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 "pipe/p_context.h"
24 #include "pipe/p_state.h"
25 #include "util/u_inlines.h"
26 #include "util/u_format.h"
27 #include "translate/translate.h"
29 #include "nv50_context.h"
30 #include "nv50_resource.h"
32 #include "nv50_3d.xml.h"
35 nv50_vertex_state_delete(struct pipe_context
*pipe
,
38 struct nv50_vertex_stateobj
*so
= hwcso
;
41 so
->translate
->release(so
->translate
);
46 nv50_vertex_state_create(struct pipe_context
*pipe
,
47 unsigned num_elements
,
48 const struct pipe_vertex_element
*elements
)
50 struct nv50_vertex_stateobj
*so
;
51 struct translate_key transkey
;
54 so
= MALLOC(sizeof(*so
) +
55 num_elements
* sizeof(struct nv50_vertex_element
));
58 so
->num_elements
= num_elements
;
59 so
->instance_elts
= 0;
60 so
->instance_bufs
= 0;
61 so
->need_conversion
= FALSE
;
63 memset(so
->vb_access_size
, 0, sizeof(so
->vb_access_size
));
65 for (i
= 0; i
< PIPE_MAX_ATTRIBS
; ++i
)
66 so
->min_instance_div
[i
] = 0xffffffff;
68 transkey
.nr_elements
= 0;
69 transkey
.output_stride
= 0;
71 for (i
= 0; i
< num_elements
; ++i
) {
72 const struct pipe_vertex_element
*ve
= &elements
[i
];
73 const unsigned vbi
= ve
->vertex_buffer_index
;
75 enum pipe_format fmt
= ve
->src_format
;
77 so
->element
[i
].pipe
= elements
[i
];
78 so
->element
[i
].state
= nv50_format_table
[fmt
].vtx
;
80 if (!so
->element
[i
].state
) {
81 switch (util_format_get_nr_components(fmt
)) {
82 case 1: fmt
= PIPE_FORMAT_R32_FLOAT
; break;
83 case 2: fmt
= PIPE_FORMAT_R32G32_FLOAT
; break;
84 case 3: fmt
= PIPE_FORMAT_R32G32B32_FLOAT
; break;
85 case 4: fmt
= PIPE_FORMAT_R32G32B32A32_FLOAT
; break;
91 so
->element
[i
].state
= nv50_format_table
[fmt
].vtx
;
92 so
->need_conversion
= TRUE
;
94 so
->element
[i
].state
|= i
;
96 size
= util_format_get_blocksize(fmt
);
97 if (so
->vb_access_size
[vbi
] < (ve
->src_offset
+ size
))
98 so
->vb_access_size
[vbi
] = ve
->src_offset
+ size
;
101 unsigned j
= transkey
.nr_elements
++;
103 transkey
.element
[j
].type
= TRANSLATE_ELEMENT_NORMAL
;
104 transkey
.element
[j
].input_format
= ve
->src_format
;
105 transkey
.element
[j
].input_buffer
= vbi
;
106 transkey
.element
[j
].input_offset
= ve
->src_offset
;
107 transkey
.element
[j
].instance_divisor
= ve
->instance_divisor
;
109 transkey
.element
[j
].output_format
= fmt
;
110 transkey
.element
[j
].output_offset
= transkey
.output_stride
;
111 transkey
.output_stride
+= (util_format_get_stride(fmt
, 1) + 3) & ~3;
113 if (unlikely(ve
->instance_divisor
)) {
114 so
->instance_elts
|= 1 << i
;
115 so
->instance_bufs
|= 1 << vbi
;
116 if (ve
->instance_divisor
< so
->min_instance_div
[vbi
])
117 so
->min_instance_div
[vbi
] = ve
->instance_divisor
;
122 so
->translate
= translate_create(&transkey
);
123 so
->vertex_size
