2 * Copyright 2010 Christoph Bumiller
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 shall be included in
12 * all copies or substantial portions of the Software.
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
23 #include "util/u_format.h"
24 #include "util/u_format_s3tc.h"
25 #include "pipe/p_screen.h"
27 #include "vl/vl_decoder.h"
28 #include "vl/vl_video_buffer.h"
30 #include "nouveau/nouveau_vp3_video.h"
32 #include "nvc0_context.h"
33 #include "nvc0_screen.h"
35 #include "nvc0_graph_macros.h"
37 #ifndef NOUVEAU_GETPARAM_GRAPH_UNITS
38 # define NOUVEAU_GETPARAM_GRAPH_UNITS 13
42 nvc0_screen_is_format_supported(struct pipe_screen
*pscreen
,
43 enum pipe_format format
,
44 enum pipe_texture_target target
,
45 unsigned sample_count
,
50 if (!(0x117 & (1 << sample_count
))) /* 0, 1, 2, 4 or 8 */
53 if (!util_format_is_supported(format
, bindings
))
56 if ((bindings
& PIPE_BIND_SAMPLER_VIEW
) && (target
!= PIPE_BUFFER
))
57 if (util_format_get_blocksizebits(format
) == 3 * 32)
60 /* transfers & shared are always supported */
61 bindings
&= ~(PIPE_BIND_TRANSFER_READ
|
62 PIPE_BIND_TRANSFER_WRITE
|
65 return (nvc0_format_table
[format
].usage
& bindings
) == bindings
;
69 nvc0_screen_get_param(struct pipe_screen
*pscreen
, enum pipe_cap param
)
71 const uint16_t class_3d
= nouveau_screen(pscreen
)->class_3d
;
74 case PIPE_CAP_MAX_COMBINED_SAMPLERS
:
76 case PIPE_CAP_MAX_TEXTURE_2D_LEVELS
:
77 case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS
:
79 case PIPE_CAP_MAX_TEXTURE_3D_LEVELS
:
80 return (class_3d
>= NVE4_3D_CLASS
) ? 13 : 12;
81 case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS
:
83 case PIPE_CAP_MIN_TEXEL_OFFSET
:
85 case PIPE_CAP_MAX_TEXEL_OFFSET
:
87 case PIPE_CAP_TEXTURE_MIRROR_CLAMP
:
88 case PIPE_CAP_TEXTURE_SWIZZLE
:
89 case PIPE_CAP_TEXTURE_SHADOW_MAP
:
90 case PIPE_CAP_NPOT_TEXTURES
:
91 case PIPE_CAP_ANISOTROPIC_FILTER
:
92 case PIPE_CAP_SEAMLESS_CUBE_MAP
:
93 case PIPE_CAP_CUBE_MAP_ARRAY
:
94 case PIPE_CAP_TEXTURE_BUFFER_OBJECTS
:
95 case PIPE_CAP_TEXTURE_MULTISAMPLE
:
97 case PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE
:
99 case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE
:
100 return (class_3d
>= NVE4_3D_CLASS
) ? 1 : 0;
101 case PIPE_CAP_TWO_SIDED_STENCIL
:
102 case PIPE_CAP_DEPTH_CLIP_DISABLE
:
103 case PIPE_CAP_POINT_SPRITE
:
104 case PIPE_CAP_TGSI_TEXCOORD
:
108 case PIPE_CAP_GLSL_FEATURE_LEVEL
:
110 case PIPE_CAP_MAX_RENDER_TARGETS
:
112 case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS
:
114 case PIPE_CAP_FRAGMENT_COLOR_CLAMPED
:
115 case PIPE_CAP_VERTEX_COLOR_UNCLAMPED
:
116 case PIPE_CAP_VERTEX_COLOR_CLAMPED
:
118 case PIPE_CAP_QUERY_TIMESTAMP
:
119 case PIPE_CAP_QUERY_TIME_ELAPSED
:
120 case PIPE_CAP_OCCLUSION_QUERY
:
121 case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME
:
122 case PIPE_CAP_QUERY_PIPELINE_STATISTICS
:
124 case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS
:
126 case PIPE_CAP_MAX_STREAM_OUTPUT_SEPARATE_COMPONENTS
:
127 case PIPE_CAP_MAX_STREAM_OUTPUT_INTERLEAVED_COMPONENTS
:
129 case PIPE_CAP_BLEND_EQUATION_SEPARATE
:
130 case PIPE_CAP_INDEP_BLEND_ENABLE
:
131 case PIPE_CAP_INDEP_BLEND_FUNC
:
133 case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT
:
134 case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER
:
136 case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT
:
