swscale_template.c
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * This file is part of Libav.
5  *
6  * Libav is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * Libav is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with Libav; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 #undef REAL_MOVNTQ
22 #undef MOVNTQ
23 #undef PREFETCH
24 
25 #if COMPILE_TEMPLATE_MMXEXT
26 #define PREFETCH "prefetchnta"
27 #else
28 #define PREFETCH " # nop"
29 #endif
30 
31 #if COMPILE_TEMPLATE_MMXEXT
32 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
33 #else
34 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
35 #endif
36 #define MOVNTQ(a,b) REAL_MOVNTQ(a,b)
37 
38 #define YSCALEYUV2PACKEDX_UV \
39  __asm__ volatile(\
40  "xor %%"REG_a", %%"REG_a" \n\t"\
41  ".p2align 4 \n\t"\
42  "nop \n\t"\
43  "1: \n\t"\
44  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
45  "mov (%%"REG_d"), %%"REG_S" \n\t"\
46  "movq "VROUNDER_OFFSET"(%0), %%mm3 \n\t"\
47  "movq %%mm3, %%mm4 \n\t"\
48  ".p2align 4 \n\t"\
49  "2: \n\t"\
50  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
51  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\
52  "add %6, %%"REG_S" \n\t" \
53  "movq (%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\
54  "add $16, %%"REG_d" \n\t"\
55  "mov (%%"REG_d"), %%"REG_S" \n\t"\
56  "pmulhw %%mm0, %%mm2 \n\t"\
57  "pmulhw %%mm0, %%mm5 \n\t"\
58  "paddw %%mm2, %%mm3 \n\t"\
59  "paddw %%mm5, %%mm4 \n\t"\
60  "test %%"REG_S", %%"REG_S" \n\t"\
61  " jnz 2b \n\t"\
62 
63 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
64  "lea "offset"(%0), %%"REG_d" \n\t"\
65  "mov (%%"REG_d"), %%"REG_S" \n\t"\
66  "movq "VROUNDER_OFFSET"(%0), "#dst1" \n\t"\
67  "movq "#dst1", "#dst2" \n\t"\
68  ".p2align 4 \n\t"\
69  "2: \n\t"\
70  "movq 8(%%"REG_d"), "#coeff" \n\t" /* filterCoeff */\
71  "movq (%%"REG_S", %%"REG_a", 2), "#src1" \n\t" /* Y1srcData */\
72  "movq 8(%%"REG_S", %%"REG_a", 2), "#src2" \n\t" /* Y2srcData */\
73  "add $16, %%"REG_d" \n\t"\
74  "mov (%%"REG_d"), %%"REG_S" \n\t"\
75  "pmulhw "#coeff", "#src1" \n\t"\
76  "pmulhw "#coeff", "#src2" \n\t"\
77  "paddw "#src1", "#dst1" \n\t"\
78  "paddw "#src2", "#dst2" \n\t"\
79  "test %%"REG_S", %%"REG_S" \n\t"\
80  " jnz 2b \n\t"\
81 
82 #define YSCALEYUV2PACKEDX \
83  YSCALEYUV2PACKEDX_UV \
84  YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
85 
86 #define YSCALEYUV2PACKEDX_END \
87  :: "r" (&c->redDither), \
88  "m" (dummy), "m" (dummy), "m" (dummy),\
89  "r" (dest), "m" (dstW_reg), "m"(uv_off) \
90  : "%"REG_a, "%"REG_d, "%"REG_S \
91  );
92 
93 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
94  __asm__ volatile(\
95  "xor %%"REG_a", %%"REG_a" \n\t"\
96  ".p2align 4 \n\t"\
97  "nop \n\t"\
98  "1: \n\t"\
99  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
100  "mov (%%"REG_d"), %%"REG_S" \n\t"\
101  "pxor %%mm4, %%mm4 \n\t"\
102  "pxor %%mm5, %%mm5 \n\t"\
103  "pxor %%mm6, %%mm6 \n\t"\
104  "pxor %%mm7, %%mm7 \n\t"\
105  ".p2align 4 \n\t"\
106  "2: \n\t"\
107  "movq (%%"REG_S", %%"REG_a"), %%mm0 \n\t" /* UsrcData */\
108  "add %6, %%"REG_S" \n\t" \
109  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* VsrcData */\
110  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
111  "movq (%%"REG_S", %%"REG_a"), %%mm1 \n\t" /* UsrcData */\
112  "movq %%mm0, %%mm3 \n\t"\
113  "punpcklwd %%mm1, %%mm0 \n\t"\
114  "punpckhwd %%mm1, %%mm3 \n\t"\
115  "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1 \n\t" /* filterCoeff */\
116  "pmaddwd %%mm1, %%mm0 \n\t"\
117  "pmaddwd %%mm1, %%mm3 \n\t"\
118  "paddd %%mm0, %%mm4 \n\t"\
119  "paddd %%mm3, %%mm5 \n\t"\
120  "add %6, %%"REG_S" \n\t" \
121  "movq (%%"REG_S", %%"REG_a"), %%mm3 \n\t" /* VsrcData */\
122  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
123  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
124  "test %%"REG_S", %%"REG_S" \n\t"\
125  "movq %%mm2, %%mm0 \n\t"\
126  "punpcklwd %%mm3, %%mm2 \n\t"\
127  "punpckhwd %%mm3, %%mm0 \n\t"\
128  "pmaddwd %%mm1, %%mm2 \n\t"\
129  "pmaddwd %%mm1, %%mm0 \n\t"\
130  "paddd %%mm2, %%mm6 \n\t"\
131  "paddd %%mm0, %%mm7 \n\t"\
132  " jnz 2b \n\t"\
133  "psrad $16, %%mm4 \n\t"\
134  "psrad $16, %%mm5 \n\t"\
135  "psrad $16, %%mm6 \n\t"\
136  "psrad $16, %%mm7 \n\t"\
137  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
138  "packssdw %%mm5, %%mm4 \n\t"\
139  "packssdw %%mm7, %%mm6 \n\t"\
140  "paddw %%mm0, %%mm4 \n\t"\
141  "paddw %%mm0, %%mm6 \n\t"\
142  "movq %%mm4, "U_TEMP"(%0) \n\t"\
143  "movq %%mm6, "V_TEMP"(%0) \n\t"\
144 
145 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
146  "lea "offset"(%0), %%"REG_d" \n\t"\
147  "mov (%%"REG_d"), %%"REG_S" \n\t"\
148  "pxor %%mm1, %%mm1 \n\t"\
149  "pxor %%mm5, %%mm5 \n\t"\
150  "pxor %%mm7, %%mm7 \n\t"\
151  "pxor %%mm6, %%mm6 \n\t"\
152  ".p2align 4 \n\t"\
153  "2: \n\t"\
154  "movq (%%"REG_S", %%"REG_a", 2), %%mm0 \n\t" /* Y1srcData */\
155  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y2srcData */\
156  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
157  "movq (%%"REG_S", %%"REG_a", 2), %%mm4 \n\t" /* Y1srcData */\
158  "movq %%mm0, %%mm3 \n\t"\
159  "punpcklwd %%mm4, %%mm0 \n\t"\
160  "punpckhwd %%mm4, %%mm3 \n\t"\
161  "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4 \n\t" /* filterCoeff */\
162  "pmaddwd %%mm4, %%mm0 \n\t"\
163  "pmaddwd %%mm4, %%mm3 \n\t"\
164  "paddd %%mm0, %%mm1 \n\t"\
165  "paddd %%mm3, %%mm5 \n\t"\
166  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3 \n\t" /* Y2srcData */\
167  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
168  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
169  "test %%"REG_S", %%"REG_S" \n\t"\
170  "movq %%mm2, %%mm0 \n\t"\
171  "punpcklwd %%mm3, %%mm2 \n\t"\
172  "punpckhwd %%mm3, %%mm0 \n\t"\
173  "pmaddwd %%mm4, %%mm2 \n\t"\
174  "pmaddwd %%mm4, %%mm0 \n\t"\
175  "paddd %%mm2, %%mm7 \n\t"\
176  "paddd %%mm0, %%mm6 \n\t"\
177  " jnz 2b \n\t"\
178  "psrad $16, %%mm1 \n\t"\
179  "psrad $16, %%mm5 \n\t"\
180  "psrad $16, %%mm7 \n\t"\
181  "psrad $16, %%mm6 \n\t"\
182  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
183  "packssdw %%mm5, %%mm1 \n\t"\
184  "packssdw %%mm6, %%mm7 \n\t"\
185  "paddw %%mm0, %%mm1 \n\t"\
186  "paddw %%mm0, %%mm7 \n\t"\
187  "movq "U_TEMP"(%0), %%mm3 \n\t"\
188  "movq "V_TEMP"(%0), %%mm4 \n\t"\
189 
190 #define YSCALEYUV2PACKEDX_ACCURATE \
191  YSCALEYUV2PACKEDX_ACCURATE_UV \
192  YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
193 
194 #define YSCALEYUV2RGBX \
195  "psubw "U_OFFSET"(%0), %%mm3 \n\t" /* (U-128)8*/\
196  "psubw "V_OFFSET"(%0), %%mm4 \n\t" /* (V-128)8*/\
197  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
198  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
199  "pmulhw "UG_COEFF"(%0), %%mm3 \n\t"\
200  "pmulhw "VG_COEFF"(%0), %%mm4 \n\t"\
201  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
202  "pmulhw "UB_COEFF"(%0), %%mm2 \n\t"\
203  "pmulhw "VR_COEFF"(%0), %%mm5 \n\t"\
204  "psubw "Y_OFFSET"(%0), %%mm1 \n\t" /* 8(Y-16)*/\
205  "psubw "Y_OFFSET"(%0), %%mm7 \n\t" /* 8(Y-16)*/\
206  "pmulhw "Y_COEFF"(%0), %%mm1 \n\t"\
207  "pmulhw "Y_COEFF"(%0), %%mm7 \n\t"\
208  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
209  "paddw %%mm3, %%mm4 \n\t"\
210  "movq %%mm2, %%mm0 \n\t"\
