FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
swscale_template.c
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * This file is part of FFmpeg.
5  *
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.
10  *
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.
15  *
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
19  */
20 
21 #include <stdint.h>
22 
23 #include "libavutil/x86/asm.h"
24 #include "libswscale/swscale_internal.h"
25 
26 #undef REAL_MOVNTQ
27 #undef MOVNTQ
28 #undef MOVNTQ2
29 #undef PREFETCH
30 
31 
32 #if COMPILE_TEMPLATE_MMXEXT
33 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
34 #define MOVNTQ2 "movntq "
35 #else
36 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
37 #define MOVNTQ2 "movq "
38 #endif
39 #define MOVNTQ(a,b) REAL_MOVNTQ(a,b)
40 
41 #if !COMPILE_TEMPLATE_MMXEXT
42 static av_always_inline void
43 dither_8to16(const uint8_t *srcDither, int rot)
44 {
45  if (rot) {
46  __asm__ volatile("pxor %%mm0, %%mm0\n\t"
47  "movq (%0), %%mm3\n\t"
48  "movq %%mm3, %%mm4\n\t"
49  "psrlq $24, %%mm3\n\t"
50  "psllq $40, %%mm4\n\t"
51  "por %%mm4, %%mm3\n\t"
52  "movq %%mm3, %%mm4\n\t"
53  "punpcklbw %%mm0, %%mm3\n\t"
54  "punpckhbw %%mm0, %%mm4\n\t"
55  :: "r"(srcDither)
56  );
57  } else {
58  __asm__ volatile("pxor %%mm0, %%mm0\n\t"
59  "movq (%0), %%mm3\n\t"
60  "movq %%mm3, %%mm4\n\t"
61  "punpcklbw %%mm0, %%mm3\n\t"
62  "punpckhbw %%mm0, %%mm4\n\t"
63  :: "r"(srcDither)
64  );
65  }
66 }
67 #endif
68 
69 static void RENAME(yuv2yuvX)(const int16_t *filter, int filterSize,
70  const int16_t **src, uint8_t *dest, int dstW,
71  const uint8_t *dither, int offset)
72 {
73  dither_8to16(dither, offset);
74  filterSize--;
75  __asm__ volatile(
76  "movd %0, %%mm1\n\t"
77  "punpcklwd %%mm1, %%mm1\n\t"
78  "punpckldq %%mm1, %%mm1\n\t"
79  "psllw $3, %%mm1\n\t"
80  "paddw %%mm1, %%mm3\n\t"
81  "paddw %%mm1, %%mm4\n\t"
82  "psraw $4, %%mm3\n\t"
83  "psraw $4, %%mm4\n\t"
84  ::"m"(filterSize)
85  );
86 
87  __asm__ volatile(\
88  "movq %%mm3, %%mm6\n\t"
89  "movq %%mm4, %%mm7\n\t"
90  "movl %3, %%ecx\n\t"
91  "mov %0, %%"REG_d" \n\t"\
92  "mov (%%"REG_d"), %%"REG_S" \n\t"\
93  ".p2align 4 \n\t" /* FIXME Unroll? */\
94  "1: \n\t"\
95  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
96  "movq (%%"REG_S", %%"REG_c", 2), %%mm2 \n\t" /* srcData */\
97  "movq 8(%%"REG_S", %%"REG_c", 2), %%mm5 \n\t" /* srcData */\
98  "add $16, %%"REG_d" \n\t"\
99  "mov (%%"REG_d"), %%"REG_S" \n\t"\
100  "test %%"REG_S", %%"REG_S" \n\t"\
101  "pmulhw %%mm0, %%mm2 \n\t"\
102  "pmulhw %%mm0, %%mm5 \n\t"\
103  "paddw %%mm2, %%mm3 \n\t"\
104  "paddw %%mm5, %%mm4 \n\t"\
105  " jnz 1b \n\t"\
106  "psraw $3, %%mm3 \n\t"\
107  "psraw $3, %%mm4 \n\t"\
108  "packuswb %%mm4, %%mm3 \n\t"
109  MOVNTQ2 " %%mm3, (%1, %%"REG_c")\n\t"
110  "add $8, %%"REG_c" \n\t"\
111  "cmp %2, %%"REG_c" \n\t"\
112  "movq %%mm6, %%mm3\n\t"
113  "movq %%mm7, %%mm4\n\t"
114  "mov %0, %%"REG_d" \n\t"\
115  "mov (%%"REG_d"), %%"REG_S" \n\t"\
116  "jb 1b \n\t"\
117  :: "g" (filter),
118  "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset)
119  : "%"REG_d, "%"REG_S, "%"REG_c
120  );
121 }
122 
123 #define YSCALEYUV2PACKEDX_UV \
124  __asm__ volatile(\
125  "xor %%"REG_a", %%"REG_a" \n\t"\
126  ".p2align 4 \n\t"\
127  "nop \n\t"\
128  "1: \n\t"\
129  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
130  "mov (%%"REG_d"), %%"REG_S" \n\t"\
131  "movq "VROUNDER_OFFSET"(%0), %%mm3 \n\t"\
132  "movq %%mm3, %%mm4 \n\t"\
133  ".p2align 4 \n\t"\
134  "2: \n\t"\
135  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
136  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\
137  "add %6, %%"REG_S" \n\t" \
138  "movq (%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\
139  "add $16, %%"REG_d" \n\t"\
140  "mov (%%"REG_d"), %%"REG_S" \n\t"\
141  "pmulhw %%mm0, %%mm2 \n\t"\
142  "pmulhw %%mm0, %%mm5 \n\t"\
143  "paddw %%mm2, %%mm3 \n\t"\
144  "paddw %%mm5, %%mm4 \n\t"\
145  "test %%"REG_S", %%"REG_S" \n\t"\
146  " jnz 2b \n\t"\
147 
148 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
149  "lea "offset"(%0), %%"REG_d" \n\t"\
150  "mov (%%"REG_d"), %%"REG_S" \n\t"\
151  "movq "VROUNDER_OFFSET"(%0), "#dst1" \n\t"\
152  "movq "#dst1", "#dst2" \n\t"\
153  ".p2align 4 \n\t"\
154  "2: \n\t"\
155  "movq 8(%%"REG_d"), "#coeff" \n\t" /* filterCoeff */\
156  "movq (%%"REG_S", %%"REG_a", 2), "#src1" \n\t" /* Y1srcData */\
157  "movq 8(%%"REG_S", %%"REG_a", 2), "#src2" \n\t" /* Y2srcData */\
158  "add $16, %%"REG_d" \n\t"\
159  "mov (%%"REG_d"), %%"REG_S" \n\t"\
160  "pmulhw "#coeff", "#src1" \n\t"\
161  "pmulhw "#coeff", "#src2" \n\t"\
162  "paddw "#src1", "#dst1" \n\t"\
163  "paddw "#src2", "#dst2" \n\t"\
164  "test %%"REG_S", %%"REG_S" \n\t"\
165  " jnz 2b \n\t"\
166 
167 #define YSCALEYUV2PACKEDX \
168  YSCALEYUV2PACKEDX_UV \
169  YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
170 
171 #define YSCALEYUV2PACKEDX_END \
172  :: "r" (&c->redDither), \
173  "m" (dummy), "m" (dummy), "m" (dummy),\
174  "r" (dest), "m" (dstW_reg), "m"(uv_off) \
175  NAMED_CONSTRAINTS_ADD(bF8,bFC) \
176  : "%"REG_a, "%"REG_d, "%"REG_S \
177  );
178 
179 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
180  __asm__ volatile(\
181  "xor %%"REG_a", %%"REG_a" \n\t"\
182  ".p2align 4 \n\t"\
183  "nop \n\t"\
184  "1: \n\t"\
185  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
186  "mov (%%"REG_d"), %%"REG_S" \n\t"\
187  "pxor %%mm4, %%mm4 \n\t"\
188  "pxor %%mm5, %%mm5 \n\t"\
189  "pxor %%mm6, %%mm6 \n\t"\
190  "pxor %%mm7, %%mm7 \n\t"\
191  ".p2align 4 \n\t"\
192  "2: \n\t"\
193  "movq (%%"REG_S", %%"REG_a"), %%mm0 \n\t" /* UsrcData */\
194  "add %6, %%"REG_S" \n\t" \
195  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* VsrcData */\
196  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
197  "movq (%%"REG_S", %%"REG_a"), %%mm1 \n\t" /* UsrcData */\
198  "movq %%mm0, %%mm3 \n\t"\
199  "punpcklwd %%mm1, %%mm0 \n\t"\
200  "punpckhwd %%mm1, %%mm3 \n\t"\
201  "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1 \n\t" /* filterCoeff */\
202  "pmaddwd %%mm1, %%mm0 \n\t"\
203  "pmaddwd %%mm1, %%mm3 \n\t"\
204  "paddd %%mm0, %%mm4 \n\t"\
205  "paddd %%mm3, %%mm5 \n\t"\
206  "add %6, %%"REG_S" \n\t" \
207  "movq (%%"REG_S", %%"REG_a"), %%mm3 \n\t" /* VsrcData */\
208  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
209  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
210  "test %%"REG_S", %%"REG_S" \n\t"\
211  "movq %%mm2, %%mm0 \n\t"\
212  "punpcklwd %%mm3, %%mm2 \n\t"\
213  "punpckhwd %%mm3, %%mm0 \n\t"\
214  "pmaddwd %%mm1, %%mm2 \n\t"\
215  "pmaddwd %%mm1, %%mm0 \n\t"\
216  "paddd %%mm2, %%mm6 \n\t"\
217  "paddd %%mm0, %%mm7 \n\t"\
218  " jnz 2b \n\t"\
219  "psrad $16, %%mm4 \n\t"\
220  "psrad $16, %%mm5 \n\t"\
221  "psrad $16, %%mm6 \n\t"\
222  "psrad $16, %%mm7 \n\t"\
223  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
224  "packssdw %%mm5, %%mm4 \n\t"\
225  "packssdw %%mm7, %%mm6 \n\t"\
226  "paddw %%mm0, %%mm4 \n\t"\
227  "paddw %%mm0, %%mm6 \n\t"\
228  "movq %%mm4, "U_TEMP"(%0) \n\t"\
229  "movq %%mm6, "V_TEMP"(%0) \n\t"\
230 
231 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
232  "lea "offset"(%0), %%"REG_d" \n\t"\
