1 /*
2 * Copyright 2009 Advanced Micro Devices, Inc.
3 *
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 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
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
13 * Software.
14 *
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 NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
19 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
20 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
21 * DEALINGS IN THE SOFTWARE.
22 *
23 * Authors:
24 * Alex Deucher <alexander.deucher@amd.com>
25 *
26 * ------------------------ This file is DEPRECATED! -------------------------
27 */
28 #include <drm/drmP.h>
29 #include <drm/radeon_drm.h>
30 #include "radeon_drv.h"
31
32 #include "r600_blit_shaders.h"
33
34 /* 23 bits of float fractional data */
35 #define I2F_FRAC_BITS 23
36 #define I2F_MASK ((1 << I2F_FRAC_BITS) - 1)
37
38 /*
39 * Converts unsigned integer into 32-bit IEEE floating point representation.
40 * Will be exact from 0 to 2^24. Above that, we round towards zero
41 * as the fractional bits will not fit in a float. (It would be better to
42 * round towards even as the fpu does, but that is slower.)
43 */
int2float(uint32_t x)44 static __pure uint32_t int2float(uint32_t x)
45 {
46 uint32_t msb, exponent, fraction;
47
48 /* Zero is special */
49 if (!x) return 0;
50
51 /* Get location of the most significant bit */
52 msb = __fls(x);
53
54 /*
55 * Use a rotate instead of a shift because that works both leftwards
56 * and rightwards due to the mod(32) behaviour. This means we don't
57 * need to check to see if we are above 2^24 or not.
58 */
59 fraction = ror32(x, (msb - I2F_FRAC_BITS) & 0x1f) & I2F_MASK;
60 exponent = (127 + msb) << I2F_FRAC_BITS;
61
62 return fraction + exponent;
63 }
64
65 #define DI_PT_RECTLIST 0x11
66 #define DI_INDEX_SIZE_16_BIT 0x0
67 #define DI_SRC_SEL_AUTO_INDEX 0x2
68
69 #define FMT_8 0x1
70 #define FMT_5_6_5 0x8
71 #define FMT_8_8_8_8 0x1a
72 #define COLOR_8 0x1
73 #define COLOR_5_6_5 0x8
74 #define COLOR_8_8_8_8 0x1a
75
76 static void
set_render_target(drm_radeon_private_t * dev_priv,int format,int w,int h,u64 gpu_addr)77 set_render_target(drm_radeon_private_t *dev_priv, int format, int w, int h, u64 gpu_addr)
78 {
79 u32 cb_color_info;
80 int pitch, slice;
81 RING_LOCALS;
82 DRM_DEBUG("\n");
83
84 h = ALIGN(h, 8);
85 if (h < 8)
86 h = 8;
87
88 cb_color_info = ((format << 2) | (1 << 27));
89 pitch = (w / 8) - 1;
90 slice = ((w * h) / 64) - 1;
91
92 if (((dev_priv->flags & RADEON_FAMILY_MASK) > CHIP_R600) &&
93 ((dev_priv->flags & RADEON_FAMILY_MASK) < CHIP_RV770)) {
94 BEGIN_RING(21 + 2);
95 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
96 OUT_RING((R600_CB_COLOR0_BASE - R600_SET_CONTEXT_REG_OFFSET) >> 2);
97 OUT_RING(gpu_addr >> 8);
98 OUT_RING(CP_PACKET3(R600_IT_SURFACE_BASE_UPDATE, 0));
99 OUT_RING(2 << 0);
100 } else {
101 BEGIN_RING(21);
102 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
103 OUT_RING((R600_CB_COLOR0_BASE - R600_SET_CONTEXT_REG_OFFSET) >> 2);
104 OUT_RING(gpu_addr >> 8);
105 }
106
107 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
