1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
|
/* SPDX-License-Identifier: GPL-2.0-or-later
* Copyright 2016 by Mike Erwin. All rights reserved. */
/** \file
* \ingroup gpu
*
* GL implementation of GPUBatch.
* The only specificity of GL here is that it caches a list of
* Vertex Array Objects based on the bound shader interface.
*/
#include "BLI_assert.h"
#include "gpu_batch_private.hh"
#include "gpu_shader_private.hh"
#include "gl_context.hh"
#include "gl_debug.hh"
#include "gl_index_buffer.hh"
#include "gl_primitive.hh"
#include "gl_storage_buffer.hh"
#include "gl_vertex_array.hh"
#include "gl_batch.hh"
using namespace blender::gpu;
/* -------------------------------------------------------------------- */
/** \name VAO Cache
*
* Each #GLBatch has a small cache of VAO objects that are used to avoid VAO reconfiguration.
* TODO(fclem): Could be revisited to avoid so much cross references.
* \{ */
GLVaoCache::GLVaoCache()
{
init();
}
GLVaoCache::~GLVaoCache()
{
this->clear();
}
void GLVaoCache::init()
{
context_ = nullptr;
interface_ = nullptr;
is_dynamic_vao_count = false;
for (int i = 0; i < GPU_VAO_STATIC_LEN; i++) {
static_vaos.interfaces[i] = nullptr;
static_vaos.vao_ids[i] = 0;
}
vao_base_instance_ = 0;
base_instance_ = 0;
vao_id_ = 0;
}
void GLVaoCache::insert(const GLShaderInterface *interface, GLuint vao)
{
/* Now insert the cache. */
if (!is_dynamic_vao_count) {
int i; /* find first unused slot */
for (i = 0; i < GPU_VAO_STATIC_LEN; i++) {
if (static_vaos.vao_ids[i] == 0) {
break;
}
}
if (i < GPU_VAO_STATIC_LEN) {
static_vaos.interfaces[i] = interface;
static_vaos.vao_ids[i] = vao;
}
else {
/* Erase previous entries, they will be added back if drawn again. */
for (int i = 0; i < GPU_VAO_STATIC_LEN; i++) {
if (static_vaos.interfaces[i] != nullptr) {
const_cast<GLShaderInterface *>(static_vaos.interfaces[i])->ref_remove(this);
context_->vao_free(static_vaos.vao_ids[i]);
}
}
/* Not enough place switch to dynamic. */
is_dynamic_vao_count = true;
/* Init dynamic arrays and let the branch below set the values. */
dynamic_vaos.count = GPU_BATCH_VAO_DYN_ALLOC_COUNT;
dynamic_vaos.interfaces = (const GLShaderInterface **)MEM_callocN(
dynamic_vaos.count * sizeof(GLShaderInterface *), "dyn vaos interfaces");
dynamic_vaos.vao_ids = (GLuint *)MEM_callocN(dynamic_vaos.count * sizeof(GLuint),
"dyn vaos ids");
}
}
if (is_dynamic_vao_count) {
int i; /* find first unused slot */
for (i = 0; i < dynamic_vaos.count; i++) {
if (dynamic_vaos.vao_ids[i] == 0) {
break;
}
}
if (i == dynamic_vaos.count) {
/* Not enough place, realloc the array. */
i = dynamic_vaos.count;
dynamic_vaos.count += GPU_BATCH_VAO_DYN_ALLOC_COUNT;
dynamic_vaos.interfaces = (const GLShaderInterface **)MEM_recallocN(
(void *)dynamic_vaos.interfaces, sizeof(GLShaderInterface *) * dynamic_vaos.count);
dynamic_vaos.vao_ids = (GLuint *)MEM_recallocN(dynamic_vaos.vao_ids,
sizeof(GLuint) * dynamic_vaos.count);
}
dynamic_vaos.interfaces[i] = interface;
dynamic_vaos.vao_ids[i] = vao;
}
