splited MaterialRenderList and MaterialRenderMap, RenderExtraBuffer

inc/hgl/graph/MaterialRenderList.h

@@ -0,0 +1,75 @@
+#pragma once
+#include<hgl/graph/RenderNode.h>
+
+VK_NAMESPACE_BEGIN
+struct RenderNodeExtraBuffer;
+
+/**
+* 同一材质的对象渲染列表
+*/
+class MaterialRenderList
+{
+    GPUDevice *device;
+    RenderCmdBuffer *cmd_buf;
+
+    Material *mtl;
+
+    RenderNodeList rn_list;
+
+private:
+
+    RenderNodeExtraBuffer *extra_buffer;
+
+    struct RenderItem
+    {
+                uint32_t            first;
+                uint32_t            count;
+
+                Pipeline *          pipeline;
+                MaterialInstance *  mi;
+        const   VertexInputData *   vid;
+
+    public:
+
+        void Set(Renderable *);
+    };
+
+    SortedSets<MaterialInstance *> mi_set;
+    List<RenderItem> ri_list;
+    uint ri_count;
+
+    void Stat();
+
+protected:
+
+    uint32_t binding_count;
+    VkBuffer *buffer_list;
+    VkDeviceSize *buffer_offset;
+
+            MaterialInstance *  last_mi;
+            Pipeline *          last_pipeline;
+    const   VertexInputData *   last_vid;
+            uint                last_index;
+
+    void Bind(MaterialInstance *);
+    bool Bind(const VertexInputData *,const uint);
+
+    void Render(RenderItem *);
+
+public:
+
+    MaterialRenderList(GPUDevice *d,Material *m);
+    ~MaterialRenderList();
+
+    void Add(Renderable *ri,const Matrix4f &mat);
+
+    void ClearData()
+    {
+        rn_list.ClearData();
+    }
+
+    void End();
+
+    void Render(RenderCmdBuffer *);
+};//class MaterialRenderList
+VK_NAMESPACE_END

inc/hgl/graph/MaterialRenderMap.h

@@ -0,0 +1,32 @@
+#pragma once
+#include<hgl/graph/MaterialRenderList.h>
+
+VK_NAMESPACE_BEGIN
+class MaterialRenderMap:public ObjectMap<Material *,MaterialRenderList>
+{
+public:
+
+    MaterialRenderMap()=default;
+    virtual ~MaterialRenderMap()=default;
+
+    void Begin()
+    {
+        for(auto *it:data_list)
+            it->value->ClearData();
+    }
+
+    void End()
+    {
+        for(auto *it:data_list)
+            it->value->End();
+    }
+
+    void Render(RenderCmdBuffer *rcb)
+    {
+        if(!rcb)return;
+
+        for(auto *it:data_list)
+            it->value->Render(rcb);
+    }
+};//class MaterialRenderMap
+VK_NAMESPACE_END

inc/hgl/graph/RenderList.h

@@ -3,7 +3,7 @@
 
 #include<hgl/graph/VK.h>
 #include<hgl/graph/SceneNode.h>
-#include<hgl/graph/RenderNode.h>
+#include<hgl/graph/MaterialRenderMap.h>
 #include<hgl/graph/VKArrayBuffer.h>
 #include<hgl/graph/VKMaterial.h>
 namespace hgl

inc/hgl/graph/RenderNode.h

@@ -1,9 +1,6 @@
 #ifndef HGL_GRAPH_RENDER_NODE_INCLUDE
 #define HGL_GRAPH_RENDER_NODE_INCLUDE
 
-#include<hgl/math/Math.h>
-#include<hgl/type/Map.h>
-#include<hgl/type/SortedSets.h>
 #include<hgl/graph/VK.h>
 namespace hgl
 {
@@ -23,105 +20,6 @@ namespace hgl
         };
 
