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 "nv50/nv50_context.h"
28 #include "nv50/nv50_screen.h"
30 #include "nouveau_vp3_video.h"
32 #include "nv_object.xml.h"
35 #ifndef NOUVEAU_GETPARAM_GRAPH_UNITS
36 # define NOUVEAU_GETPARAM_GRAPH_UNITS 13
39 /* affected by LOCAL_WARPS_LOG_ALLOC / LOCAL_WARPS_NO_CLAMP */
40 #define LOCAL_WARPS_ALLOC 32
41 /* affected by STACK_WARPS_LOG_ALLOC / STACK_WARPS_NO_CLAMP */
42 #define STACK_WARPS_ALLOC 32
44 #define THREADS_IN_WARP 32
46 #define ONE_TEMP_SIZE (4/*vector*/ * sizeof(float))
49 nv50_screen_is_format_supported(struct pipe_screen
*pscreen
,
50 enum pipe_format format
,
51 enum pipe_texture_target target
,
52 unsigned sample_count
,
57 if (!(0x117 & (1 << sample_count
))) /* 0, 1, 2, 4 or 8 */
59 if (sample_count
== 8 && util_format_get_blocksizebits(format
) >= 128)
62 if (!util_format_is_supported(format
, bindings
))
66 case PIPE_FORMAT_Z16_UNORM
:
67 if (nv50_screen(pscreen
)->tesla
->oclass
< NVA0_3D_CLASS
)
74 /* transfers & shared are always supported */
75 bindings
&= ~(PIPE_BIND_TRANSFER_READ
|
76 PIPE_BIND_TRANSFER_WRITE
|
79 return (nv50_format_table
[format
].usage
& bindings
) == bindings
;
83 nv50_screen_get_param(struct pipe_screen
*pscreen
, enum pipe_cap param
)
85 const uint16_t class_3d
= nouveau_screen(pscreen
)->class_3d
;
88 case PIPE_CAP_MAX_COMBINED_SAMPLERS
:
90 case PIPE_CAP_MAX_TEXTURE_2D_LEVELS
:
92 case PIPE_CAP_MAX_TEXTURE_3D_LEVELS
:
94 case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS
:
96 case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS
:
98 case PIPE_CAP_MIN_TEXEL_OFFSET
:
100 case PIPE_CAP_MAX_TEXEL_OFFSET
:
102 case PIPE_CAP_TEXTURE_MIRROR_CLAMP
:
103 case PIPE_CAP_TEXTURE_SWIZZLE
:
104 case PIPE_CAP_TEXTURE_SHADOW_MAP
:
105 case PIPE_CAP_NPOT_TEXTURES
:
106 case PIPE_CAP_MIXED_FRAMEBUFFER_SIZES
:
107 case PIPE_CAP_ANISOTROPIC_FILTER
:
108 case PIPE_CAP_SCALED_RESOLVE
:
109 case PIPE_CAP_TEXTURE_BUFFER_OBJECTS
:
111 case PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE
:
113 case PIPE_CAP_SEAMLESS_CUBE_MAP
:
114 return nv50_screen(pscreen
)->tesla
->oclass
>= NVA0_3D_CLASS
;
115 case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE
:
117 case PIPE_CAP_CUBE_MAP_ARRAY
:
120 return nv50_screen(pscreen)->tesla->oclass >= NVA3_3D_CLASS;
122 case PIPE_CAP_TWO_SIDED_STENCIL
:
123 case PIPE_CAP_DEPTH_CLIP_DISABLE
:
124 case PIPE_CAP_POINT_SPRITE
:
128 case PIPE_CAP_GLSL_FEATURE_LEVEL
:
130 case PIPE_CAP_MAX_RENDER_TARGETS
:
132 case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS
:
134 case PIPE_CAP_FRAGMENT_COLOR_CLAMPED
:
135 case PIPE_CAP_VERTEX_COLOR_UNCLAMPED
:
136 case PIPE_CAP_VERTEX_COLOR_CLAMPED
:
138 case PIPE_CAP_QUERY_TIMESTAMP
:
139 case PIPE_CAP_QUERY_TIME_ELAPSED
:
140 case PIPE_CAP_OCCLUSION_QUERY
:
142 case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS
:
144 case PIPE_CAP_MAX_STREAM_OUTPUT_INTERLEAVED_COMPONENTS
:
145 case PIPE_CAP_MAX_STREAM_OUTPUT_SEPARATE_COMPONENTS
:
147 case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME
:
148 return (class_3d
>= NVA0_3D_CLASS
