aboutsummaryrefslogtreecommitdiffhomepage
path: root/src/gpu/GrProgramDesc.cpp
blob: b0396f9c378a02a9f613f4adbee5bd9984b8f972 (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
/*
 * Copyright 2016 Google Inc.
 *
 * Use of this source code is governed by a BSD-style license that can be
 * found in the LICENSE file.
 */

#include "GrProgramDesc.h"

#include "GrPipeline.h"
#include "GrPrimitiveProcessor.h"
#include "GrProcessor.h"
#include "GrRenderTargetPriv.h"
#include "GrShaderCaps.h"
#include "GrTexturePriv.h"
#include "SkChecksum.h"
#include "SkTo.h"
#include "glsl/GrGLSLFragmentProcessor.h"
#include "glsl/GrGLSLFragmentShaderBuilder.h"

enum {
    kSamplerOrImageTypeKeyBits = 4
};

static inline uint16_t texture_type_key(GrTextureType type) {
    int value = UINT16_MAX;
    switch (type) {
        case GrTextureType::k2D:
            value = 0;
            break;
        case GrTextureType::kExternal:
            value = 1;
            break;
        case GrTextureType::kRectangle:
            value = 2;
            break;
    }
    SkASSERT((value & ((1 << kSamplerOrImageTypeKeyBits) - 1)) == value);
    return SkToU16(value);
}

static uint16_t sampler_key(GrTextureType textureType, GrPixelConfig config,
                            const GrShaderCaps& caps) {
    int samplerTypeKey = texture_type_key(textureType);

    GR_STATIC_ASSERT(1 == sizeof(caps.configTextureSwizzle(config).asKey()));
    return SkToU16(samplerTypeKey |
                   caps.configTextureSwizzle(config).asKey() << kSamplerOrImageTypeKeyBits |
                   (GrSLSamplerPrecision(config) << (8 + kSamplerOrImageTypeKeyBits)));
}

static void add_sampler_and_image_keys(GrProcessorKeyBuilder* b, const GrResourceIOProcessor& proc,
                                       const GrShaderCaps& caps) {
    int numTextureSamplers = proc.numTextureSamplers();
    // Need two bytes per key.
    int word32Count = (numTextureSamplers + 1) / 2;
    if (0 == word32Count) {
        return;
    }
    uint16_t* k16 = reinterpret_cast<uint16_t*>(b->add32n(word32Count));
    int j = 0;
    for (int i = 0; i < numTextureSamplers; ++i, ++j) {
        const GrResourceIOProcessor::TextureSampler& sampler = proc.textureSampler(i);
        const GrTexture* tex = sampler.peekTexture();

        k16[j] = sampler_key(tex->texturePriv().textureType(), tex->config(), caps);
    }
    // zero the last 16 bits if the number of uniforms for samplers is odd.
    if (numTextureSamplers & 0x1) {
        k16[numTextureSamplers] = 0;
    }
}

/**
 * A function which emits a meta key into the key builder.  This is required because shader code may
 * be dependent on properties of the effect that the effect itself doesn't use
 * in its key (e.g. the pixel format of textures used). So we create a meta-key for
 * every effect using this function. It is also responsible for inserting the effect's class ID
 * which must be different for every GrProcessor subclass. It can fail if an effect uses too many
 * transforms, etc, for the space allotted in the meta-key.  NOTE, both FPs and GPs share this
 * function because it is hairy, though FPs do not have attribs, and GPs do not have transforms
 */
static bool gen_meta_key(const GrResourceIOProcessor& proc,
                         const GrShaderCaps& shaderCaps,
                         uint32_t transformKey,
                         GrProcessorKeyBuilder* b) {
    size_t processorKeySize = b->size();
    uint32_t classID = proc.classID();

    // Currently we allow 16 bits for the class id and the overall processor key size.
    static const uint32_t kMetaKeyInvalidMask = ~((uint32_t)UINT16_MAX);
    if ((processorKeySize | classID) & kMetaKeyInvalidMask) {
        return false;
    }

    add_sampler_and_image_keys(b, proc, shaderCaps);

    uint32_t* key = b->add32n(2);
    key[0] = (classID << 16) | SkToU32(processorKeySize);
    key[1] = transformKey;
    return true;
}

static bool gen_meta_key(const GrXferProcessor& xp,
                         const GrShaderCaps& shaderCaps,
                         GrProcessorKeyBuilder* b) {
    size_t processorKeySize = b->size();
    uint32_t classID = xp.classID();

