Mercurial > sdl-ios-xcode
comparison src/video/SDL_yuv_mmx.c @ 1148:63fb2da89a4b
Patched inline assembly to compile on gcc 4.0.1. Details are here:
http://bugs.gentoo.org/show_bug.cgi?id=104437
I believe the patch is from Harald van Dijk (comment #16 at the above url).
author | Ryan C. Gordon <icculus@icculus.org> |
---|---|
date | Wed, 28 Sep 2005 02:26:29 +0000 |
parents | 29d7db09776e |
children | c9b51268668f |
comparison
equal
deleted
inserted
replaced
1147:b580f7201543 | 1148:63fb2da89a4b |
---|---|
28 | 28 |
29 #if defined(i386) && defined(__GNUC__) && defined(USE_ASMBLIT) | 29 #if defined(i386) && defined(__GNUC__) && defined(USE_ASMBLIT) |
30 | 30 |
31 #include "SDL_types.h" | 31 #include "SDL_types.h" |
32 | 32 |
33 #if __GNUC__ > 2 | 33 #define ASM_ARRAY(x) x[] __asm__("_" #x) __attribute__((used)) |
34 # undef GCC2_HACK | |
35 #else | |
36 # define GCC2_HACK | |
37 #endif | |
38 | |
39 | 34 |
40 #if defined(GCC2_HACK) && defined (__ELF__) | 35 static unsigned int ASM_ARRAY(MMX_0080w) = {0x00800080, 0x00800080}; |
41 #define ASM_VAR(X) _##X | 36 static unsigned int ASM_ARRAY(MMX_00FFw) = {0x00ff00ff, 0x00ff00ff}; |
42 #else | 37 static unsigned int ASM_ARRAY(MMX_FF00w) = {0xff00ff00, 0xff00ff00}; |
43 #define ASM_VAR(X) X | 38 |
44 #endif | 39 static unsigned short ASM_ARRAY(MMX_Ycoeff) = {0x004a, 0x004a, 0x004a, 0x004a}; |
45 | 40 |
46 static volatile unsigned int ASM_VAR(MMX_0080w)[] = {0x00800080, 0x00800080}; | 41 static unsigned short ASM_ARRAY(MMX_UbluRGB) = {0x0072, 0x0072, 0x0072, 0x0072}; |
47 static volatile unsigned int ASM_VAR(MMX_00FFw)[] = {0x00ff00ff, 0x00ff00ff}; | 42 static unsigned short ASM_ARRAY(MMX_VredRGB) = {0x0059, 0x0059, 0x0059, 0x0059}; |
48 static volatile unsigned int ASM_VAR(MMX_FF00w)[] = {0xff00ff00, 0xff00ff00}; | 43 static unsigned short ASM_ARRAY(MMX_UgrnRGB) = {0xffea, 0xffea, 0xffea, 0xffea}; |
49 | 44 static unsigned short ASM_ARRAY(MMX_VgrnRGB) = {0xffd2, 0xffd2, 0xffd2, 0xffd2}; |
50 static volatile unsigned short ASM_VAR(MMX_Ycoeff)[] = {0x004a, 0x004a, 0x004a, 0x004a}; | 45 |
51 | 46 static unsigned short ASM_ARRAY(MMX_Ublu5x5) = {0x0081, 0x0081, 0x0081, 0x0081}; |
52 static volatile unsigned short ASM_VAR(MMX_UbluRGB)[] = {0x0072, 0x0072, 0x0072, 0x0072}; | 47 static unsigned short ASM_ARRAY(MMX_Vred5x5) = {0x0066, 0x0066, 0x0066, 0x0066}; |
53 static volatile unsigned short ASM_VAR(MMX_VredRGB)[] = {0x0059, 0x0059, 0x0059, 0x0059}; | 48 static unsigned short ASM_ARRAY(MMX_Ugrn555) = {0xffe7, 0xffe7, 0xffe7, 0xffe7}; |
