projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Merge branch 'glsl-pp-rework-2'
[mesa.git] / src / gallium / auxiliary / util / u_gen_mipmap.c
1 /**************************************************************************
2 *
3 * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
4 * All Rights Reserved.
5 * Copyright 2008 VMware, Inc. All rights reserved.
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the
9 * "Software"), to deal in the Software without restriction, including
10 * without limitation the rights to use, copy, modify, merge, publish,
11 * distribute, sub license, and/or sell copies of the Software, and to
12 * permit persons to whom the Software is furnished to do so, subject to
13 * the following conditions:
14 *
15 * The above copyright notice and this permission notice (including the
16 * next paragraph) shall be included in all copies or substantial portions
17 * of the Software.
18 *
19 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
20 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
21 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
22 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
23 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
24 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
25 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 *
27 **************************************************************************/
28
29 /**
30 * @file
31 * Mipmap generation utility
32 *
33 * @author Brian Paul
34 */
35
36
37 #include "pipe/p_context.h"
38 #include "util/u_debug.h"
39 #include "pipe/p_defines.h"
40 #include "pipe/p_inlines.h"
41 #include "pipe/p_shader_tokens.h"
42 #include "pipe/p_state.h"
43
44 #include "util/u_memory.h"
45 #include "util/u_draw_quad.h"
46 #include "util/u_gen_mipmap.h"
47 #include "util/u_simple_shaders.h"
48 #include "util/u_math.h"
49 #include "util/u_texture.h"
50
51 #include "cso_cache/cso_context.h"
52
53
54 struct gen_mipmap_state
55 {
56 struct pipe_context *pipe;
57 struct cso_context *cso;
58
59 struct pipe_blend_state blend;
60 struct pipe_depth_stencil_alpha_state depthstencil;
61 struct pipe_rasterizer_state rasterizer;
62 struct pipe_sampler_state sampler;
63
64 void *vs;
65 void *fs;
66
67 struct pipe_buffer *vbuf; /**< quad vertices */
68 unsigned vbuf_slot;
69
70 float vertices[4][2][4]; /**< vertex/texcoords for quad */
71 };
72
73
74
75 enum dtype
76 {
77 DTYPE_UBYTE,
78 DTYPE_UBYTE_3_3_2,
79 DTYPE_USHORT,
80 DTYPE_USHORT_4_4_4_4,
81 DTYPE_USHORT_5_6_5,
82 DTYPE_USHORT_1_5_5_5_REV,
83 DTYPE_UINT,
84 DTYPE_FLOAT,
85 DTYPE_HALF_FLOAT
86 };
87
88
89 typedef ushort half_float;
90
91
92 static half_float
93 float_to_half(float f)
94 {
95 /* XXX fix this */
96 return 0;
97 }
98
99 static float
100 half_to_float(half_float h)
101 {
102 /* XXX fix this */
103 return 0.0f;
104 }
105
106
107
108
109 /**
110 * \name Support macros for do_row and do_row_3d
111 *
112 * The macro madness is here for two reasons. First, it compacts the code
113 * slightly. Second, it makes it much easier to adjust the specifics of the
114 * filter to tune the rounding characteristics.
115 */
116 /*@{*/
117 #define DECLARE_ROW_POINTERS(t, e) \
118 const t(*rowA)[e] = (const t(*)[e]) srcRowA; \
119 const t(*rowB)[e] = (const t(*)[e]) srcRowB; \
120 const t(*rowC)[e] = (const t(*)[e]) srcRowC; \
121 const t(*rowD)[e] = (const t(*)[e]) srcRowD; \
122 t(*dst)[e] = (t(*)[e]) dstRow
123
124 #define DECLARE_ROW_POINTERS0(t) \
125 const t *rowA = (const t *) srcRowA; \
126 const t *rowB = (const t *) srcRowB; \
127 const t *rowC = (const t *) srcRowC; \
128 const t *rowD = (const t *) srcRowD; \
129 t *dst = (t *) dstRow
130
131 #define FILTER_SUM_3D(Aj, Ak, Bj, Bk, Cj, Ck, Dj, Dk) \
132 ((unsigned) Aj + (unsigned) Ak \
133 + (unsigned) Bj + (unsigned) Bk \
134 + (unsigned) Cj + (unsigned) Ck \
135 + (unsigned) Dj + (unsigned) Dk \
136 + 4) >> 3
137
138 #define FILTER_3D(e) \
139 do { \
140 dst[i][e] = FILTER_SUM_3D(rowA[j][e], rowA[k][e], \
141 rowB[j][e], rowB[k][e], \
142 rowC[j][e], rowC[k][e], \
143 rowD[j][e], rowD[k][e]); \
144 } while(0)
145
146 #define FILTER_F_3D(e) \
147 do { \
148 dst[i][e] = (rowA[j][e] + rowA[k][e] \
149 + rowB[j][e] + rowB[k][e] \
150 + rowC[j][e] + rowC[k][e] \
151 + rowD[j][e] + rowD[k][e]) * 0.125F; \
152 } while(0)
153
154 #define FILTER_HF_3D(e) \
155 do { \
156 const float aj = half_to_float(rowA[j][e]); \
157 const float ak = half_to_float(rowA[k][e]); \
158 const float bj = half_to_float(rowB[j][e]); \
159 const float bk = half_to_float(rowB[k][e]); \
160 const float cj = half_to_float(rowC[j][e]); \
161 const float ck = half_to_float(rowC[k][e]); \
162 const float dj = half_to_float(rowD[j][e]); \
163 const float dk = half_to_float(rowD[k][e]); \
164 dst[i][e] = float_to_half((aj + ak + bj + bk + cj + ck + dj + dk) \
165 * 0.125F); \
166 } while(0)
167 /*@}*/
168
169
170 /**
171 * Average together two rows of a source image to produce a single new
172 * row in the dest image. It's legal for the two source rows to point
173 * to the same data. The source width must be equal to either the
174 * dest width or two times the dest width.
175 * \param datatype GL_UNSIGNED_BYTE, GL_UNSIGNED_SHORT, GL_FLOAT, etc.
176 * \param comps number of components per pixel (1..4)
177 */
178 static void
179 do_row(enum dtype datatype, uint comps, int srcWidth,
180 const void *srcRowA, const void *srcRowB,
181 int dstWidth, void *dstRow)
182 {
183 const uint k0 = (srcWidth == dstWidth) ? 0 : 1;
184 const uint colStride = (srcWidth == dstWidth) ? 1 : 2;
185
186 assert(comps >= 1);
187 assert(comps <= 4);
188
189 /* This assertion is no longer valid with non-power-of-2 textures
190 assert(srcWidth == dstWidth || srcWidth == 2 * dstWidth);
191 */
192
193 if (datatype == DTYPE_UBYTE && comps == 4) {
194 uint i, j, k;
195 const ubyte(*rowA)[4] = (const ubyte(*)[4]) srcRowA;
196 const ubyte(*rowB)[4] = (const ubyte(*)[4]) srcRowB;
197 ubyte(*dst)[4] = (ubyte(*)[4]) dstRow;
198 for (i = j = 0, k = k0; i < (uint) dstWidth;
199 i++, j += colStride, k += colStride) {
200 dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
201 dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
202 dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4;
203 dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4;
204 }
205 }
206 else if (datatype == DTYPE_UBYTE && comps == 3) {
207 uint i, j, k;
208 const ubyte(*rowA)[3] = (const ubyte(*)[3]) srcRowA;
209 const ubyte(*rowB)[3] = (const ubyte(*)[3]) srcRowB;
210 ubyte(*dst)[3] = (ubyte(*)[3]) dstRow;
211 for (i = j = 0, k = k0; i < (uint) dstWidth;
212 i++, j += colStride, k += colStride) {
213 dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
