diff --git a/llvm/tools/llvm-profgen/PerfReader.cpp b/llvm/tools/llvm-profgen/PerfReader.cpp
--- a/llvm/tools/llvm-profgen/PerfReader.cpp
+++ b/llvm/tools/llvm-profgen/PerfReader.cpp
@@ -18,6 +18,11 @@
                                         cl::ZeroOrMore,
                                         cl::desc("Print unwinder output"));
 
+static cl::opt<int> CSProfProbeStackDepth(
+    "csprof-probe-stack-depth", cl::init(-1), cl::ZeroOrMore,
+    cl::desc("Cut off probe stack from the bottom and keep it up to a given "
+             "depth. No cut is made if the input is -1."));
+
 extern cl::opt<bool> ShowDisassemblyOnly;
 extern cl::opt<bool> ShowSourceLocations;
 
@@ -104,11 +109,18 @@
 std::shared_ptr<ProbeBasedCtxKey> ProbeStack::getContextKey() {
   std::shared_ptr<ProbeBasedCtxKey> ProbeBasedKey =
       std::make_shared<ProbeBasedCtxKey>();
+  auto &P = ProbeBasedKey->Probes;
   for (auto CallProbe : Stack) {
-    ProbeBasedKey->Probes.emplace_back(CallProbe);
+    P.emplace_back(CallProbe);
+  }
+  CSProfileGenerator::compressRecursionContext<const MCDecodedPseudoProbe *>(P);
+  if (CSProfProbeStackDepth >= 0 &&
+      static_cast<size_t>(CSProfProbeStackDepth) < P.size()) {
+    std::copy(P.begin() + P.size() - static_cast<size_t>(CSProfProbeStackDepth),
+              P.end(), P.begin());
+    P.resize(CSProfProbeStackDepth);
   }
-  CSProfileGenerator::compressRecursionContext<const MCDecodedPseudoProbe *>(
-      ProbeBasedKey->Probes);
+
   ProbeBasedKey->genHashCode();
   return ProbeBasedKey;
 }