233  "mov (%%"REG_d"), %%"REG_S" \n\t"\
234  "pxor %%mm1, %%mm1 \n\t"\
235  "pxor %%mm5, %%mm5 \n\t"\
236  "pxor %%mm7, %%mm7 \n\t"\
237  "pxor %%mm6, %%mm6 \n\t"\
238  ".p2align 4 \n\t"\
239  "2: \n\t"\
240  "movq (%%"REG_S", %%"REG_a", 2), %%mm0 \n\t" /* Y1srcData */\
241  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y2srcData */\
242  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
243  "movq (%%"REG_S", %%"REG_a", 2), %%mm4 \n\t" /* Y1srcData */\
244  "movq %%mm0, %%mm3 \n\t"\
245  "punpcklwd %%mm4, %%mm0 \n\t"\
246  "punpckhwd %%mm4, %%mm3 \n\t"\
247  "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4 \n\t" /* filterCoeff */\
248  "pmaddwd %%mm4, %%mm0 \n\t"\
249  "pmaddwd %%mm4, %%mm3 \n\t"\
250  "paddd %%mm0, %%mm1 \n\t"\
251  "paddd %%mm3, %%mm5 \n\t"\
252  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3 \n\t" /* Y2srcData */\
253  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
254  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
255  "test %%"REG_S", %%"REG_S" \n\t"\
256  "movq %%mm2, %%mm0 \n\t"\
257  "punpcklwd %%mm3, %%mm2 \n\t"\
258  "punpckhwd %%mm3, %%mm0 \n\t"\
259  "pmaddwd %%mm4, %%mm2 \n\t"\
260  "pmaddwd %%mm4, %%mm0 \n\t"\
261  "paddd %%mm2, %%mm7 \n\t"\
262  "paddd %%mm0, %%mm6 \n\t"\
263  " jnz 2b \n\t"\
264  "psrad $16, %%mm1 \n\t"\
265  "psrad $16, %%mm5 \n\t"\
266  "psrad $16, %%mm7 \n\t"\
267  "psrad $16, %%mm6 \n\t"\
268  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
269  "packssdw %%mm5, %%mm1 \n\t"\
270  "packssdw %%mm6, %%mm7 \n\t"\
271  "paddw %%mm0, %%mm1 \n\t"\
272  "paddw %%mm0, %%mm7 \n\t"\
273  "movq "U_TEMP"(%0), %%mm3 \n\t"\
274  "movq "V_TEMP"(%0), %%mm4 \n\t"\
275 
276 #define YSCALEYUV2PACKEDX_ACCURATE \
277  YSCALEYUV2PACKEDX_ACCURATE_UV \
278  YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
279 
280 #define YSCALEYUV2RGBX \
281  "psubw "U_OFFSET"(%0), %%mm3 \n\t" /* (U-128)8*/\
282  "psubw "V_OFFSET"(%0), %%mm4 \n\t" /* (V-128)8*/\
283  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
284  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
285  "pmulhw "UG_COEFF"(%0), %%mm3 \n\t"\
286  "pmulhw "VG_COEFF"(%0), %%mm4 \n\t"\
287  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
288  "pmulhw "UB_COEFF"(%0), %%mm2 \n\t"\
289  "pmulhw "VR_COEFF"(%0), %%mm5 \n\t"\
290  "psubw "Y_OFFSET"(%0), %%mm1 \n\t" /* 8(Y-16)*/\
291  "psubw "Y_OFFSET"(%0), %%mm7 \n\t" /* 8(Y-16)*/\
292  "pmulhw "Y_COEFF"(%0), %%mm1 \n\t"\
293  "pmulhw "Y_COEFF"(%0), %%mm7 \n\t"\
294  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
295  "paddw %%mm3, %%mm4 \n\t"\
296  "movq %%mm2, %%mm0 \n\t"\
297  "movq %%mm5, %%mm6 \n\t"\
298  "movq %%mm4, %%mm3 \n\t"\
299  "punpcklwd %%mm2, %%mm2 \n\t"\
300  "punpcklwd %%mm5, %%mm5 \n\t"\
301  "punpcklwd %%mm4, %%mm4 \n\t"\
302  "paddw %%mm1, %%mm2 \n\t"\
303  "paddw %%mm1, %%mm5 \n\t"\
304  "paddw %%mm1, %%mm4 \n\t"\
305  "punpckhwd %%mm0, %%mm0 \n\t"\
306  "punpckhwd %%mm6, %%mm6 \n\t"\
307  "punpckhwd %%mm3, %%mm3 \n\t"\
308  "paddw %%mm7, %%mm0 \n\t"\
309  "paddw %%mm7, %%mm6 \n\t"\
310  "paddw %%mm7, %%mm3 \n\t"\
311  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
312  "packuswb %%mm0, %%mm2 \n\t"\
313  "packuswb %%mm6, %%mm5 \n\t"\
314  "packuswb %%mm3, %%mm4 \n\t"\
315 
316 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
317  "movq "#b", "#q2" \n\t" /* B */\
318  "movq "#r", "#t" \n\t" /* R */\
319  "punpcklbw "#g", "#b" \n\t" /* GBGBGBGB 0 */\
320  "punpcklbw "#a", "#r" \n\t" /* ARARARAR 0 */\
321  "punpckhbw "#g", "#q2" \n\t" /* GBGBGBGB 2 */\
322  "punpckhbw "#a", "#t" \n\t" /* ARARARAR 2 */\
323  "movq "#b", "#q0" \n\t" /* GBGBGBGB 0 */\
324  "movq "#q2", "#q3" \n\t" /* GBGBGBGB 2 */\
325  "punpcklwd "#r", "#q0" \n\t" /* ARGBARGB 0 */\
326  "punpckhwd "#r", "#b" \n\t" /* ARGBARGB 1 */\
327  "punpcklwd "#t", "#q2" \n\t" /* ARGBARGB 2 */\
328  "punpckhwd "#t", "#q3" \n\t" /* ARGBARGB 3 */\
329 \
330  MOVNTQ( q0, (dst, index, 4))\
331  MOVNTQ( b, 8(dst, index, 4))\
332  MOVNTQ( q2, 16(dst, index, 4))\
333  MOVNTQ( q3, 24(dst, index, 4))\
334 \
335  "add $8, "#index" \n\t"\
336  "cmp "dstw", "#index" \n\t"\
337  " jb 1b \n\t"
338 #define WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t)
339 
340 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
341  const int16_t **lumSrc, int lumFilterSize,
342  const int16_t *chrFilter, const int16_t **chrUSrc,
343  const int16_t **chrVSrc,
344  int chrFilterSize, const int16_t **alpSrc,
345  uint8_t *dest, int dstW, int dstY)
346 {
347  x86_reg dummy=0;
348  x86_reg dstW_reg = dstW;
349  x86_reg uv_off = c->uv_offx2;
350 
351  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
352  YSCALEYUV2PACKEDX_ACCURATE
353  YSCALEYUV2RGBX
354  "movq %%mm2, "U_TEMP"(%0) \n\t"
355  "movq %%mm4, "V_TEMP"(%0) \n\t"
356  "movq %%mm5, "Y_TEMP"(%0) \n\t"
357  YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)
358  "movq "Y_TEMP"(%0), %%mm5 \n\t"
359  "psraw $3, %%mm1 \n\t"
360  "psraw $3, %%mm7 \n\t"
361  "packuswb %%mm7, %%mm1 \n\t"
362  WRITEBGR32(%4, "%5", %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
363  YSCALEYUV2PACKEDX_END
364  } else {
365  YSCALEYUV2PACKEDX_ACCURATE
366  YSCALEYUV2RGBX
367  "pcmpeqd %%mm7, %%mm7 \n\t"
368  WRITEBGR32(%4, "%5", %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
369  YSCALEYUV2PACKEDX_END
370  }
371 }
372 
373 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
374  const int16_t **lumSrc, int lumFilterSize,
375  const int16_t *chrFilter, const int16_t **chrUSrc,
376  const int16_t **chrVSrc,
377  int chrFilterSize, const int16_t **alpSrc,
378  uint8_t *dest, int dstW, int dstY)
379 {
380  x86_reg dummy=0;
381  x86_reg dstW_reg = dstW;
382  x86_reg uv_off = c->uv_offx2;
383 
384  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
385  YSCALEYUV2PACKEDX
386  YSCALEYUV2RGBX
387  YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
388  "psraw $3, %%mm1 \n\t"
389  "psraw $3, %%mm7 \n\t"
390  "packuswb %%mm7, %%mm1 \n\t"
391  WRITEBGR32(%4, "%5", %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
392  YSCALEYUV2PACKEDX_END
393  } else {
394  YSCALEYUV2PACKEDX
395  YSCALEYUV2RGBX
396  "pcmpeqd %%mm7, %%mm7 \n\t"
397  WRITEBGR32(%4, "%5", %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
398  YSCALEYUV2PACKEDX_END
399  }
400 }
401 
402 #define REAL_WRITERGB16(dst, dstw, index) \
403  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
404  "pand "MANGLE(bFC)", %%mm4 \n\t" /* G */\
405  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
406  "psrlq $3, %%mm2 \n\t"\
407 \
408  "movq %%mm2, %%mm1 \n\t"\
409  "movq %%mm4, %%mm3 \n\t"\
410 \
411  "punpcklbw %%mm7, %%mm3 \n\t"\
412  "punpcklbw %%mm5, %%mm2 \n\t"\
413  "punpckhbw %%mm7, %%mm4 \n\t"\
414  "punpckhbw %%mm5, %%mm1 \n\t"\
415 \
416  "psllq $3, %%mm3 \n\t"\
417  "psllq $3, %%mm4 \n\t"\
418 \
419  "por %%mm3, %%mm2 \n\t"\
420  "por %%mm4, %%mm1 \n\t"\
421 \
422  MOVNTQ(%%mm2, (dst, index, 2))\
423  MOVNTQ(%%mm1, 8(dst, index, 2))\
424 \
425  "add $8, "#index" \n\t"\
426  "cmp "dstw", "#index" \n\t"\
427  " jb 1b \n\t"
428 #define WRITERGB16(dst, dstw, index) REAL_WRITERGB16(dst, dstw, index)
429 
430 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
