Contents of /trunk/libsdl/patches/libsdl-1.2.9-gcc-mmx.patch
Parent Directory | Revision Log
Revision 144 -
(show annotations)
(download)
Tue May 8 20:06:05 2007 UTC (17 years, 4 months ago) by niro
File size: 15320 byte(s)
Tue May 8 20:06:05 2007 UTC (17 years, 4 months ago) by niro
File size: 15320 byte(s)
-import
1 | http://bugs.gentoo.org/104437 |
2 | |
3 | get this mmx stuff to build with gcc2/gcc3/gcc4 |
4 | |
5 | Index: src/video/SDL_yuv_mmx.c |
6 | =================================================================== |
7 | RCS file: /home/sdlweb/libsdl.org/cvs/SDL12/src/video/SDL_yuv_mmx.c,v |
8 | retrieving revision 1.9 |
9 | retrieving revision 1.10 |
10 | diff -u -p -r1.9 -r1.10 |
11 | --- src/video/SDL_yuv_mmx.c 13 Feb 2005 07:10:02 -0000 1.9 |
12 | +++ src/video/SDL_yuv_mmx.c 28 Sep 2005 02:26:29 -0000 1.10 |
13 | @@ -30,42 +30,31 @@ static char rcsid = |
14 | |
15 | #include "SDL_types.h" |
16 | |
17 | -#if __GNUC__ > 2 |
18 | -# undef GCC2_HACK |
19 | -#else |
20 | -# define GCC2_HACK |
21 | -#endif |
22 | - |
23 | - |
24 | -#if defined(GCC2_HACK) && defined (__ELF__) |
25 | -#define ASM_VAR(X) _##X |
26 | -#else |
27 | -#define ASM_VAR(X) X |
28 | -#endif |
29 | +#define ASM_ARRAY(x) x[] __asm__("_" #x) __attribute__((used)) |
30 | |
31 | -static volatile unsigned int ASM_VAR(MMX_0080w)[] = {0x00800080, 0x00800080}; |
32 | -static volatile unsigned int ASM_VAR(MMX_00FFw)[] = {0x00ff00ff, 0x00ff00ff}; |
33 | -static volatile unsigned int ASM_VAR(MMX_FF00w)[] = {0xff00ff00, 0xff00ff00}; |
34 | - |
35 | -static volatile unsigned short ASM_VAR(MMX_Ycoeff)[] = {0x004a, 0x004a, 0x004a, 0x004a}; |
36 | - |
37 | -static volatile unsigned short ASM_VAR(MMX_UbluRGB)[] = {0x0072, 0x0072, 0x0072, 0x0072}; |
38 | -static volatile unsigned short ASM_VAR(MMX_VredRGB)[] = {0x0059, 0x0059, 0x0059, 0x0059}; |
39 | -static volatile unsigned short ASM_VAR(MMX_UgrnRGB)[] = {0xffea, 0xffea, 0xffea, 0xffea}; |
40 | -static volatile unsigned short ASM_VAR(MMX_VgrnRGB)[] = {0xffd2, 0xffd2, 0xffd2, 0xffd2}; |
41 | - |
42 | -static volatile unsigned short ASM_VAR(MMX_Ublu5x5)[] = {0x0081, 0x0081, 0x0081, 0x0081}; |
43 | -static volatile unsigned short ASM_VAR(MMX_Vred5x5)[] = {0x0066, 0x0066, 0x0066, 0x0066}; |
44 | -static volatile unsigned short ASM_VAR(MMX_Ugrn555)[] = {0xffe7, 0xffe7, 0xffe7, 0xffe7}; |
45 | -static volatile unsigned short ASM_VAR(MMX_Vgrn555)[] = {0xffcc, 0xffcc, 0xffcc, 0xffcc}; |
46 | -static volatile unsigned short ASM_VAR(MMX_Ugrn565)[] = {0xffe8, 0xffe8, 0xffe8, 0xffe8}; |
47 | -static volatile unsigned short ASM_VAR(MMX_Vgrn565)[] = {0xffcd, 0xffcd, 0xffcd, 0xffcd}; |
48 | - |
49 | -static volatile unsigned short ASM_VAR(MMX_red555)[] = {0x7c00, 0x7c00, 0x7c00, 0x7c00}; |
50 | -static volatile unsigned short ASM_VAR(MMX_red565)[] = {0xf800, 0xf800, 0xf800, 0xf800}; |
51 | -static volatile unsigned short ASM_VAR(MMX_grn555)[] = {0x03e0, 0x03e0, 0x03e0, 0x03e0}; |