214 dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
215 dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4;
216 }
217 }
218 else if (datatype == DTYPE_UBYTE && comps == 2) {
219 uint i, j, k;
220 const ubyte(*rowA)[2] = (const ubyte(*)[2]) srcRowA;
221 const ubyte(*rowB)[2] = (const ubyte(*)[2]) srcRowB;
222 ubyte(*dst)[2] = (ubyte(*)[2]) dstRow;
223 for (i = j = 0, k = k0; i < (uint) dstWidth;
224 i++, j += colStride, k += colStride) {
225 dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) >> 2;
226 dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) >> 2;
227 }
228 }
229 else if (datatype == DTYPE_UBYTE && comps == 1) {
230 uint i, j, k;
231 const ubyte *rowA = (const ubyte *) srcRowA;
232 const ubyte *rowB = (const ubyte *) srcRowB;
233 ubyte *dst = (ubyte *) dstRow;
234 for (i = j = 0, k = k0; i < (uint) dstWidth;
235 i++, j += colStride, k += colStride) {
236 dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) >> 2;
237 }
238 }
239
240 else if (datatype == DTYPE_USHORT && comps == 4) {
241 uint i, j, k;
242 const ushort(*rowA)[4] = (const ushort(*)[4]) srcRowA;
243 const ushort(*rowB)[4] = (const ushort(*)[4]) srcRowB;
244 ushort(*dst)[4] = (ushort(*)[4]) dstRow;
245 for (i = j = 0, k = k0; i < (uint) dstWidth;
246 i++, j += colStride, k += colStride) {
247 dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
248 dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
249 dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4;
250 dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4;
251 }
252 }
253 else if (datatype == DTYPE_USHORT && comps == 3) {
254 uint i, j, k;
255 const ushort(*rowA)[3] = (const ushort(*)[3]) srcRowA;
256 const ushort(*rowB)[3] = (const ushort(*)[3]) srcRowB;
257 ushort(*dst)[3] = (ushort(*)[3]) dstRow;
258 for (i = j = 0, k = k0; i < (uint) dstWidth;
259 i++, j += colStride, k += colStride) {
260 dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
261 dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
262 dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4;
263 }
264 }
265 else if (datatype == DTYPE_USHORT && comps == 2) {
266 uint i, j, k;
267 const ushort(*rowA)[2] = (const ushort(*)[2]) srcRowA;
268 const ushort(*rowB)[2] = (const ushort(*)[2]) srcRowB;
269 ushort(*dst)[2] = (ushort(*)[2]) dstRow;
270 for (i = j = 0, k = k0; i < (uint) dstWidth;
271 i++, j += colStride, k += colStride) {
272 dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
273 dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
274 }
275 }
276 else if (datatype == DTYPE_USHORT && comps == 1) {
277 uint i, j, k;
278 const ushort *rowA = (const ushort *) srcRowA;
279 const ushort *rowB = (const ushort *) srcRowB;
280 ushort *dst = (ushort *) dstRow;
281 for (i = j = 0, k = k0; i < (uint) dstWidth;
282 i++, j += colStride, k += colStride) {
283 dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4;
284 }
285 }
286
287 else if (datatype == DTYPE_FLOAT && comps == 4) {
288 uint i, j, k;
289 const float(*rowA)[4] = (const float(*)[4]) srcRowA;
290 const float(*rowB)[4] = (const float(*)[4]) srcRowB;
291 float(*dst)[4] = (float(*)[4]) dstRow;
292 for (i = j = 0, k = k0; i < (uint) dstWidth;
293 i++, j += colStride, k += colStride) {
294 dst[i][0] = (rowA[j][0] + rowA[k][0] +
295 rowB[j][0] + rowB[k][0]) * 0.25F;
296 dst[i][1] = (rowA[j][1] + rowA[k][1] +
297 rowB[j][1] + rowB[k][1]) * 0.25F;
298 dst[i][2] = (rowA[j][2] + rowA[k][2] +
299 rowB[j][2] + rowB[k][2]) * 0.25F;
300 dst[i][3] = (rowA[j][3] + rowA[k][3] +
301 rowB[j][3] + rowB[k][3]) * 0.25F;
302 }
303 }
304 else if (datatype == DTYPE_FLOAT && comps == 3) {
305 uint i, j, k;
306 const float(*rowA)[3] = (const float(*)[3]) srcRowA;
307 const float(*rowB)[3] = (const float(*)[3]) srcRowB;
308 float(*dst)[3] = (float(*)[3]) dstRow;
309 for (i = j = 0, k = k0; i < (uint) dstWidth;
310 i++, j += colStride, k += colStride) {
311 dst[i][0] = (rowA[j][0] + rowA[k][0] +
312 rowB[j][0] + rowB[k][0]) * 0.25F;
313 dst[i][1] = (rowA[j][1] + rowA[k][1] +
314 rowB[j][1] + rowB[k][1]) * 0.25F;
315 dst[i][2] = (rowA[j][2] + rowA[k][2] +
316 rowB[j][2] + rowB[k][2]) * 0.25F;
317 }
318 }
319 else if (datatype == DTYPE_FLOAT && comps == 2) {
320 uint i, j, k;
321 const float(*rowA)[2] = (const float(*)[2]) srcRowA;
322 const float(*rowB)[2] = (const float(*)[2]) srcRowB;
323 float(*dst)[2] = (float(*)[2]) dstRow;
324 for (i = j = 0, k = k0; i < (uint) dstWidth;
325 i++, j += colStride, k += colStride) {
326 dst[i][0] = (rowA[j][0] + rowA[k][0] +
327 rowB[j][0] + rowB[k][0]) * 0.25F;
328 dst[i][1] = (rowA[j][1] + rowA[k][1] +
329 rowB[j][1] + rowB[k][1]) * 0.25F;
330 }
331 }
332 else if (datatype == DTYPE_FLOAT && comps == 1) {
333 uint i, j, k;
334 const float *rowA = (const float *) srcRowA;
335 const float *rowB = (const float *) srcRowB;
336 float *dst = (float *) dstRow;
337 for (i = j = 0, k = k0; i < (uint) dstWidth;
338 i++, j += colStride, k += colStride) {
339 dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) * 0.25F;
340 }
341 }
342
343 #if 0
344 else if (datatype == HALF_DTYPE_FLOAT && comps == 4) {
345 uint i, j, k, comp;
346 const half_float(*rowA)[4] = (const half_float(*)[4]) srcRowA;
347 const half_float(*rowB)[4] = (const half_float(*)[4]) srcRowB;
348 half_float(*dst)[4] = (half_float(*)[4]) dstRow;
349 for (i = j = 0, k = k0; i < (uint) dstWidth;
350 i++, j += colStride, k += colStride) {
351 for (comp = 0; comp < 4; comp++) {
352 float aj, ak, bj, bk;
353 aj = half_to_float(rowA[j][comp]);
354 ak = half_to_float(rowA[k][comp]);
355 bj = half_to_float(rowB[j][comp]);
356 bk = half_to_float(rowB[k][comp]);
357 dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F);
358 }
359 }
360 }
361 else if (datatype == DTYPE_HALF_FLOAT && comps == 3) {
362 uint i, j, k, comp;
363 const half_float(*rowA)[3] = (const half_float(*)[3]) srcRowA;
364 const half_float(*rowB)[3] = (const half_float(*)[3]) srcRowB;
365 half_float(*dst)[3] = (half_float(*)[3]) dstRow;
366 for (i = j = 0, k = k0; i < (uint) dstWidth;
367 i++, j += colStride, k += colStride) {
368 for (comp = 0; comp < 3; comp++) {
369 float aj, ak, bj, bk;
370 aj = half_to_float(rowA[j][comp]);
371 ak = half_to_float(rowA[k][comp]);
372 bj = half_to_float(rowB[j][comp]);
373 bk = half_to_float(rowB[k][comp]);
374 dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F);
375 }
376 }
377 }
378 else if (datatype == DTYPE_HALF_FLOAT && comps == 2) {
379 uint i, j, k, comp;
380 const half_float(*rowA)[2] = (const half_float(*)[2]) srcRowA;
381 const half_float(*rowB)[2] = (const half_float(*)[2]) srcRowB;
382 half_float(*dst)[2] = (half_float(*)[2]) dstRow;
