2 * Copyright 2012 Red Hat 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 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.
27 #include <nouveau_drm.h>
28 #include "util/u_format.h"
29 #include "util/u_format_s3tc.h"
31 #include "nv_object.xml.h"
32 #include "nv_m2mf.xml.h"
33 #include "nv30/nv30-40_3d.xml.h"
34 #include "nv30/nv01_2d.xml.h"
36 #include "nouveau_fence.h"
37 #include "nv30/nv30_screen.h"
38 #include "nv30/nv30_context.h"
39 #include "nv30/nv30_resource.h"
40 #include "nv30/nv30_format.h"
42 #define RANKINE_0397_CHIPSET 0x00000003
43 #define RANKINE_0497_CHIPSET 0x000001e0
44 #define RANKINE_0697_CHIPSET 0x00000010
45 #define CURIE_4097_CHIPSET 0x00000baf
46 #define CURIE_4497_CHIPSET 0x00005450
47 #define CURIE_4497_CHIPSET6X 0x00000088
50 nv30_screen_get_param(struct pipe_screen
*pscreen
, enum pipe_cap param
)
52 struct nv30_screen
*screen
= nv30_screen(pscreen
);
53 struct nouveau_object
*eng3d
= screen
->eng3d
;
54 struct nouveau_device
*dev
= nouveau_screen(pscreen
)->device
;
57 /* non-boolean capabilities */
58 case PIPE_CAP_MAX_RENDER_TARGETS
:
59 return (eng3d
->oclass
>= NV40_3D_CLASS
) ? 4 : 1;
60 case PIPE_CAP_MAX_TEXTURE_2D_LEVELS
:
62 case PIPE_CAP_MAX_TEXTURE_3D_LEVELS
:
64 case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS
:
66 case PIPE_CAP_GLSL_FEATURE_LEVEL
:
68 case PIPE_CAP_ENDIANNESS
:
69 return PIPE_ENDIAN_LITTLE
;
70 case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT
:
72 case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT
:
73 return NOUVEAU_MIN_BUFFER_MAP_ALIGN
;
74 case PIPE_CAP_MAX_VIEWPORTS
:
76 case PIPE_CAP_MAX_VERTEX_ATTRIB_STRIDE
:
78 /* supported capabilities */
79 case PIPE_CAP_TWO_SIDED_STENCIL
:
80 case PIPE_CAP_ANISOTROPIC_FILTER
:
81 case PIPE_CAP_POINT_SPRITE
:
82 case PIPE_CAP_OCCLUSION_QUERY
:
83 case PIPE_CAP_QUERY_TIME_ELAPSED
:
84 case PIPE_CAP_QUERY_TIMESTAMP
:
85 case PIPE_CAP_TEXTURE_SHADOW_MAP
:
86 case PIPE_CAP_TEXTURE_SWIZZLE
:
87 case PIPE_CAP_DEPTH_CLIP_DISABLE
:
88 case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT
:
89 case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT
:
90 case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER
:
91 case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER
:
92 case PIPE_CAP_TGSI_TEXCOORD
:
93 case PIPE_CAP_USER_CONSTANT_BUFFERS
:
94 case PIPE_CAP_USER_INDEX_BUFFERS
:
95 case PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT
:
96 case PIPE_CAP_VERTEX_BUFFER_OFFSET_4BYTE_ALIGNED_ONLY
:
97 case PIPE_CAP_VERTEX_BUFFER_STRIDE_4BYTE_ALIGNED_ONLY
:
98 case PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY
:
99 case PIPE_CAP_PREFER_BLIT_BASED_TEXTURE_TRANSFER
:
101 /* nv35 capabilities */
102 case PIPE_CAP_DEPTH_BOUNDS_TEST
:
103 return eng3d
->oclass
== NV35_3D_CLASS
|| eng3d
->oclass
>= NV40_3D_CLASS
;
104 /* nv4x capabilities */
105 case PIPE_CAP_BLEND_EQUATION_SEPARATE
:
106 case PIPE_CAP_NPOT_TEXTURES
:
107 case PIPE_CAP_CONDITIONAL_RENDER
:
108 case PIPE_CAP_TEXTURE_MIRROR_CLAMP
:
109 case PIPE_CAP_PRIMITIVE_RESTART
:
110 return (eng3d
->oclass
>= NV40_3D_CLASS
) ? 1 : 0;
112 case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS
:
114 case PIPE_CAP_INDEP_BLEND_ENABLE
:
115 case PIPE_CAP_INDEP_BLEND_FUNC
:
116 case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS
:
117 case PIPE_CAP_SHADER_STENCIL_EXPORT
:
118 case PIPE_CAP_TGSI_INSTANCEID
:
119 case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR
: /* XXX: yes? */
120 case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS
:
121 case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME
:
122 case PIPE_CAP_MIN_TEXEL_OFFSET
:
123 case PIPE_CAP_MAX_TEXEL_OFFSET
:
124 case PIPE_CAP_MIN_TEXTURE_GATHER_OFFSET
:
125 case PIPE_CAP_MAX_TEXTURE_GATHER_OFFSET
:
126 case PIPE_CAP_MAX_STREAM_OUTPUT_SEPARATE_COMPONENTS
:
127 case PIPE_CAP_MAX_STREAM_OUTPUT_INTERLEAVED_COMPONENTS
:
128 case PIPE_CAP_MAX_GEOMETRY_OUTPUT_VERTICES
:
129 case PIPE_CAP_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS
:
130 case PIPE_CAP_MAX_VERTEX_STREAMS
:
131 case PIPE_CAP_TGSI_CAN_COMPACT_CONSTANTS
:
132 case PIPE_CAP_TEXTURE_BARRIER
:
133 case PIPE_CAP_SEAMLESS_CUBE_MAP
:
134 case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE
:
135 case PIPE_CAP_CUBE_MAP_ARRAY
:
136 case PIPE_CAP_VERTEX_COLOR_UNCLAMPED
:
137 case PIPE_CAP_FRAGMENT_COLOR_CLAMPED
:
138 case PIPE_CAP_VERTEX_COLOR_CLAMPED
:
139 case PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION
:
140 case PIPE_CAP_MIXED_COLORBUFFER_FORMATS
:
141 case PIPE_CAP_START_INSTANCE
:
142 case PIPE_CAP_TEXTURE_MULTISAMPLE
:
143 case PIPE_CAP_TEXTURE_BUFFER_OBJECTS
:
144 case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT
:
145 case PIPE_CAP_QUERY_PIPELINE_STATISTICS
:
146 case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK
:
147 case PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE
:
148 case PIPE_CAP_MIXED_FRAMEBUFFER_SIZES
:
149 case PIPE_CAP_TGSI_VS_LAYER_VIEWPORT
:
150 case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS
:
151 case PIPE_CAP_TEXTURE_GATHER_SM5
:
152 case PIPE_CAP_FAKE_SW_MSAA
:
153 case PIPE_CAP_TEXTURE_QUERY_LOD
:
154 case PIPE_CAP_SAMPLE_SHADING
:
155 case PIPE_CAP_TEXTURE_GATHER_OFFSETS
:
156 case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION
:
157 case PIPE_CAP_USER_VERTEX_BUFFERS
:
158 case PIPE_CAP_COMPUTE
:
159 case PIPE_CAP_DRAW_INDIRECT
:
160 case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE
:
161 case PIPE_CAP_CONDITIONAL_RENDER_INVERTED
:
162 case PIPE_CAP_SAMPLER_VIEW_TARGET
:
163 case PIPE_CAP_CLIP_HALFZ
:
164 case PIPE_CAP_VERTEXID_NOBASE
:
165 case PIPE_CAP_POLYGON_OFFSET_CLAMP
:
166 case PIPE_CAP_MULTISAMPLE_Z_RESOLVE
:
167 case PIPE_CAP_RESOURCE_FROM_USER_MEMORY
:
168 case PIPE_CAP_DEVICE_RESET_STATUS_QUERY
:
169 case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS
:
170 case PIPE_CAP_TEXTURE_FLOAT_LINEAR
:
171 case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR
:
172 case PIPE_CAP_TGSI_TXQS
:
173 case PIPE_CAP_FORCE_PERSAMPLE_INTERP
:
176 case PIPE_CAP_VENDOR_ID
:
178 case PIPE_CAP_DEVICE_ID
: {
180 if (nouveau_getparam(dev
, NOUVEAU_GETPARAM_PCI_DEVICE
, &device_id
)) {
181 NOUVEAU_ERR("NOUVEAU_GETPARAM_PCI_DEVICE failed.\n");
186 case PIPE_CAP_ACCELERATED
:
188 case PIPE_CAP_VIDEO_MEMORY
:
189 return dev
->vram_size
>> 20;
194 debug_printf("unknown param %d\n", param
);
199 nv30_screen_get_paramf(struct pipe_screen
*pscreen
, enum pipe_capf param
)
201 struct nv30_screen
*screen
= nv30_screen(pscreen
);
202 struct nouveau_object
*eng3d
= screen
->eng3d
;
205 case PIPE_CAPF_MAX_LINE_WIDTH
:
206 case PIPE_CAPF_MAX_LINE_WIDTH_AA
:
208 case PIPE_CAPF_MAX_POINT_WIDTH
:
209 case PIPE_CAPF_MAX_POINT_WIDTH_AA
:
211 case PIPE_CAPF_MAX_TEXTURE_ANISOTROPY
:
212 return (eng3d
->oclass
>= NV40_3D_CLASS
) ? 16.0 : 8.0;
213 case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS
:
216 debug_printf("unknown paramf %d\n", param
);
222 nv30_screen_get_shader_param(struct pipe_screen
*pscreen
, unsigned shader
,
223 enum pipe_shader_cap param
)
225 struct nv30_screen
*screen
= nv30_screen(pscreen
);
226 struct nouveau_object
*eng3d
= screen
->eng3d
;
229 case PIPE_SHADER_VERTEX
:
