[PATCH] D12603: Use fixed-point representation for BranchProbability

[Xinliang David Li via llvm-commits]

UINT32_MAX vs 1<<31 choice is still under discussion, so please be patient
:)

David


On Fri, Sep 11, 2015 at 5:20 PM, Cong Hou via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

> congh updated this revision to Diff 34607.
> congh added a comment.
>
> Update the patch by using 1<<31 instead of UINT32_MAX as the constant
> denominator.
>
>
> http://reviews.llvm.org/D12603
>
> Files:
>   include/llvm/Analysis/BlockFrequencyInfoImpl.h
>   include/llvm/Support/BranchProbability.h
>   lib/Support/BranchProbability.cpp
>   lib/Target/ARM/ARMBaseInstrInfo.cpp
>   test/Analysis/BlockFrequencyInfo/basic.ll
>   test/Analysis/BlockFrequencyInfo/loops_with_profile_info.ll
>   test/Analysis/BranchProbabilityInfo/basic.ll
>   test/Analysis/BranchProbabilityInfo/loop.ll
>   test/Analysis/BranchProbabilityInfo/noreturn.ll
>   test/Analysis/BranchProbabilityInfo/pr18705.ll
>   test/Analysis/BranchProbabilityInfo/pr22718.ll
>   test/CodeGen/AArch64/fast-isel-branch-cond-split.ll
>   test/CodeGen/ARM/2013-10-11-select-stalls.ll
>   test/CodeGen/ARM/ifcvt4.ll
>   test/CodeGen/ARM/sjlj-prepare-critical-edge.ll
>   test/CodeGen/ARM/tail-merge-branch-weight.ll
>   test/CodeGen/ARM/test-sharedidx.ll
>   test/CodeGen/Thumb2/thumb2-ifcvt1.ll
>   test/Transforms/SampleProfile/branch.ll
>   test/Transforms/SampleProfile/calls.ll
>   test/Transforms/SampleProfile/discriminator.ll
>   test/Transforms/SampleProfile/fnptr.ll
>   test/Transforms/SampleProfile/propagate.ll
>   unittests/Support/BlockFrequencyTest.cpp
>   unittests/Support/BranchProbabilityTest.cpp
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150911/562efbc6/attachment.html>