[AArch64] Assembly support for the Checked Pointer Arithmetic Extension

clang/test/Driver/aarch64-v95a.c

@@ -13,3 +13,8 @@
 // RUN: %clang -target aarch64_be -mbig-endian -march=armv9.5-a -### -c %s 2>&1 | FileCheck -check-prefix=GENERICV95A-BE %s
 // GENERICV95A-BE: "-cc1"{{.*}} "-triple" "aarch64_be{{.*}}" "-target-cpu" "generic" "-target-feature" "+neon" "-target-feature" "+v9.5a"
 
+// ===== Features supported on aarch64 =====
+
+// RUN: %clang -target aarch64 -march=armv9.5a+cpa -### -c %s 2>&1 | FileCheck -check-prefix=V95A-CPA %s
+// RUN: %clang -target aarch64 -march=armv9.5-a+cpa -### -c %s 2>&1 | FileCheck -check-prefix=V95A-CPA %s
+// V95A-CPA: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-cpu" "generic" "-target-feature" "+neon" "-target-feature" "+v9.5a" "-target-feature" "+cpa"

llvm/include/llvm/TargetParser/AArch64TargetParser.h

@@ -173,6 +173,7 @@ enum ArchExtKind : unsigned {
   AEK_SMEF8F16 =      69, // FEAT_SME_F8F16
   AEK_SMEF8F32 =      70, // FEAT_SME_F8F32
   AEK_SMEFA64 =       71, // FEAT_SME_FA64
+  AEK_CPA =           72, // FEAT_CPA
   AEK_NUM_EXTENSIONS
 };
 using ExtensionBitset = Bitset<AEK_NUM_EXTENSIONS>;
@@ -295,6 +296,7 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"sme-f8f16", AArch64::AEK_SMEF8F16, "+sme-f8f16", "-sme-f8f16", FEAT_INIT, "+sme2,+fp8", 0},
     {"sme-f8f32", AArch64::AEK_SMEF8F32, "+sme-f8f32", "-sme-f8f32", FEAT_INIT, "+sme2,+fp8", 0},
     {"sme-fa64",  AArch64::AEK_SMEFA64,  "+sme-fa64", "-sme-fa64",  FEAT_INIT, "", 0},
+    {"cpa", AArch64::AEK_CPA, "+cpa", "-cpa", FEAT_INIT, "", 0},
     // Special cases
     {"none", AArch64::AEK_NONE, {}, {}, FEAT_INIT, "", ExtensionInfo::MaxFMVPriority},
 };
@@ -378,7 +380,8 @@ inline constexpr ArchInfo ARMV9_3A  = { VersionTuple{9, 3}, AProfile, "armv9.3-a
                                         AArch64::ExtensionBitset({AArch64::AEK_MOPS, AArch64::AEK_HBC}))};
 inline constexpr ArchInfo ARMV9_4A  = { VersionTuple{9, 4}, AProfile, "armv9.4-a", "+v9.4a", (ARMV9_3A.DefaultExts |
                                         AArch64::ExtensionBitset({AArch64::AEK_SPECRES2, AArch64::AEK_CSSC, AArch64::AEK_RASv2}))};
-inline constexpr ArchInfo ARMV9_5A  = { VersionTuple{9, 5}, AProfile, "armv9.5-a", "+v9.5a", (ARMV9_4A.DefaultExts)};
+inline constexpr ArchInfo ARMV9_5A  = { VersionTuple{9, 5}, AProfile, "armv9.5-a", "+v9.5a", (ARMV9_4A.DefaultExts |
+                                        AArch64::ExtensionBitset({AArch64::AEK_CPA}))};
 // For v8-R, we do not enable crypto and align with GCC that enables a more minimal set of optional architecture extensions.
 inline constexpr ArchInfo ARMV8R    = { VersionTuple{8, 0}, RProfile, "armv8-r", "+v8r", (ARMV8_5A.DefaultExts |
                                         AArch64::ExtensionBitset({AArch64::AEK_SSBS,

llvm/lib/Target/AArch64/AArch64.td

@@ -622,6 +622,9 @@ def FeatureLdpAlignedOnly : SubtargetFeature<"ldp-aligned-only", "HasLdpAlignedO
 def FeatureStpAlignedOnly : SubtargetFeature<"stp-aligned-only", "HasStpAlignedOnly",
     "true", "In order to emit stp, first check if the store will be aligned to 2 * element_size">;
 