const_cast<GLShaderInterface *>(interface)->ref_add(this);
}
void GLVaoCache::remove(const GLShaderInterface *interface)
{
const int count = (is_dynamic_vao_count) ? dynamic_vaos.count : GPU_VAO_STATIC_LEN;
GLuint *vaos = (is_dynamic_vao_count) ? dynamic_vaos.vao_ids : static_vaos.vao_ids;
const GLShaderInterface **interfaces = (is_dynamic_vao_count) ? dynamic_vaos.interfaces :
static_vaos.interfaces;
for (int i = 0; i < count; i++) {
if (interfaces[i] == interface) {
context_->vao_free(vaos[i]);
vaos[i] = 0;
interfaces[i] = nullptr;
break; /* cannot have duplicates */
}
}
}
void GLVaoCache::clear()
{
GLContext *ctx = GLContext::get();
const int count = (is_dynamic_vao_count) ? dynamic_vaos.count : GPU_VAO_STATIC_LEN;
GLuint *vaos = (is_dynamic_vao_count) ? dynamic_vaos.vao_ids : static_vaos.vao_ids;
const GLShaderInterface **interfaces = (is_dynamic_vao_count) ? dynamic_vaos.interfaces :
static_vaos.interfaces;
/* Early out, nothing to free. */
if (context_ == nullptr) {
return;
}
if (context_ == ctx) {
glDeleteVertexArrays(count, vaos);
glDeleteVertexArrays(1, &vao_base_instance_);
}
else {
/* TODO(fclem): Slow way. Could avoid multiple mutex lock here */
for (int i = 0; i < count; i++) {
context_->vao_free(vaos[i]);
}
context_->vao_free(vao_base_instance_);
}
for (int i = 0; i < count; i++) {
if (interfaces[i] != nullptr) {
const_cast<GLShaderInterface *>(interfaces[i])->ref_remove(this);
}
}
if (is_dynamic_vao_count) {
MEM_freeN((void *)dynamic_vaos.interfaces);
MEM_freeN(dynamic_vaos.vao_ids);
}
if (context_) {
context_->vao_cache_unregister(this);
}
/* Reinit. */
this->init();
}
GLuint GLVaoCache::lookup(const GLShaderInterface *interface)
{
const int count = (is_dynamic_vao_count) ? dynamic_vaos.count : GPU_VAO_STATIC_LEN;
const GLShaderInterface **interfaces = (is_dynamic_vao_count) ? dynamic_vaos.interfaces :
static_vaos.interfaces;
for (int i = 0; i < count; i++) {
if (interfaces[i] == interface) {
return (is_dynamic_vao_count) ? dynamic_vaos.vao_ids[i] : static_vaos.vao_ids[i];
}
}
return 0;
}
void GLVaoCache::context_check()
{
GLContext *ctx = GLContext::get();
BLI_assert(ctx);
if (context_ != ctx) {
if (context_ != nullptr) {
/* IMPORTANT: Trying to draw a batch in multiple different context will trash the VAO cache.
* This has major performance impact and should be avoided in most cases. */
context_->vao_cache_unregister(this);
}
this->clear();
context_ = ctx;
context_->vao_cache_register(this);
}
}
GLuint GLVaoCache::base_instance_vao_get(GPUBatch *batch, int i_first)
{
this->context_check();
/* Make sure the interface is up to date. */
Shader *shader = GLContext::get()->shader;
GLShaderInterface *interface = static_cast<GLShaderInterface *>(shader->interface);
if (interface_ != interface) {
vao_get(batch);
/* Trigger update. */
base_instance_ = 0;
}
/**
* There seems to be a nasty bug when drawing using the same VAO reconfiguring (T71147).
* We just use a throwaway VAO for that. Note that this is likely to degrade performance.
