X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=include%2Fllvm%2FIntrinsicsARM.td;h=93b1ae1dc887b1985d672e0f301cf30b26daef50;hb=3756e70af69096a82b367ee9667e7720ca2201e4;hp=e9a7e2e91da7e2f99aff06ea8493d033462adb79;hpb=9e699cc72639f25b20e73f51ab06ab7a0884ea62;p=oota-llvm.git diff --git a/include/llvm/IntrinsicsARM.td b/include/llvm/IntrinsicsARM.td index e9a7e2e91da..93b1ae1dc88 100644 --- a/include/llvm/IntrinsicsARM.td +++ b/include/llvm/IntrinsicsARM.td @@ -1,10 +1,10 @@ //===- IntrinsicsARM.td - Defines ARM intrinsics -----------*- tablegen -*-===// -// +// // The LLVM Compiler Infrastructure // // This file is distributed under the University of Illinois Open Source // License. See LICENSE.TXT for details. -// +// //===----------------------------------------------------------------------===// // // This file defines all of the ARM-specific intrinsics. @@ -16,67 +16,136 @@ // TLS let TargetPrefix = "arm" in { // All intrinsics start with "llvm.arm.". - def int_arm_thread_pointer : GCCBuiltin<"__builtin_thread_pointer">, - Intrinsic<[llvm_ptr_ty], [], [IntrNoMem]>; -} + +def int_arm_thread_pointer : GCCBuiltin<"__builtin_thread_pointer">, + Intrinsic<[llvm_ptr_ty], [], [IntrNoMem]>; //===----------------------------------------------------------------------===// -// Advanced SIMD (NEON) +// Saturating Arithmentic + +def int_arm_qadd : GCCBuiltin<"__builtin_arm_qadd">, + Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], + [IntrNoMem, Commutative]>; +def int_arm_qsub : GCCBuiltin<"__builtin_arm_qsub">, + Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; +def int_arm_ssat : GCCBuiltin<"__builtin_arm_ssat">, + Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; +def int_arm_usat : GCCBuiltin<"__builtin_arm_usat">, + Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; -let TargetPrefix = "arm" in { // All intrinsics start with "llvm.arm.". +//===----------------------------------------------------------------------===// +// Load and Store exclusive doubleword - // The following classes do not correspond directly to GCC builtins. - class Neon_1Arg_Intrinsic - : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [IntrNoMem]>; - class Neon_1Arg_Float_Intrinsic - : Intrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem]>; - class Neon_1Arg_Narrow_Intrinsic - : Intrinsic<[llvm_anyint_ty], - [LLVMExtendedElementVectorType<0>], [IntrNoMem]>; - class Neon_1Arg_Long_Intrinsic - : Intrinsic<[llvm_anyint_ty], - [LLVMTruncatedElementVectorType<0>], [IntrNoMem]>; - class Neon_2Arg_Intrinsic - : Intrinsic<[llvm_anyint_ty], [LLVMMatchType<0>, LLVMMatchType<0>], - [IntrNoMem]>; - class Neon_2Arg_Float_Intrinsic - : Intrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>, LLVMMatchType<0>], - [IntrNoMem]>; - class Neon_2Arg_Narrow_Intrinsic - : Intrinsic<[llvm_anyint_ty], - [LLVMExtendedElementVectorType<0>, - LLVMExtendedElementVectorType<0>], - [IntrNoMem]>; - class Neon_2Arg_Long_Intrinsic - : Intrinsic<[llvm_anyint_ty], - [LLVMTruncatedElementVectorType<0>, - LLVMTruncatedElementVectorType<0>], - [IntrNoMem]>; - class Neon_2Arg_Wide_Intrinsic - : Intrinsic<[llvm_anyint_ty], - [LLVMMatchType<0>, LLVMTruncatedElementVectorType<0>], - [IntrNoMem]>; - class Neon_3Arg_Intrinsic - : Intrinsic<[llvm_anyint_ty], - [LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>], - [IntrNoMem]>; - class Neon_3Arg_Long_Intrinsic - : Intrinsic<[llvm_anyint_ty], - [LLVMMatchType<0>, - LLVMTruncatedElementVectorType<0>, - LLVMTruncatedElementVectorType<0>], - [IntrNoMem]>; - class Neon_2Result_Intrinsic - : Intrinsic<[llvm_anyint_ty, LLVMMatchType<0>], - [LLVMMatchType<0>, LLVMMatchType<0>], [IntrNoMem]>; - class Neon_2Result_Float_Intrinsic - : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>], - [LLVMMatchType<0>, LLVMMatchType<0>], [IntrNoMem]>; - class Neon_CvtFxToFP_Intrinsic - : Intrinsic<[llvm_anyfloat_ty], [llvm_anyint_ty, llvm_i32_ty], [IntrNoMem]>; - class Neon_CvtFPToFx_Intrinsic - : Intrinsic<[llvm_anyint_ty], [llvm_anyfloat_ty, llvm_i32_ty], [IntrNoMem]>; -} +def int_arm_strexd : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, + llvm_ptr_ty], [IntrReadWriteArgMem]>; +def int_arm_ldrexd : Intrinsic<[llvm_i32_ty, llvm_i32_ty], [llvm_ptr_ty], + [IntrReadArgMem]>; + +//===----------------------------------------------------------------------===// +// VFP + +def int_arm_get_fpscr : GCCBuiltin<"__builtin_arm_get_fpscr">, + Intrinsic<[llvm_i32_ty], [], [IntrNoMem]>; +def int_arm_set_fpscr : GCCBuiltin<"__builtin_arm_set_fpscr">, + Intrinsic<[], [llvm_i32_ty], []>; +def int_arm_vcvtr : Intrinsic<[llvm_float_ty], [llvm_anyfloat_ty], + [IntrNoMem]>; +def int_arm_vcvtru : Intrinsic<[llvm_float_ty], [llvm_anyfloat_ty], + [IntrNoMem]>; + +//===----------------------------------------------------------------------===// +// Coprocessor + +// Move to coprocessor +def int_arm_mcr : GCCBuiltin<"__builtin_arm_mcr">, + Intrinsic<[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, + llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], []>; +def int_arm_mcr2 : GCCBuiltin<"__builtin_arm_mcr2">, + Intrinsic<[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, + llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], []>; + +// Move from coprocessor +def int_arm_mrc : GCCBuiltin<"__builtin_arm_mrc">, + Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, + llvm_i32_ty, llvm_i32_ty], []>; +def int_arm_mrc2 : GCCBuiltin<"__builtin_arm_mrc2">, + Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, + llvm_i32_ty, llvm_i32_ty], []>; + +// Coprocessor data processing +def int_arm_cdp : GCCBuiltin<"__builtin_arm_cdp">, + Intrinsic<[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, + llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], []>; +def int_arm_cdp2 : GCCBuiltin<"__builtin_arm_cdp2">, + Intrinsic<[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, + llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], []>; + +// Move from two registers to coprocessor +def int_arm_mcrr : GCCBuiltin<"__builtin_arm_mcrr">, + Intrinsic<[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, + llvm_i32_ty, llvm_i32_ty], []>; +def int_arm_mcrr2 : GCCBuiltin<"__builtin_arm_mcrr2">, + Intrinsic<[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, + llvm_i32_ty, llvm_i32_ty], []>; + +//===----------------------------------------------------------------------===// +// Advanced