X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=include%2Fllvm%2FIR%2FIntrinsicsAArch64.td;h=578f259aae145173edd144d5685de075270b977a;hp=4d54a2303d62affbc063249ebd33011eb7d00f36;hb=7b1f1048e86ebf78bb2f04232261813077b541ae;hpb=4c433cf673199528b601f664be3d4c121991a7e2 diff --git a/include/llvm/IR/IntrinsicsAArch64.td b/include/llvm/IR/IntrinsicsAArch64.td index 4d54a2303d6..578f259aae1 100644 --- a/include/llvm/IR/IntrinsicsAArch64.td +++ b/include/llvm/IR/IntrinsicsAArch64.td @@ -1,4 +1,4 @@ -//===- IntrinsicsAArch64.td - Defines AArch64 intrinsics -----------*- tablegen -*-===// +//===- IntrinsicsAARCH64.td - Defines AARCH64 intrinsics ---*- tablegen -*-===// // // The LLVM Compiler Infrastructure // @@ -7,288 +7,652 @@ // //===----------------------------------------------------------------------===// // -// This file defines all of the AArch64-specific intrinsics. +// This file defines all of the AARCH64-specific intrinsics. // //===----------------------------------------------------------------------===// +let TargetPrefix = "aarch64" in { + +def int_aarch64_thread_pointer : GCCBuiltin<"__builtin_thread_pointer">, + Intrinsic<[llvm_ptr_ty], [], [IntrNoMem]>; + +def int_aarch64_ldxr : Intrinsic<[llvm_i64_ty], [llvm_anyptr_ty]>; +def int_aarch64_ldaxr : Intrinsic<[llvm_i64_ty], [llvm_anyptr_ty]>; +def int_aarch64_stxr : Intrinsic<[llvm_i32_ty], [llvm_i64_ty, llvm_anyptr_ty]>; +def int_aarch64_stlxr : Intrinsic<[llvm_i32_ty], [llvm_i64_ty, llvm_anyptr_ty]>; + +def int_aarch64_ldxp : Intrinsic<[llvm_i64_ty, llvm_i64_ty], [llvm_ptr_ty]>; +def int_aarch64_ldaxp : Intrinsic<[llvm_i64_ty, llvm_i64_ty], [llvm_ptr_ty]>; +def int_aarch64_stxp : Intrinsic<[llvm_i32_ty], + [llvm_i64_ty, llvm_i64_ty, llvm_ptr_ty]>; +def int_aarch64_stlxp : Intrinsic<[llvm_i32_ty], + [llvm_i64_ty, llvm_i64_ty, llvm_ptr_ty]>; + +def int_aarch64_clrex : Intrinsic<[]>; + +def int_aarch64_sdiv : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>, + LLVMMatchType<0>], [IntrNoMem]>; +def int_aarch64_udiv : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>, + LLVMMatchType<0>], [IntrNoMem]>; + +//===----------------------------------------------------------------------===// +// HINT + +def int_aarch64_hint : Intrinsic<[], [llvm_i32_ty]>; + +//===----------------------------------------------------------------------===// +// RBIT + +def int_aarch64_rbit : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], + [IntrNoMem]>; + +//===----------------------------------------------------------------------===// +// Data Barrier Instructions + +def int_aarch64_dmb : GCCBuiltin<"__builtin_arm_dmb">, Intrinsic<[], [llvm_i32_ty]>; +def int_aarch64_dsb : GCCBuiltin<"__builtin_arm_dsb">, Intrinsic<[], [llvm_i32_ty]>; +def int_aarch64_isb : GCCBuiltin<"__builtin_arm_isb">, Intrinsic<[], [llvm_i32_ty]>; + +} + //===----------------------------------------------------------------------===// // Advanced SIMD (NEON) let TargetPrefix = "aarch64" in { // All intrinsics start with "llvm.aarch64.". + class AdvSIMD_2Scalar_Float_Intrinsic + : Intrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem]>; + + class AdvSIMD_FPToIntRounding_Intrinsic + : Intrinsic<[llvm_anyint_ty], [llvm_anyfloat_ty], [IntrNoMem]>; + + class AdvSIMD_1IntArg_Intrinsic + : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [IntrNoMem]>; + class AdvSIMD_1FloatArg_Intrinsic + : Intrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem]>; + class AdvSIMD_1VectorArg_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [LLVMMatchType<0>], [IntrNoMem]>; + class AdvSIMD_1VectorArg_Expand_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty], [IntrNoMem]>; + class AdvSIMD_1VectorArg_Long_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [LLVMTruncatedType<0>], [IntrNoMem]>; + class AdvSIMD_1IntArg_Narrow_Intrinsic + : Intrinsic<[llvm_anyint_ty], [llvm_anyint_ty], [IntrNoMem]>; + class AdvSIMD_1VectorArg_Narrow_Intrinsic + : Intrinsic<[llvm_anyint_ty], [LLVMExtendedType<0>], [IntrNoMem]>; + class AdvSIMD_1VectorArg_Int_Across_Intrinsic + : Intrinsic<[llvm_anyint_ty], [llvm_anyvector_ty], [IntrNoMem]>; + class AdvSIMD_1VectorArg_Float_Across_Intrinsic + : Intrinsic<[llvm_anyfloat_ty], [llvm_anyvector_ty], [IntrNoMem]>; + + class AdvSIMD_2IntArg_Intrinsic + : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_2FloatArg_Intrinsic + : Intrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Compare_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty, LLVMMatchType<1>], + [IntrNoMem]>; + class AdvSIMD_2Arg_FloatCompare_Intrinsic + : Intrinsic<[llvm_anyint_ty], [llvm_anyfloat_ty, LLVMMatchType<1>], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Long_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMTruncatedType<0>, LLVMTruncatedType<0>], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Wide_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMTruncatedType<0>], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Narrow_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMExtendedType<0>, LLVMExtendedType<0>], + [IntrNoMem]>; + class AdvSIMD_2Arg_Scalar_Narrow_Intrinsic + : Intrinsic<[llvm_anyint_ty], + [LLVMExtendedType<0>, llvm_i32_ty], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Scalar_Expand_BySize_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [llvm_anyvector_ty], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Scalar_Wide_BySize_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMTruncatedType<0>], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Scalar_Wide_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMTruncatedType<0>, llvm_i32_ty], + [IntrNoMem]>; + class AdvSIMD_2VectorArg_Tied_Narrow_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMHalfElementsVectorType<0>, llvm_anyvector_ty], + [IntrNoMem]>; + + class AdvSIMD_3VectorArg_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_3VectorArg_Scalar_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty], + [IntrNoMem]>; + class AdvSIMD_3VectorArg_Tied_Narrow_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMHalfElementsVectorType<0>, llvm_anyvector_ty, + LLVMMatchType<1>], [IntrNoMem]>; + class AdvSIMD_3VectorArg_Scalar_Tied_Narrow_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMHalfElementsVectorType<0>, llvm_anyvector_ty, llvm_i32_ty], + [IntrNoMem]>; + class AdvSIMD_CvtFxToFP_Intrinsic + : Intrinsic<[llvm_anyfloat_ty], [llvm_anyint_ty, llvm_i32_ty], + [IntrNoMem]>; + class AdvSIMD_CvtFPToFx_Intrinsic + : Intrinsic<[llvm_anyint_ty], [llvm_anyfloat_ty, llvm_i32_ty], + [IntrNoMem]>; +} + +// Arithmetic ops + +let Properties = [IntrNoMem] in { + // Vector Add Across Lanes + def int_aarch64_neon_saddv : AdvSIMD_1VectorArg_Int_Across_Intrinsic; + def int_aarch64_neon_uaddv : AdvSIMD_1VectorArg_Int_Across_Intrinsic; + def int_aarch64_neon_faddv : AdvSIMD_1VectorArg_Float_Across_Intrinsic; + + // Vector Long Add Across Lanes + def int_aarch64_neon_saddlv : AdvSIMD_1VectorArg_Int_Across_Intrinsic; + def int_aarch64_neon_uaddlv : AdvSIMD_1VectorArg_Int_Across_Intrinsic; + + // Vector Halving Add + def int_aarch64_neon_shadd : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_uhadd : AdvSIMD_2VectorArg_Intrinsic; + + // Vector Rounding Halving Add + def int_aarch64_neon_srhadd : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_urhadd : AdvSIMD_2VectorArg_Intrinsic; + + // Vector Saturating Add + def int_aarch64_neon_sqadd : AdvSIMD_2IntArg_Intrinsic; + def int_aarch64_neon_suqadd : AdvSIMD_2IntArg_Intrinsic; + def int_aarch64_neon_usqadd : AdvSIMD_2IntArg_Intrinsic; + def int_aarch64_neon_uqadd : AdvSIMD_2IntArg_Intrinsic; + + // Vector Add High-Half + // FIXME: this is a legacy intrinsic for aarch64_simd.h. Remove it when that + // header is no longer supported. + def int_aarch64_neon_addhn : AdvSIMD_2VectorArg_Narrow_Intrinsic; + + // Vector Rounding Add High-Half + def int_aarch64_neon_raddhn : AdvSIMD_2VectorArg_Narrow_Intrinsic; + + // Vector Saturating Doubling Multiply High + def int_aarch64_neon_sqdmulh : AdvSIMD_2IntArg_Intrinsic; + + // Vector Saturating Rounding Doubling Multiply High + def int_aarch64_neon_sqrdmulh : AdvSIMD_2IntArg_Intrinsic; + + // Vector Polynominal Multiply + def int_aarch64_neon_pmul : AdvSIMD_2VectorArg_Intrinsic; + + // Vector Long Multiply + def int_aarch64_neon_smull : AdvSIMD_2VectorArg_Long_Intrinsic; + def int_aarch64_neon_umull : AdvSIMD_2VectorArg_Long_Intrinsic; + def int_aarch64_neon_pmull : AdvSIMD_2VectorArg_Long_Intrinsic; + + // 64-bit polynomial multiply really returns an i128, which is not legal. Fake + // it with a v16i8. + def int_aarch64_neon_pmull64 : + Intrinsic<[llvm_v16i8_ty], [llvm_i64_ty, llvm_i64_ty], [IntrNoMem]>; + + // Vector Extending Multiply + def int_aarch64_neon_fmulx : AdvSIMD_2FloatArg_Intrinsic { + let Properties = [IntrNoMem, Commutative]; + } + + // Vector Saturating Doubling Long Multiply + def int_aarch64_neon_sqdmull : AdvSIMD_2VectorArg_Long_Intrinsic; + def int_aarch64_neon_sqdmulls_scalar + : Intrinsic<[llvm_i64_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; + + // Vector Halving Subtract + def int_aarch64_neon_shsub : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_uhsub : AdvSIMD_2VectorArg_Intrinsic; + + // Vector Saturating Subtract + def int_aarch64_neon_sqsub : AdvSIMD_2IntArg_Intrinsic; + def