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
|
/*
* Copyright 2011 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include "GrGpuGL.h"
#include "builders/GrGLProgramBuilder.h"
#include "GrProcessor.h"
#include "GrGLProcessor.h"
#include "GrGLPathRendering.h"
#include "GrOptDrawState.h"
#include "SkRTConf.h"
#include "SkTSearch.h"
#ifdef PROGRAM_CACHE_STATS
SK_CONF_DECLARE(bool, c_DisplayCache, "gpu.displayCache", false,
"Display program cache usage.");
#endif
typedef GrGLProgramDataManager::UniformHandle UniformHandle;
struct GrGpuGL::ProgramCache::Entry {
SK_DECLARE_INST_COUNT_ROOT(Entry);
Entry() : fProgram(NULL), fLRUStamp(0) {}
SkAutoTUnref<GrGLProgram> fProgram;
unsigned int fLRUStamp;
};
struct GrGpuGL::ProgramCache::ProgDescLess {
bool operator() (const GrProgramDesc& desc, const Entry* entry) {
SkASSERT(entry->fProgram.get());
return GrProgramDesc::Less(desc, entry->fProgram->getDesc());
}
bool operator() (const Entry* entry, const GrProgramDesc& desc) {
SkASSERT(entry->fProgram.get());
return GrProgramDesc::Less(entry->fProgram->getDesc(), desc);
}
};
GrGpuGL::ProgramCache::ProgramCache(GrGpuGL* gpu)
: fCount(0)
, fCurrLRUStamp(0)
, fGpu(gpu)
#ifdef PROGRAM_CACHE_STATS
, fTotalRequests(0)
, fCacheMisses(0)
, fHashMisses(0)
#endif
{
for (int i = 0; i < 1 << kHashBits; ++i) {
fHashTable[i] = NULL;
}
}
GrGpuGL::ProgramCache::~ProgramCache() {
for (int i = 0; i < fCount; ++i){
SkDELETE(fEntries[i]);
}
// dump stats
#ifdef PROGRAM_CACHE_STATS
if (c_DisplayCache) {
SkDebugf("--- Program Cache ---\n");
SkDebugf("Total requests: %d\n", fTotalRequests);
SkDebugf("Cache misses: %d\n", fCacheMisses);
SkDebugf("Cache miss %%: %f\n", (fTotalRequests > 0) ?
100.f * fCacheMisses / fTotalRequests :
0.f);
int cacheHits = fTotalRequests - fCacheMisses;
SkDebugf("Hash miss %%: %f\n", (cacheHits > 0) ? 100.f * fHashMisses / cacheHits : 0.f);
SkDebugf("---------------------\n");
}
#endif
}
void GrGpuGL::ProgramCache::abandon() {
for (int i = 0; i < fCount; ++i) {
SkASSERT(fEntries[i]->fProgram.get());
fEntries[i]->fProgram->abandon();
SkDELETE(fEntries[i]);
}
fCount = 0;
}
int GrGpuGL::ProgramCache::search(const GrProgramDesc& desc) const {
ProgDescLess less;
return SkTSearch(fEntries, fCount, desc, sizeof(Entry*), less);
}
GrGLProgram* GrGpuGL::ProgramCache::getProgram(const GrOptDrawState& optState, DrawType type) {
#ifdef PROGRAM_CACHE_STATS
++fTotalRequests;
#endif
Entry* entry = NULL;
uint32_t hashIdx = optState.programDesc().getChecksum();
hashIdx ^= hashIdx >> 16;
if (kHashBits <= 8) {
hashIdx ^= hashIdx >> 8;
}
hashIdx &=((1 << kHashBits) - 1);
Entry* hashedEntry = fHashTable[hashIdx];
if (hashedEntry && hashedEntry->fProgram->getDesc() == optState.programDesc()) {
SkASSERT(hashedEntry->fProgram);
entry = hashedEntry;
}