137 case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER
:
139 case PIPE_CAP_SHADER_STENCIL_EXPORT
:
141 case PIPE_CAP_PRIMITIVE_RESTART
:
142 case PIPE_CAP_TGSI_INSTANCEID
:
143 case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR
:
144 case PIPE_CAP_MIXED_COLORBUFFER_FORMATS
:
145 case PIPE_CAP_CONDITIONAL_RENDER
:
146 case PIPE_CAP_TEXTURE_BARRIER
:
147 case PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION
:
148 case PIPE_CAP_START_INSTANCE
:
150 case PIPE_CAP_TGSI_CAN_COMPACT_CONSTANTS
:
151 return 0; /* state trackers will know better */
152 case PIPE_CAP_USER_CONSTANT_BUFFERS
:
153 case PIPE_CAP_USER_INDEX_BUFFERS
:
154 case PIPE_CAP_USER_VERTEX_BUFFERS
:
156 case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT
:
158 case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT
:
159 return 1; /* 256 for binding as RT, but that's not possible in GL */
160 case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT
:
161 return NOUVEAU_MIN_BUFFER_MAP_ALIGN
;
162 case PIPE_CAP_VERTEX_BUFFER_OFFSET_4BYTE_ALIGNED_ONLY
:
163 case PIPE_CAP_VERTEX_BUFFER_STRIDE_4BYTE_ALIGNED_ONLY
:
164 case PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY
:
166 case PIPE_CAP_COMPUTE
:
167 return (class_3d
>= NVE4_3D_CLASS
) ? 1 : 0;
168 case PIPE_CAP_PREFER_BLIT_BASED_TEXTURE_TRANSFER
:
170 case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK
:
171 return PIPE_QUIRK_TEXTURE_BORDER_COLOR_SWIZZLE_NV50
;
172 case PIPE_CAP_ENDIANNESS
:
173 return PIPE_ENDIAN_LITTLE
;
175 NOUVEAU_ERR("unknown PIPE_CAP %d\n", param
);
181 nvc0_screen_get_shader_param(struct pipe_screen
*pscreen
, unsigned shader
,
182 enum pipe_shader_cap param
)
184 const uint16_t class_3d
= nouveau_screen(pscreen
)->class_3d
;
187 case PIPE_SHADER_VERTEX
:
189 case PIPE_SHADER_TESSELLATION_CONTROL:
190 case PIPE_SHADER_TESSELLATION_EVALUATION:
192 case PIPE_SHADER_GEOMETRY
:
193 case PIPE_SHADER_FRAGMENT
:
195 case PIPE_SHADER_COMPUTE
:
196 if (class_3d
< NVE4_3D_CLASS
)
204 case PIPE_SHADER_CAP_PREFERRED_IR
:
205 return PIPE_SHADER_IR_TGSI
;
206 case PIPE_SHADER_CAP_MAX_INSTRUCTIONS
:
207 case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS
:
208 case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS
:
209 case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS
:
211 case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH
:
213 case PIPE_SHADER_CAP_MAX_INPUTS
:
214 if (shader
== PIPE_SHADER_VERTEX
)
216 /* NOTE: These only count our slots for GENERIC varyings.
217 * The address space may be larger, but the actual hard limit seems to be
218 * less than what the address space layout permits, so don't add TEXCOORD,
221 if (shader
== PIPE_SHADER_FRAGMENT
)
223 /* Actually this counts CLIPVERTEX, which occupies the last generic slot,
224 * and excludes 0x60 per-patch inputs.
227 case PIPE_SHADER_CAP_MAX_CONSTS
:
229 case PIPE_SHADER_CAP_MAX_CONST_BUFFERS
:
230 if (shader
== PIPE_SHADER_COMPUTE
&& class_3d
>= NVE4_3D_CLASS
)
231 return NVE4_MAX_PIPE_CONSTBUFS_COMPUTE
;
232 return NVC0_MAX_PIPE_CONSTBUFS
;
233 case PIPE_SHADER_CAP_MAX_ADDRS
:
235 case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR
:
236 case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR
:
237 return shader
!= PIPE_SHADER_FRAGMENT
;
238 case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR
:
239 case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR
:
