• Advertisement
Sign in to follow this  

DX11 Seperation of Render and SceneSystem.

This topic is 2032 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hi,
im try to write my own render engine as a hobby so its nothing fancy. Im trying to seperate scence graphics and graphics objects from the Rendering part of engine.

Lets say i have a object that contains the Geometry, lets call the one DX11Geometry. It contains stuff like DX11buffers and other DX11 related stuff. This object does only exist in the rendersystem..

in the SceneManagment i have a scenobject that repesent the actuall object in the ending, it contain the stuff like translation, rotation..


Here's my question, how do I best link the SceneObject to the DX11Geometry object. Remeber i do want the RenderSystem completet independent of the ScenceObject, so SceneObject dont know what a DX11Geometryobject is..

I was think that SceneObjectcould have some kind uID that points to a DX11Gemotry object in the RenderSystem..
like RenderSystem->RenderGeometry(int uid). Im not sure how efficent this methody is, when because i would need some kind of array in the RendetrSystem to store all the DX11Gemotry object..

any thoughts on this are more then welcome.. im quite new to this kind of graphics programing.

Share this post


Link to post
Share on other sites
Advertisement
In my engine I do this:

class Scene
Object[] objects
class Object
vec3 pos
Quaterion rot
Model *model
virtual void draw()
class Renderer
vec3 cameraPos
Scene *scene //draws scene at cameraPos
virtual void draw()

class Model
vec3 verts,norms,etc
virtual void process()
void load(string path)
do stuff
process()


I then make custom implementations of Model and Renderer for each API (and even rendering technique, deferred vs forward, etc). You can have a load that's API independant and then subclass Model to make a DX11Model that would store any necessary info and override process() to handle uploading the data to the GPU, and have different subclasses of Renderer that override draw() to handle the different APIs, etc

Share this post


Link to post
Share on other sites

Here's my question, how do I best link the SceneObject to the DX11Geometry object. Remeber i do want the RenderSystem completet independent of the ScenceObject, so SceneObject dont know what a DX11Geometryobject is..


Create a data bridge type that links a SceneObject to a DX11GeometryObject, SceneObject doesnt know about DX11GeoObject and vice-versa. std::pair<SceneObject*,DX11GeometryObject*> fits well for that. Then you have a process that goes through the pairs and copies the appropriate data across from SceneObject (transforms etc) into your DX11 constant buffers. The object is a two way association, but a 1 way data pipeline.

Share this post


Link to post
Share on other sites
Hi, thx a lot for two excellent answers. Think i will actually look in to both solutions two my problem. Have to say that i totally forgot about the std::pair.
I like this forum already people that actually know there stuff! wink.png

Share this post


Link to post
Share on other sites

[quote name='korvax' timestamp='1340221764' post='4951096']
Here's my question, how do I best link the SceneObject to the DX11Geometry object. Remeber i do want the RenderSystem completet independent of the ScenceObject, so SceneObject dont know what a DX11Geometryobject is..


Create a data bridge type that links a SceneObject to a DX11GeometryObject, SceneObject doesnt know about DX11GeoObject and vice-versa. std::pair<SceneObject*,DX11GeometryObject*> fits well for that. Then you have a process that goes through the pairs and copies the appropriate data across from SceneObject (transforms etc) into your DX11 constant buffers. The object is a two way association, but a 1 way data pipeline.
[/quote]

Is this a better way of doing it than using a int ID thats mapping them, and then storing DX11GeometryObject in a array structure, or is it faster to use pair and then map it with objects?

Share this post


Link to post
Share on other sites

I then make custom implementations of Model and Renderer for each API (and even rendering technique, deferred vs forward, etc). You can have a load that's API independant and then subclass Model to make a DX11Model that would store any necessary info and override process() to handle uploading the data to the GPU, and have different subclasses of Renderer that override draw() to handle the different APIs, etc

In this case do you end up doing a a lot of static_cast in each frame in the RenderSystem between the Model and the DX11Model? If so isn't it rather costly to cast ever model each fram?