         using RenderNodeList=List<RenderNode>;
-
-        struct RenderNodeExtraBuffer;
-
-        /**
-         * 同一材质的对象渲染列表
-         */
-        class MaterialRenderList
-        {
-            GPUDevice *device;
-            RenderCmdBuffer *cmd_buf;
-
-            Material *mtl;
-
-            RenderNodeList rn_list;
-
-        private:
-
-            RenderNodeExtraBuffer *extra_buffer;
-
-            struct RenderItem
-            {
-                        uint32_t            first;
-                        uint32_t            count;
-
-                        Pipeline *          pipeline;
-                        MaterialInstance *  mi;
-                const   VertexInputData *   vid;
-
-            public:
-
-                void Set(Renderable *);
-            };
-
-            SortedSets<MaterialInstance *> mi_set;
-            List<RenderItem> ri_list;
-            uint ri_count;
-
-            void Stat();
-
-        protected:
-
-            uint32_t binding_count;
-            VkBuffer *buffer_list;
-            VkDeviceSize *buffer_offset;
-
-                  MaterialInstance *  last_mi;
-                  Pipeline *          last_pipeline;
-            const VertexInputData *   last_vid;
-                  uint                last_index;
-
-            void Bind(MaterialInstance *);
-            bool Bind(const VertexInputData *,const uint);
-
-            void Render(RenderItem *);
-
-        public:
-
-            MaterialRenderList(GPUDevice *d,Material *m);
-            ~MaterialRenderList();
-
-            void Add(Renderable *ri,const Matrix4f &mat);
-
-            void ClearData()
-            {
-                rn_list.ClearData();
-            }
-
-            void End();
-
-            void Render(RenderCmdBuffer *);
-        };
-
-        class MaterialRenderMap:public ObjectMap<Material *,MaterialRenderList>
-        {
-        public:
-
-            MaterialRenderMap()=default;
-            virtual ~MaterialRenderMap()=default;
-
-            void Begin()
-            {
-                for(auto *it:data_list)
-                    it->value->ClearData();
-            }
-
-            void End()
-            {
-                for(auto *it:data_list)
-                    it->value->End();
-            }
-
-            void Render(RenderCmdBuffer *rcb)
-            {
-                if(!rcb)return;
-
-                for(auto *it:data_list)
-                    it->value->Render(rcb);
-            }
-        };
     }//namespace graph
 }//namespace hgl
 #endif//HGL_GRAPH_RENDER_NODE_INCLUDE

src/SceneGraph/CMakeLists.txt

@@ -38,11 +38,13 @@ SET(SCENE_GRAPH_HEADER  ${SG_INCLUDE_PATH}/SceneInfo.h
                         ${SG_INCLUDE_PATH}/RenderNode.h
                         ${SG_INCLUDE_PATH}/SceneOrient.h
                         ${SG_INCLUDE_PATH}/RenderList.h
+                        ${SG_INCLUDE_PATH}/MaterialRenderList.h
+                        ${SG_INCLUDE_PATH}/MaterialRenderMap.h
                         )
 
-SET(SCENE_GRAPH_SOURCE  
+SET(SCENE_GRAPH_SOURCE  RenderList.cpp
-                        RenderList.cpp
+                        MaterialRenderList.cpp
-                        RenderNode.cpp
+                        RenderExtraBuffer.h
                         SceneNode.cpp
                         SceneOrient.cpp)
 