241 case PIPE_SHADER_CAP_MAX_PREDS
:
243 case PIPE_SHADER_CAP_MAX_TEMPS
:
244 return NVC0_CAP_MAX_PROGRAM_TEMPS
;
245 case PIPE_SHADER_CAP_TGSI_CONT_SUPPORTED
:
247 case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED
:
249 case PIPE_SHADER_CAP_SUBROUTINES
:
251 case PIPE_SHADER_CAP_INTEGERS
:
253 case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS
:
254 return 16; /* would be 32 in linked (OpenGL-style) mode */
256 case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLER_VIEWS:
260 NOUVEAU_ERR("unknown PIPE_SHADER_CAP %d\n", param
);
266 nvc0_screen_get_paramf(struct pipe_screen
*pscreen
, enum pipe_capf param
)
269 case PIPE_CAPF_MAX_LINE_WIDTH
:
270 case PIPE_CAPF_MAX_LINE_WIDTH_AA
:
272 case PIPE_CAPF_MAX_POINT_WIDTH
:
274 case PIPE_CAPF_MAX_POINT_WIDTH_AA
:
276 case PIPE_CAPF_MAX_TEXTURE_ANISOTROPY
:
278 case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS
:
281 NOUVEAU_ERR("unknown PIPE_CAP %d\n", param
);
287 nvc0_screen_get_compute_param(struct pipe_screen
*pscreen
,
288 enum pipe_compute_cap param
, void *data
)
290 uint64_t *data64
= (uint64_t *)data
;
291 const uint16_t obj_class
= nvc0_screen(pscreen
)->compute
->oclass
;
294 case PIPE_COMPUTE_CAP_GRID_DIMENSION
:
297 case PIPE_COMPUTE_CAP_MAX_GRID_SIZE
:
298 data64
[0] = (obj_class
>= NVE4_COMPUTE_CLASS
) ? 0x7fffffff : 65535;
302 case PIPE_COMPUTE_CAP_MAX_BLOCK_SIZE
:
307 case PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK
:
310 case PIPE_COMPUTE_CAP_MAX_GLOBAL_SIZE
: /* g[] */
311 data64
[0] = (uint64_t)1 << 40;
313 case PIPE_COMPUTE_CAP_MAX_LOCAL_SIZE
: /* s[] */
314 data64
[0] = 48 << 10;
316 case PIPE_COMPUTE_CAP_MAX_PRIVATE_SIZE
: /* l[] */
317 data64
[0] = 512 << 10;
319 case PIPE_COMPUTE_CAP_MAX_INPUT_SIZE
: /* c[], arbitrary limit */
328 nvc0_screen_destroy(struct pipe_screen
*pscreen
)
330 struct nvc0_screen
*screen
= nvc0_screen(pscreen
);
332 if (screen
->base
.fence
.current
) {
333 nouveau_fence_wait(screen
->base
.fence
.current
);
334 nouveau_fence_ref(NULL
, &screen
->base
.fence
.current
);
336 if (screen
->base
.pushbuf
)
337 screen
->base
.pushbuf
->user_priv
= NULL
;
340 nvc0_blitter_destroy(screen
);
341 if (screen
->pm
.prog
) {
342 screen
->pm
.prog
->code
= NULL
; /* hardcoded, don't FREE */
343 nvc0_program_destroy(NULL
, screen
->pm
.prog
);
346 nouveau_bo_ref(NULL
, &screen
->text
);
347 nouveau_bo_ref(NULL
, &screen
->uniform_bo
);
348 nouveau_bo_ref(NULL
, &screen
->tls
);
349 nouveau_bo_ref(NULL
, &screen
->txc
);
350 nouveau_bo_ref(NULL
, &screen
->fence
.bo
);
351 nouveau_bo_ref(NULL
, &screen
->poly_cache
);
352 nouveau_bo_ref(NULL
, &screen
->parm
);
354 nouveau_heap_destroy(&screen
->lib_code
);
355 nouveau_heap_destroy(&screen
->text_heap
);
357 FREE(screen
->tic
.entries
);
359 nouveau_mm_destroy(screen
->mm_VRAM_fe0
);
361 nouveau_object_del(&screen
->eng3d
);
362 nouveau_object_del(&screen
->eng2d
);
363 nouveau_object_del(&screen
->m2mf
);
365 nouveau_screen_fini(&screen
->base
);
371 nvc0_graph_set_macro(struct nvc0_screen
*screen
, uint32_t m
, unsigned pos
,
372 unsigned size
, const uint32_t *data
)
374 struct nouveau_pushbuf
*push
= screen
->base
.pushbuf
;
378 BEGIN_NVC0(push
, SUBC_3D(NVC0_GRAPH_MACRO_ID
), 2);
379 PUSH_DATA (push
, (m
- 0x3800) / 8);
380 PUSH_DATA (push
, pos
);
381 BEGIN_1IC0(push
, SUBC_3D(NVC0_GRAPH_MACRO_UPLOAD_POS
), size
+ 1);
382 PUSH_DATA (push
, pos
);
383 PUSH_DATAp(push
