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.
26 #include "util/format/u_format.h"
27 #include "util/u_inlines.h"
28 #include "util/u_prim.h"
29 #include "translate/translate.h"
31 #include "nouveau_fence.h"
32 #include "nv_object.xml.h"
33 #include "nv30/nv30-40_3d.xml.h"
34 #include "nv30/nv30_context.h"
35 #include "nv30/nv30_format.h"
38 nv30_emit_vtxattr(struct nv30_context
*nv30
, struct pipe_vertex_buffer
*vb
,
39 struct pipe_vertex_element
*ve
, unsigned attr
)
41 const unsigned nc
= util_format_get_nr_components(ve
->src_format
);
42 struct nouveau_pushbuf
*push
= nv30
->base
.pushbuf
;
43 struct nv04_resource
*res
= nv04_resource(vb
->buffer
.resource
);
47 data
= nouveau_resource_map_offset(&nv30
->base
, res
, vb
->buffer_offset
+
48 ve
->src_offset
, NOUVEAU_BO_RD
);
50 util_format_unpack_rgba(ve
->src_format
, v
, data
, 1);
54 BEGIN_NV04(push
, NV30_3D(VTX_ATTR_4F(attr
)), 4);
55 PUSH_DATAf(push
, v
[0]);
56 PUSH_DATAf(push
, v
[1]);
57 PUSH_DATAf(push
, v
[2]);
58 PUSH_DATAf(push
, v
[3]);
61 BEGIN_NV04(push
, NV30_3D(VTX_ATTR_3F(attr
)), 3);
62 PUSH_DATAf(push
, v
[0]);
63 PUSH_DATAf(push
, v
[1]);
64 PUSH_DATAf(push
, v
[2]);
67 BEGIN_NV04(push
, NV30_3D(VTX_ATTR_2F(attr
)), 2);
68 PUSH_DATAf(push
, v
[0]);
69 PUSH_DATAf(push
, v
[1]);
72 BEGIN_NV04(push
, NV30_3D(VTX_ATTR_1F(attr
)), 1);
73 PUSH_DATAf(push
, v
[0]);
82 nv30_vbuf_range(struct nv30_context
*nv30
, int vbi
,
83 uint32_t *base
, uint32_t *size
)
85 assert(nv30
->vbo_max_index
!= ~0);
86 *base
= nv30
->vbo_min_index
* nv30
->vtxbuf
[vbi
].stride
;
87 *size
= (nv30
->vbo_max_index
-
88 nv30
->vbo_min_index
+ 1) * nv30
->vtxbuf
[vbi
].stride
;
92 nv30_prevalidate_vbufs(struct nv30_context
*nv30
)
94 struct pipe_vertex_buffer
*vb
;
95 struct nv04_resource
*buf
;
99 nv30
->vbo_fifo
= nv30
->vbo_user
= 0;
101 for (i
= 0; i
< nv30
->num_vtxbufs
; i
++) {
102 vb
= &nv30
->vtxbuf
[i
];
103 if (!vb
->stride
|| !vb
->buffer
.resource
) /* NOTE: user_buffer not implemented */
105 buf
= nv04_resource(vb
->buffer
.resource
);
107 /* NOTE: user buffers with temporary storage count as mapped by GPU */
108 if (!nouveau_resource_mapped_by_gpu(vb
->buffer
.resource
)) {
109 if (nv30
->vbo_push_hint
) {
113 if (buf
->status
& NOUVEAU_BUFFER_STATUS_USER_MEMORY
) {
114 nv30
->vbo_user
|= 1 << i
;
115 assert(vb
->stride
> vb
->buffer_offset
);
116 nv30_vbuf_range(nv30
, i
, &base
, &size
);
117 nouveau_user_buffer_upload(&nv30
->base
, buf
, base
, size
);
119 nouveau_buffer_migrate(&nv30
->base
, buf
, NOUVEAU_BO_GART
);
121 nv30
->base
.vbo_dirty
= true;