src/SceneGraph/MaterialRenderList.cpp

@@ -0,0 +1,338 @@
+#include<hgl/graph/MaterialRenderList.h>
+#include<hgl/graph/VKRenderable.h>
+#include<hgl/graph/VKDevice.h>
+#include<hgl/graph/VKCommandBuffer.h>
+#include<hgl/graph/VKVertexInput.h>
+#include<hgl/util/sort/Sort.h>
+#include"RenderExtraBuffer.h"
+
+/**
+* 
+* 理论上讲，我们需要按以下顺序排序
+*
+*   for(material)
+*       for(pipeline)
+*           for(material_instance)
+*               for(vbo)
+*/
+
+template<> 
+int Comparator<hgl::graph::RenderNode>::compare(const hgl::graph::RenderNode &obj_one,const hgl::graph::RenderNode &obj_two) const
+{
+    int off;
+
+    hgl::graph::Renderable *ri_one=obj_one.ri;
+    hgl::graph::Renderable *ri_two=obj_two.ri;
+
+    //比较管线
+    {
+        off=ri_one->GetPipeline()
+           -ri_two->GetPipeline();
+
+        if(off)
+            return off;
+    }
+
+    //比较材质实例
+    {
+        off=ri_one->GetMaterialInstance()
+           -ri_two->GetMaterialInstance();
+
+        if(off)
+            return off;
+    }
+
+    //比较模型
+    {
+        off=ri_one->GetPrimitive()
+           -ri_two->GetPrimitive();
+
+        if(off)
+            return off;
+    }
+
+    return 0;
+}
+
+VK_NAMESPACE_BEGIN
+MaterialRenderList::MaterialRenderList(GPUDevice *d,Material *m)
+{
+    device=d;
+    cmd_buf=nullptr;
+    mtl=m;
+    extra_buffer=nullptr;
+
+    const VertexInput *vi=mtl->GetVertexInput();
+
+    binding_count=vi->GetCount();
+    buffer_list=new VkBuffer[binding_count];
+    buffer_offset=new VkDeviceSize[binding_count];                        
+}
+
+MaterialRenderList::~MaterialRenderList()
+{
+    delete[] buffer_offset;
+    delete[] buffer_list;
+
+    SAFE_CLEAR(extra_buffer)
+}
+
+void MaterialRenderList::Add(Renderable *ri,const Matrix4f &mat)
+{
+    RenderNode rn;
+
+    rn.local_to_world=mat;
+    rn.ri=ri;
+
+    rn_list.Add(rn);
+}
+
+void MaterialRenderList::End()
+{
+    //排序
+    {
+        Comparator<hgl::graph::RenderNode> rnc;
+
+        Sort(rn_list,&rnc);
+    }
+
+    //写入LocalToWorld数据
+    {
+        uint count=rn_list.GetCount();
+
+        if(count<=0)return;
+
+        if(!extra_buffer)
+            extra_buffer=new RenderNodeExtraBuffer;
+
+        if(extra_buffer->node_count<count)
+            extra_buffer->NodeAlloc(device,count);
+
+        //写入数据
+        extra_buffer->WriteData(rn_list.GetData(),count);
+    }
+
+    Stat();
+}
+
+void MaterialRenderList::RenderItem::Set(Renderable *ri)
+{
+    pipeline    =ri->GetPipeline();
+    mi          =ri->GetMaterialInstance();
+    vid         =ri->GetVertexInputData();
+}
+
+void MaterialRenderList::Stat()
+{
+    const uint count=rn_list.GetCount();
+    RenderNode *rn=rn_list.GetData();
+
+    ri_list.ClearData();
+    ri_list.PreMalloc(count);
+
+    mi_set.ClearData();
+
+    RenderItem *ri=ri_list.GetData();
+            
+    ri_count=1;
+
+    ri->first=0;
+    ri->count=1;
+    ri->Set(rn->ri);
+
+    last_pipeline   =ri->pipeline;
+    last_mi         =ri->mi;
+    last_vid        =ri->vid;
+
+    mi_set.Add(last_mi);
+
+    ++rn;
+
+    for(uint i=1;i<count;i++)
+    {   
+        if(last_pipeline==rn->ri->GetPipeline())
+            if(last_mi==rn->ri->GetMaterialInstance())
+                if(last_vid==rn->ri->GetVertexInputData())
+                {
+                    ++ri->count;
+                    ++rn;
+                    continue;
+                }
+
+        ++ri_count;
+        ++ri;
+
+        ri->first=i;
+        ri->count=1;
+        ri->Set(rn->ri);
+
+        if(last_mi!=ri->mi)
+            mi_set.Add(ri->mi);
+
+        last_pipeline   =ri->pipeline;
+        last_mi         =ri->mi;
+        last_vid        =ri->vid;
+
+        ++rn;
+    }
+}
+
+void MaterialRenderList::Bind(MaterialInstance *mi)
+{
+}
+
+bool MaterialRenderList::Bind(const VertexInputData *vid,const uint first)