108 OUT_RING((R600_CB_COLOR0_SIZE - R600_SET_CONTEXT_REG_OFFSET) >> 2);
109 OUT_RING((pitch << 0) | (slice << 10));
110
111 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
112 OUT_RING((R600_CB_COLOR0_VIEW - R600_SET_CONTEXT_REG_OFFSET) >> 2);
113 OUT_RING(0);
114
115 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
116 OUT_RING((R600_CB_COLOR0_INFO - R600_SET_CONTEXT_REG_OFFSET) >> 2);
117 OUT_RING(cb_color_info);
118
119 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
120 OUT_RING((R600_CB_COLOR0_TILE - R600_SET_CONTEXT_REG_OFFSET) >> 2);
121 OUT_RING(0);
122
123 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
124 OUT_RING((R600_CB_COLOR0_FRAG - R600_SET_CONTEXT_REG_OFFSET) >> 2);
125 OUT_RING(0);
126
127 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
128 OUT_RING((R600_CB_COLOR0_MASK - R600_SET_CONTEXT_REG_OFFSET) >> 2);
129 OUT_RING(0);
130
131 ADVANCE_RING();
132 }
133
134 static void
cp_set_surface_sync(drm_radeon_private_t * dev_priv,u32 sync_type,u32 size,u64 mc_addr)135 cp_set_surface_sync(drm_radeon_private_t *dev_priv,
136 u32 sync_type, u32 size, u64 mc_addr)
137 {
138 u32 cp_coher_size;
139 RING_LOCALS;
140 DRM_DEBUG("\n");
141
142 if (size == 0xffffffff)
143 cp_coher_size = 0xffffffff;
144 else
145 cp_coher_size = ((size + 255) >> 8);
146
147 BEGIN_RING(5);
148 OUT_RING(CP_PACKET3(R600_IT_SURFACE_SYNC, 3));
149 OUT_RING(sync_type);
150 OUT_RING(cp_coher_size);
151 OUT_RING((mc_addr >> 8));
152 OUT_RING(10); /* poll interval */
153 ADVANCE_RING();
154 }
155
156 static void
set_shaders(struct drm_device * dev)157 set_shaders(struct drm_device *dev)
158 {
159 drm_radeon_private_t *dev_priv = dev->dev_private;
160 u64 gpu_addr;
161 int i;
162 u32 *vs, *ps;
163 uint32_t sq_pgm_resources;
164 RING_LOCALS;
165 DRM_DEBUG("\n");
166
167 /* load shaders */
168 vs = (u32 *) ((char *)dev->agp_buffer_map->handle + dev_priv->blit_vb->offset);
169 ps = (u32 *) ((char *)dev->agp_buffer_map->handle + dev_priv->blit_vb->offset + 256);
170
171 for (i = 0; i < r6xx_vs_size; i++)
172 vs[i] = cpu_to_le32(r6xx_vs[i]);
173 for (i = 0; i < r6xx_ps_size; i++)
174 ps[i] = cpu_to_le32(r6xx_ps[i]);
175
176 dev_priv->blit_vb->used = 512;
177
178 gpu_addr = dev_priv->gart_buffers_offset + dev_priv->blit_vb->offset;
179
180 /* setup shader regs */
181 sq_pgm_resources = (1 << 0);
182
183 BEGIN_RING(9 + 12);
184 /* VS */
185 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
186 OUT_RING((R600_SQ_PGM_START_VS - R600_SET_CONTEXT_REG_OFFSET) >> 2);
187 OUT_RING(gpu_addr >> 8);
188
189 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
190 OUT_RING((R600_SQ_PGM_RESOURCES_VS - R600_SET_CONTEXT_REG_OFFSET) >> 2);
191 OUT_RING(sq_pgm_resources);
192
193 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
194 OUT_RING((R600_SQ_PGM_CF_OFFSET_VS - R600_SET_CONTEXT_REG_OFFSET) >> 2);
195 OUT_RING(0);
196
197 /* PS */
198 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
199 OUT_RING((R600_SQ_PGM_START_PS - R600_SET_CONTEXT_REG_OFFSET) >> 2);
200 OUT_RING((gpu_addr + 256) >> 8);
201
202 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
203 OUT_RING((R600_SQ_PGM_RESOURCES_PS - R600_SET_CONTEXT_REG_OFFSET) >> 2);