128 nv30_update_user_vbufs(struct nv30_context
*nv30
)
130 struct nouveau_pushbuf
*push
= nv30
->base
.pushbuf
;
131 uint32_t base
, offset
, size
;
133 uint32_t written
= 0;
135 for (i
= 0; i
< nv30
->vertex
->num_elements
; i
++) {
136 struct pipe_vertex_element
*ve
= &nv30
->vertex
->pipe
[i
];
137 const int b
= ve
->vertex_buffer_index
;
138 struct pipe_vertex_buffer
*vb
= &nv30
->vtxbuf
[b
];
139 struct nv04_resource
*buf
= nv04_resource(vb
->buffer
.resource
);
141 if (!(nv30
->vbo_user
& (1 << b
)))
145 nv30_emit_vtxattr(nv30
, vb
, ve
, i
);
148 nv30_vbuf_range(nv30
, b
, &base
, &size
);
150 if (!(written
& (1 << b
))) {
152 nouveau_user_buffer_upload(&nv30
->base
, buf
, base
, size
);
155 offset
= vb
->buffer_offset
+ ve
->src_offset
;
157 BEGIN_NV04(push
, NV30_3D(VTXBUF(i
)), 1);
158 PUSH_RESRC(push
, NV30_3D(VTXBUF(i
)), BUFCTX_VTXTMP
, buf
, offset
,
159 NOUVEAU_BO_LOW
| NOUVEAU_BO_RD
,
160 0, NV30_3D_VTXBUF_DMA1
);
162 nv30
->base
.vbo_dirty
= true;
166 nv30_release_user_vbufs(struct nv30_context
*nv30
)
168 uint32_t vbo_user
= nv30
->vbo_user
;
171 int i
= ffs(vbo_user
) - 1;
172 vbo_user
&= ~(1 << i
);
174 nouveau_buffer_release_gpu_storage(nv04_resource(nv30
->vtxbuf
[i
].buffer
.resource
));
177 nouveau_bufctx_reset(nv30
->bufctx
, BUFCTX_VTXTMP
);
181 nv30_vbo_validate(struct nv30_context
*nv30
)
183 struct nouveau_pushbuf
*push
= nv30
->base
.pushbuf
;
184 struct nv30_vertex_stateobj
*vertex
= nv30
->vertex
;
185 struct pipe_vertex_element
*ve
;
186 struct pipe_vertex_buffer
*vb
;
187 unsigned i
, redefine
;
189 nouveau_bufctx_reset(nv30
->bufctx
, BUFCTX_VTXBUF
);
190 if (!nv30
->vertex
|| nv30
->draw_flags
)
193 #if UTIL_ARCH_BIG_ENDIAN
194 if (1) { /* Figure out where the buffers are getting messed up */
196 if (unlikely(vertex
->need_conversion
)) {
201 nv30_prevalidate_vbufs(nv30
);
204 if (!PUSH_SPACE(push
, 128))
207 redefine
= MAX2(vertex
->num_elements
, nv30
->state
.num_vtxelts
);
211 BEGIN_NV04(push
, NV30_3D(VTXFMT(0)), redefine
);
213 for (i
= 0; i
< vertex
->num_elements
; i
++) {
214 ve
= &vertex
->pipe
[i
];
215 vb
= &nv30
->vtxbuf
[ve
->vertex_buffer_index
];
217 if (likely(vb
->stride
) || nv30
->vbo_fifo
)
218 PUSH_DATA (push
, (vb
->stride
<< 8) | vertex
->element
[i
].state
);
220 PUSH_DATA (push
, NV30_3D_VTXFMT_TYPE_V32_FLOAT
);
223 for (; i
< nv30
->state
.num_vtxelts
; i
++) {
224 PUSH_DATA (push
, NV30_3D_VTXFMT_TYPE_V32_FLOAT
);
227 for (i
= 0; i
< vertex
->num_elements
; i
++) {
228 struct nv04_resource
*res
;
232 ve
= &vertex
->pipe
[i
];
233 vb
= &nv30
->vtxbuf
[ve
->vertex_buffer_index
];
234 user
= (nv30