) ? 1 : 0;
149 case PIPE_CAP_BLEND_EQUATION_SEPARATE
:
150 case PIPE_CAP_INDEP_BLEND_ENABLE
:
152 case PIPE_CAP_INDEP_BLEND_FUNC
:
153 return nv50_screen(pscreen
)->tesla
->oclass
>= NVA3_3D_CLASS
;
154 case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT
:
155 case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER
:
157 case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT
:
158 case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER
:
160 case PIPE_CAP_SHADER_STENCIL_EXPORT
:
162 case PIPE_CAP_PRIMITIVE_RESTART
:
163 case PIPE_CAP_TGSI_INSTANCEID
:
164 case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR
:
165 case PIPE_CAP_MIXED_COLORBUFFER_FORMATS
:
166 case PIPE_CAP_CONDITIONAL_RENDER
:
167 case PIPE_CAP_TEXTURE_BARRIER
:
168 case PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION
:
169 case PIPE_CAP_START_INSTANCE
:
171 case PIPE_CAP_TGSI_CAN_COMPACT_CONSTANTS
:
172 return 0; /* state trackers will know better */
173 case PIPE_CAP_USER_CONSTANT_BUFFERS
:
174 case PIPE_CAP_USER_INDEX_BUFFERS
:
175 case PIPE_CAP_USER_VERTEX_BUFFERS
:
177 case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT
:
179 case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT
:
180 return 1; /* 256 for binding as RT, but that's not possible in GL */
181 case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT
:
182 return NOUVEAU_MIN_BUFFER_MAP_ALIGN
;
183 case PIPE_CAP_VERTEX_BUFFER_OFFSET_4BYTE_ALIGNED_ONLY
:
184 case PIPE_CAP_VERTEX_BUFFER_STRIDE_4BYTE_ALIGNED_ONLY
:
185 case PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY
:
186 case PIPE_CAP_TGSI_TEXCOORD
:
187 case PIPE_CAP_TEXTURE_MULTISAMPLE
:
189 case PIPE_CAP_PREFER_BLIT_BASED_TEXTURE_TRANSFER
:
191 case PIPE_CAP_QUERY_PIPELINE_STATISTICS
:
193 case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK
:
194 return PIPE_QUIRK_TEXTURE_BORDER_COLOR_SWIZZLE_NV50
;
195 case PIPE_CAP_ENDIANNESS
:
196 return PIPE_ENDIAN_LITTLE
;
198 NOUVEAU_ERR("unknown PIPE_CAP %d\n", param
);
204 nv50_screen_get_shader_param(struct pipe_screen
*pscreen
, unsigned shader
,
205 enum pipe_shader_cap param
)
208 case PIPE_SHADER_VERTEX
:
209 case PIPE_SHADER_GEOMETRY
:
210 case PIPE_SHADER_FRAGMENT
:
217 case PIPE_SHADER_CAP_MAX_INSTRUCTIONS
:
218 case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS
:
219 case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS
:
220 case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS
:
222 case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH
:
224 case PIPE_SHADER_CAP_MAX_INPUTS
:
225 if (shader
== PIPE_SHADER_VERTEX
)
228 case PIPE_SHADER_CAP_MAX_CONSTS
:
230 case PIPE_SHADER_CAP_MAX_CONST_BUFFERS
:
231 return NV50_MAX_PIPE_CONSTBUFS
;
232 case PIPE_SHADER_CAP_MAX_ADDRS
:
234 case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR
:
235 case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR
:
236 return shader
!= PIPE_SHADER_FRAGMENT
;
237 case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR
:
238 case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR
:
240 case PIPE_SHADER_CAP_MAX_PREDS
