AArch64: support the Apple NEON syntax for v8.2 crypto instructions. (31c25fad) · Commits · llvm-doe / llvm-project

llvm/lib/Target/AArch64/AArch64InstrFormats.td

+15 −11

Original line number	Diff line number	Diff line
		@@ -10407,9 +10407,9 @@ class CryptoRRTied<bits<1>op0, bits<2>op1, string asm, string asmops>
		let Inst{11-10} = op1;
		}
		class CryptoRRTied_2D<bits<1>op0, bits<2>op1, string asm>
		: CryptoRRTied<op0, op1, asm, "{\t$Vd.2d, $Vn.2d}">;
		: CryptoRRTied<op0, op1, asm, "{\t$Vd.2d, $Vn.2d\|.2d\t$Vd, $Vn}">;
		class CryptoRRTied_4S<bits<1>op0, bits<2>op1, string asm>
		: CryptoRRTied<op0, op1, asm, "{\t$Vd.4s, $Vn.4s}">;
		: CryptoRRTied<op0, op1, asm, "{\t$Vd.4s, $Vn.4s\|.4s\t$Vd, $Vn}">;

		class CryptoRRR<bits<1> op0, bits<2>op1, dag oops, dag iops, string asm,
		string asmops, string cst>
		@@ -10424,19 +10424,19 @@ class CryptoRRR<bits<1> op0, bits<2>op1, dag oops, dag iops, string asm,
		}
		class CryptoRRR_2D<bits<1> op0, bits<2>op1, string asm>
		: CryptoRRR<op0, op1, (outs V128:$Vd), (ins V128:$Vn, V128:$Vm), asm,
		"{\t$Vd.2d, $Vn.2d, $Vm.2d}", "">;
		"{\t$Vd.2d, $Vn.2d, $Vm.2d\|.2d\t$Vd, $Vn, $Vm}", "">;
		class CryptoRRRTied_2D<bits<1> op0, bits<2>op1, string asm>
		: CryptoRRR<op0, op1, (outs V128:$Vdst), (ins V128:$Vd, V128:$Vn, V128:$Vm), asm,
		"{\t$Vd.2d, $Vn.2d, $Vm.2d}", "$Vd = $Vdst">;
		"{\t$Vd.2d, $Vn.2d, $Vm.2d\|.2d\t$Vd, $Vn, $Vm}", "$Vd = $Vdst">;
		class CryptoRRR_4S<bits<1> op0, bits<2>op1, string asm>
		: CryptoRRR<op0, op1, (outs V128:$Vd), (ins V128:$Vn, V128:$Vm), asm,
		"{\t$Vd.4s, $Vn.4s, $Vm.4s}", "">;
		"{\t$Vd.4s, $Vn.4s, $Vm.4s\|.4s\t$Vd, $Vn, $Vm}", "">;
		class CryptoRRRTied_4S<bits<1> op0, bits<2>op1, string asm>
		: CryptoRRR<op0, op1, (outs V128:$Vdst), (ins V128:$Vd, V128:$Vn, V128:$Vm), asm,
		"{\t$Vd.4s, $Vn.4s, $Vm.4s}", "$Vd = $Vdst">;
		"{\t$Vd.4s, $Vn.4s, $Vm.4s\|.4s\t$Vd, $Vn, $Vm}", "$Vd = $Vdst">;
		class CryptoRRRTied<bits<1> op0, bits<2>op1, string asm>
		: CryptoRRR<op0, op1, (outs FPR128:$Vdst), (ins FPR128:$Vd, FPR128:$Vn, V128:$Vm),
		asm, "{\t$Vd, $Vn, $Vm.2d}", "$Vd = $Vdst">;
		asm, "{\t$Vd, $Vn, $Vm.2d\|.2d\t$Vd, $Vn, $Vm}", "$Vd = $Vdst">;

		class CryptoRRRR<bits<2>op0, string asm, string asmops>
		: BaseCryptoV82<(outs V128:$Vd), (ins V128:$Vn, V128:$Vm, V128:$Va), asm,
		@@ -10450,15 +10450,18 @@ class CryptoRRRR<bits<2>op0, string asm, string asmops>
		let Inst{14-10} = Va;
		}
		class CryptoRRRR_16B<bits<2>op0, string asm>
		: CryptoRRRR<op0, asm, "{\t$Vd.16b, $Vn.16b, $Vm.16b, $Va.16b}"> {
		: CryptoRRRR<op0, asm, "{\t$Vd.16b, $Vn.16b, $Vm.16b, $Va.16b" #
		"\|.16b\t$Vd, $Vn, $Vm, $Va}"> {
		}
		class CryptoRRRR_4S<bits<2>op0, string asm>
		: CryptoRRRR<op0, asm, "{\t$Vd.4s, $Vn.4s, $Vm.4s, $Va.4s}"> {
		: CryptoRRRR<op0, asm, "{\t$Vd.4s, $Vn.4s, $Vm.4s, $Va.4s" #
		"\|.4s\t$Vd, $Vn, $Vm, $Va}"> {
		}

