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 #if COMPILE_TEMPLATE_MMXEXT
32 #define PREFETCH "prefetchnta"
33 #else
34 #define PREFETCH " # nop"
35 #endif
36 
37 #if COMPILE_TEMPLATE_MMXEXT
38 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
39 #define MOVNTQ2 "movntq "
40 #else
41 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
42 #define MOVNTQ2 "movq "
43 #endif
44 #define MOVNTQ(a,b) REAL_MOVNTQ(a,b)
45 
46 #if !COMPILE_TEMPLATE_MMXEXT
47 static av_always_inline void
48 dither_8to16(const uint8_t *srcDither, int rot)
49 {
50  if (rot) {
51  __asm__ volatile("pxor %%mm0, %%mm0\n\t"
52  "movq (%0), %%mm3\n\t"
53  "movq %%mm3, %%mm4\n\t"
54  "psrlq $24, %%mm3\n\t"
55  "psllq $40, %%mm4\n\t"
56  "por %%mm4, %%mm3\n\t"
57  "movq %%mm3, %%mm4\n\t"
58  "punpcklbw %%mm0, %%mm3\n\t"
59  "punpckhbw %%mm0, %%mm4\n\t"
60  :: "r"(srcDither)
61  );
62  } else {
63  __asm__ volatile("pxor %%mm0, %%mm0\n\t"
64  "movq (%0), %%mm3\n\t"
65  "movq %%mm3, %%mm4\n\t"
66  "punpcklbw %%mm0, %%mm3\n\t"
67  "punpckhbw %%mm0, %%mm4\n\t"
68  :: "r"(srcDither)
69  );
70  }
71 }
72 #endif
73 
74 static void RENAME(yuv2yuvX)(const int16_t *filter, int filterSize,
75  const int16_t **src, uint8_t *dest, int dstW,
76  const uint8_t *dither, int offset)
77 {
78  dither_8to16(dither, offset);
79  filterSize--;
80  __asm__ volatile(
81  "movd %0, %%mm1\n\t"
82  "punpcklwd %%mm1, %%mm1\n\t"
83  "punpckldq %%mm1, %%mm1\n\t"
84  "psllw $3, %%mm1\n\t"
85  "paddw %%mm1, %%mm3\n\t"
86  "paddw %%mm1, %%mm4\n\t"
87  "psraw $4, %%mm3\n\t"
88  "psraw $4, %%mm4\n\t"
89  ::"m"(filterSize)
90  );
91 
92  __asm__ volatile(\
93  "movq %%mm3, %%mm6\n\t"
94  "movq %%mm4, %%mm7\n\t"
95  "movl %3, %%ecx\n\t"
96  "mov %0, %%"REG_d" \n\t"\
97  "mov (%%"REG_d"), %%"REG_S" \n\t"\
98  ".p2align 4 \n\t" /* FIXME Unroll? */\
99  "1: \n\t"\
100  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
101  "movq (%%"REG_S", %%"REG_c", 2), %%mm2 \n\t" /* srcData */\
102  "movq 8(%%"REG_S", %%"REG_c", 2), %%mm5 \n\t" /* srcData */\
103  "add $16, %%"REG_d" \n\t"\
104  "mov (%%"REG_d"), %%"REG_S" \n\t"\
105  "test %%"REG_S", %%"REG_S" \n\t"\
106  "pmulhw %%mm0, %%mm2 \n\t"\
107  "pmulhw %%mm0, %%mm5 \n\t"\
108  "paddw %%mm2, %%mm3 \n\t"\
109  "paddw %%mm5, %%mm4 \n\t"\
110  " jnz 1b \n\t"\
111  "psraw $3, %%mm3 \n\t"\
112  "psraw $3, %%mm4 \n\t"\
113  "packuswb %%mm4, %%mm3 \n\t"
114  MOVNTQ2 " %%mm3, (%1, %%"REG_c")\n\t"
115  "add $8, %%"REG_c" \n\t"\
116  "cmp %2, %%"REG_c" \n\t"\
117  "movq %%mm6, %%mm3\n\t"
118  "movq %%mm7, %%mm4\n\t"
119  "mov %0, %%"REG_d" \n\t"\
120  "mov (%%"REG_d"), %%"REG_S" \n\t"\
121  "jb 1b \n\t"\
122  :: "g" (filter),
123  "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset)
124  : "%"REG_d, "%"REG_S, "%"REG_c
125  );
126 }
127 
128 #define YSCALEYUV2PACKEDX_UV \
129  __asm__ volatile(\
130  "xor %%"REG_a", %%"REG_a" \n\t"\
131  ".p2align 4 \n\t"\
132  "nop \n\t"\
133  "1: \n\t"\
134  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
135  "mov (%%"REG_d"), %%"REG_S" \n\t"\
136  "movq "VROUNDER_OFFSET"(%0), %%mm3 \n\t"\
137  "movq %%mm3, %%mm4 \n\t"\
138  ".p2align 4 \n\t"\
139  "2: \n\t"\
140  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
141  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\
142  "add %6, %%"REG_S" \n\t" \
143  "movq (%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\
144  "add $16, %%"REG_d" \n\t"\
145  "mov (%%"REG_d"), %%"REG_S" \n\t"\
146  "pmulhw %%mm0, %%mm2 \n\t"\
147  "pmulhw %%mm0, %%mm5 \n\t"\
148  "paddw %%mm2, %%mm3 \n\t"\
149  "paddw %%mm5, %%mm4 \n\t"\
150  "test %%"REG_S", %%"REG_S" \n\t"\
151  " jnz 2b \n\t"\
152 
153 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
154  "lea "offset"(%0), %%"REG_d" \n\t"\
155  "mov (%%"REG_d"), %%"REG_S" \n\t"\
156  "movq "VROUNDER_OFFSET"(%0), "#dst1" \n\t"\
157  "movq "#dst1", "#dst2" \n\t"\
158  ".p2align 4 \n\t"\
159  "2: \n\t"\
160  "movq 8(%%"REG_d"), "#coeff" \n\t" /* filterCoeff */\
161  "movq (%%"REG_S", %%"REG_a", 2), "#src1" \n\t" /* Y1srcData */\
162  "movq 8(%%"REG_S", %%"REG_a", 2), "#src2" \n\t" /* Y2srcData */\
163  "add $16, %%"REG_d" \n\t"\
164  "mov (%%"REG_d"), %%"REG_S" \n\t"\
165  "pmulhw "#coeff", "#src1" \n\t"\
166  "pmulhw "#coeff", "#src2" \n\t"\
167  "paddw "#src1", "#dst1" \n\t"\
168  "paddw "#src2", "#dst2" \n\t"\
169  "test %%"REG_S", %%"REG_S" \n\t"\
170  " jnz 2b \n\t"\
171 
172 #define YSCALEYUV2PACKEDX \
173  YSCALEYUV2PACKEDX_UV \
174  YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
175 
176 #define YSCALEYUV2PACKEDX_END \
177  :: "r" (&c->redDither), \
178  "m" (dummy), "m" (dummy), "m" (dummy),\
179  "r" (dest), "m" (dstW_reg), "m"(uv_off) \
180  NAMED_CONSTRAINTS_ADD(bF8,bFC) \
181  : "%"REG_a, "%"REG_d, "%"REG_S \
182  );
183 
184 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
185  __asm__ volatile(\
186  "xor %%"REG_a", %%"REG_a" \n\t"\
187  ".p2align 4 \n\t"\
188  "nop \n\t"\
189  "1: \n\t"\
190  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
191  "mov (%%"REG_d"), %%"REG_S" \n\t"\
192  "pxor %%mm4, %%mm4 \n\t"\
193  "pxor %%mm5, %%mm5 \n\t"\
194  "pxor %%mm6, %%mm6 \n\t"\
195  "pxor %%mm7, %%mm7 \n\t"\
196  ".p2align 4 \n\t"\
197  "2: \n\t"\
198  "movq (%%"REG_S", %%"REG_a"), %%mm0 \n\t" /* UsrcData */\
199  "add %6, %%"REG_S" \n\t" \
200  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* VsrcData */\
201  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
202  "movq (%%"REG_S", %%"REG_a"), %%mm1 \n\t" /* UsrcData */\
203  "movq %%mm0, %%mm3 \n\t"\
204  "punpcklwd %%mm1, %%mm0 \n\t"\
205  "punpckhwd %%mm1, %%mm3 \n\t"\
206  "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1 \n\t" /* filterCoeff */\
207  "pmaddwd %%mm1, %%mm0 \n\t"\
208  "pmaddwd %%mm1, %%mm3 \n\t"\
209  "paddd %%mm0, %%mm4 \n\t"\
210  "paddd %%mm3, %%mm5 \n\t"\
211  "add %6, %%"REG_S" \n\t" \
212  "movq (%%"REG_S", %%"REG_a"), %%mm3 \n\t" /* VsrcData */\
213  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
214  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
215  "test %%"REG_S", %%"REG_S" \n\t"\
216  "movq %%mm2, %%mm0 \n\t"\
217  "punpcklwd %%mm3, %%mm2 \n\t"\
218  "punpckhwd %%mm3, %%mm0 \n\t"\
219  "pmaddwd %%mm1, %%mm2 \n\t"\
220  "pmaddwd %%mm1, %%mm0 \n\t"\
221  "paddd %%mm2, %%mm6 \n\t"\
222  "paddd %%mm0, %%mm7 \n\t"\
223  " jnz 2b \n\t"\
224  "psrad $16, %%mm4 \n\t"\
225  "psrad $16, %%mm5 \n\t"\
226  "psrad $16, %%mm6 \n\t"\
227  "psrad $16, %%mm7 \n\t"\
228  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
229  "packssdw %%mm5, %%mm4 \n\t"\
230  "packssdw %%mm7, %%mm6 \n\t"\
231  "paddw %%mm0, %%mm4 \n\t"\
232  "paddw %%mm0, %%mm6 \n\t"\
233  "movq %%mm4, "U_TEMP"(%0) \n\t"\
234  "movq %%mm6, "V_TEMP"(%0) \n\t"\
235 
236 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
237  "lea "offset"(%0), %%"REG_d" \n\t"\
238  "mov (%%"REG_d"), %%"REG_S" \n\t"\
239  "pxor %%mm1, %%mm1 \n\t"\
240  "pxor %%mm5, %%mm5 \n\t"\
241  "pxor %%mm7, %%mm7 \n\t"\
242  "pxor %%mm6, %%mm6 \n\t"\
243  ".p2align 4 \n\t"\
244  "2: \n\t"\
245  "movq (%%"REG_S", %%"REG_a", 2), %%mm0 \n\t" /* Y1srcData */\
246  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y2srcData */\
247  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
248  "movq (%%"REG_S", %%"REG_a", 2), %%mm4 \n\t" /* Y1srcData */\
249  "movq %%mm0, %%mm3 \n\t"\
250  "punpcklwd %%mm4, %%mm0 \n\t"\
251  "punpckhwd %%mm4, %%mm3 \n\t"\
252  "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4 \n\t" /* filterCoeff */\
253  "pmaddwd %%mm4, %%mm0 \n\t"\
254  "pmaddwd %%mm4, %%mm3 \n\t"\
255  "paddd %%mm0, %%mm1 \n\t"\
256  "paddd %%mm3, %%mm5 \n\t"\
257  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3 \n\t" /* Y2srcData */\
258  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
259  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