383 for (i = j = 0, k = k0; i < (uint) dstWidth;
384 i++, j += colStride, k += colStride) {
385 for (comp = 0; comp < 2; comp++) {
386 float aj, ak, bj, bk;
387 aj = half_to_float(rowA[j][comp]);
388 ak = half_to_float(rowA[k][comp]);
389 bj = half_to_float(rowB[j][comp]);
390 bk = half_to_float(rowB[k][comp]);
391 dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F);
392 }
393 }
394 }
395 else if (datatype == DTYPE_HALF_FLOAT && comps == 1) {
396 uint i, j, k;
397 const half_float *rowA = (const half_float *) srcRowA;
398 const half_float *rowB = (const half_float *) srcRowB;
399 half_float *dst = (half_float *) dstRow;
400 for (i = j = 0, k = k0; i < (uint) dstWidth;
401 i++, j += colStride, k += colStride) {
402 float aj, ak, bj, bk;
403 aj = half_to_float(rowA[j]);
404 ak = half_to_float(rowA[k]);
405 bj = half_to_float(rowB[j]);
406 bk = half_to_float(rowB[k]);
407 dst[i] = float_to_half((aj + ak + bj + bk) * 0.25F);
408 }
409 }
410 #endif
411
412 else if (datatype == DTYPE_UINT && comps == 1) {
413 uint i, j, k;
414 const uint *rowA = (const uint *) srcRowA;
415 const uint *rowB = (const uint *) srcRowB;
416 uint *dst = (uint *) dstRow;
417 for (i = j = 0, k = k0; i < (uint) dstWidth;
418 i++, j += colStride, k += colStride) {
419 dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4;
420 }
421 }
422
423 else if (datatype == DTYPE_USHORT_5_6_5 && comps == 3) {
424 uint i, j, k;
425 const ushort *rowA = (const ushort *) srcRowA;
426 const ushort *rowB = (const ushort *) srcRowB;
427 ushort *dst = (ushort *) dstRow;
428 for (i = j = 0, k = k0; i < (uint) dstWidth;
429 i++, j += colStride, k += colStride) {
430 const int rowAr0 = rowA[j] & 0x1f;
431 const int rowAr1 = rowA[k] & 0x1f;
432 const int rowBr0 = rowB[j] & 0x1f;
433 const int rowBr1 = rowB[k] & 0x1f;
434 const int rowAg0 = (rowA[j] >> 5) & 0x3f;
435 const int rowAg1 = (rowA[k] >> 5) & 0x3f;
436 const int rowBg0 = (rowB[j] >> 5) & 0x3f;
437 const int rowBg1 = (rowB[k] >> 5) & 0x3f;
438 const int rowAb0 = (rowA[j] >> 11) & 0x1f;
439 const int rowAb1 = (rowA[k] >> 11) & 0x1f;
440 const int rowBb0 = (rowB[j] >> 11) & 0x1f;
441 const int rowBb1 = (rowB[k] >> 11) & 0x1f;
442 const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2;
443 const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2;
444 const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2;
445 dst[i] = (blue << 11) | (green << 5) | red;
446 }
447 }
448 else if (datatype == DTYPE_USHORT_4_4_4_4 && comps == 4) {
449 uint i, j, k;
450 const ushort *rowA = (const ushort *) srcRowA;
451 const ushort *rowB = (const ushort *) srcRowB;
452 ushort *dst = (ushort *) dstRow;
453 for (i = j = 0, k = k0; i < (uint) dstWidth;
454 i++, j += colStride, k += colStride) {
455 const int rowAr0 = rowA[j] & 0xf;
456 const int rowAr1 = rowA[k] & 0xf;
457 const int rowBr0 = rowB[j] & 0xf;
458 const int rowBr1 = rowB[k] & 0xf;
459 const int rowAg0 = (rowA[j] >> 4) & 0xf;
460 const int rowAg1 = (rowA[k] >> 4) & 0xf;
461 const int rowBg0 = (rowB[j] >> 4) & 0xf;
462 const int rowBg1 = (rowB[k] >> 4) & 0xf;
463 const int rowAb0 = (rowA[j] >> 8) & 0xf;
464 const int rowAb1 = (rowA[k] >> 8) & 0xf;
465 const int rowBb0 = (rowB[j] >> 8) & 0xf;
466 const int rowBb1 = (rowB[k] >> 8) & 0xf;
467 const int rowAa0 = (rowA[j] >> 12) & 0xf;
468 const int rowAa1 = (rowA[k] >> 12) & 0xf;
469 const int rowBa0 = (rowB[j] >> 12) & 0xf;
470 const int rowBa1 = (rowB[k] >> 12) & 0xf;
471 const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2;
472 const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2;
473 const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2;
474 const int alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2;
475 dst[i] = (alpha << 12) | (blue << 8) | (green << 4) | red;
476 }
477 }
478 else if (datatype == DTYPE_USHORT_1_5_5_5_REV && comps == 4) {
479 uint i, j, k;
480 const ushort *rowA = (const ushort *) srcRowA;
481 const ushort *rowB = (const ushort *) srcRowB;
482 ushort *dst = (ushort *) dstRow;
483 for (i = j = 0, k = k0; i < (uint) dstWidth;
484 i++, j += colStride, k += colStride) {
485 const int rowAr0 = rowA[j] & 0x1f;
486 const int rowAr1 = rowA[k] & 0x1f;
487 const int rowBr0 = rowB[j] & 0x1f;
488 const int rowBr1 = rowB[k] & 0x1f;
489 const int rowAg0 = (rowA[j] >> 5) & 0x1f;
490 const int rowAg1 = (rowA[k] >> 5) & 0x1f;
491 const int rowBg0 = (rowB[j] >> 5) & 0x1f;
492 const int rowBg1 = (rowB[k] >> 5) & 0x1f;
493 const int rowAb0 = (rowA[j] >> 10) & 0x1f;
494 const int rowAb1 = (rowA[k] >> 10) & 0x1f;
495 const int rowBb0 = (rowB[j] >> 10) & 0x1f;
496 const int rowBb1 = (rowB[k] >> 10) & 0x1f;
497 const int rowAa0 = (rowA[j] >> 15) & 0x1;
498 const int rowAa1 = (rowA[k] >> 15) & 0x1;
499 const int rowBa0 = (rowB[j] >> 15) & 0x1;
500 const int rowBa1 = (rowB[k] >> 15) & 0x1;
501 const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2;
502 const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2;
503 const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2;
504 const int alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2;
505 dst[i] = (alpha << 15) | (blue << 10) | (green << 5) | red;
506 }
507 }
508 else if (datatype == DTYPE_UBYTE_3_3_2 && comps == 3) {
509 uint i, j, k;
510 const ubyte *rowA = (const ubyte *) srcRowA;
511 const ubyte *rowB = (const ubyte *) srcRowB;
512 ubyte *dst = (ubyte *) dstRow;
513 for (i = j = 0, k = k0; i < (uint) dstWidth;
514 i++, j += colStride, k += colStride) {
515 const int rowAr0 = rowA[j] & 0x3;
516 const int rowAr1 = rowA[k] & 0x3;
517 const int rowBr0 = rowB[j] & 0x3;
518 const int rowBr1 = rowB[k] & 0x3;
519 const int rowAg0 = (rowA[j] >> 2) & 0x7;
520 const int rowAg1 = (rowA[k] >> 2) & 0x7;
521 const int rowBg0 = (rowB[j] >> 2) & 0x7;
522 const int rowBg1 = (rowB[k] >> 2) & 0x7;
523 const int rowAb0 = (rowA[j] >> 5) & 0x7;
524 const int rowAb1 = (rowA[k] >> 5) & 0x7;
525 const int rowBb0 = (rowB[j] >> 5) & 0x7;
526 const int rowBb1 = (rowB[k] >> 5) & 0x7;
527 const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2;
528 const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2;
529 const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2;
530 dst[i] = (blue << 5) | (green << 2) | red;
531 }
532 }
533 else {
534 debug_printf("bad format in do_row()");
535 }
536 }
537
538
539 /**
540 * Average together four rows of a source image to produce a single new
541 * row in the dest image. It's legal for the two source rows to point
542 * to the same data. The source width must be equal to either the
543 * dest width or two times the dest width.
544 *
545 * \param datatype GL pixel type \c GL_UNSIGNED_BYTE, \c GL_UNSIGNED_SHORT,
546 * \c GL_FLOAT, etc.
