FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/libavcodec/x86/cavsdsp.c
Date: 2025-03-25 06:04:38
Exec Total Coverage
Lines: 25 43 58.1%
Functions: 3 45 6.7%
Branches: 8 16 50.0%
Line Branch Exec Source
1 /*
2 * Chinese AVS video (AVS1-P2, JiZhun profile) decoder.
3 * Copyright (c) 2006 Stefan Gehrer <stefan.gehrer@gmx.de>
4 *
5 * MMX-optimized DSP functions, based on H.264 optimizations by
6 * Michael Niedermayer and Loren Merritt
7 *
8 * This file is part of FFmpeg.
9 *
10 * FFmpeg is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
14 *
15 * FFmpeg is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
19 *
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 */
24
25 #include "libavutil/attributes.h"
26 #include "libavutil/common.h"
27 #include "libavutil/cpu.h"
28 #include "libavutil/mem_internal.h"
29 #include "libavutil/x86/asm.h"
30 #include "libavutil/x86/cpu.h"
31 #include "libavcodec/cavsdsp.h"
32 #include "libavcodec/idctdsp.h"
33 #include "constants.h"
34 #include "fpel.h"
35 #include "idctdsp.h"
36 #include "config.h"
37
38
39 #if HAVE_SSE2_EXTERNAL
40
41 void ff_cavs_idct8_sse2(int16_t *out, const int16_t *in);
42
43 4921 static void cavs_idct8_add_sse2(uint8_t *dst, int16_t *block, ptrdiff_t stride)
44 {
45 4921 LOCAL_ALIGNED(16, int16_t, b2, [64]);
46 4921 ff_cavs_idct8_sse2(b2, block);
47 4921 ff_add_pixels_clamped_sse2(b2, dst, stride);
48 4921 }
49
50 #endif /* HAVE_SSE2_EXTERNAL */
51
52 #if HAVE_MMXEXT_INLINE
53
54 DECLARE_ASM_CONST(8, uint64_t, pw_42) = 0x002A002A002A002AULL;
55 DECLARE_ASM_CONST(8, uint64_t, pw_96) = 0x0060006000600060ULL;
56
57 /*****************************************************************************
58 *
59 * motion compensation
60 *
61 ****************************************************************************/
62
63 /* vertical filter [-1 -2 96 42 -7 0] */
64 #define QPEL_CAVSV1(A,B,C,D,E,F,OP,ADD, MUL1, MUL2) \
65 "movd (%0), "#F" \n\t"\
66 "movq "#C", %%mm6 \n\t"\
67 "pmullw "MANGLE(MUL1)", %%mm6\n\t"\
68 "movq "#D", %%mm7 \n\t"\
69 "pmullw "MANGLE(MUL2)", %%mm7\n\t"\
70 "psllw $3, "#E" \n\t"\
71 "psubw "#E", %%mm6 \n\t"\
72 "psraw $3, "#E" \n\t"\
73 "paddw %%mm7, %%mm6 \n\t"\
74 "paddw "#E", %%mm6 \n\t"\
75 "paddw "#B", "#B" \n\t"\
76 "pxor %%mm7, %%mm7 \n\t"\
77 "add %2, %0 \n\t"\
78 "punpcklbw %%mm7, "#F" \n\t"\
79 "psubw "#B", %%mm6 \n\t"\
80 "psraw $1, "#B" \n\t"\
81 "psubw "#A", %%mm6 \n\t"\
82 "paddw "MANGLE(ADD)", %%mm6 \n\t"\
83 "psraw $7, %%mm6 \n\t"\
84 "packuswb %%mm6, %%mm6 \n\t"\
85 OP(%%mm6, (%1), A, d) \
86 "add %3, %1 \n\t"
87
88 /* vertical filter [ 0 -1 5 5 -1 0] */
89 #define QPEL_CAVSV2(A,B,C,D,E,F,OP,ADD, MUL1, MUL2) \
90 "movd (%0), "#F" \n\t"\
91 "movq "#C", %%mm6 \n\t"\
92 "paddw "#D", %%mm6 \n\t"\
93 "pmullw "MANGLE(MUL1)", %%mm6\n\t"\