:
242 case PIPE_SHADER_CAP_MAX_TEMPS
:
243 return nv50_screen(pscreen
)->max_tls_space
/ ONE_TEMP_SIZE
;
244 case PIPE_SHADER_CAP_TGSI_CONT_SUPPORTED
:
246 case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED
:
248 case PIPE_SHADER_CAP_SUBROUTINES
:
249 return 0; /* please inline, or provide function declarations */
250 case PIPE_SHADER_CAP_INTEGERS
:
252 case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS
:
253 /* The chip could handle more sampler views than samplers */
254 case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS
:
255 return MIN2(32, PIPE_MAX_SAMPLERS
);
257 NOUVEAU_ERR("unknown PIPE_SHADER_CAP %d\n", param
);
263 nv50_screen_get_paramf(struct pipe_screen
*pscreen
, enum pipe_capf param
)
266 case PIPE_CAPF_MAX_LINE_WIDTH
:
267 case PIPE_CAPF_MAX_LINE_WIDTH_AA
:
269 case PIPE_CAPF_MAX_POINT_WIDTH
:
270 case PIPE_CAPF_MAX_POINT_WIDTH_AA
:
272 case PIPE_CAPF_MAX_TEXTURE_ANISOTROPY
:
274 case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS
:
277 NOUVEAU_ERR("unknown PIPE_CAP %d\n", param
);
283 nv50_screen_destroy(struct pipe_screen
*pscreen
)
285 struct nv50_screen
*screen
= nv50_screen(pscreen
);
287 if (screen
->base
.fence
.current
) {
288 nouveau_fence_wait(screen
->base
.fence
.current
);
289 nouveau_fence_ref (NULL
, &screen
->base
.fence
.current
);
291 if (screen
->base
.pushbuf
)
292 screen
->base
.pushbuf
->user_priv
= NULL
;
295 nv50_blitter_destroy(screen
);
297 nouveau_bo_ref(NULL
, &screen
->code
);
298 nouveau_bo_ref(NULL
, &screen
->tls_bo
);
299 nouveau_bo_ref(NULL
, &screen
->stack_bo
);
300 nouveau_bo_ref(NULL
, &screen
->txc
);
301 nouveau_bo_ref(NULL
, &screen
->uniforms
);
302 nouveau_bo_ref(NULL
, &screen
->fence
.bo
);
304 nouveau_heap_destroy(&screen
->vp_code_heap
);
305 nouveau_heap_destroy(&screen
->gp_code_heap
);
306 nouveau_heap_destroy(&screen
->fp_code_heap
);
308 FREE(screen
->tic
.entries
);
310 nouveau_object_del(&screen
->tesla
);
311 nouveau_object_del(&screen
->eng2d
);
312 nouveau_object_del(&screen
->m2mf
);
313 nouveau_object_del(&screen
->sync
);
315 nouveau_screen_fini(&screen
->base
);
321 nv50_screen_fence_emit(struct pipe_screen
*pscreen
, u32
*sequence
)
323 struct nv50_screen
*screen
= nv50_screen(pscreen
);
324 struct nouveau_pushbuf
*push
= screen
->base
.pushbuf
;
326 /* we need to do it after possible flush in MARK_RING */
327 *sequence
= ++screen
->base
.fence
.sequence
;
329 PUSH_DATA (push
, NV50_FIFO_PKHDR(NV50_3D(QUERY_ADDRESS_HIGH
), 4));
330 PUSH_DATAh(push
, screen
->fence
.bo
->offset
);
331 PUSH_DATA (push
, screen
->fence
.bo
->offset
);
332 PUSH_DATA (push
, *sequence
);
333 PUSH_DATA (push
, NV50_3D_QUERY_GET_MODE_WRITE_UNK0
|
334 NV50_3D_QUERY_GET_UNK4
|
335 NV50_3D_QUERY_GET_UNIT_CROP
|
336 NV50_3D_QUERY_GET_TYPE_QUERY
|
337 NV50_3D_QUERY_GET_QUERY_SELECT_ZERO
|
338 NV50_3D_QUERY_GET_SHORT
);
342 nv50_screen_fence_update(struct pipe_screen
*pscreen
)
344 return nv50_screen(pscreen
)->fence
.map
[0];
348 nv50_screen_init_hwctx(struct nv50_screen
*screen
)
350 struct nouveau_pushbuf
*push
= screen
->base
.pushbuf