431  const int16_t **lumSrc, int lumFilterSize,
432  const int16_t *chrFilter, const int16_t **chrUSrc,
433  const int16_t **chrVSrc,
434  int chrFilterSize, const int16_t **alpSrc,
435  uint8_t *dest, int dstW, int dstY)
436 {
437  x86_reg dummy=0;
438  x86_reg dstW_reg = dstW;
439  x86_reg uv_off = c->uv_offx2;
440 
441  YSCALEYUV2PACKEDX_ACCURATE
442  YSCALEYUV2RGBX
443  "pxor %%mm7, %%mm7 \n\t"
444  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
445 #ifdef DITHER1XBPP
446  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
447  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
448  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
449 #endif
450  WRITERGB16(%4, "%5", %%REGa)
451  YSCALEYUV2PACKEDX_END
452 }
453 
454 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
455  const int16_t **lumSrc, int lumFilterSize,
456  const int16_t *chrFilter, const int16_t **chrUSrc,
457  const int16_t **chrVSrc,
458  int chrFilterSize, const int16_t **alpSrc,
459  uint8_t *dest, int dstW, int dstY)
460 {
461  x86_reg dummy=0;
462  x86_reg dstW_reg = dstW;
463  x86_reg uv_off = c->uv_offx2;
464 
465  YSCALEYUV2PACKEDX
466  YSCALEYUV2RGBX
467  "pxor %%mm7, %%mm7 \n\t"
468  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
469 #ifdef DITHER1XBPP
470  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
471  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
472  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
473 #endif
474  WRITERGB16(%4, "%5", %%REGa)
475  YSCALEYUV2PACKEDX_END
476 }
477 
478 #define REAL_WRITERGB15(dst, dstw, index) \
479  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
480  "pand "MANGLE(bF8)", %%mm4 \n\t" /* G */\
481  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
482  "psrlq $3, %%mm2 \n\t"\
483  "psrlq $1, %%mm5 \n\t"\
484 \
485  "movq %%mm2, %%mm1 \n\t"\
486  "movq %%mm4, %%mm3 \n\t"\
487 \
488  "punpcklbw %%mm7, %%mm3 \n\t"\
489  "punpcklbw %%mm5, %%mm2 \n\t"\
490  "punpckhbw %%mm7, %%mm4 \n\t"\
491  "punpckhbw %%mm5, %%mm1 \n\t"\
492 \
493  "psllq $2, %%mm3 \n\t"\
494  "psllq $2, %%mm4 \n\t"\
495 \
496  "por %%mm3, %%mm2 \n\t"\
497  "por %%mm4, %%mm1 \n\t"\
498 \
499  MOVNTQ(%%mm2, (dst, index, 2))\
500  MOVNTQ(%%mm1, 8(dst, index, 2))\
501 \
502  "add $8, "#index" \n\t"\
503  "cmp "dstw", "#index" \n\t"\
504  " jb 1b \n\t"
505 #define WRITERGB15(dst, dstw, index) REAL_WRITERGB15(dst, dstw, index)
506 
507 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
508  const int16_t **lumSrc, int lumFilterSize,
509  const int16_t *chrFilter, const int16_t **chrUSrc,
510  const int16_t **chrVSrc,
511  int chrFilterSize, const int16_t **alpSrc,
512  uint8_t *dest, int dstW, int dstY)
513 {
514  x86_reg dummy=0;
515  x86_reg dstW_reg = dstW;
516  x86_reg uv_off = c->uv_offx2;
517 
518  YSCALEYUV2PACKEDX_ACCURATE
519  YSCALEYUV2RGBX
520  "pxor %%mm7, %%mm7 \n\t"
521  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
522 #ifdef DITHER1XBPP
523  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
524  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
525  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
526 #endif
527  WRITERGB15(%4, "%5", %%REGa)
528  YSCALEYUV2PACKEDX_END
529 }
530 
531 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
532  const int16_t **lumSrc, int lumFilterSize,
533  const int16_t *chrFilter, const int16_t **chrUSrc,
534  const int16_t **chrVSrc,
535  int chrFilterSize, const int16_t **alpSrc,
536  uint8_t *dest, int dstW, int dstY)
537 {
538  x86_reg dummy=0;
539  x86_reg dstW_reg = dstW;
540  x86_reg uv_off = c->uv_offx2;
541 
542  YSCALEYUV2PACKEDX
543  YSCALEYUV2RGBX
544  "pxor %%mm7, %%mm7 \n\t"
545  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
546 #ifdef DITHER1XBPP
547  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
548  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
549  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
550 #endif
551  WRITERGB15(%4, "%5", %%REGa)
552  YSCALEYUV2PACKEDX_END
553 }
554 
555 #define WRITEBGR24MMX(dst, dstw, index) \
556  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
557  "movq %%mm2, %%mm1 \n\t" /* B */\
558  "movq %%mm5, %%mm6 \n\t" /* R */\
559  "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\
560  "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\
561  "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\
562  "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\
563  "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\
564  "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\
565  "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\
566  "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\
567  "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\
568  "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\
569 \
570  "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\
571  "movq %%mm2, %%mm6 \n\t" /* 0RGB0RGB 1 */\
572  "movq %%mm1, %%mm5 \n\t" /* 0RGB0RGB 2 */\
573  "movq %%mm3, %%mm7 \n\t" /* 0RGB0RGB 3 */\
574 \
575  "psllq $40, %%mm0 \n\t" /* RGB00000 0 */\
576  "psllq $40, %%mm2 \n\t" /* RGB00000 1 */\
577  "psllq $40, %%mm1 \n\t" /* RGB00000 2 */\
578  "psllq $40, %%mm3 \n\t" /* RGB00000 3 */\
579 \
580  "punpckhdq %%mm4, %%mm0 \n\t" /* 0RGBRGB0 0 */\
581  "punpckhdq %%mm6, %%mm2 \n\t" /* 0RGBRGB0 1 */\
582  "punpckhdq %%mm5, %%mm1 \n\t" /* 0RGBRGB0 2 */\
583  "punpckhdq %%mm7, %%mm3 \n\t" /* 0RGBRGB0 3 */\
584 \
585  "psrlq $8, %%mm0 \n\t" /* 00RGBRGB 0 */\
586  "movq %%mm2, %%mm6 \n\t" /* 0RGBRGB0 1 */\
587  "psllq $40, %%mm2 \n\t" /* GB000000 1 */\
588  "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\
589  MOVNTQ(%%mm0, (dst))\
590 \
591  "psrlq $24, %%mm6 \n\t" /* 0000RGBR 1 */\
592  "movq %%mm1, %%mm5 \n\t" /* 0RGBRGB0 2 */\
593  "psllq $24, %%mm1 \n\t" /* BRGB0000 2 */\
594  "por %%mm1, %%mm6 \n\t" /* BRGBRGBR 1 */\
595  MOVNTQ(%%mm6, 8(dst))\
596 \
597  "psrlq $40, %%mm5 \n\t" /* 000000RG 2 */\
598  "psllq $8, %%mm3 \n\t" /* RGBRGB00 3 */\
599  "por %%mm3, %%mm5 \n\t" /* RGBRGBRG 2 */\
600  MOVNTQ(%%mm5, 16(dst))\
601 \
602  "add $24, "#dst" \n\t"\
603 \
604  "add $8, "#index" \n\t"\
605  "cmp "dstw", "#index" \n\t"\
606  " jb 1b \n\t"
607 
608 #define WRITEBGR24MMXEXT(dst, dstw, index) \
609  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
610  "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
611  "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
612  "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2 B1 B0 B1 B0 */\
613  "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2 G1 G0 G1 G0 */\
614  "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0 R1 R0 R1 R0 */\
615 \
616  "pand %%mm0, %%mm1 \n\t" /* B2 B1 B0 */\
617  "pand %%mm0, %%mm3 \n\t" /* G2 G1 G0 */\
618  "pand %%mm7, %%mm6 \n\t" /* R1 R0 */\
619 \
620  "psllq $8, %%mm3 \n\t" /* G2 G1 G0 */\
621  "por %%mm1, %%mm6 \n\t"\
622  "por %%mm3, %%mm6 \n\t"\
623  MOVNTQ(%%mm6, (dst))\
624 \
625  "psrlq $8, %%mm4 \n\t" /* 00 G7 G6 G5 G4 G3 G2 G1 */\
626  "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4 B3 B2 B3 B2 */\
627  "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3 G4 G3 G4 G3 */\
628  "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4 R3 R2 R3 R2 */\
