On 22/12/2014 16:44, Sebastian Schaetz wrote:
Kyle Lutz
writes: Yeah, these are all interesting ideas. While I'm a bit hesitant to bake this behavior into the low-level command_queue interface, I'd be more than happy to support this type of API at a higher level.
I have the exact same problem, I'm hesitant to include these thoughts into Aura because I haven't found the best way to do it. I guess to find the correct API, a few variants have to be implemented and we have to see what makes most sense.
I've also been meaning to look at Aura again more closely. Last time was many months ago and it looks like you've done quite a lot of work since then.
You are of course free to look but I have to disappoint your: these thoughts are up until now just thoughts. I'll let you know once I have something.
I think Aura and Boost.Compute have mertis as low level, C++ styled proper wrapper on top of accelerator technology. It's important for them to be flexible and not too fancy so multiple fancier tool can reuse them.