54 static volatile unsigned short ASM_VAR(MMX_UgrnRGB)[] = {0xffea, 0xffea, 0xffea, 0xffea}; | 49 static unsigned short ASM_ARRAY(MMX_Vgrn555) = {0xffcc, 0xffcc, 0xffcc, 0xffcc}; |
55 static volatile unsigned short ASM_VAR(MMX_VgrnRGB)[] = {0xffd2, 0xffd2, 0xffd2, 0xffd2}; | 50 static unsigned short ASM_ARRAY(MMX_Ugrn565) = {0xffe8, 0xffe8, 0xffe8, 0xffe8}; |
56 | 51 static unsigned short ASM_ARRAY(MMX_Vgrn565) = {0xffcd, 0xffcd, 0xffcd, 0xffcd}; |
57 static volatile unsigned short ASM_VAR(MMX_Ublu5x5)[] = {0x0081, 0x0081, 0x0081, 0x0081}; | 52 |
58 static volatile unsigned short ASM_VAR(MMX_Vred5x5)[] = {0x0066, 0x0066, 0x0066, 0x0066}; | 53 static unsigned short ASM_ARRAY(MMX_red555) = {0x7c00, 0x7c00, 0x7c00, 0x7c00}; |
59 static volatile unsigned short ASM_VAR(MMX_Ugrn555)[] = {0xffe7, 0xffe7, 0xffe7, 0xffe7}; | 54 static unsigned short ASM_ARRAY(MMX_red565) = {0xf800, 0xf800, 0xf800, 0xf800}; |
60 static volatile unsigned short ASM_VAR(MMX_Vgrn555)[] = {0xffcc, 0xffcc, 0xffcc, 0xffcc}; | 55 static unsigned short ASM_ARRAY(MMX_grn555) = {0x03e0, 0x03e0, 0x03e0, 0x03e0}; |
61 static volatile unsigned short ASM_VAR(MMX_Ugrn565)[] = {0xffe8, 0xffe8, 0xffe8, 0xffe8}; | 56 static unsigned short ASM_ARRAY(MMX_grn565) = {0x07e0, 0x07e0, 0x07e0, 0x07e0}; |
62 static volatile unsigned short ASM_VAR(MMX_Vgrn565)[] = {0xffcd, 0xffcd, 0xffcd, 0xffcd}; | 57 static unsigned short ASM_ARRAY(MMX_blu5x5) = {0x001f, 0x001f, 0x001f, 0x001f}; |
63 | |
64 static volatile unsigned short ASM_VAR(MMX_red555)[] = {0x7c00, 0x7c00, 0x7c00, 0x7c00}; | |
65 static volatile unsigned short ASM_VAR(MMX_red565)[] = {0xf800, 0xf800, 0xf800, 0xf800}; | |
66 static volatile unsigned short ASM_VAR(MMX_grn555)[] = {0x03e0, 0x03e0, 0x03e0, 0x03e0}; | |
67 static volatile unsigned short ASM_VAR(MMX_grn565)[] = {0x07e0, 0x07e0, 0x07e0, 0x07e0}; | |
68 static volatile unsigned short ASM_VAR(MMX_blu5x5)[] = {0x001f, 0x001f, 0x001f, 0x001f}; | |
69 | 58 |
70 /** | 59 /** |
71 This MMX assembler is my first assembler/MMX program ever. | 60 This MMX assembler is my first assembler/MMX program ever. |
72 Thus it maybe buggy. | 61 Thus it maybe buggy. |
73 Send patches to: | 62 Send patches to: |
125 "movd (%%ebx), %%mm1\n" // 0 0 0 0 v3 v2 v1 v0 | 114 "movd (%%ebx), %%mm1\n" // 0 0 0 0 v3 v2 v1 v0 |
126 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00 | 115 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00 |
127 "movd (%2), %%mm2\n" // 0 0 0 0 l3 l2 l1 l0 | 116 "movd (%2), %%mm2\n" // 0 0 0 0 l3 l2 l1 l0 |
128 "punpcklbw %%mm7,%%mm1\n" // 0 v3 0 v2 00 v1 00 v0 | 117 "punpcklbw %%mm7,%%mm1\n" // 0 v3 0 v2 00 v1 00 v0 |
