summaryrefslogtreecommitdiffstats
path: root/gfx/angle/src/compiler/translator/ScalarizeVecAndMatConstructorArgs.cpp
blob: 5afa0d3086251212651c05020cf6e753f6a5dc0f (plain)
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
//
// Copyright (c) 2002-2014 The ANGLE Project Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
//
// Scalarize vector and matrix constructor args, so that vectors built from components don't have
// matrix arguments, and matrices built from components don't have vector arguments. This avoids
// driver bugs around vector and matrix constructors.
//

#include "common/debug.h"
#include "compiler/translator/ScalarizeVecAndMatConstructorArgs.h"

#include <algorithm>

#include "angle_gl.h"
#include "common/angleutils.h"
#include "compiler/translator/IntermNode.h"

namespace sh
{

namespace
{

bool ContainsMatrixNode(const TIntermSequence &sequence)
{
    for (size_t ii = 0; ii < sequence.size(); ++ii)
    {
        TIntermTyped *node = sequence[ii]->getAsTyped();
        if (node && node->isMatrix())
            return true;
    }
    return false;
}

bool ContainsVectorNode(const TIntermSequence &sequence)
{
    for (size_t ii = 0; ii < sequence.size(); ++ii)
    {
        TIntermTyped *node = sequence[ii]->getAsTyped();
        if (node && node->isVector())
            return true;
    }
    return false;
}

TIntermBinary *ConstructVectorIndexBinaryNode(TIntermSymbol *symbolNode, int index)
{
    return new TIntermBinary(EOpIndexDirect, symbolNode, TIntermTyped::CreateIndexNode(index));
}

TIntermBinary *ConstructMatrixIndexBinaryNode(
    TIntermSymbol *symbolNode, int colIndex, int rowIndex)
{
    TIntermBinary *colVectorNode =
        ConstructVectorIndexBinaryNode(symbolNode, colIndex);

    return new TIntermBinary(EOpIndexDirect, colVectorNode,
                             TIntermTyped::CreateIndexNode(rowIndex));
}

class ScalarizeArgsTraverser : public TIntermTraverser
{
  public:
    ScalarizeArgsTraverser(sh::GLenum shaderType,
                           bool fragmentPrecisionHigh,
                           unsigned int *temporaryIndex)
        : TIntermTraverser(true, false, false),
          mShaderType(shaderType),
          mFragmentPrecisionHigh(fragmentPrecisionHigh)
    {
        useTemporaryIndex(temporaryIndex);
    }

  protected:
    bool visitAggregate(Visit visit, TIntermAggregate *node) override;
    bool visitBlock(Visit visit, TIntermBlock *node) override;

  private:
    void scalarizeArgs(TIntermAggregate *aggregate, bool scalarizeVector, bool scalarizeMatrix);

    // If we have the following code:
    //   mat4 m(0);
    //   vec4 v(1, m);
    // We will rewrite to:
    //   mat4 m(0);
    //   mat4 s0 = m;
    //   vec4 v(1, s0[0][0], s0[0][1], s0[0][2]);
    // This function is to create nodes for "mat4 s0 = m;" and insert it to the code sequence. This
    // way the possible side effects of the constructor argument will only be evaluated once.
    void createTempVariable(TIntermTyped *original);

    std::vector<TIntermSequence> mBlockStack;

    sh::GLenum mShaderType;
    bool mFragmentPrecisionHigh;
};

bool ScalarizeArgsTraverser::visitAggregate(Visit visit, TIntermAggregate *node)
{
    if (visit == PreVisit)
    {
        switch (node->getOp())
        {
          case EOpConstructVec2:
          case EOpConstructVec3:
          case EOpConstructVec4:
          case EOpConstructBVec2:
          case EOpConstructBVec3:
          case EOpConstructBVec4:
          case EOpConstructIVec2:
          case EOpConstructIVec3:
          case EOpConstructIVec4:
            if (ContainsMatrixNode(*(node->getSequence())))
                scalarizeArgs(node, false, true);
            break;
          case EOpConstructMat2:
          case EOpConstructMat2x3:
          case EOpConstructMat2x4:
          case EOpConstructMat3x2:
          case EOpConstructMat3:
          case EOpConstructMat3x4:
          case EOpConstructMat4x2:
          case EOpConstructMat4x3:
          case EOpConstructMat4:
            if (ContainsVectorNode(*(node->getSequence())))
                scalarizeArgs(node, true, false);
            break;
          default:
            break;
        }
    }
    return true;
}

bool ScalarizeArgsTraverser::visitBlock(Visit visit, TIntermBlock *node)
{
    mBlockStack.push_back(TIntermSequence());
    {
        for (TIntermNode *child : *node->getSequence())
        {
            ASSERT(child != nullptr);
            child->traverse(this);
            mBlockStack.back().push_back(child);
        }
    }
    if (mBlockStack.back().size() > node->getSequence()->size())
    {
        node->getSequence()->clear();
        *(node->getSequence()) = mBlockStack.back();
    }
    mBlockStack.pop_back();
    return false;
}