204 OUT_RING(sq_pgm_resources | (1 << 28));
205
206 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
207 OUT_RING((R600_SQ_PGM_EXPORTS_PS - R600_SET_CONTEXT_REG_OFFSET) >> 2);
208 OUT_RING(2);
209
210 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1));
211 OUT_RING((R600_SQ_PGM_CF_OFFSET_PS - R600_SET_CONTEXT_REG_OFFSET) >> 2);
212 OUT_RING(0);
213 ADVANCE_RING();
214
215 cp_set_surface_sync(dev_priv,
216 R600_SH_ACTION_ENA, 512, gpu_addr);
217 }
218
219 static void
set_vtx_resource(drm_radeon_private_t * dev_priv,u64 gpu_addr)220 set_vtx_resource(drm_radeon_private_t *dev_priv, u64 gpu_addr)
221 {
222 uint32_t sq_vtx_constant_word2;
223 RING_LOCALS;
224 DRM_DEBUG("\n");
225
226 sq_vtx_constant_word2 = (((gpu_addr >> 32) & 0xff) | (16 << 8));
227 #ifdef __BIG_ENDIAN
228 sq_vtx_constant_word2 |= (2 << 30);
229 #endif
230
231 BEGIN_RING(9);
232 OUT_RING(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
233 OUT_RING(0x460);
234 OUT_RING(gpu_addr & 0xffffffff);
235 OUT_RING(48 - 1);
236 OUT_RING(sq_vtx_constant_word2);
237 OUT_RING(1 << 0);
238 OUT_RING(0);
239 OUT_RING(0);
240 OUT_RING(R600_SQ_TEX_VTX_VALID_BUFFER << 30);
241 ADVANCE_RING();
242
243 if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV610) ||
244 ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV620) ||
245 ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS780) ||
246 ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS880) ||
247 ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV710))
248 cp_set_surface_sync(dev_priv,
249 R600_TC_ACTION_ENA, 48, gpu_addr);
250 else
251 cp_set_surface_sync(dev_priv,
252 R600_VC_ACTION_ENA, 48, gpu_addr);
253 }
254
255 static void
set_tex_resource(drm_radeon_private_t * dev_priv,int format,int w,int h,int pitch,u64 gpu_addr)256 set_tex_resource(drm_radeon_private_t *dev_priv,
257 int format, int w, int h, int pitch, u64 gpu_addr)
258 {
259 uint32_t sq_tex_resource_word0, sq_tex_resource_word1, sq_tex_resource_word4;
260 RING_LOCALS;
261 DRM_DEBUG("\n");
262
263 if (h < 1)
264 h = 1;
265
266 sq_tex_resource_word0 = (1 << 0);
267 sq_tex_resource_word0 |= ((((pitch >> 3) - 1) << 8) |
268 ((w - 1) << 19));
269
270 sq_tex_resource_word1 = (format << 26);
271 sq_tex_resource_word1 |= ((h - 1) << 0);
272
273 sq_tex_resource_word4 = ((1 << 14) |
274 (0 << 16) |
275 (1 << 19) |
276 (2 << 22) |
277 (3 << 25));
278
279 BEGIN_RING(9);
280 OUT_RING(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
281 OUT_RING(0);
282 OUT_RING(sq_tex_resource_word0);
283 OUT_RING(sq_tex_resource_word1);
284 OUT_RING(gpu_addr >> 8);
285 OUT_RING(gpu_addr >> 8);
286 OUT_RING(sq_tex_resource_word4);
287 OUT_RING(0);
288 OUT_RING(R600_SQ_TEX_VTX_VALID_TEXTURE << 30);
289 ADVANCE_RING();
290
291 }
292
293 static void
set_scissors(drm_radeon_private_t * dev_priv,int x1,int y1,int x2,int y2)294 set_scissors(drm_radeon_private_t *dev_priv, int x1, int y1, int x2, int y2)
295 {
296 RING_LOCALS;
297 DRM_DEBUG("\n");
298
299 BEGIN_RING(12);
300 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 2));
301 OUT_RING((R600_PA_SC_SCREEN_SCISSOR_TL - R600_SET_CONTEXT_REG_OFFSET) >> 2);
302 OUT_RING((x1 << 0) | (y1 << 16));
