2 * Copyright 2003 VMware, Inc.
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the
7 * "Software"), to deal in the Software without restriction, including
8 * without limitation the rights to use, copy, modify, merge, publish,
9 * distribute, sublicense, and/or sell copies of the Software, and to
10 * permit persons to whom the Software is furnished to do so, subject to
11 * the following conditions:
13 * The above copyright notice and this permission notice (including the
14 * next paragraph) shall be included in all copies or substantial portions
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
19 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
20 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
21 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
22 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
23 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 #include "main/arrayobj.h"
27 #include "main/bufferobj.h"
28 #include "main/context.h"
29 #include "main/enums.h"
30 #include "main/macros.h"
31 #include "main/glformats.h"
35 #include "brw_defines.h"
36 #include "brw_context.h"
37 #include "brw_state.h"
39 #include "intel_batchbuffer.h"
40 #include "intel_buffer_objects.h"
42 static const GLuint double_types_float
[5] = {
45 ISL_FORMAT_R64G64_FLOAT
,
46 ISL_FORMAT_R64G64B64_FLOAT
,
47 ISL_FORMAT_R64G64B64A64_FLOAT
50 static const GLuint double_types_passthru
[5] = {
52 ISL_FORMAT_R64_PASSTHRU
,
53 ISL_FORMAT_R64G64_PASSTHRU
,
54 ISL_FORMAT_R64G64B64_PASSTHRU
,
55 ISL_FORMAT_R64G64B64A64_PASSTHRU
58 static const GLuint float_types
[5] = {
61 ISL_FORMAT_R32G32_FLOAT
,
62 ISL_FORMAT_R32G32B32_FLOAT
,
63 ISL_FORMAT_R32G32B32A32_FLOAT
66 static const GLuint half_float_types
[5] = {
69 ISL_FORMAT_R16G16_FLOAT
,
70 ISL_FORMAT_R16G16B16_FLOAT
,
71 ISL_FORMAT_R16G16B16A16_FLOAT
74 static const GLuint fixed_point_types
[5] = {
76 ISL_FORMAT_R32_SFIXED
,
77 ISL_FORMAT_R32G32_SFIXED
,
78 ISL_FORMAT_R32G32B32_SFIXED
,
79 ISL_FORMAT_R32G32B32A32_SFIXED
,
82 static const GLuint uint_types_direct
[5] = {
85 ISL_FORMAT_R32G32_UINT
,
86 ISL_FORMAT_R32G32B32_UINT
,
87 ISL_FORMAT_R32G32B32A32_UINT
90 static const GLuint uint_types_norm
[5] = {
93 ISL_FORMAT_R32G32_UNORM
,
94 ISL_FORMAT_R32G32B32_UNORM
,
95 ISL_FORMAT_R32G32B32A32_UNORM
98 static const GLuint uint_types_scale
[5] = {
100 ISL_FORMAT_R32_USCALED
,
101 ISL_FORMAT_R32G32_USCALED
,
102 ISL_FORMAT_R32G32B32_USCALED
,
103 ISL_FORMAT_R32G32B32A32_USCALED
106 static const GLuint int_types_direct
[5] = {
109 ISL_FORMAT_R32G32_SINT
,
110 ISL_FORMAT_R32G32B32_SINT
,
111 ISL_FORMAT_R32G32B32A32_SINT
114 static const GLuint int_types_norm
[5] = {
116 ISL_FORMAT_R32_SNORM
,
117 ISL_FORMAT_R32G32_SNORM
,
118 ISL_FORMAT_R32G32B32_SNORM
,
119 ISL_FORMAT_R32G32B32A32_SNORM
122 static const GLuint int_types_scale
[5] = {
124 ISL_FORMAT_R32_SSCALED
,
125 ISL_FORMAT_R32G32_SSCALED