211  "movq %%mm5, %%mm6 \n\t"\
212  "movq %%mm4, %%mm3 \n\t"\
213  "punpcklwd %%mm2, %%mm2 \n\t"\
214  "punpcklwd %%mm5, %%mm5 \n\t"\
215  "punpcklwd %%mm4, %%mm4 \n\t"\
216  "paddw %%mm1, %%mm2 \n\t"\
217  "paddw %%mm1, %%mm5 \n\t"\
218  "paddw %%mm1, %%mm4 \n\t"\
219  "punpckhwd %%mm0, %%mm0 \n\t"\
220  "punpckhwd %%mm6, %%mm6 \n\t"\
221  "punpckhwd %%mm3, %%mm3 \n\t"\
222  "paddw %%mm7, %%mm0 \n\t"\
223  "paddw %%mm7, %%mm6 \n\t"\
224  "paddw %%mm7, %%mm3 \n\t"\
225  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
226  "packuswb %%mm0, %%mm2 \n\t"\
227  "packuswb %%mm6, %%mm5 \n\t"\
228  "packuswb %%mm3, %%mm4 \n\t"\
229 
230 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
231  "movq "#b", "#q2" \n\t" /* B */\
232  "movq "#r", "#t" \n\t" /* R */\
233  "punpcklbw "#g", "#b" \n\t" /* GBGBGBGB 0 */\
234  "punpcklbw "#a", "#r" \n\t" /* ARARARAR 0 */\
235  "punpckhbw "#g", "#q2" \n\t" /* GBGBGBGB 2 */\
236  "punpckhbw "#a", "#t" \n\t" /* ARARARAR 2 */\
237  "movq "#b", "#q0" \n\t" /* GBGBGBGB 0 */\
238  "movq "#q2", "#q3" \n\t" /* GBGBGBGB 2 */\
239  "punpcklwd "#r", "#q0" \n\t" /* ARGBARGB 0 */\
240  "punpckhwd "#r", "#b" \n\t" /* ARGBARGB 1 */\
241  "punpcklwd "#t", "#q2" \n\t" /* ARGBARGB 2 */\
242  "punpckhwd "#t", "#q3" \n\t" /* ARGBARGB 3 */\
243 \
244  MOVNTQ( q0, (dst, index, 4))\
245  MOVNTQ( b, 8(dst, index, 4))\
246  MOVNTQ( q2, 16(dst, index, 4))\
247  MOVNTQ( q3, 24(dst, index, 4))\
248 \
249  "add $8, "#index" \n\t"\
250  "cmp "#dstw", "#index" \n\t"\
251  " jb 1b \n\t"
252 #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)
253 
254 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
255  const int16_t **lumSrc, int lumFilterSize,
256  const int16_t *chrFilter, const int16_t **chrUSrc,
257  const int16_t **chrVSrc,
258  int chrFilterSize, const int16_t **alpSrc,
259  uint8_t *dest, int dstW, int dstY)
260 {
261  x86_reg dummy=0;
262  x86_reg dstW_reg = dstW;
263  x86_reg uv_off = c->uv_off_byte;
264 
265  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
268  "movq %%mm2, "U_TEMP"(%0) \n\t"
269  "movq %%mm4, "V_TEMP"(%0) \n\t"
270  "movq %%mm5, "Y_TEMP"(%0) \n\t"
272  "movq "Y_TEMP"(%0), %%mm5 \n\t"
273  "psraw $3, %%mm1 \n\t"
274  "psraw $3, %%mm7 \n\t"
275  "packuswb %%mm7, %%mm1 \n\t"
276  WRITEBGR32(%4, %5, %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
278  } else {
281  "pcmpeqd %%mm7, %%mm7 \n\t"
282  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
284  }
285 }
286 
287 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
288  const int16_t **lumSrc, int lumFilterSize,
289  const int16_t *chrFilter, const int16_t **chrUSrc,
290  const int16_t **chrVSrc,
291  int chrFilterSize, const int16_t **alpSrc,
292  uint8_t *dest, int dstW, int dstY)
293 {
294  x86_reg dummy=0;
295  x86_reg dstW_reg = dstW;
296  x86_reg uv_off = c->uv_off_byte;
297 
298  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
301  YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
302  "psraw $3, %%mm1 \n\t"
303  "psraw $3, %%mm7 \n\t"
304  "packuswb %%mm7, %%mm1 \n\t"
305  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
307  } else {
310  "pcmpeqd %%mm7, %%mm7 \n\t"
311  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
313  }
314 }
315 
316 #define REAL_WRITERGB16(dst, dstw, index) \
317  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
318  "pand "MANGLE(bFC)", %%mm4 \n\t" /* G */\
319  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
320  "psrlq $3, %%mm2 \n\t"\
321 \
322  "movq %%mm2, %%mm1 \n\t"\
323  "movq %%mm4, %%mm3 \n\t"\
324 \
325  "punpcklbw %%mm7, %%mm3 \n\t"\
326  "punpcklbw %%mm5, %%mm2 \n\t"\
327  "punpckhbw %%mm7, %%mm4 \n\t"\
328  "punpckhbw %%mm5, %%mm1 \n\t"\
329 \
330  "psllq $3, %%mm3 \n\t"\
331  "psllq $3, %%mm4 \n\t"\
332 \
333  "por %%mm3, %%mm2 \n\t"\
334  "por %%mm4, %%mm1 \n\t"\
335 \
336  MOVNTQ(%%mm2, (dst, index, 2))\
337  MOVNTQ(%%mm1, 8(dst, index, 2))\
338 \
339  "add $8, "#index" \n\t"\
340  "cmp "#dstw", "#index" \n\t"\
341  " jb 1b \n\t"
342 #define WRITERGB16(dst, dstw, index) REAL_WRITERGB16(dst, dstw, index)
343 
344 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
345  const int16_t **lumSrc, int lumFilterSize,
346  const int16_t *chrFilter, const int16_t **chrUSrc,
347  const int16_t **chrVSrc,
348  int chrFilterSize, const int16_t **alpSrc,
349  uint8_t *dest, int dstW, int dstY)
350 {
351  x86_reg dummy=0;
352  x86_reg dstW_reg = dstW;
353  x86_reg uv_off = c->uv_off_byte;
354 
357  "pxor %%mm7, %%mm7 \n\t"
358  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
359 #ifdef DITHER1XBPP
360  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
361  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
362  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
363 #endif
364  WRITERGB16(%4, %5, %%REGa)
366 }
367 
368 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
369  const int16_t **lumSrc, int lumFilterSize,
370  const int16_t *chrFilter, const int16_t **chrUSrc,
371  const int16_t **chrVSrc,
372  int chrFilterSize, const int16_t **alpSrc,
373  uint8_t *dest, int dstW, int dstY)
374 {
375  x86_reg dummy=0;
376  x86_reg dstW_reg = dstW;
377  x86_reg uv_off = c->uv_off_byte;
378 
381  "pxor %%mm7, %%mm7 \n\t"
382  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
383 #ifdef DITHER1XBPP
384  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
385  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
386  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
387 #endif
388  WRITERGB16(%4, %5, %%REGa)
390 }
391 
392 #define REAL_WRITERGB15(dst, dstw, index) \
393  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
394  "pand "MANGLE(bF8)", %%mm4 \n\t" /* G */\
395  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
396  "psrlq $3, %%mm2 \n\t"\
397  "psrlq $1, %%mm5 \n\t"\
398 \
399  "movq %%mm2, %%mm1 \n\t"\
400  "movq %%mm4, %%mm3 \n\t"\
401 \
402  "punpcklbw %%mm7, %%mm3 \n\t"\
403  "punpcklbw %%mm5, %%mm2 \n\t"\
404  "punpckhbw %%mm7, %%mm4 \n\t"\
405  "punpckhbw %%mm5, %%mm1 \n\t"\
406 \
407  "psllq $2, %%mm3 \n\t"\
408  "psllq $2, %%mm4 \n\t"\
409 \
410  "por %%mm3, %%mm2 \n\t"\
411  "por %%mm4, %%mm1 \n\t"\
412 \
413  MOVNTQ(%%mm2, (dst, index, 2))\
414  MOVNTQ(%%mm1, 8(dst, index, 2))\
415 \
416  "add $8, "#index" \n\t"\
417  "cmp "#dstw", "#index" \n\t"\
418  " jb 1b \n\t"
419 #define WRITERGB15(dst, dstw, index) REAL_WRITERGB15(dst, dstw, index)
420 
421 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
422  const int16_t **lumSrc, int lumFilterSize,
423  const int16_t *chrFilter, const int16_t **chrUSrc,
424  const int16_t **chrVSrc,
425  int chrFilterSize, const int16_t **alpSrc,
426  uint8_t *dest, int dstW, int dstY)
427 {
428  x86_reg dummy=0;
429  x86_reg dstW_reg = dstW;
430  x86_reg uv_off = c->uv_off_byte;
431 
434  "pxor %%mm7, %%mm7 \n\t"
435  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
436 #ifdef DITHER1XBPP
437  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
438  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
439  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
440 #endif
441  WRITERGB15(%4, %5, %%REGa)
443 }
444 
445 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
446  const int16_t **lumSrc, int lumFilterSize,
447  const int16_t *chrFilter, const int16_t **chrUSrc,
