1//===--- arm_sve.td - ARM SVE compiler interface ------------------------===// 2// 3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4// See https://llvm.org/LICENSE.txt for license information. 5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6// 7//===----------------------------------------------------------------------===// 8// 9// This file defines the TableGen definitions from which the ARM SVE header 10// file will be generated. See: 11// 12// https://developer.arm.com/architectures/system-architectures/software-standards/acle 13// 14//===----------------------------------------------------------------------===// 15 16include "arm_sve_sme_incl.td" 17 18//////////////////////////////////////////////////////////////////////////////// 19// Loads 20 21// Load one vector (scalar base) 22def SVLD1 : MInst<"svld1[_{2}]", "dPc", "csilUcUsUiUlhfd", [IsLoad, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_ld1">; 23def SVLD1SB : MInst<"svld1sb_{d}", "dPS", "silUsUiUl", [IsLoad, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_ld1">; 24def SVLD1UB : MInst<"svld1ub_{d}", "dPW", "silUsUiUl", [IsLoad, IsZExtReturn, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_ld1">; 25def SVLD1SH : MInst<"svld1sh_{d}", "dPT", "ilUiUl", [IsLoad, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_ld1">; 26def SVLD1UH : MInst<"svld1uh_{d}", "dPX", "ilUiUl", [IsLoad, IsZExtReturn, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_ld1">; 27def SVLD1SW : MInst<"svld1sw_{d}", "dPU", "lUl", [IsLoad, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_ld1">; 28def SVLD1UW : MInst<"svld1uw_{d}", "dPY", "lUl", [IsLoad, IsZExtReturn, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_ld1">; 29 30let TargetGuard = "sve,bf16" in { 31 def SVLD1_BF : MInst<"svld1[_{2}]", "dPc", "b", [IsLoad, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_ld1">; 32 def SVLD1_VNUM_BF : MInst<"svld1_vnum[_{2}]", "dPcl", "b", [IsLoad, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_ld1">; 33} 34 35// Load one vector (scalar base, VL displacement) 36def SVLD1_VNUM : MInst<"svld1_vnum[_{2}]", "dPcl", "csilUcUsUiUlhfd", [IsLoad, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_ld1">; 37def SVLD1SB_VNUM : MInst<"svld1sb_vnum_{d}", "dPSl", "silUsUiUl", [IsLoad, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_ld1">; 38def SVLD1UB_VNUM : MInst<"svld1ub_vnum_{d}", "dPWl", "silUsUiUl", [IsLoad, IsZExtReturn, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_ld1">; 39def SVLD1SH_VNUM : MInst<"svld1sh_vnum_{d}", "dPTl", "ilUiUl", [IsLoad, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_ld1">; 40def SVLD1UH_VNUM : MInst<"svld1uh_vnum_{d}", "dPXl", "ilUiUl", [IsLoad, IsZExtReturn, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_ld1">; 41def SVLD1SW_VNUM : MInst<"svld1sw_vnum_{d}", "dPUl", "lUl", [IsLoad, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_ld1">; 42def SVLD1UW_VNUM : MInst<"svld1uw_vnum_{d}", "dPYl", "lUl", [IsLoad, IsZExtReturn, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_ld1">; 43 44// Load one vector (vector base) 45def SVLD1_GATHER_BASES_U : MInst<"svld1_gather[_{2}base]_{d}", "dPu", "ilUiUlfd", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ld1_gather_scalar_offset">; 46def SVLD1SB_GATHER_BASES_U : MInst<"svld1sb_gather[_{2}base]_{d}", "dPu", "ilUiUl", [IsGatherLoad], MemEltTyInt8, "aarch64_sve_ld1_gather_scalar_offset">; 47def SVLD1UB_GATHER_BASES_U : MInst<"svld1ub_gather[_{2}base]_{d}", "dPu", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ld1_gather_scalar_offset">; 48def SVLD1SH_GATHER_BASES_U : MInst<"svld1sh_gather[_{2}base]_{d}", "dPu", "ilUiUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ld1_gather_scalar_offset">; 49def SVLD1UH_GATHER_BASES_U : MInst<"svld1uh_gather[_{2}base]_{d}", "dPu", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_scalar_offset">; 50def SVLD1SW_GATHER_BASES_U : MInst<"svld1sw_gather[_{2}base]_{d}", "dPu", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ld1_gather_scalar_offset">; 51def SVLD1UW_GATHER_BASES_U : MInst<"svld1uw_gather[_{2}base]_{d}", "dPu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ld1_gather_scalar_offset">; 52 53// Load one vector (scalar base, signed vector offset in bytes) 54def SVLD1_GATHER_64B_OFFSETS_S : MInst<"svld1_gather_[{3}]offset[_{d}]", "dPcx", "lUld", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ld1_gather">; 55def SVLD1SB_GATHER_64B_OFFSETS_S : MInst<"svld1sb_gather_[{3}]offset_{d}", "dPSx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ld1_gather">; 56def SVLD1UB_GATHER_64B_OFFSETS_S : MInst<"svld1ub_gather_[{3}]offset_{d}", "dPWx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ld1_gather">; 57def SVLD1SH_GATHER_64B_OFFSETS_S : MInst<"svld1sh_gather_[{3}]offset_{d}", "dPTx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ld1_gather">; 58def SVLD1UH_GATHER_64B_OFFSETS_S : MInst<"svld1uh_gather_[{3}]offset_{d}", "dPXx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather">; 59def SVLD1SW_GATHER_64B_OFFSETS_S : MInst<"svld1sw_gather_[{3}]offset_{d}", "dPUx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ld1_gather">; 60def SVLD1UW_GATHER_64B_OFFSETS_S : MInst<"svld1uw_gather_[{3}]offset_{d}", "dPYx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ld1_gather">; 61 62def SVLD1_GATHER_32B_OFFSETS_S : MInst<"svld1_gather_[{3}]offset[_{d}]", "dPcx", "iUif", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ld1_gather_sxtw">; 63def SVLD1SB_GATHER_32B_OFFSETS_S : MInst<"svld1sb_gather_[{3}]offset_{d}", "dPSx", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ld1_gather_sxtw">; 64def SVLD1UB_GATHER_32B_OFFSETS_S : MInst<"svld1ub_gather_[{3}]offset_{d}", "dPWx", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ld1_gather_sxtw">; 65def SVLD1SH_GATHER_32B_OFFSETS_S : MInst<"svld1sh_gather_[{3}]offset_{d}", "dPTx", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ld1_gather_sxtw">; 66def SVLD1UH_GATHER_32B_OFFSETS_S : MInst<"svld1uh_gather_[{3}]offset_{d}", "dPXx", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_sxtw">; 67 68// Load one vector (scalar base, unsigned vector offset in bytes) 69def SVLD1_GATHER_64B_OFFSETS_U : MInst<"svld1_gather_[{3}]offset[_{d}]", "dPcu", "lUld", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ld1_gather">; 70def SVLD1SB_GATHER_64B_OFFSETS_U : MInst<"svld1sb_gather_[{3}]offset_{d}", "dPSu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ld1_gather">; 71def SVLD1UB_GATHER_64B_OFFSETS_U : MInst<"svld1ub_gather_[{3}]offset_{d}", "dPWu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ld1_gather">; 72def SVLD1SH_GATHER_64B_OFFSETS_U : MInst<"svld1sh_gather_[{3}]offset_{d}", "dPTu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ld1_gather">; 73def SVLD1UH_GATHER_64B_OFFSETS_U : MInst<"svld1uh_gather_[{3}]offset_{d}", "dPXu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather">; 74def SVLD1SW_GATHER_64B_OFFSETS_U : MInst<"svld1sw_gather_[{3}]offset_{d}", "dPUu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ld1_gather">; 75def SVLD1UW_GATHER_64B_OFFSETS_U : MInst<"svld1uw_gather_[{3}]offset_{d}", "dPYu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ld1_gather">; 76 77def SVLD1_GATHER_32B_OFFSETS_U : MInst<"svld1_gather_[{3}]offset[_{d}]", "dPcu", "iUif", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ld1_gather_uxtw">; 78def SVLD1SB_GATHER_32B_OFFSETS_U : MInst<"svld1sb_gather_[{3}]offset_{d}", "dPSu", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ld1_gather_uxtw">; 79def SVLD1UB_GATHER_32B_OFFSETS_U : MInst<"svld1ub_gather_[{3}]offset_{d}", "dPWu", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ld1_gather_uxtw">; 80def SVLD1SH_GATHER_32B_OFFSETS_U : MInst<"svld1sh_gather_[{3}]offset_{d}", "dPTu", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ld1_gather_uxtw">; 81def SVLD1UH_GATHER_32B_OFFSETS_U : MInst<"svld1uh_gather_[{3}]offset_{d}", "dPXu", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_uxtw">; 82 83// Load one vector (vector base, signed scalar offset in bytes) 84def SVLD1_GATHER_OFFSET_S : MInst<"svld1_gather[_{2}base]_offset_{d}", "dPul", "ilUiUlfd", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ld1_gather_scalar_offset">; 85def SVLD1SB_GATHER_OFFSET_S : MInst<"svld1sb_gather[_{2}base]_offset_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ld1_gather_scalar_offset">; 86def SVLD1UB_GATHER_OFFSET_S : MInst<"svld1ub_gather[_{2}base]_offset_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ld1_gather_scalar_offset">; 87def SVLD1SH_GATHER_OFFSET_S : MInst<"svld1sh_gather[_{2}base]_offset_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ld1_gather_scalar_offset">; 88def SVLD1UH_GATHER_OFFSET_S : MInst<"svld1uh_gather[_{2}base]_offset_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_scalar_offset">; 89def SVLD1SW_GATHER_OFFSET_S : MInst<"svld1sw_gather[_{2}base]_offset_{d}", "dPul", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ld1_gather_scalar_offset">; 90def SVLD1UW_GATHER_OFFSET_S : MInst<"svld1uw_gather[_{2}base]_offset_{d}", "dPul", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ld1_gather_scalar_offset">; 91 92// Load one vector (scalar base, signed vector index) 93def SVLD1_GATHER_64B_INDICES_S : MInst<"svld1_gather_[{3}]index[_{d}]", "dPcx", "lUld", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ld1_gather_index">; 94def SVLD1SH_GATHER_64B_INDICES_S : MInst<"svld1sh_gather_[{3}]index_{d}", "dPTx", "lUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ld1_gather_index">; 95def SVLD1UH_GATHER_64B_INDICES_S : MInst<"svld1uh_gather_[{3}]index_{d}", "dPXx", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_index">; 96def SVLD1SW_GATHER_64B_INDICES_S : MInst<"svld1sw_gather_[{3}]index_{d}", "dPUx", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ld1_gather_index">; 97def SVLD1UW_GATHER_64B_INDICES_S : MInst<"svld1uw_gather_[{3}]index_{d}", "dPYx", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ld1_gather_index">; 98 99def SVLD1_GATHER_32B_INDICES_S : MInst<"svld1_gather_[{3}]index[_{d}]", "dPcx", "iUif", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ld1_gather_sxtw_index">; 100def SVLD1SH_GATHER_32B_INDICES_S : MInst<"svld1sh_gather_[{3}]index_{d}", "dPTx", "iUi", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ld1_gather_sxtw_index">; 101def SVLD1UH_GATHER_32B_INDICES_S : MInst<"svld1uh_gather_[{3}]index_{d}", "dPXx", "iUi", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_sxtw_index">; 102 103// Load one vector (scalar base, unsigned vector index) 104def SVLD1_GATHER_64B_INDICES_U : MInst<"svld1_gather_[{3}]index[_{d}]", "dPcu", "lUld", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ld1_gather_index">; 105def SVLD1SH_GATHER_64B_INDICES_U : MInst<"svld1sh_gather_[{3}]index_{d}", "dPTu", "lUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ld1_gather_index">; 106def SVLD1UH_GATHER_64B_INDICES_U : MInst<"svld1uh_gather_[{3}]index_{d}", "dPXu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_index">; 107def SVLD1SW_GATHER_64B_INDICES_U : MInst<"svld1sw_gather_[{3}]index_{d}", "dPUu", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ld1_gather_index">; 108def SVLD1UW_GATHER_64B_INDICES_U : MInst<"svld1uw_gather_[{3}]index_{d}", "dPYu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ld1_gather_index">; 109 110def SVLD1_GATHER_32B_INDICES_U : MInst<"svld1_gather_[{3}]index[_{d}]", "dPcu", "iUif", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ld1_gather_uxtw_index">; 111def SVLD1SH_GATHER_32B_INDICES_U : MInst<"svld1sh_gather_[{3}]index_{d}", "dPTu", "iUi", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ld1_gather_uxtw_index">; 112def SVLD1UH_GATHER_32B_INDICES_U : MInst<"svld1uh_gather_[{3}]index_{d}", "dPXu", "iUi", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_uxtw_index">; 113 114// Load one vector (vector base, signed scalar index) 115def SVLD1_GATHER_INDEX_S : MInst<"svld1_gather[_{2}base]_index_{d}", "dPul", "ilUiUlfd", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ld1_gather_scalar_offset">; 116def SVLD1SH_GATHER_INDEX_S : MInst<"svld1sh_gather[_{2}base]_index_{d}", "dPul", "ilUiUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ld1_gather_scalar_offset">; 117def SVLD1UH_GATHER_INDEX_S : MInst<"svld1uh_gather[_{2}base]_index_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ld1_gather_scalar_offset">; 118def SVLD1SW_GATHER_INDEX_S : MInst<"svld1sw_gather[_{2}base]_index_{d}", "dPul", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ld1_gather_scalar_offset">; 119def SVLD1UW_GATHER_INDEX_S : MInst<"svld1uw_gather[_{2}base]_index_{d}", "dPul", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ld1_gather_scalar_offset">; 120 121 122// First-faulting load one vector (scalar base) 123def SVLDFF1 : MInst<"svldff1[_{2}]", "dPc", "csilUcUsUiUlhfd", [IsLoad], MemEltTyDefault, "aarch64_sve_ldff1">; 124def SVLDFF1SB : MInst<"svldff1sb_{d}", "dPS", "silUsUiUl", [IsLoad], MemEltTyInt8, "aarch64_sve_ldff1">; 125def SVLDFF1UB : MInst<"svldff1ub_{d}", "dPW", "silUsUiUl", [IsLoad, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldff1">; 126def SVLDFF1SH : MInst<"svldff1sh_{d}", "dPT", "ilUiUl", [IsLoad], MemEltTyInt16, "aarch64_sve_ldff1">; 127def SVLDFF1UH : MInst<"svldff1uh_{d}", "dPX", "ilUiUl", [IsLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1">; 128def SVLDFF1SW : MInst<"svldff1sw_{d}", "dPU", "lUl", [IsLoad], MemEltTyInt32, "aarch64_sve_ldff1">; 129def SVLDFF1UW : MInst<"svldff1uw_{d}", "dPY", "lUl", [IsLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1">; 130 131// First-faulting load one vector (scalar base, VL displacement) 132def SVLDFF1_VNUM : MInst<"svldff1_vnum[_{2}]", "dPcl", "csilUcUsUiUlhfd", [IsLoad], MemEltTyDefault, "aarch64_sve_ldff1">; 133def SVLDFF1SB_VNUM : MInst<"svldff1sb_vnum_{d}", "dPSl", "silUsUiUl", [IsLoad], MemEltTyInt8, "aarch64_sve_ldff1">; 134def SVLDFF1UB_VNUM : MInst<"svldff1ub_vnum_{d}", "dPWl", "silUsUiUl", [IsLoad, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldff1">; 135def SVLDFF1SH_VNUM : MInst<"svldff1sh_vnum_{d}", "dPTl", "ilUiUl", [IsLoad], MemEltTyInt16, "aarch64_sve_ldff1">; 136def SVLDFF1UH_VNUM : MInst<"svldff1uh_vnum_{d}", "dPXl", "ilUiUl", [IsLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1">; 137def SVLDFF1SW_VNUM : MInst<"svldff1sw_vnum_{d}", "dPUl", "lUl", [IsLoad], MemEltTyInt32, "aarch64_sve_ldff1">; 138def SVLDFF1UW_VNUM : MInst<"svldff1uw_vnum_{d}", "dPYl", "lUl", [IsLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1">; 139 140let TargetGuard = "sve,bf16" in { 141 def SVLDFF1_BF : MInst<"svldff1[_{2}]", "dPc", "b", [IsLoad], MemEltTyDefault, "aarch64_sve_ldff1">; 142 def SVLDFF1_VNUM_BF : MInst<"svldff1_vnum[_{2}]", "dPcl", "b", [IsLoad], MemEltTyDefault, "aarch64_sve_ldff1">; 143} 144 145// First-faulting load one vector (vector base) 146def SVLDFF1_GATHER_BASES_U : MInst<"svldff1_gather[_{2}base]_{d}", "dPu", "ilUiUlfd", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldff1_gather_scalar_offset">; 147def SVLDFF1SB_GATHER_BASES_U : MInst<"svldff1sb_gather[_{2}base]_{d}", "dPu", "ilUiUl", [IsGatherLoad], MemEltTyInt8, "aarch64_sve_ldff1_gather_scalar_offset">; 148def SVLDFF1UB_GATHER_BASES_U : MInst<"svldff1ub_gather[_{2}base]_{d}", "dPu", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldff1_gather_scalar_offset">; 149def SVLDFF1SH_GATHER_BASES_U : MInst<"svldff1sh_gather[_{2}base]_{d}", "dPu", "ilUiUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldff1_gather_scalar_offset">; 150def SVLDFF1UH_GATHER_BASES_U : MInst<"svldff1uh_gather[_{2}base]_{d}", "dPu", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_scalar_offset">; 151def SVLDFF1SW_GATHER_BASES_U : MInst<"svldff1sw_gather[_{2}base]_{d}", "dPu", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ldff1_gather_scalar_offset">; 152def SVLDFF1UW_GATHER_BASES_U : MInst<"svldff1uw_gather[_{2}base]_{d}", "dPu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1_gather_scalar_offset">; 153 154// First-faulting load one vector (scalar base, signed vector offset in bytes) 155def SVLDFF1_GATHER_64B_OFFSETS_S : MInst<"svldff1_gather_[{3}]offset[_{d}]", "dPcx", "lUld", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldff1_gather">; 156def SVLDFF1SB_GATHER_64B_OFFSETS_S : MInst<"svldff1sb_gather_[{3}]offset_{d}", "dPSx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldff1_gather">; 157def SVLDFF1UB_GATHER_64B_OFFSETS_S : MInst<"svldff1ub_gather_[{3}]offset_{d}", "dPWx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldff1_gather">; 158def SVLDFF1SH_GATHER_64B_OFFSETS_S : MInst<"svldff1sh_gather_[{3}]offset_{d}", "dPTx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldff1_gather">; 159def SVLDFF1UH_GATHER_64B_OFFSETS_S : MInst<"svldff1uh_gather_[{3}]offset_{d}", "dPXx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather">; 160def SVLDFF1SW_GATHER_64B_OFFSETS_S : MInst<"svldff1sw_gather_[{3}]offset_{d}", "dPUx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ldff1_gather">; 161def SVLDFF1UW_GATHER_64B_OFFSETS_S : MInst<"svldff1uw_gather_[{3}]offset_{d}", "dPYx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1_gather">; 162 163def SVLDFF1_GATHER_32B_OFFSETS_S : MInst<"svldff1_gather_[{3}]offset[_{d}]", "dPcx", "iUif", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldff1_gather_sxtw">; 164def SVLDFF1SB_GATHER_32B_OFFSETS_S : MInst<"svldff1sb_gather_[{3}]offset_{d}", "dPSx", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldff1_gather_sxtw">; 165def SVLDFF1UB_GATHER_32B_OFFSETS_S : MInst<"svldff1ub_gather_[{3}]offset_{d}", "dPWx", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldff1_gather_sxtw">; 166def SVLDFF1SH_GATHER_32B_OFFSETS_S : MInst<"svldff1sh_gather_[{3}]offset_{d}", "dPTx", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldff1_gather_sxtw">; 167def SVLDFF1UH_GATHER_32B_OFFSETS_S : MInst<"svldff1uh_gather_[{3}]offset_{d}", "dPXx", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_sxtw">; 168 169// First-faulting load one vector (scalar base, unsigned vector offset in bytes) 170def SVLDFF1_GATHER_64B_OFFSETS_U : MInst<"svldff1_gather_[{3}]offset[_{d}]", "dPcu", "lUld", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldff1_gather">; 171def SVLDFF1SB_GATHER_64B_OFFSETS_U : MInst<"svldff1sb_gather_[{3}]offset_{d}", "dPSu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldff1_gather">; 172def SVLDFF1UB_GATHER_64B_OFFSETS_U : MInst<"svldff1ub_gather_[{3}]offset_{d}", "dPWu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldff1_gather">; 173def SVLDFF1SH_GATHER_64B_OFFSETS_U : MInst<"svldff1sh_gather_[{3}]offset_{d}", "dPTu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldff1_gather">; 174def SVLDFF1UH_GATHER_64B_OFFSETS_U : MInst<"svldff1uh_gather_[{3}]offset_{d}", "dPXu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather">; 175def SVLDFF1SW_GATHER_64B_OFFSETS_U : MInst<"svldff1sw_gather_[{3}]offset_{d}", "dPUu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ldff1_gather">; 176def SVLDFF1UW_GATHER_64B_OFFSETS_U : MInst<"svldff1uw_gather_[{3}]offset_{d}", "dPYu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1_gather">; 177 178def SVLDFF1_GATHER_32B_OFFSETS_U : MInst<"svldff1_gather_[{3}]offset[_{d}]", "dPcu", "iUif", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldff1_gather_uxtw">; 179def SVLDFF1SB_GATHER_32B_OFFSETS_U : MInst<"svldff1sb_gather_[{3}]offset_{d}", "dPSu", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldff1_gather_uxtw">; 180def SVLDFF1UB_GATHER_32B_OFFSETS_U : MInst<"svldff1ub_gather_[{3}]offset_{d}", "dPWu", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldff1_gather_uxtw">; 181def SVLDFF1SH_GATHER_32B_OFFSETS_U : MInst<"svldff1sh_gather_[{3}]offset_{d}", "dPTu", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldff1_gather_uxtw">; 182def SVLDFF1UH_GATHER_32B_OFFSETS_U : MInst<"svldff1uh_gather_[{3}]offset_{d}", "dPXu", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_uxtw">; 183 184// First-faulting load one vector (vector base, signed scalar offset in bytes) 185def SVLDFF1_GATHER_OFFSET_S : MInst<"svldff1_gather[_{2}base]_offset_{d}", "dPul", "ilUiUlfd", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldff1_gather_scalar_offset">; 186def SVLDFF1SB_GATHER_OFFSET_S : MInst<"svldff1sb_gather[_{2}base]_offset_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldff1_gather_scalar_offset">; 187def SVLDFF1UB_GATHER_OFFSET_S : MInst<"svldff1ub_gather[_{2}base]_offset_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldff1_gather_scalar_offset">; 188def SVLDFF1SH_GATHER_OFFSET_S : MInst<"svldff1sh_gather[_{2}base]_offset_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldff1_gather_scalar_offset">; 189def SVLDFF1UH_GATHER_OFFSET_S : MInst<"svldff1uh_gather[_{2}base]_offset_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_scalar_offset">; 190def SVLDFF1SW_GATHER_OFFSET_S : MInst<"svldff1sw_gather[_{2}base]_offset_{d}", "dPul", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ldff1_gather_scalar_offset">; 191def SVLDFF1UW_GATHER_OFFSET_S : MInst<"svldff1uw_gather[_{2}base]_offset_{d}", "dPul", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1_gather_scalar_offset">; 192 193// First-faulting load one vector (scalar base, signed vector index) 194def SVLDFF1_GATHER_64B_INDICES_S : MInst<"svldff1_gather_[{3}]index[_{d}]", "dPcx", "lUld", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldff1_gather_index">; 195def SVLDFF1SH_GATHER_64B_INDICES_S : MInst<"svldff1sh_gather_[{3}]index_{d}", "dPTx", "lUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldff1_gather_index">; 196def