Nvidia SDK OpenCL CPU support NVIDIA GPU Computing SDK support for compile OpenCL kernels for CPUs

Can I compile OpenCL kernels for x86(_64) with Nvidia GPU Computing SDK?
I have to compare CPU and GPU execution (better if the code is the same) for my thesis. At now I can’t try it because I haven’t an NVIDIA OpenCL capable card at home but I have to use remotely university computers with Tesla and NVIDIA OpenCL implementation but they are not yet confgured for OpenCL remote execution (I opened another topic about that). If I could know now if CPU compilation and execution are supported I would know if I can use that implementation for CPU execution or I must ask to the University the installation of Intel or AMD/ATI implementations for that.

Can I compile OpenCL kernels for x86(_64) with Nvidia GPU Computing SDK?
I have to compare CPU and GPU execution (better if the code is the same) for my thesis. At now I can’t try it because I haven’t an NVIDIA OpenCL capable card at home but I have to use remotely university computers with Tesla and NVIDIA OpenCL implementation but they are not yet confgured for OpenCL remote execution (I opened another topic about that). If I could know now if CPU compilation and execution are supported I would know if I can use that implementation for CPU execution or I must ask to the University the installation of Intel or AMD/ATI implementations for that.

You can have multiple OpenCL SDK’s installed. (Read more about ICD.) So you can easily test on NVidia GPUs and also on CPU thanks to AMD implementation (as long as you won’t use textures). Good luck.

You can have multiple OpenCL SDK’s installed. (Read more about ICD.) So you can easily test on NVidia GPUs and also on CPU thanks to AMD implementation (as long as you won’t use textures). Good luck.

I know I can have multiple implementations of OpenCL with ICD but do you confirm NVIDIA doesn’t support CPU and I MUST install other implementations to use OpenCL on CPU?

I know I can have multiple implementations of OpenCL with ICD but do you confirm NVIDIA doesn’t support CPU and I MUST install other implementations to use OpenCL on CPU?

Yes, that’s right. Nvidia doesn’t support the CPU implementation. As far as I know, only AMD does.

Yes, that’s right. Nvidia doesn’t support the CPU implementation. As far as I know, only AMD does.

Intel just released an alpha SDK for their newer CPUs.

I tried the AMD SDK and couldn’t get it to work in conjunction with my NV SDK.

Intel just released an alpha SDK for their newer CPUs.

I tried the AMD SDK and couldn’t get it to work in conjunction with my NV SDK.

BTW: WRT to the AMD SDK. I had to change my kernel to use __constant sampler instead of const sampler and it will compile with the AMD CPU compiler. But it won’t link. There is no indication as to why it won’t link though.

Companies have had OCL implementations for over a year now - it should be easier than this.

BTW: WRT to the AMD SDK. I had to change my kernel to use __constant sampler instead of const sampler and it will compile with the AMD CPU compiler. But it won’t link. There is no indication as to why it won’t link though.

Companies have had OCL implementations for over a year now - it should be easier than this.

I already knew Intel implmeentation, in the first post I cited also Intel for CPU implementation. If you have multiple implementation you have to choose the implementation you want to use in a way similar to this

http://developer.amd.com/support/KnowledgeBase/Lists/KnowledgeBase/DispForm.aspx?ID=71

I already knew Intel implmeentation, in the first post I cited also Intel for CPU implementation. If you have multiple implementation you have to choose the implementation you want to use in a way similar to this

http://developer.amd.com/support/KnowledgeBase/Lists/KnowledgeBase/DispForm.aspx?ID=71

With ATI Stream SDK 2.2 and Catalyst driver 10.10 (8.783…) I used in the parameters of my kernel function “const” keyword and the program compile and run either on GPU (ATI) or CPU (tried on AMD and also on Intel CPU).

With ATI Stream SDK 2.2 and Catalyst driver 10.10 (8.783…) I used in the parameters of my kernel function “const” keyword and the program compile and run either on GPU (ATI) or CPU (tried on AMD and also on Intel CPU).