,
126 ISL_FORMAT_R32G32B32_SSCALED
,
127 ISL_FORMAT_R32G32B32A32_SSCALED
130 static const GLuint ushort_types_direct
[5] = {
133 ISL_FORMAT_R16G16_UINT
,
134 ISL_FORMAT_R16G16B16_UINT
,
135 ISL_FORMAT_R16G16B16A16_UINT
138 static const GLuint ushort_types_norm
[5] = {
140 ISL_FORMAT_R16_UNORM
,
141 ISL_FORMAT_R16G16_UNORM
,
142 ISL_FORMAT_R16G16B16_UNORM
,
143 ISL_FORMAT_R16G16B16A16_UNORM
146 static const GLuint ushort_types_scale
[5] = {
148 ISL_FORMAT_R16_USCALED
,
149 ISL_FORMAT_R16G16_USCALED
,
150 ISL_FORMAT_R16G16B16_USCALED
,
151 ISL_FORMAT_R16G16B16A16_USCALED
154 static const GLuint short_types_direct
[5] = {
157 ISL_FORMAT_R16G16_SINT
,
158 ISL_FORMAT_R16G16B16_SINT
,
159 ISL_FORMAT_R16G16B16A16_SINT
162 static const GLuint short_types_norm
[5] = {
164 ISL_FORMAT_R16_SNORM
,
165 ISL_FORMAT_R16G16_SNORM
,
166 ISL_FORMAT_R16G16B16_SNORM
,
167 ISL_FORMAT_R16G16B16A16_SNORM
170 static const GLuint short_types_scale
[5] = {
172 ISL_FORMAT_R16_SSCALED
,
173 ISL_FORMAT_R16G16_SSCALED
,
174 ISL_FORMAT_R16G16B16_SSCALED
,
175 ISL_FORMAT_R16G16B16A16_SSCALED
178 static const GLuint ubyte_types_direct
[5] = {
181 ISL_FORMAT_R8G8_UINT
,
182 ISL_FORMAT_R8G8B8_UINT
,
183 ISL_FORMAT_R8G8B8A8_UINT
186 static const GLuint ubyte_types_norm
[5] = {
189 ISL_FORMAT_R8G8_UNORM
,
190 ISL_FORMAT_R8G8B8_UNORM
,
191 ISL_FORMAT_R8G8B8A8_UNORM
194 static const GLuint ubyte_types_scale
[5] = {
196 ISL_FORMAT_R8_USCALED
,
197 ISL_FORMAT_R8G8_USCALED
,
198 ISL_FORMAT_R8G8B8_USCALED
,
199 ISL_FORMAT_R8G8B8A8_USCALED
202 static const GLuint byte_types_direct
[5] = {
205 ISL_FORMAT_R8G8_SINT
,
206 ISL_FORMAT_R8G8B8_SINT
,
207 ISL_FORMAT_R8G8B8A8_SINT
210 static const GLuint byte_types_norm
[5] = {
213 ISL_FORMAT_R8G8_SNORM
,
214 ISL_FORMAT_R8G8B8_SNORM
,
215 ISL_FORMAT_R8G8B8A8_SNORM
218 static const GLuint byte_types_scale
[5] = {
220 ISL_FORMAT_R8_SSCALED
,
221 ISL_FORMAT_R8G8_SSCALED
,
222 ISL_FORMAT_R8G8B8_SSCALED
,
223 ISL_FORMAT_R8G8B8A8_SSCALED
227 double_types(int size
, GLboolean doubles
)
229 /* From the BDW PRM, Volume 2d, page 588 (VERTEX_ELEMENT_STATE):
230 * "When SourceElementFormat is set to one of the *64*_PASSTHRU formats,
231 * 64-bit components are stored in the URB without any conversion."
232 * Also included on BDW PRM, Volume 7, page 470, table "Source Element
233 * Formats Supported in VF Unit"
235 * Previous PRMs don't include those references, so for gen7 we can't use
236 * PASSTHRU formats directly. But in any case, we prefer to return passthru
237 * even in that case, because that reflects what we want to achieve, even
238 * if we would need to workaround on gen < 8.
241 ? double_types_passthru
[size
]
242 : double_types_float
[size
]);
246 * Given vertex array type/size/format/normalized info, return
247 * the appopriate hardware surface type.
248 * Format will be GL_RGBA or possibly GL_BGRA for GLubyte[4] color arrays.