SVLDFF1UH_GATHER_64B_INDICES_S : MInst<"svldff1uh_gather_[{3}]index_{d}", "dPXx", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_index">; 197def SVLDFF1SW_GATHER_64B_INDICES_S : MInst<"svldff1sw_gather_[{3}]index_{d}", "dPUx", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ldff1_gather_index">; 198def SVLDFF1UW_GATHER_64B_INDICES_S : MInst<"svldff1uw_gather_[{3}]index_{d}", "dPYx", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1_gather_index">; 199 200def SVLDFF1_GATHER_32B_INDICES_S : MInst<"svldff1_gather_[{3}]index[_{d}]", "dPcx", "iUif", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldff1_gather_sxtw_index">; 201def SVLDFF1SH_GATHER_32B_INDICES_S : MInst<"svldff1sh_gather_[{3}]index_{d}", "dPTx", "iUi", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldff1_gather_sxtw_index">; 202def SVLDFF1UH_GATHER_32B_INDICES_S : MInst<"svldff1uh_gather_[{3}]index_{d}", "dPXx", "iUi", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_sxtw_index">; 203 204// First-faulting load one vector (scalar base, unsigned vector index) 205def SVLDFF1_GATHER_64B_INDICES_U : MInst<"svldff1_gather_[{3}]index[_{d}]", "dPcu", "lUld", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldff1_gather_index">; 206def SVLDFF1SH_GATHER_64B_INDICES_U : MInst<"svldff1sh_gather_[{3}]index_{d}", "dPTu", "lUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldff1_gather_index">; 207def SVLDFF1UH_GATHER_64B_INDICES_U : MInst<"svldff1uh_gather_[{3}]index_{d}", "dPXu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_index">; 208def SVLDFF1SW_GATHER_64B_INDICES_U : MInst<"svldff1sw_gather_[{3}]index_{d}", "dPUu", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ldff1_gather_index">; 209def SVLDFF1UW_GATHER_64B_INDICES_U : MInst<"svldff1uw_gather_[{3}]index_{d}", "dPYu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1_gather_index">; 210 211def SVLDFF1_GATHER_32B_INDICES_U : MInst<"svldff1_gather_[{3}]index[_{d}]", "dPcu", "iUif", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldff1_gather_uxtw_index">; 212def SVLDFF1SH_GATHER_32B_INDICES_U : MInst<"svldff1sh_gather_[{3}]index_{d}", "dPTu", "iUi", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldff1_gather_uxtw_index">; 213def SVLDFF1UH_GATHER_32B_INDICES_U : MInst<"svldff1uh_gather_[{3}]index_{d}", "dPXu", "iUi", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_uxtw_index">; 214 215// First-faulting load one vector (vector base, signed scalar index) 216def SVLDFF1_GATHER_INDEX_S : MInst<"svldff1_gather[_{2}base]_index_{d}", "dPul", "ilUiUlfd", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldff1_gather_scalar_offset">; 217def SVLDFF1SH_GATHER_INDEX_S : MInst<"svldff1sh_gather[_{2}base]_index_{d}", "dPul", "ilUiUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldff1_gather_scalar_offset">; 218def SVLDFF1UH_GATHER_INDEX_S : MInst<"svldff1uh_gather[_{2}base]_index_{d}", "dPul", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldff1_gather_scalar_offset">; 219def SVLDFF1SW_GATHER_INDEX_S : MInst<"svldff1sw_gather[_{2}base]_index_{d}", "dPul", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ldff1_gather_scalar_offset">; 220def SVLDFF1UW_GATHER_INDEX_S : MInst<"svldff1uw_gather[_{2}base]_index_{d}", "dPul", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldff1_gather_scalar_offset">; 221 222// Non-faulting load one vector (scalar base) 223def SVLDNF1 : MInst<"svldnf1[_{2}]", "dPc", "csilUcUsUiUlhfd", [IsLoad], MemEltTyDefault, "aarch64_sve_ldnf1">; 224def SVLDNF1SB : MInst<"svldnf1sb_{d}", "dPS", "silUsUiUl", [IsLoad], MemEltTyInt8, "aarch64_sve_ldnf1">; 225def SVLDNF1UB : MInst<"svldnf1ub_{d}", "dPW", "silUsUiUl", [IsLoad, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldnf1">; 226def SVLDNF1SH : MInst<"svldnf1sh_{d}", "dPT", "ilUiUl", [IsLoad], MemEltTyInt16, "aarch64_sve_ldnf1">; 227def SVLDNF1UH : MInst<"svldnf1uh_{d}", "dPX", "ilUiUl", [IsLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnf1">; 228def SVLDNF1SW : MInst<"svldnf1sw_{d}", "dPU", "lUl", [IsLoad], MemEltTyInt32, "aarch64_sve_ldnf1">; 229def SVLDNF1UW : MInst<"svldnf1uw_{d}", "dPY", "lUl", [IsLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnf1">; 230 231// Non-faulting load one vector (scalar base, VL displacement) 232def SVLDNF1_VNUM : MInst<"svldnf1_vnum[_{2}]", "dPcl", "csilUcUsUiUlhfd", [IsLoad], MemEltTyDefault, "aarch64_sve_ldnf1">; 233def SVLDNF1SB_VNUM : MInst<"svldnf1sb_vnum_{d}", "dPSl", "silUsUiUl", [IsLoad], MemEltTyInt8, "aarch64_sve_ldnf1">; 234def SVLDNF1UB_VNUM : MInst<"svldnf1ub_vnum_{d}", "dPWl", "silUsUiUl", [IsLoad, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldnf1">; 235def SVLDNF1SH_VNUM : MInst<"svldnf1sh_vnum_{d}", "dPTl", "ilUiUl", [IsLoad], MemEltTyInt16, "aarch64_sve_ldnf1">; 236def SVLDNF1UH_VNUM : MInst<"svldnf1uh_vnum_{d}", "dPXl", "ilUiUl", [IsLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnf1">; 237def SVLDNF1SW_VNUM : MInst<"svldnf1sw_vnum_{d}", "dPUl", "lUl", [IsLoad], MemEltTyInt32, "aarch64_sve_ldnf1">; 238def SVLDNF1UW_VNUM : MInst<"svldnf1uw_vnum_{d}", "dPYl", "lUl", [IsLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnf1">; 239 240let TargetGuard = "sve,bf16" in { 241 def SVLDNF1_BF : MInst<"svldnf1[_{2}]", "dPc", "b", [IsLoad], MemEltTyDefault, "aarch64_sve_ldnf1">; 242 def SVLDNF1_VNUM_BF : MInst<"svldnf1_vnum[_{2}]", "dPcl", "b", [IsLoad], MemEltTyDefault, "aarch64_sve_ldnf1">; 243} 244 245// Load one vector, unextended load, non-temporal (scalar base) 246def SVLDNT1 : MInst<"svldnt1[_{2}]", "dPc", "csilUcUsUiUlhfd", [IsLoad, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_ldnt1">; 247 248// Load one vector, unextended load, non-temporal (scalar base, VL displacement) 249def SVLDNT1_VNUM : MInst<"svldnt1_vnum[_{2}]", "dPcl", "csilUcUsUiUlhfd", [IsLoad, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_ldnt1">; 250 251let TargetGuard = "sve,bf16" in { 252 def SVLDNT1_BF : MInst<"svldnt1[_{2}]", "dPc", "b", [IsLoad, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_ldnt1">; 253 def SVLDNT1_VNUM_BF : MInst<"svldnt1_vnum[_{2}]", "dPcl", "b", [IsLoad, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_ldnt1">; 254} 255 256// Load one quadword and replicate (scalar base) 257def SVLD1RQ : SInst<"svld1rq[_{2}]", "dPc", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_ld1rq", [IsStreamingCompatible]>; 258 259let TargetGuard = "sve,bf16" in { 260 def SVLD1RQ_BF : SInst<"svld1rq[_{2}]", "dPc", "b", MergeNone, "aarch64_sve_ld1rq", [IsStreamingCompatible]>; 261} 262 263multiclass StructLoad<string name, string proto, string i> { 264 def : SInst<name, proto, "csilUcUsUiUlhfd", MergeNone, i, [IsStructLoad, IsStreamingCompatible]>; 265 let TargetGuard = "sve,bf16" in { 266 def: SInst<name, proto, "b", MergeNone, i, [IsStructLoad, IsStreamingCompatible]>; 267 } 268} 269 270// Load N-element structure into N vectors (scalar base) 271defm SVLD2 : StructLoad<"svld2[_{2}]", "2Pc", "aarch64_sve_ld2_sret">; 272defm SVLD3 : StructLoad<"svld3[_{2}]", "3Pc", "aarch64_sve_ld3_sret">; 273defm SVLD4 : StructLoad<"svld4[_{2}]", "4Pc", "aarch64_sve_ld4_sret">; 274 275// Load N-element structure into N vectors (scalar base, VL displacement) 276defm SVLD2_VNUM : StructLoad<"svld2_vnum[_{2}]", "2Pcl", "aarch64_sve_ld2_sret">; 277defm SVLD3_VNUM : StructLoad<"svld3_vnum[_{2}]", "3Pcl", "aarch64_sve_ld3_sret">; 278defm SVLD4_VNUM : StructLoad<"svld4_vnum[_{2}]", "4Pcl", "aarch64_sve_ld4_sret">; 279 280// Load one octoword and replicate (scalar base) 281let TargetGuard = "sve,f64mm" in { 282 def SVLD1RO : SInst<"svld1ro[_{2}]", "dPc", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_ld1ro">; 283} 284let TargetGuard = "sve,f64mm,bf16" in { 285 def SVLD1RO_BF16 : SInst<"svld1ro[_{2}]", "dPc", "b", MergeNone, "aarch64_sve_ld1ro">; 286} 287 288let TargetGuard = "sve,bf16" in { 289 def SVBFDOT : SInst<"svbfdot[_{0}]", "MMdd", "b", MergeNone, "aarch64_sve_bfdot", [IsOverloadNone, IsStreamingCompatible]>; 290 def SVBFMLALB : SInst<"svbfmlalb[_{0}]", "MMdd", "b", MergeNone, "aarch64_sve_bfmlalb", [IsOverloadNone, IsStreamingCompatible]>; 291 def SVBFMLALT : SInst<"svbfmlalt[_{0}]", "MMdd", "b", MergeNone, "aarch64_sve_bfmlalt", [IsOverloadNone, IsStreamingCompatible]>; 292 def SVBFMMLA : SInst<"svbfmmla[_{0}]", "MMdd", "b", MergeNone, "aarch64_sve_bfmmla", [IsOverloadNone, IsStreamingCompatible]>; 293 def SVBFDOT_N : SInst<"svbfdot[_n_{0}]", "MMda", "b", MergeNone, "aarch64_sve_bfdot", [IsOverloadNone, IsStreamingCompatible]>; 294 def SVBFMLAL_N : SInst<"svbfmlalb[_n_{0}]", "MMda", "b", MergeNone, "aarch64_sve_bfmlalb", [IsOverloadNone, IsStreamingCompatible]>; 295 def SVBFMLALT_N : SInst<"svbfmlalt[_n_{0}]", "MMda", "b", MergeNone, "aarch64_sve_bfmlalt", [IsOverloadNone, IsStreamingCompatible]>; 296 def SVBFDOT_LANE : SInst<"svbfdot_lane[_{0}]", "MMddi", "b", MergeNone, "aarch64_sve_bfdot_lane_v2", [IsOverloadNone, IsStreamingCompatible], [ImmCheck<3, ImmCheck0_3>]>; 297 def SVBFMLALB_LANE : SInst<"svbfmlalb_lane[_{0}]", "MMddi", "b", MergeNone, "aarch64_sve_bfmlalb_lane_v2", [IsOverloadNone, IsStreamingCompatible], [ImmCheck<3, ImmCheck0_7>]>; 298 def SVBFMLALT_LANE : SInst<"svbfmlalt_lane[_{0}]", "MMddi", "b", MergeNone, "aarch64_sve_bfmlalt_lane_v2", [IsOverloadNone, IsStreamingCompatible], [ImmCheck<3, ImmCheck0_7>]>; 299} 300 301let TargetGuard = "sve2p1" in { 302 // Contiguous zero-extending load to quadword (single vector). 303 def SVLD1UWQ : MInst<"svld1uwq[_{d}]", "dPc", "iUif", [IsLoad], MemEltTyInt32, "aarch64_sve_ld1uwq">; 304 def SVLD1UWQ_VNUM : MInst<"svld1uwq_vnum[_{d}]", "dPcl", "iUif", [IsLoad], MemEltTyInt32, "aarch64_sve_ld1uwq">; 305 306 def SVLD1UDQ : MInst<"svld1udq[_{d}]", "dPc", "lUld", [IsLoad], MemEltTyInt64, "aarch64_sve_ld1udq">; 307 def SVLD1UDQ_VNUM : MInst<"svld1udq_vnum[_{d}]", "dPcl", "lUld", [IsLoad], MemEltTyInt64, "aarch64_sve_ld1udq">; 308 309 // Load one vector (vector base + scalar offset) 310 def SVLD1Q_GATHER_U64BASE_OFFSET : MInst<"svld1q_gather[_{2}base]_offset_{d}", "dPgl", "cUcsUsiUilUlfhdb", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ld1q_gather_scalar_offset">; 311 def SVLD1Q_GATHER_U64BASE : MInst<"svld1q_gather[_{2}base]_{d}", "dPg", "cUcsUsiUilUlfhdb", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ld1q_gather_scalar_offset">; 312 313 // Load one vector (scalar base + vector offset) 314 def SVLD1Q_GATHER_U64OFFSET : MInst<"svld1q_gather_[{3}]offset[_{d}]", "dPcg", "cUcsUsiUilUlfhdb", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ld1q_gather_vector_offset">; 315 316 // Load N-element structure into N vectors (scalar base) 317 defm SVLD2Q : StructLoad<"svld2q[_{2}]", "2Pc", "aarch64_sve_ld2q_sret">; 318 defm SVLD3Q : StructLoad<"svld3q[_{2}]", "3Pc", "aarch64_sve_ld3q_sret">; 319 defm SVLD4Q : StructLoad<"svld4q[_{2}]", "4Pc", "aarch64_sve_ld4q_sret">; 320 321 // Load N-element structure into N vectors (scalar base, VL displacement) 322 defm SVLD2Q_VNUM : StructLoad<"svld2q_vnum[_{2}]", "2Pcl", "aarch64_sve_ld2q_sret">; 323 defm SVLD3Q_VNUM : StructLoad<"svld3q_vnum[_{2}]", "3Pcl", "aarch64_sve_ld3q_sret">; 324 defm SVLD4Q_VNUM : StructLoad<"svld4q_vnum[_{2}]", "4Pcl", "aarch64_sve_ld4q_sret">; 325 326 // Load quadwords (scalar base + vector index) 327 def SVLD1Q_GATHER_INDICES_U : MInst<"svld1q_gather_[{3}]index[_{d}]", "dPcg", "sUsiUilUlbhfd", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ld1q_gather_index">; 328 329 // Load quadwords (vector base + scalar index) 330 def SVLD1Q_GATHER_INDEX_S : MInst<"svld1q_gather[_{2}base]_index_{d}", "dPgl", "sUsiUilUlbhfd", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ld1q_gather_scalar_offset">; 331} 332 333//////////////////////////////////////////////////////////////////////////////// 334// Stores 335 336// Store one vector (scalar base) 337def SVST1 : MInst<"svst1[_{d}]", "vPpd", "csilUcUsUiUlhfd", [IsStore, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_st1">; 338def SVST1B_S : MInst<"svst1b[_{d}]", "vPAd", "sil", [IsStore, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_st1">; 339def SVST1B_U : MInst<"svst1b[_{d}]", "vPEd", "UsUiUl", [IsStore, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_st1">; 340def SVST1H_S : MInst<"svst1h[_{d}]", "vPBd", "il", [IsStore, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_st1">; 341def SVST1H_U : MInst<"svst1h[_{d}]", "vPFd", "UiUl", [IsStore, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_st1">; 342def SVST1W_S : MInst<"svst1w[_{d}]", "vPCd", "l", [IsStore, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_st1">; 343def SVST1W_U : MInst<"svst1w[_{d}]", "vPGd", "Ul", [IsStore, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_st1">; 344 345// Store one vector (scalar base, VL displacement) 346def SVST1_VNUM : MInst<"svst1_vnum[_{d}]", "vPpld", "csilUcUsUiUlhfd", [IsStore, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_st1">; 347def SVST1B_VNUM_S : MInst<"svst1b_vnum[_{d}]", "vPAld", "sil", [IsStore, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_st1">; 348def SVST1B_VNUM_U : MInst<"svst1b_vnum[_{d}]", "vPEld", "UsUiUl", [IsStore, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_st1">; 349def SVST1H_VNUM_S : MInst<"svst1h_vnum[_{d}]", "vPBld", "il", [IsStore, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_st1">; 350def SVST1H_VNUM_U : MInst<"svst1h_vnum[_{d}]", "vPFld", "UiUl", [IsStore, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_st1">; 351def SVST1W_VNUM_S : MInst<"svst1w_vnum[_{d}]", "vPCld", "l", [IsStore, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_st1">; 352def SVST1W_VNUM_U : MInst<"svst1w_vnum[_{d}]", "vPGld", "Ul", [IsStore, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_st1">; 353 354let TargetGuard = "sve,bf16" in { 355 def SVST1_BF : MInst<"svst1[_{d}]", "vPpd", "b", [IsStore, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_st1">; 356 def SVST1_VNUM_BF : MInst<"svst1_vnum[_{d}]", "vPpld", "b", [IsStore, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_st1">; 357} 358 359// Store one vector (vector base) 360def SVST1_SCATTER_BASES_U : MInst<"svst1_scatter[_{2}base_{d}]", "vPud", "ilUiUlfd", [IsScatterStore], MemEltTyDefault, "aarch64_sve_st1_scatter_scalar_offset">; 361def SVST1B_SCATTER_BASES_U : MInst<"svst1b_scatter[_{2}base_{d}]", "vPud", "ilUiUl", [IsScatterStore], MemEltTyInt8, "aarch64_sve_st1_scatter_scalar_offset">; 362def SVST1H_SCATTER_BASES_U : MInst<"svst1h_scatter[_{2}base_{d}]", "vPud", "ilUiUl", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_scalar_offset">; 363def SVST1W_SCATTER_BASES_U : MInst<"svst1w_scatter[_{2}base_{d}]", "vPud", "lUl", [IsScatterStore], MemEltTyInt32, "aarch64_sve_st1_scatter_scalar_offset">; 364 365// Store one vector (scalar base, signed vector offset in bytes) 366def SVST1_SCATTER_64B_OFFSETS_S : MInst<"svst1_scatter_[{3}]offset[_{d}]", "vPpxd", "lUld", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_st1_scatter">; 367def SVST1B_SCATTER_64B_OFFSETS_SS : MInst<"svst1b_scatter_[{3}]offset[_{d}]", "vPAxd", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter">; 368def SVST1B_SCATTER_64B_OFFSETS_SU : MInst<"svst1b_scatter_[{3}]offset[_{d}]", "vPExd", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter">; 369def SVST1H_SCATTER_64B_OFFSETS_SS : MInst<"svst1h_scatter_[{3}]offset[_{d}]", "vPBxd", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter">; 370def SVST1H_SCATTER_64B_OFFSETS_SU : MInst<"svst1h_scatter_[{3}]offset[_{d}]", "vPFxd", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter">; 371def SVST1W_SCATTER_64B_OFFSETS_SS : MInst<"svst1w_scatter_[{3}]offset[_{d}]", "vPCxd", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_st1_scatter">; 372def SVST1W_SCATTER_64B_OFFSETS_SU : MInst<"svst1w_scatter_[{3}]offset[_{d}]", "vPGxd", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_st1_scatter">; 373 374def SVST1_SCATTER_32B_OFFSETS_S : MInst<"svst1_scatter_[{3}]offset[_{d}]", "vPpxd", "iUif", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_st1_scatter_sxtw">; 375def SVST1B_SCATTER_32B_OFFSETS_SS : MInst<"svst1b_scatter_[{3}]offset[_{d}]", "vPAxd", "i", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter_sxtw">; 376def SVST1B_SCATTER_32B_OFFSETS_SU : MInst<"svst1b_scatter_[{3}]offset[_{d}]", "vPExd", "Ui", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter_sxtw">; 377def SVST1H_SCATTER_32B_OFFSETS_SS : MInst<"svst1h_scatter_[{3}]offset[_{d}]", "vPBxd", "i", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter_sxtw">; 378def SVST1H_SCATTER_32B_OFFSETS_SU : MInst<"svst1h_scatter_[{3}]offset[_{d}]", "vPFxd", "Ui", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter_sxtw">; 379 380// Store one vector (scalar base, unsigned vector offset in bytes) 381def SVST1_SCATTER_64B_OFFSETS_U : MInst<"svst1_scatter_[{3}]offset[_{d}]", "vPpud", "lUld", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_st1_scatter">; 382def SVST1B_SCATTER_64B_OFFSETS_US : MInst<"svst1b_scatter_[{3}]offset[_{d}]", "vPAud", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter">; 383def SVST1B_SCATTER_64B_OFFSETS_UU : MInst<"svst1b_scatter_[{3}]offset[_{d}]", "vPEud", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter">; 384def SVST1H_SCATTER_64B_OFFSETS_US : MInst<"svst1h_scatter_[{3}]offset[_{d}]", "vPBud", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter">; 385def SVST1H_SCATTER_64B_OFFSETS_UU : MInst<"svst1h_scatter_[{3}]offset[_{d}]", "vPFud", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter">; 386def SVST1W_SCATTER_64B_OFFSETS_US : MInst<"svst1w_scatter_[{3}]offset[_{d}]", "vPCud", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_st1_scatter">; 387def SVST1W_SCATTER_64B_OFFSETS_UU : MInst<"svst1w_scatter_[{3}]offset[_{d}]", "vPGud", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_st1_scatter">; 388 389def SVST1_SCATTER_32B_OFFSETS_U : MInst<"svst1_scatter_[{3}]offset[_{d}]", "vPpud", "iUif", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_st1_scatter_uxtw">; 390def SVST1B_SCATTER_32B_OFFSETS_US : MInst<"svst1b_scatter_[{3}]offset[_{d}]", "vPAud", "i", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter_uxtw">; 391def SVST1B_SCATTER_32B_OFFSETS_UU : MInst<"svst1b_scatter_[{3}]offset[_{d}]", "vPEud", "Ui", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter_uxtw">; 392def SVST1H_SCATTER_32B_OFFSETS_US : MInst<"svst1h_scatter_[{3}]offset[_{d}]", "vPBud", "i", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter_uxtw">; 393def SVST1H_SCATTER_32B_OFFSETS_UU : MInst<"svst1h_scatter_[{3}]offset[_{d}]", "vPFud", "Ui", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter_uxtw">; 394 395// Store one vector (vector base, signed scalar offset in bytes) 396def SVST1_SCATTER_OFFSET_S : MInst<"svst1_scatter[_{2}base]_offset[_{d}]", "vPuld", "ilUiUlfd", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_st1_scatter_scalar_offset">; 397def SVST1B_SCATTER_OFFSET_S : MInst<"svst1b_scatter[_{2}base]_offset[_{d}]", "vPuld", "ilUiUl", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_st1_scatter_scalar_offset">; 398def SVST1H_SCATTER_OFFSET_S : MInst<"svst1h_scatter[_{2}base]_offset[_{d}]", "vPuld", "ilUiUl", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_st1_scatter_scalar_offset">; 399def SVST1W_SCATTER_OFFSET_S : MInst<"svst1w_scatter[_{2}base]_offset[_{d}]", "vPuld", "lUl", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_st1_scatter_scalar_offset">; 400 401// Store one vector (scalar base, signed vector index) 402def SVST1_SCATTER_64B_INDICES_S : MInst<"svst1_scatter_[{3}]index[_{d}]", "vPpxd", "lUld", [IsScatterStore], MemEltTyDefault, "aarch64_sve_st1_scatter_index">; 403def SVST1H_SCATTER_64B_INDICES_SS : MInst<"svst1h_scatter_[{3}]index[_{d}]", "vPBxd", "l", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_index">; 404def SVST1H_SCATTER_64B_INDICES_SU : MInst<"svst1h_scatter_[{3}]index[_{d}]", "vPFxd", "Ul", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_index">; 405def SVST1W_SCATTER_64B_INDICES_SS : MInst<"svst1w_scatter_[{3}]index[_{d}]", "vPCxd", "l", [IsScatterStore], MemEltTyInt32, "aarch64_sve_st1_scatter_index">; 406def SVST1W_SCATTER_64B_INDICES_SU : MInst<"svst1w_scatter_[{3}]index[_{d}]", "vPGxd", "Ul", [IsScatterStore], MemEltTyInt32, "aarch64_sve_st1_scatter_index">; 407 408def SVST1_SCATTER_32B_INDICES_S : MInst<"svst1_scatter_[{3}]index[_{d}]", "vPpxd", "iUif", [IsScatterStore], MemEltTyDefault, "aarch64_sve_st1_scatter_sxtw_index">; 409def SVST1H_SCATTER_32B_INDICES_SS : MInst<"svst1h_scatter_[{3}]index[_{d}]", "vPBxd", "i", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_sxtw_index">; 410def SVST1H_SCATTER_32B_INDICES_SU : MInst<"svst1h_scatter_[{3}]index[_{d}]", "vPFxd", "Ui", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_sxtw_index">; 411 412// Store one vector (scalar base, unsigned vector index) 413def SVST1_SCATTER_64B_INDICES_U : MInst<"svst1_scatter_[{3}]index[_{d}]", "vPpud", "lUld", [IsScatterStore], MemEltTyDefault, "aarch64_sve_st1_scatter_index">; 414def SVST1H_SCATTER_64B_INDICES_US : MInst<"svst1h_scatter_[{3}]index[_{d}]", "vPBud", "l", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_index">; 415def SVST1H_SCATTER_64B_INDICES_UU : MInst<"svst1h_scatter_[{3}]index[_{d}]", "vPFud", "Ul", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_index">; 416def SVST1W_SCATTER_64B_INDICES_US : MInst<"svst1w_scatter_[{3}]index[_{d}]", "vPCud", "l", [IsScatterStore], MemEltTyInt32, "aarch64_sve_st1_scatter_index">; 417def SVST1W_SCATTER_64B_INDICES_UU : MInst<"svst1w_scatter_[{3}]index[_{d}]", "vPGud", "Ul", [IsScatterStore], MemEltTyInt32, "aarch64_sve_st1_scatter_index">; 418 419def SVST1_SCATTER_32B_INDICES_U : MInst<"svst1_scatter_[{3}]index[_{d}]", "vPpud", "iUif", [IsScatterStore], MemEltTyDefault, "aarch64_sve_st1_scatter_uxtw_index">; 420def SVST1H_SCATTER_32B_INDICES_US : MInst<"svst1h_scatter_[{3}]index[_{d}]", "vPBud", "i", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_uxtw_index">; 421def SVST1H_SCATTER_32B_INDICES_UU : MInst<"svst1h_scatter_[{3}]index[_{d}]", "vPFud", "Ui", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_uxtw_index">; 422 423// Store one vector (vector base, signed scalar index) 424def SVST1_SCATTER_INDEX_S : MInst<"svst1_scatter[_{2}base]_index[_{d}]", "vPuld", "ilUiUlfd", [IsScatterStore], MemEltTyDefault, "aarch64_sve_st1_scatter_scalar_offset">; 425def SVST1H_SCATTER_INDEX_S : MInst<"svst1h_scatter[_{2}base]_index[_{d}]", "vPuld", "ilUiUl", [IsScatterStore], MemEltTyInt16, "aarch64_sve_st1_scatter_scalar_offset">; 426def SVST1W_SCATTER_INDEX_S : MInst<"svst1w_scatter[_{2}base]_index[_{d}]", "vPuld", "lUl", [IsScatterStore], MemEltTyInt32, "aarch64_sve_st1_scatter_scalar_offset">; 427 428multiclass StructStore<string name, string proto, string i> { 429 def : SInst<name, proto, "csilUcUsUiUlhfd", MergeNone, i, [IsStructStore, IsStreamingCompatible]>; 430 let TargetGuard = "sve,bf16" in { 431 def: SInst<name, proto, "b", MergeNone, i, [IsStructStore, IsStreamingCompatible]>; 432 } 433} 434// Store N vectors into N-element structure (scalar base) 435defm SVST2 : StructStore<"svst2[_{d}]", "vPp2", "aarch64_sve_st2">; 436defm SVST3 : StructStore<"svst3[_{d}]", "vPp3", "aarch64_sve_st3">; 437defm SVST4 : StructStore<"svst4[_{d}]", "vPp4", "aarch64_sve_st4">; 438 439// Store N vectors into N-element structure (scalar base, VL displacement) 440defm SVST2_VNUM : StructStore<"svst2_vnum[_{d}]", "vPpl2", "aarch64_sve_st2">; 441defm SVST3_VNUM : StructStore<"svst3_vnum[_{d}]", "vPpl3", "aarch64_sve_st3">; 442defm SVST4_VNUM : StructStore<"svst4_vnum[_{d}]", "vPpl4", "aarch64_sve_st4">; 443 444// Store one vector, with no truncation, non-temporal (scalar base) 445def SVSTNT1 : MInst<"svstnt1[_{d}]", "vPpd", "csilUcUsUiUlhfd", [IsStore, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_stnt1">; 446 447// Store one vector, with no truncation, non-temporal (scalar base, VL displacement) 448def SVSTNT1_VNUM : MInst<"svstnt1_vnum[_{d}]", "vPpld", "csilUcUsUiUlhfd", [IsStore, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_stnt1">; 449 450let TargetGuard = "sve,bf16" in { 451 def SVSTNT1_BF : MInst<"svstnt1[_{d}]", "vPpd", "b", [IsStore, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_stnt1">; 452 def SVSTNT1_VNUM_BF : MInst<"svstnt1_vnum[_{d}]", "vPpld", "b", [IsStore, IsStreamingCompatible], MemEltTyDefault, "aarch64_sve_stnt1">; 453} 454 455let TargetGuard = "sve2p1" in { 456 // Contiguous truncating store from quadword (single vector). 