+{
+    //binding号都是在VertexInput::CreateVIL时连续紧密排列生成的，所以bind时first_binding写0就行了。
+
+    const VIL *vil=last_mi->GetVIL();
+
+    if(vil->GetCount(VertexInputGroup::Basic)!=vid->binding_count)
+        return(false);                                                  //这里基本不太可能，因为CreateRenderable时就会检查值是否一样
+
+    uint count=0;
+
+    //Basic组，它所有的VBO信息均来自于Primitive，由vid参数传递进来
+    {
+        hgl_cpy(buffer_list,vid->buffer_list,vid->binding_count);
+        hgl_cpy(buffer_offset,vid->buffer_offset,vid->binding_count);
+
+        count=vid->binding_count;
+    }
+
+    if(count<binding_count) //材质组
+    {
+        const uint mtl_binding_count=vil->GetCount(VertexInputGroup::MaterialInstanceID);
+
+        if(mtl_binding_count>0)
+        {
+            if(mtl_binding_count!=1)                                    //只有MaterialInstanceID
+                return(false);
+
+            count+=mtl_binding_count;
+        }
+    }
+
+    if(count<binding_count) //Joint组，暂未支持
+    {
+        const uint joint_id_binding_count=vil->GetCount(VertexInputGroup::JointID);
+
+        if(joint_id_binding_count>0)                                        //有矩阵信息
+        {
+            count+=joint_id_binding_count;
+
+            if(count<binding_count) //JointWeight组
+            {
+                const uint joing_weight_binding_count=vil->GetCount(VertexInputGroup::JointWeight);
+
+                if(joing_weight_binding_count!=1)
+                {
+                    ++count;
+                }
+                else //JointWieght不为1是个bug，除非未来支持8权重
+                {
+                    return(false);
+                }
+            }
+            else //有JointID没有JointWeight? 这是个BUG
+            {
+                return(false);
+            }
+        }
+    }
+
+    if(count<binding_count)//LocalToWorld组，由RenderList合成
+    {
+        const uint l2w_binding_count=vil->GetCount(VertexInputGroup::LocalToWorld);
+
+        if(l2w_binding_count>0)                                         //有变换矩阵信息
+        {
+            if(l2w_binding_count!=4)
+                return(false);
+
+            hgl_cpy(buffer_list+count,extra_buffer->l2w_buffer,4);
+
+            for(uint i=0;i<4;i++)
+                buffer_offset[count+i]=first*16;                        //mat4每列都是rgba32f，自然是16字节
+
+            count+=l2w_binding_count;
+        }
+    }
+
+    if(count!=binding_count)
+    {
+        //还有没支持的绑定组？？？？
+
+        return(false);
+    }
+
+    cmd_buf->BindVBO(0,count,buffer_list,buffer_offset);
+
+    return(true);
+}
+
+void MaterialRenderList::Render(RenderItem *ri)
+{
+    if(last_pipeline!=ri->pipeline)
+    {
+        cmd_buf->BindPipeline(ri->pipeline);
+        last_pipeline=ri->pipeline;
+
+        last_mi=nullptr;
+        last_vid=nullptr;
+
+        //这里未来尝试换pipeline同时不换mi/primitive是否需要重新绑定mi/primitive
+    }
+
+    if(last_mi!=ri->mi)
+    {
+        Bind(ri->mi);
+        last_mi=ri->mi;
+
+        last_vid=nullptr;
+    }
+
+    if(!ri->vid->Comp(last_vid))
+    {
+        Bind(ri->vid,ri->first);
+        last_vid=ri->vid;
+    }
+
+    const IndexBufferData *ibd=last_vid->index_buffer;
+
+    if(ibd->buffer)
+    {
+        cmd_buf->BindIBO(ibd);
+
+        cmd_buf->DrawIndexed(ibd->buffer->GetCount(),ri->count);
+    }
+    else
+    {
+        cmd_buf->Draw(last_vid->vertex_count,ri->count);
+    }
+}
+
+void MaterialRenderList::Render(RenderCmdBuffer *rcb)
+{
+    if(!rcb)return;
+    const uint count=rn_list.GetCount();
+
+    if(count<=0)return;
+
+    if(ri_count<=0)return;
+
+    cmd_buf=rcb;
+
+    RenderItem *ri=ri_list.GetData();
+
+    last_pipeline   =nullptr;
+    last_mi         =nullptr;
+    last_vid        =nullptr;
+
+    for(uint i=0;i<ri_count;i++)
+    {
+        Render(ri);
+        ++ri;
+    }
+}
+VK_NAMESPACE_END