SIMD (NEON) + +// The following classes do not correspond directly to GCC builtins. +class Neon_1Arg_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [LLVMMatchType<0>], [IntrNoMem]>; +class Neon_1Arg_Narrow_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMExtendedElementVectorType<0>], [IntrNoMem]>; +class Neon_2Arg_Intrinsic + : Intrinsic<[llvm_anyvector_ty], [LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem]>; +class Neon_2Arg_Narrow_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMExtendedElementVectorType<0>, + LLVMExtendedElementVectorType<0>], + [IntrNoMem]>; +class Neon_2Arg_Long_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMTruncatedElementVectorType<0>, + LLVMTruncatedElementVectorType<0>], + [IntrNoMem]>; +class Neon_3Arg_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem]>; +class Neon_3Arg_Long_Intrinsic + : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, + LLVMTruncatedElementVectorType<0>, + LLVMTruncatedElementVectorType<0>], + [IntrNoMem]>; +class Neon_CvtFxToFP_Intrinsic + : Intrinsic<[llvm_anyfloat_ty], [llvm_anyint_ty, llvm_i32_ty], [IntrNoMem]>; +class Neon_CvtFPToFx_Intrinsic + : Intrinsic<[llvm_anyint_ty], [llvm_anyfloat_ty, llvm_i32_ty], [IntrNoMem]>; + +// The table operands for VTBL and VTBX consist of 1 to 4 v8i8 vectors. +// Besides the table, VTBL has one other v8i8 argument and VTBX has two. +// Overall, the classes range from 2 to 6 v8i8 arguments. +class Neon_Tbl2Arg_Intrinsic + : Intrinsic<[llvm_v8i8_ty], + [llvm_v8i8_ty, llvm_v8i8_ty], [IntrNoMem]>; +class Neon_Tbl3Arg_Intrinsic + : Intrinsic<[llvm_v8i8_ty], + [llvm_v8i8_ty, llvm_v8i8_ty, llvm_v8i8_ty], [IntrNoMem]>; +class Neon_Tbl4Arg_Intrinsic + : Intrinsic<[llvm_v8i8_ty], + [llvm_v8i8_ty, llvm_v8i8_ty, llvm_v8i8_ty, llvm_v8i8_ty], + [IntrNoMem]>; +class Neon_Tbl5Arg_Intrinsic + : Intrinsic<[llvm_v8i8_ty], + [llvm_v8i8_ty, llvm_v8i8_ty, llvm_v8i8_ty, llvm_v8i8_ty, + llvm_v8i8_ty], [IntrNoMem]>; +class Neon_Tbl6Arg_Intrinsic + : Intrinsic<[llvm_v8i8_ty], + [llvm_v8i8_ty, llvm_v8i8_ty, llvm_v8i8_ty, llvm_v8i8_ty, + llvm_v8i8_ty, llvm_v8i8_ty], [IntrNoMem]>; // Arithmetic ops @@ -91,10 +160,6 @@ let Properties = [IntrNoMem, Commutative] in { def int_arm_neon_vqaddu : Neon_2Arg_Intrinsic; def int_arm_neon_vaddhn : Neon_2Arg_Narrow_Intrinsic; def int_arm_neon_vraddhn : Neon_2Arg_Narrow_Intrinsic; - def int_arm_neon_vaddls : Neon_2Arg_Long_Intrinsic; - def int_arm_neon_vaddlu : Neon_2Arg_Long_Intrinsic; - def int_arm_neon_vaddws : Neon_2Arg_Wide_Intrinsic; - def int_arm_neon_vaddwu : Neon_2Arg_Wide_Intrinsic; // Vector Multiply. def int_arm_neon_vmulp : Neon_2Arg_Intrinsic; @@ -106,28 +171,22 @@ let Properties = [IntrNoMem, Commutative] in { def int_arm_neon_vqdmull : Neon_2Arg_Long_Intrinsic; // Vector Multiply and Accumulate/Subtract. - def int_arm_neon_vmlals : Neon_3Arg_Long_Intrinsic; - def int_arm_neon_vmlalu : Neon_3Arg_Long_Intrinsic; - def int_arm_neon_vmlsls : Neon_3Arg_Long_Intrinsic; - def int_arm_neon_vmlslu : Neon_3Arg_Long_Intrinsic; def int_arm_neon_vqdmlal : Neon_3Arg_Long_Intrinsic; def int_arm_neon_vqdmlsl : Neon_3Arg_Long_Intrinsic; // Vector Maximum. def int_arm_neon_vmaxs : Neon_2Arg_Intrinsic; def int_arm_neon_vmaxu : Neon_2Arg_Intrinsic; - def int_arm_neon_vmaxf : Neon_2Arg_Float_Intrinsic; // Vector Minimum. def int_arm_neon_vmins : Neon_2Arg_Intrinsic; def int_arm_neon_vminu : Neon_2Arg_Intrinsic; - def int_arm_neon_vminf : Neon_2Arg_Float_Intrinsic; // Vector Reciprocal Step. - def int_arm_neon_vrecps : Neon_2Arg_Float_Intrinsic; + def int_arm_neon_vrecps : Neon_2Arg_Intrinsic; // Vector Reciprocal Square Root Step. - def int_arm_neon_vrsqrts : Neon_2Arg_Float_Intrinsic; + def int_arm_neon_vrsqrts : Neon_2Arg_Intrinsic; } // Vector Subtract. @@ -137,74 +196,52 @@ def int_arm_neon_vqsubs : Neon_2Arg_Intrinsic; def int_arm_neon_vqsubu : Neon_2Arg_Intrinsic; def int_arm_neon_vsubhn : Neon_2Arg_Narrow_Intrinsic; def int_arm_neon_vrsubhn : Neon_2Arg_Narrow_Intrinsic; -def int_arm_neon_vsubls : Neon_2Arg_Long_Intrinsic; -def int_arm_neon_vsublu : Neon_2Arg_Long_Intrinsic; -def int_arm_neon_vsubws : Neon_2Arg_Wide_Intrinsic; -def int_arm_neon_vsubwu : Neon_2Arg_Wide_Intrinsic; // Vector Absolute Compare. -let TargetPrefix = "arm" in { - def int_arm_neon_vacged : Intrinsic<[llvm_v2i32_ty], - [llvm_v2f32_ty, llvm_v2f32_ty], - [IntrNoMem]>; - def int_arm_neon_vacgeq : Intrinsic<[llvm_v4i32_ty], - [llvm_v4f32_ty, llvm_v4f32_ty], - [IntrNoMem]>; - def int_arm_neon_vacgtd : Intrinsic<[llvm_v2i32_ty], - [llvm_v2f32_ty, llvm_v2f32_ty], - [IntrNoMem]>; - def int_arm_neon_vacgtq : Intrinsic<[llvm_v4i32_ty], - [llvm_v4f32_ty, llvm_v4f32_ty], - [IntrNoMem]>; -} +def int_arm_neon_vacged : Intrinsic<[llvm_v2i32_ty], + [llvm_v2f32_ty, llvm_v2f32_ty], + [IntrNoMem]>; +def int_arm_neon_vacgeq : Intrinsic<[llvm_v4i32_ty], + [llvm_v4f32_ty, llvm_v4f32_ty], + [IntrNoMem]>; +def int_arm_neon_vacgtd : Intrinsic<[llvm_v2i32_ty], + [llvm_v2f32_ty, llvm_v2f32_ty], + [IntrNoMem]>; +def int_arm_neon_vacgtq : Intrinsic<[llvm_v4i32_ty], + [llvm_v4f32_ty, llvm_v4f32_ty], + [IntrNoMem]>; // Vector Absolute Differences. def int_arm_neon_vabds : Neon_2Arg_Intrinsic; def int_arm_neon_vabdu : Neon_2Arg_Intrinsic; -def int_arm_neon_vabdf : Neon_2Arg_Float_Intrinsic; -def int_arm_neon_vabdls : Neon_2Arg_Long_Intrinsic; -def int_arm_neon_vabdlu : Neon_2Arg_Long_Intrinsic; - -// Vector Absolute Difference and Accumulate. -def int_arm_neon_vabas : Neon_3Arg_Intrinsic; -def int_arm_neon_vabau : Neon_3Arg_Intrinsic; -def int_arm_neon_vabals : Neon_3Arg_Long_Intrinsic; -def int_arm_neon_vabalu : Neon_3Arg_Long_Intrinsic; // Vector Pairwise Add. -def int_arm_neon_vpaddi : Neon_2Arg_Intrinsic; -def int_arm_neon_vpaddf : Neon_2Arg_Float_Intrinsic; +def int_arm_neon_vpadd : Neon_2Arg_Intrinsic; // Vector Pairwise Add Long. // Note: This is different than the other "long" NEON intrinsics because // the result vector has half as many elements as the source vector. // The source and destination vector types must be specified separately. -let TargetPrefix = "arm" in { - def int_arm_neon_vpaddls : Intrinsic<[llvm_anyint_ty], [llvm_anyint_ty], - [IntrNoMem]>; - def int_arm_neon_vpaddlu : Intrinsic<[llvm_anyint_ty], [llvm_anyint_ty], - [IntrNoMem]>; -} +def int_arm_neon_vpaddls : Intrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty], + [IntrNoMem]>; +def int_arm_neon_vpaddlu : Intrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty], + [IntrNoMem]>; // Vector Pairwise Add and Accumulate Long. // Note: This is similar to vpaddl but the destination vector also appears // as the first argument. -let TargetPrefix = "arm" in { - def int_arm_neon_vpadals : Intrinsic<[llvm_anyint_ty], - [LLVMMatchType<0>, llvm_anyint_ty], - [IntrNoMem]>; - def int_arm_neon_vpadalu : Intrinsic<[llvm_anyint_ty], - [LLVMMatchType<0>, llvm_anyint_ty], - [IntrNoMem]>; -} +def int_arm_neon_vpadals : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, llvm_anyvector_ty], + [IntrNoMem]>; +def int_arm_neon_vpadalu : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, llvm_anyvector_ty], + [IntrNoMem]>; // Vector Pairwise Maximum and Minimum. def int_arm_neon_vpmaxs : Neon_2Arg_Intrinsic; def int_arm_neon_vpmaxu : Neon_2Arg_Intrinsic; -def int_arm_neon_vpmaxf : Neon_2Arg_Float_Intrinsic; def int_arm_neon_vpmins : Neon_2Arg_Intrinsic; def int_arm_neon_vpminu : Neon_2Arg_Intrinsic; -def int_arm_neon_vpminf : Neon_2Arg_Float_Intrinsic; // Vector Shifts: // @@ -259,7 +296,6 @@ def int_arm_neon_vshiftins : Neon_3Arg_Intrinsic; // Vector Absolute Value and Saturating Absolute Value. def int_arm_neon_vabs : Neon_1Arg_Intrinsic; -def int_arm_neon_vabsf : Neon_1Arg_Float_Intrinsic; def int_arm_neon_vqabs : Neon_1Arg_Intrinsic; // Vector Saturating Negate. @@ -274,11 +310,9 @@ def int_arm_neon_vcnt : Neon_1Arg_Intrinsic; // Vector Reciprocal Estimate. def int_arm_neon_vrecpe : Neon_1Arg_Intrinsic; -def int_arm_neon_vrecpef : Neon_1Arg_Float_Intrinsic; // Vector Reciprocal Square Root Estimate. def int_arm_neon_vrsqrte : Neon_1Arg_Intrinsic; -def int_arm_neon_vrsqrtef : Neon_1Arg_Float_Intrinsic; // Vector Conversions Between Floating-point and Fixed-point. def int_arm_neon_vcvtfp2fxs : Neon_CvtFPToFx_Intrinsic; @@ -286,86 +320,110 @@ def int_arm_neon_vcvtfp2fxu : Neon_CvtFPToFx_Intrinsic; def int_arm_neon_vcvtfxs2fp : Neon_CvtFxToFP_Intrinsic; def int_arm_neon_vcvtfxu2fp : Neon_CvtFxToFP_Intrinsic; -// Narrowing and Lengthening Vector Moves. -def int_arm_neon_vmovn : Neon_1Arg_Narrow_Intrinsic; +// Vector Conversions Between Half-Precision and Single-Precision. +def int_arm_neon_vcvtfp2hf + : Intrinsic<[llvm_v4i16_ty], [llvm_v4f32_ty], [IntrNoMem]>; +def int_arm_neon_vcvthf2fp + : Intrinsic<[llvm_v4f32_ty], [llvm_v4i16_ty], [IntrNoMem]>; + +// Narrowing Saturating Vector Moves. def int_arm_neon_vqmovns : Neon_1Arg_Narrow_Intrinsic; def int_arm_neon_vqmovnu : Neon_1Arg_Narrow_Intrinsic; def int_arm_neon_vqmovnsu : Neon_1Arg_Narrow_Intrinsic; -def int_arm_neon_vmovls : Neon_1Arg_Long_Intrinsic; -def int_arm_neon_vmovlu : Neon_1Arg_Long_Intrinsic; - -// Vector Transpose. -def int_arm_neon_vtrni : Neon_2Result_Intrinsic; -def int_arm_neon_vtrnf : Neon_2Result_Float_Intrinsic; - -// Vector Interleave (vzip). -def int_arm_neon_vzipi : Neon_2Result_Intrinsic; -def int_arm_neon_vzipf : Neon_2Result_Float_Intrinsic; - -// Vector