diff --git a/llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp b/llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp --- a/llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp +++ b/llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp @@ -423,6 +423,8 @@ bool trySelectCastFixedLengthToScalableVector(SDNode *N); bool trySelectCastScalableToFixedLengthVector(SDNode *N); + bool SelectXAR(SDNode *N); + // Include the pieces autogenerated from the target description. #include "AArch64GenDAGISel.inc" @@ -4144,6 +4146,35 @@ return true; } +bool AArch64DAGToDAGISel::SelectXAR(SDNode *N) { + assert(N->getOpcode() == ISD::OR && "Expected OR instruction"); + + SDNode *N0 = cast(N->getOperand(0)); + SDNode *N1 = cast(N->getOperand(1)); + + if (!(N0->getOpcode() == AArch64ISD::VSHL) || !(N1->getOpcode() == AArch64ISD::VLSHR)) + return false; + + if (!(N0->getOperand(0)->getOpcode() == ISD::XOR) || !(N1->getOperand(0)->getOpcode() == ISD::XOR)) + return false; + + SDNode *XOR = cast(N0->getOperand(0)); + SDValue R1 = XOR->getOperand(0); + SDValue R2 = XOR->getOperand(1); + SDValue imm = N1->getOperand(1); + + unsigned HsAmt = cast(N0->getOperand(1))->getZExtValue(); + unsigned ShAmt = cast(N1->getOperand(1))->getZExtValue(); + + if (!(ShAmt + HsAmt == 64)) + return false; + + SDValue Ops[] = { R1, R2, imm }; + CurDAG->SelectNodeTo(N, AArch64::XAR, MVT::i64, Ops); + + return true; +} + void AArch64DAGToDAGISel::Select(SDNode *Node) { // If we have a custom node, we already have selected! if (Node->isMachineOpcode()) { @@ -4212,6 +4243,10 @@ case ISD::OR: if (tryBitfieldInsertOp(Node)) return; + if (Subtarget->hasSHA3()) + // Try matching XAR instruction here + if (SelectXAR(Node)) + return; break; case ISD::EXTRACT_SUBVECTOR: { diff --git a/llvm/test/CodeGen/AArch64/xar.ll b/llvm/test/CodeGen/AArch64/xar.ll new file mode 100644 --- /dev/null +++ b/llvm/test/CodeGen/AArch64/xar.ll @@ -0,0 +1,25 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2 + +; RUN: llc -mtriple=aarch64-none-eabi -mattr=+sha3 < %s | FileCheck --check-prefix=SHA3 %s +; RUN: llc -mtriple=aarch64-none-eabi -mattr=-sha3 < %s | FileCheck --check-prefix=NOSHA3 %s + +define <2 x i64> @xar_54(<2 x i64> %x, <2 x i64> %y) { +; SHA3-LABEL: xar_54: +; SHA3: // %bb.0: +; SHA3-NEXT: mov w8, #54 // =0x36 +; SHA3-NEXT: xar v0.2d, v0.2d, v1.2d, w8 +; SHA3-NEXT: ret +; +; NOSHA3-LABEL: xar_54: +; NOSHA3: // %bb.0: +; NOSHA3-NEXT: eor v0.16b, v0.16b, v1.16b +; NOSHA3-NEXT: ushr v1.2d, v0.2d, #54 +; NOSHA3-NEXT: shl v0.2d, v0.2d, #10 +; NOSHA3-NEXT: orr v0.16b, v0.16b, v1.16b +; NOSHA3-NEXT: ret + %a = xor <2 x i64> %x, %y + %b = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a, <2 x i64> %a, <2 x i64> ) + ret <2 x i64> %b +} + +declare <2 x i64> @llvm.fshl.v2i64(<2 x i64>, <2 x i64>, <2 x i64>)