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
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
#include "config.h"

#include "Headers/GFXDevice.h"

#include "Core/Headers/ParamHandler.h"
#include "Core/Time/Headers/ProfileTimer.h"
#include "Managers/Headers/SceneManager.h"

#include "Rendering/Headers/Renderer.h"
#include "Rendering/PostFX/Headers/PostFX.h"
#include "Rendering/Camera/Headers/FreeFlyCamera.h"
#include "Rendering/RenderPass/Headers/RenderPass.h"

#include "Platform/Video/Headers/IMPrimitive.h"
#include "Platform/Video/Textures/Headers/Texture.h"
#include "Platform/Video/Shaders/Headers/ShaderProgram.h"
#include "Platform/Video/Buffers/ShaderBuffer/Headers/ShaderBuffer.h"

namespace Divide {

std::array<VertexBuffer::AttribFlags, to_const_uint(RenderStage::COUNT)> VertexBuffer::_attribMaskPerStage;

GFXDevice::GFXDevice()
    : _api(nullptr), 
    _renderStage(RenderStage::DISPLAY),
    _prevRenderStage(RenderStage::COUNT),
    _commandBuildTimer(Time::ADD_TIMER("Command Generation Timer"))
{
    // Hash values
    _state2DRenderingHash = 0;
    _defaultStateBlockHash = 0;
    _currentStateBlockHash = 0;
    _previousStateBlockHash = 0;
    _defaultStateNoDepthHash = 0;
    _stateDepthOnlyRenderingHash = 0;
    // Pointers
    _axisGizmo = nullptr;
    _imShader = nullptr;
    _imShaderLines = nullptr;
    _gfxDataBuffer = nullptr;
    _HIZConstructProgram = nullptr;
    _HIZCullProgram = nullptr;
    _framebufferDraw = nullptr;
    _previewDepthMapShader = nullptr;
    _displayShader = nullptr;
    // Integers
    FRAME_COUNT = 0;
    FRAME_DRAW_CALLS = 0;
    FRAME_DRAW_CALLS_PREV = FRAME_DRAW_CALLS;
    _imShaderTextureFlag = -1;
    _imShaderWorldMatrix = -1;
    // Floats
    _interpolationFactor = 1.0;
    // Cameras
    _2DCamera = nullptr;
    _cubeCamera = nullptr;
    _dualParaboloidCamera = nullptr;
    // Booleans
    _2DRendering = false;
    _drawDebugAxis = false;
    _viewportUpdate = false;
    _rasterizationEnabled = true;
    _zWriteEnabled = true;
    // Enumerated Types
    _shadowDetailLevel = RenderDetailLevel::HIGH;
    _GPUVendor = GPUVendor::COUNT;
    _API_ID = RenderAPI::COUNT;
    // Clipping planes
    _clippingPlanes.resize(Config::MAX_CLIP_PLANES, Plane<F32>(0, 0, 0, 0));
    // Render targets
    for (RenderTarget& renderTarget : _renderTarget) {
        renderTarget._buffer = nullptr;
    }
    for (RenderTarget& renderTarget : _reflectionTarget) {
        renderTarget._buffer = nullptr;
    }
    for (RenderTarget& renderTarget : _refractionTarget) {
        renderTarget._buffer = nullptr;
    }
    // To allow calls to "setBaseViewport"
    
    _viewport.push(vec4<I32>(-1));

    _lastCommandCount.fill(0);
    _lastNodeCount.fill(0);
    // Red X-axis
    _axisLines.push_back(
        Line(VECTOR3_ZERO, WORLD_X_AXIS * 2, vec4<U8>(255, 0, 0, 255), 3.0f));
    // Green Y-axis
    _axisLines.push_back(
        Line(VECTOR3_ZERO, WORLD_Y_AXIS * 2, vec4<U8>(0, 255, 0, 255), 3.0f));
    // Blue Z-axis
    _axisLines.push_back(
        Line(VECTOR3_ZERO, WORLD_Z_AXIS * 2, vec4<U8>(0, 0, 255, 255), 3.0f));

    VertexBuffer::AttribFlags flags;
    flags.fill(true);
    VertexBuffer::setAttribMasks(flags);

    // Don't (currently) need these for shadow passes
    flags[to_const_uint(VertexBuffer::VertexAttribute::ATTRIB_COLOR)] = false;
    flags[to_const_uint(VertexBuffer::VertexAttribute::ATTRIB_TANGENT)] = false;
    VertexBuffer::setAttribMask(RenderStage::Z_PRE_PASS, flags);
    flags[to_const_uint(VertexBuffer::VertexAttribute::ATTRIB_NORMAL)] = false;
    VertexBuffer::setAttribMask(RenderStage::SHADOW, flags);
}

