2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
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
23 #include "libswscale/swscale.h"
24 #include "libswscale/swscale_internal.h"
25 #include "libavutil/intreadwrite.h"
26 #include "libavutil/x86_cpu.h"
27 #include "libavutil/cpu.h"
28 #include "libavutil/pixdesc.h"
30 DECLARE_ASM_CONST(8, uint64_t, bF8)= 0xF8F8F8F8F8F8F8F8LL;
31 DECLARE_ASM_CONST(8, uint64_t, bFC)= 0xFCFCFCFCFCFCFCFCLL;
32 DECLARE_ASM_CONST(8, uint64_t, w10)= 0x0010001000100010LL;
33 DECLARE_ASM_CONST(8, uint64_t, w02)= 0x0002000200020002LL;
34 DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
35 DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
36 DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
37 DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
39 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
41 0x0200020002000200LL,};
43 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
45 0x0004000400040004LL,};
47 DECLARE_ASM_CONST(8, uint64_t, b16Mask)= 0x001F001F001F001FLL;
48 DECLARE_ASM_CONST(8, uint64_t, g16Mask)= 0x07E007E007E007E0LL;
49 DECLARE_ASM_CONST(8, uint64_t, r16Mask)= 0xF800F800F800F800LL;
50 DECLARE_ASM_CONST(8, uint64_t, b15Mask)= 0x001F001F001F001FLL;
51 DECLARE_ASM_CONST(8, uint64_t, g15Mask)= 0x03E003E003E003E0LL;
52 DECLARE_ASM_CONST(8, uint64_t, r15Mask)= 0x7C007C007C007C00LL;
54 DECLARE_ALIGNED(8, const uint64_t, ff_M24A) = 0x00FF0000FF0000FFLL;
55 DECLARE_ALIGNED(8, const uint64_t, ff_M24B) = 0xFF0000FF0000FF00LL;
56 DECLARE_ALIGNED(8, const uint64_t, ff_M24C) = 0x0000FF0000FF0000LL;
59 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000000210041000DULL;
60 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000FFEEFFDC0038ULL;
61 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00000038FFD2FFF8ULL;
63 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000020E540830C8BULL;
64 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000ED0FDAC23831ULL;
65 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00003831D0E6F6EAULL;
66 #endif /* FAST_BGR2YV12 */
67 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL;
68 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
69 DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
71 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
72 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
73 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
74 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
75 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008010000080100ULL;
77 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
78 {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
79 {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
82 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040010000400100ULL;
87 #define COMPILE_TEMPLATE_MMX2 0
88 #define RENAME(a) a ## _MMX
89 #include "swscale_template.c"
95 #undef COMPILE_TEMPLATE_MMX2
96 #define COMPILE_TEMPLATE_MMX2 1
97 #define RENAME(a) a ## _MMX2
98 #include "swscale_template.c"
101 void updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufIndex,
102 int lastInLumBuf, int lastInChrBuf)
104 const int dstH= c->dstH;
