2 * Copyright 2010 Advanced Micro Devices, Inc.
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:
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 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.
24 * Alex Deucher <alexander.deucher@amd.com>
29 #include "radeon_drm.h"
32 #include "evergreend.h"
33 #include "evergreen_blit_shaders.h"
34 #include "cayman_blit_shaders.h"
36 #define DI_PT_RECTLIST 0x11
37 #define DI_INDEX_SIZE_16_BIT 0x0
38 #define DI_SRC_SEL_AUTO_INDEX 0x2
42 #define FMT_8_8_8_8 0x1a
44 #define COLOR_5_6_5 0x8
45 #define COLOR_8_8_8_8 0x1a
49 set_render_target(struct radeon_device
*rdev
, int format
,
50 int w
, int h
, u64 gpu_addr
)
59 cb_color_info
= ((format
<< 2) | (1 << 24) | (1 << 8));
61 slice
= ((w
* h
) / 64) - 1;
63 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONTEXT_REG
, 15));
64 radeon_ring_write(rdev
, (CB_COLOR0_BASE
- PACKET3_SET_CONTEXT_REG_START
) >> 2);
65 radeon_ring_write(rdev
, gpu_addr
>> 8);
66 radeon_ring_write(rdev
, pitch
);
67 radeon_ring_write(rdev
, slice
);
68 radeon_ring_write(rdev
, 0);
69 radeon_ring_write(rdev
, cb_color_info
);
70 radeon_ring_write(rdev
, (1 << 4));
71 radeon_ring_write(rdev
, (w
- 1) | ((h
- 1) << 16));
72 radeon_ring_write(rdev
, 0);
73 radeon_ring_write(rdev
, 0);
74 radeon_ring_write(rdev
, 0);
75 radeon_ring_write(rdev
, 0);
76 radeon_ring_write(rdev
, 0);
77 radeon_ring_write(rdev
, 0);
78 radeon_ring_write(rdev
, 0);
79 radeon_ring_write(rdev
, 0);
84 cp_set_surface_sync(struct radeon_device
*rdev
,
85 u32 sync_type
, u32 size
,
90 if (size
== 0xffffffff)
91 cp_coher_size
= 0xffffffff;
93 cp_coher_size
= ((size
+ 255) >> 8);
95 radeon_ring_write(rdev
, PACKET3(PACKET3_SURFACE_SYNC
, 3));
96 radeon_ring_write(rdev
, sync_type
);
97 radeon_ring_write(rdev
, cp_coher_size
);
98 radeon_ring_write(rdev
, mc_addr
>> 8);
99 radeon_ring_write(rdev
, 10); /* poll interval */
102 /* emits 11dw + 1 surface sync = 16dw */
104 set_shaders(struct radeon_device
*rdev
)
109 gpu_addr
= rdev
->r600_blit
.shader_gpu_addr
+ rdev
->r600_blit
.vs_offset
;
110 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONTEXT_REG
, 3));
111 radeon_ring_write(rdev
, (SQ_PGM_START_VS
- PACKET3_SET_CONTEXT_REG_START
) >> 2);
112 radeon_ring_write(rdev
, gpu_addr
>> 8);
113 radeon_ring_write(rdev
, 2);
114 radeon_ring_write(rdev
, 0);
117 gpu_addr
= rdev
->r600_blit
.shader_gpu_addr
+ rdev
->r600_blit
.ps_offset
;
118 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONTEXT_REG
, 4));
119 radeon_ring_write(rdev
, (SQ_PGM_START_PS
- PACKET3_SET_CONTEXT_REG_START
) >> 2);
120 radeon_ring_write(rdev
, gpu_addr
>> 8);
121 radeon_ring_write(rdev
, 1);
122 radeon_ring_write(rdev
, 0);
123 radeon_ring_write(rdev
, 2);
125 gpu_addr
= rdev
->r600_blit
.shader_gpu_addr
+ rdev