251 brw_get_vertex_surface_type(struct brw_context
*brw
,
252 const struct gl_vertex_format
*glformat
)
254 int size
= glformat
->Size
;
255 const struct gen_device_info
*devinfo
= &brw
->screen
->devinfo
;
256 const bool is_ivybridge_or_older
=
257 devinfo
->gen
<= 7 && !devinfo
->is_baytrail
&& !devinfo
->is_haswell
;
259 if (unlikely(INTEL_DEBUG
& DEBUG_VERTS
))
260 fprintf(stderr
, "type %s size %d normalized %d\n",
261 _mesa_enum_to_string(glformat
->Type
),
262 glformat
->Size
, glformat
->Normalized
);
264 if (glformat
->Integer
) {
265 assert(glformat
->Format
== GL_RGBA
); /* sanity check */
266 switch (glformat
->Type
) {
267 case GL_INT
: return int_types_direct
[size
];
269 if (is_ivybridge_or_older
&& size
== 3)
270 return short_types_direct
[4];
272 return short_types_direct
[size
];
274 if (is_ivybridge_or_older
&& size
== 3)
275 return byte_types_direct
[4];
277 return byte_types_direct
[size
];
278 case GL_UNSIGNED_INT
: return uint_types_direct
[size
];
279 case GL_UNSIGNED_SHORT
:
280 if (is_ivybridge_or_older
&& size
== 3)
281 return ushort_types_direct
[4];
283 return ushort_types_direct
[size
];
284 case GL_UNSIGNED_BYTE
:
285 if (is_ivybridge_or_older
&& size
== 3)
286 return ubyte_types_direct
[4];
288 return ubyte_types_direct
[size
];
289 default: unreachable("not reached");
291 } else if (glformat
->Type
== GL_UNSIGNED_INT_10F_11F_11F_REV
) {
292 return ISL_FORMAT_R11G11B10_FLOAT
;
293 } else if (glformat
->Normalized
) {
294 switch (glformat
->Type
) {
295 case GL_DOUBLE
: return double_types(size
, glformat
->Doubles
);
296 case GL_FLOAT
: return float_types
[size
];
298 case GL_HALF_FLOAT_OES
:
299 if (devinfo
->gen
< 6 && size
== 3)
300 return half_float_types
[4];
302 return half_float_types
[size
];
303 case GL_INT
: return int_types_norm
[size
];
304 case GL_SHORT
: return short_types_norm
[size
];
305 case GL_BYTE
: return byte_types_norm
[size
];
306 case GL_UNSIGNED_INT
: return uint_types_norm
[size
];
307 case GL_UNSIGNED_SHORT
: return ushort_types_norm
[size
];
308 case GL_UNSIGNED_BYTE
:
309 if (glformat
->Format
== GL_BGRA
) {
310 /* See GL_EXT_vertex_array_bgra */
312 return ISL_FORMAT_B8G8R8A8_UNORM
;
315 return ubyte_types_norm
[size
];
318 if (devinfo
->gen
>= 8 || devinfo
->is_haswell
)
319 return fixed_point_types
[size
];
321 /* This produces GL_FIXED inputs as values between INT32_MIN and
322 * INT32_MAX, which will be scaled down by 1/65536 by the VS.
324 return int_types_scale
[size
];
325 /* See GL_ARB_vertex_type_2_10_10_10_rev.
326 * W/A: Pre-Haswell, the hardware doesn't really support the formats we'd
327 * like to use here, so upload everything as UINT and fix
330 case GL_INT_2_10_10_10_REV
:
332 if (devinfo
->gen
>= 8 || devinfo
->is_haswell
) {
333 return glformat
->Format
== GL_BGRA
334 ? ISL_FORMAT_B10G10R10A2_SNORM
335 : ISL_FORMAT_R10G10B10A2_SNORM
;
337 return ISL_FORMAT_R10G10B10A2_UINT
;
338 case GL_UNSIGNED_INT_2_10_10_10_REV
:
340 if (devinfo
->gen
>= 8 || devinfo
->is_haswell
) {
341 return glformat
->Format
== GL_BGRA
342 ? ISL_FORMAT_B10G10R10A2_UNORM
343 : ISL_FORMAT_R10G10B10A2_UNORM
;
345 return ISL_FORMAT_R10G10B10A2_UINT
;
346 default: unreachable("not reached");
350 /* See GL_ARB_vertex_type_2_10_10_10_rev.
