From: Nick Lewycky Date: Tue, 9 Dec 2008 07:25:04 +0000 (+0000) Subject: It's easy to handle SLE/SGE when the loop has a unit stride. X-Git-Url: http://demsky.eecs.uci.edu/git/?a=commitdiff_plain;h=0bed364307fa5c009239e65352ed36cb451953a7;p=oota-llvm.git It's easy to handle SLE/SGE when the loop has a unit stride. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60748 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/lib/Analysis/ScalarEvolution.cpp b/lib/Analysis/ScalarEvolution.cpp index 2b714de3b3f..c9af6de79d9 100644 --- a/lib/Analysis/ScalarEvolution.cpp +++ b/lib/Analysis/ScalarEvolution.cpp @@ -2924,8 +2924,12 @@ bool ScalarEvolutionsImpl::potentialInfiniteLoop(SCEV *Stride, SCEV *RHS, if (!R) return true; - if (isSigned) + if (isSigned) { + if (SC->getValue()->isOne()) + return R->getValue()->isMaxValue(true); + return true; // XXX: because we don't have an sdiv scev. + } // If negative, it wraps around every iteration, but we don't care about that. APInt S = SC->getValue()->getValue().abs(); diff --git a/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll b/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll new file mode 100644 index 00000000000..a9a7c056585 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll @@ -0,0 +1,24 @@ +; RUN: llvm-as < %s | opt -analyze -scalar-evolution | grep {255 iterations} + +define i32 @foo(i32 %x, i32 %y, i32* %lam, i32* %alp) nounwind { +bb1.thread: + br label %bb1 + +bb1: ; preds = %bb1, %bb1.thread + %indvar = phi i32 [ 0, %bb1.thread ], [ %indvar.next, %bb1 ] ; [#uses=4] + %i.0.reg2mem.0 = sub i32 255, %indvar ; [#uses=2] + %0 = getelementptr i32* %alp, i32 %i.0.reg2mem.0 ; [#uses=1] + %1 = load i32* %0, align 4 ; [#uses=1] + %2 = getelementptr i32* %lam, i32 %i.0.reg2mem.0 ; [#uses=1] + store i32 %1, i32* %2, align 4 + %3 = sub i32 254, %indvar ; [#uses=1] + %4 = icmp slt i32 %3, 0 ; [#uses=1] + %indvar.next = add i32 %indvar, 1 ; [#uses=1] + br i1 %4, label %bb2, label %bb1 + +bb2: ; preds = %bb1 + %tmp10 = mul i32 %indvar, %x ; [#uses=1] + %z.0.reg2mem.0 = add i32 %tmp10, %y ; [#uses=1] + %5 = add i32 %z.0.reg2mem.0, %x ; [#uses=1] + ret i32 %5 +}