->vbo_user
& (1 << ve
->vertex_buffer_index
));
236 res
= nv04_resource(vb
->buffer
.resource
);
238 if (nv30
->vbo_fifo
|| unlikely(vb
->stride
== 0)) {
240 nv30_emit_vtxattr(nv30
, vb
, ve
, i
);
244 offset
= ve
->src_offset
+ vb
->buffer_offset
;
246 BEGIN_NV04(push
, NV30_3D(VTXBUF(i
)), 1);
247 PUSH_RESRC(push
, NV30_3D(VTXBUF(i
)), user
? BUFCTX_VTXTMP
: BUFCTX_VTXBUF
,
248 res
, offset
, NOUVEAU_BO_LOW
| NOUVEAU_BO_RD
,
249 0, NV30_3D_VTXBUF_DMA1
);
252 nv30
->state
.num_vtxelts
= vertex
->num_elements
;
256 nv30_vertex_state_create(struct pipe_context
*pipe
, unsigned num_elements
,
257 const struct pipe_vertex_element
*elements
)
259 struct nv30_vertex_stateobj
*so
;
260 struct translate_key transkey
;
263 so
= MALLOC(sizeof(*so
) + sizeof(*so
->element
) * num_elements
);
266 memcpy(so
->pipe
, elements
, sizeof(*elements
) * num_elements
);
267 so
->num_elements
= num_elements
;
268 so
->need_conversion
= false;
270 transkey
.nr_elements
= 0;
271 transkey
.output_stride
= 0;
273 for (i
= 0; i
< num_elements
; i
++) {
274 const struct pipe_vertex_element
*ve
= &elements
[i
];
275 const unsigned vbi
= ve
->vertex_buffer_index
;
276 enum pipe_format fmt
= ve
->src_format
;
278 so
->element
[i
].state
= nv30_vtxfmt(pipe
->screen
, fmt
)->hw
;
279 if (!so
->element
[i
].state
) {
280 switch (util_format_get_nr_components(fmt
)) {
281 case 1: fmt
= PIPE_FORMAT_R32_FLOAT
; break;
282 case 2: fmt
= PIPE_FORMAT_R32G32_FLOAT
; break;
283 case 3: fmt
= PIPE_FORMAT_R32G32B32_FLOAT
; break;
284 case 4: fmt
= PIPE_FORMAT_R32G32B32A32_FLOAT
; break;
290 so
->element
[i
].state
= nv30_vtxfmt(pipe
->screen
, fmt
)->hw
;
291 so
->need_conversion
= true;
295 unsigned j
= transkey
.nr_elements
++;
297 transkey
.element
[j
].type
= TRANSLATE_ELEMENT_NORMAL
;
298 transkey
.element
[j
].input_format
= ve
->src_format
;
299 transkey
.element
[j
].input_buffer
= vbi
;
300 transkey
.element
[j
].input_offset
= ve
->src_offset
;
301 transkey
.element
[j
].instance_divisor
= ve
->instance_divisor
;
303 transkey
.element
[j
].output_format
= fmt
;
304 transkey
.element
[j
].output_offset
= transkey
.output_stride
;
305 transkey
.output_stride
+= (util_format_get_stride(fmt
, 1) + 3) & ~3;
309 so
->translate
= translate_create(&transkey
);
310 so
->vtx_size
= transkey
.output_stride
/ 4;
311 so
->vtx_per_packet_max
= NV04_PFIFO_MAX_PACKET_LEN
/ MAX2(so
->vtx_size
, 1);
316 nv30_vertex_state_delete(struct pipe_context
*pipe
, void *hwcso
)
318 struct nv30_vertex_stateobj
*so
= hwcso
;
321 so
->translate
->release(so
->translate
);
326 nv30_vertex_state_bind(struct pipe_context
