2 * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
23 #include "r600_formats.h"
26 #include "pipe/p_shader_tokens.h"
27 #include "util/u_pack_color.h"
28 #include "util/u_memory.h"
29 #include "util/u_framebuffer.h"
30 #include "util/u_dual_blend.h"
32 static uint32_t r600_translate_blend_function(int blend_func
)
36 return V_028804_COMB_DST_PLUS_SRC
;
37 case PIPE_BLEND_SUBTRACT
:
38 return V_028804_COMB_SRC_MINUS_DST
;
39 case PIPE_BLEND_REVERSE_SUBTRACT
:
40 return V_028804_COMB_DST_MINUS_SRC
;
42 return V_028804_COMB_MIN_DST_SRC
;
44 return V_028804_COMB_MAX_DST_SRC
;
46 R600_ERR("Unknown blend function %d\n", blend_func
);
53 static uint32_t r600_translate_blend_factor(int blend_fact
)
56 case PIPE_BLENDFACTOR_ONE
:
57 return V_028804_BLEND_ONE
;
58 case PIPE_BLENDFACTOR_SRC_COLOR
:
59 return V_028804_BLEND_SRC_COLOR
;
60 case PIPE_BLENDFACTOR_SRC_ALPHA
:
61 return V_028804_BLEND_SRC_ALPHA
;
62 case PIPE_BLENDFACTOR_DST_ALPHA
:
63 return V_028804_BLEND_DST_ALPHA
;
64 case PIPE_BLENDFACTOR_DST_COLOR
:
65 return V_028804_BLEND_DST_COLOR
;
66 case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE
:
67 return V_028804_BLEND_SRC_ALPHA_SATURATE
;
68 case PIPE_BLENDFACTOR_CONST_COLOR
:
69 return V_028804_BLEND_CONST_COLOR
;
70 case PIPE_BLENDFACTOR_CONST_ALPHA
:
71 return V_028804_BLEND_CONST_ALPHA
;
72 case PIPE_BLENDFACTOR_ZERO
:
73 return V_028804_BLEND_ZERO
;
74 case PIPE_BLENDFACTOR_INV_SRC_COLOR
:
75 return V_028804_BLEND_ONE_MINUS_SRC_COLOR
;
76 case PIPE_BLENDFACTOR_INV_SRC_ALPHA
:
77 return V_028804_BLEND_ONE_MINUS_SRC_ALPHA
;
78 case PIPE_BLENDFACTOR_INV_DST_ALPHA
:
79 return V_028804_BLEND_ONE_MINUS_DST_ALPHA
;
80 case PIPE_BLENDFACTOR_INV_DST_COLOR
:
81 return V_028804_BLEND_ONE_MINUS_DST_COLOR
;
82 case PIPE_BLENDFACTOR_INV_CONST_COLOR
:
83 return V_028804_BLEND_ONE_MINUS_CONST_COLOR
;
84 case PIPE_BLENDFACTOR_INV_CONST_ALPHA
:
85 return V_028804_BLEND_ONE_MINUS_CONST_ALPHA
;
86 case PIPE_BLENDFACTOR_SRC1_COLOR
:
87 return V_028804_BLEND_SRC1_COLOR
;
88 case PIPE_BLENDFACTOR_SRC1_ALPHA
:
89 return V_028804_BLEND_SRC1_ALPHA
;
90 case PIPE_BLENDFACTOR_INV_SRC1_COLOR
:
91 return V_028804_BLEND_INV_SRC1_COLOR
;
92 case PIPE_BLENDFACTOR_INV_SRC1_ALPHA
:
93 return V_028804_BLEND_INV_SRC1_ALPHA
;
95 R600_ERR("Bad blend factor %d not supported!\n", blend_fact
);
102 static unsigned r600_tex_dim(unsigned dim
)
106 case PIPE_TEXTURE_1D
:
107 return V_038000_SQ_TEX_DIM_1D
;
108 case PIPE_TEXTURE_1D_ARRAY
:
109 return V_038000_SQ_TEX_DIM_1D_ARRAY
;
110 case PIPE_TEXTURE_2D
:
111 case PIPE_TEXTURE_RECT
:
112 return V_038000_SQ_TEX_DIM_2D
;
113 case PIPE_TEXTURE_2D_ARRAY
:
114 return V_038000_SQ_TEX_DIM_2D_ARRAY
;
115 case PIPE_TEXTURE_3D
:
116 return V_038000_SQ_TEX_DIM_3D
;
117 case PIPE_TEXTURE_CUBE
:
118 return V_038000_SQ_TEX_DIM_CUBEMAP
;
122 static uint32_t r600_translate_dbformat(enum pipe_format format
)
125 case PIPE_FORMAT_Z16_UNORM
:
126 return V_028010_DEPTH_16
;
127 case PIPE_FORMAT_Z24X8_UNORM
:
128 return V_028010_DEPTH_X8_24
;
129 case PIPE_FORMAT_Z24_UNORM_S8_UINT
:
130 return V_028010_DEPTH_8_24
;
131 case PIPE_FORMAT_Z32_FLOAT
:
132 return V_028010_DEPTH_32_FLOAT
;
133 case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT
:
134 return V_028010_DEPTH_X24_8_32_FLOAT
;
140 static uint32_t r600_translate_colorswap(enum pipe_format format
)
144 case PIPE_FORMAT_A8_UNORM
:
145 case PIPE_FORMAT_A8_SNORM
:
146 case PIPE_FORMAT_A8_UINT
:
147 case PIPE_FORMAT_A8_SINT
:
148 case PIPE_FORMAT_A16_UNORM
:
149 case PIPE_FORMAT_A16_SNORM
:
150 case PIPE_FORMAT_A16_UINT
:
151 case PIPE_FORMAT_A16_SINT
:
152 case PIPE_FORMAT_A16_FLOAT
:
153 case PIPE_FORMAT_A32_UINT
:
154 case PIPE_FORMAT_A32_SINT
:
155 case PIPE_FORMAT_A32_FLOAT
:
156 case PIPE_FORMAT_R4A4_UNORM
:
157 return V_0280A0_SWAP_ALT_REV
;
158 case PIPE_FORMAT_I8_UNORM
:
159 case PIPE_FORMAT_I8_SNORM
:
160 case PIPE_FORMAT_I8_UINT
:
161 case PIPE_FORMAT_I8_SINT
:
162 case PIPE_FORMAT_L8_UNORM
:
163 case PIPE_FORMAT_L8_SNORM
:
164 case PIPE_FORMAT_L8_UINT
:
165 case PIPE_FORMAT_L8_SINT
:
166 case PIPE_FORMAT_L8_SRGB
:
167 case PIPE_FORMAT_L16_UNORM
:
168 case PIPE_FORMAT_L16_SNORM
:
169 case PIPE_FORMAT_L16_UINT
:
170 case PIPE_FORMAT_L16_SINT
:
171 case PIPE_FORMAT_L16_FLOAT
:
172 case PIPE_FORMAT_L32_UINT
:
173 case PIPE_FORMAT_L32_SINT
:
174 case PIPE_FORMAT_L32_FLOAT
:
175 case PIPE_FORMAT_I16_UNORM
:
176 case PIPE_FORMAT_I16_SNORM
:
177 case PIPE_FORMAT_I16_UINT
:
178 case PIPE_FORMAT_I16_SINT
:
179 case PIPE_FORMAT_I16_FLOAT
:
180 case PIPE_FORMAT_I32_UINT
:
181 case PIPE_FORMAT_I32_SINT
:
182 case PIPE_FORMAT_I32_FLOAT
:
183 case PIPE_FORMAT_R8_UNORM
:
184 case PIPE_FORMAT_R8_SNORM
:
185 case PIPE_FORMAT_R8_UINT
:
186 case PIPE_FORMAT_R8_SINT
:
187 return V_0280A0_SWAP_STD
;
189 case PIPE_FORMAT_L4A4_UNORM
:
190 case PIPE_FORMAT_A4R4_UNORM
:
191 return V_0280A0_SWAP_ALT
;
193 /* 16-bit buffers. */
194 case PIPE_FORMAT_B5G6R5_UNORM
:
195 return V_0280A0_SWAP_STD_REV
;
197 case PIPE_FORMAT_B5G5R5A1_UNORM
:
198 case PIPE_FORMAT_B5G5R5X1_UNORM
:
199 return V_0280A0_SWAP_ALT
;
201 case PIPE_FORMAT_B4G4R4A4_UNORM
:
202 case PIPE_FORMAT_B4G4R4X4_UNORM
:
203 return V_0280A0_SWAP_ALT
;
205 case PIPE_FORMAT_Z16_UNORM
:
206 return V_0280A0_SWAP_STD
;
208 case PIPE_FORMAT_L8A8_UNORM
:
209 case PIPE_FORMAT_L8A8_SNORM
:
210 case PIPE_FORMAT_L8A8_UINT
:
211 case PIPE_FORMAT_L8A8_SINT
:
212 case PIPE_FORMAT_L8A8_SRGB
:
213 case PIPE_FORMAT_L16A16_UNORM
:
214 case PIPE_FORMAT_L16A16_SNORM
:
215 case PIPE_FORMAT_L16A16_UINT
:
216 case PIPE_FORMAT_L16A16_SINT
:
217 case PIPE_FORMAT_L16A16_FLOAT
:
218 case PIPE_FORMAT_L32A32_UINT
:
219 case PIPE_FORMAT_L32A32_SINT
:
220 case PIPE_FORMAT_L32A32_FLOAT
:
221 return V_0280A0_SWAP_ALT
;
222 case PIPE_FORMAT_R8G8_UNORM
:
223 case PIPE_FORMAT_R8G8_SNORM
:
224 case PIPE_FORMAT_R8G8_UINT
:
225 case PIPE_FORMAT_R8G8_SINT
:
226 return V_0280A0_SWAP_STD
;
228 case PIPE_FORMAT_R16_UNORM
:
229 case PIPE_FORMAT_R16_SNORM
:
230 case PIPE_FORMAT_R16_UINT
:
231 case PIPE_FORMAT_R16_SINT
:
232 case PIPE_FORMAT_R16_FLOAT
:
233 return V_0280A0_SWAP_STD
;
235 /* 32-bit buffers. */
237 case PIPE_FORMAT_A8B8G8R8_SRGB
:
238 return V_0280A0_SWAP_STD_REV
;
239 case PIPE_FORMAT_B8G8R8A8_SRGB
:
240 return V_0280A0_SWAP_ALT
;
242 case PIPE_FORMAT_B8G8R8A8_UNORM
:
243 case PIPE_FORMAT_B8G8R8X8_UNORM
:
244 return V_0280A0_SWAP_ALT
;
246 case PIPE_FORMAT_A8R8G8B8_UNORM
:
247 case PIPE_FORMAT_X8R8G8B8_UNORM
:
248 return V_0280A0_SWAP_ALT_REV
;
249 case PIPE_FORMAT_R8G8B8A8_SNORM
:
250 case PIPE_FORMAT_R8G8B8A8_UNORM
:
251 case PIPE_FORMAT_R8G8B8X8_UNORM
:
252 case PIPE_FORMAT_R8G8B8A8_SINT
:
253 case PIPE_FORMAT_R8G8B8A8_UINT
:
254 return V_0280A0_SWAP_STD
;
256 case PIPE_FORMAT_A8B8G8R8_UNORM
:
257 case PIPE_FORMAT_X8B8G8R8_UNORM
:
258 /* case PIPE_FORMAT_R8SG8SB8UX8U_NORM: */
259 return V_0280A0_SWAP_STD_REV
;
261 case PIPE_FORMAT_Z24X8_UNORM
:
262 case PIPE_FORMAT_Z24_UNORM_S8_UINT
:
263 return V_0280A0_SWAP_STD
;
265 case PIPE_FORMAT_X8Z24_UNORM
:
266 case PIPE_FORMAT_S8_UINT_Z24_UNORM
:
267 return V_0280A0_SWAP_STD
;
269 case PIPE_FORMAT_R10G10B10A2_UNORM
:
270 case PIPE_FORMAT_R10G10B10X2_SNORM
:
271 case PIPE_FORMAT_R10SG10SB10SA2U_NORM
:
272 return V_0280A0_SWAP_STD
;
274 case PIPE_FORMAT_B10G10R10A2_UNORM
:
275 case PIPE_FORMAT_B10G10R10A2_UINT
:
276 return V_0280A0_SWAP_ALT
;
278 case PIPE_FORMAT_R11G11B10_FLOAT
:
279 case PIPE_FORMAT_R16G16_UNORM
:
280 case PIPE_FORMAT_R16G16_SNORM
:
281 case PIPE_FORMAT_R16G16_FLOAT
:
282 case PIPE_FORMAT_R16G16_UINT
:
283 case PIPE_FORMAT_R16G16_SINT
:
284 case PIPE_FORMAT_R16G16B16_FLOAT
:
285 case PIPE_FORMAT_R32G32B32_FLOAT
:
286 case PIPE_FORMAT_R32_UINT
:
287 case PIPE_FORMAT_R32_SINT
:
288 case PIPE_FORMAT_R32_FLOAT
:
289 case PIPE_FORMAT_Z32_FLOAT
:
290 return V_0280A0_SWAP_STD
;
292 /* 64-bit buffers. */
293 case PIPE_FORMAT_R32G32_FLOAT
:
294 case PIPE_FORMAT_R32G32_UINT
:
295 case PIPE_FORMAT_R32G32_SINT
:
296 case PIPE_FORMAT_R16G16B16A16_UNORM
:
297 case PIPE_FORMAT_R16G16B16A16_SNORM
:
298 case PIPE_FORMAT_R16G16B16A16_UINT
:
299 case PIPE_FORMAT_R16G16B16A16_SINT
:
300 case PIPE_FORMAT_R16G16B16A16_FLOAT
:
301 case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT
:
303 /* 128-bit buffers. */
304 case PIPE_FORMAT_R32G32B32A32_FLOAT
:
305 case PIPE_FORMAT_R32G32B32A32_SNORM
:
306 case PIPE_FORMAT_R32G32B32A32_UNORM
:
307 case PIPE_FORMAT_R32G32B32A32_SINT
:
308 case PIPE_FORMAT_R32G32B32A32_UINT
:
309 return V_0280A0_SWAP_STD
;
311 R600_ERR("unsupported colorswap format %d\n", format
);
317 static uint32_t r600_translate_colorformat(enum pipe_format format
)
320 case PIPE_FORMAT_L4A4_UNORM
:
321 case PIPE_FORMAT_R4A4_UNORM
:
322 case PIPE_FORMAT_A4R4_UNORM
:
323 return V_0280A0_COLOR_4_4
;
326 case PIPE_FORMAT_A8_UNORM
:
327 case PIPE_FORMAT_A8_SNORM
:
328 case PIPE_FORMAT_A8_UINT
:
329 case PIPE_FORMAT_A8_SINT
:
330 case PIPE_FORMAT_I8_UNORM
:
331 case PIPE_FORMAT_I8_SNORM
:
332 case PIPE_FORMAT_I8_UINT
:
333 case PIPE_FORMAT_I8_SINT
:
334 case PIPE_FORMAT_L8_UNORM
:
335 case PIPE_FORMAT_L8_SNORM
:
336 case PIPE_FORMAT_L8_UINT
:
337 case PIPE_FORMAT_L8_SINT
:
338 case PIPE_FORMAT_L8_SRGB
:
339 case PIPE_FORMAT_R8_UNORM
:
340 case PIPE_FORMAT_R8_SNORM
:
341 case PIPE_FORMAT_R8_UINT
:
342 case PIPE_FORMAT_R8_SINT
:
343 return V_0280A0_COLOR_8
;
345 /* 16-bit buffers. */
346 case PIPE_FORMAT_B5G6R5_UNORM
:
347 return V_0280A0_COLOR_5_6_5
;
349 case PIPE_FORMAT_B5G5R5A1_UNORM
:
350 case PIPE_FORMAT_B5G5R5X1_UNORM
:
351 return V_0280A0_COLOR_1_5_5_5
;
353 case PIPE_FORMAT_B4G4R4A4_UNORM
:
354 case PIPE_FORMAT_B4G4R4X4_UNORM
:
355 return V_0280A0_COLOR_4_4_4_4
;
357 case PIPE_FORMAT_Z16_UNORM
:
358 return V_0280A0_COLOR_16
;
360 case PIPE_FORMAT_L8A8_UNORM
:
361 case PIPE_FORMAT_L8A8_SNORM
:
362 case PIPE_FORMAT_L8A8_UINT
:
363 case PIPE_FORMAT_L8A8_SINT
:
364 case PIPE_FORMAT_L8A8_SRGB
:
365 case PIPE_FORMAT_R8G8_UNORM
:
366 case PIPE_FORMAT_R8G8_SNORM
:
367 case PIPE_FORMAT_R8G8_UINT
:
368 case PIPE_FORMAT_R8G8_SINT
:
369 return V_0280A0_COLOR_8_8
;
371 case PIPE_FORMAT_R16_UNORM
:
372 case PIPE_FORMAT_R16_SNORM
:
373 case PIPE_FORMAT_R16_UINT
:
374 case PIPE_FORMAT_R16_SINT
:
375 case PIPE_FORMAT_A16_UNORM
:
376 case PIPE_FORMAT_A16_SNORM
:
377 case PIPE_FORMAT_A16_UINT
:
378 case PIPE_FORMAT_A16_SINT
:
379 case PIPE_FORMAT_L16_UNORM
:
380 case PIPE_FORMAT_L16_SNORM
:
381 case PIPE_FORMAT_L16_UINT
:
382 case PIPE_FORMAT_L16_SINT
:
383 case PIPE_FORMAT_I16_UNORM
:
384 case PIPE_FORMAT_I16_SNORM
:
385 case PIPE_FORMAT_I16_UINT
:
386 case PIPE_FORMAT_I16_SINT
:
387 return V_0280A0_COLOR_16
;
389 case PIPE_FORMAT_R16_FLOAT
:
390 case PIPE_FORMAT_A16_FLOAT
:
391 case PIPE_FORMAT_L16_FLOAT
