diff --git a/llvm/lib/Transforms/IPO/HotColdSplitting.cpp b/llvm/lib/Transforms/IPO/HotColdSplitting.cpp
--- a/llvm/lib/Transforms/IPO/HotColdSplitting.cpp
+++ b/llvm/lib/Transforms/IPO/HotColdSplitting.cpp
@@ -119,6 +119,7 @@
   // in the case of a longjmp, if the block is cold according to
   // profile information, we mark it as unlikely to be executed as well.
   if (blockEndsInUnreachable(BB)) {
+    LLVM_DEBUG(dbgs() << "Block ends in unreachable: " << BB.getName() << "\n");
     if (auto *CI =
             dyn_cast_or_null<CallInst>(BB.getTerminator()->getPrevNode()))
       if (CI->hasFnAttr(Attribute::NoReturn)) {
@@ -126,6 +127,7 @@
           return (II->getIntrinsicID() != Intrinsic::eh_sjlj_longjmp) ||
                  (BFI && PSI->isColdBlock(&BB, BFI));
         return !CI->getCalledFunction()->getName().startswith("longjmp") ||
+               !CI->getCalledFunction()->getName().endswith("longjmp") ||
                (BFI && PSI->isColdBlock(&BB, BFI));
       }
     return true;
diff --git a/llvm/test/Transforms/HotColdSplit/outline-unless-longjmp.ll b/llvm/test/Transforms/HotColdSplit/outline-unless-longjmp.ll
deleted file mode 100644
--- a/llvm/test/Transforms/HotColdSplit/outline-unless-longjmp.ll
+++ /dev/null
@@ -1,66 +0,0 @@
-; REQUIRES: asserts
-; RUN: opt -S -hotcoldsplit -debug < %s 2>&1 | FileCheck %s
-; RUN: opt -hotcoldsplit %s -o /dev/null
-
-target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
-target triple = "x86_64-unknown-linux-gnu"
-
-%"class.std::ios_base::Init" = type { i8 }
-%struct.__jmp_buf_tag = type { [8 x i64], i32, %struct.__sigset_t }
-%struct.__sigset_t = type { [16 x i64] }
-
-@jbuf = dso_local global [1 x %struct.__jmp_buf_tag] zeroinitializer, align 16
-
-; CHECK: @f
-; CHECK-NOT: {{.*}}@f.cold.1
-
-
-declare dso_local i32 @__cxa_atexit(void (i8*)*, i8*, i8*) #3
-
-define dso_local void @_Z1ai(i32 %i) #4 {
-entry:
-  %i.addr = alloca i32, align 4
-  store i32 %i, i32* %i.addr, align 4
-  %0 = load i32, i32* %i.addr, align 4
-  %add = add nsw i32 %0, 1
-  call void @longjmp(%struct.__jmp_buf_tag* getelementptr inbounds ([1 x %struct.__jmp_buf_tag], [1 x %struct.__jmp_buf_tag]* @jbuf, i64 0, i64 0), i32 %add) #8
-  unreachable
-}
-
-declare dso_local void @longjmp(%struct.__jmp_buf_tag*, i32) #5
-
-define dso_local i32 @f() #6 {
-entry:
-  %retval = alloca i32, align 4
-  %i = alloca i32, align 4
-  store i32 0, i32* %retval, align 4
-  store volatile i32 0, i32* %i, align 4
-  %call = call i32 @_setjmp(%struct.__jmp_buf_tag* getelementptr inbounds ([1 x %struct.__jmp_buf_tag], [1 x %struct.__jmp_buf_tag]* @jbuf, i64 0, i64 0)) #9
-  %cmp = icmp ne i32 %call, 3
-  br i1 %cmp, label %if.then, label %if.end
-
-if.then:                                          ; preds = %entry
-  %0 = load volatile i32, i32* %i, align 4
-  %inc = add nsw i32 %0, 1
-  store volatile i32 %inc, i32* %i, align 4
-  call void @_Z1ai(i32 %0) #10