231 case PIPE_SHADER_CAP_MAX_INSTRUCTIONS
:
232 case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS
:
233 return (eng3d
->oclass
>= NV40_3D_CLASS
) ? 512 : 256;
234 case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS
:
235 case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS
:
236 return (eng3d
->oclass
>= NV40_3D_CLASS
) ? 512 : 0;
237 case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH
:
239 case PIPE_SHADER_CAP_MAX_INPUTS
:
240 case PIPE_SHADER_CAP_MAX_OUTPUTS
:
242 case PIPE_SHADER_CAP_MAX_CONST_BUFFER_SIZE
:
243 return ((eng3d
->oclass
>= NV40_3D_CLASS
) ? (468 - 6): (256 - 6)) * sizeof(float[4]);
244 case PIPE_SHADER_CAP_MAX_CONST_BUFFERS
:
246 case PIPE_SHADER_CAP_MAX_TEMPS
:
247 return (eng3d
->oclass
>= NV40_3D_CLASS
) ? 32 : 13;
248 case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS
:
249 case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS
:
251 case PIPE_SHADER_CAP_MAX_PREDS
:
252 case PIPE_SHADER_CAP_TGSI_CONT_SUPPORTED
:
253 case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED
:
254 case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR
:
255 case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR
:
256 case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR
:
257 case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR
:
258 case PIPE_SHADER_CAP_SUBROUTINES
:
259 case PIPE_SHADER_CAP_INTEGERS
:
260 case PIPE_SHADER_CAP_DOUBLES
:
261 case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED
:
262 case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED
:
263 case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED
:
264 case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE
:
267 debug_printf("unknown vertex shader param %d\n", param
);
271 case PIPE_SHADER_FRAGMENT
:
273 case PIPE_SHADER_CAP_MAX_INSTRUCTIONS
:
274 case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS
:
275 case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS
:
276 case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS
:
278 case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH
:
280 case PIPE_SHADER_CAP_MAX_INPUTS
:
281 return 8; /* should be possible to do 10 with nv4x */
282 case PIPE_SHADER_CAP_MAX_OUTPUTS
:
284 case PIPE_SHADER_CAP_MAX_CONST_BUFFER_SIZE
:
285 return ((eng3d
->oclass
>= NV40_3D_CLASS
) ? 224 : 32) * sizeof(float[4]);
286 case PIPE_SHADER_CAP_MAX_CONST_BUFFERS
:
288 case PIPE_SHADER_CAP_MAX_TEMPS
:
290 case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS
:
291 case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS
:
293 case PIPE_SHADER_CAP_MAX_PREDS
:
294 case PIPE_SHADER_CAP_TGSI_CONT_SUPPORTED
:
295 case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED
:
296 case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR
:
297 case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR
:
298 case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR
:
299 case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR
:
300 case PIPE_SHADER_CAP_SUBROUTINES
:
301 case PIPE_SHADER_CAP_DOUBLES
:
302 case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED
:
303 case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED
:
304 case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED
:
305 case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE
:
308 debug_printf("unknown fragment shader param %d\n", param
);
318 nv30_screen_is_format_supported(struct pipe_screen
*pscreen
,
319 enum pipe_format format
,
320 enum pipe_texture_target target
,
321 unsigned sample_count
,
324 if (sample_count
> nv30_screen(pscreen
)->max_sample_count
)