int entryIdx;
if (NULL == entry) {
entryIdx = this->search(optState.programDesc());
if (entryIdx >= 0) {
entry = fEntries[entryIdx];
#ifdef PROGRAM_CACHE_STATS
++fHashMisses;
#endif
}
}
if (NULL == entry) {
// We have a cache miss
#ifdef PROGRAM_CACHE_STATS
++fCacheMisses;
#endif
GrGLProgram* program = GrGLProgramBuilder::CreateProgram(optState, type, fGpu);
if (NULL == program) {
return NULL;
}
int purgeIdx = 0;
if (fCount < kMaxEntries) {
entry = SkNEW(Entry);
purgeIdx = fCount++;
fEntries[purgeIdx] = entry;
} else {
SkASSERT(fCount == kMaxEntries);
purgeIdx = 0;
for (int i = 1; i < kMaxEntries; ++i) {
if (fEntries[i]->fLRUStamp < fEntries[purgeIdx]->fLRUStamp) {
purgeIdx = i;
}
}
entry = fEntries[purgeIdx];
int purgedHashIdx = entry->fProgram->getDesc().getChecksum() & ((1 << kHashBits) - 1);
if (fHashTable[purgedHashIdx] == entry) {
fHashTable[purgedHashIdx] = NULL;
}
}
SkASSERT(fEntries[purgeIdx] == entry);
entry->fProgram.reset(program);
// We need to shift fEntries around so that the entry currently at purgeIdx is placed
// just before the entry at ~entryIdx (in order to keep fEntries sorted by descriptor).
entryIdx = ~entryIdx;
if (entryIdx < purgeIdx) {
// Let E and P be the entries at index entryIdx and purgeIdx, respectively.
// If the entries array looks like this:
// aaaaEbbbbbPccccc
// we rearrange it to look like this:
// aaaaPEbbbbbccccc
size_t copySize = (purgeIdx - entryIdx) * sizeof(Entry*);
memmove(fEntries + entryIdx + 1, fEntries + entryIdx, copySize);
fEntries[entryIdx] = entry;
} else if (purgeIdx < entryIdx) {
// If the entries array looks like this:
// aaaaPbbbbbEccccc
// we rearrange it to look like this:
// aaaabbbbbPEccccc
size_t copySize = (entryIdx - purgeIdx - 1) * sizeof(Entry*);
memmove(fEntries + purgeIdx, fEntries + purgeIdx + 1, copySize);
fEntries[entryIdx - 1] = entry;
}
#ifdef SK_DEBUG
SkASSERT(fEntries[0]->fProgram.get());
for (int i = 0; i < fCount - 1; ++i) {
SkASSERT(fEntries[i + 1]->fProgram.get());
const GrProgramDesc& a = fEntries[i]->fProgram->getDesc();
const GrProgramDesc& b = fEntries[i + 1]->fProgram->getDesc();
SkASSERT(GrProgramDesc::Less(a, b));
SkASSERT(!GrProgramDesc::Less(b, a));
}
#endif
}
fHashTable[hashIdx] = entry;
entry->fLRUStamp = fCurrLRUStamp;
if (SK_MaxU32 == fCurrLRUStamp) {
// wrap around! just trash our LRU, one time hit.
for (int i = 0; i < fCount; ++i) {
fEntries[i]->fLRUStamp = 0;
}
}
++fCurrLRUStamp;
return entry->fProgram;
}
////////////////////////////////////////////////////////////////////////////////
#define GL_CALL(X) GR_GL_CALL(this->glInterface(), X)
bool GrGpuGL::flushGraphicsState(const GrOptDrawState& optState,
DrawType type,
const GrDeviceCoordTexture* dstCopy) {
// GrGpu::setupClipAndFlushState should have already checked this and bailed if not true.