129 "punpckldq %%mm1,%%mm1\n" // 00 v1 00 v0 00 v1 00 v0 | 118 "punpckldq %%mm1,%%mm1\n" // 00 v1 00 v0 00 v1 00 v0 |
130 #ifdef GCC2_HACK | |
131 "psubw _MMX_0080w,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 | 119 "psubw _MMX_0080w,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 |
132 #else | |
133 "psubw %[_MMX_0080w],%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 | |
134 #endif | |
135 | 120 |
136 // create Cr_g (result in mm0) | 121 // create Cr_g (result in mm0) |
137 "movq %%mm1,%%mm0\n" // r1 r1 r0 r0 r1 r1 r0 r0 | 122 "movq %%mm1,%%mm0\n" // r1 r1 r0 r0 r1 r1 r0 r0 |
138 #ifdef GCC2_HACK | |
139 "pmullw _MMX_VgrnRGB,%%mm0\n"// red*-46dec=0.7136*64 | 123 "pmullw _MMX_VgrnRGB,%%mm0\n"// red*-46dec=0.7136*64 |
140 "pmullw _MMX_VredRGB,%%mm1\n"// red*89dec=1.4013*64 | 124 "pmullw _MMX_VredRGB,%%mm1\n"// red*89dec=1.4013*64 |
141 #else | |
142 "pmullw %[_MMX_VgrnRGB],%%mm0\n"// red*-46dec=0.7136*64 | |
143 "pmullw %[_MMX_VredRGB],%%mm1\n"// red*89dec=1.4013*64 | |
144 #endif | |
145 "psraw $6, %%mm0\n" // red=red/64 | 125 "psraw $6, %%mm0\n" // red=red/64 |
146 "psraw $6, %%mm1\n" // red=red/64 | 126 "psraw $6, %%mm1\n" // red=red/64 |
147 | 127 |
148 // create L1 L2 (result in mm2,mm4) | 128 // create L1 L2 (result in mm2,mm4) |
149 // L2=lum+cols | 129 // L2=lum+cols |
150 "movq (%2,%4),%%mm3\n" // 0 0 0 0 L3 L2 L1 L0 | 130 "movq (%2,%4),%%mm3\n" // 0 0 0 0 L3 L2 L1 L0 |
151 "punpckldq %%mm3,%%mm2\n" // L3 L2 L1 L0 l3 l2 l1 l0 | 131 "punpckldq %%mm3,%%mm2\n" // L3 L2 L1 L0 l3 l2 l1 l0 |
152 "movq %%mm2,%%mm4\n" // L3 L2 L1 L0 l3 l2 l1 l0 | 132 "movq %%mm2,%%mm4\n" // L3 L2 L1 L0 l3 l2 l1 l0 |
153 #ifdef GCC2_HACK | |
154 "pand _MMX_FF00w,%%mm2\n" // L3 0 L1 0 l3 0 l1 0 | 133 "pand _MMX_FF00w,%%mm2\n" // L3 0 L1 0 l3 0 l1 0 |
155 "pand _MMX_00FFw,%%mm4\n" // 0 L2 0 L0 0 l2 0 l0 | 134 "pand _MMX_00FFw,%%mm4\n" // 0 L2 0 L0 0 l2 0 l0 |
156 #else | |
157 "pand %[_MMX_FF00w],%%mm2\n" // L3 0 L1 0 l3 0 l1 0 | |
158 "pand %[_MMX_00FFw],%%mm4\n" // 0 L2 0 L0 0 l2 0 l0 | |
159 #endif | |
160 "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1 | 135 "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1 |
161 | 136 |
162 // create R (result in mm6) | 137 // create R (result in mm6) |
163 "movq %%mm2,%%mm5\n" // 0 L3 0 L1 0 l3 0 l1 | 138 "movq %%mm2,%%mm5\n" // 0 L3 0 L1 0 l3 0 l1 |
164 "movq %%mm4,%%mm6\n" // 0 L2 0 L0 0 l2 0 l0 | 139 "movq %%mm4,%%mm6\n" // 0 L2 0 L0 0 l2 0 l0 |
171 | 146 |
172 // create Cb (result in mm1) | 147 // create Cb (result in mm1) |
