Feature or enhancement

Proposal:

When the tier 2 optimizer can prove that an operand to a float binary operation is uniquely referenced, we can mutate it in place instead of allocating a new PyFloatObject. This avoids a PyFloat_FromDouble allocation and the subsequent deallocation of the consumed operand on every iteration.

The unique reference tracking infrastructure added in #144300 can be used for the float operations as well. Also see #90530 where a similar optimization was proposed for tier 1 using runtime refcount checks.

Has this already been discussed elsewhere?

No response given

Links to previous discussion of this feature:

No response

Linked PRs

• gh-146306: Optimize float operations by mutating uniquely-referenced operands in place (JIT only) #146307