;
351 struct nv04_fifo
*fifo
;
354 fifo
= (struct nv04_fifo
*)screen
->base
.channel
->data
;
356 BEGIN_NV04(push
, SUBC_M2MF(NV01_SUBCHAN_OBJECT
), 1);
357 PUSH_DATA (push
, screen
->m2mf
->handle
);
358 BEGIN_NV04(push
, SUBC_M2MF(NV03_M2MF_DMA_NOTIFY
), 3);
359 PUSH_DATA (push
, screen
->sync
->handle
);
360 PUSH_DATA (push
, fifo
->vram
);
361 PUSH_DATA (push
, fifo
->vram
);
363 BEGIN_NV04(push
, SUBC_2D(NV01_SUBCHAN_OBJECT
), 1);
364 PUSH_DATA (push
, screen
->eng2d
->handle
);
365 BEGIN_NV04(push
, NV50_2D(DMA_NOTIFY
), 4);
366 PUSH_DATA (push
, screen
->sync
->handle
);
367 PUSH_DATA (push
, fifo
->vram
);
368 PUSH_DATA (push
, fifo
->vram
);
369 PUSH_DATA (push
, fifo
->vram
);
370 BEGIN_NV04(push
, NV50_2D(OPERATION
), 1);
371 PUSH_DATA (push
, NV50_2D_OPERATION_SRCCOPY
);
372 BEGIN_NV04(push
, NV50_2D(CLIP_ENABLE
), 1);
374 BEGIN_NV04(push
, NV50_2D(COLOR_KEY_ENABLE
), 1);
376 BEGIN_NV04(push
, SUBC_2D(0x0888), 1);
379 BEGIN_NV04(push
, SUBC_3D(NV01_SUBCHAN_OBJECT
), 1);
380 PUSH_DATA (push
, screen
->tesla
->handle
);
382 BEGIN_NV04(push
, NV50_3D(COND_MODE
), 1);
383 PUSH_DATA (push
, NV50_3D_COND_MODE_ALWAYS
);
385 BEGIN_NV04(push
, NV50_3D(DMA_NOTIFY
), 1);
386 PUSH_DATA (push
, screen
->sync
->handle
);
387 BEGIN_NV04(push
, NV50_3D(DMA_ZETA
), 11);
388 for (i
= 0; i
< 11; ++i
)
389 PUSH_DATA(push
, fifo
->vram
);
390 BEGIN_NV04(push
, NV50_3D(DMA_COLOR(0)), NV50_3D_DMA_COLOR__LEN
);
391 for (i
= 0; i
< NV50_3D_DMA_COLOR__LEN
; ++i
)
392 PUSH_DATA(push
, fifo
->vram
);
394 BEGIN_NV04(push
, NV50_3D(REG_MODE
), 1);
395 PUSH_DATA (push
, NV50_3D_REG_MODE_STRIPED
);
396 BEGIN_NV04(push
, NV50_3D(UNK1400_LANES
), 1);
397 PUSH_DATA (push
, 0xf);
399 if (debug_get_bool_option("NOUVEAU_SHADER_WATCHDOG", TRUE
)) {
400 BEGIN_NV04(push
, NV50_3D(WATCHDOG_TIMER
), 1);
401 PUSH_DATA (push
, 0x18);
404 BEGIN_NV04(push
, NV50_3D(RT_CONTROL
), 1);
407 BEGIN_NV04(push
, NV50_3D(CSAA_ENABLE
), 1);
409 BEGIN_NV04(push
, NV50_3D(MULTISAMPLE_ENABLE
), 1);
411 BEGIN_NV04(push
, NV50_3D(MULTISAMPLE_MODE
), 1);
412 PUSH_DATA (push
, NV50_3D_MULTISAMPLE_MODE_MS1
);
413 BEGIN_NV04(push
, NV50_3D(MULTISAMPLE_CTRL
), 1);
415 BEGIN_NV04(push
, NV50_3D(LINE_LAST_PIXEL
), 1);
417 BEGIN_NV04(push
, NV50_3D(BLEND_SEPARATE_ALPHA
), 1);
420 if (screen
->tesla
->oclass
>= NVA0_3D_CLASS
) {
421 BEGIN_NV04(push
, SUBC_3D(NVA0_3D_TEX_MISC
), 1);
422 PUSH_DATA (push
, NVA0_3D_TEX_MISC_SEAMLESS_CUBE_MAP
);
425 BEGIN_NV04(push
, NV50_3D(SCREEN_Y_CONTROL
), 1);
427 BEGIN_NV04(push
, NV50_3D(WINDOW_OFFSET_X
), 2);
430 BEGIN_NV04(push
, NV50_3D(ZCULL_REGION
), 1);
431 PUSH_DATA (push
, 0x3f);
433 BEGIN_NV04(push
, NV50_3D(VP_ADDRESS_HIGH
), 2);
434 PUSH_DATAh(push
, screen
->code
->offset
+ (0 << NV50_CODE_BO_SIZE_LOG2
));
435 PUSH_DATA (push
, screen
->code
->offset
+ (0 << NV50_CODE_BO_SIZE_LOG2
));
437 BEGIN_NV04(push
, NV50_3D(FP_ADDRESS_HIGH
), 2);
438 PUSH_DATAh(push
, screen
->code
->offset
+ (1 << NV50_CODE_BO_SIZE_LOG2
));
439 PUSH_DATA (push
, screen
->code
->offset
+ (1 << NV50_CODE_BO_SIZE_LOG2
));
441 BEGIN_NV04(push
, NV50_3D(GP_ADDRESS_HIGH
