm_ldv, perm_stv, sel);
231 }
232
233 static av_always_inline void h264_idct_dc_add_internal(uint8_t *dst, int16_t *block, int stride, int size)
234 {
235     vec_s16 dc16;
236     vec_u8 dcplus, dcminus, v0, v1, v2, v3, aligner;
237     LOAD_ZERO;
238     DECLARE_ALIGNED(16, int, dc);
239     int i;
240
241     dc = (block[0] + 32) >> 6;
242     block[0] = 0;
243     dc16 = vec_splat((vec_s16) vec_lde(0, &dc), 1);
244
245     if (size == 4)
246         dc16 = vec_sld(dc16, zero_s16v, 8);
247     dcplus = vec_packsu(dc16, zero_s16v);
248     dcminus = vec_packsu(vec_sub(zero_s16v, dc16), zero_s16v);
249
250     aligner = vec_lvsr(0, dst);
251     dcplus = vec_perm(dcplus, dcplus, aligner);
252     dcminus = vec_perm(dcminus, dcminus, aligner);
253
254     for (i = 0; i < size; i += 4) {
255         v0 = vec_ld(0, dst+0*stride);
256         v1 = vec_ld(0, dst+1*stride);
257         v2 = vec_ld(0, dst+2*stride);
258         v3 = vec_ld(0, dst+3*stride);
259
260         v0 = vec_adds(v0, dcplus);
261         v1 = vec_adds(v1, dcplus);
262         v2 = vec_adds(v2, dcplus);
263         v3 = vec_adds(v3, dcplus);
264
265         v0 = vec_subs(v0, dcminus);
266         v1 = vec_subs(v1, dcminus);
267         v2 = vec_subs(v2, dcminus);
268         v3 = vec_subs(v3, dcminus);
269
270         vec_st(v0, 0, dst+0*stride);
271         vec_st(v1, 0, dst+1*stride);
272         vec_st(v2, 0, dst+2*stride);
273         vec_st(v3, 0, dst+3*stride);
274
275         dst += 4*stride;
276     }
277 }
278
279 static void h264_idct_dc_add_altivec(uint8_t *dst, int16_t *block, int stride)
280 {
281     h264_idct_dc_add_internal(dst, block, stride, 4);
282 }
283
284 static void ff_h264_idct8_dc_add_altivec(uint8_t *dst, int16_t *block, int stride)
285 {
286     h264_idct_dc_add_internal(dst, block, stride, 8);
287 }
288
289 static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
290     int i;
291     for(i=0; i<16; i++){
292         int nnz = nnzc[ scan8[i] ];
293         if(nnz){
294             if(nnz==1 && block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
295             else                      ff_h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
296         }
297     }
298 }
299
300 static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
301     int i;
302     for(i=0; i<16; i++){
303         if(nnzc[ scan8[i] ]) ff_h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
304         else if(block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
305     }
306 }
307
308 static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
309     int i;
310     for(i=0; i<16; i+=4){
311         int nnz = nnzc[ scan8[i] ];
312         if(nnz){
313             if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
314             else                      ff_h264_idct8_add_altivec   (dst + block_offset[i], block + i*16, stride);
315         }
316     }
317 }
318
319 static void ff_h264_idct_add8_altivec(uint8_t **dest, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
320     int i, j;
321     for (j = 1; j < 3; j++) {
322         for(i = j * 16; i < j * 16 + 4; i++){
323             if(nnzc[ scan8[i] ])
324                 ff_h264_idct_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
325             else if(block[i*16])
326                 h264_idct_dc_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
327         }
328     }
329 }
330
331 #define transpose4x16(r0, r1, r2, r3) {      \
332     register vec_u8 r4;                    \
333     register vec_u8 r5;                    \
334     register vec_u8 r6;                    \
335     register vec_u8 r7;                    \
336                                              \
337     r4 = vec_mergeh(r0, r2);  /*0, 2 set 0*/ \
338     r5 = vec_mergel(r0, r2);  /*0, 2 set 1*/ \
339     r6 = vec_mergeh(r1, r3);  /*1, 3 set 0*/ \
340     r7 = vec_mergel(r1, r3);  /*1, 3 set 1*/ \
341                                              \
342     r0 = vec_mergeh(r4, r6);  /*all set 0*/  \
343     r1 = vec_mergel(r4, r6);  /*all set 1*/  \
344     r2 = vec_mergeh(r5, r7);  /*all set 2*/  \
345     r3 = vec_mergel(r5, r7);  /*all set 3*/  \
346 }
347
348 static inline void write16x4(uint8_t *dst, int dst_stride,
349                              register vec_u8 r0, register vec_u8 r1,
350                              register vec_u8 r2, register vec_u8 r3) {
351     DECLARE_ALIGNED(16, unsigned char, result)[64];
352     uint32_t *src_int = (uint32_t *)result, *dst_int = (uint32_t *)dst;
