diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 1169efce3123f..f9f52d628d8f3 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -153,6 +153,14 @@ cl::opt<bool> EnableSVEGISel(
     cl::desc("Enable / disable SVE scalable vectors in Global ISel"),
     cl::init(false));
 
+// TODO: This option should be removed once we switch to always using PTRADD in
+// the SelectionDAG.
+static cl::opt<bool> UseFEATCPACodegen(
+    "aarch64-use-featcpa-codegen", cl::Hidden,
+    cl::desc("Generate ISD::PTRADD nodes for pointer arithmetic in "
+             "SelectionDAG for FEAT_CPA"),
+    cl::init(false));
+
 /// Value type used for condition codes.
 static const MVT MVT_CC = MVT::i32;
 
@@ -30416,3 +30424,8 @@ bool AArch64TargetLowering::isTypeDesirableForOp(unsigned Opc, EVT VT) const {
 
   return TargetLowering::isTypeDesirableForOp(Opc, VT);
 }
+
+bool AArch64TargetLowering::shouldPreservePtrArith(const Function &F,
+                                                   EVT VT) const {
+  return Subtarget->hasCPA() && UseFEATCPACodegen;
+}
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.h b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
index e0b6c1b8c0baf..89f90ee2b7707 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
@@ -539,6 +539,10 @@ class AArch64TargetLowering : public TargetLowering {
   /// True if stack clash protection is enabled for this functions.
   bool hasInlineStackProbe(const MachineFunction &MF) const override;
 
+  /// In AArch64, true if FEAT_CPA is present. Allows pointer arithmetic
+  /// semantics to be preserved for instruction selection.
+  bool shouldPreservePtrArith(const Function &F, EVT PtrVT) const override;
+
 private:
   /// Keep a pointer to the AArch64Subtarget around so that we can
   /// make the right decision when generating code for different targets.
diff --git a/llvm/lib/Target/AArch64/AArch64InstrFormats.td b/llvm/lib/Target/AArch64/AArch64InstrFormats.td
index 9078748c14834..ba7cbccc0bcd6 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrFormats.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrFormats.td
@@ -13124,7 +13124,7 @@ def LSLImm3ShiftOperand : AsmOperandClass {
   let DiagnosticType = "AddSubLSLImm3ShiftLarge";
 }
 
-def lsl_imm3_shift_operand : Operand<i32> {
+def lsl_imm3_shift_operand : Operand<i64> {
   let PrintMethod = "printShifter";
   let ParserMatchClass = LSLImm3ShiftOperand;
 }
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.td b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
index f90f12b5ac3c7..47f88ee5f05b0 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
@@ -10716,6 +10716,21 @@ let Predicates = [HasCPA] in {
   // Scalar multiply-add/subtract
   def MADDPT : MulAccumCPA<0, "maddpt">;
   def MSUBPT : MulAccumCPA<1, "msubpt">;
+
+  def : Pat<(ptradd GPR64sp:$Rn, GPR64sp:$Rm),
+            (ADDPT_shift GPR64sp:$Rn, GPR64sp:$Rm, (i64 0))>;
+  def : Pat<(ptradd GPR64sp:$Rn, (shl GPR64sp:$Rm, (i64 imm0_7:$imm))),
+            (ADDPT_shift GPR64sp:$Rn, GPR64sp:$Rm,
+                         (i64 imm0_7:$imm))>;
+  def : Pat<(ptradd GPR64sp:$Rn, (ineg GPR64sp:$Rm)),
+            (SUBPT_shift GPR64sp:$Rn, GPR64sp:$Rm, (i64 0))>;
+  def : Pat<(ptradd GPR64sp:$Rn, (ineg (shl GPR64sp:$Rm, (i64 imm0_7:$imm)))),
+            (SUBPT_shift GPR64sp:$Rn, GPR64sp:$Rm,
+                         (i64 imm0_7:$imm))>;
+  def : Pat<(ptradd GPR64:$Ra, (mul GPR64:$Rn, GPR64:$Rm)),
+            (MADDPT GPR64:$Rn, GPR64:$Rm, GPR64:$Ra)>;
+  def : Pat<(ptradd GPR64:$Ra, (mul GPR64:$Rn, (ineg GPR64:$Rm))),
+            (MSUBPT GPR64:$Rn, GPR64:$Rm, GPR64:$Ra)>;
 }
 
 def round_v4fp32_to_v4bf16 :
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp b/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
index 51b42325ef842..5081cc4bba144 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
@@ -2197,8 +2197,14 @@ bool AArch64InstructionSelector::preISelLower(MachineInstr &I) {
     }
     return Changed;
   }
-  case TargetOpcode::G_PTR_ADD:
+  case TargetOpcode::G_PTR_ADD: {
+    // If Checked Pointer Arithmetic (FEAT_CPA) is present, preserve the pointer
+    // arithmetic semantics instead of falling back to regular arithmetic.
+    const auto &TL = STI.getTargetLowering();
+    if (TL->shouldPreservePtrArith(MF.getFunction(), EVT()))
+      return false;
     return convertPtrAddToAdd(I, MRI);
+  }
   case TargetOpcode::G_LOAD: {
     // For scalar loads of pointers, we try to convert the dest type from p0
     // to s64 so that our imported patterns can match. Like with the G_PTR_ADD
diff --git a/llvm/test/CodeGen/AArch64/cpa-globalisel.ll b/llvm/test/CodeGen/AArch64/cpa-globalisel.ll
new file mode 100644
index 0000000000000..c9b48b9685df3
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/cpa-globalisel.ll
@@ -0,0 +1,776 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=aarch64 -verify-machineinstrs --mattr=+cpa -aarch64-use-featcpa-codegen=true -O0 -global-isel=1 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK-CPA-O0
+; RUN: llc -mtriple=aarch64 -verify-machineinstrs --mattr=+cpa -aarch64-use-featcpa-codegen=true -O3 -global-isel=1 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK-CPA-O3
+; RUN: llc -mtriple=aarch64 -verify-machineinstrs --mattr=-cpa -O0 -global-isel=1 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK-NOCPA-O0
+; RUN: llc -mtriple=aarch64 -verify-machineinstrs --mattr=-cpa -O3 -global-isel=1 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK-NOCPA-O3
+
+%struct.my_type = type { i64, i64 }
+%struct.my_type2 = type { i64, i64, i64, i64, i64, i64 }
+
+@array = external dso_local global [10 x %struct.my_type], align 8
+@array2 = external dso_local global [10 x %struct.my_type2], align 8
+
+define void @addpt1(i64 %index, i64 %arg) {
+; CHECK-CPA-O0-LABEL: addpt1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    adrp x8, array
+; CHECK-CPA-O0-NEXT:    add x8, x8, :lo12:array
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x0, lsl #4
+; CHECK-CPA-O0-NEXT:    str x1, [x8, #8]
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: addpt1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    adrp x8, array
+; CHECK-CPA-O3-NEXT:    add x8, x8, :lo12:array
+; CHECK-CPA-O3-NEXT:    addpt x8, x8, x0, lsl #4
+; CHECK-CPA-O3-NEXT:    str x1, [x8, #8]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: addpt1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    adrp x8, array
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, :lo12:array
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, x0, lsl #4
+; CHECK-NOCPA-O0-NEXT:    str x1, [x8, #8]
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: addpt1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    adrp x8, array
+; CHECK-NOCPA-O3-NEXT:    add x8, x8, :lo12:array
+; CHECK-NOCPA-O3-NEXT:    add x8, x8, x0, lsl #4
+; CHECK-NOCPA-O3-NEXT:    str x1, [x8, #8]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %e2 = getelementptr inbounds %struct.my_type, ptr @array, i64 %index, i32 1
+  store i64 %arg, ptr %e2, align 8
+  ret void
+}
+
+define void @maddpt1(i32 %pos, ptr %val) {
+; CHECK-CPA-O0-LABEL: maddpt1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    // implicit-def: $x8
+; CHECK-CPA-O0-NEXT:    mov w8, w0
+; CHECK-CPA-O0-NEXT:    sxtw x8, w8
+; CHECK-CPA-O0-NEXT:    mov w9, #48 // =0x30
+; CHECK-CPA-O0-NEXT:    // kill: def $x9 killed $w9
+; CHECK-CPA-O0-NEXT:    adrp x10, array2
+; CHECK-CPA-O0-NEXT:    add x10, x10, :lo12:array2
+; CHECK-CPA-O0-NEXT:    maddpt x0, x8, x9, x10
+; CHECK-CPA-O0-NEXT:    mov w8, #48 // =0x30
+; CHECK-CPA-O0-NEXT:    mov w2, w8
+; CHECK-CPA-O0-NEXT:    b memcpy
+;
+; CHECK-CPA-O3-LABEL: maddpt1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-CPA-O3-NEXT:    sxtw x9, w0
+; CHECK-CPA-O3-NEXT:    mov w8, #48 // =0x30
+; CHECK-CPA-O3-NEXT:    ldr q0, [x1]
+; CHECK-CPA-O3-NEXT:    adrp x10, array2
+; CHECK-CPA-O3-NEXT:    add x10, x10, :lo12:array2
+; CHECK-CPA-O3-NEXT:    maddpt x8, x9, x8, x10