= transkey
.output_stride
/ 4;
124 so
->packet_vertex_limit
= NV04_PFIFO_MAX_PACKET_LEN
/
125 MAX2(so
->vertex_size
, 1);
130 #define NV50_3D_VERTEX_ATTRIB_INACTIVE \
131 NV50_3D_VERTEX_ARRAY_ATTRIB_TYPE_FLOAT | \
132 NV50_3D_VERTEX_ARRAY_ATTRIB_FORMAT_32_32_32_32 | \
133 NV50_3D_VERTEX_ARRAY_ATTRIB_CONST
136 nv50_emit_vtxattr(struct nv50_context
*nv50
, struct pipe_vertex_buffer
*vb
,
137 struct pipe_vertex_element
*ve
, unsigned attr
)
139 struct nouveau_pushbuf
*push
= nv50
->base
.pushbuf
;
140 const void *data
= (const uint8_t *)vb
->user_buffer
+ ve
->src_offset
;
142 const unsigned nc
= util_format_get_nr_components(ve
->src_format
);
143 const struct util_format_description
*desc
=
144 util_format_description(ve
->src_format
);
146 assert(vb
->user_buffer
);
148 if (desc
->channel
[0].pure_integer
) {
149 if (desc
->channel
[0].type
== UTIL_FORMAT_TYPE_SIGNED
) {
150 desc
->unpack_rgba_sint((int32_t *)v
, 0, data
, 0, 1, 1);
152 desc
->unpack_rgba_uint((uint32_t *)v
, 0, data
, 0, 1, 1);
155 desc
->unpack_rgba_float(v
, 0, data
, 0, 1, 1);
160 BEGIN_NV04(push
, NV50_3D(VTX_ATTR_4F_X(attr
)), 4);
161 PUSH_DATAf(push
, v
[0]);
162 PUSH_DATAf(push
, v
[1]);
163 PUSH_DATAf(push
, v
[2]);
164 PUSH_DATAf(push
, v
[3]);
167 BEGIN_NV04(push
, NV50_3D(VTX_ATTR_3F_X(attr
)), 3);
168 PUSH_DATAf(push
, v
[0]);
169 PUSH_DATAf(push
, v
[1]);
170 PUSH_DATAf(push
, v
[2]);
173 BEGIN_NV04(push
, NV50_3D(VTX_ATTR_2F_X(attr
)), 2);
174 PUSH_DATAf(push
, v
[0]);
175 PUSH_DATAf(push
, v
[1]);
178 if (attr
== nv50
->vertprog
->vp
.edgeflag
) {
179 BEGIN_NV04(push
, NV50_3D(EDGEFLAG
), 1);
180 PUSH_DATA (push
, v
[0] ? 1 : 0);
182 BEGIN_NV04(push
, NV50_3D(VTX_ATTR_1F(attr
)), 1);
183 PUSH_DATAf(push
, v
[0]);
192 nv50_user_vbuf_range(struct nv50_context
*nv50
, int vbi
,
193 uint32_t *base
, uint32_t *size
)
195 if (unlikely(nv50
->vertex
->instance_bufs
& (1 << vbi
))) {
196 /* TODO: use min and max instance divisor to get a proper range */
198 *size
= nv50
->vtxbuf
[vbi
].buffer
->width0
;
200 /* NOTE: if there are user buffers, we *must* have index bounds */
201 assert(nv50
->vb_elt_limit
!= ~0);
202 *base
= nv50
->vb_elt_first
* nv50
->vtxbuf
[vbi
].stride
;
203 *size
= nv50
->vb_elt_limit
* nv50
->vtxbuf
[vbi
].stride
+
204 nv50
->vertex
->vb_access_size
[vbi
];
209 nv50_upload_user_buffers(struct nv50_context
*nv50
,
210 uint64_t addrs
[], uint32_t limits
[])
214 for (b
= 0; b
< nv50
->num_vtxbufs
; ++b
) {
215 struct nouveau_bo
*bo
;
216 const struct pipe_vertex_buffer
*vb
= &nv50
->vtxbuf
[b
];
219 if (!(nv50
->vbo_user
& (1 << b
)) || !vb
->stride
)
221 nv50_user_vbuf_range(nv50
, b
, &base
, &size
);
223 limits
[b
] = base
+ size
- 1;
224 addrs
[b
] = nouveau_scratch_data(&nv50
->base
, vb
->user_buffer
, base
, size
,
227 BCTX_REFN_bo(nv50
->bufctx_3d
