b-sumner

Does it happen if you compile with an explicit `--offload-arch=gfxNNN --offload-arch=gfxMMM ...` ? I recommend always being explicit about what GPUs you want to target when not using the runtime...

As I understand it, rCUDA was developed by Technical University of Valencia. Have you contacted them with this request?

HIP does not currently support malloc/free or non-placement new/delete.

The necessary attribute was added to the device library when needed. See https://github.com/ROCm/llvm-project/blob/amd-staging/amd/device-libs/ockl/src/cg.cl#L87 . I think the issue here is that the device library is not consistent with the compiler....

@sjnewbury since the AMD fork's amd-staging branch is as close to upstream main as we can get, it's likely that tip device libraries amd-staging will be OK with tip LLVM....

Good grief! Thanks for pointing this out, and apologies for any headaches it caused. Will fix ASAP.

Fix will not be in ROCm 5.7.

Have you reported this issue to blackmagicdesign? To start fixing, the preproceeed OpenCL C program source is the minimum needed. Do you get anything if you set environment AMD_OCL_BUILD_OPTIONS_APPEND=-save-temps;C:\path\to\temps\prefix ?

Sorry for the delay. I wasn't expecting compiler concerns to be reported here. Can you provide the sources for the kernel and a standalone app which drives the kernel and...

I downloaded the link and installed everything needed to build. But the build doesn't work because kernel.sh is not found, but cmake expects it: CMakeLists.txt: COMMAND ${CMAKE_CURRENT_SOURCE_DIR}/kernel.sh