+; CHECK-CPA-O3-NEXT:    str q0, [x8]
+; CHECK-CPA-O3-NEXT:    ldr q0, [x1, #16]
+; CHECK-CPA-O3-NEXT:    str q0, [x8, #16]
+; CHECK-CPA-O3-NEXT:    ldr q0, [x1, #32]
+; CHECK-CPA-O3-NEXT:    str q0, [x8, #32]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: maddpt1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    adrp x9, array2
+; CHECK-NOCPA-O0-NEXT:    add x9, x9, :lo12:array2
+; CHECK-NOCPA-O0-NEXT:    mov w8, #48 // =0x30
+; CHECK-NOCPA-O0-NEXT:    smaddl x0, w0, w8, x9
+; CHECK-NOCPA-O0-NEXT:    mov w8, #48 // =0x30
+; CHECK-NOCPA-O0-NEXT:    mov w2, w8
+; CHECK-NOCPA-O0-NEXT:    b memcpy
+;
+; CHECK-NOCPA-O3-LABEL: maddpt1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    mov w8, #48 // =0x30
+; CHECK-NOCPA-O3-NEXT:    adrp x9, array2
+; CHECK-NOCPA-O3-NEXT:    add x9, x9, :lo12:array2
+; CHECK-NOCPA-O3-NEXT:    smaddl x8, w0, w8, x9
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x1]
+; CHECK-NOCPA-O3-NEXT:    str q0, [x8]
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x1, #16]
+; CHECK-NOCPA-O3-NEXT:    str q0, [x8, #16]
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x1, #32]
+; CHECK-NOCPA-O3-NEXT:    str q0, [x8, #32]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %idxprom = sext i32 %pos to i64
+  %arrayidx = getelementptr inbounds [10 x %struct.my_type2], ptr @array2, i64 0, i64 %idxprom
+  tail call void @llvm.memcpy.p0.p0.i64(ptr align 8 dereferenceable(48) %arrayidx, ptr align 8 dereferenceable(48) %val, i64 48, i1 false)
+  ret void
+}
+
+define void @msubpt1(i32 %index, i32 %elem) {
+; CHECK-CPA-O0-LABEL: msubpt1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    // implicit-def: $x8
+; CHECK-CPA-O0-NEXT:    mov w8, w0
+; CHECK-CPA-O0-NEXT:    sxtw x10, w8
+; CHECK-CPA-O0-NEXT:    mov w8, #48 // =0x30
+; CHECK-CPA-O0-NEXT:    mov w9, w8
+; CHECK-CPA-O0-NEXT:    mov w8, #288 // =0x120
+; CHECK-CPA-O0-NEXT:    mov w11, w8
+; CHECK-CPA-O0-NEXT:    adrp x8, array2
+; CHECK-CPA-O0-NEXT:    add x8, x8, :lo12:array2
+; CHECK-CPA-O0-NEXT:    addpt x11, x8, x11
+; CHECK-CPA-O0-NEXT:    msubpt x0, x9, x10, x11
+; CHECK-CPA-O0-NEXT:    mov w9, #48 // =0x30
+; CHECK-CPA-O0-NEXT:    mov w2, w9
+; CHECK-CPA-O0-NEXT:    mov w9, #96 // =0x60
+; CHECK-CPA-O0-NEXT:    // kill: def $x9 killed $w9
+; CHECK-CPA-O0-NEXT:    addpt x1, x8, x9
+; CHECK-CPA-O0-NEXT:    b memcpy
+;
+; CHECK-CPA-O3-LABEL: msubpt1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-CPA-O3-NEXT:    sxtw x9, w0
+; CHECK-CPA-O3-NEXT:    adrp x10, array2+96
+; CHECK-CPA-O3-NEXT:    add x10, x10, :lo12:array2+96
+; CHECK-CPA-O3-NEXT:    mov w8, #48 // =0x30
+; CHECK-CPA-O3-NEXT:    ldr q0, [x10]
+; CHECK-CPA-O3-NEXT:    msubpt x8, x8, x9, x10
+; CHECK-CPA-O3-NEXT:    str q0, [x8, #192]
+; CHECK-CPA-O3-NEXT:    ldr q0, [x10, #16]
+; CHECK-CPA-O3-NEXT:    str q0, [x8, #208]
+; CHECK-CPA-O3-NEXT:    ldr q0, [x10, #32]
+; CHECK-CPA-O3-NEXT:    str q0, [x8, #224]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: msubpt1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    // implicit-def: $x8
+; CHECK-NOCPA-O0-NEXT:    mov w8, w0
+; CHECK-NOCPA-O0-NEXT:    sxtw x8, w8
+; CHECK-NOCPA-O0-NEXT:    mov w9, #48 // =0x30
+; CHECK-NOCPA-O0-NEXT:    // kill: def $x9 killed $w9
+; CHECK-NOCPA-O0-NEXT:    mneg x10, x8, x9
+; CHECK-NOCPA-O0-NEXT:    adrp x8, array2
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, :lo12:array2
+; CHECK-NOCPA-O0-NEXT:    add x9, x8, #288
+; CHECK-NOCPA-O0-NEXT:    add x0, x9, x10
+; CHECK-NOCPA-O0-NEXT:    mov w9, #48 // =0x30
+; CHECK-NOCPA-O0-NEXT:    mov w2, w9
+; CHECK-NOCPA-O0-NEXT:    add x1, x8, #96
+; CHECK-NOCPA-O0-NEXT:    b memcpy
+;
+; CHECK-NOCPA-O3-LABEL: msubpt1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-NOCPA-O3-NEXT:    sxtw x8, w0
+; CHECK-NOCPA-O3-NEXT:    mov w9, #48 // =0x30
+; CHECK-NOCPA-O3-NEXT:    mneg x8, x8, x9
+; CHECK-NOCPA-O3-NEXT:    adrp x9, array2+96
+; CHECK-NOCPA-O3-NEXT:    add x9, x9, :lo12:array2+96
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x9]
+; CHECK-NOCPA-O3-NEXT:    add x8, x9, x8
+; CHECK-NOCPA-O3-NEXT:    str q0, [x8, #192]
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x9, #16]
+; CHECK-NOCPA-O3-NEXT:    str q0, [x8, #208]
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x9, #32]
+; CHECK-NOCPA-O3-NEXT:    str q0, [x8, #224]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %idx.ext = sext i32 %index to i64
+  %idx.neg = sub nsw i64 0, %idx.ext
+  %add.ptr = getelementptr inbounds %struct.my_type2, ptr getelementptr inbounds ([10 x %struct.my_type2], ptr @array2, i64 0, i64 6), i64 %idx.neg
+  tail call void @llvm.memcpy.p0.p0.i64(ptr align 8 dereferenceable(48) %add.ptr, ptr align 8 dereferenceable(48) getelementptr inbounds ([10 x %struct.my_type2], ptr @array2, i64 0, i64 2), i64 48, i1 false), !tbaa.struct !6
+  ret void
+}
+
+define void @subpt1(i32 %index, i32 %elem) {
+; CHECK-CPA-O0-LABEL: subpt1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    mov w8, #-16 // =0xfffffff0
+; CHECK-CPA-O0-NEXT:    smull x9, w0, w8
+; CHECK-CPA-O0-NEXT:    adrp x8, array
+; CHECK-CPA-O0-NEXT:    add x8, x8, :lo12:array
+; CHECK-CPA-O0-NEXT:    ldr q0, [x8, #32]
+; CHECK-CPA-O0-NEXT:    mov w10, #96 // =0x60
+; CHECK-CPA-O0-NEXT:    // kill: def $x10 killed $w10
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x10
+; CHECK-CPA-O0-NEXT:    str q0, [x8, x9, lsl #4]
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subpt1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    mov w8, #-16 // =0xfffffff0
+; CHECK-CPA-O3-NEXT:    adrp x9, array+32
+; CHECK-CPA-O3-NEXT:    add x9, x9, :lo12:array+32
+; CHECK-CPA-O3-NEXT:    smull x8, w0, w8
+; CHECK-CPA-O3-NEXT:    ldr q0, [x9]
+; CHECK-CPA-O3-NEXT:    addpt x8, x9, x8, lsl #4
+; CHECK-CPA-O3-NEXT:    str q0, [x8, #64]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subpt1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    mov w8, #-16 // =0xfffffff0
+; CHECK-NOCPA-O0-NEXT:    smull x9, w0, w8
+; CHECK-NOCPA-O0-NEXT:    adrp x8, array
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, :lo12:array
+; CHECK-NOCPA-O0-NEXT:    ldr q0, [x8, #32]
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, #96
+; CHECK-NOCPA-O0-NEXT:    str q0, [x8, x9, lsl #4]
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subpt1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    mov w8, #-16 // =0xfffffff0
+; CHECK-NOCPA-O3-NEXT:    adrp x9, array+32
+; CHECK-NOCPA-O3-NEXT:    add x9, x9, :lo12:array+32
+; CHECK-NOCPA-O3-NEXT:    smull x8, w0, w8
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x9]
+; CHECK-NOCPA-O3-NEXT:    add x8, x9, x8, lsl #4
+; CHECK-NOCPA-O3-NEXT:    str q0, [x8, #64]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %conv = sext i32 %index to i64
+  %mul.neg = mul nsw i64 %conv, -16
+  %add.ptr = getelementptr inbounds %struct.my_type, ptr getelementptr inbounds ([10 x %struct.my_type], ptr @array, i64 0, i64 6), i64 %mul.neg
+  tail call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(16) %add.ptr, ptr noundef nonnull align 8 dereferenceable(16) getelementptr inbounds ([10 x %struct.my_type], ptr @array, i64 0, i64 2), i64 16, i1 false), !tbaa.struct !6
+  ret void
+}
+
+define void @subpt2(i32 %index, i32 %elem) {
+; CHECK-CPA-O0-LABEL: subpt2:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    mov x8, xzr
+; CHECK-CPA-O0-NEXT:    subs x9, x8, w0, sxtw
+; CHECK-CPA-O0-NEXT:    adrp x8, array
+; CHECK-CPA-O0-NEXT:    add x8, x8, :lo12:array
+; CHECK-CPA-O0-NEXT:    ldr q0, [x8, #32]
+; CHECK-CPA-O0-NEXT:    mov w10, #96 // =0x60
+; CHECK-CPA-O0-NEXT:    // kill: def $x10 killed $w10
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x10
+; CHECK-CPA-O0-NEXT:    str q0, [x8, x9, lsl #4]
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subpt2:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    mov x8, xzr
+; CHECK-CPA-O3-NEXT:    adrp x9, array+32
+; CHECK-CPA-O3-NEXT:    add x9, x9, :lo12:array+32
+; CHECK-CPA-O3-NEXT:    sub x8, x8, w0, sxtw
