Improve performance of sum of products with nonlinear expressions #3470
Conversation
|
Looking into an example where it gives an improvement to implement |
| push!(x.args, y) | ||
| arg = *(args...) | ||
| if _MA.add_sub_op(op) != + | ||
| arg = _MA.add_sub_op(op)(arg) |
There was a problem hiding this comment.
I really think we should avoid this. If the user writes x + y - 2 * z, we should make -(+(x, y), *(2, z)), not `+(x, y, -(*(2, z))).
Julia has a parsing precedence order. Let's follow that where possible.
There was a problem hiding this comment.
I thought flatten! was doing that anyway
|
I also don't see how this fixes your issue. It just defines a separate method so that the other issue isn't hit. We'd still need the |
| x::GenericNonlinearExpr, | ||
| y::AbstractJuMPScalar, |
There was a problem hiding this comment.
I don't remember why I only wrote the binary case 😆
Closes #3469