52 | -static volatile unsigned short ASM_VAR(MMX_grn565)[] = {0x07e0, 0x07e0, 0x07e0, 0x07e0}; |
53 | -static volatile unsigned short ASM_VAR(MMX_blu5x5)[] = {0x001f, 0x001f, 0x001f, 0x001f}; |
54 | +static unsigned int ASM_ARRAY(MMX_0080w) = {0x00800080, 0x00800080}; |
55 | +static unsigned int ASM_ARRAY(MMX_00FFw) = {0x00ff00ff, 0x00ff00ff}; |
56 | +static unsigned int ASM_ARRAY(MMX_FF00w) = {0xff00ff00, 0xff00ff00}; |
57 | + |
58 | +static unsigned short ASM_ARRAY(MMX_Ycoeff) = {0x004a, 0x004a, 0x004a, 0x004a}; |
59 | + |
60 | +static unsigned short ASM_ARRAY(MMX_UbluRGB) = {0x0072, 0x0072, 0x0072, 0x0072}; |
61 | +static unsigned short ASM_ARRAY(MMX_VredRGB) = {0x0059, 0x0059, 0x0059, 0x0059}; |
62 | +static unsigned short ASM_ARRAY(MMX_UgrnRGB) = {0xffea, 0xffea, 0xffea, 0xffea}; |
63 | +static unsigned short ASM_ARRAY(MMX_VgrnRGB) = {0xffd2, 0xffd2, 0xffd2, 0xffd2}; |
64 | + |
65 | +static unsigned short ASM_ARRAY(MMX_Ublu5x5) = {0x0081, 0x0081, 0x0081, 0x0081}; |
66 | +static unsigned short ASM_ARRAY(MMX_Vred5x5) = {0x0066, 0x0066, 0x0066, 0x0066}; |
67 | +static unsigned short ASM_ARRAY(MMX_Ugrn555) = {0xffe7, 0xffe7, 0xffe7, 0xffe7}; |
68 | +static unsigned short ASM_ARRAY(MMX_Vgrn555) = {0xffcc, 0xffcc, 0xffcc, 0xffcc}; |
69 | +static unsigned short ASM_ARRAY(MMX_Ugrn565) = {0xffe8, 0xffe8, 0xffe8, 0xffe8}; |
70 | +static unsigned short ASM_ARRAY(MMX_Vgrn565) = {0xffcd, 0xffcd, 0xffcd, 0xffcd}; |
71 | + |
72 | +static unsigned short ASM_ARRAY(MMX_red555) = {0x7c00, 0x7c00, 0x7c00, 0x7c00}; |
73 | +static unsigned short ASM_ARRAY(MMX_red565) = {0xf800, 0xf800, 0xf800, 0xf800}; |
74 | +static unsigned short ASM_ARRAY(MMX_grn555) = {0x03e0, 0x03e0, 0x03e0, 0x03e0}; |
75 | +static unsigned short ASM_ARRAY(MMX_grn565) = {0x07e0, 0x07e0, 0x07e0, 0x07e0}; |
76 | +static unsigned short ASM_ARRAY(MMX_blu5x5) = {0x001f, 0x001f, 0x001f, 0x001f}; |
77 | |
78 | /** |
79 | This MMX assembler is my first assembler/MMX program ever. |
80 | @@ -127,21 +116,12 @@ void ColorRGBDitherYV12MMX1X( int *color |
81 | "movd (%2), %%mm2\n" // 0 0 0 0 l3 l2 l1 l0 |
82 | "punpcklbw %%mm7,%%mm1\n" // 0 v3 0 v2 00 v1 00 v0 |
83 | "punpckldq %%mm1,%%mm1\n" // 00 v1 00 v0 00 v1 00 v0 |
84 | -#ifdef GCC2_HACK |
85 | "psubw _MMX_0080w,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 |
86 | -#else |
87 | - "psubw %[_MMX_0080w],%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 |
88 | -#endif |
89 | |
90 | // create Cr_g (result in mm0) |
91 | "movq %%mm1,%%mm0\n" // r1 r1 r0 r0 r1 r1 r0 r0 |
92 | -#ifdef GCC2_HACK |
93 | "pmullw _MMX_VgrnRGB,%%mm0\n"// red*-46dec=0.7136*64 |
94 | "pmullw _MMX_VredRGB,%%mm1\n"// red*89dec=1.4013*64 |
95 | -#else |
96 | - "pmullw %[_MMX_VgrnRGB],%%mm0\n"// red*-46dec=0.7136*64 |
97 | - "pmullw %[_MMX_VredRGB],%%mm1\n"// red*89dec=1.4013*64 |
98 | -#endif |
99 | "psraw $6, %%mm0\n" // red=red/64 |
100 | "psraw $6, %%mm1\n" // red=red/64 |
