[FIX] Don't consider reductions which are partially outside the SCoP

+ Test case

llvm-svn: 212080
This commit is contained in:
Johannes Doerfert 2014-07-01 00:32:29 +00:00
parent 0e2cc2a519
commit 9890a05287
2 changed files with 71 additions and 2 deletions

View File

@ -732,6 +732,10 @@ void ScopStmt::collectCandiateReductionLoads(
if (!BinOp->isCommutative() || !BinOp->isAssociative())
return;
// Skip if the binary operator is outside the current SCoP
if (BinOp->getParent() != Store->getParent())
return;
// Skip if it is a multiplicative reduction and we disabled them
if (DisableMultiplicativeReductions &&
(BinOp->getOpcode() == Instruction::Mul ||
@ -746,8 +750,10 @@ void ScopStmt::collectCandiateReductionLoads(
// A load is only a candidate if it cannot escape (thus has only this use)
if (PossibleLoad0 && PossibleLoad0->getNumUses() == 1)
if (PossibleLoad0->getParent() == Store->getParent())
Loads.push_back(lookupAccessFor(PossibleLoad0));
if (PossibleLoad1 && PossibleLoad1->getNumUses() == 1)
if (PossibleLoad1->getParent() == Store->getParent())
Loads.push_back(lookupAccessFor(PossibleLoad1));
}

View File

@ -0,0 +1,63 @@
; RUN: opt %loadPolly -polly-scops -analyze < %s | FileCheck %s
;
; CHECK-NOT: Reduction like: 1
;
; int c, d;
; void f(int *sum) {
; for (int i = 0; i < 1024; i++)
; *sum = c + d;
; }
;
target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-n32-S64"
@c = common global i32 0, align 4
@d = common global i32 0, align 4
define void @loads_outside_scop(i32* %sum) {
entry:
%tmp = load i32* @c, align 4
%tmp1 = load i32* @d, align 4
br label %for.cond
for.cond: ; preds = %for.inc, %entry
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
%exitcond = icmp ne i32 %i.0, 1024
br i1 %exitcond, label %for.body, label %for.end
for.body: ; preds = %for.cond
%add = add nsw i32 %tmp, %tmp1
store i32 %add, i32* %sum, align 4
br label %for.inc
for.inc: ; preds = %for.body
%inc = add nsw i32 %i.0, 1
br label %for.cond
for.end: ; preds = %for.cond
ret void
}
define void @binop_outside_scop(i32* %sum) {
entry:
%tmp = load i32* @c, align 4
%tmp1 = load i32* @d, align 4
%add = add nsw i32 %tmp, %tmp1
br label %for.cond
for.cond: ; preds = %for.inc, %entry
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
%exitcond = icmp ne i32 %i.0, 1024
br i1 %exitcond, label %for.body, label %for.end
for.body: ; preds = %for.cond
store i32 %add, i32* %sum, align 4
br label %for.inc
for.inc: ; preds = %for.body
%inc = add nsw i32 %i.0, 1
br label %for.cond
for.end: ; preds = %for.cond
ret void
}