Share this post


Link to post
Share on other sites
I use something similar to what I've read in a presentation from the halo guys.
my logic side requests an ID from the renderer when it creates an object, based on strings to a mesh+material.
this ID is 32bit, lower 16bit represent an index into an array of object references in the renderer, and upper 16bit represent a counter that identifies the object in the array slot.

int createObject(std::string Mesh,std::string Material)
{
int ID=find(Mesh,Material);
if(ID>=0)
return ID;
int ID=FindEmptySlot();
QueueObjectForStreaming(ID,Mesh,Material);
ID|=(m_CounterArray[ID]+1)<<16;
return ID;
}

every time I create an object I try to find a free slot in the array of obiect, increase that counter and return (Counter<<16)|ObjectIndex;
if the array is full, I obviously overwrite some old object. when the logic side want to render, it just passes the ID + world matrix and the renderer indexes into the object array, if the upper 16bit match with the counter, the object is rendered, if not, it's obviously not the right one.
if the counter is < of the ID, then it means the streaming system is trying to load it in a queue, then I drop to some default object (box in my case), if the counter is higher, the object needs to be recreated. there are various ways the handle it, e.g. your "draw" call to the renderer could return that this object is invalid.

the nice part is that the rendering and logic are really completely separated (ok, not fully true, my collision checks try to access the mesh to share data), you can have a lot of references to the same object, it's transparently streamed in and also it can be removed (if not replaced), if memory is getting critical. Edited by Krypt0n

Share this post


Link to post
Share on other sites

I use something similar to what I've read in a presentation from the halo guys.
my logic side requests an ID from the renderer when it creates an object, based on strings to a mesh+material.
this ID is 32bit, lower 16bit represent an index into an array of object references in the renderer, and upper 16bit represent a counter that identifies the object in the array slot.

int createObject(std::string Mesh,std::string Material)
{
int ID=find(Mesh,Material);
if(ID>=0)
return ID;
int ID=FindEmptySlot();
QueueObjectForStreaming(ID,Mesh,Material);
ID|=(m_CounterArray[ID]+1)<<16;
return ID;
}

every time I create an object I try to find a free slot in the array of obiect, increase that counter and return (Counter<<16)|ObjectIndex;
if the array is full, I obviously overwrite some old object. when the logic side want to render, it just passes the ID + world matrix and the renderer indexes into the object array, if the upper 16bit match with the counter, the object is rendered, if not, it's obviously not the right one.
if the counter is < of the ID, then it means the streaming system is trying to load it in a queue, then I drop to some default object (box in my case), if the counter is higher, the object needs to be recreated. there are various ways the handle it, e.g. your "draw" call to the renderer could return that this object is invalid.

the nice part is that the rendering and logic are really completely separated (ok, not fully true, my collision checks try to access the mesh to share data), you can have a lot of references to the same object, it's transparently streamed in and also it can be removed (if not replaced), if memory is getting critical.


That looks like a good solution but I have some questions...

How is the logic side able generate the ID? Since it is separated from the renderer it has no way of knowing where the object is stored right? the createObject() function is part of the renderer?, if so the scene system shouldn't be able to call it. Edited by TiagoCosta

Share this post


Link to post
Share on other sites
at some place you need to have an interface where you pass data. so yes, "CreateObject" is a renderer interface that the logic side uses. but it's not a connection, it's still separate. if you e.g. want to run on a server without a GPU, you just don't run the renderer thread (or update, if single threaded), the logic still can get some handles and pass them for rendering, they would have no meaning, they also have no meaning of the engine, it's just a handle that could potentially represent something. if you want, you can create a compile time flag (aka #define) that does not create any rendering code at all and "CreateObject" always just immediately returns 0;

(my system is actually a bit more complicated and it's not the renderer that is handling that part, but the resource system that is shared between game logic, scene'graph', renderer, streaming, physics etc.).

Share this post


Link to post
Share on other sites

I use something similar to what I've read in a presentation from the halo guys.


Do you remeber where you found that articlepls?

thx for you help thought.

Share this post


Link to post
Share on other sites

[quote name='Krypt0n' timestamp='1340280800' post='4951311']
I use something similar to what I've read in a presentation from the halo guys.