+def FeatureCPA : SubtargetFeature<"cpa", "HasCPA", "true",
+  "Enable ARMv9.5-A Checked Pointer Arithmetic (FEAT_CPA)">;
+
 //===----------------------------------------------------------------------===//
 // Architectures.
 //
@@ -692,7 +695,7 @@ def HasV9_4aOps : SubtargetFeature<
 
 def HasV9_5aOps : SubtargetFeature<
   "v9.5a", "HasV9_5aOps", "true", "Support ARM v9.5a instructions",
-  [HasV9_4aOps]>;
+  [HasV9_4aOps, FeatureCPA]>;
 
 def HasV8_0rOps : SubtargetFeature<
   "v8r", "HasV8_0rOps", "true", "Support ARM v8r instructions",

llvm/lib/Target/AArch64/AArch64InstrFormats.td

@@ -12446,6 +12446,58 @@ class SystemPXtI<bit L, string asm> :
   BaseSYSPEncoding<L, asm, "\t$op1, $Cn, $Cm, $op2, $Rt", (outs),
   (ins imm0_7:$op1, sys_cr_op:$Cn, sys_cr_op:$Cm, imm0_7:$op2, XSeqPairClassOperand:$Rt)>;
 
+//----------------------------------------------------------------------------
+// 2023 Armv9.5 Extensions
+//----------------------------------------------------------------------------
+
+//---
+// Checked Pointer Arithmetic (FEAT_CPA)
+//---
+
+def LSLImm3ShiftOperand : AsmOperandClass {
+  let SuperClasses = [ExtendOperandLSL64];
+  let Name = "LSLImm3Shift";
+  let RenderMethod = "addLSLImm3ShifterOperands";
+  let DiagnosticType = "AddSubLSLImm3ShiftLarge";
+}
+
+def lsl_imm3_shift_operand : Operand<i32> {
+  let PrintMethod = "printShifter";
+  let ParserMatchClass = LSLImm3ShiftOperand;
+}
+
+// Base CPA scalar add/subtract with lsl #imm3 shift
+class BaseAddSubCPA<bit isSub, string asm> : I<(outs GPR64sp:$Rd),
+    (ins GPR64sp:$Rn, GPR64:$Rm, lsl_imm3_shift_operand:$shift_imm),
+    asm, "\t$Rd, $Rn, $Rm$shift_imm", "", []>, Sched<[]> {
+  bits<5> Rd;
+  bits<5> Rn;
+  bits<5> Rm;
+  bits<3> shift_imm;
+  let Inst{31} = 0b1;
+  let Inst{30} = isSub;
+  let Inst{29-21} = 0b011010000;
+  let Inst{20-16} = Rm;
+  let Inst{15-13} = 0b001;
+  let Inst{12-10} = shift_imm;
+  let Inst{9-5} = Rn;
+  let Inst{4-0} = Rd;
+}
+
+// Alias for CPA scalar add/subtract with no shift
+class AddSubCPAAlias<string asm, Instruction inst>
+    : InstAlias<asm#"\t$Rd, $Rn, $Rm",
+                (inst GPR64sp:$Rd, GPR64sp:$Rn, GPR64:$Rm, 0)>;
+
+multiclass AddSubCPA<bit isSub, string asm> {
+  def _shift : BaseAddSubCPA<isSub, asm>;
+  def _noshift : AddSubCPAAlias<asm, !cast<Instruction>(NAME#"_shift")>;
+}
+
+class MulAccumCPA<bit isSub, string asm>
+  : BaseMulAccum<isSub, 0b011, GPR64, GPR64, asm, []>, Sched<[]> {
+  let Inst{31} = 0b1;
+}
 
 //----------------------------------------------------------------------------
 // Allow the size specifier tokens to be upper case, not just lower.

llvm/lib/Target/AArch64/AArch64InstrInfo.td

@@ -289,6 +289,8 @@ def HasCHK           : Predicate<"Subtarget->hasCHK()">,
                        AssemblerPredicateWithAll<(all_of FeatureCHK), "chk">;
 def HasGCS           : Predicate<"Subtarget->hasGCS()">,
                        AssemblerPredicateWithAll<(all_of FeatureGCS), "gcs">;
+def HasCPA           : Predicate<"Subtarget->hasCPA()">,
+                       AssemblerPredicateWithAll<(all_of FeatureCPA), "cpa">;
 def IsLE             : Predicate<"Subtarget->isLittleEndian()">;
 def IsBE             : Predicate<"!Subtarget->isLittleEndian()">;
 def IsWindows        : Predicate<"Subtarget->isTargetWindows()">;
@@ -9367,6 +9369,10 @@ let Predicates = [HasD128] in {
   }
 }
 