629 \
630  "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5 B4 B3 */\
631  "pand %%mm7, %%mm3 \n\t" /* G4 G3 */\
632  "pand %%mm0, %%mm6 \n\t" /* R4 R3 R2 */\
633 \
634  "por %%mm1, %%mm3 \n\t" /* B5 G4 B4 G3 B3 */\
635  "por %%mm3, %%mm6 \n\t"\
636  MOVNTQ(%%mm6, 8(dst))\
637 \
638  "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6 B7 B6 B6 B7 */\
639  "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7 G6 G5 G6 G5 */\
640  "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6 R5 R4 R5 R4 */\
641 \
642  "pand %%mm7, %%mm1 \n\t" /* B7 B6 */\
643  "pand %%mm0, %%mm3 \n\t" /* G7 G6 G5 */\
644  "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7 R6 R5 */\
645 \
646  "por %%mm1, %%mm3 \n\t"\
647  "por %%mm3, %%mm6 \n\t"\
648  MOVNTQ(%%mm6, 16(dst))\
649 \
650  "add $24, "#dst" \n\t"\
651 \
652  "add $8, "#index" \n\t"\
653  "cmp "dstw", "#index" \n\t"\
654  " jb 1b \n\t"
655 
656 #if COMPILE_TEMPLATE_MMXEXT
657 #undef WRITEBGR24
658 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMXEXT(dst, dstw, index)
659 #else
660 #undef WRITEBGR24
661 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX(dst, dstw, index)
662 #endif
663 
664 #if HAVE_6REGS
665 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
666  const int16_t **lumSrc, int lumFilterSize,
667  const int16_t *chrFilter, const int16_t **chrUSrc,
668  const int16_t **chrVSrc,
669  int chrFilterSize, const int16_t **alpSrc,
670  uint8_t *dest, int dstW, int dstY)
671 {
672  x86_reg dummy=0;
673  x86_reg dstW_reg = dstW;
674  x86_reg uv_off = c->uv_offx2;
675 
676  YSCALEYUV2PACKEDX_ACCURATE
677  YSCALEYUV2RGBX
678  "pxor %%mm7, %%mm7 \n\t"
679  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
680  "add %4, %%"REG_c" \n\t"
681  WRITEBGR24(%%REGc, "%5", %%REGa)
682  :: "r" (&c->redDither),
683  "m" (dummy), "m" (dummy), "m" (dummy),
684  "r" (dest), "m" (dstW_reg), "m"(uv_off)
685  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
686  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
687  );
688 }
689 
690 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
691  const int16_t **lumSrc, int lumFilterSize,
692  const int16_t *chrFilter, const int16_t **chrUSrc,
693  const int16_t **chrVSrc,
694  int chrFilterSize, const int16_t **alpSrc,
695  uint8_t *dest, int dstW, int dstY)
696 {
697  x86_reg dummy=0;
698  x86_reg dstW_reg = dstW;
699  x86_reg uv_off = c->uv_offx2;
700 
701  YSCALEYUV2PACKEDX
702  YSCALEYUV2RGBX
703  "pxor %%mm7, %%mm7 \n\t"
704  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c" \n\t" //FIXME optimize
705  "add %4, %%"REG_c" \n\t"
706  WRITEBGR24(%%REGc, "%5", %%REGa)
707  :: "r" (&c->redDither),
708  "m" (dummy), "m" (dummy), "m" (dummy),
709  "r" (dest), "m" (dstW_reg), "m"(uv_off)
710  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
711  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
712  );
713 }
714 #endif /* HAVE_6REGS */
715 
716 #define REAL_WRITEYUY2(dst, dstw, index) \
717  "packuswb %%mm3, %%mm3 \n\t"\
718  "packuswb %%mm4, %%mm4 \n\t"\
719  "packuswb %%mm7, %%mm1 \n\t"\
720  "punpcklbw %%mm4, %%mm3 \n\t"\
721  "movq %%mm1, %%mm7 \n\t"\
722  "punpcklbw %%mm3, %%mm1 \n\t"\
723  "punpckhbw %%mm3, %%mm7 \n\t"\
724 \
725  MOVNTQ(%%mm1, (dst, index, 2))\
726  MOVNTQ(%%mm7, 8(dst, index, 2))\
727 \
728  "add $8, "#index" \n\t"\
729  "cmp "dstw", "#index" \n\t"\
730  " jb 1b \n\t"
731 #define WRITEYUY2(dst, dstw, index) REAL_WRITEYUY2(dst, dstw, index)
732 
733 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
734  const int16_t **lumSrc, int lumFilterSize,
735  const int16_t *chrFilter, const int16_t **chrUSrc,
736  const int16_t **chrVSrc,
737  int chrFilterSize, const int16_t **alpSrc,
738  uint8_t *dest, int dstW, int dstY)
739 {
740  x86_reg dummy=0;
741  x86_reg dstW_reg = dstW;
742  x86_reg uv_off = c->uv_offx2;
743 
744  YSCALEYUV2PACKEDX_ACCURATE
745  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
746  "psraw $3, %%mm3 \n\t"
747  "psraw $3, %%mm4 \n\t"
748  "psraw $3, %%mm1 \n\t"
749  "psraw $3, %%mm7 \n\t"
750  WRITEYUY2(%4, "%5", %%REGa)
751  YSCALEYUV2PACKEDX_END
752 }
753 
754 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
755  const int16_t **lumSrc, int lumFilterSize,
756  const int16_t *chrFilter, const int16_t **chrUSrc,
757  const int16_t **chrVSrc,
758  int chrFilterSize, const int16_t **alpSrc,
759  uint8_t *dest, int dstW, int dstY)
760 {
761  x86_reg dummy=0;
762  x86_reg dstW_reg = dstW;
763  x86_reg uv_off = c->uv_offx2;
764 
765  YSCALEYUV2PACKEDX
766  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
767  "psraw $3, %%mm3 \n\t"
768  "psraw $3, %%mm4 \n\t"
769  "psraw $3, %%mm1 \n\t"
770  "psraw $3, %%mm7 \n\t"
771  WRITEYUY2(%4, "%5", %%REGa)
772  YSCALEYUV2PACKEDX_END
773 }
774 
775 #define REAL_YSCALEYUV2RGB_UV(index, c) \
776  "xor "#index", "#index" \n\t"\
777  ".p2align 4 \n\t"\
778  "1: \n\t"\
779  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
780  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
781  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
782  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
783  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
784  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
785  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
786  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
787  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
788  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
789  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
790  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
791  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
792  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
793  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
794  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
795  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
796  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
797  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
798  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
799  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
800  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
801 
802 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
803  "movq ("#b1", "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
804  "movq ("#b2", "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
805  "movq 8("#b1", "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
806  "movq 8("#b2", "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
807  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
808  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
809  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
810  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
811  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
812  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
813  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
814  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
815 
816 #define REAL_YSCALEYUV2RGB_COEFF(c) \
817  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
818  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
