The SDK doesnt come with emulation environment and has dependencies on the hardware, the graphics driver (if you want to run on GPU) etc. So you cannot install it on non-Intel platforms. For example, when targetting CPU devices, how will the compiler know what vector instructions to generate without knowing what the target CPU device the app will be running on? We can argue similarly in case of GPU device also.
I am curious - does Nvidia's SDK allow you to build your app even though you do not have their driver or graphics card installed?
Thanks,
Raghu




Why not possible to do cross-development ?
I build OpenCL apps for different hardware (ATi, nVidia) on single AMD APU-based netbook. nVidia SDK had no problems to install and apps built work OK on hosts with nVidia GPU installed. There is no need to have actual hardware when one linking app with corresponding OpenCL library (nVidia or AMD in my case).
But with Intel's OpenCL SDk situation is very different. It just refused to install on my development machine at all complaining about not compatible hardware. But I don't want to run app there, I want to build app there cause it's my dev host, it has VS and other tools installed.
So, why Intel refused to allow cross-compilation ?