Closed toelli-msft closed 2 years ago
Yes, makes sense. aten has a huge number of signatures for each function, with different optional arguments. We supply a separate implementation for each concrete signature. Of course we can commmon up our implementations, but when we do this for real, we will just autogenerate like in https://github.com/nunoplopes/torchy/blob/a8aab1ae5a04c0c1bcbb96b0b02dc6470fb19500/autogen/dispatch_wrappers.h#L7236
in a place where the two have the same effect but sum is likely to be faster. There is a lot of noise in the benchmarks so it's not completely obvious it is faster.
Fixes https://github.com/microsoft/knossos-ksc/issues/967
I don't understand why a version of
aten::sum
with an extra(Tuple)
argument is required.