Deinterleave (vuzp). -def int_arm_neon_vuzpi : Neon_2Result_Intrinsic; -def int_arm_neon_vuzpf : Neon_2Result_Float_Intrinsic; - -let TargetPrefix = "arm" in { - - // De-interleaving vector loads from N-element structures. - def int_arm_neon_vld1i : Intrinsic<[llvm_anyint_ty], - [llvm_ptr_ty], [IntrReadArgMem]>; - def int_arm_neon_vld1f : Intrinsic<[llvm_anyfloat_ty], - [llvm_ptr_ty], [IntrReadArgMem]>; - def int_arm_neon_vld2i : Intrinsic<[llvm_anyint_ty, LLVMMatchType<0>], - [llvm_ptr_ty], [IntrReadArgMem]>; - def int_arm_neon_vld2f : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>], - [llvm_ptr_ty], [IntrReadArgMem]>; - def int_arm_neon_vld3i : Intrinsic<[llvm_anyint_ty, LLVMMatchType<0>, - LLVMMatchType<0>], - [llvm_ptr_ty], [IntrReadArgMem]>; - def int_arm_neon_vld3f : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>, - LLVMMatchType<0>], - [llvm_ptr_ty], [IntrReadArgMem]>; - def int_arm_neon_vld4i : Intrinsic<[llvm_anyint_ty, LLVMMatchType<0>, - LLVMMatchType<0>, LLVMMatchType<0>], - [llvm_ptr_ty], [IntrReadArgMem]>; - def int_arm_neon_vld4f : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>, - LLVMMatchType<0>, LLVMMatchType<0>], - [llvm_ptr_ty], [IntrReadArgMem]>; - - // Interleaving vector stores from N-element structures. - def int_arm_neon_vst1i : Intrinsic<[llvm_void_ty], - [llvm_ptr_ty, llvm_anyint_ty], - [IntrWriteArgMem]>; - def int_arm_neon_vst1f : Intrinsic<[llvm_void_ty], - [llvm_ptr_ty, llvm_anyfloat_ty], - [IntrWriteArgMem]>; - def int_arm_neon_vst2i : Intrinsic<[llvm_void_ty], - [llvm_ptr_ty, llvm_anyint_ty, - LLVMMatchType<0>], [IntrWriteArgMem]>; - def int_arm_neon_vst2f : Intrinsic<[llvm_void_ty], - [llvm_ptr_ty, llvm_anyfloat_ty, - LLVMMatchType<0>], [IntrWriteArgMem]>; - def int_arm_neon_vst3i : Intrinsic<[llvm_void_ty], - [llvm_ptr_ty, llvm_anyint_ty, - LLVMMatchType<0>, LLVMMatchType<0>], - [IntrWriteArgMem]>; - def int_arm_neon_vst3f : Intrinsic<[llvm_void_ty], - [llvm_ptr_ty, llvm_anyfloat_ty, - LLVMMatchType<0>, LLVMMatchType<0>], - [IntrWriteArgMem]>; - def int_arm_neon_vst4i : Intrinsic<[llvm_void_ty], - [llvm_ptr_ty, llvm_anyint_ty, - LLVMMatchType<0>, LLVMMatchType<0>, - LLVMMatchType<0>], [IntrWriteArgMem]>; - def int_arm_neon_vst4f : Intrinsic<[llvm_void_ty], - [llvm_ptr_ty, llvm_anyfloat_ty, - LLVMMatchType<0>, LLVMMatchType<0>, - LLVMMatchType<0>], [IntrWriteArgMem]>; - - // Vector Table Lookup - def int_arm_neon_vtbl : Intrinsic<[llvm_v8i8_ty], - [llvm_anyint_ty, llvm_v8i8_ty], - [IntrNoMem]>; - // Vector Table Extension - def int_arm_neon_vtbx : Intrinsic<[llvm_v8i8_ty], - [llvm_v8i8_ty, llvm_anyint_ty, - llvm_v8i8_ty], [IntrNoMem]>; -} + +// Vector Table Lookup. +// The first 1-4 arguments are the table. +def int_arm_neon_vtbl1 : Neon_Tbl2Arg_Intrinsic; +def int_arm_neon_vtbl2 : Neon_Tbl3Arg_Intrinsic; +def int_arm_neon_vtbl3 : Neon_Tbl4Arg_Intrinsic; +def int_arm_neon_vtbl4 : Neon_Tbl5Arg_Intrinsic; + +// Vector Table Extension. +// Some elements of the destination vector may not be updated, so the original +// value of that vector is passed as the first argument. The next 1-4 +// arguments after that are the table. +def