[llvm-dev] Fwd: Assertion failing on LLVM intrinsics instruction

Cristianno Martins via llvm-dev llvm-dev at lists.llvm.org
Mon May 30 13:28:38 PDT 2016


Hi there,

I was trying to see what kind of informations I can get on dependence/alias
analysis from the trunk llvm, and stumbled upon two potencial problems.
First of all, the most recent documentation I could find online
<http://llvm.org/docs/AliasAnalysis.html> suggests the use of options not
available on the code (like enabling globalsmodref-aa, or count-aa, for
example). Is there a more recent one, or is just the documentation not
being updated?

Second, after trying "opt --help" and scanning for everything that remotely
resembled alias or dependency analysis, I ended up trying to run opt over
part of the code of bzip2 from SPEC 2006, but unfortunately an assertion
fail prevents it from finishing the analysis:

Assertion failed: (dep.isUnknown() && "unexpected dependence type"),
function getInstTypePair, file
/Developer/llvm/lib/Analysis/MemDepPrinter.cpp, line 70.
Stack dump:
0. Program arguments: /Developer/llvm/build/bin/opt -aa -aa-eval -basicaa
-cfl-aa -external-aa -globals-aa -objc-arc-aa -print-alias-sets -scev-aa
-scoped-noalias -tbaa -da -memdep -print-memdeps
build_base_yosemite.0000/blocksort.bc -debug-pass=Structure -disable-output
-S
1. Running pass 'Function Pass Manager' on module
'build_base_yosemite.0000/blocksort.bc'.
2. Running pass 'Print MemDeps of function' on function '@BZ2_blockSort'
Illegal instruction: 4


I was kinda curious about which instruction was causing this assertion to
be thrown, and I added a print to the MemDepPrinter class to see that the
instruction was actually an llvm intrinsic:

Instruction:   call void @llvm.memset.p0i8.i32(i8* %bigDone152.i, i8 0, i32
256, i32 1, i1 false) #6


So, my question is: is this the expected behavior for MemDepPrinter? Is
there anything I can do to avoid failing this assertion?

Thanks in advance,

--
Cristianno Martins
<cristiannomartins at hotmail.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160530/25f598fd/attachment.html>


More information about the llvm-dev mailing list