260  "test %%"REG_S", %%"REG_S" \n\t"\
261  "movq %%mm2, %%mm0 \n\t"\
262  "punpcklwd %%mm3, %%mm2 \n\t"\
263  "punpckhwd %%mm3, %%mm0 \n\t"\
264  "pmaddwd %%mm4, %%mm2 \n\t"\
265  "pmaddwd %%mm4, %%mm0 \n\t"\
266  "paddd %%mm2, %%mm7 \n\t"\
267  "paddd %%mm0, %%mm6 \n\t"\
268  " jnz 2b \n\t"\
269  "psrad $16, %%mm1 \n\t"\
270  "psrad $16, %%mm5 \n\t"\
271  "psrad $16, %%mm7 \n\t"\
272  "psrad $16, %%mm6 \n\t"\
273  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
274  "packssdw %%mm5, %%mm1 \n\t"\
275  "packssdw %%mm6, %%mm7 \n\t"\
276  "paddw %%mm0, %%mm1 \n\t"\
277  "paddw %%mm0, %%mm7 \n\t"\
278  "movq "U_TEMP"(%0), %%mm3 \n\t"\
279  "movq "V_TEMP"(%0), %%mm4 \n\t"\
280 
281 #define YSCALEYUV2PACKEDX_ACCURATE \
282  YSCALEYUV2PACKEDX_ACCURATE_UV \
283  YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
284 
285 #define YSCALEYUV2RGBX \
286  "psubw "U_OFFSET"(%0), %%mm3 \n\t" /* (U-128)8*/\
287  "psubw "V_OFFSET"(%0), %%mm4 \n\t" /* (V-128)8*/\
288  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
289  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
290  "pmulhw "UG_COEFF"(%0), %%mm3 \n\t"\
291  "pmulhw "VG_COEFF"(%0), %%mm4 \n\t"\
292  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
293  "pmulhw "UB_COEFF"(%0), %%mm2 \n\t"\
294  "pmulhw "VR_COEFF"(%0), %%mm5 \n\t"\
295  "psubw "Y_OFFSET"(%0), %%mm1 \n\t" /* 8(Y-16)*/\
296  "psubw "Y_OFFSET"(%0), %%mm7 \n\t" /* 8(Y-16)*/\
297  "pmulhw "Y_COEFF"(%0), %%mm1 \n\t"\
298  "pmulhw "Y_COEFF"(%0), %%mm7 \n\t"\
299  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
300  "paddw %%mm3, %%mm4 \n\t"\
301  "movq %%mm2, %%mm0 \n\t"\
302  "movq %%mm5, %%mm6 \n\t"\
303  "movq %%mm4, %%mm3 \n\t"\
304  "punpcklwd %%mm2, %%mm2 \n\t"\
305  "punpcklwd %%mm5, %%mm5 \n\t"\
306  "punpcklwd %%mm4, %%mm4 \n\t"\
307  "paddw %%mm1, %%mm2 \n\t"\
308  "paddw %%mm1, %%mm5 \n\t"\
309  "paddw %%mm1, %%mm4 \n\t"\
310  "punpckhwd %%mm0, %%mm0 \n\t"\
311  "punpckhwd %%mm6, %%mm6 \n\t"\
312  "punpckhwd %%mm3, %%mm3 \n\t"\
313  "paddw %%mm7, %%mm0 \n\t"\
314  "paddw %%mm7, %%mm6 \n\t"\
315  "paddw %%mm7, %%mm3 \n\t"\
316  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
317  "packuswb %%mm0, %%mm2 \n\t"\
318  "packuswb %%mm6, %%mm5 \n\t"\
319  "packuswb %%mm3, %%mm4 \n\t"\
320 
321 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
322  "movq "#b", "#q2" \n\t" /* B */\
323  "movq "#r", "#t" \n\t" /* R */\
324  "punpcklbw "#g", "#b" \n\t" /* GBGBGBGB 0 */\
325  "punpcklbw "#a", "#r" \n\t" /* ARARARAR 0 */\
326  "punpckhbw "#g", "#q2" \n\t" /* GBGBGBGB 2 */\
327  "punpckhbw "#a", "#t" \n\t" /* ARARARAR 2 */\
328  "movq "#b", "#q0" \n\t" /* GBGBGBGB 0 */\
329  "movq "#q2", "#q3" \n\t" /* GBGBGBGB 2 */\
330  "punpcklwd "#r", "#q0" \n\t" /* ARGBARGB 0 */\
331  "punpckhwd "#r", "#b" \n\t" /* ARGBARGB 1 */\
332  "punpcklwd "#t", "#q2" \n\t" /* ARGBARGB 2 */\
333  "punpckhwd "#t", "#q3" \n\t" /* ARGBARGB 3 */\
334 \
335  MOVNTQ( q0, (dst, index, 4))\
336  MOVNTQ( b, 8(dst, index, 4))\
337  MOVNTQ( q2, 16(dst, index, 4))\
338  MOVNTQ( q3, 24(dst, index, 4))\
339 \
340  "add $8, "#index" \n\t"\
341  "cmp "dstw", "#index" \n\t"\
342  " jb 1b \n\t"
343 #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)
344 
345 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
346  const int16_t **lumSrc, int lumFilterSize,
347  const int16_t *chrFilter, const int16_t **chrUSrc,
348  const int16_t **chrVSrc,
349  int chrFilterSize, const int16_t **alpSrc,
350  uint8_t *dest, int dstW, int dstY)
351 {
352  x86_reg dummy=0;
353  x86_reg dstW_reg = dstW;
354  x86_reg uv_off = c->uv_offx2;
355 
356  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
357  YSCALEYUV2PACKEDX_ACCURATE
358  YSCALEYUV2RGBX
359  "movq %%mm2, "U_TEMP"(%0) \n\t"
360  "movq %%mm4, "V_TEMP"(%0) \n\t"
361  "movq %%mm5, "Y_TEMP"(%0) \n\t"
362  YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)
363  "movq "Y_TEMP"(%0), %%mm5 \n\t"
364  "psraw $3, %%mm1 \n\t"
365  "psraw $3, %%mm7 \n\t"
366  "packuswb %%mm7, %%mm1 \n\t"
367  WRITEBGR32(%4, "%5", %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
368  YSCALEYUV2PACKEDX_END
369  } else {
370  YSCALEYUV2PACKEDX_ACCURATE
371  YSCALEYUV2RGBX
372  "pcmpeqd %%mm7, %%mm7 \n\t"
373  WRITEBGR32(%4, "%5", %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
374  YSCALEYUV2PACKEDX_END
375  }
376 }
377 
378 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
379  const int16_t **lumSrc, int lumFilterSize,
380  const int16_t *chrFilter, const int16_t **chrUSrc,
381  const int16_t **chrVSrc,
382  int chrFilterSize, const int16_t **alpSrc,
383  uint8_t *dest, int dstW, int dstY)
384 {
385  x86_reg dummy=0;
386  x86_reg dstW_reg = dstW;
387  x86_reg uv_off = c->uv_offx2;
388 
389  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
390  YSCALEYUV2PACKEDX
391  YSCALEYUV2RGBX
392  YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
393  "psraw $3, %%mm1 \n\t"
394  "psraw $3, %%mm7 \n\t"
395  "packuswb %%mm7, %%mm1 \n\t"
396  WRITEBGR32(%4, "%5", %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
397  YSCALEYUV2PACKEDX_END
398  } else {
399  YSCALEYUV2PACKEDX
400  YSCALEYUV2RGBX
401  "pcmpeqd %%mm7, %%mm7 \n\t"
402  WRITEBGR32(%4, "%5", %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
403  YSCALEYUV2PACKEDX_END
404  }
405 }
406 
407 #define REAL_WRITERGB16(dst, dstw, index) \
408  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
409  "pand "MANGLE(bFC)", %%mm4 \n\t" /* G */\
410  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
411  "psrlq $3, %%mm2 \n\t"\
412 \
413  "movq %%mm2, %%mm1 \n\t"\
414  "movq %%mm4, %%mm3 \n\t"\
415 \
416  "punpcklbw %%mm7, %%mm3 \n\t"\
417  "punpcklbw %%mm5, %%mm2 \n\t"\
418  "punpckhbw %%mm7, %%mm4 \n\t"\
419  "punpckhbw %%mm5, %%mm1 \n\t"\
420 \
421  "psllq $3, %%mm3 \n\t"\
422  "psllq $3, %%mm4 \n\t"\
423 \
424  "por %%mm3, %%mm2 \n\t"\
425  "por %%mm4, %%mm1 \n\t"\
426 \
427  MOVNTQ(%%mm2, (dst, index, 2))\
428  MOVNTQ(%%mm1, 8(dst, index, 2))\
429 \
430  "add $8, "#index" \n\t"\
431  "cmp "dstw", "#index" \n\t"\
432  " jb 1b \n\t"
433 #define WRITERGB16(dst, dstw, index) REAL_WRITERGB16(dst, dstw, index)
434 
435 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
436  const int16_t **lumSrc, int lumFilterSize,
437  const int16_t *chrFilter, const int16_t **chrUSrc,
438  const int16_t **chrVSrc,
439  int chrFilterSize, const int16_t **alpSrc,
440  uint8_t *dest, int dstW, int dstY)
441 {
442  x86_reg dummy=0;
443  x86_reg dstW_reg = dstW;
444  x86_reg uv_off = c->uv_offx2;
445 
446  YSCALEYUV2PACKEDX_ACCURATE
447  YSCALEYUV2RGBX
448  "pxor %%mm7, %%mm7 \n\t"
449  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
450 #ifdef DITHER1XBPP
451  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
452  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
453  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
454 #endif
455  WRITERGB16(%4, "%5", %%REGa)
456  YSCALEYUV2PACKEDX_END
457 }
458 
459 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
460  const int16_t **lumSrc, int lumFilterSize,
461  const int16_t *chrFilter, const int16_t **chrUSrc,
462  const int16_t **chrVSrc,
463  int chrFilterSize, const int16_t **alpSrc,
464  uint8_t *dest, int dstW, int dstY)
465 {
466  x86_reg dummy=0;
467  x86_reg dstW_reg = dstW;
468  x86_reg uv_off = c->uv_offx2;
469 
470  YSCALEYUV2PACKEDX
471  YSCALEYUV2RGBX
472  "pxor %%mm7, %%mm7 \n\t"
473  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
474 #ifdef DITHER1XBPP
475  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
476  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
477  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
478 #endif
479  WRITERGB16(%4, "%5", %%REGa)
480  YSCALEYUV2PACKEDX_END
481 }
482 
