Julian Oppermann

Moving the discussion on #12393 to here: We (the kernel fusion team) could re-use a `std::string` replacement class in our [effort ](https://github.com/intel/llvm/pulls?q=is%3Apr+%5BNoSTL%5D) to remove STL classes from the interface between...

You can run ``` clang -fclangir-enable -S -emit-llvm --target=spirv64 some-opencl-kernel.cl ``` on OpenCL C kernel code and look at the failing assertions. We've collected a bit more context in our...

`sycl/test-e2e/KernelFusion/internalize_array_wrapper.cpp` is addressed in #10882.

Two high-level questions: - Is this safe to use this together with MLIR's multithreading? - Could there be a situation in which a single compiler instance would need to handle...

@seven-mile and I discussed offline whether the check can be made earlier in the compilation. The issue is that we need to query the `CIRAllocaAddressSpace`, the `DataLayout` and the `TargetLoweringInfo`...

Thanks for reporting @leothaud, I think your analysis is correct, but on the other hand the limitatation is documented for the `ChainingCyclicProblem` in the design document: https://github.com/llvm/circt/blob/aa0a07dc05fbba6da311718c0d4b8efed4912535/docs/Scheduling.md?plain=1#L276-L283 NB: The class...

I'll look into the `KernelCompiler/sycl.cpp` failure.

Thanks for the PR! At a first glance, I'm a bit lukewarm on making the simplex solver part of the API. Can you elaborate a bit on your use-case and...

> I considered modifying the Problem directly rather than modifying the SSP module and recreating a new Problem for each schedule. Yes, modifying the `Problem` is certainly the way to...

> > Do you need additional APIs there, e.g. to reset the solution properties or so? > > Scheduling an already scheduled problem overwrites the previous solution data, such as...