SkASSERT(optState.getRenderTarget());
if (kStencilPath_DrawType == type) {
const GrRenderTarget* rt = optState.getRenderTarget();
SkISize size;
size.set(rt->width(), rt->height());
this->glPathRendering()->setProjectionMatrix(optState.getViewMatrix(), size, rt->origin());
} else {
this->flushMiscFixedFunctionState(optState);
GrBlendCoeff srcCoeff = optState.getSrcBlendCoeff();
GrBlendCoeff dstCoeff = optState.getDstBlendCoeff();
// In these blend coeff's we end up drawing nothing so we can skip draw all together
if (kZero_GrBlendCoeff == srcCoeff && kOne_GrBlendCoeff == dstCoeff &&
!optState.getStencil().doesWrite()) {
return false;
}
fCurrentProgram.reset(fProgramCache->getProgram(optState, type));
if (NULL == fCurrentProgram.get()) {
SkDEBUGFAIL("Failed to create program!");
return false;
}
fCurrentProgram.get()->ref();
GrGLuint programID = fCurrentProgram->programID();
if (fHWProgramID != programID) {
GL_CALL(UseProgram(programID));
fHWProgramID = programID;
}
this->flushBlend(optState, kDrawLines_DrawType == type, srcCoeff, dstCoeff);
fCurrentProgram->setData(optState, type, dstCopy);
}
GrGLRenderTarget* glRT = static_cast<GrGLRenderTarget*>(optState.getRenderTarget());
this->flushStencil(optState.getStencil(), type);
this->flushScissor(optState.getScissorState(), glRT->getViewport(), glRT->origin());
this->flushAAState(optState, type);
// This must come after textures are flushed because a texture may need
// to be msaa-resolved (which will modify bound FBO state).
this->flushRenderTarget(glRT, NULL);
return true;
}
void GrGpuGL::setupGeometry(const GrOptDrawState& optState,
const GrDrawTarget::DrawInfo& info,
size_t* indexOffsetInBytes) {
GrGLsizei stride = static_cast<GrGLsizei>(optState.getVertexStride());
size_t vertexOffsetInBytes = stride * info.startVertex();
GrGLVertexBuffer* vbuf;
vbuf = (GrGLVertexBuffer*) info.vertexBuffer();
SkASSERT(vbuf);
SkASSERT(!vbuf->isMapped());
vertexOffsetInBytes += vbuf->baseOffset();
GrGLIndexBuffer* ibuf = NULL;
if (info.isIndexed()) {
SkASSERT(indexOffsetInBytes);
*indexOffsetInBytes = 0;
ibuf = (GrGLIndexBuffer*)info.indexBuffer();
SkASSERT(ibuf);
SkASSERT(!ibuf->isMapped());
*indexOffsetInBytes += ibuf->baseOffset();
}
GrGLAttribArrayState* attribState =
fHWGeometryState.bindArrayAndBuffersToDraw(this, vbuf, ibuf);
if (fCurrentProgram->hasVertexShader()) {
int vertexAttribCount = optState.getVertexAttribCount();
uint32_t usedAttribArraysMask = 0;
const GrVertexAttrib* vertexAttrib = optState.getVertexAttribs();
for (int vertexAttribIndex = 0; vertexAttribIndex < vertexAttribCount;
++vertexAttribIndex, ++vertexAttrib) {
usedAttribArraysMask |= (1 << vertexAttribIndex);
GrVertexAttribType attribType = vertexAttrib->fType;
attribState->set(this,
vertexAttribIndex,
vbuf,
GrGLAttribTypeToLayout(attribType).fCount,
GrGLAttribTypeToLayout(attribType).fType,
GrGLAttribTypeToLayout(attribType).fNormalized,
stride,
reinterpret_cast<GrGLvoid*>(
vertexOffsetInBytes + vertexAttrib->fOffset));
}
attribState->disableUnusedArrays(this, usedAttribArraysMask);
}
}
void GrGpuGL::buildProgramDesc(const GrOptDrawState& optState,
const GrProgramDesc::DescInfo& descInfo,
GrGpu::DrawType drawType,
const GrDeviceCoordTexture* dstCopy,
GrProgramDesc* desc) {
if (!GrGLProgramDescBuilder::Build(optState, descInfo, drawType, this, dstCopy, desc)) {
SkDEBUGFAIL("Failed to generate GL program descriptor");
}
}
|