src/SceneGraph/RenderExtraBuffer.h

@@ -0,0 +1,117 @@
+#pragma once
+#include<hgl/graph/VKVertexAttribBuffer.h>
+
+VK_NAMESPACE_BEGIN
+// ubo_range大致分为三档:
+//
+//  16k: Mali-T系列或更早、Mali-G71、nVidia GeForce RTX 3070 Laptop为16k
+// 
+//  64k: 大部分手机与PC均为64k
+// 
+// >64k: Intel 核显与 PowerVR 为128MB，AMD显卡为4GB，可视为随显存无上限。
+// 
+// 我们使用uint8类型在vertex input中保存MaterialInstance ID，表示范围0-255。
+// 所以MaterialInstance结构容量按16k/64k分为两个档次，64字节和256字节
+
+// 如果一定要使用超过16K/64K硬件限制的容量，有两种办法
+// 一、分多次渲染，使用UBO Offset偏移UBO数据区。
+// 二、使用SSBO，但这样会导致性能下降，所以不推荐使用。
+
+// 但我们不解决这个问题
+// 我们天然要求将材质实例数据分为两个等级，同时要求一次渲染不能超过256种材质实例。
+// 所以 UBO Range为16k时，实例数据不能超过64字节。UBO Range为64k时，实例数据不能超过256字节。
+
+/*
+* 渲染节点额外提供的数据
+*/
+struct RenderNodeExtraBuffer
+{
+    uint node_count;                            ///<渲染节点数量
+    uint mi_count;                              ///<材质实例数量
+
+    uint mi_size;                               ///<单个材质实例数量长度
+    DeviceBuffer *mi_data_buffer;               ///<材质实例数据UBO/SSBO
+
+    VBO *mi_id;
+    VkBuffer mi_id_buffer;
+
+    VBO *bone_id,*bone_weight;
+    VkBuffer bone_id_buffer,bone_weight_buffer;
+
+    VBO *l2w_vbo[4];
+    VkBuffer l2w_buffer[4];
+
+public:
+
+    RenderNodeExtraBuffer()
+    {
+        hgl_zero(*this);
+    }
+
+    ~RenderNodeExtraBuffer()
+    {
+        Clear();
+    }
+
+    void Clear()
+    {
+        SAFE_CLEAR(mi_id)
+        SAFE_CLEAR(mi_data_buffer);
+
+        SAFE_CLEAR(bone_id)
+        SAFE_CLEAR(bone_weight)
+
+        SAFE_CLEAR(l2w_vbo[0])
+        SAFE_CLEAR(l2w_vbo[1])
+        SAFE_CLEAR(l2w_vbo[2])
+        SAFE_CLEAR(l2w_vbo[3])
+        node_count=0;
+    }
+
+    void NodeAlloc(GPUDevice *dev,const uint c)
+    {
+        Clear();
+        node_count=power_to_2(c);
+
+        for(uint i=0;i<4;i++)
+        {
+            l2w_vbo[i]=dev->CreateVBO(VF_V4F,node_count);
+            l2w_buffer[i]=l2w_vbo[i]->GetBuffer();
+        }
+    }
+
+    void MIAlloc(GPUDevice *dev,const uint c)
+    {
+        if(c<=0||mi_size<=0)return;
+    
+        mi_count=power_to_2(c);
+
+        mi_id=dev->CreateVBO(VF_V1U8,mi_count);
+        mi_id_buffer=mi_id->GetBuffer();
+
+        mi_data_buffer=dev->CreateUBO(mi_count*mi_size);
+    }
+
+    void WriteData(RenderNode *render_node,const uint count)
+    {
+        RenderNode *rn;
+        glm::vec4 *tp;
+
+        for(uint col=0;col<4;col++)
+        {
+            tp=(glm::vec4 *)(l2w_vbo[col]->Map());
+
+            rn=render_node;
+
+            for(uint i=0;i<count;i++)
+            {
+                *tp=rn->local_to_world[col];
+                ++tp;
+                ++rn;
+            }
+
+            l2w_vbo[col]->Unmap();
+        }
+    }
+};//struct RenderNodeExtraBuffer
+VK_NAMESPACE_END