351 * W/A: the hardware doesn't really support the formats we'd
352 * like to use here, so upload everything as UINT and fix
355 if (glformat
->Type
== GL_INT_2_10_10_10_REV
) {
357 if (devinfo
->gen
>= 8 || devinfo
->is_haswell
) {
358 return glformat
->Format
== GL_BGRA
359 ? ISL_FORMAT_B10G10R10A2_SSCALED
360 : ISL_FORMAT_R10G10B10A2_SSCALED
;
362 return ISL_FORMAT_R10G10B10A2_UINT
;
363 } else if (glformat
->Type
== GL_UNSIGNED_INT_2_10_10_10_REV
) {
365 if (devinfo
->gen
>= 8 || devinfo
->is_haswell
) {
366 return glformat
->Format
== GL_BGRA
367 ? ISL_FORMAT_B10G10R10A2_USCALED
368 : ISL_FORMAT_R10G10B10A2_USCALED
;
370 return ISL_FORMAT_R10G10B10A2_UINT
;
372 assert(glformat
->Format
== GL_RGBA
); /* sanity check */
373 switch (glformat
->Type
) {
374 case GL_DOUBLE
: return double_types(size
, glformat
->Doubles
);
375 case GL_FLOAT
: return float_types
[size
];
377 case GL_HALF_FLOAT_OES
:
378 if (devinfo
->gen
< 6 && size
== 3)
379 return half_float_types
[4];
381 return half_float_types
[size
];
382 case GL_INT
: return int_types_scale
[size
];
383 case GL_SHORT
: return short_types_scale
[size
];
384 case GL_BYTE
: return byte_types_scale
[size
];
385 case GL_UNSIGNED_INT
: return uint_types_scale
[size
];
386 case GL_UNSIGNED_SHORT
: return ushort_types_scale
[size
];
387 case GL_UNSIGNED_BYTE
: return ubyte_types_scale
[size
];
389 if (devinfo
->gen
>= 8 || devinfo
->is_haswell
)
390 return fixed_point_types
[size
];
392 /* This produces GL_FIXED inputs as values between INT32_MIN and
393 * INT32_MAX, which will be scaled down by 1/65536 by the VS.
395 return int_types_scale
[size
];
396 default: unreachable("not reached");
402 copy_array_to_vbo_array(struct brw_context
*brw
,
403 const uint8_t *const ptr
, const int src_stride
,
405 struct brw_vertex_buffer
*buffer
,
408 const unsigned char *src
= ptr
+ min
* src_stride
;
409 int count
= max
- min
+ 1;
410 GLuint size
= count
* dst_stride
;
411 uint8_t *dst
= brw_upload_space(&brw
->upload
, size
, dst_stride
,
412 &buffer
->bo
, &buffer
->offset
);
414 /* The GL 4.5 spec says:
415 * "If any enabled array’s buffer binding is zero when DrawArrays or
416 * one of the other drawing commands defined in section 10.4 is called,
417 * the result is undefined."
419 * In this case, let's the dst with undefined values
422 if (dst_stride
== src_stride
) {
423 memcpy(dst
, src
, size
);
426 memcpy(dst
, src
, dst_stride
);
432 buffer
->stride
= dst_stride
;
437 brw_prepare_vertices(struct brw_context
*brw
)
439 const struct gen_device_info
*devinfo
= &brw
->screen
->devinfo
;
440 struct gl_context
*ctx
= &brw
->ctx
;
441 /* BRW_NEW_VERTEX_PROGRAM */
442 const struct gl_program
*vp
= brw
->programs
[MESA_SHADER_VERTEX
];
443 /* BRW_NEW_VS_PROG_DATA */
444 const struct brw_vs_prog_data
*vs_prog_data
=
445 brw_vs_prog_data(brw
->vs
.base
.prog_data
);
446 const uint64_t vs_inputs64
=
447 nir_get_single_slot_attribs_mask(vs_prog_data
->inputs_read
,
449 assert((vs_inputs64
& ~(uint64_t)VERT_BIT_ALL
) == 0);
450 unsigned vs_inputs
= (unsigned)vs_inputs64
;
451 unsigned int min_index
= brw
->vb
.min_index
+ brw
->basevertex
;
452 unsigned int max_index
= brw
->vb
.max_index
+ brw
->basevertex
;
457 * On gen6+, edge flags don't end up in the VUE (either in or out of the
458 * VS). Instead, they're uploaded as the last vertex element, and the data
459 * is passed sideband through the fixed function units. So, we need to
460 * prepare the vertex buffer for it, but it's not present in inputs_read.
