[PATCH] D32666: [PartialInlining] Add support in partial inliner to use cost analysis in inliner
Xinliang David Li via llvm-commits
llvm-commits at lists.llvm.org
Fri Apr 28 17:26:10 PDT 2017
On Fri, Apr 28, 2017 at 5:07 PM, Davide Italiano via Phabricator <
reviews at reviews.llvm.org> wrote:
> davide added inline comments.
>
>
> ================
> Comment at: lib/Transforms/IPO/PartialInlining.cpp:111
> + AU.addRequired<ProfileSummaryInfoWrapperPass>();
> + AU.addRequired<TargetTransformInfoWrapperPass>();
> }
> ----------------
> Why don't we preserve TTI? (not that it matters for performances, but)
>
>
> ================
> Comment at: lib/Transforms/IPO/PartialInlining.cpp:302-306
> + if (IC.isAlways()) {
> + ORE.emit(OptimizationRemarkAnalysis(DEBUG_TYPE, "AlwaysInline", Call)
> + << NV("Callee", Callee)
> + << " should always be fully inlined, not partially");
> + return false;
> ----------------
> Can you add a test that triggers this?
>
>
> ================
> Comment at: lib/Transforms/IPO/PartialInlining.cpp:309-323
> + if (IC.isNever()) {
> + ORE.emit(OptimizationRemarkMissed(DEBUG_TYPE, "NeverInline", Call)
> + << NV("Callee", Callee) << " not partially inlined into "
> + << NV("Caller", Caller)
> + << " because it should never be inlined (cost=never)");
> + return false;
> + }
> ----------------
> and these as well, maybe
>
>
>
There is a test case for noinline case.
David
> https://reviews.llvm.org/D32666
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170428/6b6b3e94/attachment.html>
More information about the llvm-commits
mailing list