, data
, size
);
389 nvc0_magic_3d_init(struct nouveau_pushbuf
*push
, uint16_t obj_class
)
391 BEGIN_NVC0(push
, SUBC_3D(0x10cc), 1);
392 PUSH_DATA (push
, 0xff);
393 BEGIN_NVC0(push
, SUBC_3D(0x10e0), 2);
394 PUSH_DATA (push
, 0xff);
395 PUSH_DATA (push
, 0xff);
396 BEGIN_NVC0(push
, SUBC_3D(0x10ec), 2);
397 PUSH_DATA (push
, 0xff);
398 PUSH_DATA (push
, 0xff);
399 BEGIN_NVC0(push
, SUBC_3D(0x074c), 1);
400 PUSH_DATA (push
, 0x3f);
402 BEGIN_NVC0(push
, SUBC_3D(0x16a8), 1);
403 PUSH_DATA (push
, (3 << 16) | 3);
404 BEGIN_NVC0(push
, SUBC_3D(0x1794), 1);
405 PUSH_DATA (push
, (2 << 16) | 2);
406 BEGIN_NVC0(push
, SUBC_3D(0x0de8), 1);
409 BEGIN_NVC0(push
, SUBC_3D(0x12ac), 1);
411 BEGIN_NVC0(push
, SUBC_3D(0x0218), 1);
412 PUSH_DATA (push
, 0x10);
413 BEGIN_NVC0(push
, SUBC_3D(0x10fc), 1);
414 PUSH_DATA (push
, 0x10);
415 BEGIN_NVC0(push
, SUBC_3D(0x1290), 1);
416 PUSH_DATA (push
, 0x10);
417 BEGIN_NVC0(push
, SUBC_3D(0x12d8), 2);
418 PUSH_DATA (push
, 0x10);
419 PUSH_DATA (push
, 0x10);
420 BEGIN_NVC0(push
, SUBC_3D(0x1140), 1);
421 PUSH_DATA (push
, 0x10);
422 BEGIN_NVC0(push
, SUBC_3D(0x1610), 1);
423 PUSH_DATA (push
, 0xe);
425 BEGIN_NVC0(push
, SUBC_3D(0x164c), 1);
426 PUSH_DATA (push
, 1 << 12);
427 BEGIN_NVC0(push
, SUBC_3D(0x030c), 1);
429 BEGIN_NVC0(push
, SUBC_3D(0x0300), 1);
432 BEGIN_NVC0(push
, SUBC_3D(0x02d0), 1);
433 PUSH_DATA (push
, 0x3fffff);
434 BEGIN_NVC0(push
, SUBC_3D(0x0fdc), 1);
436 BEGIN_NVC0(push
, SUBC_3D(0x19c0), 1);
438 BEGIN_NVC0(push
, SUBC_3D(0x075c), 1);
441 if (obj_class
>= NVE4_3D_CLASS
) {
442 BEGIN_NVC0(push
, SUBC_3D(0x07fc), 1);
446 /* TODO: find out what software methods 0x1528, 0x1280 and (on nve4) 0x02dc
447 * are supposed to do */
451 nvc0_screen_fence_emit(struct pipe_screen
*pscreen
, u32
*sequence
)
453 struct nvc0_screen
*screen
= nvc0_screen(pscreen
);
454 struct nouveau_pushbuf
*push
= screen
->base
.pushbuf
;
456 /* we need to do it after possible flush in MARK_RING */
457 *sequence
= ++screen
->base
.fence
.sequence
;
459 BEGIN_NVC0(push
, NVC0_3D(QUERY_ADDRESS_HIGH
), 4);
460 PUSH_DATAh(push
, screen
->fence
.bo
->offset
);
461 PUSH_DATA (push
, screen
->fence
.bo
->offset
);
462 PUSH_DATA (push
, *sequence
);
463 PUSH_DATA (push
, NVC0_3D_QUERY_GET_FENCE
| NVC0_3D_QUERY_GET_SHORT
|
464 (0xf << NVC0_3D_QUERY_GET_UNIT__SHIFT
));
468 nvc0_screen_fence_update(struct pipe_screen
*pscreen
)
470 struct nvc0_screen
*screen
= nvc0_screen(pscreen
);
471 return screen
->fence
.map
[0];
475 nvc0_screen_init_compute(struct nvc0_screen
*screen
)
477 screen
->base
.base
.get_compute_param
= nvc0_screen_get_compute_param
;
479 switch (screen
->base
.device
->chipset
& 0xf0) {
482 /* Using COMPUTE has weird effects on 3D state, we need to
483 * investigate this further before enabling it by default.
485 if (debug_get_bool_option("NVC0_COMPUTE", FALSE
))
486 return nvc0_screen_compute_setup(screen
, screen
->base
.pushbuf
);
490 return nve4_screen_compute_setup(screen
, screen
->base
.pushbuf
);
497 nvc0_screen_resize_tls_area(struct nvc0_screen
*screen
,
498 uint32_t lpos
, uint32_t lneg
, uint32_t cstack
)
500 struct nouveau_bo
*bo
= NULL
;
502 uint64_t size
= (lpos
+ lneg
) * 32 + cstack
;
504 if (size
>= (1 << 20)) {
505 NOUVEAU_ERR("requested TLS size too large: 0x%"PRIx64
"\n", size
);
509 size
*= (screen
->base
.device
->chipset
>= 0xe0) ? 64 : 48; /* max warps */
510 size
