Subversion Repository Public Repository

Divide-Framework

This repository has no backups
This repository's network speed is throttled to 100KB/sec

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
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
#include "Headers/glVertexArray.h"

#include "Platform/Video/Headers/GFXDevice.h"
#include "Platform/Video/OpenGL/Headers/GLWrapper.h"

#include "Core/Headers/Console.h"
#include "Utility/Headers/Localization.h"

namespace Divide {

glVertexArray::VAOMap glVertexArray::_VAOMap;

IMPLEMENT_CUSTOM_ALLOCATOR(glVertexArray, 0, 0)

void glVertexArray::cleanup() {
    clearVaos();
}


GLuint glVertexArray::getVao(size_t hash) {
    VAOMap::const_iterator result = _VAOMap.find(hash);
    return result != std::end(_VAOMap) ? result->second  : 0;
}

void glVertexArray::setVao(size_t hash, GLuint id) {
    std::pair<VAOMap::const_iterator, bool> result =
        hashAlg::emplace(_VAOMap, hash, id);
    assert(result.second);
    ACKNOWLEDGE_UNUSED(result);
}

void glVertexArray::clearVaos() {
    for (VAOMap::value_type& value : _VAOMap) {
        if (value.second != 0) {
            GLUtil::_vaoPool.deallocate(value.second);
        }
    }
    _VAOMap.clear();
}

/// Default destructor
glVertexArray::glVertexArray(GFXDevice& context)
    : VertexBuffer(context),
      _refreshQueued(false),
      _formatInternal(GL_NONE)
{
    // We assume everything is static draw
    _usage = GL_STATIC_DRAW;
    _prevSize = -1;
    _prevSizeIndices = -1;
    _effectiveEntrySize = -1;
    _IBid = 0;
    _vaoCaches.fill(0);
    _vaoHashes.fill(0);

    _useAttribute.fill(false);
    _attributeOffset.fill(0);
}

glVertexArray::~glVertexArray()
{
    destroy();
}

void glVertexArray::reset() {
    _usage = GL_STATIC_DRAW;
    _prevSize = -1;
    _prevSizeIndices = -1;
    _vaoHashes.fill(0);

    _useAttribute.fill(false);
    _attributeOffset.fill(0);
    VertexBuffer::reset();
}

/// Delete buffer
void glVertexArray::destroy() {
    GLUtil::releaseVBO(_VBHandle._id, _VBHandle._offset);
}

/// Trim down the Vertex vector to only upload the minimal ammount of data to the GPU
std::pair<bufferPtr, size_t> glVertexArray::getMinimalData() {
    bool useColour     = _useAttribute[to_const_uint(VertexAttribute::ATTRIB_COLOR)];
    bool useNormals   = _useAttribute[to_const_uint(VertexAttribute::ATTRIB_NORMAL)];
    bool useTangents  = _useAttribute[to_const_uint(VertexAttribute::ATTRIB_TANGENT)];
    bool useTexcoords = _useAttribute[to_const_uint(VertexAttribute::ATTRIB_TEXCOORD)];
    bool useBoneData  = _useAttribute[to_const_uint(VertexAttribute::ATTRIB_BONE_INDICE)];

    size_t prevOffset = sizeof(vec3<F32>);
    if (useNormals) {
        _attributeOffset[to_const_uint(VertexAttribute::ATTRIB_NORMAL)] = to_uint(prevOffset);
        prevOffset += sizeof(F32);
    }

    if (useTangents) {
        _attributeOffset[to_const_uint(VertexAttribute::ATTRIB_TANGENT)] = to_uint(prevOffset);
        prevOffset += sizeof(F32);
    }

    if (useColour) {
        _attributeOffset[to_const_uint(VertexAttribute::ATTRIB_COLOR)] = to_uint(prevOffset);
        prevOffset += sizeof(vec4<U8>);
    }

    if (useTexcoords) {
        _attributeOffset[to_const_uint(VertexAttribute::ATTRIB_TEXCOORD)] = to_uint(prevOffset);
        prevOffset += sizeof(vec2<F32>);
    }

    if (useBoneData) {
        _attributeOffset[to_const_uint(VertexAttribute::ATTRIB_BONE_WEIGHT)] = to_uint(prevOffset);
        prevOffset += sizeof(U32);
        _attributeOffset[to_const_uint(VertexAttribute::ATTRIB_BONE_INDICE)] = to_uint(prevOffset);
        prevOffset += sizeof(U32);
    }

