Index: include/llvm/Analysis/TargetTransformInfoImpl.h
===================================================================
--- include/llvm/Analysis/TargetTransformInfoImpl.h
+++ include/llvm/Analysis/TargetTransformInfoImpl.h
@@ -807,15 +807,9 @@
       return static_cast<T *>(this)->getCallCost(F, Arguments);
     }
 
-    if (const CastInst *CI = dyn_cast<CastInst>(U)) {
-      // Result of a cmp instruction is often extended (to be used by other
-      // cmp instructions, logical or return instructions). These are usually
-      // nop on most sane targets.
-      if (isa<CmpInst>(CI->getOperand(0)))
-        return TTI::TCC_Free;
-      if (isa<SExtInst>(CI) || isa<ZExtInst>(CI) || isa<FPExtInst>(CI))
-        return static_cast<T *>(this)->getExtCost(CI, Operands.back());
-    }
+    if (isa<SExtInst>(U) || isa<ZExtInst>(U) || isa<FPExtInst>(U))
+      return static_cast<T *>(this)->getExtCost(cast<Instruction>(U),
+                                                Operands.back());
 
     return static_cast<T *>(this)->getOperationCost(
         Operator::getOpcode(U), U->getType(),
Index: test/Analysis/CostModel/SystemZ/getUserCost_ext_i1.ll
===================================================================
--- /dev/null
+++ test/Analysis/CostModel/SystemZ/getUserCost_ext_i1.ll
@@ -0,0 +1,95 @@
+; RUN: opt < %s -mtriple=s390x-unknown-linux -mcpu=z13 -o - -loop-unroll \
+; RUN:  -debug-only=loop-unroll 2>&1 | FileCheck %s
+; REQUIRES: asserts
+;
+; Check that getUserCost() does not return TCC_Free (0 cost) for extensions
+; of i1. This function is called by CodeMetrics and seems to be a poor-mans
+; alternative to the methods CostModel is using. There is no direct way of
+; testing the costs for the instructions, but one way is to make a small loop
+; and run loop-unroller and see what the loop size is.
+;
+; Phis are free, but the other instructions in for.body are not. Test that
+; extensions of i1 is counted as 1 (and not 0) by checking that the sum is 5.
+
+; CHECK: fun0
+; CHECK: Loop Size = 5
+; CHECK: fun1
+; CHECK: Loop Size = 5
+; CHECK: fun2
+; CHECK: Loop Size = 5
+; CHECK: fun3
+; CHECK: Loop Size = 5
+
+define i64 @fun0(i64 %n, i32 %v) {
+entry:
+  %cmp0 = icmp slt i32 %v, 0
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %S  = phi i64 [ 0, %entry], [ %a, %for.body ]
+  %conv = zext i1 %cmp0 to i64
+  %a = add i64 %S, %conv
+  %iv.next = add nuw nsw i64 %iv, 1
+  %cmp = icmp slt i64 %iv.next, %n
+  br i1 %cmp, label %for.body, label %for.end
+
+for.end:
+  ret i64 %S
+}
+
+define i64 @fun1(i64 %n, i32 %v) {
+entry:
+  %cmp0 = icmp slt i32 %v, 0
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %S  = phi i64 [ 0, %entry], [ %a, %for.body ]
+  %conv = sext i1 %cmp0 to i64
+  %a = add i64 %S, %conv
+  %iv.next = add nuw nsw i64 %iv, 1
+  %cmp = icmp slt i64 %iv.next, %n
+  br i1 %cmp, label %for.body, label %for.end
+
+for.end:
+  ret i64 %S
+}
+
+define double @fun2(i64 %n, i32 %v) {
+entry:
+  %cmp0 = icmp slt i32 %v, 0
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %S  = phi double [ 0.0, %entry], [ %fpadd, %for.body ]
+  %conv = uitofp i1 %cmp0 to double
+  %fpadd = fadd double %S, %conv
+  %iv.next = add nuw nsw i64 %iv, 1
+  %cmp = icmp slt i64 %iv.next, %n
+  br i1 %cmp, label %for.body, label %for.end
+
+for.end:
+  ret double %S
+}
+
+define double @fun3(i64 %n, i32 %v) {
+entry:
+  %cmp0 = icmp slt i32 %v, 0
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %S  = phi double [ 0.0, %entry], [ %fpadd, %for.body ]
+  %conv = sitofp i1 %cmp0 to double
+  %fpadd = fadd double %S, %conv
+  %iv.next = add nuw nsw i64 %iv, 1
+  %cmp = icmp slt i64 %iv.next, %n
+  br i1 %cmp, label %for.body, label %for.end
+
+for.end:
+  ret double %S
+}
+
+