), 2);
442 PUSH_DATAh(push
, screen
->code
->offset
+ (2 << NV50_CODE_BO_SIZE_LOG2
));
443 PUSH_DATA (push
, screen
->code
->offset
+ (2 << NV50_CODE_BO_SIZE_LOG2
));
445 BEGIN_NV04(push
, NV50_3D(LOCAL_ADDRESS_HIGH
), 3);
446 PUSH_DATAh(push
, screen
->tls_bo
->offset
);
447 PUSH_DATA (push
, screen
->tls_bo
->offset
);
448 PUSH_DATA (push
, util_logbase2(screen
->cur_tls_space
/ 8));
450 BEGIN_NV04(push
, NV50_3D(STACK_ADDRESS_HIGH
), 3);
451 PUSH_DATAh(push
, screen
->stack_bo
->offset
);
452 PUSH_DATA (push
, screen
->stack_bo
->offset
);
455 BEGIN_NV04(push
, NV50_3D(CB_DEF_ADDRESS_HIGH
), 3);
456 PUSH_DATAh(push
, screen
->uniforms
->offset
+ (0 << 16));
457 PUSH_DATA (push
, screen
->uniforms
->offset
+ (0 << 16));
458 PUSH_DATA (push
, (NV50_CB_PVP
<< 16) | 0x0000);
460 BEGIN_NV04(push
, NV50_3D(CB_DEF_ADDRESS_HIGH
), 3);
461 PUSH_DATAh(push
, screen
->uniforms
->offset
+ (1 << 16));
462 PUSH_DATA (push
, screen
->uniforms
->offset
+ (1 << 16));
463 PUSH_DATA (push
, (NV50_CB_PGP
<< 16) | 0x0000);
465 BEGIN_NV04(push
, NV50_3D(CB_DEF_ADDRESS_HIGH
), 3);
466 PUSH_DATAh(push
, screen
->uniforms
->offset
+ (2 << 16));
467 PUSH_DATA (push
, screen
->uniforms
->offset
+ (2 << 16));
468 PUSH_DATA (push
, (NV50_CB_PFP
<< 16) | 0x0000);
470 BEGIN_NV04(push
, NV50_3D(CB_DEF_ADDRESS_HIGH
), 3);
471 PUSH_DATAh(push
, screen
->uniforms
->offset
+ (3 << 16));
472 PUSH_DATA (push
, screen
->uniforms
->offset
+ (3 << 16));
473 PUSH_DATA (push
, (NV50_CB_AUX
<< 16) | 0x0200);
475 BEGIN_NI04(push
, NV50_3D(SET_PROGRAM_CB
), 3);
476 PUSH_DATA (push
, (NV50_CB_AUX
<< 12) | 0xf01);
477 PUSH_DATA (push
, (NV50_CB_AUX
<< 12) | 0xf21);
478 PUSH_DATA (push
, (NV50_CB_AUX
<< 12) | 0xf31);
480 /* return { 0.0, 0.0, 0.0, 0.0 } on out-of-bounds vtxbuf access */
481 BEGIN_NV04(push
, NV50_3D(CB_ADDR
), 1);
482 PUSH_DATA (push
, ((1 << 9) << 6) | NV50_CB_AUX
);
483 BEGIN_NI04(push
, NV50_3D(CB_DATA(0)), 4);
484 PUSH_DATAf(push
, 0.0f
);
485 PUSH_DATAf(push
, 0.0f
);
486 PUSH_DATAf(push
, 0.0f
);
487 PUSH_DATAf(push
, 0.0f
);
488 BEGIN_NV04(push
, NV50_3D(VERTEX_RUNOUT_ADDRESS_HIGH
), 2);
489 PUSH_DATAh(push
, screen
->uniforms
->offset
+ (3 << 16) + (1 << 9));
490 PUSH_DATA (push
, screen
->uniforms
->offset
+ (3 << 16) + (1 << 9));
492 /* max TIC (bits 4:8) & TSC bindings, per program type */
493 for (i
= 0; i
< 3; ++i
) {
494 BEGIN_NV04(push
, NV50_3D(TEX_LIMITS(i
)), 1);
495 PUSH_DATA (push
, 0x54);
498 BEGIN_NV04(push
, NV50_3D(TIC_ADDRESS_HIGH
), 3);
499 PUSH_DATAh(push
, screen
->txc
->offset
);
500 PUSH_DATA (push
, screen
->txc
->offset
);
501 PUSH_DATA (push
, NV50_TIC_MAX_ENTRIES
- 1);
503 BEGIN_NV04(push
, NV50_3D(TSC_ADDRESS_HIGH
), 3);
504 PUSH_DATAh(push
, screen
->txc
->offset
+ 65536);
505 PUSH_DATA (push
, screen
->txc
->offset
+ 65536);
506 PUSH_DATA (push
, NV50_TSC_MAX_ENTRIES
- 1);
508 BEGIN_NV04(push
, NV50_3D(LINKED_TSC
), 1);
511 BEGIN_NV04(push
, NV50_3D(CLIP_RECTS_EN
), 1);
513 BEGIN_NV04(push
, NV50_3D(CLIP_RECTS_MODE
), 1);
514 PUSH_DATA (push
, NV50_3D_CLIP_RECTS_MODE_INSIDE_ANY
);
515 BEGIN_NV04(push