->r600_blit
.vs_offset
;
126 cp_set_surface_sync(rdev
, PACKET3_SH_ACTION_ENA
, 512, gpu_addr
);
129 /* emits 10 + 1 sync (5) = 15 */
131 set_vtx_resource(struct radeon_device
*rdev
, u64 gpu_addr
)
133 u32 sq_vtx_constant_word2
, sq_vtx_constant_word3
;
135 /* high addr, stride */
136 sq_vtx_constant_word2
= ((upper_32_bits(gpu_addr
) & 0xff) | (16 << 8));
138 sq_vtx_constant_word2
|= (2 << 30);
141 sq_vtx_constant_word3
= (0 << 3) | (1 << 6) | (2 << 9) | (3 << 12);
143 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_RESOURCE
, 8));
144 radeon_ring_write(rdev
, 0x580);
145 radeon_ring_write(rdev
, gpu_addr
& 0xffffffff);
146 radeon_ring_write(rdev
, 48 - 1); /* size */
147 radeon_ring_write(rdev
, sq_vtx_constant_word2
);
148 radeon_ring_write(rdev
, sq_vtx_constant_word3
);
149 radeon_ring_write(rdev
, 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_CEDAR
) ||
155 (rdev
->family
== CHIP_PALM
) ||
156 (rdev
->family
== CHIP_SUMO
) ||
157 (rdev
->family
== CHIP_SUMO2
) ||
158 (rdev
->family
== CHIP_CAICOS
))
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
);
169 set_tex_resource(struct radeon_device
*rdev
,
170 int format
, int w
, int h
, int pitch
,
173 u32 sq_tex_resource_word0
, sq_tex_resource_word1
;
174 u32 sq_tex_resource_word4
, sq_tex_resource_word7
;
179 sq_tex_resource_word0
= (1 << 0); /* 2D */
180 sq_tex_resource_word0
|= ((((pitch
>> 3) - 1) << 6) |
182 sq_tex_resource_word1
= ((h
- 1) << 0) | (1 << 28);
184 sq_tex_resource_word4
= (0 << 16) | (1 << 19) | (2 << 22) | (3 << 25);
186 sq_tex_resource_word7
= format
| (SQ_TEX_VTX_VALID_TEXTURE
<< 30);
188 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_RESOURCE
, 8));
189 radeon_ring_write(rdev
, 0);
190 radeon_ring_write(rdev
, sq_tex_resource_word0
);
191 radeon_ring_write(rdev
, sq_tex_resource_word1
);
192 radeon_ring_write(rdev
, gpu_addr
>> 8);
193 radeon_ring_write(rdev
, gpu_addr
>> 8);
194 radeon_ring_write(rdev
, sq_tex_resource_word4
);
195 radeon_ring_write(rdev
, 0);
196 radeon_ring_write(rdev
, 0);
197 radeon_ring_write(rdev
, sq_tex_resource_word7
);
202 set_scissors(struct radeon_device
*rdev
, int x1
, int y1
,
205 /* workaround some hw bugs */
210 if (rdev
->family
== CHIP_CAYMAN
) {
211 if ((x2
== 1) && (y2
== 1))
215 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONTEXT_REG
, 2));
216 radeon_ring_write(rdev
, (PA_SC_SCREEN_SCISSOR_TL
- PACKET3_SET_CONTEXT_REG_START
) >> 2);
217 radeon_ring_write(rdev
, (x1
<< 0) | (y1
<< 16));
218 radeon_ring_write(rdev
, (x2
<< 0) | (y2
<< 16));
220 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONTEXT_REG
, 2));
221 radeon_ring_write(rdev
, (PA_SC_GENERIC_SCISSOR_TL
- PACKET3_SET_CONTEXT_REG_START
) >> 2);
222 radeon_ring_write(rdev
, (x1
<< 0) | (y1
<< 16) | (1 << 31));
