[LLVMbugs] [Bug 3323] New: AliasAnalysis pass removed before function pass can use it
bugzilla-daemon at cs.uiuc.edu
bugzilla-daemon at cs.uiuc.edu
Wed Jan 14 02:54:49 PST 2009
http://llvm.org/bugs/show_bug.cgi?id=3323
Summary: AliasAnalysis pass removed before function pass can use
it
Product: libraries
Version: 2.3
Platform: PC
OS/Version: Linux
Status: NEW
Severity: normal
Priority: P2
Component: Core LLVM classes
AssignedTo: unassignedbugs at nondot.org
ReportedBy: hvdieren at elis.ugent.be
CC: hvdieren at elis.ugent.be, llvmbugs at cs.uiuc.edu
The following command line suggests that the LICM pass can use the Andersens
Alias Analysis (for instance) to do its alias analysis:
$ opt -debug-pass=Structure -disable-opt -basicaa -anders-aa -licm in.bc -f -o
out.bc
But as the output shows, the AndersensAA, which is a ModulePass, is destroyed
before the FunctionPasses are run!
Pass Arguments: -anders-aa -domtree -loops -loopsimplify -domfrontier
-scalar-evolution -licm -preverify -domtree -verify
Target Data Layout
Basic Alias Analysis (default AA impl)
ModulePass Manager
Andersen's Interprocedural Alias Analysis
-- Andersen's Interprocedural Alias Analysis
FunctionPass Manager
Dominator Tree Construction
Natural Loop Construction
Canonicalize natural loops
Dominance Frontier Construction
Scalar Evolution Analysis
Loop Pass Manager
Loop Invariant Code Motion
-- Loop Invariant Code Motion
-- Canonicalize natural loops
-- Basic Alias Analysis (default AA impl)
-- Natural Loop Construction
-- Dominator Tree Construction
-- Dominance Frontier Construction
-- Scalar Evolution Analysis
Preliminary module verification
Dominator Tree Construction
Module Verifier
-- Preliminary module verification
-- Dominator Tree Construction
-- Module Verifier
Bitcode Writer
-- Bitcode Writer
The problem is thus that FunctionPasses (and other finer-grain passes) can
never benefit from detailed AliasAnalysis passes, which are invariably
ModulePasses.
--
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
More information about the llvm-bugs
mailing list