173 "movd (%1), %%mm1\n" // 0 0 0 0 u3 u2 u1 u0 | 148 "movd (%1), %%mm1\n" // 0 0 0 0 u3 u2 u1 u0 |
174 "punpcklbw %%mm7,%%mm1\n" // 0 u3 0 u2 00 u1 00 u0 | 149 "punpcklbw %%mm7,%%mm1\n" // 0 u3 0 u2 00 u1 00 u0 |
175 "punpckldq %%mm1,%%mm1\n" // 00 u1 00 u0 00 u1 00 u0 | 150 "punpckldq %%mm1,%%mm1\n" // 00 u1 00 u0 00 u1 00 u0 |
176 #ifdef GCC2_HACK | |
177 "psubw _MMX_0080w,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 | 151 "psubw _MMX_0080w,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 |
178 #else | |
179 "psubw %[_MMX_0080w],%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 | |
180 #endif | |
181 // create Cb_g (result in mm5) | 152 // create Cb_g (result in mm5) |
182 "movq %%mm1,%%mm5\n" // u1 u1 u0 u0 u1 u1 u0 u0 | 153 "movq %%mm1,%%mm5\n" // u1 u1 u0 u0 u1 u1 u0 u0 |
183 #ifdef GCC2_HACK | |
184 "pmullw _MMX_UgrnRGB,%%mm5\n" // blue*-109dec=1.7129*64 | 154 "pmullw _MMX_UgrnRGB,%%mm5\n" // blue*-109dec=1.7129*64 |
185 "pmullw _MMX_UbluRGB,%%mm1\n" // blue*114dec=1.78125*64 | 155 "pmullw _MMX_UbluRGB,%%mm1\n" // blue*114dec=1.78125*64 |
186 #else | |
187 "pmullw %[_MMX_UgrnRGB],%%mm5\n" // blue*-109dec=1.7129*64 | |
188 "pmullw %[_MMX_UbluRGB],%%mm1\n" // blue*114dec=1.78125*64 | |
189 #endif | |
190 "psraw $6, %%mm5\n" // blue=red/64 | 156 "psraw $6, %%mm5\n" // blue=red/64 |
191 "psraw $6, %%mm1\n" // blue=blue/64 | 157 "psraw $6, %%mm1\n" // blue=blue/64 |
192 | 158 |
193 // create G (result in mm7) | 159 // create G (result in mm7) |
194 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1 | 160 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1 |
266 "jl 1b\n" | 232 "jl 1b\n" |
267 "emms\n" | 233 "emms\n" |
268 "popl %%ebx\n" | 234 "popl %%ebx\n" |
269 : | 235 : |
270 : "m" (cr), "r"(cb),"r"(lum), | 236 : "m" (cr), "r"(cb),"r"(lum), |
271 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod) | 237 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)); |
272 #ifndef GCC2_HACK | |
273 ,[_MMX_0080w] "m" (*MMX_0080w), | |
274 [_MMX_00FFw] "m" (*MMX_00FFw), | |
275 [_MMX_FF00w] "m" (*MMX_FF00w), | |
276 [_MMX_VgrnRGB] "m" (*MMX_VgrnRGB), | |
277 [_MMX_VredRGB] "m" (*MMX_VredRGB), | |
278 [_MMX_UgrnRGB] "m" (*MMX_UgrnRGB), | |
279 [_MMX_UbluRGB] "m" (*MMX_UbluRGB) | |
280 #endif | |
281 ); | |
282 } | 238 } |
283 | 239 |
284 void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, | 240 void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, |
285 unsigned char *lum, unsigned char *cr, | 241 unsigned char *lum, unsigned char *cr, |
286 unsigned char *cb, unsigned char *out, | 242 unsigned char *cb, unsigned char *out, |