223 radeon_ring_write(rdev
, (x2
<< 0) | (y2
<< 16));
225 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONTEXT_REG
, 2));
226 radeon_ring_write(rdev
, (PA_SC_WINDOW_SCISSOR_TL
- PACKET3_SET_CONTEXT_REG_START
) >> 2);
227 radeon_ring_write(rdev
, (x1
<< 0) | (y1
<< 16) | (1 << 31));
228 radeon_ring_write(rdev
, (x2
<< 0) | (y2
<< 16));
233 draw_auto(struct radeon_device
*rdev
)
235 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONFIG_REG
, 1));
236 radeon_ring_write(rdev
, (VGT_PRIMITIVE_TYPE
- PACKET3_SET_CONFIG_REG_START
) >> 2);
237 radeon_ring_write(rdev
, DI_PT_RECTLIST
);
239 radeon_ring_write(rdev
, PACKET3(PACKET3_INDEX_TYPE
, 0));
240 radeon_ring_write(rdev
,
244 DI_INDEX_SIZE_16_BIT
);
246 radeon_ring_write(rdev
, PACKET3(PACKET3_NUM_INSTANCES
, 0));
247 radeon_ring_write(rdev
, 1);
249 radeon_ring_write(rdev
, PACKET3(PACKET3_DRAW_INDEX_AUTO
, 1));
250 radeon_ring_write(rdev
, 3);
251 radeon_ring_write(rdev
, DI_SRC_SEL_AUTO_INDEX
);
257 set_default_state(struct radeon_device
*rdev
)
259 u32 sq_config
, sq_gpr_resource_mgmt_1
, sq_gpr_resource_mgmt_2
, sq_gpr_resource_mgmt_3
;
260 u32 sq_thread_resource_mgmt
, sq_thread_resource_mgmt_2
;
261 u32 sq_stack_resource_mgmt_1
, sq_stack_resource_mgmt_2
, sq_stack_resource_mgmt_3
;
262 int num_ps_gprs
, num_vs_gprs
, num_temp_gprs
;
263 int num_gs_gprs
, num_es_gprs
, num_hs_gprs
, num_ls_gprs
;
264 int num_ps_threads
, num_vs_threads
, num_gs_threads
, num_es_threads
;
265 int num_hs_threads
, num_ls_threads
;
266 int num_ps_stack_entries
, num_vs_stack_entries
, num_gs_stack_entries
, num_es_stack_entries
;
267 int num_hs_stack_entries
, num_ls_stack_entries
;
271 /* set clear context state */
272 radeon_ring_write(rdev
, PACKET3(PACKET3_CLEAR_STATE
, 0));
273 radeon_ring_write(rdev
, 0);
275 if (rdev
->family
< CHIP_CAYMAN
) {
276 switch (rdev
->family
) {
292 num_ps_stack_entries
= 42;
293 num_vs_stack_entries
= 42;
294 num_gs_stack_entries
= 42;
295 num_es_stack_entries
= 42;
296 num_hs_stack_entries
= 42;
297 num_ls_stack_entries
= 42;
307 num_ps_threads
= 128;
313 num_ps_stack_entries
= 42;
314 num_vs_stack_entries
= 42;
315 num_gs_stack_entries
= 42;
316 num_es_stack_entries
= 42;
317 num_hs_stack_entries
= 42;
318 num_ls_stack_entries
= 42;
328 num_ps_threads
= 128;
334 num_ps_stack_entries
= 85;
335 num_vs_stack_entries
= 85;
336 num_gs_stack_entries
= 85;
337 num_es_stack_entries
= 85;
338 num_hs_stack_entries
= 85;
339 num_ls_stack_entries
= 85;
350 num_ps_threads
= 128;
356 num_ps_stack_entries
= 85;
357 num_vs_stack_entries
= 85;
358 num_gs_stack_entries
= 85;
359 num_es_stack_entries
= 85;
360 num_hs_stack_entries
= 85;
361 num_ls_stack_entries
= 85;
377 num_ps_stack_entries
= 42;
378 num_vs_stack_entries
= 42;
379 num_gs_stack_entries
= 42;
380 num_es_stack_entries