353     int int_dst_stride = dst_stride/4;
354
355     vec_st(r0, 0, result);
356     vec_st(r1, 16, result);
357     vec_st(r2, 32, result);
358     vec_st(r3, 48, result);
359     /* FIXME: there has to be a better way!!!! */
360     *dst_int = *src_int;
361     *(dst_int+   int_dst_stride) = *(src_int + 1);
362     *(dst_int+ 2*int_dst_stride) = *(src_int + 2);
363     *(dst_int+ 3*int_dst_stride) = *(src_int + 3);
364     *(dst_int+ 4*int_dst_stride) = *(src_int + 4);
365     *(dst_int+ 5*int_dst_stride) = *(src_int + 5);
366     *(dst_int+ 6*int_dst_stride) = *(src_int + 6);
367     *(dst_int+ 7*int_dst_stride) = *(src_int + 7);
368     *(dst_int+ 8*int_dst_stride) = *(src_int + 8);
369     *(dst_int+ 9*int_dst_stride) = *(src_int + 9);
370     *(dst_int+10*int_dst_stride) = *(src_int + 10);
371     *(dst_int+11*int_dst_stride) = *(src_int + 11);
372     *(dst_int+12*int_dst_stride) = *(src_int + 12);
373     *(dst_int+13*int_dst_stride) = *(src_int + 13);
374     *(dst_int+14*int_dst_stride) = *(src_int + 14);
375     *(dst_int+15*int_dst_stride) = *(src_int + 15);
376 }
377
378 /** @brief performs a 6x16 transpose of data in src, and stores it to dst
379     @todo FIXME: see if we can't spare some vec_lvsl() by them factorizing
380     out of unaligned_load() */
381 #define readAndTranspose16x6(src, src_stride, r8, r9, r10, r11, r12, r13) {\
382     register vec_u8 r0  = unaligned_load(0,             src);            \
383     register vec_u8 r1  = unaligned_load(   src_stride, src);            \
384     register vec_u8 r2  = unaligned_load(2* src_stride, src);            \
385     register vec_u8 r3  = unaligned_load(3* src_stride, src);            \
386     register vec_u8 r4  = unaligned_load(4* src_stride, src);            \
387     register vec_u8 r5  = unaligned_load(5* src_stride, src);            \
388     register vec_u8 r6  = unaligned_load(6* src_stride, src);            \
389     register vec_u8 r7  = unaligned_load(7* src_stride, src);            \
390     register vec_u8 r14 = unaligned_load(14*src_stride, src);            \
391     register vec_u8 r15 = unaligned_load(15*src_stride, src);            \
392                                                                            \
393     r8  = unaligned_load( 8*src_stride, src);                              \
394     r9  = unaligned_load( 9*src_stride, src);                              \
395     r10 = unaligned_load(10*src_stride, src);                              \
396     r11 = unaligned_load(11*src_stride, src);                              \
397     r12 = unaligned_load(12*src_stride, src);                              \
398     r13 = unaligned_load(13*src_stride, src);                              \
399                                                                            \
400     /*Merge first pairs*/                                                  \
401     r0 = vec_mergeh(r0, r8);    /*0, 8*/                                   \
402     r1 = vec_mergeh(r1, r9);    /*1, 9*/                                   \
403     r2 = vec_mergeh(r2, r10);   /*2,10*/                                   \
404     r3 = vec_mergeh(r3, r11);   /*3,11*/                                   \
405     r4 = vec_mergeh(r4, r12);   /*4,12*/                                   \
406     r5 = vec_mergeh(r5, r13);   /*5,13*/                                   \
407     r6 = vec_mergeh(r6, r14);   /*6,14*/                                   \
408     r7 = vec_mergeh(r7, r15);   /*7,15*/                                   \
409                                                                            \
410     /*Merge second pairs*/                                                 \
411     r8  = vec_mergeh(r0, r4);   /*0,4, 8,12 set 0*/                        \
412     r9  = vec_mergel(r0, r4);   /*0,4, 8,12 set 1*/                        \
413     r10 = vec_mergeh(r1, r5);   /*1,5, 9,13 set 0*/                        \
414     r11 = vec_mergel(r1, r5);   /*1,5, 9,13 set 1*/                        \
415     r12 = vec_mergeh(r2, r6);   /*2,6,10,14 set 0*/                        \
416     r13 = vec_mergel(r2, r6);   /*2,6,10,14 set 1*/                        \
417     r14 = vec_mergeh(r3, r7);   /*3,7,11,15 set 0*/                        \
418     r15 = vec_mergel(r3, r7);   /*3,7,11,15 set 1*/                        \
419                                                                            \
420     /*Third merge*/                                                        \
