This topic is 2876 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

## Recommended Posts

##### Share on other sites
You're misunderstanding how the multithreading functionality works. The point isn't to allow to have two threads using the device simultaneously, the point is to have multiple threads building up portions of a command list so that your main rendering thread can combine the portions into one big command list that can be submitted to the GPU. You're incorrect when you say that this would be like single-threaded rendering...using a deferred context allows you to spread the overhead of building the command list over multiple threads. The final part where you combine command lists is relatively cheap.

Ultimately your deferred contexts need to be serialized into a single command list. This is because the GPU only reads one command at a time, so you need to make sure that the command lists from your deferred contexts are combined in the order needed for the GPU to properly draw your scene.

If you want to support multiple output windows, you need to do it the old-fashioned way: create a swap chain for each window, render each view seperately, and present with the corresponding swap chain.

##### Share on other sites
Thank you for you comment MJP. I figured it out (with your help), why my way didn't work.

In my opinion, the deferred contexts are only simple software lists. So it shouldn't matter, how the lists are pushed to the immediate contexts. If you use a 3rd thread, you automatically serializes the lists. And this was my problem: the synchronized executing of the command lists. Because I render all my stuff in renderings threads, it doesn't matter, that this threads waits for the (synchronized) immediate context to push their commands. In my opinion, this is faster, than using a 3rd thread.

Now, instead of using a 3rd thread, I simply synchronize my immediate context like this:
for every rendering thread:  render to deferred context  build the command list  lock (immediateContext) {    immediateContext.ExecuteCommandList(commandList, false);  }  dispose the command list

Only the immediate contexts has to be synchronized. Other things, like the device itself are already synchronized by SlimDX or DX. And this was something, I misunderstud in the DX-API :) Now, all the threads render, render, render, and if one of them becomes ready, it renders to the device. Only, if two threads finishes the work simultanously, one of them has to wait for the other to execute the command list. This should be more efficient than using a third thread, which implies synchronized passing of command list from the rendering threads to this 3rd thread, etc.

Thank you for your help :)

• 10
• 11
• 9
• 16
• 19
• ### Similar Content

• 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.

• 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 )
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:

• Hi, can someone please explain why this is giving an assertion EyePosition!=0 exception?

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.
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.

• 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