[llvm] Add ARM:r12 and ARM:LR as scratch regs for ARM (PR #82454)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-arm

Author: AtariDreams (AtariDreams)

<details>
<summary>Changes</summary>

r12 is the intra-procedural scratch register and lr is clobbered across every call.

---
Full diff: https://github.com/llvm/llvm-project/pull/82454.diff


2 Files Affected:

- (modified) llvm/lib/Target/ARM/ARMISelLowering.cpp (+8) 
- (modified) llvm/lib/Target/ARM/ARMISelLowering.h (+2) 


``````````diff
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index b98006ed0cb3f4..91b3b224aa1679 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -13789,6 +13789,14 @@ static SDValue PerformADDVecReduce(SDNode *N, SelectionDAG &DAG,
   return SDValue();
 }
 
+const MCPhysReg *getScratchRegisters(CallingConv::ID) const {
+  // LR is a not a scratch register, but we must treat it as clobbered by any
+  // call site. Hence we include LR in the scratch registers, which are in turn
+  // added as implicit-defs for stackmaps and patchpoints.
+  static const MCPhysReg ScratchRegs[] = {ARM::R12, ARM::LR, 0};
+  return ScratchRegs;
+}
+
 bool
 ARMTargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
                                                  CombineLevel Level) const {
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.h b/llvm/lib/Target/ARM/ARMISelLowering.h
index b13ddf697cb806..bca55b5c95a235 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.h
+++ b/llvm/lib/Target/ARM/ARMISelLowering.h
@@ -738,6 +738,8 @@ class VectorType;
     Align getABIAlignmentForCallingConv(Type *ArgTy,
                                         const DataLayout &DL) const override;
 
+    const MCPhysReg *getScratchRegisters(CallingConv::ID CC) const override;
+
     bool isDesirableToCommuteWithShift(const SDNode *N,
                                        CombineLevel Level) const override;
 

``````````

</details>


https://github.com/llvm/llvm-project/pull/82454