421     r0 = vec_mergeh(r8,  r12);  /*0,2,4,6,8,10,12,14 set 0*/               \
422     r1 = vec_mergel(r8,  r12);  /*0,2,4,6,8,10,12,14 set 1*/               \
423     r2 = vec_mergeh(r9,  r13);  /*0,2,4,6,8,10,12,14 set 2*/               \
424     r4 = vec_mergeh(r10, r14);  /*1,3,5,7,9,11,13,15 set 0*/               \
425     r5 = vec_mergel(r10, r14);  /*1,3,5,7,9,11,13,15 set 1*/               \
426     r6 = vec_mergeh(r11, r15);  /*1,3,5,7,9,11,13,15 set 2*/               \
427     /* Don't need to compute 3 and 7*/                                     \
428                                                                            \
429     /*Final merge*/                                                        \
430     r8  = vec_mergeh(r0, r4);   /*all set 0*/                              \
431     r9  = vec_mergel(r0, r4);   /*all set 1*/                              \
432     r10 = vec_mergeh(r1, r5);   /*all set 2*/                              \
433     r11 = vec_mergel(r1, r5);   /*all set 3*/                              \
434     r12 = vec_mergeh(r2, r6);   /*all set 4*/                              \
435     r13 = vec_mergel(r2, r6);   /*all set 5*/                              \
436     /* Don't need to compute 14 and 15*/                                   \
437                                                                            \
438 }
439
440 // out: o = |x-y| < a
441 static inline vec_u8 diff_lt_altivec ( register vec_u8 x,
442                                          register vec_u8 y,
443                                          register vec_u8 a) {
444
445     register vec_u8 diff = vec_subs(x, y);
446     register vec_u8 diffneg = vec_subs(y, x);
447     register vec_u8 o = vec_or(diff, diffneg); /* |x-y| */
448     o = (vec_u8)vec_cmplt(o, a);
449     return o;
450 }
451
452 static inline vec_u8 h264_deblock_mask ( register vec_u8 p0,
453                                            register vec_u8 p1,
454                                            register vec_u8 q0,
455                                            register vec_u8 q1,
456                                            register vec_u8 alpha,
457                                            register vec_u8 beta) {
458
459     register vec_u8 mask;
460     register vec_u8 tempmask;
461
462     mask = diff_lt_altivec(p0, q0, alpha);
463     tempmask = diff_lt_altivec(p1, p0, beta);
464     mask = vec_and(mask, tempmask);
465     tempmask = diff_lt_altivec(q1, q0, beta);
466     mask = vec_and(mask, tempmask);
467
468     return mask;
469 }
470
471 // out: newp1 = clip((p2 + ((p0 + q0 + 1) >> 1)) >> 1, p1-tc0, p1+tc0)
472 static inline vec_u8 h264_deblock_q1(register vec_u8 p0,
473                                        register vec_u8 p1,
474                                        register vec_u8 p2,
475                                        register vec_u8 q0,
476                                        register vec_u8 tc0) {
477
478     register vec_u8 average = vec_avg(p0, q0);
479     register vec_u8 temp;
480     register vec_u8 uncliped;
481     register vec_u8 ones;
482     register vec_u8 max;
483     register vec_u8 min;
484     register vec_u8 newp1;
485
486     temp = vec_xor(average, p2);
487     average = vec_avg(average, p2);     /*avg(p2, avg(p0, q0)) */
488     ones = vec_splat_u8(1);
489     temp = vec_and(temp, ones);         /*(p2^avg(p0, q0)) & 1 */
490     uncliped = vec_subs(average, temp); /*(p2+((p0+q0+1)>>1))>>1 */
491     max = vec_adds(p1, tc0);
492     min = vec_subs(p1, tc0);
493     newp1 = vec_max(min, uncliped);
494     newp1 = vec_min(max, newp1);
495     return newp1;
496 }
497
498 #define h264_deblock_p0_q0(p0, p1, q0, q1, tc0masked) {                                           \
499                                                                                                   \
500     const vec_u8 A0v = vec_sl(vec_splat_u8(10), vec_splat_u8(4));                               \
501                                                                                                   \
502     register vec_u8 pq0bit = vec_xor(p0,q0);                                                    \
503     register vec_u8 q1minus;                                                                    \
504     register vec_u8 p0minus;                                                                    \
505     register vec_u8 stage1;                                                                     \
506     register vec_u8 stage2;                                                                     \
