From: Brendon Cahoon <bcahoon@codeaurora.org>
Date: Mon, 20 Apr 2015 16:03:28 +0000 (+0000)
Subject: Recognize n/1 in the SCEV divide function
X-Git-Tag: llvmorg-3.7.0-rc1~6408
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=a57cc8bc817f3ff7a48bfd8221562e3cc2a2bc10;p=platform%2Fupstream%2Fllvm.git

Recognize n/1 in the SCEV divide function

n/1 generates a quotient equal to n and a remainder of 0.
If this case is not recognized, then the SCEV divide() function
can return a remainder that is greater than or equal to the
denominator, which means the delinearized subscripts for the
test case will be incorrect.

Differential Revision: http://reviews.llvm.org/D9003

llvm-svn: 235311
---

diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 37377f0..d88b026 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -726,6 +726,13 @@ public:
       return;
     }
 
+    // A simple case when N/1. The quotient is N.
+    if (Denominator->isOne()) {
+      *Quotient = Numerator;
+      *Remainder = D.Zero;
+      return;
+    }
+
     // Split the Denominator when it is a product.
     if (const SCEVMulExpr *T = dyn_cast<const SCEVMulExpr>(Denominator)) {
       const SCEV *Q, *R;
diff --git a/llvm/test/Analysis/Delinearization/divide_by_one.ll b/llvm/test/Analysis/Delinearization/divide_by_one.ll
new file mode 100644
index 0000000..c6e0b13
--- /dev/null
+++ b/llvm/test/Analysis/Delinearization/divide_by_one.ll
@@ -0,0 +1,70 @@
+; RUN: opt < %s -analyze -delinearize | FileCheck %s
+
+target datalayout = "e-m:e-p:32:32-i1:32-i64:64-a:0-n32"
+
+; Check that division by 1 can be delinearized.
+;
+; void test1(unsigned char *dst, int stride, int bs) {
+;   for (int r = bs; r >= 0; --r)
+;     for (int c = 0; c < bs; ++c)
+;       dst[r * stride + c] = dst[(r + 1) * stride + c - 1];
+; }
+
+; AddRec: {{(-1 + ((1 + %bs) * %stride)),+,(-1 * %stride)}<%for.cond1.preheader>,+,1}<nw><%for.body3>
+; CHECK: Inst:  %0 = load i8, i8* %arrayidx, align 1
+; CHECK: Base offset: %dst
+; CHECK: ArrayDecl[UnknownSize][%stride] with elements of 1 bytes.
+; CHECK: ArrayRef[{(1 + %bs),+,-1}<nw><%for.cond1.preheader>][{-1,+,1}<nw><%for.body3>]
+
+; AddRec: {{(%stride * %bs),+,(-1 * %stride)}<%for.cond1.preheader>,+,1}<nw><%for.body3>
+; CHECK: Inst:  store i8 %0, i8* %arrayidx7, align 1
+; CHECK: Base offset: %dst
+; CHECK: ArrayDecl[UnknownSize][%stride] with elements of 1 bytes.
+; CHECK: ArrayRef[{%bs,+,-1}<nsw><%for.cond1.preheader>][{0,+,1}<nuw><nsw><%for.body3>]
+
+define void @test(i8* nocapture %dst, i32 %stride, i32 %bs) {
+entry:
+  %cmp20 = icmp sgt i32 %bs, -1
+  br i1 %cmp20, label %for.cond1.preheader.lr.ph, label %for.end9
+
+for.cond1.preheader.lr.ph:
+  %cmp218 = icmp slt i32 0, %bs
+  br label %for.cond1.preheader
+
+for.cond1.preheader:
+  %r.021 = phi i32 [ %bs, %for.cond1.preheader.lr.ph ], [ %dec, %for.inc8 ]
+  br i1 %cmp218, label %for.body3.lr.ph, label %for.inc8
+
+for.body3.lr.ph:
+  %add = add nsw i32 %r.021, 1
+  %mul = mul nsw i32 %add, %stride
+  %add4 = add i32 %mul, -1
+  %mul5 = mul nsw i32 %r.021, %stride
+  br label %for.body3
+
+for.body3:
+  %c.019 = phi i32 [ 0, %for.body3.lr.ph ], [ %inc, %for.body3 ]
+  %sub = add i32 %add4, %c.019
+  %arrayidx = getelementptr inbounds i8, i8* %dst, i32 %sub
+  %0 = load i8, i8* %arrayidx, align 1
+  %add6 = add nsw i32 %c.019, %mul5
+  %arrayidx7 = getelementptr inbounds i8, i8* %dst, i32 %add6
+  store i8 %0, i8* %arrayidx7, align 1
+  %inc = add nsw i32 %c.019, 1
+  %cmp2 = icmp slt i32 %inc, %bs
+  br i1 %cmp2, label %for.body3, label %for.cond1.for.inc8_crit_edge
+
+for.cond1.for.inc8_crit_edge:
+  br label %for.inc8
+
+for.inc8:
+  %dec = add nsw i32 %r.021, -1
+  %cmp = icmp sgt i32 %dec, -1
+  br i1 %cmp, label %for.cond1.preheader, label %for.cond.for.end9_crit_edge
+
+for.cond.for.end9_crit_edge:
+  br label %for.end9
+
+for.end9:
+  ret void
+}