:
392 case PIPE_FORMAT_I16_FLOAT
:
393 return V_0280A0_COLOR_16_FLOAT
;
395 /* 32-bit buffers. */
396 case PIPE_FORMAT_A8B8G8R8_SRGB
:
397 case PIPE_FORMAT_A8B8G8R8_UNORM
:
398 case PIPE_FORMAT_A8R8G8B8_UNORM
:
399 case PIPE_FORMAT_B8G8R8A8_SRGB
:
400 case PIPE_FORMAT_B8G8R8A8_UNORM
:
401 case PIPE_FORMAT_B8G8R8X8_UNORM
:
402 case PIPE_FORMAT_R8G8B8A8_SNORM
:
403 case PIPE_FORMAT_R8G8B8A8_UNORM
:
404 case PIPE_FORMAT_R8G8B8X8_UNORM
:
405 case PIPE_FORMAT_R8SG8SB8UX8U_NORM
:
406 case PIPE_FORMAT_X8B8G8R8_UNORM
:
407 case PIPE_FORMAT_X8R8G8B8_UNORM
:
408 case PIPE_FORMAT_R8G8B8_UNORM
:
409 case PIPE_FORMAT_R8G8B8A8_SINT
:
410 case PIPE_FORMAT_R8G8B8A8_UINT
:
411 return V_0280A0_COLOR_8_8_8_8
;
413 case PIPE_FORMAT_R10G10B10A2_UNORM
:
414 case PIPE_FORMAT_R10G10B10X2_SNORM
:
415 case PIPE_FORMAT_B10G10R10A2_UNORM
:
416 case PIPE_FORMAT_B10G10R10A2_UINT
:
417 case PIPE_FORMAT_R10SG10SB10SA2U_NORM
:
418 return V_0280A0_COLOR_2_10_10_10
;
420 case PIPE_FORMAT_Z24X8_UNORM
:
421 case PIPE_FORMAT_Z24_UNORM_S8_UINT
:
422 return V_0280A0_COLOR_8_24
;
424 case PIPE_FORMAT_X8Z24_UNORM
:
425 case PIPE_FORMAT_S8_UINT_Z24_UNORM
:
426 return V_0280A0_COLOR_24_8
;
428 case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT
:
429 return V_0280A0_COLOR_X24_8_32_FLOAT
;
431 case PIPE_FORMAT_R32_UINT
:
432 case PIPE_FORMAT_R32_SINT
:
433 case PIPE_FORMAT_A32_UINT
:
434 case PIPE_FORMAT_A32_SINT
:
435 case PIPE_FORMAT_L32_UINT
:
436 case PIPE_FORMAT_L32_SINT
:
437 case PIPE_FORMAT_I32_UINT
:
438 case PIPE_FORMAT_I32_SINT
:
439 return V_0280A0_COLOR_32
;
441 case PIPE_FORMAT_R32_FLOAT
:
442 case PIPE_FORMAT_A32_FLOAT
:
443 case PIPE_FORMAT_L32_FLOAT
:
444 case PIPE_FORMAT_I32_FLOAT
:
445 case PIPE_FORMAT_Z32_FLOAT
:
446 return V_0280A0_COLOR_32_FLOAT
;
448 case PIPE_FORMAT_R16G16_FLOAT
:
449 case PIPE_FORMAT_L16A16_FLOAT
:
450 return V_0280A0_COLOR_16_16_FLOAT
;
452 case PIPE_FORMAT_R16G16_UNORM
:
453 case PIPE_FORMAT_R16G16_SNORM
:
454 case PIPE_FORMAT_R16G16_UINT
:
455 case PIPE_FORMAT_R16G16_SINT
:
456 case PIPE_FORMAT_L16A16_UNORM
:
457 case PIPE_FORMAT_L16A16_SNORM
:
458 case PIPE_FORMAT_L16A16_UINT
:
459 case PIPE_FORMAT_L16A16_SINT
:
460 return V_0280A0_COLOR_16_16
;
462 case PIPE_FORMAT_R11G11B10_FLOAT
:
463 return V_0280A0_COLOR_10_11_11_FLOAT
;
465 /* 64-bit buffers. */
466 case PIPE_FORMAT_R16G16B16A16_UINT
:
467 case PIPE_FORMAT_R16G16B16A16_SINT
:
468 case PIPE_FORMAT_R16G16B16A16_UNORM
:
469 case PIPE_FORMAT_R16G16B16A16_SNORM
:
470 return V_0280A0_COLOR_16_16_16_16
;
472 case PIPE_FORMAT_R16G16B16_FLOAT
:
473 case PIPE_FORMAT_R16G16B16A16_FLOAT
:
474 return V_0280A0_COLOR_16_16_16_16_FLOAT
;
476 case PIPE_FORMAT_R32G32_FLOAT
:
477 case PIPE_FORMAT_L32A32_FLOAT
:
478 return V_0280A0_COLOR_32_32_FLOAT
;
480 case PIPE_FORMAT_R32G32_SINT
:
481 case PIPE_FORMAT_R32G32_UINT
:
482 case PIPE_FORMAT_L32A32_UINT
:
483 case PIPE_FORMAT_L32A32_SINT
:
484 return V_0280A0_COLOR_32_32
;
486 /* 96-bit buffers. */
487 case PIPE_FORMAT_R32G32B32_FLOAT
:
488 return V_0280A0_COLOR_32_32_32_FLOAT
;
490 /* 128-bit buffers. */
491 case PIPE_FORMAT_R32G32B32A32_FLOAT
:
492 return V_0280A0_COLOR_32_32_32_32_FLOAT
;
493 case PIPE_FORMAT_R32G32B32A32_SNORM
:
494 case PIPE_FORMAT_R32G32B32A32_UNORM
:
495 case PIPE_FORMAT_R32G32B32A32_SINT
:
496 case PIPE_FORMAT_R32G32B32A32_UINT
:
497 return V_0280A0_COLOR_32_32_32_32
;
500 case PIPE_FORMAT_UYVY
:
501 case PIPE_FORMAT_YUYV
:
503 return ~0U; /* Unsupported. */
507 static uint32_t r600_colorformat_endian_swap(uint32_t colorformat
)
509 if (R600_BIG_ENDIAN
) {
510 switch(colorformat
) {
511 case V_0280A0_COLOR_4_4
:
515 case V_0280A0_COLOR_8
:
518 /* 16-bit buffers. */
519 case V_0280A0_COLOR_5_6_5
:
520 case V_0280A0_COLOR_1_5_5_5
:
521 case V_0280A0_COLOR_4_4_4_4
:
522 case V_0280A0_COLOR_16
:
523 case V_0280A0_COLOR_8_8
:
526 /* 32-bit buffers. */
527 case V_0280A0_COLOR_8_8_8_8
:
528 case V_0280A0_COLOR_2_10_10_10
:
529 case V_0280A0_COLOR_8_24
:
530 case V_0280A0_COLOR_24_8
:
531 case V_0280A0_COLOR_32_FLOAT
:
532 case V_0280A0_COLOR_16_16_FLOAT
:
533 case V_0280A0_COLOR_16_16
:
536 /* 64-bit buffers. */
537 case V_0280A0_COLOR_16_16_16_16
:
538 case V_0280A0_COLOR_16_16_16_16_FLOAT
:
541 case V_0280A0_COLOR_32_32_FLOAT
:
542 case V_0280A0_COLOR_32_32
:
543 case V_0280A0_COLOR_X24_8_32_FLOAT
:
546 /* 128-bit buffers. */
547 case V_0280A0_COLOR_32_32_32_FLOAT
:
548 case V_0280A0_COLOR_32_32_32_32_FLOAT
:
549 case V_0280A0_COLOR_32_32_32_32
:
552 return ENDIAN_NONE
; /* Unsupported. */
559 static bool r600_is_sampler_format_supported(struct pipe_screen
*screen
, enum pipe_format format
)
561 return r600_translate_texformat(screen
, format
, NULL
, NULL
, NULL
) != ~0U;
564 static bool r600_is_colorbuffer_format_supported(enum pipe_format format
)
566 return r600_translate_colorformat(format
) != ~0U &&
567 r600_translate_colorswap(format
) != ~0U;
570 static bool r600_is_zs_format_supported(enum pipe_format format
)
572 return r600_translate_dbformat(format
) != ~0U;
575 boolean
r600_is_format_supported(struct pipe_screen
*screen
,
576 enum pipe_format format
,
577 enum pipe_texture_target target
,
578 unsigned sample_count
,
583 if (target
>= PIPE_MAX_TEXTURE_TYPES
) {
584 R600_ERR("r600: unsupported texture type %d\n", target
);
588 if (!util_format_is_supported(format
, usage
))
592 if (sample_count
> 1)
595 if ((usage
& PIPE_BIND_SAMPLER_VIEW
) &&
596 r600_is_sampler_format_supported(screen
, format
)) {
597 retval
|= PIPE_BIND_SAMPLER_VIEW
;
600 if ((usage
& (PIPE_BIND_RENDER_TARGET
|
601 PIPE_BIND_DISPLAY_TARGET
|
603 PIPE_BIND_SHARED
)) &&
604 r600_is_colorbuffer_format_supported(format
)) {
606 (PIPE_BIND_RENDER_TARGET
|
607 PIPE_BIND_DISPLAY_TARGET
|
612 if ((usage
& PIPE_BIND_DEPTH_STENCIL
) &&
613 r600_is_zs_format_supported(format
)) {
614 retval
|= PIPE_BIND_DEPTH_STENCIL
;
617 if ((usage
& PIPE_BIND_VERTEX_BUFFER
) &&
618 r600_is_vertex_format_supported(format
)) {
619 retval
|= PIPE_BIND_VERTEX_BUFFER
;
622 if (usage
& PIPE_BIND_TRANSFER_READ
)
623 retval
|= PIPE_BIND_TRANSFER_READ
;
624 if (usage
& PIPE_BIND_TRANSFER_WRITE
)
625 retval
|= PIPE_BIND_TRANSFER_WRITE
;
627 return retval
== usage
;
630 void r600_polygon_offset_update(struct r600_context
*rctx
)
632 struct r600_pipe_state state
;
634 state
.id
= R600_PIPE_STATE_POLYGON_OFFSET
;
636 if (rctx
->rasterizer
&& rctx
->framebuffer
.zsbuf
) {
637 float offset_units
= rctx
->rasterizer
->offset_units
;
638 unsigned offset_db_fmt_cntl
= 0, depth
;
640 switch (rctx
->framebuffer
.zsbuf
->format
) {
641 case PIPE_FORMAT_Z24X8_UNORM
:
642 case PIPE_FORMAT_Z24_UNORM_S8_UINT
:
644 offset_units
*= 2.0f
;
646 case PIPE_FORMAT_Z32_FLOAT
:
647 case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT
:
649 offset_units
*= 1.0f
;
650 offset_db_fmt_cntl
|= S_028DF8_POLY_OFFSET_DB_IS_FLOAT_FMT(1);
652 case PIPE_FORMAT_Z16_UNORM
:
654 offset_units
*= 4.0f
;
659 /* XXX some of those reg can be computed with cso */
660 offset_db_fmt_cntl
|= S_028DF8_POLY_OFFSET_NEG_NUM_DB_BITS(depth
);
661 r600_pipe_state_add_reg(&state
,
662 R_028E00_PA_SU_POLY_OFFSET_FRONT_SCALE
,
663 fui(rctx
->rasterizer
->offset_scale
));
664 r600_pipe_state_add_reg(&state
,
665 R_028E04_PA_SU_POLY_OFFSET_FRONT_OFFSET
,
667 r600_pipe_state_add_reg(&state
,
668 R_028E08_PA_SU_POLY_OFFSET_BACK_SCALE
,
669 fui(rctx
->rasterizer
->offset_scale
));
670 r600_pipe_state_add_reg(&state
,
671 R_028E0C_PA_SU_POLY_OFFSET_BACK_OFFSET
,
673 r600_pipe_state_add_reg(&state
,
674 R_028DF8_PA_SU_POLY_OFFSET_DB_FMT_CNTL
,
676 r600_context_pipe_state_set(rctx
, &state
);
680 static void *r600_create_blend_state(struct pipe_context
*ctx
,
681 const struct pipe_blend_state
*state
)
683 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
684 struct r600_pipe_blend
*blend
= CALLOC_STRUCT(r600_pipe_blend
);
685 struct r600_pipe_state
*rstate
;
686 uint32_t color_control
= 0, target_mask
;
691 rstate
= &blend
->rstate
;
693 rstate
->id
= R600_PIPE_STATE_BLEND
;
697 /* R600 does not support per-MRT blends */
698 if (rctx
->family
> CHIP_R600
)
699 color_control
|= S_028808_PER_MRT_BLEND(1);
700 if (state
->logicop_enable
) {
701 color_control
|= (state
->logicop_func
<< 16) | (state
->logicop_func
<< 20);
703 color_control
|= (0xcc << 16);
705 /* we pretend 8 buffer are used, CB_SHADER_MASK will disable unused one */
706 if (state
->independent_blend_enable
) {
707 for (int i
= 0; i
< 8; i
++) {
708 if (state
->rt
[i
].blend_enable
) {
709 color_control
|= S_028808_TARGET_BLEND_ENABLE(1 << i
);
711 target_mask
|= (state
->rt
[i
].colormask
<< (4 * i
));
714 for (int i
= 0; i
< 8; i
++) {
715 if (state
->rt
[0].blend_enable
) {
716 color_control
|= S_028808_TARGET_BLEND_ENABLE(1 << i
);
718 target_mask
|= (state
->rt
[0].colormask
<< (4 * i
));
723 color_control
|= S_028808_SPECIAL_OP(V_028808_NORMAL
);
725 color_control
|= S_028808_SPECIAL_OP(V_028808_DISABLE
);
727 blend
->cb_target_mask
= target_mask
;
728 blend
->cb_color_control
= color_control
;
729 /* only MRT0 has dual src blend */
730 blend
->dual_src_blend
= util_blend_state_is_dual(state
, 0);
731 for (int i
= 0; i
< 8; i
++) {
732 /* state->rt entries > 0 only written if independent blending */
733 const int j
= state
->independent_blend_enable
? i
: 0;
735 unsigned eqRGB
= state
->rt
[j
].rgb_func
;
736 unsigned srcRGB
= state
->rt
[j
].rgb_src_factor
;
737 unsigned dstRGB
= state
->rt
[j
].rgb_dst_factor
;
739 unsigned eqA
= state
->rt
[j
].alpha_func
;
740 unsigned srcA
= state
->rt
[j
].alpha_src_factor
;
741 unsigned dstA
= state
->rt
[j
].alpha_dst_factor
;
744 if (!state
->rt
[j
].blend_enable
)
747 bc
|= S_028804_COLOR_COMB_FCN(r600_translate_blend_function(eqRGB
));
748 bc
|= S_028804_COLOR_SRCBLEND(r600_translate_blend_factor(srcRGB
));
749 bc
|= S_028804_COLOR_DESTBLEND(r600_translate_blend_factor(dstRGB
));
751 if (srcA
!= srcRGB
|| dstA
!= dstRGB
|| eqA
!= eqRGB
) {
752 bc
|= S_028804_SEPARATE_ALPHA_BLEND(1);
753 bc
|= S_028804_ALPHA_COMB_FCN(r600_translate_blend_function(eqA
));
754 bc
|= S_028804_ALPHA_SRCBLEND(r600_translate_blend_factor(srcA
));
755 bc
|= S_028804_ALPHA_DESTBLEND(r600_translate_blend_factor(dstA
));
758 /* R600 does not support per-MRT blends */
759 if (rctx
->family
> CHIP_R600
)
760 r600_pipe_state_add_reg(rstate
, R_028780_CB_BLEND0_CONTROL
+ i
* 4, bc
);
762 r600_pipe_state_add_reg(rstate
, R_028804_CB_BLEND_CONTROL
, bc
);
767 static void *r600_create_dsa_state(struct pipe_context
*ctx
,
768 const struct pipe_depth_stencil_alpha_state
*state
)
770 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
771 struct r600_pipe_dsa
*dsa
= CALLOC_STRUCT(r600_pipe_dsa
);