105 const int flags= c->flags;
106 int16_t **lumPixBuf= c->lumPixBuf;
107 int16_t **chrUPixBuf= c->chrUPixBuf;
108 int16_t **alpPixBuf= c->alpPixBuf;
109 const int vLumBufSize= c->vLumBufSize;
110 const int vChrBufSize= c->vChrBufSize;
111 int16_t *vLumFilterPos= c->vLumFilterPos;
112 int16_t *vChrFilterPos= c->vChrFilterPos;
113 int16_t *vLumFilter= c->vLumFilter;
114 int16_t *vChrFilter= c->vChrFilter;
115 int32_t *lumMmxFilter= c->lumMmxFilter;
116 int32_t *chrMmxFilter= c->chrMmxFilter;
117 int32_t av_unused *alpMmxFilter= c->alpMmxFilter;
118 const int vLumFilterSize= c->vLumFilterSize;
119 const int vChrFilterSize= c->vChrFilterSize;
120 const int chrDstY= dstY>>c->chrDstVSubSample;
121 const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input
122 const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input
124 c->blueDither= ff_dither8[dstY&1];
125 if (c->dstFormat == PIX_FMT_RGB555 || c->dstFormat == PIX_FMT_BGR555)
126 c->greenDither= ff_dither8[dstY&1];
128 c->greenDither= ff_dither4[dstY&1];
129 c->redDither= ff_dither8[(dstY+1)&1];
130 if (dstY < dstH - 2) {
131 const int16_t **lumSrcPtr= (const int16_t **) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
132 const int16_t **chrUSrcPtr= (const int16_t **) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
133 const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
135 if (flags & SWS_ACCURATE_RND) {
136 int s= APCK_SIZE / 8;
137 for (i=0; i<vLumFilterSize; i+=2) {
138 *(const void**)&lumMmxFilter[s*i ]= lumSrcPtr[i ];
139 *(const void**)&lumMmxFilter[s*i+APCK_PTR2/4 ]= lumSrcPtr[i+(vLumFilterSize>1)];
140 lumMmxFilter[s*i+APCK_COEF/4 ]=
141 lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i ]
142 + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0);
143 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
144 *(const void**)&alpMmxFilter[s*i ]= alpSrcPtr[i ];
145 *(const void**)&alpMmxFilter[s*i+APCK_PTR2/4 ]= alpSrcPtr[i+(vLumFilterSize>1)];
146 alpMmxFilter[s*i+APCK_COEF/4 ]=
147 alpMmxFilter[s*i+APCK_COEF/4+1]= lumMmxFilter[s*i+APCK_COEF/4 ];
150 for (i=0; i<vChrFilterSize; i+=2) {
151 *(const void**)&chrMmxFilter[s*i ]= chrUSrcPtr[i ];
152 *(const void**)&chrMmxFilter[s*i+APCK_PTR2/4 ]= chrUSrcPtr[i+(vChrFilterSize>1)];
153 chrMmxFilter[s*i+APCK_COEF/4 ]=
154 chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i ]
155 + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0);
158 for (i=0; i<vLumFilterSize; i++) {
159 *(const void**)&lumMmxFilter[4*i+0]= lumSrcPtr[i];
162 ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001;
163 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
164 *(const void**)&alpMmxFilter[4*i+0]= alpSrcPtr[i];
166 alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2];
169 for (i=0; i<vChrFilterSize; i++) {
170 *(const void**)&chrMmxFilter[4*i+0]= chrUSrcPtr[i];
173 ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001;
179 static void yuv2yuvX_sse3(const int16_t *filter, int filterSize,
180 const int16_t **src, uint8_t *dest, int dstW,
181 const uint8_t *dither, int offset)
184 __asm__ volatile("movq (%0), %%xmm3\n\t"
185 "movdqa %%xmm3, %%xmm4\n\t"
186 "psrlq $24, %%xmm3\n\t"
