[AArch64][SVE] Add SVE index intrinsic

Summary:
Implements the @llvm.aarch64.sve.index intrinsic, which
takes a scalar base and step value.

This patch also adds the printSImm function to AArch64InstPrinter
to ensure that immediates of type i8 & i16 are printed correctly.

Reviewers: sdesmalen, andwar, efriedma, dancgr, cameron.mcinally, rengolin

Reviewed By: cameron.mcinally

Subscribers: tatyana-krasnukha, tschuett, kristof.beyls, hiraditya, rkruppe, arphaman, psnobl, cfe-commits, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74550

llvm/include/llvm/IR/IntrinsicsAArch64.td

@@ -796,6 +796,12 @@ let TargetPrefix = "aarch64" in {  // All intrinsics start with "llvm.aarch64.".
                  LLVMPointerTo<0>],
                 [IntrArgMemOnly, NoCapture<2>]>;
 
+  class AdvSIMD_SVE_Index_Intrinsic
+    : Intrinsic<[llvm_anyvector_ty],
+                [LLVMVectorElementType<0>,
+                 LLVMVectorElementType<0>],
+                [IntrNoMem]>;
+
   class AdvSIMD_Merged1VectorArg_Intrinsic
     : Intrinsic<[llvm_anyvector_ty],
                 [LLVMMatchType<0>,
@@ -1237,6 +1243,8 @@ def int_aarch64_sve_stnt1 : AdvSIMD_1Vec_PredStore_Intrinsic;
 
 def int_aarch64_sve_dup : AdvSIMD_SVE_DUP_Intrinsic;
 
+def int_aarch64_sve_index : AdvSIMD_SVE_Index_Intrinsic;
+
 //
 // Integer arithmetic
 //

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -1426,6 +1426,7 @@ const char *AArch64TargetLowering::getTargetNodeName(unsigned Opcode) const {
   case AArch64ISD::STP:               return "AArch64ISD::STP";
   case AArch64ISD::STNP:              return "AArch64ISD::STNP";
   case AArch64ISD::DUP_PRED:          return "AArch64ISD::DUP_PRED";
+  case AArch64ISD::INDEX_VECTOR:      return "AArch64ISD::INDEX_VECTOR";
   }
   return nullptr;
 }
@@ -10918,6 +10919,21 @@ static SDValue LowerSVEIntReduction(SDNode *N, unsigned Opc,
   return SDValue();
 }
 
+static SDValue LowerSVEIntrinsicIndex(SDNode *N, SelectionDAG &DAG) {
+  SDLoc DL(N);
+  SDValue Op1 = N->getOperand(1);
+  SDValue Op2 = N->getOperand(2);
+  EVT ScalarTy = Op1.getValueType();
+
+  if ((ScalarTy == MVT::i8) || (ScalarTy == MVT::i16)) {
+    Op1 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i32, Op1);
+    Op2 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i32, Op2);
+  }
+
+  return DAG.getNode(AArch64ISD::INDEX_VECTOR, DL, N->getValueType(0),
+                     Op1, Op2);
+}
+
 static SDValue LowerSVEIntrinsicDUP(SDNode *N, SelectionDAG &DAG) {
   SDLoc dl(N);
   SDValue Scalar = N->getOperand(3);
@@ -11118,6 +11134,8 @@ static SDValue performIntrinsicCombine(SDNode *N,
     return LowerSVEIntReduction(N, AArch64ISD::EORV_PRED, DAG);
   case Intrinsic::aarch64_sve_andv:
     return LowerSVEIntReduction(N, AArch64ISD::ANDV_PRED, DAG);
+  case Intrinsic::aarch64_sve_index:
+    return LowerSVEIntrinsicIndex(N, DAG);
   case Intrinsic::aarch64_sve_dup:
     return LowerSVEIntrinsicDUP(N, DAG);
   case Intrinsic::aarch64_sve_ext:

llvm/lib/Target/AArch64/AArch64ISelLowering.h

@@ -216,6 +216,7 @@ enum NodeType : unsigned {
   PTRUE,
 
   DUP_PRED,
+  INDEX_VECTOR,
 
   LDNF1,
   LDNF1S,

llvm/lib/Target/AArch64/AArch64InstrFormats.td

@@ -330,6 +330,18 @@ def simm5_32b : Operand<i32>, ImmLeaf<i32, [{ return Imm >= -16 && Imm < 16; }]>
   let DecoderMethod = "DecodeSImm<5>";
 }
 