303 OUT_RING((x2 << 0) | (y2 << 16));
304
305 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 2));
306 OUT_RING((R600_PA_SC_GENERIC_SCISSOR_TL - R600_SET_CONTEXT_REG_OFFSET) >> 2);
307 OUT_RING((x1 << 0) | (y1 << 16) | (1 << 31));
308 OUT_RING((x2 << 0) | (y2 << 16));
309
310 OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 2));
311 OUT_RING((R600_PA_SC_WINDOW_SCISSOR_TL - R600_SET_CONTEXT_REG_OFFSET) >> 2);
312 OUT_RING((x1 << 0) | (y1 << 16) | (1 << 31));
313 OUT_RING((x2 << 0) | (y2 << 16));
314 ADVANCE_RING();
315 }
316
317 static void
draw_auto(drm_radeon_private_t * dev_priv)318 draw_auto(drm_radeon_private_t *dev_priv)
319 {
320 RING_LOCALS;
321 DRM_DEBUG("\n");
322
323 BEGIN_RING(10);
324 OUT_RING(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
325 OUT_RING((R600_VGT_PRIMITIVE_TYPE - R600_SET_CONFIG_REG_OFFSET) >> 2);
326 OUT_RING(DI_PT_RECTLIST);
327
328 OUT_RING(CP_PACKET3(R600_IT_INDEX_TYPE, 0));
329 #ifdef __BIG_ENDIAN
330 OUT_RING((2 << 2) | DI_INDEX_SIZE_16_BIT);
331 #else
332 OUT_RING(DI_INDEX_SIZE_16_BIT);
333 #endif
334
335 OUT_RING(CP_PACKET3(R600_IT_NUM_INSTANCES, 0));
336 OUT_RING(1);
337
338 OUT_RING(CP_PACKET3(R600_IT_DRAW_INDEX_AUTO, 1));
339 OUT_RING(3);
340 OUT_RING(DI_SRC_SEL_AUTO_INDEX);
341
342 ADVANCE_RING();
343 COMMIT_RING();
344 }
345
346 static void
set_default_state(drm_radeon_private_t * dev_priv)347 set_default_state(drm_radeon_private_t *dev_priv)
348 {
349 int i;
350 u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2;
351 u32 sq_thread_resource_mgmt, sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2;
352 int num_ps_gprs, num_vs_gprs, num_temp_gprs, num_gs_gprs, num_es_gprs;
353 int num_ps_threads, num_vs_threads, num_gs_threads, num_es_threads;
354 int num_ps_stack_entries, num_vs_stack_entries, num_gs_stack_entries, num_es_stack_entries;
355 RING_LOCALS;
356
357 switch ((dev_priv->flags & RADEON_FAMILY_MASK)) {
358 case CHIP_R600:
359 num_ps_gprs = 192;
360 num_vs_gprs = 56;
361 num_temp_gprs = 4;
362 num_gs_gprs = 0;
363 num_es_gprs = 0;
364 num_ps_threads = 136;
365 num_vs_threads = 48;
366 num_gs_threads = 4;
367 num_es_threads = 4;
368 num_ps_stack_entries = 128;
369 num_vs_stack_entries = 128;
370 num_gs_stack_entries = 0;
371 num_es_stack_entries = 0;
372 break;
373 case CHIP_RV630:
374 case CHIP_RV635:
375 num_ps_gprs = 84;
376 num_vs_gprs = 36;
377 num_temp_gprs = 4;
378 num_gs_gprs = 0;
379 num_es_gprs = 0;
380 num_ps_threads = 144;
381 num_vs_threads = 40;
382 num_gs_threads = 4;
383 num_es_threads = 4;
384 num_ps_stack_entries = 40;
385 num_vs_stack_entries = 40;
386 num_gs_stack_entries = 32;
387 num_es_stack_entries = 16;
388 break;
389 case CHIP_RV610:
390 case CHIP_RV620:
391 case CHIP_RS780:
392 case CHIP_RS880:
393 default:
394 num_ps_gprs = 84;
395 num_vs_gprs = 36;
396 num_temp_gprs = 4;
397 num_gs_gprs = 0;
398 num_es_gprs = 0;
399 num_ps_threads = 136;
400 num_vs_threads = 48;
401 num_gs_threads = 4;
402 num_es_threads = 4;
403 num_ps_stack_entries = 40;
404 num_vs_stack_entries = 40;
405 num_gs_stack_entries = 32;
406 num_es_stack_entries = 16;
407 break;
408 case CHIP_RV670:
409 num_ps_gprs = 144;
