[LLVMbugs] [Bug 4775] New: -indvars pass turning static alloca into dynamic alloca
bugzilla-daemon at cs.uiuc.edu
bugzilla-daemon at cs.uiuc.edu
Tue Aug 25 10:16:50 PDT 2009
http://llvm.org/bugs/show_bug.cgi?id=4775
Summary: -indvars pass turning static alloca into dynamic alloca
Product: libraries
Version: 1.0
Platform: PC
OS/Version: All
Status: NEW
Keywords: code-quality, regression
Severity: normal
Priority: P2
Component: Scalar Optimizations
AssignedTo: unassignedbugs at nondot.org
ReportedBy: clattner at apple.com
CC: gohman at apple.com, llvmbugs at cs.uiuc.edu
This seems pretty bad: opt -indvars turns this:
; 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-darwin10.0"
@llvm.used = appending global [1 x i8*] [i8* bitcast (i32 ()* @main to i8*)],
section "llvm.metadata" ; <[1 x i8*]*> [#uses=0]
define i32 @main() nounwind ssp {
entry:
%result.i = alloca i32, align 4 ; <i32*> [#uses=2]
br label %while.cond
while.cond: ; preds = %while.cond, %entry
%call = call i32 @bar() nounwind ssp ; <i32> [#uses=1]
%tobool = icmp eq i32 %call, 0 ; <i1> [#uses=1]
br i1 %tobool, label %while.end, label %while.cond
while.end: ; preds = %while.cond
volatile store i32 0, i32* %result.i
%tmp.i = volatile load i32* %result.i ; <i32> [#uses=0]
ret i32 0
}
declare i32 @bar() ssp
into:
...
br i1 %tobool, label %while.end, label %while.cond
while.end: ; preds = %while.cond
%result.i = alloca i32, align 4 ; <i32*> [#uses=2]
volatile store i32 0, i32* %result.i
%tmp.i = volatile load i32* %result.i ; <i32> [#uses=0]
ret i32 0
Dynamic allocas are *much* more expensive than static ones.
--
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