94 "add %2, %0 \n\t"\
95 "punpcklbw %%mm7, "#F" \n\t"\
96 "psubw "#B", %%mm6 \n\t"\
97 "psubw "#E", %%mm6 \n\t"\
98 "paddw "MANGLE(ADD)", %%mm6 \n\t"\
99 "psraw $3, %%mm6 \n\t"\
100 "packuswb %%mm6, %%mm6 \n\t"\
101 OP(%%mm6, (%1), A, d) \
102 "add %3, %1 \n\t"
103
104 /* vertical filter [ 0 -7 42 96 -2 -1] */
105 #define QPEL_CAVSV3(A,B,C,D,E,F,OP,ADD, MUL1, MUL2) \
106 "movd (%0), "#F" \n\t"\
107 "movq "#C", %%mm6 \n\t"\
108 "pmullw "MANGLE(MUL2)", %%mm6\n\t"\
109 "movq "#D", %%mm7 \n\t"\
110 "pmullw "MANGLE(MUL1)", %%mm7\n\t"\
111 "psllw $3, "#B" \n\t"\
112 "psubw "#B", %%mm6 \n\t"\
113 "psraw $3, "#B" \n\t"\
114 "paddw %%mm7, %%mm6 \n\t"\
115 "paddw "#B", %%mm6 \n\t"\
116 "paddw "#E", "#E" \n\t"\
117 "pxor %%mm7, %%mm7 \n\t"\
118 "add %2, %0 \n\t"\
119 "punpcklbw %%mm7, "#F" \n\t"\
120 "psubw "#E", %%mm6 \n\t"\
121 "psraw $1, "#E" \n\t"\
122 "psubw "#F", %%mm6 \n\t"\
123 "paddw "MANGLE(ADD)", %%mm6 \n\t"\
124 "psraw $7, %%mm6 \n\t"\
125 "packuswb %%mm6, %%mm6 \n\t"\
126 OP(%%mm6, (%1), A, d) \
127 "add %3, %1 \n\t"
128
129
130 #define QPEL_CAVSVNUM(VOP,OP,ADD,MUL1,MUL2)\
131 int w= 2;\
132 src -= 2*srcStride;\
133 \
134 while(w--){\
135 __asm__ volatile(\
136 "pxor %%mm7, %%mm7 \n\t"\
137 "movd (%0), %%mm0 \n\t"\
138 "add %2, %0 \n\t"\
139 "movd (%0), %%mm1 \n\t"\
140 "add %2, %0 \n\t"\
141 "movd (%0), %%mm2 \n\t"\
142 "add %2, %0 \n\t"\
143 "movd (%0), %%mm3 \n\t"\
144 "add %2, %0 \n\t"\
145 "movd (%0), %%mm4 \n\t"\
146 "add %2, %0 \n\t"\
147 "punpcklbw %%mm7, %%mm0 \n\t"\
148 "punpcklbw %%mm7, %%mm1 \n\t"\
149 "punpcklbw %%mm7, %%mm2 \n\t"\
150 "punpcklbw %%mm7, %%mm3 \n\t"\
151 "punpcklbw %%mm7, %%mm4 \n\t"\
152 VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, ADD, MUL1, MUL2)\
153 VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, ADD, MUL1, MUL2)\
154 VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, ADD, MUL1, MUL2)\
155 VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, ADD, MUL1, MUL2)\
156 VOP(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP, ADD, MUL1, MUL2)\
157 VOP(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP, ADD, MUL1, MUL2)\
158 VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, ADD, MUL1, MUL2)\
159 VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, ADD, MUL1, MUL2)\
160 \
161 : "+a"(src), "+c"(dst)\
162 : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride)\
163 NAMED_CONSTRAINTS_ADD(ADD,MUL1,MUL2)\
164 : "memory"\
165 );\
166 if(h==16){\
167 __asm__ volatile(\
168 VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, ADD, MUL1, MUL2)\
169 VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, ADD, MUL1, MUL2)\
170 VOP(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP, ADD, MUL1, MUL2)\
171 VOP(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP, ADD, MUL1, MUL2)\