= align(size
, 0x8000);
511 size
*= screen
->mp_count
;
513 size
= align(size
, 1 << 17);
515 ret
= nouveau_bo_new(screen
->base
.device
, NOUVEAU_BO_VRAM
, 1 << 17, size
,
518 NOUVEAU_ERR("failed to allocate TLS area, size: 0x%"PRIx64
"\n", size
);
521 nouveau_bo_ref(NULL
, &screen
->tls
);
526 #define FAIL_SCREEN_INIT(str, err) \
528 NOUVEAU_ERR(str, err); \
529 nvc0_screen_destroy(pscreen); \
534 nvc0_screen_create(struct nouveau_device
*dev
)
536 struct nvc0_screen
*screen
;
537 struct pipe_screen
*pscreen
;
538 struct nouveau_object
*chan
;
539 struct nouveau_pushbuf
*push
;
544 union nouveau_bo_config mm_config
;
546 switch (dev
->chipset
& ~0xf) {
556 screen
= CALLOC_STRUCT(nvc0_screen
);
559 pscreen
= &screen
->base
.base
;
561 ret
= nouveau_screen_init(&screen
->base
, dev
);
563 nvc0_screen_destroy(pscreen
);
566 chan
= screen
->base
.channel
;
567 push
= screen
->base
.pushbuf
;
568 push
->user_priv
= screen
;
571 screen
->base
.vidmem_bindings
|= PIPE_BIND_CONSTANT_BUFFER
|
572 PIPE_BIND_VERTEX_BUFFER
| PIPE_BIND_INDEX_BUFFER
;
573 screen
->base
.sysmem_bindings
|=
574 PIPE_BIND_VERTEX_BUFFER
| PIPE_BIND_INDEX_BUFFER
;
576 pscreen
->destroy
= nvc0_screen_destroy
;
577 pscreen
->context_create
= nvc0_create
;
578 pscreen
->is_format_supported
= nvc0_screen_is_format_supported
;
579 pscreen
->get_param
= nvc0_screen_get_param
;
580 pscreen
->get_shader_param
= nvc0_screen_get_shader_param
;
581 pscreen
->get_paramf
= nvc0_screen_get_paramf
;
582 pscreen
->get_driver_query_info
= nvc0_screen_get_driver_query_info
;
584 nvc0_screen_init_resource_functions(pscreen
);
586 screen
->base
.base
.get_video_param
= nouveau_vp3_screen_get_video_param
;
587 screen
->base
.base
.is_video_format_supported
= nouveau_vp3_screen_video_supported
;
589 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_GART
| NOUVEAU_BO_MAP
, 0, 4096, NULL
,
593 nouveau_bo_map(screen
->fence
.bo
, 0, NULL
);
594 screen
->fence
.map
= screen
->fence
.bo
->map
;
595 screen
->base
.fence
.emit
= nvc0_screen_fence_emit
;
596 screen
->base
.fence
.update
= nvc0_screen_fence_update
;
598 switch (dev
->chipset
& 0xf0) {
600 obj_class
= NVF0_P2MF_CLASS
;
603 obj_class
= NVE4_P2MF_CLASS
;
606 obj_class
= NVC0_M2MF_CLASS
;
609 ret
= nouveau_object_new(chan
, 0xbeef323f, obj_class
, NULL
, 0,
612 FAIL_SCREEN_INIT("Error allocating PGRAPH context for M2MF: %d\n", ret
);
614 BEGIN_NVC0(push
, SUBC_M2MF(NV01_SUBCHAN_OBJECT
), 1);
615 PUSH_DATA (push
, screen
->m2mf
->oclass
);
616 if (screen
->m2mf
->oclass
== NVE4_P2MF_CLASS
) {
617 BEGIN_NVC0(push
, SUBC_COPY(NV01_SUBCHAN_OBJECT
), 1);
618 PUSH_DATA (push
, 0xa0b5);
621 ret
= nouveau_object_new(chan
, 0xbeef902d, NVC0_2D_CLASS
, NULL
, 0,
624 FAIL_SCREEN_INIT("Error allocating PGRAPH context for 2D: %d\n", ret
);
626 BEGIN_NVC0(push
, SUBC_2D(NV01_SUBCHAN_OBJECT
), 1);
627 PUSH_DATA (push
, screen
->eng2d
->oclass
);
628 BEGIN_NVC0(push
, NVC0_2D(SINGLE_GPC
), 1);
630 BEGIN_NVC0(push
, NVC0_2D(OPERATION
), 1);
631 PUSH_DATA (push
, NVC0_2D_OPERATION_SRCCOPY
);
632 BEGIN_NVC0(push
, NVC0_2D(CLIP_ENABLE
), 1);
634 BEGIN_NVC0(push
, NVC0_2D(COLOR_KEY_ENABLE
), 1);
636 BEGIN_NVC0(push
, SUBC_2D(0x0884), 1);
637 PUSH_DATA (push
, 0x3f);
638 BEGIN_NVC0(push
, SUBC_2D(0x0888), 1);
641 BEGIN_NVC0(push
, SUBC_2D(NVC0_GRAPH_NOTIFY_ADDRESS_HIGH
), 2);
642 PUSH_DATAh(push
, screen
->fence
.bo
->offset
+ 16);