448  const int16_t **chrVSrc,
449  int chrFilterSize, const int16_t **alpSrc,
450  uint8_t *dest, int dstW, int dstY)
451 {
452  x86_reg dummy=0;
453  x86_reg dstW_reg = dstW;
454  x86_reg uv_off = c->uv_off_byte;
455 
458  "pxor %%mm7, %%mm7 \n\t"
459  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
460 #ifdef DITHER1XBPP
461  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
462  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
463  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
464 #endif
465  WRITERGB15(%4, %5, %%REGa)
467 }
468 
469 #define WRITEBGR24MMX(dst, dstw, index) \
470  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
471  "movq %%mm2, %%mm1 \n\t" /* B */\
472  "movq %%mm5, %%mm6 \n\t" /* R */\
473  "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\
474  "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\
475  "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\
476  "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\
477  "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\
478  "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\
479  "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\
480  "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\
481  "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\
482  "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\
483 \
484  "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\
485  "movq %%mm2, %%mm6 \n\t" /* 0RGB0RGB 1 */\
486  "movq %%mm1, %%mm5 \n\t" /* 0RGB0RGB 2 */\
487  "movq %%mm3, %%mm7 \n\t" /* 0RGB0RGB 3 */\
488 \
489  "psllq $40, %%mm0 \n\t" /* RGB00000 0 */\
490  "psllq $40, %%mm2 \n\t" /* RGB00000 1 */\
491  "psllq $40, %%mm1 \n\t" /* RGB00000 2 */\
492  "psllq $40, %%mm3 \n\t" /* RGB00000 3 */\
493 \
494  "punpckhdq %%mm4, %%mm0 \n\t" /* 0RGBRGB0 0 */\
495  "punpckhdq %%mm6, %%mm2 \n\t" /* 0RGBRGB0 1 */\
496  "punpckhdq %%mm5, %%mm1 \n\t" /* 0RGBRGB0 2 */\
497  "punpckhdq %%mm7, %%mm3 \n\t" /* 0RGBRGB0 3 */\
498 \
499  "psrlq $8, %%mm0 \n\t" /* 00RGBRGB 0 */\
500  "movq %%mm2, %%mm6 \n\t" /* 0RGBRGB0 1 */\
501  "psllq $40, %%mm2 \n\t" /* GB000000 1 */\
502  "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\
503  MOVNTQ(%%mm0, (dst))\
504 \
505  "psrlq $24, %%mm6 \n\t" /* 0000RGBR 1 */\
506  "movq %%mm1, %%mm5 \n\t" /* 0RGBRGB0 2 */\
507  "psllq $24, %%mm1 \n\t" /* BRGB0000 2 */\
508  "por %%mm1, %%mm6 \n\t" /* BRGBRGBR 1 */\
509  MOVNTQ(%%mm6, 8(dst))\
510 \
511  "psrlq $40, %%mm5 \n\t" /* 000000RG 2 */\
512  "psllq $8, %%mm3 \n\t" /* RGBRGB00 3 */\
513  "por %%mm3, %%mm5 \n\t" /* RGBRGBRG 2 */\
514  MOVNTQ(%%mm5, 16(dst))\
515 \
516  "add $24, "#dst" \n\t"\
517 \
518  "add $8, "#index" \n\t"\
519  "cmp "#dstw", "#index" \n\t"\
520  " jb 1b \n\t"
521 
522 #define WRITEBGR24MMXEXT(dst, dstw, index) \
523  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
524  "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
525  "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
526  "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2 B1 B0 B1 B0 */\
527  "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2 G1 G0 G1 G0 */\
528  "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0 R1 R0 R1 R0 */\
529 \
530  "pand %%mm0, %%mm1 \n\t" /* B2 B1 B0 */\
531  "pand %%mm0, %%mm3 \n\t" /* G2 G1 G0 */\
532  "pand %%mm7, %%mm6 \n\t" /* R1 R0 */\
533 \
534  "psllq $8, %%mm3 \n\t" /* G2 G1 G0 */\
535  "por %%mm1, %%mm6 \n\t"\
536  "por %%mm3, %%mm6 \n\t"\
537  MOVNTQ(%%mm6, (dst))\
538 \
539  "psrlq $8, %%mm4 \n\t" /* 00 G7 G6 G5 G4 G3 G2 G1 */\
540  "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4 B3 B2 B3 B2 */\
541  "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3 G4 G3 G4 G3 */\
542  "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4 R3 R2 R3 R2 */\
543 \
544  "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5 B4 B3 */\
545  "pand %%mm7, %%mm3 \n\t" /* G4 G3 */\
546  "pand %%mm0, %%mm6 \n\t" /* R4 R3 R2 */\
547 \
548  "por %%mm1, %%mm3 \n\t" /* B5 G4 B4 G3 B3 */\
549  "por %%mm3, %%mm6 \n\t"\
550  MOVNTQ(%%mm6, 8(dst))\
551 \
552  "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6 B7 B6 B6 B7 */\
553  "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7 G6 G5 G6 G5 */\
554  "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6 R5 R4 R5 R4 */\
555 \
556  "pand %%mm7, %%mm1 \n\t" /* B7 B6 */\
557  "pand %%mm0, %%mm3 \n\t" /* G7 G6 G5 */\
558  "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7 R6 R5 */\
559 \
560  "por %%mm1, %%mm3 \n\t"\
561  "por %%mm3, %%mm6 \n\t"\
562  MOVNTQ(%%mm6, 16(dst))\
563 \
564  "add $24, "#dst" \n\t"\
565 \
566  "add $8, "#index" \n\t"\
567  "cmp "#dstw", "#index" \n\t"\
568  " jb 1b \n\t"
569 
570 #if COMPILE_TEMPLATE_MMXEXT
571 #undef WRITEBGR24
572 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMXEXT(dst, dstw, index)
573 #else
574 #undef WRITEBGR24
575 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX(dst, dstw, index)
576 #endif
577 
578 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
579  const int16_t **lumSrc, int lumFilterSize,
580  const int16_t *chrFilter, const int16_t **chrUSrc,
581  const int16_t **chrVSrc,
582  int chrFilterSize, const int16_t **alpSrc,
583  uint8_t *dest, int dstW, int dstY)
584 {
585  x86_reg dummy=0;
586  x86_reg dstW_reg = dstW;
587  x86_reg uv_off = c->uv_off_byte;
588 
591  "pxor %%mm7, %%mm7 \n\t"
592  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
593  "add %4, %%"REG_c" \n\t"
594  WRITEBGR24(%%REGc, %5, %%REGa)
595  :: "r" (&c->redDither),
596  "m" (dummy), "m" (dummy), "m" (dummy),
597  "r" (dest), "m" (dstW_reg), "m"(uv_off)
598  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
599  );
600 }
601 
602 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
603  const int16_t **lumSrc, int lumFilterSize,
604  const int16_t *chrFilter, const int16_t **chrUSrc,
605  const int16_t **chrVSrc,
606  int chrFilterSize, const int16_t **alpSrc,
607  uint8_t *dest, int dstW, int dstY)
608 {
609  x86_reg dummy=0;
610  x86_reg dstW_reg = dstW;
611  x86_reg uv_off = c->uv_off_byte;
612 
615  "pxor %%mm7, %%mm7 \n\t"
616  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c" \n\t" //FIXME optimize
617  "add %4, %%"REG_c" \n\t"
618  WRITEBGR24(%%REGc, %5, %%REGa)
619  :: "r" (&c->redDither),
620  "m" (dummy), "m" (dummy), "m" (dummy),
621  "r" (dest), "m" (dstW_reg), "m"(uv_off)
622  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
623  );
624 }
625 
626 #define REAL_WRITEYUY2(dst, dstw, index) \
627  "packuswb %%mm3, %%mm3 \n\t"\
628  "packuswb %%mm4, %%mm4 \n\t"\
629  "packuswb %%mm7, %%mm1 \n\t"\
630  "punpcklbw %%mm4, %%mm3 \n\t"\
631  "movq %%mm1, %%mm7 \n\t"\
632  "punpcklbw %%mm3, %%mm1 \n\t"\
633  "punpckhbw %%mm3, %%mm7 \n\t"\
634 \
635  MOVNTQ(%%mm1, (dst, index, 2))\
636  MOVNTQ(%%mm7, 8(dst, index, 2))\
637 \
638  "add $8, "#index" \n\t"\
639  "cmp "#dstw", "#index" \n\t"\
640  " jb 1b \n\t"
641 #define WRITEYUY2(dst, dstw, index) REAL_WRITEYUY2(dst, dstw, index)
642 
643 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
644  const int16_t **lumSrc, int lumFilterSize,
645  const int16_t *chrFilter, const int16_t **chrUSrc,
646  const int16_t **chrVSrc,
647  int chrFilterSize, const int16_t **alpSrc,