+; CHECK-CPA-O3-NEXT:    ldr q0, [x9]
+; CHECK-CPA-O3-NEXT:    addpt x8, x9, x8, lsl #4
+; CHECK-CPA-O3-NEXT:    str q0, [x8, #64]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subpt2:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    mov x8, xzr
+; CHECK-NOCPA-O0-NEXT:    subs x9, x8, w0, sxtw
+; CHECK-NOCPA-O0-NEXT:    adrp x8, array
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, :lo12:array
+; CHECK-NOCPA-O0-NEXT:    ldr q0, [x8, #32]
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, #96
+; CHECK-NOCPA-O0-NEXT:    str q0, [x8, x9, lsl #4]
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subpt2:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    mov x8, xzr
+; CHECK-NOCPA-O3-NEXT:    adrp x9, array+32
+; CHECK-NOCPA-O3-NEXT:    add x9, x9, :lo12:array+32
+; CHECK-NOCPA-O3-NEXT:    sub x8, x8, w0, sxtw
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x9]
+; CHECK-NOCPA-O3-NEXT:    add x8, x9, x8, lsl #4
+; CHECK-NOCPA-O3-NEXT:    str q0, [x8, #64]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %idx.ext = sext i32 %index to i64
+  %idx.neg = sub nsw i64 0, %idx.ext
+  %add.ptr = getelementptr inbounds %struct.my_type, ptr getelementptr inbounds ([10 x %struct.my_type], ptr @array, i64 0, i64 6), i64 %idx.neg
+  tail call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(16) %add.ptr, ptr noundef nonnull align 8 dereferenceable(16) getelementptr inbounds ([10 x %struct.my_type], ptr @array, i64 0, i64 2), i64 16, i1 false), !tbaa.struct !11
+  ret void
+}
+
+define ptr @subpt3(ptr %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: subpt3:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    mov x8, #-8 // =0xfffffffffffffff8
+; CHECK-CPA-O0-NEXT:    addpt x0, x0, x8
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subpt3:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    mov x8, #-8 // =0xfffffffffffffff8
+; CHECK-CPA-O3-NEXT:    addpt x0, x0, x8
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subpt3:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    subs x0, x0, #8
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subpt3:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub x0, x0, #8
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = getelementptr inbounds i64, ptr %ptr, i64 -1
+  ret ptr %incdec.ptr.i.i.i
+}
+
+define i64 @subi64(i64 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: subi64:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    subs x0, x0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subi64:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub x0, x0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subi64:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    subs x0, x0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subi64:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub x0, x0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i64 %ptr, -1
+  ret i64 %incdec.ptr.i.i.i
+}
+
+define i32 @subi32(i32 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: subi32:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    subs w0, w0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subi32:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub w0, w0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subi32:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    subs w0, w0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subi32:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub w0, w0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i32 %ptr, -1
+  ret i32 %incdec.ptr.i.i.i
+}
+
+define i16 @subi16(i16 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: subi16:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    subs w0, w0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subi16:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub w0, w0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subi16:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    subs w0, w0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subi16:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub w0, w0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i16 %ptr, -1
+  ret i16 %incdec.ptr.i.i.i
+}
+
+define i64 @addi64(i64 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: addi64:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    add x0, x0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: addi64:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    add x0, x0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: addi64:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    add x0, x0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: addi64:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    add x0, x0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i64 %ptr, 1
+  ret i64 %incdec.ptr.i.i.i
+}
+
+define i32 @addi32(i32 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: addi32:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    add w0, w0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: addi32:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    add w0, w0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: addi32:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    add w0, w0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: addi32:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    add w0, w0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i32 %ptr, 1
+  ret i32 %incdec.ptr.i.i.i
+}
+
+define i16 @addi16(i16 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: addi16:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    add w0, w0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: addi16:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    add w0, w0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: addi16:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    add w0, w0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: addi16:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    add w0, w0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i16 %ptr, 1
+  ret i16 %incdec.ptr.i.i.i
+}
+
+define i64 @arith1(i64 noundef %0, i64 noundef %1, i64 noundef %2) {
+; CHECK-CPA-O0-LABEL: arith1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    sub sp, sp, #32
+; CHECK-CPA-O0-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-CPA-O0-NEXT:    str x0, [sp, #24]
+; CHECK-CPA-O0-NEXT:    str x1, [sp, #16]
+; CHECK-CPA-O0-NEXT:    str x2, [sp, #8]
+; CHECK-CPA-O0-NEXT:    ldr x8, [sp, #24]
+; CHECK-CPA-O0-NEXT:    ldr x9, [sp, #16]
+; CHECK-CPA-O0-NEXT:    ldr x10, [sp, #8]
+; CHECK-CPA-O0-NEXT:    mul x9, x9, x10
+; CHECK-CPA-O0-NEXT:    add x0, x8, x9
+; CHECK-CPA-O0-NEXT:    add sp, sp, #32
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: arith1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub sp, sp, #32
+; CHECK-CPA-O3-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-CPA-O3-NEXT:    stp x2, x1, [sp, #8]
+; CHECK-CPA-O3-NEXT:    str x0, [sp, #24]
+; CHECK-CPA-O3-NEXT:    madd x0, x1, x2, x0
+; CHECK-CPA-O3-NEXT:    add sp, sp, #32
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: arith1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    sub sp, sp, #32
+; CHECK-NOCPA-O0-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NOCPA-O0-NEXT:    str x0, [sp, #24]
+; CHECK-NOCPA-O0-NEXT:    str x1, [sp, #16]
+; CHECK-NOCPA-O0-NEXT:    str x2, [sp, #8]
+; CHECK-NOCPA-O0-NEXT:    ldr x8, [sp, #24]
+; CHECK-NOCPA-O0-NEXT:    ldr x9, [sp, #16]
+; CHECK-NOCPA-O0-NEXT:    ldr x10, [sp, #8]
+; CHECK-NOCPA-O0-NEXT:    mul x9, x9, x10
+; CHECK-NOCPA-O0-NEXT:    add x0, x8, x9
+; CHECK-NOCPA-O0-NEXT:    add sp, sp, #32
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: arith1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub sp, sp, #32
+; CHECK-NOCPA-O3-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NOCPA-O3-NEXT:    stp x2, x1, [sp, #8]
+; CHECK-NOCPA-O3-NEXT:    str x0, [sp, #24]
+; CHECK-NOCPA-O3-NEXT:    madd x0, x1, x2, x0
