3 #include "radeon_drm.h"
7 #include "r600_blit_shaders.h"
9 #define DI_PT_RECTLIST 0x11
10 #define DI_INDEX_SIZE_16_BIT 0x0
11 #define DI_SRC_SEL_AUTO_INDEX 0x2
15 #define FMT_8_8_8_8 0x1a
17 #define COLOR_5_6_5 0x8
18 #define COLOR_8_8_8_8 0x1a
20 /* emits 21 on rv770+, 23 on r600 */
22 set_render_target(struct radeon_device *rdev, int format,
23 int w, int h, u64 gpu_addr)
32 cb_color_info = ((format << 2) | (1 << 27));
34 slice = ((w * h) / 64) - 1;
36 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
37 radeon_ring_write(rdev, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
38 radeon_ring_write(rdev, gpu_addr >> 8);
40 if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770) {
41 radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_BASE_UPDATE, 0));
42 radeon_ring_write(rdev, 2 << 0);
45 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
46 radeon_ring_write(rdev, (CB_COLOR0_SIZE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
47 radeon_ring_write(rdev, (pitch << 0) | (slice << 10));
49 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
50 radeon_ring_write(rdev, (CB_COLOR0_VIEW - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
51 radeon_ring_write(rdev, 0);
53 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
54 radeon_ring_write(rdev, (CB_COLOR0_INFO - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
55 radeon_ring_write(rdev, cb_color_info);
57 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
58 radeon_ring_write(rdev, (CB_COLOR0_TILE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
59 radeon_ring_write(rdev, 0);
61 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
62 radeon_ring_write(rdev, (CB_COLOR0_FRAG - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
63 radeon_ring_write(rdev, 0);
65 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
66 radeon_ring_write(rdev, (CB_COLOR0_MASK - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
67 radeon_ring_write(rdev, 0);
72 cp_set_surface_sync(struct radeon_device *rdev,
73 u32 sync_type, u32 size,
78 if (size == 0xffffffff)
79 cp_coher_size = 0xffffffff;
81 cp_coher_size = ((size + 255) >> 8);
83 radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_SYNC, 3));
84 radeon_ring_write(rdev, sync_type);
85 radeon_ring_write(rdev, cp_coher_size);
86 radeon_ring_write(rdev, mc_addr >> 8);
87 radeon_ring_write(rdev, 10); /* poll interval */
90 /* emits 21dw + 1 surface sync = 26dw */
92 set_shaders(struct radeon_device *rdev)
97 /* setup shader regs */
98 sq_pgm_resources = (1 << 0);
101 gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
102 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
103 radeon_ring_write(rdev, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
104 radeon_ring_write(rdev, gpu_addr >> 8);
106 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
107 radeon_ring_write(rdev, (SQ_PGM_RESOURCES_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
108 radeon_ring_write(rdev, sq_pgm_resources);
110 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
111 radeon_ring_write(rdev, (SQ_PGM_CF_OFFSET_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
112 radeon_ring_write(rdev, 0);
115 gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.ps_offset;
116 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
117 radeon_ring_write(rdev, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
118 radeon_ring_write(rdev, gpu_addr >> 8);
120 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
121 radeon_ring_write(rdev, (SQ_PGM_RESOURCES_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
122 radeon_ring_write(rdev, sq_pgm_resources | (1 << 28));
124 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