483 #define REAL_WRITERGB15(dst, dstw, index) \
484  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
485  "pand "MANGLE(bF8)", %%mm4 \n\t" /* G */\
486  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
487  "psrlq $3, %%mm2 \n\t"\
488  "psrlq $1, %%mm5 \n\t"\
489 \
490  "movq %%mm2, %%mm1 \n\t"\
491  "movq %%mm4, %%mm3 \n\t"\
492 \
493  "punpcklbw %%mm7, %%mm3 \n\t"\
494  "punpcklbw %%mm5, %%mm2 \n\t"\
495  "punpckhbw %%mm7, %%mm4 \n\t"\
496  "punpckhbw %%mm5, %%mm1 \n\t"\
497 \
498  "psllq $2, %%mm3 \n\t"\
499  "psllq $2, %%mm4 \n\t"\
500 \
501  "por %%mm3, %%mm2 \n\t"\
502  "por %%mm4, %%mm1 \n\t"\
503 \
504  MOVNTQ(%%mm2, (dst, index, 2))\
505  MOVNTQ(%%mm1, 8(dst, index, 2))\
506 \
507  "add $8, "#index" \n\t"\
508  "cmp "dstw", "#index" \n\t"\
509  " jb 1b \n\t"
510 #define WRITERGB15(dst, dstw, index) REAL_WRITERGB15(dst, dstw, index)
511 
512 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
513  const int16_t **lumSrc, int lumFilterSize,
514  const int16_t *chrFilter, const int16_t **chrUSrc,
515  const int16_t **chrVSrc,
516  int chrFilterSize, const int16_t **alpSrc,
517  uint8_t *dest, int dstW, int dstY)
518 {
519  x86_reg dummy=0;
520  x86_reg dstW_reg = dstW;
521  x86_reg uv_off = c->uv_offx2;
522 
523  YSCALEYUV2PACKEDX_ACCURATE
524  YSCALEYUV2RGBX
525  "pxor %%mm7, %%mm7 \n\t"
526  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
527 #ifdef DITHER1XBPP
528  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
529  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
530  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
531 #endif
532  WRITERGB15(%4, "%5", %%REGa)
533  YSCALEYUV2PACKEDX_END
534 }
535 
536 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
537  const int16_t **lumSrc, int lumFilterSize,
538  const int16_t *chrFilter, const int16_t **chrUSrc,
539  const int16_t **chrVSrc,
540  int chrFilterSize, const int16_t **alpSrc,
541  uint8_t *dest, int dstW, int dstY)
542 {
543  x86_reg dummy=0;
544  x86_reg dstW_reg = dstW;
545  x86_reg uv_off = c->uv_offx2;
546 
547  YSCALEYUV2PACKEDX
548  YSCALEYUV2RGBX
549  "pxor %%mm7, %%mm7 \n\t"
550  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
551 #ifdef DITHER1XBPP
552  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
553  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
554  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
555 #endif
556  WRITERGB15(%4, "%5", %%REGa)
557  YSCALEYUV2PACKEDX_END
558 }
559 
560 #define WRITEBGR24MMX(dst, dstw, index) \
561  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
562  "movq %%mm2, %%mm1 \n\t" /* B */\
563  "movq %%mm5, %%mm6 \n\t" /* R */\
564  "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\
565  "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\
566  "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\
567  "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\
568  "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\
569  "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\
570  "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\
571  "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\
572  "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\
573  "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\
574 \
575  "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\
576  "movq %%mm2, %%mm6 \n\t" /* 0RGB0RGB 1 */\
577  "movq %%mm1, %%mm5 \n\t" /* 0RGB0RGB 2 */\
578  "movq %%mm3, %%mm7 \n\t" /* 0RGB0RGB 3 */\
579 \
580  "psllq $40, %%mm0 \n\t" /* RGB00000 0 */\
581  "psllq $40, %%mm2 \n\t" /* RGB00000 1 */\
582  "psllq $40, %%mm1 \n\t" /* RGB00000 2 */\
583  "psllq $40, %%mm3 \n\t" /* RGB00000 3 */\
584 \
585  "punpckhdq %%mm4, %%mm0 \n\t" /* 0RGBRGB0 0 */\
586  "punpckhdq %%mm6, %%mm2 \n\t" /* 0RGBRGB0 1 */\
587  "punpckhdq %%mm5, %%mm1 \n\t" /* 0RGBRGB0 2 */\
588  "punpckhdq %%mm7, %%mm3 \n\t" /* 0RGBRGB0 3 */\
589 \
590  "psrlq $8, %%mm0 \n\t" /* 00RGBRGB 0 */\
591  "movq %%mm2, %%mm6 \n\t" /* 0RGBRGB0 1 */\
592  "psllq $40, %%mm2 \n\t" /* GB000000 1 */\
593  "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\
594  MOVNTQ(%%mm0, (dst))\
595 \
596  "psrlq $24, %%mm6 \n\t" /* 0000RGBR 1 */\
597  "movq %%mm1, %%mm5 \n\t" /* 0RGBRGB0 2 */\
598  "psllq $24, %%mm1 \n\t" /* BRGB0000 2 */\
599  "por %%mm1, %%mm6 \n\t" /* BRGBRGBR 1 */\
600  MOVNTQ(%%mm6, 8(dst))\
601 \
602  "psrlq $40, %%mm5 \n\t" /* 000000RG 2 */\
603  "psllq $8, %%mm3 \n\t" /* RGBRGB00 3 */\
604  "por %%mm3, %%mm5 \n\t" /* RGBRGBRG 2 */\
605  MOVNTQ(%%mm5, 16(dst))\
606 \
607  "add $24, "#dst" \n\t"\
608 \
609  "add $8, "#index" \n\t"\
610  "cmp "dstw", "#index" \n\t"\
611  " jb 1b \n\t"
612 
613 #define WRITEBGR24MMXEXT(dst, dstw, index) \
614  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
615  "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
616  "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
617  "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2 B1 B0 B1 B0 */\
618  "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2 G1 G0 G1 G0 */\
619  "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0 R1 R0 R1 R0 */\
620 \
621  "pand %%mm0, %%mm1 \n\t" /* B2 B1 B0 */\
622  "pand %%mm0, %%mm3 \n\t" /* G2 G1 G0 */\
623  "pand %%mm7, %%mm6 \n\t" /* R1 R0 */\
624 \
625  "psllq $8, %%mm3 \n\t" /* G2 G1 G0 */\
626  "por %%mm1, %%mm6 \n\t"\
627  "por %%mm3, %%mm6 \n\t"\
628  MOVNTQ(%%mm6, (dst))\
629 \
630  "psrlq $8, %%mm4 \n\t" /* 00 G7 G6 G5 G4 G3 G2 G1 */\
631  "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4 B3 B2 B3 B2 */\
632  "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3 G4 G3 G4 G3 */\
633  "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4 R3 R2 R3 R2 */\
634 \
635  "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5 B4 B3 */\
636  "pand %%mm7, %%mm3 \n\t" /* G4 G3 */\
637  "pand %%mm0, %%mm6 \n\t" /* R4 R3 R2 */\
638 \
639  "por %%mm1, %%mm3 \n\t" /* B5 G4 B4 G3 B3 */\
640  "por %%mm3, %%mm6 \n\t"\
641  MOVNTQ(%%mm6, 8(dst))\
642 \
643  "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6 B7 B6 B6 B7 */\
644  "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7 G6 G5 G6 G5 */\
645  "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6 R5 R4 R5 R4 */\
646 \
647  "pand %%mm7, %%mm1 \n\t" /* B7 B6 */\
648  "pand %%mm0, %%mm3 \n\t" /* G7 G6 G5 */\
649  "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7 R6 R5 */\
650 \
651  "por %%mm1, %%mm3 \n\t"\
652  "por %%mm3, %%mm6 \n\t"\
653  MOVNTQ(%%mm6, 16(dst))\
654 \
655  "add $24, "#dst" \n\t"\
656 \
657  "add $8, "#index" \n\t"\
658  "cmp "dstw", "#index" \n\t"\
659  " jb 1b \n\t"
660 
661 #if COMPILE_TEMPLATE_MMXEXT
662 #undef WRITEBGR24
663 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMXEXT(dst, dstw, index)
664 #else
665 #undef WRITEBGR24
666 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX(dst, dstw, index)
667 #endif
668 
669 #if HAVE_6REGS
670 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
671  const int16_t **lumSrc, int lumFilterSize,
672  const int16_t *chrFilter, const int16_t **chrUSrc,
673  const int16_t **chrVSrc,
674  int chrFilterSize, const int16_t **alpSrc,
675  uint8_t *dest, int dstW, int dstY)
676 {
677  x86_reg dummy=0;
678  x86_reg dstW_reg = dstW;
679  x86_reg uv_off = c->uv_offx2;
680 
681  YSCALEYUV2PACKEDX_ACCURATE
682  YSCALEYUV2RGBX
683  "pxor %%mm7, %%mm7 \n\t"
684  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
685  "add %4, %%"REG_c" \n\t"
686  WRITEBGR24(%%REGc, "%5", %%REGa)
687  :: "r" (&c->redDither),
688  "m" (dummy), "m" (dummy), "m" (dummy),
689  "r" (dest), "m" (dstW_reg), "m"(uv_off)
690  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
691  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
692  );
693 }
694 
695 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
696  const int16_t **lumSrc, int lumFilterSize,