+; CHECK-NOCPA-O3-NEXT:    add sp, sp, #32
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %4 = alloca i64, align 8
+  %5 = alloca i64, align 8
+  %6 = alloca i64, align 8
+  store i64 %0, ptr %4, align 8
+  store i64 %1, ptr %5, align 8
+  store i64 %2, ptr %6, align 8
+  %7 = load i64, ptr %4, align 8
+  %8 = load i64, ptr %5, align 8
+  %9 = load i64, ptr %6, align 8
+  %10 = mul nsw i64 %8, %9
+  %11 = add nsw i64 %7, %10
+  ret i64 %11
+}
+
+define i64 @arith2(ptr noundef %0, i64 noundef %1, i64 noundef %2, i32 noundef %3) {
+; CHECK-CPA-O0-LABEL: arith2:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    sub sp, sp, #32
+; CHECK-CPA-O0-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-CPA-O0-NEXT:    str x0, [sp, #24]
+; CHECK-CPA-O0-NEXT:    str x1, [sp, #16]
+; CHECK-CPA-O0-NEXT:    str x2, [sp, #8]
+; CHECK-CPA-O0-NEXT:    str w3, [sp, #4]
+; CHECK-CPA-O0-NEXT:    ldr x10, [sp, #24]
+; CHECK-CPA-O0-NEXT:    ldrsw x8, [sp, #4]
+; CHECK-CPA-O0-NEXT:    mov w9, #48 // =0x30
+; CHECK-CPA-O0-NEXT:    // kill: def $x9 killed $w9
+; CHECK-CPA-O0-NEXT:    maddpt x8, x8, x9, x10
+; CHECK-CPA-O0-NEXT:    ldr x8, [x8, #24]
+; CHECK-CPA-O0-NEXT:    ldr x10, [sp, #16]
+; CHECK-CPA-O0-NEXT:    ldr x9, [sp, #8]
+; CHECK-CPA-O0-NEXT:    mul x10, x10, x9
+; CHECK-CPA-O0-NEXT:    add x8, x8, x10
+; CHECK-CPA-O0-NEXT:    subs x0, x8, x9
+; CHECK-CPA-O0-NEXT:    add sp, sp, #32
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: arith2:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub sp, sp, #32
+; CHECK-CPA-O3-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-CPA-O3-NEXT:    str w3, [sp, #4]
+; CHECK-CPA-O3-NEXT:    mov w10, #48 // =0x30
+; CHECK-CPA-O3-NEXT:    ldrsw x9, [sp, #4]
+; CHECK-CPA-O3-NEXT:    stp x1, x0, [sp, #16]
+; CHECK-CPA-O3-NEXT:    maddpt x8, x9, x10, x0
+; CHECK-CPA-O3-NEXT:    str x2, [sp, #8]
+; CHECK-CPA-O3-NEXT:    ldr x8, [x8, #24]
+; CHECK-CPA-O3-NEXT:    madd x8, x1, x2, x8
+; CHECK-CPA-O3-NEXT:    sub x0, x8, x2
+; CHECK-CPA-O3-NEXT:    add sp, sp, #32
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: arith2:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    sub sp, sp, #32
+; CHECK-NOCPA-O0-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NOCPA-O0-NEXT:    str x0, [sp, #24]
+; CHECK-NOCPA-O0-NEXT:    str x1, [sp, #16]
+; CHECK-NOCPA-O0-NEXT:    str x2, [sp, #8]
+; CHECK-NOCPA-O0-NEXT:    str w3, [sp, #4]
+; CHECK-NOCPA-O0-NEXT:    ldr x8, [sp, #24]
+; CHECK-NOCPA-O0-NEXT:    ldrsw x9, [sp, #4]
+; CHECK-NOCPA-O0-NEXT:    mov w10, #48 // =0x30
+; CHECK-NOCPA-O0-NEXT:    // kill: def $x10 killed $w10
+; CHECK-NOCPA-O0-NEXT:    mul x9, x9, x10
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, x9
+; CHECK-NOCPA-O0-NEXT:    ldr x8, [x8, #24]
+; CHECK-NOCPA-O0-NEXT:    ldr x10, [sp, #16]
+; CHECK-NOCPA-O0-NEXT:    ldr x9, [sp, #8]
+; CHECK-NOCPA-O0-NEXT:    mul x10, x10, x9
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, x10
+; CHECK-NOCPA-O0-NEXT:    subs x0, x8, x9
+; CHECK-NOCPA-O0-NEXT:    add sp, sp, #32
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: arith2:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub sp, sp, #32
+; CHECK-NOCPA-O3-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NOCPA-O3-NEXT:    str w3, [sp, #4]
+; CHECK-NOCPA-O3-NEXT:    mov w9, #48 // =0x30
+; CHECK-NOCPA-O3-NEXT:    ldrsw x8, [sp, #4]
+; CHECK-NOCPA-O3-NEXT:    stp x1, x0, [sp, #16]
+; CHECK-NOCPA-O3-NEXT:    str x2, [sp, #8]
+; CHECK-NOCPA-O3-NEXT:    madd x8, x8, x9, x0
+; CHECK-NOCPA-O3-NEXT:    ldr x8, [x8, #24]
+; CHECK-NOCPA-O3-NEXT:    madd x8, x1, x2, x8
+; CHECK-NOCPA-O3-NEXT:    sub x0, x8, x2
+; CHECK-NOCPA-O3-NEXT:    add sp, sp, #32
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %5 = alloca ptr, align 8
+  %6 = alloca i64, align 8
+  %7 = alloca i64, align 8
+  %8 = alloca i32, align 4
+  store ptr %0, ptr %5, align 8
+  store i64 %1, ptr %6, align 8
+  store i64 %2, ptr %7, align 8
+  store i32 %3, ptr %8, align 4
+  %9 = load ptr, ptr %5, align 8
+  %10 = load i32, ptr %8, align 4
+  %11 = sext i32 %10 to i64
+  %12 = getelementptr inbounds %struct.my_type2, ptr %9, i64 %11
+  %13 = getelementptr inbounds %struct.my_type2, ptr %12, i32 0, i32 3
+  %14 = load i64, ptr %13, align 8
+  %15 = load i64, ptr %6, align 8
+  %16 = load i64, ptr %7, align 8
+  %17 = mul nsw i64 %15, %16
+  %18 = add nsw i64 %14, %17
+  %19 = sub nsw i64 %18, %16
+  ret i64 %19
+}
+
+@a = hidden global [2 x [1 x [2 x i8]]] [[1 x [2 x i8]] [[2 x i8] c"\01\01"], [1 x [2 x i8]] [[2 x i8] c"\01\01"]], align 1
+@b = hidden global i16 0, align 2
+@.str = private unnamed_addr constant [7 x i8] c"hello\0A\00", align 1
+
+define hidden void @multidim() {
+; CHECK-CPA-O0-LABEL: multidim:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-CPA-O0-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-CPA-O0-NEXT:    .cfi_offset w30, -16
+; CHECK-CPA-O0-NEXT:    adrp x8, b
+; CHECK-CPA-O0-NEXT:    ldrh w9, [x8, :lo12:b]
+; CHECK-CPA-O0-NEXT:    mov w10, w9
+; CHECK-CPA-O0-NEXT:    ldrh w8, [x8, :lo12:b]
+; CHECK-CPA-O0-NEXT:    add w9, w8, #1
+; CHECK-CPA-O0-NEXT:    // implicit-def: $x8
+; CHECK-CPA-O0-NEXT:    mov w8, w9
+; CHECK-CPA-O0-NEXT:    sxtw x9, w8
+; CHECK-CPA-O0-NEXT:    mov w8, #2 // =0x2
+; CHECK-CPA-O0-NEXT:    mov w11, w8
+; CHECK-CPA-O0-NEXT:    adrp x8, a
+; CHECK-CPA-O0-NEXT:    add x8, x8, :lo12:a
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x11
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x10, lsl #1
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x9
+; CHECK-CPA-O0-NEXT:    ldrb w8, [x8]
+; CHECK-CPA-O0-NEXT:    cbz w8, .LBB14_2
+; CHECK-CPA-O0-NEXT:    b .LBB14_1
+; CHECK-CPA-O0-NEXT:  .LBB14_1:
+; CHECK-CPA-O0-NEXT:    adrp x0, .L.str
+; CHECK-CPA-O0-NEXT:    add x0, x0, :lo12:.L.str
+; CHECK-CPA-O0-NEXT:    bl printf
+; CHECK-CPA-O0-NEXT:    b .LBB14_2
+; CHECK-CPA-O0-NEXT:  .LBB14_2:
+; CHECK-CPA-O0-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: multidim:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    adrp x8, b
+; CHECK-CPA-O3-NEXT:    adrp x10, a+2
+; CHECK-CPA-O3-NEXT:    add x10, x10, :lo12:a+2
+; CHECK-CPA-O3-NEXT:    ldrh w9, [x8, :lo12:b]
+; CHECK-CPA-O3-NEXT:    ldrh w8, [x8, :lo12:b]
+; CHECK-CPA-O3-NEXT:    addpt x9, x10, x9, lsl #1
+; CHECK-CPA-O3-NEXT:    addpt x8, x9, x8
+; CHECK-CPA-O3-NEXT:    ldrb w8, [x8, #1]
+; CHECK-CPA-O3-NEXT:    cbz w8, .LBB14_2
+; CHECK-CPA-O3-NEXT:  // %bb.1:
+; CHECK-CPA-O3-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-CPA-O3-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-CPA-O3-NEXT:    .cfi_offset w30, -16
+; CHECK-CPA-O3-NEXT:    adrp x0, .L.str
+; CHECK-CPA-O3-NEXT:    add x0, x0, :lo12:.L.str
+; CHECK-CPA-O3-NEXT:    bl printf
+; CHECK-CPA-O3-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-CPA-O3-NEXT:  .LBB14_2:
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: multidim:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NOCPA-O0-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NOCPA-O0-NEXT:    .cfi_offset w30, -16
+; CHECK-NOCPA-O0-NEXT:    adrp x8, b
+; CHECK-NOCPA-O0-NEXT:    ldrh w9, [x8, :lo12:b]
+; CHECK-NOCPA-O0-NEXT:    mov w10, w9
+; CHECK-NOCPA-O0-NEXT:    ldrh w8, [x8, :lo12:b]
+; CHECK-NOCPA-O0-NEXT:    add w9, w8, #1
+; CHECK-NOCPA-O0-NEXT:    adrp x8, a
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, :lo12:a
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, #2
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, x10, lsl #1
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, w9, sxtw
+; CHECK-NOCPA-O0-NEXT:    ldrb w8, [x8]
+; CHECK-NOCPA-O0-NEXT:    cbz w8, .LBB14_2
+; CHECK-NOCPA-O0-NEXT:    b .LBB14_1
+; CHECK-NOCPA-O0-NEXT:  .LBB14_1:
+; CHECK-NOCPA-O0-NEXT:    adrp x0, .L.str
+; CHECK-NOCPA-O0-NEXT:    add x0, x0, :lo12:.L.str
+; CHECK-NOCPA-O0-NEXT:    bl printf
+; CHECK-NOCPA-O0-NEXT:    b .LBB14_2
+; CHECK-NOCPA-O0-NEXT:  .LBB14_2:
+; CHECK-NOCPA-O0-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: multidim:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    adrp x8, b