772 unsigned db_depth_control
, alpha_test_control
, alpha_ref
;
773 struct r600_pipe_state
*rstate
;
779 dsa
->valuemask
[0] = state
->stencil
[0].valuemask
;
780 dsa
->valuemask
[1] = state
->stencil
[1].valuemask
;
781 dsa
->writemask
[0] = state
->stencil
[0].writemask
;
782 dsa
->writemask
[1] = state
->stencil
[1].writemask
;
784 rstate
= &dsa
->rstate
;
786 rstate
->id
= R600_PIPE_STATE_DSA
;
787 db_depth_control
= S_028800_Z_ENABLE(state
->depth
.enabled
) |
788 S_028800_Z_WRITE_ENABLE(state
->depth
.writemask
) |
789 S_028800_ZFUNC(state
->depth
.func
);
792 if (state
->stencil
[0].enabled
) {
793 db_depth_control
|= S_028800_STENCIL_ENABLE(1);
794 db_depth_control
|= S_028800_STENCILFUNC(state
->stencil
[0].func
); /* translates straight */
795 db_depth_control
|= S_028800_STENCILFAIL(r600_translate_stencil_op(state
->stencil
[0].fail_op
));
796 db_depth_control
|= S_028800_STENCILZPASS(r600_translate_stencil_op(state
->stencil
[0].zpass_op
));
797 db_depth_control
|= S_028800_STENCILZFAIL(r600_translate_stencil_op(state
->stencil
[0].zfail_op
));
799 if (state
->stencil
[1].enabled
) {
800 db_depth_control
|= S_028800_BACKFACE_ENABLE(1);
801 db_depth_control
|= S_028800_STENCILFUNC_BF(state
->stencil
[1].func
); /* translates straight */
802 db_depth_control
|= S_028800_STENCILFAIL_BF(r600_translate_stencil_op(state
->stencil
[1].fail_op
));
803 db_depth_control
|= S_028800_STENCILZPASS_BF(r600_translate_stencil_op(state
->stencil
[1].zpass_op
));
804 db_depth_control
|= S_028800_STENCILZFAIL_BF(r600_translate_stencil_op(state
->stencil
[1].zfail_op
));
809 alpha_test_control
= 0;
811 if (state
->alpha
.enabled
) {
812 alpha_test_control
= S_028410_ALPHA_FUNC(state
->alpha
.func
);
813 alpha_test_control
|= S_028410_ALPHA_TEST_ENABLE(1);
814 alpha_ref
= fui(state
->alpha
.ref_value
);
816 dsa
->sx_alpha_test_control
= alpha_test_control
& 0xff;
817 dsa
->alpha_ref
= alpha_ref
;
819 r600_pipe_state_add_reg(rstate
, R_028800_DB_DEPTH_CONTROL
, db_depth_control
);
823 static void *r600_create_rs_state(struct pipe_context
*ctx
,
824 const struct pipe_rasterizer_state
*state
)
826 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
827 struct r600_pipe_rasterizer
*rs
= CALLOC_STRUCT(r600_pipe_rasterizer
);
828 struct r600_pipe_state
*rstate
;
830 unsigned prov_vtx
= 1, polygon_dual_mode
;
831 unsigned sc_mode_cntl
;
832 float psize_min
, psize_max
;
838 polygon_dual_mode
= (state
->fill_front
!= PIPE_POLYGON_MODE_FILL
||
839 state
->fill_back
!= PIPE_POLYGON_MODE_FILL
);
841 if (state
->flatshade_first
)
844 rstate
= &rs
->rstate
;
845 rs
->flatshade
= state
->flatshade
;
846 rs
->sprite_coord_enable
= state
->sprite_coord_enable
;
847 rs
->two_side
= state
->light_twoside
;
848 rs
->clip_plane_enable
= state
->clip_plane_enable
;
849 rs
->pa_sc_line_stipple
= state
->line_stipple_enable
?
850 S_028A0C_LINE_PATTERN(state
->line_stipple_pattern
) |
851 S_028A0C_REPEAT_COUNT(state
->line_stipple_factor
) : 0;
852 rs
->pa_cl_clip_cntl
=
853 S_028810_PS_UCP_MODE(3) |
854 S_028810_ZCLIP_NEAR_DISABLE(!state
->depth_clip
) |
855 S_028810_ZCLIP_FAR_DISABLE(!state
->depth_clip
) |
856 S_028810_DX_LINEAR_ATTR_CLIP_ENA(1);
859 rs
->offset_units
= state
->offset_units
;
860 rs
->offset_scale
= state
->offset_scale
* 12.0f
;
862 rstate
->id
= R600_PIPE_STATE_RASTERIZER
;
863 tmp
= S_0286D4_FLAT_SHADE_ENA(1);
864 if (state
->sprite_coord_enable
) {
865 tmp
|= S_0286D4_PNT_SPRITE_ENA(1) |
866 S_0286D4_PNT_SPRITE_OVRD_X(2) |
867 S_0286D4_PNT_SPRITE_OVRD_Y(3) |
868 S_0286D4_PNT_SPRITE_OVRD_Z(0) |
869 S_0286D4_PNT_SPRITE_OVRD_W(1);
870 if (state
->sprite_coord_mode
!= PIPE_SPRITE_COORD_UPPER_LEFT
) {
871 tmp
|= S_0286D4_PNT_SPRITE_TOP_1(1);
874 r600_pipe_state_add_reg(rstate
, R_0286D4_SPI_INTERP_CONTROL_0
, tmp
);
876 /* point size 12.4 fixed point */
877 tmp
= r600_pack_float_12p4(state
->point_size
/2);
878 r600_pipe_state_add_reg(rstate
, R_028A00_PA_SU_POINT_SIZE
, S_028A00_HEIGHT(tmp
) | S_028A00_WIDTH(tmp
));
880 if (state
->point_size_per_vertex
) {
881 psize_min
= util_get_min_point_size(state
);
884 /* Force the point size to be as if the vertex output was disabled. */
885 psize_min
= state
->point_size
;
886 psize_max
= state
->point_size
;
888 /* Divide by two, because 0.5 = 1 pixel. */
889 r600_pipe_state_add_reg(rstate
, R_028A04_PA_SU_POINT_MINMAX
,
890 S_028A04_MIN_SIZE(r600_pack_float_12p4(psize_min
/2)) |
891 S_028A04_MAX_SIZE(r600_pack_float_12p4(psize_max
/2)));
893 tmp
= r600_pack_float_12p4(state
->line_width
/2);
894 r600_pipe_state_add_reg(rstate
, R_028A08_PA_SU_LINE_CNTL
, S_028A08_WIDTH(tmp
));
896 if (rctx
->chip_class
>= R700
) {
898 S_028A4C_FORCE_EOV_CNTDWN_ENABLE(1) |
899 S_028A4C_FORCE_EOV_REZ_ENABLE(1) |
900 S_028A4C_R700_ZMM_LINE_OFFSET(1) |
901 S_028A4C_R700_VPORT_SCISSOR_ENABLE(state
->scissor
);
904 S_028A4C_WALK_ALIGN8_PRIM_FITS_ST(1) |
905 S_028A4C_FORCE_EOV_CNTDWN_ENABLE(1);
906 rs
->scissor_enable
= state
->scissor
;
908 sc_mode_cntl
|= S_028A4C_LINE_STIPPLE_ENABLE(state
->line_stipple_enable
);
910 r600_pipe_state_add_reg(rstate
, R_028A4C_PA_SC_MODE_CNTL
, sc_mode_cntl
);
912 r600_pipe_state_add_reg(rstate
, R_028C08_PA_SU_VTX_CNTL
,
913 S_028C08_PIX_CENTER_HALF(state
->gl_rasterization_rules
));
915 r600_pipe_state_add_reg(rstate
, R_028DFC_PA_SU_POLY_OFFSET_CLAMP
, fui(state
->offset_clamp
));
916 r600_pipe_state_add_reg(rstate
, R_028814_PA_SU_SC_MODE_CNTL
,
917 S_028814_PROVOKING_VTX_LAST(prov_vtx
) |
918 S_028814_CULL_FRONT(state
->cull_face
& PIPE_FACE_FRONT
? 1 : 0) |
919 S_028814_CULL_BACK(state
->cull_face
& PIPE_FACE_BACK
? 1 : 0) |
920 S_028814_FACE(!state
->front_ccw
) |
921 S_028814_POLY_OFFSET_FRONT_ENABLE(state
->offset_tri
) |
922 S_028814_POLY_OFFSET_BACK_ENABLE(state
->offset_tri
) |
923 S_028814_POLY_OFFSET_PARA_ENABLE(state
->offset_tri
) |
924 S_028814_POLY_MODE(polygon_dual_mode
) |
925 S_028814_POLYMODE_FRONT_PTYPE(r600_translate_fill(state
->fill_front
)) |
926 S_028814_POLYMODE_BACK_PTYPE(r600_translate_fill(state
->fill_back
)));
927 r600_pipe_state_add_reg(rstate
, R_028350_SX_MISC
, S_028350_MULTIPASS(state
->rasterizer_discard
));
931 static void *r600_create_sampler_state(struct pipe_context
*ctx
,
932 const struct pipe_sampler_state
*state
)
934 struct r600_pipe_sampler_state
*ss
= CALLOC_STRUCT(r600_pipe_sampler_state
);
935 struct r600_pipe_state
*rstate
;
937 unsigned aniso_flag_offset
= state
->max_anisotropy
> 1 ? 4 : 0;
943 ss
->seamless_cube_map
= state
->seamless_cube_map
;
944 rstate
= &ss
->rstate
;
945 rstate
->id
= R600_PIPE_STATE_SAMPLER
;
946 util_pack_color(state
->border_color
.f
, PIPE_FORMAT_B8G8R8A8_UNORM
, &uc
);
947 r600_pipe_state_add_reg_noblock(rstate
, R_03C000_SQ_TEX_SAMPLER_WORD0_0
,
948 S_03C000_CLAMP_X(r600_tex_wrap(state
->wrap_s
)) |
949 S_03C000_CLAMP_Y(r600_tex_wrap(state
->wrap_t
)) |
950 S_03C000_CLAMP_Z(r600_tex_wrap(state
->wrap_r
)) |
951 S_03C000_XY_MAG_FILTER(r600_tex_filter(state
->mag_img_filter
) | aniso_flag_offset
) |
952 S_03C000_XY_MIN_FILTER(r600_tex_filter(state
->min_img_filter
) | aniso_flag_offset
) |
953 S_03C000_MIP_FILTER(r600_tex_mipfilter(state
->min_mip_filter
)) |
954 S_03C000_MAX_ANISO(r600_tex_aniso_filter(state
->max_anisotropy
)) |
955 S_03C000_DEPTH_COMPARE_FUNCTION(r600_tex_compare(state
->compare_func
)) |
956 S_03C000_BORDER_COLOR_TYPE(uc
.ui
? V_03C000_SQ_TEX_BORDER_COLOR_REGISTER
: 0), NULL
, 0);
957 r600_pipe_state_add_reg_noblock(rstate
, R_03C004_SQ_TEX_SAMPLER_WORD1_0
,
958 S_03C004_MIN_LOD(S_FIXED(CLAMP(state
->min_lod
, 0, 15), 6)) |
959 S_03C004_MAX_LOD(S_FIXED(CLAMP(state
->max_lod
, 0, 15), 6)) |
960 S_03C004_LOD_BIAS(S_FIXED(CLAMP(state
->lod_bias
, -16, 16), 6)), NULL
, 0);
961 r600_pipe_state_add_reg_noblock(rstate
, R_03C008_SQ_TEX_SAMPLER_WORD2_0
, S_03C008_TYPE(1), NULL
, 0);
963 r600_pipe_state_add_reg_noblock(rstate
, R_00A400_TD_PS_SAMPLER0_BORDER_RED
, fui(state
->border_color
.f
[0]), NULL
, 0);
964 r600_pipe_state_add_reg_noblock(rstate
, R_00A404_TD_PS_SAMPLER0_BORDER_GREEN
, fui(state
->border_color
.f
[1]), NULL
, 0);
965 r600_pipe_state_add_reg_noblock(rstate
, R_00A408_TD_PS_SAMPLER0_BORDER_BLUE
, fui(state
->border_color
.f
[2]), NULL
, 0);
966 r600_pipe_state_add_reg_noblock(rstate
, R_00A40C_TD_PS_SAMPLER0_BORDER_ALPHA
, fui(state
->border_color
.f
[3]), NULL
, 0);
971 static struct pipe_sampler_view
*r600_create_sampler_view(struct pipe_context
*ctx
,
972 struct pipe_resource
*texture
,
973 const struct pipe_sampler_view
*state
)
975 struct r600_screen
*rscreen
= (struct r600_screen
*)ctx
->screen
;
976 struct r600_pipe_sampler_view
*view
= CALLOC_STRUCT(r600_pipe_sampler_view
);
977 struct r600_resource_texture
*tmp
= (struct r600_resource_texture
*)texture
;
978 unsigned format
, endian
;
979 uint32_t word4
= 0, yuv_format
= 0, pitch
= 0;
980 unsigned char swizzle
[4], array_mode
= 0, tile_type
= 0;
981 unsigned width
, height
, depth
, offset_level
, last_level
;
986 /* initialize base object */
988 view
->base
.texture
= NULL
;
989 pipe_reference(NULL
, &texture
->reference
);
990 view
->base
.texture
= texture
;
991 view
->base
.reference
.count
= 1;
992 view
->base
.context
= ctx
;
994 swizzle
[0] = state
->swizzle_r
;
995 swizzle
[1] = state
->swizzle_g
;
996 swizzle
[2] = state
->swizzle_b
;
997 swizzle
[3] = state
->swizzle_a
;
999 format
= r600_translate_texformat(ctx
->screen
, state
->format
,
1001 &word4
, &yuv_format
);
1002 assert(format
!= ~0);
1008 if (tmp
->is_depth
&& !tmp
->is_flushing_texture
) {
1009 if (!r600_init_flushed_depth_texture(ctx
, texture
, NULL
)) {
1013 tmp
= tmp
->flushed_depth_texture
;
1016 endian
= r600_colorformat_endian_swap(format
);
1018 offset_level
= state
->u
.tex
.first_level
;
1019 last_level
= state
->u
.tex
.last_level
- offset_level
;
1020 if (!rscreen
->use_surface_alloc
) {
1021 width
= u_minify(texture
->width0
, offset_level
);
1022 height
= u_minify(texture
->height0
, offset_level
);
1023 depth
= u_minify(texture
->depth0
, offset_level
);
1025 pitch
= align(tmp
->pitch_in_blocks
[offset_level
] *
1026 util_format_get_blockwidth(state
->format
), 8);
1027 array_mode
= tmp
->array_mode
[offset_level
];
1028 tile_type
= tmp
->tile_type
;
1030 if (texture
->target
== PIPE_TEXTURE_1D_ARRAY
) {
1032 depth
= texture
->array_size
;
1033 } else if (texture
->target
== PIPE_TEXTURE_2D_ARRAY
) {
1034 depth
= texture
->array_size
;
1037 view
->tex_resource
= &tmp
->resource
;
1038 view
->tex_resource_words
[0] = (S_038000_DIM(r600_tex_dim(texture
->target
)) |
1039 S_038000_TILE_MODE(array_mode
) |
1040 S_038000_TILE_TYPE(tile_type
) |
1041 S_038000_PITCH((pitch
/ 8) - 1) |
1042 S_038000_TEX_WIDTH(width
- 1));
1043 view
