[LLVMbugs] [Bug 86] [basicaa] Basic AA misses a large number of simple testcases

[bugzilla-daemon at zion.cs.uiuc.edu]

http://llvm.cs.uiuc.edu/bugs/show_bug.cgi?id=86

sabre at nondot.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
           Platform|PC                          |All
         Resolution|                            |FIXED
   Target Milestone|---                         |1.1



------- Additional Comments From sabre at nondot.org  2003-12-11 17:26 -------
These two patches implement this PR:
http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20031208/010094.html
http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20031208/010097.html

With these patches, we correctly/perfectly resolve all aliases from
BasicAA/2003-11-04-SimpleCases.ll as well as 2003-12-11-ConstExprGEP.ll.

An example from crafty that improves a lot is the MakeMove function (there may
be others, this is just one I was debugging other bugs in recently.  It mainly
references globals and arrays):

Before:

$ extract -func MakeMove 186.crafty.llvm.bc | opt -aa-eval -disable-output
===== Alias Analysis Evaluator Report =====
  30135 Total Alias Queries Performed
  23879 no alias responses (79%)
  6041 may alias responses (20%)
  215 must alias responses (0%)
  Alias Analysis Evaluator Summary: 79%/20%/0%

Now:

$ extract -func MakeMove 186.crafty.llvm.bc | opt -aa-eval -disable-output
===== Alias Analysis Evaluator Report =====
  30135 Total Alias Queries Performed
  28471 no alias responses (94%)
  1448 may alias responses (4%)
  216 must alias responses (0%)
  Alias Analysis Evaluator Summary: 94%/4%/0%

Not bad for a purely local algorithm :)

-Chris



------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.