= 42;
381 num_hs_stack_entries
= 42;
382 num_ls_stack_entries
= 42;
398 num_ps_stack_entries
= 42;
399 num_vs_stack_entries
= 42;
400 num_gs_stack_entries
= 42;
401 num_es_stack_entries
= 42;
402 num_hs_stack_entries
= 42;
403 num_ls_stack_entries
= 42;
419 num_ps_stack_entries
= 85;
420 num_vs_stack_entries
= 85;
421 num_gs_stack_entries
= 85;
422 num_es_stack_entries
= 85;
423 num_hs_stack_entries
= 85;
424 num_ls_stack_entries
= 85;
434 num_ps_threads
= 128;
440 num_ps_stack_entries
= 85;
441 num_vs_stack_entries
= 85;
442 num_gs_stack_entries
= 85;
443 num_es_stack_entries
= 85;
444 num_hs_stack_entries
= 85;
445 num_ls_stack_entries
= 85;
455 num_ps_threads
= 128;
461 num_ps_stack_entries
= 42;
462 num_vs_stack_entries
= 42;
463 num_gs_stack_entries
= 42;
464 num_es_stack_entries
= 42;
465 num_hs_stack_entries
= 42;
466 num_ls_stack_entries
= 42;
476 num_ps_threads
= 128;
482 num_ps_stack_entries
= 42;
483 num_vs_stack_entries
= 42;
484 num_gs_stack_entries
= 42;
485 num_es_stack_entries
= 42;
486 num_hs_stack_entries
= 42;
487 num_ls_stack_entries
= 42;
491 if ((rdev
->family
== CHIP_CEDAR
) ||
492 (rdev
->family
== CHIP_PALM
) ||
493 (rdev
->family
== CHIP_SUMO
) ||
494 (rdev
->family
== CHIP_SUMO2
) ||
495 (rdev
->family
== CHIP_CAICOS
))
498 sq_config
= VC_ENABLE
;
500 sq_config
|= (EXPORT_SRC_C
|
509 sq_gpr_resource_mgmt_1
= (NUM_PS_GPRS(num_ps_gprs
) |
510 NUM_VS_GPRS(num_vs_gprs
) |
511 NUM_CLAUSE_TEMP_GPRS(num_temp_gprs
));
512 sq_gpr_resource_mgmt_2
= (NUM_GS_GPRS(num_gs_gprs
) |
513 NUM_ES_GPRS(num_es_gprs
));
514 sq_gpr_resource_mgmt_3
= (NUM_HS_GPRS(num_hs_gprs
) |
515 NUM_LS_GPRS(num_ls_gprs
));
516 sq_thread_resource_mgmt
= (NUM_PS_THREADS(num_ps_threads
) |
517 NUM_VS_THREADS(num_vs_threads
) |
518 NUM_GS_THREADS(num_gs_threads
) |
519 NUM_ES_THREADS(num_es_threads
));
520 sq_thread_resource_mgmt_2
= (NUM_HS_THREADS(num_hs_threads
) |
521 NUM_LS_THREADS(num_ls_threads
));
522 sq_stack_resource_mgmt_1
= (NUM_PS_STACK_ENTRIES(num_ps_stack_entries
) |
523 NUM_VS_STACK_ENTRIES(num_vs_stack_entries
));
524 sq_stack_resource_mgmt_2
= (NUM_GS_STACK_ENTRIES(num_gs_stack_entries
) |
525 NUM_ES_STACK_ENTRIES(num_es_stack_entries
));
526 sq_stack_resource_mgmt_3
= (NUM_HS_STACK_ENTRIES(num_hs_stack_entries
) |
527 NUM_LS_STACK_ENTRIES(num_ls_stack_entries
));
529 /* disable dyn gprs */
530 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONFIG_REG
, 1));
531 radeon_ring_write(rdev
, (SQ_DYN_GPR_CNTL_PS_FLUSH_REQ
- PACKET3_SET_CONFIG_REG_START
) >> 2);
532 radeon_ring_write(rdev
, 0);
535 radeon_ring_write(rdev
, PACKET3(PACKET3_SET_CONFIG_REG
, 11));
536 radeon_ring_write(rdev
, (SQ_CONFIG
- PACKET3_SET_CONFIG_REG_START
) >> 2);
537 radeon_ring_write(rdev
, sq_config
);
538 radeon_ring_write(rdev
, sq_gpr_resource_mgmt_1