187 "psllq $40, %%xmm4\n\t"
188 "por %%xmm4, %%xmm3\n\t"
192 __asm__ volatile("movq (%0), %%xmm3\n\t"
197 "pxor %%xmm0, %%xmm0\n\t"
198 "punpcklbw %%xmm0, %%xmm3\n\t"
199 "psraw $4, %%xmm3\n\t"
200 "movdqa %%xmm3, %%xmm4\n\t"
201 "movdqa %%xmm3, %%xmm7\n\t"
203 "mov %0, %%"REG_d" \n\t"\
204 "mov (%%"REG_d"), %%"REG_S" \n\t"\
205 ".p2align 4 \n\t" /* FIXME Unroll? */\
207 "movddup 8(%%"REG_d"), %%xmm0 \n\t" /* filterCoeff */\
208 "movdqa (%%"REG_S", %%"REG_c", 2), %%xmm2 \n\t" /* srcData */\
209 "movdqa 16(%%"REG_S", %%"REG_c", 2), %%xmm5 \n\t" /* srcData */\
210 "add $16, %%"REG_d" \n\t"\
211 "mov (%%"REG_d"), %%"REG_S" \n\t"\
212 "test %%"REG_S", %%"REG_S" \n\t"\
213 "pmulhw %%xmm0, %%xmm2 \n\t"\
214 "pmulhw %%xmm0, %%xmm5 \n\t"\
215 "paddw %%xmm2, %%xmm3 \n\t"\
216 "paddw %%xmm5, %%xmm4 \n\t"\
218 "psraw $3, %%xmm3 \n\t"\
219 "psraw $3, %%xmm4 \n\t"\
220 "packuswb %%xmm4, %%xmm3 \n\t"
221 "movntdq %%xmm3, (%1, %%"REG_c")\n\t"
222 "add $16, %%"REG_c" \n\t"\
223 "cmp %2, %%"REG_c" \n\t"\
224 "movdqa %%xmm7, %%xmm3\n\t"
225 "movdqa %%xmm7, %%xmm4\n\t"
226 "mov %0, %%"REG_d" \n\t"\
227 "mov (%%"REG_d"), %%"REG_S" \n\t"\
230 "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset)
231 : "%"REG_d, "%"REG_S, "%"REG_c
235 #define SCALE_FUNC(filter_n, from_bpc, to_bpc, opt) \
236 extern void ff_hscale ## from_bpc ## to ## to_bpc ## _ ## filter_n ## _ ## opt( \
237 SwsContext *c, int16_t *data, \
238 int dstW, const uint8_t *src, \
239 const int16_t *filter, \
240 const int16_t *filterPos, int filterSize);
242 #define SCALE_FUNCS(filter_n, opt) \
243 SCALE_FUNC(filter_n, 8, 15, opt); \
244 SCALE_FUNC(filter_n, 9, 15, opt); \
245 SCALE_FUNC(filter_n, 10, 15, opt); \
246 SCALE_FUNC(filter_n, 14, 15, opt); \
247 SCALE_FUNC(filter_n, 16, 15, opt); \
248 SCALE_FUNC(filter_n, 8, 19, opt); \
249 SCALE_FUNC(filter_n, 9, 19, opt); \
250 SCALE_FUNC(filter_n, 10, 19, opt); \
251 SCALE_FUNC(filter_n, 14, 19, opt); \
252 SCALE_FUNC(filter_n, 16, 19, opt)
254 #define SCALE_FUNCS_MMX(opt) \
255 SCALE_FUNCS(4, opt); \
256 SCALE_FUNCS(8, opt); \
259 #define SCALE_FUNCS_SSE(opt) \
260 SCALE_FUNCS(4, opt); \
261 SCALE_FUNCS(8, opt); \
262 SCALE_FUNCS(X4, opt); \
266 SCALE_FUNCS_MMX(mmx);
268 SCALE_FUNCS_SSE(sse2);
269 SCALE_FUNCS_SSE(ssse3);
270 SCALE_FUNCS_SSE(sse4);
272 #define VSCALEX_FUNC(size, opt) \
273 extern void ff_yuv2planeX_ ## size ## _ ## opt(const int16_t *filter, int filterSize, \
274 const int16_t **src, uint8_t *dest, int dstW, \
275 const uint8_t *dither, int offset)
276 #define VSCALEX_FUNCS(opt1, opt2) \
277 VSCALEX_FUNC(8, opt1); \
278 VSCALEX_FUNC(9, opt2); \
279 VSCALEX_FUNC(10, opt2)
282 VSCALEX_FUNCS(mmx, mmx2);
284 VSCALEX_FUNCS(sse2, sse2);
285 VSCALEX_FUNCS(sse4, sse4);
286 VSCALEX_FUNC(16, sse4);
287 VSCALEX_FUNCS(avx, avx);
289 void ff_sws_init_swScale_mmx(SwsContext *c)
291 int cpu_flags = av_get_cpu_flags();
293 if (cpu_flags & AV_CPU_FLAG_MMX)
294 sws_init_swScale_MMX(c);
296 if (cpu_flags & AV_CPU_FLAG_MMX2)
297 sws_init_swScale_MMX2(c);
298 if (cpu_flags & AV_CPU_FLAG_SSE3){
299 if(c->use_mmx_vfilter && !(c->flags & SWS_ACCURATE_RND))
300 c->yuv2planeX = yuv2yuvX_sse3;