457 def SVST1UWQ : MInst<"svst1wq[_{d}]", "vPcd", "iUif", [IsStore], MemEltTyInt32, "aarch64_sve_st1wq">; 458 def SVST1UWQ_VNUM : MInst<"svst1wq_vnum[_{d}]", "vPcld", "iUif", [IsStore], MemEltTyInt32, "aarch64_sve_st1wq">; 459 460 def SVST1UDQ : MInst<"svst1dq[_{d}]", "vPcd", "lUld", [IsStore], MemEltTyInt64, "aarch64_sve_st1dq">; 461 def SVST1UDQ_VNUM : MInst<"svst1dq_vnum[_{d}]", "vPcld", "lUld", [IsStore], MemEltTyInt64, "aarch64_sve_st1dq">; 462 463 // Store one vector (vector base + scalar offset) 464 def SVST1Q_SCATTER_U64BASE_OFFSET : MInst<"svst1q_scatter[_{2}base]_offset[_{d}]", "vPgld", "cUcsUsiUilUlfhdb", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_st1q_scatter_scalar_offset">; 465 def SVST1Q_SCATTER_U64BASE : MInst<"svst1q_scatter[_{2}base][_{d}]", "vPgd", "cUcsUsiUilUlfhdb", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_st1q_scatter_scalar_offset">; 466 467 // Store one vector (scalar base + vector offset) 468 def SVST1Q_SCATTER_U64OFFSET : MInst<"svst1q_scatter_[{3}]offset[_{d}]", "vPpgd", "cUcsUsiUilUlfhdb", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_st1q_scatter_vector_offset">; 469 470 // Store N vectors into N-element structure (scalar base) 471 defm SVST2Q : StructStore<"svst2q[_{d}]", "vPc2", "aarch64_sve_st2q">; 472 defm SVST3Q : StructStore<"svst3q[_{d}]", "vPc3", "aarch64_sve_st3q">; 473 defm SVST4Q : StructStore<"svst4q[_{d}]", "vPc4", "aarch64_sve_st4q">; 474 475 // Store N vectors into N-element structure (scalar base, VL displacement) 476 defm SVST2Q_VNUM : StructStore<"svst2q_vnum[_{d}]", "vPcl2", "aarch64_sve_st2q">; 477 defm SVST3Q_VNUM : StructStore<"svst3q_vnum[_{d}]", "vPcl3", "aarch64_sve_st3q">; 478 defm SVST4Q_VNUM : StructStore<"svst4q_vnum[_{d}]", "vPcl4", "aarch64_sve_st4q">; 479 480 // Scatter store quadwords (scalar base + vector index) 481 def SVST1Q_SCATTER_INDICES_U : MInst<"svst1q_scatter_[{3}]index[_{d}]", "vPpgd", "sUsiUilUlbhfd", [IsScatterStore], MemEltTyDefault, "aarch64_sve_st1q_scatter_index">; 482 483 // Scatter store quadwords (vector base + scalar index) 484 def SVST1Q_SCATTER_INDEX_S : MInst<"svst1q_scatter[_{2}base]_index[_{d}]", "vPgld", "sUsiUilUlbhfd", [IsScatterStore], MemEltTyDefault, "aarch64_sve_st1q_scatter_scalar_offset">; 485} 486 487//////////////////////////////////////////////////////////////////////////////// 488// Prefetches 489 490// Prefetch (Scalar base) 491def SVPRFB : MInst<"svprfb", "vPQJ", "c", [IsPrefetch, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_prf">; 492def SVPRFH : MInst<"svprfh", "vPQJ", "s", [IsPrefetch, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_prf">; 493def SVPRFW : MInst<"svprfw", "vPQJ", "i", [IsPrefetch, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_prf">; 494def SVPRFD : MInst<"svprfd", "vPQJ", "l", [IsPrefetch, IsStreamingCompatible], MemEltTyInt64, "aarch64_sve_prf">; 495 496// Prefetch (Scalar base, VL displacement) 497def SVPRFB_VNUM : MInst<"svprfb_vnum", "vPQlJ", "c", [IsPrefetch, IsStreamingCompatible], MemEltTyInt8, "aarch64_sve_prf">; 498def SVPRFH_VNUM : MInst<"svprfh_vnum", "vPQlJ", "s", [IsPrefetch, IsStreamingCompatible], MemEltTyInt16, "aarch64_sve_prf">; 499def SVPRFW_VNUM : MInst<"svprfw_vnum", "vPQlJ", "i", [IsPrefetch, IsStreamingCompatible], MemEltTyInt32, "aarch64_sve_prf">; 500def SVPRFD_VNUM : MInst<"svprfd_vnum", "vPQlJ", "l", [IsPrefetch, IsStreamingCompatible], MemEltTyInt64, "aarch64_sve_prf">; 501 502// Prefetch (Vector bases) 503def SVPRFB_GATHER_BASES : MInst<"svprfb_gather[_{2}base]", "vPdJ", "UiUl", [IsGatherPrefetch], MemEltTyInt8, "aarch64_sve_prfb_gather_scalar_offset">; 504def SVPRFH_GATHER_BASES : MInst<"svprfh_gather[_{2}base]", "vPdJ", "UiUl", [IsGatherPrefetch], MemEltTyInt16, "aarch64_sve_prfh_gather_scalar_offset">; 505def SVPRFW_GATHER_BASES : MInst<"svprfw_gather[_{2}base]", "vPdJ", "UiUl", [IsGatherPrefetch], MemEltTyInt32, "aarch64_sve_prfw_gather_scalar_offset">; 506def SVPRFD_GATHER_BASES : MInst<"svprfd_gather[_{2}base]", "vPdJ", "UiUl", [IsGatherPrefetch], MemEltTyInt64, "aarch64_sve_prfd_gather_scalar_offset">; 507 508// Prefetch (Scalar base, Vector offsets) 509def SVPRFB_GATHER_32B_OFFSETS_S : MInst<"svprfb_gather_[{3}]offset", "vPQdJ", "i", [IsGatherPrefetch], MemEltTyInt8, "aarch64_sve_prfb_gather_sxtw_index">; 510def SVPRFH_GATHER_32B_OFFSETS_S : MInst<"svprfh_gather_[{3}]index", "vPQdJ", "i", [IsGatherPrefetch], MemEltTyInt16, "aarch64_sve_prfh_gather_sxtw_index">; 511def SVPRFW_GATHER_32B_OFFSETS_S : MInst<"svprfw_gather_[{3}]index", "vPQdJ", "i", [IsGatherPrefetch], MemEltTyInt32, "aarch64_sve_prfw_gather_sxtw_index">; 512def SVPRFD_GATHER_32B_OFFSETS_S : MInst<"svprfd_gather_[{3}]index", "vPQdJ", "i", [IsGatherPrefetch], MemEltTyInt64, "aarch64_sve_prfd_gather_sxtw_index">; 513 514def SVPRFB_GATHER_64B_OFFSETS_S : MInst<"svprfb_gather_[{3}]offset", "vPQdJ", "l", [IsGatherPrefetch], MemEltTyInt8, "aarch64_sve_prfb_gather_index">; 515def SVPRFH_GATHER_64B_OFFSETS_S : MInst<"svprfh_gather_[{3}]index", "vPQdJ", "l", [IsGatherPrefetch], MemEltTyInt16, "aarch64_sve_prfh_gather_index">; 516def SVPRFW_GATHER_64B_OFFSETS_S : MInst<"svprfw_gather_[{3}]index", "vPQdJ", "l", [IsGatherPrefetch], MemEltTyInt32, "aarch64_sve_prfw_gather_index">; 517def SVPRFD_GATHER_64B_OFFSETS_S : MInst<"svprfd_gather_[{3}]index", "vPQdJ", "l", [IsGatherPrefetch], MemEltTyInt64, "aarch64_sve_prfd_gather_index">; 518 519def SVPRFB_GATHER_32B_OFFSETS_U : MInst<"svprfb_gather_[{3}]offset", "vPQdJ", "Ui", [IsGatherPrefetch], MemEltTyInt8, "aarch64_sve_prfb_gather_uxtw_index">; 520def SVPRFH_GATHER_32B_OFFSETS_U : MInst<"svprfh_gather_[{3}]index", "vPQdJ", "Ui", [IsGatherPrefetch], MemEltTyInt16, "aarch64_sve_prfh_gather_uxtw_index">; 521def SVPRFW_GATHER_32B_OFFSETS_U : MInst<"svprfw_gather_[{3}]index", "vPQdJ", "Ui", [IsGatherPrefetch], MemEltTyInt32, "aarch64_sve_prfw_gather_uxtw_index">; 522def SVPRFD_GATHER_32B_OFFSETS_U : MInst<"svprfd_gather_[{3}]index", "vPQdJ", "Ui", [IsGatherPrefetch], MemEltTyInt64, "aarch64_sve_prfd_gather_uxtw_index">; 523 524def SVPRFB_GATHER_64B_OFFSETS_U : MInst<"svprfb_gather_[{3}]offset", "vPQdJ", "Ul", [IsGatherPrefetch], MemEltTyInt8, "aarch64_sve_prfb_gather_index">; 525def SVPRFH_GATHER_64B_OFFSETS_U : MInst<"svprfh_gather_[{3}]index", "vPQdJ", "Ul", [IsGatherPrefetch], MemEltTyInt16, "aarch64_sve_prfh_gather_index">; 526def SVPRFW_GATHER_64B_OFFSETS_U : MInst<"svprfw_gather_[{3}]index", "vPQdJ", "Ul", [IsGatherPrefetch], MemEltTyInt32, "aarch64_sve_prfw_gather_index">; 527def SVPRFD_GATHER_64B_OFFSETS_U : MInst<"svprfd_gather_[{3}]index", "vPQdJ", "Ul", [IsGatherPrefetch], MemEltTyInt64, "aarch64_sve_prfd_gather_index">; 528 529// Prefetch (Vector bases, scalar offset) 530def SVPRFB_GATHER_BASES_OFFSET : MInst<"svprfb_gather[_{2}base]_offset", "vPdlJ", "UiUl", [IsGatherPrefetch], MemEltTyInt8, "aarch64_sve_prfb_gather_scalar_offset">; 531def SVPRFH_GATHER_BASES_OFFSET : MInst<"svprfh_gather[_{2}base]_index", "vPdlJ", "UiUl", [IsGatherPrefetch], MemEltTyInt16, "aarch64_sve_prfh_gather_scalar_offset">; 532def SVPRFW_GATHER_BASES_OFFSET : MInst<"svprfw_gather[_{2}base]_index", "vPdlJ", "UiUl", [IsGatherPrefetch], MemEltTyInt32, "aarch64_sve_prfw_gather_scalar_offset">; 533def SVPRFD_GATHER_BASES_OFFSET : MInst<"svprfd_gather[_{2}base]_index", "vPdlJ", "UiUl", [IsGatherPrefetch], MemEltTyInt64, "aarch64_sve_prfd_gather_scalar_offset">; 534 535//////////////////////////////////////////////////////////////////////////////// 536// Address calculations 537 538def SVADRB : SInst<"svadrb[_{0}base]_[{2}]offset", "uud", "ilUiUl", MergeNone, "aarch64_sve_adrb">; 539def SVADRH : SInst<"svadrh[_{0}base]_[{2}]index", "uud", "ilUiUl", MergeNone, "aarch64_sve_adrh">; 540def SVADRW : SInst<"svadrw[_{0}base]_[{2}]index", "uud", "ilUiUl", MergeNone, "aarch64_sve_adrw">; 541def SVADRD : SInst<"svadrd[_{0}base]_[{2}]index", "uud", "ilUiUl", MergeNone, "aarch64_sve_adrd">; 542 543//////////////////////////////////////////////////////////////////////////////// 544// Scalar to vector 545 546def SVDUPQ_8 : SInst<"svdupq[_n]_{d}", "dssssssssssssssss", "cUc", MergeNone, "", [IsStreamingCompatible]>; 547def SVDUPQ_16 : SInst<"svdupq[_n]_{d}", "dssssssss", "sUsh", MergeNone, "", [IsStreamingCompatible]>; 548let TargetGuard = "sve,bf16" in { 549 def SVDUPQ_BF16 : SInst<"svdupq[_n]_{d}", "dssssssss", "b", MergeNone, "", [IsStreamingCompatible]>; 550} 551def SVDUPQ_32 : SInst<"svdupq[_n]_{d}", "dssss", "iUif", MergeNone, "", [IsStreamingCompatible]>; 552def SVDUPQ_64 : SInst<"svdupq[_n]_{d}", "dss", "lUld", MergeNone, "", [IsStreamingCompatible]>; 553 554multiclass svdup_base<string n, string p, MergeType mt, string i> { 555 def NAME : SInst<n, p, "csilUcUsUiUlhfd", mt, i, [IsStreamingCompatible]>; 556 let TargetGuard = "sve,bf16" in { 557 def _BF16: SInst<n, p, "b", mt, i, [IsStreamingCompatible]>; 558 } 559} 560 561defm SVDUP : svdup_base<"svdup[_n]_{d}", "ds", MergeNone, "aarch64_sve_dup_x">; 562defm SVDUP_M : svdup_base<"svdup[_n]_{d}", "ddPs", MergeOp1, "aarch64_sve_dup">; 563defm SVDUP_X : svdup_base<"svdup[_n]_{d}", "dPs", MergeAnyExp, "aarch64_sve_dup">; 564defm SVDUP_Z : svdup_base<"svdup[_n]_{d}", "dPs", MergeZeroExp, "aarch64_sve_dup">; 565 566def SVINDEX : SInst<"svindex_{d}", "dss", "csilUcUsUiUl", MergeNone, "aarch64_sve_index", [IsStreamingCompatible]>; 567 568// Integer arithmetic 569 570multiclass SInstZPZ<string name, string types, string intrinsic> { 571 def _M : SInst<name # "[_{d}]", "ddPd", types, MergeOp1, intrinsic, [IsStreamingCompatible]>; 572 def _X : SInst<name # "[_{d}]", "dPd", types, MergeAnyExp, intrinsic, [IsStreamingCompatible]>; 573 def _Z : SInst<name # "[_{d}]", "dPd", types, MergeZeroExp, intrinsic, [IsStreamingCompatible]>; 574} 575 576defm SVABS : SInstZPZ<"svabs", "csil", "aarch64_sve_abs">; 577defm SVNEG : SInstZPZ<"svneg", "csil", "aarch64_sve_neg">; 578 579//------------------------------------------------------------------------------ 580 581multiclass SInstZPZZ<string name, string types, string m_intrinsic, string x_intrinsic, list<FlagType> flags=[]> { 582 def _M : SInst<name # "[_{d}]", "dPdd", types, MergeOp1, m_intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 583 def _X : SInst<name # "[_{d}]", "dPdd", types, MergeAny, x_intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 584 def _Z : SInst<name # "[_{d}]", "dPdd", types, MergeZero, m_intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 585 586 def _N_M : SInst<name # "[_n_{d}]", "dPda", types, MergeOp1, m_intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 587 def _N_X : SInst<name # "[_n_{d}]", "dPda", types, MergeAny, x_intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 588 def _N_Z : SInst<name # "[_n_{d}]", "dPda", types, MergeZero, m_intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 589} 590 591defm SVABD_S : SInstZPZZ<"svabd", "csil", "aarch64_sve_sabd", "aarch64_sve_sabd_u">; 592defm SVABD_U : SInstZPZZ<"svabd", "UcUsUiUl", "aarch64_sve_uabd", "aarch64_sve_uabd_u">; 593defm SVADD : SInstZPZZ<"svadd", "csilUcUsUiUl", "aarch64_sve_add", "aarch64_sve_add_u">; 594defm SVDIV_S : SInstZPZZ<"svdiv", "il", "aarch64_sve_sdiv", "aarch64_sve_sdiv_u">; 595defm SVDIV_U : SInstZPZZ<"svdiv", "UiUl", "aarch64_sve_udiv", "aarch64_sve_udiv_u">; 596defm SVDIVR_S : SInstZPZZ<"svdivr", "il", "aarch64_sve_sdivr", "aarch64_sve_sdiv_u", [ReverseMergeAnyBinOp]>; 597defm SVDIVR_U : SInstZPZZ<"svdivr", "UiUl", "aarch64_sve_udivr", "aarch64_sve_udiv_u", [ReverseMergeAnyBinOp]>; 598defm SVMAX_S : SInstZPZZ<"svmax", "csil", "aarch64_sve_smax", "aarch64_sve_smax_u">; 599defm SVMAX_U : SInstZPZZ<"svmax", "UcUsUiUl", "aarch64_sve_umax", "aarch64_sve_umax_u">; 600defm SVMIN_S : SInstZPZZ<"svmin", "csil", "aarch64_sve_smin", "aarch64_sve_smin_u">; 601defm SVMIN_U : SInstZPZZ<"svmin", "UcUsUiUl", "aarch64_sve_umin", "aarch64_sve_umin_u">; 602defm SVMUL : SInstZPZZ<"svmul", "csilUcUsUiUl", "aarch64_sve_mul", "aarch64_sve_mul_u">; 603defm SVMULH_S : SInstZPZZ<"svmulh", "csil", "aarch64_sve_smulh", "aarch64_sve_smulh_u">; 604defm SVMULH_U : SInstZPZZ<"svmulh", "UcUsUiUl", "aarch64_sve_umulh", "aarch64_sve_umulh_u">; 605defm SVSUB : SInstZPZZ<"svsub", "csilUcUsUiUl", "aarch64_sve_sub", "aarch64_sve_sub_u">; 606defm SVSUBR : SInstZPZZ<"svsubr", "csilUcUsUiUl", "aarch64_sve_subr", "aarch64_sve_sub_u", [ReverseMergeAnyBinOp]>; 607 608//------------------------------------------------------------------------------ 609 610multiclass SInstZPZZZ<string name, string types, string m_intrinsic, string x_intrinsic, list<FlagType> flags=[]> { 611 def _M : SInst<name # "[_{d}]", "dPddd", types, MergeOp1, m_intrinsic, flags>; 612 def _X : SInst<name # "[_{d}]", "dPddd", types, MergeAny, x_intrinsic, flags>; 613 def _Z : SInst<name # "[_{d}]", "dPddd", types, MergeZero, m_intrinsic, flags>; 614 615 def _N_M : SInst<name # "[_n_{d}]", "dPdda", types, MergeOp1, m_intrinsic, flags>; 616 def _N_X : SInst<name # "[_n_{d}]", "dPdda", types, MergeAny, x_intrinsic, flags>; 617 def _N_Z : SInst<name # "[_n_{d}]", "dPdda", types, MergeZero, m_intrinsic, flags>; 618} 619 620defm SVMAD : SInstZPZZZ<"svmad", "csilUcUsUiUl", "aarch64_sve_mad", "aarch64_sve_mla_u", [ReverseMergeAnyAccOp, IsStreamingCompatible]>; 621defm SVMLA : SInstZPZZZ<"svmla", "csilUcUsUiUl", "aarch64_sve_mla", "aarch64_sve_mla_u", [IsStreamingCompatible]>; 622defm SVMLS : SInstZPZZZ<"svmls", "csilUcUsUiUl", "aarch64_sve_mls", "aarch64_sve_mls_u", [IsStreamingCompatible]>; 623defm SVMSB : SInstZPZZZ<"svmsb", "csilUcUsUiUl", "aarch64_sve_msb", "aarch64_sve_mls_u", [ReverseMergeAnyAccOp, IsStreamingCompatible]>; 624 625//------------------------------------------------------------------------------ 626 627def SVDOT_S : SInst<"svdot[_{0}]", "ddqq", "il", MergeNone, "aarch64_sve_sdot", [IsStreamingCompatible]>; 628def SVDOT_U : SInst<"svdot[_{0}]", "ddqq", "UiUl", MergeNone, "aarch64_sve_udot", [IsStreamingCompatible]>; 629def SVQADD_S : SInst<"svqadd[_{d}]", "ddd", "csil", MergeNone, "aarch64_sve_sqadd_x", [IsStreamingCompatible]>; 630def SVQADD_U : SInst<"svqadd[_{d}]", "ddd", "UcUsUiUl", MergeNone, "aarch64_sve_uqadd_x", [IsStreamingCompatible]>; 631def SVQSUB_S : SInst<"svqsub[_{d}]", "ddd", "csil", MergeNone, "aarch64_sve_sqsub_x", [IsStreamingCompatible]>; 632def SVQSUB_U : SInst<"svqsub[_{d}]", "ddd", "UcUsUiUl", MergeNone, "aarch64_sve_uqsub_x", [IsStreamingCompatible]>; 633 634def SVDOT_N_S : SInst<"svdot[_n_{0}]", "ddqr", "il", MergeNone, "aarch64_sve_sdot", [IsStreamingCompatible]>; 635def SVDOT_N_U : SInst<"svdot[_n_{0}]", "ddqr", "UiUl", MergeNone, "aarch64_sve_udot", [IsStreamingCompatible]>; 636def SVQADD_N_S : SInst<"svqadd[_n_{d}]", "dda", "csil", MergeNone, "aarch64_sve_sqadd_x", [IsStreamingCompatible]>; 637def SVQADD_N_U : SInst<"svqadd[_n_{d}]", "dda", "UcUsUiUl", MergeNone, "aarch64_sve_uqadd_x", [IsStreamingCompatible]>; 638def SVQSUB_N_S : SInst<"svqsub[_n_{d}]", "dda", "csil", MergeNone, "aarch64_sve_sqsub_x", [IsStreamingCompatible]>; 639def SVQSUB_N_U : SInst<"svqsub[_n_{d}]", "dda", "UcUsUiUl", MergeNone, "aarch64_sve_uqsub_x", [IsStreamingCompatible]>; 640 641def SVDOT_LANE_S : SInst<"svdot_lane[_{d}]", "ddqqi", "il", MergeNone, "aarch64_sve_sdot_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndexDot, 2>]>; 642def SVDOT_LANE_U : SInst<"svdot_lane[_{d}]", "ddqqi", "UiUl", MergeNone, "aarch64_sve_udot_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndexDot, 2>]>; 643 644//////////////////////////////////////////////////////////////////////////////// 645// Logical operations 646 647defm SVAND : SInstZPZZ<"svand", "csilUcUsUiUl", "aarch64_sve_and", "aarch64_sve_and_u">; 648defm SVBIC : SInstZPZZ<"svbic", "csilUcUsUiUl", "aarch64_sve_bic", "aarch64_sve_bic_u">; 649defm SVEOR : SInstZPZZ<"sveor", "csilUcUsUiUl", "aarch64_sve_eor", "aarch64_sve_eor_u">; 650defm SVORR : SInstZPZZ<"svorr", "csilUcUsUiUl", "aarch64_sve_orr", "aarch64_sve_orr_u">; 651 652defm SVCNOT : SInstZPZ<"svcnot", "csilUcUsUiUl", "aarch64_sve_cnot">; 653defm SVNOT : SInstZPZ<"svnot", "csilUcUsUiUl", "aarch64_sve_not">; 654 655//////////////////////////////////////////////////////////////////////////////// 656// Shifts 657 658multiclass SInst_SHIFT<string name, string intrinsic, string ts, string wide_ts> { 659 def _M : SInst<name # "[_{d}]", "dPdu", ts, MergeOp1, intrinsic, [IsStreamingCompatible]>; 660 def _X : SInst<name # "[_{d}]", "dPdu", ts, MergeAny, intrinsic # _u, [IsStreamingCompatible]>; 661 def _Z : SInst<name # "[_{d}]", "dPdu", ts, MergeZero, intrinsic, [IsStreamingCompatible]>; 662 663 def _N_M : SInst<name # "[_n_{d}]", "dPdL", ts, MergeOp1, intrinsic, [IsStreamingCompatible]>; 664 def _N_X : SInst<name # "[_n_{d}]", "dPdL", ts, MergeAny, intrinsic # _u, [IsStreamingCompatible]>; 665 def _N_Z : SInst<name # "[_n_{d}]", "dPdL", ts, MergeZero, intrinsic, [IsStreamingCompatible]>; 666 667 def _WIDE_M : SInst<name # _wide # "[_{d}]", "dPdg", wide_ts, MergeOp1, intrinsic # _wide, [IsStreamingCompatible]>; 668 def _WIDE_X : SInst<name # _wide # "[_{d}]", "dPdg", wide_ts, MergeAny, intrinsic # _wide, [IsStreamingCompatible]>; 669 def _WIDE_Z : SInst<name # _wide # "[_{d}]", "dPdg", wide_ts, MergeZero, intrinsic # _wide, [IsStreamingCompatible]>; 670 671 def _WIDE_N_M : SInst<name # _wide # "[_n_{d}]", "dPdf", wide_ts, MergeOp1, intrinsic # _wide, [IsStreamingCompatible]>; 672 def _WIDE_N_X : SInst<name # _wide # "[_n_{d}]", "dPdf", wide_ts, MergeAny, intrinsic # _wide, [IsStreamingCompatible]>; 673 def _WIDE_N_Z : SInst<name # _wide # "[_n_{d}]", "dPdf", wide_ts, MergeZero, intrinsic # _wide, [IsStreamingCompatible]>; 674} 675 676defm SVASR : SInst_SHIFT<"svasr", "aarch64_sve_asr", "csil", "csi">; 677defm SVLSL : SInst_SHIFT<"svlsl", "aarch64_sve_lsl", "csilUcUsUiUl", "csiUcUsUi">; 678defm SVLSR : SInst_SHIFT<"svlsr", "aarch64_sve_lsr", "UcUsUiUl", "UcUsUi">; 679 680def SVASRD_M : SInst<"svasrd[_n_{d}]", "dPdi", "csil", MergeOp1, "aarch64_sve_asrd", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 681def SVASRD_X : SInst<"svasrd[_n_{d}]", "dPdi", "csil", MergeAny, "aarch64_sve_asrd", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 682def SVASRD_Z : SInst<"svasrd[_n_{d}]", "dPdi", "csil", MergeZero, "aarch64_sve_asrd", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 683 684def SVINSR : SInst<"svinsr[_n_{d}]", "dds", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_insr", [IsStreamingCompatible]>; 685let TargetGuard = "sve,bf16" in { 686 def SVINSR_BF16 : SInst<"svinsr[_n_{d}]", "dds", "b", MergeNone, "aarch64_sve_insr", [IsStreamingCompatible]>; 687} 688 689//////////////////////////////////////////////////////////////////////////////// 690// Integer reductions 691 692def SVADDV_S : SInst<"svaddv[_{d}]", "lPd", "csil", MergeNone, "aarch64_sve_saddv", [IsStreamingCompatible]>; 693def SVADDV_U : SInst<"svaddv[_{d}]", "nPd", "UcUsUiUl", MergeNone, "aarch64_sve_uaddv", [IsStreamingCompatible]>; 694def SVANDV : SInst<"svandv[_{d}]", "sPd", "csilUcUsUiUl", MergeNone, "aarch64_sve_andv", [IsStreamingCompatible]>; 695def SVEORV : SInst<"sveorv[_{d}]", "sPd", "csilUcUsUiUl", MergeNone, "aarch64_sve_eorv", [IsStreamingCompatible]>; 696def SVMAXV_S : SInst<"svmaxv[_{d}]", "sPd", "csil", MergeNone, "aarch64_sve_smaxv", [IsStreamingCompatible]>; 697def SVMAXV_U : SInst<"svmaxv[_{d}]", "sPd", "UcUsUiUl", MergeNone, "aarch64_sve_umaxv", [IsStreamingCompatible]>; 698def SVMINV_S : SInst<"svminv[_{d}]", "sPd", "csil", MergeNone, "aarch64_sve_sminv", [IsStreamingCompatible]>; 699def SVMINV_U : SInst<"svminv[_{d}]", "sPd", "UcUsUiUl", MergeNone, "aarch64_sve_uminv", [IsStreamingCompatible]>; 700def SVORV : SInst<"svorv[_{d}]", "sPd", "csilUcUsUiUl", MergeNone, "aarch64_sve_orv", [IsStreamingCompatible]>; 701 702//////////////////////////////////////////////////////////////////////////////// 703// Integer comparisons 704 705def SVCMPEQ : SInst<"svcmpeq[_{d}]", "PPdd", "csilUcUsUiUl", MergeNone, "aarch64_sve_cmpeq", [IsStreamingCompatible]>; 706def SVCMPNE : SInst<"svcmpne[_{d}]", "PPdd", "csilUcUsUiUl", MergeNone, "aarch64_sve_cmpne", [IsStreamingCompatible]>; 707def SVCMPGE : SInst<"svcmpge[_{d}]", "PPdd", "csil", MergeNone, "aarch64_sve_cmpge", [IsStreamingCompatible]>; 708def SVCMPGT : SInst<"svcmpgt[_{d}]", "PPdd", "csil", MergeNone, "aarch64_sve_cmpgt", [IsStreamingCompatible]>; 709def SVCMPLE : SInst<"svcmple[_{d}]", "PPdd", "csil", MergeNone, "aarch64_sve_cmpge", [ReverseCompare, IsStreamingCompatible]>; 710def SVCMPLT : SInst<"svcmplt[_{d}]", "PPdd", "csil", MergeNone, "aarch64_sve_cmpgt", [ReverseCompare, IsStreamingCompatible]>; 711def SVCMPHI : SInst<"svcmpgt[_{d}]", "PPdd", "UcUsUiUl", MergeNone, "aarch64_sve_cmphi", [IsStreamingCompatible]>; 712def SVCMPHS : SInst<"svcmpge[_{d}]", "PPdd", "UcUsUiUl", MergeNone, "aarch64_sve_cmphs", [IsStreamingCompatible]>; 713def SVCMPLO : SInst<"svcmplt[_{d}]", "PPdd", "UcUsUiUl", MergeNone, "aarch64_sve_cmphi", [ReverseCompare, IsStreamingCompatible]>; 714def SVCMPLS : SInst<"svcmple[_{d}]", "PPdd", "UcUsUiUl", MergeNone, "aarch64_sve_cmphs", [ReverseCompare, IsStreamingCompatible]>; 715 716def SVCMPEQ_N : SInst<"svcmpeq[_n_{d}]", "PPda", "csilUcUsUiUl", MergeNone, "aarch64_sve_cmpeq", [IsStreamingCompatible]>; 717def SVCMPNE_N : SInst<"svcmpne[_n_{d}]", "PPda", "csilUcUsUiUl", MergeNone, "aarch64_sve_cmpne", [IsStreamingCompatible]>; 718def SVCMPGE_N : SInst<"svcmpge[_n_{d}]", "PPda", "csil", MergeNone, "aarch64_sve_cmpge", [IsStreamingCompatible]>; 719def SVCMPGT_N : SInst<"svcmpgt[_n_{d}]", "PPda", "csil", MergeNone, "aarch64_sve_cmpgt", [IsStreamingCompatible]>; 720def SVCMPLE_N : SInst<"svcmple[_n_{d}]", "PPda", "csil", MergeNone, "aarch64_sve_cmpge", [ReverseCompare, IsStreamingCompatible]>; 721def SVCMPLT_N : SInst<"svcmplt[_n_{d}]", "PPda", "csil", MergeNone, "aarch64_sve_cmpgt", [ReverseCompare, IsStreamingCompatible]>; 722def SVCMPHS_N : SInst<"svcmpge[_n_{d}]", "PPda", "UcUsUiUl", MergeNone, "aarch64_sve_cmphs", [IsStreamingCompatible]>; 723def SVCMPHI_N : SInst<"svcmpgt[_n_{d}]", "PPda", "UcUsUiUl", MergeNone, "aarch64_sve_cmphi", [IsStreamingCompatible]>; 724def SVCMPLS_N : SInst<"svcmple[_n_{d}]", "PPda", "UcUsUiUl", MergeNone, "aarch64_sve_cmphs", [ReverseCompare, IsStreamingCompatible]>; 725def SVCMPLO_N : SInst<"svcmplt[_n_{d}]", "PPda", "UcUsUiUl", MergeNone, "aarch64_sve_cmphi", [ReverseCompare, IsStreamingCompatible]>; 726 727def SVCMPEQ_WIDE : SInst<"svcmpeq_wide[_{d}]", "PPdw", "csi", MergeNone, "aarch64_sve_cmpeq_wide", [IsStreamingCompatible]>; 728def SVCMPNE_WIDE : SInst<"svcmpne_wide[_{d}]", "PPdw", "csi", MergeNone, "aarch64_sve_cmpne_wide", [IsStreamingCompatible]>; 729def SVCMPGE_WIDE : SInst<"svcmpge_wide[_{d}]", "PPdw", "csi", MergeNone, "aarch64_sve_cmpge_wide", [IsStreamingCompatible]>; 730def SVCMPGT_WIDE : SInst<"svcmpgt_wide[_{d}]", "PPdw", "csi", MergeNone, "aarch64_sve_cmpgt_wide", [IsStreamingCompatible]>; 731def SVCMPLE_WIDE : SInst<"svcmple_wide[_{d}]", "PPdw", "csi", MergeNone, "aarch64_sve_cmple_wide", [IsStreamingCompatible]>; 732def SVCMPLT_WIDE : SInst<"svcmplt_wide[_{d}]", "PPdw", "csi", MergeNone, "aarch64_sve_cmplt_wide", [IsStreamingCompatible]>; 733def SVCMPHI_WIDE : SInst<"svcmpgt_wide[_{d}]", "PPdw", "UcUsUi", MergeNone, "aarch64_sve_cmphi_wide", [IsStreamingCompatible]>; 734def SVCMPHS_WIDE : SInst<"svcmpge_wide[_{d}]", "PPdw", "UcUsUi", MergeNone, "aarch64_sve_cmphs_wide", [IsStreamingCompatible]>; 735def SVCMPLO_WIDE : SInst<"svcmplt_wide[_{d}]", "PPdw", "UcUsUi", MergeNone, "aarch64_sve_cmplo_wide", [IsStreamingCompatible]>; 736def SVCMPLS_WIDE : SInst<"svcmple_wide[_{d}]", "PPdw", "UcUsUi", MergeNone, "aarch64_sve_cmpls_wide", [IsStreamingCompatible]>; 737 738def SVCMPEQ_WIDE_N : SInst<"svcmpeq_wide[_n_{d}]", "PPdj", "csi", MergeNone, "aarch64_sve_cmpeq_wide", [IsStreamingCompatible]>; 739def SVCMPNE_WIDE_N : SInst<"svcmpne_wide[_n_{d}]", "PPdj", "csi", MergeNone, "aarch64_sve_cmpne_wide", [IsStreamingCompatible]>; 740def SVCMPGE_WIDE_N : SInst<"svcmpge_wide[_n_{d}]", "PPdj", "csi", MergeNone, "aarch64_sve_cmpge_wide", [IsStreamingCompatible]>; 741def SVCMPGT_WIDE_N : SInst<"svcmpgt_wide[_n_{d}]", "PPdj", "csi", MergeNone, "aarch64_sve_cmpgt_wide", [IsStreamingCompatible]>; 742def SVCMPLE_WIDE_N : SInst<"svcmple_wide[_n_{d}]", "PPdj", "csi", MergeNone, "aarch64_sve_cmple_wide", [IsStreamingCompatible]>; 743def SVCMPLT_WIDE_N : SInst<"svcmplt_wide[_n_{d}]", "PPdj", "csi", MergeNone, "aarch64_sve_cmplt_wide", [IsStreamingCompatible]>; 744def SVCMPHS_WIDE_N : SInst<"svcmpge_wide[_n_{d}]", "PPdj", "UcUsUi", MergeNone, "aarch64_sve_cmphs_wide", [IsStreamingCompatible]>; 745def SVCMPHI_WIDE_N : SInst<"svcmpgt_wide[_n_{d}]", "PPdj", "UcUsUi", MergeNone, "aarch64_sve_cmphi_wide", [IsStreamingCompatible]>; 746def SVCMPLO_WIDE_N : SInst<"svcmplt_wide[_n_{d}]", "PPdj", "UcUsUi", MergeNone, "aarch64_sve_cmplo_wide", [IsStreamingCompatible]>; 747def SVCMPLS_WIDE_N : SInst<"svcmple_wide[_n_{d}]", "PPdj", "UcUsUi", MergeNone, "aarch64_sve_cmpls_wide", [IsStreamingCompatible]>; 748 749//////////////////////////////////////////////////////////////////////////////// 750// While comparisons 751 752def SVWHILELE_S32 : SInst<"svwhilele_{d}[_{1}]", "Pkk", "PcPsPiPl", MergeNone, "aarch64_sve_whilele", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 753def SVWHILELE_S64 : SInst<"svwhilele_{d}[_{1}]", "Pll", "PcPsPiPl", MergeNone, "aarch64_sve_whilele", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 754def SVWHILELO_U32 : SInst<"svwhilelt_{d}[_{1}]", "Pmm", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilelo", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 755def SVWHILELO_U64 : SInst<"svwhilelt_{d}[_{1}]", "Pnn", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilelo", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 756def SVWHILELS_U32 : SInst<"svwhilele_{d}[_{1}]", "Pmm", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilels", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 757def SVWHILELS_U64 : SInst<"svwhilele_{d}[_{1}]", "Pnn", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilels", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 758def SVWHILELT_S32 : SInst<"svwhilelt_{d}[_{1}]", "Pkk", "PcPsPiPl", MergeNone, "aarch64_sve_whilelt", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 759def SVWHILELT_S64 : SInst<"svwhilelt_{d}[_{1}]", "Pll", "PcPsPiPl", MergeNone, "aarch64_sve_whilelt", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 760 761//////////////////////////////////////////////////////////////////////////////// 762// Counting bit 763 764multiclass SInstCLS<string name, string types, string intrinsic, list<FlagType> flags=[]> { 765 def _M : SInst<name # "[_{d}]", "uuPd", types, MergeOp1, intrinsic, flags>; 766 def _X : SInst<name # "[_{d}]", "uPd", types, MergeAnyExp, intrinsic, flags>; 767 def _Z : SInst<name # "[_{d}]", "uPd", types, MergeZeroExp, intrinsic, flags>; 768} 769 770defm SVCLS : SInstCLS<"svcls", "csil", "aarch64_sve_cls", [IsStreamingCompatible]>; 771defm SVCLZ : SInstCLS<"svclz", "csilUcUsUiUl", "aarch64_sve_clz", [IsStreamingCompatible]>; 772defm SVCNT : SInstCLS<"svcnt", "csilUcUsUiUlhfd", "aarch64_sve_cnt", [IsStreamingCompatible]>; 773 774let TargetGuard = "sve,bf16" in { 775 defm SVCNT_BF16 : SInstCLS<"svcnt", "b", "aarch64_sve_cnt", [IsStreamingCompatible]>; 776} 777 778//////////////////////////////////////////////////////////////////////////////// 779// Conversion 780 781defm SVEXTB_S : SInstZPZ<"svextb", "sil", "aarch64_sve_sxtb">; 782defm SVEXTB_U : SInstZPZ<"svextb", "UsUiUl", "aarch64_sve_uxtb">; 783defm SVEXTH_S : SInstZPZ<"svexth", "il", "aarch64_sve_sxth">; 784defm SVEXTH_U : SInstZPZ<"svexth", "UiUl", "aarch64_sve_uxth">; 785defm SVEXTW_S : SInstZPZ<"svextw", "l", "aarch64_sve_sxtw">; 786defm SVEXTW_U : SInstZPZ<"svextw", "Ul", "aarch64_sve_uxtw">; 787 788//////////////////////////////////////////////////////////////////////////////// 789// Reversal 790 791defm SVRBIT : SInstZPZ<"svrbit", "csilUcUsUiUl", "aarch64_sve_rbit">; 792defm SVREVB : SInstZPZ<"svrevb", "silUsUiUl", "aarch64_sve_revb">; 793defm SVREVH : SInstZPZ<"svrevh", "ilUiUl", "aarch64_sve_revh">; 794defm SVREVW : SInstZPZ<"svrevw", "lUl", "aarch64_sve_revw">; 795 796//////////////////////////////////////////////////////////////////////////////// 797// Floating-point arithmetic 798 799defm SVABS_F : SInstZPZ<"svabs", "hfd", "aarch64_sve_fabs">; 800defm SVNEG_F : SInstZPZ<"svneg", "hfd", "aarch64_sve_fneg">; 801 802defm SVABD_F : SInstZPZZ<"svabd", "hfd", "aarch64_sve_fabd", "aarch64_sve_fabd_u">; 803defm SVADD_F : SInstZPZZ<"svadd", "hfd", "aarch64_sve_fadd", "aarch64_sve_fadd_u">; 804defm SVDIV_F : SInstZPZZ<"svdiv", "hfd", "aarch64_sve_fdiv", "aarch64_sve_fdiv_u">; 805defm SVDIVR_F : SInstZPZZ<"svdivr", "hfd", "aarch64_sve_fdivr", "aarch64_sve_fdiv_u", [ReverseMergeAnyBinOp]>; 806defm SVMAX_F : SInstZPZZ<"svmax", "hfd", "aarch64_sve_fmax", "aarch64_sve_fmax_u">; 807defm SVMAXNM : SInstZPZZ<"svmaxnm","hfd", "aarch64_sve_fmaxnm", "aarch64_sve_fmaxnm_u">; 808defm SVMIN_F : SInstZPZZ<"svmin", "hfd", "aarch64_sve_fmin", "aarch64_sve_fmin_u">; 809defm SVMINNM : SInstZPZZ<"svminnm","hfd", "aarch64_sve_fminnm", "aarch64_sve_fminnm_u">; 810defm SVMUL_F : SInstZPZZ<"svmul", "hfd", "aarch64_sve_fmul", "aarch64_sve_fmul_u">; 811defm SVMULX : SInstZPZZ<"svmulx", "hfd", "aarch64_sve_fmulx", "aarch64_sve_fmulx_u">; 812defm SVSUB_F : SInstZPZZ<"svsub", "hfd", "aarch64_sve_fsub", "aarch64_sve_fsub_u">; 813defm SVSUBR_F : SInstZPZZ<"svsubr", "hfd", "aarch64_sve_fsubr", "aarch64_sve_fsub_u", [ReverseMergeAnyBinOp]>; 814 815defm SVRECPX : SInstZPZ<"svrecpx", "hfd", "aarch64_sve_frecpx">; 816defm SVRINTA : SInstZPZ<"svrinta", "hfd", "aarch64_sve_frinta">; 817defm SVRINTI : SInstZPZ<"svrinti", "hfd", "aarch64_sve_frinti">; 818defm SVRINTM : SInstZPZ<"svrintm", "hfd", "aarch64_sve_frintm">; 819defm SVRINTN : SInstZPZ<"svrintn", "hfd", "aarch64_sve_frintn">; 820defm SVRINTP : SInstZPZ<"svrintp", "hfd", "aarch64_sve_frintp">; 821defm SVRINTX : SInstZPZ<"svrintx", "hfd", "aarch64_sve_frintx">; 822defm SVRINTZ : SInstZPZ<"svrintz", "hfd", "aarch64_sve_frintz">; 823defm SVSQRT : SInstZPZ<"svsqrt", "hfd", "aarch64_sve_fsqrt">; 824 825def SVEXPA : SInst<"svexpa[_{d}]", "du", "hfd", MergeNone, "aarch64_sve_fexpa_x">; 826def SVTMAD : SInst<"svtmad[_{d}]", "dddi", "hfd", MergeNone, "aarch64_sve_ftmad_x", [], [ImmCheck<2, ImmCheck0_7>]>; 827def SVTSMUL : SInst<"svtsmul[_{d}]", "ddu", "hfd", MergeNone, "aarch64_sve_ftsmul_x">; 828def SVTSSEL : SInst<"svtssel[_{d}]", "ddu", "hfd", MergeNone, "aarch64_sve_ftssel_x">; 829 830def SVSCALE_M : SInst<"svscale[_{d}]", "dPdx", "hfd", MergeOp1, "aarch64_sve_fscale", [IsStreamingCompatible]>; 831def SVSCALE_X : SInst<"svscale[_{d}]", "dPdx", "hfd", MergeAny, "aarch64_sve_fscale", [IsStreamingCompatible]>; 832def SVSCALE_Z : SInst<"svscale[_{d}]", "dPdx", "hfd", MergeZero, "aarch64_sve_fscale", [IsStreamingCompatible]>; 833 834def SVSCALE_N_M : SInst<"svscale[_n_{d}]", "dPdK", "hfd", MergeOp1, "aarch64_sve_fscale", [IsStreamingCompatible]>; 835def SVSCALE_N_X : SInst<"svscale[_n_{d}]", "dPdK", "hfd", MergeAny, "aarch64_sve_fscale", [IsStreamingCompatible]>; 836def SVSCALE_N_Z : SInst<"svscale[_n_{d}]", "dPdK", "hfd", MergeZero, "aarch64_sve_fscale", [IsStreamingCompatible]>; 837 838defm SVMAD_F : SInstZPZZZ<"svmad", "hfd", "aarch64_sve_fmad", "aarch64_sve_fmla_u", [IsStreamingCompatible, ReverseMergeAnyAccOp]>; 839defm SVMLA_F : SInstZPZZZ<"svmla", "hfd", "aarch64_sve_fmla", "aarch64_sve_fmla_u", [IsStreamingCompatible]>; 840defm SVMLS_F : SInstZPZZZ<"svmls", "hfd", "aarch64_sve_fmls", "aarch64_sve_fmls_u", [IsStreamingCompatible]>; 841defm SVMSB_F : SInstZPZZZ<"svmsb", "hfd", "aarch64_sve_fmsb", "aarch64_sve_fmls_u", [IsStreamingCompatible, ReverseMergeAnyAccOp]>; 842defm SVNMAD_F : SInstZPZZZ<"svnmad", "hfd", "aarch64_sve_fnmad", "aarch64_sve_fnmla_u", [IsStreamingCompatible, ReverseMergeAnyAccOp]>; 843defm SVNMLA_F : SInstZPZZZ<"svnmla", "hfd", "aarch64_sve_fnmla", "aarch64_sve_fnmla_u", [IsStreamingCompatible]>; 844defm SVNMLS_F : SInstZPZZZ<"svnmls", "hfd", "aarch64_sve_fnmls", "aarch64_sve_fnmls_u", [IsStreamingCompatible]>; 845defm SVNMSB_F : SInstZPZZZ<"svnmsb", "hfd", "aarch64_sve_fnmsb", "aarch64_sve_fnmls_u", [IsStreamingCompatible, ReverseMergeAnyAccOp]>; 846 847def SVCADD_M : SInst<"svcadd[_{d}]", "dPddi", "hfd", MergeOp1, "aarch64_sve_fcadd", [IsStreamingCompatible], [ImmCheck<3, ImmCheckComplexRot90_270>]>; 848def SVCADD_X : SInst<"svcadd[_{d}]", "dPddi", "hfd", MergeAny, "aarch64_sve_fcadd", [IsStreamingCompatible], [ImmCheck<3, ImmCheckComplexRot90_270>]>; 849def SVCADD_Z : SInst<"svcadd[_{d}]", "dPddi", "hfd", MergeZero, "aarch64_sve_fcadd", [IsStreamingCompatible], [ImmCheck<3, ImmCheckComplexRot90_270>]>; 850def SVCMLA_M : SInst<"svcmla[_{d}]", "dPdddi", "hfd", MergeOp1, "aarch64_sve_fcmla", [IsStreamingCompatible], [ImmCheck<4, ImmCheckComplexRotAll90>]>; 851def SVCMLA_X : SInst<"svcmla[_{d}]", "dPdddi", "hfd", MergeAny, "aarch64_sve_fcmla", [IsStreamingCompatible], [ImmCheck<4, ImmCheckComplexRotAll90>]>; 852def SVCMLA_Z : SInst<"svcmla[_{d}]", "dPdddi", "hfd", MergeZero, "aarch64_sve_fcmla", [IsStreamingCompatible], [ImmCheck<4, ImmCheckComplexRotAll90>]>; 853 854def SVCMLA_LANE : SInst<"svcmla_lane[_{d}]", "ddddii", "hf", MergeNone, "aarch64_sve_fcmla_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndexCompRotate, 2>, 855 ImmCheck<4, ImmCheckComplexRotAll90>]>; 856def SVMLA_LANE : SInst<"svmla_lane[_{d}]", "ddddi", "hfd", MergeNone, "aarch64_sve_fmla_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 857def SVMLS_LANE : SInst<"svmls_lane[_{d}]", "ddddi", "hfd", MergeNone, "aarch64_sve_fmls_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 858def SVMUL_LANE : SInst<"svmul_lane[_{d}]", "dddi", "hfd", MergeNone, "aarch64_sve_fmul_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 859 860def SVRECPE : SInst<"svrecpe[_{d}]", "dd", "hfd", MergeNone, "aarch64_sve_frecpe_x", [IsStreamingCompatible]>; 861def SVRECPS : SInst<"svrecps[_{d}]", "ddd", "hfd", MergeNone, "aarch64_sve_frecps_x", [IsStreamingCompatible]>; 862def SVRSQRTE : SInst<"svrsqrte[_{d}]", "dd", "hfd", MergeNone, "aarch64_sve_frsqrte_x", [IsStreamingCompatible]>; 863def SVRSQRTS : SInst<"svrsqrts[_{d}]", "ddd", "hfd", MergeNone, "aarch64_sve_frsqrts_x", [IsStreamingCompatible]>; 864 865//////////////////////////////////////////////////////////////////////////////// 866// Floating-point reductions 867 868def SVFADDA : SInst<"svadda[_{d}]", "sPsd", "hfd", MergeNone, "aarch64_sve_fadda", [IsStreamingCompatible]>; 869def SVFADDV : SInst<"svaddv[_{d}]", "sPd", "hfd", MergeNone, "aarch64_sve_faddv", [IsStreamingCompatible]>; 870def SVFMAXV : SInst<"svmaxv[_{d}]", "sPd", "hfd", MergeNone, "aarch64_sve_fmaxv", [IsStreamingCompatible]>; 871def SVFMAXNMV : SInst<"svmaxnmv[_{d}]", "sPd", "hfd", MergeNone, "aarch64_sve_fmaxnmv", [IsStreamingCompatible]>; 872def SVFMINV : SInst<"svminv[_{d}]", "sPd", "hfd", MergeNone, "aarch64_sve_fminv", [IsStreamingCompatible]>; 873def SVFMINNMV : SInst<"svminnmv[_{d}]", "sPd", "hfd", MergeNone, "aarch64_sve_fminnmv", [IsStreamingCompatible]>; 874 875//////////////////////////////////////////////////////////////////////////////// 876// Floating-point comparisons 877 878def SVACGE : SInst<"svacge[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_facge", [IsStreamingCompatible]>; 879def SVACGT : SInst<"svacgt[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_facgt", [IsStreamingCompatible]>; 880def SVACLE : SInst<"svacle[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_facge", [ReverseCompare, IsStreamingCompatible]>; 881def SVACLT : SInst<"svaclt[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_facgt", [ReverseCompare, IsStreamingCompatible]>; 882def SVCMPUO : SInst<"svcmpuo[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_fcmpuo", [IsStreamingCompatible]>; 883 884def SVACGE_N : SInst<"svacge[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_facge", [IsStreamingCompatible]>; 885def SVACGT_N : SInst<"svacgt[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_facgt", [IsStreamingCompatible]>; 886def SVACLE_N : SInst<"svacle[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_facge", [ReverseCompare, IsStreamingCompatible]>; 887def SVACLT_N : SInst<"svaclt[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_facgt", [ReverseCompare, IsStreamingCompatible]>; 888def SVCMPUO_N : SInst<"svcmpuo[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_fcmpuo", [IsStreamingCompatible]>; 889 890def SVCMPEQ_F : SInst<"svcmpeq[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_fcmpeq", [IsStreamingCompatible]>; 891def SVCMPNE_F : SInst<"svcmpne[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_fcmpne", [IsStreamingCompatible]>; 892def SVCMPGE_F : SInst<"svcmpge[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_fcmpge", [IsStreamingCompatible]>; 893def SVCMPGT_F : SInst<"svcmpgt[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_fcmpgt", [IsStreamingCompatible]>; 894def SVCMPLE_F : SInst<"svcmple[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_fcmpge", [ReverseCompare, IsStreamingCompatible]>; 895def SVCMPLT_F : SInst<"svcmplt[_{d}]", "PPdd", "hfd", MergeNone, "aarch64_sve_fcmpgt", [ReverseCompare, IsStreamingCompatible]>; 896 897def SVCMPEQ_F_N : SInst<"svcmpeq[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_fcmpeq", [IsStreamingCompatible]>; 898def SVCMPNE_F_N : SInst<"svcmpne[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_fcmpne", [IsStreamingCompatible]>; 899def SVCMPGE_F_N : SInst<"svcmpge[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_fcmpge", [IsStreamingCompatible]>; 900def SVCMPGT_F_N : SInst<"svcmpgt[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_fcmpgt", [IsStreamingCompatible]>; 901def SVCMPLE_F_N : SInst<"svcmple[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_fcmpge", [ReverseCompare, IsStreamingCompatible]>; 902def SVCMPLT_F_N : SInst<"svcmplt[_n_{d}]", "PPda", "hfd", MergeNone, "aarch64_sve_fcmpgt", [ReverseCompare, IsStreamingCompatible]>; 903 904//////////////////////////////////////////////////////////////////////////////// 905// Floating-point conversions 906 907multiclass SInstCvtMXZ< 908 string name, string m_types, string xz_types, string types, 909 string intrinsic, list<FlagType> flags = [IsOverloadNone]> { 910 def _M : SInst<name, m_types, types, MergeOp1, intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 911 def _X : SInst<name, xz_types, types, MergeAnyExp, intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 912 def _Z : SInst<name, xz_types, types, MergeZeroExp, intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 913} 914 915multiclass SInstCvtMX<string name, string m_types, string xz_types, 916 string types, string intrinsic, 917 list<FlagType> flags = [IsOverloadNone]> { 918 def _M : SInst<name, m_types, types, MergeOp1, intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 919 def _X : SInst<name, xz_types, types, MergeAnyExp, intrinsic, !listconcat(flags, [IsStreamingCompatible])>; 920} 921 922// svcvt_s##_f16 923defm SVFCVTZS_S16_F16 : SInstCvtMXZ<"svcvt_s16[_f16]", "ddPO", "dPO", "s", "aarch64_sve_fcvtzs", [IsOverloadCvt]>; 924defm SVFCVTZS_S32_F16 : SInstCvtMXZ<"svcvt_s32[_f16]", "ddPO", "dPO", "i", "aarch64_sve_fcvtzs_i32f16">; 925defm SVFCVTZS_S64_F16 : SInstCvtMXZ<"svcvt_s64[_f16]", "ddPO", "dPO", "l", "aarch64_sve_fcvtzs_i64f16">; 926 927// svcvt_s##_f32 928defm SVFCVTZS_S32_F32 : SInstCvtMXZ<"svcvt_s32[_f32]", "ddPM", "dPM", "i", "aarch64_sve_fcvtzs", [IsOverloadCvt]>; 929defm SVFCVTZS_S64_F32 : SInstCvtMXZ<"svcvt_s64[_f32]", "ddPM", "dPM", "l", "aarch64_sve_fcvtzs_i64f32">; 930 931let TargetGuard = "sve,bf16" in { 932 defm SVCVT_BF16_F32 : SInstCvtMXZ<"svcvt_bf16[_f32]", "ddPM", "dPM", "b", "aarch64_sve_fcvt_bf16f32">; 933 def SVCVTNT_BF16_F32 : SInst<"svcvtnt_bf16[_f32]", "ddPM", "b", MergeOp1, "aarch64_sve_fcvtnt_bf16f32", [IsOverloadNone, IsStreamingCompatible]>; 934} 935 936// svcvt_s##_f64 937defm SVFCVTZS_S32_F64 : SInstCvtMXZ<"svcvt_s32[_f64]", "ttPd", "tPd", "d", "aarch64_sve_fcvtzs_i32f64">; 938defm SVFCVTZS_S64_F64 : SInstCvtMXZ<"svcvt_s64[_f64]", "ddPN", "dPN", "l", "aarch64_sve_fcvtzs", [IsOverloadCvt]>; 939 940// svcvt_u##_f16 941defm SVFCVTZU_U16_F16 : SInstCvtMXZ<"svcvt_u16[_f16]", "ddPO", "dPO", "Us", "aarch64_sve_fcvtzu", [IsOverloadCvt]>; 942defm SVFCVTZU_U32_F16 : SInstCvtMXZ<"svcvt_u32[_f16]", "ddPO", "dPO", "Ui", "aarch64_sve_fcvtzu_i32f16">; 