-  unreachable
-
-if.end:                                           ; preds = %entry
-  ret i32 0
-}
-
-; Function Attrs: nounwind returns_twice
-declare dso_local i32 @_setjmp(%struct.__jmp_buf_tag*) #7
-
-
-attributes #3 = { nounwind }
-attributes #4 = { noinline noreturn nounwind uwtable }
-attributes #5 = { noreturn nounwind }
-attributes #6 = { noinline norecurse nounwind uwtable }
-attributes #7 = { nounwind returns_twice }
-attributes #8 = { noreturn nounwind }
-attributes #9 = { nounwind returns_twice }
-attributes #10 = { noreturn }
-
-
diff --git a/llvm/test/Transforms/HotColdSplit/sjlj-nosplit.ll b/llvm/test/Transforms/HotColdSplit/sjlj-nosplit.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/Transforms/HotColdSplit/sjlj-nosplit.ll
@@ -0,0 +1,103 @@
+; RUN:  opt -hotcoldsplit -S < %s | FileCheck %s
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@c = dso_local global i32 1, align 4
+@buf = dso_local global [20 x i8*] zeroinitializer, align 16
+
+; CHECK-LABEL: @f
+; CHECK-NOT: f.cold.1
+define dso_local void @f() #0 {
+entry:
+  %i = alloca i32, align 4
+  %j = alloca i32, align 4
+  %k = alloca i32, align 4
+  %0 = load i32, i32* @c, align 4
+  %tobool = icmp ne i32 %0, 0
+  br i1 %tobool, label %if.then, label %if.else
+
+if.then:                                          ; preds = %entry
+  ret void
+
+if.else:                                          ; preds = %entry
+  %1 = load i32, i32* @c, align 4
+  %inc = add nsw i32 %1, 1
+  store i32 %inc, i32* @c, align 4
+  %2 = load i32, i32* @c, align 4
+  %inc1 = add nsw i32 %2, 1
+  store i32 %inc1, i32* @c, align 4
+  %3 = load i32, i32* @c, align 4
+  %inc2 = add nsw i32 %3, 1
+  store i32 %inc2, i32* @c, align 4
+  %4 = load i32, i32* @c, align 4
+  %inc3 = add nsw i32 %4, 1
+  store i32 %inc3, i32* @c, align 4
+  %5 = load i32, i32* @c, align 4
+  %dec = add nsw i32 %5, -1
+  store i32 %dec, i32* @c, align 4
+  %6 = load i32, i32* @c, align 4
+  %dec4 = add nsw i32 %6, -1
+  store i32 %dec4, i32* @c, align 4
+  %7 = load i32, i32* @c, align 4
+  %inc5 = add nsw i32 %7, 1
+  store i32 %inc5, i32* @c, align 4
+  %8 = load i32, i32* @c, align 4
+  %inc6 = add nsw i32 %8, 1
+  store i32 %inc6, i32* @c, align 4
+  %9 = load i32, i32* @c, align 4
+  %add = add nsw i32 %9, 1
+  store i32 %add, i32* %i, align 4
+  %10 = load i32, i32* %i, align 4
+  %sub = sub nsw i32 %10, 1
+  store i32 %sub, i32* %j, align 4
+  %11 = load i32, i32* %i, align 4
+  %add7 = add nsw i32 %11, 2
+  store i32 %add7, i32* %k, align 4
+  call void @llvm.eh.sjlj.longjmp(i8* bitcast ([20 x i8*]* @buf to i8*))
+  unreachable
+}
+
+declare void @llvm.eh.sjlj.longjmp(i8*) #1
+
+; CHECK-LABEL: @main
+; CHECK-NOT: main.cold.1
+define dso_local i32 @main() #0 {
+entry:
+  %retval = alloca i32, align 4
+  %i = alloca i32, align 4
+  store i32 0, i32* %retval, align 4
+  store i32 0, i32* %i, align 4
+  %0 = call i8* @llvm.frameaddress.p0i8(i32 0)
