There are several places in the code that are currently broken as
they assume an Instruction always has a parent Function when
attempting to get the vscale_range attribute. This patch adds checks
that an Instruction has a parent.
Details
Details
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Comment Actions
I don't think all these cases are currently broken, because InstCombine traverses nodes that are in the IR. I'm not against having an extra guard for it though.
Is it worth adding a test for the ValueTracking case to avoid regression?
Comment Actions
Hi @sdesmalen, yeah I agree it would be good to have a test for the ValueTracking.cpp fix, but I wasn't sure how to create a simple, reliable test case where I is guaranteed not to have a parent?