101 | |
102 | @@ -150,13 +130,8 @@ void ColorRGBDitherYV12MMX1X( int *color |
103 | "movq (%2,%4),%%mm3\n" // 0 0 0 0 L3 L2 L1 L0 |
104 | "punpckldq %%mm3,%%mm2\n" // L3 L2 L1 L0 l3 l2 l1 l0 |
105 | "movq %%mm2,%%mm4\n" // L3 L2 L1 L0 l3 l2 l1 l0 |
106 | -#ifdef GCC2_HACK |
107 | "pand _MMX_FF00w,%%mm2\n" // L3 0 L1 0 l3 0 l1 0 |
108 | "pand _MMX_00FFw,%%mm4\n" // 0 L2 0 L0 0 l2 0 l0 |
109 | -#else |
110 | - "pand %[_MMX_FF00w],%%mm2\n" // L3 0 L1 0 l3 0 l1 0 |
111 | - "pand %[_MMX_00FFw],%%mm4\n" // 0 L2 0 L0 0 l2 0 l0 |
112 | -#endif |
113 | "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1 |
114 | |
115 | // create R (result in mm6) |
116 | @@ -173,20 +148,11 @@ void ColorRGBDitherYV12MMX1X( int *color |
117 | "movd (%1), %%mm1\n" // 0 0 0 0 u3 u2 u1 u0 |
118 | "punpcklbw %%mm7,%%mm1\n" // 0 u3 0 u2 00 u1 00 u0 |
119 | "punpckldq %%mm1,%%mm1\n" // 00 u1 00 u0 00 u1 00 u0 |
120 | -#ifdef GCC2_HACK |
121 | "psubw _MMX_0080w,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 |
122 | -#else |
123 | - "psubw %[_MMX_0080w],%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 |
124 | -#endif |
125 | // create Cb_g (result in mm5) |
126 | "movq %%mm1,%%mm5\n" // u1 u1 u0 u0 u1 u1 u0 u0 |
127 | -#ifdef GCC2_HACK |
128 | "pmullw _MMX_UgrnRGB,%%mm5\n" // blue*-109dec=1.7129*64 |
129 | "pmullw _MMX_UbluRGB,%%mm1\n" // blue*114dec=1.78125*64 |
130 | -#else |
131 | - "pmullw %[_MMX_UgrnRGB],%%mm5\n" // blue*-109dec=1.7129*64 |
132 | - "pmullw %[_MMX_UbluRGB],%%mm1\n" // blue*114dec=1.78125*64 |
133 | -#endif |
134 | "psraw $6, %%mm5\n" // blue=red/64 |
135 | "psraw $6, %%mm1\n" // blue=blue/64 |
136 | |
137 | @@ -268,17 +234,7 @@ void ColorRGBDitherYV12MMX1X( int *color |
138 | "popl %%ebx\n" |
139 | : |
140 | : "m" (cr), "r"(cb),"r"(lum), |
141 | - "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod) |
142 | -#ifndef GCC2_HACK |
143 | - ,[_MMX_0080w] "m" (*MMX_0080w), |
144 | - [_MMX_00FFw] "m" (*MMX_00FFw), |
145 | - [_MMX_FF00w] "m" (*MMX_FF00w), |
146 | - [_MMX_VgrnRGB] "m" (*MMX_VgrnRGB), |
147 | - [_MMX_VredRGB] "m" (*MMX_VredRGB), |
148 | - [_MMX_UgrnRGB] "m" (*MMX_UgrnRGB), |
149 | - [_MMX_UbluRGB] "m" (*MMX_UbluRGB) |
150 | -#endif |
151 | - ); |
152 | + "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)); |
153 | } |
154 | |
155 | void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, |
156 | @@ -307,48 +263,21 @@ void Color565DitherYV12MMX1X( int *color |
157 | "movd (%%ebx), %%mm1\n" // 4 Cr 0 0 0 0 v3 v2 v1 v0 |
158 | "punpcklbw %%mm7, %%mm0\n" // 4 W cb 0 u3 0 u2 0 u1 0 u0 |
159 | "punpcklbw %%mm7, %%mm1\n" // 4 W cr 0 v3 0 v2 0 v1 0 v0 |
160 | -#ifdef GCC2_HACK |
161 | "psubw _MMX_0080w, %%mm0\n" |
162 | "psubw _MMX_0080w, %%mm1\n" |
163 | -#else |
164 | - "psubw %[_MMX_0080w], %%mm0\n" |
165 | - "psubw %[_MMX_0080w], %%mm1\n" |
166 | -#endif |
167 | "movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0 |
168 | "movq %%mm1, %%mm3\n" // Cr |
