ScalarEvolution support for <= and >= loops.

Also, generalize ScalarEvolutions's min and max recognition to handle
some new forms of min and max that this change makes more common.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102234 91177308-0d34-0410-b5e6-96231b3b80d8

5 files changed, 138 insertions, 24 deletions

diff --git a/lib/Analysis/ScalarEvolution.cpp b/lib/Analysis/ScalarEvolution.cpp
index d62eef7..0ef5d84 100644
--- a/lib/Analysis/ScalarEvolution.cpp
+++ b/lib/Analysis/ScalarEvolution.cpp
@@ -3416,10 +3416,22 @@ const SCEV *ScalarEvolution::createSCEV(Value *V) {
         // fall through
       case ICmpInst::ICMP_SGT:
       case ICmpInst::ICMP_SGE:
-        if (LHS == U->getOperand(1) && RHS == U->getOperand(2))
-          return getSMaxExpr(getSCEV(LHS), getSCEV(RHS));
-        else if (LHS == U->getOperand(2) && RHS == U->getOperand(1))
-          return getSMinExpr(getSCEV(LHS), getSCEV(RHS));
+        // a >s b ? a+x : b+x  ->  smax(a, b)+x
+        // a >s b ? b+x : a+x  ->  smin(a, b)+x
+        if (LHS->getType() == U->getType()) {
+          const SCEV *LS = getSCEV(LHS);
+          const SCEV *RS = getSCEV(RHS);
+          const SCEV *LA = getSCEV(U->getOperand(1));
+          const SCEV *RA = getSCEV(U->getOperand(2));
+          const SCEV *LDiff = getMinusSCEV(LA, LS);
+          const SCEV *RDiff = getMinusSCEV(RA, RS);
+          if (LDiff == RDiff)
+            return getAddExpr(getSMaxExpr(LS, RS), LDiff);
+          LDiff = getMinusSCEV(LA, RS);
+          RDiff = getMinusSCEV(RA, LS);
+          if (LDiff == RDiff)
+            return getAddExpr(getSMinExpr(LS, RS), LDiff);
+        }
         break;
       case ICmpInst::ICMP_ULT:
       case ICmpInst::ICMP_ULE:
@@ -3427,28 +3439,52 @@ const SCEV *ScalarEvolution::createSCEV(Value *V) {
         // fall through
       case ICmpInst::ICMP_UGT:
       case ICmpInst::ICMP_UGE:
-        if (LHS == U->getOperand(1) && RHS == U->getOperand(2))
-          return getUMaxExpr(getSCEV(LHS), getSCEV(RHS));
-        else if (LHS == U->getOperand(2) && RHS == U->getOperand(1))
-          return getUMinExpr(getSCEV(LHS), getSCEV(RHS));
+        // a >u b ? a+x : b+x  ->  umax(a, b)+x
+        // a >u b ? b+x : a+x  ->  umin(a, b)+x
+        if (LHS->getType() == U->getType()) {
+          const SCEV *LS = getSCEV(LHS);
+          const SCEV *RS = getSCEV(RHS);
+          const SCEV *LA = getSCEV(U->getOperand(1));
+          const SCEV *RA = getSCEV(U->getOperand(2));
+          const SCEV *LDiff = getMinusSCEV(LA, LS);
+          const SCEV *RDiff = getMinusSCEV(RA, RS);
+          if (LDiff == RDiff)
+            return getAddExpr(getUMaxExpr(LS, RS), LDiff);
+          LDiff = getMinusSCEV(LA, RS);
+          RDiff = getMinusSCEV(RA, LS);
+          if (LDiff == RDiff)
+            return getAddExpr(getUMinExpr(LS, RS), LDiff);
+        }
         break;
       case ICmpInst::ICMP_NE:
-        // n != 0 ? n : 1  ->  umax(n, 1)
-        if (LHS == U->getOperand(1) &&
-            isa<ConstantInt>(U->getOperand(2)) &&
-            cast<ConstantInt>(U->getOperand(2))->isOne() &&
+        // n != 0 ? n+x : 1+x  ->  umax(n, 1)+x
+        if (LHS->getType() == U->getType() &&
             isa<ConstantInt>(RHS) &&
-            cast<ConstantInt>(RHS)->isZero())
-          return getUMaxExpr(getSCEV(LHS), getSCEV(U->getOperand(2)));
+            cast<ConstantInt>(RHS)->isZero()) {
+          const SCEV *One = getConstant(LHS->getType(), 1);
+          const SCEV *LS = getSCEV(LHS);
+          const SCEV *LA = getSCEV(U->getOperand(1));
+          const SCEV *RA = getSCEV(U->getOperand(2));
+          const SCEV *LDiff = getMinusSCEV(LA, LS);
+          const SCEV *RDiff = getMinusSCEV(RA, One);
+          if (LDiff == RDiff)
+            return getAddExpr(getUMaxExpr(LS, One), LDiff);
+        }
         break;
       case ICmpInst::ICMP_EQ:
-        // n == 0 ? 1 : n  ->  umax(n, 1)
-        if (LHS == U->getOperand(2) &&
-            isa<ConstantInt>(U->getOperand(1)) &&
-            cast<ConstantInt>(U->getOperand(1))->isOne() &&
+        // n == 0 ? 1+x : n+x  ->  umax(n, 1)+x
+        if (LHS->getType() == U->getType() &&
             isa<ConstantInt>(RHS) &&
-            cast<ConstantInt>(RHS)->isZero())
-          return getUMaxExpr(getSCEV(LHS), getSCEV(U->getOperand(1)));
+            cast<ConstantInt>(RHS)->isZero()) {
+          const SCEV *One = getConstant(LHS->getType(), 1);
+          const SCEV *LS = getSCEV(LHS);
+          const SCEV *LA = getSCEV(U->getOperand(1));
+          const SCEV *RA = getSCEV(U->getOperand(2));
+          const SCEV *LDiff = getMinusSCEV(LA, One);
+          const SCEV *RDiff = getMinusSCEV(RA, LS);
+          if (LDiff == RDiff)
+            return getAddExpr(getUMaxExpr(LS, One), LDiff);
+        }
         break;
       default:
         break;
@@ -3885,6 +3921,57 @@ ScalarEvolution::ComputeBackedgeTakenCountFromExitCondICmp(const Loop *L,
         if (!isa<SCEVCouldNotCompute>(Ret)) return Ret;
       }
 