    _effectiveEntrySize = static_cast<GLsizei>(prevOffset);

    _smallData.reserve(_data.size() * _effectiveEntrySize);

    for (const Vertex& data : _data) {
        _smallData << data._position;

        if (useNormals) {
            _smallData << data._normal;
        }

        if (useTangents) {
            _smallData << data._tangent;
        }

        if (useColour) {
            _smallData << data._colour.r;
            _smallData << data._colour.g;
            _smallData << data._colour.b;
            _smallData << data._colour.a;
        }

        if (useTexcoords) {
            _smallData << data._texcoord.s;
            _smallData << data._texcoord.t;
        }

        if (useBoneData) {
            _smallData << data._weights.i;
            _smallData << data._indices.i;
        }
    }

    if (_staticBuffer && !keepData()) {
        _data.clear();
    } else {
        vectorAlg::shrinkToFit(_data);
    }

    return std::make_pair((bufferPtr)_smallData.contents(), _smallData.size());
}

/// Create a dynamic or static VB
bool glVertexArray::create(bool staticDraw) {
    // If we want a dynamic buffer, then we are doing something outdated, such
    // as software skinning, or software water rendering
    if (!staticDraw) {
        // OpenGLES support isn't added, but this check doesn't break anything,
        // so I'll just leave it here -Ionut
        GLenum usage = (_context.getAPI() == RenderAPI::OpenGLES)
                           ? GL_STREAM_DRAW
                           : GL_DYNAMIC_DRAW;
        if (usage != _usage) {
            _usage = usage;
            _refreshQueued = true;
        }
    }

    return VertexBuffer::create(staticDraw);
}

/// Update internal data based on construction method (dynamic/static)
bool glVertexArray::refresh() {
    // Dynamic LOD elements (such as terrain) need dynamic indices
    // We can manually override index usage (again, used by the Terrain
    // rendering system)
    assert(!_hardwareIndicesL.empty() || !_hardwareIndicesS.empty() &&
           "glVertexArray::refresh error: Invalid index data on Refresh()!");

    GLsizei nSizeIndices =
        (GLsizei)(usesLargeIndices() ? _hardwareIndicesL.size() * sizeof(U32)
                                     : _hardwareIndicesS.size() * sizeof(U16));

    bool indicesChanged = (nSizeIndices != _prevSizeIndices);
    _prevSizeIndices = nSizeIndices;
    _refreshQueued = indicesChanged;

    /// Can only add attributes for now. No removal support. -Ionut
    for (U8 i = 0; i < to_const_uint(VertexAttribute::COUNT); ++i) {
        _useAttribute[i] = _useAttribute[i] || _attribDirty[i];
        if (!_refreshQueued) {
           if (_attribDirty[i]) {
               _refreshQueued = true;
           }
        }
    }
    _attribDirty.fill(false);


    Console::printfn("VAO HASHES: ");
    std::array<bool, to_const_uint(RenderStage::COUNT)> vaoCachesDirty;
    vaoCachesDirty.fill(false);
    std::array<AttribFlags, to_const_uint(RenderStage::COUNT)> attributesPerStage;
    for (U8 i = 0; i < to_const_uint(RenderStage::COUNT); ++i) {
        const AttribFlags& stageMask = _attribMaskPerStage[i];

        AttribFlags& stageUsage = attributesPerStage[i];
        for (U8 j = 0; j < to_const_uint(VertexAttribute::COUNT); ++j) {
            stageUsage[j] = _useAttribute[j] && stageMask[j];
        }
        size_t crtHash = std::hash<AttribFlags>()(stageUsage);
        _vaoHashes[i] = crtHash;
        _vaoCaches[i] = getVao(crtHash);
        if (_vaoCaches[i] == 0) {
            // Generate a "Vertex Array Object"
            _vaoCaches[i] = GLUtil::_vaoPool.allocate();
            assert(_vaoCaches[i] != 0 && Locale::get(_ID("ERROR_VAO_INIT")));
            setVao(crtHash, _vaoCaches[i]);
            vaoCachesDirty[i] = true;
        }

        Console::printfn("      %d : %d", i, to_uint(crtHash));
    }


    std::pair<bufferPtr, size_t> bufferData = getMinimalData();
    // If any of the VBO's components changed size, we need to recreate the
    // entire buffer.

