Gary Miguel

This is fixed in my fork: https://github.com/garymm/plex-for-kodi/commit/787b361bf58ff95e60e6625734a74776709017b5 Anyone should feel free to open a PR to fix it here too.

Got merged into master, but not released: https://github.com/plexinc/plex-for-kodi/commit/ff3b4f233dd456beb3a440a234f7eec6a65ce976

Thanks for the quick reply! > Please try with int instead of float Tried uint32_t and int, same results. > Verify that the GPU can execute kernels of that size...

Oh also this program works correctly with the DPC++ compiler on the same machine and GPU, so I do suspect a bug in either OpenSYCL or nvc++.

`--opensycl-targets=cuda:sm_86` works correctly.

Uploaded the PTX. [vector_add_sycl.n001.ptx.txt](https://github.com/OpenSYCL/OpenSYCL/files/11537997/vector_add_sycl.n001.ptx.txt) Produced by modifying the command printed out by `--opensycle-dryrun`, changing: `-gpu=cc86` to `-gpu=cc86,keep`.

Is there a way to compare it to the PTX that's produced by the `cuda` target?

Tried all of these to no avail: * Older nvc++ 22.11 (same bug) and 20.11 (doesn't compile) * `-gpu=cc70` * -O0 * -std=c++20

Did you mean to CC someone else? I have no particular expertise or interest in this bug.

I'm seeing this again. I tried upgradng to 1.64.2 (latest helm chart) and deleting all the pods and this time I can't figure out a way to fix it.