GFXDevice::~GFXDevice()
{
}

/// Generate a cube texture and store it in the provided framebuffer
void GFXDevice::generateCubeMap(Framebuffer& cubeMap,
                                const U32 arrayOffset,
                                const vec3<F32>& pos,
                                const vec2<F32>& zPlanes,
                                RenderStage renderStage) {
    // Only the first color attachment or the depth attachment is used for now
    // and it must be a cube map texture
    const Texture_ptr& colorAttachment = cubeMap.getAttachment(TextureDescriptor::AttachmentType::Color0, false);
    const Texture_ptr& depthAttachment = cubeMap.getAttachment(TextureDescriptor::AttachmentType::Depth, false);
    // Color attachment takes precedent over depth attachment
    bool hasColor = (colorAttachment != nullptr);
    bool hasDepth = (depthAttachment != nullptr);
    // Everyone's innocent until proven guilty
    bool isValidFB = true;
    if (hasColor) {
        // We only need the color attachment
        isValidFB = (colorAttachment->getTextureType() == TextureType::TEXTURE_CUBE_MAP) ||
                    (colorAttachment->getTextureType() == TextureType::TEXTURE_CUBE_ARRAY);
    } else {
        // We don't have a color attachment, so we require a cube map depth
        // attachment
        isValidFB = hasDepth && (depthAttachment->getTextureType() == TextureType::TEXTURE_CUBE_MAP ||
                                 depthAttachment->getTextureType() == TextureType::TEXTURE_CUBE_ARRAY);
    }
    // Make sure we have a proper render target to draw to
    if (!isValidFB) {
        // Future formats must be added later (e.g. cube map arrays)
        Console::errorfn(Locale::get(_ID("ERROR_GFX_DEVICE_INVALID_FB_CUBEMAP")));
        return;
    }
    // No dual-paraboloid rendering here. Just draw once for each face.
    static vec3<F32> TabUp[6] = {WORLD_Y_NEG_AXIS, WORLD_Y_NEG_AXIS,
                                 WORLD_Z_AXIS,     WORLD_Z_NEG_AXIS,
                                 WORLD_Y_NEG_AXIS, WORLD_Y_NEG_AXIS};
    // Get the center and up vectors for each cube face
    vec3<F32> TabCenter[6] = {vec3<F32>(pos.x + 1.0f, pos.y, pos.z),
                              vec3<F32>(pos.x - 1.0f, pos.y, pos.z),
                              vec3<F32>(pos.x, pos.y + 1.0f, pos.z),
                              vec3<F32>(pos.x, pos.y - 1.0f, pos.z),
                              vec3<F32>(pos.x, pos.y, pos.z + 1.0f),
                              vec3<F32>(pos.x, pos.y, pos.z - 1.0f)};

    Kernel& kernel = Application::instance().kernel();
    // Set a 90 degree vertical FoV perspective projection
    _cubeCamera->setProjection(1.0f, 90.0f, zPlanes);
    // Set the cube camera as the currently active one
    kernel.getCameraMgr().pushActiveCamera(_cubeCamera);
    // Set the desired render stage, remembering the previous one
    RenderStage prevRenderStage = setRenderStage(renderStage);
    // Enable our render target
    cubeMap.begin(Framebuffer::defaultPolicy());
    // For each of the environment's faces (TOP, DOWN, NORTH, SOUTH, EAST, WEST)
    for (U8 i = 0; i < 6; ++i) {
        // Draw to the current cubemap face
        cubeMap.drawToFace(hasColor ? TextureDescriptor::AttachmentType::Color0
                                    : TextureDescriptor::AttachmentType::Depth,
                           i + arrayOffset);
        // Point our camera to the correct face
        _cubeCamera->lookAt(pos, TabCenter[i], TabUp[i]);
        // And generated required matrices
        _cubeCamera->renderLookAt();
        // Pass our render function to the renderer
        SceneManager::instance().renderVisibleNodes(renderStage, true, i);
    }
    // Resolve our render target
    cubeMap.end();
    // Return to our previous rendering stage
    setRenderStage(prevRenderStage);
    // Restore our previous camera
    kernel.getCameraMgr().popActiveCamera();
}