+def simm5_8b : Operand<i32>, ImmLeaf<i32, [{ return (int8_t)Imm >= -16 && (int8_t)Imm < 16; }]> {
+  let ParserMatchClass = SImm5Operand;
+  let DecoderMethod = "DecodeSImm<5>";
+  let PrintMethod = "printSImm<8>";
+}
+
+def simm5_16b : Operand<i32>, ImmLeaf<i32, [{ return (int16_t)Imm >= -16 && (int16_t)Imm < 16; }]> {
+  let ParserMatchClass = SImm5Operand;
+  let DecoderMethod = "DecodeSImm<5>";
+  let PrintMethod = "printSImm<16>";
+}
+
 // simm7sN predicate - True if the immediate is a multiple of N in the range
 // [-64 * N, 63 * N].
 

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

@@ -99,6 +99,9 @@ def AArch64ptest     : SDNode<"AArch64ISD::PTEST", SDT_AArch64PTest>;
 def SDT_AArch64DUP_PRED  : SDTypeProfile<1, 3, [SDTCisVec<0>, SDTCisSameAs<0,1>, SDTCisVec<2>, SDTCVecEltisVT<2,i1>]>;
 def AArch64dup_pred : SDNode<"AArch64ISD::DUP_PRED", SDT_AArch64DUP_PRED>;
 
+def SDT_IndexVector : SDTypeProfile<1, 2, [SDTCisVec<0>, SDTCisSameAs<1, 2>, SDTCisInt<2>]>;
+def index_vector : SDNode<"AArch64ISD::INDEX_VECTOR", SDT_IndexVector, []>;
+
 let Predicates = [HasSVE] in {
 
   defm RDFFR_PPz  : sve_int_rdffr_pred<0b0, "rdffr", int_aarch64_sve_rdffr_z>;
@@ -961,10 +964,10 @@ let Predicates = [HasSVE] in {
   defm INCP_ZP     : sve_int_count_v<0b10000, "incp">;
   defm DECP_ZP     : sve_int_count_v<0b10100, "decp">;
 
-  defm INDEX_RR : sve_int_index_rr<"index">;
-  defm INDEX_IR : sve_int_index_ir<"index">;
-  defm INDEX_RI : sve_int_index_ri<"index">;
-  defm INDEX_II : sve_int_index_ii<"index">;
+  defm INDEX_RR : sve_int_index_rr<"index", index_vector>;
+  defm INDEX_IR : sve_int_index_ir<"index", index_vector>;
+  defm INDEX_RI : sve_int_index_ri<"index", index_vector>;
+  defm INDEX_II : sve_int_index_ii<"index", index_vector>;
 
   // Unpredicated shifts
   defm ASR_ZZI : sve_int_bin_cons_shift_imm_right<0b00, "asr">;

llvm/lib/Target/AArch64/MCTargetDesc/AArch64InstPrinter.cpp

@@ -900,6 +900,19 @@ void AArch64InstPrinter::printImmHex(const MCInst *MI, unsigned OpNo,
   O << format("#%#llx", Op.getImm());
 }
 