327 if (!(0x00000017 & (1 << sample_count
)))
330 if (!util_format_is_supported(format
, bindings
)) {
334 /* transfers & shared are always supported */
335 bindings
&= ~(PIPE_BIND_TRANSFER_READ
|
336 PIPE_BIND_TRANSFER_WRITE
|
339 return (nv30_format_info(pscreen
, format
)->bindings
& bindings
) == bindings
;
343 nv30_screen_fence_emit(struct pipe_screen
*pscreen
, uint32_t *sequence
)
345 struct nv30_screen
*screen
= nv30_screen(pscreen
);
346 struct nouveau_pushbuf
*push
= screen
->base
.pushbuf
;
348 *sequence
= ++screen
->base
.fence
.sequence
;
350 assert(PUSH_AVAIL(push
) >= 3);
351 PUSH_DATA (push
, NV30_3D_FENCE_OFFSET
|
352 (2 /* size */ << 18) | (7 /* subchan */ << 13));
354 PUSH_DATA (push
, *sequence
);
358 nv30_screen_fence_update(struct pipe_screen
*pscreen
)
360 struct nv30_screen
*screen
= nv30_screen(pscreen
);
361 struct nv04_notify
*fence
= screen
->fence
->data
;
362 return *(uint32_t *)((char *)screen
->notify
->map
+ fence
->offset
);
366 nv30_screen_destroy(struct pipe_screen
*pscreen
)
368 struct nv30_screen
*screen
= nv30_screen(pscreen
);
370 if (!nouveau_drm_screen_unref(&screen
->base
))
373 if (screen
->base
.fence
.current
) {
374 struct nouveau_fence
*current
= NULL
;
376 /* nouveau_fence_wait will create a new current fence, so wait on the
377 * _current_ one, and remove both.
379 nouveau_fence_ref(screen
->base
.fence
.current
, ¤t
);
380 nouveau_fence_wait(current
);
381 nouveau_fence_ref(NULL
, ¤t
);
382 nouveau_fence_ref(NULL
, &screen
->base
.fence
.current
);
385 nouveau_bo_ref(NULL
, &screen
->notify
);
387 nouveau_heap_destroy(&screen
->query_heap
);
388 nouveau_heap_destroy(&screen
->vp_exec_heap
);
389 nouveau_heap_destroy(&screen
->vp_data_heap
);
391 nouveau_object_del(&screen
->query
);
392 nouveau_object_del(&screen
->fence
);
393 nouveau_object_del(&screen
->ntfy
);
395 nouveau_object_del(&screen
->sifm
);
396 nouveau_object_del(&screen
->swzsurf
);
397 nouveau_object_del(&screen
->surf2d
);
398 nouveau_object_del(&screen
->m2mf
);
399 nouveau_object_del(&screen
->eng3d
);
400 nouveau_object_del(&screen
->null
);
402 nouveau_screen_fini(&screen
->base
);
406 #define FAIL_SCREEN_INIT(str, err) \
408 NOUVEAU_ERR(str, err); \
409 nv30_screen_destroy(pscreen); \
414 nv30_screen_create(struct nouveau_device
*dev
)
416 struct nv30_screen
*screen
= CALLOC_STRUCT(nv30_screen
);
417 struct pipe_screen
*pscreen
;
418 struct nouveau_pushbuf
*push
;
419 struct nv04_fifo
*fifo
;
426 switch (dev
->chipset
& 0xf0) {
428 if (RANKINE_0397_CHIPSET
& (1 << (dev
->chipset
& 0x0f)))
429 oclass
= NV30_3D_CLASS
;
431 if (RANKINE_0697_CHIPSET
& (1 << (dev
->chipset
& 0x0f)))
432 oclass
= NV34_3D_CLASS
;
434 if (RANKINE_0497_CHIPSET
& (1 << (dev
->chipset
& 0x0f)))
435 oclass
= NV35_3D_CLASS
;
438 if (CURIE_4097_CHIPSET
& (1 << (dev
->chipset
& 0x0f)))
439 oclass
= NV40_3D_CLASS
;
441 if (CURIE_4497_CHIPSET
& (1 << (dev
->chipset
& 0x0f)))
442 oclass
= NV44_3D_CLASS
;
445 if (CURIE_4497_CHIPSET6X
& (1 << (dev
->chipset
& 0x0f)))
446 oclass
= NV44_3D_CLASS
;
453 NOUVEAU_ERR("unknown 3d class for 0x%02x\n", dev
->chipset
);
459 * Some modern apps try to use msaa without keeping in mind the
460 * restrictions on videomem of older cards. Resulting in dmesg saying:
461 * [ 1197.850642] nouveau E[soffice.bin[3785]] fail ttm_validate
462 * [ 1197.850648] nouveau E[soffice.bin[3785]] validating bo list
463 * [ 1197.850654] nouveau E[soffice.bin[3785]] validate: -12
465 * Because we are running out of video memory, after which the program
466 * using the msaa visual freezes, and eventually the entire system freezes.
