Currently, via the GPU Arch entry, only 1 cc is supported. We designed the entry to be a list of string, so that we actually want to support multiple compute capability generation. The way how this could manifest is pending discussion. Especially:
Should the result contains 1 binding file per CC?
Is there a way to compress the generation in to a binding file?
Currently, via the
GPU Archentry, only 1 cc is supported. We designed the entry to be a list of string, so that we actually want to support multiple compute capability generation. The way how this could manifest is pending discussion. Especially:Should the result contains 1 binding file per CC?
Is there a way to compress the generation in to a binding file?