643 PUSH_DATA (push
, screen
->fence
.bo
->offset
+ 16);
645 switch (dev
->chipset
& 0xf0) {
647 obj_class
= NVF0_3D_CLASS
;
650 obj_class
= NVE4_3D_CLASS
;
655 switch (dev
->chipset
) {
658 obj_class
= NVC8_3D_CLASS
;
661 obj_class
= NVC1_3D_CLASS
;
664 obj_class
= NVC0_3D_CLASS
;
669 ret
= nouveau_object_new(chan
, 0xbeef003d, obj_class
, NULL
, 0,
672 FAIL_SCREEN_INIT("Error allocating PGRAPH context for 3D: %d\n", ret
);
673 screen
->base
.class_3d
= obj_class
;
675 BEGIN_NVC0(push
, SUBC_3D(NV01_SUBCHAN_OBJECT
), 1);
676 PUSH_DATA (push
, screen
->eng3d
->oclass
);
678 BEGIN_NVC0(push
, NVC0_3D(COND_MODE
), 1);
679 PUSH_DATA (push
, NVC0_3D_COND_MODE_ALWAYS
);
681 if (debug_get_bool_option("NOUVEAU_SHADER_WATCHDOG", TRUE
)) {
682 /* kill shaders after about 1 second (at 100 MHz) */
683 BEGIN_NVC0(push
, NVC0_3D(WATCHDOG_TIMER
), 1);
684 PUSH_DATA (push
, 0x17);
687 IMMED_NVC0(push
, NVC0_3D(ZETA_COMP_ENABLE
), dev
->drm_version
>= 0x01000101);
688 BEGIN_NVC0(push
, NVC0_3D(RT_COMP_ENABLE(0)), 8);
689 for (i
= 0; i
< 8; ++i
)
690 PUSH_DATA(push
, dev
->drm_version
>= 0x01000101);
692 BEGIN_NVC0(push
, NVC0_3D(RT_CONTROL
), 1);
695 BEGIN_NVC0(push
, NVC0_3D(CSAA_ENABLE
), 1);
697 BEGIN_NVC0(push
, NVC0_3D(MULTISAMPLE_ENABLE
), 1);
699 BEGIN_NVC0(push
, NVC0_3D(MULTISAMPLE_MODE
), 1);
700 PUSH_DATA (push
, NVC0_3D_MULTISAMPLE_MODE_MS1
);
701 BEGIN_NVC0(push
, NVC0_3D(MULTISAMPLE_CTRL
), 1);
703 BEGIN_NVC0(push
, NVC0_3D(LINE_WIDTH_SEPARATE
), 1);
705 BEGIN_NVC0(push
, NVC0_3D(LINE_LAST_PIXEL
), 1);
707 BEGIN_NVC0(push
, NVC0_3D(BLEND_SEPARATE_ALPHA
), 1);
709 BEGIN_NVC0(push
, NVC0_3D(BLEND_ENABLE_COMMON
), 1);
711 if (screen
->eng3d
->oclass
< NVE4_3D_CLASS
) {
712 BEGIN_NVC0(push
, NVC0_3D(TEX_MISC
), 1);
713 PUSH_DATA (push
, NVC0_3D_TEX_MISC_SEAMLESS_CUBE_MAP
);
715 BEGIN_NVC0(push
, NVE4_3D(TEX_CB_INDEX
), 1);
716 PUSH_DATA (push
, 15);
718 BEGIN_NVC0(push
, NVC0_3D(CALL_LIMIT_LOG
), 1);
719 PUSH_DATA (push
, 8); /* 128 */
720 BEGIN_NVC0(push
, NVC0_3D(ZCULL_STATCTRS_ENABLE
), 1);
722 if (screen
->eng3d
->oclass
>= NVC1_3D_CLASS
) {
723 BEGIN_NVC0(push
, NVC0_3D(CACHE_SPLIT
), 1);
724 PUSH_DATA (push
, NVC0_3D_CACHE_SPLIT_48K_SHARED_16K_L1
);
727 nvc0_magic_3d_init(push
, screen
->eng3d
->oclass
);
729 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 17, 1 << 20, NULL
,
734 /* XXX: getting a page fault at the end of the code buffer every few
735 * launches, don't use the last 256 bytes to work around them - prefetch ?
737 nouveau_heap_init(&screen
->text_heap
, 0, (1 << 20) - 0x100);
739 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 12, 6 << 16, NULL
,
740 &screen
->uniform_bo
);
744 for (i
= 0; i
< 5; ++i
) {
745 /* TIC and TSC entries for each unit (nve4+ only) */
746 /* auxiliary constants (6 user clip planes, base instance id) */
747 BEGIN_NVC0(push
, NVC0_3D(CB_SIZE
), 3);
748 PUSH_DATA (push
, 512);
749 PUSH_DATAh(push
, screen
->uniform_bo
->offset
+ (5 << 16) + (i
<< 9));
750 PUSH_DATA (push
, screen
->uniform_bo
->offset
+ (5 << 16) + (i
<< 9));
751 BEGIN_NVC0(push
, NVC0_3D(CB_BIND(i
)), 1);
752 PUSH_DATA (push
, (15 << 4) | 1);
753 if (screen
->eng3d
->oclass
>= NVE4_3D_CLASS
) {
755 BEGIN_1IC0(push
, NVC0_3D(CB_POS
), 9);
757 for (j
= 0; j
< 8; ++j
)
760 BEGIN_NVC0(push
, NVC0_3D(TEX_LIMITS(i
)), 1);
761 PUSH_DATA (push
, 0x54);
764 BEGIN_NVC0(push
, NVC0_3D(LINKED_TSC
), 1);