648  uint8_t *dest, int dstW, int dstY)
649 {
650  x86_reg dummy=0;
651  x86_reg dstW_reg = dstW;
652  x86_reg uv_off = c->uv_off_byte;
653 
655  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
656  "psraw $3, %%mm3 \n\t"
657  "psraw $3, %%mm4 \n\t"
658  "psraw $3, %%mm1 \n\t"
659  "psraw $3, %%mm7 \n\t"
660  WRITEYUY2(%4, %5, %%REGa)
662 }
663 
664 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
665  const int16_t **lumSrc, int lumFilterSize,
666  const int16_t *chrFilter, const int16_t **chrUSrc,
667  const int16_t **chrVSrc,
668  int chrFilterSize, const int16_t **alpSrc,
669  uint8_t *dest, int dstW, int dstY)
670 {
671  x86_reg dummy=0;
672  x86_reg dstW_reg = dstW;
673  x86_reg uv_off = c->uv_off_byte;
674 
676  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
677  "psraw $3, %%mm3 \n\t"
678  "psraw $3, %%mm4 \n\t"
679  "psraw $3, %%mm1 \n\t"
680  "psraw $3, %%mm7 \n\t"
681  WRITEYUY2(%4, %5, %%REGa)
683 }
684 
685 #define REAL_YSCALEYUV2RGB_UV(index, c) \
686  "xor "#index", "#index" \n\t"\
687  ".p2align 4 \n\t"\
688  "1: \n\t"\
689  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
690  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
691  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
692  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
693  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
694  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
695  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
696  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
697  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
698  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
699  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
700  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
701  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
702  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
703  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
704  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
705  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
706  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
707  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
708  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
709  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
710  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
711 
712 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
713  "movq ("#b1", "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
714  "movq ("#b2", "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
715  "movq 8("#b1", "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
716  "movq 8("#b2", "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
717  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
718  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
719  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
720  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
721  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
722  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
723  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
724  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
725 
726 #define REAL_YSCALEYUV2RGB_COEFF(c) \
727  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
728  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
729  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
730  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
731  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
732  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
733  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
734  "paddw %%mm3, %%mm4 \n\t"\
735  "movq %%mm2, %%mm0 \n\t"\
736  "movq %%mm5, %%mm6 \n\t"\
737  "movq %%mm4, %%mm3 \n\t"\
738  "punpcklwd %%mm2, %%mm2 \n\t"\
739  "punpcklwd %%mm5, %%mm5 \n\t"\
740  "punpcklwd %%mm4, %%mm4 \n\t"\
741  "paddw %%mm1, %%mm2 \n\t"\
742  "paddw %%mm1, %%mm5 \n\t"\
743  "paddw %%mm1, %%mm4 \n\t"\
744  "punpckhwd %%mm0, %%mm0 \n\t"\
745  "punpckhwd %%mm6, %%mm6 \n\t"\
746  "punpckhwd %%mm3, %%mm3 \n\t"\
747  "paddw %%mm7, %%mm0 \n\t"\
748  "paddw %%mm7, %%mm6 \n\t"\
749  "paddw %%mm7, %%mm3 \n\t"\
750  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
751  "packuswb %%mm0, %%mm2 \n\t"\
752  "packuswb %%mm6, %%mm5 \n\t"\
753  "packuswb %%mm3, %%mm4 \n\t"\
754 
755 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
756 
757 #define YSCALEYUV2RGB(index, c) \
758  REAL_YSCALEYUV2RGB_UV(index, c) \
759  REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
760  REAL_YSCALEYUV2RGB_COEFF(c)
761 
765 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
766  const int16_t *ubuf[2], const int16_t *vbuf[2],
767  const int16_t *abuf[2], uint8_t *dest,
768  int dstW, int yalpha, int uvalpha, int y)
769 {
770  const int16_t *buf0 = buf[0], *buf1 = buf[1],
771  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
772 
773  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
774  const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
775 #if ARCH_X86_64
776  __asm__ volatile(
777  YSCALEYUV2RGB(%%r8, %5)
778  YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
779  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
780  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
781  "packuswb %%mm7, %%mm1 \n\t"
782  WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
783  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
784  "a" (&c->redDither),
785  "r" (abuf0), "r" (abuf1)
786  : "%r8"
787  );
788 #else
789  *(const uint16_t **)(&c->u_temp)=abuf0;
790  *(const uint16_t **)(&c->v_temp)=abuf1;
791  __asm__ volatile(
792  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
793  "mov %4, %%"REG_b" \n\t"
794  "push %%"REG_BP" \n\t"
795  YSCALEYUV2RGB(%%REGBP, %5)
796  "push %0 \n\t"
797  "push %1 \n\t"
798  "mov "U_TEMP"(%5), %0 \n\t"
799  "mov "V_TEMP"(%5), %1 \n\t"
800  YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
801  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
802  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
803  "packuswb %%mm7, %%mm1 \n\t"
804  "pop %1 \n\t"
805  "pop %0 \n\t"
806  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
807  "pop %%"REG_BP" \n\t"
808  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
809  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
810  "a" (&c->redDither)
811  );
812 #endif
813  } else {
814  __asm__ volatile(
815  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
816  "mov %4, %%"REG_b" \n\t"
817  "push %%"REG_BP" \n\t"
818  YSCALEYUV2RGB(%%REGBP, %5)
819  "pcmpeqd %%mm7, %%mm7 \n\t"
820  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
821  "pop %%"REG_BP" \n\t"
822  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
823  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
824  "a" (&c->redDither)
825  );
826  }
827 }
828 
829 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
830  const int16_t *ubuf[2], const int16_t *vbuf[2],
831  const int16_t *abuf[2], uint8_t *dest,
832  int dstW, int yalpha, int uvalpha, int y)
833 {
834  const int16_t *buf0 = buf[0], *buf1 = buf[1],
835  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
836 
837  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
