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
#include "Headers/RenderBin.h"
#include "Graphs/Headers/SceneGraphNode.h"
#include "Managers/Headers/LightManager.h"
#include "Geometry/Shapes/Headers/Object3D.h"
#include "Geometry/Material/Headers/Material.h"
#include "Hardware/Video/Headers/GFXDevice.h"
#include "Hardware/Video/Headers/RenderStateBlock.h"

namespace Divide {

RenderBinItem::RenderBinItem(I32 sortKeyA,
                             I32 sortKeyB, 
                             F32 distToCamSq, 
                             SceneGraphNode* const node) : _node(node),
                                                           _sortKeyA( sortKeyA ),
                                                           _sortKeyB( sortKeyB ),
                                                           _distanceToCameraSq(distToCamSq)
{
    _stateHash = 0;
    RenderingComponent* const renderable = _node->getComponent<RenderingComponent>();
    if (renderable) {
        Material* nodeMaterial = renderable->getMaterialInstance();
        // If we do not have a material, no need to continue
        if (!nodeMaterial) {
            return;
        }
        // Sort by state hash depending on the current rendering stage
        // Save the render state hash value for sorting
        _stateHash = GFX_DEVICE.getStateBlockDescriptor(nodeMaterial->getRenderStateBlock(GFX_DEVICE.getRenderStage())).getHash();
    }
}

struct RenderQueueDistanceBacktoFront{
    bool operator()( const RenderBinItem &a, const RenderBinItem &b) const {
        return a._distanceToCameraSq < b._distanceToCameraSq;
    }
};

struct RenderQueueDistanceFrontToBack{
    bool operator()( const RenderBinItem &a, const RenderBinItem &b) const {
        return a._distanceToCameraSq > b._distanceToCameraSq;
    }
};

/// Sorting opaque items is a 3 step process:
/// 1: sort by shaders
/// 2: if the shader is identical, sort by state hash
/// 3: if shader is identical and state hash is identical, sort by albedo ID
struct RenderQueueKeyCompare{
    //Sort
    bool operator()(const RenderBinItem &a, const RenderBinItem &b) const{
        //Sort by shader in all states The sort key is the shader id (for now)
        if (a._sortKeyA < b._sortKeyA) {
            return true;
        }

        if (a._sortKeyA > b._sortKeyA) {
            return false;
        }
         
        // If the shader values are the same, we use the state hash for sorting
        // The _stateHash is a CRC value created based on the RenderState.
        if (a._stateHash < b._stateHash) {
            return true;
        }

        if (a._stateHash > b._stateHash) {
            return false;
        }

        // If both the shader are the same and the state hashes match, 
        // we sort by the secondary key (usually the texture id)
        return (a._sortKeyB < b._sortKeyB);
    }
};

RenderBin::RenderBin(const RenderBinType& rbType,
                     const RenderingOrder::List& renderOrder,
                     D32 drawKey) : _rbType(rbType),
                                    _renderOrder(renderOrder),
                                    _drawKey(drawKey)
{
    _renderBinStack.reserve(125);
    renderBinTypeToNameMap[RBT_PLACEHOLDER] = "Invalid Bin";
    renderBinTypeToNameMap[RBT_MESH]        = "Mesh Bin";
    renderBinTypeToNameMap[RBT_IMPOSTOR]    = "Impostor Bin";
    renderBinTypeToNameMap[RBT_DELEGATE]    = "Delegate Bin";
    renderBinTypeToNameMap[RBT_TRANSLUCENT] = "Translucent Bin";
    renderBinTypeToNameMap[RBT_SKY]         = "Sky Bin";
    renderBinTypeToNameMap[RBT_WATER]       = "Water Bin";
    renderBinTypeToNameMap[RBT_TERRAIN]     = "Terrain Bin";
    renderBinTypeToNameMap[RBT_PARTICLES]   = "Particle Bin";
    renderBinTypeToNameMap[RBT_VEGETATION_GRASS]   = "Grass Bin";
    renderBinTypeToNameMap[RBT_VEGETATION_TREES]   = "Trees Bin";
    renderBinTypeToNameMap[RBT_DECALS]      = "Decals Bin";
    renderBinTypeToNameMap[RBT_SHADOWS]     = "Shadow Bin";
}