->tex_resource_words
[1] = (S_038004_TEX_HEIGHT(height
- 1) |
1044 S_038004_TEX_DEPTH(depth
- 1) |
1045 S_038004_DATA_FORMAT(format
));
1046 view
->tex_resource_words
[2] = tmp
->offset
[offset_level
] >> 8;
1047 view
->tex_resource_words
[3] = tmp
->offset
[offset_level
+1] >> 8;
1048 view
->tex_resource_words
[4] = (word4
|
1049 S_038010_SRF_MODE_ALL(V_038010_SRF_MODE_ZERO_CLAMP_MINUS_ONE
) |
1050 S_038010_REQUEST_SIZE(1) |
1051 S_038010_ENDIAN_SWAP(endian
) |
1052 S_038010_BASE_LEVEL(0));
1053 view
->tex_resource_words
[5] = (S_038014_LAST_LEVEL(last_level
) |
1054 S_038014_BASE_ARRAY(state
->u
.tex
.first_layer
) |
1055 S_038014_LAST_ARRAY(state
->u
.tex
.last_layer
));
1056 view
->tex_resource_words
[6] = (S_038018_TYPE(V_038010_SQ_TEX_VTX_VALID_TEXTURE
) |
1057 S_038018_MAX_ANISO(4 /* max 16 samples */));
1059 width
= tmp
->surface
.level
[offset_level
].npix_x
;
1060 height
= tmp
->surface
.level
[offset_level
].npix_y
;
1061 depth
= tmp
->surface
.level
[offset_level
].npix_z
;
1062 pitch
= tmp
->surface
.level
[offset_level
].nblk_x
* util_format_get_blockwidth(state
->format
);
1063 tile_type
= tmp
->tile_type
;
1065 if (texture
->target
== PIPE_TEXTURE_1D_ARRAY
) {
1067 depth
= texture
->array_size
;
1068 } else if (texture
->target
== PIPE_TEXTURE_2D_ARRAY
) {
1069 depth
= texture
->array_size
;
1071 switch (tmp
->surface
.level
[offset_level
].mode
) {
1072 case RADEON_SURF_MODE_LINEAR_ALIGNED
:
1073 array_mode
= V_038000_ARRAY_LINEAR_ALIGNED
;
1075 case RADEON_SURF_MODE_1D
:
1076 array_mode
= V_038000_ARRAY_1D_TILED_THIN1
;
1078 case RADEON_SURF_MODE_2D
:
1079 array_mode
= V_038000_ARRAY_2D_TILED_THIN1
;
1081 case RADEON_SURF_MODE_LINEAR
:
1083 array_mode
= V_038000_ARRAY_LINEAR_GENERAL
;
1087 view
->tex_resource
= &tmp
->resource
;
1088 view
->tex_resource_words
[0] = (S_038000_DIM(r600_tex_dim(texture
->target
)) |
1089 S_038000_TILE_MODE(array_mode
) |
1090 S_038000_TILE_TYPE(tile_type
) |
1091 S_038000_PITCH((pitch
/ 8) - 1) |
1092 S_038000_TEX_WIDTH(width
- 1));
1093 view
->tex_resource_words
[1] = (S_038004_TEX_HEIGHT(height
- 1) |
1094 S_038004_TEX_DEPTH(depth
- 1) |
1095 S_038004_DATA_FORMAT(format
));
1096 view
->tex_resource_words
[2] = tmp
->surface
.level
[offset_level
].offset
>> 8;
1097 if (offset_level
>= tmp
->surface
.last_level
) {
1098 view
->tex_resource_words
[3] = tmp
->surface
.level
[offset_level
].offset
>> 8;
1100 view
->tex_resource_words
[3] = tmp
->surface
.level
[offset_level
+ 1].offset
>> 8;
1102 view
->tex_resource_words
[4] = (word4
|
1103 S_038010_SRF_MODE_ALL(V_038010_SRF_MODE_ZERO_CLAMP_MINUS_ONE
) |
1104 S_038010_REQUEST_SIZE(1) |
1105 S_038010_ENDIAN_SWAP(endian
) |
1106 S_038010_BASE_LEVEL(0));
1107 view
->tex_resource_words
[5] = (S_038014_LAST_LEVEL(last_level
) |
1108 S_038014_BASE_ARRAY(state
->u
.tex
.first_layer
) |
1109 S_038014_LAST_ARRAY(state
->u
.tex
.last_layer
));
1110 view
->tex_resource_words
[6] = (S_038018_TYPE(V_038010_SQ_TEX_VTX_VALID_TEXTURE
) |
1111 S_038018_MAX_ANISO(4 /* max 16 samples */));
1116 static void r600_set_vs_sampler_views(struct pipe_context
*ctx
, unsigned count
,
1117 struct pipe_sampler_view
**views
)
1119 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
1120 r600_set_sampler_views(rctx
, &rctx
->vs_samplers
, count
, views
);
1123 static void r600_set_ps_sampler_views(struct pipe_context
*ctx
, unsigned count
,
1124 struct pipe_sampler_view
**views
)
1126 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
1127 r600_set_sampler_views(rctx
, &rctx
->ps_samplers
, count
, views
);
1130 static void r600_set_seamless_cubemap(struct r600_context
*rctx
, boolean enable
)
1132 struct r600_pipe_state
*rstate
= CALLOC_STRUCT(r600_pipe_state
);
1136 rstate
->id
= R600_PIPE_STATE_SEAMLESS_CUBEMAP
;
1137 r600_pipe_state_add_reg(rstate
, R_009508_TA_CNTL_AUX
,
1138 (enable
? 0 : S_009508_DISABLE_CUBE_WRAP(1)) |
1139 S_009508_DISABLE_CUBE_ANISO(1) |
1140 S_009508_SYNC_GRADIENT(1) |
1141 S_009508_SYNC_WALKER(1) |
1142 S_009508_SYNC_ALIGNER(1));
1144 free(rctx
->states
[R600_PIPE_STATE_SEAMLESS_CUBEMAP
]);
1145 rctx
->states
[R600_PIPE_STATE_SEAMLESS_CUBEMAP
] = rstate
;
1146 r600_context_pipe_state_set(rctx
, rstate
);
1149 static void r600_bind_samplers(struct r600_context
*rctx
,
1150 struct r600_textures_info
*dst
,
1151 unsigned count
, void **states
)
1153 memcpy(dst
->samplers
, states
, sizeof(void*) * count
);
1154 dst
->n_samplers
= count
;
1155 dst
->samplers_dirty
= true;
1158 static void r600_bind_vs_samplers(struct pipe_context
*ctx
, unsigned count
, void **states
)
1160 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
1161 r600_bind_samplers(rctx
, &rctx
->vs_samplers
, count
, states
);
1164 static void r600_bind_ps_samplers(struct pipe_context
*ctx
, unsigned count
, void **states
)
1166 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
1167 r600_bind_samplers(rctx
, &rctx
->ps_samplers
, count
, states
);
1170 static void r600_update_samplers(struct r600_context
*rctx
,
1171 struct r600_textures_info
*tex
,
1172 void (*set_sampler
)(struct r600_context
*, struct r600_pipe_state
*, unsigned))
1176 if (tex
->samplers_dirty
) {
1178 for (i
= 0; i
< tex
->n_samplers
; i
++) {
1179 if (!tex
->samplers
[i
])
1182 /* TEX_ARRAY_OVERRIDE must be set for array textures to disable
1183 * filtering between layers.
1184 * Don't update TEX_ARRAY_OVERRIDE if we don't have the sampler view. */
1185 if (tex
->views
.views
[i
]) {
1186 if (tex
->views
.views
[i
]->base
.texture
->target
== PIPE_TEXTURE_1D_ARRAY
||
1187 tex
->views
.views
[i
]->base
.texture
->target
== PIPE_TEXTURE_2D_ARRAY
) {
1188 tex
->samplers
[i
]->rstate
.regs
[0].value
|= S_03C000_TEX_ARRAY_OVERRIDE(1);
1189 tex
->is_array_sampler
[i
] = true;
1191 tex
->samplers
[i
]->rstate
.regs
[0].value
&= C_03C000_TEX_ARRAY_OVERRIDE
;
1192 tex
->is_array_sampler
[i
] = false;
1196 set_sampler(rctx
, &tex
->samplers
[i
]->rstate
, i
);
1198 if (tex
->samplers
[i
])
1199 seamless
= tex
->samplers
[i
]->seamless_cube_map
;
1203 r600_set_seamless_cubemap(rctx
, seamless
);
1205 tex
->samplers_dirty
= false;
1209 void r600_update_sampler_states(struct r600_context
*rctx
)
1211 r600_update_samplers(rctx
, &rctx
->vs_samplers
,
1212 r600_context_pipe_state_set_vs_sampler
);
1213 r600_update_samplers(rctx
, &rctx
->ps_samplers
,
1214 r600_context_pipe_state_set_ps_sampler
);
1217 static void r600_set_clip_state(struct pipe_context
*ctx
,
1218 const struct pipe_clip_state
*state
)
1220 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
1221 struct r600_pipe_state
*rstate
= CALLOC_STRUCT(r600_pipe_state
);
1222 struct pipe_constant_buffer cb
;
1227 rctx
->clip
= *state
;
1228 rstate
->id
= R600_PIPE_STATE_CLIP
;
1229 for (int i
= 0; i
< 6; i
++) {
1230 r600_pipe_state_add_reg(rstate
,
1231 R_028E20_PA_CL_UCP0_X
+ i
* 16,
1232 fui(state
->ucp
[i
][0]));
1233 r600_pipe_state_add_reg(rstate
,
1234 R_028E24_PA_CL_UCP0_Y
+ i
* 16,
1235 fui(state
->ucp
[i
][1]) );
1236 r600_pipe_state_add_reg(rstate
,
1237 R_028E28_PA_CL_UCP0_Z
+ i
* 16,
1238 fui(state
->ucp
[i
][2]));
1239 r600_pipe_state_add_reg(rstate
,
1240 R_028E2C_PA_CL_UCP0_W
+ i
* 16,
1241 fui(state
->ucp
[i
][3]));
1244 free(rctx
->states
[R600_PIPE_STATE_CLIP
]);
1245 rctx
->states
[R600_PIPE_STATE_CLIP
] = rstate
;
1246 r600_context_pipe_state_set(rctx
, rstate
);
1249 cb
.user_buffer
= state
->ucp
;
1250 cb
.buffer_offset
= 0;
1251 cb
.buffer_size
= 4*4*8;
1252 r600_set_constant_buffer(ctx
, PIPE_SHADER_VERTEX
, 1, &cb
);
1253 pipe_resource_reference(&cb
.buffer
, NULL
);
1256 static void r600_set_polygon_stipple(struct pipe_context
*ctx
,
1257 const struct pipe_poly_stipple
*state
)
1261 static void r600_set_sample_mask(struct pipe_context
*pipe
, unsigned sample_mask
)
1265 void r600_set_scissor_state(struct r600_context
*rctx
,
1266 const struct pipe_scissor_state
*state
)
1268 struct r600_pipe_state
*rstate
= CALLOC_STRUCT(r600_pipe_state
);
1274 rstate
->id
= R600_PIPE_STATE_SCISSOR
;
1275 tl
= S_028240_TL_X(state
->minx
) | S_028240_TL_Y(state
->miny
) | S_028240_WINDOW_OFFSET_DISABLE(1);
1276 br
= S_028244_BR_X(state
->maxx
) | S_028244_BR_Y(state
->maxy
);
1277 r600_pipe_state_add_reg(rstate
,
1278 R_028250_PA_SC_VPORT_SCISSOR_0_TL
, tl
);
1279 r600_pipe_state_add_reg(rstate
,
1280 R_028254_PA_SC_VPORT_SCISSOR_0_BR
, br
);
1282 free(rctx
->states
[R600_PIPE_STATE_SCISSOR
]);
1283 rctx
->states
[R600_PIPE_STATE_SCISSOR
] = rstate
;
1284 r600_context_pipe_state_set(rctx
, rstate
);
1287 static void r600_pipe_set_scissor_state(struct pipe_context
*ctx
,
1288 const struct pipe_scissor_state
*state
)
1290 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
1292 if (rctx
->chip_class
== R600
) {
1293 rctx
->scissor_state
= *state
;
1295 if (!rctx
->scissor_enable
)
1299 r600_set_scissor_state(rctx
, state
);
1302 static void r600_set_viewport_state(struct pipe_context
*ctx
,
1303 const struct pipe_viewport_state
*state
)
1305 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
1306 struct r600_pipe_state
*rstate
= CALLOC_STRUCT(r600_pipe_state
);
1311 rctx
->viewport
= *state
;
1312 rstate
->id
= R600_PIPE_STATE_VIEWPORT
;
1313 r600_pipe_state_add_reg(rstate
, R_02843C_PA_CL_VPORT_XSCALE_0
, fui(state
->scale
[0]));
1314 r600_pipe_state_add_reg(rstate
, R_028444_PA_CL_VPORT_YSCALE_0
, fui(state
->scale
[1]));
1315 r600_pipe_state_add_reg(rstate
, R_02844C_PA_CL_VPORT_ZSCALE_0
, fui(state
->scale
[2]));
1316 r600_pipe_state_add_reg(rstate
, R_028440_PA_CL_VPORT_XOFFSET_0
, fui(state
->translate
[0]));
1317 r600_pipe_state_add_reg(rstate
, R_028448_PA_CL_VPORT_YOFFSET_0
, fui(state
->translate
[1]));
1318 r600_pipe_state_add_reg(rstate
, R_028450_PA_CL_VPORT_ZOFFSET_0
, fui(state
->translate
[2]));
1320 free(rctx
->states
[R600_PIPE_STATE_VIEWPORT
]);
1321 rctx
->states
[R600_PIPE_STATE_VIEWPORT
] = rstate
;
1322 r600_context_pipe_state_set(rctx
, rstate
);
1325 static void r600_cb(struct r600_context
*rctx
, struct r600_pipe_state
*rstate
,
1326 const struct pipe_framebuffer_state
*state
, int cb
)
1328 struct r600_screen
*rscreen
= rctx
->screen
;
1329 struct r600_resource_texture
*rtex
;
1330 struct r600_surface
*surf
;
1331 unsigned level
= state
->cbufs
[cb
]->u
.tex
.level
;
1332 unsigned pitch
, slice
;
1333 unsigned color_info
;
1334 unsigned format
, swap
, ntype
, endian
;
1336 const struct util_format_description
*desc
;
1338 bool blend_bypass
= 0, blend_clamp
= 1, alphatest_bypass
;
1340 surf
= (struct r600_surface
*)state
->cbufs
[cb
];
1341 rtex
= (struct r600_resource_texture
*)state
->cbufs
[cb
]->texture
;
1343 if (rtex
->is_depth
&& !rtex
->is_flushing_texture
) {
1344 rtex
= rtex