462 if (devinfo
->gen
>= 6 && (ctx
->Polygon
.FrontMode
!= GL_FILL
||
463 ctx
->Polygon
.BackMode
!= GL_FILL
)) {
464 vs_inputs
|= VERT_BIT_EDGEFLAG
;
468 fprintf(stderr
, "%s %d..%d\n", __func__
, min_index
, max_index
);
470 /* Accumulate the list of enabled arrays. */
471 brw
->vb
.nr_enabled
= 0;
473 unsigned mask
= vs_inputs
;
475 const gl_vert_attrib attr
= u_bit_scan(&mask
);
476 struct brw_vertex_element
*input
= &brw
->vb
.inputs
[attr
];
477 brw
->vb
.enabled
[brw
->vb
.nr_enabled
++] = input
;
479 assert(brw
->vb
.nr_enabled
<= VERT_ATTRIB_MAX
);
481 if (brw
->vb
.nr_enabled
== 0)
484 if (brw
->vb
.nr_buffers
)
488 const struct gl_vertex_array_object
*vao
= ctx
->Array
._DrawVAO
;
490 unsigned vbomask
= vs_inputs
& _mesa_draw_vbo_array_bits(ctx
);
492 const struct gl_vertex_buffer_binding
*const glbinding
=
493 _mesa_draw_buffer_binding(vao
, ffs(vbomask
) - 1);
494 const GLsizei stride
= glbinding
->Stride
;
496 assert(_mesa_is_bufferobj(glbinding
->BufferObj
));
498 /* Accumulate the range of a single vertex, start with inverted range */
499 uint32_t vertex_range_start
= ~(uint32_t)0;
500 uint32_t vertex_range_end
= 0;
502 const unsigned boundmask
= _mesa_draw_bound_attrib_bits(glbinding
);
503 unsigned attrmask
= vbomask
& boundmask
;
504 /* Mark the those attributes as processed */
506 /* We can assume that we have an array for the binding */
508 /* Walk attributes belonging to the binding */
510 const gl_vert_attrib attr
= u_bit_scan(&attrmask
);
511 const struct gl_array_attributes
*const glattrib
=
512 _mesa_draw_array_attrib(vao
, attr
);
513 const uint32_t rel_offset
=
514 _mesa_draw_attributes_relative_offset(glattrib
);
515 const uint32_t rel_end
= rel_offset
+ glattrib
->Format
._ElementSize
;
517 vertex_range_start
= MIN2(vertex_range_start
, rel_offset
);
518 vertex_range_end
= MAX2(vertex_range_end
, rel_end
);
520 struct brw_vertex_element
*input
= &brw
->vb
.inputs
[attr
];
521 input
->glattrib
= glattrib
;
523 input
->is_dual_slot
= (vp
->DualSlotInputs
& BITFIELD64_BIT(attr
)) != 0;
524 input
->offset
= rel_offset
;
526 assert(vertex_range_start
<= vertex_range_end
);
528 struct intel_buffer_object
*intel_buffer
=
529 intel_buffer_object(glbinding
->BufferObj
);
530 struct brw_vertex_buffer
*buffer
= &brw
->vb
.buffers
[j
];
532 const uint32_t offset
= _mesa_draw_binding_offset(glbinding
);
534 /* If nothing else is known take the buffer size and offset as a bound */
535 uint32_t start
= vertex_range_start
;
536 uint32_t range
= intel_buffer
->Base
.Size
- offset
- vertex_range_start
;
537 /* Check if we can get a more narrow range */
538 if (glbinding
->InstanceDivisor
) {
539 if (brw
->num_instances
) {
540 const uint32_t vertex_size
= vertex_range_end
- vertex_range_start
;
541 start
= vertex_range_start
+ stride
* brw
->baseinstance
;
542 range
= (stride
* ((brw
->num_instances
- 1) /
543 glbinding
->InstanceDivisor
) +
547 if (brw
->vb
.index_bounds_valid
) {
548 const uint32_t vertex_size
= vertex_range_end
- vertex_range_start
;
549 start
= vertex_range_start
+ stride
* min_index
;
550 range
= (stride
* (max_index
- min_index
) +
555 buffer
->offset
= offset
;
556 buffer
->size
= start
+ range
;
557 buffer
->stride
= stride
;
558 buffer
->step_rate
= glbinding
->InstanceDivisor
;
560 buffer
->bo
= intel_bufferobj_buffer(brw
, intel_buffer
, offset
+ start
,
562 brw_bo_reference(buffer
->bo
);
567 /* If we need to upload all the arrays, then we can trim those arrays to
568 * only the used elements [min_index, max_index] so long as we adjust all
569 * the values used in the 3DPRIMITIVE i.e. by setting the vertex bias.