, VERTEX_TMP
, NOUVEAU_BO_GART
|
230 nv50
->base
.vbo_dirty
= TRUE
;
234 nv50_update_user_vbufs(struct nv50_context
*nv50
)
236 uint64_t address
[PIPE_MAX_ATTRIBS
];
237 struct nouveau_pushbuf
*push
= nv50
->base
.pushbuf
;
239 uint32_t written
= 0;
241 for (i
= 0; i
< nv50
->vertex
->num_elements
; ++i
) {
242 struct pipe_vertex_element
*ve
= &nv50
->vertex
->element
[i
].pipe
;
243 const unsigned b
= ve
->vertex_buffer_index
;
244 struct pipe_vertex_buffer
*vb
= &nv50
->vtxbuf
[b
];
247 if (!(nv50
->vbo_user
& (1 << b
)))
251 nv50_emit_vtxattr(nv50
, vb
, ve
, i
);
254 nv50_user_vbuf_range(nv50
, b
, &base
, &size
);
256 if (!(written
& (1 << b
))) {
257 struct nouveau_bo
*bo
;
258 const uint32_t bo_flags
= NOUVEAU_BO_GART
| NOUVEAU_BO_RD
;
260 address
[b
] = nouveau_scratch_data(&nv50
->base
, vb
->user_buffer
,
263 BCTX_REFN_bo(nv50
->bufctx_3d
, VERTEX_TMP
, bo_flags
, bo
);
266 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_LIMIT_HIGH(i
)), 2);
267 PUSH_DATAh(push
, address
[b
] + base
+ size
- 1);
268 PUSH_DATA (push
, address
[b
] + base
+ size
- 1);
269 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_START_HIGH(i
)), 2);
270 PUSH_DATAh(push
, address
[b
] + ve
->src_offset
);
271 PUSH_DATA (push
, address
[b
] + ve
->src_offset
);
273 nv50
->base
.vbo_dirty
= TRUE
;
277 nv50_release_user_vbufs(struct nv50_context
*nv50
)
279 if (nv50
->vbo_user
) {
280 nouveau_bufctx_reset(nv50
->bufctx_3d
, NV50_BIND_VERTEX_TMP
);
281 nouveau_scratch_done(&nv50
->base
);
286 nv50_vertex_arrays_validate(struct nv50_context
*nv50
)
288 uint64_t addrs
[PIPE_MAX_ATTRIBS
];
289 uint32_t limits
[PIPE_MAX_ATTRIBS
];
290 struct nouveau_pushbuf
*push
= nv50
->base
.pushbuf
;
291 struct nv50_vertex_stateobj
*vertex
= nv50
->vertex
;
292 struct pipe_vertex_buffer
*vb
;
293 struct nv50_vertex_element
*ve
;
297 const unsigned n
= MAX2(vertex
->num_elements
, nv50
->state
.num_vtxelts
);
299 if (unlikely(vertex
->need_conversion
))
302 if (nv50
->vbo_user
& ~nv50
->vbo_constant
)
303 nv50
->vbo_fifo
= nv50
->vbo_push_hint
? ~0 : 0;
307 if (!nv50
->vbo_fifo
) {
308 /* if vertex buffer was written by GPU - flush VBO cache */
309 for (i
= 0; i
< nv50
->num_vtxbufs
; ++i
) {
310 struct nv04_resource
*buf
= nv04_resource(nv50
->vtxbuf
[i
].buffer
);
311 if (buf
&& buf
->status
& NOUVEAU_BUFFER_STATUS_GPU_WRITING
) {
312 buf
->status
&= ~NOUVEAU_BUFFER_STATUS_GPU_WRITING
;
313 nv50
->base
.vbo_dirty
= TRUE
;
319 /* update vertex format state */
320 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_ATTRIB(0)), n
);
321 if (nv50
->vbo_fifo
) {
322 nv50
->state
.num_vtxelts
= vertex
->num_elements
;
323 for (i
= 0; i
< vertex
->num_elements
; ++i
)
324 PUSH_DATA (push
, vertex
->element
[i
].state
);
326 PUSH_DATA (push
, NV50_3D_VERTEX_ATTRIB_INACTIVE