+  // If possible, canonicalize GE/LE comparisons to GT/LT comparisons, by
+  // adding or subtracting 1 from one of the operands.
+  switch (Cond) {
+  case ICmpInst::ICMP_SLE:
+    if (!getSignedRange(RHS).getSignedMax().isMaxSignedValue()) {
+      RHS = getAddExpr(getConstant(RHS->getType(), 1, true), RHS,
+                       /*HasNUW=*/false, /*HasNSW=*/true);
+      Cond = ICmpInst::ICMP_SLT;
+    } else if (!getSignedRange(LHS).getSignedMin().isMinSignedValue()) {
+      LHS = getAddExpr(getConstant(RHS->getType(), -1, true), LHS,
+                       /*HasNUW=*/false, /*HasNSW=*/true);
+      Cond = ICmpInst::ICMP_SLT;
+    }
+    break;
+  case ICmpInst::ICMP_SGE:
+    if (!getSignedRange(RHS).getSignedMin().isMinSignedValue()) {
+      RHS = getAddExpr(getConstant(RHS->getType(), -1, true), RHS,
+                       /*HasNUW=*/false, /*HasNSW=*/true);
+      Cond = ICmpInst::ICMP_SGT;
+    } else if (!getSignedRange(LHS).getSignedMax().isMaxSignedValue()) {
+      LHS = getAddExpr(getConstant(RHS->getType(), 1, true), LHS,
+                       /*HasNUW=*/false, /*HasNSW=*/true);
+      Cond = ICmpInst::ICMP_SGT;
+    }
+    break;
+  case ICmpInst::ICMP_ULE:
+    if (!getUnsignedRange(RHS).getUnsignedMax().isMaxValue()) {
+      RHS = getAddExpr(getConstant(RHS->getType(), 1, false), RHS,
+                       /*HasNUW=*/true, /*HasNSW=*/false);
+      Cond = ICmpInst::ICMP_ULT;
+    } else if (!getUnsignedRange(LHS).getUnsignedMin().isMinValue()) {
+      LHS = getAddExpr(getConstant(RHS->getType(), -1, false), LHS,
+                       /*HasNUW=*/true, /*HasNSW=*/false);
+      Cond = ICmpInst::ICMP_ULT;
+    }
+    break;
+  case ICmpInst::ICMP_UGE:
+    if (!getUnsignedRange(RHS).getUnsignedMin().isMinValue()) {
+      RHS = getAddExpr(getConstant(RHS->getType(), -1, false), RHS,
+                       /*HasNUW=*/true, /*HasNSW=*/false);
+      Cond = ICmpInst::ICMP_UGT;
+    } else if (!getUnsignedRange(LHS).getUnsignedMax().isMaxValue()) {
+      LHS = getAddExpr(getConstant(RHS->getType(), 1, false), LHS,
+                       /*HasNUW=*/true, /*HasNSW=*/false);
+      Cond = ICmpInst::ICMP_UGT;
+    }
+    break;
+  default:
+    break;
+  }
+
   switch (Cond) {
   case ICmpInst::ICMP_NE: {                     // while (X != Y)
     // Convert to: while (X-Y != 0)
diff --git a/test/Analysis/ScalarEvolution/2008-11-18-LessThanOrEqual.ll b/test/Analysis/ScalarEvolution/2008-11-18-LessThanOrEqual.ll
index 06637b5..335bbaf 100644
--- a/test/Analysis/ScalarEvolution/2008-11-18-LessThanOrEqual.ll
+++ b/test/Analysis/ScalarEvolution/2008-11-18-LessThanOrEqual.ll
@@ -1,6 +1,5 @@
 ; RUN: opt < %s -analyze -scalar-evolution |& \
 ; RUN: grep {Loop %bb: backedge-taken count is (7 + (-1 \\* %argc))}
-; XFAIL: *
 