819  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
820  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
821  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
822  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
823  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
824  "paddw %%mm3, %%mm4 \n\t"\
825  "movq %%mm2, %%mm0 \n\t"\
826  "movq %%mm5, %%mm6 \n\t"\
827  "movq %%mm4, %%mm3 \n\t"\
828  "punpcklwd %%mm2, %%mm2 \n\t"\
829  "punpcklwd %%mm5, %%mm5 \n\t"\
830  "punpcklwd %%mm4, %%mm4 \n\t"\
831  "paddw %%mm1, %%mm2 \n\t"\
832  "paddw %%mm1, %%mm5 \n\t"\
833  "paddw %%mm1, %%mm4 \n\t"\
834  "punpckhwd %%mm0, %%mm0 \n\t"\
835  "punpckhwd %%mm6, %%mm6 \n\t"\
836  "punpckhwd %%mm3, %%mm3 \n\t"\
837  "paddw %%mm7, %%mm0 \n\t"\
838  "paddw %%mm7, %%mm6 \n\t"\
839  "paddw %%mm7, %%mm3 \n\t"\
840  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
841  "packuswb %%mm0, %%mm2 \n\t"\
842  "packuswb %%mm6, %%mm5 \n\t"\
843  "packuswb %%mm3, %%mm4 \n\t"\
844 
845 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
846 
847 #define YSCALEYUV2RGB(index, c) \
848  REAL_YSCALEYUV2RGB_UV(index, c) \
849  REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
850  REAL_YSCALEYUV2RGB_COEFF(c)
851 
852 /**
853  * vertical bilinear scale YV12 to RGB
854  */
855 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
856  const int16_t *ubuf[2], const int16_t *vbuf[2],
857  const int16_t *abuf[2], uint8_t *dest,
858  int dstW, int yalpha, int uvalpha, int y)
859 {
860  const int16_t *buf0 = buf[0], *buf1 = buf[1],
861  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
862 
863  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
864  const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
865 #if ARCH_X86_64
866  __asm__ volatile(
867  YSCALEYUV2RGB(%%r8, %5)
868  YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
869  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
870  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
871  "packuswb %%mm7, %%mm1 \n\t"
872  WRITEBGR32(%4, DSTW_OFFSET"(%5)", %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
873  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
874  "a" (&c->redDither),
875  "r" (abuf0), "r" (abuf1)
876  : "%r8"
877  );
878 #else
879  c->u_temp=(intptr_t)abuf0;
880  c->v_temp=(intptr_t)abuf1;
881  __asm__ volatile(
882  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
883  "mov %4, %%"REG_b" \n\t"
884  "push %%"REG_BP" \n\t"
885  YSCALEYUV2RGB(%%REGBP, %5)
886  "push %0 \n\t"
887  "push %1 \n\t"
888  "mov "U_TEMP"(%5), %0 \n\t"
889  "mov "V_TEMP"(%5), %1 \n\t"
890  YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
891  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
892  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
893  "packuswb %%mm7, %%mm1 \n\t"
894  "pop %1 \n\t"
895  "pop %0 \n\t"
896  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
897  "pop %%"REG_BP" \n\t"
898  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
899  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
900  "a" (&c->redDither)
901  );
902 #endif
903  } else {
904  __asm__ volatile(
905  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
906  "mov %4, %%"REG_b" \n\t"
907  "push %%"REG_BP" \n\t"
908  YSCALEYUV2RGB(%%REGBP, %5)
909  "pcmpeqd %%mm7, %%mm7 \n\t"
910  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
911  "pop %%"REG_BP" \n\t"
912  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
913  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
914  "a" (&c->redDither)
915  );
916  }
917 }
918 
919 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
920  const int16_t *ubuf[2], const int16_t *vbuf[2],
921  const int16_t *abuf[2], uint8_t *dest,
922  int dstW, int yalpha, int uvalpha, int y)
923 {
924  const int16_t *buf0 = buf[0], *buf1 = buf[1],
925  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
926 
927  __asm__ volatile(
928  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
929  "mov %4, %%"REG_b" \n\t"
930  "push %%"REG_BP" \n\t"
931  YSCALEYUV2RGB(%%REGBP, %5)
932  "pxor %%mm7, %%mm7 \n\t"
933  WRITEBGR24(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
934  "pop %%"REG_BP" \n\t"
935  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
936  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
937  "a" (&c->redDither)
938  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
939  );
940 }
941 
942 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
943  const int16_t *ubuf[2], const int16_t *vbuf[2],
944  const int16_t *abuf[2], uint8_t *dest,
945  int dstW, int yalpha, int uvalpha, int y)
946 {
947  const int16_t *buf0 = buf[0], *buf1 = buf[1],
948  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
949 
950  __asm__ volatile(
951  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
952  "mov %4, %%"REG_b" \n\t"
953  "push %%"REG_BP" \n\t"
954  YSCALEYUV2RGB(%%REGBP, %5)
955  "pxor %%mm7, %%mm7 \n\t"
956  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
957 #ifdef DITHER1XBPP
958  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
959  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
960  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
961 #endif
962  WRITERGB15(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
963  "pop %%"REG_BP" \n\t"
964  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
965  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
966  "a" (&c->redDither)
967  NAMED_CONSTRAINTS_ADD(bF8)
968  );
969 }
970 
971 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
972  const int16_t *ubuf[2], const int16_t *vbuf[2],
973  const int16_t *abuf[2], uint8_t *dest,
974  int dstW, int yalpha, int uvalpha, int y)
975 {
976  const int16_t *buf0 = buf[0], *buf1 = buf[1],
977  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
978 
979  __asm__ volatile(
980  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
981  "mov %4, %%"REG_b" \n\t"
982  "push %%"REG_BP" \n\t"
983  YSCALEYUV2RGB(%%REGBP, %5)
984  "pxor %%mm7, %%mm7 \n\t"
985  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
986 #ifdef DITHER1XBPP
987  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
988  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
989  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
990 #endif
991  WRITERGB16(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
992  "pop %%"REG_BP" \n\t"
993  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
994  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
995  "a" (&c->redDither)
996  NAMED_CONSTRAINTS_ADD(bF8,bFC)
997  );
998 }
999 
1000 #define REAL_YSCALEYUV2PACKED(index, c) \
1001  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
1002  "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1 \n\t"\
1003  "psraw $3, %%mm0 \n\t"\
1004  "psraw $3, %%mm1 \n\t"\
1005  "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1006  "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1007  "xor "#index", "#index" \n\t"\
1008  ".p2align 4 \n\t"\
1009  "1: \n\t"\