);
327 for (i
= 0; i
< n
; ++i
) {
328 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_FETCH(i
)), 1);
333 for (i
= 0; i
< vertex
->num_elements
; ++i
) {
334 const unsigned b
= vertex
->element
[i
].pipe
.vertex_buffer_index
;
335 ve
= &vertex
->element
[i
];
336 vb
= &nv50
->vtxbuf
[b
];
338 if (likely(vb
->stride
) || !(nv50
->vbo_user
& (1 << b
)))
339 PUSH_DATA(push
, ve
->state
);
341 PUSH_DATA(push
, ve
->state
| NV50_3D_VERTEX_ARRAY_ATTRIB_CONST
);
344 PUSH_DATA(push
, NV50_3D_VERTEX_ATTRIB_INACTIVE
);
346 /* update per-instance enables */
347 mask
= vertex
->instance_elts
^ nv50
->state
.instance_elts
;
349 const int i
= ffs(mask
) - 1;
351 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_PER_INSTANCE(i
)), 1);
352 PUSH_DATA (push
, (vertex
->instance_elts
>> i
) & 1);
354 nv50
->state
.instance_elts
= vertex
->instance_elts
;
356 if (nv50
->vbo_user
& ~nv50
->vbo_constant
)
357 nv50_upload_user_buffers(nv50
, addrs
, limits
);
359 /* update buffers and set constant attributes */
360 for (i
= 0; i
< vertex
->num_elements
; ++i
) {
361 uint64_t address
, limit
;
362 const unsigned b
= vertex
->element
[i
].pipe
.vertex_buffer_index
;
363 ve
= &vertex
->element
[i
];
364 vb
= &nv50
->vtxbuf
[b
];
366 if (unlikely(nv50
->vbo_constant
& (1 << b
))) {
367 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_FETCH(i
)), 1);
369 nv50_emit_vtxattr(nv50
, vb
, &ve
->pipe
, i
);
372 if (nv50
->vbo_user
& (1 << b
)) {
373 address
= addrs
[b
] + ve
->pipe
.src_offset
;
374 limit
= addrs
[b
] + limits
[b
];
376 struct nv04_resource
*buf
= nv04_resource(vb
->buffer
);
377 if (!(refd
& (1 << b
))) {
379 BCTX_REFN(nv50
->bufctx_3d
, VERTEX
, buf
, RD
);
381 address
= buf
->address
+ vb
->buffer_offset
+ ve
->pipe
.src_offset
;
382 limit
= buf
->address
+ buf
->base
.width0
- 1;
385 if (unlikely(ve
->pipe
.instance_divisor
)) {
386 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_FETCH(i
)), 4);
387 PUSH_DATA (push
, NV50_3D_VERTEX_ARRAY_FETCH_ENABLE
| vb
->stride
);
388 PUSH_DATAh(push
, address
);
389 PUSH_DATA (push
, address
);
390 PUSH_DATA (push
, ve
->pipe
.instance_divisor
);
392 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_FETCH(i
)), 3);
393 PUSH_DATA (push
, NV50_3D_VERTEX_ARRAY_FETCH_ENABLE
| vb
->stride
);
394 PUSH_DATAh(push
, address
);
395 PUSH_DATA (push
, address
);
397 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_LIMIT_HIGH(i
)), 2);
398 PUSH_DATAh(push
, limit
);
399 PUSH_DATA (push
, limit
);
401 for (; i
< nv50
->state
.num_vtxelts
; ++i
) {
402 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_FETCH(i
)), 1);
405 nv50
->state
.num_vtxelts
= vertex
->num_elements
;
408 #define NV50_PRIM_GL_CASE(n) \
409 case PIPE_PRIM_##n: return NV50_3D_VERTEX_BEGIN_GL_PRIMITIVE_##n
411 static INLINE
unsigned