);
539 radeon_ring_write(rdev
, sq_gpr_resource_mgmt_2
);
540 radeon_ring_write(rdev
, sq_gpr_resource_mgmt_3
);
541 radeon_ring_write(rdev
, 0);
542 radeon_ring_write(rdev
, 0);
543 radeon_ring_write(rdev
, sq_thread_resource_mgmt
);
544 radeon_ring_write(rdev
, sq_thread_resource_mgmt_2
);
545 radeon_ring_write(rdev
, sq_stack_resource_mgmt_1
);
546 radeon_ring_write(rdev
, sq_stack_resource_mgmt_2
);
547 radeon_ring_write(rdev
, sq_stack_resource_mgmt_3
);
550 /* CONTEXT_CONTROL */
551 radeon_ring_write(rdev
, 0xc0012800);
552 radeon_ring_write(rdev
, 0x80000000);
553 radeon_ring_write(rdev
, 0x80000000);
555 /* SQ_VTX_BASE_VTX_LOC */
556 radeon_ring_write(rdev
, 0xc0026f00);
557 radeon_ring_write(rdev
, 0x00000000);
558 radeon_ring_write(rdev
, 0x00000000);
559 radeon_ring_write(rdev
, 0x00000000);
562 radeon_ring_write(rdev
, 0xc0036e00);
563 radeon_ring_write(rdev
, 0x00000000);
564 radeon_ring_write(rdev
, 0x00000012);
565 radeon_ring_write(rdev
, 0x00000000);
566 radeon_ring_write(rdev
, 0x00000000);
568 /* set to DX10/11 mode */
569 radeon_ring_write(rdev
, PACKET3(PACKET3_MODE_CONTROL
, 0));
570 radeon_ring_write(rdev
, 1);
572 /* emit an IB pointing at default state */
573 dwords
= ALIGN(rdev
->r600_blit
.state_len
, 0x10);
574 gpu_addr
= rdev
->r600_blit
.shader_gpu_addr
+ rdev
->r600_blit
.state_offset
;
575 radeon_ring_write(rdev
, PACKET3(PACKET3_INDIRECT_BUFFER
, 2));
576 radeon_ring_write(rdev
, gpu_addr
& 0xFFFFFFFC);
577 radeon_ring_write(rdev
, upper_32_bits(gpu_addr
) & 0xFF);
578 radeon_ring_write(rdev
, dwords
);
582 static inline uint32_t i2f(uint32_t input
)
584 u32 result
, i
, exponent
, fraction
;
586 if ((input
& 0x3fff) == 0)
587 result
= 0; /* 0 is a special case */
589 exponent
= 140; /* exponent biased by 127; */
590 fraction
= (input
& 0x3fff) << 10; /* cheat and only
591 handle numbers below 2^^15 */
592 for (i
= 0; i
< 14; i
++) {
593 if (fraction
& 0x800000)
596 fraction
= fraction
<< 1; /* keep
597 shifting left until top bit = 1 */
598 exponent
= exponent
- 1;
601 result
= exponent
<< 23 | (fraction
& 0x7fffff); /* mask
602 off top bit; assumed 1 */
607 int evergreen_blit_init(struct radeon_device
*rdev
)
613 int num_packet2s
= 0;
615 /* pin copy shader into vram if already initialized */
616 if (rdev
->r600_blit
.shader_obj
)
619 mutex_init(&rdev
->r600_blit
.mutex
);
620 rdev
->r600_blit
.state_offset
= 0;
622 if (rdev
->family
< CHIP_CAYMAN
)
623 rdev
->r600_blit
.state_len
= evergreen_default_size
;
625 rdev
->r600_blit
.state_len
= cayman_default_size
;
627 dwords
= rdev
->r600_blit
.state_len
;
628 while (dwords
& 0xf) {
629 packet2s
[num_packet2s
++] = cpu_to_le32(PACKET2(0));
633 obj_size
= dwords
* 4;
634 obj_size
= ALIGN(obj_size
, 256);
636 rdev
->r600_blit
.vs_offset