void RenderBin::sort(const RenderStage& currentRenderStage){
    //WriteLock w_lock(_renderBinGetMutex);
    switch(_renderOrder){
        default :
        case RenderingOrder::BY_STATE : {
            if (GFX_DEVICE.isCurrentRenderStage(DEPTH_STAGE)) {
                std::sort(_renderBinStack.begin(), _renderBinStack.end(), 
                          RenderQueueDistanceFrontToBack());
            } else {
                std::sort(_renderBinStack.begin(), _renderBinStack.end(),
                          RenderQueueKeyCompare());
            }
        } break;
        case RenderingOrder::BACK_TO_FRONT : {
            std::sort(_renderBinStack.begin(), _renderBinStack.end(),
                      RenderQueueDistanceBacktoFront());
        } break;
        case RenderingOrder::FRONT_TO_BACK : {
            std::sort(_renderBinStack.begin(), _renderBinStack.end(),
                      RenderQueueDistanceFrontToBack());
        } break;
        case RenderingOrder::NONE : {
            //no need to sort
        } break;
        case RenderingOrder::ORDER_PLACEHOLDER : {
            ERROR_FN(Locale::get("ERROR_INVALID_RENDER_BIN_SORT_ORDER"), 
                     renderBinTypeToNameMap[_rbType]);
        } break;
    };
}

void RenderBin::refresh(){
    //WriteLock w_lock(_renderBinGetMutex);
    _renderBinStack.clear();
    _renderBinStack.reserve(128);
}

void RenderBin::addNodeToBin(SceneGraphNode* const sgn, const vec3<F32>& eyePos){
    I32 keyA = (U32)_renderBinStack.size() + 1;
    I32 keyB = keyA;
    RenderingComponent* const renderable = sgn->getComponent<RenderingComponent>();
    if (renderable) {
        Material* nodeMaterial = renderable->getMaterialInstance();
        if (nodeMaterial){
            nodeMaterial->getSortKeys(keyA, keyB);
        }
    }
	_renderBinStack.push_back(RenderBinItem(keyA,
                                            keyB, 
                                            sgn->getBoundingBoxConst().nearestDistanceFromPointSquared(eyePos), 
                                            sgn));
}

void RenderBin::preRender(const RenderStage& currentRenderStage){
}

void RenderBin::render(const SceneRenderState& renderState, const RenderStage& currentRenderStage){
    //We need to apply different materials for each stage. As nodes are sorted, this should be very fast
    for(U16 j = 0; j < getBinSize(); ++j){
        //Call render and the stage exclusion mask should do the rest
        RenderingComponent* const renderable = getItem(j)._node->getComponent<RenderingComponent>();
        if (renderable) {
            renderable->render(renderState, currentRenderStage);
        }
    }
}

void RenderBin::postRender(const RenderStage& currentRenderStage){

}

};

Commits for Divide-Framework/trunk/Source Code/Rendering/RenderPass/RenderBin.cpp

Diff revisions: vs.
Revision Author Commited Message
331 Diff Diff IonutCava picture IonutCava Sat 06 Dec, 2014 20:53:45 +0000

[Ionut]
- Limited line length to 132 characters to improve readability and diff-comparisons
- Refactored memory allocation/deallocation functions
- Fixed a few compatibility issues with HashMap.h
- Fixed a bug in GPU Skinning shaders (cast a float to int)

330 Diff Diff IonutCava picture IonutCava Fri 10 Oct, 2014 17:19:11 +0000

[Ionut]
- New rendering algorithm :
— Perform second pass for visible nodes gathering all render commands and uploading them to an indirect draw buffer
— Render geometry with indirect draw commands
— Removed per-node range binding of the node buffer. Instead, bind the entire buffer once and index data in shaders
— Use “baseInstance” parameter from IndirectDrawCommand as a node buffer index
— Prefer DSA extensions for updating buffer data where applicable
- Moved all rendering specific code from SceneGraphNode to a new RenderingComponent
- Optimized “toString” utility function

