[llvm] [AArch64][SME] Remove unused ZA lazy-save (PR #81648)

[Sander de Smalen via llvm-commits]

================
@@ -2929,6 +2931,82 @@ AArch64TargetLowering::EmitZero(MachineInstr &MI, MachineBasicBlock *BB) const {
   return BB;
 }
 
+MachineBasicBlock *
+AArch64TargetLowering::EmitInitTPIDR2Object(MachineInstr &MI,
+                                            MachineBasicBlock *BB) const {
+  MachineFunction *MF = BB->getParent();
+  MachineFrameInfo &MFI = MF->getFrameInfo();
+  AArch64FunctionInfo *FuncInfo = MF->getInfo<AArch64FunctionInfo>();
+  TPIDR2Object &TPIDR2 = FuncInfo->getTPIDR2Obj();
+  if (TPIDR2.Uses > 0) {
+    const TargetInstrInfo *TII = Subtarget->getInstrInfo();
+    // Store the buffer pointer to the TPIDR2 stack object.
+    BuildMI(*BB, MI, MI.getDebugLoc(), TII->get(AArch64::STRXui))
+        .addReg(MI.getOperand(0).getReg())
+        .addFrameIndex(TPIDR2.FrameIndex)
+        .addImm(0);
+    // Set the reserved bytes (10-15) to zero
+    BuildMI(*BB, MI, MI.getDebugLoc(), TII->get(AArch64::STRHHui))
+        .addReg(AArch64::WZR)
+        .addFrameIndex(TPIDR2.FrameIndex)
+        .addImm(5);
+    BuildMI(*BB, MI, MI.getDebugLoc(), TII->get(AArch64::STRWui))
+        .addReg(AArch64::WZR)
+        .addFrameIndex(TPIDR2.FrameIndex)
+        .addImm(3);
+  } else
+    MFI.RemoveStackObject(TPIDR2.FrameIndex);
+
+  BB->remove_instr(&MI);
+  return BB;
+}
+
+MachineBasicBlock *
+AArch64TargetLowering::EmitExpandZABuffer(MachineInstr &MI,
+                                          MachineBasicBlock *BB) const {
+  MachineFunction *MF = BB->getParent();
+  MachineFrameInfo &MFI = MF->getFrameInfo();
+  AArch64FunctionInfo *FuncInfo = MF->getInfo<AArch64FunctionInfo>();
+  // TODO This function grows the stack with a subtraction, which doesn't work
+  // on Windows. Some refactoring to share the functionality in
+  // LowerWindowsDYNAMIC_STACKALLOC will be required once the Windows ABI
+  // supports SME
+  assert(!MF->getSubtarget<AArch64Subtarget>().isTargetWindows() &&
+         "Lazy ZA save is not yet supported on Windows");
+
+  TPIDR2Object &TPIDR2 = FuncInfo->getTPIDR2Obj();
+
+  if (TPIDR2.Uses > 0) {
+    const TargetInstrInfo *TII = Subtarget->getInstrInfo();
+    MachineRegisterInfo &MRI = MF->getRegInfo();
+
+    // The SUBXrs below won't always be emitted in a form that accepts SP
+    // directly
+    Register SP = MRI.createVirtualRegister(&AArch64::GPR64RegClass);
+    BuildMI(*BB, MI, MI.getDebugLoc(), TII->get(TargetOpcode::COPY), SP)
+        .addReg(AArch64::SP);
+
+    // Allocate a lazy-save buffer object of the size given, normally SVL * SVL
+    Register BufferAddr = MRI.createVirtualRegister(&AArch64::GPR64RegClass);
+    BuildMI(*BB, MI, MI.getDebugLoc(), TII->get(AArch64::SUBXrs), BufferAddr)
+        .addReg(SP)
+        .add(MI.getOperand(1))
+        .addImm(0);
+    BuildMI(*BB, MI, MI.getDebugLoc(), TII->get(TargetOpcode::COPY),
+            AArch64::SP)
+        .addReg(BufferAddr);
+    BuildMI(*BB, MI, MI.getDebugLoc(), TII->get(TargetOpcode::COPY),
+            MI.getOperand(0).getReg())
+        .addReg(BufferAddr);
----------------
sdesmalen-arm wrote:

`BufferAddr` and this COPY are superfluous, since you could have stored the result of the SUBXrs directly to `MI.getOperand(0).getReg()` (the destination reg)

https://github.com/llvm/llvm-project/pull/81648