    GLsizei size = static_cast<GLsizei>(bufferData.second);
    bool sizeChanged = size != _prevSize;
    bool needsReallocation = false;
    U32 countRequirement = GLUtil::VBO::getChunkCountForSize(size);
    if (sizeChanged) {
        needsReallocation = countRequirement > GLUtil::VBO::getChunkCountForSize(_prevSize);
        _prevSize = size;
    }

    if (!_refreshQueued && !sizeChanged) {
        return false;
    }

    if (_VBHandle._id == 0 || needsReallocation) {
        if (_VBHandle._id != 0) {
            GLUtil::releaseVBO(_VBHandle._id, _VBHandle._offset);
        }
        // Generate a new Vertex Buffer Object
        GLUtil::commitVBO(countRequirement,_usage, _VBHandle._id, _VBHandle._offset);
        assert(_VBHandle._id != 0 && Locale::get(_ID("ERROR_VB_INIT")));
    }
    // Refresh buffer data (if this is the first call to refresh, this will be true)
    if (sizeChanged) {
        Console::d_printfn(Locale::get(_ID("DISPLAY_VB_METRICS")),
                           _VBHandle._id,
                           size,
                           countRequirement * GLUtil::VBO::MAX_VBO_CHUNK_SIZE_BYTES,
                           GLUtil::getVBOMemUsage(_VBHandle._id),
                           GLUtil::getVBOCount());
    }

    // Allocate sufficient space in our buffer
    glNamedBufferSubData(_VBHandle._id, _VBHandle._offset * GLUtil::VBO::MAX_VBO_CHUNK_SIZE_BYTES, size, bufferData.first);

    _smallData.clear();

    // Check if we need to update the IBO (will be true for the first Refresh() call)
    if (indicesChanged) {
        bufferPtr data = usesLargeIndices()
                             ? static_cast<bufferPtr>(_hardwareIndicesL.data())
                             : static_cast<bufferPtr>(_hardwareIndicesS.data());
        // Update our IB
        glNamedBufferData(_IBid, nSizeIndices, data, GL_STATIC_DRAW);
    }
    for (U8 i = 0; i < to_const_uint(RenderStage::COUNT); ++i) {
        if (vaoCachesDirty[i]) {
            // Set vertex attribute pointers
            uploadVBAttributes(i);
        }
    }
    vaoCachesDirty.fill(false);
    // Validate the buffer
    checkStatus();
    // Possibly clear client-side buffer for all non-required attributes?
    // foreach attribute if !required then delete else skip ?
    _refreshQueued = false;
    return true;
}

/// This method creates the initial VAO and VB OpenGL objects and queues a
/// Refresh call
bool glVertexArray::createInternal() {
    // Avoid double create calls
    assert(_VBHandle._id == 0 && "glVertexArray error: Attempted to double create a VB");
    // Position data is a minim requirement
    if (_data.empty()) {
        Console::errorfn(Locale::get(_ID("ERROR_VB_POSITION")));
        return false;
    }
    VertexBuffer::createInternal();

    _formatInternal = GLUtil::glDataFormat[to_uint(_format)];
    // Generate an "Index Buffer Object"
    glCreateBuffers(1, &_IBid);
    // Validate buffer creation
    // Assert if the IB creation failed
    assert(_IBid != 0 && Locale::get(_ID("ERROR_IB_INIT")));
    // Calling refresh updates all stored information and sends it to the GPU
    return queueRefresh();
}

/// Render the current buffer data using the specified command
void glVertexArray::draw(const GenericDrawCommand& command, bool useCmdBuffer) {
    // Make sure the buffer is current
    // Make sure we have valid data (buffer creation is deferred to the first activate call)
    if (_IBid == 0) {
        if (!createInternal()) {
            return;
        }
    }

    // Check if we have a refresh request queued up
    if (_refreshQueued) {
        if (!refresh()) {
            return;
        }
    }

