|
Lines 29-34
Link Here
|
| 29 |
|
29 |
|
| 30 |
#ifdef RENDER |
30 |
#ifdef RENDER |
| 31 |
|
31 |
|
|
|
32 |
#include <mmintrin.h> |
| 33 |
|
| 32 |
#include "picturestr.h" |
34 |
#include "picturestr.h" |
| 33 |
#include "mipict.h" |
35 |
#include "mipict.h" |
| 34 |
#include "fbpict.h" |
36 |
#include "fbpict.h" |
|
Lines 48-53
Link Here
|
| 48 |
#define CHECKPOINT() |
50 |
#define CHECKPOINT() |
| 49 |
#endif |
51 |
#endif |
| 50 |
|
52 |
|
|
|
53 |
#define mmx_and(a,b) ((Vector1x64)_mm_and_si64((__m64)(a),(__m64)(b))) |
| 54 |
#define mmx_or(a,b) ((Vector1x64)_mm_or_si64((__m64)(a),(__m64)(b))) |
| 55 |
#define mmx_xor(a,b) ((Vector1x64)_mm_xor_si64((__m64)(a),(__m64)(b))) |
| 56 |
|
| 51 |
typedef struct |
57 |
typedef struct |
| 52 |
{ |
58 |
{ |
| 53 |
ullong mmx_zero; |
59 |
ullong mmx_zero; |
|
Lines 102-108
Link Here
|
| 102 |
static __inline__ Vector4x16 |
108 |
static __inline__ Vector4x16 |
| 103 |
negate (Vector4x16 mask) |
109 |
negate (Vector4x16 mask) |
| 104 |
{ |
110 |
{ |
| 105 |
return (Vector4x16)__builtin_ia32_pxor ( |
111 |
return (Vector4x16)mmx_xor ( |
| 106 |
(Vector1x64)mask, |
112 |
(Vector1x64)mask, |
| 107 |
(Vector1x64)c.mmx_4x00ff); |
113 |
(Vector1x64)c.mmx_4x00ff); |
| 108 |
} |
114 |
} |
|
Lines 163-171
Link Here
|
| 163 |
|
169 |
|
| 164 |
t1 = shift ((Vector1x64)pixel, -48); |
170 |
t1 = shift ((Vector1x64)pixel, -48); |
| 165 |
t2 = shift (t1, 16); |
171 |
t2 = shift (t1, 16); |
| 166 |
t1 = __builtin_ia32_por (t1, t2); |
172 |
t1 = mmx_or (t1, t2); |
| 167 |
t2 = shift (t1, 32); |
173 |
t2 = shift (t1, 32); |
| 168 |
t1 = __builtin_ia32_por (t1, t2); |
174 |
t1 = mmx_or (t1, t2); |
| 169 |
|
175 |
|
| 170 |
return (Vector4x16)t1; |
176 |
return (Vector4x16)t1; |
| 171 |
} |
177 |
} |
|
Lines 178-186
Link Here
|
| 178 |
t1 = shift ((Vector1x64)pixel, 48); |
184 |
t1 = shift ((Vector1x64)pixel, 48); |
| 179 |
t1 = shift (t1, -48); |
185 |
t1 = shift (t1, -48); |
| 180 |
t2 = shift (t1, 16); |
186 |
t2 = shift (t1, 16); |
| 181 |
t1 = __builtin_ia32_por (t1, t2); |
187 |
t1 = mmx_or (t1, t2); |
| 182 |
t2 = shift (t1, 32); |
188 |
t2 = shift (t1, 32); |
| 183 |
t1 = __builtin_ia32_por (t1, t2); |
189 |
t1 = mmx_or (t1, t2); |
| 184 |
|
190 |
|
| 185 |
return (Vector4x16)t1; |
191 |
return (Vector4x16)t1; |
| 186 |
} |
192 |
} |
|
Lines 192-206
Link Here
|
| 192 |
|
198 |
|
| 193 |
x = y = z = (Vector1x64)pixel; |
199 |
x = y = z = (Vector1x64)pixel; |
| 194 |
|
200 |
|
| 195 |
x = __builtin_ia32_pand (x, (Vector1x64)c.mmx_ffff0000ffff0000); |
201 |
x = mmx_and (x, (Vector1x64)c.mmx_ffff0000ffff0000); |
| 196 |
y = __builtin_ia32_pand (y, (Vector1x64)c.mmx_000000000000ffff); |
202 |
y = mmx_and (y, (Vector1x64)c.mmx_000000000000ffff); |
| 197 |
z = __builtin_ia32_pand (z, (Vector1x64)c.mmx_0000ffff00000000); |
203 |
z = mmx_and (z, (Vector1x64)c.mmx_0000ffff00000000); |
| 198 |
|
204 |
|
| 199 |
y = shift (y, 32); |
205 |
y = shift (y, 32); |
| 200 |
z = shift (z, -32); |
206 |
z = shift (z, -32); |
| 201 |
|
207 |
|
| 202 |
x = __builtin_ia32_por (x, y); |
208 |
x = mmx_or (x, y); |
| 203 |
x = __builtin_ia32_por (x, z); |
209 |
x = mmx_or (x, z); |
| 204 |
|
210 |
|
| 205 |
return (Vector4x16)x; |
211 |
return (Vector4x16)x; |
| 206 |
} |
212 |
} |
|
Lines 234-240
Link Here
|
| 234 |
