[llvm-bugs] [Bug 32519] New: ScopDetection::isMaxRegionInScop updates DetectionContext, but does not need to invalidate it.

via llvm-bugs llvm-bugs at lists.llvm.org
Tue Apr 4 06:18:14 PDT 2017


http://bugs.llvm.org/show_bug.cgi?id=32519

            Bug ID: 32519
           Summary: ScopDetection::isMaxRegionInScop updates
                    DetectionContext, but does not need to invalidate it.
           Product: Polly
           Version: unspecified
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Other
          Assignee: polly-dev at googlegroups.com
          Reporter: llvm at meinersbur.de
                CC: llvm-bugs at lists.llvm.org

The method ScopDetection::isMaxRegionInScop() is supposed to find out whether a
region is still a valid SCoP after code generation of another. This is tested
by test\Isl\CodeGen\multiple-scops-in-a-row.ll.

As a side-effect of this check, it also updates the DetectionContext. After
this update multiple-scops-in-a-row.ll is again a valid SCoP. Logically,
code-generation of another SCoP should never invalidate another. However, it
remains an invalid SCoP in the following LLVM test-suite programs:

External/Povray/povray
External/SPEC/CINT2006/400.perlbench/400.perlbench
MultiSource/Applications/lemon/lemon
MultiSource/Benchmarks/MallocBench/cfrac/cfrac
MultiSource/Benchmarks/Prolangs-C/TimberWolfMC/timberwolfmc
MultiSource/Benchmarks/mafft/pairlocalalign
SingleSource/Benchmarks/Polybench/linear-algebra/solvers/gramschmidt/gramschmidt

This update of DetectionContext should be explicit, not in isMaxRegionInScop()
which should be side-effect-free.

Ideally, we identify the reason why some SCoPs are not valid SCoPs after code
generation anymore and fix this.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20170404/f1f5f0cd/attachment.html>


More information about the llvm-bugs mailing list