410 num_vs_gprs = 40;
411 num_temp_gprs = 4;
412 num_gs_gprs = 0;
413 num_es_gprs = 0;
414 num_ps_threads = 136;
415 num_vs_threads = 48;
416 num_gs_threads = 4;
417 num_es_threads = 4;
418 num_ps_stack_entries = 40;
419 num_vs_stack_entries = 40;
420 num_gs_stack_entries = 32;
421 num_es_stack_entries = 16;
422 break;
423 case CHIP_RV770:
424 num_ps_gprs = 192;
425 num_vs_gprs = 56;
426 num_temp_gprs = 4;
427 num_gs_gprs = 0;
428 num_es_gprs = 0;
429 num_ps_threads = 188;
430 num_vs_threads = 60;
431 num_gs_threads = 0;
432 num_es_threads = 0;
433 num_ps_stack_entries = 256;
434 num_vs_stack_entries = 256;
435 num_gs_stack_entries = 0;
436 num_es_stack_entries = 0;
437 break;
438 case CHIP_RV730:
439 case CHIP_RV740:
440 num_ps_gprs = 84;
441 num_vs_gprs = 36;
442 num_temp_gprs = 4;
443 num_gs_gprs = 0;
444 num_es_gprs = 0;
445 num_ps_threads = 188;
446 num_vs_threads = 60;
447 num_gs_threads = 0;
448 num_es_threads = 0;
449 num_ps_stack_entries = 128;
450 num_vs_stack_entries = 128;
451 num_gs_stack_entries = 0;
452 num_es_stack_entries = 0;
453 break;
454 case CHIP_RV710:
455 num_ps_gprs = 192;
456 num_vs_gprs = 56;
457 num_temp_gprs = 4;
458 num_gs_gprs = 0;
459 num_es_gprs = 0;
460 num_ps_threads = 144;
461 num_vs_threads = 48;
462 num_gs_threads = 0;
463 num_es_threads = 0;
464 num_ps_stack_entries = 128;
465 num_vs_stack_entries = 128;
466 num_gs_stack_entries = 0;
467 num_es_stack_entries = 0;
468 break;
469 }
470
471 if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV610) ||
472 ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV620) ||
473 ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS780) ||
474 ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS880) ||
475 ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV710))
476 sq_config = 0;
477 else
478 sq_config = R600_VC_ENABLE;
479
480 sq_config |= (R600_DX9_CONSTS |
481 R600_ALU_INST_PREFER_VECTOR |
482 R600_PS_PRIO(0) |
483 R600_VS_PRIO(1) |
484 R600_GS_PRIO(2) |
485 R600_ES_PRIO(3));
486
487 sq_gpr_resource_mgmt_1 = (R600_NUM_PS_GPRS(num_ps_gprs) |
488 R600_NUM_VS_GPRS(num_vs_gprs) |
489 R600_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs));
490 sq_gpr_resource_mgmt_2 = (R600_NUM_GS_GPRS(num_gs_gprs) |
491 R600_NUM_ES_GPRS(num_es_gprs));
492 sq_thread_resource_mgmt = (R600_NUM_PS_THREADS(num_ps_threads) |
493 R600_NUM_VS_THREADS(num_vs_threads) |
494 R600_NUM_GS_THREADS(num_gs_threads) |
495 R600_NUM_ES_THREADS(num_es_threads));
496 sq_stack_resource_mgmt_1 = (R600_NUM_PS_STACK_ENTRIES(num_ps_stack_entries) |
497 R600_NUM_VS_STACK_ENTRIES(num_vs_stack_entries));
498 sq_stack_resource_mgmt_2 = (R600_NUM_GS_STACK_ENTRIES(num_gs_stack_entries) |
499 R600_NUM_ES_STACK_ENTRIES(num_es_stack_entries));
500
501 if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV770) {
502 BEGIN_RING(r7xx_default_size + 10);
503 for (i = 0; i < r7xx_default_size; i++)
504 OUT_RING(r7xx_default_state[i]);
505 } else {
506 BEGIN_RING(r6xx_default_size + 10);
507 for (i = 0; i < r6xx_default_size; i++)
508 OUT_RING(r6xx_default_state[i]);
509 }
510 OUT_RING(CP_PACKET3(R600_IT_EVENT_WRITE, 0));
511 OUT_RING(R600_CACHE_FLUSH_AND_INV_EVENT);