+; CHECK-NOCPA-O3-NEXT:    adrp x10, a+2
+; CHECK-NOCPA-O3-NEXT:    add x10, x10, :lo12:a+2
+; CHECK-NOCPA-O3-NEXT:    ldrh w9, [x8, :lo12:b]
+; CHECK-NOCPA-O3-NEXT:    ldrh w8, [x8, :lo12:b]
+; CHECK-NOCPA-O3-NEXT:    add x9, x10, x9, lsl #1
+; CHECK-NOCPA-O3-NEXT:    add x8, x9, x8
+; CHECK-NOCPA-O3-NEXT:    ldrb w8, [x8, #1]
+; CHECK-NOCPA-O3-NEXT:    cbz w8, .LBB14_2
+; CHECK-NOCPA-O3-NEXT:  // %bb.1:
+; CHECK-NOCPA-O3-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NOCPA-O3-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NOCPA-O3-NEXT:    .cfi_offset w30, -16
+; CHECK-NOCPA-O3-NEXT:    adrp x0, .L.str
+; CHECK-NOCPA-O3-NEXT:    add x0, x0, :lo12:.L.str
+; CHECK-NOCPA-O3-NEXT:    bl printf
+; CHECK-NOCPA-O3-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-NOCPA-O3-NEXT:  .LBB14_2:
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %1 = load i16, ptr @b, align 2
+  %2 = zext i16 %1 to i64
+  %3 = getelementptr inbounds nuw [1 x [2 x i8]], ptr getelementptr inbounds ([2 x [1 x [2 x i8]]], ptr @a, i64 0, i64 1), i64 0, i64 %2
+  %4 = load i16, ptr @b, align 2
+  %5 = zext i16 %4 to i32
+  %6 = add nsw i32 %5, 1
+  %7 = sext i32 %6 to i64
+  %8 = getelementptr inbounds [2 x i8], ptr %3, i64 0, i64 %7
+  %9 = load i8, ptr %8, align 1
+  %10 = icmp ne i8 %9, 0
+  br i1 %10, label %11, label %13
+
+11:
+  %12 = call i32 (ptr, ...) @printf(ptr noundef @.str)
+  br label %13
+
+13:
+  ret void
+}
+
+declare void @llvm.memcpy.p0.p0.i64(ptr, ptr, i64, i1)
+declare i32 @printf(ptr noundef, ...) #1
+
+!6 = !{i64 0, i64 8, !7, i64 8, i64 8, !7, i64 16, i64 8, !7, i64 24, i64 8, !7, i64 32, i64 8, !7, i64 40, i64 8, !7}
+!7 = !{!8, !8, i64 0}
+!8 = !{!"long", !9, i64 0}
+!9 = !{!"omnipotent char", !10, i64 0}
+!10 = !{!"Simple C++ TBAA"}
+!11 = !{i64 0, i64 8, !7, i64 8, i64 8, !7}
diff --git a/llvm/test/CodeGen/AArch64/cpa-selectiondag.ll b/llvm/test/CodeGen/AArch64/cpa-selectiondag.ll
new file mode 100644
index 0000000000000..69fd6c4de78bc
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/cpa-selectiondag.ll
@@ -0,0 +1,796 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=aarch64 -verify-machineinstrs --mattr=+cpa -aarch64-use-featcpa-codegen=true -O0 -global-isel=0 -fast-isel=0 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK-CPA-O0
+; RUN: llc -mtriple=aarch64 -verify-machineinstrs --mattr=+cpa -aarch64-use-featcpa-codegen=true -O3 -global-isel=0 -fast-isel=0 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK-CPA-O3
+; RUN: llc -mtriple=aarch64 -verify-machineinstrs --mattr=-cpa -O0 -global-isel=0 -fast-isel=0 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK-NOCPA-O0
+; RUN: llc -mtriple=aarch64 -verify-machineinstrs --mattr=-cpa -O3 -global-isel=0 -fast-isel=0 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK-NOCPA-O3
+
+%struct.my_type = type { i64, i64 }
+%struct.my_type2 = type { i64, i64, i64, i64, i64, i64 }
+
+@array = external dso_local global [10 x %struct.my_type], align 8
+@array2 = external dso_local global [10 x %struct.my_type2], align 8
+
+define void @addpt1(i64 %index, i64 %arg) {
+; CHECK-CPA-O0-LABEL: addpt1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    adrp x8, array
+; CHECK-CPA-O0-NEXT:    add x8, x8, :lo12:array
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x0, lsl #4
+; CHECK-CPA-O0-NEXT:    str x1, [x8, #8]
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: addpt1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    adrp x8, array
+; CHECK-CPA-O3-NEXT:    add x8, x8, :lo12:array
+; CHECK-CPA-O3-NEXT:    addpt x8, x8, x0, lsl #4
+; CHECK-CPA-O3-NEXT:    str x1, [x8, #8]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: addpt1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    adrp x8, array
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, :lo12:array
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, x0, lsl #4
+; CHECK-NOCPA-O0-NEXT:    str x1, [x8, #8]
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: addpt1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    adrp x8, array
+; CHECK-NOCPA-O3-NEXT:    add x8, x8, :lo12:array
+; CHECK-NOCPA-O3-NEXT:    add x8, x8, x0, lsl #4
+; CHECK-NOCPA-O3-NEXT:    str x1, [x8, #8]
+; CHECK-NOCPA-O3-NEXT:    ret
+
+entry:
+  %e2 = getelementptr inbounds %struct.my_type, ptr @array, i64 %index, i32 1
+  store i64 %arg, ptr %e2, align 8
+  ret void
+}
+
+define void @maddpt1(i32 %pos, ptr %val) {
+; CHECK-CPA-O0-LABEL: maddpt1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    // implicit-def: $x8
+; CHECK-CPA-O0-NEXT:    mov w8, w0
+; CHECK-CPA-O0-NEXT:    sxtw x8, w8
+; CHECK-CPA-O0-NEXT:    adrp x10, array2
+; CHECK-CPA-O0-NEXT:    add x10, x10, :lo12:array2
+; CHECK-CPA-O0-NEXT:    mov w9, #48 // =0x30
+; CHECK-CPA-O0-NEXT:    // kill: def $x9 killed $w9
+; CHECK-CPA-O0-NEXT:    maddpt x8, x8, x9, x10
+; CHECK-CPA-O0-NEXT:    ldr q0, [x1]
+; CHECK-CPA-O0-NEXT:    ldr q1, [x1, #16]
+; CHECK-CPA-O0-NEXT:    ldr q2, [x1, #32]
+; CHECK-CPA-O0-NEXT:    str q2, [x8, #32]
+; CHECK-CPA-O0-NEXT:    str q1, [x8, #16]
+; CHECK-CPA-O0-NEXT:    str q0, [x8]
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: maddpt1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-CPA-O3-NEXT:    sxtw x9, w0
+; CHECK-CPA-O3-NEXT:    ldp q1, q0, [x1, #16]
+; CHECK-CPA-O3-NEXT:    mov w8, #48 // =0x30
+; CHECK-CPA-O3-NEXT:    adrp x10, array2
+; CHECK-CPA-O3-NEXT:    add x10, x10, :lo12:array2
+; CHECK-CPA-O3-NEXT:    ldr q2, [x1]
+; CHECK-CPA-O3-NEXT:    maddpt x8, x9, x8, x10
+; CHECK-CPA-O3-NEXT:    stp q1, q0, [x8, #16]
+; CHECK-CPA-O3-NEXT:    str q2, [x8]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: maddpt1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    adrp x9, array2
+; CHECK-NOCPA-O0-NEXT:    add x9, x9, :lo12:array2
+; CHECK-NOCPA-O0-NEXT:    mov w8, #48 // =0x30
+; CHECK-NOCPA-O0-NEXT:    smaddl x8, w0, w8, x9
+; CHECK-NOCPA-O0-NEXT:    ldr q0, [x1]
+; CHECK-NOCPA-O0-NEXT:    ldr q1, [x1, #16]
+; CHECK-NOCPA-O0-NEXT:    ldr q2, [x1, #32]
+; CHECK-NOCPA-O0-NEXT:    str q2, [x8, #32]
+; CHECK-NOCPA-O0-NEXT:    str q1, [x8, #16]
+; CHECK-NOCPA-O0-NEXT:    str q0, [x8]
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: maddpt1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    mov w8, #48 // =0x30
+; CHECK-NOCPA-O3-NEXT:    adrp x9, array2
+; CHECK-NOCPA-O3-NEXT:    add x9, x9, :lo12:array2
+; CHECK-NOCPA-O3-NEXT:    smaddl x8, w0, w8, x9
+; CHECK-NOCPA-O3-NEXT:    ldp q1, q0, [x1, #16]
+; CHECK-NOCPA-O3-NEXT:    ldr q2, [x1]
+; CHECK-NOCPA-O3-NEXT:    stp q1, q0, [x8, #16]
+; CHECK-NOCPA-O3-NEXT:    str q2, [x8]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %idxprom = sext i32 %pos to i64
+  %arrayidx = getelementptr inbounds [10 x %struct.my_type2], ptr @array2, i64 0, i64 %idxprom
+  tail call void @llvm.memcpy.p0.p0.i64(ptr align 8 dereferenceable(48) %arrayidx, ptr align 8 dereferenceable(48) %val, i64 48, i1 false)
+  ret void
+}
+
+define void @msubpt1(i32 %index, i32 %elem) {
+; CHECK-CPA-O0-LABEL: msubpt1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    // implicit-def: $x8
+; CHECK-CPA-O0-NEXT:    mov w8, w0
+; CHECK-CPA-O0-NEXT:    sxtw x9, w8
+; CHECK-CPA-O0-NEXT:    mov x8, xzr
+; CHECK-CPA-O0-NEXT:    subs x8, x8, x9
+; CHECK-CPA-O0-NEXT:    lsl x8, x8, #1
+; CHECK-CPA-O0-NEXT:    subs x10, x8, x9
+; CHECK-CPA-O0-NEXT:    adrp x9, array2
+; CHECK-CPA-O0-NEXT:    add x9, x9, :lo12:array2
+; CHECK-CPA-O0-NEXT:    mov w8, #288 // =0x120
+; CHECK-CPA-O0-NEXT:    // kill: def $x8 killed $w8
+; CHECK-CPA-O0-NEXT:    addpt x8, x9, x8
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x10, lsl #4
+; CHECK-CPA-O0-NEXT:    mov w10, #96 // =0x60
+; CHECK-CPA-O0-NEXT:    // kill: def $x10 killed $w10
+; CHECK-CPA-O0-NEXT:    addpt x10, x9, x10
+; CHECK-CPA-O0-NEXT:    ldr q1, [x10, #16]
+; CHECK-CPA-O0-NEXT:    ldr q2, [x10, #32]
+; CHECK-CPA-O0-NEXT:    ldr q0, [x9, #96]
+; CHECK-CPA-O0-NEXT:    str q2, [x8, #32]