void GFXDevice::generateDualParaboloidMap(Framebuffer& targetBuffer,
                                          const U32 arrayOffset,
                                          const vec3<F32>& pos,
                                          const vec2<F32>& zPlanes,
                                          RenderStage renderStage)
{
    const Texture_ptr& colorAttachment = targetBuffer.getAttachment(TextureDescriptor::AttachmentType::Color0, false);
    const Texture_ptr& depthAttachment = targetBuffer.getAttachment(TextureDescriptor::AttachmentType::Depth, false);
    // Color attachment takes precedent over depth attachment
    bool hasColor = (colorAttachment != nullptr);
    bool hasDepth = (depthAttachment != nullptr);
    bool isValidFB = true;
    if (hasColor) {
        // We only need the color attachment
        isValidFB = colorAttachment->getTextureType() == TextureType::TEXTURE_2D_ARRAY;
    } else {
        // We don't have a color attachment, so we require a cube map depth   // attachment
        isValidFB = hasDepth && depthAttachment->getTextureType() == TextureType::TEXTURE_2D_ARRAY;
    }
    // Make sure we have a proper render target to draw to
    if (!isValidFB) {
        // Future formats must be added later (e.g. cube map arrays)
        Console::errorfn(Locale::get(_ID("ERROR_GFX_DEVICE_INVALID_FB_DP")));
        return;
    }
    Kernel& kernel = Application::instance().kernel();
    // Set a 90 degree vertical FoV perspective projection
    _dualParaboloidCamera->setProjection(1.0f, 180.0f, zPlanes);
    // Set the cube camera as the currently active one
    kernel.getCameraMgr().pushActiveCamera(_dualParaboloidCamera);
    // Set the desired render stage, remembering the previous one
    RenderStage prevRenderStage = setRenderStage(renderStage);
    // Enable our render target
    targetBuffer.begin(Framebuffer::defaultPolicy());
        for (U8 i = 0; i < 2; ++i) {
            targetBuffer.drawToLayer(hasColor ? TextureDescriptor::AttachmentType::Color0
                                              : TextureDescriptor::AttachmentType::Depth,
                                     i + arrayOffset);
            // Point our camera to the correct face
            _dualParaboloidCamera->lookAt(pos, (i == 0 ? WORLD_Z_NEG_AXIS : WORLD_Z_AXIS) + pos, WORLD_Y_AXIS);
            // And generated required matrices
            _dualParaboloidCamera->renderLookAt();
            // Pass our render function to the renderer
            SceneManager::instance().renderVisibleNodes(renderStage, true, i);
        }
    targetBuffer.end();
    // Return to our previous rendering stage
    setRenderStage(prevRenderStage);
    // Restore our previous camera
    kernel.getCameraMgr().popActiveCamera();
}

/// If the stateBlock doesn't exist in the state block map, add it for future reference
bool GFXDevice::registerRenderStateBlock(const RenderStateBlock& descriptor) {
    // Each combination of render states has a unique hash value
    size_t hashValue = descriptor.getHash();
    // Find the corresponding render state block
    // Create a new one if none are found. The GFXDevice class is
    // responsible for deleting these!
    std::pair<RenderStateMap::iterator, bool> result =
        hashAlg::emplace(_stateBlockMap, hashValue, descriptor);
    // Return true if registration was successful 
    return result.second;
}

/// Activate the render state block described by the specified hash value (0 == default state block)
size_t GFXDevice::setStateBlock(size_t stateBlockHash) {
    // Passing 0 is a perfectly acceptable way of enabling the default render state block
    if (stateBlockHash == 0) {
        stateBlockHash = _defaultStateBlockHash;
    }

    // If the new state hash is different from the previous one
    if (stateBlockHash != _currentStateBlockHash) {
        // Remember the previous state hash
        _previousStateBlockHash = _currentStateBlockHash;
        // Update the current state hash
        _currentStateBlockHash = stateBlockHash;
        RenderStateMap::const_iterator currentStateIt = _stateBlockMap.find(_currentStateBlockHash);
        RenderStateMap::const_iterator previousStateIt = _stateBlockMap.find(_previousStateBlockHash);

        DIVIDE_ASSERT(currentStateIt != previousStateIt &&
                      currentStateIt != std::cend(_stateBlockMap) &&
                      previousStateIt != std::cend(_stateBlockMap),
                      "GFXDevice error: Invalid state blocks detected on activation!");

        // Activate the new render state block in an rendering API dependent way
        _api->activateStateBlock(currentStateIt->second, previousStateIt->second);
    }
    // Return the previous state hash
    return _previousStateBlockHash;
}