547 * \param comps number of components per pixel (1..4)
548 * \param srcWidth Width of a row in the source data
549 * \param srcRowA Pointer to one of the rows of source data
550 * \param srcRowB Pointer to one of the rows of source data
551 * \param srcRowC Pointer to one of the rows of source data
552 * \param srcRowD Pointer to one of the rows of source data
553 * \param dstWidth Width of a row in the destination data
554 * \param srcRowA Pointer to the row of destination data
555 */
556 static void
557 do_row_3D(enum dtype datatype, uint comps, int srcWidth,
558 const void *srcRowA, const void *srcRowB,
559 const void *srcRowC, const void *srcRowD,
560 int dstWidth, void *dstRow)
561 {
562 const uint k0 = (srcWidth == dstWidth) ? 0 : 1;
563 const uint colStride = (srcWidth == dstWidth) ? 1 : 2;
564 uint i, j, k;
565
566 assert(comps >= 1);
567 assert(comps <= 4);
568
569 if ((datatype == DTYPE_UBYTE) && (comps == 4)) {
570 DECLARE_ROW_POINTERS(ubyte, 4);
571
572 for (i = j = 0, k = k0; i < (uint) dstWidth;
573 i++, j += colStride, k += colStride) {
574 FILTER_3D(0);
575 FILTER_3D(1);
576 FILTER_3D(2);
577 FILTER_3D(3);
578 }
579 }
580 else if ((datatype == DTYPE_UBYTE) && (comps == 3)) {
581 DECLARE_ROW_POINTERS(ubyte, 3);
582
583 for (i = j = 0, k = k0; i < (uint) dstWidth;
584 i++, j += colStride, k += colStride) {
585 FILTER_3D(0);
586 FILTER_3D(1);
587 FILTER_3D(2);
588 }
589 }
590 else if ((datatype == DTYPE_UBYTE) && (comps == 2)) {
591 DECLARE_ROW_POINTERS(ubyte, 2);
592
593 for (i = j = 0, k = k0; i < (uint) dstWidth;
594 i++, j += colStride, k += colStride) {
595 FILTER_3D(0);
596 FILTER_3D(1);
597 }
598 }
599 else if ((datatype == DTYPE_UBYTE) && (comps == 1)) {
600 DECLARE_ROW_POINTERS(ubyte, 1);
601
602 for (i = j = 0, k = k0; i < (uint) dstWidth;
603 i++, j += colStride, k += colStride) {
604 FILTER_3D(0);
605 }
606 }
607 else if ((datatype == DTYPE_USHORT) && (comps == 4)) {
608 DECLARE_ROW_POINTERS(ushort, 4);
609
610 for (i = j = 0, k = k0; i < (uint) dstWidth;
611 i++, j += colStride, k += colStride) {
612 FILTER_3D(0);
613 FILTER_3D(1);
614 FILTER_3D(2);
615 FILTER_3D(3);
616 }
617 }
618 else if ((datatype == DTYPE_USHORT) && (comps == 3)) {
619 DECLARE_ROW_POINTERS(ushort, 3);
620
621 for (i = j = 0, k = k0; i < (uint) dstWidth;
622 i++, j += colStride, k += colStride) {
623 FILTER_3D(0);
624 FILTER_3D(1);
625 FILTER_3D(2);
626 }
627 }
628 else if ((datatype == DTYPE_USHORT) && (comps == 2)) {
629 DECLARE_ROW_POINTERS(ushort, 2);
630
631 for (i = j = 0, k = k0; i < (uint) dstWidth;
632 i++, j += colStride, k += colStride) {
633 FILTER_3D(0);
634 FILTER_3D(1);
635 }
636 }
637 else if ((datatype == DTYPE_USHORT) && (comps == 1)) {
638 DECLARE_ROW_POINTERS(ushort, 1);
639
640 for (i = j = 0, k = k0; i < (uint) dstWidth;
641 i++, j += colStride, k += colStride) {
642 FILTER_3D(0);
643 }
644 }
645 else if ((datatype == DTYPE_FLOAT) && (comps == 4)) {
646 DECLARE_ROW_POINTERS(float, 4);
647
648 for (i = j = 0, k = k0; i < (uint) dstWidth;
649 i++, j += colStride, k += colStride) {
650 FILTER_F_3D(0);
651 FILTER_F_3D(1);
652 FILTER_F_3D(2);
653 FILTER_F_3D(3);
654 }
655 }
656 else if ((datatype == DTYPE_FLOAT) && (comps == 3)) {
657 DECLARE_ROW_POINTERS(float, 3);
658
659 for (i = j = 0, k = k0; i < (uint) dstWidth;
660 i++, j += colStride, k += colStride) {
661 FILTER_F_3D(0);
662 FILTER_F_3D(1);
663 FILTER_F_3D(2);
664 }
665 }
666 else if ((datatype == DTYPE_FLOAT) && (comps == 2)) {
667 DECLARE_ROW_POINTERS(float, 2);
668
669 for (i = j = 0, k = k0; i < (uint) dstWidth;
670 i++, j += colStride, k += colStride) {
671 FILTER_F_3D(0);
672 FILTER_F_3D(1);
673 }
674 }
675 else if ((datatype == DTYPE_FLOAT) && (comps == 1)) {
676 DECLARE_ROW_POINTERS(float, 1);
677
678 for (i = j = 0, k = k0; i < (uint) dstWidth;
679 i++, j += colStride, k += colStride) {
680 FILTER_F_3D(0);
681 }
682 }
683 else if ((datatype == DTYPE_HALF_FLOAT) && (comps == 4)) {
684 DECLARE_ROW_POINTERS(half_float, 4);
685
686 for (i = j = 0, k = k0; i < (uint) dstWidth;
687 i++, j += colStride, k += colStride) {
688 FILTER_HF_3D(0);
689 FILTER_HF_3D(1);
690 FILTER_HF_3D(2);
691 FILTER_HF_3D(3);
692 }
693 }
694 else if ((datatype == DTYPE_HALF_FLOAT) && (comps == 3)) {
695 DECLARE_ROW_POINTERS(half_float, 4);
696
697 for (i = j = 0, k = k0; i < (uint) dstWidth;
698 i++, j += colStride, k += colStride) {
699 FILTER_HF_3D(0);
700 FILTER_HF_3D(1);
701 FILTER_HF_3D(2);
702 }
703 }
704 else if ((datatype == DTYPE_HALF_FLOAT) && (comps == 2)) {
705 DECLARE_ROW_POINTERS(half_float, 4);
706
707 for (i = j = 0, k = k0; i < (uint) dstWidth;
708 i++, j += colStride, k += colStride) {
709 FILTER_HF_3D(0);
710 FILTER_HF_3D(1);
711 }
712 }
713 else if ((datatype == DTYPE_HALF_FLOAT) && (comps == 1)) {
714 DECLARE_ROW_POINTERS(half_float, 4);
715
716 for (i = j = 0, k = k0; i < (uint) dstWidth;
717 i++, j += colStride, k += colStride) {
718 FILTER_HF_3D(0);
719 }
720 }
721 else if ((datatype == DTYPE_UINT) && (comps == 1)) {
722 const uint *rowA = (const uint *) srcRowA;
723 const uint *rowB = (const uint *) srcRowB;
724 const uint *rowC = (const uint *) srcRowC;
725 const uint *rowD = (const uint *) srcRowD;
726 float *dst = (float *) dstRow;
727
728 for (i = j = 0, k = k0; i < (uint) dstWidth;
729 i++, j += colStride, k += colStride) {
730 const uint64_t tmp = (((uint64_t) rowA[j] + (uint64_t) rowA[k])
731 + ((uint64_t) rowB[j] + (uint64_t) rowB[k])
732 + ((uint64_t) rowC[j] + (uint64_t) rowC[k])
733 + ((uint64_t) rowD[j] + (uint64_t) rowD[k]));
734 dst[i] = (float)((double) tmp * 0.125);
735 }
736 }
737 else if ((datatype == DTYPE_USHORT_5_6_5) && (comps == 3)) {
738 DECLARE_ROW_POINTERS0(ushort);
739
740 for (i = j = 0, k = k0; i < (uint) dstWidth;
741 i++, j += colStride, k += colStride) {
742 const int rowAr0 = rowA[j] & 0x1f;
743 const int rowAr1 = rowA[k] & 0x1f;
744 const int rowBr0 = rowB[j] & 0x1f;
745 const int rowBr1 = rowB[k] & 0x1f;
746 const int rowCr0 = rowC[j] & 0x1f;
747 const int rowCr1 = rowC[k] & 0x1f;
748 const int rowDr0 = rowD[j] & 0x1f;
749 const int rowDr1 = rowD[k] & 0x1f;
750 const int rowAg0 = (rowA[j] >> 5) & 0x3f;
751 const int rowAg1 = (rowA[k] >> 5) & 0x3f;
752 const int rowBg0 = (rowB[j] >> 5) & 0x3f;
753 const int rowBg1 = (rowB[k] >> 5) & 0x3f;
754 const int rowCg0 = (rowC[j] >> 5) & 0x3f;
755 const int rowCg1 = (rowC[k] >> 5) & 0x3f;
756 const int rowDg0 = (rowD[j] >> 5) & 0x3f;
757 const int rowDg1 = (rowD[k] >> 5) & 0x3f;