125 radeon_ring_write(rdev, (SQ_PGM_EXPORTS_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
126 radeon_ring_write(rdev, 2);
128 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
129 radeon_ring_write(rdev, (SQ_PGM_CF_OFFSET_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
130 radeon_ring_write(rdev, 0);
132 gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
133 cp_set_surface_sync(rdev, PACKET3_SH_ACTION_ENA, 512, gpu_addr);
136 /* emits 9 + 1 sync (5) = 14*/
138 set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
140 u32 sq_vtx_constant_word2;
142 sq_vtx_constant_word2 = ((upper_32_bits(gpu_addr) & 0xff) | (16 << 8));
144 radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 7));
145 radeon_ring_write(rdev, 0x460);
146 radeon_ring_write(rdev, gpu_addr & 0xffffffff);
147 radeon_ring_write(rdev, 48 - 1);
148 radeon_ring_write(rdev, sq_vtx_constant_word2);
149 radeon_ring_write(rdev, 1 << 0);
150 radeon_ring_write(rdev, 0);
151 radeon_ring_write(rdev, 0);
152 radeon_ring_write(rdev, SQ_TEX_VTX_VALID_BUFFER << 30);
154 if ((rdev->family == CHIP_RV610) ||
155 (rdev->family == CHIP_RV620) ||
156 (rdev->family == CHIP_RS780) ||
157 (rdev->family == CHIP_RS880) ||
158 (rdev->family == CHIP_RV710))
159 cp_set_surface_sync(rdev,
160 PACKET3_TC_ACTION_ENA, 48, gpu_addr);
162 cp_set_surface_sync(rdev,
163 PACKET3_VC_ACTION_ENA, 48, gpu_addr);
168 set_tex_resource(struct radeon_device *rdev,
169 int format, int w, int h, int pitch,
172 uint32_t sq_tex_resource_word0, sq_tex_resource_word1, sq_tex_resource_word4;
177 sq_tex_resource_word0 = (1 << 0);
178 sq_tex_resource_word0 |= ((((pitch >> 3) - 1) << 8) |
181 sq_tex_resource_word1 = (format << 26);
182 sq_tex_resource_word1 |= ((h - 1) << 0);
184 sq_tex_resource_word4 = ((1 << 14) |
190 radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 7));
191 radeon_ring_write(rdev, 0);
192 radeon_ring_write(rdev, sq_tex_resource_word0);
193 radeon_ring_write(rdev, sq_tex_resource_word1);
194 radeon_ring_write(rdev, gpu_addr >> 8);
195 radeon_ring_write(rdev, gpu_addr >> 8);
196 radeon_ring_write(rdev, sq_tex_resource_word4);
197 radeon_ring_write(rdev, 0);
198 radeon_ring_write(rdev, SQ_TEX_VTX_VALID_TEXTURE << 30);
203 set_scissors(struct radeon_device *rdev, int x1, int y1,
206 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
207 radeon_ring_write(rdev, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
208 radeon_ring_write(rdev, (x1 << 0) | (y1 << 16));
209 radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
211 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
212 radeon_ring_write(rdev, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
213 radeon_ring_write(rdev, (x1 << 0) | (y1 << 16) | (1 << 31));
214 radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
216 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
217 radeon_ring_write(rdev, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
218 radeon_ring_write(rdev, (x1 << 0) | (y1 << 16) | (1 << 31));
219 radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
224 draw_auto(struct radeon_device *rdev)
226 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
227 radeon_ring_write(rdev, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
228 radeon_ring_write(rdev, DI_PT_RECTLIST);
230 radeon_ring_write(rdev, PACKET3(PACKET3_INDEX_TYPE, 0));
231 radeon_ring_write(rdev, DI_INDEX_SIZE_16_BIT);
233 radeon_ring_write(rdev, PACKET3(PACKET3_NUM_INSTANCES, 0));
234 radeon_ring_write(rdev, 1);
236 radeon_ring_write(rdev, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1));
237 radeon_ring_write(rdev, 3);
238 radeon_ring_write(rdev, DI_SRC_SEL_AUTO_INDEX);
244 set_default_state(struct radeon_device *rdev)