412 nv50_prim_gl(unsigned prim
)
415 NV50_PRIM_GL_CASE(POINTS
);
416 NV50_PRIM_GL_CASE(LINES
);
417 NV50_PRIM_GL_CASE(LINE_LOOP
);
418 NV50_PRIM_GL_CASE(LINE_STRIP
);
419 NV50_PRIM_GL_CASE(TRIANGLES
);
420 NV50_PRIM_GL_CASE(TRIANGLE_STRIP
);
421 NV50_PRIM_GL_CASE(TRIANGLE_FAN
);
422 NV50_PRIM_GL_CASE(QUADS
);
423 NV50_PRIM_GL_CASE(QUAD_STRIP
);
424 NV50_PRIM_GL_CASE(POLYGON
);
425 NV50_PRIM_GL_CASE(LINES_ADJACENCY
);
426 NV50_PRIM_GL_CASE(LINE_STRIP_ADJACENCY
);
427 NV50_PRIM_GL_CASE(TRIANGLES_ADJACENCY
);
428 NV50_PRIM_GL_CASE(TRIANGLE_STRIP_ADJACENCY
);
430 return NV50_3D_VERTEX_BEGIN_GL_PRIMITIVE_POINTS
;
435 /* For pre-nva0 transform feedback. */
436 static const uint8_t nv50_pipe_prim_to_prim_size
[PIPE_PRIM_MAX
+ 1] =
438 [PIPE_PRIM_POINTS
] = 1,
439 [PIPE_PRIM_LINES
] = 2,
440 [PIPE_PRIM_LINE_LOOP
] = 2,
441 [PIPE_PRIM_LINE_STRIP
] = 2,
442 [PIPE_PRIM_TRIANGLES
] = 3,
443 [PIPE_PRIM_TRIANGLE_STRIP
] = 3,
444 [PIPE_PRIM_TRIANGLE_FAN
] = 3,
445 [PIPE_PRIM_QUADS
] = 3,
446 [PIPE_PRIM_QUAD_STRIP
] = 3,
447 [PIPE_PRIM_POLYGON
] = 3,
448 [PIPE_PRIM_LINES_ADJACENCY
] = 2,
449 [PIPE_PRIM_LINE_STRIP_ADJACENCY
] = 2,
450 [PIPE_PRIM_TRIANGLES_ADJACENCY
] = 3,
451 [PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY
] = 3
455 nv50_draw_arrays(struct nv50_context
*nv50
,
456 unsigned mode
, unsigned start
, unsigned count
,
457 unsigned instance_count
)
459 struct nouveau_pushbuf
*push
= nv50
->base
.pushbuf
;
462 if (nv50
->state
.index_bias
) {
463 BEGIN_NV04(push
, NV50_3D(VB_ELEMENT_BASE
), 1);
465 nv50
->state
.index_bias
= 0;
468 prim
= nv50_prim_gl(mode
);
470 while (instance_count
--) {
471 BEGIN_NV04(push
, NV50_3D(VERTEX_BEGIN_GL
), 1);
472 PUSH_DATA (push
, prim
);
473 BEGIN_NV04(push
, NV50_3D(VERTEX_BUFFER_FIRST
), 2);
474 PUSH_DATA (push
, start
);
475 PUSH_DATA (push
, count
);
476 BEGIN_NV04(push
, NV50_3D(VERTEX_END_GL
), 1);
479 prim
|= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT
;
484 nv50_draw_elements_inline_u08(struct nouveau_pushbuf
*push
, const uint8_t *map
,
485 unsigned start
, unsigned count
)
491 BEGIN_NI04(push
, NV50_3D(VB_ELEMENT_U32
), count
& 3);
492 for (i
= 0; i
< (count
& 3); ++i
)
493 PUSH_DATA(push
, *map
++);
497 unsigned i
, nr
= MIN2(count
, NV04_PFIFO_MAX_PACKET_LEN
* 4) / 4;
499 BEGIN_NI04(push
, NV50_3D(VB_ELEMENT_U8
), nr
);
500 for (i
= 0; i
< nr
; ++i
) {
502 (map
[3] << 24) | (map
[2] << 16) | (map
[1] << 8) | map
[0]);
510 nv50_draw_elements_inline_u16(struct nouveau_pushbuf
*push
, const uint16_t *map
,
511 unsigned start
, unsigned count
)
517 BEGIN_NV04(push
, NV50_3D(VB_ELEMENT_U32
), 1);
518 PUSH_DATA (push
, *map
++);
521 unsigned i
, nr
= MIN2(count
, NV04_PFIFO_MAX_PACKET_LEN
* 2) / 2;
523 BEGIN_NI04(push
, NV50_3D(VB_ELEMENT_U16
), nr
);
524 for (i
= 0; i
< nr
; ++i
) {