758 const int rowAb0 = (rowA[j] >> 11) & 0x1f;
759 const int rowAb1 = (rowA[k] >> 11) & 0x1f;
760 const int rowBb0 = (rowB[j] >> 11) & 0x1f;
761 const int rowBb1 = (rowB[k] >> 11) & 0x1f;
762 const int rowCb0 = (rowC[j] >> 11) & 0x1f;
763 const int rowCb1 = (rowC[k] >> 11) & 0x1f;
764 const int rowDb0 = (rowD[j] >> 11) & 0x1f;
765 const int rowDb1 = (rowD[k] >> 11) & 0x1f;
766 const int r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
767 rowCr0, rowCr1, rowDr0, rowDr1);
768 const int g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
769 rowCg0, rowCg1, rowDg0, rowDg1);
770 const int b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
771 rowCb0, rowCb1, rowDb0, rowDb1);
772 dst[i] = (b << 11) | (g << 5) | r;
773 }
774 }
775 else if ((datatype == DTYPE_USHORT_4_4_4_4) && (comps == 4)) {
776 DECLARE_ROW_POINTERS0(ushort);
777
778 for (i = j = 0, k = k0; i < (uint) dstWidth;
779 i++, j += colStride, k += colStride) {
780 const int rowAr0 = rowA[j] & 0xf;
781 const int rowAr1 = rowA[k] & 0xf;
782 const int rowBr0 = rowB[j] & 0xf;
783 const int rowBr1 = rowB[k] & 0xf;
784 const int rowCr0 = rowC[j] & 0xf;
785 const int rowCr1 = rowC[k] & 0xf;
786 const int rowDr0 = rowD[j] & 0xf;
787 const int rowDr1 = rowD[k] & 0xf;
788 const int rowAg0 = (rowA[j] >> 4) & 0xf;
789 const int rowAg1 = (rowA[k] >> 4) & 0xf;
790 const int rowBg0 = (rowB[j] >> 4) & 0xf;
791 const int rowBg1 = (rowB[k] >> 4) & 0xf;
792 const int rowCg0 = (rowC[j] >> 4) & 0xf;
793 const int rowCg1 = (rowC[k] >> 4) & 0xf;
794 const int rowDg0 = (rowD[j] >> 4) & 0xf;
795 const int rowDg1 = (rowD[k] >> 4) & 0xf;
796 const int rowAb0 = (rowA[j] >> 8) & 0xf;
797 const int rowAb1 = (rowA[k] >> 8) & 0xf;
798 const int rowBb0 = (rowB[j] >> 8) & 0xf;
799 const int rowBb1 = (rowB[k] >> 8) & 0xf;
800 const int rowCb0 = (rowC[j] >> 8) & 0xf;
801 const int rowCb1 = (rowC[k] >> 8) & 0xf;
802 const int rowDb0 = (rowD[j] >> 8) & 0xf;
803 const int rowDb1 = (rowD[k] >> 8) & 0xf;
804 const int rowAa0 = (rowA[j] >> 12) & 0xf;
805 const int rowAa1 = (rowA[k] >> 12) & 0xf;
806 const int rowBa0 = (rowB[j] >> 12) & 0xf;
807 const int rowBa1 = (rowB[k] >> 12) & 0xf;
808 const int rowCa0 = (rowC[j] >> 12) & 0xf;
809 const int rowCa1 = (rowC[k] >> 12) & 0xf;
810 const int rowDa0 = (rowD[j] >> 12) & 0xf;
811 const int rowDa1 = (rowD[k] >> 12) & 0xf;
812 const int r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
813 rowCr0, rowCr1, rowDr0, rowDr1);
814 const int g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
815 rowCg0, rowCg1, rowDg0, rowDg1);
816 const int b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
817 rowCb0, rowCb1, rowDb0, rowDb1);
818 const int a = FILTER_SUM_3D(rowAa0, rowAa1, rowBa0, rowBa1,
819 rowCa0, rowCa1, rowDa0, rowDa1);
820
821 dst[i] = (a << 12) | (b << 8) | (g << 4) | r;
822 }
823 }
824 else if ((datatype == DTYPE_USHORT_1_5_5_5_REV) && (comps == 4)) {
825 DECLARE_ROW_POINTERS0(ushort);
826
827 for (i = j = 0, k = k0; i < (uint) dstWidth;
828 i++, j += colStride, k += colStride) {
829 const int rowAr0 = rowA[j] & 0x1f;
830 const int rowAr1 = rowA[k] & 0x1f;
831 const int rowBr0 = rowB[j] & 0x1f;
832 const int rowBr1 = rowB[k] & 0x1f;
833 const int rowCr0 = rowC[j] & 0x1f;
834 const int rowCr1 = rowC[k] & 0x1f;
835 const int rowDr0 = rowD[j] & 0x1f;
836 const int rowDr1 = rowD[k] & 0x1f;
837 const int rowAg0 = (rowA[j] >> 5) & 0x1f;
838 const int rowAg1 = (rowA[k] >> 5) & 0x1f;
839 const int rowBg0 = (rowB[j] >> 5) & 0x1f;
840 const int rowBg1 = (rowB[k] >> 5) & 0x1f;
841 const int rowCg0 = (rowC[j] >> 5) & 0x1f;
842 const int rowCg1 = (rowC[k] >> 5) & 0x1f;
843 const int rowDg0 = (rowD[j] >> 5) & 0x1f;
844 const int rowDg1 = (rowD[k] >> 5) & 0x1f;
845 const int rowAb0 = (rowA[j] >> 10) & 0x1f;
846 const int rowAb1 = (rowA[k] >> 10) & 0x1f;
847 const int rowBb0 = (rowB[j] >> 10) & 0x1f;
848 const int rowBb1 = (rowB[k] >> 10) & 0x1f;
849 const int rowCb0 = (rowC[j] >> 10) & 0x1f;
850 const int rowCb1 = (rowC[k] >> 10) & 0x1f;
851 const int rowDb0 = (rowD[j] >> 10) & 0x1f;
852 const int rowDb1 = (rowD[k] >> 10) & 0x1f;
853 const int rowAa0 = (rowA[j] >> 15) & 0x1;
854 const int rowAa1 = (rowA[k] >> 15) & 0x1;
855 const int rowBa0 = (rowB[j] >> 15) & 0x1;
856 const int rowBa1 = (rowB[k] >> 15) & 0x1;
857 const int rowCa0 = (rowC[j] >> 15) & 0x1;
858 const int rowCa1 = (rowC[k] >> 15) & 0x1;
859 const int rowDa0 = (rowD[j] >> 15) & 0x1;
860 const int rowDa1 = (rowD[k] >> 15) & 0x1;
861 const int r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
862 rowCr0, rowCr1, rowDr0, rowDr1);
863 const int g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
864 rowCg0, rowCg1, rowDg0, rowDg1);
865 const int b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
866 rowCb0, rowCb1, rowDb0, rowDb1);
867 const int a = FILTER_SUM_3D(rowAa0, rowAa1, rowBa0, rowBa1,
868 rowCa0, rowCa1, rowDa0, rowDa1);
869
870 dst[i] = (a << 15) | (b << 10) | (g << 5) | r;
871 }
872 }
873 else if ((datatype == DTYPE_UBYTE_3_3_2) && (comps == 3)) {
874 DECLARE_ROW_POINTERS0(ushort);
875
876 for (i = j = 0, k = k0; i < (uint) dstWidth;
877 i++, j += colStride, k += colStride) {
878 const int rowAr0 = rowA[j] & 0x3;
879 const int rowAr1 = rowA[k] & 0x3;
880 const int rowBr0 = rowB[j] & 0x3;
881 const int rowBr1 = rowB[k] & 0x3;
882 const int rowCr0 = rowC[j] & 0x3;
883 const int rowCr1 = rowC[k] & 0x3;
884 const int rowDr0 = rowD[j] & 0x3;
885 const int rowDr1 = rowD[k] & 0x3;
886 const int rowAg0 = (rowA[j] >> 2) & 0x7;
887 const int rowAg1 = (rowA[k] >> 2) & 0x7;
888 const int rowBg0 = (rowB[j] >> 2) & 0x7;
889 const int rowBg1 = (rowB[k] >> 2) & 0x7;
890 const int rowCg0 = (rowC[j] >> 2) & 0x7;
891 const int rowCg1 = (rowC[k] >> 2) & 0x7;
892 const int rowDg0 = (rowD[j] >> 2) & 0x7;
893 const int rowDg1 = (rowD[k] >> 2) & 0x7;
894 const int rowAb0 = (rowA[j] >> 5) & 0x7;
895 const int rowAb1 = (rowA[k] >> 5) & 0x7;
896 const int rowBb0 = (rowB[j] >> 5) & 0x7;
897 const int rowBb1 = (rowB[k] >> 5) & 0x7;
898 const int rowCb0 = (rowC[j] >> 5) & 0x7;
899 const int rowCb1 = (rowC[k] >> 5) & 0x7;
900 const int rowDb0 = (rowD[j] >> 5) & 0x7;
901 const int rowDb1 = (rowD[k] >> 5) & 0x7;
902 const int r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
903 rowCr0, rowCr1, rowDr0, rowDr1);
904 const int g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
905 rowCg0, rowCg1, rowDg0, rowDg1);
906 const int b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
907 rowCb0, rowCb1, rowDb0, rowDb1);
908 dst[i] = (b << 5) | (g << 2) | r;
