[AArch64] Use SUBXrx64 for dynamic stack to refer to sp

When we lower dynamic stack, we need to substract pattern `x15 << 4`  from sp.
Subtract instruction with arith shifted register(SUBXrs) can't refer to sp. So for now we need two extra mov like:

```
mov x0, sp
sub x0, x0, x15, lsl #4
mov sp, x0
```
If we want to refer to sp in subtract instruction like this:
```
sub	sp, sp, x15, lsl #4
```
We must use arith extended register version(SUBXrx).
So in this patch when we find sub have sp operand on src0, try to select to SubXrx64.

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D129932

llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

@@ -69,6 +69,7 @@ public:
   bool tryMLAV64LaneV128(SDNode *N);
   bool tryMULLV64LaneV128(unsigned IntNo, SDNode *N);
   bool SelectArithExtendedRegister(SDValue N, SDValue &Reg, SDValue &Shift);
+  bool SelectArithUXTXRegister(SDValue N, SDValue &Reg, SDValue &Shift);
   bool SelectArithImmed(SDValue N, SDValue &Val, SDValue &Shift);
   bool SelectNegArithImmed(SDValue N, SDValue &Val, SDValue &Shift);
   bool SelectArithShiftedRegister(SDValue N, SDValue &Reg, SDValue &Shift) {
@@ -893,6 +894,30 @@ bool AArch64DAGToDAGISel::SelectArithExtendedRegister(SDValue N, SDValue &Reg,
   return isWorthFolding(N);
 }
 
+/// SelectArithUXTXRegister - Select a "UXTX register" operand. This
+/// operand is refered by the instructions have SP operand
+bool AArch64DAGToDAGISel::SelectArithUXTXRegister(SDValue N, SDValue &Reg,
+                                                  SDValue &Shift) {
+  unsigned ShiftVal = 0;
+  AArch64_AM::ShiftExtendType Ext;
+
+  if (N.getOpcode() != ISD::SHL)
+    return false;
+
+  ConstantSDNode *CSD = dyn_cast<ConstantSDNode>(N.getOperand(1));
+  if (!CSD)
+    return false;
+  ShiftVal = CSD->getZExtValue();
+  if (ShiftVal > 4)
+    return false;
+
+  Ext = AArch64_AM::UXTX;
+  Reg = N.getOperand(0);
+  Shift = CurDAG->getTargetConstant(getArithExtendImm(Ext, ShiftVal), SDLoc(N),
+                                    MVT::i32);
+  return isWorthFolding(N);
+}
+
 /// If there's a use of this ADDlow that's not itself a load/store then we'll
 /// need to create a real ADD instruction from it anyway and there's no point in
 /// folding it into the mem op. Theoretically, it shouldn't matter, but there's

llvm/lib/Target/AArch64/AArch64InstrFormats.td

@@ -1168,6 +1168,8 @@ def gi_arith_extended_reg32to64_i64 :
     GIComplexOperandMatcher<s64, "selectArithExtendedRegister">,
     GIComplexPatternEquiv<arith_extended_reg32to64_i64>;
 
+def arith_uxtx : ComplexPattern<i64, 2, "SelectArithUXTXRegister", []>;
+
 // Floating-point immediate.
 
 def fpimm16XForm : SDNodeXForm<fpimm, [{

llvm/lib/Target/AArch64/AArch64InstrInfo.td

@@ -1691,6 +1691,11 @@ def : InstAlias<"mov $dst, $src",
 defm ADDS : AddSubS<0, "adds", AArch64add_flag, "cmn", "subs", "cmp">;
 defm SUBS : AddSubS<1, "subs", AArch64sub_flag, "cmp", "adds", "cmn">;
 
+def copyFromSP: PatLeaf<(i64 GPR64:$src), [{
+  return N->getOpcode() == ISD::CopyFromReg &&
+         cast<RegisterSDNode>(N->getOperand(1))->getReg() == AArch64::SP;
+}]>;
+
 // Use SUBS instead of SUB to enable CSE between SUBS and SUB.
 def : Pat<(sub GPR32sp:$Rn, addsub_shifted_imm32:$imm),
           (SUBSWri GPR32sp:$Rn, addsub_shifted_imm32:$imm)>;
@@ -1709,6 +1714,8 @@ def : Pat<(sub GPR32sp:$R2, arith_extended_reg32_i32:$R3),
           (SUBSWrx GPR32sp:$R2, arith_extended_reg32_i32:$R3)>;
 def : Pat<(sub GPR64sp:$R2, arith_extended_reg32to64_i64:$R3),
           (SUBSXrx GPR64sp:$R2, arith_extended_reg32to64_i64:$R3)>;
+def : Pat<(sub copyFromSP:$R2, (arith_uxtx GPR64:$R3, arith_extendlsl64:$imm)),
+          (SUBXrx64 GPR64sp:$R2, GPR64:$R3, arith_extendlsl64:$imm)>;
 }
 
 // Because of the immediate format for add/sub-imm instructions, the

llvm/test/CodeGen/AArch64/win-alloca.ll

@@ -17,7 +17,6 @@ declare void @func2(i8*)
 ; CHECK: add [[REG1:x[0-9]+]], x0, #15
 ; CHECK-OPT: lsr x15, [[REG1]], #4
 ; CHECK: bl __chkstk
-; CHECK: mov [[REG2:x[0-9]+]], sp
-; CHECK-OPT: sub [[REG3:x[0-9]+]], [[REG2]], x15, lsl #4
+; CHECK-OPT: sub [[REG3:x[0-9]+]], sp, x15, lsl #4
 ; CHECK-OPT: mov sp, [[REG3]]
 ; CHECK: bl func2

llvm/test/CodeGen/AArch64/win64_vararg.ll

@@ -197,8 +197,7 @@ define void @vla(i32, i8*, ...) local_unnamed_addr {
 ; CHECK-NEXT:    stp x5, x6, [x29, #48]
 ; CHECK-NEXT:    str x7, [x29, #64]
 ; CHECK-NEXT:    bl __chkstk
-; CHECK-NEXT:    mov x8, sp
-; CHECK-NEXT:    sub x20, x8, x15, lsl #4
+; CHECK-NEXT:    sub x20, sp, x15, lsl #4
 ; CHECK-NEXT:    mov sp, x20
 ; CHECK-NEXT:    ldr x21, [x29, #16]
 ; CHECK-NEXT:    sxtw x22, w0