838  __asm__ volatile(
839  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
840  "mov %4, %%"REG_b" \n\t"
841  "push %%"REG_BP" \n\t"
842  YSCALEYUV2RGB(%%REGBP, %5)
843  "pxor %%mm7, %%mm7 \n\t"
844  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
845  "pop %%"REG_BP" \n\t"
846  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
847  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
848  "a" (&c->redDither)
849  );
850 }
851 
852 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
853  const int16_t *ubuf[2], const int16_t *vbuf[2],
854  const int16_t *abuf[2], uint8_t *dest,
855  int dstW, int yalpha, int uvalpha, int y)
856 {
857  const int16_t *buf0 = buf[0], *buf1 = buf[1],
858  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
859 
860  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
861  __asm__ volatile(
862  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
863  "mov %4, %%"REG_b" \n\t"
864  "push %%"REG_BP" \n\t"
865  YSCALEYUV2RGB(%%REGBP, %5)
866  "pxor %%mm7, %%mm7 \n\t"
867  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
868 #ifdef DITHER1XBPP
869  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
870  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
871  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
872 #endif
873  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
874  "pop %%"REG_BP" \n\t"
875  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
876  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
877  "a" (&c->redDither)
878  );
879 }
880 
881 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
882  const int16_t *ubuf[2], const int16_t *vbuf[2],
883  const int16_t *abuf[2], uint8_t *dest,
884  int dstW, int yalpha, int uvalpha, int y)
885 {
886  const int16_t *buf0 = buf[0], *buf1 = buf[1],
887  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
888 
889  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
890  __asm__ volatile(
891  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
892  "mov %4, %%"REG_b" \n\t"
893  "push %%"REG_BP" \n\t"
894  YSCALEYUV2RGB(%%REGBP, %5)
895  "pxor %%mm7, %%mm7 \n\t"
896  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
897 #ifdef DITHER1XBPP
898  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
899  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
900  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
901 #endif
902  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
903  "pop %%"REG_BP" \n\t"
904  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
905  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
906  "a" (&c->redDither)
907  );
908 }
909 
910 #define REAL_YSCALEYUV2PACKED(index, c) \
911  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
912  "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1 \n\t"\
913  "psraw $3, %%mm0 \n\t"\
914  "psraw $3, %%mm1 \n\t"\
915  "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
916  "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
917  "xor "#index", "#index" \n\t"\
918  ".p2align 4 \n\t"\
919  "1: \n\t"\
920  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
921  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
922  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
923  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
924  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
925  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
926  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
927  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
928  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
929  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
930  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
931  "psraw $7, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
932  "psraw $7, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
933  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
934  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
935  "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
936  "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
937  "movq 8(%0, "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
938  "movq 8(%1, "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
939  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
940  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
941  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
942  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
943  "psraw $7, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
944  "psraw $7, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
945  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
946  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
947 
948 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c)
949 
950 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
951  const int16_t *ubuf[2], const int16_t *vbuf[2],
952  const int16_t *abuf[2], uint8_t *dest,
953  int dstW, int yalpha, int uvalpha, int y)
954 {
955  const int16_t *buf0 = buf[0], *buf1 = buf[1],
956  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
957 
958  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
959  __asm__ volatile(
960  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
961  "mov %4, %%"REG_b" \n\t"
962  "push %%"REG_BP" \n\t"
963  YSCALEYUV2PACKED(%%REGBP, %5)
964  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
965  "pop %%"REG_BP" \n\t"
966  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
967  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
968  "a" (&c->redDither)
969  );
970 }
971 
972 #define REAL_YSCALEYUV2RGB1(index, c) \
973  "xor "#index", "#index" \n\t"\
974  ".p2align 4 \n\t"\
975  "1: \n\t"\
976  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
977  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
978  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
979  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
980  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
981  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
982  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
983  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
984  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
985  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
986  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
987  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
988  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
989  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
990  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
991  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
992  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
993  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
994  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
995  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
996  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
997  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
998  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
999  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1000  "paddw %%mm3, %%mm4 \n\t"\
1001  "movq %%mm2, %%mm0 \n\t"\
1002  "movq %%mm5, %%mm6 \n\t"\