909 }
910 }
911 else {
912 debug_printf("bad format in do_row_3D()");
913 }
914 }
915
916
917
918 static void
919 format_to_type_comps(enum pipe_format pformat,
920 enum dtype *datatype, uint *comps)
921 {
922 /* XXX I think this could be implemented in terms of the pf_*() functions */
923 switch (pformat) {
924 case PIPE_FORMAT_A8R8G8B8_UNORM:
925 case PIPE_FORMAT_X8R8G8B8_UNORM:
926 case PIPE_FORMAT_B8G8R8A8_UNORM:
927 case PIPE_FORMAT_B8G8R8X8_UNORM:
928 case PIPE_FORMAT_R8G8B8A8_SRGB:
929 case PIPE_FORMAT_R8G8B8X8_SRGB:
930 case PIPE_FORMAT_A8R8G8B8_SRGB:
931 case PIPE_FORMAT_X8R8G8B8_SRGB:
932 case PIPE_FORMAT_B8G8R8A8_SRGB:
933 case PIPE_FORMAT_B8G8R8X8_SRGB:
934 case PIPE_FORMAT_R8G8B8_SRGB:
935 *datatype = DTYPE_UBYTE;
936 *comps = 4;
937 return;
938 case PIPE_FORMAT_A1R5G5B5_UNORM:
939 *datatype = DTYPE_USHORT_1_5_5_5_REV;
940 *comps = 4;
941 return;
942 case PIPE_FORMAT_A4R4G4B4_UNORM:
943 *datatype = DTYPE_USHORT_4_4_4_4;
944 *comps = 4;
945 return;
946 case PIPE_FORMAT_R5G6B5_UNORM:
947 *datatype = DTYPE_USHORT_5_6_5;
948 *comps = 3;
949 return;
950 case PIPE_FORMAT_L8_UNORM:
951 case PIPE_FORMAT_L8_SRGB:
952 case PIPE_FORMAT_A8_UNORM:
953 case PIPE_FORMAT_I8_UNORM:
954 *datatype = DTYPE_UBYTE;
955 *comps = 1;
956 return;
957 case PIPE_FORMAT_A8L8_UNORM:
958 case PIPE_FORMAT_A8L8_SRGB:
959 *datatype = DTYPE_UBYTE;
960 *comps = 2;
961 return;
962 default:
963 assert(0);
964 *datatype = DTYPE_UBYTE;
965 *comps = 0;
966 break;
967 }
968 }
969
970
971 static void
972 reduce_1d(enum pipe_format pformat,
973 int srcWidth, const ubyte *srcPtr,
974 int dstWidth, ubyte *dstPtr)
975 {
976 enum dtype datatype;
977 uint comps;
978
979 format_to_type_comps(pformat, &datatype, &comps);
980
981 /* we just duplicate the input row, kind of hack, saves code */
982 do_row(datatype, comps,
983 srcWidth, srcPtr, srcPtr,
984 dstWidth, dstPtr);
985 }
986
987
988 /**
989 * Strides are in bytes. If zero, it'll be computed as width * bpp.
990 */
991 static void
992 reduce_2d(enum pipe_format pformat,
993 int srcWidth, int srcHeight,
994 int srcRowStride, const ubyte *srcPtr,
995 int dstWidth, int dstHeight,
996 int dstRowStride, ubyte *dstPtr)
997 {
998 enum dtype datatype;
999 uint comps;
1000 const int bpt = pf_get_blocksize(pformat);
1001 const ubyte *srcA, *srcB;
1002 ubyte *dst;
1003 int row;
1004
1005 format_to_type_comps(pformat, &datatype, &comps);
1006
1007 if (!srcRowStride)
1008 srcRowStride = bpt * srcWidth;
1009
1010 if (!dstRowStride)
1011 dstRowStride = bpt * dstWidth;
1012
1013 /* Compute src and dst pointers */
1014 srcA = srcPtr;
1015 if (srcHeight > 1)
1016 srcB = srcA + srcRowStride;
1017 else
1018 srcB = srcA;
1019 dst = dstPtr;
1020
1021 for (row = 0; row < dstHeight; row++) {
1022 do_row(datatype, comps,
1023 srcWidth, srcA, srcB,
1024 dstWidth, dst);
1025 srcA += 2 * srcRowStride;
1026 srcB += 2 * srcRowStride;
1027 dst += dstRowStride;
1028 }
1029 }
1030
1031
1032 static void
1033 reduce_3d(enum pipe_format pformat,
1034 int srcWidth, int srcHeight, int srcDepth,
1035 int srcRowStride, const ubyte *srcPtr,
1036 int dstWidth, int dstHeight, int dstDepth,
1037 int dstRowStride, ubyte *dstPtr)
1038 {
1039 const int bpt = pf_get_blocksize(pformat);
1040 const int border = 0;
1041 int img, row;
1042 int bytesPerSrcImage, bytesPerDstImage;
1043 int bytesPerSrcRow, bytesPerDstRow;
1044 int srcImageOffset, srcRowOffset;
1045 enum dtype datatype;
1046 uint comps;
1047
1048 format_to_type_comps(pformat, &datatype, &comps);
1049
1050 bytesPerSrcImage = srcWidth * srcHeight * bpt;
1051 bytesPerDstImage = dstWidth * dstHeight * bpt;
1052
1053 bytesPerSrcRow = srcWidth * bpt;
1054 bytesPerDstRow = dstWidth * bpt;
1055
1056 /* Offset between adjacent src images to be averaged together */
1057 srcImageOffset = (srcDepth == dstDepth) ? 0 : bytesPerSrcImage;
1058
1059 /* Offset between adjacent src rows to be averaged together */
1060 srcRowOffset = (srcHeight == dstHeight) ? 0 : srcWidth * bpt;
1061
1062 /*
1063 * Need to average together up to 8 src pixels for each dest pixel.
1064 * Break that down into 3 operations:
1065 * 1. take two rows from source image and average them together.
1066 * 2. take two rows from next source image and average them together.
1067 * 3. take the two averaged rows and average them for the final dst row.
1068 */
1069
1070 /*
1071 _mesa_printf("mip3d %d x %d x %d -> %d x %d x %d\n",
1072 srcWidth, srcHeight, srcDepth, dstWidth, dstHeight, dstDepth);
1073 */
1074
1075 for (img = 0; img < dstDepth; img++) {
1076 /* first source image pointer, skipping border */
1077 const ubyte *imgSrcA = srcPtr
1078 + (bytesPerSrcImage + bytesPerSrcRow + border) * bpt * border
1079 + img * (bytesPerSrcImage + srcImageOffset);
1080 /* second source image pointer, skipping border */
1081 const ubyte *imgSrcB = imgSrcA + srcImageOffset;
1082 /* address of the dest image, skipping border */
1083 ubyte *imgDst = dstPtr
1084 + (bytesPerDstImage + bytesPerDstRow + border) * bpt * border
1085 + img * bytesPerDstImage;
1086
1087 /* setup the four source row pointers and the dest row pointer */
1088 const ubyte *srcImgARowA = imgSrcA;
1089 const ubyte *srcImgARowB = imgSrcA + srcRowOffset;
1090 const ubyte *srcImgBRowA = imgSrcB;
1091 const ubyte *srcImgBRowB = imgSrcB + srcRowOffset;
1092 ubyte *dstImgRow = imgDst;
1093
1094 for (row = 0; row < dstHeight; row++) {
1095 do_row_3D(datatype, comps, srcWidth,
1096 srcImgARowA, srcImgARowB,
1097 srcImgBRowA, srcImgBRowB,
1098 dstWidth, dstImgRow);
1099
1100 /* advance to next rows */
1101 srcImgARowA += bytesPerSrcRow + srcRowOffset;
1102 srcImgARowB += bytesPerSrcRow + srcRowOffset;
1103 srcImgBRowA += bytesPerSrcRow + srcRowOffset;
1104 srcImgBRowB += bytesPerSrcRow + srcRowOffset;
1105 dstImgRow += bytesPerDstRow;
1106 }
1107 }
1108 }
1109
1110
1111
1112
1113 static void
1114 make_1d_mipmap(struct gen_mipmap_state *ctx,
1115 struct pipe_texture *pt,
1116 uint face, uint baseLevel, uint lastLevel)
1117 {
1118 struct pipe_context *pipe = ctx->pipe;
1119 struct pipe_screen *screen = pipe->screen;
1120 const uint zslice = 0;
1121 uint dstLevel;
1122
1123 for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
1124 const uint srcLevel = dstLevel - 1;
1125 struct pipe_transfer *srcTrans, *dstTrans;
1126 void *srcMap, *dstMap;
1127
1128 srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice,
1129 PIPE_TRANSFER_READ, 0, 0,
1130 u_minify(pt->width0, srcLevel),
1131 u_minify(pt->height0, srcLevel));
