Ewan Crawford
Ewan Crawford
Tried this out on the `cl.xml` file from an internal docs MR, and it worked nicely at generating the `cl_ext.h` header. Thanks 😃 A minor nice-to-have would be a `requirements.txt`...
> > @EwanC asked me to elaborate my feedback from IWOCL, so here goes: > > Fundamentally, I am not sure whether we need such a large new API surface,...
> ```c++ > auto exec_handle = std::make_shared(); > > { > // The property could also be tied to/referenced as forward-progress progress behavior > // and also allow to define...
I tried the `cmdbuf-emu-test-queues` branch out with the SYCL-Graph tests we had which motivated this issue, and setting `g_cEnhancedErrorChecking` does indeed fix the issues. See * https://github.com/intel/llvm/blob/sycl/sycl/test-e2e/Graph/RecordReplay/work_group_size_prop.cpp * https://github.com/intel/llvm/blob/sycl/sycl/test-e2e/Graph/Explicit/work_group_size_prop.cpp *...
Want to confirm I understand this proposal because I agree having clear messaging to users on how to deal with API changes is a nice improvement. The goal is that...
That's a good point, a runtime could return a different signature from [clGetExtensionFunctionAddressForPlatform](https://registry.khronos.org/OpenCL/specs/3.0-unified/html/OpenCL_API.html#clGetExtensionFunctionAddressForPlatform) than the signature that the app was built with. We do have an existing runtime query [CL_DEVICE_EXTENSIONS_WITH_VERSION](https://registry.khronos.org/OpenCL/specs/3.0-unified/html/OpenCL_API.html#CL_DEVICE_EXTENSIONS_WITH_VERSION)...
Assigning myself to try draft this up, because I think it effectively blocks https://github.com/KhronosGroup/OpenCL-Docs/pull/1045 if want users to have a good experience of handling that API break.
> To help with the review, I tried to create a minimal diff between the current tip of master and the headers generated her. I did this using `clang-format` and...
Taking this out of draft now that the implementation has been merged in https://github.com/intel/llvm/pull/12840
> This looks good! Just a couple small comments below. @gmlueck Can I confirm you're good to merge this? The implementation of the API defined in this PR is merged...