void ScalarizeArgsTraverser::scalarizeArgs(TIntermAggregate *aggregate,
                                           bool scalarizeVector,
                                           bool scalarizeMatrix)
{
    ASSERT(aggregate);
    int size = 0;
    switch (aggregate->getOp())
    {
      case EOpConstructVec2:
      case EOpConstructBVec2:
      case EOpConstructIVec2:
        size = 2;
        break;
      case EOpConstructVec3:
      case EOpConstructBVec3:
      case EOpConstructIVec3:
        size = 3;
        break;
      case EOpConstructVec4:
      case EOpConstructBVec4:
      case EOpConstructIVec4:
      case EOpConstructMat2:
        size = 4;
        break;
      case EOpConstructMat2x3:
      case EOpConstructMat3x2:
        size = 6;
        break;
      case EOpConstructMat2x4:
      case EOpConstructMat4x2:
        size = 8;
        break;
      case EOpConstructMat3:
        size = 9;
        break;
      case EOpConstructMat3x4:
      case EOpConstructMat4x3:
        size = 12;
        break;
      case EOpConstructMat4:
        size = 16;
        break;
      default:
        break;
    }
    TIntermSequence *sequence = aggregate->getSequence();
    TIntermSequence original(*sequence);
    sequence->clear();
    for (size_t ii = 0; ii < original.size(); ++ii)
    {
        ASSERT(size > 0);
        TIntermTyped *node = original[ii]->getAsTyped();
        ASSERT(node);
        createTempVariable(node);
        if (node->isScalar())
        {
            sequence->push_back(createTempSymbol(node->getType()));
            size--;
        }
        else if (node->isVector())
        {
            if (scalarizeVector)
            {
                int repeat = std::min(size, node->getNominalSize());
                size -= repeat;
                for (int index = 0; index < repeat; ++index)
                {
                    TIntermSymbol *symbolNode = createTempSymbol(node->getType());
                    TIntermBinary *newNode = ConstructVectorIndexBinaryNode(
                        symbolNode, index);
                    sequence->push_back(newNode);
                }
            }
            else
            {
                TIntermSymbol *symbolNode = createTempSymbol(node->getType());
                sequence->push_back(symbolNode);
                size -= node->getNominalSize();
            }
        }
        else
        {
            ASSERT(node->isMatrix());
            if (scalarizeMatrix)
            {
                int colIndex = 0, rowIndex = 0;
                int repeat = std::min(size, node->getCols() * node->getRows());
                size -= repeat;
                while (repeat > 0)
                {
                    TIntermSymbol *symbolNode = createTempSymbol(node->getType());
                    TIntermBinary *newNode = ConstructMatrixIndexBinaryNode(
                        symbolNode, colIndex, rowIndex);
                    sequence->push_back(newNode);
                    rowIndex++;
                    if (rowIndex >= node->getRows())
                    {
                        rowIndex = 0;
                        colIndex++;
                    }
                    repeat--;
                }
            }
            else
            {
                TIntermSymbol *symbolNode = createTempSymbol(node->getType());
                sequence->push_back(symbolNode);
                size -= node->getCols() * node->getRows();
            }
        }
    }
}

void ScalarizeArgsTraverser::createTempVariable(TIntermTyped *original)
{
    ASSERT(original);
    nextTemporaryIndex();
    TIntermDeclaration *decl = createTempInitDeclaration(original);

    TType type = original->getType();
    if (mShaderType == GL_FRAGMENT_SHADER &&
        type.getBasicType() == EbtFloat &&
        type.getPrecision() == EbpUndefined)
    {
        // We use the highest available precision for the temporary variable
        // to avoid computing the actual precision using the rules defined
        // in GLSL ES 1.0 Section 4.5.2.
        TIntermBinary *init = decl->getSequence()->at(0)->getAsBinaryNode();
        init->getTypePointer()->setPrecision(mFragmentPrecisionHigh ? EbpHigh : EbpMedium);
        init->getLeft()->getTypePointer()->setPrecision(mFragmentPrecisionHigh ? EbpHigh
                                                                               : EbpMedium);
    }

    ASSERT(mBlockStack.size() > 0);
    TIntermSequence &sequence = mBlockStack.back();
    sequence.push_back(decl);
}

}  // namespace anonymous

void ScalarizeVecAndMatConstructorArgs(TIntermBlock *root,
                                       sh::GLenum shaderType,
                                       bool fragmentPrecisionHigh,
                                       unsigned int *temporaryIndex)
{
    ScalarizeArgsTraverser scalarizer(shaderType, fragmentPrecisionHigh, temporaryIndex);
    root->traverse(&scalarizer);
}

}  // namespace sh