int_arm_neon_vtbx1 : Neon_Tbl3Arg_Intrinsic; +def int_arm_neon_vtbx2 : Neon_Tbl4Arg_Intrinsic; +def int_arm_neon_vtbx3 : Neon_Tbl5Arg_Intrinsic; +def int_arm_neon_vtbx4 : Neon_Tbl6Arg_Intrinsic; + +// De-interleaving vector loads from N-element structures. +// Source operands are the address and alignment. +def int_arm_neon_vld1 : Intrinsic<[llvm_anyvector_ty], + [llvm_ptr_ty, llvm_i32_ty], + [IntrReadArgMem]>; +def int_arm_neon_vld2 : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>], + [llvm_ptr_ty, llvm_i32_ty], + [IntrReadArgMem]>; +def int_arm_neon_vld3 : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>], + [llvm_ptr_ty, llvm_i32_ty], + [IntrReadArgMem]>; +def int_arm_neon_vld4 : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>], + [llvm_ptr_ty, llvm_i32_ty], + [IntrReadArgMem]>; + +// Vector load N-element structure to one lane. +// Source operands are: the address, the N input vectors (since only one +// lane is assigned), the lane number, and the alignment. +def int_arm_neon_vld2lane : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>], + [llvm_ptr_ty, LLVMMatchType<0>, + LLVMMatchType<0>, llvm_i32_ty, + llvm_i32_ty], [IntrReadArgMem]>; +def int_arm_neon_vld3lane : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>], + [llvm_ptr_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>, + llvm_i32_ty, llvm_i32_ty], + [IntrReadArgMem]>; +def int_arm_neon_vld4lane : Intrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>], + [llvm_ptr_ty, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>, + LLVMMatchType<0>, llvm_i32_ty, + llvm_i32_ty], [IntrReadArgMem]>; + +// Interleaving vector stores from N-element structures. +// Source operands are: the address, the N vectors, and the alignment. +def int_arm_neon_vst1 : Intrinsic<[], + [llvm_ptr_ty, llvm_anyvector_ty, + llvm_i32_ty], [IntrReadWriteArgMem]>; +def int_arm_neon_vst2 : Intrinsic<[], + [llvm_ptr_ty, llvm_anyvector_ty, + LLVMMatchType<0>, llvm_i32_ty], + [IntrReadWriteArgMem]>; +def int_arm_neon_vst3 : Intrinsic<[], + [llvm_ptr_ty, llvm_anyvector_ty, + LLVMMatchType<0>, LLVMMatchType<0>, + llvm_i32_ty], [IntrReadWriteArgMem]>; +def int_arm_neon_vst4 : Intrinsic<[], + [llvm_ptr_ty, llvm_anyvector_ty, + LLVMMatchType<0>, LLVMMatchType<0>, + LLVMMatchType<0>, llvm_i32_ty], + [IntrReadWriteArgMem]>; + +// Vector store N-element structure from one lane. +// Source operands are: the address, the N vectors, the lane number, and +// the alignment. +def int_arm_neon_vst2lane : Intrinsic<[], + [llvm_ptr_ty, llvm_anyvector_ty, + LLVMMatchType<0>, llvm_i32_ty, + llvm_i32_ty], [IntrReadWriteArgMem]>; +def int_arm_neon_vst3lane : Intrinsic<[], + [llvm_ptr_ty, llvm_anyvector_ty, + LLVMMatchType<0>, LLVMMatchType<0>, + llvm_i32_ty, llvm_i32_ty], + [IntrReadWriteArgMem]>; +def int_arm_neon_vst4lane : Intrinsic<[], + [llvm_ptr_ty, llvm_anyvector_ty, + LLVMMatchType<0>, LLVMMatchType<0>, + LLVMMatchType<0>, llvm_i32_ty, + llvm_i32_ty], [IntrReadWriteArgMem]>; + +// Vector bitwise select. +def int_arm_neon_vbsl : Intrinsic<[llvm_anyvector_ty], + [LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem]>; + +} // end TargetPrefix