Ben Ashbaugh

I found an odd case where enabling the int64 atomic extensions is required to use the 64-bit C11 atomic functions while investigating a separate issue: https://godbolt.org/z/Kb3Pxf My main reason for...

I don't want to feature-creep this PR, but I think the observations in this comment are good and we should fix them, perhaps by refactoring the intro section in the...

> Btw I have covered this in my change to the extension spec: > Or do you think we need to add something more? I think we're a lot closer...

Hello, could you please take another look at my comments here https://github.com/KhronosGroup/OpenCL-Docs/pull/355#pullrequestreview-541524654 and ensure they're addressed? I'm especially concerned about: > Extension pragma directives are optional and they are only...

> I am trying to understand what is blocking this PR or general progress on #82 at the moment? What would be helpful if someone provides any feedback of what...

Since this PR came up again in today's teleconference: If we want to get this over the finish line (and it still looks like most of the changes are really...

I took an action item in today's Khronos call to propose spec text to clarify this behavior. Here's what I have so far: > === Attribute Qualifiers > > The...

> So we should be able to always split large work-item into whatever chunk size who are is available by the backend (assuming the local-group size specified fit ofc) ....

> To be honest, I don't know... I guess my hand-wavy answer is "as much as they are used running on NVIDIA". More than 32-bit, this is for sure. And...

Hi, good question! It looks like our implementation will return `CL_INVALID_VALUE` in this case but we will not update `binary_status[i]`, though reading through the spec it's not clear to me...