src/SceneGraph/RenderNode.cpp

@@ -1,455 +0,0 @@
-#include<hgl/graph/RenderNode.h>
-#include<hgl/graph/VKRenderable.h>
-#include<hgl/graph/VKDevice.h>
-#include<hgl/graph/VKVertexAttribBuffer.h>
-#include<hgl/graph/VKVertexInput.h>
-#include<hgl/graph/VKCommandBuffer.h>
-#include<hgl/util/sort/Sort.h>
-
-/**
-* 
-* 理论上讲，我们需要按以下顺序排序
-*
-*   for(material)
-*       for(pipeline)
-*           for(material_instance)
-*               for(vbo)
-*/
-
-template<> 
-int Comparator<hgl::graph::RenderNode>::compare(const hgl::graph::RenderNode &obj_one,const hgl::graph::RenderNode &obj_two) const
-{
-    int off;
-
-    hgl::graph::Renderable *ri_one=obj_one.ri;
-    hgl::graph::Renderable *ri_two=obj_two.ri;
-
-    //比较管线
-    {
-        off=ri_one->GetPipeline()
-           -ri_two->GetPipeline();
-
-        if(off)
-            return off;
-    }
-
-    //比较材质实例
-    {
-        off=ri_one->GetMaterialInstance()
-           -ri_two->GetMaterialInstance();
-
-        if(off)
-            return off;
-    }
-
-    //比较模型
-    {
-        off=ri_one->GetPrimitive()
-           -ri_two->GetPrimitive();
-
-        if(off)
-            return off;
-    }
-
-    return 0;
-}
-
-namespace hgl
-{
-    namespace graph
-    {
-    // ubo_range大致分为三档:
-    //
-    //  16k: Mali-T系列或更早、Mali-G71、nVidia GeForce RTX 3070 Laptop为16k
-    // 
-    //  64k: 大部分手机与PC均为64k
-    // 
-    // >64k: Intel 核显与 PowerVR 为128MB，AMD显卡为4GB，可视为随显存无上限。
-    // 
-    // 我们使用uint8类型在vertex input中保存MaterialInstance ID，表示范围0-255。
-    // 所以MaterialInstance结构容量按16k/64k分为两个档次，64字节和256字节
-
-    // 如果一定要使用超过16K/64K硬件限制的容量，有两种办法
-    // 一、分多次渲染，使用UBO Offset偏移UBO数据区。
-    // 二、使用SSBO，但这样会导致性能下降，所以不推荐使用。
-
-    // 但我们不解决这个问题
-    // 我们天然要求将材质实例数据分为两个等级，同时要求一次渲染不能超过256种材质实例。
-    // 所以 UBO Range为16k时，实例数据不能超过64字节。UBO Range为64k时，实例数据不能超过256字节。
-
-        /*
-         * 渲染节点额外提供的数据
-         */
-        struct RenderNodeExtraBuffer
-        {
-            uint node_count;                            ///<渲染节点数量
-            uint mi_count;                              ///<材质实例数量
-
-            uint mi_size;                               ///<单个材质实例数量长度
-            DeviceBuffer *mi_data_buffer;               ///<材质实例数据UBO/SSBO
-
-            VBO *mi_id;
-            VkBuffer mi_id_buffer;
-
-            VBO *bone_id,*bone_weight;
-            VkBuffer bone_id_buffer,bone_weight_buffer;
-
-            VBO *l2w_vbo[4];
-            VkBuffer l2w_buffer[4];
-
-        public:
-
-            RenderNodeExtraBuffer()
-            {
-                hgl_zero(*this);
-            }
-
-            ~RenderNodeExtraBuffer()
-            {
-                Clear();
-            }
-
-            void Clear()
-            {
-                SAFE_CLEAR(mi_id)
-                SAFE_CLEAR(mi_data_buffer);
-
-                SAFE_CLEAR(bone_id)
-                SAFE_CLEAR(bone_weight)
-
-                SAFE_CLEAR(l2w_vbo[0])
-                SAFE_CLEAR(l2w_vbo[1])
-                SAFE_CLEAR(l2w_vbo[2])
-                SAFE_CLEAR(l2w_vbo[3])
-                node_count=0;
-            }
-
-            void NodeAlloc(GPUDevice *dev,const uint c)
-            {
-                Clear();
-                node_count=power_to_2(c);
-
-                for(uint i=0;i<4;i++)
-                {
-                    l2w_vbo[i]=dev->CreateVBO(VF_V4F,node_count);
-                    l2w_buffer[i]=l2w_vbo[i]->GetBuffer();
-                }
-            }
-
-            void MIAlloc(GPUDevice *dev,const uint c)
-            {
-                if(c<=0||mi_size<=0)return;
-    
-                mi_count=power_to_2(c);
-
-                mi_id=dev->CreateVBO(VF_V1U8,mi_count);
-                mi_id_buffer=mi_id->GetBuffer();
-
-                mi_data_buffer=dev->CreateUBO(mi_count*mi_size);
-            }
-
-            void WriteData(RenderNode *render_node,const uint count)
-            {
-                RenderNode *rn;
-                glm::vec4 *tp;
-
-                for(uint col=0;col<4;col++)
-                {
-                    tp=(glm::vec4 *)(l2w_vbo[col]->Map());
-
-                    rn=render_node;
-
-                    for(uint i=0;i<count;i++)
-                    {
