[PATCH] D48283: [SCEV] Properly solve quadratic equations with coefficients in i1

[Krzysztof Parzyszek via Phabricator via llvm-commits]

kparzysz created this revision.
kparzysz added a reviewer: sanjoy.
kparzysz edited the summary of this revision.

Commit r334318 exposed a problem in SolveQuadraticEquation: an add-rec was created with coefficients in i1, and division by 2 became equivalent to division by zero.  See http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20180611/561033.html.

This patch adds special treatment of i1 to avoid such complications.


Repository:
  rL LLVM

https://reviews.llvm.org/D48283

Files:
  lib/Analysis/ScalarEvolution.cpp
  test/Analysis/ScalarEvolution/solve-quadratic-i1.ll


Index: test/Analysis/ScalarEvolution/solve-quadratic-i1.ll
===================================================================
--- /dev/null
+++ test/Analysis/ScalarEvolution/solve-quadratic-i1.ll
@@ -0,0 +1,39 @@
+; RUN: opt -analyze -scalar-evolution -S < %s | FileCheck %s
+
+; CHECK: Printing analysis 'Scalar Evolution Analysis' for function 'f1':
+; CHECK-NEXT: Classifying expressions for: @f1
+; CHECK-NEXT:   %a.0 = phi i16 [ 2, %entry ], [ %inc, %lbl1 ]
+; CHECK-NEXT:   -->  {2,+,1}<nuw><nsw><%lbl1> U: [2,-32768) S: [2,-32768) Exits: 3 LoopDispositions: { %lbl1: Computable }
+; CHECK-NEXT:   %b.0 = phi i16 [ 1, %entry ], [ %add, %lbl1 ]
+; CHECK-NEXT:   -->  {1,+,2,+,1}<%lbl1> U: full-set S: full-set Exits: 3 LoopDispositions: { %lbl1: Computable }
+; CHECK-NEXT:   %inc = add nsw i16 %a.0, 1
+; CHECK-NEXT:   -->  {3,+,1}<nuw><%lbl1> U: [3,0) S: [3,0) Exits: 4 LoopDispositions: { %lbl1: Computable }
+; CHECK-NEXT:   %add = add nsw i16 %b.0, %a.0
+; CHECK-NEXT:   -->  {3,+,3,+,1}<%lbl1> U: full-set S: full-set Exits: 6 LoopDispositions: { %lbl1: Computable }
+; CHECK-NEXT:   %and = and i16 %add, 1
+; CHECK-NEXT:   -->  (zext i1 {true,+,true,+,true}<%lbl1> to i16) U: [0,2) S: [0,2) Exits: 0 LoopDispositions: { %lbl1: Computable }
+; CHECK-NEXT: Determining loop execution counts for: @f1
+; CHECK-NEXT: Loop %lbl1: backedge-taken count is 1
+; CHECK-NEXT: Loop %lbl1: max backedge-taken count is 1
+; CHECK-NEXT: Loop %lbl1: Predicated backedge-taken count is 1
+; CHECK-NEXT:  Predicates:
+; CHECK: Loop %lbl1: Trip multiple is 2
+
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @f1() {
+entry:
+  br label %lbl1
+
+lbl1:                                             ; preds = %lbl1, %entry
+  %a.0 = phi i16 [ 2, %entry ], [ %inc, %lbl1 ]
+  %b.0 = phi i16 [ 1, %entry ], [ %add, %lbl1 ]
+  %inc = add nsw i16 %a.0, 1
+  %add = add nsw i16 %b.0, %a.0
+  %and = and i16 %add, 1
+  %tobool = icmp ne i16 %and, 0
+  br i1 %tobool, label %lbl1, label %if.end
+
+if.end:                                           ; preds = %lbl1
+  ret void
+}
Index: lib/Analysis/ScalarEvolution.cpp
===================================================================
--- lib/Analysis/ScalarEvolution.cpp
+++ lib/Analysis/ScalarEvolution.cpp
@@ -8278,6 +8278,30 @@
   const APInt &L = LC->getAPInt();
   const APInt &M = MC->getAPInt();
   const APInt &N = NC->getAPInt();
+
+  // If the bit width is 1, then the arithmetic is a bit different.
+  if (BitWidth == 1) {
+    // Normally, the increments are M+N, M+2N, ...
+    // Here they are M+N, M, M+N, etc., so the accumulated values are
+    // L+M+N, L+N, L+M, L, and these four values will repeat.
+    bool BL = !L.isNullValue(), BM = !M.isNullValue(), BN = !N.isNullValue();
+    int Iter;
+
+    if (!(BL ^ BM ^ BN))
+      Iter = 0;   // e.g. L=N, M=0
+    else if (!(BL ^ BN))
+      Iter = 1;   // e.g. L=N, M=1
+    else if (!(BL ^ BM))
+      Iter = 2;   // e.g. L=0, M=0, N=1
+    else if (!BL)
+      Iter = 3;   // e.g. L=0, M=1, N=1
+    else
+      return None;
+
+    const auto *Root = cast<SCEVConstant>(SE.getConstant(APInt(32, Iter)));
+    return std::make_pair(Root, Root);
+  }
+
   APInt Two(BitWidth, 2);
 
   // Convert from chrec coefficients to polynomial coefficients AX^2+BX+C


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D48283.151726.patch
Type: text/x-patch
Size: 3276 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180618/89bc9621/attachment.bin>