over_rev_non_pre (Vector4x16 src, Vector4x16 dest) |
240 |
over_rev_non_pre (Vector4x16 src, Vector4x16 dest) |
| 235 |
{ |
241 |
{ |
| 236 |
Vector4x16 srca = expand_alpha (src); |
242 |
Vector4x16 srca = expand_alpha (src); |
| 237 |
Vector4x16 srcfaaa = (Vector4x16)__builtin_ia32_por((Vector1x64)srca, (Vector1x64)c.mmx_full_alpha); |
243 |
Vector4x16 srcfaaa = (Vector4x16)mmx_or((Vector1x64)srca, (Vector1x64)c.mmx_full_alpha); |
| 238 |
|
244 |
|
| 239 |
return over(pix_multiply(invert_colors(src), srcfaaa), srca, dest); |
245 |
return over(pix_multiply(invert_colors(src), srcfaaa), srca, dest); |
| 240 |
} |
246 |
} |
|
Lines 300-308
Link Here
|
| 300 |
Vector1x64 t1 = shift (p, 36 - 11); |
306 |
Vector1x64 t1 = shift (p, 36 - 11); |
| 301 |
Vector1x64 t2 = shift (p, 16 - 5); |
307 |
Vector1x64 t2 = shift (p, 16 - 5); |
| 302 |
|
308 |
|
| 303 |
p = __builtin_ia32_por (t1, p); |
309 |
p = mmx_or (t1, p); |
| 304 |
p = __builtin_ia32_por (t2, p); |
310 |
p = mmx_or (t2, p); |
| 305 |
p = __builtin_ia32_pand (p, (Vector1x64)c.mmx_565_rgb); |
311 |
p = mmx_and (p, (Vector1x64)c.mmx_565_rgb); |
| 306 |
|
312 |
|
| 307 |
pixel = __builtin_ia32_pmullw ((Vector4x16)p, (Vector4x16)c.mmx_565_unpack_multiplier); |
313 |
pixel = __builtin_ia32_pmullw ((Vector4x16)p, (Vector4x16)c.mmx_565_unpack_multiplier); |
| 308 |
return __builtin_ia32_psrlw (pixel, 8); |
314 |
return __builtin_ia32_psrlw (pixel, 8); |
|
Lines 324-350
Link Here
|
| 324 |
Vector1x64 t = (Vector1x64)target; |
330 |
Vector1x64 t = (Vector1x64)target; |
| 325 |
Vector1x64 r, g, b; |
331 |
Vector1x64 r, g, b; |
| 326 |
|
332 |
|
| 327 |
r = __builtin_ia32_pand (p, (Vector1x64)c.mmx_565_r); |
333 |
r = mmx_and (p, (Vector1x64)c.mmx_565_r); |
| 328 |
g = __builtin_ia32_pand (p, (Vector1x64)c.mmx_565_g); |
334 |
g = mmx_and (p, (Vector1x64)c.mmx_565_g); |
| 329 |
b = __builtin_ia32_pand (p, (Vector1x64)c.mmx_565_b); |
335 |
b = mmx_and (p, (Vector1x64)c.mmx_565_b); |
| 330 |
|
336 |
|
| 331 |
r = shift (r, - (32 - 8) + pos * 16); |
337 |
r = shift (r, - (32 - 8) + pos * 16); |
| 332 |
g = shift (g, - (16 - 3) + pos * 16); |
338 |
g = shift (g, - (16 - 3) + pos * 16); |
| 333 |
b = shift (b, - (0 + 3) + pos * 16); |
339 |
b = shift (b, - (0 + 3) + pos * 16); |
| 334 |
|
340 |
|
| 335 |
if (pos == 0) |
341 |
if (pos == 0) |
| 336 |
t = __builtin_ia32_pand (t, (Vector1x64)c.mmx_mask_0); |
342 |
t = mmx_and (t, (Vector1x64)c.mmx_mask_0); |
| 337 |
else if (pos == 1) |
343 |
else if (pos == 1) |
| 338 |
t = __builtin_ia32_pand (t, (Vector1x64)c.mmx_mask_1); |
344 |
t = mmx_and (t, (Vector1x64)c.mmx_mask_1); |
| 339 |
else if (pos == 2) |
345 |
else if (pos == 2) |
| 340 |
t = __builtin_ia32_pand (t, (Vector1x64)c.mmx_mask_2); |
346 |
t = mmx_and (t, (Vector1x64)c.mmx_mask_2); |
| 341 |
else if (pos == 3) |
347 |
else if (pos == 3) |
| 342 |
t = __builtin_ia32_pand (t, (Vector1x64)c.mmx_mask_3); |
348 |
t = mmx_and (t, (Vector1x64)c.mmx_mask_3); |
| 343 |
|
349 |
|
| 344 |
p = __builtin_ia32_por (r, t); |
350 |
p = mmx_or (r, t); |
| 345 |
p = __builtin_ia32_por (g, p); |
351 |
p = mmx_or (g, p); |
| 346 |
|
352 |
|
| 347 |
return (Vector4x16)__builtin_ia32_por (b, p); |
353 |
return (Vector4x16)mmx_or (b, p); |
| 348 |
} |
354 |
} |
| 349 |
|
355 |
|
| 350 |
static __inline__ void |
356 |
static __inline__ void |