-                        *tp=rn->local_to_world[col];
-                        ++tp;
-                        ++rn;
-                    }
-
-                    l2w_vbo[col]->Unmap();
-                }
-            }
-        };//struct RenderNodeExtraBuffer
-
-        MaterialRenderList::MaterialRenderList(GPUDevice *d,Material *m)
-        {
-            device=d;
-            cmd_buf=nullptr;
-            mtl=m;
-            extra_buffer=nullptr;
-
-            const VertexInput *vi=mtl->GetVertexInput();
-
-            binding_count=vi->GetCount();
-            buffer_list=new VkBuffer[binding_count];
-            buffer_offset=new VkDeviceSize[binding_count];                        
-        }
-
-        MaterialRenderList::~MaterialRenderList()
-        {
-            delete[] buffer_offset;
-            delete[] buffer_list;
-
-            SAFE_CLEAR(extra_buffer)
-        }
-
-        void MaterialRenderList::Add(Renderable *ri,const Matrix4f &mat)
-        {
-            RenderNode rn;
-
-            rn.local_to_world=mat;
-            rn.ri=ri;
-
-            rn_list.Add(rn);
-        }
-
-        void MaterialRenderList::End()
-        {
-            //排序
-            {
-                Comparator<hgl::graph::RenderNode> rnc;
-
-                Sort(rn_list,&rnc);
-            }
-
-            //写入LocalToWorld数据
-            {
-                uint count=rn_list.GetCount();
-
-                if(count<=0)return;
-
-                if(!extra_buffer)
-                    extra_buffer=new RenderNodeExtraBuffer;
-
-                if(extra_buffer->node_count<count)
-                    extra_buffer->NodeAlloc(device,count);
-
-                //写入数据
-                extra_buffer->WriteData(rn_list.GetData(),count);
-            }
-
-            Stat();
-        }
-
-        void MaterialRenderList::RenderItem::Set(Renderable *ri)
-        {
-            pipeline    =ri->GetPipeline();
-            mi          =ri->GetMaterialInstance();
-            vid         =ri->GetVertexInputData();
-        }
-
-        void MaterialRenderList::Stat()
-        {
-            const uint count=rn_list.GetCount();
-            RenderNode *rn=rn_list.GetData();
-
-            ri_list.ClearData();
-            ri_list.PreMalloc(count);
-
-            mi_set.ClearData();
-
-            RenderItem *ri=ri_list.GetData();
-            
-            ri_count=1;
-
-            ri->first=0;
-            ri->count=1;
-            ri->Set(rn->ri);
-
-            last_pipeline   =ri->pipeline;
-            last_mi         =ri->mi;
-            last_vid        =ri->vid;
-
-            mi_set.Add(last_mi);
-
-            ++rn;
-
-            for(uint i=1;i<count;i++)
-            {   
-                if(last_pipeline==rn->ri->GetPipeline())
-                    if(last_mi==rn->ri->GetMaterialInstance())
-                        if(last_vid==rn->ri->GetVertexInputData())
-                        {
-                            ++ri->count;
-                            ++rn;
-                            continue;
-                        }
-
-                ++ri_count;
-                ++ri;
-
-                ri->first=i;
-                ri->count=1;
-                ri->Set(rn->ri);
-
-                if(last_mi!=ri->mi)
-                    mi_set.Add(ri->mi);
-
-                last_pipeline   =ri->pipeline;
-                last_mi         =ri->mi;
-                last_vid        =ri->vid;
-
-                ++rn;
-            }
-        }
-
-        void MaterialRenderList::Bind(MaterialInstance *mi)
-        {
-        }
-
-        bool MaterialRenderList::Bind(const VertexInputData *vid,const uint first)
-        {
-            //binding号都是在VertexInput::CreateVIL时连续紧密排列生成的，所以bind时first_binding写0就行了。
-
-            const VIL *vil=last_mi->GetVIL();
-
-            if(vil->GetCount(VertexInputGroup::Basic)!=vid->binding_count)
-                return(false);                                                  //这里基本不太可能，因为CreateRenderable时就会检查值是否一样
-
-            uint count=0;
-
-            //Basic组，它所有的VBO信息均来自于Primitive，由vid参数传递进来
-            {
-                hgl_cpy(buffer_list,vid->buffer_list,vid->binding_count);