328 Diff Diff IonutCava picture IonutCava Sat 04 Oct, 2014 18:03:36 +0000

[Ionut]
- Added per SceneGraphNode material instances. (Cloned from the SceneNode’s material)
- Improved shader loading for materials

321 Diff Diff IonutCava picture IonutCava Wed 03 Sep, 2014 22:05:15 +0000

[Ionut]
- Added support (and enabled by default) for the EASTL library (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2271.html / https://github.com/paulhodge/EASTL)
— Strings, vectors and hash_maps can be combined from EASTL, STL and Boost (not all combinations work. e.g. EASTL strings with STL/Boost containers due to the lack of a proper hash function)

319 Diff Diff IonutCava picture IonutCava Sat 30 Aug, 2014 21:37:57 +0000

[Ionut]
- Performance tweaks based on profiling data

318 Diff Diff IonutCava picture IonutCava Sat 30 Aug, 2014 17:35:53 +0000

[Ionut]
- Wrapped the entire code in a “Divide” namespace
- VertexBuffers now call “shrink_to_fit” on all internal data storage
- Improved some vector performance by preferring “emplace_back” instead of “push_back” + proepr usage of reserve / resize
- Wrapped OIS specific types and classes in Divide::Input namespace
- Added the messageBox.layout file (forgot it in the previous few commits)

304 Diff Diff IonutCava picture IonutCava Sun 10 Aug, 2014 21:10:54 +0000

[Ionut]
- Redesigned the WorldEditor (dummy functions added for every button/option)
- Improved debug rendering performance by using static IMPrimitives for axis gizmo rendering
- Simplified skeleton rendering

303 Diff Diff IonutCava picture IonutCava Sat 09 Aug, 2014 19:12:32 +0000

[Ionut]
- Fixed mouse movement (change resolution callback properly informs inputInterface of the new window’s width and height)
- Fixed immediate mode emulation rendering (destroyed index buffers for some reason)
- Added axis gizmos for Scene / Scene + Selected Object / Scene + All object cycled via the F5 key in debug

296 Diff Diff k1ngp1n picture k1ngp1n Sun 03 Aug, 2014 14:54:21 +0000

[Ionut]
- nVidia compatibility fixes (GUI, shadows, post-processing)
- Texture bind slots are now part of ShaderProgram class
- Renamed ForwardRenderer class to ForwardPlusRenderer and moved ligh grid creation to it
- More code cleanups and comments

286 IonutCava picture IonutCava Fri 18 Jul, 2014 16:32:03 +0000

[Ionut] [[BR]]

    • very buggy commit ** [[BR]]

- New rendering pipeline: [[BR]]
— Bumped OpenGL version to 4.4 [[BR]]
— Command based rendering. GFXDevice dispatches rendering commands to their respective buffers [[BR]]
— State changing is exclusively controlled by the GFXDevice class [[BR]]
— Material properties and transform matrices are stored in shader storage buffers that are updated after culling and uploaded once per RenderPass. Objects use drawIDs to index data in shaders [[BR]]
— Removed all custom matrices and other, per shader, bind data [[BR]]
— Removed per node light data. Light culling and shader buffers will replace the old light management system [[BR]]
— Framebuffers now use the Texture class for attachments [[BR]]
— Sampler objects now work like RenderStateBlocks. GL_API handles them and stores them in a hash indexed map [[BR]]
— Removed “prepareMaterial” calls. Pre-draw material setup is now only limited to texture binding [[BR]]
— Moved immediate mode emulation and primitive rendering from API level to GFXDevice level [[BR]]
— Framebuffer rendering must be completed with a call to end() to ensure proper MSAA resolve (asserts, otherwise) [[BR]]
- Fixed hash code generation for RenderStateBlock (and now, glSamplerObject) using boost::hash_combine [[BR]]
- Temporarily disabled XML caching for materials [[BR]]