    // Bind the vertex array object that in turn activates all of the bindings
    // and pointers set on creation
    GLuint vao = _vaoCaches[to_uint(_context.getRenderStage())];
    if (GL_API::setActiveVAO(vao)) {
        // If this is the first time the VAO is bound in the current loop, check
        // for primitive restart requests
        GL_API::togglePrimitiveRestart(_primitiveRestartEnabled);
    }
    // Bind the the vertex buffer and index buffer
    GL_API::bindActiveBuffer(vao,
                             0,
                             _VBHandle._id,
                             _VBHandle._offset * GLUtil::VBO::MAX_VBO_CHUNK_SIZE_BYTES,
                             _effectiveEntrySize);

    GLUtil::submitRenderCommand(command, useCmdBuffer, _formatInternal, _IBid);
}

/// Activate and set all of the required vertex attributes.
void glVertexArray::uploadVBAttributes(U8 vaoIndex) {
    // Bind the current VAO to save our attributes
    GL_API::setActiveVAO(_vaoCaches[vaoIndex]);
    static const U32 positionLoc = to_const_uint(AttribLocation::VERTEX_POSITION);
    static const U32 colourLoc = to_const_uint(AttribLocation::VERTEX_COLOR);
    static const U32 normalLoc = to_const_uint(AttribLocation::VERTEX_NORMAL);
    static const U32 texCoordLoc = to_const_uint(AttribLocation::VERTEX_TEXCOORD);
    static const U32 tangentLoc = to_const_uint(AttribLocation::VERTEX_TANGENT);
    static const U32 boneWeightLoc = to_const_uint(AttribLocation::VERTEX_BONE_WEIGHT);
    static const U32 boneIndiceLoc = to_const_uint(AttribLocation::VERTEX_BONE_INDICE);
    
    glEnableVertexAttribArray(positionLoc);
    glVertexAttribFormat(positionLoc, 3, GL_FLOAT, GL_FALSE, _attributeOffset[positionLoc]);
    glVertexAttribBinding(positionLoc, 0);

    if (_useAttribute[colourLoc]) {
        glEnableVertexAttribArray(colourLoc);
        glVertexAttribFormat(colourLoc, 4, GL_UNSIGNED_BYTE, GL_TRUE, _attributeOffset[colourLoc]);
        glVertexAttribBinding(colourLoc, 0);
    } else {
        glDisableVertexAttribArray(colourLoc);
    }

    if (_useAttribute[normalLoc]) {
        glEnableVertexAttribArray(normalLoc);
        glVertexAttribFormat(normalLoc, 1, GL_FLOAT, GL_FALSE, _attributeOffset[normalLoc]);
        glVertexAttribBinding(normalLoc, 0);
    } else {
        glDisableVertexAttribArray(normalLoc);
    }

    if (_useAttribute[texCoordLoc]) {
        glEnableVertexAttribArray(texCoordLoc);
        glVertexAttribFormat(texCoordLoc, 2, GL_FLOAT, GL_FALSE, _attributeOffset[texCoordLoc]);
        glVertexAttribBinding(texCoordLoc, 0);
    } else {
        glDisableVertexAttribArray(texCoordLoc);
    }

    if (_useAttribute[tangentLoc]) {
        glEnableVertexAttribArray(tangentLoc);
        glVertexAttribFormat(tangentLoc, 1, GL_FLOAT, GL_FALSE, _attributeOffset[tangentLoc]);
        glVertexAttribBinding(tangentLoc, 0);
    } else {
        glDisableVertexAttribArray(tangentLoc);
    }

    if (_useAttribute[boneWeightLoc]) {
        assert(_useAttribute[boneIndiceLoc]);

        glEnableVertexAttribArray(boneWeightLoc);
        glEnableVertexAttribArray(boneIndiceLoc);
        glVertexAttribFormat(boneWeightLoc, 4, GL_UNSIGNED_BYTE, GL_TRUE, _attributeOffset[boneWeightLoc]);
        glVertexAttribIFormat(boneIndiceLoc, 4, GL_UNSIGNED_BYTE, _attributeOffset[boneIndiceLoc]);
        glVertexAttribBinding(boneWeightLoc, 0);
        glVertexAttribBinding(boneIndiceLoc, 0);
    } else {
        glDisableVertexAttribArray(boneWeightLoc);
        glDisableVertexAttribArray(boneIndiceLoc);
    }
}

/// Various post-create checks
void glVertexArray::checkStatus() {
}
};

Commits for Divide-Framework/trunk/Source Code/Platform/Video/OpenGL/Buffers/VertexBuffer/glVertexArray.cpp

