[SCEV] Make howFarToZero use a simpler formula for max backedge-taken count.

This is both easier to understand, and produces a tighter bound in certain
cases.

Differential Revision: https://reviews.llvm.org/D28393

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@291701 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Analysis/ScalarEvolution.cpp b/lib/Analysis/ScalarEvolution.cpp
index 44f1a6dde0d2195ca62adcb0d46b3845da2a0fbd..bd7e0e832f68e42096e93ed674d680566a81f689 100644 (file)
--- a/lib/Analysis/ScalarEvolution.cpp
+++ b/lib/Analysis/ScalarEvolution.cpp
@@ -7206,17 +7206,8 @@ ScalarEvolution::howFarToZero(const SCEV *V, const Loop *L, bool ControlsExit,
   // 1*N = -Start; -1*N = Start (mod 2^BW), so:
   //   N = Distance (as unsigned)
   if (StepC->getValue()->equalsInt(1) || StepC->getValue()->isAllOnesValue()) {
-    ConstantRange CR = getUnsignedRange(Start);
-    const SCEV *MaxBECount;
-    if (!CountDown && CR.getUnsignedMin().isMinValue())
-      // When counting up, the worst starting value is 1, not 0.
-      MaxBECount = CR.getUnsignedMax().isMinValue()
-        ? getConstant(APInt::getMinValue(CR.getBitWidth()))
-        : getConstant(APInt::getMaxValue(CR.getBitWidth()));
-    else
-      MaxBECount = getConstant(CountDown ? CR.getUnsignedMax()
-                                         : -CR.getUnsignedMin());
-    return ExitLimit(Distance, MaxBECount, false, Predicates);
+    APInt MaxBECount = getUnsignedRange(Distance).getUnsignedMax();
+    return ExitLimit(Distance, getConstant(MaxBECount), false, Predicates);
   }
 
   // As a special case, handle the instance where Step is a positive power of

diff --git a/test/Analysis/ScalarEvolution/max-trip-count.ll b/test/Analysis/ScalarEvolution/max-trip-count.ll
index 614e9b265ab0978d37ebad75596822f4b306a4ba..bee0ac90f05e6e69234421fdbe118650fe65e11d 100644 (file)
--- a/test/Analysis/ScalarEvolution/max-trip-count.ll
+++ b/test/Analysis/ScalarEvolution/max-trip-count.ll
@@ -207,3 +207,86 @@ for.cond.i:                                       ; preds = %for.body.i
 bar.exit:                                         ; preds = %for.cond.i, %for.body.i
   ret i32 0
 }
+
+; CHECK-LABEL: @ne_max_trip_count_1
+; CHECK: Loop %for.body: max backedge-taken count is 7
+define i32 @ne_max_trip_count_1(i32 %n) {
+entry:
+  %masked = and i32 %n, 7
+  br label %for.body
+
+for.body:
+  %i = phi i32 [ 0, %entry ], [ %add, %for.body ]
+  %add = add nsw i32 %i, 1
+  %cmp = icmp ne i32 %i, %masked
+  br i1 %cmp, label %for.body, label %bar.exit
+
+bar.exit:
+  ret i32 0
+}
+
+; CHECK-LABEL: @ne_max_trip_count_2
+; CHECK: Loop %for.body: max backedge-taken count is -1
+define i32 @ne_max_trip_count_2(i32 %n) {
+entry:
+  %masked = and i32 %n, 7
+  br label %for.body
+
+for.body:
+  %i = phi i32 [ 0, %entry ], [ %add, %for.body ]
+  %add = add nsw i32 %i, 1
+  %cmp = icmp ne i32 %add, %masked
+  br i1 %cmp, label %for.body, label %bar.exit
+
+bar.exit:
+  ret i32 0
+}
+
+; TODO: Improve count based on guard.
+; CHECK-LABEL: @ne_max_trip_count_3
+; CHECK: Loop %for.body: max backedge-taken count is -1
+define i32 @ne_max_trip_count_3(i32 %n) {
+entry:
+  %masked = and i32 %n, 7
+  %guard = icmp eq i32 %masked, 0
+  br i1 %guard, label %exit, label %for.preheader
+
+for.preheader:
+  br label %for.body
+
+for.body:
+  %i = phi i32 [ 0, %for.preheader ], [ %add, %for.body ]
+  %add = add nsw i32 %i, 1
+  %cmp = icmp ne i32 %add, %masked
+  br i1 %cmp, label %for.body, label %loop.exit
+
+loop.exit:
+  br label %exit
+
+exit:
+  ret i32 0
+}
+
+; TODO: Improve count based on guard.
+; CHECK-LABEL: @ne_max_trip_count_4
+; CHECK: Loop %for.body: max backedge-taken count is -1
+define i32 @ne_max_trip_count_4(i32 %n) {
+entry:
+  %guard = icmp eq i32 %n, 0
+  br i1 %guard, label %exit, label %for.preheader
+
+for.preheader:
+  br label %for.body
+
+for.body:
+  %i = phi i32 [ 0, %for.preheader ], [ %add, %for.body ]
+  %add = add nsw i32 %i, 1
+  %cmp = icmp ne i32 %add, %n
+  br i1 %cmp, label %for.body, label %loop.exit
+
+loop.exit:
+  br label %exit
+
+exit:
+  ret i32 0
+}