305 "movd (%1), %%mm0\n" // 4 Cb 0 0 0 0 u3 u2 u1 u0 | 261 "movd (%1), %%mm0\n" // 4 Cb 0 0 0 0 u3 u2 u1 u0 |
306 "pxor %%mm7, %%mm7\n" | 262 "pxor %%mm7, %%mm7\n" |
307 "movd (%%ebx), %%mm1\n" // 4 Cr 0 0 0 0 v3 v2 v1 v0 | 263 "movd (%%ebx), %%mm1\n" // 4 Cr 0 0 0 0 v3 v2 v1 v0 |
308 "punpcklbw %%mm7, %%mm0\n" // 4 W cb 0 u3 0 u2 0 u1 0 u0 | 264 "punpcklbw %%mm7, %%mm0\n" // 4 W cb 0 u3 0 u2 0 u1 0 u0 |
309 "punpcklbw %%mm7, %%mm1\n" // 4 W cr 0 v3 0 v2 0 v1 0 v0 | 265 "punpcklbw %%mm7, %%mm1\n" // 4 W cr 0 v3 0 v2 0 v1 0 v0 |
310 #ifdef GCC2_HACK | |
311 "psubw _MMX_0080w, %%mm0\n" | 266 "psubw _MMX_0080w, %%mm0\n" |
312 "psubw _MMX_0080w, %%mm1\n" | 267 "psubw _MMX_0080w, %%mm1\n" |
313 #else | |
314 "psubw %[_MMX_0080w], %%mm0\n" | |
315 "psubw %[_MMX_0080w], %%mm1\n" | |
316 #endif | |
317 "movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0 | 268 "movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0 |
318 "movq %%mm1, %%mm3\n" // Cr | 269 "movq %%mm1, %%mm3\n" // Cr |
319 #ifdef GCC2_HACK | |
320 "pmullw _MMX_Ugrn565, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0 | 270 "pmullw _MMX_Ugrn565, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0 |
321 #else | |
322 "pmullw %[_MMX_Ugrn565], %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0 | |
323 #endif | |
324 "movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0 | 271 "movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0 |
325 #ifdef GCC2_HACK | |
326 "pmullw _MMX_Ublu5x5, %%mm0\n" // Cb2blue | 272 "pmullw _MMX_Ublu5x5, %%mm0\n" // Cb2blue |
327 "pand _MMX_00FFw, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0 | 273 "pand _MMX_00FFw, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0 |
328 "pmullw _MMX_Vgrn565, %%mm3\n" // Cr2green | 274 "pmullw _MMX_Vgrn565, %%mm3\n" // Cr2green |
329 #else | |
330 "pmullw %[_MMX_Ublu5x5], %%mm0\n" // Cb2blue | |
331 "pand %[_MMX_00FFw], %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0 | |
332 "pmullw %[_MMX_Vgrn565], %%mm3\n" // Cr2green | |
333 #endif | |
334 "movq (%2), %%mm7\n" // L2 | 275 "movq (%2), %%mm7\n" // L2 |
335 #ifdef GCC2_HACK | |
336 "pmullw _MMX_Vred5x5, %%mm1\n" // Cr2red | 276 "pmullw _MMX_Vred5x5, %%mm1\n" // Cr2red |
337 #else | |
338 "pmullw %[_MMX_Vred5x5], %%mm1\n" // Cr2red | |
339 #endif | |
340 "psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1 | 277 "psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1 |
341 #ifdef GCC2_HACK | |
342 "pmullw _MMX_Ycoeff, %%mm6\n" // lum1 | 278 "pmullw _MMX_Ycoeff, %%mm6\n" // lum1 |