697  const int16_t *chrFilter, const int16_t **chrUSrc,
698  const int16_t **chrVSrc,
699  int chrFilterSize, const int16_t **alpSrc,
700  uint8_t *dest, int dstW, int dstY)
701 {
702  x86_reg dummy=0;
703  x86_reg dstW_reg = dstW;
704  x86_reg uv_off = c->uv_offx2;
705 
706  YSCALEYUV2PACKEDX
707  YSCALEYUV2RGBX
708  "pxor %%mm7, %%mm7 \n\t"
709  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c" \n\t" //FIXME optimize
710  "add %4, %%"REG_c" \n\t"
711  WRITEBGR24(%%REGc, "%5", %%REGa)
712  :: "r" (&c->redDither),
713  "m" (dummy), "m" (dummy), "m" (dummy),
714  "r" (dest), "m" (dstW_reg), "m"(uv_off)
715  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
716  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
717  );
718 }
719 #endif /* HAVE_6REGS */
720 
721 #define REAL_WRITEYUY2(dst, dstw, index) \
722  "packuswb %%mm3, %%mm3 \n\t"\
723  "packuswb %%mm4, %%mm4 \n\t"\
724  "packuswb %%mm7, %%mm1 \n\t"\
725  "punpcklbw %%mm4, %%mm3 \n\t"\
726  "movq %%mm1, %%mm7 \n\t"\
727  "punpcklbw %%mm3, %%mm1 \n\t"\
728  "punpckhbw %%mm3, %%mm7 \n\t"\
729 \
730  MOVNTQ(%%mm1, (dst, index, 2))\
731  MOVNTQ(%%mm7, 8(dst, index, 2))\
732 \
733  "add $8, "#index" \n\t"\
734  "cmp "dstw", "#index" \n\t"\
735  " jb 1b \n\t"
736 #define WRITEYUY2(dst, dstw, index) REAL_WRITEYUY2(dst, dstw, index)
737 
738 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
739  const int16_t **lumSrc, int lumFilterSize,
740  const int16_t *chrFilter, const int16_t **chrUSrc,
741  const int16_t **chrVSrc,
742  int chrFilterSize, const int16_t **alpSrc,
743  uint8_t *dest, int dstW, int dstY)
744 {
745  x86_reg dummy=0;
746  x86_reg dstW_reg = dstW;
747  x86_reg uv_off = c->uv_offx2;
748 
749  YSCALEYUV2PACKEDX_ACCURATE
750  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
751  "psraw $3, %%mm3 \n\t"
752  "psraw $3, %%mm4 \n\t"
753  "psraw $3, %%mm1 \n\t"
754  "psraw $3, %%mm7 \n\t"
755  WRITEYUY2(%4, "%5", %%REGa)
756  YSCALEYUV2PACKEDX_END
757 }
758 
759 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
760  const int16_t **lumSrc, int lumFilterSize,
761  const int16_t *chrFilter, const int16_t **chrUSrc,
762  const int16_t **chrVSrc,
763  int chrFilterSize, const int16_t **alpSrc,
764  uint8_t *dest, int dstW, int dstY)
765 {
766  x86_reg dummy=0;
767  x86_reg dstW_reg = dstW;
768  x86_reg uv_off = c->uv_offx2;
769 
770  YSCALEYUV2PACKEDX
771  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
772  "psraw $3, %%mm3 \n\t"
773  "psraw $3, %%mm4 \n\t"
774  "psraw $3, %%mm1 \n\t"
775  "psraw $3, %%mm7 \n\t"
776  WRITEYUY2(%4, "%5", %%REGa)
777  YSCALEYUV2PACKEDX_END
778 }
779 
780 #define REAL_YSCALEYUV2RGB_UV(index, c) \
781  "xor "#index", "#index" \n\t"\
782  ".p2align 4 \n\t"\
783  "1: \n\t"\
784  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
785  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
786  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
787  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
788  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
789  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
790  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
791  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
792  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
793  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
794  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
795  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
796  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
797  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
798  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
799  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
800  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
801  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
802  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
803  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
804  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
805  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
806 
807 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
808  "movq ("#b1", "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
809  "movq ("#b2", "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
810  "movq 8("#b1", "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
811  "movq 8("#b2", "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
812  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
813  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
814  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
815  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
816  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
817  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
818  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
819  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
820 
821 #define REAL_YSCALEYUV2RGB_COEFF(c) \
822  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
823  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
824  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
825  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
826  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
827  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
828  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
829  "paddw %%mm3, %%mm4 \n\t"\
830  "movq %%mm2, %%mm0 \n\t"\
831  "movq %%mm5, %%mm6 \n\t"\
832  "movq %%mm4, %%mm3 \n\t"\
833  "punpcklwd %%mm2, %%mm2 \n\t"\
834  "punpcklwd %%mm5, %%mm5 \n\t"\
835  "punpcklwd %%mm4, %%mm4 \n\t"\
836  "paddw %%mm1, %%mm2 \n\t"\
837  "paddw %%mm1, %%mm5 \n\t"\
838  "paddw %%mm1, %%mm4 \n\t"\
839  "punpckhwd %%mm0, %%mm0 \n\t"\
840  "punpckhwd %%mm6, %%mm6 \n\t"\
841  "punpckhwd %%mm3, %%mm3 \n\t"\
842  "paddw %%mm7, %%mm0 \n\t"\
843  "paddw %%mm7, %%mm6 \n\t"\
844  "paddw %%mm7, %%mm3 \n\t"\
845  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
846  "packuswb %%mm0, %%mm2 \n\t"\
847  "packuswb %%mm6, %%mm5 \n\t"\
848  "packuswb %%mm3, %%mm4 \n\t"\
849 
850 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
851 
852 #define YSCALEYUV2RGB(index, c) \
853  REAL_YSCALEYUV2RGB_UV(index, c) \
854  REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
855  REAL_YSCALEYUV2RGB_COEFF(c)
856 
857 /**
858  * vertical bilinear scale YV12 to RGB
859  */
860 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
861  const int16_t *ubuf[2], const int16_t *vbuf[2],
862  const int16_t *abuf[2], uint8_t *dest,
863  int dstW, int yalpha, int uvalpha, int y)
864 {
865  const int16_t *buf0 = buf[0], *buf1 = buf[1],
866  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
867 
868  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
869  const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
870 #if ARCH_X86_64
871  __asm__ volatile(
872  YSCALEYUV2RGB(%%r8, %5)
873  YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
874  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
875  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
876  "packuswb %%mm7, %%mm1 \n\t"
877  WRITEBGR32(%4, DSTW_OFFSET"(%5)", %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
878  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
879  "a" (&c->redDither),
880  "r" (abuf0), "r" (abuf1)
881  : "%r8"
882  );
883 #else
884  c->u_temp=(intptr_t)abuf0;
885  c->v_temp=(intptr_t)abuf1;
886  __asm__ volatile(
887  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
888  "mov %4, %%"REG_b" \n\t"
889  "push %%"REG_BP" \n\t"
890  YSCALEYUV2RGB(%%REGBP, %5)
891  "push %0 \n\t"
892  "push %1 \n\t"
893  "mov "U_TEMP"(%5), %0 \n\t"
894  "mov "V_TEMP"(%5), %1 \n\t"
895  YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
896  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