172 VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, ADD, MUL1, MUL2)\
173 VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, ADD, MUL1, MUL2)\
174 VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, ADD, MUL1, MUL2)\
175 VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, ADD, MUL1, MUL2)\
176 \
177 : "+a"(src), "+c"(dst)\
178 : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride)\
179 NAMED_CONSTRAINTS_ADD(ADD,MUL1,MUL2)\
180 : "memory"\
181 );\
182 }\
183 src += 4-(h+5)*srcStride;\
184 dst += 4-h*dstStride;\
185 }
186
187 #define QPEL_CAVS(OPNAME, OP, MMX)\
188 static void OPNAME ## cavs_qpel8_h_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
189 {\
190 int h=8;\
191 __asm__ volatile(\
192 "pxor %%mm7, %%mm7 \n\t"\
193 "movq "MANGLE(ff_pw_5)", %%mm6\n\t"\
194 "1: \n\t"\
195 "movq (%0), %%mm0 \n\t"\
196 "movq 1(%0), %%mm2 \n\t"\
197 "movq %%mm0, %%mm1 \n\t"\
198 "movq %%mm2, %%mm3 \n\t"\
199 "punpcklbw %%mm7, %%mm0 \n\t"\
200 "punpckhbw %%mm7, %%mm1 \n\t"\
201 "punpcklbw %%mm7, %%mm2 \n\t"\
202 "punpckhbw %%mm7, %%mm3 \n\t"\
203 "paddw %%mm2, %%mm0 \n\t"\
204 "paddw %%mm3, %%mm1 \n\t"\
205 "pmullw %%mm6, %%mm0 \n\t"\
206 "pmullw %%mm6, %%mm1 \n\t"\
207 "movq -1(%0), %%mm2 \n\t"\
208 "movq 2(%0), %%mm4 \n\t"\
209 "movq %%mm2, %%mm3 \n\t"\
210 "movq %%mm4, %%mm5 \n\t"\
211 "punpcklbw %%mm7, %%mm2 \n\t"\
212 "punpckhbw %%mm7, %%mm3 \n\t"\
213 "punpcklbw %%mm7, %%mm4 \n\t"\
214 "punpckhbw %%mm7, %%mm5 \n\t"\
215 "paddw %%mm4, %%mm2 \n\t"\
216 "paddw %%mm3, %%mm5 \n\t"\
217 "psubw %%mm2, %%mm0 \n\t"\
218 "psubw %%mm5, %%mm1 \n\t"\
219 "movq "MANGLE(ff_pw_4)", %%mm5\n\t"\
220 "paddw %%mm5, %%mm0 \n\t"\
221 "paddw %%mm5, %%mm1 \n\t"\
222 "psraw $3, %%mm0 \n\t"\
223 "psraw $3, %%mm1 \n\t"\
224 "packuswb %%mm1, %%mm0 \n\t"\
225 OP(%%mm0, (%1),%%mm5, q) \
226 "add %3, %0 \n\t"\
227 "add %4, %1 \n\t"\
228 "decl %2 \n\t"\
229 " jnz 1b \n\t"\
230 : "+a"(src), "+c"(dst), "+m"(h)\
231 : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride)\
232 NAMED_CONSTRAINTS_ADD(ff_pw_4,ff_pw_5)\
233 : "memory"\
234 );\
235 }\
236 \
237 static inline void OPNAME ## cavs_qpel8or16_v1_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride, int h)\
238 { \
239 QPEL_CAVSVNUM(QPEL_CAVSV1,OP,ff_pw_64,pw_96,pw_42) \
240 }\
241 \
242 static inline void OPNAME ## cavs_qpel8or16_v2_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride, int h)\
243 { \
244 QPEL_CAVSVNUM(QPEL_CAVSV2,OP,ff_pw_4,ff_pw_5,pw_42) \
245 }\
246 \
247 static inline void OPNAME ## cavs_qpel8or16_v3_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride, int h)\
248 { \
249 QPEL_CAVSVNUM(QPEL_CAVSV3,OP,ff_pw_64,pw_96,pw_42) \
250 }\
251 \
252 static void OPNAME ## cavs_qpel8_v1_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