246 u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2;
247 u32 sq_thread_resource_mgmt, sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2;
248 int num_ps_gprs, num_vs_gprs, num_temp_gprs, num_gs_gprs, num_es_gprs;
249 int num_ps_threads, num_vs_threads, num_gs_threads, num_es_threads;
250 int num_ps_stack_entries, num_vs_stack_entries, num_gs_stack_entries, num_es_stack_entries;
254 switch (rdev->family) {
261 num_ps_threads = 136;
265 num_ps_stack_entries = 128;
266 num_vs_stack_entries = 128;
267 num_gs_stack_entries = 0;
268 num_es_stack_entries = 0;
277 num_ps_threads = 144;
281 num_ps_stack_entries = 40;
282 num_vs_stack_entries = 40;
283 num_gs_stack_entries = 32;
284 num_es_stack_entries = 16;
296 num_ps_threads = 136;
300 num_ps_stack_entries = 40;
301 num_vs_stack_entries = 40;
302 num_gs_stack_entries = 32;
303 num_es_stack_entries = 16;
311 num_ps_threads = 136;
315 num_ps_stack_entries = 40;
316 num_vs_stack_entries = 40;
317 num_gs_stack_entries = 32;
318 num_es_stack_entries = 16;
326 num_ps_threads = 188;
330 num_ps_stack_entries = 256;
331 num_vs_stack_entries = 256;
332 num_gs_stack_entries = 0;
333 num_es_stack_entries = 0;
342 num_ps_threads = 188;
346 num_ps_stack_entries = 128;
347 num_vs_stack_entries = 128;
348 num_gs_stack_entries = 0;
349 num_es_stack_entries = 0;
357 num_ps_threads = 144;
361 num_ps_stack_entries = 128;
362 num_vs_stack_entries = 128;
363 num_gs_stack_entries = 0;
364 num_es_stack_entries = 0;
368 if ((rdev->family == CHIP_RV610) ||
369 (rdev->family == CHIP_RV620) ||
370 (rdev->family == CHIP_RS780) ||
371 (rdev->family == CHIP_RS880) ||
372 (rdev->family == CHIP_RV710))
375 sq_config = VC_ENABLE;
377 sq_config |= (DX9_CONSTS |
378 ALU_INST_PREFER_VECTOR |
384 sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(num_ps_gprs) |
385 NUM_VS_GPRS(num_vs_gprs) |
386 NUM_CLAUSE_TEMP_GPRS(num_temp_gprs));
387 sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(num_gs_gprs) |
388 NUM_ES_GPRS(num_es_gprs));
389 sq_thread_resource_mgmt = (NUM_PS_THREADS(num_ps_threads) |
390 NUM_VS_THREADS(num_vs_threads) |
391 NUM_GS_THREADS(num_gs_threads) |
392 NUM_ES_THREADS(num_es_threads));
393 sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(num_ps_stack_entries) |
394 NUM_VS_STACK_ENTRIES(num_vs_stack_entries));
395 sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(num_gs_stack_entries) |
396 NUM_ES_STACK_ENTRIES(num_es_stack_entries));
398 /* emit an IB pointing at default state */
399 dwords = ALIGN(rdev->r600_blit.state_len, 0x10);
400 gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset;
401 radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
402 radeon_ring_write(rdev, gpu_addr & 0xFFFFFFFC);
403 radeon_ring_write(rdev, upper_32_bits(gpu_addr) & 0xFF);
404 radeon_ring_write(rdev, dwords);
407 radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 6));
408 radeon_ring_write(rdev, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
409 radeon_ring_write(rdev, sq_config);
410 radeon_ring_write(rdev, sq_gpr_resource_mgmt_1);
411 radeon_ring_write(rdev, sq_gpr_resource_mgmt_2);
412 radeon_ring_write(rdev, sq_thread_resource_mgmt);
413 radeon_ring_write(rdev, sq_stack_resource_mgmt_1);
414 radeon_ring_write(rdev, sq_stack_resource_mgmt_2);
417 static inline uint32_t i2f(uint32_t input)
419 u32 result, i, exponent, fraction;
421 if ((input & 0x3fff) == 0)
422 result = 0; /* 0 is a special case */
424 exponent = 140; /* exponent biased by 127; */
425 fraction = (input & 0x3fff) << 10; /* cheat and only
426 handle numbers below 2^^15 */
427 for (i = 0; i < 14; i++) {
428 if (fraction & 0x800000)
431 fraction = fraction << 1; /* keep
432 shifting left until top bit = 1 */
433 exponent = exponent - 1;
