AMDGPU: Make m0 unallocatable

m0 may need to be written for spill code, so
we don't want general code uses relying on the
value stored in it.

This introduces a few code quality regressions where copies
from m0 are not coalesced into copies of a copy of m0.

llvm-svn: 287841

6 files changed, 23 insertions, 16 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
index 4d55a663d48..bf493c9fd38 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
@@ -253,7 +253,7 @@ SDNode *AMDGPUDAGToDAGISel::glueCopyToM0(SDNode *N) const {
 static unsigned selectSGPRVectorRegClassID(unsigned NumVectorElts) {
   switch (NumVectorElts) {
   case 1:
-    return AMDGPU::SReg_32RegClassID;
+    return AMDGPU::SReg_32_XM0RegClassID;
   case 2:
     return AMDGPU::SReg_64RegClassID;
   case 4:
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 6deee51a064..04edc91f0a7 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -59,7 +59,7 @@ SITargetLowering::SITargetLowering(const TargetMachine &TM,
   addRegisterClass(MVT::i1, &AMDGPU::VReg_1RegClass);
   addRegisterClass(MVT::i64, &AMDGPU::SReg_64RegClass);
 
-  addRegisterClass(MVT::i32, &AMDGPU::SReg_32RegClass);
+  addRegisterClass(MVT::i32, &AMDGPU::SReg_32_XM0RegClass);
   addRegisterClass(MVT::f32, &AMDGPU::VGPR_32RegClass);
 
   addRegisterClass(MVT::f64, &AMDGPU::VReg_64RegClass);
@@ -79,8 +79,8 @@ SITargetLowering::SITargetLowering(const TargetMachine &TM,
   addRegisterClass(MVT::v16f32, &AMDGPU::VReg_512RegClass);
 
   if (Subtarget->has16BitInsts()) {
-    addRegisterClass(MVT::i16, &AMDGPU::SReg_32RegClass);
-    addRegisterClass(MVT::f16, &AMDGPU::SReg_32RegClass);
+    addRegisterClass(MVT::i16, &AMDGPU::SReg_32_XM0RegClass);
+    addRegisterClass(MVT::f16, &AMDGPU::SReg_32_XM0RegClass);
   }
 
   computeRegisterProperties(STI.getRegisterInfo());
@@ -941,25 +941,25 @@ SDValue SITargetLowering::LowerFormalArguments(
   // Start adding system SGPRs.
   if (Info->hasWorkGroupIDX()) {
     unsigned Reg = Info->addWorkGroupIDX();
-    MF.addLiveIn(Reg, &AMDGPU::SReg_32RegClass);
+    MF.addLiveIn(Reg, &AMDGPU::SReg_32_XM0RegClass);
     CCInfo.AllocateReg(Reg);
   }
 
   if (Info->hasWorkGroupIDY()) {
     unsigned Reg = Info->addWorkGroupIDY();
-    MF.addLiveIn(Reg, &AMDGPU::SReg_32RegClass);
+    MF.addLiveIn(Reg, &AMDGPU::SReg_32_XM0RegClass);
     CCInfo.AllocateReg(Reg);
   }
 
   if (Info->hasWorkGroupIDZ()) {
     unsigned Reg = Info->addWorkGroupIDZ();
-    MF.addLiveIn(Reg, &AMDGPU::SReg_32RegClass);
+    MF.addLiveIn(Reg, &AMDGPU::SReg_32_XM0RegClass);
     CCInfo.AllocateReg(Reg);
   }
 
   if (Info->hasWorkGroupInfo()) {
     unsigned Reg = Info->addWorkGroupInfo();
-    MF.addLiveIn(Reg, &AMDGPU::SReg_32RegClass);
+    MF.addLiveIn(Reg, &AMDGPU::SReg_32_XM0RegClass);
     CCInfo.AllocateReg(Reg);
   }
 