Do you remeber where you found that articlepls?

thx for you help thought.
[/quote]
took me my lunch time, thought it was was halo 3 paper, didn't remember it was from halo2, so it was kind of the last one I checked, lol, yet, here it is:
http://www.bungie.net/images/Inside/publications/presentations/publicationsdes/engineering/butcher_gametech04.pdf
page 29

cheers

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
  • Advertisement
  • Popular Tags

  • Advertisement
  • Popular Now

  • Similar Content

    • By AxeGuywithanAxe
      I wanted to see how others are currently handling descriptor heap updates and management.
      I've read a few articles and there tends to be three major strategies :
      1 ) You split up descriptor heaps per shader stage ( i.e one for vertex shader , pixel , hull, etc)
      2) You have one descriptor heap for an entire pipeline
      3) You split up descriptor heaps for update each update frequency (i.e EResourceSet_PerInstance , EResourceSet_PerPass , EResourceSet_PerMaterial, etc)
      The benefits of the first two approaches is that it makes it easier to port current code, and descriptor / resource descriptor management and updating tends to be easier to manage, but it seems to be not as efficient.
      The benefits of the third approach seems to be that it's the most efficient because you only manage and update objects when they change.
    • By evelyn4you
      hi,
      until now i use typical vertexshader approach for skinning with a Constantbuffer containing the transform matrix for the bones and an the vertexbuffer containing bone index and bone weight.
      Now i have implemented realtime environment  probe cubemaping so i have to render my scene from many point of views and the time for skinning takes too long because it is recalculated for every side of the cubemap.
      For Info i am working on Win7 an therefore use one Shadermodel 5.0 not 5.x that have more options, or is there a way to use 5.x in Win 7
      My Graphic Card is Directx 12 compatible NVidia GTX 960
      the member turanszkij has posted a good for me understandable compute shader. ( for Info: in his engine he uses an optimized version of it )
      https://turanszkij.wordpress.com/2017/09/09/skinning-in-compute-shader/
      Now my questions
       is it possible to feed the compute shader with my orignial vertexbuffer or do i have to copy it in several ByteAdressBuffers as implemented in the following code ?
        the same question is about the constant buffer of the matrixes
       my more urgent question is how do i feed my normal pipeline with the result of the compute Shader which are 2 RWByteAddressBuffers that contain position an normal
      for example i could use 2 vertexbuffer bindings
      1 containing only the uv coordinates
      2.containing position and normal
      How do i copy from the RWByteAddressBuffers to the vertexbuffer ?
       
      (Code from turanszkij )
      Here is my shader implementation for skinning a mesh in a compute shader:
      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 struct Bone { float4x4 pose; }; StructuredBuffer<Bone> boneBuffer;   ByteAddressBuffer vertexBuffer_POS; // T-Pose pos ByteAddressBuffer vertexBuffer_NOR; // T-Pose normal ByteAddressBuffer vertexBuffer_WEI; // bone weights ByteAddressBuffer vertexBuffer_BON; // bone indices   RWByteAddressBuffer streamoutBuffer_POS; // skinned pos RWByteAddressBuffer streamoutBuffer_NOR; // skinned normal RWByteAddressBuffer streamoutBuffer_PRE; // previous frame skinned pos   inline void Skinning(inout float4 pos, inout float4 nor, in float4 inBon, in float4 inWei) {  float4 p = 0, pp = 0;  float3 n = 0;  float4x4 m;  float3x3 m3;  float weisum = 0;   // force loop to reduce register pressure  // though this way we can not interleave TEX - ALU operations  [loop]  for (uint i = 0; ((i &lt; 4) &amp;&amp; (weisum&lt;1.0f)); ++i)  {  m = boneBuffer[(uint)inBon].pose;  m3 = (float3x3)m;   p += mul(float4(pos.xyz, 1), m)*inWei;  n += mul(nor.xyz, m3)*inWei;   weisum += inWei;  }   bool w = any(inWei);  pos.xyz = w ? p.xyz : pos.xyz;  nor.xyz = w ? n : nor.xyz; }   [numthreads(1024, 1, 1)] void main( uint3 DTid : SV_DispatchThreadID ) {  const uint fetchAddress = DTid.x * 16; // stride is 16 bytes for each vertex buffer now...   uint4 pos_u = vertexBuffer_POS.Load4(fetchAddress);  uint4 nor_u = vertexBuffer_NOR.Load4(fetchAddress);  uint4 wei_u = vertexBuffer_WEI.Load4(fetchAddress);  uint4 bon_u = vertexBuffer_BON.Load4(fetchAddress);   float4 pos = asfloat(pos_u);  float4 nor = asfloat(nor_u);  float4 wei = asfloat(wei_u);  float4 bon = asfloat(bon_u);   Skinning(pos, nor, bon, wei);   pos_u = asuint(pos);  nor_u = asuint(nor);   // copy prev frame current pos to current frame prev pos streamoutBuffer_PRE.Store4(fetchAddress, streamoutBuffer_POS.Load4(fetchAddress)); // write out skinned props:  streamoutBuffer_POS.Store4(fetchAddress, pos_u);  streamoutBuffer_NOR.Store4(fetchAddress, nor_u); }  
    • By mister345
      Hi, can someone please explain why this is giving an assertion EyePosition!=0 exception?
       