436 result = exponent << 23 | (fraction & 0x7fffff); /* mask
437 off top bit; assumed 1 */
442 int r600_blit_init(struct radeon_device *rdev)
448 int num_packet2s = 0;
450 mutex_init(&rdev->r600_blit.mutex);
451 rdev->r600_blit.state_offset = 0;
453 if (rdev->family >= CHIP_RV770)
454 rdev->r600_blit.state_len = r7xx_default_size;
456 rdev->r600_blit.state_len = r6xx_default_size;
458 dwords = rdev->r600_blit.state_len;
459 while (dwords & 0xf) {
460 packet2s[num_packet2s++] = PACKET2(0);
464 obj_size = dwords * 4;
465 obj_size = ALIGN(obj_size, 256);
467 rdev->r600_blit.vs_offset = obj_size;
468 obj_size += r6xx_vs_size * 4;
469 obj_size = ALIGN(obj_size, 256);
471 rdev->r600_blit.ps_offset = obj_size;
472 obj_size += r6xx_ps_size * 4;
473 obj_size = ALIGN(obj_size, 256);
475 r = radeon_bo_create(rdev, NULL, obj_size, true, RADEON_GEM_DOMAIN_VRAM,
476 &rdev->r600_blit.shader_obj);
478 DRM_ERROR("r600 failed to allocate shader\n");
482 DRM_DEBUG("r6xx blit allocated bo %08x vs %08x ps %08x\n",
484 rdev->r600_blit.vs_offset, rdev->r600_blit.ps_offset);
486 r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
487 if (unlikely(r != 0))
489 r = radeon_bo_kmap(rdev->r600_blit.shader_obj, &ptr);
491 DRM_ERROR("failed to map blit object %d\n", r);
494 if (rdev->family >= CHIP_RV770)
495 memcpy_toio(ptr + rdev->r600_blit.state_offset,
496 r7xx_default_state, rdev->r600_blit.state_len * 4);
498 memcpy_toio(ptr + rdev->r600_blit.state_offset,
499 r6xx_default_state, rdev->r600_blit.state_len * 4);
501 memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4),
502 packet2s, num_packet2s * 4);
503 memcpy(ptr + rdev->r600_blit.vs_offset, r6xx_vs, r6xx_vs_size * 4);
504 memcpy(ptr + rdev->r600_blit.ps_offset, r6xx_ps, r6xx_ps_size * 4);
505 radeon_bo_kunmap(rdev->r600_blit.shader_obj);
506 radeon_bo_unreserve(rdev->r600_blit.shader_obj);
510 void r600_blit_fini(struct radeon_device *rdev)
514 if (rdev->r600_blit.shader_obj == NULL)
516 /* If we can't reserve the bo, unref should be enough to destroy
517 * it when it becomes idle.
519 r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
521 radeon_bo_unpin(rdev->r600_blit.shader_obj);
522 radeon_bo_unreserve(rdev->r600_blit.shader_obj);
524 radeon_bo_unref(&rdev->r600_blit.shader_obj);
527 int r600_vb_ib_get(struct radeon_device *rdev)
530 r = radeon_ib_get(rdev, &rdev->r600_blit.vb_ib);
532 DRM_ERROR("failed to get IB for vertex buffer\n");
536 rdev->r600_blit.vb_total = 64*1024;
537 rdev->r600_blit.vb_used = 0;
541 void r600_vb_ib_put(struct radeon_device *rdev)
543 radeon_fence_emit(rdev, rdev->r600_blit.vb_ib->fence);
544 radeon_ib_free(rdev, &rdev->r600_blit.vb_ib);
547 int r600_blit_prepare_copy(struct radeon_device *rdev, int size_bytes)
550 int ring_size, line_size;
552 /* loops of emits 64 + fence emit possible */
553 int dwords_per_loop = 76, num_loops;
555 r = r600_vb_ib_get(rdev);
559 /* set_render_target emits 2 extra dwords on rv6xx */
560 if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770)
561 dwords_per_loop += 2;
563 /* 8 bpp vs 32 bpp for xfer unit */
569 max_size = 8192 * line_size;
571 /* major loops cover the max size transfer */
572 num_loops = ((size_bytes + max_size) / max_size);
573 /* minor loops cover the extra non aligned bits */
574 num_loops += ((size_bytes % line_size) ? 1 : 0);
575 /* calculate number of loops correctly */
576 ring_size = num_loops * dwords_per_loop;
577 /* set default + shaders */
578 ring_size += 40; /* shaders + def state */
579 ring_size += 10; /* fence emit for VB IB */
580 ring_size += 5; /* done copy */
581 ring_size += 10; /* fence emit for done copy */
582 r = radeon_ring_lock(rdev, ring_size);
586 set_default_state(rdev); /* 14 */
