Is the OpenCL platform eventually going to be used to support multiple hosts? Because I had the platform error in my C++ program I wanted to farther research it and see just exactly what it is. The specification was pretty vague as to what it is being used for. Right now you will never really have more then one platform, is that correct? But I was thinking that it would be really cool to be able to pull in multiple hosts like in cluster computing and have access to all of the node’s resources as well. Is this the goal?
You’ll have all the ICD compatible installed platforms on your computer.
For example I installed, under linux, nVidia and AMD platforms. I have a GPU device with the nVidia platform and an intel CPU device with AMD’s.
You can get the platform list and the device list for each platform with OpenCL API.
Ah, ok. So the platform is just a list of the OpenCL compatible device drivers that are installed on the system? Not as powerful as I was envisioning, but it’s all good. Thanks,
I guess the platform concept is a enough generic to be really powerful :
A platform provide access to all devices it can get… so we can imagine a platform that try to query the network to find devices and, so, permits the user to enqueue command over the network.
Often a platform is installed by a driver, but we can imagine a driver which install multiple platforms.
Excellent! Thanks Matrem, that makes sense. Too bad my programming skills weren’t better, although I’m sure someone will come up with something cool eventually.