Nol Moonen

- Improve the configuration generation script, it now supports all benchmarks that have config autotuning. - Added support for large indices in `device_reduce_by_key`, `device_partition`, and `device_unique`. - Improved block sort...

Fix bug in device partition for large problem sizes

2

Previously, a check was performed for the last block of each launch which should have been performed only at the last block across all launches. This gave unintended effects, so...

Compile error in hip_cooperative_groups.h with ROCm 5.5

1

With ROCm 5.5, the inclusion of `hip/hip_cooperative_groups.h` results in a compile error on debug builds: ```c++ In file included from /opt/rocm-5.5.0/include/hip/hip_cooperative_groups.h:38: /opt/rocm-5.5.0/include/hip/amd_detail/amd_hip_cooperative_groups.h:455:7: error: invalid instruction, did you mean: s_trap? __hip_assert(false...

ROCm 5.3 gfx1030 hang with hipStreamCreate and hipStreamDestroy

3

The following test hangs with ROCm 5.3 on the `gfx1030` architecture (AMD Radeon PRO V620). ```c++ #include #include int main() { printf("starting..\n"); hipStream_t stream; hipStreamCreate(&stream); hipStreamDestroy(stream); hipStream_t stream2; hipStreamCreateWithFlags(&stream2, hipStreamNonBlocking);...