[llvm] r284378 - [SCEV] Consider delinearization pattern with extension with identity factor

[Tobias Grosser via llvm-commits]

Author: grosser
Date: Mon Oct 17 06:56:26 2016
New Revision: 284378

URL: http://llvm.org/viewvc/llvm-project?rev=284378&view=rev
Log:
[SCEV] Consider delinearization pattern with extension with identity factor

Summary: The delinearization algorithm did not consider terms which had an extension without a multiply factor, i.e. a identify factor. We lose cases where size is char type where there will no multiply factor.

Reviewers: sanjoy, grosser

Subscribers: mzolotukhin, Eugene.Zelenko, llvm-commits, mssimpso, sanjoy, grosser

Differential Revision: https://reviews.llvm.org/D16492

Added:
    llvm/trunk/test/Analysis/Delinearization/terms_with_identity_factor.ll
Modified:
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=284378&r1=284377&r2=284378&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon Oct 17 06:56:26 2016
@@ -8948,7 +8948,8 @@ struct SCEVCollectTerms {
       : Terms(T) {}
 
   bool follow(const SCEV *S) {
-    if (isa<SCEVUnknown>(S) || isa<SCEVMulExpr>(S)) {
+    if (isa<SCEVUnknown>(S) || isa<SCEVMulExpr>(S) ||
+        isa<SCEVSignExtendExpr>(S)) {
       if (!containsUndefs(S))
         Terms.push_back(S);
 

Added: llvm/trunk/test/Analysis/Delinearization/terms_with_identity_factor.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/Delinearization/terms_with_identity_factor.ll?rev=284378&view=auto
==============================================================================
--- llvm/trunk/test/Analysis/Delinearization/terms_with_identity_factor.ll (added)
+++ llvm/trunk/test/Analysis/Delinearization/terms_with_identity_factor.ll Mon Oct 17 06:56:26 2016
@@ -0,0 +1,64 @@
+; REQUIRES: asserts
+; RUN: opt < %s -analyze -delinearize -debug 2>&1 | FileCheck %s
+; void foo (int m, int n, char *A) {
+;    for (int i=0; i < m; i++)
+;      for(int j=0; j< n; j++)
+;        A[i*n+j] += 1;
+;}
+
+; ModuleID = 'delin.cpp'
+;target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
+;target triple = "aarch64--linux-gnu"
+
+; CHECK-LABEL: Delinearization on function foo
+; CHECK: Inst:  %4 = load i8, i8* %arrayidx.us, align 1
+; CHECK: Subscripts
+; CHECK-NEXT: {0,+,1}<nuw><nsw><%for.body3.lr.ph.us>
+; CHECK-NEXT: {0,+,1}<nuw><nsw><%for.body3.us>
+; CHECK: succeeded to delinearize
+
+define void @foo(i32 %m, i32 %n, i8* nocapture %A) #0 {
+entry:
+  br label %entry.split
+
+entry.split:                                      ; preds = %entry
+  %cmp15 = icmp sgt i32 %m, 0
+  %cmp213 = icmp sgt i32 %n, 0
+  %or.cond = and i1 %cmp15, %cmp213
+  br i1 %or.cond, label %for.cond1.preheader.lr.ph.split.us, label %for.end8
+
+for.cond1.preheader.lr.ph.split.us:               ; preds = %entry.split
+  %0 = add i32 %n, -1
+  %1 = sext i32 %n to i64
+  %2 = add i32 %m, -1
+  br label %for.body3.lr.ph.us
+
+for.body3.us:                                     ; preds = %for.body3.us, %for.body3.lr.ph.us
+  %indvars.iv = phi i64 [ 0, %for.body3.lr.ph.us ], [ %indvars.iv.next, %for.body3.us ]
+  %3 = add nsw i64 %indvars.iv, %5
+  %arrayidx.us = getelementptr inbounds i8, i8* %A, i64 %3
+  %4 = load i8, i8* %arrayidx.us, align 1
+  %add4.us = add i8 %4, 1
+  store i8 %add4.us, i8* %arrayidx.us, align 1
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, %0
+  br i1 %exitcond, label %for.cond1.for.inc6_crit_edge.us, label %for.body3.us
+
+for.body3.lr.ph.us:                               ; preds = %for.cond1.for.inc6_crit_edge.us, %for.cond1.preheader.lr.ph.split.us
+  %indvars.iv19 = phi i64 [ %indvars.iv.next20, %for.cond1.for.inc6_crit_edge.us ], [ 0, %for.cond1.preheader.lr.ph.split.us ]
+  %5 = mul nsw i64 %indvars.iv19, %1
+  br label %for.body3.us
+
+for.cond1.for.inc6_crit_edge.us:                  ; preds = %for.body3.us
+  %indvars.iv.next20 = add nuw nsw i64 %indvars.iv19, 1
+  %lftr.wideiv22 = trunc i64 %indvars.iv19 to i32
+  %exitcond23 = icmp eq i32 %lftr.wideiv22, %2
+  br i1 %exitcond23, label %for.end8.loopexit, label %for.body3.lr.ph.us
+
+for.end8.loopexit:                                ; preds = %for.cond1.for.inc6_crit_edge.us
+  br label %for.end8
+
+for.end8:                                         ; preds = %for.end8.loopexit, %entry.split
+  ret void
+}