*pipe
, void *hwcso
)
328 struct nv30_context
*nv30
= nv30_context(pipe
);
330 nv30
->vertex
= hwcso
;
331 nv30
->dirty
|= NV30_NEW_VERTEX
;
335 nv30_draw_arrays(struct nv30_context
*nv30
,
336 unsigned mode
, unsigned start
, unsigned count
,
337 unsigned instance_count
)
339 struct nouveau_pushbuf
*push
= nv30
->base
.pushbuf
;
342 prim
= nv30_prim_gl(mode
);
344 BEGIN_NV04(push
, NV30_3D(VERTEX_BEGIN_END
), 1);
345 PUSH_DATA (push
, prim
);
347 const unsigned mpush
= 2047 * 256;
348 unsigned npush
= (count
> mpush
) ? mpush
: count
;
349 unsigned wpush
= ((npush
+ 255) & ~255) >> 8;
353 BEGIN_NI04(push
, NV30_3D(VB_VERTEX_BATCH
), wpush
);
354 while (npush
>= 256) {
355 PUSH_DATA (push
, 0xff000000 | start
);
361 PUSH_DATA (push
, ((npush
- 1) << 24) | start
);
363 BEGIN_NV04(push
, NV30_3D(VERTEX_BEGIN_END
), 1);
364 PUSH_DATA (push
, NV30_3D_VERTEX_BEGIN_END_STOP
);
368 nv30_draw_elements_inline_u08(struct nouveau_pushbuf
*push
, const uint8_t *map
,
369 unsigned start
, unsigned count
)
374 BEGIN_NV04(push
, NV30_3D(VB_ELEMENT_U32
), 1);
375 PUSH_DATA (push
, *map
++);
380 unsigned npush
= MIN2(count
, NV04_PFIFO_MAX_PACKET_LEN
);
383 BEGIN_NI04(push
, NV30_3D(VB_ELEMENT_U16
), npush
);
385 PUSH_DATA (push
, (map
[1] << 16) | map
[0]);
393 nv30_draw_elements_inline_u16(struct nouveau_pushbuf
*push
, const uint16_t *map
,
394 unsigned start
, unsigned count
)
399 BEGIN_NV04(push
, NV30_3D(VB_ELEMENT_U32
), 1);
400 PUSH_DATA (push
, *map
++);
405 unsigned npush
= MIN2(count
, NV04_PFIFO_MAX_PACKET_LEN
);
408 BEGIN_NI04(push
, NV30_3D(VB_ELEMENT_U16
), npush
);
410 PUSH_DATA (push
, (map
[1] << 16) | map
[0]);
417 nv30_draw_elements_inline_u32(struct nouveau_pushbuf
*push
, const uint32_t *map
,
418 unsigned start
, unsigned count
)
423 const unsigned nr
= MIN2(count
, NV04_PFIFO_MAX_PACKET_LEN
);
425 BEGIN_NI04(push
, NV30_3D(VB_ELEMENT_U32
), nr
);
426 PUSH_DATAp(push
, map
, nr
);
434 nv30_draw_elements_inline_u32_short(struct nouveau_pushbuf
*push
,
436 unsigned start
, unsigned count
)
441 BEGIN_NV04(push
, NV30_3D(VB_ELEMENT_U32
), 1);
442 PUSH_DATA (push
, *map
++);
447 unsigned npush
= MIN2(count
, NV04_PFIFO_MAX_PACKET_LEN
);
450 BEGIN_NI04(push
, NV30_3D(VB_ELEMENT_U16
), npush
);
452 PUSH_DATA (push
, (map
[1] << 16) | map
[0]);
459 nv30_draw_elements(struct nv30_context
*nv30
, bool shorten
,
460 const struct pipe_draw_info
*info
,
461 unsigned mode
, unsigned start
, unsigned count
,
462 unsigned instance_count
, int32_t index_bias
,
465 struct nouveau_pushbuf
*push
= nv30
->base
.pushbuf
;
466 struct nouveau_object