+template<int Size>
+void AArch64InstPrinter::printSImm(const MCInst *MI, unsigned OpNo,
+                                  const MCSubtargetInfo &STI,
+                                  raw_ostream &O) {
+  const MCOperand &Op = MI->getOperand(OpNo);
+  if (Size == 8)
+    O << "#" << formatImm((signed char)Op.getImm());
+  else if (Size == 16)
+    O << "#" << formatImm((signed short)Op.getImm());
+  else
+    O << "#" << formatImm(Op.getImm());
+}
+
 void AArch64InstPrinter::printPostIncOperand(const MCInst *MI, unsigned OpNo,
                                              unsigned Imm, raw_ostream &O) {
   const MCOperand &Op = MI->getOperand(OpNo);

llvm/lib/Target/AArch64/MCTargetDesc/AArch64InstPrinter.h

@@ -56,6 +56,9 @@ class AArch64InstPrinter : public MCInstPrinter {
                 raw_ostream &O);
   void printImmHex(const MCInst *MI, unsigned OpNo, const MCSubtargetInfo &STI,
                    raw_ostream &O);
+  template <int Size>
+  void printSImm(const MCInst *MI, unsigned OpNo, const MCSubtargetInfo &STI,
+                 raw_ostream &O);
   template <typename T> void printImmSVE(T Value, raw_ostream &O);
   void printPostIncOperand(const MCInst *MI, unsigned OpNo, unsigned Imm,
                            raw_ostream &O);

llvm/lib/Target/AArch64/SVEInstrFormats.td

@@ -4386,11 +4386,20 @@ class sve_int_index_ii<bits<2> sz8_64, string asm, ZPRRegOp zprty,
   let Inst{4-0}   = Zd;
 }
 
-multiclass sve_int_index_ii<string asm> {
-  def _B : sve_int_index_ii<0b00, asm, ZPR8, simm5_32b>;
-  def _H : sve_int_index_ii<0b01, asm, ZPR16, simm5_32b>;
+multiclass sve_int_index_ii<string asm, SDPatternOperator op> {
+  def _B : sve_int_index_ii<0b00, asm, ZPR8, simm5_8b>;
+  def _H : sve_int_index_ii<0b01, asm, ZPR16, simm5_16b>;
   def _S : sve_int_index_ii<0b10, asm, ZPR32, simm5_32b>;
   def _D : sve_int_index_ii<0b11, asm, ZPR64, simm5_64b>;
+
+  def : Pat<(nxv16i8 (op simm5_8b:$imm5, simm5_8b:$imm5b)),
+            (!cast<Instruction>(NAME # "_B") simm5_8b:$imm5, simm5_8b:$imm5b)>;
+  def : Pat<(nxv8i16 (op simm5_16b:$imm5, simm5_16b:$imm5b)),
+            (!cast<Instruction>(NAME # "_H") simm5_16b:$imm5, simm5_16b:$imm5b)>;
+  def : Pat<(nxv4i32 (op simm5_32b:$imm5, simm5_32b:$imm5b)),
+            (!cast<Instruction>(NAME # "_S") simm5_32b:$imm5, simm5_32b:$imm5b)>;
+  def : Pat<(nxv2i64 (op simm5_64b:$imm5, simm5_64b:$imm5b)),
+            (!cast<Instruction>(NAME # "_D") simm5_64b:$imm5, simm5_64b:$imm5b)>;
 }
 
 class sve_int_index_ir<bits<2> sz8_64, string asm, ZPRRegOp zprty,
@@ -4410,11 +4419,20 @@ class sve_int_index_ir<bits<2> sz8_64, string asm, ZPRRegOp zprty,
   let Inst{4-0}   = Zd;
 }
 
-multiclass sve_int_index_ir<string asm> {
-  def _B : sve_int_index_ir<0b00, asm, ZPR8, GPR32, simm5_32b>;
-  def _H : sve_int_index_ir<0b01, asm, ZPR16, GPR32, simm5_32b>;
+multiclass sve_int_index_ir<string asm, SDPatternOperator op> {
+  def _B : sve_int_index_ir<0b00, asm, ZPR8, GPR32, simm5_8b>;
+  def _H : sve_int_index_ir<0b01, asm, ZPR16, GPR32, simm5_16b>;
   def _S : sve_int_index_ir<0b10, asm, ZPR32, GPR32, simm5_32b>;
   def _D : sve_int_index_ir<0b11, asm, ZPR64, GPR64, simm5_64b>;
+
+  def : Pat<(nxv16i8 (op simm5_8b:$imm5, GPR32:$Rm)),
+            (!cast<Instruction>(NAME # "_B") simm5_8b:$imm5, GPR32:$Rm)>;
+  def : Pat<(nxv8i16 (op simm5_16b:$imm5, GPR32:$Rm)),
+            (!cast<Instruction>(NAME # "_H") simm5_16b:$imm5, GPR32:$Rm)>;
+  def : Pat<(nxv4i32 (op simm5_32b:$imm5, GPR32:$Rm)),
+            (!cast<Instruction>(NAME # "_S") simm5_32b:$imm5, GPR32:$Rm)>;
+  def : Pat<(nxv2i64 (op simm5_64b:$imm5, GPR64:$Rm)),
+            (!cast<Instruction>(NAME # "_D") simm5_64b:$imm5, GPR64:$Rm)>;
 }
 
