Using AMD SDK with Nvidia driver and Nvidia samples with AMD driver..

Includes some fixes to run some samples and insight about failing example

Nvidia samples AMD drivers

[url=“GPU computing: Nvidia OpenCL samples with AMD OpenCL drivers!”]http://oscarbg.blogspot.com/2009/11/nvidia...amd-opencl.html[/url]

Nvidia samples Nvidia 195 driver

http://oscarbg.blogspot.com/2009/11/amd-op…-opencl_05.html

AMD samples Nvidia 195 driver

[url=“GPU computing: AMD OpenCL samples on Nvidia 195 OpenCL drivers!!”]http://oscarbg.blogspot.com/2009/11/amd-op...195-opencl.html[/url]

AMD and Nvidia SDK in VS 2010

[url=“GPU computing: Optix and OpenCL SDKs with Visual Studio 2010”]http://oscarbg.blogspot.com/2009/11/optix-...ual-studio.html[/url]

I don’t have time to test those but it’s a cool thing you’re doing, testing portability across vendors. Good work and keep us informed!

By the way, could you post a short conclusion, keypoints to your findings? Are samples generally portable, save for minor issues or is there still much work to be done?

Yes, thanks for posting your findings. With a new standard like OpenCL there are bound to be a few bugs and differences in interpretation of the spec, so it’s good to see someone testing compatibility like this. Our goal is to have the most compliant and fully featured OpenCL implementation on the market.

R.e. the issues you bring up here:
[url=“GPU computing: AMD OpenCL samples on Nvidia 195 OpenCL drivers!!”]http://oscarbg.blogspot.com/2009/11/amd-op...195-opencl.html[/url]

  1. Kernels without parameters e.g. __kernel void foo() - are legal OpenCL, but our compiler doesn’t currently allow this. We’ll try and fix this.

  2. Explicit casts between different vector types (e.g. uint4 and float4) are not allowed according to OpenCL spec and we throw an error. However AMD and Apple seem to allow this. We are trying to get a clarification from OpenCL Working Group on this issue.