571 brw
->vb
.start_vertex_bias
= 0;
573 if ((vs_inputs
& _mesa_draw_vbo_array_bits(ctx
)) == 0) {
574 brw
->vb
.start_vertex_bias
= -delta
;
578 unsigned usermask
= vs_inputs
& _mesa_draw_user_array_bits(ctx
);
580 const struct gl_vertex_buffer_binding
*const glbinding
=
581 _mesa_draw_buffer_binding(vao
, ffs(usermask
) - 1);
582 const GLsizei stride
= glbinding
->Stride
;
584 assert(!_mesa_is_bufferobj(glbinding
->BufferObj
));
585 assert(brw
->vb
.index_bounds_valid
);
587 /* Accumulate the range of a single vertex, start with inverted range */
588 uint32_t vertex_range_start
= ~(uint32_t)0;
589 uint32_t vertex_range_end
= 0;
591 const unsigned boundmask
= _mesa_draw_bound_attrib_bits(glbinding
);
592 unsigned attrmask
= usermask
& boundmask
;
593 /* Mark the those attributes as processed */
594 usermask
^= attrmask
;
595 /* We can assume that we have an array for the binding */
597 /* Walk attributes belonging to the binding */
599 const gl_vert_attrib attr
= u_bit_scan(&attrmask
);
600 const struct gl_array_attributes
*const glattrib
=
601 _mesa_draw_array_attrib(vao
, attr
);
602 const uint32_t rel_offset
=
603 _mesa_draw_attributes_relative_offset(glattrib
);
604 const uint32_t rel_end
= rel_offset
+ glattrib
->Format
._ElementSize
;
606 vertex_range_start
= MIN2(vertex_range_start
, rel_offset
);
607 vertex_range_end
= MAX2(vertex_range_end
, rel_end
);
609 struct brw_vertex_element
*input
= &brw
->vb
.inputs
[attr
];
610 input
->glattrib
= glattrib
;
612 input
->is_dual_slot
= (vp
->DualSlotInputs
& BITFIELD64_BIT(attr
)) != 0;
613 input
->offset
= rel_offset
;
615 assert(vertex_range_start
<= vertex_range_end
);
617 struct brw_vertex_buffer
*buffer
= &brw
->vb
.buffers
[j
];
619 const uint8_t *ptr
= (const uint8_t*)_mesa_draw_binding_offset(glbinding
);
620 ptr
+= vertex_range_start
;
621 const uint32_t vertex_size
= vertex_range_end
- vertex_range_start
;
622 if (glbinding
->Stride
== 0) {
623 /* If the source stride is zero, we just want to upload the current
624 * attribute once and set the buffer's stride to 0. There's no need
625 * to replicate it out.
627 copy_array_to_vbo_array(brw
, ptr
, 0, 0, 0, buffer
, vertex_size
);
628 } else if (glbinding
->InstanceDivisor
== 0) {
629 copy_array_to_vbo_array(brw
, ptr
, stride
, min_index
,
630 max_index
, buffer
, vertex_size
);
632 /* This is an instanced attribute, since its InstanceDivisor
633 * is not zero. Therefore, its data will be stepped after the
634 * instanced draw has been run InstanceDivisor times.
636 uint32_t instanced_attr_max_index
=
637 (brw
->num_instances
- 1) / glbinding
->InstanceDivisor
;
638 copy_array_to_vbo_array(brw
, ptr
, stride
, 0,
639 instanced_attr_max_index
, buffer
, vertex_size
);
641 buffer
->offset
-= delta
* buffer
->stride
+ vertex_range_start
;
642 buffer
->size
+= delta
* buffer
->stride
+ vertex_range_start
;
643 buffer
->step_rate
= glbinding
->InstanceDivisor
;
648 /* Upload the current values */
649 unsigned curmask
= vs_inputs
& _mesa_draw_current_bits(ctx
);
651 /* For each attribute, upload the maximum possible size. */
652 uint8_t data
[VERT_ATTRIB_MAX
* sizeof(GLdouble
) * 4];
653 uint8_t *cursor
= data
;
656 const gl_vert_attrib attr
= u_bit_scan(&curmask
);
657 const struct gl_array_attributes
*const glattrib
=
658 _mesa_draw_current_attrib(ctx
, attr
);
659 const unsigned size
= glattrib
->Format
