Support matching `max` and `min` reductions
Opened this issue · 0 comments
kaushikcfd commented
With the additional assumption that all the operands are non-negative, the transformations (both algebraic and loop) would be sound (I think) to translate from their Einstein summation equivalents.