169 | -#ifdef GCC2_HACK |
170 | "pmullw _MMX_Ugrn565, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0 |
171 | -#else |
172 | - "pmullw %[_MMX_Ugrn565], %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0 |
173 | -#endif |
174 | "movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0 |
175 | -#ifdef GCC2_HACK |
176 | "pmullw _MMX_Ublu5x5, %%mm0\n" // Cb2blue |
177 | "pand _MMX_00FFw, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0 |
178 | "pmullw _MMX_Vgrn565, %%mm3\n" // Cr2green |
179 | -#else |
180 | - "pmullw %[_MMX_Ublu5x5], %%mm0\n" // Cb2blue |
181 | - "pand %[_MMX_00FFw], %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0 |
182 | - "pmullw %[_MMX_Vgrn565], %%mm3\n" // Cr2green |
183 | -#endif |
184 | "movq (%2), %%mm7\n" // L2 |
185 | -#ifdef GCC2_HACK |
186 | "pmullw _MMX_Vred5x5, %%mm1\n" // Cr2red |
187 | -#else |
188 | - "pmullw %[_MMX_Vred5x5], %%mm1\n" // Cr2red |
189 | -#endif |
190 | "psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1 |
191 | -#ifdef GCC2_HACK |
192 | "pmullw _MMX_Ycoeff, %%mm6\n" // lum1 |
193 | -#else |
194 | - "pmullw %[_MMX_Ycoeff], %%mm6\n" // lum1 |
195 | -#endif |
196 | "paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green |
197 | -#ifdef GCC2_HACK |
198 | "pmullw _MMX_Ycoeff, %%mm7\n" // lum2 |
199 | -#else |
200 | - "pmullw %[_MMX_Ycoeff], %%mm7\n" // lum2 |
201 | -#endif |
202 | |
203 | "movq %%mm6, %%mm4\n" // lum1 |
204 | "paddw %%mm0, %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0 |
205 | @@ -366,20 +295,11 @@ void Color565DitherYV12MMX1X( int *color |
206 | "punpcklbw %%mm4, %%mm4\n" |
207 | "punpcklbw %%mm5, %%mm5\n" |
208 | |
209 | -#ifdef GCC2_HACK |
210 | "pand _MMX_red565, %%mm4\n" |
211 | -#else |
212 | - "pand %[_MMX_red565], %%mm4\n" |
213 | -#endif |
214 | "psllw $3, %%mm5\n" // GREEN 1 |
215 | "punpcklbw %%mm6, %%mm6\n" |
216 | -#ifdef GCC2_HACK |
217 | "pand _MMX_grn565, %%mm5\n" |
218 | "pand _MMX_red565, %%mm6\n" |
219 | -#else |
220 | - "pand %[_MMX_grn565], %%mm5\n" |
221 | - "pand %[_MMX_red565], %%mm6\n" |
222 | -#endif |
223 | "por %%mm5, %%mm4\n" // |
224 | "psrlw $11, %%mm6\n" // BLUE 1 |
225 | "movq %%mm3, %%mm5\n" // lum2 |
226 | @@ -393,44 +313,23 @@ void Color565DitherYV12MMX1X( int *color |
227 | "packuswb %%mm3, %%mm3\n" |
228 | "packuswb %%mm5, %%mm5\n" |
229 | "packuswb %%mm7, %%mm7\n" |
230 | -#ifdef GCC2_HACK |
231 | "pand _MMX_00FFw, %%mm6\n" // L3 |
232 | -#else |
233 | - "pand %[_MMX_00FFw], %%mm6\n" // L3 |
234 | -#endif |
235 | "punpcklbw %%mm3, %%mm3\n" |
236 | "punpcklbw %%mm5, %%mm5\n" |
237 | -#ifdef GCC2_HACK |
238 | "pmullw _MMX_Ycoeff, %%mm6\n" // lum3 |
239 | -#else |
240 | - "pmullw %[_MMX_Ycoeff], %%mm6\n" // lum3 |
241 | -#endif |
242 | "punpcklbw %%mm7, %%mm7\n" |
243 | "psllw $3, %%mm5\n" // GREEN 2 |
244 | -#ifdef GCC2_HACK |
245 | "pand _MMX_red565, %%mm7\n" |
246 | "pand _MMX_red565, %%mm3\n" |
247 | -#else |