->flushed_depth_texture
;
1347 /* XXX quite sure for dx10+ hw don't need any offset hacks */
1348 if (!rscreen
->use_surface_alloc
) {
1349 offset
= r600_texture_get_offset(rtex
,
1350 level
, state
->cbufs
[cb
]->u
.tex
.first_layer
);
1351 pitch
= rtex
->pitch_in_blocks
[level
] / 8 - 1;
1352 slice
= rtex
->pitch_in_blocks
[level
] * surf
->aligned_height
/ 64;
1356 color_info
= S_0280A0_ARRAY_MODE(rtex
->array_mode
[level
]);
1358 offset
= rtex
->surface
.level
[level
].offset
;
1359 if (rtex
->surface
.level
[level
].mode
< RADEON_SURF_MODE_1D
) {
1360 offset
+= rtex
->surface
.level
[level
].slice_size
*
1361 state
->cbufs
[cb
]->u
.tex
.first_layer
;
1363 pitch
= rtex
->surface
.level
[level
].nblk_x
/ 8 - 1;
1364 slice
= (rtex
->surface
.level
[level
].nblk_x
* rtex
->surface
.level
[level
].nblk_y
) / 64;
1369 switch (rtex
->surface
.level
[level
].mode
) {
1370 case RADEON_SURF_MODE_LINEAR_ALIGNED
:
1371 color_info
= S_0280A0_ARRAY_MODE(V_038000_ARRAY_LINEAR_ALIGNED
);
1373 case RADEON_SURF_MODE_1D
:
1374 color_info
= S_0280A0_ARRAY_MODE(V_038000_ARRAY_1D_TILED_THIN1
);
1376 case RADEON_SURF_MODE_2D
:
1377 color_info
= S_0280A0_ARRAY_MODE(V_038000_ARRAY_2D_TILED_THIN1
);
1379 case RADEON_SURF_MODE_LINEAR
:
1381 color_info
= S_0280A0_ARRAY_MODE(V_038000_ARRAY_LINEAR_GENERAL
);
1385 desc
= util_format_description(surf
->base
.format
);
1387 for (i
= 0; i
< 4; i
++) {
1388 if (desc
->channel
[i
].type
!= UTIL_FORMAT_TYPE_VOID
) {
1393 ntype
= V_0280A0_NUMBER_UNORM
;
1394 if (desc
->colorspace
== UTIL_FORMAT_COLORSPACE_SRGB
)
1395 ntype
= V_0280A0_NUMBER_SRGB
;
1396 else if (desc
->channel
[i
].type
== UTIL_FORMAT_TYPE_SIGNED
) {
1397 if (desc
->channel
[i
].normalized
)
1398 ntype
= V_0280A0_NUMBER_SNORM
;
1399 else if (desc
->channel
[i
].pure_integer
)
1400 ntype
= V_0280A0_NUMBER_SINT
;
1401 } else if (desc
->channel
[i
].type
== UTIL_FORMAT_TYPE_UNSIGNED
) {
1402 if (desc
->channel
[i
].normalized
)
1403 ntype
= V_0280A0_NUMBER_UNORM
;
1404 else if (desc
->channel
[i
].pure_integer
)
1405 ntype
= V_0280A0_NUMBER_UINT
;
1408 format
= r600_translate_colorformat(surf
->base
.format
);
1409 assert(format
!= ~0);
1411 swap
= r600_translate_colorswap(surf
->base
.format
);
1414 if (rtex
->resource
.b
.b
.usage
== PIPE_USAGE_STAGING
) {
1415 endian
= ENDIAN_NONE
;
1417 endian
= r600_colorformat_endian_swap(format
);
1420 /* set blend bypass according to docs if SINT/UINT or
1421 8/24 COLOR variants */
1422 if (ntype
== V_0280A0_NUMBER_UINT
|| ntype
== V_0280A0_NUMBER_SINT
||
1423 format
== V_0280A0_COLOR_8_24
|| format
== V_0280A0_COLOR_24_8
||
1424 format
== V_0280A0_COLOR_X24_8_32_FLOAT
) {
1429 /* Alpha-test is done on the first colorbuffer only. */
1431 alphatest_bypass
= ntype
== V_0280A0_NUMBER_UINT
|| ntype
== V_0280A0_NUMBER_SINT
;
1432 if (rctx
->alphatest_state
.bypass
!= alphatest_bypass
) {
1433 rctx
->alphatest_state
.bypass
= alphatest_bypass
;
1434 r600_atom_dirty(rctx
, &rctx
->alphatest_state
.atom
);
1438 color_info
|= S_0280A0_FORMAT(format
) |
1439 S_0280A0_COMP_SWAP(swap
) |
1440 S_0280A0_BLEND_BYPASS(blend_bypass
) |
1441 S_0280A0_BLEND_CLAMP(blend_clamp
) |
1442 S_0280A0_NUMBER_TYPE(ntype
) |
1443 S_0280A0_ENDIAN(endian
);
1445 /* EXPORT_NORM is an optimzation that can be enabled for better
1446 * performance in certain cases
1448 if (rctx
->chip_class
== R600
) {
1449 /* EXPORT_NORM can be enabled if:
1450 * - 11-bit or smaller UNORM/SNORM/SRGB
1451 * - BLEND_CLAMP is enabled
1452 * - BLEND_FLOAT32 is disabled
1454 if (desc
->colorspace
!= UTIL_FORMAT_COLORSPACE_ZS
&&
1455 (desc
->channel
[i
].size
< 12 &&
1456 desc
->channel
[i
].type
!= UTIL_FORMAT_TYPE_FLOAT
&&
1457 ntype
!= V_0280A0_NUMBER_UINT
&&
1458 ntype
!= V_0280A0_NUMBER_SINT
) &&
1459 G_0280A0_BLEND_CLAMP(color_info
) &&
1460 !G_0280A0_BLEND_FLOAT32(color_info
)) {
1461 color_info
|= S_0280A0_SOURCE_FORMAT(V_0280A0_EXPORT_NORM
);
1463 rctx
->export_16bpc
= false;
1466 /* EXPORT_NORM can be enabled if:
1467 * - 11-bit or smaller UNORM/SNORM/SRGB
1468 * - 16-bit or smaller FLOAT
1470 if (desc
->colorspace
!= UTIL_FORMAT_COLORSPACE_ZS
&&
1471 ((desc
->channel
[i
].size
< 12 &&
1472 desc
->channel
[i
].type
!= UTIL_FORMAT_TYPE_FLOAT
&&
1473 ntype
!= V_0280A0_NUMBER_UINT
&& ntype
!= V_0280A0_NUMBER_SINT
) ||
1474 (desc
->channel
[i
].size
< 17 &&
1475 desc
->channel
[i
].type
== UTIL_FORMAT_TYPE_FLOAT
))) {
1476 color_info
|= S_0280A0_SOURCE_FORMAT(V_0280A0_EXPORT_NORM
);
1478 rctx
->export_16bpc
= false;
1482 /* for possible dual-src MRT write color info 1 */
1483 if (cb
== 0 && rctx
->framebuffer
.nr_cbufs
== 1) {
1484 r600_pipe_state_add_reg_bo(rstate
,
1485 R_0280A0_CB_COLOR0_INFO
+ 1 * 4,
1486 color_info
, &rtex
->resource
, RADEON_USAGE_READWRITE
);
1489 r600_pipe_state_add_reg_bo(rstate
,
1490 R_028040_CB_COLOR0_BASE
+ cb
* 4,
1491 offset
>> 8, &rtex
->resource
, RADEON_USAGE_READWRITE
);
1492 r600_pipe_state_add_reg_bo(rstate
,
1493 R_0280A0_CB_COLOR0_INFO
+ cb
* 4,
1494 color_info
, &rtex
->resource
, RADEON_USAGE_READWRITE
);
1495 r600_pipe_state_add_reg(rstate
,
1496 R_028060_CB_COLOR0_SIZE
+ cb
* 4,
1497 S_028060_PITCH_TILE_MAX(pitch
) |
1498 S_028060_SLICE_TILE_MAX(slice
));
1499 if (!rscreen
->use_surface_alloc
) {
1500 r600_pipe_state_add_reg(rstate
,
1501 R_028080_CB_COLOR0_VIEW
+ cb
* 4,
1504 if (rtex
->surface
.level
[level
].mode
< RADEON_SURF_MODE_1D
) {
1505 r600_pipe_state_add_reg(rstate
,
1506 R_028080_CB_COLOR0_VIEW
+ cb
* 4,
1509 r600_pipe_state_add_reg(rstate
,
1510 R_028080_CB_COLOR0_VIEW
+ cb
* 4,
1511 S_028080_SLICE_START(state
->cbufs
[cb
]->u
.tex
.first_layer
) |
1512 S_028080_SLICE_MAX(state
->cbufs
[cb
]->u
.tex
.last_layer
));
1515 r600_pipe_state_add_reg_bo(rstate
,
1516 R_0280E0_CB_COLOR0_FRAG
+ cb
* 4,
1517 0, &rtex
->resource
, RADEON_USAGE_READWRITE
);
1518 r600_pipe_state_add_reg_bo(rstate
,
1519 R_0280C0_CB_COLOR0_TILE
+ cb
* 4,
1520 0, &rtex
->resource
, RADEON_USAGE_READWRITE
);
1523 static void r600_db(struct r600_context
*rctx
, struct r600_pipe_state
*rstate
,
1524 const struct pipe_framebuffer_state
*state
)
1526 struct r600_screen
*rscreen
= rctx
->screen
;
1527 struct r600_resource_texture
*rtex
;
1528 struct r600_surface
*surf
;
1529 unsigned level
, pitch
, slice
, format
, offset
, array_mode
;
1531 if (state
->zsbuf
== NULL
)
1534 level
= state
->zsbuf
->u
.tex
.level
;
1536 surf
= (struct r600_surface
*)state
->zsbuf
;
1537 rtex
= (struct r600_resource_texture
*)state
->zsbuf
->texture
;
1539 if (!rscreen
->use_surface_alloc
) {
1540 /* XXX remove this once tiling is properly supported */
1541 array_mode
= rtex
->array_mode
[level
] ? rtex
->array_mode
[level
] :
1542 V_0280A0_ARRAY_1D_TILED_THIN1
;
1544 /* XXX quite sure for dx10+ hw don't need any offset hacks */
1545 offset
= r600_texture_get_offset((struct r600_resource_texture
*)state
->zsbuf
->texture
,
1546 level
, state
->zsbuf
->u
.tex
.first_layer
);
1547 pitch
= rtex
->pitch_in_blocks
[level
] / 8 - 1;
1548 slice
= rtex
->pitch_in_blocks
[level
] * surf
->aligned_height
/ 64;
1553 offset
= rtex
->surface
.level
[level
].offset
;
1554 pitch
= rtex
->surface
.level
[level
].nblk_x
/ 8 - 1;
1555 slice
= (rtex
->surface
.level
[level
].nblk_x
* rtex
->surface
.level
[level
].nblk_y
) / 64;
1559 switch (rtex
->surface
.level
[level
].mode
) {
1560 case RADEON_SURF_MODE_2D
:
1561 array_mode
= V_0280A0_ARRAY_2D_TILED_THIN1
;
1563 case RADEON_SURF_MODE_1D
:
1564 case RADEON_SURF_MODE_LINEAR_ALIGNED
:
1565 case RADEON_SURF_MODE_LINEAR
:
1567 array_mode
= V_0280A0_ARRAY_1D_TILED_THIN1
;
1572 format
= r600_translate_dbformat(state
->zsbuf
->format
);
1573 assert(format
!= ~0);
1575 r600_pipe_state_add_reg_bo(rstate
, R_02800C_DB_DEPTH_BASE
,
1576 offset
>> 8, &rtex
->resource
, RADEON_USAGE_READWRITE
);
1577 r600_pipe_state_add_reg(rstate
, R_028000_DB_DEPTH_SIZE
,
1578 S_028000_PITCH_TILE_MAX(pitch
) | S_028000_SLICE_TILE_MAX(slice
));
1579 if (!rscreen
->use_surface_alloc
) {
1580 r600_pipe_state_add_reg(rstate
, R_028004_DB_DEPTH_VIEW
, 0x00000000);
1582 r600_pipe_state_add_reg(rstate
, R_028004_DB_DEPTH_VIEW
,
1583 S_028004_SLICE_START(state
->zsbuf
->u
.tex
.first_layer
) |
1584 S_028004_SLICE_MAX(state
->zsbuf
->u
.tex
.last_layer
));
1586 r600_pipe_state_add_reg_bo(rstate
, R_028010_DB_DEPTH_INFO
,
1587 S_028010_ARRAY_MODE(array_mode
) | S_028010_FORMAT(format
),
1588 &rtex
->resource
, RADEON_USAGE_READWRITE
);
1589 r600_pipe_state_add_reg(rstate
, R_028D34_DB_PREFETCH_LIMIT
,
1590 (surf
->aligned_height
/ 8) - 1);
1593 static void r600_set_framebuffer_state(struct pipe_context
*ctx
,
1594 const struct pipe_framebuffer_state
*state
)
1596 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
1597 struct r600_pipe_state
*rstate
= CALLOC_STRUCT(r600_pipe_state
);
1603 r600_flush_framebuffer(rctx
, false);
1605 /* unreference old buffer and reference new one */
1606 rstate
->id
= R600_PIPE_STATE_FRAMEBUFFER
;
1608 util_copy_framebuffer_state(&rctx
->framebuffer
, state
);
1611 rctx
->export_16bpc
= true;
1612 rctx
->nr_cbufs
= state
->nr_cbufs
;
1614 for (int i
= 0; i
< state
->nr_cbufs
; i
++) {
1615 r600_cb(rctx
, rstate
, state
, i
);
1618 r600_db(rctx
, rstate
, state
);
1621 tl
= S_028240_TL_X(0) | S_028240_TL_Y(0) | S_028240_WINDOW_OFFSET_DISABLE(1);
1622 br
= S_028244_BR_X(state
->width
) | S_028244_BR_Y(state
->height
);
1624 r600_pipe_state_add_reg(rstate
,
1625 R_028204_PA_SC_WINDOW_SCISSOR_TL
, tl
);
1626 r600_pipe_state_add_reg(rstate
,
1627 R_028208_PA_SC_WINDOW_SCISSOR_BR
, br
);
1629 /* Always enable the first colorbuffer in CB_SHADER_CONTROL. This
1630 * will assure that the alpha-test will work even if there is
1631 * no colorbuffer bound. */
1632 r600_pipe_state_add_reg(rstate
, R_0287A0_CB_SHADER_CONTROL
,
1633 (1ull << MAX2(state
->nr_cbufs
, 1)) - 1);
1635 if (state
->nr_cbufs
== 0 && rctx
->alphatest_state
.bypass
) {
1636 rctx
->alphatest_state
.bypass
= false;
1637 r600_atom_dirty(rctx
, &rctx
->alphatest_state
.atom
);
1640 free(rctx
->states
[R600_PIPE_STATE_FRAMEBUFFER
]);
1641 rctx
->states
[R600_PIPE_STATE_FRAMEBUFFER
] = rstate
;
1642 r600_context_pipe_state_set(rctx
, rstate
);
1645 r600_polygon_offset_update(rctx
);
1648 if (rctx
->cb_misc_state
.nr_cbufs
!= state
->nr_cbufs
) {
1649 rctx
->cb_misc_state
.nr_cbufs
= state
->nr_cbufs
;
1650 r600_atom_dirty(rctx
, &rctx
->cb_misc_state
.atom
);
1654 static void r600_emit_cb_misc_state(struct r600_context
*rctx
, struct r600_atom
*atom
)
1656 struct radeon_winsys_cs
*cs
= rctx
->cs
;
1657 struct r600_cb_misc_state
*a
= (struct r600_cb_misc_state
*)atom
;
1658 unsigned fb_colormask
= (1ULL << ((unsigned)a
->nr_cbufs
* 4)) - 1;