-                hgl_cpy(buffer_offset,vid->buffer_offset,vid->binding_count);
-
-                count=vid->binding_count;
-            }
-
-            if(count<binding_count) //材质组
-            {
-                const uint mtl_binding_count=vil->GetCount(VertexInputGroup::MaterialInstanceID);
-
-                if(mtl_binding_count>0)
-                {
-                    if(mtl_binding_count!=1)                                    //只有MaterialInstanceID
-                        return(false);
-
-                    count+=mtl_binding_count;
-                }
-            }
-
-            if(count<binding_count) //Joint组，暂未支持
-            {
-                const uint joint_id_binding_count=vil->GetCount(VertexInputGroup::JointID);
-
-                if(joint_id_binding_count>0)                                        //有矩阵信息
-                {
-                    count+=joint_id_binding_count;
-
-                    if(count<binding_count) //JointWeight组
-                    {
-                        const uint joing_weight_binding_count=vil->GetCount(VertexInputGroup::JointWeight);
-
-                        if(joing_weight_binding_count!=1)
-                        {
-                            ++count;
-                        }
-                        else //JointWieght不为1是个bug，除非未来支持8权重
-                        {
-                            return(false);
-                        }
-                    }
-                    else //有JointID没有JointWeight? 这是个BUG
-                    {
-                        return(false);
-                    }
-                }
-            }
-
-            if(count<binding_count)//LocalToWorld组，由RenderList合成
-            {
-                const uint l2w_binding_count=vil->GetCount(VertexInputGroup::LocalToWorld);
-
-                if(l2w_binding_count>0)                                         //有变换矩阵信息
-                {
-                    if(l2w_binding_count!=4)
-                        return(false);
-
-                    hgl_cpy(buffer_list+count,extra_buffer->l2w_buffer,4);
-
-                    for(uint i=0;i<4;i++)
-                        buffer_offset[count+i]=first*16;                        //mat4每列都是rgba32f，自然是16字节
-
-                    count+=l2w_binding_count;
-                }
-            }
-
-            if(count!=binding_count)
-            {
-                //还有没支持的绑定组？？？？
-
-                return(false);
-            }
-
-            cmd_buf->BindVBO(0,count,buffer_list,buffer_offset);
-
-            return(true);
-        }
-
-        void MaterialRenderList::Render(RenderItem *ri)
-        {
-            if(last_pipeline!=ri->pipeline)
-            {
-                cmd_buf->BindPipeline(ri->pipeline);
-                last_pipeline=ri->pipeline;
-
-                last_mi=nullptr;
-                last_vid=nullptr;
-
-                //这里未来尝试换pipeline同时不换mi/primitive是否需要重新绑定mi/primitive
-            }
-
-            if(last_mi!=ri->mi)
-            {
-                Bind(ri->mi);
-                last_mi=ri->mi;
-
-                last_vid=nullptr;
-            }
-
-            if(!ri->vid->Comp(last_vid))
-            {
-                Bind(ri->vid,ri->first);
-                last_vid=ri->vid;
-            }
-
-            const IndexBufferData *ibd=last_vid->index_buffer;
-
-            if(ibd->buffer)
-            {
-                cmd_buf->BindIBO(ibd);
-
-                cmd_buf->DrawIndexed(ibd->buffer->GetCount(),ri->count);
-            }
-            else
-            {
-                cmd_buf->Draw(last_vid->vertex_count,ri->count);
-            }
-        }
-
-        void MaterialRenderList::Render(RenderCmdBuffer *rcb)
-        {
-            if(!rcb)return;
-            const uint count=rn_list.GetCount();
-
-            if(count<=0)return;
-
-            if(ri_count<=0)return;
-
-            cmd_buf=rcb;
-
-            RenderItem *ri=ri_list.GetData();
-
-            last_pipeline   =nullptr;
-            last_mi         =nullptr;
-            last_vid        =nullptr;
-
-            for(uint i=0;i<ri_count;i++)
-            {
-                Render(ri);
-                ++ri;
-            }
-        }
-    }//namespace graph
-}//namespace hgl