897  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
898  "packuswb %%mm7, %%mm1 \n\t"
899  "pop %1 \n\t"
900  "pop %0 \n\t"
901  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
902  "pop %%"REG_BP" \n\t"
903  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
904  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
905  "a" (&c->redDither)
906  );
907 #endif
908  } else {
909  __asm__ volatile(
910  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
911  "mov %4, %%"REG_b" \n\t"
912  "push %%"REG_BP" \n\t"
913  YSCALEYUV2RGB(%%REGBP, %5)
914  "pcmpeqd %%mm7, %%mm7 \n\t"
915  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
916  "pop %%"REG_BP" \n\t"
917  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
918  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
919  "a" (&c->redDither)
920  );
921  }
922 }
923 
924 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
925  const int16_t *ubuf[2], const int16_t *vbuf[2],
926  const int16_t *abuf[2], uint8_t *dest,
927  int dstW, int yalpha, int uvalpha, int y)
928 {
929  const int16_t *buf0 = buf[0], *buf1 = buf[1],
930  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
931 
932  __asm__ volatile(
933  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
934  "mov %4, %%"REG_b" \n\t"
935  "push %%"REG_BP" \n\t"
936  YSCALEYUV2RGB(%%REGBP, %5)
937  "pxor %%mm7, %%mm7 \n\t"
938  WRITEBGR24(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
939  "pop %%"REG_BP" \n\t"
940  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
941  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
942  "a" (&c->redDither)
943  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
944  );
945 }
946 
947 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
948  const int16_t *ubuf[2], const int16_t *vbuf[2],
949  const int16_t *abuf[2], uint8_t *dest,
950  int dstW, int yalpha, int uvalpha, int y)
951 {
952  const int16_t *buf0 = buf[0], *buf1 = buf[1],
953  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
954 
955  __asm__ volatile(
956  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
957  "mov %4, %%"REG_b" \n\t"
958  "push %%"REG_BP" \n\t"
959  YSCALEYUV2RGB(%%REGBP, %5)
960  "pxor %%mm7, %%mm7 \n\t"
961  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
962 #ifdef DITHER1XBPP
963  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
964  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
965  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
966 #endif
967  WRITERGB15(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
968  "pop %%"REG_BP" \n\t"
969  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
970  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
971  "a" (&c->redDither)
972  NAMED_CONSTRAINTS_ADD(bF8)
973  );
974 }
975 
976 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
977  const int16_t *ubuf[2], const int16_t *vbuf[2],
978  const int16_t *abuf[2], uint8_t *dest,
979  int dstW, int yalpha, int uvalpha, int y)
980 {
981  const int16_t *buf0 = buf[0], *buf1 = buf[1],
982  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
983 
984  __asm__ volatile(
985  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
986  "mov %4, %%"REG_b" \n\t"
987  "push %%"REG_BP" \n\t"
988  YSCALEYUV2RGB(%%REGBP, %5)
989  "pxor %%mm7, %%mm7 \n\t"
990  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
991 #ifdef DITHER1XBPP
992  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
993  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
994  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
995 #endif
996  WRITERGB16(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
997  "pop %%"REG_BP" \n\t"
998  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
999  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1000  "a" (&c->redDither)
1001  NAMED_CONSTRAINTS_ADD(bF8,bFC)
1002  );
1003 }
1004 
1005 #define REAL_YSCALEYUV2PACKED(index, c) \
1006  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
1007  "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1 \n\t"\
1008  "psraw $3, %%mm0 \n\t"\
1009  "psraw $3, %%mm1 \n\t"\
1010  "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1011  "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1012  "xor "#index", "#index" \n\t"\
1013  ".p2align 4 \n\t"\
1014  "1: \n\t"\
1015  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1016  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1017  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1018  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1019  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1020  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1021  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
1022  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
1023  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
1024  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
1025  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
1026  "psraw $7, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1027  "psraw $7, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1028  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
1029  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
1030  "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
1031  "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
1032  "movq 8(%0, "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
1033  "movq 8(%1, "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
1034  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
1035  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
1036  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1037  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1038  "psraw $7, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1039  "psraw $7, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1040  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1041  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1042 
1043 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c)
1044 
1045 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
1046  const int16_t *ubuf[2], const int16_t *vbuf[2],
1047  const int16_t *abuf[2], uint8_t *dest,
1048  int dstW, int yalpha, int uvalpha, int y)
1049 {
1050  const int16_t *buf0 = buf[0], *buf1 = buf[1],
1051  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1052 
1053  __asm__ volatile(
1054  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1055  "mov %4, %%"REG_b" \n\t"
1056  "push %%"REG_BP" \n\t"
1057  YSCALEYUV2PACKED(%%REGBP, %5)
1058  WRITEYUY2(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1059  "pop %%"REG_BP" \n\t"
1060  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1061  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1062  "a" (&c->redDither)
1063  );
1064 }
1065 
1066 #define REAL_YSCALEYUV2RGB1(index, c) \
1067  "xor "#index", "#index" \n\t"\
1068  ".p2align 4 \n\t"\
1069  "1: \n\t"\
1070  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1071  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1072  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1073  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1074  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1075  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1076  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1077  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1078  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1079  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1080  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1081  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1082  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1083  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1084  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1085  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1086  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1087  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1088  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1089  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1090  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1091  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1092  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1093  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1094  "paddw %%mm3, %%mm4 \n\t"\
