Simplify [su]max(MAX, n) to MAX. This comes up in loop tripcount

computations in loops with multiple exits.

Adjust the testcase for PR4436 so that the relevant portion isn't
optimized away.

llvm-svn: 74073

3 files changed, 76 insertions, 5 deletions

diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 3bbeb95212e..6ac6efa1b85 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -1690,10 +1690,14 @@ ScalarEvolution::getSMaxExpr(SmallVectorImpl<const SCEV*> &Ops) {
       LHSC = cast<SCEVConstant>(Ops[0]);
     }
 
-    // If we are left with a constant -inf, strip it off.
+    // If we are left with a constant minimum-int, strip it off.
     if (cast<SCEVConstant>(Ops[0])->getValue()->isMinValue(true)) {
       Ops.erase(Ops.begin());
       --Idx;
+    } else if (cast<SCEVConstant>(Ops[0])->getValue()->isMaxValue(true)) {
+      // If we have an smax with a constant maximum-int, it will always be
+      // maximum-int.
+      return Ops[0];
     }
   }
 
@@ -1777,10 +1781,14 @@ ScalarEvolution::getUMaxExpr(SmallVectorImpl<const SCEV*> &Ops) {
       LHSC = cast<SCEVConstant>(Ops[0]);
     }
 
-    // If we are left with a constant zero, strip it off.
+    // If we are left with a constant minimum-int, strip it off.
     if (cast<SCEVConstant>(Ops[0])->getValue()->isMinValue(false)) {
       Ops.erase(Ops.begin());
       --Idx;
+    } else if (cast<SCEVConstant>(Ops[0])->getValue()->isMaxValue(false)) {
+      // If we have an umax with a constant maximum-int, it will always be
+      // maximum-int.
+      return Ops[0];
     }
   }
 
diff --git a/llvm/test/Transforms/IndVarSimplify/loop_evaluate7.ll b/llvm/test/Transforms/IndVarSimplify/loop_evaluate7.ll
index bce65e047da..6e31c55d526 100644
--- a/llvm/test/Transforms/IndVarSimplify/loop_evaluate7.ll
+++ b/llvm/test/Transforms/IndVarSimplify/loop_evaluate7.ll
@@ -4,7 +4,7 @@
 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32"
 target triple = "i386-pc-linux-gnu"
 
-define i8* @string_expandtabs() nounwind {
+define i8* @string_expandtabs(i32 %n, i8* %m) nounwind {
 entry:
 	br i1 undef, label %bb33, label %bb1
 
@@ -22,11 +22,11 @@ bb19:		; preds = %bb30
 
 bb20:		; preds = %bb19
 	%0 = load i32* undef, align 4		; <i32> [#uses=1]
-	%1 = sub i32 %0, undef		; <i32> [#uses=1]
+	%1 = sub i32 %0, %n		; <i32> [#uses=1]
 	br label %bb23
 
 bb21:		; preds = %bb23
-	%2 = icmp ult i8* %q.0, undef		; <i1> [#uses=1]
+	%2 = icmp ult i8* %q.0, %m		; <i1> [#uses=1]
 	br i1 %2, label %bb22, label %overflow2
 
 bb22:		; preds = %bb21
diff --git a/llvm/test/Transforms/IndVarSimplify/loop_evaluate8.ll b/llvm/test/Transforms/IndVarSimplify/loop_evaluate8.ll
new file mode 100644
index 00000000000..fa2f9e57f07
--- /dev/null
+++ b/llvm/test/Transforms/IndVarSimplify/loop_evaluate8.ll
@@ -0,0 +1,63 @@
+; RUN: llvm-as < %s | opt -indvars | llvm-dis | not grep select
+
+; This loop has backedge-taken-count zero. Indvars shouldn't expand any
+; instructions to compute a trip count.
+
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32"
+target triple = "i386-pc-linux-gnu"
+
+define i8* @string_expandtabs() nounwind {
+entry:
+	br i1 undef, label %bb33, label %bb1
+
+bb1:		; preds = %entry
+	br i1 undef, label %overflow1, label %bb15
+
+bb15:		; preds = %bb1
+	br i1 undef, label %bb33, label %bb17
+
+bb17:		; preds = %bb15
+	br label %bb30
+
+bb19:		; preds = %bb30
+	br i1 undef, label %bb20, label %bb29
+
+bb20:		; preds = %bb19
+	%0 = load i32* undef, align 4		; <i32> [#uses=1]
+	%1 = sub i32 %0, undef		; <i32> [#uses=1]
+	br label %bb23
+
+bb21:		; preds = %bb23
+	%2 = icmp ult i8* %q.0, undef		; <i1> [#uses=1]
+	br i1 %2, label %bb22, label %overflow2
+
+bb22:		; preds = %bb21
+	%3 = getelementptr i8* %q.0, i32 1		; <i8*> [#uses=1]
+	br label %bb23
+
+bb23:		; preds = %bb22, %bb20
+	%i.2 = phi i32 [ %1, %bb20 ], [ %4, %bb22 ]		; <i32> [#uses=1]
+	%q.0 = phi i8* [ undef, %bb20 ], [ %3, %bb22 ]		; <i8*> [#uses=3]
+	%4 = add i32 %i.2, -1		; <i32> [#uses=2]
+	%5 = icmp eq i32 %4, -1		; <i1> [#uses=1]
+	br i1 %5, label %bb29, label %bb21
+
+bb29:		; preds = %bb23, %bb19
+	%q.1 = phi i8* [ undef, %bb19 ], [ %q.0, %bb23 ]		; <i8*> [#uses=0]
+	br label %bb30
+
+bb30:		; preds = %bb29, %bb17
+	br i1 undef, label %bb19, label %bb33
+
+overflow2:		; preds = %bb21
+	br i1 undef, label %bb32, label %overflow1
+
+bb32:		; preds = %overflow2
+	br label %overflow1
+
+overflow1:		; preds = %bb32, %overflow2, %bb1
+	ret i8* null
+
+bb33:		; preds = %bb30, %bb15, %entry
+	ret i8* undef
+}