468 * To work around this we do not allow msaa visauls by default and allow
469 * the user to override this via NV30_MAX_MSAA.
471 screen
->max_sample_count
= debug_get_num_option("NV30_MAX_MSAA", 0);
472 if (screen
->max_sample_count
> 4)
473 screen
->max_sample_count
= 4;
475 pscreen
= &screen
->base
.base
;
476 pscreen
->destroy
= nv30_screen_destroy
;
477 pscreen
->get_param
= nv30_screen_get_param
;
478 pscreen
->get_paramf
= nv30_screen_get_paramf
;
479 pscreen
->get_shader_param
= nv30_screen_get_shader_param
;
480 pscreen
->context_create
= nv30_context_create
;
481 pscreen
->is_format_supported
= nv30_screen_is_format_supported
;
482 nv30_resource_screen_init(pscreen
);
483 nouveau_screen_init_vdec(&screen
->base
);
485 screen
->base
.fence
.emit
= nv30_screen_fence_emit
;
486 screen
->base
.fence
.update
= nv30_screen_fence_update
;
488 ret
= nouveau_screen_init(&screen
->base
, dev
);
490 FAIL_SCREEN_INIT("nv30_screen_init failed: %d\n", ret
);
492 screen
->base
.vidmem_bindings
|= PIPE_BIND_VERTEX_BUFFER
;
493 screen
->base
.sysmem_bindings
|= PIPE_BIND_VERTEX_BUFFER
;
494 if (oclass
== NV40_3D_CLASS
) {
495 screen
->base
.vidmem_bindings
|= PIPE_BIND_INDEX_BUFFER
;
496 screen
->base
.sysmem_bindings
|= PIPE_BIND_INDEX_BUFFER
;
499 fifo
= screen
->base
.channel
->data
;
500 push
= screen
->base
.pushbuf
;
501 push
->rsvd_kick
= 16;
503 ret
= nouveau_object_new(screen
->base
.channel
, 0x00000000, NV01_NULL_CLASS
,
504 NULL
, 0, &screen
->null
);
506 FAIL_SCREEN_INIT("error allocating null object: %d\n", ret
);
508 /* DMA_FENCE refuses to accept DMA objects with "adjust" filled in,
509 * this means that the address pointed at by the DMA object must
510 * be 4KiB aligned, which means this object needs to be the first
511 * one allocated on the channel.