._ElementSize
;
660 const unsigned alignment
= align(size
, sizeof(GLdouble
));
661 memcpy(cursor
, glattrib
->Ptr
, size
);
662 if (alignment
!= size
)
663 memset(cursor
+ size
, 0, alignment
- size
);
665 struct brw_vertex_element
*input
= &brw
->vb
.inputs
[attr
];
666 input
->glattrib
= glattrib
;
668 input
->is_dual_slot
= (vp
->DualSlotInputs
& BITFIELD64_BIT(attr
)) != 0;
669 input
->offset
= cursor
- data
;
674 struct brw_vertex_buffer
*buffer
= &brw
->vb
.buffers
[j
];
675 const unsigned size
= cursor
- data
;
676 brw_upload_data(&brw
->upload
, data
, size
, size
,
677 &buffer
->bo
, &buffer
->offset
);
680 buffer
->step_rate
= 0;
684 brw
->vb
.nr_buffers
= j
;
688 brw_prepare_shader_draw_parameters(struct brw_context
*brw
)
690 const struct brw_vs_prog_data
*vs_prog_data
=
691 brw_vs_prog_data(brw
->vs
.base
.prog_data
);
693 /* For non-indirect draws, upload the shader draw parameters */
694 if ((vs_prog_data
->uses_firstvertex
|| vs_prog_data
->uses_baseinstance
) &&
695 brw
->draw
.draw_params_bo
== NULL
) {
696 brw_upload_data(&brw
->upload
,
697 &brw
->draw
.params
, sizeof(brw
->draw
.params
), 4,
698 &brw
->draw
.draw_params_bo
,
699 &brw
->draw
.draw_params_offset
);
702 if (vs_prog_data
->uses_drawid
|| vs_prog_data
->uses_is_indexed_draw
) {
703 brw_upload_data(&brw
->upload
,
704 &brw
->draw
.derived_params
, sizeof(brw
->draw
.derived_params
), 4,
705 &brw
->draw
.derived_draw_params_bo
,
706 &brw
->draw
.derived_draw_params_offset
);
711 brw_upload_indices(struct brw_context
*brw
)
713 const struct _mesa_index_buffer
*index_buffer
= brw
->ib
.ib
;
715 struct brw_bo
*old_bo
= brw
->ib
.bo
;
716 struct gl_buffer_object
*bufferobj
;
720 if (index_buffer
== NULL
)
723 ib_type_size
= 1 << index_buffer
->index_size_shift
;
724 ib_size
= index_buffer
->count
? ib_type_size
* index_buffer
->count
:
725 index_buffer
->obj
->Size
;
726 bufferobj
= index_buffer
->obj
;
728 /* Turn into a proper VBO:
730 if (!_mesa_is_bufferobj(bufferobj
)) {
731 /* Get new bufferobj, offset:
733 brw_upload_data(&brw
->upload
, index_buffer
->ptr
, ib_size
, ib_type_size
,
734 &brw
->ib
.bo
, &offset
);
735 brw
->ib
.size
= brw
->ib
.bo
->size
;
737 offset
= (GLuint
) (unsigned long) index_buffer
->ptr
;
740 intel_bufferobj_buffer(brw
, intel_buffer_object(bufferobj
),
741 offset
, ib_size
, false);
742 if (bo
!= brw
->ib
.bo
) {
743 brw_bo_unreference(brw
->ib
.bo
);
745 brw
->ib
.size
= bufferobj
->Size
;
746 brw_bo_reference(bo
);
750 /* Use 3DPRIMITIVE's start_vertex_offset to avoid re-uploading
751 * the index buffer state when we're just moving the start index
754 brw
->ib
.start_vertex_offset
= offset
/ ib_type_size
;
756 if (brw
->ib
.bo
!= old_bo
)
757 brw
->ctx
.NewDriverState
|= BRW_NEW_INDEX_BUFFER
;
759 unsigned index_size
= 1 << index_buffer
->index_size_shift
;
760 if (index_size
!= brw
->ib
.index_size
) {
761 brw
->ib
.index_size
= index_size
;
762 brw
->ctx
.NewDriverState
|= BRW_NEW_INDEX_BUFFER
;
765 /* We need to re-emit an index buffer state each time
766 * when cut index flag is changed
768 if (brw
->prim_restart
.enable_cut_index
!= brw
->ib
.enable_cut_index
) {
769 brw
->ib
.enable_cut_index
= brw
->prim_restart
.enable_cut_index
;
770 brw
->ctx
.NewDriverState
|= BRW_NEW_INDEX_BUFFER
;
774 const struct brw_tracked_state brw_indices
= {
777 .brw
= BRW_NEW_BLORP
|
780 .emit
= brw_upload_indices
,