Instead of manually creating output IDs and directly constructing Split and Merge nodes, IterDomain::split and IterDomain::merge are used. This would reduce duplication of logic.
There's some slight change in iteration types of IDs generated when TensorView::rFactor is used. See the inline comment.
Instead of manually creating output IDs and directly constructing Split and Merge nodes, IterDomain::split and IterDomain::merge are used. This would reduce duplication of logic.
There's some slight change in iteration types of IDs generated when
TensorView::rFactor
is used. See the inline comment.As of 8321cfd, everything seems fine.
Closes #2520