1010  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1011  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1012  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1013  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1014  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1015  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1016  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
1017  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
1018  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
1019  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
1020  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
1021  "psraw $7, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1022  "psraw $7, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1023  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
1024  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
1025  "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
1026  "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
1027  "movq 8(%0, "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
1028  "movq 8(%1, "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
1029  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
1030  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
1031  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1032  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1033  "psraw $7, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1034  "psraw $7, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1035  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1036  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1037 
1038 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c)
1039 
1040 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
1041  const int16_t *ubuf[2], const int16_t *vbuf[2],
1042  const int16_t *abuf[2], uint8_t *dest,
1043  int dstW, int yalpha, int uvalpha, int y)
1044 {
1045  const int16_t *buf0 = buf[0], *buf1 = buf[1],
1046  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1047 
1048  __asm__ volatile(
1049  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1050  "mov %4, %%"REG_b" \n\t"
1051  "push %%"REG_BP" \n\t"
1052  YSCALEYUV2PACKED(%%REGBP, %5)
1053  WRITEYUY2(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1054  "pop %%"REG_BP" \n\t"
1055  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1056  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1057  "a" (&c->redDither)
1058  );
1059 }
1060 
1061 #define REAL_YSCALEYUV2RGB1(index, c) \
1062  "xor "#index", "#index" \n\t"\
1063  ".p2align 4 \n\t"\
1064  "1: \n\t"\
1065  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1066  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1067  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1068  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1069  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1070  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1071  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1072  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1073  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1074  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1075  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1076  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1077  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1078  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1079  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1080  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1081  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1082  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1083  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1084  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1085  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1086  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1087  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1088  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1089  "paddw %%mm3, %%mm4 \n\t"\
1090  "movq %%mm2, %%mm0 \n\t"\
1091  "movq %%mm5, %%mm6 \n\t"\
1092  "movq %%mm4, %%mm3 \n\t"\
1093  "punpcklwd %%mm2, %%mm2 \n\t"\
1094  "punpcklwd %%mm5, %%mm5 \n\t"\
1095  "punpcklwd %%mm4, %%mm4 \n\t"\
1096  "paddw %%mm1, %%mm2 \n\t"\
1097  "paddw %%mm1, %%mm5 \n\t"\
1098  "paddw %%mm1, %%mm4 \n\t"\
1099  "punpckhwd %%mm0, %%mm0 \n\t"\
1100  "punpckhwd %%mm6, %%mm6 \n\t"\
1101  "punpckhwd %%mm3, %%mm3 \n\t"\
1102  "paddw %%mm7, %%mm0 \n\t"\
1103  "paddw %%mm7, %%mm6 \n\t"\
1104  "paddw %%mm7, %%mm3 \n\t"\
1105  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1106  "packuswb %%mm0, %%mm2 \n\t"\
1107  "packuswb %%mm6, %%mm5 \n\t"\
1108  "packuswb %%mm3, %%mm4 \n\t"\
1109 
1110 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c)
1111 
1112 // do vertical chrominance interpolation
1113 #define REAL_YSCALEYUV2RGB1b(index, c) \
1114  "xor "#index", "#index" \n\t"\
1115  ".p2align 4 \n\t"\
1116  "1: \n\t"\
1117  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1118  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1119  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1120  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1121  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1122  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1123  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1124  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1125  "psrlw $5, %%mm3 \n\t" /*FIXME might overflow*/\
1126  "psrlw $5, %%mm4 \n\t" /*FIXME might overflow*/\
1127  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1128  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1129  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1130  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1131  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1132  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1133  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1134  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1135  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1136  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1137  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1138  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1139  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1140  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1141  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1142  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1143  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1144  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1145  "paddw %%mm3, %%mm4 \n\t"\
1146  "movq %%mm2, %%mm0 \n\t"\
1147  "movq %%mm5, %%mm6 \n\t"\
1148  "movq %%mm4, %%mm3 \n\t"\
1149  "punpcklwd %%mm2, %%mm2 \n\t"\
1150  "punpcklwd %%mm5, %%mm5 \n\t"\
1151  "punpcklwd %%mm4, %%mm4 \n\t"\
1152  "paddw %%mm1, %%mm2 \n\t"\
1153  "paddw %%mm1, %%mm5 \n\t"\
1154  "paddw %%mm1, %%mm4 \n\t"\
1155  "punpckhwd %%mm0, %%mm0 \n\t"\
1156  "punpckhwd %%mm6, %%mm6 \n\t"\
1157  "punpckhwd %%mm3, %%mm3 \n\t"\
1158  "paddw %%mm7, %%mm0 \n\t"\
1159  "paddw %%mm7, %%mm6 \n\t"\
1160  "paddw %%mm7, %%mm3 \n\t"\
1161  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1162  "packuswb %%mm0, %%mm2 \n\t"\
1163  "packuswb %%mm6, %%mm5 \n\t"\