767 /* return { 0.0, 0.0, 0.0, 0.0 } for out-of-bounds vtxbuf access */
768 BEGIN_NVC0(push
, NVC0_3D(CB_SIZE
), 3);
769 PUSH_DATA (push
, 256);
770 PUSH_DATAh(push
, screen
->uniform_bo
->offset
+ (5 << 16) + (6 << 9));
771 PUSH_DATA (push
, screen
->uniform_bo
->offset
+ (5 << 16) + (6 << 9));
772 BEGIN_1IC0(push
, NVC0_3D(CB_POS
), 5);
774 PUSH_DATAf(push
, 0.0f
);
775 PUSH_DATAf(push
, 0.0f
);
776 PUSH_DATAf(push
, 0.0f
);
777 PUSH_DATAf(push
, 0.0f
);
778 BEGIN_NVC0(push
, NVC0_3D(VERTEX_RUNOUT_ADDRESS_HIGH
), 2);
779 PUSH_DATAh(push
, screen
->uniform_bo
->offset
+ (5 << 16) + (6 << 9));
780 PUSH_DATA (push
, screen
->uniform_bo
->offset
+ (5 << 16) + (6 << 9));
782 if (dev
->drm_version
>= 0x01000101) {
783 ret
= nouveau_getparam(dev
, NOUVEAU_GETPARAM_GRAPH_UNITS
, &value
);
785 NOUVEAU_ERR("NOUVEAU_GETPARAM_GRAPH_UNITS failed.\n");
789 if (dev
->chipset
>= 0xe0 && dev
->chipset
< 0xf0)
790 value
= (8 << 8) | 4;
792 value
= (16 << 8) | 4;
794 screen
->mp_count
= value
>> 8;
795 screen
->mp_count_compute
= screen
->mp_count
;
797 nvc0_screen_resize_tls_area(screen
, 128 * 16, 0, 0x200);
799 BEGIN_NVC0(push
, NVC0_3D(CODE_ADDRESS_HIGH
), 2);
800 PUSH_DATAh(push
, screen
->text
->offset
);
801 PUSH_DATA (push
, screen
->text
->offset
);
802 BEGIN_NVC0(push
, NVC0_3D(TEMP_ADDRESS_HIGH
), 4);
803 PUSH_DATAh(push
, screen
->tls
->offset
);
804 PUSH_DATA (push
, screen
->tls
->offset
);
805 PUSH_DATA (push
, screen
->tls
->size
>> 32);
806 PUSH_DATA (push
, screen
->tls
->size
);
807 BEGIN_NVC0(push
, NVC0_3D(WARP_TEMP_ALLOC
), 1);
809 BEGIN_NVC0(push
, NVC0_3D(LOCAL_BASE
), 1);
812 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 17, 1 << 20, NULL
,
813 &screen
->poly_cache
);
817 BEGIN_NVC0(push
, NVC0_3D(VERTEX_QUARANTINE_ADDRESS_HIGH
), 3);
818 PUSH_DATAh(push
, screen
->poly_cache
->offset
);
819 PUSH_DATA (push
, screen
->poly_cache
->offset
);
822 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 17, 1 << 17, NULL
,
827 BEGIN_NVC0(push
, NVC0_3D(TIC_ADDRESS_HIGH
), 3);
828 PUSH_DATAh(push
, screen
->txc
->offset
);
829 PUSH_DATA (push
, screen
->txc
->offset
);
830 PUSH_DATA (push
, NVC0_TIC_MAX_ENTRIES
- 1);
832 BEGIN_NVC0(push
, NVC0_3D(TSC_ADDRESS_HIGH
), 3);
833 PUSH_DATAh(push
, screen
->txc
->offset
+ 65536);
834 PUSH_DATA (push
, screen
->txc
->offset
+ 65536);
835 PUSH_DATA (push
, NVC0_TSC_MAX_ENTRIES
- 1);
837 BEGIN_NVC0(push
, NVC0_3D(SCREEN_Y_CONTROL
), 1);
839 BEGIN_NVC0(push
, NVC0_3D(WINDOW_OFFSET_X
), 2);
842 BEGIN_NVC0(push
, NVC0_3D(ZCULL_REGION
), 1); /* deactivate ZCULL */
843 PUSH_DATA (push
, 0x3f);
845 BEGIN_NVC0(push
, NVC0_3D(CLIP_RECTS_MODE
), 1);
846 PUSH_DATA (push
, NVC0_3D_CLIP_RECTS_MODE_INSIDE_ANY
);
847 BEGIN_NVC0(push
, NVC0_3D(CLIP_RECT_HORIZ(0)), 8 * 2);
848 for (i
= 0; i
< 8 * 2; ++i
)
850 BEGIN_NVC0(push
, NVC0_3D(CLIP_RECTS_EN
), 1);
852 BEGIN_NVC0(push
, NVC0_3D(CLIPID_ENABLE
), 1);
855 /* neither scissors, viewport nor stencil mask should affect clears */
856 BEGIN_NVC0(push
, NVC0_3D(CLEAR_FLAGS
), 1);
859 BEGIN_NVC0(push
, NVC0_3D(VIEWPORT_TRANSFORM_EN
), 1);
861 BEGIN_NVC0(push
, NVC0_3D(DEPTH_RANGE_NEAR(0)), 2);
862 PUSH_DATAf(push
, 0.0f
);
863 PUSH_DATAf(push
, 1.0f
);
864 BEGIN_NVC0(push
, NVC0_3D(VIEW_VOLUME_CLIP_CTRL
), 1);
865 PUSH_DATA (push
, NVC0_3D_VIEW_VOLUME_CLIP_CTRL_UNK1_UNK1
);
867 /* We use scissors instead of exact view volume clipping,
868 * so they're always enabled.