@@ -2414,15 +2414,15 @@ SDValue SITargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
                                   SI::KernelInputOffsets::LOCAL_SIZE_Z);
   case Intrinsic::amdgcn_workgroup_id_x:
   case Intrinsic::r600_read_tgid_x:
-    return CreateLiveInRegister(DAG, &AMDGPU::SReg_32RegClass,
+    return CreateLiveInRegister(DAG, &AMDGPU::SReg_32_XM0RegClass,
       TRI->getPreloadedValue(MF, SIRegisterInfo::WORKGROUP_ID_X), VT);
   case Intrinsic::amdgcn_workgroup_id_y:
   case Intrinsic::r600_read_tgid_y:
-    return CreateLiveInRegister(DAG, &AMDGPU::SReg_32RegClass,
+    return CreateLiveInRegister(DAG, &AMDGPU::SReg_32_XM0RegClass,
       TRI->getPreloadedValue(MF, SIRegisterInfo::WORKGROUP_ID_Y), VT);
   case Intrinsic::amdgcn_workgroup_id_z:
   case Intrinsic::r600_read_tgid_z:
-    return CreateLiveInRegister(DAG, &AMDGPU::SReg_32RegClass,
+    return CreateLiveInRegister(DAG, &AMDGPU::SReg_32_XM0RegClass,
       TRI->getPreloadedValue(MF, SIRegisterInfo::WORKGROUP_ID_Z), VT);
   case Intrinsic::amdgcn_workitem_id_x:
   case Intrinsic::r600_read_tidig_x:
@@ -4182,7 +4182,7 @@ SITargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI,
       default:
         return std::make_pair(0U, nullptr);
       case 32:
-        return std::make_pair(0U, &AMDGPU::SReg_32RegClass);
+        return std::make_pair(0U, &AMDGPU::SReg_32_XM0RegClass);
       case 64:
         return std::make_pair(0U, &AMDGPU::SGPR_64RegClass);
       case 128:
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index e9fbde16e2a..85eca558f25 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -364,7 +364,8 @@ void SIInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
     return;
   }
 
-  if (RC == &AMDGPU::SReg_32RegClass) {
+  if (RC == &AMDGPU::SReg_32_XM0RegClass ||
+      RC == &AMDGPU::SReg_32RegClass) {
     if (SrcReg == AMDGPU::SCC) {
       BuildMI(MBB, MI, DL, get(AMDGPU::S_CSELECT_B32), DestReg)
           .addImm(-1)
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
index fb42130767a..37bda5337f4 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -233,7 +233,7 @@ void SIRegisterInfo::materializeFrameBaseRegister(MachineBasicBlock *MBB,
 
   MachineRegisterInfo &MRI = MF->getRegInfo();
   unsigned UnusedCarry = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-  unsigned OffsetReg = MRI.createVirtualRegister(&AMDGPU::SReg_32RegClass);
+  unsigned OffsetReg = MRI.createVirtualRegister(&AMDGPU::SReg_32_XM0RegClass);
 
   BuildMI(*MBB, Ins, DL, TII->get(AMDGPU::S_MOV_B32), OffsetReg)
     .addImm(Offset);
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
index d1907d16aba..5bdd8bec54e 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
@@ -120,6 +120,11 @@ def SCC_CLASS : RegisterClass<"AMDGPU", [i1], 1, (add SCC)> {
   let isAllocatable = 0;
 }
 
+def M0_CLASS : RegisterClass<"AMDGPU", [i32], 32, (add M0)> {
+  let CopyCost = 1;
+  let isAllocatable = 0;
+}
+
 // TODO: Do we need to set DwarfRegAlias on register tuples?
 
 // SGPR 32-bit registers
@@ -259,8 +264,9 @@ def SReg_32_XM0 : RegisterClass<"AMDGPU", [i32, f32, i16, f16], 32,
 
 // Register class for all scalar registers (SGPRs + Special Registers)
 def SReg_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16], 32,
-  (add SReg_32_XM0, M0, VCC_LO, VCC_HI, EXEC_LO, EXEC_HI, FLAT_SCR_LO, FLAT_SCR_HI)> {
+  (add SReg_32_XM0, M0_CLASS)> {
   let AllocationPriority = 1;
+  let isAllocatable = 0;
 }
 
 def SGPR_64 : RegisterClass<"AMDGPU", [v2i32, i64, f64], 32, (add SGPR_64Regs)> {
diff --git a/llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp b/llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp
index 21518389b8b..76b0b4573fc 100644
--- a/llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp
+++ b/llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp
@@ -437,7 +437,7 @@ bool SIWholeQuadMode::requiresCorrectState(const MachineInstr &MI) const {
 MachineBasicBlock::iterator
 SIWholeQuadMode::saveSCC(MachineBasicBlock &MBB,
                          MachineBasicBlock::iterator Before) {
-  unsigned SaveReg = MRI->createVirtualRegister(&AMDGPU::SReg_32RegClass);
+  unsigned SaveReg = MRI->createVirtualRegister(&AMDGPU::SReg_32_XM0RegClass);
 
   MachineInstr *Save =
       BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::COPY), SaveReg)