512 /* SQ config */
513 OUT_RING(CP_PACKET3(R600_IT_SET_CONFIG_REG, 6));
514 OUT_RING((R600_SQ_CONFIG - R600_SET_CONFIG_REG_OFFSET) >> 2);
515 OUT_RING(sq_config);
516 OUT_RING(sq_gpr_resource_mgmt_1);
517 OUT_RING(sq_gpr_resource_mgmt_2);
518 OUT_RING(sq_thread_resource_mgmt);
519 OUT_RING(sq_stack_resource_mgmt_1);
520 OUT_RING(sq_stack_resource_mgmt_2);
521 ADVANCE_RING();
522 }
523
r600_nomm_get_vb(struct drm_device * dev)524 static int r600_nomm_get_vb(struct drm_device *dev)
525 {
526 drm_radeon_private_t *dev_priv = dev->dev_private;
527 dev_priv->blit_vb = radeon_freelist_get(dev);
528 if (!dev_priv->blit_vb) {
529 DRM_ERROR("Unable to allocate vertex buffer for blit\n");
530 return -EAGAIN;
531 }
532 return 0;
533 }
534
r600_nomm_put_vb(struct drm_device * dev)535 static void r600_nomm_put_vb(struct drm_device *dev)
536 {
537 drm_radeon_private_t *dev_priv = dev->dev_private;
538
539 dev_priv->blit_vb->used = 0;
540 radeon_cp_discard_buffer(dev, dev_priv->blit_vb->file_priv->master, dev_priv->blit_vb);
541 }
542
r600_nomm_get_vb_ptr(struct drm_device * dev)543 static void *r600_nomm_get_vb_ptr(struct drm_device *dev)
544 {
545 drm_radeon_private_t *dev_priv = dev->dev_private;
546 return (((char *)dev->agp_buffer_map->handle +
547 dev_priv->blit_vb->offset + dev_priv->blit_vb->used));
548 }
549
550 int
r600_prepare_blit_copy(struct drm_device * dev,struct drm_file * file_priv)551 r600_prepare_blit_copy(struct drm_device *dev, struct drm_file *file_priv)
552 {
553 drm_radeon_private_t *dev_priv = dev->dev_private;
554 int ret;
555 DRM_DEBUG("\n");
556
557 ret = r600_nomm_get_vb(dev);
558 if (ret)
559 return ret;
560
561 dev_priv->blit_vb->file_priv = file_priv;
562
563 set_default_state(dev_priv);
564 set_shaders(dev);
565
566 return 0;
567 }
568
569
570 void
r600_done_blit_copy(struct drm_device * dev)571 r600_done_blit_copy(struct drm_device *dev)
572 {
573 drm_radeon_private_t *dev_priv = dev->dev_private;
574 RING_LOCALS;
575 DRM_DEBUG("\n");
576
577 BEGIN_RING(5);
578 OUT_RING(CP_PACKET3(R600_IT_EVENT_WRITE, 0));
579 OUT_RING(R600_CACHE_FLUSH_AND_INV_EVENT);
580 /* wait for 3D idle clean */
581 OUT_RING(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
582 OUT_RING((R600_WAIT_UNTIL - R600_SET_CONFIG_REG_OFFSET) >> 2);
583 OUT_RING(RADEON_WAIT_3D_IDLE | RADEON_WAIT_3D_IDLECLEAN);
584
585 ADVANCE_RING();
586 COMMIT_RING();
587
588 r600_nomm_put_vb(dev);
589 }
590
591 void
r600_blit_copy(struct drm_device * dev,uint64_t src_gpu_addr,uint64_t dst_gpu_addr,int size_bytes)592 r600_blit_copy(struct drm_device *dev,
593 uint64_t src_gpu_addr, uint64_t dst_gpu_addr,
594 int size_bytes)
595 {
596 drm_radeon_private_t *dev_priv = dev->dev_private;
597 int max_bytes;
598 u64 vb_addr;
599 u32 *vb;
600
601 vb = r600_nomm_get_vb_ptr(dev);
602
603 if ((size_bytes & 3) || (src_gpu_addr & 3) || (dst_gpu_addr & 3)) {
604 max_bytes = 8192;
605
606 while (size_bytes) {
607 int cur_size = size_bytes;
608 int src_x = src_gpu_addr & 255;
609 int dst_x = dst_gpu_addr & 255;
610 int h = 1;
611 src_gpu_addr = src_gpu_addr & ~255;
612 dst_gpu_addr = dst_gpu_addr & ~255;
613
614 if (!src_x && !dst_x) {