253 { \
254 OPNAME ## cavs_qpel8or16_v1_ ## MMX(dst , src , dstStride, srcStride, 8);\
255 }\
256 static void OPNAME ## cavs_qpel16_v1_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
257 { \
258 OPNAME ## cavs_qpel8or16_v1_ ## MMX(dst , src , dstStride, srcStride, 16);\
259 OPNAME ## cavs_qpel8or16_v1_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
260 }\
261 \
262 static void OPNAME ## cavs_qpel8_v2_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
263 { \
264 OPNAME ## cavs_qpel8or16_v2_ ## MMX(dst , src , dstStride, srcStride, 8);\
265 }\
266 static void OPNAME ## cavs_qpel16_v2_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
267 { \
268 OPNAME ## cavs_qpel8or16_v2_ ## MMX(dst , src , dstStride, srcStride, 16);\
269 OPNAME ## cavs_qpel8or16_v2_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
270 }\
271 \
272 static void OPNAME ## cavs_qpel8_v3_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
273 { \
274 OPNAME ## cavs_qpel8or16_v3_ ## MMX(dst , src , dstStride, srcStride, 8);\
275 }\
276 static void OPNAME ## cavs_qpel16_v3_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
277 { \
278 OPNAME ## cavs_qpel8or16_v3_ ## MMX(dst , src , dstStride, srcStride, 16);\
279 OPNAME ## cavs_qpel8or16_v3_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
280 }\
281 \
282 static void OPNAME ## cavs_qpel16_h_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
283 { \
284 OPNAME ## cavs_qpel8_h_ ## MMX(dst , src , dstStride, srcStride);\
285 OPNAME ## cavs_qpel8_h_ ## MMX(dst+8, src+8, dstStride, srcStride);\
286 src += 8*srcStride;\
287 dst += 8*dstStride;\
288 OPNAME ## cavs_qpel8_h_ ## MMX(dst , src , dstStride, srcStride);\
289 OPNAME ## cavs_qpel8_h_ ## MMX(dst+8, src+8, dstStride, srcStride);\
290 }\
291
292 #define CAVS_MC(OPNAME, SIZE, MMX) \
293 static void OPNAME ## cavs_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
294 {\
295 OPNAME ## cavs_qpel ## SIZE ## _h_ ## MMX(dst, src, stride, stride);\
296 }\
297 \
298 static void OPNAME ## cavs_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
299 {\
300 OPNAME ## cavs_qpel ## SIZE ## _v1_ ## MMX(dst, src, stride, stride);\
301 }\
302 \
303 static void OPNAME ## cavs_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
304 {\
305 OPNAME ## cavs_qpel ## SIZE ## _v2_ ## MMX(dst, src, stride, stride);\
306 }\
307 \
308 static void OPNAME ## cavs_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
309 {\
310 OPNAME ## cavs_qpel ## SIZE ## _v3_ ## MMX(dst, src, stride, stride);\
311 }\
312
313 #define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b " \n\t"
314 #define AVG_MMXEXT_OP(a, b, temp, size) \
315 "mov" #size " " #b ", " #temp " \n\t"\
316 "pavgb " #temp ", " #a " \n\t"\
317 "mov" #size " " #a ", " #b " \n\t"