/// Return the the render state block defined by the specified hash value.
const RenderStateBlock& GFXDevice::getRenderStateBlock(size_t renderStateBlockHash) const {
    // Find the render state block associated with the received hash value
    RenderStateMap::const_iterator it = _stateBlockMap.find(renderStateBlockHash);
    // Assert if it doesn't exist. Avoids programming errors.
    DIVIDE_ASSERT(it != std::cend(_stateBlockMap),
                  "GFXDevice error: Invalid render state block hash specified "
                  "for getRenderStateBlock!");
    // Return the state block's descriptor
    return it->second;
}

void GFXDevice::increaseResolution() {
    const WindowManager& winManager = Application::instance().windowManager();
    const vec2<U16>& resolution = winManager.getActiveWindow().getDimensions();
    const vectorImpl<GPUState::GPUVideoMode>& displayModes = _state.getDisplayModes(winManager.targetDisplay());

    vectorImpl<GPUState::GPUVideoMode>::const_reverse_iterator it;
    for (it = std::rbegin(displayModes); it != std::rend(displayModes); ++it) {
        const vec2<U16>& tempResolution = it->_resolution;

        if (resolution.width < tempResolution.width &&
            resolution.height < tempResolution.height) {
            WindowManager& winMgr = Application::instance().windowManager();
            winMgr.handleWindowEvent(WindowEvent::RESOLUTION_CHANGED,
                                     winMgr.getActiveWindow().getGUID(),
                                     to_int(tempResolution.width),
                                     to_int(tempResolution.height));
            return;
        }
    }
}

void GFXDevice::decreaseResolution() {
    const WindowManager& winManager = Application::instance().windowManager();
    const vec2<U16>& resolution = winManager.getActiveWindow().getDimensions();
    const vectorImpl<GPUState::GPUVideoMode>& displayModes = _state.getDisplayModes(winManager.targetDisplay());
    
    vectorImpl<GPUState::GPUVideoMode>::const_iterator it;
    for (it = std::begin(displayModes); it != std::end(displayModes); ++it) {
        const vec2<U16>& tempResolution = it->_resolution;
        if (resolution.width > tempResolution.width &&
            resolution.height > tempResolution.height) {
            WindowManager& winMgr = Application::instance().windowManager();
            winMgr.handleWindowEvent(WindowEvent::RESOLUTION_CHANGED,
                                     winMgr.getActiveWindow().getGUID(),
                                     to_int(tempResolution.width),
                                     to_int(tempResolution.height));
            return;
        }
    }
}

void GFXDevice::toggleFullScreen() {
    WindowManager& winManager = Application::instance().windowManager();
    switch (winManager.getActiveWindow().type()) {
        case WindowType::WINDOW:
        case WindowType::SPLASH:
            winManager.getActiveWindow().type(WindowType::FULLSCREEN_WINDOWED);
            break;
        case WindowType::FULLSCREEN_WINDOWED:
            winManager.getActiveWindow().type(WindowType::FULLSCREEN);
            break;
        case WindowType::FULLSCREEN:
            winManager.getActiveWindow().type(WindowType::WINDOW);
            break;
    };
}

/// The main entry point for any resolution change request
void GFXDevice::onChangeResolution(U16 w, U16 h) {
    // Make sure we are in a valid state that allows resolution updates
    if (_renderTarget[to_const_uint(RenderTargetID::SCREEN)]._buffer != nullptr) {
        // Update resolution only if it's different from the current one.
        // Avoid resolution change on minimize so we don't thrash render targets
        if (vec2<U16>(w, h) ==  _renderTarget[to_const_uint(RenderTargetID::SCREEN)]._buffer->getResolution() ||
           !(w > 1 && h > 1)) {
            return;
        }
        // Update render targets with the new resolution
        for (U32 i = 0; i < to_const_uint(RenderTargetID::COUNT); ++i) {
            Framebuffer* renderTarget = _renderTarget[i]._buffer;
            if (renderTarget) {
                renderTarget->create(w, h);
            }
        }
    }

    // Update post-processing render targets and buffers
    PostFX::instance().updateResolution(w, h);
    _gpuBlock._data._invScreenDimension.xy(1.0f / w, 1.0f / h);
    _gpuBlock._updated = true;
    // Update the 2D camera so it matches our new rendering viewport
    _2DCamera->setProjection(vec4<F32>(0, to_float(w), 0, to_float(h)), vec2<F32>(-1, 1));
}

