Ivan Baev via llvm-dev
2015-Aug-12 00:46 UTC
[llvm-dev] RFC: PGO Late instrumentation for LLVM
It could be a bit confusing to have two ways for PGO instrumentation at the same time. If small functions like get/set are responsible for most of the current overhead, could we inline these in clang? What will be the replacement for Function entry counts under PGO late instrumentation: these are valuable for many inter-procedural profile-based optimizations? Ivan
Xinliang David Li via llvm-dev
2015-Aug-12 05:35 UTC
[llvm-dev] RFC: PGO Late instrumentation for LLVM
On Tue, Aug 11, 2015 at 5:46 PM, Ivan Baev via llvm-dev < llvm-dev at lists.llvm.org> wrote:> It could be a bit confusing to have two ways for PGO instrumentation at > the same time. If small functions like get/set are responsible for most of > the current overhead, could we inline these in clang? > > What will be the replacement for Function entry counts under PGO late > instrumentation: these are valuable for many inter-procedural > profile-based optimizations? > >Block counts (including entry count) can be fully reconstructed with late instrumentation. No information is lost. David> Ivan > > > > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org http://llvm.cs.uiuc.edu > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150811/73fc89ba/attachment.html>