int_aarch64_neon_uqsub : AdvSIMD_2IntArg_Intrinsic; + + // Vector Subtract High-Half + // FIXME: this is a legacy intrinsic for aarch64_simd.h. Remove it when that + // header is no longer supported. + def int_aarch64_neon_subhn : AdvSIMD_2VectorArg_Narrow_Intrinsic; + + // Vector Rounding Subtract High-Half + def int_aarch64_neon_rsubhn : AdvSIMD_2VectorArg_Narrow_Intrinsic; + + // Vector Compare Absolute Greater-than-or-equal + def int_aarch64_neon_facge : AdvSIMD_2Arg_FloatCompare_Intrinsic; + + // Vector Compare Absolute Greater-than + def int_aarch64_neon_facgt : AdvSIMD_2Arg_FloatCompare_Intrinsic; + + // Vector Absolute Difference + def int_aarch64_neon_sabd : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_uabd : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_fabd : AdvSIMD_2VectorArg_Intrinsic; + + // Scalar Absolute Difference + def int_aarch64_sisd_fabd : AdvSIMD_2Scalar_Float_Intrinsic; + + // Vector Max + def int_aarch64_neon_smax : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_umax : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_fmax : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_fmaxnmp : AdvSIMD_2VectorArg_Intrinsic; + + // Vector Max Across Lanes + def int_aarch64_neon_smaxv : AdvSIMD_1VectorArg_Int_Across_Intrinsic; + def int_aarch64_neon_umaxv : AdvSIMD_1VectorArg_Int_Across_Intrinsic; + def int_aarch64_neon_fmaxv : AdvSIMD_1VectorArg_Float_Across_Intrinsic; + def int_aarch64_neon_fmaxnmv : AdvSIMD_1VectorArg_Float_Across_Intrinsic; + + // Vector Min + def int_aarch64_neon_smin : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_umin : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_fmin : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_fminnmp : AdvSIMD_2VectorArg_Intrinsic; + + // Vector Min/Max Number + def int_aarch64_neon_fminnm : AdvSIMD_2FloatArg_Intrinsic; + def int_aarch64_neon_fmaxnm : AdvSIMD_2FloatArg_Intrinsic; + + // Vector Min Across Lanes + def int_aarch64_neon_sminv : AdvSIMD_1VectorArg_Int_Across_Intrinsic; + def int_aarch64_neon_uminv : AdvSIMD_1VectorArg_Int_Across_Intrinsic; + def int_aarch64_neon_fminv : AdvSIMD_1VectorArg_Float_Across_Intrinsic; + def int_aarch64_neon_fminnmv : AdvSIMD_1VectorArg_Float_Across_Intrinsic; + + // Pairwise Add + def int_aarch64_neon_addp : AdvSIMD_2VectorArg_Intrinsic; + + // Long Pairwise Add + // FIXME: In theory, we shouldn't need intrinsics for saddlp or + // uaddlp, but tblgen's type inference currently can't handle the + // pattern fragments this ends up generating. + def int_aarch64_neon_saddlp : AdvSIMD_1VectorArg_Expand_Intrinsic; + def int_aarch64_neon_uaddlp : AdvSIMD_1VectorArg_Expand_Intrinsic; + + // Folding Maximum + def int_aarch64_neon_smaxp : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_umaxp : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_fmaxp : AdvSIMD_2VectorArg_Intrinsic; + + // Folding Minimum + def int_aarch64_neon_sminp : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_uminp : AdvSIMD_2VectorArg_Intrinsic; + def int_aarch64_neon_fminp : AdvSIMD_2VectorArg_Intrinsic; + + // Reciprocal Estimate/Step + def int_aarch64_neon_frecps : AdvSIMD_2FloatArg_Intrinsic; + def int_aarch64_neon_frsqrts : AdvSIMD_2FloatArg_Intrinsic; + + // Reciprocal Exponent + def int_aarch64_neon_frecpx : AdvSIMD_1FloatArg_Intrinsic; + + // Vector Saturating Shift Left + def int_aarch64_neon_sqshl : AdvSIMD_2IntArg_Intrinsic; + def int_aarch64_neon_uqshl : AdvSIMD_2IntArg_Intrinsic; + + // Vector Rounding Shift Left + def int_aarch64_neon_srshl : AdvSIMD_2IntArg_Intrinsic; + def int_aarch64_neon_urshl : AdvSIMD_2IntArg_Intrinsic; + + // Vector Saturating Rounding Shift Left + def int_aarch64_neon_sqrshl : AdvSIMD_2IntArg_Intrinsic; + def int_aarch64_neon_uqrshl : AdvSIMD_2IntArg_Intrinsic; + + // Vector Signed->Unsigned Shift Left by Constant + def int_aarch64_neon_sqshlu : AdvSIMD_2IntArg_Intrinsic; + + // Vector Signed->Unsigned Narrowing Saturating Shift Right by Constant + def int_aarch64_neon_sqshrun : AdvSIMD_2Arg_Scalar_Narrow_Intrinsic; + + // Vector Signed->Unsigned Rounding Narrowing Saturating Shift Right by Const + def int_aarch64_neon_sqrshrun : AdvSIMD_2Arg_Scalar_Narrow_Intrinsic; + + // Vector Narrowing Shift Right by Constant + def int_aarch64_neon_sqshrn : AdvSIMD_2Arg_Scalar_Narrow_Intrinsic; + def int_aarch64_neon_uqshrn : AdvSIMD_2Arg_Scalar_Narrow_Intrinsic; + + // Vector Rounding Narrowing Shift Right by Constant + def int_aarch64_neon_rshrn : AdvSIMD_2Arg_Scalar_Narrow_Intrinsic; + + // Vector Rounding Narrowing Saturating Shift Right by Constant + def int_aarch64_neon_sqrshrn : AdvSIMD_2Arg_Scalar_Narrow_Intrinsic; + def int_aarch64_neon_uqrshrn : AdvSIMD_2Arg_Scalar_Narrow_Intrinsic; + + // Vector Shift Left + def int_aarch64_neon_sshl : AdvSIMD_2IntArg_Intrinsic; + def int_aarch64_neon_ushl : AdvSIMD_2IntArg_Intrinsic; + + // Vector Widening Shift Left by Constant + def int_aarch64_neon_shll : AdvSIMD_2VectorArg_Scalar_Wide_BySize_Intrinsic; + def int_aarch64_neon_sshll : AdvSIMD_2VectorArg_Scalar_Wide_Intrinsic; + def int_aarch64_neon_ushll : AdvSIMD_2VectorArg_Scalar_Wide_Intrinsic; + + // Vector Shift Right by Constant and Insert + def int_aarch64_neon_vsri : AdvSIMD_3VectorArg_Scalar_Intrinsic; + + // Vector Shift Left by Constant and Insert + def int_aarch64_neon_vsli : AdvSIMD_3VectorArg_Scalar_Intrinsic; + + // Vector Saturating Narrow + def int_aarch64_neon_scalar_sqxtn: AdvSIMD_1IntArg_Narrow_Intrinsic; + def int_aarch64_neon_scalar_uqxtn : AdvSIMD_1IntArg_Narrow_Intrinsic; + def int_aarch64_neon_sqxtn : AdvSIMD_1VectorArg_Narrow_Intrinsic; + def int_aarch64_neon_uqxtn : AdvSIMD_1VectorArg_Narrow_Intrinsic; + + // Vector Saturating Extract and Unsigned Narrow + def int_aarch64_neon_scalar_sqxtun : AdvSIMD_1IntArg_Narrow_Intrinsic; + def int_aarch64_neon_sqxtun : AdvSIMD_1VectorArg_Narrow_Intrinsic; + + // Vector Absolute Value + def int_aarch64_neon_abs : AdvSIMD_1IntArg_Intrinsic; + + // Vector Saturating Absolute Value + def int_aarch64_neon_sqabs : AdvSIMD_1IntArg_Intrinsic; + + // Vector Saturating Negation + def int_aarch64_neon_sqneg : AdvSIMD_1IntArg_Intrinsic; + + // Vector Count Leading Sign Bits + def int_aarch64_neon_cls : AdvSIMD_1VectorArg_Intrinsic; + + // Vector Reciprocal Estimate + def int_aarch64_neon_urecpe : AdvSIMD_1VectorArg_Intrinsic; + def int_aarch64_neon_frecpe : AdvSIMD_1FloatArg_Intrinsic; + + // Vector Square Root Estimate + def int_aarch64_neon_ursqrte : AdvSIMD_1VectorArg_Intrinsic; + def int_aarch64_neon_frsqrte : AdvSIMD_1FloatArg_Intrinsic; + + // Vector Bitwise Reverse + def int_aarch64_neon_rbit : AdvSIMD_1VectorArg_Intrinsic; + + // Vector Conversions Between Half-Precision and Single-Precision. + def int_aarch64_neon_vcvtfp2hf + : Intrinsic<[llvm_v4i16_ty], [llvm_v4f32_ty], [IntrNoMem]>; + def int_aarch64_neon_vcvthf2fp + : Intrinsic<[llvm_v4f32_ty], [llvm_v4i16_ty], [IntrNoMem]>; + + // Vector Conversions Between Floating-point and Fixed-point. + def int_aarch64_neon_vcvtfp2fxs : AdvSIMD_CvtFPToFx_Intrinsic; + def int_aarch64_neon_vcvtfp2fxu : AdvSIMD_CvtFPToFx_Intrinsic; + def int_aarch64_neon_vcvtfxs2fp : AdvSIMD_CvtFxToFP_Intrinsic; + def int_aarch64_neon_vcvtfxu2fp : AdvSIMD_CvtFxToFP_Intrinsic; + + // Vector FP->Int Conversions + def int_aarch64_neon_fcvtas : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtau : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtms : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtmu : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtns : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtnu : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtps : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtpu : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtzs : AdvSIMD_FPToIntRounding_Intrinsic; + def int_aarch64_neon_fcvtzu : AdvSIMD_FPToIntRounding_Intrinsic; + + // Vector FP Rounding: only ties to even is unrepresented by a normal + // intrinsic. + def int_aarch64_neon_frintn : AdvSIMD_1FloatArg_Intrinsic; + + // Scalar FP->Int conversions + + // Vector FP Inexact Narrowing + def int_aarch64_neon_fcvtxn : AdvSIMD_1VectorArg_Expand_Intrinsic; + + // Scalar FP Inexact Narrowing + def int_aarch64_sisd_fcvtxn : Intrinsic<[llvm_float_ty], [llvm_double_ty], + [IntrNoMem]>; +} + +let TargetPrefix = "aarch64" in { // All intrinsics start with "llvm.aarch64.". + class AdvSIMD_2Vector2Index_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [llvm_anyvector_ty, llvm_i64_ty, LLVMMatchType<0>, llvm_i64_ty], + [IntrNoMem]>; +} -// Vector Absolute Compare (Floating Point) -def int_aarch64_neon_vacgeq : - Intrinsic<[llvm_v2i64_ty], [llvm_v2f64_ty, llvm_v2f64_ty], [IntrNoMem]>; -def int_aarch64_neon_vacgtq : - Intrinsic<[llvm_v2i64_ty], [llvm_v2f64_ty, llvm_v2f64_ty], [IntrNoMem]>; - -// Vector maxNum (Floating Point) -def int_aarch64_neon_vmaxnm : Neon_2Arg_Intrinsic; - -// Vector minNum (Floating Point) -def int_aarch64_neon_vminnm : Neon_2Arg_Intrinsic; - -// Vector Pairwise maxNum (Floating Point) -def int_aarch64_neon_vpmaxnm : Neon_2Arg_Intrinsic; - -// Vector Pairwise minNum (Floating Point) -def int_aarch64_neon_vpminnm : Neon_2Arg_Intrinsic; - -// Vector Multiply Extended (Floating Point) -def int_aarch64_neon_vmulx : Neon_2Arg_Intrinsic; - -class