 class sve_int_index_ri<bits<2> sz8_64, string asm, ZPRRegOp zprty,
@@ -4434,11 +4452,20 @@ class sve_int_index_ri<bits<2> sz8_64, string asm, ZPRRegOp zprty,
   let Inst{4-0}   = Zd;
 }
 
-multiclass sve_int_index_ri<string asm> {
-  def _B : sve_int_index_ri<0b00, asm, ZPR8, GPR32, simm5_32b>;
-  def _H : sve_int_index_ri<0b01, asm, ZPR16, GPR32, simm5_32b>;
+multiclass sve_int_index_ri<string asm, SDPatternOperator op> {
+  def _B : sve_int_index_ri<0b00, asm, ZPR8, GPR32, simm5_8b>;
+  def _H : sve_int_index_ri<0b01, asm, ZPR16, GPR32, simm5_16b>;
   def _S : sve_int_index_ri<0b10, asm, ZPR32, GPR32, simm5_32b>;
   def _D : sve_int_index_ri<0b11, asm, ZPR64, GPR64, simm5_64b>;
+
+  def : Pat<(nxv16i8 (op GPR32:$Rm, simm5_8b:$imm5)),
+            (!cast<Instruction>(NAME # "_B") GPR32:$Rm, simm5_8b:$imm5)>;
+  def : Pat<(nxv8i16 (op GPR32:$Rm, simm5_16b:$imm5)),
+            (!cast<Instruction>(NAME # "_H") GPR32:$Rm, simm5_16b:$imm5)>;
+  def : Pat<(nxv4i32 (op GPR32:$Rm, simm5_32b:$imm5)),
+            (!cast<Instruction>(NAME # "_S") GPR32:$Rm, simm5_32b:$imm5)>;
+  def : Pat<(nxv2i64 (op GPR64:$Rm, simm5_64b:$imm5)),
+            (!cast<Instruction>(NAME # "_D") GPR64:$Rm, simm5_64b:$imm5)>;
 }
 
 class sve_int_index_rr<bits<2> sz8_64, string asm, ZPRRegOp zprty,
@@ -4458,11 +4485,16 @@ class sve_int_index_rr<bits<2> sz8_64, string asm, ZPRRegOp zprty,
   let Inst{4-0}   = Zd;
 }
 
-multiclass sve_int_index_rr<string asm> {
+multiclass sve_int_index_rr<string asm, SDPatternOperator op> {
   def _B : sve_int_index_rr<0b00, asm, ZPR8, GPR32>;
   def _H : sve_int_index_rr<0b01, asm, ZPR16, GPR32>;
   def _S : sve_int_index_rr<0b10, asm, ZPR32, GPR32>;
   def _D : sve_int_index_rr<0b11, asm, ZPR64, GPR64>;
+
+  def : SVE_2_Op_Pat<nxv16i8, op, i32, i32, !cast<Instruction>(NAME # _B)>;
+  def : SVE_2_Op_Pat<nxv8i16, op, i32, i32, !cast<Instruction>(NAME # _H)>;
+  def : SVE_2_Op_Pat<nxv4i32, op, i32, i32, !cast<Instruction>(NAME # _S)>;
+  def : SVE_2_Op_Pat<nxv2i64, op, i64, i64, !cast<Instruction>(NAME # _D)>;
 }
 //
 //===----------------------------------------------------------------------===//