/// Return a GFXDevice specific matrix or a derivative of it
void GFXDevice::getMatrix(const MATRIX& mode, mat4<F32>& mat) {
    // The matrix names are self-explanatory
    if (mode == MATRIX::VIEW_PROJECTION) {
        mat.set(_gpuBlock._data._ViewProjectionMatrix);
    } else if (mode == MATRIX::VIEW) {
        mat.set(_gpuBlock._data._ViewMatrix);
    } else if (mode == MATRIX::PROJECTION) {
        mat.set(_gpuBlock._data._ProjectionMatrix);
    } else if (mode == MATRIX::TEXTURE) {
        mat.identity();
        Console::errorfn(Locale::get(_ID("ERROR_TEXTURE_MATRIX_ACCESS")));
    } else if (mode == MATRIX::VIEW_INV) {
        _gpuBlock._data._ViewMatrix.getInverse(mat);
    } else if (mode == MATRIX::PROJECTION_INV) {
        _gpuBlock._data._ProjectionMatrix.getInverse(mat);
    } else if (mode == MATRIX::VIEW_PROJECTION_INV) {
        _gpuBlock._data._ViewProjectionMatrix.getInverse(mat);
    } else {
        DIVIDE_ASSERT(
            false,
            "GFXDevice error: attempted to query an invalid matrix target!");
    }
}

/// Update the internal GPU data buffer with the clip plane values
void GFXDevice::updateClipPlanes() {
    GPUBlock::GPUData& data = _gpuBlock._data;
    for (U8 i = 0; i < Config::MAX_CLIP_PLANES; ++i) {
        data._clipPlanes[i] = _clippingPlanes[i].getEquation();
    }
    _gpuBlock._updated = true;
}

/// Update the internal GPU data buffer with the updated viewport dimensions
void GFXDevice::updateViewportInternal(const vec4<I32>& viewport) {
    // Change the viewport on the Rendering API level
    _api->changeViewport(viewport);
    // Update the buffer with the new value
    _gpuBlock._data._ViewPort.set(viewport.x, viewport.y, viewport.z, viewport.w);
    _gpuBlock._updated = true;
}

/// Update the virtual camera's matrices and upload them to the GPU
F32* GFXDevice::lookAt(const mat4<F32>& viewMatrix, const vec3<F32>& eyePos) {
    bool updated = false;

    GPUBlock::GPUData& data = _gpuBlock._data;

    if (eyePos != _gpuBlock._data._cameraPosition.xyz()) {
        data._cameraPosition.xyz(eyePos);
        updated = true;
    }

    if (viewMatrix != _gpuBlock._data._ViewMatrix) {
        data._ViewMatrix.set(viewMatrix);
        updated = true;
    }

    if (updated) {
        mat4<F32>::Multiply(data._ViewMatrix, data._ProjectionMatrix, data._ViewProjectionMatrix);
        computeFrustumPlanes();
        _gpuBlock._updated = true;
    }

    return data._ViewMatrix.mat;
}

/// Enable an orthographic projection and upload the corresponding matrices to
/// the GPU
F32* GFXDevice::setProjection(const vec4<F32>& rect, const vec2<F32>& planes) {
    GPUBlock::GPUData& data = _gpuBlock._data;

    data._ProjectionMatrix.ortho(rect.x, rect.y, rect.z, rect.w, planes.x, planes.y);

    data._ZPlanesCombined.xy(planes);

    mat4<F32>::Multiply(data._ViewMatrix, data._ProjectionMatrix, data._ViewProjectionMatrix);

    computeFrustumPlanes();

    _gpuBlock._updated = true;

    return data._ProjectionMatrix.mat;
}

/// Enable a perspective projection and upload the corresponding matrices to the
/// GPU
F32* GFXDevice::setProjection(F32 FoV, F32 aspectRatio,  const vec2<F32>& planes) {
    GPUBlock::GPUData& data = _gpuBlock._data;

    data._ProjectionMatrix.perspective(Angle::DegreesToRadians(FoV),
                                       aspectRatio,
                                       planes.x, planes.y);

    data._ZPlanesCombined.xy(planes);
    data._cameraPosition.w = aspectRatio;
    data._renderProperties.z = FoV;
    data._renderProperties.w = std::tan(FoV * 0.5f);
    mat4<F32>::Multiply(data._ViewMatrix, data._ProjectionMatrix, data._ViewProjectionMatrix);
    computeFrustumPlanes();
    _gpuBlock._updated = true;