+//===----------------------------===//
+// 2023 Architecture Extensions:
+//===----------------------------===//
+
 let Predicates = [HasFP8] in {
   defm F1CVTL  : SIMDMixedTwoVectorFP8<0b00, "f1cvtl">;
   defm F2CVTL  : SIMDMixedTwoVectorFP8<0b01, "f2cvtl">;
@@ -9408,6 +9414,19 @@ let Predicates = [HasFP8DOT4] in {
  defm FDOT : SIMDThreeSameVectorDOT4<"fdot">;
 } // End let Predicates = [HasFP8DOT4]
 
+//===----------------------------------------------------------------------===//
+// Checked Pointer Arithmetic (FEAT_CPA)
+//===----------------------------------------------------------------------===//
+let Predicates = [HasCPA] in {
+  // Scalar add/subtract
+  defm ADDPT : AddSubCPA<0, "addpt">;
+  defm SUBPT : AddSubCPA<1, "subpt">;
+
+  // Scalar multiply-add/subtract
+  def MADDPT : MulAccumCPA<0, "maddpt">;
+  def MSUBPT : MulAccumCPA<1, "msubpt">;
+}
+
 include "AArch64InstrAtomics.td"
 include "AArch64SVEInstrInfo.td"
 include "AArch64SMEInstrInfo.td"

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

@@ -4163,3 +4163,24 @@ let Predicates = [HasSVE2orSME2, HasLUT] in {
 // LUTI4 (two contiguous registers)
   defm LUTI4_Z2ZZI  : sve2_luti4_vector_vg2_index<"luti4">;
 } // End HasSVE2orSME2, HasLUT
+
+//===----------------------------------------------------------------------===//
+// Checked Pointer Arithmetic (FEAT_CPA)
+//===----------------------------------------------------------------------===//
+let Predicates = [HasSVEorSME, HasCPA] in {
+  // Add/subtract (vectors, unpredicated)
+  def ADD_ZZZ_CPA : sve_int_bin_cons_arit_0<0b11, 0b010, "addpt", ZPR64>;
+  def SUB_ZZZ_CPA : sve_int_bin_cons_arit_0<0b11, 0b011, "subpt", ZPR64>;
+
+  // Add/subtract (vectors, predicated)
+  let DestructiveInstType = DestructiveBinaryComm in {
+    def ADD_ZPmZ_CPA : sve_int_bin_pred_arit_log<0b11, 0b00, 0b100, "addpt", ZPR64>;
+    def SUB_ZPmZ_CPA : sve_int_bin_pred_arit_log<0b11, 0b00, 0b101, "subpt", ZPR64>;
+  }
+
+  // Multiply-add vectors, writing multiplicand
+  def MAD_CPA : sve_int_mad_cpa<"madpt">;
+
+  // Multiply-add vectors, writing addend
+  def MLA_CPA : sve_int_mla_cpa<"mlapt">;
+}

llvm/lib/Target/AArch64/AArch64SchedA64FX.td

@@ -24,7 +24,7 @@ def A64FXModel : SchedMachineModel {
     [HasSVE2, HasSVE2AES, HasSVE2SM4, HasSVE2SHA3, HasSVE2BitPerm, HasPAuth,
      HasSVE2orSME, HasMTE, HasMatMulInt8, HasBF16, HasSME2, HasSME2p1, HasSVE2p1,
      HasSVE2p1_or_HasSME2p1, HasSMEF16F16, HasSSVE_FP8FMA, HasSMEF8F16, HasSMEF8F32,
-     HasSMEFA64];
+     HasSMEFA64, HasCPA];
 
   let FullInstRWOverlapCheck = 0;
 }

llvm/lib/Target/AArch64/AArch64SchedNeoverseN2.td

@@ -19,7 +19,7 @@ def NeoverseN2Model : SchedMachineModel {
   let CompleteModel         =   1;
 
   list<Predicate> UnsupportedFeatures = !listconcat(SMEUnsupported.F,
-                                                    [HasSVE2p1]);
+                                                    [HasSVE2p1, HasCPA]);
 }
 
 //===----------------------------------------------------------------------===//

llvm/lib/Target/AArch64/AArch64SchedNeoverseV1.td

@@ -28,7 +28,7 @@ def NeoverseV1Model : SchedMachineModel {
 
   list<Predicate> UnsupportedFeatures = !listconcat(SVE2Unsupported.F,
                                                     SMEUnsupported.F,
-                                                    [HasMTE]);
+                                                    [HasMTE, HasCPA]);
 }
 