*/
#ifdef __APPLE__
glDeleteVertexArrays(1, &vao_base_instance_);
vao_base_instance_ = 0;
base_instance_ = 0;
#endif
if (vao_base_instance_ == 0) {
glGenVertexArrays(1, &vao_base_instance_);
}
if (base_instance_ != i_first) {
base_instance_ = i_first;
GLVertArray::update_bindings(vao_base_instance_, batch, interface_, i_first);
}
return vao_base_instance_;
}
GLuint GLVaoCache::vao_get(GPUBatch *batch)
{
this->context_check();
Shader *shader = GLContext::get()->shader;
GLShaderInterface *interface = static_cast<GLShaderInterface *>(shader->interface);
if (interface_ != interface) {
interface_ = interface;
vao_id_ = this->lookup(interface_);
if (vao_id_ == 0) {
/* Cache miss, create a new VAO. */
glGenVertexArrays(1, &vao_id_);
this->insert(interface_, vao_id_);
GLVertArray::update_bindings(vao_id_, batch, interface_, 0);
}
}
return vao_id_;
}
/** \} */
/* -------------------------------------------------------------------- */
/** \name Drawing
* \{ */
void GLBatch::bind(int i_first)
{
GLContext::get()->state_manager->apply_state();
if (flag & GPU_BATCH_DIRTY) {
flag &= ~GPU_BATCH_DIRTY;
vao_cache_.clear();
}
#if GPU_TRACK_INDEX_RANGE
/* Can be removed if GL 4.3 is required. */
if (!GLContext::fixed_restart_index_support && (elem != nullptr)) {
glPrimitiveRestartIndex(this->elem_()->restart_index());
}
#endif
/* Can be removed if GL 4.2 is required. */
if (!GLContext::base_instance_support && (i_first > 0)) {
glBindVertexArray(vao_cache_.base_instance_vao_get(this, i_first));
}
else {
glBindVertexArray(vao_cache_.vao_get(this));
}
}
void GLBatch::draw(int v_first, int v_count, int i_first, int i_count)
{
GL_CHECK_RESOURCES("Batch");
this->bind(i_first);
BLI_assert(v_count > 0 && i_count > 0);
GLenum gl_type = to_gl(prim_type);
if (elem) {
const GLIndexBuf *el = this->elem_();
GLenum index_type = to_gl(el->index_type_);
GLint base_index = el->index_base_;
void *v_first_ofs = el->offset_ptr(v_first);
if (GLContext::base_instance_support) {
glDrawElementsInstancedBaseVertexBaseInstance(
gl_type, v_count, index_type, v_first_ofs, i_count, base_index, i_first);
}
else {
glDrawElementsInstancedBaseVertex(
gl_type, v_count, index_type, v_first_ofs, i_count, base_index);
}
}
else {
#ifdef __APPLE__
glDisable(GL_PRIMITIVE_RESTART);
#endif
if (GLContext::base_instance_support) {
glDrawArraysInstancedBaseInstance(gl_type, v_first, v_count, i_count, i_first);
}
else {
glDrawArraysInstanced(gl_type, v_first, v_count, i_count);
}
#ifdef __APPLE__
glEnable(GL_PRIMITIVE_RESTART);
#endif
}
}
void GLBatch::draw_indirect(GPUStorageBuf *indirect_buf)
{
GL_CHECK_RESOURCES("Batch");
this->bind(0);
dynamic_cast<GLStorageBuf *>(unwrap(indirect_buf))->bind_as(GL_DRAW_INDIRECT_BUFFER);
/* This barrier needs to be here as it only work on the currently bound indirect buffer. */
glMemoryBarrier(GL_COMMAND_BARRIER_BIT);
GLenum gl_type = to_gl(prim_type);
if (elem) {
const GLIndexBuf *el = this->elem_();
GLenum index_type = to_gl(el->index_type_);
glDrawElementsIndirect(gl_type, index_type, (GLvoid *)nullptr);
}
else {
glDrawArraysIndirect(gl_type, (GLvoid *)nullptr);
}
/* Unbind. */
glBindBuffer(GL_DRAW_INDIRECT_BUFFER, 0);
}
/** \} */
|