Index: include/llvm/Target/TargetRegisterInfo.h
===================================================================
--- include/llvm/Target/TargetRegisterInfo.h
+++ include/llvm/Target/TargetRegisterInfo.h
@@ -813,6 +813,13 @@
     return false;
   }
 
+  /// Returns true if the target requires using the RegScavenger directly for
+  /// frame elimination despite using requiresFrameIndexScavenging.
+  virtual bool requiresFrameIndexReplacementScavenging(
+    const MachineFunction &MF) const {
+    return false;
+  }
+
   /// Returns true if the target wants the LocalStackAllocation pass to be run
   /// and virtual base registers used for more efficient stack access.
   virtual bool requiresVirtualBaseRegisters(const MachineFunction &MF) const {
Index: lib/CodeGen/PrologEpilogInserter.cpp
===================================================================
--- lib/CodeGen/PrologEpilogInserter.cpp
+++ lib/CodeGen/PrologEpilogInserter.cpp
@@ -117,6 +117,10 @@
   // TRI->requiresFrameIndexScavenging() for the current function.
   bool FrameIndexVirtualScavenging;
 
+  // Flag to control whether the scavenger should be passed even though
+  // FrameIndexVirtualScavenging is used.
+  bool FrameIndexEliminationScavenging;
+
   void calculateCallFrameInfo(MachineFunction &Fn);
   void calculateSaveRestoreBlocks(MachineFunction &Fn);
 
@@ -176,6 +180,8 @@
 
   RS = TRI->requiresRegisterScavenging(Fn) ? new RegScavenger() : nullptr;
   FrameIndexVirtualScavenging = TRI->requiresFrameIndexScavenging(Fn);
+  FrameIndexEliminationScavenging
+    = TRI->requiresFrameIndexReplacementScavenging(Fn);
 
   // Calculate the MaxCallFrameSize and AdjustsStack variables for the
   // function's frame information. Also eliminates call frame pseudo
@@ -1046,7 +1052,8 @@
   unsigned FrameSetupOpcode = TII.getCallFrameSetupOpcode();
   unsigned FrameDestroyOpcode = TII.getCallFrameDestroyOpcode();
 
-  if (RS && !FrameIndexVirtualScavenging) RS->enterBasicBlock(*BB);
+  if (RS && FrameIndexEliminationScavenging)
+    RS->enterBasicBlock(*BB);
 
   bool InsideCallSequence = false;
 
@@ -1115,7 +1122,7 @@
       // use that target machine register info object to eliminate
       // it.
       TRI.eliminateFrameIndex(MI, SPAdj, i,
-                              FrameIndexVirtualScavenging ?  nullptr : RS);
+                              FrameIndexEliminationScavenging ?  RS : nullptr);
 
       // Reset the iterator if we were at the beginning of the BB.
       if (AtBeginning) {
@@ -1131,7 +1138,7 @@
     // the SP adjustment made by each instruction in the sequence.
     // This includes both the frame setup/destroy pseudos (handled above),
     // as well as other instructions that have side effects w.r.t the SP.
-    // Note that this must come after eliminateFrameIndex, because 
+    // Note that this must come after eliminateFrameIndex, because
     // if I itself referred to a frame index, we shouldn't count its own
     // adjustment.
     if (DidFinishLoop && InsideCallSequence)
@@ -1140,7 +1147,7 @@
     if (DoIncr && I != BB->end()) ++I;
 
     // Update register states.
-    if (RS && !FrameIndexVirtualScavenging && DidFinishLoop)
+    if (RS && FrameIndexEliminationScavenging && DidFinishLoop)
       RS->forward(MI);
   }
 }
Index: lib/Target/AMDGPU/SIRegisterInfo.h
===================================================================
--- lib/Target/AMDGPU/SIRegisterInfo.h
+++ lib/Target/AMDGPU/SIRegisterInfo.h
@@ -50,8 +50,9 @@
 
   bool requiresRegisterScavenging(const MachineFunction &Fn) const override;
 
-
   bool requiresFrameIndexScavenging(const MachineFunction &MF) const override;
+  bool requiresFrameIndexReplacementScavenging(
+    const MachineFunction &MF) const override;
   bool requiresVirtualBaseRegisters(const MachineFunction &Fn) const override;
   bool trackLivenessAfterRegAlloc(const MachineFunction &MF) const override;
 
Index: lib/Target/AMDGPU/SIRegisterInfo.cpp
===================================================================
--- lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -182,6 +182,16 @@
   return MF.getFrameInfo().hasStackObjects();
 }
 
+bool SIRegisterInfo::requiresFrameIndexReplacementScavenging(
+  const MachineFunction &MF) const {
+  // m0 is needed for the scalar store offset. m0 is unallocatable, so we can't
+  // create a virtual register for it during frame index elimination, so the
+  // scavenger is directly needed.
+  return MF.getFrameInfo().hasStackObjects() &&
+         MF.getSubtarget<SISubtarget>().hasScalarStores() &&
+         MF.getInfo<SIMachineFunctionInfo>()->hasSpilledSGPRs();
+}
+
 bool SIRegisterInfo::requiresVirtualBaseRegisters(
   const MachineFunction &) const {
   // There are no special dedicated stack or frame pointers.