      _lightBufferVS->viewMatrix = DirectX::XMMatrixLookAtLH(XMLoadFloat3(&_lightBufferVS->position), XMLoadFloat3(&_lookAt), XMLoadFloat3(&up));
      It looks like DirectX doesnt want the 2nd parameter to be a zero vector in the assertion, but I passed in a zero vector with this exact same code in another program and it ran just fine. (Here is the version of the code that worked - note XMLoadFloat3(&m_lookAt) parameter value is (0,0,0) at runtime - I debugged it - but it throws no exceptions.
          m_viewMatrix = DirectX::XMMatrixLookAtLH(XMLoadFloat3(&m_position), XMLoadFloat3(&m_lookAt), XMLoadFloat3(&up)); Here is the repo for the broken code (See LightClass) https://github.com/mister51213/DirectX11Engine/blob/master/DirectX11Engine/LightClass.cpp
      and here is the repo with the alternative version of the code that is working with a value of (0,0,0) for the second parameter.
      https://github.com/mister51213/DX11Port_SoftShadows/blob/master/Engine/lightclass.cpp
    • By mister345
      Hi, can somebody please tell me in clear simple steps how to debug and step through an hlsl shader file?
      I already did Debug > Start Graphics Debugging > then captured some frames from Visual Studio and
      double clicked on the frame to open it, but no idea where to go from there.
       
      I've been searching for hours and there's no information on this, not even on the Microsoft Website!
      They say "open the  Graphics Pixel History window" but there is no such window!
      Then they say, in the "Pipeline Stages choose Start Debugging"  but the Start Debugging option is nowhere to be found in the whole interface.
      Also, how do I even open the hlsl file that I want to set a break point in from inside the Graphics Debugger?
       
      All I want to do is set a break point in a specific hlsl file, step thru it, and see the data, but this is so unbelievably complicated
      and Microsoft's instructions are horrible! Somebody please, please help.
       
       
       

    • By mister345
      I finally ported Rastertek's tutorial # 42 on soft shadows and blur shading. This tutorial has a ton of really useful effects and there's no working version anywhere online.
      Unfortunately it just draws a black screen. Not sure what's causing it. I'm guessing the camera or ortho matrix transforms are wrong, light directions, or maybe texture resources not being properly initialized.  I didnt change any of the variables though, only upgraded all types and functions DirectX3DVector3 to XMFLOAT3, and used DirectXTK for texture loading. If anyone is willing to take a look at what might be causing the black screen, maybe something pops out to you, let me know, thanks.
      https://github.com/mister51213/DX11Port_SoftShadows
       
      Also, for reference, here's tutorial #40 which has normal shadows but no blur, which I also ported, and it works perfectly.
      https://github.com/mister51213/DX11Port_ShadowMapping
       
  • Advertisement