*eng3d
= nv30
->screen
->eng3d
;
467 unsigned prim
= nv30_prim_gl(mode
);
469 if (eng3d
->oclass
>= NV40_3D_CLASS
&& index_bias
!= nv30
->state
.index_bias
) {
470 BEGIN_NV04(push
, NV40_3D(VB_ELEMENT_BASE
), 1);
471 PUSH_DATA (push
, index_bias
);
472 nv30
->state
.index_bias
= index_bias
;
475 if (eng3d
->oclass
== NV40_3D_CLASS
&& index_size
> 1 &&
476 !info
->has_user_indices
) {
477 struct nv04_resource
*res
= nv04_resource(info
->index
.resource
);
480 assert(nouveau_resource_mapped_by_gpu(&res
->base
));
482 BEGIN_NV04(push
, NV30_3D(IDXBUF_OFFSET
), 2);
483 PUSH_RESRC(push
, NV30_3D(IDXBUF_OFFSET
), BUFCTX_IDXBUF
, res
, offset
,
484 NOUVEAU_BO_LOW
| NOUVEAU_BO_RD
, 0, 0);
485 PUSH_MTHD (push
, NV30_3D(IDXBUF_FORMAT
), BUFCTX_IDXBUF
, res
->bo
,
486 (index_size
== 2) ? 0x00000010 : 0x00000000,
487 res
->domain
| NOUVEAU_BO_RD
,
488 0, NV30_3D_IDXBUF_FORMAT_DMA1
);
489 BEGIN_NV04(push
, NV30_3D(VERTEX_BEGIN_END
), 1);
490 PUSH_DATA (push
, prim
);
492 const unsigned mpush
= 2047 * 256;
493 unsigned npush
= (count
> mpush
) ? mpush
: count
;
494 unsigned wpush
= ((npush
+ 255) & ~255) >> 8;
498 BEGIN_NI04(push
, NV30_3D(VB_INDEX_BATCH
), wpush
);
499 while (npush
>= 256) {
500 PUSH_DATA (push
, 0xff000000 | start
);
506 PUSH_DATA (push
, ((npush
- 1) << 24) | start
);
508 BEGIN_NV04(push
, NV30_3D(VERTEX_BEGIN_END
), 1);
509 PUSH_DATA (push
, NV30_3D_VERTEX_BEGIN_END_STOP
);
510 PUSH_RESET(push
, BUFCTX_IDXBUF
);
513 if (!info
->has_user_indices
)
514 data
= nouveau_resource_map_offset(&nv30
->base
,
515 nv04_resource(info
->index
.resource
),
516 start
* index_size
, NOUVEAU_BO_RD
);
518 data
= info
->index
.user
;
522 BEGIN_NV04(push
, NV30_3D(VERTEX_BEGIN_END
), 1);
523 PUSH_DATA (push
, prim
);
524 switch (index_size
) {
526 nv30_draw_elements_inline_u08(push
, data
, start
, count
);
529 nv30_draw_elements_inline_u16(push
, data
, start
, count
);
533 nv30_draw_elements_inline_u32_short(push
, data
, start
, count
);
535 nv30_draw_elements_inline_u32(push
, data
, start
, count
);
541 BEGIN_NV04(push
, NV30_3D(VERTEX_BEGIN_END
), 1);
542 PUSH_DATA (push
, NV30_3D_VERTEX_BEGIN_END_STOP
);
547 nv30_draw_vbo(struct pipe_context
*pipe
, const struct pipe_draw_info
*info
)
549 struct nv30_context
*nv30
= nv30_context(pipe
);
550 struct nouveau_pushbuf
*push
= nv30
->base
.pushbuf
;
553 if (!info
->primitive_restart
&&
554 !u_trim_pipe_prim(info
->mode
, (unsigned*)&info
->count
))
557 /* For picking only a few vertices from a large user buffer, push is better,
558 * if index count is larger and we expect repeated vertices, suggest upload.