318
319 #endif /* HAVE_MMXEXT_INLINE */
320
321 #if HAVE_MMX_EXTERNAL
322 ✗ static void put_cavs_qpel8_mc00_mmx(uint8_t *dst, const uint8_t *src,
323 ptrdiff_t stride)
324 {
325 ✗ ff_put_pixels8_mmx(dst, src, stride, 8);
326 ✗ }
327
328 ✗ static void avg_cavs_qpel8_mc00_mmxext(uint8_t *dst, const uint8_t *src,
329 ptrdiff_t stride)
330 {
331 ✗ ff_avg_pixels8_mmxext(dst, src, stride, 8);
332 ✗ }
333
334 ✗ static void put_cavs_qpel16_mc00_sse2(uint8_t *dst, const uint8_t *src,
335 ptrdiff_t stride)
336 {
337 ✗ ff_put_pixels16_sse2(dst, src, stride, 16);
338 ✗ }
339
340 ✗ static void avg_cavs_qpel16_mc00_sse2(uint8_t *dst, const uint8_t *src,
341 ptrdiff_t stride)
342 {
343 ✗ ff_avg_pixels16_sse2(dst, src, stride, 16);
344 ✗ }
345 #endif
346
347 2 static av_cold void cavsdsp_init_mmx(CAVSDSPContext *c)
348 {
349 #if HAVE_MMX_EXTERNAL
350 2 c->put_cavs_qpel_pixels_tab[1][0] = put_cavs_qpel8_mc00_mmx;
351 #endif /* HAVE_MMX_EXTERNAL */
352 2 }
353
354 #define DSPFUNC(PFX, IDX, NUM, EXT) \
355 c->PFX ## _cavs_qpel_pixels_tab[IDX][ 2] = PFX ## _cavs_qpel ## NUM ## _mc20_ ## EXT; \
356 c->PFX ## _cavs_qpel_pixels_tab[IDX][ 4] = PFX ## _cavs_qpel ## NUM ## _mc01_ ## EXT; \
357 c->PFX ## _cavs_qpel_pixels_tab[IDX][ 8] = PFX ## _cavs_qpel ## NUM ## _mc02_ ## EXT; \
358 c->PFX ## _cavs_qpel_pixels_tab[IDX][12] = PFX ## _cavs_qpel ## NUM ## _mc03_ ## EXT; \
359
360 #if HAVE_MMXEXT_INLINE
361 ✗ QPEL_CAVS(put_, PUT_OP, mmxext)
362 ✗ QPEL_CAVS(avg_, AVG_MMXEXT_OP, mmxext)
363
364 ✗ CAVS_MC(put_, 8, mmxext)
365 ✗ CAVS_MC(put_, 16, mmxext)
366 ✗ CAVS_MC(avg_, 8, mmxext)
367 ✗ CAVS_MC(avg_, 16, mmxext)
368 #endif /* HAVE_MMXEXT_INLINE */
369
370 6 av_cold void ff_cavsdsp_init_x86(CAVSDSPContext *c)
371 {
372 6 av_unused int cpu_flags = av_get_cpu_flags();
373
374
2/2
✓ Branch 0 taken 2 times.
✓ Branch 1 taken 4 times.
 6 if (X86_MMX(cpu_flags))
375 2 cavsdsp_init_mmx(c);
376
377 #if HAVE_MMXEXT_INLINE
378
2/2
✓ Branch 0 taken 2 times.
✓ Branch 1 taken 4 times.
 6 if (INLINE_MMXEXT(cpu_flags)) {
379 2 DSPFUNC(put, 0, 16, mmxext);
380 2 DSPFUNC(put, 1, 8, mmxext);
381 2 DSPFUNC(avg, 0, 16, mmxext);
382 2 DSPFUNC(avg, 1, 8, mmxext);
383 }
384 #endif
385 #if HAVE_MMX_EXTERNAL
386
2/2
✓ Branch 0 taken 2 times.
✓ Branch 1 taken 4 times.
 6 if (EXTERNAL_MMXEXT(cpu_flags)) {
387 2 c->avg_cavs_qpel_pixels_tab[1][0] = avg_cavs_qpel8_mc00_mmxext;
388 }
389 #endif
390 #if HAVE_SSE2_EXTERNAL
391
2/2
✓ Branch 0 taken 2 times.
✓ Branch 1 taken 4 times.
 6 if (EXTERNAL_SSE2(cpu_flags)) {
392 2 c->put_cavs_qpel_pixels_tab[0][0] = put_cavs_qpel16_mc00_sse2;
393 2 c->avg_cavs_qpel_pixels_tab[0][0] = avg_cavs_qpel16_mc00_sse2;
394
395 2 c->cavs_idct8_add = cavs_idct8_add_sse2;
396 2 c->idct_perm = FF_IDCT_PERM_TRANSPOSE;
397 }
398 #endif
399 6 }
400