diff --git a/llvm/include/llvm/CodeGen/BasicTTIImpl.h b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
--- a/llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -1153,12 +1153,14 @@
       // FIXME: If necessary, this should go in target-specific overrides.
       if (VF == 1 && RetVF == 1 && getTLI()->isCheapToSpeculateCttz())
         return TargetTransformInfo::TCC_Basic;
+      return TargetTransformInfo::TCC_Expensive;
       break;
 
     case Intrinsic::ctlz:
       // FIXME: If necessary, this should go in target-specific overrides.
       if (VF == 1 && RetVF == 1 && getTLI()->isCheapToSpeculateCtlz())
         return TargetTransformInfo::TCC_Basic;
+      return TargetTransformInfo::TCC_Expensive;
       break;
 
     case Intrinsic::memcpy:
diff --git a/llvm/test/Transforms/SimplifyCFG/X86/speculate-cttz-ctlz.ll b/llvm/test/Transforms/SimplifyCFG/X86/speculate-cttz-ctlz.ll
--- a/llvm/test/Transforms/SimplifyCFG/X86/speculate-cttz-ctlz.ll
+++ b/llvm/test/Transforms/SimplifyCFG/X86/speculate-cttz-ctlz.ll
@@ -407,22 +407,64 @@
 }
 
 define i16 @test9_loop(i32 %x, i16* %ptr) {
-; ALL-LABEL: @test9_loop(
-; ALL-NEXT:  entry:
-; ALL-NEXT:    br label [[LOOP_HEADER:%.*]]
-; ALL:       loop.header:
-; ALL-NEXT:    [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP_HEADER]] ]
-; ALL-NEXT:    [[TOBOOL:%.*]] = icmp eq i32 [[X:%.*]], 0
-; ALL-NEXT:    [[XOR:%.*]] = xor i32 [[X]], -1
-; ALL-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[XOR]], i1 true)
-; ALL-NEXT:    [[CAST:%.*]] = trunc i32 [[TMP0]] to i16
-; ALL-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i16 32, i16 [[CAST]]
-; ALL-NEXT:    store i16 [[COND]], i16* [[PTR:%.*]], align 2
-; ALL-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
-; ALL-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV]], 100
-; ALL-NEXT:    br i1 [[EC]], label [[LOOP_EXIT:%.*]], label [[LOOP_HEADER]]
-; ALL:       loop.exit:
-; ALL-NEXT:    ret i16 [[COND]]
+; BMI-LABEL: @test9_loop(
+; BMI-NEXT:  entry:
+; BMI-NEXT:    br label [[LOOP_HEADER:%.*]]
+; BMI:       loop.header:
+; BMI-NEXT:    [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP_HEADER]] ]
+; BMI-NEXT:    [[TOBOOL:%.*]] = icmp eq i32 [[X:%.*]], 0
+; BMI-NEXT:    [[XOR:%.*]] = xor i32 [[X]], -1
+; BMI-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[XOR]], i1 true)
+; BMI-NEXT:    [[CAST:%.*]] = trunc i32 [[TMP0]] to i16
+; BMI-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i16 32, i16 [[CAST]]
+; BMI-NEXT:    store i16 [[COND]], i16* [[PTR:%.*]], align 2
+; BMI-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
+; BMI-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV]], 100
+; BMI-NEXT:    br i1 [[EC]], label [[LOOP_EXIT:%.*]], label [[LOOP_HEADER]]
+; BMI:       loop.exit:
+; BMI-NEXT:    ret i16 [[COND]]
+;
+; LZCNT-LABEL: @test9_loop(
+; LZCNT-NEXT:  entry:
+; LZCNT-NEXT:    br label [[LOOP_HEADER:%.*]]
+; LZCNT:       loop.header:
+; LZCNT-NEXT:    [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[COND_END:%.*]] ]
+; LZCNT-NEXT:    [[TOBOOL:%.*]] = icmp eq i32 [[X:%.*]], 0
+; LZCNT-NEXT:    br i1 [[TOBOOL]], label [[COND_END]], label [[COND_TRUE:%.*]]
+; LZCNT:       cond.true:
+; LZCNT-NEXT:    [[XOR:%.*]] = xor i32 [[X]], -1
+; LZCNT-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[XOR]], i1 true)
+; LZCNT-NEXT:    [[CAST:%.*]] = trunc i32 [[TMP0]] to i16
+; LZCNT-NEXT:    br label [[COND_END]]
+; LZCNT:       cond.end:
+; LZCNT-NEXT:    [[COND:%.*]] = phi i16 [ [[CAST]], [[COND_TRUE]] ], [ 32, [[LOOP_HEADER]] ]
+; LZCNT-NEXT:    store i16 [[COND]], i16* [[PTR:%.*]], align 2
+; LZCNT-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
+; LZCNT-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV]], 100
+; LZCNT-NEXT:    br i1 [[EC]], label [[LOOP_EXIT:%.*]], label [[LOOP_HEADER]]
+; LZCNT:       loop.exit:
+; LZCNT-NEXT:    ret i16 [[COND]]
+;
+; GENERIC-LABEL: @test9_loop(
+; GENERIC-NEXT:  entry:
+; GENERIC-NEXT:    br label [[LOOP_HEADER:%.*]]
+; GENERIC:       loop.header:
+; GENERIC-NEXT:    [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[COND_END:%.*]] ]
+; GENERIC-NEXT:    [[TOBOOL:%.*]] = icmp eq i32 [[X:%.*]], 0
+; GENERIC-NEXT:    br i1 [[TOBOOL]], label [[COND_END]], label [[COND_TRUE:%.*]]
+; GENERIC:       cond.true:
+; GENERIC-NEXT:    [[XOR:%.*]] = xor i32 [[X]], -1
+; GENERIC-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[XOR]], i1 true)
+; GENERIC-NEXT:    [[CAST:%.*]] = trunc i32 [[TMP0]] to i16
+; GENERIC-NEXT:    br label [[COND_END]]
+; GENERIC:       cond.end:
+; GENERIC-NEXT:    [[COND:%.*]] = phi i16 [ [[CAST]], [[COND_TRUE]] ], [ 32, [[LOOP_HEADER]] ]
+; GENERIC-NEXT:    store i16 [[COND]], i16* [[PTR:%.*]], align 2
+; GENERIC-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
+; GENERIC-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV]], 100
+; GENERIC-NEXT:    br i1 [[EC]], label [[LOOP_EXIT:%.*]], label [[LOOP_HEADER]]
+; GENERIC:       loop.exit:
+; GENERIC-NEXT:    ret i16 [[COND]]
 ;
 entry:
   br label %loop.header