560 nv30
->vbo_push_hint
= /* the 64 is heuristic */
561 !(info
->index_size
&&
562 ((info
->max_index
- info
->min_index
+ 64) < info
->count
));
564 nv30
->vbo_min_index
= info
->min_index
;
565 nv30
->vbo_max_index
= info
->max_index
;
567 if (nv30
->vbo_push_hint
!= !!nv30
->vbo_fifo
)
568 nv30
->dirty
|= NV30_NEW_ARRAYS
;
570 push
->user_priv
= &nv30
->bufctx
;
571 if (nv30
->vbo_user
&& !(nv30
->dirty
& (NV30_NEW_VERTEX
| NV30_NEW_ARRAYS
)))
572 nv30_update_user_vbufs(nv30
);
574 nv30_state_validate(nv30
, ~0, true);
575 if (nv30
->draw_flags
) {
576 nv30_render_vbo(pipe
, info
);
579 if (nv30
->vbo_fifo
) {
580 nv30_push_vbo(nv30
, info
);
584 for (i
= 0; i
< nv30
->num_vtxbufs
&& !nv30
->base
.vbo_dirty
; ++i
) {
585 if (!nv30
->vtxbuf
[i
].buffer
.resource
)
587 if (nv30
->vtxbuf
[i
].buffer
.resource
->flags
& PIPE_RESOURCE_FLAG_MAP_COHERENT
)
588 nv30
->base
.vbo_dirty
= true;
591 if (!nv30
->base
.vbo_dirty
&& info
->index_size
&& !info
->has_user_indices
&&
592 info
->index
.resource
->flags
& PIPE_RESOURCE_FLAG_MAP_COHERENT
)
593 nv30
->base
.vbo_dirty
= true;
595 if (nv30
->base
.vbo_dirty
) {
596 BEGIN_NV04(push
, NV30_3D(VTX_CACHE_INVALIDATE_1710
), 1);
598 nv30
->base
.vbo_dirty
= false;
601 if (!info
->index_size
) {
602 nv30_draw_arrays(nv30
,
603 info
->mode
, info
->start
, info
->count
,
604 info
->instance_count
);
606 bool shorten
= info
->max_index
<= 65535;
608 if (info
->primitive_restart
!= nv30
->state
.prim_restart
) {
609 if (info
->primitive_restart
) {
610 BEGIN_NV04(push
, NV40_3D(PRIM_RESTART_ENABLE
), 2);
612 PUSH_DATA (push
, info
->restart_index
);
614 if (info
->restart_index
> 65535)
617 BEGIN_NV04(push
, NV40_3D(PRIM_RESTART_ENABLE
), 1);
620 nv30
->state
.prim_restart
= info
->primitive_restart
;
622 if (info
->primitive_restart
) {
623 BEGIN_NV04(push
, NV40_3D(PRIM_RESTART_INDEX
), 1);
624 PUSH_DATA (push
, info
->restart_index
);
626 if (info
->restart_index
> 65535)
630 nv30_draw_elements(nv30
, shorten
, info
,
631 info
->mode
, info
->start
, info
->count
,
632 info
->instance_count
, info
->index_bias
, info
->index_size
);
635 nv30_state_release(nv30
);
636 nv30_release_user_vbufs(nv30
);
640 nv30_vbo_init(struct pipe_context
*pipe
)
642 pipe
->create_vertex_elements_state
= nv30_vertex_state_create
;
643 pipe
->delete_vertex_elements_state
= nv30_vertex_state_delete
;
644 pipe
->bind_vertex_elements_state
= nv30_vertex_state_bind
;
645 pipe
->draw_vbo
= nv30_draw_vbo
;