943defm SVFCVTZU_U64_F16 : SInstCvtMXZ<"svcvt_u64[_f16]", "ddPO", "dPO", "Ul", "aarch64_sve_fcvtzu_i64f16">; 944 945// svcvt_u##_f32 946defm SVFCVTZU_U32_F32 : SInstCvtMXZ<"svcvt_u32[_f32]", "ddPM", "dPM", "Ui", "aarch64_sve_fcvtzu", [IsOverloadCvt]>; 947defm SVFCVTZU_U64_F32 : SInstCvtMXZ<"svcvt_u64[_f32]", "ddPM", "dPM", "Ul", "aarch64_sve_fcvtzu_i64f32">; 948 949// svcvt_u##_f64 950defm SVFCVTZU_U32_F64 : SInstCvtMXZ<"svcvt_u32[_f64]", "zzPd", "zPd", "d", "aarch64_sve_fcvtzu_i32f64">; 951defm SVFCVTZU_U64_F64 : SInstCvtMXZ<"svcvt_u64[_f64]", "ddPN", "dPN", "Ul", "aarch64_sve_fcvtzu", [IsOverloadCvt]>; 952 953// svcvt_f16_s## 954defm SVFCVTZS_F16_S16 : SInstCvtMXZ<"svcvt_f16[_s16]", "OOPd", "OPd", "s", "aarch64_sve_scvtf", [IsOverloadCvt]>; 955defm SVFCVTZS_F16_S32 : SInstCvtMXZ<"svcvt_f16[_s32]", "OOPd", "OPd", "i", "aarch64_sve_scvtf_f16i32">; 956defm SVFCVTZS_F16_S64 : SInstCvtMXZ<"svcvt_f16[_s64]", "OOPd", "OPd", "l", "aarch64_sve_scvtf_f16i64">; 957 958// svcvt_f32_s## 959defm SVFCVTZS_F32_S32 : SInstCvtMXZ<"svcvt_f32[_s32]", "MMPd", "MPd", "i", "aarch64_sve_scvtf", [IsOverloadCvt]>; 960defm SVFCVTZS_F32_S64 : SInstCvtMXZ<"svcvt_f32[_s64]", "MMPd", "MPd", "l", "aarch64_sve_scvtf_f32i64">; 961 962// svcvt_f64_s## 963defm SVFCVTZS_F64_S32 : SInstCvtMXZ<"svcvt_f64[_s32]", "ddPt", "dPt", "d", "aarch64_sve_scvtf_f64i32">; 964defm SVFCVTZS_F64_S64 : SInstCvtMXZ<"svcvt_f64[_s64]", "NNPd", "NPd", "l", "aarch64_sve_scvtf", [IsOverloadCvt]>; 965 966// svcvt_f16_u## 967defm SVFCVTZU_F16_U16 : SInstCvtMXZ<"svcvt_f16[_u16]", "OOPd", "OPd", "Us", "aarch64_sve_ucvtf", [IsOverloadCvt]>; 968defm SVFCVTZU_F16_U32 : SInstCvtMXZ<"svcvt_f16[_u32]", "OOPd", "OPd", "Ui", "aarch64_sve_ucvtf_f16i32">; 969defm SVFCVTZU_F16_U64 : SInstCvtMXZ<"svcvt_f16[_u64]", "OOPd", "OPd", "Ul", "aarch64_sve_ucvtf_f16i64">; 970 971// svcvt_f32_u## 972defm SVFCVTZU_F32_U32 : SInstCvtMXZ<"svcvt_f32[_u32]", "MMPd", "MPd", "Ui", "aarch64_sve_ucvtf", [IsOverloadCvt]>; 973defm SVFCVTZU_F32_U64 : SInstCvtMXZ<"svcvt_f32[_u64]", "MMPd", "MPd", "Ul", "aarch64_sve_ucvtf_f32i64">; 974 975// svcvt_f64_u## 976defm SVFCVTZU_F64_U32 : SInstCvtMXZ<"svcvt_f64[_u32]", "ddPz", "dPz", "d", "aarch64_sve_ucvtf_f64i32">; 977defm SVFCVTZU_F64_U64 : SInstCvtMXZ<"svcvt_f64[_u64]", "NNPd", "NPd", "Ul", "aarch64_sve_ucvtf", [IsOverloadCvt]>; 978 979// svcvt_f16_f## 980defm SVFCVT_F16_F32 : SInstCvtMXZ<"svcvt_f16[_f32]", "OOPd", "OPd", "f", "aarch64_sve_fcvt_f16f32">; 981defm SVFCVT_F16_F64 : SInstCvtMXZ<"svcvt_f16[_f64]", "OOPd", "OPd", "d", "aarch64_sve_fcvt_f16f64">; 982 983// svcvt_f32_f## 984defm SVFCVT_F32_F16 : SInstCvtMXZ<"svcvt_f32[_f16]", "ddPO", "dPO", "f", "aarch64_sve_fcvt_f32f16">; 985defm SVFCVT_F32_F64 : SInstCvtMXZ<"svcvt_f32[_f64]", "MMPd", "MPd", "d", "aarch64_sve_fcvt_f32f64">; 986 987// svcvt_f64_f## 988defm SVFCVT_F64_F16 : SInstCvtMXZ<"svcvt_f64[_f16]", "ddPO", "dPO", "d", "aarch64_sve_fcvt_f64f16">; 989defm SVFCVT_F64_F32 : SInstCvtMXZ<"svcvt_f64[_f32]", "ddPM", "dPM", "d", "aarch64_sve_fcvt_f64f32">; 990 991let TargetGuard = "sve2" in { 992defm SVCVTLT_F32 : SInstCvtMX<"svcvtlt_f32[_f16]", "ddPh", "dPh", "f", "aarch64_sve_fcvtlt_f32f16">; 993defm SVCVTLT_F64 : SInstCvtMX<"svcvtlt_f64[_f32]", "ddPh", "dPh", "d", "aarch64_sve_fcvtlt_f64f32">; 994 995defm SVCVTX_F32 : SInstCvtMXZ<"svcvtx_f32[_f64]", "MMPd", "MPd", "d", "aarch64_sve_fcvtx_f32f64">; 996 997def SVCVTNT_F32 : SInst<"svcvtnt_f16[_f32]", "hhPd", "f", MergeOp1, "aarch64_sve_fcvtnt_f16f32", [IsOverloadNone, IsStreamingCompatible]>; 998def SVCVTNT_F64 : SInst<"svcvtnt_f32[_f64]", "hhPd", "d", MergeOp1, "aarch64_sve_fcvtnt_f32f64", [IsOverloadNone, IsStreamingCompatible]>; 999// SVCVTNT_X : Implemented as macro by SveEmitter.cpp 1000 1001def SVCVTXNT_F32 : SInst<"svcvtxnt_f32[_f64]", "MMPd", "d", MergeOp1, "aarch64_sve_fcvtxnt_f32f64", [IsOverloadNone, IsStreamingCompatible]>; 1002// SVCVTXNT_X_F32 : Implemented as macro by SveEmitter.cpp 1003 1004} 1005 1006//////////////////////////////////////////////////////////////////////////////// 1007// Permutations and selection 1008 1009multiclass SVEPerm<string name, string proto, string i> { 1010 def : SInst<name, proto, "csilUcUsUiUlhfd", MergeNone, i, [IsStreamingCompatible]>; 1011 let TargetGuard = "sve,bf16" in { 1012 def: SInst<name, proto, "b", MergeNone, i, [IsStreamingCompatible]>; 1013 } 1014} 1015 1016defm SVCLASTA : SVEPerm<"svclasta[_{d}]", "dPdd", "aarch64_sve_clasta">; 1017defm SVCLASTA_N : SVEPerm<"svclasta[_n_{d}]", "sPsd", "aarch64_sve_clasta_n">; 1018defm SVCLASTB : SVEPerm<"svclastb[_{d}]", "dPdd", "aarch64_sve_clastb">; 1019defm SVCLASTB_N : SVEPerm<"svclastb[_n_{d}]", "sPsd", "aarch64_sve_clastb_n">; 1020 1021def SVCOMPACT : SInst<"svcompact[_{d}]", "dPd", "ilUiUlfd", MergeNone, "aarch64_sve_compact">; 1022// Note: svdup_lane is implemented using the intrinsic for TBL to represent a 1023// splat of any possible lane. It is upto LLVM to pick a more efficient 1024// instruction such as DUP (indexed) if the lane index fits the range of the 1025// instruction's immediate. 1026def SVDUP_LANE : SInst<"svdup_lane[_{d}]", "ddL", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tbl", [IsStreamingCompatible]>; 1027let TargetGuard = "sve,bf16" in { 1028def SVDUP_LANE_BF16 : 1029 SInst<"svdup_lane[_{d}]", "ddL", "b", MergeNone, "aarch64_sve_tbl", [IsStreamingCompatible]>; 1030} 1031 1032def SVDUPQ_LANE : SInst<"svdupq_lane[_{d}]", "ddn", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_dupq_lane", [IsStreamingCompatible]>; 1033let TargetGuard = "sve,bf16" in { 1034 def SVDUPQ_LANE_BF16 : SInst<"svdupq_lane[_{d}]", "ddn", "b", MergeNone, "aarch64_sve_dupq_lane", [IsStreamingCompatible]>; 1035} 1036def SVEXT : SInst<"svext[_{d}]", "dddi", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_ext", [IsStreamingCompatible], [ImmCheck<2, ImmCheckExtract, 1>]>; 1037defm SVLASTA : SVEPerm<"svlasta[_{d}]", "sPd", "aarch64_sve_lasta">; 1038defm SVLASTB : SVEPerm<"svlastb[_{d}]", "sPd", "aarch64_sve_lastb">; 1039def SVREV : SInst<"svrev[_{d}]", "dd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_rev", [IsStreamingCompatible]>; 1040def SVSEL : SInst<"svsel[_{d}]", "dPdd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_sel", [IsStreamingCompatible]>; 1041def SVSPLICE : SInst<"svsplice[_{d}]", "dPdd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_splice", [IsStreamingCompatible]>; 1042def SVTBL : SInst<"svtbl[_{d}]", "ddu", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tbl", [IsStreamingCompatible]>; 1043 1044let TargetGuard = "sve,bf16" in { 1045 def SVTBL_BF16 : SInst<"svtbl[_{d}]", "ddu", "b", MergeNone, "aarch64_sve_tbl", [IsStreamingCompatible]>; 1046} 1047 1048def SVTRN1 : SInst<"svtrn1[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_trn1", [IsStreamingCompatible]>; 1049def SVTRN2 : SInst<"svtrn2[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_trn2", [IsStreamingCompatible]>; 1050def SVUNPKHI_S : SInst<"svunpkhi[_{d}]", "dh", "sil", MergeNone, "aarch64_sve_sunpkhi", [IsStreamingCompatible]>; 1051def SVUNPKHI_U : SInst<"svunpkhi[_{d}]", "dh", "UsUiUl", MergeNone, "aarch64_sve_uunpkhi", [IsStreamingCompatible]>; 1052def SVUNPKLO_S : SInst<"svunpklo[_{d}]", "dh", "sil", MergeNone, "aarch64_sve_sunpklo", [IsStreamingCompatible]>; 1053def SVUNPKLO_U : SInst<"svunpklo[_{d}]", "dh", "UsUiUl", MergeNone, "aarch64_sve_uunpklo", [IsStreamingCompatible]>; 1054def SVUZP1 : SInst<"svuzp1[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_uzp1", [IsStreamingCompatible]>; 1055def SVUZP2 : SInst<"svuzp2[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_uzp2", [IsStreamingCompatible]>; 1056def SVZIP1 : SInst<"svzip1[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_zip1", [IsStreamingCompatible]>; 1057def SVZIP2 : SInst<"svzip2[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_zip2", [IsStreamingCompatible]>; 1058 1059let TargetGuard = "sve,bf16" in { 1060def SVEXT_BF16 : SInst<"svext[_{d}]", "dddi", "b", MergeNone, "aarch64_sve_ext", [IsStreamingCompatible], [ImmCheck<2, ImmCheckExtract, 1>]>; 1061def SVREV_BF16 : SInst<"svrev[_{d}]", "dd", "b", MergeNone, "aarch64_sve_rev", [IsStreamingCompatible]>; 1062def SVSEL_BF16 : SInst<"svsel[_{d}]", "dPdd", "b", MergeNone, "aarch64_sve_sel", [IsStreamingCompatible]>; 1063def SVSPLICE_BF16 : SInst<"svsplice[_{d}]", "dPdd", "b", MergeNone, "aarch64_sve_splice", [IsStreamingCompatible]>; 1064def SVTRN1_BF16 : SInst<"svtrn1[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_trn1", [IsStreamingCompatible]>; 1065def SVTRN2_BF16 : SInst<"svtrn2[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_trn2", [IsStreamingCompatible]>; 1066def SVUZP1_BF16 : SInst<"svuzp1[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_uzp1", [IsStreamingCompatible]>; 1067def SVUZP2_BF16 : SInst<"svuzp2[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_uzp2", [IsStreamingCompatible]>; 1068def SVZIP1_BF16 : SInst<"svzip1[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_zip1", [IsStreamingCompatible]>; 1069def SVZIP2_BF16 : SInst<"svzip2[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_zip2", [IsStreamingCompatible]>; 1070} 1071 1072def SVREV_B8 : SInst<"svrev_b8", "PP", "Pc", MergeNone, "aarch64_sve_rev", [IsStreamingCompatible]>; 1073def SVREV_B16 : SInst<"svrev_b16", "PP", "Pc", MergeNone, "aarch64_sve_rev_b16", [IsOverloadNone, IsStreamingCompatible]>; 1074def SVREV_B32 : SInst<"svrev_b32", "PP", "Pc", MergeNone, "aarch64_sve_rev_b32", [IsOverloadNone, IsStreamingCompatible]>; 1075def SVREV_B64 : SInst<"svrev_b64", "PP", "Pc", MergeNone, "aarch64_sve_rev_b64", [IsOverloadNone, IsStreamingCompatible]>; 1076def SVSEL_B : SInst<"svsel[_b]", "PPPP", "Pc", MergeNone, "aarch64_sve_sel", [IsStreamingCompatible]>; 1077def SVTRN1_B8 : SInst<"svtrn1_b8", "PPP", "Pc", MergeNone, "aarch64_sve_trn1", [IsStreamingCompatible]>; 1078def SVTRN1_B16 : SInst<"svtrn1_b16", "PPP", "Pc", MergeNone, "aarch64_sve_trn1_b16", [IsOverloadNone, IsStreamingCompatible]>; 1079def SVTRN1_B32 : SInst<"svtrn1_b32", "PPP", "Pc", MergeNone, "aarch64_sve_trn1_b32", [IsOverloadNone, IsStreamingCompatible]>; 1080def SVTRN1_B64 : SInst<"svtrn1_b64", "PPP", "Pc", MergeNone, "aarch64_sve_trn1_b64", [IsOverloadNone, IsStreamingCompatible]>; 1081def SVTRN2_B8 : SInst<"svtrn2_b8", "PPP", "Pc", MergeNone, "aarch64_sve_trn2", [IsStreamingCompatible]>; 1082def SVTRN2_B16 : SInst<"svtrn2_b16", "PPP", "Pc", MergeNone, "aarch64_sve_trn2_b16", [IsOverloadNone, IsStreamingCompatible]>; 1083def SVTRN2_B32 : SInst<"svtrn2_b32", "PPP", "Pc", MergeNone, "aarch64_sve_trn2_b32", [IsOverloadNone, IsStreamingCompatible]>; 1084def SVTRN2_B64 : SInst<"svtrn2_b64", "PPP", "Pc", MergeNone, "aarch64_sve_trn2_b64", [IsOverloadNone, IsStreamingCompatible]>; 1085def SVPUNPKHI : SInst<"svunpkhi[_b]", "PP", "Pc", MergeNone, "aarch64_sve_punpkhi", [IsStreamingCompatible]>; 1086def SVPUNPKLO : SInst<"svunpklo[_b]", "PP", "Pc", MergeNone, "aarch64_sve_punpklo", [IsStreamingCompatible]>; 1087def SVUZP1_B8 : SInst<"svuzp1_b8", "PPP", "Pc", MergeNone, "aarch64_sve_uzp1", [IsStreamingCompatible]>; 1088def SVUZP1_B16 : SInst<"svuzp1_b16", "PPP", "Pc", MergeNone, "aarch64_sve_uzp1_b16", [IsOverloadNone, IsStreamingCompatible]>; 1089def SVUZP1_B32 : SInst<"svuzp1_b32", "PPP", "Pc", MergeNone, "aarch64_sve_uzp1_b32", [IsOverloadNone, IsStreamingCompatible]>; 1090def SVUZP1_B64 : SInst<"svuzp1_b64", "PPP", "Pc", MergeNone, "aarch64_sve_uzp1_b64", [IsOverloadNone, IsStreamingCompatible]>; 1091def SVUZP2_B8 : SInst<"svuzp2_b8", "PPP", "Pc", MergeNone, "aarch64_sve_uzp2", [IsStreamingCompatible]>; 1092def SVUZP2_B16 : SInst<"svuzp2_b16", "PPP", "Pc", MergeNone, "aarch64_sve_uzp2_b16", [IsOverloadNone, IsStreamingCompatible]>; 1093def SVUZP2_B32 : SInst<"svuzp2_b32", "PPP", "Pc", MergeNone, "aarch64_sve_uzp2_b32", [IsOverloadNone, IsStreamingCompatible]>; 1094def SVUZP2_B64 : SInst<"svuzp2_b64", "PPP", "Pc", MergeNone, "aarch64_sve_uzp2_b64", [IsOverloadNone, IsStreamingCompatible]>; 1095def SVZIP1_B8 : SInst<"svzip1_b8", "PPP", "Pc", MergeNone, "aarch64_sve_zip1", [IsStreamingCompatible]>; 1096def SVZIP1_B16 : SInst<"svzip1_b16", "PPP", "Pc", MergeNone, "aarch64_sve_zip1_b16", [IsOverloadNone, IsStreamingCompatible]>; 1097def SVZIP1_B32 : SInst<"svzip1_b32", "PPP", "Pc", MergeNone, "aarch64_sve_zip1_b32", [IsOverloadNone, IsStreamingCompatible]>; 1098def SVZIP1_B64 : SInst<"svzip1_b64", "PPP", "Pc", MergeNone, "aarch64_sve_zip1_b64", [IsOverloadNone, IsStreamingCompatible]>; 1099def SVZIP2_B : SInst<"svzip2_b8", "PPP", "Pc", MergeNone, "aarch64_sve_zip2", [IsStreamingCompatible]>; 1100def SVZIP2_B16 : SInst<"svzip2_b16", "PPP", "Pc", MergeNone, "aarch64_sve_zip2_b16", [IsOverloadNone, IsStreamingCompatible]>; 1101def SVZIP2_B32 : SInst<"svzip2_b32", "PPP", "Pc", MergeNone, "aarch64_sve_zip2_b32", [IsOverloadNone, IsStreamingCompatible]>; 1102def SVZIP2_B64 : SInst<"svzip2_b64", "PPP", "Pc", MergeNone, "aarch64_sve_zip2_b64", [IsOverloadNone, IsStreamingCompatible]>; 1103 1104//////////////////////////////////////////////////////////////////////////////// 1105// Predicate creation 1106 1107def SVPFALSE : SInst<"svpfalse[_b]", "Pv", "", MergeNone, "", [IsOverloadNone, IsStreamingCompatible]>; 1108 1109def SVPTRUE_PAT : SInst<"svptrue_pat_{d}", "PI", "PcPsPiPl", MergeNone, "aarch64_sve_ptrue", [IsStreamingCompatible]>; 1110def SVPTRUE : SInst<"svptrue_{d}", "Pv", "PcPsPiPl", MergeNone, "aarch64_sve_ptrue", [IsAppendSVALL, IsStreamingCompatible]>; 1111 1112def SVDUPQ_B8 : SInst<"svdupq[_n]_{d}", "Pssssssssssssssss", "Pc", MergeNone, "", [IsStreamingCompatible]>; 1113def SVDUPQ_B16 : SInst<"svdupq[_n]_{d}", "Pssssssss", "Ps", MergeNone, "", [IsStreamingCompatible]>; 1114def SVDUPQ_B32 : SInst<"svdupq[_n]_{d}", "Pssss", "Pi", MergeNone, "", [IsStreamingCompatible]>; 1115def SVDUPQ_B64 : SInst<"svdupq[_n]_{d}", "Pss", "Pl", MergeNone, "", [IsStreamingCompatible]>; 1116def SVDUP_N_B : SInst<"svdup[_n]_{d}", "Ps", "PcPsPiPl", MergeNone, "", [IsStreamingCompatible]>; 1117 1118 1119//////////////////////////////////////////////////////////////////////////////// 1120// Predicate operations 1121 1122def SVAND_B_Z : SInst<"svand[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_and_z", [IsStreamingCompatible]>; 1123def SVBIC_B_Z : SInst<"svbic[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_bic_z", [IsStreamingCompatible]>; 1124def SVEOR_B_Z : SInst<"sveor[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_eor_z", [IsStreamingCompatible]>; 1125def SVMOV_B_Z : SInst<"svmov[_b]_z", "PPP", "Pc", MergeNone, "", [IsStreamingCompatible]>; // Uses custom expansion 1126def SVNAND_B_Z : SInst<"svnand[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_nand_z", [IsStreamingCompatible]>; 1127def SVNOR_B_Z : SInst<"svnor[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_nor_z", [IsStreamingCompatible]>; 1128def SVNOT_B_Z : SInst<"svnot[_b]_z", "PPP", "Pc", MergeNone, "", [IsStreamingCompatible]>; // Uses custom expansion 1129def SVORN_B_Z : SInst<"svorn[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_orn_z", [IsStreamingCompatible]>; 1130def SVORR_B_Z : SInst<"svorr[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_orr_z", [IsStreamingCompatible]>; 1131 1132def SVBRKA : SInst<"svbrka[_b]_m", "PPPP", "Pc", MergeNone, "aarch64_sve_brka", [IsStreamingCompatible]>; 1133def SVBRKA_Z : SInst<"svbrka[_b]_z", "PPP", "Pc", MergeNone, "aarch64_sve_brka_z", [IsStreamingCompatible]>; 1134def SVBRKB : SInst<"svbrkb[_b]_m", "PPPP", "Pc", MergeNone, "aarch64_sve_brkb", [IsStreamingCompatible]>; 1135def SVBRKB_Z : SInst<"svbrkb[_b]_z", "PPP", "Pc", MergeNone, "aarch64_sve_brkb_z", [IsStreamingCompatible]>; 1136def SVBRKN_Z : SInst<"svbrkn[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_brkn_z", [IsStreamingCompatible]>; 1137def SVBRKPA_Z : SInst<"svbrkpa[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_brkpa_z", [IsStreamingCompatible]>; 1138def SVBRKPB_Z : SInst<"svbrkpb[_b]_z", "PPPP", "Pc", MergeNone, "aarch64_sve_brkpb_z", [IsStreamingCompatible]>; 1139 1140def SVPFIRST : SInst<"svpfirst[_b]", "PPP", "Pc", MergeNone, "aarch64_sve_pfirst", [IsStreamingCompatible]>; 1141def SVPNEXT : SInst<"svpnext_{d}", "PPP", "PcPsPiPl", MergeNone, "aarch64_sve_pnext", [IsStreamingCompatible]>; 1142 1143//////////////////////////////////////////////////////////////////////////////// 1144// Testing predicates 1145 1146def SVPTEST_ANY : SInst<"svptest_any", "sPP", "Pc", MergeNone, "aarch64_sve_ptest_any", [IsStreamingCompatible]>; 1147def SVPTEST_FIRST : SInst<"svptest_first", "sPP", "Pc", MergeNone, "aarch64_sve_ptest_first", [IsStreamingCompatible]>; 1148def SVPTEST_LAST : SInst<"svptest_last", "sPP", "Pc", MergeNone, "aarch64_sve_ptest_last", [IsStreamingCompatible]>; 1149 1150//////////////////////////////////////////////////////////////////////////////// 1151// FFR manipulation 1152 1153def SVRDFFR : SInst<"svrdffr", "Pv", "Pc", MergeNone, "", [IsOverloadNone]>; 1154def SVRDFFR_Z : SInst<"svrdffr_z", "PP", "Pc", MergeNone, "", [IsOverloadNone]>; 1155def SVSETFFR : SInst<"svsetffr", "vv", "", MergeNone, "", [IsOverloadNone]>; 1156def SVWRFFR : SInst<"svwrffr", "vP", "Pc", MergeNone, "", [IsOverloadNone]>; 1157 1158//////////////////////////////////////////////////////////////////////////////// 1159// Counting elements 1160 1161def SVCNTB_PAT : SInst<"svcntb_pat", "nI", "", MergeNone, "aarch64_sve_cntb", [IsOverloadNone, IsStreamingCompatible]>; 1162def SVCNTH_PAT : SInst<"svcnth_pat", "nI", "", MergeNone, "aarch64_sve_cnth", [IsOverloadNone, IsStreamingCompatible]>; 1163def SVCNTW_PAT : SInst<"svcntw_pat", "nI", "", MergeNone, "aarch64_sve_cntw", [IsOverloadNone, IsStreamingCompatible]>; 1164def SVCNTD_PAT : SInst<"svcntd_pat", "nI", "", MergeNone, "aarch64_sve_cntd", [IsOverloadNone, IsStreamingCompatible]>; 1165 1166def SVCNTB : SInst<"svcntb", "nv", "", MergeNone, "aarch64_sve_cntb", [IsAppendSVALL, IsOverloadNone, IsStreamingCompatible]>; 1167def SVCNTH : SInst<"svcnth", "nv", "", MergeNone, "aarch64_sve_cnth", [IsAppendSVALL, IsOverloadNone, IsStreamingCompatible]>; 1168def SVCNTW : SInst<"svcntw", "nv", "", MergeNone, "aarch64_sve_cntw", [IsAppendSVALL, IsOverloadNone, IsStreamingCompatible]>; 1169def SVCNTD : SInst<"svcntd", "nv", "", MergeNone, "aarch64_sve_cntd", [IsAppendSVALL, IsOverloadNone, IsStreamingCompatible]>; 1170 1171def SVCNTP : SInst<"svcntp_{d}", "nPP", "PcPsPiPl", MergeNone, "aarch64_sve_cntp", [IsStreamingCompatible]>; 1172def SVLEN : SInst<"svlen[_{d}]", "nd", "csilUcUsUiUlhfd", MergeNone, "", [IsStreamingCompatible]>; 1173 1174let TargetGuard = "sve,bf16" in { 1175def SVLEN_BF16 : SInst<"svlen[_{d}]", "nd", "b", MergeNone, "", [IsStreamingCompatible]>; 1176} 1177 1178//////////////////////////////////////////////////////////////////////////////// 1179// Saturating scalar arithmetic 1180 1181class sat_type<string u, string t> { string U = u; string T = t; } 1182def SignedByte : sat_type<"", "c">; 1183def SignedHalf : sat_type<"", "s">; 1184def SignedWord : sat_type<"", "i">; 1185def SignedDoubleWord : sat_type<"", "l">; 1186def UnsignedByte : sat_type<"U", "Uc">; 1187def UnsignedHalf : sat_type<"U", "Us">; 1188def UnsignedWord : sat_type<"U", "Ui">; 1189def UnsignedDoubleWord : sat_type<"U", "Ul">; 1190 1191multiclass SInst_SAT1<string name, string intrinsic, sat_type type> { 1192 def _N32 : SInst<name # "_pat[_n_{d}]", "ssIi", type.U # "i", MergeNone, intrinsic # "_n32", [IsOverloadNone, IsStreamingCompatible], [ImmCheck<2, ImmCheck1_16>]>; 1193 def _N64 : SInst<name # "_pat[_n_{d}]", "ssIi", type.U # "l", MergeNone, intrinsic # "_n64", [IsOverloadNone, IsStreamingCompatible], [ImmCheck<2, ImmCheck1_16>]>; 1194 def _N32_ALL : SInst<name # "[_n_{d}]", "ssi", type.U # "i", MergeNone, intrinsic # "_n32", [IsOverloadNone, IsInsertOp1SVALL, IsStreamingCompatible], [ImmCheck<1, ImmCheck1_16>]>; 1195 def _N64_ALL : SInst<name # "[_n_{d}]", "ssi", type.U # "l", MergeNone, intrinsic # "_n64", [IsOverloadNone, IsInsertOp1SVALL, IsStreamingCompatible], [ImmCheck<1, ImmCheck1_16>]>; 1196} 1197 1198multiclass SInst_SAT2<string name, string intrinsic, sat_type type> { 1199 def "" : SInst<name # "_pat[_{d}]", "ddIi", type.T, MergeNone, intrinsic, [IsStreamingCompatible], [ImmCheck<2, ImmCheck1_16>]>; 1200 def _ALL : SInst<name # "[_{d}]", "ddi", type.T, MergeNone, intrinsic, [IsInsertOp1SVALL, IsStreamingCompatible], [ImmCheck<1, ImmCheck1_16>]>; 1201 1202 def _N32 : SInst<name # "_pat[_n_{d}]", "ssIi", type.U # "i", MergeNone, intrinsic # "_n32", [IsOverloadNone, IsStreamingCompatible], [ImmCheck<2, ImmCheck1_16>]>; 1203 def _N64 : SInst<name # "_pat[_n_{d}]", "ssIi", type.U # "l", MergeNone, intrinsic # "_n64", [IsOverloadNone, IsStreamingCompatible], [ImmCheck<2, ImmCheck1_16>]>; 1204 def _N32_ALL : SInst<name # "[_n_{d}]", "ssi", type.U # "i", MergeNone, intrinsic # "_n32", [IsOverloadNone, IsInsertOp1SVALL, IsStreamingCompatible], [ImmCheck<1, ImmCheck1_16>]>; 1205 def _N64_ALL : SInst<name # "[_n_{d}]", "ssi", type.U # "l", MergeNone, intrinsic # "_n64", [IsOverloadNone, IsInsertOp1SVALL, IsStreamingCompatible], [ImmCheck<1, ImmCheck1_16>]>; 1206} 1207 1208defm SVQDECB_S : SInst_SAT1<"svqdecb", "aarch64_sve_sqdecb", SignedByte>; 1209defm SVQDECB_U : SInst_SAT1<"svqdecb", "aarch64_sve_uqdecb", UnsignedByte>; 1210defm SVQDECH_S : SInst_SAT2<"svqdech", "aarch64_sve_sqdech", SignedHalf>; 1211defm SVQDECH_U : SInst_SAT2<"svqdech", "aarch64_sve_uqdech", UnsignedHalf>; 1212defm SVQDECW_S : SInst_SAT2<"svqdecw", "aarch64_sve_sqdecw", SignedWord>; 1213defm SVQDECW_U : SInst_SAT2<"svqdecw", "aarch64_sve_uqdecw", UnsignedWord>; 