1132 dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice,
1133 PIPE_TRANSFER_WRITE, 0, 0,
1134 u_minify(pt->width0, dstLevel),
1135 u_minify(pt->height0, dstLevel));
1136
1137 srcMap = (ubyte *) screen->transfer_map(screen, srcTrans);
1138 dstMap = (ubyte *) screen->transfer_map(screen, dstTrans);
1139
1140 reduce_1d(pt->format,
1141 srcTrans->width, srcMap,
1142 dstTrans->width, dstMap);
1143
1144 screen->transfer_unmap(screen, srcTrans);
1145 screen->transfer_unmap(screen, dstTrans);
1146
1147 screen->tex_transfer_destroy(srcTrans);
1148 screen->tex_transfer_destroy(dstTrans);
1149 }
1150 }
1151
1152
1153 static void
1154 make_2d_mipmap(struct gen_mipmap_state *ctx,
1155 struct pipe_texture *pt,
1156 uint face, uint baseLevel, uint lastLevel)
1157 {
1158 struct pipe_context *pipe = ctx->pipe;
1159 struct pipe_screen *screen = pipe->screen;
1160 const uint zslice = 0;
1161 uint dstLevel;
1162
1163 assert(pf_get_blockwidth(pt->format) == 1);
1164 assert(pf_get_blockheight(pt->format) == 1);
1165
1166 for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
1167 const uint srcLevel = dstLevel - 1;
1168 struct pipe_transfer *srcTrans, *dstTrans;
1169 ubyte *srcMap, *dstMap;
1170
1171 srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice,
1172 PIPE_TRANSFER_READ, 0, 0,
1173 u_minify(pt->width0, srcLevel),
1174 u_minify(pt->height0, srcLevel));
1175 dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice,
1176 PIPE_TRANSFER_WRITE, 0, 0,
1177 u_minify(pt->width0, dstLevel),
1178 u_minify(pt->height0, dstLevel));
1179
1180 srcMap = (ubyte *) screen->transfer_map(screen, srcTrans);
1181 dstMap = (ubyte *) screen->transfer_map(screen, dstTrans);
1182
1183 reduce_2d(pt->format,
1184 srcTrans->width, srcTrans->height,
1185 srcTrans->stride, srcMap,
1186 dstTrans->width, dstTrans->height,
1187 dstTrans->stride, dstMap);
1188
1189 screen->transfer_unmap(screen, srcTrans);
1190 screen->transfer_unmap(screen, dstTrans);
1191
1192 screen->tex_transfer_destroy(srcTrans);
1193 screen->tex_transfer_destroy(dstTrans);
1194 }
1195 }
1196
1197
1198 static void
1199 make_3d_mipmap(struct gen_mipmap_state *ctx,
1200 struct pipe_texture *pt,
1201 uint face, uint baseLevel, uint lastLevel)
1202 {
1203 #if 0
1204 struct pipe_context *pipe = ctx->pipe;
1205 struct pipe_screen *screen = pipe->screen;
1206 uint dstLevel, zslice = 0;
1207
1208 assert(pf_get_blockwidth(pt->format) == 1);
1209 assert(pf_get_blockheight(pt->format) == 1);
1210
1211 for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
1212 const uint srcLevel = dstLevel - 1;
1213 struct pipe_transfer *srcTrans, *dstTrans;
1214 ubyte *srcMap, *dstMap;
1215
1216 srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice,
1217 PIPE_TRANSFER_READ, 0, 0,
1218 u_minify(pt->width0, srcLevel),
1219 u_minify(pt->height0, srcLevel));
1220 dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice,
1221 PIPE_TRANSFER_WRITE, 0, 0,
1222 u_minify(pt->width0, dstLevel),
1223 u_minify(pt->height0, dstLevel));
1224
1225 srcMap = (ubyte *) screen->transfer_map(screen, srcTrans);
1226 dstMap = (ubyte *) screen->transfer_map(screen, dstTrans);
1227
1228 reduce_3d(pt->format,
1229 srcTrans->width, srcTrans->height,
1230 srcTrans->stride, srcMap,
1231 dstTrans->width, dstTrans->height,
1232 dstTrans->stride, dstMap);
1233
1234 screen->transfer_unmap(screen, srcTrans);
1235 screen->transfer_unmap(screen, dstTrans);
1236
1237 screen->tex_transfer_destroy(srcTrans);
1238 screen->tex_transfer_destroy(dstTrans);
1239 }
1240 #else
1241 (void) reduce_3d;
1242 #endif
1243 }
1244
1245
1246 static void
1247 fallback_gen_mipmap(struct gen_mipmap_state *ctx,
1248 struct pipe_texture *pt,
1249 uint face, uint baseLevel, uint lastLevel)
1250 {
1251 switch (pt->target) {
1252 case PIPE_TEXTURE_1D:
1253 make_1d_mipmap(ctx, pt, face, baseLevel, lastLevel);
1254 break;
1255 case PIPE_TEXTURE_2D:
1256 case PIPE_TEXTURE_CUBE:
1257 make_2d_mipmap(ctx, pt, face, baseLevel, lastLevel);
1258 break;
1259 case PIPE_TEXTURE_3D:
1260 make_3d_mipmap(ctx, pt, face, baseLevel, lastLevel);
1261 break;
1262 default:
1263 assert(0);
1264 }
1265 }
1266
1267
1268 /**
1269 * Create a mipmap generation context.
1270 * The idea is to create one of these and re-use it each time we need to
1271 * generate a mipmap.
1272 */
1273 struct gen_mipmap_state *
1274 util_create_gen_mipmap(struct pipe_context *pipe,
1275 struct cso_context *cso)
1276 {
1277 struct gen_mipmap_state *ctx;
1278 uint i;
1279
1280 ctx = CALLOC_STRUCT(gen_mipmap_state);
1281 if (!ctx)
1282 return NULL;
1283
1284 ctx->pipe = pipe;
1285 ctx->cso = cso;
1286
1287 /* disabled blending/masking */
1288 memset(&ctx->blend, 0, sizeof(ctx->blend));
1289 ctx->blend.colormask = PIPE_MASK_RGBA;
1290
1291 /* no-op depth/stencil/alpha */
1292 memset(&ctx->depthstencil, 0, sizeof(ctx->depthstencil));
1293
1294 /* rasterizer */
1295 memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer));
1296 ctx->rasterizer.front_winding = PIPE_WINDING_CW;
1297 ctx->rasterizer.cull_mode = PIPE_WINDING_NONE;
1298 ctx->rasterizer.bypass_vs_clip_and_viewport = 1;
1299 ctx->rasterizer.gl_rasterization_rules = 1;
1300
1301 /* sampler state */
1302 memset(&ctx->sampler, 0, sizeof(ctx->sampler));
1303 ctx->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
1304 ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
1305 ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
1306 ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST;
1307 ctx->sampler.normalized_coords = 1;
1308
1309 /* vertex shader - still needed to specify mapping from fragment
1310 * shader input semantics to vertex elements
1311 */
1312 {
1313 const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
1314 TGSI_SEMANTIC_GENERIC };
1315 const uint semantic_indexes[] = { 0, 0 };
1316 ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names,
1317 semantic_indexes);
1318 }
1319
1320 /* fragment shader */
1321 ctx->fs = util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D);
1322
1323 /* vertex data that doesn't change */
1324 for (i = 0; i < 4; i++) {
1325 ctx->vertices[i][0][2] = 0.0f; /* z */
1326 ctx->vertices[i][0][3] = 1.0f; /* w */
1327 ctx->vertices[i][1][3] = 1.0f; /* q */
1328 }
1329
1330 /* Note: the actual vertex buffer is allocated as needed below */
1331
1332 return ctx;
1333 }
1334
1335
1336 /**
1337 * Get next "slot" of vertex space in the vertex buffer.
1338 * We're allocating one large vertex buffer and using it piece by piece.