, NV50_3D(CLIP_RECT_HORIZ(0)), 8 * 2);
516 for (i
= 0; i
< 8 * 2; ++i
)
518 BEGIN_NV04(push
, NV50_3D(CLIPID_ENABLE
), 1);
521 BEGIN_NV04(push
, NV50_3D(VIEWPORT_TRANSFORM_EN
), 1);
523 BEGIN_NV04(push
, NV50_3D(DEPTH_RANGE_NEAR(0)), 2);
524 PUSH_DATAf(push
, 0.0f
);
525 PUSH_DATAf(push
, 1.0f
);
527 BEGIN_NV04(push
, NV50_3D(VIEW_VOLUME_CLIP_CTRL
), 1);
528 #ifdef NV50_SCISSORS_CLIPPING
529 PUSH_DATA (push
, 0x0000);
531 PUSH_DATA (push
, 0x1080);
534 BEGIN_NV04(push
, NV50_3D(CLEAR_FLAGS
), 1);
535 PUSH_DATA (push
, NV50_3D_CLEAR_FLAGS_CLEAR_RECT_VIEWPORT
);
537 /* We use scissors instead of exact view volume clipping,
538 * so they're always enabled.
540 BEGIN_NV04(push
, NV50_3D(SCISSOR_ENABLE(0)), 3);
542 PUSH_DATA (push
, 8192 << 16);
543 PUSH_DATA (push
, 8192 << 16);
545 BEGIN_NV04(push
, NV50_3D(RASTERIZE_ENABLE
), 1);
547 BEGIN_NV04(push
, NV50_3D(POINT_RASTER_RULES
), 1);
548 PUSH_DATA (push
, NV50_3D_POINT_RASTER_RULES_OGL
);
549 BEGIN_NV04(push
, NV50_3D(FRAG_COLOR_CLAMP_EN
), 1);
550 PUSH_DATA (push
, 0x11111111);
551 BEGIN_NV04(push
, NV50_3D(EDGEFLAG
), 1);
557 static int nv50_tls_alloc(struct nv50_screen
*screen
, unsigned tls_space
,
560 struct nouveau_device
*dev
= screen
->base
.device
;
563 screen
->cur_tls_space
= util_next_power_of_two(tls_space
/ ONE_TEMP_SIZE
) *
565 if (nouveau_mesa_debug
)
566 debug_printf("allocating space for %u temps\n",
567 util_next_power_of_two(tls_space
/ ONE_TEMP_SIZE
));
568 *tls_size
= screen
->cur_tls_space
* util_next_power_of_two(screen
->TPs
) *
569 screen
->MPsInTP
* LOCAL_WARPS_ALLOC
* THREADS_IN_WARP
;
571 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 16,
572 *tls_size
, NULL
, &screen
->tls_bo
);
574 NOUVEAU_ERR("Failed to allocate local bo: %d\n", ret
);
581 int nv50_tls_realloc(struct nv50_screen
*screen
, unsigned tls_space
)
583 struct nouveau_pushbuf
*push
= screen
->base
.pushbuf
;
587 if (tls_space
< screen
->cur_tls_space
)
589 if (tls_space
> screen
->max_tls_space
) {
590 /* fixable by limiting number of warps (LOCAL_WARPS_LOG_ALLOC /
591 * LOCAL_WARPS_NO_CLAMP) */
592 NOUVEAU_ERR("Unsupported number of temporaries (%u > %u). Fixable if someone cares.\n",
593 (unsigned)(tls_space
/ ONE_TEMP_SIZE
),
594 (unsigned)(screen
->max_tls_space
/ ONE_TEMP_SIZE
));
598 nouveau_bo_ref(NULL
, &screen
->tls_bo
);
599 ret
= nv50_tls_alloc(screen
, tls_space
, &tls_size
);
603 BEGIN_NV04(push
, NV50_3D(LOCAL_ADDRESS_HIGH
), 3);
604 PUSH_DATAh(push
, screen
->tls_bo
->offset
);
605 PUSH_DATA (push
, screen
->tls_bo
->offset
);
606 PUSH_DATA (push
, util_logbase2(screen
->cur_tls_space
/ 8));
612 nv50_screen_create(struct nouveau_device
*dev
)
614 struct nv50_screen
*screen
;
615 struct pipe_screen
*pscreen
;
616 struct nouveau_object
*chan
;
618 uint32_t tesla_class
;
622 screen
= CALLOC_STRUCT(nv50_screen
);
625 pscreen
= &screen
->base
.base
;
627 ret
= nouveau_screen_init(&screen
->base
, dev
);
629 NOUVEAU_ERR("nouveau_screen_init failed: %d\n", ret
);
633 /* TODO: Prevent FIFO prefetch before transfer of index buffers and
634 * admit them to VRAM.