1214defm SVQDECD_S : SInst_SAT2<"svqdecd", "aarch64_sve_sqdecd", SignedDoubleWord>; 1215defm SVQDECD_U : SInst_SAT2<"svqdecd", "aarch64_sve_uqdecd", UnsignedDoubleWord>; 1216 1217defm SVQINCB_S : SInst_SAT1<"svqincb", "aarch64_sve_sqincb", SignedByte>; 1218defm SVQINCB_U : SInst_SAT1<"svqincb", "aarch64_sve_uqincb", UnsignedByte>; 1219defm SVQINCH_S : SInst_SAT2<"svqinch", "aarch64_sve_sqinch", SignedHalf>; 1220defm SVQINCH_U : SInst_SAT2<"svqinch", "aarch64_sve_uqinch", UnsignedHalf>; 1221defm SVQINCW_S : SInst_SAT2<"svqincw", "aarch64_sve_sqincw", SignedWord>; 1222defm SVQINCW_U : SInst_SAT2<"svqincw", "aarch64_sve_uqincw", UnsignedWord>; 1223defm SVQINCD_S : SInst_SAT2<"svqincd", "aarch64_sve_sqincd", SignedDoubleWord>; 1224defm SVQINCD_U : SInst_SAT2<"svqincd", "aarch64_sve_uqincd", UnsignedDoubleWord>; 1225 1226def SVQDECP_S : SInst<"svqdecp[_{d}]", "ddP", "sil", MergeNone, "aarch64_sve_sqdecp", [IsStreamingCompatible]>; 1227def SVQDECP_U : SInst<"svqdecp[_{d}]", "ddP", "UsUiUl", MergeNone, "aarch64_sve_uqdecp", [IsStreamingCompatible]>; 1228def SVQINCP_S : SInst<"svqincp[_{d}]", "ddP", "sil", MergeNone, "aarch64_sve_sqincp", [IsStreamingCompatible]>; 1229def SVQINCP_U : SInst<"svqincp[_{d}]", "ddP", "UsUiUl", MergeNone, "aarch64_sve_uqincp", [IsStreamingCompatible]>; 1230 1231def SVQDECP_N_S32 : SInst<"svqdecp[_n_s32]_{d}", "kkP", "PcPsPiPl", MergeNone, "aarch64_sve_sqdecp_n32", [IsStreamingCompatible]>; 1232def SVQDECP_N_S64 : SInst<"svqdecp[_n_s64]_{d}", "llP", "PcPsPiPl", MergeNone, "aarch64_sve_sqdecp_n64", [IsStreamingCompatible]>; 1233def SVQDECP_N_U32 : SInst<"svqdecp[_n_u32]_{d}", "mmP", "PcPsPiPl", MergeNone, "aarch64_sve_uqdecp_n32", [IsStreamingCompatible]>; 1234def SVQDECP_N_U64 : SInst<"svqdecp[_n_u64]_{d}", "nnP", "PcPsPiPl", MergeNone, "aarch64_sve_uqdecp_n64", [IsStreamingCompatible]>; 1235def SVQINCP_N_S32 : SInst<"svqincp[_n_s32]_{d}", "kkP", "PcPsPiPl", MergeNone, "aarch64_sve_sqincp_n32", [IsStreamingCompatible]>; 1236def SVQINCP_N_S64 : SInst<"svqincp[_n_s64]_{d}", "llP", "PcPsPiPl", MergeNone, "aarch64_sve_sqincp_n64", [IsStreamingCompatible]>; 1237def SVQINCP_N_U32 : SInst<"svqincp[_n_u32]_{d}", "mmP", "PcPsPiPl", MergeNone, "aarch64_sve_uqincp_n32", [IsStreamingCompatible]>; 1238def SVQINCP_N_U64 : SInst<"svqincp[_n_u64]_{d}", "nnP", "PcPsPiPl", MergeNone, "aarch64_sve_uqincp_n64", [IsStreamingCompatible]>; 1239 1240let TargetGuard = "sve,i8mm" in { 1241def SVMLLA_S32 : SInst<"svmmla[_s32]", "ddqq","i", MergeNone, "aarch64_sve_smmla">; 1242def SVMLLA_U32 : SInst<"svmmla[_u32]", "ddqq","Ui", MergeNone, "aarch64_sve_ummla">; 1243def SVUSMLLA_S32 : SInst<"svusmmla[_s32]", "ddbq","i", MergeNone, "aarch64_sve_usmmla">; 1244 1245def SVUSDOT_S : SInst<"svusdot[_s32]", "ddbq", "i", MergeNone, "aarch64_sve_usdot", [IsStreamingCompatible]>; 1246def SVUSDOT_N_S : SInst<"svusdot[_n_s32]", "ddbr", "i", MergeNone, "aarch64_sve_usdot", [IsStreamingCompatible]>; 1247def SVSUDOT_S : SInst<"svsudot[_s32]", "ddqb", "i", MergeNone, "aarch64_sve_usdot", [ReverseUSDOT, IsStreamingCompatible]>; 1248def SVSUDOT_N_S : SInst<"svsudot[_n_s32]", "ddq@", "i", MergeNone, "aarch64_sve_usdot", [ReverseUSDOT, IsStreamingCompatible]>; 1249 1250def SVUSDOT_LANE_S : SInst<"svusdot_lane[_s32]", "ddbqi", "i", MergeNone, "aarch64_sve_usdot_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndexDot, 2>]>; 1251def SVSUDOT_LANE_S : SInst<"svsudot_lane[_s32]", "ddqbi", "i", MergeNone, "aarch64_sve_sudot_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndexDot, 2>]>; 1252} 1253 1254let TargetGuard = "sve,f32mm" in { 1255def SVMLLA_F32 : SInst<"svmmla[_f32]", "dddd","f", MergeNone, "aarch64_sve_fmmla">; 1256} 1257 1258let TargetGuard = "sve,f64mm" in { 1259def SVMLLA_F64 : SInst<"svmmla[_f64]", "dddd","d", MergeNone, "aarch64_sve_fmmla">; 1260def SVTRN1Q : SInst<"svtrn1q[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_trn1q", [IsStreamingCompatible]>; 1261def SVTRN2Q : SInst<"svtrn2q[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_trn2q", [IsStreamingCompatible]>; 1262def SVUZP1Q : SInst<"svuzp1q[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_uzp1q", [IsStreamingCompatible]>; 1263def SVUZP2Q : SInst<"svuzp2q[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_uzp2q", [IsStreamingCompatible]>; 1264def SVZIP1Q : SInst<"svzip1q[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_zip1q", [IsStreamingCompatible]>; 1265def SVZIP2Q : SInst<"svzip2q[_{d}]", "ddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_zip2q", [IsStreamingCompatible]>; 1266} 1267 1268let TargetGuard = "sve,bf16,f64mm" in { 1269def SVTRN1Q_BF16 : SInst<"svtrn1q[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_trn1q">; 1270def SVTRN2Q_BF16 : SInst<"svtrn2q[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_trn2q">; 1271def SVUZP1Q_BF16 : SInst<"svuzp1q[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_uzp1q", [IsStreamingCompatible]>; 1272def SVUZP2Q_BF16 : SInst<"svuzp2q[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_uzp2q", [IsStreamingCompatible]>; 1273def SVZIP1Q_BF16 : SInst<"svzip1q[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_zip1q", [IsStreamingCompatible]>; 1274def SVZIP2Q_BF16 : SInst<"svzip2q[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_zip2q", [IsStreamingCompatible]>; 1275} 1276 1277//////////////////////////////////////////////////////////////////////////////// 1278// Vector creation 1279def SVUNDEF_1 : SInst<"svundef_{d}", "dv", "csilUcUsUiUlhfd", MergeNone, "", [IsUndef, IsStreamingCompatible]>; 1280def SVUNDEF_2 : SInst<"svundef2_{d}", "2v", "csilUcUsUiUlhfd", MergeNone, "", [IsUndef, IsStreamingCompatible]>; 1281def SVUNDEF_3 : SInst<"svundef3_{d}", "3v", "csilUcUsUiUlhfd", MergeNone, "", [IsUndef, IsStreamingCompatible]>; 1282def SVUNDEF_4 : SInst<"svundef4_{d}", "4v", "csilUcUsUiUlhfd", MergeNone, "", [IsUndef, IsStreamingCompatible]>; 1283 1284def SVCREATE_2 : SInst<"svcreate2[_{d}]", "2dd", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleCreate, IsStreamingCompatible]>; 1285def SVCREATE_3 : SInst<"svcreate3[_{d}]", "3ddd", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleCreate, IsStreamingCompatible]>; 1286def SVCREATE_4 : SInst<"svcreate4[_{d}]", "4dddd", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleCreate, IsStreamingCompatible]>; 1287 1288let TargetGuard = "sve,bf16" in { 1289def SVUNDEF_1_BF16 : SInst<"svundef_{d}", "dv", "b", MergeNone, "", [IsUndef, IsStreamingCompatible]>; 1290def SVUNDEF_2_BF16 : SInst<"svundef2_{d}", "2v", "b", MergeNone, "", [IsUndef, IsStreamingCompatible]>; 1291def SVUNDEF_3_BF16 : SInst<"svundef3_{d}", "3v", "b", MergeNone, "", [IsUndef, IsStreamingCompatible]>; 1292def SVUNDEF_4_BF16 : SInst<"svundef4_{d}", "4v", "b", MergeNone, "", [IsUndef, IsStreamingCompatible]>; 1293 1294def SVCREATE_2_BF16 : SInst<"svcreate2[_{d}]", "2dd", "b", MergeNone, "", [IsTupleCreate, IsStreamingCompatible]>; 1295def SVCREATE_3_BF16 : SInst<"svcreate3[_{d}]", "3ddd", "b", MergeNone, "", [IsTupleCreate, IsStreamingCompatible]>; 1296def SVCREATE_4_BF16 : SInst<"svcreate4[_{d}]", "4dddd", "b", MergeNone, "", [IsTupleCreate, IsStreamingCompatible]>; 1297} 1298 1299let TargetGuard = "sve2p1|sme2" in { 1300 def SVCREATE_2_B : SInst<"svcreate2[_b]", "2dd", "Pc", MergeNone, "", [IsTupleCreate, IsStreamingCompatible]>; 1301 def SVCREATE_4_B : SInst<"svcreate4[_b]", "4dddd", "Pc", MergeNone, "", [IsTupleCreate, IsStreamingCompatible]>; 1302} 1303 1304//////////////////////////////////////////////////////////////////////////////// 1305// Vector insertion and extraction 1306def SVGET_2 : SInst<"svget2[_{d}]", "d2i", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleGet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_1>]>; 1307def SVGET_3 : SInst<"svget3[_{d}]", "d3i", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleGet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_2>]>; 1308def SVGET_4 : SInst<"svget4[_{d}]", "d4i", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleGet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_3>]>; 1309 1310def SVSET_2 : SInst<"svset2[_{d}]", "22id", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleSet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_1>]>; 1311def SVSET_3 : SInst<"svset3[_{d}]", "33id", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleSet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_2>]>; 1312def SVSET_4 : SInst<"svset4[_{d}]", "44id", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleSet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_3>]>; 1313 1314let TargetGuard = "sve,bf16" in { 1315def SVGET_2_BF16 : SInst<"svget2[_{d}]", "d2i", "b", MergeNone, "", [IsTupleGet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_1>]>; 1316def SVGET_3_BF16 : SInst<"svget3[_{d}]", "d3i", "b", MergeNone, "", [IsTupleGet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_2>]>; 1317def SVGET_4_BF16 : SInst<"svget4[_{d}]", "d4i", "b", MergeNone, "", [IsTupleGet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_3>]>; 1318 1319def SVSET_2_BF16 : SInst<"svset2[_{d}]", "22id", "b", MergeNone, "", [IsTupleSet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_1>]>; 1320def SVSET_3_BF16 : SInst<"svset3[_{d}]", "33id", "b", MergeNone, "", [IsTupleSet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_2>]>; 1321def SVSET_4_BF16 : SInst<"svset4[_{d}]", "44id", "b", MergeNone, "", [IsTupleSet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_3>]>; 1322} 1323 1324let TargetGuard = "sve2p1|sme2" in { 1325 def SVGET_2_B : SInst<"svget2[_b]", "d2i", "Pc", MergeNone, "", [IsTupleGet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_1>]>; 1326 def SVGET_4_B : SInst<"svget4[_b]", "d4i", "Pc", MergeNone, "", [IsTupleGet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_3>]>; 1327 1328 def SVSET_2_B : SInst<"svset2[_b]", "22id", "Pc", MergeNone, "", [IsTupleSet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_1>]>; 1329 def SVSET_4_B : SInst<"svset4[_b]", "44id", "Pc", MergeNone, "", [IsTupleSet, IsStreamingCompatible], [ImmCheck<1, ImmCheck0_3>]>; 1330} 1331 1332let TargetGuard = "sve2p1|sme2" in { 1333 def SVUNDEF_2_B: Inst<"svundef2_b", "2", "Pc", MergeNone, "", [IsUndef, IsStreamingCompatible], []>; 1334 def SVUNDEF_4_B: Inst<"svundef4_b", "4", "Pc", MergeNone, "", [IsUndef, IsStreamingCompatible], []>; 1335} 1336//////////////////////////////////////////////////////////////////////////////// 1337// SVE2 WhileGE/GT 1338let TargetGuard = "sve2" in { 1339def SVWHILEGE_S32 : SInst<"svwhilege_{d}[_{1}]", "Pkk", "PcPsPiPl", MergeNone, "aarch64_sve_whilege", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 1340def SVWHILEGE_S64 : SInst<"svwhilege_{d}[_{1}]", "Pll", "PcPsPiPl", MergeNone, "aarch64_sve_whilege", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 1341def SVWHILEGT_S32 : SInst<"svwhilegt_{d}[_{1}]", "Pkk", "PcPsPiPl", MergeNone, "aarch64_sve_whilegt", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 1342def SVWHILEGT_S64 : SInst<"svwhilegt_{d}[_{1}]", "Pll", "PcPsPiPl", MergeNone, "aarch64_sve_whilegt", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 1343def SVWHILEHI_U32 : SInst<"svwhilegt_{d}[_{1}]", "Pmm", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilehi", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 1344def SVWHILEHI_U64 : SInst<"svwhilegt_{d}[_{1}]", "Pnn", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilehi", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 1345def SVWHILEHS_U32 : SInst<"svwhilege_{d}[_{1}]", "Pmm", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilehs", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 1346def SVWHILEHS_U64 : SInst<"svwhilege_{d}[_{1}]", "Pnn", "PUcPUsPUiPUl", MergeNone, "aarch64_sve_whilehs", [IsOverloadWhileOrMultiVecCvt, IsStreamingCompatible]>; 1347} 1348 1349let TargetGuard = "sve2p1|sme2" in { 1350 def SVWHILEGE_S64_X2 : SInst<"svwhilege_{d}[_{1}]_x2", "2ll", "PcPsPiPl", MergeNone, "aarch64_sve_whilege_x2", [IsStreamingOrSVE2p1]>; 1351 def SVWHILEGT_S64_X2 : SInst<"svwhilegt_{d}[_{1}]_x2", "2ll", "PcPsPiPl", MergeNone, "aarch64_sve_whilegt_x2", [IsStreamingOrSVE2p1]>; 1352 def SVWHILEHI_U64_X2 : SInst<"svwhilegt_{d}[_{1}]_x2", "2nn", "PcPsPiPl", MergeNone, "aarch64_sve_whilehi_x2", [IsStreamingOrSVE2p1]>; 1353 def SVWHILEHS_U64_X2 : SInst<"svwhilege_{d}[_{1}]_x2", "2nn", "PcPsPiPl", MergeNone, "aarch64_sve_whilehs_x2", [IsStreamingOrSVE2p1]>; 1354 def SVWHILELE_S64_X2 : SInst<"svwhilele_{d}[_{1}]_x2", "2ll", "PcPsPiPl", MergeNone, "aarch64_sve_whilele_x2", [IsStreamingOrSVE2p1]>; 1355 def SVWHILELT_S64_X2 : SInst<"svwhilelt_{d}[_{1}]_x2", "2ll", "PcPsPiPl", MergeNone, "aarch64_sve_whilelt_x2", [IsStreamingOrSVE2p1]>; 1356 def SVWHILELO_U64_X2 : SInst<"svwhilelt_{d}[_{1}]_x2", "2nn", "PcPsPiPl", MergeNone, "aarch64_sve_whilelo_x2", [IsStreamingOrSVE2p1]>; 1357 def SVWHILELS_U64_X2 : SInst<"svwhilele_{d}[_{1}]_x2", "2nn", "PcPsPiPl", MergeNone, "aarch64_sve_whilels_x2", [IsStreamingOrSVE2p1]>; 1358 1359} 1360 1361//////////////////////////////////////////////////////////////////////////////// 1362// SVE2 - Uniform DSP operations 1363 1364let TargetGuard = "sve2" in { 1365defm SVQADD_S : SInstZPZZ<"svqadd", "csli", "aarch64_sve_sqadd", "aarch64_sve_sqadd">; 1366defm SVQADD_U : SInstZPZZ<"svqadd", "UcUsUiUl", "aarch64_sve_uqadd", "aarch64_sve_uqadd">; 1367defm SVHADD_S : SInstZPZZ<"svhadd", "csli", "aarch64_sve_shadd", "aarch64_sve_shadd">; 1368defm SVHADD_U : SInstZPZZ<"svhadd", "UcUsUiUl", "aarch64_sve_uhadd", "aarch64_sve_uhadd">; 1369defm SVRHADD_S : SInstZPZZ<"svrhadd", "csli", "aarch64_sve_srhadd", "aarch64_sve_srhadd">; 1370defm SVRHADD_U : SInstZPZZ<"svrhadd", "UcUsUiUl", "aarch64_sve_urhadd", "aarch64_sve_urhadd">; 1371 1372defm SVQSUB_S : SInstZPZZ<"svqsub", "csli", "aarch64_sve_sqsub", "aarch64_sve_sqsub_u">; 1373defm SVQSUB_U : SInstZPZZ<"svqsub", "UcUsUiUl", "aarch64_sve_uqsub", "aarch64_sve_uqsub_u">; 1374defm SVQSUBR_S : SInstZPZZ<"svqsubr", "csli", "aarch64_sve_sqsubr", "aarch64_sve_sqsub_u", [ReverseMergeAnyBinOp]>; 1375defm SVQSUBR_U : SInstZPZZ<"svqsubr", "UcUsUiUl", "aarch64_sve_uqsubr", "aarch64_sve_uqsub_u", [ReverseMergeAnyBinOp]>; 1376defm SVHSUB_S : SInstZPZZ<"svhsub", "csli", "aarch64_sve_shsub", "aarch64_sve_shsub">; 1377defm SVHSUB_U : SInstZPZZ<"svhsub", "UcUsUiUl", "aarch64_sve_uhsub", "aarch64_sve_uhsub">; 1378defm SVHSUBR_S : SInstZPZZ<"svhsubr", "csli", "aarch64_sve_shsubr", "aarch64_sve_shsubr">; 1379defm SVHSUBR_U : SInstZPZZ<"svhsubr", "UcUsUiUl", "aarch64_sve_uhsubr", "aarch64_sve_uhsubr">; 1380 1381defm SVQABS : SInstZPZ<"svqabs", "csil", "aarch64_sve_sqabs">; 1382defm SVQNEG : SInstZPZ<"svqneg", "csil", "aarch64_sve_sqneg">; 1383defm SVRECPE : SInstZPZ<"svrecpe", "Ui", "aarch64_sve_urecpe">; 1384defm SVRSQRTE : SInstZPZ<"svrsqrte", "Ui", "aarch64_sve_ursqrte">; 1385} 1386 1387//------------------------------------------------------------------------------ 1388 1389multiclass SInstZPZxZ<string name, string types, string pat_v, string pat_n, string intrinsic, list<FlagType> flags=[]> { 1390 def _M : SInst<name # "[_{d}]", pat_v, types, MergeOp1, intrinsic, flags>; 1391 def _X : SInst<name # "[_{d}]", pat_v, types, MergeAny, intrinsic, flags>; 1392 def _Z : SInst<name # "[_{d}]", pat_v, types, MergeZero, intrinsic, flags>; 1393 1394 def _N_M : SInst<name # "[_n_{d}]", pat_n, types, MergeOp1, intrinsic, flags>; 1395 def _N_X : SInst<name # "[_n_{d}]", pat_n, types, MergeAny, intrinsic, flags>; 1396 def _N_Z : SInst<name # "[_n_{d}]", pat_n, types, MergeZero, intrinsic, flags>; 1397} 1398 1399let TargetGuard = "sve2" in { 1400defm SVQRSHL_S : SInstZPZxZ<"svqrshl", "csil", "dPdx", "dPdK", "aarch64_sve_sqrshl", [IsStreamingCompatible]>; 1401defm SVQRSHL_U : SInstZPZxZ<"svqrshl", "UcUsUiUl", "dPdx", "dPdK", "aarch64_sve_uqrshl", [IsStreamingCompatible]>; 1402defm SVQSHL_S : SInstZPZxZ<"svqshl", "csil", "dPdx", "dPdK", "aarch64_sve_sqshl", [IsStreamingCompatible]>; 1403defm SVQSHL_U : SInstZPZxZ<"svqshl", "UcUsUiUl", "dPdx", "dPdK", "aarch64_sve_uqshl", [IsStreamingCompatible]>; 1404defm SVRSHL_S : SInstZPZxZ<"svrshl", "csil", "dPdx", "dPdK", "aarch64_sve_srshl", [IsStreamingCompatible]>; 1405defm SVRSHL_U : SInstZPZxZ<"svrshl", "UcUsUiUl", "dPdx", "dPdK", "aarch64_sve_urshl", [IsStreamingCompatible]>; 1406defm SVSQADD : SInstZPZxZ<"svsqadd", "UcUsUiUl", "dPdx", "dPdK", "aarch64_sve_usqadd", [IsStreamingCompatible]>; 1407defm SVUQADD : SInstZPZxZ<"svuqadd", "csil", "dPdu", "dPdL", "aarch64_sve_suqadd", [IsStreamingCompatible]>; 1408 1409def SVABA_S : SInst<"svaba[_{d}]", "dddd", "csil" , MergeNone, "aarch64_sve_saba", [IsStreamingCompatible]>; 1410def SVABA_U : SInst<"svaba[_{d}]", "dddd", "UcUsUiUl", MergeNone, "aarch64_sve_uaba", [IsStreamingCompatible]>; 1411def SVQDMULH : SInst<"svqdmulh[_{d}]", "ddd", "csil", MergeNone, "aarch64_sve_sqdmulh", [IsStreamingCompatible]>; 1412def SVQRDMULH : SInst<"svqrdmulh[_{d}]", "ddd", "csil", MergeNone, "aarch64_sve_sqrdmulh", [IsStreamingCompatible]>; 1413def SVQRDMLAH : SInst<"svqrdmlah[_{d}]", "dddd", "csil", MergeNone, "aarch64_sve_sqrdmlah", [IsStreamingCompatible]>; 1414def SVQRDMLSH : SInst<"svqrdmlsh[_{d}]", "dddd", "csil", MergeNone, "aarch64_sve_sqrdmlsh", [IsStreamingCompatible]>; 1415 1416def SVABA_S_N : SInst<"svaba[_n_{d}]", "ddda", "csil", MergeNone, "aarch64_sve_saba", [IsStreamingCompatible]>; 1417def SVABA_U_N : SInst<"svaba[_n_{d}]", "ddda", "UcUsUiUl", MergeNone, "aarch64_sve_uaba", [IsStreamingCompatible]>; 1418def SVQDMULH_N : SInst<"svqdmulh[_n_{d}]", "dda", "csil", MergeNone, "aarch64_sve_sqdmulh", [IsStreamingCompatible]>; 1419def SVQRDMULH_N : SInst<"svqrdmulh[_n_{d}]", "dda", "csil", MergeNone, "aarch64_sve_sqrdmulh", [IsStreamingCompatible]>; 1420def SVQRDMLAH_N : SInst<"svqrdmlah[_n_{d}]", "ddda", "csil", MergeNone, "aarch64_sve_sqrdmlah", [IsStreamingCompatible]>; 1421def SVQRDMLSH_N : SInst<"svqrdmlsh[_n_{d}]", "ddda", "csil", MergeNone, "aarch64_sve_sqrdmlsh", [IsStreamingCompatible]>; 1422 1423def SVQDMULH_LANE : SInst<"svqdmulh_lane[_{d}]", "dddi", "sil", MergeNone, "aarch64_sve_sqdmulh_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1424def SVQRDMULH_LANE : SInst<"svqrdmulh_lane[_{d}]", "dddi", "sil", MergeNone, "aarch64_sve_sqrdmulh_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1425def SVQRDMLAH_LANE : SInst<"svqrdmlah_lane[_{d}]", "ddddi", "sil", MergeNone, "aarch64_sve_sqrdmlah_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1426def SVQRDMLSH_LANE : SInst<"svqrdmlsh_lane[_{d}]", "ddddi", "sil", MergeNone, "aarch64_sve_sqrdmlsh_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1427 1428def SVQSHLU_M : SInst<"svqshlu[_n_{d}]", "uPdi", "csil", MergeOp1, "aarch64_sve_sqshlu", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftLeft, 1>]>; 1429def SVQSHLU_X : SInst<"svqshlu[_n_{d}]", "uPdi", "csil", MergeAny, "aarch64_sve_sqshlu", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftLeft, 1>]>; 1430def SVQSHLU_Z : SInst<"svqshlu[_n_{d}]", "uPdi", "csil", MergeZero, "aarch64_sve_sqshlu", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftLeft, 1>]>; 1431def SVRSHR_M_S : SInst<"svrshr[_n_{d}]", "dPdi", "csil", MergeOp1, "aarch64_sve_srshr", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1432def SVRSHR_M_U : SInst<"svrshr[_n_{d}]", "dPdi", "UcUsUiUl", MergeOp1, "aarch64_sve_urshr", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1433def SVRSHR_X_S : SInst<"svrshr[_n_{d}]", "dPdi", "csil", MergeAny, "aarch64_sve_srshr", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1434def SVRSHR_X_U : SInst<"svrshr[_n_{d}]", "dPdi", "UcUsUiUl", MergeAny, "aarch64_sve_urshr", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1435def SVRSHR_Z_S : SInst<"svrshr[_n_{d}]", "dPdi", "csil", MergeZero, "aarch64_sve_srshr", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1436def SVRSHR_Z_U : SInst<"svrshr[_n_{d}]", "dPdi", "UcUsUiUl", MergeZero, "aarch64_sve_urshr", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1437def SVRSRA_S : SInst<"svrsra[_n_{d}]", "dddi", "csil", MergeNone, "aarch64_sve_srsra", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1438def SVRSRA_U : SInst<"svrsra[_n_{d}]", "dddi", "UcUsUiUl", MergeNone, "aarch64_sve_ursra", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1439def SVSLI : SInst<"svsli[_n_{d}]", "dddi", "csilUcUsUiUl", MergeNone, "aarch64_sve_sli", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftLeft, 1>]>; 1440def SVSRA_S : SInst<"svsra[_n_{d}]", "dddi", "csil", MergeNone, "aarch64_sve_ssra", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1441def SVSRA_U : SInst<"svsra[_n_{d}]", "dddi", "UcUsUiUl", MergeNone, "aarch64_sve_usra", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1442def SVSRI : SInst<"svsri[_n_{d}]", "dddi", "csilUcUsUiUl", MergeNone, "aarch64_sve_sri", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1443} 1444 1445//////////////////////////////////////////////////////////////////////////////// 1446// SVE2 - Non-widening pairwise arithmetic 1447 1448multiclass SInstPairwise<string name, string types, string intrinsic, list<FlagType> flags=[]> { 1449 def _M : SInst<name # "[_{d}]", "dPdd", types, MergeOp1, intrinsic, flags>; 1450 def _X : SInst<name # "[_{d}]", "dPdd", types, MergeAny, intrinsic, flags>; 1451} 1452 1453let TargetGuard = "sve2" in { 1454defm SVADDP : SInstPairwise<"svaddp", "csliUcUsUiUl", "aarch64_sve_addp", [IsStreamingCompatible]>; 1455defm SVADDP_F : SInstPairwise<"svaddp", "hfd", "aarch64_sve_faddp", [IsStreamingCompatible]>; 1456defm SVMAXNMP : SInstPairwise<"svmaxnmp", "hfd", "aarch64_sve_fmaxnmp", [IsStreamingCompatible]>; 1457defm SVMAXP_F : SInstPairwise<"svmaxp", "hfd", "aarch64_sve_fmaxp", [IsStreamingCompatible]>; 1458defm SVMAXP_S : SInstPairwise<"svmaxp", "csli", "aarch64_sve_smaxp", [IsStreamingCompatible]>; 1459defm SVMAXP_U : SInstPairwise<"svmaxp", "UcUsUiUl", "aarch64_sve_umaxp", [IsStreamingCompatible]>; 1460defm SVMINNMP : SInstPairwise<"svminnmp", "hfd", "aarch64_sve_fminnmp", [IsStreamingCompatible]>; 1461defm SVMINP_F : SInstPairwise<"svminp", "hfd", "aarch64_sve_fminp", [IsStreamingCompatible]>; 1462defm SVMINP_S : SInstPairwise<"svminp", "csli", "aarch64_sve_sminp", [IsStreamingCompatible]>; 1463defm SVMINP_U : SInstPairwise<"svminp", "UcUsUiUl", "aarch64_sve_uminp", [IsStreamingCompatible]>; 1464} 1465 1466//////////////////////////////////////////////////////////////////////////////// 1467// SVE2 - Widening pairwise arithmetic 1468 1469let TargetGuard = "sve2" in { 1470def SVADALP_S_M : SInst<"svadalp[_{d}]", "dPdh", "sil", MergeOp1, "aarch64_sve_sadalp", [IsStreamingCompatible]>; 1471def SVADALP_S_X : SInst<"svadalp[_{d}]", "dPdh", "sil", MergeAny, "aarch64_sve_sadalp", [IsStreamingCompatible]>; 1472def SVADALP_S_Z : SInst<"svadalp[_{d}]", "dPdh", "sil", MergeZero, "aarch64_sve_sadalp", [IsStreamingCompatible]>; 1473 1474def SVADALP_U_M : SInst<"svadalp[_{d}]", "dPdh", "UsUiUl", MergeOp1, "aarch64_sve_uadalp", [IsStreamingCompatible]>; 1475def SVADALP_U_X : SInst<"svadalp[_{d}]", "dPdh", "UsUiUl", MergeAny, "aarch64_sve_uadalp", [IsStreamingCompatible]>; 1476def SVADALP_U_Z : SInst<"svadalp[_{d}]", "dPdh", "UsUiUl", MergeZero, "aarch64_sve_uadalp", [IsStreamingCompatible]>; 1477} 1478 1479//////////////////////////////////////////////////////////////////////////////// 1480// SVE2 - Bitwise ternary logical instructions 1481// 1482 1483let TargetGuard = "sve2" in { 1484def SVBCAX : SInst<"svbcax[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_bcax", [IsStreamingCompatible]>; 1485def SVBSL : SInst<"svbsl[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_bsl", [IsStreamingCompatible]>; 1486def SVBSL1N : SInst<"svbsl1n[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_bsl1n", [IsStreamingCompatible]>; 1487def SVBSL2N : SInst<"svbsl2n[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_bsl2n", [IsStreamingCompatible]>; 1488def SVEOR3 : SInst<"sveor3[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_eor3", [IsStreamingCompatible]>; 1489def SVNBSL : SInst<"svnbsl[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_nbsl", [IsStreamingCompatible]>; 1490 1491def SVBCAX_N : SInst<"svbcax[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_bcax", [IsStreamingCompatible]>; 1492def SVBSL_N : SInst<"svbsl[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_bsl", [IsStreamingCompatible]>; 1493def SVBSL1N_N : SInst<"svbsl1n[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_bsl1n", [IsStreamingCompatible]>; 1494def SVBSL2N_N : SInst<"svbsl2n[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_bsl2n", [IsStreamingCompatible]>; 1495def SVEOR3_N : SInst<"sveor3[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_eor3", [IsStreamingCompatible]>; 1496def SVNBSL_N : SInst<"svnbsl[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_nbsl", [IsStreamingCompatible]>; 1497def SVXAR_N : SInst<"svxar[_n_{d}]", "dddi", "csilUcUsUiUl", MergeNone, "aarch64_sve_xar", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRight, 1>]>; 1498} 1499 1500//////////////////////////////////////////////////////////////////////////////// 1501// SVE2 - Large integer arithmetic 1502 1503let TargetGuard = "sve2" in { 1504def SVADCLB : SInst<"svadclb[_{d}]", "dddd", "UiUl", MergeNone, "aarch64_sve_adclb", [IsStreamingCompatible]>; 1505def SVADCLT : SInst<"svadclt[_{d}]", "dddd", "UiUl", MergeNone, "aarch64_sve_adclt", [IsStreamingCompatible]>; 1506def SVSBCLB : SInst<"svsbclb[_{d}]", "dddd", "UiUl", MergeNone, "aarch64_sve_sbclb", [IsStreamingCompatible]>; 1507def SVSBCLT : SInst<"svsbclt[_{d}]", "dddd", "UiUl", MergeNone, "aarch64_sve_sbclt", [IsStreamingCompatible]>; 1508 1509def SVADCLB_N : SInst<"svadclb[_n_{d}]", "ddda", "UiUl", MergeNone, "aarch64_sve_adclb", [IsStreamingCompatible]>; 1510def SVADCLT_N : SInst<"svadclt[_n_{d}]", "ddda", "UiUl", MergeNone, "aarch64_sve_adclt", [IsStreamingCompatible]>; 1511def SVSBCLB_N : SInst<"svsbclb[_n_{d}]", "ddda", "UiUl", MergeNone, "aarch64_sve_sbclb", [IsStreamingCompatible]>; 1512def SVSBCLT_N : SInst<"svsbclt[_n_{d}]", "ddda", "UiUl", MergeNone, "aarch64_sve_sbclt", [IsStreamingCompatible]>; 1513} 1514 1515//////////////////////////////////////////////////////////////////////////////// 1516// SVE2 - Multiplication by indexed elements 1517 1518let TargetGuard = "sve2" in { 1519def SVMLA_LANE_2 : SInst<"svmla_lane[_{d}]", "ddddi", "silUsUiUl", MergeNone, "aarch64_sve_mla_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1520def SVMLS_LANE_2 : SInst<"svmls_lane[_{d}]", "ddddi", "silUsUiUl", MergeNone, "aarch64_sve_mls_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1521def SVMUL_LANE_2 : SInst<"svmul_lane[_{d}]", "dddi", "silUsUiUl", MergeNone, "aarch64_sve_mul_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1522} 1523 1524//////////////////////////////////////////////////////////////////////////////// 1525// SVE2 - Uniform complex integer arithmetic 1526let TargetGuard = "sve2" in { 1527def SVCADD : SInst<"svcadd[_{d}]", "dddi", "csilUcUsUiUl", MergeNone, "aarch64_sve_cadd_x", [IsStreamingCompatible], [ImmCheck<2, ImmCheckComplexRot90_270>]>; 1528def SVSQCADD : SInst<"svqcadd[_{d}]", "dddi", "csil", MergeNone, "aarch64_sve_sqcadd_x", [IsStreamingCompatible], [ImmCheck<2, ImmCheckComplexRot90_270>]>; 1529def SVCMLA : SInst<"svcmla[_{d}]", "ddddi", "csilUcUsUiUl", MergeNone, "aarch64_sve_cmla_x", [IsStreamingCompatible], [ImmCheck<3, ImmCheckComplexRotAll90>]>; 1530def SVCMLA_LANE_X : SInst<"svcmla_lane[_{d}]", "ddddii", "siUsUi", MergeNone, "aarch64_sve_cmla_lane_x", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndexCompRotate, 2>, 1531 ImmCheck<4, ImmCheckComplexRotAll90>]>; 1532def SVSQRDCMLAH_X : SInst<"svqrdcmlah[_{d}]", "ddddi", "csil", MergeNone, "aarch64_sve_sqrdcmlah_x", [IsStreamingCompatible], [ImmCheck<3, ImmCheckComplexRotAll90>]>; 1533def SVSQRDCMLAH_LANE_X : SInst<"svqrdcmlah_lane[_{d}]", "ddddii", "si", MergeNone, "aarch64_sve_sqrdcmlah_lane_x", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndexCompRotate, 2>, 1534 ImmCheck<4, ImmCheckComplexRotAll90>]>; 1535} 1536 1537//////////////////////////////////////////////////////////////////////////////// 1538// SVE2 - Widening DSP operations 1539 1540multiclass SInstWideDSPAcc<string name, string types, string intrinsic> { 1541 def : SInst<name # "[_{d}]", "ddhh", types, MergeNone, intrinsic, [IsStreamingCompatible]>; 1542 def _N : SInst<name # "[_n_{d}]", "ddhR", types, MergeNone, intrinsic, [IsStreamingCompatible]>; 1543} 1544 1545multiclass SInstWideDSPLong<string name, string types, string intrinsic> { 1546 def : SInst<name # "[_{d}]", "dhh", types, MergeNone, intrinsic, [IsStreamingCompatible]>; 1547 def _N : SInst<name # "[_n_{d}]", "dhR", types, MergeNone, intrinsic, [IsStreamingCompatible]>; 1548} 1549 1550multiclass SInstWideDSPWide<string name, string types, string intrinsic> { 1551 def : SInst<name # "[_{d}]", "ddh", types, MergeNone, intrinsic, [IsStreamingCompatible]>; 1552 def _N : SInst<name # "[_n_{d}]", "ddR", types, MergeNone, intrinsic, [IsStreamingCompatible]>; 1553} 1554 1555let TargetGuard = "sve2" in { 1556defm SVABALB_S : SInstWideDSPAcc<"svabalb", "sil", "aarch64_sve_sabalb">; 1557defm SVABALB_U : SInstWideDSPAcc<"svabalb", "UsUiUl", "aarch64_sve_uabalb">; 1558defm SVABALT_S : SInstWideDSPAcc<"svabalt", "sil", "aarch64_sve_sabalt">; 1559defm SVABALT_U : SInstWideDSPAcc<"svabalt", "UsUiUl", "aarch64_sve_uabalt">; 1560defm SVMLALB_S : SInstWideDSPAcc<"svmlalb", "sil", "aarch64_sve_smlalb">; 1561defm SVMLALB_U : SInstWideDSPAcc<"svmlalb", "UsUiUl", "aarch64_sve_umlalb">; 1562defm SVMLALT_S : SInstWideDSPAcc<"svmlalt", "sil", "aarch64_sve_smlalt">; 1563defm SVMLALT_U : SInstWideDSPAcc<"svmlalt", "UsUiUl", "aarch64_sve_umlalt">; 1564defm SVMLSLB_S : SInstWideDSPAcc<"svmlslb", "sil", "aarch64_sve_smlslb">; 1565defm SVMLSLB_U : SInstWideDSPAcc<"svmlslb", "UsUiUl", "aarch64_sve_umlslb">; 1566defm SVMLSLT_S : SInstWideDSPAcc<"svmlslt", "sil", "aarch64_sve_smlslt">; 1567defm SVMLSLT_U : SInstWideDSPAcc<"svmlslt", "UsUiUl", "aarch64_sve_umlslt">; 1568defm SVQDMLALB : SInstWideDSPAcc<"svqdmlalb", "sil", "aarch64_sve_sqdmlalb">; 1569defm SVQDMLALT : SInstWideDSPAcc<"svqdmlalt", "sil", "aarch64_sve_sqdmlalt">; 1570defm SVQDMLSLB : SInstWideDSPAcc<"svqdmlslb", "sil", "aarch64_sve_sqdmlslb">; 1571defm SVQDMLSLT : SInstWideDSPAcc<"svqdmlslt", "sil", "aarch64_sve_sqdmlslt">; 1572 1573defm SVABDLB_S : SInstWideDSPLong<"svabdlb", "sil", "aarch64_sve_sabdlb">; 1574defm SVABDLB_U : SInstWideDSPLong<"svabdlb", "UsUiUl", "aarch64_sve_uabdlb">; 1575defm SVABDLT_S : SInstWideDSPLong<"svabdlt", "sil", "aarch64_sve_sabdlt">; 1576defm SVABDLT_U : SInstWideDSPLong<"svabdlt", "UsUiUl", "aarch64_sve_uabdlt">; 1577defm SVADDLB_S : SInstWideDSPLong<"svaddlb", "sil", "aarch64_sve_saddlb">; 1578defm SVADDLB_U : SInstWideDSPLong<"svaddlb", "UsUiUl", "aarch64_sve_uaddlb">; 1579defm SVADDLT_S : SInstWideDSPLong<"svaddlt", "sil", "aarch64_sve_saddlt">; 1580defm SVADDLT_U : SInstWideDSPLong<"svaddlt", "UsUiUl", "aarch64_sve_uaddlt">; 1581defm SVMULLB_S : SInstWideDSPLong<"svmullb", "sil", "aarch64_sve_smullb">; 1582defm SVMULLB_U : SInstWideDSPLong<"svmullb", "UsUiUl", "aarch64_sve_umullb">; 1583defm SVMULLT_S : SInstWideDSPLong<"svmullt", "sil", "aarch64_sve_smullt">; 1584defm SVMULLT_U : SInstWideDSPLong<"svmullt", "UsUiUl", "aarch64_sve_umullt">; 1585defm SVQDMULLB : SInstWideDSPLong<"svqdmullb", "sil", "aarch64_sve_sqdmullb">; 1586defm SVQDMULLT : SInstWideDSPLong<"svqdmullt", "sil", "aarch64_sve_sqdmullt">; 1587defm SVSUBLB_S : SInstWideDSPLong<"svsublb", "sil", "aarch64_sve_ssublb">; 1588defm SVSUBLB_U : SInstWideDSPLong<"svsublb", "UsUiUl", "aarch64_sve_usublb">; 1589defm SVSUBLT_S : SInstWideDSPLong<"svsublt", "sil", "aarch64_sve_ssublt">; 1590defm SVSUBLT_U : SInstWideDSPLong<"svsublt", "UsUiUl", "aarch64_sve_usublt">; 1591 1592defm SVADDWB_S : SInstWideDSPWide<"svaddwb", "sil", "aarch64_sve_saddwb">; 1593defm SVADDWB_U : SInstWideDSPWide<"svaddwb", "UsUiUl", "aarch64_sve_uaddwb">; 1594defm SVADDWT_S : SInstWideDSPWide<"svaddwt", "sil", "aarch64_sve_saddwt">; 1595defm SVADDWT_U : SInstWideDSPWide<"svaddwt", "UsUiUl", "aarch64_sve_uaddwt">; 1596defm SVSUBWB_S : SInstWideDSPWide<"svsubwb", "sil", "aarch64_sve_ssubwb">; 1597defm SVSUBWB_U : SInstWideDSPWide<"svsubwb", "UsUiUl", "aarch64_sve_usubwb">; 1598defm SVSUBWT_S : SInstWideDSPWide<"svsubwt", "sil", "aarch64_sve_ssubwt">; 1599defm SVSUBWT_U : SInstWideDSPWide<"svsubwt", "UsUiUl", "aarch64_sve_usubwt">; 1600 1601def SVSHLLB_S_N : SInst<"svshllb[_n_{d}]", "dhi", "sil", MergeNone, "aarch64_sve_sshllb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftLeft, 0>]>; 1602def SVSHLLB_U_N : SInst<"svshllb[_n_{d}]", "dhi", "UsUiUl", MergeNone, "aarch64_sve_ushllb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftLeft, 0>]>; 1603def SVSHLLT_S_N : SInst<"svshllt[_n_{d}]", "dhi", "sil", MergeNone, "aarch64_sve_sshllt", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftLeft, 0>]>; 1604def SVSHLLT_U_N : SInst<"svshllt[_n_{d}]", "dhi", "UsUiUl", MergeNone, "aarch64_sve_ushllt", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftLeft, 0>]>; 1605 1606def SVMOVLB_S_N : SInst<"svmovlb[_{d}]", "dh", "sil", MergeNone, "", [IsStreamingCompatible]>; 1607def SVMOVLB_U_N : SInst<"svmovlb[_{d}]", "dh", "UsUiUl", MergeNone, "", [IsStreamingCompatible]>; 1608def SVMOVLT_S_N : SInst<"svmovlt[_{d}]", "dh", "sil", MergeNone, "", [IsStreamingCompatible]>; 1609def SVMOVLT_U_N : SInst<"svmovlt[_{d}]", "dh", "UsUiUl", MergeNone, "", [IsStreamingCompatible]>; 1610 1611def SVMLALB_S_LANE : SInst<"svmlalb_lane[_{d}]", "ddhhi", "il", MergeNone, "aarch64_sve_smlalb_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1612def SVMLALB_U_LANE : SInst<"svmlalb_lane[_{d}]", "ddhhi", "UiUl", MergeNone, "aarch64_sve_umlalb_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1613def SVMLALT_S_LANE : SInst<"svmlalt_lane[_{d}]", "ddhhi", "il", MergeNone, "aarch64_sve_smlalt_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1614def SVMLALT_U_LANE : SInst<"svmlalt_lane[_{d}]", "ddhhi", "UiUl", MergeNone, "aarch64_sve_umlalt_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1615def SVMLSLB_S_LANE : SInst<"svmlslb_lane[_{d}]", "ddhhi", "il", MergeNone, "aarch64_sve_smlslb_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1616def SVMLSLB_U_LANE : SInst<"svmlslb_lane[_{d}]", "ddhhi", "UiUl", MergeNone, "aarch64_sve_umlslb_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1617def SVMLSLT_S_LANE : SInst<"svmlslt_lane[_{d}]", "ddhhi", "il", MergeNone, "aarch64_sve_smlslt_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1618def SVMLSLT_U_LANE : SInst<"svmlslt_lane[_{d}]", "ddhhi", "UiUl", MergeNone, "aarch64_sve_umlslt_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1619def SVMULLB_S_LANE : SInst<"svmullb_lane[_{d}]", "dhhi", "il", MergeNone, "aarch64_sve_smullb_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1620def SVMULLB_U_LANE : SInst<"svmullb_lane[_{d}]", "dhhi", "UiUl", MergeNone, "aarch64_sve_umullb_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1621def SVMULLT_S_LANE : SInst<"svmullt_lane[_{d}]", "dhhi", "il", MergeNone, "aarch64_sve_smullt_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1622def SVMULLT_U_LANE : SInst<"svmullt_lane[_{d}]", "dhhi", "UiUl", MergeNone, "aarch64_sve_umullt_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1623def SVQDMLALB_LANE : SInst<"svqdmlalb_lane[_{d}]", "ddhhi", "il", MergeNone, "aarch64_sve_sqdmlalb_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1624def SVQDMLALT_LANE : SInst<"svqdmlalt_lane[_{d}]", "ddhhi", "il", MergeNone, "aarch64_sve_sqdmlalt_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1625def SVQDMLSLB_LANE : SInst<"svqdmlslb_lane[_{d}]", "ddhhi", "il", MergeNone, "aarch64_sve_sqdmlslb_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1626def SVQDMLSLT_LANE : SInst<"svqdmlslt_lane[_{d}]", "ddhhi", "il", MergeNone, "aarch64_sve_sqdmlslt_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1627def SVQDMULLB_LANE : SInst<"svqdmullb_lane[_{d}]", "dhhi", "il", MergeNone, "aarch64_sve_sqdmullb_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1628def SVQDMULLT_LANE : SInst<"svqdmullt_lane[_{d}]", "dhhi", "il", MergeNone, "aarch64_sve_sqdmullt_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 1629} 1630 1631//////////////////////////////////////////////////////////////////////////////// 1632// SVE2 - Narrowing DSP operations 1633 1634let TargetGuard = "sve2" in { 1635def SVADDHNB : SInst<"svaddhnb[_{d}]", "hdd", "silUsUiUl", MergeNone, "aarch64_sve_addhnb", [IsStreamingCompatible]>; 1636def SVADDHNT : SInst<"svaddhnt[_{d}]", "hhdd", "silUsUiUl", MergeNone, "aarch64_sve_addhnt", [IsStreamingCompatible]>; 1637def SVRADDHNB : SInst<"svraddhnb[_{d}]", "hdd", "silUsUiUl", MergeNone, "aarch64_sve_raddhnb", [IsStreamingCompatible]>; 1638def SVRADDHNT : SInst<"svraddhnt[_{d}]", "hhdd", "silUsUiUl", MergeNone, "aarch64_sve_raddhnt", [IsStreamingCompatible]>; 1639def SVRSUBHNB : SInst<"svrsubhnb[_{d}]", "hdd", "silUsUiUl", MergeNone, "aarch64_sve_rsubhnb", [IsStreamingCompatible]>; 1640def SVRSUBHNT : SInst<"svrsubhnt[_{d}]", "hhdd", "silUsUiUl", MergeNone, "aarch64_sve_rsubhnt", [IsStreamingCompatible]>; 1641def SVSUBHNB : SInst<"svsubhnb[_{d}]", "hdd", "silUsUiUl", MergeNone, "aarch64_sve_subhnb", [IsStreamingCompatible]>; 1642def SVSUBHNT : SInst<"svsubhnt[_{d}]", "hhdd", "silUsUiUl", MergeNone, "aarch64_sve_subhnt", [IsStreamingCompatible]>; 1643 1644def SVADDHNB_N : SInst<"svaddhnb[_n_{d}]", "hda", "silUsUiUl", MergeNone, "aarch64_sve_addhnb", [IsStreamingCompatible]>; 1645def SVADDHNT_N : SInst<"svaddhnt[_n_{d}]", "hhda", "silUsUiUl", MergeNone, "aarch64_sve_addhnt", [IsStreamingCompatible]>; 1646def SVRADDHNB_N : SInst<"svraddhnb[_n_{d}]", "hda", "silUsUiUl", MergeNone, "aarch64_sve_raddhnb", [IsStreamingCompatible]>; 1647def SVRADDHNT_N : SInst<"svraddhnt[_n_{d}]", "hhda", "silUsUiUl", MergeNone, "aarch64_sve_raddhnt", [IsStreamingCompatible]>; 1648def SVRSUBHNB_N : SInst<"svrsubhnb[_n_{d}]", "hda", "silUsUiUl", MergeNone, "aarch64_sve_rsubhnb", [IsStreamingCompatible]>; 1649def SVRSUBHNT_N : SInst<"svrsubhnt[_n_{d}]", "hhda", "silUsUiUl", MergeNone, "aarch64_sve_rsubhnt", [IsStreamingCompatible]>; 1650def SVSUBHNB_N : SInst<"svsubhnb[_n_{d}]", "hda", "silUsUiUl", MergeNone, "aarch64_sve_subhnb", [IsStreamingCompatible]>; 1651def SVSUBHNT_N : SInst<"svsubhnt[_n_{d}]", "hhda", "silUsUiUl", MergeNone, "aarch64_sve_subhnt", [IsStreamingCompatible]>; 1652 1653def SVSHRNB : SInst<"svshrnb[_n_{d}]", "hdi", "silUsUiUl", MergeNone, "aarch64_sve_shrnb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftRightNarrow, 0>]>; 1654def SVRSHRNB : SInst<"svrshrnb[_n_{d}]", "hdi", "silUsUiUl", MergeNone, "aarch64_sve_rshrnb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftRightNarrow, 0>]>; 1655def SVQSHRUNB : SInst<"svqshrunb[_n_{d}]", "edi", "sil", MergeNone, "aarch64_sve_sqshrunb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftRightNarrow, 0>]>; 1656def SVQRSHRUNB : SInst<"svqrshrunb[_n_{d}]", "edi", "sil", MergeNone, "aarch64_sve_sqrshrunb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftRightNarrow, 0>]>; 1657def SVQSHRNB_S : SInst<"svqshrnb[_n_{d}]", "hdi", "sil", MergeNone, "aarch64_sve_sqshrnb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftRightNarrow, 0>]>; 1658def SVQSHRNB_U : SInst<"svqshrnb[_n_{d}]", "hdi", "UsUiUl", MergeNone, "aarch64_sve_uqshrnb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftRightNarrow, 0>]>; 1659def SVQRSHRNB_S : SInst<"svqrshrnb[_n_{d}]", "hdi", "sil", MergeNone, "aarch64_sve_sqrshrnb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftRightNarrow, 0>]>; 1660def SVQRSHRNB_U : SInst<"svqrshrnb[_n_{d}]", "hdi", "UsUiUl", MergeNone, "aarch64_sve_uqrshrnb", [IsStreamingCompatible], [ImmCheck<1, ImmCheckShiftRightNarrow, 0>]>; 1661 1662def SVSHRNT : SInst<"svshrnt[_n_{d}]", "hhdi", "silUsUiUl", MergeNone, "aarch64_sve_shrnt", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRightNarrow, 1>]>; 1663def SVRSHRNT : SInst<"svrshrnt[_n_{d}]", "hhdi", "silUsUiUl", MergeNone, "aarch64_sve_rshrnt", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRightNarrow, 1>]>; 1664def SVQSHRUNT : SInst<"svqshrunt[_n_{d}]", "eedi", "sil", MergeNone, "aarch64_sve_sqshrunt", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRightNarrow, 1>]>; 1665def SVQRSHRUNT : SInst<"svqrshrunt[_n_{d}]", "eedi", "sil", MergeNone, "aarch64_sve_sqrshrunt", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRightNarrow, 1>]>; 1666def SVQSHRNT_S : SInst<"svqshrnt[_n_{d}]", "hhdi", "sil", MergeNone, "aarch64_sve_sqshrnt", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRightNarrow, 1>]>; 1667def SVQSHRNT_U : SInst<"svqshrnt[_n_{d}]", "hhdi", "UsUiUl", MergeNone, "aarch64_sve_uqshrnt", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRightNarrow, 1>]>; 1668def SVQRSHRNT_S : SInst<"svqrshrnt[_n_{d}]", "hhdi", "sil", MergeNone, "aarch64_sve_sqrshrnt", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRightNarrow, 1>]>; 1669def SVQRSHRNT_U : SInst<"svqrshrnt[_n_{d}]", "hhdi", "UsUiUl", MergeNone, "aarch64_sve_uqrshrnt", [IsStreamingCompatible], [ImmCheck<2, ImmCheckShiftRightNarrow, 1>]>; 1670} 1671//////////////////////////////////////////////////////////////////////////////// 1672// SVE2 - Unary narrowing operations 1673 1674let TargetGuard = "sve2" in { 1675def SVQXTNB_S : SInst<"svqxtnb[_{d}]", "hd", "sil", MergeNone, "aarch64_sve_sqxtnb", [IsStreamingCompatible]>; 1676def SVQXTNB_U : SInst<"svqxtnb[_{d}]", "hd", "UsUiUl", MergeNone, "aarch64_sve_uqxtnb", [IsStreamingCompatible]>; 1677def SVQXTUNB_S : SInst<"svqxtunb[_{d}]", "ed", "sil", MergeNone, "aarch64_sve_sqxtunb", [IsStreamingCompatible]>; 1678 1679def SVQXTNT_S : SInst<"svqxtnt[_{d}]", "hhd", "sil", MergeNone, "aarch64_sve_sqxtnt", [IsStreamingCompatible]>; 1680def SVQXTNT_U : SInst<"svqxtnt[_{d}]", "hhd", "UsUiUl", MergeNone, "aarch64_sve_uqxtnt", [IsStreamingCompatible]>; 1681def SVQXTUNT_S : SInst<"svqxtunt[_{d}]", "eed", "sil", MergeNone, "aarch64_sve_sqxtunt", [IsStreamingCompatible]>; 1682} 1683 1684//////////////////////////////////////////////////////////////////////////////// 1685// SVE2 - Widening complex integer arithmetic 1686 1687let TargetGuard = "sve2" in { 1688defm SVADDLBT : SInstWideDSPLong<"svaddlbt", "sil", "aarch64_sve_saddlbt">; 1689defm SVSUBLBT : SInstWideDSPLong<"svsublbt", "sil", "aarch64_sve_ssublbt">; 1690defm SVSUBLTB : SInstWideDSPLong<"svsubltb", "sil", "aarch64_sve_ssubltb">; 1691 1692defm SVQDMLALBT : SInstWideDSPAcc<"svqdmlalbt", "sil", "aarch64_sve_sqdmlalbt">; 1693defm SVQDMLSLBT : SInstWideDSPAcc<"svqdmlslbt", "sil", "aarch64_sve_sqdmlslbt">; 1694} 1695 1696//////////////////////////////////////////////////////////////////////////////// 1697// SVE2 - Non-temporal gather/scatter 1698let TargetGuard = "sve2" in { 1699// Non-temporal gather load one vector (vector base) 1700def SVLDNT1_GATHER_BASES_U : MInst<"svldnt1_gather[_{2}base]_{0}", "dPu", "ilUiUlfd", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldnt1_gather_scalar_offset">; 1701def SVLDNT1SB_GATHER_BASES_U : MInst<"svldnt1sb_gather[_{2}base]_{0}", "dPu", "ilUiUl", [IsGatherLoad], MemEltTyInt8, "aarch64_sve_ldnt1_gather_scalar_offset">; 