diff --git a/llvm/lib/Target/AMDGPU/SILowerControlFlow.cpp b/llvm/lib/Target/AMDGPU/SILowerControlFlow.cpp
--- a/llvm/lib/Target/AMDGPU/SILowerControlFlow.cpp
+++ b/llvm/lib/Target/AMDGPU/SILowerControlFlow.cpp
@@ -74,6 +74,10 @@
 
 #define DEBUG_TYPE "si-lower-control-flow"
 
+static cl::opt<bool>
+RemoveRedundantEndcf("amdgpu-remove-redundant-endcf",
+    cl::init(false), cl::ReallyHidden);
+
 namespace {
 
 class SILowerControlFlow : public MachineFunctionPass {
@@ -444,14 +448,17 @@
 
   // If the only instruction immediately following this END_CF is an another
   // END_CF in the only successor we can avoid emitting exec mask restore here.
-  auto Next = skipIgnoreExecInstsTrivialSucc(MBB, std::next(MI.getIterator()));
-  if (Next != MBB.end() && (Next->getOpcode() == AMDGPU::SI_END_CF ||
-                            LoweredEndCf.count(&*Next))) {
-    LLVM_DEBUG(dbgs() << "Skip redundant "; MI.dump());
-    if (LIS)
-      LIS->RemoveMachineInstrFromMaps(MI);
-    MI.eraseFromParent();
-    return;
+  if (RemoveRedundantEndcf) {
+    auto Next =
+      skipIgnoreExecInstsTrivialSucc(MBB, std::next(MI.getIterator()));
+    if (Next != MBB.end() && (Next->getOpcode() == AMDGPU::SI_END_CF ||
+                              LoweredEndCf.count(&*Next))) {
+      LLVM_DEBUG(dbgs() << "Skip redundant "; MI.dump());
+      if (LIS)
+        LIS->RemoveMachineInstrFromMaps(MI);
+      MI.eraseFromParent();
+      return;
+    }
   }
 
   MachineBasicBlock::iterator InsPt =
diff --git a/llvm/test/CodeGen/AMDGPU/collapse-endcf.ll b/llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
--- a/llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
+++ b/llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -amdgpu-remove-redundant-endcf < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
 
 ; GCN-LABEL: {{^}}simple_nested_if:
 ; GCN:      s_and_saveexec_b64 [[SAVEEXEC:s\[[0-9:]+\]]]
diff --git a/llvm/test/CodeGen/AMDGPU/collapse-endcf.mir b/llvm/test/CodeGen/AMDGPU/collapse-endcf.mir
--- a/llvm/test/CodeGen/AMDGPU/collapse-endcf.mir
+++ b/llvm/test/CodeGen/AMDGPU/collapse-endcf.mir
@@ -1,5 +1,5 @@
 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -march=amdgcn -verify-machineinstrs -run-pass=si-lower-control-flow %s -o - | FileCheck -check-prefix=GCN %s
+# RUN: llc -march=amdgcn -verify-machineinstrs -run-pass=si-lower-control-flow -amdgpu-remove-redundant-endcf %s -o - | FileCheck -check-prefix=GCN %s
 
 # Make sure dbg_value doesn't change codeegn when collapsing end_cf
 ---
diff --git a/llvm/test/CodeGen/AMDGPU/mul24-pass-ordering.ll b/llvm/test/CodeGen/AMDGPU/mul24-pass-ordering.ll
--- a/llvm/test/CodeGen/AMDGPU/mul24-pass-ordering.ll
+++ b/llvm/test/CodeGen/AMDGPU/mul24-pass-ordering.ll
@@ -58,7 +58,7 @@
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v5
 ; GFX9-NEXT:    v_cmp_lt_u32_e64 s[4:5], v0, v1
 ; GFX9-NEXT:    s_and_saveexec_b64 s[10:11], s[4:5]
-; GFX9-NEXT:    s_cbranch_execz BB1_3
+; GFX9-NEXT:    s_cbranch_execz BB1_4
 ; GFX9-NEXT:  ; %bb.1: ; %bb19
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v7, v6
 ; GFX9-NEXT:    v_and_b32_e32 v5, 0xffffff, v6
@@ -100,7 +100,9 @@
 ; GFX9-NEXT:    v_add_u32_e32 v3, v3, v6
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[12:13]
 ; GFX9-NEXT:    s_cbranch_execnz BB1_2
-; GFX9-NEXT:  BB1_3: ; %Flow3
+; GFX9-NEXT:  ; %bb.3: ; %Flow
+; GFX9-NEXT:    s_or_b64 exec, exec, s[12:13]
+; GFX9-NEXT:  BB1_4: ; %Flow3
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[10:11]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]