Ben Barsdell

Thanks Miles, fantastic review! Lots of excellent points. Re the test failures, what CUDA version and GPU model is this running with?

Sorry it took so long to get to this. I've added nvrtc_cli to `make all`.

Maybe I misunderstood, I'm having trouble thinking this through. Some of them need to be compile time checks, and those won't gain anything by adding runtime checks. Will other parts...

Sorry I missed this. You don't have to use jitify2_preprocess, other workflows are possible. If you are starting from ptx/cubin, you can create a `CompiledProgramData` object directly from ptx or...

Apparently this is the replacement: ``` #include using pinned_allocator = thrust::mr::stateless_resource_allocator< T, thrust::system::cuda::universal_host_pinned_memory_resource>; ```

I like this idea, but I think you're right that we need to be careful to avoid patching cuda/std headers or double-patching other headers. ``` bool is_cuda_std_header = detail::get_workaround_system_headers().count(header_name); ```...

I think this should work, as long as the `-I` option has the correct path (e.g., "/usr/local/cuda/include"). If it's still not working for you, could you provide a full reproducer...

@Robadob This is actually not in response to that issue; I only just became aware of it and do not know the root cause (possibly some header hierarchy changed and...

@Robadob Thanks for the help with this. I've pushed fixes for the issues you identified as well as some more that I found with manylinux2014.

I think the `stdlib.h` error is because you're using `--pre-include` to include cuda.h. I probably need to add logic to handle pre-includes. I'm guessing the `dynamic/curve_rtc_dynamic.h` error is because a...