525 PUSH_DATA(push
, (map
[1] << 16) | map
[0]);
533 nv50_draw_elements_inline_u32(struct nouveau_pushbuf
*push
, const uint32_t *map
,
534 unsigned start
, unsigned count
)
539 const unsigned nr
= MIN2(count
, NV04_PFIFO_MAX_PACKET_LEN
);
541 BEGIN_NI04(push
, NV50_3D(VB_ELEMENT_U32
), nr
);
542 PUSH_DATAp(push
, map
, nr
);
550 nv50_draw_elements_inline_u32_short(struct nouveau_pushbuf
*push
,
552 unsigned start
, unsigned count
)
558 BEGIN_NV04(push
, NV50_3D(VB_ELEMENT_U32
), 1);
559 PUSH_DATA (push
, *map
++);
562 unsigned i
, nr
= MIN2(count
, NV04_PFIFO_MAX_PACKET_LEN
* 2) / 2;
564 BEGIN_NI04(push
, NV50_3D(VB_ELEMENT_U16
), nr
);
565 for (i
= 0; i
< nr
; ++i
) {
566 PUSH_DATA(push
, (map
[1] << 16) | map
[0]);
574 nv50_draw_elements(struct nv50_context
*nv50
, boolean shorten
,
575 unsigned mode
, unsigned start
, unsigned count
,
576 unsigned instance_count
, int32_t index_bias
)
578 struct nouveau_pushbuf
*push
= nv50
->base
.pushbuf
;
580 const unsigned index_size
= nv50
->idxbuf
.index_size
;
582 prim
= nv50_prim_gl(mode
);
584 if (index_bias
!= nv50
->state
.index_bias
) {
585 BEGIN_NV04(push
, NV50_3D(VB_ELEMENT_BASE
), 1);
586 PUSH_DATA (push
, index_bias
);
587 nv50
->state
.index_bias
= index_bias
;
590 if (nv50
->idxbuf
.buffer
) {
591 struct nv04_resource
*buf
= nv04_resource(nv50
->idxbuf
.buffer
);
594 const unsigned base
= (buf
->offset
+ nv50
->idxbuf
.offset
) & ~3;
596 start
+= ((buf
->offset
+ nv50
->idxbuf
.offset
) & 3) >> (index_size
>> 1);
598 assert(nouveau_resource_mapped_by_gpu(nv50
->idxbuf
.buffer
));
600 while (instance_count
--) {
601 BEGIN_NV04(push
, NV50_3D(VERTEX_BEGIN_GL
), 1);
602 PUSH_DATA (push
, prim
);
604 nouveau_pushbuf_space(push
, 8, 0, 1);
606 switch (index_size
) {
608 BEGIN_NL50(push
, NV50_3D(VB_ELEMENT_U32
), count
);
609 nouveau_pushbuf_data(push
, buf
->bo
, base
+ start
* 4, count
* 4);
612 pb_start
= (start
& ~1) * 2;
613 pb_bytes
= ((start
+ count
+ 1) & ~1) * 2 - pb_start
;
615 BEGIN_NV04(push
, NV50_3D(VB_ELEMENT_U16_SETUP
), 1);
616 PUSH_DATA (push
, (start
<< 31) | count
);
617 BEGIN_NL50(push
, NV50_3D(VB_ELEMENT_U16
), pb_bytes
/ 4);
618 nouveau_pushbuf_data(push
, buf
->bo
, base
+ pb_start
, pb_bytes
);
619 BEGIN_NV04(push
, NV50_3D(VB_ELEMENT_U16_SETUP
), 1);
623 assert(index_size
== 1);
624 pb_start
= start
& ~3;
625 pb_bytes
= ((start
+ count
+ 3) & ~3) - pb_start
;
627 BEGIN_NV04(push
, NV50_3D(VB_ELEMENT_U8_SETUP
), 1);
628 PUSH_DATA (push
, (start
<< 30) | count
);
629 BEGIN_NL50(push
, NV50_3D(VB_ELEMENT_U8
), pb_bytes
/ 4);
630 nouveau_pushbuf_data(push
, buf
->bo
, base
+ pb_start
, pb_bytes
);
631 BEGIN_NV04(push
, NV50_3D(VB_ELEMENT_U8_SETUP
), 1);
635 BEGIN_NV04(push
, NV50_3D(VERTEX_END_GL
), 1);
638 prim
|= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT
;