343 #else | |
344 "pmullw %[_MMX_Ycoeff], %%mm6\n" // lum1 | |
345 #endif | |
346 "paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green | 279 "paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green |
347 #ifdef GCC2_HACK | |
348 "pmullw _MMX_Ycoeff, %%mm7\n" // lum2 | 280 "pmullw _MMX_Ycoeff, %%mm7\n" // lum2 |
349 #else | |
350 "pmullw %[_MMX_Ycoeff], %%mm7\n" // lum2 | |
351 #endif | |
352 | 281 |
353 "movq %%mm6, %%mm4\n" // lum1 | 282 "movq %%mm6, %%mm4\n" // lum1 |
354 "paddw %%mm0, %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0 | 283 "paddw %%mm0, %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0 |
355 "movq %%mm4, %%mm5\n" // lum1 | 284 "movq %%mm4, %%mm5\n" // lum1 |
356 "paddw %%mm1, %%mm4\n" // lum1 +red 00 R6 00 R4 00 R2 00 R0 | 285 "paddw %%mm1, %%mm4\n" // lum1 +red 00 R6 00 R4 00 R2 00 R0 |
364 "packuswb %%mm5, %%mm5\n" // G1 G1 | 293 "packuswb %%mm5, %%mm5\n" // G1 G1 |
365 "packuswb %%mm6, %%mm6\n" // B1 B1 | 294 "packuswb %%mm6, %%mm6\n" // B1 B1 |
366 "punpcklbw %%mm4, %%mm4\n" | 295 "punpcklbw %%mm4, %%mm4\n" |
367 "punpcklbw %%mm5, %%mm5\n" | 296 "punpcklbw %%mm5, %%mm5\n" |
368 | 297 |
369 #ifdef GCC2_HACK | |
370 "pand _MMX_red565, %%mm4\n" | 298 "pand _MMX_red565, %%mm4\n" |
371 #else | |
372 "pand %[_MMX_red565], %%mm4\n" | |
373 #endif | |
374 "psllw $3, %%mm5\n" // GREEN 1 | 299 "psllw $3, %%mm5\n" // GREEN 1 |
375 "punpcklbw %%mm6, %%mm6\n" | 300 "punpcklbw %%mm6, %%mm6\n" |
376 #ifdef GCC2_HACK | |
377 "pand _MMX_grn565, %%mm5\n" | 301 "pand _MMX_grn565, %%mm5\n" |
378 "pand _MMX_red565, %%mm6\n" | 302 "pand _MMX_red565, %%mm6\n" |
379 #else | |
380 "pand %[_MMX_grn565], %%mm5\n" | |
381 "pand %[_MMX_red565], %%mm6\n" | |
382 #endif | |
383 "por %%mm5, %%mm4\n" // | 303 "por %%mm5, %%mm4\n" // |
384 "psrlw $11, %%mm6\n" // BLUE 1 | 304 "psrlw $11, %%mm6\n" // BLUE 1 |
385 "movq %%mm3, %%mm5\n" // lum2 | 305 "movq %%mm3, %%mm5\n" // lum2 |
386 "paddw %%mm1, %%mm3\n" // lum2 +red 00 R7 00 R5 00 R3 00 R1 | 306 "paddw %%mm1, %%mm3\n" // lum2 +red 00 R7 00 R5 00 R3 00 R1 |
387 "paddw %%mm2, %%mm5\n" // lum2 +green 00 G7 00 G5 00 G3 00 G1 | 307 "paddw %%mm2, %%mm5\n" // lum2 +green 00 G7 00 G5 00 G3 00 G1 |
391 "movq (%2, %4), %%mm6\n" // L3 load lum2 | 311 "movq (%2, %4), %%mm6\n" // L3 load lum2 |
392 "psraw $6, %%mm7\n" | 312 "psraw $6, %%mm7\n" |
393 "packuswb %%mm3, %%mm3\n" | 313 "packuswb %%mm3, %%mm3\n" |
394 "packuswb %%mm5, %%mm5\n" | 314 "packuswb %%mm5, %%mm5\n" |
395 "packuswb %%mm7, %%mm7\n" | 315 "packuswb %%mm7, %%mm7\n" |
396 #ifdef GCC2_HACK | |
