Scratch is only using when reordering a tensor. Investigate fast allocation/deallocation of scratch by each tensor whenever needed. This will eliminate the awkwardness of passing scratch around and make it easier to write wrappers for other tensor libraries and use them with qFlex.
Scratch is only using when reordering a tensor. Investigate fast allocation/deallocation of scratch by each tensor whenever needed. This will eliminate the awkwardness of passing scratch around and make it easier to write wrappers for other tensor libraries and use them with qFlex.