641 const void *data
= nv50
->idxbuf
.user_buffer
;
643 while (instance_count
--) {
644 BEGIN_NV04(push
, NV50_3D(VERTEX_BEGIN_GL
), 1);
645 PUSH_DATA (push
, prim
);
646 switch (index_size
) {
648 nv50_draw_elements_inline_u08(push
, data
, start
, count
);
651 nv50_draw_elements_inline_u16(push
, data
, start
, count
);
655 nv50_draw_elements_inline_u32_short(push
, data
, start
, count
);
657 nv50_draw_elements_inline_u32(push
, data
, start
, count
);
663 BEGIN_NV04(push
, NV50_3D(VERTEX_END_GL
), 1);
666 prim
|= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT
;
672 nva0_draw_stream_output(struct nv50_context
*nv50
,
673 const struct pipe_draw_info
*info
)
675 struct nouveau_pushbuf
*push
= nv50
->base
.pushbuf
;
676 struct nv50_so_target
*so
= nv50_so_target(info
->count_from_stream_output
);
677 struct nv04_resource
*res
= nv04_resource(so
->pipe
.buffer
);
678 unsigned num_instances
= info
->instance_count
;
679 unsigned mode
= nv50_prim_gl(info
->mode
);
681 if (unlikely(nv50
->screen
->base
.class_3d
< NVA0_3D_CLASS
)) {
682 /* A proper implementation without waiting doesn't seem possible,
685 NOUVEAU_ERR("draw_stream_output not supported on pre-NVA0 cards\n");
689 if (res
->status
& NOUVEAU_BUFFER_STATUS_GPU_WRITING
) {
690 res
->status
&= ~NOUVEAU_BUFFER_STATUS_GPU_WRITING
;
692 BEGIN_NV04(push
, SUBC_3D(NV50_GRAPH_SERIALIZE
), 1);
694 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_FLUSH
), 1);
698 assert(num_instances
);
701 BEGIN_NV04(push
, NV50_3D(VERTEX_BEGIN_GL
), 1);
702 PUSH_DATA (push
, mode
);
703 BEGIN_NV04(push
, NVA0_3D(DRAW_TFB_BASE
), 1);
705 BEGIN_NV04(push
, NVA0_3D(DRAW_TFB_STRIDE
), 1);
707 BEGIN_NV04(push
, NVA0_3D(DRAW_TFB_BYTES
), 1);
708 nv50_query_pushbuf_submit(push
, so
->pq
, 0x4);
709 BEGIN_NV04(push
, NV50_3D(VERTEX_END_GL
), 1);
712 mode
|= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT
;
713 } while (--num_instances
);
717 nv50_draw_vbo_kick_notify(struct nouveau_pushbuf
*chan
)
719 struct nv50_screen
*screen
= chan
->user_priv
;
721 nouveau_fence_update(&screen
->base
, TRUE
);
723 nv50_bufctx_fence(screen
->cur_ctx
->bufctx_3d
, TRUE
);
727 nv50_draw_vbo(struct pipe_context
*pipe
, const struct pipe_draw_info
*info
)
729 struct nv50_context
*nv50
= nv50_context(pipe
);
730 struct nouveau_pushbuf
*push
= nv50
->base
.pushbuf
;
732 /* NOTE: caller must ensure that (min_index + index_bias) is >= 0 */
733 nv50
->vb_elt_first
= info
->min_index
+ info
->index_bias
;
734 nv50
->vb_elt_limit
= info
->max_index
- info
->min_index
;
735 nv50
->instance_off
= info
->start_instance
;
736 nv50
->instance_max
= info
->instance_count
- 1;
738 /* For picking only a few vertices from a large user buffer, push is better,
739 * if index count is larger and we expect repeated vertices, suggest upload.