636 screen
->base
.vidmem_bindings
|= PIPE_BIND_CONSTANT_BUFFER
|
637 PIPE_BIND_VERTEX_BUFFER
;
638 screen
->base
.sysmem_bindings
|=
639 PIPE_BIND_VERTEX_BUFFER
| PIPE_BIND_INDEX_BUFFER
;
641 screen
->base
.pushbuf
->user_priv
= screen
;
642 screen
->base
.pushbuf
->rsvd_kick
= 5;
644 chan
= screen
->base
.channel
;
646 pscreen
->destroy
= nv50_screen_destroy
;
647 pscreen
->context_create
= nv50_create
;
648 pscreen
->is_format_supported
= nv50_screen_is_format_supported
;
649 pscreen
->get_param
= nv50_screen_get_param
;
650 pscreen
->get_shader_param
= nv50_screen_get_shader_param
;
651 pscreen
->get_paramf
= nv50_screen_get_paramf
;
653 nv50_screen_init_resource_functions(pscreen
);
655 if (screen
->base
.device
->chipset
< 0x84 ||
656 debug_get_bool_option("NOUVEAU_PMPEG", FALSE
)) {
658 nouveau_screen_init_vdec(&screen
->base
);
659 } else if (screen
->base
.device
->chipset
< 0x98 ||
660 screen
->base
.device
->chipset
== 0xa0) {
662 screen
->base
.base
.get_video_param
= nv84_screen_get_video_param
;
663 screen
->base
.base
.is_video_format_supported
= nv84_screen_video_supported
;
666 screen
->base
.base
.get_video_param
= nouveau_vp3_screen_get_video_param
;
667 screen
->base
.base
.is_video_format_supported
= nouveau_vp3_screen_video_supported
;
670 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_GART
| NOUVEAU_BO_MAP
, 0, 4096,
671 NULL
, &screen
->fence
.bo
);
673 NOUVEAU_ERR("Failed to allocate fence bo: %d\n", ret
);
677 nouveau_bo_map(screen
->fence
.bo
, 0, NULL
);
678 screen
->fence
.map
= screen
->fence
.bo
->map
;
679 screen
->base
.fence
.emit
= nv50_screen_fence_emit
;
680 screen
->base
.fence
.update
= nv50_screen_fence_update
;
682 ret
= nouveau_object_new(chan
, 0xbeef0301, NOUVEAU_NOTIFIER_CLASS
,
683 &(struct nv04_notify
){ .length
= 32 },
684 sizeof(struct nv04_notify
), &screen
->sync
);
686 NOUVEAU_ERR("Failed to allocate notifier: %d\n", ret
);
690 ret
= nouveau_object_new(chan
, 0xbeef5039, NV50_M2MF_CLASS
,
691 NULL
, 0, &screen
->m2mf
);
693 NOUVEAU_ERR("Failed to allocate PGRAPH context for M2MF: %d\n", ret
);
697 ret
= nouveau_object_new(chan
, 0xbeef502d, NV50_2D_CLASS
,
698 NULL
, 0, &screen
->eng2d
);
700 NOUVEAU_ERR("Failed to allocate PGRAPH context for 2D: %d\n", ret
);
704 switch (dev
->chipset
& 0xf0) {
706 tesla_class
= NV50_3D_CLASS
;
710 tesla_class
= NV84_3D_CLASS
;
713 switch (dev
->chipset
) {
717 tesla_class
= NVA0_3D_CLASS
;
720 tesla_class
= NVAF_3D_CLASS
;
723 tesla_class
= NVA3_3D_CLASS
;
728 NOUVEAU_ERR("Not a known NV50 chipset: NV%02x\n", dev
->chipset
);
731 screen
->base
.class_3d
= tesla_class
;
733 ret
= nouveau_object_new(chan
, 0xbeef5097, tesla_class
,
734 NULL
, 0, &screen
->tesla
);
736 NOUVEAU_ERR("Failed to allocate PGRAPH context for 3D: %d\n", ret
);
740 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 16,