397 "pand _MMX_00FFw, %%mm6\n" // L3 | 316 "pand _MMX_00FFw, %%mm6\n" // L3 |
398 #else | |
399 "pand %[_MMX_00FFw], %%mm6\n" // L3 | |
400 #endif | |
401 "punpcklbw %%mm3, %%mm3\n" | 317 "punpcklbw %%mm3, %%mm3\n" |
402 "punpcklbw %%mm5, %%mm5\n" | 318 "punpcklbw %%mm5, %%mm5\n" |
403 #ifdef GCC2_HACK | |
404 "pmullw _MMX_Ycoeff, %%mm6\n" // lum3 | 319 "pmullw _MMX_Ycoeff, %%mm6\n" // lum3 |
405 #else | |
406 "pmullw %[_MMX_Ycoeff], %%mm6\n" // lum3 | |
407 #endif | |
408 "punpcklbw %%mm7, %%mm7\n" | 320 "punpcklbw %%mm7, %%mm7\n" |
409 "psllw $3, %%mm5\n" // GREEN 2 | 321 "psllw $3, %%mm5\n" // GREEN 2 |
410 #ifdef GCC2_HACK | |
411 "pand _MMX_red565, %%mm7\n" | 322 "pand _MMX_red565, %%mm7\n" |
412 "pand _MMX_red565, %%mm3\n" | 323 "pand _MMX_red565, %%mm3\n" |
413 #else | |
414 "pand %[_MMX_red565], %%mm7\n" | |
415 "pand %[_MMX_red565], %%mm3\n" | |
416 #endif | |
417 "psrlw $11, %%mm7\n" // BLUE 2 | 324 "psrlw $11, %%mm7\n" // BLUE 2 |
418 #ifdef GCC2_HACK | |
419 "pand _MMX_grn565, %%mm5\n" | 325 "pand _MMX_grn565, %%mm5\n" |
420 #else | |
421 "pand %[_MMX_grn565], %%mm5\n" | |
422 #endif | |
423 "por %%mm7, %%mm3\n" | 326 "por %%mm7, %%mm3\n" |
424 "movq (%2,%4), %%mm7\n" // L4 load lum2 | 327 "movq (%2,%4), %%mm7\n" // L4 load lum2 |
425 "por %%mm5, %%mm3\n" // | 328 "por %%mm5, %%mm3\n" // |
426 "psrlw $8, %%mm7\n" // L4 | 329 "psrlw $8, %%mm7\n" // L4 |
427 "movq %%mm4, %%mm5\n" | 330 "movq %%mm4, %%mm5\n" |
428 "punpcklwd %%mm3, %%mm4\n" | 331 "punpcklwd %%mm3, %%mm4\n" |
429 #ifdef GCC2_HACK | |
430 "pmullw _MMX_Ycoeff, %%mm7\n" // lum4 | 332 "pmullw _MMX_Ycoeff, %%mm7\n" // lum4 |
431 #else | |
432 "pmullw %[_MMX_Ycoeff], %%mm7\n" // lum4 | |
433 #endif | |
434 "punpckhwd %%mm3, %%mm5\n" | 333 "punpckhwd %%mm3, %%mm5\n" |
435 | 334 |
436 "movq %%mm4, (%3)\n" // write row1 | 335 "movq %%mm4, (%3)\n" // write row1 |
437 "movq %%mm5, 8(%3)\n" // write row1 | 336 "movq %%mm5, 8(%3)\n" // write row1 |
438 | 337 |
455 "packuswb %%mm6, %%mm6\n" | 354 "packuswb %%mm6, %%mm6\n" |
456 "punpcklbw %%mm4, %%mm4\n" | 355 "punpcklbw %%mm4, %%mm4\n" |
457 "punpcklbw %%mm5, %%mm5\n" | 356 "punpcklbw %%mm5, %%mm5\n" |
458 "punpcklbw %%mm6, %%mm6\n" | 357 "punpcklbw %%mm6, %%mm6\n" |
459 "psllw $3, %%mm5\n" // GREEN 3 | 358 "psllw $3, %%mm5\n" // GREEN 3 |
460 #ifdef GCC2_HACK | |
461 "pand _MMX_red565, %%mm4\n" | 359 "pand _MMX_red565, %%mm4\n" |
462 #else | |
463 "pand %[_MMX_red565], %%mm4\n" | |
464 #endif | |
465 "psraw $6, %%mm3\n" // psr 6 | 360 "psraw $6, %%mm3\n" // psr 6 |
466 "psraw $6, %%mm0\n" | 361 "psraw $6, %%mm0\n" |
