Kyle Lutz

Could you let me know a little more about your use-case? Could you instead use `float4_`? Note that, due to padding, `sizeof(cl_float3) == sizeof(cl_float4)` and there isn't any difference in...

Thanks for the report and the links. I'll try to find some time to take a look and update the code.

Interesting. We should be able to integrate their algorithm (though perhaps as another engine, `mtgp_engine`).

Thanks for reporting this, I'll look into it. In the meantime, you can see the performance page with all the graphs here: http://boostorg.github.io/compute/boost_compute/performance.html

Hey @JablonskiMateusz! Could you provide some detail on the motivation for this change? Was this causing issues with your OpenCL library? For context, `device` is a very thin wrapper on...

Yeah, we should be able to provide ` partial_sort()` implementation. Are you interested in submitting a pull request for this?

Yeah, I'm hesitant to add more benchmark applications directly to Boost.Compute. I more like the `valhalla` approach to providing a single, separate benchmark suite which compares a variety of GPU...

Wow, thanks for making these fixes and doing all this work! It'll take me a few days to review and plan how to integrate these changes. Thanks!

So I can reproduce the segfault at size `10000000` and "Invalid Value" exception at size `5000000`. Here's the backtrace from the "Invalid Value" exception: ``` * thread #1: tid =...

The idea was to provide certain functions/classes (like the OpenGL interop APIs) but throw an `unsupported_extension_error` exception if the underyling compute device didn't support the extension (which is only detectable...