    return data._ProjectionMatrix.mat;
}

void GFXDevice::computeFrustumPlanes(const mat4<F32>& invViewProj, vec4<F32>* planesOut) {
    // Get world-space coordinates for clip-space bounds.
    vec4<F32> lbn(invViewProj * vec4<F32>(-1, -1, -1, 1));
    vec4<F32> ltn(invViewProj * vec4<F32>(-1, 1, -1, 1));
    vec4<F32> lbf(invViewProj * vec4<F32>(-1, -1, 1, 1));
    vec4<F32> rbn(invViewProj * vec4<F32>(1, -1, -1, 1));
    vec4<F32> rtn(invViewProj * vec4<F32>(1, 1, -1, 1));
    vec4<F32> rbf(invViewProj * vec4<F32>(1, -1, 1, 1));
    vec4<F32> rtf(invViewProj * vec4<F32>(1, 1, 1, 1));

    vec3<F32> lbn_pos(lbn.xyz() / lbn.w);
    vec3<F32> ltn_pos(ltn.xyz() / ltn.w);
    vec3<F32> lbf_pos(lbf.xyz() / lbf.w);
    vec3<F32> rbn_pos(rbn.xyz() / rbn.w);
    vec3<F32> rtn_pos(rtn.xyz() / rtn.w);
    vec3<F32> rbf_pos(rbf.xyz() / rbf.w);
    vec3<F32> rtf_pos(rtf.xyz() / rtf.w);

    // Get plane equations for all sides of frustum.
    vec3<F32> left_normal(Cross(lbf_pos - lbn_pos, ltn_pos - lbn_pos));
    left_normal.normalize();
    vec3<F32> right_normal(Cross(rtn_pos - rbn_pos, rbf_pos - rbn_pos));
    right_normal.normalize();
    vec3<F32> top_normal(Cross(ltn_pos - rtn_pos, rtf_pos - rtn_pos));
    top_normal.normalize();
    vec3<F32> bottom_normal(Cross(rbf_pos - rbn_pos, lbn_pos - rbn_pos));
    bottom_normal.normalize();
    vec3<F32> near_normal(Cross(ltn_pos - lbn_pos, rbn_pos - lbn_pos));
    near_normal.normalize();
    vec3<F32> far_normal(Cross(rtf_pos - rbf_pos, lbf_pos - rbf_pos));
    far_normal.normalize();

    planesOut[0].set(left_normal, -Dot(left_normal, lbn_pos)); // Left
    planesOut[1].set(right_normal, -Dot(right_normal, rbn_pos)); // Right
    planesOut[2].set(near_normal, -Dot(near_normal, lbn_pos)); // Near
    planesOut[3].set(far_normal, -Dot(far_normal, lbf_pos)); // Far
    planesOut[4].set(top_normal, -Dot(top_normal, ltn_pos)); // Top
    planesOut[5].set(bottom_normal, -Dot(bottom_normal, lbn_pos)); // Bottom
}

void GFXDevice::computeFrustumPlanes() {
    computeFrustumPlanes(getMatrix(MATRIX::VIEW_PROJECTION_INV), _gpuBlock._data._frustumPlanes);
    _gpuBlock._updated = true;
}

/// Enable or disable 2D rendering mode 
/// (orthographic projection, no depth reads)
void GFXDevice::toggle2D(bool state) {
    // Remember the previous state hash
    static size_t previousStateBlockHash = 0;
    // Prevent double 2D toggle to the same state (e.g. in a loop)
    if (state == _2DRendering) {
        return;
    }
    Kernel& kernel = Application::instance().kernel();
    _2DRendering = state;
    // If we need to enable 2D rendering
    if (state) {
        // Activate the 2D render state block
        previousStateBlockHash = setStateBlock(_state2DRenderingHash);
        // Push the 2D camera
        kernel.getCameraMgr().pushActiveCamera(_2DCamera);
        // Upload 2D camera matrices to the GPU
        _2DCamera->renderLookAt();
    } else {
        // Reverting to 3D implies popping the 2D camera
        kernel.getCameraMgr().popActiveCamera();
        // And restoring the previous state block
        setStateBlock(previousStateBlockHash);
    }
}

/// Update the rendering viewport
void GFXDevice::setViewport(const vec4<I32>& viewport) {
    // Avoid redundant changes
    _viewportUpdate = !viewport.compare(_viewport.top());

    if (_viewportUpdate) {
        // Push the new viewport
        _viewport.push(viewport);
        // Activate the new viewport
        updateViewportInternal(viewport);
    }
}

/// Restore the viewport to it's previous value
void GFXDevice::restoreViewport() {
    // If we didn't push a new viewport, there's nothing to pop
    if (!_viewportUpdate) {
        return;
    }
    // Restore the viewport
    _viewport.pop();
    // Activate the new top viewport
    updateViewportInternal(_viewport.top());
    _viewportUpdate = false;
}

