I'm doing some ComputeCpp Vs DPC++ comparisons on Windows. ComputeCpp is fine to me so far, modulo being forced to use the Professional edition to encode multiple spir/PTX targets.
I heard this limitation is not present using DPC++ but, from this post:
it seems one year ago nVidia was usable with DPC++ only under Linux, and they were not keen to add any support under Windows to that.
From your experience, is that still the case?
Thanks for your support.
You can not use DPC++ on Windows with the Nvidia GPU support. This is indicated in the CUDA back-end limitations section of Getting Started with oneAPI DPC++ instruction.