[llvm-commits] [llvm] r116958 - /llvm/trunk/lib/Analysis/TypeBasedAliasAnalysis.cpp

[Dan Gohman]

On Oct 21, 2010, at 12:57 AM, Duncan Sands wrote:

> Hi Dan,
> 
>> +/// Aliases - Test whether the type represented by A may alias the
>> +/// type represented by B.
> 
> if I understand the way you've set up TBAA right, it doesn't have to have
> anything to do with types, right?  It seems to be a completely abstract
> setup where each memory operation can be annotated with a node in an
> abstract tree, and where two memory operations will not alias if neither
> of their nodes is an ancestor of the other.  Is that correct?  If so, it
> could in theory be used for all kinds of other things and not just TBAA.

Yes. This is a design goal. I used the word "type" because that's one
way of thinking about it, but it's not tied to llvm::Type or clang::Type
or any of the other existing type systems in the compiler.

Dan