870 BEGIN_NVC0(push
, NVC0_3D(SCISSOR_ENABLE(0)), 3);
872 PUSH_DATA (push
, 8192 << 16);
873 PUSH_DATA (push
, 8192 << 16);
875 #define MK_MACRO(m, n) i = nvc0_graph_set_macro(screen, m, i, sizeof(n), n);
878 MK_MACRO(NVC0_3D_MACRO_VERTEX_ARRAY_PER_INSTANCE
, nvc0_9097_per_instance_bf
);
879 MK_MACRO(NVC0_3D_MACRO_BLEND_ENABLES
, nvc0_9097_blend_enables
);
880 MK_MACRO(NVC0_3D_MACRO_VERTEX_ARRAY_SELECT
, nvc0_9097_vertex_array_select
);
881 MK_MACRO(NVC0_3D_MACRO_TEP_SELECT
, nvc0_9097_tep_select
);
882 MK_MACRO(NVC0_3D_MACRO_GP_SELECT
, nvc0_9097_gp_select
);
883 MK_MACRO(NVC0_3D_MACRO_POLYGON_MODE_FRONT
, nvc0_9097_poly_mode_front
);
884 MK_MACRO(NVC0_3D_MACRO_POLYGON_MODE_BACK
, nvc0_9097_poly_mode_back
);
886 BEGIN_NVC0(push
, NVC0_3D(RASTERIZE_ENABLE
), 1);
888 BEGIN_NVC0(push
, NVC0_3D(RT_SEPARATE_FRAG_DATA
), 1);
890 BEGIN_NVC0(push
, NVC0_3D(MACRO_GP_SELECT
), 1);
891 PUSH_DATA (push
, 0x40);
892 BEGIN_NVC0(push
, NVC0_3D(LAYER
), 1);
894 BEGIN_NVC0(push
, NVC0_3D(MACRO_TEP_SELECT
), 1);
895 PUSH_DATA (push
, 0x30);
896 BEGIN_NVC0(push
, NVC0_3D(PATCH_VERTICES
), 1);
898 BEGIN_NVC0(push
, NVC0_3D(SP_SELECT(2)), 1);
899 PUSH_DATA (push
, 0x20);
900 BEGIN_NVC0(push
, NVC0_3D(SP_SELECT(0)), 1);
901 PUSH_DATA (push
, 0x00);
903 BEGIN_NVC0(push
, NVC0_3D(POINT_COORD_REPLACE
), 1);
905 BEGIN_NVC0(push
, NVC0_3D(POINT_RASTER_RULES
), 1);
906 PUSH_DATA (push
, NVC0_3D_POINT_RASTER_RULES_OGL
);
908 IMMED_NVC0(push
, NVC0_3D(EDGEFLAG
), 1);
910 if (nvc0_screen_init_compute(screen
))
915 screen
->tic
.entries
= CALLOC(4096, sizeof(void *));
916 screen
->tsc
.entries
= screen
->tic
.entries
+ 2048;
918 mm_config
.nvc0
.tile_mode
= 0;
919 mm_config
.nvc0
.memtype
= 0xfe0;
920 screen
->mm_VRAM_fe0
= nouveau_mm_create(dev
, NOUVEAU_BO_VRAM
, &mm_config
);
922 if (!nvc0_blitter_create(screen
))
925 nouveau_fence_new(&screen
->base
, &screen
->base
.fence
.current
, FALSE
);
930 nvc0_screen_destroy(pscreen
);
935 nvc0_screen_tic_alloc(struct nvc0_screen
*screen
, void *entry
)
937 int i
= screen
->tic
.next
;
939 while (screen
->tic
.lock
[i
/ 32] & (1 << (i
% 32)))
940 i
= (i
+ 1) & (NVC0_TIC_MAX_ENTRIES
- 1);
942 screen
->tic
.next
= (i
+ 1) & (NVC0_TIC_MAX_ENTRIES
- 1);
944 if (screen
->tic
.entries
[i
])
945 nv50_tic_entry(screen
->tic
.entries
[i
])->id
= -1;
947 screen
->tic
.entries
[i
] = entry
;
952 nvc0_screen_tsc_alloc(struct nvc0_screen
*screen
, void *entry
)
954 int i
= screen
->tsc
.next
;
956 while (screen
->tsc
.lock
[i
/ 32] & (1 << (i
% 32)))
957 i
= (i
+ 1) & (NVC0_TSC_MAX_ENTRIES
- 1);
959 screen
->tsc
.next
= (i
+ 1) & (NVC0_TSC_MAX_ENTRIES
- 1);
961 if (screen
->tsc
.entries
[i
])
962 nv50_tsc_entry(screen
->tsc
.entries
[i
])->id
= -1;
964 screen
->tsc
.entries
[i
] = entry
;