Alias Analyses

Which alias analysis algorithms in opt are worth comparing against
(for terms of precision)? I know --anders-aa seems to give me pretty
good results. Online [1] it seemed to mention steensgaard's algorithm
and some kind of data structure analysis, but neither of these are
currently accepted by opt. Is there some way to get them, or are they
subsumed by the results of --anders-aa?

Thanks,
Ben Chambers

[1] http://llvm.org/docs/AliasAnalysis.html

Hi Ben,

Which alias analysis algorithms in opt are worth comparing against
(for terms of precision)? I know --anders-aa seems to give me pretty
good results. Online [1] it seemed to mention steensgaard's algorithm
and some kind of data structure analysis, but neither of these are
currently accepted by opt. Is there some way to get them, or are they
subsumed by the results of --anders-aa?

The Steensgaard and data structure analysis were removed from LLVM
several months ago due to patent issues. Seems the documentation doesn't
reflect that. Patches welcome. :slight_smile:

Reid.