 define i32 @main(i32 %argc, i8** %argv) nounwind {
 entry:
diff --git a/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll b/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll
index 226221b..25a0434 100644
--- a/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll
+++ b/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll
@@ -1,5 +1,4 @@
 ; RUN: opt < %s -analyze -scalar-evolution | grep {backedge-taken count is 255}
-; XFAIL: *
 
 define i32 @foo(i32 %x, i32 %y, i32* %lam, i32* %alp) nounwind {
 bb1.thread:
diff --git a/test/Analysis/ScalarEvolution/2008-12-11-SMaxOverflow.ll b/test/Analysis/ScalarEvolution/2008-12-11-SMaxOverflow.ll
index 33a7479..12254e3 100644
--- a/test/Analysis/ScalarEvolution/2008-12-11-SMaxOverflow.ll
+++ b/test/Analysis/ScalarEvolution/2008-12-11-SMaxOverflow.ll
@@ -1,5 +1,7 @@
-; RUN: opt < %s -analyze -scalar-evolution | grep {0 smax}
-; XFAIL: *
+; RUN: opt < %s -analyze -scalar-evolution | FileCheck %s
+
+; CHECK: @f
+; CHECK: Loop %bb16.preheader: backedge-taken count is (-1 + %c.idx.val)
 
 define i32 @f(i32 %c.idx.val) {
 
diff --git a/test/Analysis/ScalarEvolution/sle.ll b/test/Analysis/ScalarEvolution/sle.ll
new file mode 100644
index 0000000..f38f6b6
--- /dev/null
+++ b/test/Analysis/ScalarEvolution/sle.ll
@@ -0,0 +1,27 @@
+; RUN: opt -analyze -scalar-evolution < %s | FileCheck %s
+
+; ScalarEvolution should be able to use nsw information to prove that
+; this loop has a finite trip count.
+
+; CHECK: @le
+; CHECK: Loop %for.body: backedge-taken count is %n
+; CHECK: Loop %for.body: max backedge-taken count is 9223372036854775807
+
+define void @le(i64 %n, double* nocapture %p) nounwind {
+entry:
+  %cmp6 = icmp slt i64 %n, 0                      ; <i1> [#uses=1]
+  br i1 %cmp6, label %for.end, label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+  %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ] ; <i64> [#uses=2]
+  %arrayidx = getelementptr double* %p, i64 %i    ; <double*> [#uses=2]
+  %t4 = load double* %arrayidx                    ; <double> [#uses=1]
+  %mul = fmul double %t4, 2.200000e+00            ; <double> [#uses=1]
+  store double %mul, double* %arrayidx
+  %i.next = add nsw i64 %i, 1                     ; <i64> [#uses=2]
+  %cmp = icmp sgt i64 %i.next, %n                 ; <i1> [#uses=1]
+  br i1 %cmp, label %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body, %entry
+  ret void
+}