We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.
hi,
As i anderstoud OpenGL and OpenCL use différent driver to process their command on GPU and I wonder how the GPU procees those command. Is there a pin for OpenCL command and a pin for OpenGL command ?
So openCL will use a spécific part of the GPU and OpenGL another part of the GPU. I ask this because OpencL work with matrice and openGL work with vector except is i use compute shader and image.
But using OpenGL it is hard to retreive data to CPU. So i conclude that their are two différent part in the GPU.
Could someone explain how it work on the porocessor.
thanks.
hterrolle said:using OpenGL it is hard to retreive data to CPU
Why? You can glMapBuffer, just the same as you can with a CL buffer.
Yes you are rigth , i already use it to transfert texture to java after readpixel using PBO. few years ago ;))
But the main question was about GPU architecture. Is there a part of the GPU used OpenCL and a part used for OpenGL ?
Are booth of them using the same memorie ?
Sorry for the new post. But i did some test using OpenCL and OpenGL at the same time.
I find out that it take 50% even 100% more time to procees image.
So, the rigth question would be does OpenCL and OpenGL run in parrallel on GPU or do their share something that can produce a bootle neck at some momment.
If you like i can send you a debug result.
By the way after some amount of image processing between 60 and 100. The procees time can double enven more.
May be you know why so i could improve my code.
There aren't dedicated parts of the GPU for OpenCL vs OpenGL - they can both use the shader cores / memory etc, and the drivers will use all the shader cores they need, so if the processing is large enough that it can use all the cores, then the OpenCL & OpenGL workloads will most likely go one after the other.
thanks,
But the diuble time processing appends also when using only OpenCL. For exemple using OpenCL alone after 60/100 frames i can go from 63ms per frame to 100ms. Using OpenCL and OpenGL it is from 63ms to up to 200ms.
In this case it isnt a problem of core used. So, it loock like this could be comming fron some where else. It is just my analyze. That is why i asked thequesion.
By the way when GPU start to slow donw the CPU pocesing is also slowing. It look more like mémorie acces problem or heat problem after 60/100 frames.
i did some test today and i foud out that the problem come from the masive use of multicore threading and big data processing. But no idéa why this slow donw the GPU.