+; CHECK-CPA-O0-NEXT:    str q1, [x8, #16]
+; CHECK-CPA-O0-NEXT:    str q0, [x8]
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: msubpt1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-CPA-O3-NEXT:    sxtw x9, w0
+; CHECK-CPA-O3-NEXT:    adrp x8, array2
+; CHECK-CPA-O3-NEXT:    add x8, x8, :lo12:array2
+; CHECK-CPA-O3-NEXT:    mov w11, #96 // =0x60
+; CHECK-CPA-O3-NEXT:    mov w12, #288 // =0x120
+; CHECK-CPA-O3-NEXT:    ldr q2, [x8, #96]
+; CHECK-CPA-O3-NEXT:    neg x10, x9
+; CHECK-CPA-O3-NEXT:    addpt x11, x8, x11
+; CHECK-CPA-O3-NEXT:    lsl x10, x10, #1
+; CHECK-CPA-O3-NEXT:    ldp q1, q0, [x11, #16]
+; CHECK-CPA-O3-NEXT:    sub x9, x10, x9
+; CHECK-CPA-O3-NEXT:    addpt x10, x8, x12
+; CHECK-CPA-O3-NEXT:    addpt x9, x10, x9, lsl #4
+; CHECK-CPA-O3-NEXT:    stp q1, q0, [x9, #16]
+; CHECK-CPA-O3-NEXT:    str q2, [x9]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: msubpt1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    // implicit-def: $x8
+; CHECK-NOCPA-O0-NEXT:    mov w8, w0
+; CHECK-NOCPA-O0-NEXT:    sxtw x8, w8
+; CHECK-NOCPA-O0-NEXT:    mov w9, #48 // =0x30
+; CHECK-NOCPA-O0-NEXT:    // kill: def $x9 killed $w9
+; CHECK-NOCPA-O0-NEXT:    mneg x8, x8, x9
+; CHECK-NOCPA-O0-NEXT:    adrp x9, array2
+; CHECK-NOCPA-O0-NEXT:    add x9, x9, :lo12:array2
+; CHECK-NOCPA-O0-NEXT:    add x8, x9, x8
+; CHECK-NOCPA-O0-NEXT:    ldr q0, [x9, #96]
+; CHECK-NOCPA-O0-NEXT:    ldr q1, [x9, #112]
+; CHECK-NOCPA-O0-NEXT:    ldr q2, [x9, #128]
+; CHECK-NOCPA-O0-NEXT:    str q2, [x8, #320]
+; CHECK-NOCPA-O0-NEXT:    str q1, [x8, #304]
+; CHECK-NOCPA-O0-NEXT:    str q0, [x8, #288]
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: msubpt1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-NOCPA-O3-NEXT:    sxtw x8, w0
+; CHECK-NOCPA-O3-NEXT:    mov w9, #48 // =0x30
+; CHECK-NOCPA-O3-NEXT:    mneg x8, x8, x9
+; CHECK-NOCPA-O3-NEXT:    adrp x9, array2
+; CHECK-NOCPA-O3-NEXT:    add x9, x9, :lo12:array2
+; CHECK-NOCPA-O3-NEXT:    ldp q1, q0, [x9, #112]
+; CHECK-NOCPA-O3-NEXT:    ldr q2, [x9, #96]
+; CHECK-NOCPA-O3-NEXT:    add x8, x9, x8
+; CHECK-NOCPA-O3-NEXT:    stp q1, q0, [x8, #304]
+; CHECK-NOCPA-O3-NEXT:    str q2, [x8, #288]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %idx.ext = sext i32 %index to i64
+  %idx.neg = sub nsw i64 0, %idx.ext
+  %add.ptr = getelementptr inbounds %struct.my_type2, ptr getelementptr inbounds ([10 x %struct.my_type2], ptr @array2, i64 0, i64 6), i64 %idx.neg
+  tail call void @llvm.memcpy.p0.p0.i64(ptr align 8 dereferenceable(48) %add.ptr, ptr align 8 dereferenceable(48) getelementptr inbounds ([10 x %struct.my_type2], ptr @array2, i64 0, i64 2), i64 48, i1 false), !tbaa.struct !6
+  ret void
+}
+
+define void @subpt1(i32 %index, i32 %elem) {
+; CHECK-CPA-O0-LABEL: subpt1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    adrp x9, array
+; CHECK-CPA-O0-NEXT:    add x9, x9, :lo12:array
+; CHECK-CPA-O0-NEXT:    mov w8, #96 // =0x60
+; CHECK-CPA-O0-NEXT:    // kill: def $x8 killed $w8
+; CHECK-CPA-O0-NEXT:    addpt x8, x9, x8
+; CHECK-CPA-O0-NEXT:    // implicit-def: $x10
+; CHECK-CPA-O0-NEXT:    mov w10, w0
+; CHECK-CPA-O0-NEXT:    sbfiz x10, x10, #8, #32
+; CHECK-CPA-O0-NEXT:    subpt x8, x8, x10
+; CHECK-CPA-O0-NEXT:    ldr q0, [x9, #32]
+; CHECK-CPA-O0-NEXT:    str q0, [x8]
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subpt1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-CPA-O3-NEXT:    adrp x8, array
+; CHECK-CPA-O3-NEXT:    add x8, x8, :lo12:array
+; CHECK-CPA-O3-NEXT:    mov w9, #96 // =0x60
+; CHECK-CPA-O3-NEXT:    sbfiz x10, x0, #8, #32
+; CHECK-CPA-O3-NEXT:    addpt x9, x8, x9
+; CHECK-CPA-O3-NEXT:    ldr q0, [x8, #32]
+; CHECK-CPA-O3-NEXT:    subpt x8, x9, x10
+; CHECK-CPA-O3-NEXT:    str q0, [x8]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subpt1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    // implicit-def: $x8
+; CHECK-NOCPA-O0-NEXT:    mov w8, w0
+; CHECK-NOCPA-O0-NEXT:    sxtw x8, w8
+; CHECK-NOCPA-O0-NEXT:    adrp x9, array
+; CHECK-NOCPA-O0-NEXT:    add x9, x9, :lo12:array
+; CHECK-NOCPA-O0-NEXT:    subs x8, x9, x8, lsl #8
+; CHECK-NOCPA-O0-NEXT:    ldr q0, [x9, #32]
+; CHECK-NOCPA-O0-NEXT:    str q0, [x8, #96]
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subpt1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-NOCPA-O3-NEXT:    sxtw x9, w0
+; CHECK-NOCPA-O3-NEXT:    adrp x8, array
+; CHECK-NOCPA-O3-NEXT:    add x8, x8, :lo12:array
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x8, #32]
+; CHECK-NOCPA-O3-NEXT:    sub x9, x8, x9, lsl #8
+; CHECK-NOCPA-O3-NEXT:    str q0, [x9, #96]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %conv = sext i32 %index to i64
+  %mul.neg = mul nsw i64 %conv, -16
+  %add.ptr = getelementptr inbounds %struct.my_type, ptr getelementptr inbounds ([10 x %struct.my_type], ptr @array, i64 0, i64 6), i64 %mul.neg
+  tail call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(16) %add.ptr, ptr noundef nonnull align 8 dereferenceable(16) getelementptr inbounds ([10 x %struct.my_type], ptr @array, i64 0, i64 2), i64 16, i1 false), !tbaa.struct !6
+  ret void
+}
+
+define void @subpt2(i32 %index, i32 %elem) {
+; CHECK-CPA-O0-LABEL: subpt2:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    mov x8, xzr
+; CHECK-CPA-O0-NEXT:    subs x10, x8, w0, sxtw
+; CHECK-CPA-O0-NEXT:    adrp x9, array
+; CHECK-CPA-O0-NEXT:    add x9, x9, :lo12:array
+; CHECK-CPA-O0-NEXT:    mov w8, #96 // =0x60
+; CHECK-CPA-O0-NEXT:    // kill: def $x8 killed $w8
+; CHECK-CPA-O0-NEXT:    addpt x8, x9, x8
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x10, lsl #4
+; CHECK-CPA-O0-NEXT:    ldr q0, [x9, #32]
+; CHECK-CPA-O0-NEXT:    str q0, [x8]
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subpt2:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    mov x8, xzr
+; CHECK-CPA-O3-NEXT:    mov w9, #96 // =0x60
+; CHECK-CPA-O3-NEXT:    adrp x10, array
+; CHECK-CPA-O3-NEXT:    add x10, x10, :lo12:array
+; CHECK-CPA-O3-NEXT:    sub x8, x8, w0, sxtw
+; CHECK-CPA-O3-NEXT:    addpt x9, x10, x9
+; CHECK-CPA-O3-NEXT:    ldr q0, [x10, #32]
+; CHECK-CPA-O3-NEXT:    addpt x8, x9, x8, lsl #4
+; CHECK-CPA-O3-NEXT:    str q0, [x8]
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subpt2:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    adrp x9, array
+; CHECK-NOCPA-O0-NEXT:    add x9, x9, :lo12:array
+; CHECK-NOCPA-O0-NEXT:    subs x8, x9, w0, sxtw #4
+; CHECK-NOCPA-O0-NEXT:    ldr q0, [x9, #32]
+; CHECK-NOCPA-O0-NEXT:    str q0, [x8, #96]
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subpt2:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    adrp x8, array
+; CHECK-NOCPA-O3-NEXT:    add x8, x8, :lo12:array
+; CHECK-NOCPA-O3-NEXT:    sub x9, x8, w0, sxtw #4
+; CHECK-NOCPA-O3-NEXT:    ldr q0, [x8, #32]
+; CHECK-NOCPA-O3-NEXT:    str q0, [x9, #96]
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %idx.ext = sext i32 %index to i64
+  %idx.neg = sub nsw i64 0, %idx.ext
+  %add.ptr = getelementptr inbounds %struct.my_type, ptr getelementptr inbounds ([10 x %struct.my_type], ptr @array, i64 0, i64 6), i64 %idx.neg
+  tail call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(16) %add.ptr, ptr noundef nonnull align 8 dereferenceable(16) getelementptr inbounds ([10 x %struct.my_type], ptr @array, i64 0, i64 2), i64 16, i1 false), !tbaa.struct !11
+  ret void
+}
+
+define ptr @subpt3(ptr %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: subpt3:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    mov x8, #-8 // =0xfffffffffffffff8
+; CHECK-CPA-O0-NEXT:    addpt x0, x0, x8
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subpt3:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    mov x8, #-8 // =0xfffffffffffffff8
+; CHECK-CPA-O3-NEXT:    addpt x0, x0, x8
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subpt3:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    subs x0, x0, #8
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subpt3:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub x0, x0, #8