Diff revisions: vs.
Revision Author Commited Message
750 Diff Diff IonutCava picture IonutCava Thu 07 Jul, 2016 16:02:03 +0000

[IonutCava]
- Static analysis based fixes and improvements

743 Diff Diff IonutCava picture IonutCava Tue 28 Jun, 2016 15:55:50 +0000

[IonutCava]
- More performance tuning:
— Move billboarding related camera variable calculation to shaders instead of CPU
— Replace some DIVIDE_ASSERT calls with regular assert calls in really low level code.
— Some cleanup of the GenericDrawCommand class
— glRegularBuffer does an invalidation call before updating
— Misc changes

737 Diff Diff IonutCava picture IonutCava Thu 23 Jun, 2016 16:20:05 +0000

[IonutCava]
- Improved GenericVertexData structure (separate buffer data from attribute data entirely)
- Added a VAO pool for OpenGL so that all requested VAOs are allocated on the main thread
- Added a mainThreadTask system to queue a task to run on the main thread and optionally wait for it to complete
- Added the option to specify generic geometry for the particle system with an optional index buffer

736 Diff Diff IonutCava picture IonutCava Wed 22 Jun, 2016 20:55:11 +0000

[IonutCava]
- Trying to fix rendering issues with the Particle system in regards to the usage of the Generic Vertex Data object and its buffers

727 Diff Diff IonutCava picture IonutCava Tue 14 Jun, 2016 16:01:38 +0000

[IonutCava]
- Remove node and cmd buffers from GFXDevice and add them as a new BufferData struct to RenderPass class.
— Each render pass holds its own buffers
- Improvements / fixes to the CSM code
- Added a global toggle to enable/disable custom memory allocators

721 Diff Diff IonutCava picture IonutCava Wed 08 Jun, 2016 15:47:49 +0000

[IonutCava]
- Spelling: change all references from our code of “color” to the PROPER British version of “colour” because it sounds 100x better

709 Diff Diff IonutCava picture IonutCava Thu 19 May, 2016 16:21:46 +0000

[IonutCava]
- Massive rewrite of the resource system:
— ResourceCache now uses a map of weak pointers and passes shared pointers to the call point of CreateResource
— RemoveResource is no longer needed, but resource based smart pointers use a custom deleter that calls unload()
— Keep a shared_ptr of the resource in the entity that needs it and pass shared_ptr references from getter methods
-— All of this removed the need for calls to RemoveResource, REGISTER_TRACKED_DEPENDENCY and UNREGISTER_TRACKED_DEPENDENCY reducing the number of mutex locks and atomic exchanges
- Singleton cleanup:
— Removed ShaderManager singleton and merged it’s data and responsibilities in static methods in both ShaderProgram and Shader classes.
Explanation: Due to the complex interdependency chains in the system, copying, updating, tracking the cache in a thread safe way became a lot more slower than a straight forward smart pointer based implementation. (e.g. scene graph nodes have 3d objects and material instances, 3d objects have materials, materials have textures and shaders, etc)

703 Diff Diff IonutCava picture IonutCava Wed 11 May, 2016 15:44:57 +0000

[IonutCava]
- Finish XML-based control bindings:
— Allow per bind input params (key id, mouse button index, joystick control data, etc)
— Allow mouse binding (with up to 7 buttons)
— Allow up to 4 joysticks (for now) with 120+ button, POV, Axis, Slider and Vector binding support for each
- Change all hashes in code from U32 to size_t for compatibility with STL/Boost

671 Diff Diff IonutCava picture IonutCava Fri 08 Apr, 2016 16:26:49 +0000

[Ionut]
- Custom memory allocation system part 1:
— Add http://www.codeproject.com/Articles/1089905/A-Custom-STL-std-allocator-Replacement-Improves-Pe
— Add custom allocators to strings and vectors
— Add custom allocators to a few test classes
— No support for aligned memory yet, so vec4<F32> and mat4<F32> vectors use the old allocation system
- Replace more std::async call with our custom task system

654 IonutCava picture IonutCava Tue 01 Mar, 2016 17:22:09 +0000

[Ionut]
- Moved window related code from rendering API code into proper classes:
— DisplayWindow handles SDL based window management
— WindowManager handles multiple DisplayWindow instances
— Code still needs proper extensive testing
- Small performance optimizations