<div dir="ltr">I would definitely expect to have pointer types here. The SCEVs are collected by collectConstStrideAccesses and replaceSymbolicStrideSCEV. Maybe you've uncovered a bug there somewhere?</div><div class="gmail_extra"><br><div class="gmail_quote">On Fri, Feb 17, 2017 at 3:22 AM, Karl-Johan Karlsson via llvm-commits <span dir="ltr"><<a href="mailto:llvm-commits@lists.llvm.org" target="_blank">llvm-commits@lists.llvm.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">As I wrote earlier I had to revert the change due to a buildbot error in clang-with-thin-lto-ubuntu build #1847 step build-stage3-compiler. I have tried to recreate the buildbot fault, but failed to do so.<br>
<br>
<a href="http://lab.llvm.org:8011/builders/clang-with-thin-lto-ubuntu/builds/1847/steps/build-stage3-compiler" rel="noreferrer" target="_blank">http://lab.llvm.org:8011/build<wbr>ers/clang-with-thin-lto-ubuntu<wbr>/builds/1847/steps/build-<wbr>stage3-compiler</a><br>
<br>
When looking at the stacktrace from the buildbot failure it seems like<br>
either the type of DesA.Scev or DesB.Scev is not a pointer (in InterleavedAccessInfo::analyze<wbr>Interleaving()). Isn't that a bit strange?<br>
<br>
Stack trace:<br>
ld.lld: /home/buildbot/Buildbot/Slave1<wbr>a/clang-with-thin-lto-ubuntu/<wbr>llvm.src/include/llvm/Support/<wbr>Casting.h:236: typename cast_retty<X, Y *>::ret_type llvm::cast(Y *) [X = llvm::PointerType, Y = llvm::Type]: Assertion `isa<X>(Val) && "cast<Ty>() argument of incompatible type!"' failed.<br>
#0 0x0000000000489824 PrintStackTraceSignalHandler(v<wbr>oid*) (/home/buildbot/Buildbot/Slave<wbr>1a/clang-with-thin-lto-ubuntu/<wbr>install/stage2/bin/ld.lld+<wbr>0x489824)<br>
#1 0x0000000000489b76 SignalHandler(int) (/home/buildbot/Buildbot/Slave<wbr>1a/clang-with-thin-lto-ubuntu/<wbr>install/stage2/bin/ld.lld+<wbr>0x489b76)<br>
#2 0x00007fda0bb58330 __restore_rt (/lib/x86_64-linux-gnu/libpthr<wbr>ead.so.0+0x10330)<br>
#3 0x00007fda0ab8dc37 gsignal (/lib/x86_64-linux-gnu/libc.so<wbr>.6+0x36c37)<br>
#4 0x00007fda0ab91028 abort (/lib/x86_64-linux-gnu/libc.so<wbr>.6+0x3a028)<br>
#5 0x00007fda0ab86bf6 (/lib/x86_64-linux-gnu/libc.so<wbr>.6+0x2fbf6)<br>
#6 0x00007fda0ab86ca2 (/lib/x86_64-linux-gnu/libc.so<wbr>.6+0x2fca2)<br>
#7 0x00000000019c7d97 (anonymous namespace)::LoopVectorizationL<wbr>egality::canVectorize() (/home/buildbot/Buildbot/Slave<wbr>1a/clang-with-thin-lto-ubuntu/<wbr>install/stage2/bin/ld.lld+<wbr>0x19c7d97)<br>
#8 0x00000000019bc33c llvm::LoopVectorizePass::proce<wbr>ssLoop(llvm::Loop*) (/home/buildbot/Buildbot/Slave<wbr>1a/clang-with-thin-lto-ubuntu/<wbr>install/stage2/bin/ld.lld+<wbr>0x19bc33c)<br>
#9 0x00000000019ca153 llvm::LoopVectorizePass::runIm<wbr>pl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo&, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AAResults&, llvm::AssumptionCache&, std::function<llvm::LoopAccess<wbr>Info const& (llvm::Loop&)>&, llvm::OptimizationRemarkEmitte<wbr>r&) (/home/buildbot/Buildbot/Slave<wbr>1a/clang-with-thin-lto-ubuntu/<wbr>install/stage2/bin/ld.lld+<wbr>0x19ca153)<br>
<br>
Should it really be necessary to add pointer checks in the address space check?<br>
<br>
Regards<span class="HOEnZb"><font color="#888888"><br>
/ Karl-Johan Karlsson</font></span><div class="HOEnZb"><div class="h5"><br>
<br>
On 2017-02-14 14:26, Karl-Johan Karlsson via Phabricator wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Ka-Ka added a comment.<br>
<br>
I had to revert this change due to a buildbot failure. I'm trying to recreate the fault locally ...<br>
<br>
<br>
Repository:<br>
  rL LLVM<br>
<br>
<a href="https://reviews.llvm.org/D29717" rel="noreferrer" target="_blank">https://reviews.llvm.org/D2971<wbr>7</a><br>
<br>
<br>
<br>
<br>
</blockquote>
______________________________<wbr>_________________<br>
llvm-commits mailing list<br>
<a href="mailto:llvm-commits@lists.llvm.org" target="_blank">llvm-commits@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/llvm-commits</a><br>
</div></div></blockquote></div><br></div>