305 #define ASSIGN_SCALE_FUNC2(hscalefn, filtersize, opt1, opt2) do { \
306 if (c->srcBpc == 8) { \
307 hscalefn = c->dstBpc <= 10 ? ff_hscale8to15_ ## filtersize ## _ ## opt2 : \
308 ff_hscale8to19_ ## filtersize ## _ ## opt1; \
309 } else if (c->srcBpc == 9) { \
310 hscalefn = c->dstBpc <= 10 ? ff_hscale9to15_ ## filtersize ## _ ## opt2 : \
311 ff_hscale9to19_ ## filtersize ## _ ## opt1; \
312 } else if (c->srcBpc == 10) { \
313 hscalefn = c->dstBpc <= 10 ? ff_hscale10to15_ ## filtersize ## _ ## opt2 : \
314 ff_hscale10to19_ ## filtersize ## _ ## opt1; \
315 } else if (c->srcBpc == 14 || ((c->srcFormat==PIX_FMT_PAL8||isAnyRGB(c->srcFormat)) && av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1<15)) { \
316 hscalefn = c->dstBpc <= 10 ? ff_hscale14to15_ ## filtersize ## _ ## opt2 : \
317 ff_hscale14to19_ ## filtersize ## _ ## opt1; \
318 } else { /* c->srcBpc == 16 */ \
319 hscalefn = c->dstBpc <= 10 ? ff_hscale16to15_ ## filtersize ## _ ## opt2 : \
320 ff_hscale16to19_ ## filtersize ## _ ## opt1; \
323 #define ASSIGN_MMX_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
324 switch (filtersize) { \
325 case 4: ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \
326 case 8: ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
327 default: ASSIGN_SCALE_FUNC2(hscalefn, X, opt1, opt2); break; \
329 #define ASSIGN_VSCALEX_FUNC(vscalefn, opt1, opt2, opt2chk, do_16_case) \
331 case 16: /*do_16_case;*/ break; \
332 case 10: if (!isBE(c->dstFormat) && opt2chk) /*vscalefn = ff_yuv2planeX_10_ ## opt2;*/ break; \
333 case 9: if (!isBE(c->dstFormat) && opt2chk) /*vscalefn = ff_yuv2planeX_9_ ## opt2;*/ break; \
334 default: /*vscalefn = ff_yuv2planeX_8_ ## opt1;*/ break; \
337 if (cpu_flags & AV_CPU_FLAG_MMX) {
338 ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx);
339 ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx);
340 ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMX2,);
343 #define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
344 switch (filtersize) { \
345 case 4: ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \
346 case 8: ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
347 default: if (filtersize & 4) ASSIGN_SCALE_FUNC2(hscalefn, X4, opt1, opt2); \
348 else ASSIGN_SCALE_FUNC2(hscalefn, X8, opt1, opt2); \
351 if (cpu_flags & AV_CPU_FLAG_SSE2) {
352 ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse2, sse2);
353 ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse2, sse2);
354 ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse2, sse2, 1,);
356 if (cpu_flags & AV_CPU_FLAG_SSSE3) {
357 ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, ssse3, ssse3);
358 ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, ssse3, ssse3);
360 if (cpu_flags & AV_CPU_FLAG_SSE4) {
361 /* Xto15 don't need special sse4 functions */
362 ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse4, ssse3);
363 ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse4, ssse3);
364 ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse4, sse4, 1,
365 if (!isBE(c->dstFormat)) c->yuv2planeX = ff_yuv2planeX_16_sse4);
368 if (cpu_flags & AV_CPU_FLAG_AVX) {
369 ASSIGN_VSCALEX_FUNC(c->yuv2planeX, avx, avx, 1,);