1339 */
1340 static unsigned
1341 get_next_slot(struct gen_mipmap_state *ctx)
1342 {
1343 const unsigned max_slots = 4096 / sizeof ctx->vertices;
1344
1345 if (ctx->vbuf_slot >= max_slots)
1346 util_gen_mipmap_flush( ctx );
1347
1348 if (!ctx->vbuf) {
1349 ctx->vbuf = pipe_buffer_create(ctx->pipe->screen,
1350 32,
1351 PIPE_BUFFER_USAGE_VERTEX,
1352 max_slots * sizeof ctx->vertices);
1353 }
1354
1355 return ctx->vbuf_slot++ * sizeof ctx->vertices;
1356 }
1357
1358
1359 static unsigned
1360 set_vertex_data(struct gen_mipmap_state *ctx,
1361 enum pipe_texture_target tex_target,
1362 uint face, float width, float height)
1363 {
1364 unsigned offset;
1365
1366 /* vert[0].position */
1367 ctx->vertices[0][0][0] = 0.0f; /*x*/
1368 ctx->vertices[0][0][1] = 0.0f; /*y*/
1369
1370 /* vert[1].position */
1371 ctx->vertices[1][0][0] = width;
1372 ctx->vertices[1][0][1] = 0.0f;
1373
1374 /* vert[2].position */
1375 ctx->vertices[2][0][0] = width;
1376 ctx->vertices[2][0][1] = height;
1377
1378 /* vert[3].position */
1379 ctx->vertices[3][0][0] = 0.0f;
1380 ctx->vertices[3][0][1] = height;
1381
1382 /* Setup vertex texcoords. This is a little tricky for cube maps. */
1383 if (tex_target == PIPE_TEXTURE_CUBE) {
1384 static const float st[4][2] = {
1385 {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f}
1386 };
1387
1388 util_map_texcoords2d_onto_cubemap(face, &st[0][0], 2,
1389 &ctx->vertices[0][1][0], 8);
1390 }
1391 else {
1392 /* 1D/2D */
1393 ctx->vertices[0][1][0] = 0.0f; /*s*/
1394 ctx->vertices[0][1][1] = 0.0f; /*t*/
1395 ctx->vertices[0][1][2] = 0.0f; /*r*/
1396
1397 ctx->vertices[1][1][0] = 1.0f;
1398 ctx->vertices[1][1][1] = 0.0f;
1399 ctx->vertices[1][1][2] = 0.0f;
1400
1401 ctx->vertices[2][1][0] = 1.0f;
1402 ctx->vertices[2][1][1] = 1.0f;
1403 ctx->vertices[2][1][2] = 0.0f;
1404
1405 ctx->vertices[3][1][0] = 0.0f;
1406 ctx->vertices[3][1][1] = 1.0f;
1407 ctx->vertices[3][1][2] = 0.0f;
1408 }
1409
1410 offset = get_next_slot( ctx );
1411
1412 pipe_buffer_write(ctx->pipe->screen, ctx->vbuf,
1413 offset, sizeof(ctx->vertices), ctx->vertices);
1414
1415 return offset;
1416 }
1417
1418
1419
1420 /**
1421 * Destroy a mipmap generation context
1422 */
1423 void
1424 util_destroy_gen_mipmap(struct gen_mipmap_state *ctx)
1425 {
1426 struct pipe_context *pipe = ctx->pipe;
1427
1428 pipe->delete_vs_state(pipe, ctx->vs);
1429 pipe->delete_fs_state(pipe, ctx->fs);
1430
1431 pipe_buffer_reference(&ctx->vbuf, NULL);
1432
1433 FREE(ctx);
1434 }
1435
1436
1437
1438 /* Release vertex buffer at end of frame to avoid synchronous
1439 * rendering.
1440 */
1441 void util_gen_mipmap_flush( struct gen_mipmap_state *ctx )
1442 {
1443 pipe_buffer_reference(&ctx->vbuf, NULL);
1444 ctx->vbuf_slot = 0;
1445 }
1446
1447
1448 /**
1449 * Generate mipmap images. It's assumed all needed texture memory is
1450 * already allocated.
1451 *
1452 * \param pt the texture to generate mipmap levels for
1453 * \param face which cube face to generate mipmaps for (0 for non-cube maps)
1454 * \param baseLevel the first mipmap level to use as a src
1455 * \param lastLevel the last mipmap level to generate
1456 * \param filter the minification filter used to generate mipmap levels with
1457 * \param filter one of PIPE_TEX_FILTER_LINEAR, PIPE_TEX_FILTER_NEAREST
1458 */
1459 void
1460 util_gen_mipmap(struct gen_mipmap_state *ctx,
1461 struct pipe_texture *pt,
1462 uint face, uint baseLevel, uint lastLevel, uint filter)
1463 {
1464 struct pipe_context *pipe = ctx->pipe;
1465 struct pipe_screen *screen = pipe->screen;
1466 struct pipe_framebuffer_state fb;
1467 uint dstLevel;
1468 uint zslice = 0;
1469 uint offset;
1470
1471 /* The texture object should have room for the levels which we're
1472 * about to generate.
1473 */
1474 assert(lastLevel <= pt->last_level);
1475
1476 /* If this fails, why are we here? */
1477 assert(lastLevel > baseLevel);
1478
1479 assert(filter == PIPE_TEX_FILTER_LINEAR ||
1480 filter == PIPE_TEX_FILTER_NEAREST);
1481
1482 /* check if we can render in the texture's format */
1483 if (!screen->is_format_supported(screen, pt->format, PIPE_TEXTURE_2D,
1484 PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) {
1485 fallback_gen_mipmap(ctx, pt, face, baseLevel, lastLevel);
1486 return;
1487 }
1488
1489 /* save state (restored below) */
1490 cso_save_blend(ctx->cso);
1491 cso_save_depth_stencil_alpha(ctx->cso);
1492 cso_save_rasterizer(ctx->cso);
1493 cso_save_samplers(ctx->cso);
1494 cso_save_sampler_textures(ctx->cso);
1495 cso_save_framebuffer(ctx->cso);
1496 cso_save_fragment_shader(ctx->cso);
1497 cso_save_vertex_shader(ctx->cso);
1498
1499 /* bind our state */
1500 cso_set_blend(ctx->cso, &ctx->blend);
1501 cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil);
1502 cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
1503
1504 cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
1505 cso_set_vertex_shader_handle(ctx->cso, ctx->vs);
1506
1507 /* init framebuffer state */
1508 memset(&fb, 0, sizeof(fb));
1509 fb.nr_cbufs = 1;
1510
1511 /* set min/mag to same filter for faster sw speed */
1512 ctx->sampler.mag_img_filter = filter;
1513 ctx->sampler.min_img_filter = filter;
1514
1515 /*
1516 * XXX for small mipmap levels, it may be faster to use the software
1517 * fallback path...
1518 */
1519 for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
1520 const uint srcLevel = dstLevel - 1;
1521
1522 struct pipe_surface *surf =
1523 screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
1524 PIPE_BUFFER_USAGE_GPU_WRITE);
1525
1526 /*
1527 * Setup framebuffer / dest surface
1528 */
1529 fb.cbufs[0] = surf;
1530 fb.width = u_minify(pt->width0, dstLevel);
1531 fb.height = u_minify(pt->height0, dstLevel);
1532 cso_set_framebuffer(ctx->cso, &fb);
1533
1534 /*
1535 * Setup sampler state
1536 * Note: we should only have to set the min/max LOD clamps to ensure
1537 * we grab texels from the right mipmap level. But some hardware
1538 * has trouble with min clamping so we also set the lod_bias to
1539 * try to work around that.
1540 */
1541 ctx->sampler.min_lod = ctx->sampler.max_lod = (float) srcLevel;
1542 ctx->sampler.lod_bias = (float) srcLevel;
1543 cso_single_sampler(ctx->cso, 0, &ctx->sampler);
1544 cso_single_sampler_done(ctx->cso);
1545
1546 cso_set_sampler_textures(ctx->cso, 1, &pt);
1547
1548 /* quad coords in window coords (bypassing vs, clip and viewport) */
1549 offset = set_vertex_data(ctx,
1550 pt->target,
1551 face,
1552 (float) u_minify(pt->width0, dstLevel),
1553 (float) u_minify(pt->height0, dstLevel));
1554
1555 util_draw_vertex_buffer(ctx->pipe,
1556 ctx->vbuf,
1557 offset,
1558 PIPE_PRIM_TRIANGLE_FAN,
1559 4, /* verts */
1560 2); /* attribs/vert */
1561
1562 pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
1563
1564 /* need to signal that the texture has changed _after_ rendering to it */
1565 pipe_surface_reference( &surf, NULL );
1566 }
1567
1568 /* restore state we changed */
1569 cso_restore_blend(ctx->cso);
1570 cso_restore_depth_stencil_alpha(ctx->cso);
1571 cso_restore_rasterizer(ctx->cso);
1572 cso_restore_samplers(ctx->cso);
1573 cso_restore_sampler_textures(ctx->cso);
1574 cso_restore_framebuffer(ctx->cso);
1575 cso_restore_fragment_shader(ctx->cso);
1576 cso_restore_vertex_shader(ctx->cso);
1577 }