= obj_size
;
637 if (rdev
->family
< CHIP_CAYMAN
)
638 obj_size
+= evergreen_vs_size
* 4;
640 obj_size
+= cayman_vs_size
* 4;
641 obj_size
= ALIGN(obj_size
, 256);
643 rdev
->r600_blit
.ps_offset
= obj_size
;
644 if (rdev
->family
< CHIP_CAYMAN
)
645 obj_size
+= evergreen_ps_size
* 4;
647 obj_size
+= cayman_ps_size
* 4;
648 obj_size
= ALIGN(obj_size
, 256);
650 r
= radeon_bo_create(rdev
, obj_size
, PAGE_SIZE
, true, RADEON_GEM_DOMAIN_VRAM
,
651 &rdev
->r600_blit
.shader_obj
);
653 DRM_ERROR("evergreen failed to allocate shader\n");
657 DRM_DEBUG("evergreen blit allocated bo %08x vs %08x ps %08x\n",
659 rdev
->r600_blit
.vs_offset
, rdev
->r600_blit
.ps_offset
);
661 r
= radeon_bo_reserve(rdev
->r600_blit
.shader_obj
, false);
662 if (unlikely(r
!= 0))
664 r
= radeon_bo_kmap(rdev
->r600_blit
.shader_obj
, &ptr
);
666 DRM_ERROR("failed to map blit object %d\n", r
);
670 if (rdev
->family
< CHIP_CAYMAN
) {
671 memcpy_toio(ptr
+ rdev
->r600_blit
.state_offset
,
672 evergreen_default_state
, rdev
->r600_blit
.state_len
* 4);
675 memcpy_toio(ptr
+ rdev
->r600_blit
.state_offset
+ (rdev
->r600_blit
.state_len
* 4),
676 packet2s
, num_packet2s
* 4);
677 for (i
= 0; i
< evergreen_vs_size
; i
++)
678 *(u32
*)((unsigned long)ptr
+ rdev
->r600_blit
.vs_offset
+ i
* 4) = cpu_to_le32(evergreen_vs
[i
]);
679 for (i
= 0; i
< evergreen_ps_size
; i
++)
680 *(u32
*)((unsigned long)ptr
+ rdev
->r600_blit
.ps_offset
+ i
* 4) = cpu_to_le32(evergreen_ps
[i
]);
682 memcpy_toio(ptr
+ rdev
->r600_blit
.state_offset
,
683 cayman_default_state
, rdev
->r600_blit
.state_len
* 4);
686 memcpy_toio(ptr
+ rdev
->r600_blit
.state_offset
+ (rdev
->r600_blit
.state_len
* 4),
687 packet2s
, num_packet2s
* 4);
688 for (i
= 0; i
< cayman_vs_size
; i
++)
689 *(u32
*)((unsigned long)ptr
+ rdev
->r600_blit
.vs_offset
+ i
* 4) = cpu_to_le32(cayman_vs
[i
]);
690 for (i
= 0; i
< cayman_ps_size
; i
++)
691 *(u32
*)((unsigned long)ptr
+ rdev
->r600_blit
.ps_offset
+ i
* 4) = cpu_to_le32(cayman_ps
[i
]);
693 radeon_bo_kunmap(rdev
->r600_blit
.shader_obj
);
694 radeon_bo_unreserve(rdev
->r600_blit
.shader_obj
);
697 r
= radeon_bo_reserve(rdev
->r600_blit
.shader_obj
, false);
698 if (unlikely(r
!= 0))
700 r
= radeon_bo_pin(rdev
->r600_blit
.shader_obj
, RADEON_GEM_DOMAIN_VRAM
,
701 &rdev
->r600_blit
.shader_gpu_addr
);
702 radeon_bo_unreserve(rdev
->r600_blit
.shader_obj
);
704 dev_err(rdev
->dev
, "(%d) pin blit object failed\n", r
);
707 radeon_ttm_set_active_vram_size(rdev
, rdev
->mc
.real_vram_size
);
711 void evergreen_blit_fini(struct radeon_device
*rdev
)
715 radeon_ttm_set_active_vram_size(rdev
, rdev
->mc
.visible_vram_size
);
716 if (rdev
->r600_blit
.shader_obj
== NULL
)
718 /* If we can't reserve the bo, unref should be enough to destroy
719 * it when it becomes idle.