    // Currently we allow 16 bits for the class id and the overall processor key size.
    static const uint32_t kMetaKeyInvalidMask = ~((uint32_t)UINT16_MAX);
    if ((processorKeySize | classID) & kMetaKeyInvalidMask) {
        return false;
    }

    b->add32((classID << 16) | SkToU32(processorKeySize));
    return true;
}

static bool gen_frag_proc_and_meta_keys(const GrPrimitiveProcessor& primProc,
                                        const GrFragmentProcessor& fp,
                                        const GrShaderCaps& shaderCaps,
                                        GrProcessorKeyBuilder* b) {
    for (int i = 0; i < fp.numChildProcessors(); ++i) {
        if (!gen_frag_proc_and_meta_keys(primProc, fp.childProcessor(i), shaderCaps, b)) {
            return false;
        }
    }

    fp.getGLSLProcessorKey(shaderCaps, b);

    return gen_meta_key(fp, shaderCaps, primProc.getTransformKey(fp.coordTransforms(),
                                                                 fp.numCoordTransforms()), b);
}

bool GrProgramDesc::Build(GrProgramDesc* desc,
                          const GrPrimitiveProcessor& primProc,
                          bool hasPointSize,
                          const GrPipeline& pipeline,
                          const GrShaderCaps& shaderCaps) {
    // The descriptor is used as a cache key. Thus when a field of the
    // descriptor will not affect program generation (because of the attribute
    // bindings in use or other descriptor field settings) it should be set
    // to a canonical value to avoid duplicate programs with different keys.

    GR_STATIC_ASSERT(0 == kProcessorKeysOffset % sizeof(uint32_t));
    // Make room for everything up to the effect keys.
    desc->key().reset();
    desc->key().push_back_n(kProcessorKeysOffset);

    GrProcessorKeyBuilder b(&desc->key());

    primProc.getGLSLProcessorKey(shaderCaps, &b);
    if (!gen_meta_key(primProc, shaderCaps, 0, &b)) {
        desc->key().reset();
        return false;
    }

    for (int i = 0; i < pipeline.numFragmentProcessors(); ++i) {
        const GrFragmentProcessor& fp = pipeline.getFragmentProcessor(i);
        if (!gen_frag_proc_and_meta_keys(primProc, fp, shaderCaps, &b)) {
            desc->key().reset();
            return false;
        }
    }

    const GrXferProcessor& xp = pipeline.getXferProcessor();
    const GrSurfaceOrigin* originIfDstTexture = nullptr;
    GrSurfaceOrigin origin;
    if (pipeline.dstTextureProxy()) {
        origin = pipeline.dstTextureProxy()->origin();
        originIfDstTexture = &origin;
    }
    xp.getGLSLProcessorKey(shaderCaps, &b, originIfDstTexture);
    if (!gen_meta_key(xp, shaderCaps, &b)) {
        desc->key().reset();
        return false;
    }

    // --------DO NOT MOVE HEADER ABOVE THIS LINE--------------------------------------------------
    // Because header is a pointer into the dynamic array, we can't push any new data into the key
    // below here.
    KeyHeader* header = desc->atOffset<KeyHeader, kHeaderOffset>();

    // make sure any padding in the header is zeroed.
    memset(header, 0, kHeaderSize);
    header->fOutputSwizzle = shaderCaps.configOutputSwizzle(pipeline.proxy()->config()).asKey();
    header->fSnapVerticesToPixelCenters = pipeline.snapVerticesToPixelCenters();
    header->fColorFragmentProcessorCnt = pipeline.numColorFragmentProcessors();
    header->fCoverageFragmentProcessorCnt = pipeline.numCoverageFragmentProcessors();
    // Fail if the client requested more processors than the key can fit.
    if (header->fColorFragmentProcessorCnt != pipeline.numColorFragmentProcessors() ||
        header->fCoverageFragmentProcessorCnt != pipeline.numCoverageFragmentProcessors()) {
        return false;
    }
    header->fHasPointSize = hasPointSize ? 1 : 0;
    return true;
}