[llvm-dev] [RFC] Switching to MemorySSA-backed Dead Store Elimination (aka cross-bb DSE)
Florian Hahn via llvm-dev
llvm-dev at lists.llvm.org
Tue Aug 18 14:14:38 PDT 2020
> On Aug 18, 2020, at 16:59, Michael Kruse <llvmdev at meinersbur.de> wrote:
> Thanks for all the work. The reductions in stores look promising. Do you also have performance numbers how much this improves the execution time? Did you observe any regressions where MSSA resulted in fewer removed stores?
I did not gather numbers for execution time yet, but I’ll try to share some tomorrow.
At the current state, for MultiSource/SPEC2000/SPEC2006, there are the following regressions
test-suite...6/471.omnetpp/471.omnetpp.test 321.00 316.00 -1.6%
test-suite...arks/mafft/pairlocalalign.test 64.00 61.00 -4.7%
test-suite...ks/Prolangs-C++/city/city.test 23.00 21.00 -8.7%
test-suite...oxyApps-C/miniGMG/miniGMG.test 69.00 60.00 -13.0%
I suspect those are caused by the few cases that the MemorySSA version does not yet support. Those will need some more investigating, but I think ideally we would not block the switch on them, so we can switch early and address the issues that pop up early.
More information about the llvm-dev