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = getelementptr inbounds i64, ptr %ptr, i64 -1
+  ret ptr %incdec.ptr.i.i.i
+}
+
+define i64 @subi64(i64 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: subi64:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    subs x0, x0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subi64:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub x0, x0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subi64:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    subs x0, x0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subi64:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub x0, x0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i64 %ptr, -1
+  ret i64 %incdec.ptr.i.i.i
+}
+
+define i32 @subi32(i32 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: subi32:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    subs w0, w0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subi32:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub w0, w0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subi32:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    subs w0, w0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subi32:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub w0, w0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i32 %ptr, -1
+  ret i32 %incdec.ptr.i.i.i
+}
+
+define i16 @subi16(i16 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: subi16:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    subs w0, w0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: subi16:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub w0, w0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: subi16:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    subs w0, w0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: subi16:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub w0, w0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i16 %ptr, -1
+  ret i16 %incdec.ptr.i.i.i
+}
+
+define i64 @addi64(i64 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: addi64:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    add x0, x0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: addi64:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    add x0, x0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: addi64:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    add x0, x0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: addi64:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    add x0, x0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i64 %ptr, 1
+  ret i64 %incdec.ptr.i.i.i
+}
+
+define i32 @addi32(i32 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: addi32:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    add w0, w0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: addi32:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    add w0, w0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: addi32:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    add w0, w0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: addi32:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    add w0, w0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i32 %ptr, 1
+  ret i32 %incdec.ptr.i.i.i
+}
+
+define i16 @addi16(i16 %ptr, i32 %index) {
+; CHECK-CPA-O0-LABEL: addi16:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    add w0, w0, #1
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: addi16:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    add w0, w0, #1
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: addi16:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    add w0, w0, #1
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: addi16:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    add w0, w0, #1
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %incdec.ptr.i.i.i = add i16 %ptr, 1
+  ret i16 %incdec.ptr.i.i.i
+}
+
+define i64 @arith1(i64 noundef %0, i64 noundef %1, i64 noundef %2) {
+; CHECK-CPA-O0-LABEL: arith1:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    sub sp, sp, #32
+; CHECK-CPA-O0-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-CPA-O0-NEXT:    str x0, [sp, #24]
+; CHECK-CPA-O0-NEXT:    str x1, [sp, #16]
+; CHECK-CPA-O0-NEXT:    str x2, [sp, #8]
+; CHECK-CPA-O0-NEXT:    ldr x8, [sp, #24]
+; CHECK-CPA-O0-NEXT:    ldr x9, [sp, #16]
+; CHECK-CPA-O0-NEXT:    ldr x10, [sp, #8]
+; CHECK-CPA-O0-NEXT:    mul x9, x9, x10
+; CHECK-CPA-O0-NEXT:    add x0, x8, x9
+; CHECK-CPA-O0-NEXT:    add sp, sp, #32
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: arith1:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub sp, sp, #32
+; CHECK-CPA-O3-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-CPA-O3-NEXT:    madd x8, x1, x2, x0
+; CHECK-CPA-O3-NEXT:    stp x1, x0, [sp, #16]
+; CHECK-CPA-O3-NEXT:    str x2, [sp, #8]
+; CHECK-CPA-O3-NEXT:    mov x0, x8
+; CHECK-CPA-O3-NEXT:    add sp, sp, #32
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: arith1:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    sub sp, sp, #32
+; CHECK-NOCPA-O0-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NOCPA-O0-NEXT:    str x0, [sp, #24]
+; CHECK-NOCPA-O0-NEXT:    str x1, [sp, #16]
+; CHECK-NOCPA-O0-NEXT:    str x2, [sp, #8]
+; CHECK-NOCPA-O0-NEXT:    ldr x8, [sp, #24]
+; CHECK-NOCPA-O0-NEXT:    ldr x9, [sp, #16]
+; CHECK-NOCPA-O0-NEXT:    ldr x10, [sp, #8]
+; CHECK-NOCPA-O0-NEXT:    mul x9, x9, x10
+; CHECK-NOCPA-O0-NEXT:    add x0, x8, x9
+; CHECK-NOCPA-O0-NEXT:    add sp, sp, #32
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: arith1:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub sp, sp, #32
+; CHECK-NOCPA-O3-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NOCPA-O3-NEXT:    madd x8, x1, x2, x0
+; CHECK-NOCPA-O3-NEXT:    stp x1, x0, [sp, #16]
+; CHECK-NOCPA-O3-NEXT:    str x2, [sp, #8]
+; CHECK-NOCPA-O3-NEXT:    mov x0, x8
+; CHECK-NOCPA-O3-NEXT:    add sp, sp, #32
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %4 = alloca i64, align 8
+  %5 = alloca i64, align 8
+  %6 = alloca i64, align 8
+  store i64 %0, ptr %4, align 8
+  store i64 %1, ptr %5, align 8
+  store i64 %2, ptr %6, align 8
+  %7 = load i64, ptr %4, align 8
+  %8 = load i64, ptr %5, align 8
+  %9 = load i64, ptr %6, align 8
+  %10 = mul nsw i64 %8, %9
+  %11 = add nsw i64 %7, %10
+  ret i64 %11
+}
+
+define i64 @arith2(ptr noundef %0, i64 noundef %1, i64 noundef %2, i32 noundef %3) {
+; CHECK-CPA-O0-LABEL: arith2:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    sub sp, sp, #32
+; CHECK-CPA-O0-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-CPA-O0-NEXT:    str x0, [sp, #24]
+; CHECK-CPA-O0-NEXT:    str x1, [sp, #16]
+; CHECK-CPA-O0-NEXT:    str x2, [sp, #8]
+; CHECK-CPA-O0-NEXT:    str w3, [sp, #4]
+; CHECK-CPA-O0-NEXT:    ldr x8, [sp, #24]
+; CHECK-CPA-O0-NEXT:    ldrsw x9, [sp, #4]
+; CHECK-CPA-O0-NEXT:    add x9, x9, x9, lsl #1
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x9, lsl #4
+; CHECK-CPA-O0-NEXT:    ldr x8, [x8, #24]
+; CHECK-CPA-O0-NEXT:    ldr x10, [sp, #16]
+; CHECK-CPA-O0-NEXT:    ldr x9, [sp, #8]
+; CHECK-CPA-O0-NEXT:    mul x10, x10, x9
+; CHECK-CPA-O0-NEXT:    add x8, x8, x10
+; CHECK-CPA-O0-NEXT:    subs x0, x8, x9
+; CHECK-CPA-O0-NEXT:    add sp, sp, #32
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: arith2:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    sub sp, sp, #32
+; CHECK-CPA-O3-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-CPA-O3-NEXT:    // kill: def $w3 killed $w3 def $x3
+; CHECK-CPA-O3-NEXT:    sxtw x8, w3
+; CHECK-CPA-O3-NEXT:    mov w9, #48 // =0x30
+; CHECK-CPA-O3-NEXT:    stp x1, x0, [sp, #16]
+; CHECK-CPA-O3-NEXT:    str x2, [sp, #8]
+; CHECK-CPA-O3-NEXT:    maddpt x8, x8, x9, x0
+; CHECK-CPA-O3-NEXT:    str w3, [sp, #4]
+; CHECK-CPA-O3-NEXT:    ldr x8, [x8, #24]
+; CHECK-CPA-O3-NEXT:    madd x8, x1, x2, x8
+; CHECK-CPA-O3-NEXT:    sub x0, x8, x2