/// Set a new viewport clearing the previous stack first
void GFXDevice::setBaseViewport(const vec4<I32>& viewport) {
    while (!_viewport.empty()) {
        _viewport.pop();
    }
    _viewport.push(viewport);

    // Set the new viewport
    updateViewportInternal(viewport);
    // The forced viewport can't be popped
    _viewportUpdate = false;
}

void GFXDevice::onCameraUpdate(Camera& camera) {
    if (drawDebugAxis()) {
        // We need to transform the gizmo so that it always remains axis aligned
        // Create a world matrix using a look at function with the eye position
        // backed up from the camera's view direction
        _axisGizmo->worldMatrix(
            mat4<F32>(-camera.getViewDir() * 2, VECTOR3_ZERO, camera.getUpDir()) *
            _gpuBlock._data._ViewMatrix.getInverse());
        _axisGizmo->paused(false);
    } else {
        _axisGizmo->paused(true);
    }
}

/// Depending on the context, either immediately call the function, or pass it
/// to the loading thread via a queue
bool GFXDevice::loadInContext(const CurrentContext& context,
                              const DELEGATE_CBK<>& callback) {
    // Skip invalid callbacks
    if (!callback) {
        return false;
    }
    bool loadInThread = context == CurrentContext::GFX_LOADING_CTX && _state.loadingThreadAvailable();
    // If we want and can call the function in the loading thread, add it to the
    // lock-free, single-producer, single-consumer queue
    if (loadInThread) {
        _state.addToLoadQueue(callback);
    } else {
        callback();
    }

    // The callback is valid and has been processed
    return true;
}

/// Transform our depth buffer to a HierarchicalZ buffer (for occlusion queries and screen space reflections)
void GFXDevice::constructHIZ() {
    auto setAndGetHalfViewport = [](vec4<I32>& viewportIn) -> vec4<I32>& {
        viewportIn /= 2;
        // Ensure that the viewport size is always at least 1x1
        viewportIn.z = viewportIn.z > 0 ? viewportIn.z : 1;
        viewportIn.w = viewportIn.w > 0 ? viewportIn.w : 1;
        return viewportIn;
    };

    // The depth buffer's resolution should be equal to the screen's resolution
    Framebuffer* screenTarget = _renderTarget[to_const_uint(RenderTargetID::SCREEN)]._buffer;
    vec2<U16> resolution = screenTarget->getResolution();
    // Bind the depth texture to the first texture unit
    screenTarget->bind(to_const_ubyte(ShaderProgram::TextureUsage::DEPTH), TextureDescriptor::AttachmentType::Depth);
    // We use a special shader that downsamples the buffer
    // We will use a state block that disables color writes as we will render only a depth image,
    // disables depth testing but allows depth writes
    // Set the depth buffer as the currently active render target
    Framebuffer::FramebufferTarget depthOnlyTarget;
    depthOnlyTarget._clearColorBuffersOnBind = false;
    depthOnlyTarget._clearDepthBufferOnBind = false;
    depthOnlyTarget._changeViewport = false;
    depthOnlyTarget._drawMask.fill(false);
    depthOnlyTarget._drawMask[to_const_uint(TextureDescriptor::AttachmentType::Depth)] = true;

    screenTarget->begin(depthOnlyTarget);
    // Calculate the number of mipmap levels we need to generate
    U32 numLevels = 1 + to_uint(floorf(log2f(fmaxf(to_float(resolution.width),
                                                   to_float(resolution.height)))));
    // Store the current width and height of each mip
    vec4<I32> currentViewport(0, 0, resolution.width, resolution.height);
    vec4<I32> previousViewport(_viewport.top());
    // We skip the first level as that's our full resolution image
    for (U16 i = 1; i < numLevels; ++i) {
        // Calculate next viewport size
        // Update the viewport with the new resolution
        updateViewportInternal(setAndGetHalfViewport(currentViewport));
        // Bind next mip level for rendering but first restrict fetches only to previous level
        screenTarget->setMipLevel(i - 1, i - 1, i, TextureDescriptor::AttachmentType::Depth);
        // Dummy draw command as the full screen quad is generated completely in the vertex shader
        drawTriangle(_stateDepthOnlyRenderingHash, _HIZConstructProgram);
    }
    updateViewportInternal(previousViewport);
    // Reset mipmap level range for the depth buffer
    screenTarget->resetMipLevel(TextureDescriptor::AttachmentType::Depth);
    // Unbind the render target
    screenTarget->end();
}