721 r
= radeon_bo_reserve(rdev
->r600_blit
.shader_obj
, false);
723 radeon_bo_unpin(rdev
->r600_blit
.shader_obj
);
724 radeon_bo_unreserve(rdev
->r600_blit
.shader_obj
);
726 radeon_bo_unref(&rdev
->r600_blit
.shader_obj
);
729 static int evergreen_vb_ib_get(struct radeon_device
*rdev
)
732 r
= radeon_ib_get(rdev
, &rdev
->r600_blit
.vb_ib
);
734 DRM_ERROR("failed to get IB for vertex buffer\n");
738 rdev
->r600_blit
.vb_total
= 64*1024;
739 rdev
->r600_blit
.vb_used
= 0;
743 static void evergreen_vb_ib_put(struct radeon_device
*rdev
)
745 radeon_fence_emit(rdev
, rdev
->r600_blit
.vb_ib
->fence
);
746 radeon_ib_free(rdev
, &rdev
->r600_blit
.vb_ib
);
749 int evergreen_blit_prepare_copy(struct radeon_device
*rdev
, int size_bytes
)
752 int ring_size
, line_size
;
754 /* loops of emits + fence emit possible */
755 int dwords_per_loop
= 74, num_loops
;
757 r
= evergreen_vb_ib_get(rdev
);
761 /* 8 bpp vs 32 bpp for xfer unit */
765 line_size
= 8192 * 4;
767 max_size
= 8192 * line_size
;
769 /* major loops cover the max size transfer */
770 num_loops
= ((size_bytes
+ max_size
) / max_size
);
771 /* minor loops cover the extra non aligned bits */
772 num_loops
+= ((size_bytes
% line_size
) ? 1 : 0);
773 /* calculate number of loops correctly */
774 ring_size
= num_loops
* dwords_per_loop
;
775 /* set default + shaders */
776 ring_size
+= 52; /* shaders + def state */
777 ring_size
+= 10; /* fence emit for VB IB */
778 ring_size
+= 5; /* done copy */
779 ring_size
+= 10; /* fence emit for done copy */
780 r
= radeon_ring_lock(rdev
, ring_size
);
784 set_default_state(rdev
); /* 36 */
785 set_shaders(rdev
); /* 16 */
789 void evergreen_blit_done_copy(struct radeon_device
*rdev
, struct radeon_fence
*fence
)
793 if (rdev
->r600_blit
.vb_ib
)
794 evergreen_vb_ib_put(rdev
);
797 r
= radeon_fence_emit(rdev
, fence
);
799 radeon_ring_unlock_commit(rdev
);
802 void evergreen_kms_blit_copy(struct radeon_device
*rdev
,
803 u64 src_gpu_addr
, u64 dst_gpu_addr
,
810 DRM_DEBUG("emitting copy %16llx %16llx %d %d\n", src_gpu_addr
, dst_gpu_addr
,
811 size_bytes
, rdev
->r600_blit
.vb_used
);
812 vb
= (u32
*)(rdev
->r600_blit
.vb_ib
->ptr
+ rdev
->r600_blit
.vb_used
);
813 if ((size_bytes
& 3) || (src_gpu_addr
& 3) || (dst_gpu_addr
& 3)) {
817 int cur_size
= size_bytes
;
818 int src_x
= src_gpu_addr
& 255;
819 int dst_x
= dst_gpu_addr
& 255;
821 src_gpu_addr
= src_gpu_addr
& ~255ULL;
822 dst_gpu_addr
= dst_gpu_addr
& ~255ULL;
824 if (!src_x
&& !dst_x
) {
825 h
= (cur_size
/ max_bytes
);
831 cur_size
= max_bytes
;
833 if (cur_size
> max_bytes
)
834 cur_size
= max_bytes
