[llvm] [GVN] Permit load PRE to happen in more cases (PR #79324)
David Sherwood via llvm-commits
llvm-commits at lists.llvm.org
Wed Feb 21 07:10:30 PST 2024
david-arm wrote:
Hi @nikic and @paulwalker-arm, unlike Phabricator I can't "Plan Changes" or "Request Changes" on my own PR, so I can only leave a comment here.
Downstream I have a patch that significantly restricts the maximum number of instructions we are willing to walk in total to just the amount that triggers the optimisation in omnetpp. However, it seems that this still causes significant compile-time regressions and probably the cost of doing this analysis outweighs the benefits. I'll put this PR on hold for now.
https://github.com/llvm/llvm-project/pull/79324
More information about the llvm-commits
mailing list