/// Find an unused primitive object or create a new one and return it
IMPrimitive* GFXDevice::getOrCreatePrimitive(bool allowPrimitiveRecycle) {
    IMPrimitive* tempPriv = nullptr;
    // Find an available and unused primitive (a zombie primitive)
    vectorImpl<IMPrimitive*>::iterator it;
    it = std::find_if(std::begin(_imInterfaces), std::end(_imInterfaces),
                      [](IMPrimitive* const priv) { 
                            return (priv && !priv->inUse()); 
                      });
    // If we allow resurrected primitives check if we have one available
    if (allowPrimitiveRecycle && it != std::end(_imInterfaces)) {
        tempPriv = *it;
        // If we have a valid zombie, resurrect it
        tempPriv->clear();

    } else {
        // If we do not have a valid zombie, we create a new primitive
        tempPriv = newIMP();
        // And add it to our container. The GFXDevice class is responsible for
        // deleting these!
        _imInterfaces.push_back(tempPriv);
    }
    // Toggle zombification of the primitive on or off depending on our request
    tempPriv->_canZombify = allowPrimitiveRecycle;

    return tempPriv;
}

/// Extract the pixel data from the main render target's first color attachment
/// and save it as a TGA image
void GFXDevice::Screenshot(const stringImpl& filename) {
    // Get the screen's resolution
    const vec2<U16>& resolution =
        _renderTarget[to_const_uint(RenderTargetID::SCREEN)]._buffer
            ->getResolution();
    // Allocate sufficiently large buffers to hold the pixel data
    U32 bufferSize = resolution.width * resolution.height * 4;
    U8* imageData = MemoryManager_NEW U8[bufferSize];
    // Read the pixels from the main render target (RGBA16F)
    _renderTarget[to_const_uint(RenderTargetID::SCREEN)]._buffer->readData(
        GFXImageFormat::RGBA, GFXDataFormat::UNSIGNED_BYTE, imageData);
    // Save to file
    ImageTools::SaveSeries(filename,
                           vec2<U16>(resolution.width, resolution.height), 32,
                           imageData);
    // Delete local buffers
    MemoryManager::DELETE_ARRAY(imageData);
}
};

Commits for Divide-Framework/trunk/Source Code/Platform/Video/GFXDevice.cpp

Diff revisions: vs.
Revision Author Commited Message
712 Diff Diff IonutCava picture IonutCava Tue 24 May, 2016 16:18:37 +0000

[IonutCava]
- Added the concept of “buffer” to be used by GenericVertexData and ShaderBuffer without caring if it’s persistent or not.
— Persistent buffers handle their own locking instead of relying on the parent class
- HiZ occlusion culling fixes and optimizations

710 Diff Diff IonutCava picture IonutCava Fri 20 May, 2016 16:24:40 +0000

[IonutCava]
- Code cleanup
- Initial work on Scene loading and unloading with and without unloading assets:
— Need to change AIManager from a Singleton to a per-scene component

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)

706 Diff Diff IonutCava picture IonutCava Fri 13 May, 2016 16:29:40 +0000

[IonutCava]
- Renamed LightManager to LightPool and changed it from a Singleton to a scene specific pointer
- More scene load / unload updates (changing scenes now works properly!)
- Allowed GUI interface to hold per-scene elements and only render relevant ones
- Reduced dependencies between nodes, scenemanager, scene, loaders and kernel
- Removed Reflector base class as it was useless and integrate Water class with the Reflection system already in place
- Extended RenderingComponent’s reflection system to allow refraction as well

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

698 Diff Diff IonutCava picture IonutCava Mon 09 May, 2016 20:21:34 +0000

[IonutCava]
- Remove all anaglyph rendering code. nVidia gained 1 sympathy point with me for VRWorks:
— Single Pass Stereo completely removes the need for double rendering the scene for each eye.
— LiquidVR also support similar technology
-— Once VR development starts for Divide, documentation for both SDKs should be pretty sorted.

693 Diff Diff IonutCava picture IonutCava Thu 28 Apr, 2016 16:19:35 +0000

[IonutCava]
- Cleanup Time and Profiling code

689 Diff Diff IonutCava picture IonutCava Fri 22 Apr, 2016 21:12:33 +0000

[IonutCava]
- More code cleanup (profile timers)
- Changed project build settings for better performance in both Debug and Profile targets

688 Diff Diff IonutCava picture IonutCava Fri 22 Apr, 2016 16:23:33 +0000

[IonutCava]
- Code cleanup

687 IonutCava picture IonutCava Thu 21 Apr, 2016 20:44:41 +0000

[IonutCava]
- Fix stack overflow issue with SGN relationship cache invalidation
- Code cleanup