1095  "movq %%mm2, %%mm0 \n\t"\
1096  "movq %%mm5, %%mm6 \n\t"\
1097  "movq %%mm4, %%mm3 \n\t"\
1098  "punpcklwd %%mm2, %%mm2 \n\t"\
1099  "punpcklwd %%mm5, %%mm5 \n\t"\
1100  "punpcklwd %%mm4, %%mm4 \n\t"\
1101  "paddw %%mm1, %%mm2 \n\t"\
1102  "paddw %%mm1, %%mm5 \n\t"\
1103  "paddw %%mm1, %%mm4 \n\t"\
1104  "punpckhwd %%mm0, %%mm0 \n\t"\
1105  "punpckhwd %%mm6, %%mm6 \n\t"\
1106  "punpckhwd %%mm3, %%mm3 \n\t"\
1107  "paddw %%mm7, %%mm0 \n\t"\
1108  "paddw %%mm7, %%mm6 \n\t"\
1109  "paddw %%mm7, %%mm3 \n\t"\
1110  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1111  "packuswb %%mm0, %%mm2 \n\t"\
1112  "packuswb %%mm6, %%mm5 \n\t"\
1113  "packuswb %%mm3, %%mm4 \n\t"\
1114 
1115 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c)
1116 
1117 // do vertical chrominance interpolation
1118 #define REAL_YSCALEYUV2RGB1b(index, c) \
1119  "xor "#index", "#index" \n\t"\
1120  ".p2align 4 \n\t"\
1121  "1: \n\t"\
1122  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1123  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1124  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1125  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1126  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1127  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1128  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1129  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1130  "psrlw $5, %%mm3 \n\t" /*FIXME might overflow*/\
1131  "psrlw $5, %%mm4 \n\t" /*FIXME might overflow*/\
1132  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1133  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1134  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1135  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1136  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1137  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1138  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1139  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1140  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1141  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1142  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1143  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1144  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1145  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1146  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1147  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1148  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1149  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1150  "paddw %%mm3, %%mm4 \n\t"\
1151  "movq %%mm2, %%mm0 \n\t"\
1152  "movq %%mm5, %%mm6 \n\t"\
1153  "movq %%mm4, %%mm3 \n\t"\
1154  "punpcklwd %%mm2, %%mm2 \n\t"\
1155  "punpcklwd %%mm5, %%mm5 \n\t"\
1156  "punpcklwd %%mm4, %%mm4 \n\t"\
1157  "paddw %%mm1, %%mm2 \n\t"\
1158  "paddw %%mm1, %%mm5 \n\t"\
1159  "paddw %%mm1, %%mm4 \n\t"\
1160  "punpckhwd %%mm0, %%mm0 \n\t"\
1161  "punpckhwd %%mm6, %%mm6 \n\t"\
1162  "punpckhwd %%mm3, %%mm3 \n\t"\
1163  "paddw %%mm7, %%mm0 \n\t"\
1164  "paddw %%mm7, %%mm6 \n\t"\
1165  "paddw %%mm7, %%mm3 \n\t"\
1166  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1167  "packuswb %%mm0, %%mm2 \n\t"\
1168  "packuswb %%mm6, %%mm5 \n\t"\
1169  "packuswb %%mm3, %%mm4 \n\t"\
1170 
1171 #define YSCALEYUV2RGB1b(index, c) REAL_YSCALEYUV2RGB1b(index, c)
1172 
1173 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1174  "movq (%1, "#index", 2), %%mm7 \n\t" /* abuf0[index ] */\
1175  "movq 8(%1, "#index", 2), %%mm1 \n\t" /* abuf0[index+4] */\
1176  "psraw $7, %%mm7 \n\t" /* abuf0[index ] >>7 */\
1177  "psraw $7, %%mm1 \n\t" /* abuf0[index+4] >>7 */\
1178  "packuswb %%mm1, %%mm7 \n\t"
1179 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1180 
1181 /**
1182  * YV12 to RGB without scaling or interpolating
1183  */
1184 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1185  const int16_t *ubuf[2], const int16_t *vbuf[2],
1186  const int16_t *abuf0, uint8_t *dest,
1187  int dstW, int uvalpha, int y)
1188 {
1189  const int16_t *ubuf0 = ubuf[0];
1190  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1191 
1192  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1193  const int16_t *ubuf1 = ubuf[0];
1194  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1195  __asm__ volatile(
1196  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1197  "mov %4, %%"REG_b" \n\t"
1198  "push %%"REG_BP" \n\t"
1199  YSCALEYUV2RGB1(%%REGBP, %5)
1200  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1201  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1202  "pop %%"REG_BP" \n\t"
1203  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1204  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1205  "a" (&c->redDither)
1206  );
1207  } else {
1208  __asm__ volatile(
1209  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1210  "mov %4, %%"REG_b" \n\t"
1211  "push %%"REG_BP" \n\t"
1212  YSCALEYUV2RGB1(%%REGBP, %5)
1213  "pcmpeqd %%mm7, %%mm7 \n\t"
1214  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1215  "pop %%"REG_BP" \n\t"
1216  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1217  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1218  "a" (&c->redDither)
1219  );
1220  }
1221  } else {
1222  const int16_t *ubuf1 = ubuf[1];
1223  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1224  __asm__ volatile(
1225  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1226  "mov %4, %%"REG_b" \n\t"
1227  "push %%"REG_BP" \n\t"
1228  YSCALEYUV2RGB1b(%%REGBP, %5)
1229  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1230  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1231  "pop %%"REG_BP" \n\t"
1232  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1233  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1234  "a" (&c->redDither)
1235  );
1236  } else {
1237  __asm__ volatile(
1238  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1239  "mov %4, %%"REG_b" \n\t"
1240  "push %%"REG_BP" \n\t"
1241  YSCALEYUV2RGB1b(%%REGBP, %5)
1242  "pcmpeqd %%mm7, %%mm7 \n\t"
1243  WRITEBGR32(%%REGb, DSTW_OFFSET"(%5)", %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1244  "pop %%"REG_BP" \n\t"
1245  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1246  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1247  "a" (&c->redDither)
1248  );
1249  }
1250  }
1251 }
1252 
1253 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1254  const int16_t *ubuf[2], const int16_t *vbuf[2],
1255  const int16_t *abuf0, uint8_t *dest,
1256  int dstW, int uvalpha, int y)
1257 {
1258  const int16_t *ubuf0 = ubuf[0];
1259  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1260 
1261  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1262  const int16_t *ubuf1 = ubuf[0];
1263  __asm__ volatile(
1264  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1265  "mov %4, %%"REG_b" \n\t"
1266  "push %%"REG_BP" \n\t"
1267  YSCALEYUV2RGB1(%%REGBP, %5)
1268  "pxor %%mm7, %%mm7 \n\t"
1269  WRITEBGR24(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1270  "pop %%"REG_BP" \n\t"
