Index: lib/Target/AMDGPU/SIInstrInfo.cpp
===================================================================
--- lib/Target/AMDGPU/SIInstrInfo.cpp
+++ lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -2539,6 +2539,7 @@
                                                  LiveVariables *LV) const {
   unsigned Opc = MI.getOpcode();
   bool IsF16 = false;
+  bool ConstantBus = false;
   bool IsFMA = Opc == AMDGPU::V_FMAC_F32_e32 || Opc == AMDGPU::V_FMAC_F32_e64 ||
                Opc == AMDGPU::V_FMAC_F16_e32 || Opc == AMDGPU::V_FMAC_F16_e64;
 
@@ -2582,11 +2583,11 @@
   const MachineOperand *Clamp = getNamedOperand(MI, AMDGPU::OpName::clamp);
   const MachineOperand *Omod = getNamedOperand(MI, AMDGPU::OpName::omod);
 
-  if (!Src0Mods && !Src1Mods && !Clamp && !Omod &&
+  if ((ConstantBus = !Src0Mods && !Src1Mods && !Clamp && !Omod &&
       // If we have an SGPR input, we will violate the constant bus restriction.
       (ST.getConstantBusLimit(Opc) > 1 ||
        !Src0->isReg() ||
-       !RI.isSGPRReg(MBB->getParent()->getRegInfo(), Src0->getReg()))) {
+       !RI.isSGPRReg(MBB->getParent()->getRegInfo(), Src0->getReg())))) {
     if (auto Imm = getFoldableImm(Src2)) {
       unsigned NewOpc =
          IsFMA ? (IsF16 ? AMDGPU::V_FMAAK_F16 : AMDGPU::V_FMAAK_F32)
@@ -2623,6 +2624,19 @@
 
   unsigned NewOpc = IsFMA ? (IsF16 ? AMDGPU::V_FMA_F16 : AMDGPU::V_FMA_F32)
                           : (IsF16 ? AMDGPU::V_MAD_F16 : AMDGPU::V_MAD_F32);
+
+  // If the source of the tied reg is a sgpr->vgpr copy then replace
+  // it back with the sgpr, to potentially kill the copy, if
+  // this doesn't violate the constant bus restriction
+  if (ConstantBus) {
+    const MachineFunction *MF = Src2->getParent()->getMF();
+    const MachineRegisterInfo *MRI = &MF->getRegInfo();
+    if (auto *Def = MRI->getUniqueVRegDef(Src2->getReg())) {
+      if (Def->getOpcode() == AMDGPU::COPY)
+        Src2 = &Def->getOperand(1);
+    }
+  }
+
   if (pseudoToMCOpcode(NewOpc) == -1)
     return nullptr;
 
Index: test/CodeGen/AMDGPU/fmac-fma-sgpr-copy.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AMDGPU/fmac-fma-sgpr-copy.ll
@@ -0,0 +1,19 @@
+; RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck %s
+
+; CHECK: v_fma_f32 v0, v1, v0, s0
+define amdgpu_cs float @test1(<4 x i32> inreg %a, float %b, float %y) {
+entry:
+  %buf.load = call <4 x i32> @llvm.amdgcn.s.buffer.load.v4i32(<4 x i32> %a, i32 0, i32 0)
+  %vec1 = bitcast <4 x i32> %buf.load to <4 x float>
+  %.i095 = extractelement <4 x float> %vec1, i32 0
+  %.i098 = fsub nnan arcp float %b, %.i095
+  %fma1 = call float @llvm.fma.f32(float %y, float %.i098, float %.i095) #3
+  ret float %fma1
+}
+
+declare <4 x i32> @llvm.amdgcn.s.buffer.load.v4i32(<4 x i32>, i32, i32 immarg) #2
+declare float @llvm.fma.f32(float, float, float) #1
+
+attributes #1 = { nounwind readnone speculatable willreturn }
+attributes #2 = { nounwind readnone }
+attributes #3 = { nounwind }