513 ret
= nouveau_object_new(screen
->base
.channel
, 0xbeef1e00,
514 NOUVEAU_NOTIFIER_CLASS
, &(struct nv04_notify
) {
515 .length
= 32 }, sizeof(struct nv04_notify
),
518 FAIL_SCREEN_INIT("error allocating fence notifier: %d\n", ret
);
520 /* DMA_NOTIFY object, we don't actually use this but M2MF fails without */
521 ret
= nouveau_object_new(screen
->base
.channel
, 0xbeef0301,
522 NOUVEAU_NOTIFIER_CLASS
, &(struct nv04_notify
) {
523 .length
= 32 }, sizeof(struct nv04_notify
),
526 FAIL_SCREEN_INIT("error allocating sync notifier: %d\n", ret
);
528 /* DMA_QUERY, used to implement occlusion queries, we attempt to allocate
529 * the remainder of the "notifier block" assigned by the kernel for
530 * use as query objects
532 ret
= nouveau_object_new(screen
->base
.channel
, 0xbeef0351,
533 NOUVEAU_NOTIFIER_CLASS
, &(struct nv04_notify
) {
534 .length
= 4096 - 128 }, sizeof(struct nv04_notify
),
537 FAIL_SCREEN_INIT("error allocating query notifier: %d\n", ret
);
539 ret
= nouveau_heap_init(&screen
->query_heap
, 0, 4096 - 128);
541 FAIL_SCREEN_INIT("error creating query heap: %d\n", ret
);
543 LIST_INITHEAD(&screen
->queries
);
545 /* Vertex program resources (code/data), currently 6 of the constant
546 * slots are reserved to implement user clipping planes
548 if (oclass
< NV40_3D_CLASS
) {
549 nouveau_heap_init(&screen
->vp_exec_heap
, 0, 256);
550 nouveau_heap_init(&screen
->vp_data_heap
, 6, 256 - 6);
552 nouveau_heap_init(&screen
->vp_exec_heap
, 0, 512);
553 nouveau_heap_init(&screen
->vp_data_heap
, 6, 468 - 6);
556 ret
= nouveau_bo_wrap(screen
->base
.device
, fifo
->notify
, &screen
->notify
);
558 ret
= nouveau_bo_map(screen
->notify
, 0, screen
->base
.client
);
560 FAIL_SCREEN_INIT("error mapping notifier memory: %d\n", ret
);
562 ret
= nouveau_object_new(screen
->base
.channel
, 0xbeef3097, oclass
,
563 NULL
, 0, &screen
->eng3d
);
565 FAIL_SCREEN_INIT("error allocating 3d object: %d\n", ret
);
567 BEGIN_NV04(push
, NV01_SUBC(3D
, OBJECT
), 1);
568 PUSH_DATA (push
, screen
->eng3d
->handle
);
569 BEGIN_NV04(push
, NV30_3D(DMA_NOTIFY
), 13);
570 PUSH_DATA (push
, screen
->ntfy
->handle
);
571 PUSH_DATA (push
, fifo
->vram
); /* TEXTURE0 */
572 PUSH_DATA (push
, fifo
->gart
); /* TEXTURE1 */
573 PUSH_DATA (push
, fifo
->vram
); /* COLOR1 */
574 PUSH_DATA (push
, screen
->null
->handle
); /* UNK190 */
575 PUSH_DATA (push
, fifo
->vram
); /* COLOR0 */
576 PUSH_DATA (push
, fifo
->vram
); /* ZETA */
577 PUSH_DATA (push
, fifo
->vram
); /* VTXBUF0 */
578 PUSH_DATA (push
, fifo
->gart
); /* VTXBUF1 */
579 PUSH_DATA (push
, screen
->fence
->handle
); /* FENCE */
580 PUSH_DATA (push
, screen
->query
->handle
); /* QUERY - intr 0x80 if nullobj */
581 PUSH_DATA (push
, screen
->null
->handle
); /* UNK1AC */
582 PUSH_DATA (push
, screen
->null
->handle
); /* UNK1B0 */
583 if (screen
->eng3d
->oclass
< NV40_3D_CLASS
) {
584 BEGIN_NV04(push
, SUBC_3D(0x03b0), 1);
585 PUSH_DATA (push
, 0x00100000);
586 BEGIN_NV04(push
, SUBC_3D(0x1d80), 1);
589 BEGIN_NV04(push
, SUBC_3D(0x1e98), 1);
591 BEGIN_NV04(push
, SUBC_3D(0x17e0), 3);
592 PUSH_DATA (push
, fui(0.0));
593 PUSH_DATA (push
, fui(0.0));
594 PUSH_DATA (push
, fui(1.0));