741 3 << NV50_CODE_BO_SIZE_LOG2
, NULL
, &screen
->code
);
743 NOUVEAU_ERR("Failed to allocate code bo: %d\n", ret
);
747 nouveau_heap_init(&screen
->vp_code_heap
, 0, 1 << NV50_CODE_BO_SIZE_LOG2
);
748 nouveau_heap_init(&screen
->gp_code_heap
, 0, 1 << NV50_CODE_BO_SIZE_LOG2
);
749 nouveau_heap_init(&screen
->fp_code_heap
, 0, 1 << NV50_CODE_BO_SIZE_LOG2
);
751 nouveau_getparam(dev
, NOUVEAU_GETPARAM_GRAPH_UNITS
, &value
);
753 screen
->TPs
= util_bitcount(value
& 0xffff);
754 screen
->MPsInTP
= util_bitcount((value
>> 24) & 0xf);
756 stack_size
= util_next_power_of_two(screen
->TPs
) * screen
->MPsInTP
*
757 STACK_WARPS_ALLOC
* 64 * 8;
759 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 16, stack_size
, NULL
,
762 NOUVEAU_ERR("Failed to allocate stack bo: %d\n", ret
);
766 uint64_t size_of_one_temp
= util_next_power_of_two(screen
->TPs
) *
767 screen
->MPsInTP
* LOCAL_WARPS_ALLOC
* THREADS_IN_WARP
*
769 screen
->max_tls_space
= dev
->vram_size
/ size_of_one_temp
* ONE_TEMP_SIZE
;
770 screen
->max_tls_space
/= 2; /* half of vram */
772 /* hw can address max 64 KiB */
773 screen
->max_tls_space
= MIN2(screen
->max_tls_space
, 64 << 10);
776 unsigned tls_space
= 4/*temps*/ * ONE_TEMP_SIZE
;
777 ret
= nv50_tls_alloc(screen
, tls_space
, &tls_size
);
781 if (nouveau_mesa_debug
)
782 debug_printf("TPs = %u, MPsInTP = %u, VRAM = %"PRIu64
" MiB, tls_size = %"PRIu64
" KiB\n",
783 screen
->TPs
, screen
->MPsInTP
, dev
->vram_size
>> 20, tls_size
>> 10);
785 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 16, 4 << 16, NULL
,
788 NOUVEAU_ERR("Failed to allocate uniforms bo: %d\n", ret
);
792 ret
= nouveau_bo_new(dev
, NOUVEAU_BO_VRAM
, 1 << 16, 3 << 16, NULL
,
795 NOUVEAU_ERR("Failed to allocate TIC/TSC bo: %d\n", ret
);
799 screen
->tic
.entries
= CALLOC(4096, sizeof(void *));
800 screen
->tsc
.entries
= screen
->tic
.entries
+ 2048;
802 if (!nv50_blitter_create(screen
))
805 nv50_screen_init_hwctx(screen
);
807 nouveau_fence_new(&screen
->base
, &screen
->base
.fence
.current
, FALSE
);
812 nv50_screen_destroy(pscreen
);
817 nv50_screen_tic_alloc(struct nv50_screen
*screen
, void *entry
)
819 int i
= screen
->tic
.next
;
821 while (screen
->tic
.lock
[i
/ 32] & (1 << (i
% 32)))
822 i
= (i
+ 1) & (NV50_TIC_MAX_ENTRIES
- 1);
824 screen
->tic
.next
= (i
+ 1) & (NV50_TIC_MAX_ENTRIES
- 1);
826 if (screen
->tic
.entries
[i
])
827 nv50_tic_entry(screen
->tic
.entries
[i
])->id
= -1;
829 screen
->tic
.entries
[i
] = entry
;
834 nv50_screen_tsc_alloc(struct nv50_screen
*screen
, void *entry
)
836 int i
= screen
->tsc
.next
;
838 while (screen
->tsc
.lock
[i
/ 32] & (1 << (i
% 32)))
839 i
= (i
+ 1) & (NV50_TSC_MAX_ENTRIES
- 1);
841 screen
->tsc
.next
= (i
+ 1) & (NV50_TSC_MAX_ENTRIES
- 1);
843 if (screen
->tsc
.entries
[i
])
844 nv50_tsc_entry(screen
->tsc
.entries
[i
])->id
= -1;
846 screen
->tsc
.entries
[i
] = entry
;