His other benchmarks of OmniSciDB that ran on systems with Nvidia GPUs, but I think he's pointing out that the in this case, the AMD GPU wasnt used by the DB engine even if it was part of the system config.
Just to clarify, most of the query engine is built around LLVM-based JIT compilation, and CUDA is not really used per say except for GPU-specific operators like atomic aggregates and thread synchronization, and of course we use the driver API to manage the GPUs, allocate memory, etc. Supporting AMD GPUs or the upcoming Intel Xe GPUs (or frankly anything that has an LLVM-backend) would not be particularly hard, it would just require adding similar supporting infra.
Comments
His other benchmarks of OmniSciDB that ran on systems with Nvidia GPUs, but I think he's pointing out that the in this case, the AMD GPU wasnt used by the DB engine even if it was part of the system config.
Yeah, I just checked and it appears to be a CUDA project, so neither the Intel nor AMD GPUs help here.
Still, that could have been clarified in the article.
Just to clarify, most of the query engine is built around LLVM-based JIT compilation, and CUDA is not really used per say except for GPU-specific operators like atomic aggregates and thread synchronization, and of course we use the driver API to manage the GPUs, allocate memory, etc. Supporting AMD GPUs or the upcoming Intel Xe GPUs (or frankly anything that has an LLVM-backend) would not be particularly hard, it would just require adding similar supporting infra.