1659 unsigned ps_colormask
= (1ULL << ((unsigned)a
->nr_ps_color_outputs
* 4)) - 1;
1660 unsigned multiwrite
= a
->multiwrite
&& a
->nr_cbufs
> 1;
1662 r600_write_context_reg_seq(cs
, R_028238_CB_TARGET_MASK
, 2);
1663 r600_write_value(cs
, a
->blend_colormask
& fb_colormask
); /* R_028238_CB_TARGET_MASK */
1664 r600_write_value(cs
, (a
->dual_src_blend
? ps_colormask
: 0) | fb_colormask
); /* R_02823C_CB_SHADER_MASK */
1665 r600_write_context_reg(cs
, R_028808_CB_COLOR_CONTROL
,
1666 a
->cb_color_control
|
1667 S_028808_MULTIWRITE_ENABLE(multiwrite
));
1670 static void r600_emit_db_misc_state(struct r600_context
*rctx
, struct r600_atom
*atom
)
1672 struct radeon_winsys_cs
*cs
= rctx
->cs
;
1673 struct r600_db_misc_state
*a
= (struct r600_db_misc_state
*)atom
;
1674 unsigned db_render_control
= 0;
1675 unsigned db_render_override
=
1676 S_028D10_FORCE_HIZ_ENABLE(V_028D10_FORCE_DISABLE
) |
1677 S_028D10_FORCE_HIS_ENABLE0(V_028D10_FORCE_DISABLE
) |
1678 S_028D10_FORCE_HIS_ENABLE1(V_028D10_FORCE_DISABLE
);
1680 if (a
->occlusion_query_enabled
) {
1681 if (rctx
->chip_class
>= R700
) {
1682 db_render_control
|= S_028D0C_R700_PERFECT_ZPASS_COUNTS(1);
1684 db_render_override
|= S_028D10_NOOP_CULL_DISABLE(1);
1686 if (a
->flush_depthstencil_through_cb
) {
1687 db_render_control
|= S_028D0C_DEPTH_COPY_ENABLE(1) |
1688 S_028D0C_STENCIL_COPY_ENABLE(1) |
1689 S_028D0C_COPY_CENTROID(1);
1692 r600_write_context_reg_seq(cs
, R_028D0C_DB_RENDER_CONTROL
, 2);
1693 r600_write_value(cs
, db_render_control
); /* R_028D0C_DB_RENDER_CONTROL */
1694 r600_write_value(cs
, db_render_override
); /* R_028D10_DB_RENDER_OVERRIDE */
1697 static void r600_emit_vertex_buffers(struct r600_context
*rctx
, struct r600_atom
*atom
)
1699 struct radeon_winsys_cs
*cs
= rctx
->cs
;
1700 uint32_t dirty_mask
= rctx
->vertex_buffer_state
.dirty_mask
;
1702 while (dirty_mask
) {
1703 struct pipe_vertex_buffer
*vb
;
1704 struct r600_resource
*rbuffer
;
1706 unsigned buffer_index
= u_bit_scan(&dirty_mask
);
1708 vb
= &rctx
->vertex_buffer_state
.vb
[buffer_index
];
1709 rbuffer
= (struct r600_resource
*)vb
->buffer
;
1712 offset
= vb
->buffer_offset
;
1714 /* fetch resources start at index 320 */
1715 r600_write_value(cs
, PKT3(PKT3_SET_RESOURCE
, 7, 0));
1716 r600_write_value(cs
, (320 + buffer_index
) * 7);
1717 r600_write_value(cs
, offset
); /* RESOURCEi_WORD0 */
1718 r600_write_value(cs
, rbuffer
->buf
->size
- offset
- 1); /* RESOURCEi_WORD1 */
1719 r600_write_value(cs
, /* RESOURCEi_WORD2 */
1720 S_038008_ENDIAN_SWAP(r600_endian_swap(32)) |
1721 S_038008_STRIDE(vb
->stride
));
1722 r600_write_value(cs
, 0); /* RESOURCEi_WORD3 */
1723 r600_write_value(cs
, 0); /* RESOURCEi_WORD4 */
1724 r600_write_value(cs
, 0); /* RESOURCEi_WORD5 */
1725 r600_write_value(cs
, 0xc0000000); /* RESOURCEi_WORD6 */
1727 r600_write_value(cs
, PKT3(PKT3_NOP
, 0, 0));
1728 r600_write_value(cs
, r600_context_bo_reloc(rctx
, rbuffer
, RADEON_USAGE_READ
));
1732 static void r600_emit_constant_buffers(struct r600_context
*rctx
,
1733 struct r600_constbuf_state
*state
,
1734 unsigned buffer_id_base
,
1735 unsigned reg_alu_constbuf_size
,
1736 unsigned reg_alu_const_cache
)
1738 struct radeon_winsys_cs
*cs
= rctx
->cs
;
1739 uint32_t dirty_mask
= state
->dirty_mask
;
1741 while (dirty_mask
) {
1742 struct pipe_constant_buffer
*cb
;
1743 struct r600_resource
*rbuffer
;
1745 unsigned buffer_index
= ffs(dirty_mask
) - 1;
1747 cb
= &state
->cb
[buffer_index
];
1748 rbuffer
= (struct r600_resource
*)cb
->buffer
;
1751 offset
= cb
->buffer_offset
;
1753 r600_write_context_reg(cs
, reg_alu_constbuf_size
+ buffer_index
* 4,
1754 ALIGN_DIVUP(cb
->buffer_size
>> 4, 16));
1755 r600_write_context_reg(cs
, reg_alu_const_cache
+ buffer_index
* 4, offset
>> 8);
1757 r600_write_value(cs
, PKT3(PKT3_NOP
, 0, 0));
1758 r600_write_value(cs
, r600_context_bo_reloc(rctx
, rbuffer
, RADEON_USAGE_READ
));
1760 r600_write_value(cs
, PKT3(PKT3_SET_RESOURCE
, 7, 0));
1761 r600_write_value(cs
, (buffer_id_base
+ buffer_index
) * 7);
1762 r600_write_value(cs
, offset
); /* RESOURCEi_WORD0 */
1763 r600_write_value(cs
, rbuffer
->buf
->size
- offset
- 1); /* RESOURCEi_WORD1 */
1764 r600_write_value(cs
, /* RESOURCEi_WORD2 */
1765 S_038008_ENDIAN_SWAP(r600_endian_swap(32)) |
1766 S_038008_STRIDE(16));
1767 r600_write_value(cs
, 0); /* RESOURCEi_WORD3 */
1768 r600_write_value(cs
, 0); /* RESOURCEi_WORD4 */
1769 r600_write_value(cs
, 0); /* RESOURCEi_WORD5 */
1770 r600_write_value(cs
, 0xc0000000); /* RESOURCEi_WORD6 */
1772 r600_write_value(cs
, PKT3(PKT3_NOP
, 0, 0));
1773 r600_write_value(cs
, r600_context_bo_reloc(rctx
, rbuffer
, RADEON_USAGE_READ
));
1775 dirty_mask
&= ~(1 << buffer_index
);
1777 state
->dirty_mask
= 0;
1780 static void r600_emit_vs_constant_buffers(struct r600_context
*rctx
, struct r600_atom
*atom
)
1782 r600_emit_constant_buffers(rctx
, &rctx
->vs_constbuf_state
, 160,
1783 R_028180_ALU_CONST_BUFFER_SIZE_VS_0
,
1784 R_028980_ALU_CONST_CACHE_VS_0
);
1787 static void r600_emit_ps_constant_buffers(struct r600_context
*rctx
, struct r600_atom
*atom
)
1789 r600_emit_constant_buffers(rctx
, &rctx
->ps_constbuf_state
, 0,
1790 R_028140_ALU_CONST_BUFFER_SIZE_PS_0
,
1791 R_028940_ALU_CONST_CACHE_PS_0
);
1794 static void r600_emit_sampler_views(struct r600_context
*rctx
,
1795 struct r600_samplerview_state
*state
,
1796 unsigned resource_id_base
)
1798 struct radeon_winsys_cs
*cs
= rctx
->cs
;
1799 uint32_t dirty_mask
= state
->dirty_mask
;
1801 while (dirty_mask
) {
1802 struct r600_pipe_sampler_view
*rview
;
1803 unsigned resource_index
= u_bit_scan(&dirty_mask
);
1806 rview
= state
->views
[resource_index
];
1809 r600_write_value(cs
, PKT3(PKT3_SET_RESOURCE
, 7, 0));
1810 r600_write_value(cs
, (resource_id_base
+ resource_index
) * 7);
1811 r600_write_array(cs
, 7, rview
->tex_resource_words
);
1813 /* XXX The kernel needs two relocations. This is stupid. */
1814 reloc
= r600_context_bo_reloc(rctx
, rview
->tex_resource
,
1816 r600_write_value(cs
, PKT3(PKT3_NOP
, 0, 0));
1817 r600_write_value(cs
, reloc
);
1818 r600_write_value(cs
, PKT3(PKT3_NOP
, 0, 0));
1819 r600_write_value(cs
, reloc
);
1821 state
->dirty_mask
= 0;
1824 static void r600_emit_vs_sampler_views(struct r600_context
*rctx
, struct r600_atom
*atom
)
1826 r600_emit_sampler_views(rctx
, &rctx
->vs_samplers
.views
, 160 + R600_MAX_CONST_BUFFERS
);
1829 static void r600_emit_ps_sampler_views(struct r600_context
*rctx
, struct r600_atom
*atom
)
1831 r600_emit_sampler_views(rctx
, &rctx
->ps_samplers
.views
, R600_MAX_CONST_BUFFERS
);
1834 void r600_init_state_functions(struct r600_context
*rctx
)
1836 r600_init_atom(&rctx
->cb_misc_state
.atom
, r600_emit_cb_misc_state
, 0, 0);
1837 r600_atom_dirty(rctx
, &rctx
->cb_misc_state
.atom
);
1838 r600_init_atom(&rctx
->db_misc_state
.atom
, r600_emit_db_misc_state
, 4, 0);
1839 r600_atom_dirty(rctx
, &rctx
->db_misc_state
.atom
);
1840 r600_init_atom(&rctx
->vertex_buffer_state
.atom
, r600_emit_vertex_buffers
, 0, 0);
1841 r600_init_atom(&rctx
->vs_constbuf_state
.atom
, r600_emit_vs_constant_buffers
, 0, 0);
1842 r600_init_atom(&rctx
->ps_constbuf_state
.atom
, r600_emit_ps_constant_buffers
, 0, 0);
1843 r600_init_atom(&rctx
->vs_samplers
.views
.atom
, r600_emit_vs_sampler_views
, 0, 0);
1844 r600_init_atom(&rctx
->ps_samplers
.views
.atom
, r600_emit_ps_sampler_views
, 0, 0);
1846 rctx
->context
.create_blend_state
= r600_create_blend_state
;
1847 rctx
->context
.create_depth_stencil_alpha_state
= r600_create_dsa_state
;
1848 rctx
->context
.create_fs_state
= r600_create_shader_state_ps
;
1849 rctx
->context
.create_rasterizer_state
= r600_create_rs_state
;
1850 rctx
->context
.create_sampler_state
= r600_create_sampler_state
;
1851 rctx
->context
.create_sampler_view
= r600_create_sampler_view
;
1852 rctx
->context
.create_vertex_elements_state
= r600_create_vertex_elements
;
1853 rctx
->context
.create_vs_state
= r600_create_shader_state_vs
;
1854 rctx
->context
.bind_blend_state
= r600_bind_blend_state
;
1855 rctx
->context
.bind_depth_stencil_alpha_state
= r600_bind_dsa_state
;
1856 rctx
->context
.bind_fragment_sampler_states
= r600_bind_ps_samplers
;
1857 rctx
->context
.bind_fs_state
= r600_bind_ps_shader
;
1858 rctx
->context
.bind_rasterizer_state
= r600_bind_rs_state
;
1859 rctx
->context
.bind_vertex_elements_state
= r600_bind_vertex_elements
;
1860 rctx
->context
.bind_vertex_sampler_states
= r600_bind_vs_samplers
;
1861 rctx
->context
.bind_vs_state
= r600_bind_vs_shader
;
1862 rctx
->context
.delete_blend_state
= r600_delete_state
;
1863 rctx
->context
.delete_depth_stencil_alpha_state
= r600_delete_state
;
1864 rctx
->context
.delete_fs_state
= r600_delete_ps_shader
;
1865 rctx
->context
.delete_rasterizer_state
= r600_delete_rs_state
;
1866 rctx
->context
.delete_sampler_state
= r600_delete_state
;
1867 rctx
->context
.delete_vertex_elements_state
= r600_delete_vertex_element
;
1868 rctx
->context
.delete_vs_state
= r600_delete_vs_shader
;
1869 rctx
->context
.set_blend_color
= r600_set_blend_color
;
1870 rctx
->context
.set_clip_state
= r600_set_clip_state
;
1871 rctx
->context
.set_constant_buffer
= r600_set_constant_buffer
;
1872 rctx
->context
.set_fragment_sampler_views
= r600_set_ps_sampler_views
;
1873 rctx
->context
.set_framebuffer_state
= r600_set_framebuffer_state
;
1874 rctx
->context
.set_polygon_stipple
= r600_set_polygon_stipple
;
1875 rctx
->context
.set_sample_mask
= r600_set_sample_mask
;
1876 rctx
->context
.set_scissor_state
= r600_pipe_set_scissor_state
;
1877 rctx
->context
.set_stencil_ref
= r600_set_pipe_stencil_ref
;
1878 rctx
->context
.set_vertex_buffers
= r600_set_vertex_buffers
;
1879 rctx
->context
.set_index_buffer
= r600_set_index_buffer
;
1880 rctx
->context
.set_vertex_sampler_views
= r600_set_vs_sampler_views
;
1881 rctx
->context
.set_viewport_state
= r600_set_viewport_state
;
1882 rctx
->context
.sampler_view_destroy
= r600_sampler_view_destroy
;
1883 rctx
->context
.texture_barrier
= r600_texture_barrier
;
1884 rctx
->context
.create_stream_output_target
= r600_create_so_target
;
1885 rctx
->context
.stream_output_target_destroy
= r600_so_target_destroy
;
1886 rctx
->context
.set_stream_output_targets
= r600_set_so_targets
;
1889 /* Adjust GPR allocation on R6xx/R7xx */
1890 void r600_adjust_gprs(struct r600_context
*rctx
)
1892 struct r600_pipe_state rstate
;
1893 unsigned num_ps_gprs
= rctx
->default_ps_gprs
;
1894 unsigned num_vs_gprs
= rctx
->default_vs_gprs
;
1898 /* XXX: Following call moved from r600_bind_[ps|vs]_shader,
1899 * it seems eg+ doesn't need it, r6xx/7xx probably need it only for
1900 * adjusting the GPR allocation?