+  store i8* %0, i8** getelementptr inbounds ([20 x i8*], [20 x i8*]* @buf, i64 0, i64 0), align 16
+  %1 = call i8* @llvm.stacksave()
+  store i8* %1, i8** getelementptr inbounds ([20 x i8*], [20 x i8*]* @buf, i64 0, i64 2), align 16
+  %2 = call i32 @llvm.eh.sjlj.setjmp(i8* bitcast ([20 x i8*]* @buf to i8*))
+  %tobool = icmp ne i32 %2, 0
+  br i1 %tobool, label %if.then, label %if.end
+
+if.then:                                          ; preds = %entry
+  store i32 1, i32* %retval, align 4
+  br label %return
+
+if.end:                                           ; preds = %entry
+  call void @f()
+  store i32 0, i32* %retval, align 4
+  br label %return
+
+return:                                           ; preds = %if.end, %if.then
+  %3 = load i32, i32* %retval, align 4
+  ret i32 %3
+}
+
+declare i8* @llvm.frameaddress.p0i8(i32 immarg) #2
+
+declare i8* @llvm.stacksave() #3
+
+declare i32 @llvm.eh.sjlj.setjmp(i8*) #3
+
+attributes #0 = { nounwind uwtable }
+attributes #1 = { noreturn nounwind }
+attributes #2 = { nounwind readnone }
+attributes #3 = { nounwind }
+
+
diff --git a/llvm/test/Transforms/HotColdSplit/sjlj-split.ll b/llvm/test/Transforms/HotColdSplit/sjlj-split.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/Transforms/HotColdSplit/sjlj-split.ll
@@ -0,0 +1,136 @@
+; RUN:  opt -profile-summary-cold-count=0 -hotcoldsplit -S < %s | FileCheck %s
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@c = dso_local global i32 1, align 4
+@buf = dso_local global [20 x i8*] zeroinitializer, align 16
+
+; CHECK-LABEL: @f
+; CHECK: f.cold.1
+define dso_local void @f() #0 !prof !31 {
+entry:
+  %i = alloca i32, align 4
+  %j = alloca i32, align 4
+  %k = alloca i32, align 4
+  %0 = load i32, i32* @c, align 4
+  %tobool = icmp ne i32 %0, 0
+  br i1 %tobool, label %if.then, label %if.else, !prof !32
+
+if.then:                                          ; preds = %entry
+  ret void
+
+if.else:                                          ; preds = %entry
+  %1 = load i32, i32* @c, align 4
+  %inc = add  i32 %1, 1
+  store i32 %inc, i32* @c, align 4
+  %2 = load i32, i32* @c, align 4
+  %inc1 = add  i32 %2, 1
+  store i32 %inc1, i32* @c, align 4
+  %3 = load i32, i32* @c, align 4
+  %inc2 = add  i32 %3, 1
+  store i32 %inc2, i32* @c, align 4
+  %4 = load i32, i32* @c, align 4
+  %inc3 = add  i32 %4, 1
+  store i32 %inc3, i32* @c, align 4
+  %5 = load i32, i32* @c, align 4
+  %dec = add  i32 %5, -1
+  store i32 %dec, i32* @c, align 4
+  %6 = load i32, i32* @c, align 4
+  %dec4 = add  i32 %6, -1
+  store i32 %dec4, i32* @c, align 4
+  %7 = load i32, i32* @c, align 4
+  %inc5 = add  i32 %7, 1
+  store i32 %inc5, i32* @c, align 4
+  %8 = load i32, i32* @c, align 4
+  %inc6 = add  i32 %8, 1
+  store i32 %inc6, i32* @c, align 4
+  %9 = load i32, i32* @c, align 4
+  %add = add  i32 %9, 1
+  store i32 %add, i32* %i, align 4
+  %10 = load i32, i32* %i, align 4