595 BEGIN_NV04(push
, SUBC_3D(0x1f80), 16);
596 for (i
= 0; i
< 16; i
++)
597 PUSH_DATA (push
, (i
== 8) ? 0x0000ffff : 0);
599 BEGIN_NV04(push
, NV30_3D(RC_ENABLE
), 1);
602 BEGIN_NV04(push
, NV40_3D(DMA_COLOR2
), 2);
603 PUSH_DATA (push
, fifo
->vram
);
604 PUSH_DATA (push
, fifo
->vram
); /* COLOR3 */
606 BEGIN_NV04(push
, SUBC_3D(0x1450), 1);
607 PUSH_DATA (push
, 0x00000004);
609 BEGIN_NV04(push
, SUBC_3D(0x1ea4), 3); /* ZCULL */
610 PUSH_DATA (push
, 0x00000010);
611 PUSH_DATA (push
, 0x01000100);
612 PUSH_DATA (push
, 0xff800006);
614 /* vtxprog output routing */
615 BEGIN_NV04(push
, SUBC_3D(0x1fc4), 1);
616 PUSH_DATA (push
, 0x06144321);
617 BEGIN_NV04(push
, SUBC_3D(0x1fc8), 2);
618 PUSH_DATA (push
, 0xedcba987);
619 PUSH_DATA (push
, 0x0000006f);
620 BEGIN_NV04(push
, SUBC_3D(0x1fd0), 1);
621 PUSH_DATA (push
, 0x00171615);
622 BEGIN_NV04(push
, SUBC_3D(0x1fd4), 1);
623 PUSH_DATA (push
, 0x001b1a19);
625 BEGIN_NV04(push
, SUBC_3D(0x1ef8), 1);
626 PUSH_DATA (push
, 0x0020ffff);
627 BEGIN_NV04(push
, SUBC_3D(0x1d64), 1);
628 PUSH_DATA (push
, 0x01d300d4);
630 BEGIN_NV04(push
, NV40_3D(MIPMAP_ROUNDING
), 1);
631 PUSH_DATA (push
, NV40_3D_MIPMAP_ROUNDING_MODE_DOWN
);
634 ret
= nouveau_object_new(screen
->base
.channel
, 0xbeef3901, NV03_M2MF_CLASS
,
635 NULL
, 0, &screen
->m2mf
);
637 FAIL_SCREEN_INIT("error allocating m2mf object: %d\n", ret
);
639 BEGIN_NV04(push
, NV01_SUBC(M2MF
, OBJECT
), 1);
640 PUSH_DATA (push
, screen
->m2mf
->handle
);
641 BEGIN_NV04(push
, NV03_M2MF(DMA_NOTIFY
), 1);
642 PUSH_DATA (push
, screen
->ntfy
->handle
);
644 ret
= nouveau_object_new(screen
->base
.channel
, 0xbeef6201,
645 NV10_SURFACE_2D_CLASS
, NULL
, 0, &screen
->surf2d
);
647 FAIL_SCREEN_INIT("error allocating surf2d object: %d\n", ret
);
649 BEGIN_NV04(push
, NV01_SUBC(SF2D
, OBJECT
), 1);
650 PUSH_DATA (push
, screen
->surf2d
->handle
);
651 BEGIN_NV04(push
, NV04_SF2D(DMA_NOTIFY
), 1);
652 PUSH_DATA (push
, screen
->ntfy
->handle
);
654 if (dev
->chipset
< 0x40)
655 oclass
= NV30_SURFACE_SWZ_CLASS
;
657 oclass
= NV40_SURFACE_SWZ_CLASS
;
659 ret
= nouveau_object_new(screen
->base
.channel
, 0xbeef5201, oclass
,
660 NULL
, 0, &screen
->swzsurf
);
662 FAIL_SCREEN_INIT("error allocating swizzled surface object: %d\n", ret
);
664 BEGIN_NV04(push
, NV01_SUBC(SSWZ
, OBJECT
), 1);
665 PUSH_DATA (push
, screen
->swzsurf
->handle
);
666 BEGIN_NV04(push
, NV04_SSWZ(DMA_NOTIFY
), 1);
667 PUSH_DATA (push
, screen
->ntfy
->handle
);
669 if (dev
->chipset
< 0x40)
670 oclass
= NV30_SIFM_CLASS
;
672 oclass
= NV40_SIFM_CLASS
;
674 ret
= nouveau_object_new(screen
->base
.channel
, 0xbeef7701, oclass
,
675 NULL
, 0, &screen
->sifm
);
677 FAIL_SCREEN_INIT("error allocating scaled image object: %d\n", ret
);
679 BEGIN_NV04(push
, NV01_SUBC(SIFM
, OBJECT
), 1);
680 PUSH_DATA (push
, screen
->sifm
->handle
);
681 BEGIN_NV04(push
, NV03_SIFM(DMA_NOTIFY
), 1);
682 PUSH_DATA (push
, screen
->ntfy
->handle
);
683 BEGIN_NV04(push
, NV05_SIFM(COLOR_CONVERSION
), 1);
684 PUSH_DATA (push
, NV05_SIFM_COLOR_CONVERSION_TRUNCATE
);
686 nouveau_pushbuf_kick(push
, push
->channel
);
688 nouveau_fence_new(&screen
->base
, &screen
->base
.fence
.current
, false);