1901 * Do we need this if we aren't really changing config below? */
1902 r600_inval_shader_cache(rctx
);
1904 if (rctx
->ps_shader
->current
->shader
.bc
.ngpr
> rctx
->default_ps_gprs
)
1906 diff
= rctx
->ps_shader
->current
->shader
.bc
.ngpr
- rctx
->default_ps_gprs
;
1907 num_vs_gprs
-= diff
;
1908 num_ps_gprs
+= diff
;
1911 if (rctx
->vs_shader
->current
->shader
.bc
.ngpr
> rctx
->default_vs_gprs
)
1913 diff
= rctx
->vs_shader
->current
->shader
.bc
.ngpr
- rctx
->default_vs_gprs
;
1914 num_ps_gprs
-= diff
;
1915 num_vs_gprs
+= diff
;
1919 tmp
|= S_008C04_NUM_PS_GPRS(num_ps_gprs
);
1920 tmp
|= S_008C04_NUM_VS_GPRS(num_vs_gprs
);
1921 tmp
|= S_008C04_NUM_CLAUSE_TEMP_GPRS(rctx
->r6xx_num_clause_temp_gprs
);
1923 r600_pipe_state_add_reg(&rstate
, R_008C04_SQ_GPR_RESOURCE_MGMT_1
, tmp
);
1925 r600_context_pipe_state_set(rctx
, &rstate
);
1928 void r600_init_atom_start_cs(struct r600_context
*rctx
)
1943 int num_ps_stack_entries
;
1944 int num_vs_stack_entries
;
1945 int num_gs_stack_entries
;
1946 int num_es_stack_entries
;
1947 enum radeon_family family
;
1948 struct r600_command_buffer
*cb
= &rctx
->start_cs_cmd
;
1952 r600_init_command_buffer(cb
, 256, EMIT_EARLY
);
1954 /* R6xx requires this packet at the start of each command buffer */
1955 if (rctx
->chip_class
== R600
) {
1956 r600_store_value(cb
, PKT3(PKT3_START_3D_CMDBUF
, 0, 0));
1957 r600_store_value(cb
, 0);
1959 /* All asics require this one */
1960 r600_store_value(cb
, PKT3(PKT3_CONTEXT_CONTROL
, 1, 0));
1961 r600_store_value(cb
, 0x80000000);
1962 r600_store_value(cb
, 0x80000000);
1964 family
= rctx
->family
;
1976 num_ps_threads
= 136;
1977 num_vs_threads
= 48;
1980 num_ps_stack_entries
= 128;
1981 num_vs_stack_entries
= 128;
1982 num_gs_stack_entries
= 0;
1983 num_es_stack_entries
= 0;
1992 num_ps_threads
= 144;
1993 num_vs_threads
= 40;
1996 num_ps_stack_entries
= 40;
1997 num_vs_stack_entries
= 40;
1998 num_gs_stack_entries
= 32;
1999 num_es_stack_entries
= 16;
2011 num_ps_threads
= 136;
2012 num_vs_threads
= 48;
2015 num_ps_stack_entries
= 40;
2016 num_vs_stack_entries
= 40;
2017 num_gs_stack_entries
= 32;
2018 num_es_stack_entries
= 16;
2026 num_ps_threads
= 136;
2027 num_vs_threads
= 48;
2030 num_ps_stack_entries
= 40;
2031 num_vs_stack_entries
= 40;
2032 num_gs_stack_entries
= 32;
2033 num_es_stack_entries
= 16;
2041 num_ps_threads
= 188;
2042 num_vs_threads
= 60;
2045 num_ps_stack_entries
= 256;
2046 num_vs_stack_entries
= 256;
2047 num_gs_stack_entries
= 0;
2048 num_es_stack_entries
= 0;
2057 num_ps_threads
= 188;
2058 num_vs_threads
= 60;
2061 num_ps_stack_entries
= 128;
2062 num_vs_stack_entries
= 128;
2063 num_gs_stack_entries
= 0;
2064 num_es_stack_entries
= 0;
2072 num_ps_threads
= 144;
2073 num_vs_threads
= 48;
2076 num_ps_stack_entries
= 128;
2077 num_vs_stack_entries
= 128;
2078 num_gs_stack_entries
= 0;
2079 num_es_stack_entries
= 0;
2083 rctx
->default_ps_gprs
= num_ps_gprs
;
2084 rctx
->default_vs_gprs
= num_vs_gprs
;
2085 rctx
->r6xx_num_clause_temp_gprs
= num_temp_gprs
;
2097 tmp
|= S_008C00_VC_ENABLE(1);
2100 tmp
|= S_008C00_DX9_CONSTS(0);
2101 tmp
|= S_008C00_ALU_INST_PREFER_VECTOR(1);
2102 tmp
|= S_008C00_PS_PRIO(ps_prio
);
2103 tmp
|= S_008C00_VS_PRIO(vs_prio
);
2104 tmp
|= S_008C00_GS_PRIO(gs_prio
);
2105 tmp
|= S_008C00_ES_PRIO(es_prio
);
2106 r600_store_config_reg(cb
, R_008C00_SQ_CONFIG
, tmp
);
2108 /* SQ_GPR_RESOURCE_MGMT_2 */
2109 tmp
= S_008C08_NUM_GS_GPRS(num_gs_gprs
);
2110 tmp
|= S_008C08_NUM_ES_GPRS(num_es_gprs
);
2111 r600_store_config_reg_seq(cb
, R_008C08_SQ_GPR_RESOURCE_MGMT_2
, 4);
2112 r600_store_value(cb
, tmp
);
2114 /* SQ_THREAD_RESOURCE_MGMT */
2115 tmp
= S_008C0C_NUM_PS_THREADS(num_ps_threads
);
2116 tmp
|= S_008C0C_NUM_VS_THREADS(num_vs_threads
);
2117 tmp
|= S_008C0C_NUM_GS_THREADS(num_gs_threads
);
2118 tmp
|= S_008C0C_NUM_ES_THREADS(num_es_threads
);
2119 r600_store_value(cb
, tmp
); /* R_008C0C_SQ_THREAD_RESOURCE_MGMT */
2121 /* SQ_STACK_RESOURCE_MGMT_1 */
2122 tmp
= S_008C10_NUM_PS_STACK_ENTRIES(num_ps_stack_entries
);
2123 tmp
|= S_008C10_NUM_VS_STACK_ENTRIES(num_vs_stack_entries
);
2124 r600_store_value(cb
, tmp
); /* R_008C10_SQ_STACK_RESOURCE_MGMT_1 */
2126 /* SQ_STACK_RESOURCE_MGMT_2 */
2127 tmp
= S_008C14_NUM_GS_STACK_ENTRIES(num_gs_stack_entries
);
2128 tmp
|= S_008C14_NUM_ES_STACK_ENTRIES(num_es_stack_entries
);
2129 r600_store_value(cb
, tmp
); /* R_008C14_SQ_STACK_RESOURCE_MGMT_2 */
2131 r600_store_config_reg(cb
, R_009714_VC_ENHANCE
, 0);
2133 if (rctx
->chip_class
>= R700
) {
2134 r600_store_config_reg(cb
, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ
, 0x00004000);
2135 r600_store_config_reg(cb
, R_009830_DB_DEBUG
, 0);
2136 r600_store_config_reg(cb
, R_009838_DB_WATERMARKS
, 0x00420204);
2137 r600_store_context_reg(cb
, R_0286C8_SPI_THREAD_GROUPING
, 0);
2139 r600_store_config_reg(cb
, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ
, 0);
2140 r600_store_config_reg(cb
, R_009830_DB_DEBUG
, 0x82000000);
2141 r600_store_config_reg(cb
, R_009838_DB_WATERMARKS
, 0x01020204);
2142 r600_store_context_reg(cb
, R_0286C8_SPI_THREAD_GROUPING
, 1);
2144 r600_store_context_reg_seq(cb
, R_0288A8_SQ_ESGS_RING_ITEMSIZE
, 9);
2145 r600_store_value(cb
, 0); /* R_0288A8_SQ_ESGS_RING_ITEMSIZE */
2146 r600_store_value(cb
, 0); /* R_0288AC_SQ_GSVS_RING_ITEMSIZE */
2147 r600_store_value(cb
, 0); /* R_0288B0_SQ_ESTMP_RING_ITEMSIZE */
2148 r600_store_value(cb
, 0); /* R_0288B4_SQ_GSTMP_RING_ITEMSIZE */
2149 r600_store_value(cb
, 0); /* R_0288B8_SQ_VSTMP_RING_ITEMSIZE */
2150 r600_store_value(cb
, 0); /* R_0288BC_SQ_PSTMP_RING_ITEMSIZE */
2151 r600_store_value(cb
, 0); /* R_0288C0_SQ_FBUF_RING_ITEMSIZE */
2152 r600_store_value(cb
, 0); /* R_0288C4_SQ_REDUC_RING_ITEMSIZE */
2153 r600_store_value(cb
, 0); /* R_0288C8_SQ_GS_VERT_ITEMSIZE */
2155 r600_store_context_reg_seq(cb
, R_028A10_VGT_OUTPUT_PATH_CNTL
, 13);
2156 r600_store_value(cb
, 0); /* R_028A10_VGT_OUTPUT_PATH_CNTL */
2157 r600_store_value(cb
, 0); /* R_028A14_VGT_HOS_CNTL */
2158 r600_store_value(cb
, 0); /* R_028A18_VGT_HOS_MAX_TESS_LEVEL */
2159 r600_store_value(cb
, 0); /* R_028A1C_VGT_HOS_MIN_TESS_LEVEL */
2160 r600_store_value(cb
, 0); /* R_028A20_VGT_HOS_REUSE_DEPTH */
2161 r600_store_value(cb
, 0); /* R_028A24_VGT_GROUP_PRIM_TYPE */
2162 r600_store_value(cb
, 0); /* R_028A28_VGT_GROUP_FIRST_DECR */
2163 r600_store_value(cb
, 0); /* R_028A2C_VGT_GROUP_DECR */
2164 r600_store_value(cb
, 0); /* R_028A30_VGT_GROUP_VECT_0_CNTL */
2165 r600_store_value(cb
, 0); /* R_028A34_VGT_GROUP_VECT_1_CNTL */
2166 r600_store_value(cb
, 0); /* R_028A38_VGT_GROUP_VECT_0_FMT_CNTL */
2167 r600_store_value(cb
, 0); /* R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL */
2168 r600_store_value(cb
, 0); /* R_028A40_VGT_GS_MODE, 0); */
2170 r600_store_context_reg(cb
, R_028A84_VGT_PRIMITIVEID_EN
, 0);
2171 r600_store_context_reg(cb
, R_028AA0_VGT_INSTANCE_STEP_RATE_0
, 0);
2172 r600_store_context_reg(cb
, R_028AA4_VGT_INSTANCE_STEP_RATE_1
, 0);
2174 r600_store_context_reg_seq(cb
, R_028AB0_VGT_STRMOUT_EN
, 3);
2175 r600_store_value(cb
, 0); /* R_028AB0_VGT_STRMOUT_EN */
2176 r600_store_value(cb
, 1); /* R_028AB4_VGT_REUSE_OFF */
2177 r600_store_value(cb
, 0); /* R_028AB8_VGT_VTX_CNT_EN */
2179 r600_store_context_reg(cb
, R_028B20_VGT_STRMOUT_BUFFER_EN
, 0);
2181 r600_store_context_reg_seq(cb
, R_028400_VGT_MAX_VTX_INDX
, 2);
2182 r600_store_value(cb
, ~0); /* R_028400_VGT_MAX_VTX_INDX */
2183 r600_store_value(cb
, 0); /* R_028404_VGT_MIN_VTX_INDX */
2185 r600_store_ctl_const(cb
, R_03CFF0_SQ_VTX_BASE_VTX_LOC
, 0);
2187 r600_store_context_reg_seq(cb
, R_028028_DB_STENCIL_CLEAR
, 2);
2188 r600_store_value(cb
, 0); /* R_028028_DB_STENCIL_CLEAR */
2189 r600_store_value(cb
, 0x3F800000); /* R_02802C_DB_DEPTH_CLEAR */
2191 r600_store_context_reg_seq(cb
, R_0286DC_SPI_FOG_CNTL
, 3);
2192 r600_store_value(cb
, 0); /* R_0286DC_SPI_FOG_CNTL */
2193 r600_store_value(cb
, 0); /* R_0286E0_SPI_FOG_FUNC_SCALE */
2194 r600_store_value(cb
, 0); /* R_0286E4_SPI_FOG_FUNC_BIAS */
2196 r600_store_context_reg_seq(cb
, R_028D2C_DB_SRESULTS_COMPARE_STATE1
, 2);
2197 r600_store_value(cb
, 0); /* R_028D2C_DB_SRESULTS_COMPARE_STATE1 */
2198 r600_store_value(cb
, 0); /* R_028D30_DB_PRELOAD_CONTROL */
2200 r600_store_context_reg(cb
, R_028D44_DB_ALPHA_TO_MASK
, 0xAA00);
2202 r600_store_context_reg(cb
, R_028820_PA_CL_NANINF_CNTL
, 0);
2203 r600_store_context_reg(cb
, R_028A48_PA_SC_MPASS_PS_CNTL
, 0);
2205 r600_store_context_reg_seq(cb
, R_028C00_PA_SC_LINE_CNTL
, 2);
2206 r600_store_value(cb
, 0x400); /* R_028C00_PA_SC_LINE_CNTL */
2207 r600_store_value(cb
, 0); /* R_028C04_PA_SC_AA_CONFIG */
2209 r600_store_context_reg_seq(cb
, R_028C0C_PA_CL_GB_VERT_CLIP_ADJ
, 6);
2210 r600_store_value(cb
, 0x3F800000); /* R_028C0C_PA_CL_GB_VERT_CLIP_ADJ */
2211 r600_store_value(cb
, 0x3F800000); /* R_028C10_PA_CL_GB_VERT_DISC_ADJ */
2212 r600_store_value(cb
, 0x3F800000); /* R_028C14_PA_CL_GB_HORZ_CLIP_ADJ */
2213 r600_store_value(cb
, 0x3F800000); /* R_028C18_PA_CL_GB_HORZ_DISC_ADJ */
2214 r600_store_value(cb
, 0); /* R_028C1C_PA_SC_AA_SAMPLE_LOCS_MCTX */
2215 r600_store_value(cb
, 0); /* R_028C20_PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX */
2217 r600_store_context_reg_seq(cb
, R_0282D0_PA_SC_VPORT_ZMIN_0
, 2);
2218 r600_store_value(cb
, 0); /* R_0282D0_PA_SC_VPORT_ZMIN_0 */
2219 r600_store_value(cb
, 0x3F800000); /* R_0282D4_PA_SC_VPORT_ZMAX_0 */
2221 r600_store_context_reg(cb
, R_028818_PA_CL_VTE_CNTL
, 0x43F);
2223 r600_store_context_reg_seq(cb