587 set_shaders(rdev); /* 26 */
591 void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence *fence)
595 if (rdev->r600_blit.vb_ib)
596 r600_vb_ib_put(rdev);
599 r = radeon_fence_emit(rdev, fence);
601 radeon_ring_unlock_commit(rdev);
604 void r600_kms_blit_copy(struct radeon_device *rdev,
605 u64 src_gpu_addr, u64 dst_gpu_addr,
612 DRM_DEBUG("emitting copy %16llx %16llx %d %d\n", src_gpu_addr, dst_gpu_addr,
613 size_bytes, rdev->r600_blit.vb_used);
614 vb = (u32 *)(rdev->r600_blit.vb_ib->ptr + rdev->r600_blit.vb_used);
615 if ((size_bytes & 3) || (src_gpu_addr & 3) || (dst_gpu_addr & 3)) {
619 int cur_size = size_bytes;
620 int src_x = src_gpu_addr & 255;
621 int dst_x = dst_gpu_addr & 255;
623 src_gpu_addr = src_gpu_addr & ~255;
624 dst_gpu_addr = dst_gpu_addr & ~255;
626 if (!src_x && !dst_x) {
627 h = (cur_size / max_bytes);
633 cur_size = max_bytes;
635 if (cur_size > max_bytes)
636 cur_size = max_bytes;
637 if (cur_size > (max_bytes - dst_x))
638 cur_size = (max_bytes - dst_x);
639 if (cur_size > (max_bytes - src_x))
640 cur_size = (max_bytes - src_x);
643 if ((rdev->r600_blit.vb_used + 48) > rdev->r600_blit.vb_total) {
647 r600_vb_ib_put(rdev);
649 r600_nomm_put_vb(dev);
650 r600_nomm_get_vb(dev);
651 if (!dev_priv->blit_vb)
654 vb = r600_nomm_get_vb_ptr(dev);
668 vb[8] = i2f(dst_x + cur_size);
670 vb[10] = i2f(src_x + cur_size);
674 set_tex_resource(rdev, FMT_8,
675 src_x + cur_size, h, src_x + cur_size,
679 cp_set_surface_sync(rdev,
680 PACKET3_TC_ACTION_ENA, (src_x + cur_size * h), src_gpu_addr);
683 set_render_target(rdev, COLOR_8,
688 set_scissors(rdev, dst_x, 0, dst_x + cur_size, h);
691 vb_gpu_addr = rdev->r600_blit.vb_ib->gpu_addr + rdev->r600_blit.vb_used;
692 set_vtx_resource(rdev, vb_gpu_addr);
698 cp_set_surface_sync(rdev,
699 PACKET3_CB_ACTION_ENA | PACKET3_CB0_DEST_BASE_ENA,
700 cur_size * h, dst_gpu_addr);
703 rdev->r600_blit.vb_used += 12 * 4;
705 src_gpu_addr += cur_size * h;
706 dst_gpu_addr += cur_size * h;
707 size_bytes -= cur_size * h;
710 max_bytes = 8192 * 4;
713 int cur_size = size_bytes;
714 int src_x = (src_gpu_addr & 255);
715 int dst_x = (dst_gpu_addr & 255);
717 src_gpu_addr = src_gpu_addr & ~255;
718 dst_gpu_addr = dst_gpu_addr & ~255;
720 if (!src_x && !dst_x) {
721 h = (cur_size / max_bytes);
727 cur_size = max_bytes;
729 if (cur_size > max_bytes)
730 cur_size = max_bytes;
731 if (cur_size > (max_bytes - dst_x))
732 cur_size = (max_bytes - dst_x);
733 if (cur_size > (max_bytes - src_x))
734 cur_size = (max_bytes - src_x);
737 if ((rdev->r600_blit.vb_used + 48) > rdev->r600_blit.vb_total) {
741 if ((rdev->blit_vb->used + 48) > rdev->blit_vb->total) {
742 r600_nomm_put_vb(dev);
743 r600_nomm_get_vb(dev);
748 vb = r600_nomm_get_vb_ptr(dev);
752 vb[0] = i2f(dst_x / 4);
754 vb[2] = i2f(src_x / 4);
757 vb[4] = i2f(dst_x / 4);
759 vb[6] = i2f(src_x / 4);
762 vb[8] = i2f((dst_x + cur_size) / 4);
764 vb[10] = i2f((src_x + cur_size) / 4);
768 set_tex_resource(rdev, FMT_8_8_8_8,
769 (src_x + cur_size) / 4,
770 h, (src_x + cur_size) / 4,
773 cp_set_surface_sync(rdev,
774 PACKET3_TC_ACTION_ENA, (src_x + cur_size * h), src_gpu_addr);
777 set_render_target(rdev, COLOR_8_8_8_8,
778 (dst_x + cur_size) / 4, h,
782 set_scissors(rdev, (dst_x / 4), 0, (dst_x + cur_size / 4), h);
784 /* Vertex buffer setup 14 */
785 vb_gpu_addr = rdev->r600_blit.vb_ib->gpu_addr + rdev->r600_blit.vb_used;
786 set_vtx_resource(rdev, vb_gpu_addr);
792 cp_set_surface_sync(rdev,
793 PACKET3_CB_ACTION_ENA | PACKET3_CB0_DEST_BASE_ENA,
794 cur_size * h, dst_gpu_addr);
796 /* 78 ring dwords per loop */
798 rdev->r600_blit.vb_used += 12 * 4;
800 src_gpu_addr += cur_size * h;
801 dst_gpu_addr += cur_size * h;
802 size_bytes -= cur_size * h;