[llvm] [DebugInfo] Don't apply is_stmt on MBB branches that preserve lines (PR #108251)
Jeremy Morse via llvm-commits
llvm-commits at lists.llvm.org
Mon Nov 4 08:55:14 PST 2024
https://github.com/jmorse commented:
This is looking good, although I reckon all the new code should be factored into a new named method so that readers of `beginFunctionImpl` get a better overview of what's going on.
Is there a reason for examining successors of interesting blocks found, instead of examining the predecessor of all blocks? Somehow the latter feels more natural to me; I don't think there's a technical reason for doing it like that though.
I wonder whether there's a topological way of avoiding some of the work, i.e. if we explore in RPO then we might have already found a block that needs is_stmt and can skip analysing it again for other predecessors? That might be overcomplicating it if this doesn't fire very often though.
It pains me to say this but... we must still explicitly skip meta-instructions post-isel to avoid variable location info leaking into the linetables.
https://github.com/llvm/llvm-project/pull/108251
More information about the llvm-commits
mailing list