+  %sub = sub  i32 %10, 1
+  store i32 %sub, i32* %j, align 4
+  %11 = load i32, i32* %i, align 4
+  %add7 = add  i32 %11, 2
+  store i32 %add7, i32* %k, align 4
+  call void @llvm.eh.sjlj.longjmp(i8* bitcast ([20 x i8*]* @buf to i8*))
+  unreachable
+}
+
+declare void @llvm.eh.sjlj.longjmp(i8*) #1
+
+define dso_local i32 @main() #0 !prof !31 {
+entry:
+  %retval = alloca i32, align 4
+  %i = alloca i32, align 4
+  store i32 0, i32* %retval, align 4
+  store i32 0, i32* %i, align 4
+  %0 = call i8* @llvm.frameaddress.p0i8(i32 0)
+  store i8* %0, i8** getelementptr inbounds ([20 x i8*], [20 x i8*]* @buf, i64 0, i64 0), align 16
+  %1 = call i8* @llvm.stacksave()
+  store i8* %1, i8** getelementptr inbounds ([20 x i8*], [20 x i8*]* @buf, i64 0, i64 2), align 16
+  %2 = call i32 @llvm.eh.sjlj.setjmp(i8* bitcast ([20 x i8*]* @buf to i8*))
+  %tobool = icmp ne i32 %2, 0
+  br i1 %tobool, label %if.then, label %if.end, !prof !33
+
+if.then:                                          ; preds = %entry
+  store i32 1, i32* %retval, align 4
+  br label %return
+
+if.end:                                           ; preds = %entry
+  call void @f()
+  store i32 0, i32* %retval, align 4
+  br label %return
+
+return:                                           ; preds = %if.end, %if.then
+  %3 = load i32, i32* %retval, align 4
+  ret i32 %3
+}
+
+declare i8* @llvm.frameaddress.p0i8(i32 immarg) #2
+
+declare i8* @llvm.stacksave() #3
+
+declare i32 @llvm.eh.sjlj.setjmp(i8*) #3
+
+attributes #0 = { inlinehint nounwind uwtable }
+attributes #1 = { noreturn nounwind }
+attributes #2 = { nounwind readnone }
+attributes #3 = { nounwind }
+
+!llvm.module.flags = !{!0, !1}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 1, !"ProfileSummary", !2}
+!2 = !{!3, !4, !5, !6, !7, !8, !9, !10, !11, !12}
+!3 = !{!"ProfileFormat", !"InstrProf"}
+!4 = !{!"TotalCount", i64 2}
+!5 = !{!"MaxCount", i64 1}
+!6 = !{!"MaxInternalCount", i64 1}
+!7 = !{!"MaxFunctionCount", i64 1}
+!8 = !{!"NumCounts", i64 4}
+!9 = !{!"NumFunctions", i64 2}
+!10 = !{!"IsPartialProfile", i64 0}
+!11 = !{!"PartialProfileRatio", double 0.000000e+00}
+!12 = !{!"DetailedSummary", !13}
+!13 = !{!14, !15, !16, !17, !18, !19, !20, !21, !22, !23, !24, !25, !26, !27, !28, !29}
+!14 = !{i32 10000, i64 0, i32 0}
+!15 = !{i32 100000, i64 0, i32 0}
+!16 = !{i32 200000, i64 0, i32 0}
+!17 = !{i32 300000, i64 0, i32 0}
+!18 = !{i32 400000, i64 0, i32 0}
+!19 = !{i32 500000, i64 1, i32 2}
+!20 = !{i32 600000, i64 1, i32 2}
+!21 = !{i32 700000, i64 1, i32 2}
+!22 = !{i32 800000, i64 1, i32 2}
+!23 = !{i32 900000, i64 1, i32 2}
+!24 = !{i32 950000, i64 1, i32 2}
+!25 = !{i32 990000, i64 1, i32 2}
+!26 = !{i32 999000, i64 1, i32 2}
+!27 = !{i32 999900, i64 1, i32 2}
+!28 = !{i32 999990, i64 1, i32 2}
+!29 = !{i32 999999, i64 1, i32 2}
+!31 = !{!"function_entry_count", i64 1}
+!32 = !{!"branch_weights", i32 1, i32 0}
+!33 = !{!"branch_weights", i32 0, i32 1}
+