 //===----------------------------------------------------------------------===//

llvm/lib/Target/AArch64/AArch64SchedNeoverseV2.td

@@ -22,7 +22,7 @@ def NeoverseV2Model : SchedMachineModel {
   let CompleteModel         =   1;
 
   list<Predicate> UnsupportedFeatures = !listconcat(SMEUnsupported.F,
-                                                    [HasSVE2p1]);
+                                                    [HasSVE2p1, HasCPA]);
 }
 
 //===----------------------------------------------------------------------===//

llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp

@@ -1541,6 +1541,13 @@ class AArch64Operand : public MCParsedAsmOperand {
            getShiftExtendAmount() <= 4;
   }
 
+  bool isLSLImm3Shift() const {
+    if (!isShiftExtend())
+      return false;
+    AArch64_AM::ShiftExtendType ET = getShiftExtendType();
+    return ET == AArch64_AM::LSL && getShiftExtendAmount() <= 7;
+  }
+
   template<int Width> bool isMemXExtend() const {
     if (!isExtend())
       return false;
@@ -2091,6 +2098,12 @@ class AArch64Operand : public MCParsedAsmOperand {
     Inst.addOperand(MCOperand::createImm(Imm));
   }
 
+  void addLSLImm3ShifterOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    unsigned Imm = getShiftExtendAmount();
+    Inst.addOperand(MCOperand::createImm(Imm));
+  }
+
   void addSyspXzrPairOperand(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
 
@@ -3664,6 +3677,7 @@ static const struct Extension {
     {"sme-f8f16", {AArch64::FeatureSMEF8F16}},
     {"sme-f8f32", {AArch64::FeatureSMEF8F32}},
     {"sme-fa64",  {AArch64::FeatureSMEFA64}},
+    {"cpa", {AArch64::FeatureCPA}},
 };
 
 static void setRequiredFeatureString(FeatureBitset FBS, std::string &Str) {
@@ -6064,6 +6078,9 @@ bool AArch64AsmParser::showMatchError(SMLoc Loc, unsigned ErrCode,
         "Invalid vector list, expected list with each SVE vector in the list "
         "4 registers apart, and the first register in the range [z0, z3] or "
         "[z16, z19] and with correct element type");
+  case Match_AddSubLSLImm3ShiftLarge:
+    return Error(Loc,
+      "expected 'lsl' with optional integer in range [0, 7]");
   default:
     llvm_unreachable("unexpected error code!");
   }
@@ -6448,6 +6465,7 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
   case Match_InvalidMemoryIndexed8:
   case Match_InvalidMemoryIndexed16:
   case Match_InvalidCondCode:
+  case Match_AddSubLSLImm3ShiftLarge:
   case Match_AddSubRegExtendSmall:
   case Match_AddSubRegExtendLarge:
   case Match_AddSubSecondSource:

llvm/lib/Target/AArch64/SVEInstrFormats.td

@@ -10424,3 +10424,34 @@ multiclass sve2_luti4_vector_vg2_index<string mnemonic> {
     let Inst{23-22} = idx;
   }
 }
+
+//===----------------------------------------------------------------------===//
+// Checked Pointer Arithmetic (FEAT_CPA)
+//===----------------------------------------------------------------------===//
+class sve_int_mad_cpa<string asm>
+    : I<(outs ZPR64:$Zdn), (ins ZPR64:$_Zdn, ZPR64:$Zm, ZPR64:$Za),
+        asm, "\t$Zdn, $Zm, $Za", "", []>, Sched<[]> {
+  bits<5> Zdn;
+  bits<5> Zm;
+  bits<5> Za;
+  let Inst{31-24} = 0b01000100;
+  let Inst{23-22} = 0b11; // sz
+  let Inst{21}    = 0b0;
+  let Inst{20-16} = Zm;
+  let Inst{15}    = 0b1;
+  let Inst{14-10} = 0b10110; // opc
+  let Inst{9-5}   = Za;
+  let Inst{4-0}   = Zdn;
+
+  let Constraints = "$Zdn = $_Zdn";
+  let DestructiveInstType = DestructiveOther;
+  let ElementSize = ZPR64.ElementSize;
+  let hasSideEffects = 0;
+}
+
+class sve_int_mla_cpa<string asm>
+    : sve2_int_mla<0b11, 0b10100, asm, ZPR64, ZPR64> {
+  let Inst{15} = 0b1;
+
+  let ElementSize = ZPR64.ElementSize;
+}