Neon_N2V_Intrinsic - : Intrinsic<[llvm_anyvector_ty], [LLVMMatchType<0>, llvm_i32_ty], - [IntrNoMem]>; -class Neon_N3V_Intrinsic - : Intrinsic<[llvm_anyvector_ty], - [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty], - [IntrNoMem]>; -class Neon_N2V_Narrow_Intrinsic - : Intrinsic<[llvm_anyvector_ty], - [LLVMExtendedElementVectorType<0>, llvm_i32_ty], - [IntrNoMem]>; - -// Vector rounding shift right by immediate (Signed) -def int_aarch64_neon_vsrshr : Neon_N2V_Intrinsic; -def int_aarch64_neon_vurshr : Neon_N2V_Intrinsic; -def int_aarch64_neon_vsqshlu : Neon_N2V_Intrinsic; - -def int_aarch64_neon_vsri : Neon_N3V_Intrinsic; -def int_aarch64_neon_vsli : Neon_N3V_Intrinsic; - -def int_aarch64_neon_vsqshrun : Neon_N2V_Narrow_Intrinsic; -def int_aarch64_neon_vrshrn : Neon_N2V_Narrow_Intrinsic; -def int_aarch64_neon_vsqrshrun : Neon_N2V_Narrow_Intrinsic; -def int_aarch64_neon_vsqshrn : Neon_N2V_Narrow_Intrinsic; -def int_aarch64_neon_vuqshrn : Neon_N2V_Narrow_Intrinsic; -def int_aarch64_neon_vsqrshrn : Neon_N2V_Narrow_Intrinsic; -def int_aarch64_neon_vuqrshrn : Neon_N2V_Narrow_Intrinsic; - -// Vector across -class Neon_Across_Intrinsic - : Intrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty], [IntrNoMem]>; - -class Neon_2Arg_Across_Float_Intrinsic - : Intrinsic<[llvm_anyvector_ty], [llvm_v4f32_ty], [IntrNoMem]>; - -def int_aarch64_neon_saddlv : Neon_Across_Intrinsic; -def int_aarch64_neon_uaddlv : Neon_Across_Intrinsic; -def int_aarch64_neon_smaxv : Neon_Across_Intrinsic; -def int_aarch64_neon_umaxv : Neon_Across_Intrinsic; -def int_aarch64_neon_sminv : Neon_Across_Intrinsic; -def int_aarch64_neon_uminv : Neon_Across_Intrinsic; -def int_aarch64_neon_vaddv : Neon_Across_Intrinsic; -def int_aarch64_neon_vmaxv : Neon_Across_Intrinsic; -def int_aarch64_neon_vminv : Neon_Across_Intrinsic; -def int_aarch64_neon_vmaxnmv : Neon_Across_Intrinsic; -def int_aarch64_neon_vminnmv : Neon_Across_Intrinsic; - -// Scalar Add -def int_aarch64_neon_vaddds : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty], [IntrNoMem]>; -def int_aarch64_neon_vadddu : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty], [IntrNoMem]>; - -// Scalar Saturating Add (Signed, Unsigned) -def int_aarch64_neon_vqadds : Neon_2Arg_Intrinsic; -def int_aarch64_neon_vqaddu : Neon_2Arg_Intrinsic; - -// Scalar Sub -def int_aarch64_neon_vsubds : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty], [IntrNoMem]>; -def int_aarch64_neon_vsubdu : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty], [IntrNoMem]>; - -// Scalar Saturating Sub (Signed, Unsigned) -def int_aarch64_neon_vqsubs : Neon_2Arg_Intrinsic; -def int_aarch64_neon_vqsubu : Neon_2Arg_Intrinsic; - -// Scalar Shift -// Scalar Shift Left -def int_aarch64_neon_vshlds : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty], [IntrNoMem]>; -def int_aarch64_neon_vshldu : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty], [IntrNoMem]>; - -// Scalar Saturating Shift Left -def int_aarch64_neon_vqshls : Neon_2Arg_Intrinsic; -def int_aarch64_neon_vqshlu : Neon_2Arg_Intrinsic; - -// Scalar Shift Rouding Left -def int_aarch64_neon_vrshlds : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty], [IntrNoMem]>; -def int_aarch64_neon_vrshldu : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty], [IntrNoMem]>; - -// Scalar Saturating Rounding Shift Left -def int_aarch64_neon_vqrshls : Neon_2Arg_Intrinsic; -def int_aarch64_neon_vqrshlu : Neon_2Arg_Intrinsic; - -// Scalar Reduce Pairwise Add. -def int_aarch64_neon_vpadd : - Intrinsic<[llvm_v1i64_ty], [llvm_v2i64_ty],[IntrNoMem]>; -def int_aarch64_neon_vpfadd : - Intrinsic<[llvm_v1f32_ty], [llvm_v2f32_ty], [IntrNoMem]>; -def int_aarch64_neon_vpfaddq : - Intrinsic<[llvm_v1f64_ty], [llvm_v2f64_ty], [IntrNoMem]>; - -// Scalar Reduce Pairwise Floating Point Max/Min. -def int_aarch64_neon_vpmax : - Intrinsic<[llvm_v1f32_ty], [llvm_v2f32_ty], [IntrNoMem]>; -def int_aarch64_neon_vpmaxq : - Intrinsic<[llvm_v1f64_ty], [llvm_v2f64_ty], [IntrNoMem]>; -def int_aarch64_neon_vpmin : - Intrinsic<[llvm_v1f32_ty], [llvm_v2f32_ty], [IntrNoMem]>; -def int_aarch64_neon_vpminq : - Intrinsic<[llvm_v1f64_ty], [llvm_v2f64_ty], [IntrNoMem]>; - -// Scalar Reduce Pairwise Floating Point Maxnm/Minnm. -def int_aarch64_neon_vpfmaxnm : - Intrinsic<[llvm_v1f32_ty], [llvm_v2f32_ty], [IntrNoMem]>; -def int_aarch64_neon_vpfmaxnmq : - Intrinsic<[llvm_v1f64_ty], [llvm_v2f64_ty], [IntrNoMem]>; -def int_aarch64_neon_vpfminnm : - Intrinsic<[llvm_v1f32_ty], [llvm_v2f32_ty], [IntrNoMem]>; -def int_aarch64_neon_vpfminnmq : - Intrinsic<[llvm_v1f64_ty], [llvm_v2f64_ty], [IntrNoMem]>; - -// Scalar Signed Integer Convert To Floating-point -def int_aarch64_neon_vcvtf32_s32 : - Intrinsic<[llvm_v1f32_ty], [llvm_v1i32_ty], [IntrNoMem]>; -def int_aarch64_neon_vcvtf64_s64 : - Intrinsic<[llvm_v1f64_ty], [llvm_v1i64_ty], [IntrNoMem]>; - -// Scalar Unsigned Integer Convert To Floating-point -def int_aarch64_neon_vcvtf32_u32 : - Intrinsic<[llvm_v1f32_ty], [llvm_v1i32_ty], [IntrNoMem]>; -def int_aarch64_neon_vcvtf64_u64 : - Intrinsic<[llvm_v1f64_ty], [llvm_v1i64_ty], [IntrNoMem]>; - -// Scalar Floating-point Reciprocal Exponent -def int_aarch64_neon_vrecpx : Neon_1Arg_Intrinsic; - -class Neon_Cmp_Intrinsic - : Intrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty, llvm_anyvector_ty], - [IntrNoMem]>; - -// Scalar Compare Equal -def int_aarch64_neon_vceq : Neon_Cmp_Intrinsic; - -// Scalar Compare Greater-Than or Equal -def int_aarch64_neon_vcge : Neon_Cmp_Intrinsic; -def int_aarch64_neon_vchs : Neon_Cmp_Intrinsic; - -// Scalar Compare Less-Than or Equal -def int_aarch64_neon_vclez : Neon_Cmp_Intrinsic; - -// Scalar Compare Less-Than -def int_aarch64_neon_vcltz : Neon_Cmp_Intrinsic; - -// Scalar Compare Greater-Than -def int_aarch64_neon_vcgt : Neon_Cmp_Intrinsic; -def int_aarch64_neon_vchi : Neon_Cmp_Intrinsic; - -// Scalar Compare Bitwise Test Bits -def int_aarch64_neon_vtstd : Neon_Cmp_Intrinsic; - -// Scalar Floating-point Absolute Compare Greater Than Or Equal -def int_aarch64_neon_vcage : Neon_Cmp_Intrinsic; - -// Scalar Floating-point Absolute Compare Greater Than -def int_aarch64_neon_vcagt : Neon_Cmp_Intrinsic; - -// Scalar Signed Saturating Accumulated of Unsigned Value -def int_aarch64_neon_vuqadd : Neon_2Arg_Intrinsic; - -// Scalar Unsigned Saturating Accumulated of Signed Value -def int_aarch64_neon_vsqadd : Neon_2Arg_Intrinsic; - -// Scalar Absolute Value -def int_aarch64_neon_vabs : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty], [IntrNoMem]>; - -// Scalar Negate Value -def int_aarch64_neon_vneg : - Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty], [IntrNoMem]>; - -// Signed Saturating Doubling Multiply-Add Long -def int_aarch64_neon_vqdmlal : Neon_3Arg_Long_Intrinsic; - -// Signed Saturating Doubling Multiply-Subtract Long -def int_aarch64_neon_vqdmlsl : Neon_3Arg_Long_Intrinsic; - -// Signed Saturating Doubling Multiply Long -def int_aarch64_neon_vqdmull : Neon_2Arg_Long_Intrinsic; - -class Neon_2Arg_ShiftImm_Intrinsic - : Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_i32_ty], [IntrNoMem]>; - -class Neon_3Arg_ShiftImm_Intrinsic - : Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty, llvm_i32_ty], - [IntrNoMem]>; - -// Scalar Shift Right (Immediate) -def int_aarch64_neon_vshrds_n : Neon_2Arg_ShiftImm_Intrinsic; -def int_aarch64_neon_vshrdu_n : Neon_2Arg_ShiftImm_Intrinsic; - -// Scalar Rounding Shift Right (Immediate) -def int_aarch64_neon_vrshrds_n : Neon_2Arg_ShiftImm_Intrinsic; -def int_aarch64_neon_vrshrdu_n : Neon_2Arg_ShiftImm_Intrinsic; - -// Scalar Shift Right and Accumulate (Immediate) -def int_aarch64_neon_vsrads_n : Neon_3Arg_ShiftImm_Intrinsic; -def int_aarch64_neon_vsradu_n : Neon_3Arg_ShiftImm_Intrinsic; - -// Scalar Rounding Shift Right and Accumulate (Immediate) -def int_aarch64_neon_vrsrads_n : Neon_3Arg_ShiftImm_Intrinsic; -def int_aarch64_neon_vrsradu_n : Neon_3Arg_ShiftImm_Intrinsic; - -// Scalar Shift Left (Immediate) -def int_aarch64_neon_vshld_n : Neon_2Arg_ShiftImm_Intrinsic; - -// Scalar Saturating Shift Left (Immediate) -def int_aarch64_neon_vqshls_n : Neon_N2V_Intrinsic; -def int_aarch64_neon_vqshlu_n : Neon_N2V_Intrinsic; +// Vector element to element moves +def int_aarch64_neon_vcopy_lane: AdvSIMD_2Vector2Index_Intrinsic; -// Scalar Signed Saturating Shift Left Unsigned (Immediate) -def int_aarch64_neon_vqshlus_n : Neon_N2V_Intrinsic; +let TargetPrefix = "aarch64" in { // All intrinsics start with "llvm.aarch64.". + class AdvSIMD_1Vec_Load_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [LLVMAnyPointerType>], + [IntrReadArgMem]>; + class AdvSIMD_1Vec_Store_Lane_Intrinsic + : Intrinsic<[], [llvm_anyvector_ty, llvm_i64_ty, llvm_anyptr_ty], + [IntrReadWriteArgMem, NoCapture<2>]>; + + class AdvSIMD_2Vec_Load_Intrinsic + : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>], + [LLVMAnyPointerType>], + [IntrReadArgMem]>; + class AdvSIMD_2Vec_Load_Lane_Intrinsic + : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>], + [LLVMMatchType<0>, LLVMMatchType<0>, + llvm_i64_ty, llvm_anyptr_ty], + [IntrReadArgMem]>; + class AdvSIMD_2Vec_Store_Intrinsic + : Intrinsic<[], [llvm_anyvector_ty, LLVMMatchType<0>, + LLVMAnyPointerType>], + [IntrReadWriteArgMem, NoCapture<2>]>; + class AdvSIMD_2Vec_Store_Lane_Intrinsic + : Intrinsic<[], [llvm_anyvector_ty, LLVMMatchType<0>, + llvm_i64_ty, llvm_anyptr_ty], + [IntrReadWriteArgMem, NoCapture<3>]>; + + class AdvSIMD_3Vec_Load_Intrinsic + : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, LLVMMatchType<0>], + [LLVMAnyPointerType>], + [IntrReadArgMem]>; + class AdvSIMD_3Vec_Load_Lane_Intrinsic + : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, LLVMMatchType<0>], + [LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>, + llvm_i64_ty, llvm_anyptr_ty], + [IntrReadArgMem]>; + class AdvSIMD_3Vec_Store_Intrinsic + : Intrinsic<[], [llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMAnyPointerType>], + [IntrReadWriteArgMem, NoCapture<3>]>; + class AdvSIMD_3Vec_Store_Lane_Intrinsic + : Intrinsic<[], [llvm_anyvector_ty, + LLVMMatchType<0>, LLVMMatchType<0>, + llvm_i64_ty, llvm_anyptr_ty], + [IntrReadWriteArgMem, NoCapture<4>]>; + + class AdvSIMD_4Vec_Load_Intrinsic + : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>], + [LLVMAnyPointerType>], + [IntrReadArgMem]>; + class AdvSIMD_4Vec_Load_Lane_Intrinsic + : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>], + [LLVMMatchType<0>, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>, + llvm_i64_ty, llvm_anyptr_ty], + [IntrReadArgMem]>; + class AdvSIMD_4Vec_Store_Intrinsic + : Intrinsic<[], [llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>, + LLVMAnyPointerType>], + [IntrReadWriteArgMem, NoCapture<4>]>; + class AdvSIMD_4Vec_Store_Lane_Intrinsic + : Intrinsic<[], [llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>, + llvm_i64_ty, llvm_anyptr_ty], + [IntrReadWriteArgMem, NoCapture<5>]>; +} -// Shift Right And Insert (Immediate) -def int_aarch64_neon_vsrid_n : Neon_3Arg_ShiftImm_Intrinsic; +// Memory ops -// Shift Left And Insert (Immediate) -def int_aarch64_neon_vslid_n : Neon_3Arg_ShiftImm_Intrinsic; +def int_aarch64_neon_ld1x2 : AdvSIMD_2Vec_Load_Intrinsic; +def int_aarch64_neon_ld1x3 : AdvSIMD_3Vec_Load_Intrinsic; +def int_aarch64_neon_ld1x4 : AdvSIMD_4Vec_Load_Intrinsic; -// Scalar Signed Fixed-point Convert To Floating-Point (Immediate) -def int_aarch64_neon_vcvtf32_n_s32 : - Intrinsic<[llvm_v1f32_ty], [llvm_v1i32_ty, llvm_i32_ty], [IntrNoMem]>; -def int_aarch64_neon_vcvtf64_n_s64 : - Intrinsic<[llvm_v1f64_ty], [llvm_v1i64_ty, llvm_i32_ty], [IntrNoMem]>; +def int_aarch64_neon_st1x2 : AdvSIMD_2Vec_Store_Intrinsic; +def int_aarch64_neon_st1x3 : AdvSIMD_3Vec_Store_Intrinsic; +def int_aarch64_neon_st1x4 : AdvSIMD_4Vec_Store_Intrinsic; -// Scalar Unsigned Fixed-point Convert To Floating-Point (Immediate) -def int_aarch64_neon_vcvtf32_n_u32 : - Intrinsic<[llvm_v1f32_ty], [llvm_v1i32_ty, llvm_i32_ty], [IntrNoMem]>; -def int_aarch64_neon_vcvtf64_n_u64 : - Intrinsic<[llvm_v1f64_ty], [llvm_v1i64_ty, llvm_i32_ty], [IntrNoMem]>; +def int_aarch64_neon_ld2 : AdvSIMD_2Vec_Load_Intrinsic; +def int_aarch64_neon_ld3 : AdvSIMD_3Vec_Load_Intrinsic; +def int_aarch64_neon_ld4 : AdvSIMD_4Vec_Load_Intrinsic; -// Scalar Floating-point Convert To Signed Fixed-point (Immediate) -def int_aarch64_neon_vcvts_n_s32_f32 : - Intrinsic<[llvm_v1i32_ty], [llvm_v1f32_ty, llvm_i32_ty], [IntrNoMem]>; -def int_aarch64_neon_vcvtd_n_s64_f64 : - Intrinsic<[llvm_v1i64_ty], [llvm_v1f64_ty, llvm_i32_ty], [IntrNoMem]>; +def int_aarch64_neon_ld2lane : AdvSIMD_2Vec_Load_Lane_Intrinsic; +def int_aarch64_neon_ld3lane : AdvSIMD_3Vec_Load_Lane_Intrinsic; +def int_aarch64_neon_ld4lane : AdvSIMD_4Vec_Load_Lane_Intrinsic; -// Scalar Floating-point Convert To Unsigned Fixed-point (Immediate) -def int_aarch64_neon_vcvts_n_u32_f32 : - Intrinsic<[llvm_v1i32_ty], [llvm_v1f32_ty, llvm_i32_ty], [IntrNoMem]>; -def int_aarch64_neon_vcvtd_n_u64_f64 : - Intrinsic<[llvm_v1i64_ty], [llvm_v1f64_ty, llvm_i32_ty], [IntrNoMem]>; +def int_aarch64_neon_ld2r : AdvSIMD_2Vec_Load_Intrinsic; +def int_aarch64_neon_ld3r : AdvSIMD_3Vec_Load_Intrinsic; +def int_aarch64_neon_ld4r : AdvSIMD_4Vec_Load_Intrinsic; -class Neon_SHA_Intrinsic - : Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v1i32_ty, llvm_v4i32_ty], - [IntrNoMem]>; +def int_aarch64_neon_st2 : AdvSIMD_2Vec_Store_Intrinsic; +def int_aarch64_neon_st3 : AdvSIMD_3Vec_Store_Intrinsic; +def int_aarch64_neon_st4 : AdvSIMD_4Vec_Store_Intrinsic; -def int_aarch64_neon_sha1c : Neon_SHA_Intrinsic; -def int_aarch64_neon_sha1m : Neon_SHA_Intrinsic; -def int_aarch64_neon_sha1p : Neon_SHA_Intrinsic; +def int_aarch64_neon_st2lane : AdvSIMD_2Vec_Store_Lane_Intrinsic; +def int_aarch64_neon_st3lane : AdvSIMD_3Vec_Store_Lane_Intrinsic; +def int_aarch64_neon_st4lane : AdvSIMD_4Vec_Store_Lane_Intrinsic; + +let TargetPrefix = "aarch64" in { // All intrinsics start with "llvm.aarch64.". + class