, R_028100_CB_COLOR0_MASK
, 8);
2224 for (i
= 0; i
< 8; i
++) {
2225 r600_store_value(cb
, 0);
2228 r600_store_context_reg(cb
, R_028200_PA_SC_WINDOW_OFFSET
, 0);
2229 r600_store_context_reg(cb
, R_02820C_PA_SC_CLIPRECT_RULE
, 0xFFFF);
2231 if (rctx
->chip_class
>= R700
) {
2232 r600_store_context_reg(cb
, R_028230_PA_SC_EDGERULE
, 0xAAAAAAAA);
2235 r600_store_context_reg_seq(cb
, R_028C30_CB_CLRCMP_CONTROL
, 4);
2236 r600_store_value(cb
, 0x1000000); /* R_028C30_CB_CLRCMP_CONTROL */
2237 r600_store_value(cb
, 0); /* R_028C34_CB_CLRCMP_SRC */
2238 r600_store_value(cb
, 0xFF); /* R_028C38_CB_CLRCMP_DST */
2239 r600_store_value(cb
, 0xFFFFFFFF); /* R_028C3C_CB_CLRCMP_MSK */
2241 r600_store_context_reg(cb
, R_028C48_PA_SC_AA_MASK
, 0xFFFFFFFF);
2243 r600_store_context_reg_seq(cb
, R_028030_PA_SC_SCREEN_SCISSOR_TL
, 2);
2244 r600_store_value(cb
, 0); /* R_028030_PA_SC_SCREEN_SCISSOR_TL */
2245 r600_store_value(cb
, S_028034_BR_X(8192) | S_028034_BR_Y(8192)); /* R_028034_PA_SC_SCREEN_SCISSOR_BR */
2247 r600_store_context_reg_seq(cb
, R_028240_PA_SC_GENERIC_SCISSOR_TL
, 2);
2248 r600_store_value(cb
, 0); /* R_028240_PA_SC_GENERIC_SCISSOR_TL */
2249 r600_store_value(cb
, S_028244_BR_X(8192) | S_028244_BR_Y(8192)); /* R_028244_PA_SC_GENERIC_SCISSOR_BR */
2251 r600_store_context_reg_seq(cb
, R_0288CC_SQ_PGM_CF_OFFSET_PS
, 2);
2252 r600_store_value(cb
, 0); /* R_0288CC_SQ_PGM_CF_OFFSET_PS */
2253 r600_store_value(cb
, 0); /* R_0288D0_SQ_PGM_CF_OFFSET_VS */
2255 r600_store_context_reg(cb
, R_0288A4_SQ_PGM_RESOURCES_FS
, 0);
2256 r600_store_context_reg(cb
, R_0288DC_SQ_PGM_CF_OFFSET_FS
, 0);
2258 if (rctx
->chip_class
== R700
&& rctx
->screen
->has_streamout
)
2259 r600_store_context_reg(cb
, R_028354_SX_SURFACE_SYNC
, S_028354_SURFACE_SYNC_MASK(0xf));
2260 r600_store_context_reg(cb
, R_028800_DB_DEPTH_CONTROL
, 0);
2261 if (rctx
->screen
->has_streamout
) {
2262 r600_store_context_reg(cb
, R_028B28_VGT_STRMOUT_DRAW_OPAQUE_OFFSET
, 0);
2265 r600_store_loop_const(cb
, R_03E200_SQ_LOOP_CONST_0
, 0x1000FFF);
2266 r600_store_loop_const(cb
, R_03E200_SQ_LOOP_CONST_0
+ (32 * 4), 0x1000FFF);
2269 void r600_pipe_shader_ps(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
2271 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
2272 struct r600_pipe_state
*rstate
= &shader
->rstate
;
2273 struct r600_shader
*rshader
= &shader
->shader
;
2274 unsigned i
, exports_ps
, num_cout
, spi_ps_in_control_0
, spi_input_z
, spi_ps_in_control_1
, db_shader_control
;
2275 int pos_index
= -1, face_index
= -1;
2276 unsigned tmp
, sid
, ufi
= 0;
2277 int need_linear
= 0;
2278 unsigned z_export
= 0, stencil_export
= 0;
2282 for (i
= 0; i
< rshader
->ninput
; i
++) {
2283 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_POSITION
)
2285 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_FACE
)
2288 sid
= rshader
->input
[i
].spi_sid
;
2290 tmp
= S_028644_SEMANTIC(sid
);
2292 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_POSITION
||
2293 rshader
->input
[i
].interpolate
== TGSI_INTERPOLATE_CONSTANT
||
2294 (rshader
->input
[i
].interpolate
== TGSI_INTERPOLATE_COLOR
&&
2295 rctx
->rasterizer
&& rctx
->rasterizer
->flatshade
))
2296 tmp
|= S_028644_FLAT_SHADE(1);
2298 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_GENERIC
&&
2299 rctx
->sprite_coord_enable
& (1 << rshader
->input
[i
].sid
)) {
2300 tmp
|= S_028644_PT_SPRITE_TEX(1);
2303 if (rshader
->input
[i
].centroid
)
2304 tmp
|= S_028644_SEL_CENTROID(1);
2306 if (rshader
->input
[i
].interpolate
== TGSI_INTERPOLATE_LINEAR
) {
2308 tmp
|= S_028644_SEL_LINEAR(1);
2311 r600_pipe_state_add_reg(rstate
, R_028644_SPI_PS_INPUT_CNTL_0
+ i
* 4,
2315 db_shader_control
= S_02880C_Z_ORDER(V_02880C_EARLY_Z_THEN_LATE_Z
);
2316 for (i
= 0; i
< rshader
->noutput
; i
++) {
2317 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
)
2319 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_STENCIL
)
2322 db_shader_control
|= S_02880C_Z_EXPORT_ENABLE(z_export
);
2323 db_shader_control
|= S_02880C_STENCIL_REF_EXPORT_ENABLE(stencil_export
);
2324 if (rshader
->uses_kill
)
2325 db_shader_control
|= S_02880C_KILL_ENABLE(1);
2328 for (i
= 0; i
< rshader
->noutput
; i
++) {
2329 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
||
2330 rshader
->output
[i
].name
== TGSI_SEMANTIC_STENCIL
) {
2334 num_cout
= rshader
->nr_ps_color_exports
;
2335 exports_ps
|= S_028854_EXPORT_COLORS(num_cout
);
2337 /* always at least export 1 component per pixel */
2341 shader
->nr_ps_color_outputs
= num_cout
;
2343 spi_ps_in_control_0
= S_0286CC_NUM_INTERP(rshader
->ninput
) |
2344 S_0286CC_PERSP_GRADIENT_ENA(1)|
2345 S_0286CC_LINEAR_GRADIENT_ENA(need_linear
);
2347 if (pos_index
!= -1) {
2348 spi_ps_in_control_0
|= (S_0286CC_POSITION_ENA(1) |
2349 S_0286CC_POSITION_CENTROID(rshader
->input
[pos_index
].centroid
) |
2350 S_0286CC_POSITION_ADDR(rshader
->input
[pos_index
].gpr
) |
2351 S_0286CC_BARYC_SAMPLE_CNTL(1));
2355 spi_ps_in_control_1
= 0;
2356 if (face_index
!= -1) {
2357 spi_ps_in_control_1
|= S_0286D0_FRONT_FACE_ENA(1) |
2358 S_0286D0_FRONT_FACE_ADDR(rshader
->input
[face_index
].gpr
);
2361 /* HW bug in original R600 */
2362 if (rctx
->family
== CHIP_R600
)
2365 r600_pipe_state_add_reg(rstate
, R_0286CC_SPI_PS_IN_CONTROL_0
, spi_ps_in_control_0
);
2366 r600_pipe_state_add_reg(rstate
, R_0286D0_SPI_PS_IN_CONTROL_1
, spi_ps_in_control_1
);
2367 r600_pipe_state_add_reg(rstate
, R_0286D8_SPI_INPUT_Z
, spi_input_z
);
2368 r600_pipe_state_add_reg_bo(rstate
,
2369 R_028840_SQ_PGM_START_PS
,
2370 0, shader
->bo
, RADEON_USAGE_READ
);
2371 r600_pipe_state_add_reg(rstate
,
2372 R_028850_SQ_PGM_RESOURCES_PS
,
2373 S_028850_NUM_GPRS(rshader
->bc
.ngpr
) |
2374 S_028850_STACK_SIZE(rshader
->bc
.nstack
) |
2375 S_028850_UNCACHED_FIRST_INST(ufi
));
2376 r600_pipe_state_add_reg(rstate
,
2377 R_028854_SQ_PGM_EXPORTS_PS
,
2379 /* only set some bits here, the other bits are set in the dsa state */
2380 shader
->db_shader_control
= db_shader_control
;
2381 shader
->ps_depth_export
= z_export
| stencil_export
;
2383 shader
->sprite_coord_enable
= rctx
->sprite_coord_enable
;
2384 if (rctx
->rasterizer
)
2385 shader
->flatshade
= rctx
->rasterizer
->flatshade
;
2388 void r600_pipe_shader_vs(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
2390 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
2391 struct r600_pipe_state
*rstate
= &shader
->rstate
;
2392 struct r600_shader
*rshader
= &shader
->shader
;
2393 unsigned spi_vs_out_id
[10] = {};
2394 unsigned i
, tmp
, nparams
= 0;
2396 /* clear previous register */
2399 for (i
= 0; i
< rshader
->noutput
; i
++) {
2400 if (rshader
->output
[i
].spi_sid
) {
2401 tmp
= rshader
->output
[i
].spi_sid
<< ((nparams
& 3) * 8);
2402 spi_vs_out_id
[nparams
/ 4] |= tmp
;
2407 for (i
= 0; i
< 10; i
++) {
2408 r600_pipe_state_add_reg(rstate
,
2409 R_028614_SPI_VS_OUT_ID_0
+ i
* 4,
2413 /* Certain attributes (position, psize, etc.) don't count as params.
2414 * VS is required to export at least one param and r600_shader_from_tgsi()
2415 * takes care of adding a dummy export.
2420 r600_pipe_state_add_reg(rstate
,
2421 R_0286C4_SPI_VS_OUT_CONFIG
,
2422 S_0286C4_VS_EXPORT_COUNT(nparams
- 1));
2423 r600_pipe_state_add_reg(rstate
,
2424 R_028868_SQ_PGM_RESOURCES_VS
,
2425 S_028868_NUM_GPRS(rshader
->bc
.ngpr
) |
2426 S_028868_STACK_SIZE(rshader
->bc
.nstack
));
2427 r600_pipe_state_add_reg_bo(rstate
,
2428 R_028858_SQ_PGM_START_VS
,
2429 0, shader
->bo
, RADEON_USAGE_READ
);
2431 shader
->pa_cl_vs_out_cntl
=
2432 S_02881C_VS_OUT_CCDIST0_VEC_ENA((rshader
->clip_dist_write
& 0x0F) != 0) |
2433 S_02881C_VS_OUT_CCDIST1_VEC_ENA((rshader
->clip_dist_write
& 0xF0) != 0) |
2434 S_02881C_VS_OUT_MISC_VEC_ENA(rshader
->vs_out_misc_write
) |
2435 S_02881C_USE_VTX_POINT_SIZE(rshader
->vs_out_point_size
);
2438 void r600_fetch_shader(struct pipe_context
*ctx
,
2439 struct r600_vertex_element
*ve
)
2441 struct r600_pipe_state
*rstate
;
2442 struct r600_context
*rctx
= (struct r600_context
*)ctx
;
2444 rstate
= &ve
->rstate
;
2445 rstate
->id
= R600_PIPE_STATE_FETCH_SHADER
;
2447 r600_pipe_state_add_reg_bo(rstate
, R_028894_SQ_PGM_START_FS
,
2449 ve
->fetch_shader
, RADEON_USAGE_READ
);
2452 void *r600_create_db_flush_dsa(struct r600_context
*rctx
)
2454 struct pipe_depth_stencil_alpha_state dsa
;
2455 boolean quirk
= false;
2457 if (rctx
->family
== CHIP_RV610
|| rctx
->family
== CHIP_RV630
||
2458 rctx
->family
== CHIP_RV620
|| rctx
->family
== CHIP_RV635
)
2461 memset(&dsa
, 0, sizeof(dsa
));
2464 dsa
.depth
.enabled
= 1;
2465 dsa
.depth
.func
= PIPE_FUNC_LEQUAL
;
2466 dsa
.stencil
[0].enabled
= 1;
2467 dsa
.stencil
[0].func
= PIPE_FUNC_ALWAYS
;
2468 dsa
.stencil
[0].zpass_op
= PIPE_STENCIL_OP_KEEP
;
2469 dsa
.stencil
[0].zfail_op
= PIPE_STENCIL_OP_INCR
;
2470 dsa
.stencil
[0].writemask
= 0xff;
2473 return rctx
->context
.create_depth_stencil_alpha_state(&rctx
->context
, &dsa
);
2476 void r600_update_dual_export_state(struct r600_context
* rctx
)
2478 unsigned dual_export
= rctx
->export_16bpc
&& rctx
->nr_cbufs
&&
2479 !rctx
->ps_shader
->current
->ps_depth_export
;
2480 unsigned db_shader_control
= rctx
->ps_shader
->current
->db_shader_control
|
2481 S_02880C_DUAL_EXPORT_ENABLE(dual_export
);
2483 if (db_shader_control
!= rctx
->db_shader_control
) {
2484 struct r600_pipe_state rstate
;
2486 rctx
->db_shader_control
= db_shader_control
;
2488 r600_pipe_state_add_reg(&rstate
, R_02880C_DB_SHADER_CONTROL
, db_shader_control
);
2489 r600_context_pipe_state_set(rctx
, &rstate
);