+; CHECK-CPA-O3-NEXT:    add sp, sp, #32
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: arith2:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    sub sp, sp, #32
+; CHECK-NOCPA-O0-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NOCPA-O0-NEXT:    str x0, [sp, #24]
+; CHECK-NOCPA-O0-NEXT:    str x1, [sp, #16]
+; CHECK-NOCPA-O0-NEXT:    str x2, [sp, #8]
+; CHECK-NOCPA-O0-NEXT:    str w3, [sp, #4]
+; CHECK-NOCPA-O0-NEXT:    ldr x10, [sp, #24]
+; CHECK-NOCPA-O0-NEXT:    ldrsw x8, [sp, #4]
+; CHECK-NOCPA-O0-NEXT:    mov w9, #48 // =0x30
+; CHECK-NOCPA-O0-NEXT:    mov w0, w9
+; CHECK-NOCPA-O0-NEXT:    mov w9, w0
+; CHECK-NOCPA-O0-NEXT:    // kill: def $w8 killed $w8 killed $x8
+; CHECK-NOCPA-O0-NEXT:    smaddl x8, w8, w9, x10
+; CHECK-NOCPA-O0-NEXT:    ldr x8, [x8, #24]
+; CHECK-NOCPA-O0-NEXT:    ldr x10, [sp, #16]
+; CHECK-NOCPA-O0-NEXT:    ldr x9, [sp, #8]
+; CHECK-NOCPA-O0-NEXT:    mul x10, x10, x9
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, x10
+; CHECK-NOCPA-O0-NEXT:    subs x0, x8, x9
+; CHECK-NOCPA-O0-NEXT:    add sp, sp, #32
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: arith2:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    sub sp, sp, #32
+; CHECK-NOCPA-O3-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NOCPA-O3-NEXT:    mov w8, #48 // =0x30
+; CHECK-NOCPA-O3-NEXT:    stp x1, x0, [sp, #16]
+; CHECK-NOCPA-O3-NEXT:    smaddl x8, w3, w8, x0
+; CHECK-NOCPA-O3-NEXT:    str x2, [sp, #8]
+; CHECK-NOCPA-O3-NEXT:    str w3, [sp, #4]
+; CHECK-NOCPA-O3-NEXT:    ldr x8, [x8, #24]
+; CHECK-NOCPA-O3-NEXT:    madd x8, x1, x2, x8
+; CHECK-NOCPA-O3-NEXT:    sub x0, x8, x2
+; CHECK-NOCPA-O3-NEXT:    add sp, sp, #32
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %5 = alloca ptr, align 8
+  %6 = alloca i64, align 8
+  %7 = alloca i64, align 8
+  %8 = alloca i32, align 4
+  store ptr %0, ptr %5, align 8
+  store i64 %1, ptr %6, align 8
+  store i64 %2, ptr %7, align 8
+  store i32 %3, ptr %8, align 4
+  %9 = load ptr, ptr %5, align 8
+  %10 = load i32, ptr %8, align 4
+  %11 = sext i32 %10 to i64
+  %12 = getelementptr inbounds %struct.my_type2, ptr %9, i64 %11
+  %13 = getelementptr inbounds %struct.my_type2, ptr %12, i32 0, i32 3
+  %14 = load i64, ptr %13, align 8
+  %15 = load i64, ptr %6, align 8
+  %16 = load i64, ptr %7, align 8
+  %17 = mul nsw i64 %15, %16
+  %18 = add nsw i64 %14, %17
+  %19 = sub nsw i64 %18, %16
+  ret i64 %19
+}
+
+@a = hidden global [2 x [1 x [2 x i8]]] [[1 x [2 x i8]] [[2 x i8] c"\01\01"], [1 x [2 x i8]] [[2 x i8] c"\01\01"]], align 1
+@b = hidden global i16 0, align 2
+@.str = private unnamed_addr constant [7 x i8] c"hello\0A\00", align 1
+
+define hidden void @multidim() {
+; CHECK-CPA-O0-LABEL: multidim:
+; CHECK-CPA-O0:       // %bb.0: // %entry
+; CHECK-CPA-O0-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-CPA-O0-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-CPA-O0-NEXT:    .cfi_offset w30, -16
+; CHECK-CPA-O0-NEXT:    adrp x8, b
+; CHECK-CPA-O0-NEXT:    ldrh w9, [x8, :lo12:b]
+; CHECK-CPA-O0-NEXT:    mov w8, w9
+; CHECK-CPA-O0-NEXT:    mov w10, w8
+; CHECK-CPA-O0-NEXT:    adrp x8, a
+; CHECK-CPA-O0-NEXT:    add x8, x8, :lo12:a
+; CHECK-CPA-O0-NEXT:    mov w11, #2 // =0x2
+; CHECK-CPA-O0-NEXT:    // kill: def $x11 killed $w11
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x11
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x10, lsl #1
+; CHECK-CPA-O0-NEXT:    add w9, w9, #1
+; CHECK-CPA-O0-NEXT:    mov w9, w9
+; CHECK-CPA-O0-NEXT:    // kill: def $x9 killed $w9
+; CHECK-CPA-O0-NEXT:    addpt x8, x8, x9
+; CHECK-CPA-O0-NEXT:    ldrb w8, [x8]
+; CHECK-CPA-O0-NEXT:    cbz w8, .LBB14_2
+; CHECK-CPA-O0-NEXT:    b .LBB14_1
+; CHECK-CPA-O0-NEXT:  .LBB14_1:
+; CHECK-CPA-O0-NEXT:    adrp x0, .L.str
+; CHECK-CPA-O0-NEXT:    add x0, x0, :lo12:.L.str
+; CHECK-CPA-O0-NEXT:    bl printf
+; CHECK-CPA-O0-NEXT:    b .LBB14_2
+; CHECK-CPA-O0-NEXT:  .LBB14_2:
+; CHECK-CPA-O0-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-CPA-O0-NEXT:    ret
+;
+; CHECK-CPA-O3-LABEL: multidim:
+; CHECK-CPA-O3:       // %bb.0: // %entry
+; CHECK-CPA-O3-NEXT:    adrp x8, b
+; CHECK-CPA-O3-NEXT:    mov w9, #2 // =0x2
+; CHECK-CPA-O3-NEXT:    adrp x10, a
+; CHECK-CPA-O3-NEXT:    add x10, x10, :lo12:a
+; CHECK-CPA-O3-NEXT:    ldrh w8, [x8, :lo12:b]
+; CHECK-CPA-O3-NEXT:    addpt x9, x10, x9
+; CHECK-CPA-O3-NEXT:    addpt x9, x9, x8, lsl #1
+; CHECK-CPA-O3-NEXT:    add x8, x8, #1
+; CHECK-CPA-O3-NEXT:    addpt x8, x9, x8
+; CHECK-CPA-O3-NEXT:    ldrb w8, [x8]
+; CHECK-CPA-O3-NEXT:    cbz w8, .LBB14_2
+; CHECK-CPA-O3-NEXT:  // %bb.1:
+; CHECK-CPA-O3-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-CPA-O3-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-CPA-O3-NEXT:    .cfi_offset w30, -16
+; CHECK-CPA-O3-NEXT:    adrp x0, .L.str
+; CHECK-CPA-O3-NEXT:    add x0, x0, :lo12:.L.str
+; CHECK-CPA-O3-NEXT:    bl printf
+; CHECK-CPA-O3-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-CPA-O3-NEXT:  .LBB14_2:
+; CHECK-CPA-O3-NEXT:    ret
+;
+; CHECK-NOCPA-O0-LABEL: multidim:
+; CHECK-NOCPA-O0:       // %bb.0: // %entry
+; CHECK-NOCPA-O0-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NOCPA-O0-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NOCPA-O0-NEXT:    .cfi_offset w30, -16
+; CHECK-NOCPA-O0-NEXT:    adrp x8, b
+; CHECK-NOCPA-O0-NEXT:    ldrh w9, [x8, :lo12:b]
+; CHECK-NOCPA-O0-NEXT:    adrp x8, a
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, :lo12:a
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, w9, uxtw #1
+; CHECK-NOCPA-O0-NEXT:    add w9, w9, #1
+; CHECK-NOCPA-O0-NEXT:    mov w9, w9
+; CHECK-NOCPA-O0-NEXT:    // kill: def $x9 killed $w9
+; CHECK-NOCPA-O0-NEXT:    add x8, x8, x9
+; CHECK-NOCPA-O0-NEXT:    ldrb w8, [x8, #2]
+; CHECK-NOCPA-O0-NEXT:    cbz w8, .LBB14_2
+; CHECK-NOCPA-O0-NEXT:    b .LBB14_1
+; CHECK-NOCPA-O0-NEXT:  .LBB14_1:
+; CHECK-NOCPA-O0-NEXT:    adrp x0, .L.str
+; CHECK-NOCPA-O0-NEXT:    add x0, x0, :lo12:.L.str
+; CHECK-NOCPA-O0-NEXT:    bl printf
+; CHECK-NOCPA-O0-NEXT:    b .LBB14_2
+; CHECK-NOCPA-O0-NEXT:  .LBB14_2:
+; CHECK-NOCPA-O0-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-NOCPA-O0-NEXT:    ret
+;
+; CHECK-NOCPA-O3-LABEL: multidim:
+; CHECK-NOCPA-O3:       // %bb.0: // %entry
+; CHECK-NOCPA-O3-NEXT:    adrp x8, b
+; CHECK-NOCPA-O3-NEXT:    adrp x9, a
+; CHECK-NOCPA-O3-NEXT:    add x9, x9, :lo12:a
+; CHECK-NOCPA-O3-NEXT:    ldrh w8, [x8, :lo12:b]
+; CHECK-NOCPA-O3-NEXT:    add x9, x9, x8, lsl #1
+; CHECK-NOCPA-O3-NEXT:    add x8, x9, x8
+; CHECK-NOCPA-O3-NEXT:    ldrb w8, [x8, #3]
+; CHECK-NOCPA-O3-NEXT:    cbz w8, .LBB14_2
+; CHECK-NOCPA-O3-NEXT:  // %bb.1:
+; CHECK-NOCPA-O3-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NOCPA-O3-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NOCPA-O3-NEXT:    .cfi_offset w30, -16
+; CHECK-NOCPA-O3-NEXT:    adrp x0, .L.str
+; CHECK-NOCPA-O3-NEXT:    add x0, x0, :lo12:.L.str
+; CHECK-NOCPA-O3-NEXT:    bl printf
+; CHECK-NOCPA-O3-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-NOCPA-O3-NEXT:  .LBB14_2:
+; CHECK-NOCPA-O3-NEXT:    ret
+entry:
+  %1 = load i16, ptr @b, align 2
+  %2 = zext i16 %1 to i64
+  %3 = getelementptr inbounds nuw [1 x [2 x i8]], ptr getelementptr inbounds ([2 x [1 x [2 x i8]]], ptr @a, i64 0, i64 1), i64 0, i64 %2
+  %4 = load i16, ptr @b, align 2
+  %5 = zext i16 %4 to i32
+  %6 = add nsw i32 %5, 1
+  %7 = sext i32 %6 to i64
+  %8 = getelementptr inbounds [2 x i8], ptr %3, i64 0, i64 %7
+  %9 = load i8, ptr %8, align 1
+  %10 = icmp ne i8 %9, 0
+  br i1 %10, label %11, label %13
+
+11:
+  %12 = call i32 (ptr, ...) @printf(ptr noundef @.str)
+  br label %13
+
+13:
+  ret void
+}
+
+declare void @llvm.memcpy.p0.p0.i64(ptr, ptr, i64, i1)
+declare i32 @printf(ptr noundef, ...) #1
+
+!6 = !{i64 0, i64 8, !7, i64 8, i64 8, !7, i64 16, i64 8, !7, i64 24, i64 8, !7, i64 32, i64 8, !7, i64 40, i64 8, !7}
+!7 = !{!8, !8, i64 0}
+!8 = !{!"long", !9, i64 0}
+!9 = !{!"omnipotent char", !10, i64 0}
+!10 = !{!"Simple C++ TBAA"}
+!11 = !{i64 0, i64 8, !7, i64 8, i64 8, !7}