1271  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1272  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1273  "a" (&c->redDither)
1274  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
1275  );
1276  } else {
1277  const int16_t *ubuf1 = ubuf[1];
1278  __asm__ volatile(
1279  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1280  "mov %4, %%"REG_b" \n\t"
1281  "push %%"REG_BP" \n\t"
1282  YSCALEYUV2RGB1b(%%REGBP, %5)
1283  "pxor %%mm7, %%mm7 \n\t"
1284  WRITEBGR24(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1285  "pop %%"REG_BP" \n\t"
1286  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1287  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1288  "a" (&c->redDither)
1289  NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
1290  );
1291  }
1292 }
1293 
1294 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1295  const int16_t *ubuf[2], const int16_t *vbuf[2],
1296  const int16_t *abuf0, uint8_t *dest,
1297  int dstW, int uvalpha, int y)
1298 {
1299  const int16_t *ubuf0 = ubuf[0];
1300  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1301 
1302  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1303  const int16_t *ubuf1 = ubuf[0];
1304  __asm__ volatile(
1305  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1306  "mov %4, %%"REG_b" \n\t"
1307  "push %%"REG_BP" \n\t"
1308  YSCALEYUV2RGB1(%%REGBP, %5)
1309  "pxor %%mm7, %%mm7 \n\t"
1310  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1311 #ifdef DITHER1XBPP
1312  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1313  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1314  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1315 #endif
1316  WRITERGB15(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1317  "pop %%"REG_BP" \n\t"
1318  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1319  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1320  "a" (&c->redDither)
1321  NAMED_CONSTRAINTS_ADD(bF8)
1322  );
1323  } else {
1324  const int16_t *ubuf1 = ubuf[1];
1325  __asm__ volatile(
1326  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1327  "mov %4, %%"REG_b" \n\t"
1328  "push %%"REG_BP" \n\t"
1329  YSCALEYUV2RGB1b(%%REGBP, %5)
1330  "pxor %%mm7, %%mm7 \n\t"
1331  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1332 #ifdef DITHER1XBPP
1333  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1334  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1335  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1336 #endif
1337  WRITERGB15(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1338  "pop %%"REG_BP" \n\t"
1339  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1340  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1341  "a" (&c->redDither)
1342  NAMED_CONSTRAINTS_ADD(bF8)
1343  );
1344  }
1345 }
1346 
1347 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1348  const int16_t *ubuf[2], const int16_t *vbuf[2],
1349  const int16_t *abuf0, uint8_t *dest,
1350  int dstW, int uvalpha, int y)
1351 {
1352  const int16_t *ubuf0 = ubuf[0];
1353  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1354 
1355  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1356  const int16_t *ubuf1 = ubuf[0];
1357  __asm__ volatile(
1358  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1359  "mov %4, %%"REG_b" \n\t"
1360  "push %%"REG_BP" \n\t"
1361  YSCALEYUV2RGB1(%%REGBP, %5)
1362  "pxor %%mm7, %%mm7 \n\t"
1363  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1364 #ifdef DITHER1XBPP
1365  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1366  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1367  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1368 #endif
1369  WRITERGB16(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1370  "pop %%"REG_BP" \n\t"
1371  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1372  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1373  "a" (&c->redDither)
1374  NAMED_CONSTRAINTS_ADD(bF8,bFC)
1375  );
1376  } else {
1377  const int16_t *ubuf1 = ubuf[1];
1378  __asm__ volatile(
1379  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1380  "mov %4, %%"REG_b" \n\t"
1381  "push %%"REG_BP" \n\t"
1382  YSCALEYUV2RGB1b(%%REGBP, %5)
1383  "pxor %%mm7, %%mm7 \n\t"
1384  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1385 #ifdef DITHER1XBPP
1386  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1387  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1388  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1389 #endif
1390  WRITERGB16(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1391  "pop %%"REG_BP" \n\t"
1392  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1393  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1394  "a" (&c->redDither)
1395  NAMED_CONSTRAINTS_ADD(bF8,bFC)
1396  );
1397  }
1398 }
1399 
1400 #define REAL_YSCALEYUV2PACKED1(index, c) \
1401  "xor "#index", "#index" \n\t"\
1402  ".p2align 4 \n\t"\
1403  "1: \n\t"\
1404  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1405  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1406  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1407  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1408  "psraw $7, %%mm3 \n\t" \
1409  "psraw $7, %%mm4 \n\t" \
1410  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1411  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1412  "psraw $7, %%mm1 \n\t" \
1413  "psraw $7, %%mm7 \n\t" \
1414 
1415 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c)
1416 
1417 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1418  "xor "#index", "#index" \n\t"\
1419  ".p2align 4 \n\t"\
1420  "1: \n\t"\
1421  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1422  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1423  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1424  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1425  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1426  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1427  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1428  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1429  "psrlw $8, %%mm3 \n\t" \
1430  "psrlw $8, %%mm4 \n\t" \
1431  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1432  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1433  "psraw $7, %%mm1 \n\t" \
1434  "psraw $7, %%mm7 \n\t"
1435 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c)
1436 
1437 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1438  const int16_t *ubuf[2], const int16_t *vbuf[2],
1439  const int16_t *abuf0, uint8_t *dest,
1440  int dstW, int uvalpha, int y)
1441 {
1442  const int16_t *ubuf0 = ubuf[0];
1443  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1444 
1445  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1446  const int16_t *ubuf1 = ubuf[0];
1447  __asm__ volatile(
1448  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1449  "mov %4, %%"REG_b" \n\t"
1450  "push %%"REG_BP" \n\t"
1451  YSCALEYUV2PACKED1(%%REGBP, %5)
1452  WRITEYUY2(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1453  "pop %%"REG_BP" \n\t"
1454  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1455  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1456  "a" (&c->redDither)
1457  );
1458  } else {
1459  const int16_t *ubuf1 = ubuf[1];
1460  __asm__ volatile(
1461  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1462  "mov %4, %%"REG_b" \n\t"
1463  "push %%"REG_BP" \n\t"
1464  YSCALEYUV2PACKED1b(%%REGBP, %5)
1465  WRITEYUY2(%%REGb, DSTW_OFFSET"(%5)", %%REGBP)
1466  "pop %%"REG_BP" \n\t"
1467  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1468  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1469  "a" (&c->redDither)
1470  );
1471  }
1472 }
1473 
1474 #if COMPILE_TEMPLATE_MMXEXT
1475 static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