467 #ifdef GCC2_HACK | |
468 "pand _MMX_red565, %%mm6\n" // BLUE | 362 "pand _MMX_red565, %%mm6\n" // BLUE |
469 "pand _MMX_grn565, %%mm5\n" | 363 "pand _MMX_grn565, %%mm5\n" |
470 #else | |
471 "pand %[_MMX_red565], %%mm6\n" // BLUE | |
472 "pand %[_MMX_grn565], %%mm5\n" | |
473 #endif | |
474 "psrlw $11, %%mm6\n" // BLUE 3 | 364 "psrlw $11, %%mm6\n" // BLUE 3 |
475 "por %%mm5, %%mm4\n" | 365 "por %%mm5, %%mm4\n" |
476 "psraw $6, %%mm7\n" | 366 "psraw $6, %%mm7\n" |
477 "por %%mm6, %%mm4\n" | 367 "por %%mm6, %%mm4\n" |
478 "packuswb %%mm3, %%mm3\n" | 368 "packuswb %%mm3, %%mm3\n" |
479 "packuswb %%mm0, %%mm0\n" | 369 "packuswb %%mm0, %%mm0\n" |
480 "packuswb %%mm7, %%mm7\n" | 370 "packuswb %%mm7, %%mm7\n" |
481 "punpcklbw %%mm3, %%mm3\n" | 371 "punpcklbw %%mm3, %%mm3\n" |
482 "punpcklbw %%mm0, %%mm0\n" | 372 "punpcklbw %%mm0, %%mm0\n" |
483 "punpcklbw %%mm7, %%mm7\n" | 373 "punpcklbw %%mm7, %%mm7\n" |
484 #ifdef GCC2_HACK | |
485 "pand _MMX_red565, %%mm3\n" | 374 "pand _MMX_red565, %%mm3\n" |
486 "pand _MMX_red565, %%mm7\n" // BLUE | 375 "pand _MMX_red565, %%mm7\n" // BLUE |
487 #else | |
488 "pand %[_MMX_red565], %%mm3\n" | |
489 "pand %[_MMX_red565], %%mm7\n" // BLUE | |
490 #endif | |
491 "psllw $3, %%mm0\n" // GREEN 4 | 376 "psllw $3, %%mm0\n" // GREEN 4 |
492 "psrlw $11, %%mm7\n" | 377 "psrlw $11, %%mm7\n" |
493 #ifdef GCC2_HACK | |
494 "pand _MMX_grn565, %%mm0\n" | 378 "pand _MMX_grn565, %%mm0\n" |
495 #else | |
496 "pand %[_MMX_grn565], %%mm0\n" | |
497 #endif | |
498 "por %%mm7, %%mm3\n" | 379 "por %%mm7, %%mm3\n" |
499 "por %%mm0, %%mm3\n" | 380 "por %%mm0, %%mm3\n" |
500 | 381 |
501 "movq %%mm4, %%mm5\n" | 382 "movq %%mm4, %%mm5\n" |
502 | 383 |
524 "jl 1b\n" | 405 "jl 1b\n" |
525 "emms\n" | 406 "emms\n" |
526 "popl %%ebx\n" | 407 "popl %%ebx\n" |
527 : | 408 : |
528 :"m" (cr), "r"(cb),"r"(lum), | 409 :"m" (cr), "r"(cb),"r"(lum), |
529 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod) | 410 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)); |
530 #ifndef GCC2_HACK | |
531 ,[_MMX_0080w] "m" (*MMX_0080w), | |
532 [_MMX_Ugrn565] "m" (*MMX_Ugrn565), | |
533 [_MMX_Ublu5x5] "m" (*MMX_Ublu5x5), | |
534 [_MMX_00FFw] "m" (*MMX_00FFw), | |
535 [_MMX_Vgrn565] "m" (*MMX_Vgrn565), | |
536 [_MMX_Vred5x5] "m" (*MMX_Vred5x5), | |
537 [_MMX_Ycoeff] "m" (*MMX_Ycoeff), | |
538 [_MMX_red565] "m" (*MMX_red565), | |
539 [_MMX_grn565] "m" (*MMX_grn565) | |
540 #endif | |
541 ); | |
542 } | 411 } |
543 | 412 |
544 #undef GCC2_HACK | |
545 | |
546 #endif /* GCC i386 inline assembly */ | 413 #endif /* GCC i386 inline assembly */ |