507     register vec_u8 vec160;                                                                     \
508     register vec_u8 delta;                                                                      \
509     register vec_u8 deltaneg;                                                                   \
510                                                                                                   \
511     q1minus = vec_nor(q1, q1);                 /* 255 - q1 */                                     \
512     stage1 = vec_avg(p1, q1minus);             /* (p1 - q1 + 256)>>1 */                           \
513     stage2 = vec_sr(stage1, vec_splat_u8(1));  /* (p1 - q1 + 256)>>2 = 64 + (p1 - q1) >> 2 */     \
514     p0minus = vec_nor(p0, p0);                 /* 255 - p0 */                                     \
515     stage1 = vec_avg(q0, p0minus);             /* (q0 - p0 + 256)>>1 */                           \
516     pq0bit = vec_and(pq0bit, vec_splat_u8(1));                                                    \
517     stage2 = vec_avg(stage2, pq0bit);          /* 32 + ((q0 - p0)&1 + (p1 - q1) >> 2 + 1) >> 1 */ \
518     stage2 = vec_adds(stage2, stage1);         /* 160 + ((p0 - q0) + (p1 - q1) >> 2 + 1) >> 1 */  \
519     vec160 = vec_ld(0, &A0v);                                                                     \
520     deltaneg = vec_subs(vec160, stage2);       /* -d */                                           \
521     delta = vec_subs(stage2, vec160);          /* d */                                            \
522     deltaneg = vec_min(tc0masked, deltaneg);                                                      \
523     delta = vec_min(tc0masked, delta);                                                            \
524     p0 = vec_subs(p0, deltaneg);                                                                  \
525     q0 = vec_subs(q0, delta);                                                                     \
526     p0 = vec_adds(p0, delta);                                                                     \
527     q0 = vec_adds(q0, deltaneg);                                                                  \
528 }
529
530 #define h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0) {            \
531     DECLARE_ALIGNED(16, unsigned char, temp)[16];                                             \
532     register vec_u8 alphavec;                                                              \
533     register vec_u8 betavec;                                                               \
534     register vec_u8 mask;                                                                  \
535     register vec_u8 p1mask;                                                                \
536     register vec_u8 q1mask;                                                                \
537     register vector signed   char tc0vec;                                                    \
538     register vec_u8 finaltc0;                                                              \
539     register vec_u8 tc0masked;                                                             \
540     register vec_u8 newp1;                                                                 \
541     register vec_u8 newq1;                                                                 \
542                                                                                              \
543     temp[0] = alpha;                                                                         \
544     temp[1] = beta;                                                                          \
545     alphavec = vec_ld(0, temp);                                                              \
546     betavec = vec_splat(alphavec, 0x1);                                                      \
547     alphavec = vec_splat(alphavec, 0x0);                                                     \
548     mask = h264_deblock_mask(p0, p1, q0, q1, alphavec, betavec); /*if in block */            \
549                                                                                              \
550     AV_COPY32(temp, tc0);                                                                    \
551     tc0vec = vec_ld(0, (signed char*)temp);                                                  \
552     tc0vec = vec_mergeh(tc0vec, tc0vec);                                                     \
553     tc0vec = vec_mergeh(tc0vec, tc0vec);                                                     \