1003  "movq %%mm4, %%mm3 \n\t"\
1004  "punpcklwd %%mm2, %%mm2 \n\t"\
1005  "punpcklwd %%mm5, %%mm5 \n\t"\
1006  "punpcklwd %%mm4, %%mm4 \n\t"\
1007  "paddw %%mm1, %%mm2 \n\t"\
1008  "paddw %%mm1, %%mm5 \n\t"\
1009  "paddw %%mm1, %%mm4 \n\t"\
1010  "punpckhwd %%mm0, %%mm0 \n\t"\
1011  "punpckhwd %%mm6, %%mm6 \n\t"\
1012  "punpckhwd %%mm3, %%mm3 \n\t"\
1013  "paddw %%mm7, %%mm0 \n\t"\
1014  "paddw %%mm7, %%mm6 \n\t"\
1015  "paddw %%mm7, %%mm3 \n\t"\
1016  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1017  "packuswb %%mm0, %%mm2 \n\t"\
1018  "packuswb %%mm6, %%mm5 \n\t"\
1019  "packuswb %%mm3, %%mm4 \n\t"\
1020 
1021 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c)
1022 
1023 // do vertical chrominance interpolation
1024 #define REAL_YSCALEYUV2RGB1b(index, c) \
1025  "xor "#index", "#index" \n\t"\
1026  ".p2align 4 \n\t"\
1027  "1: \n\t"\
1028  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1029  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1030  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1031  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1032  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1033  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1034  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1035  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1036  "psrlw $5, %%mm3 \n\t" /*FIXME might overflow*/\
1037  "psrlw $5, %%mm4 \n\t" /*FIXME might overflow*/\
1038  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1039  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1040  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1041  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1042  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1043  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1044  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1045  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1046  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1047  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1048  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1049  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1050  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1051  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1052  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1053  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1054  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1055  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1056  "paddw %%mm3, %%mm4 \n\t"\
1057  "movq %%mm2, %%mm0 \n\t"\
1058  "movq %%mm5, %%mm6 \n\t"\
1059  "movq %%mm4, %%mm3 \n\t"\
1060  "punpcklwd %%mm2, %%mm2 \n\t"\
1061  "punpcklwd %%mm5, %%mm5 \n\t"\
1062  "punpcklwd %%mm4, %%mm4 \n\t"\
1063  "paddw %%mm1, %%mm2 \n\t"\
1064  "paddw %%mm1, %%mm5 \n\t"\
1065  "paddw %%mm1, %%mm4 \n\t"\
1066  "punpckhwd %%mm0, %%mm0 \n\t"\
1067  "punpckhwd %%mm6, %%mm6 \n\t"\
1068  "punpckhwd %%mm3, %%mm3 \n\t"\
1069  "paddw %%mm7, %%mm0 \n\t"\
1070  "paddw %%mm7, %%mm6 \n\t"\
1071  "paddw %%mm7, %%mm3 \n\t"\
1072  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1073  "packuswb %%mm0, %%mm2 \n\t"\
1074  "packuswb %%mm6, %%mm5 \n\t"\
1075  "packuswb %%mm3, %%mm4 \n\t"\
1076 
1077 #define YSCALEYUV2RGB1b(index, c) REAL_YSCALEYUV2RGB1b(index, c)
1078 
1079 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1080  "movq (%1, "#index", 2), %%mm7 \n\t" /* abuf0[index ] */\
1081  "movq 8(%1, "#index", 2), %%mm1 \n\t" /* abuf0[index+4] */\
1082  "psraw $7, %%mm7 \n\t" /* abuf0[index ] >>7 */\
1083  "psraw $7, %%mm1 \n\t" /* abuf0[index+4] >>7 */\
1084  "packuswb %%mm1, %%mm7 \n\t"
1085 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1086 
1090 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1091  const int16_t *ubuf[2], const int16_t *vbuf[2],
1092  const int16_t *abuf0, uint8_t *dest,
1093  int dstW, int uvalpha, int y)
1094 {
1095  const int16_t *ubuf0 = ubuf[0];
1096  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1097 
1098  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1099  const int16_t *ubuf1 = ubuf[0];
1100  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1101  __asm__ volatile(
1102  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1103  "mov %4, %%"REG_b" \n\t"
1104  "push %%"REG_BP" \n\t"
1105  YSCALEYUV2RGB1(%%REGBP, %5)
1106  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1107  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1108  "pop %%"REG_BP" \n\t"
1109  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1110  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1111  "a" (&c->redDither)
1112  );
1113  } else {
1114  __asm__ volatile(
1115  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1116  "mov %4, %%"REG_b" \n\t"
1117  "push %%"REG_BP" \n\t"
1118  YSCALEYUV2RGB1(%%REGBP, %5)
1119  "pcmpeqd %%mm7, %%mm7 \n\t"
1120  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1121  "pop %%"REG_BP" \n\t"
1122  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1123  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1124  "a" (&c->redDither)
1125  );
1126  }
1127  } else {
1128  const int16_t *ubuf1 = ubuf[1];
1129  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1130  __asm__ volatile(
1131  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1132  "mov %4, %%"REG_b" \n\t"
1133  "push %%"REG_BP" \n\t"
1134  YSCALEYUV2RGB1b(%%REGBP, %5)
1135  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1136  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1137  "pop %%"REG_BP" \n\t"
1138  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1139  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1140  "a" (&c->redDither)
1141  );
1142  } else {
1143  __asm__ volatile(
1144  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1145  "mov %4, %%"REG_b" \n\t"
1146  "push %%"REG_BP" \n\t"
1147  YSCALEYUV2RGB1b(%%REGBP, %5)
1148  "pcmpeqd %%mm7, %%mm7 \n\t"
1149  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1150  "pop %%"REG_BP" \n\t"
1151  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1152  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1153  "a" (&c->redDither)
1154  );
1155  }
1156  }
1157 }
1158 
1159 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1160  const int16_t *ubuf[2], const int16_t *vbuf[2],
1161  const int16_t *abuf0, uint8_t *dest,
1162  int dstW, int uvalpha, int y)
1163 {
1164  const int16_t *ubuf0 = ubuf[0];
1165  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1166 
1167  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1168  const int16_t *ubuf1 = ubuf[0];
1169  __asm__ volatile(
1170  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1171  "mov %4, %%"REG_b" \n\t"
1172  "push %%"REG_BP" \n\t"
1173  YSCALEYUV2RGB1(%%REGBP, %5)
1174  "pxor %%mm7, %%mm7 \n\t"
1175  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1176  "pop %%"REG_BP" \n\t"
1177  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1178  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1179  "a" (&c->redDither)
1180  );
1181  } else {
1182  const int16_t *ubuf1 = ubuf[1];
1183  __asm__ volatile(
1184  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1185  "mov %4, %%"REG_b" \n\t"
1186  "push %%"REG_BP" \n\t"