;
835 if (cur_size
> (max_bytes
- dst_x
))
836 cur_size
= (max_bytes
- dst_x
);
837 if (cur_size
> (max_bytes
- src_x
))
838 cur_size
= (max_bytes
- src_x
);
841 if ((rdev
->r600_blit
.vb_used
+ 48) > rdev
->r600_blit
.vb_total
) {
855 vb
[8] = i2f(dst_x
+ cur_size
);
857 vb
[10] = i2f(src_x
+ cur_size
);
861 set_tex_resource(rdev
, FMT_8
,
862 src_x
+ cur_size
, h
, src_x
+ cur_size
,
866 cp_set_surface_sync(rdev
,
867 PACKET3_TC_ACTION_ENA
, (src_x
+ cur_size
* h
), src_gpu_addr
);
871 set_render_target(rdev
, COLOR_8
,
876 set_scissors(rdev
, dst_x
, 0, dst_x
+ cur_size
, h
);
879 vb_gpu_addr
= rdev
->r600_blit
.vb_ib
->gpu_addr
+ rdev
->r600_blit
.vb_used
;
880 set_vtx_resource(rdev
, vb_gpu_addr
);
886 cp_set_surface_sync(rdev
,
887 PACKET3_CB_ACTION_ENA
| PACKET3_CB0_DEST_BASE_ENA
,
888 cur_size
* h
, dst_gpu_addr
);
891 rdev
->r600_blit
.vb_used
+= 12 * 4;
893 src_gpu_addr
+= cur_size
* h
;
894 dst_gpu_addr
+= cur_size
* h
;
895 size_bytes
-= cur_size
* h
;
898 max_bytes
= 8192 * 4;
901 int cur_size
= size_bytes
;
902 int src_x
= (src_gpu_addr
& 255);
903 int dst_x
= (dst_gpu_addr
& 255);
905 src_gpu_addr
= src_gpu_addr
& ~255ULL;
906 dst_gpu_addr
= dst_gpu_addr
& ~255ULL;
908 if (!src_x
&& !dst_x
) {
909 h
= (cur_size
/ max_bytes
);
915 cur_size
= max_bytes
;
917 if (cur_size
> max_bytes
)
918 cur_size
= max_bytes
;
919 if (cur_size
> (max_bytes
- dst_x
))
920 cur_size
= (max_bytes
- dst_x
);
921 if (cur_size
> (max_bytes
- src_x
))
922 cur_size
= (max_bytes
- src_x
);
925 if ((rdev
->r600_blit
.vb_used
+ 48) > rdev
->r600_blit
.vb_total
) {
929 vb
[0] = i2f(dst_x
/ 4);
931 vb
[2] = i2f(src_x
/ 4);
934 vb
[4] = i2f(dst_x
/ 4);
936 vb
[6] = i2f(src_x
/ 4);
939 vb
[8] = i2f((dst_x
+ cur_size
) / 4);
941 vb
[10] = i2f((src_x
+ cur_size
) / 4);
945 set_tex_resource(rdev
, FMT_8_8_8_8
,
946 (src_x
+ cur_size
) / 4,
947 h
, (src_x
+ cur_size
) / 4,
950 cp_set_surface_sync(rdev
,
951 PACKET3_TC_ACTION_ENA
, (src_x
+ cur_size
* h
), src_gpu_addr
);
954 set_render_target(rdev
, COLOR_8_8_8_8
,
955 (dst_x
+ cur_size
) / 4, h
,
959 set_scissors(rdev
, (dst_x
/ 4), 0, (dst_x
+ cur_size
/ 4), h
);
961 /* Vertex buffer setup 15 */
962 vb_gpu_addr
= rdev
->r600_blit
.vb_ib
->gpu_addr
+ rdev
->r600_blit
.vb_used
;
963 set_vtx_resource(rdev
, vb_gpu_addr
);
969 cp_set_surface_sync(rdev
,
970 PACKET3_CB_ACTION_ENA
| PACKET3_CB0_DEST_BASE_ENA
,
971 cur_size
* h
, dst_gpu_addr
);
973 /* 74 ring dwords per loop */
975 rdev
->r600_blit
.vb_used
+= 12 * 4;
977 src_gpu_addr
+= cur_size
* h
;
978 dst_gpu_addr
+= cur_size
* h
;
979 size_bytes
-= cur_size
* h
;