llvm/test/MC/AArch64/SVE/armv9.5a-cpa.s

@@ -0,0 +1,69 @@
+// RUN: llvm-mc -triple=aarch64 -show-encoding -mattr=+sve -mattr=+cpa < %s \
+// RUN:        | FileCheck %s --check-prefixes=CHECK-ENCODING,CHECK-INST
+// RUN: llvm-mc -triple=aarch64 -show-encoding -mattr=+sme -mattr=+cpa < %s \
+// RUN:        | FileCheck %s --check-prefixes=CHECK-ENCODING,CHECK-INST
+// RUN: not llvm-mc -triple=aarch64 -show-encoding < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR
+// RUN: not llvm-mc -triple=aarch64 -show-encoding -mattr=+cpa < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR-NO-SVESME
+// RUN: not llvm-mc -triple=aarch64 -show-encoding -mattr=+sve < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR-NO-CPA
+// RUN: not llvm-mc -triple=aarch64 -show-encoding -mattr=+sme < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR-NO-CPA
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve -mattr=+cpa < %s \
+// RUN:        | llvm-objdump -d --mattr=+sve --mattr=+cpa - \
+// RUN:        | FileCheck %s --check-prefix=CHECK-INST
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve -mattr=+cpa < %s \
+// RUN:        | llvm-objdump -d --mattr=+sve --mattr=-cpa - \
+// RUN:        | FileCheck %s --check-prefix=CHECK-UNKNOWN
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve -mattr=+cpa < %s \
+// RUN:        | llvm-objdump -d --mattr=-sve --mattr=+cpa - \
+// RUN:        | FileCheck %s --check-prefix=CHECK-UNKNOWN
+
+addpt z23.d, z13.d, z8.d
+// CHECK-INST: addpt z23.d, z13.d, z8.d
+// CHECK-ENCODING: [0xb7,0x09,0xe8,0x04]
+// CHECK-ERROR: instruction requires: cpa sve or sme
+// CHECK-ERROR-NO-SVESME: instruction requires: sve or sme
+// CHECK-ERROR-NO-CPA: instruction requires: cpa
+// CHECK-UNKNOWN: 04e809b7 <unknown>
+
+addpt z23.d, p3/m, z23.d, z13.d
+// CHECK-INST: addpt z23.d, p3/m, z23.d, z13.d
+// CHECK-ENCODING: [0xb7,0x0d,0xc4,0x04]
+// CHECK-ERROR: instruction requires: cpa sve or sme
+// CHECK-ERROR-NO-SVESME: instruction requires: sve or sme
+// CHECK-ERROR-NO-CPA: instruction requires: cpa
+// CHECK-UNKNOWN: 04c40db7 <unknown>
+
+subpt z23.d, z13.d, z8.d
+// CHECK-INST: subpt z23.d, z13.d, z8.d
+// CHECK-ENCODING: [0xb7,0x0d,0xe8,0x04]
+// CHECK-ERROR: instruction requires: cpa sve or sme
+// CHECK-ERROR-NO-SVESME: instruction requires: sve or sme
+// CHECK-ERROR-NO-CPA: instruction requires: cpa
+// CHECK-UNKNOWN: 04e80db7 <unknown>
+
+subpt z23.d, p3/m, z23.d, z13.d
+// CHECK-INST: subpt z23.d, p3/m, z23.d, z13.d
+// CHECK-ENCODING: [0xb7,0x0d,0xc5,0x04]
+// CHECK-ERROR: instruction requires: cpa sve or sme
+// CHECK-ERROR-NO-SVESME: instruction requires: sve or sme
+// CHECK-ERROR-NO-CPA: instruction requires: cpa
+// CHECK-UNKNOWN: 04c50db7 <unknown>
+
+madpt z0.d, z1.d, z31.d
+// CHECK-INST: madpt z0.d, z1.d, z31.d
+// CHECK-ENCODING: [0xe0,0xdb,0xc1,0x44]
+// CHECK-ERROR: instruction requires: cpa sve or sme
+// CHECK-ERROR-NO-SVESME: instruction requires: sve or sme
+// CHECK-ERROR-NO-CPA: instruction requires: cpa
+// CHECK-UNKNOWN: 44c1dbe0 <unknown>
+
+mlapt z0.d, z1.d, z31.d
+// CHECK-INST: mlapt z0.d, z1.d, z31.d
+// CHECK-ENCODING: [0x20,0xd0,0xdf,0x44]
+// CHECK-ERROR: instruction requires: cpa sve or sme
+// CHECK-ERROR-NO-SVESME: instruction requires: sve or sme
+// CHECK-ERROR-NO-CPA: instruction requires: cpa
+// CHECK-UNKNOWN: 44dfd020 <unknown>