[LLVMbugs] [Bug 4187] New: GVN eliminates load across free?
bugzilla-daemon at cs.uiuc.edu
bugzilla-daemon at cs.uiuc.edu
Sun May 10 14:23:35 PDT 2009
http://llvm.org/bugs/show_bug.cgi?id=4187
Summary: GVN eliminates load across free?
Product: libraries
Version: trunk
Platform: PC
OS/Version: Windows XP
Status: NEW
Severity: normal
Priority: P2
Component: Scalar Optimizations
AssignedTo: unassignedbugs at nondot.org
ReportedBy: daniel at zuster.org
CC: resistor at mac.com, llvmbugs at cs.uiuc.edu
GVN eliminates the second load in this example, which I believe is invalid.
This didn't happen in 2.4, I believe.
--
; ModuleID = '2007-10-11-illegal-access-after-free-and-branch.c'
target datalayout =
"e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
target triple = "i386-apple-darwin10"
declare void @klee_make_symbolic(i8* %addr, i32 %nbytes) nounwind ssp;
define i32 @f0() nounwind ssp {
entry:
%t0 = malloc [3 x i8]
%.sub = getelementptr [3 x i8]* %t0, i32 0, i32 0
call void @klee_make_symbolic(i8* %.sub, i32 3) nounwind
%t2 = load i8* %.sub, align 4
free [3 x i8]* %t0
%t8 = load i8* %.sub, align 4
%t9 = zext i8 %t8 to i32
ret i32 %t9
}
--
Here:
--
ddunbar at ubuntu:regression$ llvm-as < load-elim.ll | opt -gvn | llvm-dis
; ModuleID = '<stdin>'
target datalayout =
"e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
target triple = "i386-apple-darwin10"
declare void @klee_make_symbolic(i8*, i32) nounwind ssp
define i32 @f0() nounwind ssp {
entry:
%t0 = malloc [3 x i8] ; <[3 x i8]*> [#uses=2]
%.sub = getelementptr [3 x i8]* %t0, i32 0, i32 0 ; <i8*>
[#uses=2]
call void @klee_make_symbolic(i8* %.sub, i32 3) nounwind
%t2 = load i8* %.sub, align 4 ; <i8> [#uses=1]
free [3 x i8]* %t0
%t9 = zext i8 %t2 to i32 ; <i32> [#uses=1]
ret i32 %t9
}
--
--
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