615 h = (cur_size / max_bytes);
616 if (h > 8192)
617 h = 8192;
618 if (h == 0)
619 h = 1;
620 else
621 cur_size = max_bytes;
622 } else {
623 if (cur_size > max_bytes)
624 cur_size = max_bytes;
625 if (cur_size > (max_bytes - dst_x))
626 cur_size = (max_bytes - dst_x);
627 if (cur_size > (max_bytes - src_x))
628 cur_size = (max_bytes - src_x);
629 }
630
631 if ((dev_priv->blit_vb->used + 48) > dev_priv->blit_vb->total) {
632
633 r600_nomm_put_vb(dev);
634 r600_nomm_get_vb(dev);
635 if (!dev_priv->blit_vb)
636 return;
637 set_shaders(dev);
638 vb = r600_nomm_get_vb_ptr(dev);
639 }
640
641 vb[0] = int2float(dst_x);
642 vb[1] = 0;
643 vb[2] = int2float(src_x);
644 vb[3] = 0;
645
646 vb[4] = int2float(dst_x);
647 vb[5] = int2float(h);
648 vb[6] = int2float(src_x);
649 vb[7] = int2float(h);
650
651 vb[8] = int2float(dst_x + cur_size);
652 vb[9] = int2float(h);
653 vb[10] = int2float(src_x + cur_size);
654 vb[11] = int2float(h);
655
656 /* src */
657 set_tex_resource(dev_priv, FMT_8,
658 src_x + cur_size, h, src_x + cur_size,
659 src_gpu_addr);
660
661 cp_set_surface_sync(dev_priv,
662 R600_TC_ACTION_ENA, (src_x + cur_size * h), src_gpu_addr);
663
664 /* dst */
665 set_render_target(dev_priv, COLOR_8,
666 dst_x + cur_size, h,
667 dst_gpu_addr);
668
669 /* scissors */
670 set_scissors(dev_priv, dst_x, 0, dst_x + cur_size, h);
671
672 /* Vertex buffer setup */
673 vb_addr = dev_priv->gart_buffers_offset +
674 dev_priv->blit_vb->offset +
675 dev_priv->blit_vb->used;
676 set_vtx_resource(dev_priv, vb_addr);
677
678 /* draw */
679 draw_auto(dev_priv);
680
681 cp_set_surface_sync(dev_priv,
682 R600_CB_ACTION_ENA | R600_CB0_DEST_BASE_ENA,
683 cur_size * h, dst_gpu_addr);
684
685 vb += 12;
686 dev_priv->blit_vb->used += 12 * 4;
687
688 src_gpu_addr += cur_size * h;
689 dst_gpu_addr += cur_size * h;
690 size_bytes -= cur_size * h;
691 }
692 } else {
693 max_bytes = 8192 * 4;
694
695 while (size_bytes) {
696 int cur_size = size_bytes;
697 int src_x = (src_gpu_addr & 255);
698 int dst_x = (dst_gpu_addr & 255);
699 int h = 1;
700 src_gpu_addr = src_gpu_addr & ~255;
701 dst_gpu_addr = dst_gpu_addr & ~255;
702
703 if (!src_x && !dst_x) {
704 h = (cur_size / max_bytes);
705 if (h > 8192)
706 h = 8192;
707 if (h == 0)
708 h = 1;
709 else
710 cur_size = max_bytes;
711 } else {
712 if (cur_size > max_bytes)
713 cur_size = max_bytes;
714 if (cur_size > (max_bytes - dst_x))
715 cur_size = (max_bytes - dst_x);
716 if (cur_size > (max_bytes - src_x))
717 cur_size = (max_bytes - src_x);
718 }
719
720 if ((dev_priv->blit_vb->used + 48) > dev_priv->blit_vb->total) {
721 r600_nomm_put_vb(dev);
722 r600_nomm_get_vb(dev);
723 if (!dev_priv->blit_vb)
724 return;
725
726 set_shaders(dev);
727 vb = r600_nomm_get_vb_ptr(dev);
728 }
729
730 vb[0] = int2float(dst_x / 4);
731 vb[1] = 0;
732 vb[2] = int2float(src_x / 4);
733 vb[3] = 0;
734
735 vb[4] = int2float(dst_x / 4);
736 vb[5] = int2float(h);
737 vb[6] = int2float(src_x / 4);
738 vb[7] = int2float(h);
739
740 vb[8] = int2float((dst_x + cur_size) / 4);
741 vb[9] = int2float(h);
742 vb[10] = int2float((src_x + cur_size) / 4);
743 vb[11] = int2float(h);
744
745 /* src */
746 set_tex_resource(dev_priv, FMT_8_8_8_8,