1476  int dstWidth, const uint8_t *src,
1477  int srcW, int xInc)
1478 {
1479  int32_t *filterPos = c->hLumFilterPos;
1480  int16_t *filter = c->hLumFilter;
1481  void *mmxextFilterCode = c->lumMmxextFilterCode;
1482  int i;
1483 #if defined(PIC)
1484  uint64_t ebxsave;
1485 #endif
1486 #if ARCH_X86_64
1487  uint64_t retsave;
1488 #endif
1489 
1490  __asm__ volatile(
1491 #if defined(PIC)
1492  "mov %%"REG_b", %5 \n\t"
1493 #if ARCH_X86_64
1494  "mov -8(%%rsp), %%"REG_a" \n\t"
1495  "mov %%"REG_a", %6 \n\t"
1496 #endif
1497 #else
1498 #if ARCH_X86_64
1499  "mov -8(%%rsp), %%"REG_a" \n\t"
1500  "mov %%"REG_a", %5 \n\t"
1501 #endif
1502 #endif
1503  "pxor %%mm7, %%mm7 \n\t"
1504  "mov %0, %%"REG_c" \n\t"
1505  "mov %1, %%"REG_D" \n\t"
1506  "mov %2, %%"REG_d" \n\t"
1507  "mov %3, %%"REG_b" \n\t"
1508  "xor %%"REG_a", %%"REG_a" \n\t" // i
1509  PREFETCH" (%%"REG_c") \n\t"
1510  PREFETCH" 32(%%"REG_c") \n\t"
1511  PREFETCH" 64(%%"REG_c") \n\t"
1512 
1513 #if ARCH_X86_64
1514 #define CALL_MMXEXT_FILTER_CODE \
1515  "movl (%%"REG_b"), %%esi \n\t"\
1516  "call *%4 \n\t"\
1517  "movl (%%"REG_b", %%"REG_a"), %%esi \n\t"\
1518  "add %%"REG_S", %%"REG_c" \n\t"\
1519  "add %%"REG_a", %%"REG_D" \n\t"\
1520  "xor %%"REG_a", %%"REG_a" \n\t"\
1521 
1522 #else
1523 #define CALL_MMXEXT_FILTER_CODE \
1524  "movl (%%"REG_b"), %%esi \n\t"\
1525  "call *%4 \n\t"\
1526  "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\
1527  "add %%"REG_a", %%"REG_D" \n\t"\
1528  "xor %%"REG_a", %%"REG_a" \n\t"\
1529 
1530 #endif /* ARCH_X86_64 */
1531 
1532  CALL_MMXEXT_FILTER_CODE
1533  CALL_MMXEXT_FILTER_CODE
1534  CALL_MMXEXT_FILTER_CODE
1535  CALL_MMXEXT_FILTER_CODE
1536  CALL_MMXEXT_FILTER_CODE
1537  CALL_MMXEXT_FILTER_CODE
1538  CALL_MMXEXT_FILTER_CODE
1539  CALL_MMXEXT_FILTER_CODE
1540 
1541 #if defined(PIC)
1542  "mov %5, %%"REG_b" \n\t"
1543 #if ARCH_X86_64
1544  "mov %6, %%"REG_a" \n\t"
1545  "mov %%"REG_a", -8(%%rsp) \n\t"
1546 #endif
1547 #else
1548 #if ARCH_X86_64
1549  "mov %5, %%"REG_a" \n\t"
1550  "mov %%"REG_a", -8(%%rsp) \n\t"
1551 #endif
1552 #endif
1553  :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),
1554  "m" (mmxextFilterCode)
1555 #if defined(PIC)
1556  ,"m" (ebxsave)
1557 #endif
1558 #if ARCH_X86_64
1559  ,"m"(retsave)
1560 #endif
1561  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1562 #if !defined(PIC)
1563  ,"%"REG_b
1564 #endif
1565  );
1566 
1567  for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
1568  dst[i] = src[srcW-1]*128;
1569 }
1570 
1571 static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
1572  int dstWidth, const uint8_t *src1,
1573  const uint8_t *src2, int srcW, int xInc)
1574 {
1575  int32_t *filterPos = c->hChrFilterPos;
1576  int16_t *filter = c->hChrFilter;
1577  void *mmxextFilterCode = c->chrMmxextFilterCode;
1578  int i;
1579 #if defined(PIC)
1580  DECLARE_ALIGNED(8, uint64_t, ebxsave);
1581 #endif
1582 #if ARCH_X86_64
1583  DECLARE_ALIGNED(8, uint64_t, retsave);
1584 #endif
1585 
1586  __asm__ volatile(
1587 #if defined(PIC)
1588  "mov %%"REG_b", %7 \n\t"
1589 #if ARCH_X86_64
1590  "mov -8(%%rsp), %%"REG_a" \n\t"
1591  "mov %%"REG_a", %8 \n\t"
1592 #endif
1593 #else
1594 #if ARCH_X86_64
1595  "mov -8(%%rsp), %%"REG_a" \n\t"
1596  "mov %%"REG_a", %7 \n\t"
1597 #endif
1598 #endif
1599  "pxor %%mm7, %%mm7 \n\t"
1600  "mov %0, %%"REG_c" \n\t"
1601  "mov %1, %%"REG_D" \n\t"
1602  "mov %2, %%"REG_d" \n\t"
1603  "mov %3, %%"REG_b" \n\t"
1604  "xor %%"REG_a", %%"REG_a" \n\t" // i
1605  PREFETCH" (%%"REG_c") \n\t"
1606  PREFETCH" 32(%%"REG_c") \n\t"
1607  PREFETCH" 64(%%"REG_c") \n\t"
1608 
1609  CALL_MMXEXT_FILTER_CODE
1610  CALL_MMXEXT_FILTER_CODE
1611  CALL_MMXEXT_FILTER_CODE
1612  CALL_MMXEXT_FILTER_CODE
1613  "xor %%"REG_a", %%"REG_a" \n\t" // i
1614  "mov %5, %%"REG_c" \n\t" // src
1615  "mov %6, %%"REG_D" \n\t" // buf2
1616  PREFETCH" (%%"REG_c") \n\t"
1617  PREFETCH" 32(%%"REG_c") \n\t"
1618  PREFETCH" 64(%%"REG_c") \n\t"
1619 
1620  CALL_MMXEXT_FILTER_CODE
1621  CALL_MMXEXT_FILTER_CODE
1622  CALL_MMXEXT_FILTER_CODE
1623  CALL_MMXEXT_FILTER_CODE
1624 
1625 #if defined(PIC)
1626  "mov %7, %%"REG_b" \n\t"
1627 #if ARCH_X86_64
1628  "mov %8, %%"REG_a" \n\t"
1629  "mov %%"REG_a", -8(%%rsp) \n\t"
1630 #endif
1631 #else
1632 #if ARCH_X86_64
1633  "mov %7, %%"REG_a" \n\t"
1634  "mov %%"REG_a", -8(%%rsp) \n\t"
1635 #endif
1636 #endif
1637  :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),
1638  "m" (mmxextFilterCode), "m" (src2), "m"(dst2)
1639 #if defined(PIC)
1640  ,"m" (ebxsave)
1641 #endif
1642 #if ARCH_X86_64
1643  ,"m"(retsave)
1644 #endif
1645  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1646 #if !defined(PIC)
1647  ,"%"REG_b
1648 #endif
1649  );
1650 
1651  for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
1652  dst1[i] = src1[srcW-1]*128;
1653  dst2[i] = src2[srcW-1]*128;
1654  }
1655 }
1656 #endif /* COMPILE_TEMPLATE_MMXEXT */
1657 
1658 static av_cold void RENAME(sws_init_swscale)(SwsContext *c)
1659 {
1660  enum AVPixelFormat dstFormat = c->dstFormat;
1661 
1662  c->use_mmx_vfilter= 0;
1663  if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) && dstFormat != AV_PIX_FMT_NV12
1664  && dstFormat != AV_PIX_FMT_NV21 && !(c->flags & SWS_BITEXACT)) {
1665  if (c->flags & SWS_ACCURATE_RND) {
1666  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1667  switch (c->dstFormat) {
1668  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X_ar); break;
1669 #if HAVE_6REGS
1670  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X_ar); break;
1671 #endif
1672  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X_ar); break;
1673  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X_ar); break;
1674  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
1675  default: break;
1676  }
1677  }
1678  } else {
1679  c->use_mmx_vfilter= 1;
1680  c->yuv2planeX = RENAME(yuv2yuvX );
1681  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1682  switch (c->dstFormat) {
1683  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X); break;
1684 #if HAVE_6REGS
1685  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X); break;
1686 #endif
1687  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X); break;
1688  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X); break;
1689  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
1690  default: break;
1691  }
1692  }
1693  }
1694  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1695  switch (c->dstFormat) {
1696  case AV_PIX_FMT_RGB32:
1697  c->yuv2packed1 = RENAME(yuv2rgb32_1);
1698  c->yuv2packed2 = RENAME(yuv2rgb32_2);
1699  break;
1700  case AV_PIX_FMT_BGR24:
1701  c->yuv2packed1 = RENAME(yuv2bgr24_1);
1702  c->yuv2packed2 = RENAME(yuv2bgr24_2);
1703  break;
1704  case AV_PIX_FMT_RGB555:
1705  c->yuv2packed1 = RENAME(yuv2rgb555_1);
1706  c->yuv2packed2 = RENAME(yuv2rgb555_2);
1707  break;
1708  case AV_PIX_FMT_RGB565:
1709  c->yuv2packed1 = RENAME(yuv2rgb565_1);
1710  c->yuv2packed2 = RENAME(yuv2rgb565_2);
1711  break;
1712  case AV_PIX_FMT_YUYV422:
1713  c->yuv2packed1 = RENAME(yuv2yuyv422_1);
1714  c->yuv2packed2 = RENAME(yuv2yuyv422_2);
1715  break;
1716  default:
1717  break;
1718  }
1719  }
1720  }
1721 
1722  if (c->srcBpc == 8 && c->dstBpc <= 14) {
1723  // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one).
1724 #if COMPILE_TEMPLATE_MMXEXT
1725  if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) {
1726  c->hyscale_fast = RENAME(hyscale_fast);
1727  c->hcscale_fast = RENAME(hcscale_fast);
1728  } else {
1729 #endif /* COMPILE_TEMPLATE_MMXEXT */
1730  c->hyscale_fast = NULL;
1731  c->hcscale_fast = NULL;
1732 #if COMPILE_TEMPLATE_MMXEXT
1733  }
1734 #endif /* COMPILE_TEMPLATE_MMXEXT */
1735  }
1736 }
Generated on Sun Jul 20 2014 23:06:09 for FFmpeg by   doxygen 1.8.2