It's really annoying to work with Arith and have it always return a Val* when we frequently use TensorView* as args. We should overload these operators to return a TensorView* when there is one in the inputs.
We should then modify our tests which frequently have syntax like:
TensorView* tv2 = static_cast<TensorView*>(add(tv1, new Float(3.0)));
to remove the static_casts.
It's really annoying to work with Arith and have it always return a
Val*when we frequently useTensorView*as args. We should overload these operators to return aTensorView*when there is one in the inputs.We should then modify our tests which frequently have syntax like:
to remove the static_casts.