554     mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_s8(-1)));  /* if tc0[i] >= 0 */         \
555     finaltc0 = vec_and((vec_u8)tc0vec, mask);     /* tc = tc0 */                           \
556                                                                                              \
557     p1mask = diff_lt_altivec(p2, p0, betavec);                                               \
558     p1mask = vec_and(p1mask, mask);                             /* if ( |p2 - p0| < beta) */ \
559     tc0masked = vec_and(p1mask, (vec_u8)tc0vec);                                           \
560     finaltc0 = vec_sub(finaltc0, p1mask);                       /* tc++ */                   \
561     newp1 = h264_deblock_q1(p0, p1, p2, q0, tc0masked);                                      \
562     /*end if*/                                                                               \
563                                                                                              \
564     q1mask = diff_lt_altivec(q2, q0, betavec);                                               \
565     q1mask = vec_and(q1mask, mask);                             /* if ( |q2 - q0| < beta ) */\
566     tc0masked = vec_and(q1mask, (vec_u8)tc0vec);                                           \
567     finaltc0 = vec_sub(finaltc0, q1mask);                       /* tc++ */                   \
568     newq1 = h264_deblock_q1(p0, q1, q2, q0, tc0masked);                                      \
569     /*end if*/                                                                               \
570                                                                                              \
571     h264_deblock_p0_q0(p0, p1, q0, q1, finaltc0);                                            \
572     p1 = newp1;                                                                              \
573     q1 = newq1;                                                                              \
574 }
575
576 static void h264_v_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) {
577
578     if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) >= 0) {
579         register vec_u8 p2 = vec_ld(-3*stride, pix);
580         register vec_u8 p1 = vec_ld(-2*stride, pix);
581         register vec_u8 p0 = vec_ld(-1*stride, pix);
582         register vec_u8 q0 = vec_ld(0, pix);
583         register vec_u8 q1 = vec_ld(stride, pix);
584         register vec_u8 q2 = vec_ld(2*stride, pix);
585         h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0);
586         vec_st(p1, -2*stride, pix);
587         vec_st(p0, -1*stride, pix);
588         vec_st(q0, 0, pix);
589         vec_st(q1, stride, pix);
590     }
591 }
592
593 static void h264_h_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) {
594
595     register vec_u8 line0, line1, line2, line3, line4, line5;
596     if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) < 0)
597         return;
598     readAndTranspose16x6(pix-3, stride, line0, line1, line2, line3, line4, line5);
599     h264_loop_filter_luma_altivec(line0, line1, line2, line3, line4, line5, alpha, beta, tc0);
600     transpose4x16(line1, line2, line3, line4);
601     write16x4(pix-2, stride, line1, line2, line3, line4);
602 }
603
604 static av_always_inline
605 void weight_h264_W_altivec(uint8_t *block, int stride, int height,
606                            int log2_denom, int weight, int offset, int w)
607 {
608     int y, aligned;
609     vec_u8 vblock;
610     vec_s16 vtemp, vweight, voffset, v0, v1;
611     vec_u16 vlog2_denom;
612     DECLARE_ALIGNED(16, int32_t, temp)[4];
613     LOAD_ZERO;
614
615     offset <<= log2_denom;
616     if(log2_denom) offset += 1<<(log2_denom-1);
617     temp[0] = log2_denom;
618     temp[1] = weight;
619     temp[2] = offset;
620
621     vtemp = (vec_s16)vec_ld(0, temp);
622     vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
623     vweight = vec_splat(vtemp, 3);
624     voffset = vec_splat(vtemp, 5);
625     aligned = !((unsigned long)block & 0xf);
626
627     for (y = 0; y < height; y++) {
628         vblock = vec_ld(0, block);
629
630         v0 = (vec_s16)vec_mergeh(zero_u8v, vblock);
631         v1 = (vec_s16)vec_mergel(zero_u8v, vblock);
632
633         if (w == 16 || aligned) {
634             v0 = vec_mladd(v0, vweight, zero_s16v);
635             v0 = vec_adds(v0, voffset);
636             v0 = vec_sra(v0, vlog2_denom);
637         }
638         if (w == 16 || !aligned) {