741 nv50
->vbo_push_hint
= /* the 64 is heuristic */
742 !(info
->indexed
&& ((nv50
->vb_elt_limit
+ 64) < info
->count
));
744 if (nv50
->vbo_user
&& !(nv50
->dirty
& (NV50_NEW_ARRAYS
| NV50_NEW_VERTEX
))) {
745 if (!!nv50
->vbo_fifo
!= nv50
->vbo_push_hint
)
746 nv50
->dirty
|= NV50_NEW_ARRAYS
;
749 nv50_update_user_vbufs(nv50
);
752 if (unlikely(nv50
->num_so_targets
&& !nv50
->gmtyprog
))
753 nv50
->state
.prim_size
= nv50_pipe_prim_to_prim_size
[info
->mode
];
755 nv50_state_validate(nv50
, ~0, 8); /* 8 as minimum, we use flush_notify */
757 push
->kick_notify
= nv50_draw_vbo_kick_notify
;
759 if (nv50
->vbo_fifo
) {
760 nv50_push_vbo(nv50
, info
);
761 push
->kick_notify
= nv50_default_kick_notify
;
762 nouveau_pushbuf_bufctx(push
, NULL
);
766 if (nv50
->state
.instance_base
!= info
->start_instance
) {
767 nv50
->state
.instance_base
= info
->start_instance
;
768 /* NOTE: this does not affect the shader input, should it ? */
769 BEGIN_NV04(push
, NV50_3D(VB_INSTANCE_BASE
), 1);
770 PUSH_DATA (push
, info
->start_instance
);
773 if (nv50
->base
.vbo_dirty
) {
774 BEGIN_NV04(push
, NV50_3D(VERTEX_ARRAY_FLUSH
), 1);
776 nv50
->base
.vbo_dirty
= FALSE
;
780 boolean shorten
= info
->max_index
<= 65535;
782 if (info
->primitive_restart
!= nv50
->state
.prim_restart
) {
783 if (info
->primitive_restart
) {
784 BEGIN_NV04(push
, NV50_3D(PRIM_RESTART_ENABLE
), 2);
786 PUSH_DATA (push
, info
->restart_index
);
788 if (info
->restart_index
> 65535)
791 BEGIN_NV04(push
, NV50_3D(PRIM_RESTART_ENABLE
), 1);
794 nv50
->state
.prim_restart
= info
->primitive_restart
;
796 if (info
->primitive_restart
) {
797 BEGIN_NV04(push
, NV50_3D(PRIM_RESTART_INDEX
), 1);
798 PUSH_DATA (push
, info
->restart_index
);
800 if (info
->restart_index
> 65535)
804 nv50_draw_elements(nv50
, shorten
,
805 info
->mode
, info
->start
, info
->count
,
806 info
->instance_count
, info
->index_bias
);
808 if (unlikely(info
->count_from_stream_output
)) {
809 nva0_draw_stream_output(nv50
, info
);
811 nv50_draw_arrays(nv50
,
812 info
->mode
, info
->start
, info
->count
,
813 info
->instance_count
);
815 push
->kick_notify
= nv50_default_kick_notify
;
817 nv50_release_user_vbufs(nv50
);
819 nouveau_pushbuf_bufctx(push
, NULL
);