AdvSIMD_Tbl1_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [llvm_v16i8_ty, LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_Tbl2_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [llvm_v16i8_ty, llvm_v16i8_ty, LLVMMatchType<0>], [IntrNoMem]>; + class AdvSIMD_Tbl3_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty, + LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_Tbl4_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty, + LLVMMatchType<0>], + [IntrNoMem]>; + + class AdvSIMD_Tbx1_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, llvm_v16i8_ty, LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_Tbx2_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, llvm_v16i8_ty, llvm_v16i8_ty, + LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_Tbx3_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, llvm_v16i8_ty, llvm_v16i8_ty, + llvm_v16i8_ty, LLVMMatchType<0>], + [IntrNoMem]>; + class AdvSIMD_Tbx4_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, llvm_v16i8_ty, llvm_v16i8_ty, + llvm_v16i8_ty, llvm_v16i8_ty, LLVMMatchType<0>], + [IntrNoMem]>; +} +def int_aarch64_neon_tbl1 : AdvSIMD_Tbl1_Intrinsic; +def int_aarch64_neon_tbl2 : AdvSIMD_Tbl2_Intrinsic; +def int_aarch64_neon_tbl3 : AdvSIMD_Tbl3_Intrinsic; +def int_aarch64_neon_tbl4 : AdvSIMD_Tbl4_Intrinsic; + +def int_aarch64_neon_tbx1 : AdvSIMD_Tbx1_Intrinsic; +def int_aarch64_neon_tbx2 : AdvSIMD_Tbx2_Intrinsic; +def int_aarch64_neon_tbx3 : AdvSIMD_Tbx3_Intrinsic; +def int_aarch64_neon_tbx4 : AdvSIMD_Tbx4_Intrinsic; + +let TargetPrefix = "aarch64" in { + class Crypto_AES_DataKey_Intrinsic + : Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty, llvm_v16i8_ty], [IntrNoMem]>; + + class Crypto_AES_Data_Intrinsic + : Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty], [IntrNoMem]>; + + // SHA intrinsic taking 5 words of the hash (v4i32, i32) and 4 of the schedule + // (v4i32). + class Crypto_SHA_5Hash4Schedule_Intrinsic + : Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_i32_ty, llvm_v4i32_ty], + [IntrNoMem]>; + + // SHA intrinsic taking 5 words of the hash (v4i32, i32) and 4 of the schedule + // (v4i32). + class Crypto_SHA_1Hash_Intrinsic + : Intrinsic<[llvm_i32_ty], [llvm_i32_ty], [IntrNoMem]>; + + // SHA intrinsic taking 8 words of the schedule + class Crypto_SHA_8Schedule_Intrinsic + : Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty], [IntrNoMem]>; + + // SHA intrinsic taking 12 words of the schedule + class Crypto_SHA_12Schedule_Intrinsic + : Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty, llvm_v4i32_ty], + [IntrNoMem]>; + + // SHA intrinsic taking 8 words of the hash and 4 of the schedule. + class Crypto_SHA_8Hash4Schedule_Intrinsic + : Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty, llvm_v4i32_ty], + [IntrNoMem]>; +} + +// AES +def int_aarch64_crypto_aese : Crypto_AES_DataKey_Intrinsic; +def int_aarch64_crypto_aesd : Crypto_AES_DataKey_Intrinsic; +def int_aarch64_crypto_aesmc : Crypto_AES_Data_Intrinsic; +def int_aarch64_crypto_aesimc : Crypto_AES_Data_Intrinsic; + +// SHA1 +def int_aarch64_crypto_sha1c : Crypto_SHA_5Hash4Schedule_Intrinsic; +def int_aarch64_crypto_sha1p : Crypto_SHA_5Hash4Schedule_Intrinsic; +def int_aarch64_crypto_sha1m : Crypto_SHA_5Hash4Schedule_Intrinsic; +def int_aarch64_crypto_sha1h : Crypto_SHA_1Hash_Intrinsic; + +def int_aarch64_crypto_sha1su0 : Crypto_SHA_12Schedule_Intrinsic; +def int_aarch64_crypto_sha1su1 : Crypto_SHA_8Schedule_Intrinsic; + +// SHA256 +def int_aarch64_crypto_sha256h : Crypto_SHA_8Hash4Schedule_Intrinsic; +def int_aarch64_crypto_sha256h2 : Crypto_SHA_8Hash4Schedule_Intrinsic; +def int_aarch64_crypto_sha256su0 : Crypto_SHA_8Schedule_Intrinsic; +def int_aarch64_crypto_sha256su1 : Crypto_SHA_12Schedule_Intrinsic; + +//===----------------------------------------------------------------------===// +// CRC32 + +let TargetPrefix = "aarch64" in { + +def int_aarch64_crc32b : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], + [IntrNoMem]>; +def int_aarch64_crc32cb : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], + [IntrNoMem]>; +def int_aarch64_crc32h : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], + [IntrNoMem]>; +def int_aarch64_crc32ch : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], + [IntrNoMem]>; +def int_aarch64_crc32w : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], + [IntrNoMem]>; +def int_aarch64_crc32cw : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], + [IntrNoMem]>; +def int_aarch64_crc32x : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i64_ty], + [IntrNoMem]>; +def int_aarch64_crc32cx : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i64_ty], + [IntrNoMem]>; }