		class CryptoRRRi6<string asm>
		: BaseCryptoV82<(outs V128:$Vd), (ins V128:$Vn, V128:$Vm, uimm6:$imm), asm,
		"{\t$Vd.2d, $Vn.2d, $Vm.2d, $imm}", "", []> {
		"{\t$Vd.2d, $Vn.2d, $Vm.2d, $imm" #
		"\|.2d\t$Vd, $Vn, $Vm, $imm}", "", []> {
		bits<6> imm;
		bits<5> Vm;
		let Inst{24-21} = 0b0100;
		@@ -10471,7 +10474,8 @@ class CryptoRRRi6<string asm>
		class CryptoRRRi2Tied<bits<1>op0, bits<2>op1, string asm>
		: BaseCryptoV82<(outs V128:$Vdst),
		(ins V128:$Vd, V128:$Vn, V128:$Vm, VectorIndexS:$imm),
		asm, "{\t$Vd.4s, $Vn.4s, $Vm.s$imm}", "$Vd = $Vdst", []> {
		asm, "{\t$Vd.4s, $Vn.4s, $Vm.s$imm" #
		"\|.4s\t$Vd, $Vn, $Vm$imm}", "$Vd = $Vdst", []> {
		bits<2> imm;
		bits<5> Vm;
		let Inst{24-21} = 0b0010;

llvm/test/MC/AArch64/armv8.2a-crypto-apple.s

0 → 100644

+41 −0

Original line number	Diff line number	Diff line
		// RUN: llvm-mc -output-asm-variant=1 -triple aarch64-apple-ios -mattr=+sha3,+sm4 -show-encoding < %s \| FileCheck %s

		sha512h.2d q0, q1, v2
		sha512h2.2d q0, q1, v2
		sha512su0.2d v11, v12
		sha512su1.2d v11, v13, v14
		eor3.16b v25, v12, v7, v2
		rax1.2d v30, v29, v26
		xar.2d v26, v21, v27, #63
		bcax.16b v31, v26, v2, v1

		//CHECK: sha512h.2d q0, q1, v2 ; encoding: [0x20,0x80,0x62,0xce]
		//CHECK: sha512h2.2d q0, q1, v2 ; encoding: [0x20,0x84,0x62,0xce]
		//CHECK: sha512su0.2d v11, v12 ; encoding: [0x8b,0x81,0xc0,0xce]
		//CHECK: sha512su1.2d v11, v13, v14 ; encoding: [0xab,0x89,0x6e,0xce]
		//CHECK: eor3.16b v25, v12, v7, v2 ; encoding: [0x99,0x09,0x07,0xce]
		//CHECK: rax1.2d v30, v29, v26 ; encoding: [0xbe,0x8f,0x7a,0xce]
		//CHECK: xar.2d v26, v21, v27, #63 ; encoding: [0xba,0xfe,0x9b,0xce]
		//CHECK: bcax.16b v31, v26, v2, v1 ; encoding: [0x5f,0x07,0x22,0xce]



		sm3ss1.4s v20, v23, v21, v22
		sm3tt1a.4s v20, v23, v21[3]
		sm3tt1b.4s v20, v23, v21[3]
		sm3tt2a.4s v20, v23, v21[3]
		sm3tt2b.4s v20, v23, v21[3]
		sm3partw1.4s v30, v29, v26
		sm3partw2.4s v30, v29, v26
		sm4ekey.4s v11, v11, v19
		sm4e.4s v2, v15

		// CHECK: sm3ss1.4s v20, v23, v21, v22 ; encoding: [0xf4,0x5a,0x55,0xce]
		// CHECK: sm3tt1a.4s v20, v23, v21[3] ; encoding: [0xf4,0xb2,0x55,0xce]
		// CHECK: sm3tt1b.4s v20, v23, v21[3] ; encoding: [0xf4,0xb6,0x55,0xce]
		// CHECK: sm3tt2a.4s v20, v23, v21[3] ; encoding: [0xf4,0xba,0x55,0xce]
		// CHECK: sm3tt2b.4s v20, v23, v21[3] ; encoding: [0xf4,0xbe,0x55,0xce]
		// CHECK: sm3partw1.4s v30, v29, v26 ; encoding: [0xbe,0xc3,0x7a,0xce]
		// CHECK: sm3partw2.4s v30, v29, v26 ; encoding: [0xbe,0xc7,0x7a,0xce]
		// CHECK: sm4ekey.4s v11, v11, v19 ; encoding: [0x6b,0xc9,0x73,0xce]
		// CHECK: sm4e.4s v2, v15 ; encoding: [0xe2,0x85,0xc0,0xce]