1187  YSCALEYUV2RGB1b(%%REGBP, %5)
1188  "pxor %%mm7, %%mm7 \n\t"
1189  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1190  "pop %%"REG_BP" \n\t"
1191  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1192  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1193  "a" (&c->redDither)
1194  );
1195  }
1196 }
1197 
1198 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1199  const int16_t *ubuf[2], const int16_t *vbuf[2],
1200  const int16_t *abuf0, uint8_t *dest,
1201  int dstW, int uvalpha, int y)
1202 {
1203  const int16_t *ubuf0 = ubuf[0];
1204  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1205 
1206  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1207  const int16_t *ubuf1 = ubuf[0];
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  "pxor %%mm7, %%mm7 \n\t"
1214  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1215 #ifdef DITHER1XBPP
1216  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1217  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1218  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1219 #endif
1220  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1221  "pop %%"REG_BP" \n\t"
1222  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1223  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1224  "a" (&c->redDither)
1225  );
1226  } else {
1227  const int16_t *ubuf1 = ubuf[1];
1228  __asm__ volatile(
1229  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1230  "mov %4, %%"REG_b" \n\t"
1231  "push %%"REG_BP" \n\t"
1232  YSCALEYUV2RGB1b(%%REGBP, %5)
1233  "pxor %%mm7, %%mm7 \n\t"
1234  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1235 #ifdef DITHER1XBPP
1236  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1237  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1238  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1239 #endif
1240  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1241  "pop %%"REG_BP" \n\t"
1242  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1243  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1244  "a" (&c->redDither)
1245  );
1246  }
1247 }
1248 
1249 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1250  const int16_t *ubuf[2], const int16_t *vbuf[2],
1251  const int16_t *abuf0, uint8_t *dest,
1252  int dstW, int uvalpha, int y)
1253 {
1254  const int16_t *ubuf0 = ubuf[0];
1255  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1256 
1257  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1258  const int16_t *ubuf1 = ubuf[0];
1259  __asm__ volatile(
1260  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1261  "mov %4, %%"REG_b" \n\t"
1262  "push %%"REG_BP" \n\t"
1263  YSCALEYUV2RGB1(%%REGBP, %5)
1264  "pxor %%mm7, %%mm7 \n\t"
1265  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1266 #ifdef DITHER1XBPP
1267  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1268  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1269  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1270 #endif
1271  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1272  "pop %%"REG_BP" \n\t"
1273  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1274  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1275  "a" (&c->redDither)
1276  );
1277  } else {
1278  const int16_t *ubuf1 = ubuf[1];
1279  __asm__ volatile(
1280  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1281  "mov %4, %%"REG_b" \n\t"
1282  "push %%"REG_BP" \n\t"
1283  YSCALEYUV2RGB1b(%%REGBP, %5)
1284  "pxor %%mm7, %%mm7 \n\t"
1285  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1286 #ifdef DITHER1XBPP
1287  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1288  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1289  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1290 #endif
1291  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1292  "pop %%"REG_BP" \n\t"
1293  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1294  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1295  "a" (&c->redDither)
1296  );
1297  }
1298 }
1299 
1300 #define REAL_YSCALEYUV2PACKED1(index, c) \
1301  "xor "#index", "#index" \n\t"\
1302  ".p2align 4 \n\t"\
1303  "1: \n\t"\
1304  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1305  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1306  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1307  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1308  "psraw $7, %%mm3 \n\t" \
1309  "psraw $7, %%mm4 \n\t" \
1310  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1311  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1312  "psraw $7, %%mm1 \n\t" \
1313  "psraw $7, %%mm7 \n\t" \
1314 
1315 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c)
1316 
1317 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1318  "xor "#index", "#index" \n\t"\
1319  ".p2align 4 \n\t"\
1320  "1: \n\t"\
1321  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1322  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1323  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1324  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1325  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1326  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1327  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1328  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1329  "psrlw $8, %%mm3 \n\t" \
1330  "psrlw $8, %%mm4 \n\t" \
1331  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1332  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1333  "psraw $7, %%mm1 \n\t" \
1334  "psraw $7, %%mm7 \n\t"
1335 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c)
1336 
1337 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1338  const int16_t *ubuf[2], const int16_t *vbuf[2],
1339  const int16_t *abuf0, uint8_t *dest,
1340  int dstW, int uvalpha, int y)
1341 {
1342  const int16_t *ubuf0 = ubuf[0];
1343  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1344 
1345  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1346  const int16_t *ubuf1 = ubuf[0];
1347  __asm__ volatile(
1348  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1349  "mov %4, %%"REG_b" \n\t"
1350  "push %%"REG_BP" \n\t"
1351  YSCALEYUV2PACKED1(%%REGBP, %5)
1352  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1353  "pop %%"REG_BP" \n\t"
1354  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1355  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1356  "a" (&c->redDither)
1357  );
1358  } else {
1359  const int16_t *ubuf1 = ubuf[1];
1360  __asm__ volatile(
1361  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1362  "mov %4, %%"REG_b" \n\t"
1363  "push %%"REG_BP" \n\t"
1364  YSCALEYUV2PACKED1b(%%REGBP, %5)
1365  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1366  "pop %%"REG_BP" \n\t"
1367  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1368  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1369  "a" (&c->redDither)
1370  );
1371  }
1372 }
1373 
1374 #if COMPILE_TEMPLATE_MMXEXT
1375 static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
1376  int dstWidth, const uint8_t *src,
1377  int srcW, int xInc)
1378 {
1379  int32_t *filterPos = c->hLumFilterPos;
1380  int16_t *filter = c->hLumFilter;
1381  void *mmxextFilterCode = c->lumMmxextFilterCode;
1382  int i;