1164  "packuswb %%mm3, %%mm4 \n\t"\
1165 
1166 #define YSCALEYUV2RGB1b(index, c) REAL_YSCALEYUV2RGB1b(index, c)
1167 
1168 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1169  "movq (%1, "#index", 2), %%mm7 \n\t" /* abuf0[index ] */\
1170  "movq 8(%1, "#index", 2), %%mm1 \n\t" /* abuf0[index+4] */\
1171  "psraw $7, %%mm7 \n\t" /* abuf0[index ] >>7 */\
1172  "psraw $7, %%mm1 \n\t" /* abuf0[index+4] >>7 */\
1173  "packuswb %%mm1, %%mm7 \n\t"
1174 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1175 
1176 /**
1177  * YV12 to RGB without scaling or interpolating
1178  */
1179 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1180  const int16_t *ubuf[2], const int16_t *vbuf[2],
1181  const int16_t *abuf0, uint8_t *dest,
1182  int dstW, int uvalpha, int y)
1183 {
1184  const int16_t *ubuf0 = ubuf[0];
1185  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1186 
1187  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1188  const int16_t *ubuf1 = ubuf[0];
1189  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1190  __asm__ volatile(
1191  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1192  "mov %4, %%"REG_b" \n\t"
1193  "push %%"REG_BP" \n\t"
1194  YSCALEYUV2RGB1(%%REGBP, %5)
1195  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1196  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1197  "pop %%"REG_BP" \n\t"
1198  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1199  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1200  "a" (&c->redDither)
1201  );
1202  } else {
1203  __asm__ volatile(
1204  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1205  "mov %4, %%"REG_b" \n\t"
1206  "push %%"REG_BP" \n\t"
1207  YSCALEYUV2RGB1(%%REGBP, %5)
1208  "pcmpeqd %%mm7, %%mm7 \n\t"
1209  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1210  "pop %%"REG_BP" \n\t"
1211  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1212  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1213  "a" (&c->redDither)
1214  );
1215  }
1216  } else {
1217  const int16_t *ubuf1 = ubuf[1];
1218  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1219  __asm__ volatile(
1220  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1221  "mov %4, %%"REG_b" \n\t"
1222  "push %%"REG_BP" \n\t"
1223  YSCALEYUV2RGB1b(%%REGBP, %5)
1224  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1225  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1226  "pop %%"REG_BP" \n\t"
1227  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1228  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1229  "a" (&c->redDither)
1230  );
1231  } else {
1232  __asm__ volatile(
1233  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1234  "mov %4, %%"REG_b" \n\t"
1235  "push %%"REG_BP" \n\t"
1236  YSCALEYUV2RGB1b(%%REGBP, %5)
1237  "pcmpeqd %%mm7, %%mm7 \n\t"
1238  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1239  "pop %%"REG_BP" \n\t"
1240  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1241  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1242  "a" (&c->redDither)
1243  );
1244  }
1245  }
1246 }
1247 
1248 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1249  const int16_t *ubuf[2], const int16_t *vbuf[2],
1250  const int16_t *abuf0, uint8_t *dest,
1251  int dstW, int uvalpha, int y)
1252 {
1253  const int16_t *ubuf0 = ubuf[0];
1254  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1255 
1256  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1257  const int16_t *ubuf1 = ubuf[0];
1258  __asm__ volatile(
1259  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1260  "mov %4, %%"REG_b" \n\t"
1261  "push %%"REG_BP" \n\t"
1262  YSCALEYUV2RGB1(%%REGBP, %5)
1263  "pxor %%mm7, %%mm7 \n\t"
1264  WRITEBGR24(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1265  "pop %%"REG_BP" \n\t"
1266  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1267  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1268  "a" (&c->redDither)
1269  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
1270  );
1271  } else {
1272  const int16_t *ubuf1 = ubuf[1];
1273  __asm__ volatile(
1274  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1275  "mov %4, %%"REG_b" \n\t"
1276  "push %%"REG_BP" \n\t"
1277  YSCALEYUV2RGB1b(%%REGBP, %5)
1278  "pxor %%mm7, %%mm7 \n\t"
1279  WRITEBGR24(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1280  "pop %%"REG_BP" \n\t"
1281  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1282  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1283  "a" (&c->redDither)
1284  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
1285  );
1286  }
1287 }
1288 
1289 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1290  const int16_t *ubuf[2], const int16_t *vbuf[2],
1291  const int16_t *abuf0, uint8_t *dest,
1292  int dstW, int uvalpha, int y)
1293 {
1294  const int16_t *ubuf0 = ubuf[0];
1295  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1296 
1297  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1298  const int16_t *ubuf1 = ubuf[0];
1299  __asm__ volatile(
1300  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1301  "mov %4, %%"REG_b" \n\t"
1302  "push %%"REG_BP" \n\t"
1303  YSCALEYUV2RGB1(%%REGBP, %5)
1304  "pxor %%mm7, %%mm7 \n\t"
1305  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1306 #ifdef DITHER1XBPP
1307  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1308  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1309  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1310 #endif
1311  WRITERGB15(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1312  "pop %%"REG_BP" \n\t"
1313  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1314  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1315  "a" (&c->redDither)
1316  NAMED_CONSTRAINTS_ADD(bF8)
1317  );
1318  } else {
1319  const int16_t *ubuf1 = ubuf[1];
1320  __asm__ volatile(
1321  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1322  "mov %4, %%"REG_b" \n\t"
1323  "push %%"REG_BP" \n\t"
1324  YSCALEYUV2RGB1b(%%REGBP, %5)
1325  "pxor %%mm7, %%mm7 \n\t"
1326  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1327 #ifdef DITHER1XBPP
1328  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1329  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1330  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1331 #endif
1332  WRITERGB15(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1333  "pop %%"REG_BP" \n\t"
1334  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1335  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1336  "a" (&c->redDither)
1337  NAMED_CONSTRAINTS_ADD(bF8)
1338  );
1339  }
1340 }
1341 
1342 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1343  const int16_t *ubuf[2], const int16_t *vbuf[2],
1344  const int16_t *abuf0, uint8_t *dest,
1345  int dstW, int uvalpha, int y)
1346 {
1347  const int16_t *ubuf0 = ubuf[0];
1348  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1349 
1350  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1351  const int16_t *ubuf1 = ubuf[0];
1352  __asm__ volatile(
1353  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1354  "mov %4, %%"REG_b" \n\t"
1355  "push %%"REG_BP" \n\t"
1356  YSCALEYUV2RGB1(%%REGBP, %5)
1357  "pxor %%mm7, %%mm7 \n\t"