639             v1 = vec_mladd(v1, vweight, zero_s16v);
640             v1 = vec_adds(v1, voffset);
641             v1 = vec_sra(v1, vlog2_denom);
642         }
643         vblock = vec_packsu(v0, v1);
644         vec_st(vblock, 0, block);
645
646         block += stride;
647     }
648 }
649
650 static av_always_inline
651 void biweight_h264_W_altivec(uint8_t *dst, uint8_t *src, int stride, int height,
652                              int log2_denom, int weightd, int weights, int offset, int w)
653 {
654     int y, dst_aligned, src_aligned;
655     vec_u8 vsrc, vdst;
656     vec_s16 vtemp, vweights, vweightd, voffset, v0, v1, v2, v3;
657     vec_u16 vlog2_denom;
658     DECLARE_ALIGNED(16, int32_t, temp)[4];
659     LOAD_ZERO;
660
661     offset = ((offset + 1) | 1) << log2_denom;
662     temp[0] = log2_denom+1;
663     temp[1] = weights;
664     temp[2] = weightd;
665     temp[3] = offset;
666
667     vtemp = (vec_s16)vec_ld(0, temp);
668     vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
669     vweights = vec_splat(vtemp, 3);
670     vweightd = vec_splat(vtemp, 5);
671     voffset = vec_splat(vtemp, 7);
672     dst_aligned = !((unsigned long)dst & 0xf);
673     src_aligned = !((unsigned long)src & 0xf);
674
675     for (y = 0; y < height; y++) {
676         vdst = vec_ld(0, dst);
677         vsrc = vec_ld(0, src);
678
679         v0 = (vec_s16)vec_mergeh(zero_u8v, vdst);
680         v1 = (vec_s16)vec_mergel(zero_u8v, vdst);
681         v2 = (vec_s16)vec_mergeh(zero_u8v, vsrc);
682         v3 = (vec_s16)vec_mergel(zero_u8v, vsrc);
683
684         if (w == 8) {
685             if (src_aligned)
686                 v3 = v2;
687             else
688                 v2 = v3;
689         }
690
691         if (w == 16 || dst_aligned) {
692             v0 = vec_mladd(v0, vweightd, zero_s16v);
693             v2 = vec_mladd(v2, vweights, zero_s16v);
694
695             v0 = vec_adds(v0, voffset);
696             v0 = vec_adds(v0, v2);
697             v0 = vec_sra(v0, vlog2_denom);
698         }
699         if (w == 16 || !dst_aligned) {
700             v1 = vec_mladd(v1, vweightd, zero_s16v);
701             v3 = vec_mladd(v3, vweights, zero_s16v);
702
703             v1 = vec_adds(v1, voffset);
704             v1 = vec_adds(v1, v3);
705             v1 = vec_sra(v1, vlog2_denom);
706         }
707         vdst = vec_packsu(v0, v1);
708         vec_st(vdst, 0, dst);
709
710         dst += stride;
711         src += stride;
712     }
713 }
714
715 #define H264_WEIGHT(W) \
716 static void ff_weight_h264_pixels ## W ## _altivec(uint8_t *block, int stride, int height, \
717                                                    int log2_denom, int weight, int offset){ \
718     weight_h264_W_altivec(block, stride, height, log2_denom, weight, offset, W); \
719 }\
720 static void ff_biweight_h264_pixels ## W ## _altivec(uint8_t *dst, uint8_t *src, int stride, int height, \
721                                                      int log2_denom, int weightd, int weights, int offset){ \
722     biweight_h264_W_altivec(dst, src, stride, height, log2_denom, weightd, weights, offset, W); \
723 }
724
725 H264_WEIGHT(16)
726 H264_WEIGHT( 8)
727 #endif /* HAVE_ALTIVEC */
728
729 av_cold void ff_h264dsp_init_ppc(H264DSPContext *c, const int bit_depth,
730                                  const int chroma_format_idc)
731 {
732 #if HAVE_ALTIVEC
733     if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) {
734     if (bit_depth == 8) {
735         c->h264_idct_add = ff_h264_idct_add_altivec;
736         if (chroma_format_idc == 1)
737             c->h264_idct_add8 = ff_h264_idct_add8_altivec;
738         c->h264_idct_add16 = ff_h264_idct_add16_altivec;
739         c->h264_idct_add16intra = ff_h264_idct_add16intra_altivec;
740         c->h264_idct_dc_add= h264_idct_dc_add_altivec;
741         c->h264_idct8_dc_add = ff_h264_idct8_dc_add_altivec;
742         c->h264_idct8_add = ff_h264_idct8_add_altivec;
743         c->h264_idct8_add4 = ff_h264_idct8_add4_altivec;
744         c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_altivec;
745         c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_altivec;
746
747         c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels16_altivec;
748         c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels8_altivec;
749         c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels16_altivec;
750         c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels8_altivec;
751     }
752     }
753 #endif /* HAVE_ALTIVEC */
754 }