1383 #if defined(PIC)
1384  uint64_t ebxsave;
1385 #endif
1386 #if ARCH_X86_64
1387  uint64_t retsave;
1388 #endif
1389 
1390  __asm__ volatile(
1391 #if defined(PIC)
1392  "mov %%"REG_b", %5 \n\t"
1393 #if ARCH_X86_64
1394  "mov -8(%%rsp), %%"REG_a" \n\t"
1395  "mov %%"REG_a", %6 \n\t"
1396 #endif
1397 #else
1398 #if ARCH_X86_64
1399  "mov -8(%%rsp), %%"REG_a" \n\t"
1400  "mov %%"REG_a", %5 \n\t"
1401 #endif
1402 #endif
1403  "pxor %%mm7, %%mm7 \n\t"
1404  "mov %0, %%"REG_c" \n\t"
1405  "mov %1, %%"REG_D" \n\t"
1406  "mov %2, %%"REG_d" \n\t"
1407  "mov %3, %%"REG_b" \n\t"
1408  "xor %%"REG_a", %%"REG_a" \n\t" // i
1409  PREFETCH" (%%"REG_c") \n\t"
1410  PREFETCH" 32(%%"REG_c") \n\t"
1411  PREFETCH" 64(%%"REG_c") \n\t"
1412 
1413 #if ARCH_X86_64
1414 #define CALL_MMXEXT_FILTER_CODE \
1415  "movl (%%"REG_b"), %%esi \n\t"\
1416  "call *%4 \n\t"\
1417  "movl (%%"REG_b", %%"REG_a"), %%esi \n\t"\
1418  "add %%"REG_S", %%"REG_c" \n\t"\
1419  "add %%"REG_a", %%"REG_D" \n\t"\
1420  "xor %%"REG_a", %%"REG_a" \n\t"\
1421 
1422 #else
1423 #define CALL_MMXEXT_FILTER_CODE \
1424  "movl (%%"REG_b"), %%esi \n\t"\
1425  "call *%4 \n\t"\
1426  "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\
1427  "add %%"REG_a", %%"REG_D" \n\t"\
1428  "xor %%"REG_a", %%"REG_a" \n\t"\
1429 
1430 #endif /* ARCH_X86_64 */
1431 
1432  CALL_MMXEXT_FILTER_CODE
1433  CALL_MMXEXT_FILTER_CODE
1434  CALL_MMXEXT_FILTER_CODE
1435  CALL_MMXEXT_FILTER_CODE
1436  CALL_MMXEXT_FILTER_CODE
1437  CALL_MMXEXT_FILTER_CODE
1438  CALL_MMXEXT_FILTER_CODE
1439  CALL_MMXEXT_FILTER_CODE
1440 
1441 #if defined(PIC)
1442  "mov %5, %%"REG_b" \n\t"
1443 #if ARCH_X86_64
1444  "mov %6, %%"REG_a" \n\t"
1445  "mov %%"REG_a", -8(%%rsp) \n\t"
1446 #endif
1447 #else
1448 #if ARCH_X86_64
1449  "mov %5, %%"REG_a" \n\t"
1450  "mov %%"REG_a", -8(%%rsp) \n\t"
1451 #endif
1452 #endif
1453  :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),
1454  "m" (mmxextFilterCode)
1455 #if defined(PIC)
1456  ,"m" (ebxsave)
1457 #endif
1458 #if ARCH_X86_64
1459  ,"m"(retsave)
1460 #endif
1461  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1462 #if !defined(PIC)
1463  ,"%"REG_b
1464 #endif
1465  );
1466 
1467  for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
1468  dst[i] = src[srcW-1]*128;
1469 }
1470 
1471 static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
1472  int dstWidth, const uint8_t *src1,
1473  const uint8_t *src2, int srcW, int xInc)
1474 {
1475  int32_t *filterPos = c->hChrFilterPos;
1476  int16_t *filter = c->hChrFilter;
1477  void *mmxextFilterCode = c->chrMmxextFilterCode;
1478  int i;
1479 #if defined(PIC)
1480  DECLARE_ALIGNED(8, uint64_t, ebxsave);
1481 #endif
1482 #if ARCH_X86_64
1483  DECLARE_ALIGNED(8, uint64_t, retsave);
1484 #endif
1485 
1486  __asm__ volatile(
1487 #if defined(PIC)
1488  "mov %%"REG_b", %7 \n\t"
1489 #if ARCH_X86_64
1490  "mov -8(%%rsp), %%"REG_a" \n\t"
1491  "mov %%"REG_a", %8 \n\t"
1492 #endif
1493 #else
1494 #if ARCH_X86_64
1495  "mov -8(%%rsp), %%"REG_a" \n\t"
1496  "mov %%"REG_a", %7 \n\t"
1497 #endif
1498 #endif
1499  "pxor %%mm7, %%mm7 \n\t"
1500  "mov %0, %%"REG_c" \n\t"
1501  "mov %1, %%"REG_D" \n\t"
1502  "mov %2, %%"REG_d" \n\t"
1503  "mov %3, %%"REG_b" \n\t"
1504  "xor %%"REG_a", %%"REG_a" \n\t" // i
1505  PREFETCH" (%%"REG_c") \n\t"
1506  PREFETCH" 32(%%"REG_c") \n\t"
1507  PREFETCH" 64(%%"REG_c") \n\t"
1508 
1509  CALL_MMXEXT_FILTER_CODE
1510  CALL_MMXEXT_FILTER_CODE
1511  CALL_MMXEXT_FILTER_CODE
1512  CALL_MMXEXT_FILTER_CODE
1513  "xor %%"REG_a", %%"REG_a" \n\t" // i
1514  "mov %5, %%"REG_c" \n\t" // src
1515  "mov %6, %%"REG_D" \n\t" // buf2
1516  PREFETCH" (%%"REG_c") \n\t"
1517  PREFETCH" 32(%%"REG_c") \n\t"
1518  PREFETCH" 64(%%"REG_c") \n\t"
1519 
1520  CALL_MMXEXT_FILTER_CODE
1521  CALL_MMXEXT_FILTER_CODE
1522  CALL_MMXEXT_FILTER_CODE
1523  CALL_MMXEXT_FILTER_CODE
1524 
1525 #if defined(PIC)
1526  "mov %7, %%"REG_b" \n\t"
1527 #if ARCH_X86_64
1528  "mov %8, %%"REG_a" \n\t"
1529  "mov %%"REG_a", -8(%%rsp) \n\t"
1530 #endif
1531 #else
1532 #if ARCH_X86_64
1533  "mov %7, %%"REG_a" \n\t"
1534  "mov %%"REG_a", -8(%%rsp) \n\t"
1535 #endif
1536 #endif
1537  :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),
1538  "m" (mmxextFilterCode), "m" (src2), "m"(dst2)
1539 #if defined(PIC)
1540  ,"m" (ebxsave)
1541 #endif
1542 #if ARCH_X86_64
1543  ,"m"(retsave)
1544 #endif
1545  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1546 #if !defined(PIC)
1547  ,"%"REG_b
1548 #endif
1549  );
1550 
1551  for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
1552  dst1[i] = src1[srcW-1]*128;
1553  dst2[i] = src2[srcW-1]*128;
1554  }
1555 }
1556 #endif /* COMPILE_TEMPLATE_MMXEXT */
1557 
1559 {
1560  enum AVPixelFormat dstFormat = c->dstFormat;
1561 
1562  if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) &&
1563  dstFormat != AV_PIX_FMT_NV12 && dstFormat != AV_PIX_FMT_NV21) {
1564  if (!(c->flags & SWS_BITEXACT)) {
1565  if (c->flags & SWS_ACCURATE_RND) {
1566  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1567  switch (c->dstFormat) {
1568  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X_ar); break;
1569  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X_ar); break;
1570  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X_ar); break;
1571  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X_ar); break;
1572  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
1573  default: break;
1574  }
1575  }
1576  } else {
1577  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1578  switch (c->dstFormat) {
1579  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X); break;
1580  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X); break;
1581  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X); break;
1582  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X); break;
1583  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
1584  default: break;
1585  }
1586  }
1587  }
1588  }
1589  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1590  switch (c->dstFormat) {
1591  case AV_PIX_FMT_RGB32:
1592  c->yuv2packed1 = RENAME(yuv2rgb32_1);
1593  c->yuv2packed2 = RENAME(yuv2rgb32_2);
1594  break;
1595  case AV_PIX_FMT_BGR24:
1596  c->yuv2packed1 = RENAME(yuv2bgr24_1);
1597  c->yuv2packed2 = RENAME(yuv2bgr24_2);
1598  break;
1599  case AV_PIX_FMT_RGB555:
1600  c->yuv2packed1 = RENAME(yuv2rgb555_1);
1601  c->yuv2packed2 = RENAME(yuv2rgb555_2);
1602  break;
1603  case AV_PIX_FMT_RGB565:
1604  c->yuv2packed1 = RENAME(yuv2rgb565_1);
1605  c->yuv2packed2 = RENAME(yuv2rgb565_2);
1606  break;
1607  case AV_PIX_FMT_YUYV422:
1608  c->yuv2packed1 = RENAME(yuv2yuyv422_1);
1609  c->yuv2packed2 = RENAME(yuv2yuyv422_2);
1610  break;
1611  default:
1612  break;
1613  }
1614  }
1615  }
1616 
1617  if (c->srcBpc == 8 && c->dstBpc <= 10) {
1618  // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one).
1619 #if COMPILE_TEMPLATE_MMXEXT
1620  if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) {
1621  c->hyscale_fast = RENAME(hyscale_fast);
1622  c->hcscale_fast = RENAME(hcscale_fast);
1623  } else {
1624 #endif /* COMPILE_TEMPLATE_MMXEXT */
1625  c->hyscale_fast = NULL;
1626  c->hcscale_fast = NULL;
1627 #if COMPILE_TEMPLATE_MMXEXT
1628  }
1629 #endif /* COMPILE_TEMPLATE_MMXEXT */
1630  }
1631 }