1358  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1359 #ifdef DITHER1XBPP
1360  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1361  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1362  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1363 #endif
1364  WRITERGB16(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1365  "pop %%"REG_BP" \n\t"
1366  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1367  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1368  "a" (&c->redDither)
1369  NAMED_CONSTRAINTS_ADD(bF8,bFC)
1370  );
1371  } else {
1372  const int16_t *ubuf1 = ubuf[1];
1373  __asm__ volatile(
1374  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1375  "mov %4, %%"REG_b" \n\t"
1376  "push %%"REG_BP" \n\t"
1377  YSCALEYUV2RGB1b(%%REGBP, %5)
1378  "pxor %%mm7, %%mm7 \n\t"
1379  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1380 #ifdef DITHER1XBPP
1381  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1382  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1383  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1384 #endif
1385  WRITERGB16(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1386  "pop %%"REG_BP" \n\t"
1387  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1388  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1389  "a" (&c->redDither)
1390  NAMED_CONSTRAINTS_ADD(bF8,bFC)
1391  );
1392  }
1393 }
1394 
1395 #define REAL_YSCALEYUV2PACKED1(index, c) \
1396  "xor "#index", "#index" \n\t"\
1397  ".p2align 4 \n\t"\
1398  "1: \n\t"\
1399  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1400  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1401  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1402  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1403  "psraw $7, %%mm3 \n\t" \
1404  "psraw $7, %%mm4 \n\t" \
1405  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1406  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1407  "psraw $7, %%mm1 \n\t" \
1408  "psraw $7, %%mm7 \n\t" \
1409 
1410 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c)
1411 
1412 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1413  "xor "#index", "#index" \n\t"\
1414  ".p2align 4 \n\t"\
1415  "1: \n\t"\
1416  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1417  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1418  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1419  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1420  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1421  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1422  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1423  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1424  "psrlw $8, %%mm3 \n\t" \
1425  "psrlw $8, %%mm4 \n\t" \
1426  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1427  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1428  "psraw $7, %%mm1 \n\t" \
1429  "psraw $7, %%mm7 \n\t"
1430 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c)
1431 
1432 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1433  const int16_t *ubuf[2], const int16_t *vbuf[2],
1434  const int16_t *abuf0, uint8_t *dest,
1435  int dstW, int uvalpha, int y)
1436 {
1437  const int16_t *ubuf0 = ubuf[0];
1438  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1439 
1440  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1441  const int16_t *ubuf1 = ubuf[0];
1442  __asm__ volatile(
1443  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1444  "mov %4, %%"REG_b" \n\t"
1445  "push %%"REG_BP" \n\t"
1446  YSCALEYUV2PACKED1(%%REGBP, %5)
1447  WRITEYUY2(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1448  "pop %%"REG_BP" \n\t"
1449  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1450  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1451  "a" (&c->redDither)
1452  );
1453  } else {
1454  const int16_t *ubuf1 = ubuf[1];
1455  __asm__ volatile(
1456  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1457  "mov %4, %%"REG_b" \n\t"
1458  "push %%"REG_BP" \n\t"
1459  YSCALEYUV2PACKED1b(%%REGBP, %5)
1460  WRITEYUY2(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1461  "pop %%"REG_BP" \n\t"
1462  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1463  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1464  "a" (&c->redDither)
1465  );
1466  }
1467 }
1468 static av_cold void RENAME(sws_init_swscale)(SwsContext *c)
1469 {
1470  enum AVPixelFormat dstFormat = c->dstFormat;
1471 
1472  c->use_mmx_vfilter= 0;
1473  if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) && dstFormat != AV_PIX_FMT_NV12
1474  && dstFormat != AV_PIX_FMT_NV21 && !(c->flags & SWS_BITEXACT)) {
1475  if (c->flags & SWS_ACCURATE_RND) {
1476  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1477  switch (c->dstFormat) {
1478  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X_ar); break;
1479 #if HAVE_6REGS
1480  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X_ar); break;
1481 #endif
1482  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X_ar); break;
1483  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X_ar); break;
1484  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
1485  default: break;
1486  }
1487  }
1488  } else {
1489  c->use_mmx_vfilter= 1;
1490  c->yuv2planeX = RENAME(yuv2yuvX );
1491  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1492  switch (c->dstFormat) {
1493  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X); break;
1494 #if HAVE_6REGS
1495  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X); break;
1496 #endif
1497  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X); break;
1498  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X); break;
1499  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
1500  default: break;
1501  }
1502  }
1503  }
1504  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1505  switch (c->dstFormat) {
1506  case AV_PIX_FMT_RGB32:
1507  c->yuv2packed1 = RENAME(yuv2rgb32_1);
1508  c->yuv2packed2 = RENAME(yuv2rgb32_2);
1509  break;
1510  case AV_PIX_FMT_BGR24:
1511  c->yuv2packed1 = RENAME(yuv2bgr24_1);
1512  c->yuv2packed2 = RENAME(yuv2bgr24_2);
1513  break;
1514  case AV_PIX_FMT_RGB555:
1515  c->yuv2packed1 = RENAME(yuv2rgb555_1);
1516  c->yuv2packed2 = RENAME(yuv2rgb555_2);
1517  break;
1518  case AV_PIX_FMT_RGB565:
1519  c->yuv2packed1 = RENAME(yuv2rgb565_1);
1520  c->yuv2packed2 = RENAME(yuv2rgb565_2);
1521  break;
1522  case AV_PIX_FMT_YUYV422:
1523  c->yuv2packed1 = RENAME(yuv2yuyv422_1);
1524  c->yuv2packed2 = RENAME(yuv2yuyv422_2);
1525  break;
1526  default:
1527  break;
1528  }
1529  }
1530  }
1531 
1532  if (c->srcBpc == 8 && c->dstBpc <= 14) {
1533  // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one).
1534 #if COMPILE_TEMPLATE_MMXEXT
1535  if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) {
1536  c->hyscale_fast = ff_hyscale_fast_mmxext;
1537  c->hcscale_fast = ff_hcscale_fast_mmxext;
1538  } else {
1539 #endif /* COMPILE_TEMPLATE_MMXEXT */
1540  c->hyscale_fast = NULL;
1541  c->hcscale_fast = NULL;
1542 #if COMPILE_TEMPLATE_MMXEXT
1543  }
1544 #endif /* COMPILE_TEMPLATE_MMXEXT */
1545  }
1546 }
Generated on Sun Mar 8 2015 02:35:15 for FFmpeg by   doxygen 1.8.2