248 | - "pand %[_MMX_red565], %%mm7\n" |
249 | - "pand %[_MMX_red565], %%mm3\n" |
250 | -#endif |
251 | "psrlw $11, %%mm7\n" // BLUE 2 |
252 | -#ifdef GCC2_HACK |
253 | "pand _MMX_grn565, %%mm5\n" |
254 | -#else |
255 | - "pand %[_MMX_grn565], %%mm5\n" |
256 | -#endif |
257 | "por %%mm7, %%mm3\n" |
258 | "movq (%2,%4), %%mm7\n" // L4 load lum2 |
259 | "por %%mm5, %%mm3\n" // |
260 | "psrlw $8, %%mm7\n" // L4 |
261 | "movq %%mm4, %%mm5\n" |
262 | "punpcklwd %%mm3, %%mm4\n" |
263 | -#ifdef GCC2_HACK |
264 | "pmullw _MMX_Ycoeff, %%mm7\n" // lum4 |
265 | -#else |
266 | - "pmullw %[_MMX_Ycoeff], %%mm7\n" // lum4 |
267 | -#endif |
268 | "punpckhwd %%mm3, %%mm5\n" |
269 | |
270 | "movq %%mm4, (%3)\n" // write row1 |
271 | @@ -457,20 +356,11 @@ void Color565DitherYV12MMX1X( int *color |
272 | "punpcklbw %%mm5, %%mm5\n" |
273 | "punpcklbw %%mm6, %%mm6\n" |
274 | "psllw $3, %%mm5\n" // GREEN 3 |
275 | -#ifdef GCC2_HACK |
276 | "pand _MMX_red565, %%mm4\n" |
277 | -#else |
278 | - "pand %[_MMX_red565], %%mm4\n" |
279 | -#endif |
280 | "psraw $6, %%mm3\n" // psr 6 |
281 | "psraw $6, %%mm0\n" |
282 | -#ifdef GCC2_HACK |
283 | "pand _MMX_red565, %%mm6\n" // BLUE |
284 | "pand _MMX_grn565, %%mm5\n" |
285 | -#else |
286 | - "pand %[_MMX_red565], %%mm6\n" // BLUE |
287 | - "pand %[_MMX_grn565], %%mm5\n" |
288 | -#endif |
289 | "psrlw $11, %%mm6\n" // BLUE 3 |
290 | "por %%mm5, %%mm4\n" |
291 | "psraw $6, %%mm7\n" |
292 | @@ -481,20 +371,11 @@ void Color565DitherYV12MMX1X( int *color |
293 | "punpcklbw %%mm3, %%mm3\n" |
294 | "punpcklbw %%mm0, %%mm0\n" |
295 | "punpcklbw %%mm7, %%mm7\n" |
296 | -#ifdef GCC2_HACK |
297 | "pand _MMX_red565, %%mm3\n" |
298 | "pand _MMX_red565, %%mm7\n" // BLUE |
299 | -#else |
300 | - "pand %[_MMX_red565], %%mm3\n" |
301 | - "pand %[_MMX_red565], %%mm7\n" // BLUE |
302 | -#endif |
303 | "psllw $3, %%mm0\n" // GREEN 4 |
304 | "psrlw $11, %%mm7\n" |
305 | -#ifdef GCC2_HACK |
306 | "pand _MMX_grn565, %%mm0\n" |
307 | -#else |
308 | - "pand %[_MMX_grn565], %%mm0\n" |
309 | -#endif |
310 | "por %%mm7, %%mm3\n" |
311 | "por %%mm0, %%mm3\n" |
312 | |
313 | @@ -526,21 +407,7 @@ void Color565DitherYV12MMX1X( int *color |
314 | "popl %%ebx\n" |
315 | : |
316 | :"m" (cr), "r"(cb),"r"(lum), |
317 | - "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod) |
318 | -#ifndef GCC2_HACK |
319 | - ,[_MMX_0080w] "m" (*MMX_0080w), |
320 | - [_MMX_Ugrn565] "m" (*MMX_Ugrn565), |
321 | - [_MMX_Ublu5x5] "m" (*MMX_Ublu5x5), |
322 | - [_MMX_00FFw] "m" (*MMX_00FFw), |
323 | - [_MMX_Vgrn565] "m" (*MMX_Vgrn565), |
324 | - [_MMX_Vred5x5] "m" (*MMX_Vred5x5), |
325 | - [_MMX_Ycoeff] "m" (*MMX_Ycoeff), |
326 | - [_MMX_red565] "m" (*MMX_red565), |
327 | - [_MMX_grn565] "m" (*MMX_grn565) |
328 | -#endif |
329 | - ); |
330 | + "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)); |
331 | } |
332 | |
333 | -#undef GCC2_HACK |
334 | - |
335 | #endif /* GCC i386 inline assembly */ |