747 (src_x + cur_size) / 4,
748 h, (src_x + cur_size) / 4,
749 src_gpu_addr);
750
751 cp_set_surface_sync(dev_priv,
752 R600_TC_ACTION_ENA, (src_x + cur_size * h), src_gpu_addr);
753
754 /* dst */
755 set_render_target(dev_priv, COLOR_8_8_8_8,
756 (dst_x + cur_size) / 4, h,
757 dst_gpu_addr);
758
759 /* scissors */
760 set_scissors(dev_priv, (dst_x / 4), 0, (dst_x + cur_size / 4), h);
761
762 /* Vertex buffer setup */
763 vb_addr = dev_priv->gart_buffers_offset +
764 dev_priv->blit_vb->offset +
765 dev_priv->blit_vb->used;
766 set_vtx_resource(dev_priv, vb_addr);
767
768 /* draw */
769 draw_auto(dev_priv);
770
771 cp_set_surface_sync(dev_priv,
772 R600_CB_ACTION_ENA | R600_CB0_DEST_BASE_ENA,
773 cur_size * h, dst_gpu_addr);
774
775 vb += 12;
776 dev_priv->blit_vb->used += 12 * 4;
777
778 src_gpu_addr += cur_size * h;
779 dst_gpu_addr += cur_size * h;
780 size_bytes -= cur_size * h;
781 }
782 }
783 }
784
785 void
r600_blit_swap(struct drm_device * dev,uint64_t src_gpu_addr,uint64_t dst_gpu_addr,int sx,int sy,int dx,int dy,int w,int h,int src_pitch,int dst_pitch,int cpp)786 r600_blit_swap(struct drm_device *dev,
787 uint64_t src_gpu_addr, uint64_t dst_gpu_addr,
788 int sx, int sy, int dx, int dy,
789 int w, int h, int src_pitch, int dst_pitch, int cpp)
790 {
791 drm_radeon_private_t *dev_priv = dev->dev_private;
792 int cb_format, tex_format;
793 int sx2, sy2, dx2, dy2;
794 u64 vb_addr;
795 u32 *vb;
796
797 if ((dev_priv->blit_vb->used + 48) > dev_priv->blit_vb->total) {
798
799 r600_nomm_put_vb(dev);
800 r600_nomm_get_vb(dev);
801 if (!dev_priv->blit_vb)
802 return;
803
804 set_shaders(dev);
805 }
806 vb = r600_nomm_get_vb_ptr(dev);
807
808 sx2 = sx + w;
809 sy2 = sy + h;
810 dx2 = dx + w;
811 dy2 = dy + h;
812
813 vb[0] = int2float(dx);
814 vb[1] = int2float(dy);
815 vb[2] = int2float(sx);
816 vb[3] = int2float(sy);
817
818 vb[4] = int2float(dx);
819 vb[5] = int2float(dy2);
820 vb[6] = int2float(sx);
821 vb[7] = int2float(sy2);
822
823 vb[8] = int2float(dx2);
824 vb[9] = int2float(dy2);
825 vb[10] = int2float(sx2);
826 vb[11] = int2float(sy2);
827
828 switch(cpp) {
829 case 4:
830 cb_format = COLOR_8_8_8_8;
831 tex_format = FMT_8_8_8_8;
832 break;
833 case 2:
834 cb_format = COLOR_5_6_5;
835 tex_format = FMT_5_6_5;
836 break;
837 default:
838 cb_format = COLOR_8;
839 tex_format = FMT_8;
840 break;
841 }
842
843 /* src */
844 set_tex_resource(dev_priv, tex_format,
845 src_pitch / cpp,
846 sy2, src_pitch / cpp,
847 src_gpu_addr);
848
849 cp_set_surface_sync(dev_priv,
850 R600_TC_ACTION_ENA, src_pitch * sy2, src_gpu_addr);
851
852 /* dst */
853 set_render_target(dev_priv, cb_format,
854 dst_pitch / cpp, dy2,
855 dst_gpu_addr);
856
857 /* scissors */
858 set_scissors(dev_priv, dx, dy, dx2, dy2);
859
860 /* Vertex buffer setup */
861 vb_addr = dev_priv->gart_buffers_offset +
862 dev_priv->blit_vb->offset +
863 dev_priv->blit_vb->used;
864 set_vtx_resource(dev_priv, vb_addr);
865
866 /* draw */
867 draw_auto(dev_priv);
868
869 cp_set_surface_sync(dev_priv,
870 R600_CB_ACTION_ENA | R600_CB0_DEST_BASE_ENA,
871 dst_pitch * dy2, dst_gpu_addr);
872
873 dev_priv->blit_vb->used += 12 * 4;
874 }
875