1702def SVLDNT1UB_GATHER_BASES_U : MInst<"svldnt1ub_gather[_{2}base]_{0}", "dPu", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldnt1_gather_scalar_offset">; 1703def SVLDNT1SH_GATHER_BASES_U : MInst<"svldnt1sh_gather[_{2}base]_{0}", "dPu", "ilUiUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldnt1_gather_scalar_offset">; 1704def SVLDNT1UH_GATHER_BASES_U : MInst<"svldnt1uh_gather[_{2}base]_{0}", "dPu", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnt1_gather_scalar_offset">; 1705def SVLDNT1SW_GATHER_BASES_U : MInst<"svldnt1sw_gather[_{2}base]_{0}", "dPu", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ldnt1_gather_scalar_offset">; 1706def SVLDNT1UW_GATHER_BASES_U : MInst<"svldnt1uw_gather[_{2}base]_{0}", "dPu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnt1_gather_scalar_offset">; 1707 1708// Non-temporal gather load one vector (scalar base, signed vector offset in bytes) 1709def SVLDNT1_GATHER_64B_OFFSETS_S : MInst<"svldnt1_gather_[{3}]offset[_{0}]", "dPcx", "lUld", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldnt1_gather">; 1710def SVLDNT1SB_GATHER_64B_OFFSETS_S : MInst<"svldnt1sb_gather_[{3}]offset_{0}", "dPSx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldnt1_gather">; 1711def SVLDNT1UB_GATHER_64B_OFFSETS_S : MInst<"svldnt1ub_gather_[{3}]offset_{0}", "dPWx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldnt1_gather">; 1712def SVLDNT1SH_GATHER_64B_OFFSETS_S : MInst<"svldnt1sh_gather_[{3}]offset_{0}", "dPTx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldnt1_gather">; 1713def SVLDNT1UH_GATHER_64B_OFFSETS_S : MInst<"svldnt1uh_gather_[{3}]offset_{0}", "dPXx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnt1_gather">; 1714def SVLDNT1SW_GATHER_64B_OFFSETS_S : MInst<"svldnt1sw_gather_[{3}]offset_{0}", "dPUx", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ldnt1_gather">; 1715def SVLDNT1UW_GATHER_64B_OFFSETS_S : MInst<"svldnt1uw_gather_[{3}]offset_{0}", "dPYx", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnt1_gather">; 1716 1717// Non-temporal gather load one vector (scalar base, unsigned vector offset in bytes) 1718def SVLDNT1_GATHER_64B_OFFSETS_U : MInst<"svldnt1_gather_[{3}]offset[_{0}]", "dPcu", "lUld", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldnt1_gather">; 1719def SVLDNT1SB_GATHER_64B_OFFSETS_U : MInst<"svldnt1sb_gather_[{3}]offset_{0}", "dPSu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldnt1_gather">; 1720def SVLDNT1UB_GATHER_64B_OFFSETS_U : MInst<"svldnt1ub_gather_[{3}]offset_{0}", "dPWu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldnt1_gather">; 1721def SVLDNT1SH_GATHER_64B_OFFSETS_U : MInst<"svldnt1sh_gather_[{3}]offset_{0}", "dPTu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldnt1_gather">; 1722def SVLDNT1UH_GATHER_64B_OFFSETS_U : MInst<"svldnt1uh_gather_[{3}]offset_{0}", "dPXu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnt1_gather">; 1723def SVLDNT1SW_GATHER_64B_OFFSETS_U : MInst<"svldnt1sw_gather_[{3}]offset_{0}", "dPUu", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ldnt1_gather">; 1724def SVLDNT1UW_GATHER_64B_OFFSETS_U : MInst<"svldnt1uw_gather_[{3}]offset_{0}", "dPYu", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnt1_gather">; 1725 1726def SVLDNT1_GATHER_32B_OFFSETS_U : MInst<"svldnt1_gather_[{3}]offset[_{0}]", "dPcu", "iUif", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldnt1_gather_uxtw">; 1727def SVLDNT1SB_GATHER_32B_OFFSETS_U : MInst<"svldnt1sb_gather_[{3}]offset_{0}", "dPSu", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldnt1_gather_uxtw">; 1728def SVLDNT1UB_GATHER_32B_OFFSETS_U : MInst<"svldnt1ub_gather_[{3}]offset_{0}", "dPWu", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldnt1_gather_uxtw">; 1729def SVLDNT1SH_GATHER_32B_OFFSETS_U : MInst<"svldnt1sh_gather_[{3}]offset_{0}", "dPTu", "iUi", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldnt1_gather_uxtw">; 1730def SVLDNT1UH_GATHER_32B_OFFSETS_U : MInst<"svldnt1uh_gather_[{3}]offset_{0}", "dPXu", "iUi", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnt1_gather_uxtw">; 1731 1732// Non-temporal gather load one vector (vector base, scalar offset in bytes) 1733def SVLDNT1_GATHER_OFFSET_S : MInst<"svldnt1_gather[_{2}base]_offset_{0}", "dPul", "ilUiUlfd", [IsGatherLoad, IsByteIndexed], MemEltTyDefault, "aarch64_sve_ldnt1_gather_scalar_offset">; 1734def SVLDNT1SB_GATHER_OFFSET_S : MInst<"svldnt1sb_gather[_{2}base]_offset_{0}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt8, "aarch64_sve_ldnt1_gather_scalar_offset">; 1735def SVLDNT1UB_GATHER_OFFSET_S : MInst<"svldnt1ub_gather[_{2}base]_offset_{0}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt8, "aarch64_sve_ldnt1_gather_scalar_offset">; 1736def SVLDNT1SH_GATHER_OFFSET_S : MInst<"svldnt1sh_gather[_{2}base]_offset_{0}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt16, "aarch64_sve_ldnt1_gather_scalar_offset">; 1737def SVLDNT1UH_GATHER_OFFSET_S : MInst<"svldnt1uh_gather[_{2}base]_offset_{0}", "dPul", "ilUiUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnt1_gather_scalar_offset">; 1738def SVLDNT1SW_GATHER_OFFSET_S : MInst<"svldnt1sw_gather[_{2}base]_offset_{0}", "dPul", "lUl", [IsGatherLoad, IsByteIndexed], MemEltTyInt32, "aarch64_sve_ldnt1_gather_scalar_offset">; 1739def SVLDNT1UW_GATHER_OFFSET_S : MInst<"svldnt1uw_gather[_{2}base]_offset_{0}", "dPul", "lUl", [IsGatherLoad, IsByteIndexed, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnt1_gather_scalar_offset">; 1740 1741// Non-temporal gather load one vector (scalar base, signed vector index) 1742def SVLDNT1_GATHER_64B_INDICES_S : MInst<"svldnt1_gather_[{3}]index[_{0}]", "dPcx", "lUld", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldnt1_gather_index">; 1743def SVLDNT1SH_GATHER_64B_INDICES_S : MInst<"svldnt1sh_gather_[{3}]index_{0}", "dPTx", "lUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldnt1_gather_index">; 1744def SVLDNT1UH_GATHER_64B_INDICES_S : MInst<"svldnt1uh_gather_[{3}]index_{0}", "dPXx", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnt1_gather_index">; 1745def SVLDNT1SW_GATHER_64B_INDICES_S : MInst<"svldnt1sw_gather_[{3}]index_{0}", "dPUx", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ldnt1_gather_index">; 1746def SVLDNT1UW_GATHER_64B_INDICES_S : MInst<"svldnt1uw_gather_[{3}]index_{0}", "dPYx", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnt1_gather_index">; 1747 1748// Non temporal gather load one vector (scalar base, unsigned vector index) 1749def SVLDNT1_GATHER_64B_INDICES_U : MInst<"svldnt1_gather_[{3}]index[_{0}]", "dPcu", "lUld", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldnt1_gather_index">; 1750def SVLDNT1SH_GATHER_64B_INDICES_U : MInst<"svldnt1sh_gather_[{3}]index_{0}", "dPTu", "lUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldnt1_gather_index">; 1751def SVLDNT1UH_GATHER_64B_INDICES_U : MInst<"svldnt1uh_gather_[{3}]index_{0}", "dPXu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnt1_gather_index">; 1752def SVLDNT1SW_GATHER_64B_INDICES_U : MInst<"svldnt1sw_gather_[{3}]index_{0}", "dPUu", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ldnt1_gather_index">; 1753def SVLDNT1UW_GATHER_64B_INDICES_U : MInst<"svldnt1uw_gather_[{3}]index_{0}", "dPYu", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnt1_gather_index">; 1754 1755// Non-temporal gather load one vector (vector base, signed scalar index) 1756def SVLDNT1_GATHER_INDEX_S : MInst<"svldnt1_gather[_{2}base]_index_{0}", "dPul", "ilUiUlfd", [IsGatherLoad], MemEltTyDefault, "aarch64_sve_ldnt1_gather_scalar_offset">; 1757def SVLDNT1SH_GATHER_INDEX_S : MInst<"svldnt1sh_gather[_{2}base]_index_{0}", "dPul", "ilUiUl", [IsGatherLoad], MemEltTyInt16, "aarch64_sve_ldnt1_gather_scalar_offset">; 1758def SVLDNT1UH_GATHER_INDEX_S : MInst<"svldnt1uh_gather[_{2}base]_index_{0}", "dPul", "ilUiUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt16, "aarch64_sve_ldnt1_gather_scalar_offset">; 1759def SVLDNT1SW_GATHER_INDEX_S : MInst<"svldnt1sw_gather[_{2}base]_index_{0}", "dPul", "lUl", [IsGatherLoad], MemEltTyInt32, "aarch64_sve_ldnt1_gather_scalar_offset">; 1760def SVLDNT1UW_GATHER_INDEX_S : MInst<"svldnt1uw_gather[_{2}base]_index_{0}", "dPul", "lUl", [IsGatherLoad, IsZExtReturn], MemEltTyInt32, "aarch64_sve_ldnt1_gather_scalar_offset">; 1761 1762// Non-temporal scatter store one vector (vector base) 1763def SVSTNT1_SCATTER_BASES_U : MInst<"svstnt1_scatter[_{2}base_{d}]", "vPud", "ilUiUlfd", [IsScatterStore], MemEltTyDefault, "aarch64_sve_stnt1_scatter_scalar_offset">; 1764def SVSTNT1B_SCATTER_BASES_U : MInst<"svstnt1b_scatter[_{2}base_{d}]", "vPud", "ilUiUl", [IsScatterStore], MemEltTyInt8, "aarch64_sve_stnt1_scatter_scalar_offset">; 1765def SVSTNT1H_SCATTER_BASES_U : MInst<"svstnt1h_scatter[_{2}base_{d}]", "vPud", "ilUiUl", [IsScatterStore], MemEltTyInt16, "aarch64_sve_stnt1_scatter_scalar_offset">; 1766def SVSTNT1W_SCATTER_BASES_U : MInst<"svstnt1w_scatter[_{2}base_{d}]", "vPud", "lUl", [IsScatterStore], MemEltTyInt32, "aarch64_sve_stnt1_scatter_scalar_offset">; 1767 1768// Non-temporal scatter store one vector (scalar base, signed vector offset in bytes) 1769def SVSTNT1_SCATTER_64B_OFFSETS_S : MInst<"svstnt1_scatter_[{3}]offset[_{d}]", "vPpxd", "lUld", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_stnt1_scatter">; 1770def SVSTNT1B_SCATTER_64B_OFFSETS_SS : MInst<"svstnt1b_scatter_[{3}]offset[_{d}]", "vPAxd", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_stnt1_scatter">; 1771def SVSTNT1B_SCATTER_64B_OFFSETS_SU : MInst<"svstnt1b_scatter_[{3}]offset[_{d}]", "vPExd", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_stnt1_scatter">; 1772def SVSTNT1H_SCATTER_64B_OFFSETS_SS : MInst<"svstnt1h_scatter_[{3}]offset[_{d}]", "vPBxd", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_stnt1_scatter">; 1773def SVSTNT1H_SCATTER_64B_OFFSETS_SU : MInst<"svstnt1h_scatter_[{3}]offset[_{d}]", "vPFxd", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_stnt1_scatter">; 1774def SVSTNT1W_SCATTER_64B_OFFSETS_SS : MInst<"svstnt1w_scatter_[{3}]offset[_{d}]", "vPCxd", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_stnt1_scatter">; 1775def SVSTNT1W_SCATTER_64B_OFFSETS_SU : MInst<"svstnt1w_scatter_[{3}]offset[_{d}]", "vPGxd", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_stnt1_scatter">; 1776 1777// Non-temporal scatter store one vector (scalar base, unsigned vector offset in bytes) 1778def SVSTNT1_SCATTER_64B_OFFSETS_U : MInst<"svstnt1_scatter_[{3}]offset[_{d}]", "vPpud", "lUld", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_stnt1_scatter">; 1779def SVSTNT1B_SCATTER_64B_OFFSETS_US : MInst<"svstnt1b_scatter_[{3}]offset[_{d}]", "vPAud", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_stnt1_scatter">; 1780def SVSTNT1B_SCATTER_64B_OFFSETS_UU : MInst<"svstnt1b_scatter_[{3}]offset[_{d}]", "vPEud", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_stnt1_scatter">; 1781def SVSTNT1H_SCATTER_64B_OFFSETS_US : MInst<"svstnt1h_scatter_[{3}]offset[_{d}]", "vPBud", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_stnt1_scatter">; 1782def SVSTNT1H_SCATTER_64B_OFFSETS_UU : MInst<"svstnt1h_scatter_[{3}]offset[_{d}]", "vPFud", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_stnt1_scatter">; 1783def SVSTNT1W_SCATTER_64B_OFFSETS_US : MInst<"svstnt1w_scatter_[{3}]offset[_{d}]", "vPCud", "l", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_stnt1_scatter">; 1784def SVSTNT1W_SCATTER_64B_OFFSETS_UU : MInst<"svstnt1w_scatter_[{3}]offset[_{d}]", "vPGud", "Ul", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_stnt1_scatter">; 1785 1786def SVSTNT1_SCATTER_32B_OFFSETS_U : MInst<"svstnt1_scatter_[{3}]offset[_{d}]", "vPpud", "iUif", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_stnt1_scatter_uxtw">; 1787def SVSTNT1B_SCATTER_32B_OFFSETS_US : MInst<"svstnt1b_scatter_[{3}]offset[_{d}]", "vPAud", "i", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_stnt1_scatter_uxtw">; 1788def SVSTNT1B_SCATTER_32B_OFFSETS_UU : MInst<"svstnt1b_scatter_[{3}]offset[_{d}]", "vPEud", "Ui", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_stnt1_scatter_uxtw">; 1789def SVSTNT1H_SCATTER_32B_OFFSETS_US : MInst<"svstnt1h_scatter_[{3}]offset[_{d}]", "vPBud", "i", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_stnt1_scatter_uxtw">; 1790def SVSTNT1H_SCATTER_32B_OFFSETS_UU : MInst<"svstnt1h_scatter_[{3}]offset[_{d}]", "vPFud", "Ui", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_stnt1_scatter_uxtw">; 1791 1792// Non-temporal scatter store one vector (vector base, scalar offset in bytes) 1793def SVSTNT1_SCATTER_OFFSET_S : MInst<"svstnt1_scatter[_{2}base]_offset[_{d}]", "vPuld", "ilUiUlfd", [IsScatterStore, IsByteIndexed], MemEltTyDefault, "aarch64_sve_stnt1_scatter_scalar_offset">; 1794def SVSTNT1B_SCATTER_OFFSET_S : MInst<"svstnt1b_scatter[_{2}base]_offset[_{d}]", "vPuld", "ilUiUl", [IsScatterStore, IsByteIndexed], MemEltTyInt8, "aarch64_sve_stnt1_scatter_scalar_offset">; 1795def SVSTNT1H_SCATTER_OFFSET_S : MInst<"svstnt1h_scatter[_{2}base]_offset[_{d}]", "vPuld", "ilUiUl", [IsScatterStore, IsByteIndexed], MemEltTyInt16, "aarch64_sve_stnt1_scatter_scalar_offset">; 1796def SVSTNT1W_SCATTER_OFFSET_S : MInst<"svstnt1w_scatter[_{2}base]_offset[_{d}]", "vPuld", "lUl", [IsScatterStore, IsByteIndexed], MemEltTyInt32, "aarch64_sve_stnt1_scatter_scalar_offset">; 1797 1798// Non-temporal scatter store one vector (scalar base, signed vector index) 1799def SVSTNT1_SCATTER_INDICES_S : MInst<"svstnt1_scatter_[{3}]index[_{d}]", "vPpxd", "lUld", [IsScatterStore], MemEltTyDefault, "aarch64_sve_stnt1_scatter_index">; 1800def SVSTNT1H_SCATTER_INDICES_SS : MInst<"svstnt1h_scatter_[{3}]index[_{d}]", "vPBxd", "l", [IsScatterStore], MemEltTyInt16, "aarch64_sve_stnt1_scatter_index">; 1801def SVSTNT1H_SCATTER_INDICES_SU : MInst<"svstnt1h_scatter_[{3}]index[_{d}]", "vPFxd", "Ul", [IsScatterStore], MemEltTyInt16, "aarch64_sve_stnt1_scatter_index">; 1802def SVSTNT1W_SCATTER_INDICES_SS : MInst<"svstnt1w_scatter_[{3}]index[_{d}]", "vPCxd", "l", [IsScatterStore], MemEltTyInt32, "aarch64_sve_stnt1_scatter_index">; 1803def SVSTNT1W_SCATTER_INDICES_SU : MInst<"svstnt1w_scatter_[{3}]index[_{d}]", "vPGxd", "Ul", [IsScatterStore], MemEltTyInt32, "aarch64_sve_stnt1_scatter_index">; 1804 1805// Non-temporal scatter store one vector (scalar base, unsigned vector index) 1806def SVSTNT1_SCATTER_INDICES_U : MInst<"svstnt1_scatter_[{3}]index[_{d}]", "vPpud", "lUld", [IsScatterStore], MemEltTyDefault, "aarch64_sve_stnt1_scatter_index">; 1807def SVSTNT1H_SCATTER_INDICES_US : MInst<"svstnt1h_scatter_[{3}]index[_{d}]", "vPBud", "l", [IsScatterStore], MemEltTyInt16, "aarch64_sve_stnt1_scatter_index">; 1808def SVSTNT1H_SCATTER_INDICES_UU : MInst<"svstnt1h_scatter_[{3}]index[_{d}]", "vPFud", "Ul", [IsScatterStore], MemEltTyInt16, "aarch64_sve_stnt1_scatter_index">; 1809def SVSTNT1W_SCATTER_INDICES_US : MInst<"svstnt1w_scatter_[{3}]index[_{d}]", "vPCud", "l", [IsScatterStore], MemEltTyInt32, "aarch64_sve_stnt1_scatter_index">; 1810def SVSTNT1W_SCATTER_INDICES_UU : MInst<"svstnt1w_scatter_[{3}]index[_{d}]", "vPGud", "Ul", [IsScatterStore], MemEltTyInt32, "aarch64_sve_stnt1_scatter_index">; 1811 1812// Non-temporal scatter store one vector (vector base, signed scalar index) 1813def SVSTNT1_SCATTER_INDEX_S : MInst<"svstnt1_scatter[_{2}base]_index[_{d}]", "vPuld", "ilUiUlfd", [IsScatterStore], MemEltTyDefault, "aarch64_sve_stnt1_scatter_scalar_offset">; 1814def SVSTNT1H_SCATTER_INDEX_S : MInst<"svstnt1h_scatter[_{2}base]_index[_{d}]", "vPuld", "ilUiUl", [IsScatterStore], MemEltTyInt16, "aarch64_sve_stnt1_scatter_scalar_offset">; 1815def SVSTNT1W_SCATTER_INDEX_S : MInst<"svstnt1w_scatter[_{2}base]_index[_{d}]", "vPuld", "lUl", [IsScatterStore], MemEltTyInt32, "aarch64_sve_stnt1_scatter_scalar_offset">; 1816} 1817 1818//////////////////////////////////////////////////////////////////////////////// 1819// SVE2 - Polynomial arithmetic 1820 1821let TargetGuard = "sve2" in { 1822def SVEORBT : SInst<"sveorbt[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_eorbt", [IsStreamingCompatible]>; 1823def SVEORBT_N : SInst<"sveorbt[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_eorbt", [IsStreamingCompatible]>; 1824def SVEORTB : SInst<"sveortb[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_eortb", [IsStreamingCompatible]>; 1825def SVEORTB_N : SInst<"sveortb[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_eortb", [IsStreamingCompatible]>; 1826def SVPMUL : SInst<"svpmul[_{d}]", "ddd", "Uc", MergeNone, "aarch64_sve_pmul", [IsStreamingCompatible]>; 1827def SVPMUL_N : SInst<"svpmul[_n_{d}]", "dda", "Uc", MergeNone, "aarch64_sve_pmul", [IsStreamingCompatible]>; 1828def SVPMULLB : SInst<"svpmullb[_{d}]", "dhh", "UsUl", MergeNone, "", [IsStreamingCompatible]>; 1829def SVPMULLB_N : SInst<"svpmullb[_n_{d}]", "dhR", "UsUl", MergeNone, "", [IsStreamingCompatible]>; 1830def SVPMULLB_PAIR : SInst<"svpmullb_pair[_{d}]", "ddd", "UcUi", MergeNone, "aarch64_sve_pmullb_pair", [IsStreamingCompatible]>; 1831def SVPMULLB_PAIR_N : SInst<"svpmullb_pair[_n_{d}]", "dda", "UcUi", MergeNone, "aarch64_sve_pmullb_pair", [IsStreamingCompatible]>; 1832def SVPMULLT : SInst<"svpmullt[_{d}]", "dhh", "UsUl", MergeNone, "", [IsStreamingCompatible]>; 1833def SVPMULLT_N : SInst<"svpmullt[_n_{d}]", "dhR", "UsUl", MergeNone, "", [IsStreamingCompatible]>; 1834def SVPMULLT_PAIR : SInst<"svpmullt_pair[_{d}]", "ddd", "UcUi", MergeNone, "aarch64_sve_pmullt_pair", [IsStreamingCompatible]>; 1835def SVPMULLT_PAIR_N : SInst<"svpmullt_pair[_n_{d}]", "dda", "UcUi", MergeNone, "aarch64_sve_pmullt_pair", [IsStreamingCompatible]>; 1836} 1837 1838//////////////////////////////////////////////////////////////////////////////// 1839// SVE2 - Complex integer dot product 1840 1841let TargetGuard = "sve2" in { 1842def SVCDOT : SInst<"svcdot[_{d}]", "ddqqi", "il", MergeNone, "aarch64_sve_cdot", [IsStreamingCompatible], [ImmCheck<3, ImmCheckComplexRotAll90>]>; 1843def SVCDOT_LANE : SInst<"svcdot_lane[_{d}]", "ddqqii", "il", MergeNone, "aarch64_sve_cdot_lane", [IsStreamingCompatible], [ImmCheck<4, ImmCheckComplexRotAll90>, 1844 ImmCheck<3, ImmCheckLaneIndexDot, 2>]>; 1845} 1846 1847//////////////////////////////////////////////////////////////////////////////// 1848// SVE2 - Floating-point widening multiply-accumulate 1849 1850let TargetGuard = "sve2" in { 1851def SVMLALB_F : SInst<"svmlalb[_{d}]", "ddhh", "f", MergeNone, "aarch64_sve_fmlalb", [IsStreamingCompatible]>; 1852def SVMLALB_F_N : SInst<"svmlalb[_n_{d}]", "ddhR", "f", MergeNone, "aarch64_sve_fmlalb", [IsStreamingCompatible]>; 1853def SVMLALB_F_LANE : SInst<"svmlalb_lane[_{d}]", "ddhhi", "f", MergeNone, "aarch64_sve_fmlalb_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1854def SVMLALT_F : SInst<"svmlalt[_{d}]", "ddhh", "f", MergeNone, "aarch64_sve_fmlalt", [IsStreamingCompatible]>; 1855def SVMLALT_F_N : SInst<"svmlalt[_n_{d}]", "ddhR", "f", MergeNone, "aarch64_sve_fmlalt", [IsStreamingCompatible]>; 1856def SVMLALT_F_LANE : SInst<"svmlalt_lane[_{d}]", "ddhhi", "f", MergeNone, "aarch64_sve_fmlalt_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1857def SVMLSLB_F : SInst<"svmlslb[_{d}]", "ddhh", "f", MergeNone, "aarch64_sve_fmlslb", [IsStreamingCompatible]>; 1858def SVMLSLB_F_N : SInst<"svmlslb[_n_{d}]", "ddhR", "f", MergeNone, "aarch64_sve_fmlslb", [IsStreamingCompatible]>; 1859def SVMLSLB_F_LANE : SInst<"svmlslb_lane[_{d}]", "ddhhi", "f", MergeNone, "aarch64_sve_fmlslb_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1860def SVMLSLT_F : SInst<"svmlslt[_{d}]", "ddhh", "f", MergeNone, "aarch64_sve_fmlslt", [IsStreamingCompatible]>; 1861def SVMLSLT_F_N : SInst<"svmlslt[_n_{d}]", "ddhR", "f", MergeNone, "aarch64_sve_fmlslt", [IsStreamingCompatible]>; 1862def SVMLSLT_F_LANE : SInst<"svmlslt_lane[_{d}]", "ddhhi", "f", MergeNone, "aarch64_sve_fmlslt_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 1863} 1864 1865//////////////////////////////////////////////////////////////////////////////// 1866// SVE2 - Floating-point integer binary logarithm 1867 1868let TargetGuard = "sve2" in { 1869def SVLOGB_M : SInst<"svlogb[_{d}]", "xxPd", "hfd", MergeOp1, "aarch64_sve_flogb", [IsStreamingCompatible]>; 1870def SVLOGB_X : SInst<"svlogb[_{d}]", "xPd", "hfd", MergeAnyExp, "aarch64_sve_flogb", [IsStreamingCompatible]>; 1871def SVLOGB_Z : SInst<"svlogb[_{d}]", "xPd", "hfd", MergeZeroExp, "aarch64_sve_flogb", [IsStreamingCompatible]>; 1872} 1873 1874//////////////////////////////////////////////////////////////////////////////// 1875// SVE2 - Vector Histogram count 1876 1877let TargetGuard = "sve2" in { 1878def SVHISTCNT : SInst<"svhistcnt[_{d}]_z", "uPdd", "ilUiUl", MergeNone, "aarch64_sve_histcnt">; 1879def SVHISTSEG : SInst<"svhistseg[_{d}]", "udd", "cUc", MergeNone, "aarch64_sve_histseg">; 1880} 1881 1882//////////////////////////////////////////////////////////////////////////////// 1883// SVE2 - Character match 1884 1885let TargetGuard = "sve2" in { 1886def SVMATCH : SInst<"svmatch[_{d}]", "PPdd", "csUcUs", MergeNone, "aarch64_sve_match">; 1887def SVNMATCH : SInst<"svnmatch[_{d}]", "PPdd", "csUcUs", MergeNone, "aarch64_sve_nmatch">; 1888} 1889 1890//////////////////////////////////////////////////////////////////////////////// 1891// SVE2 - Contiguous conflict detection 1892let TargetGuard = "sve2" in { 1893def SVWHILERW_B : SInst<"svwhilerw[_{1}]", "Pcc", "cUc", MergeNone, "aarch64_sve_whilerw_b", [IsOverloadWhileRW, IsStreamingCompatible]>; 1894def SVWHILERW_H : SInst<"svwhilerw[_{1}]", "Pcc", "sUsh", MergeNone, "aarch64_sve_whilerw_h", [IsOverloadWhileRW, IsStreamingCompatible]>; 1895def SVWHILERW_S : SInst<"svwhilerw[_{1}]", "Pcc", "iUif", MergeNone, "aarch64_sve_whilerw_s", [IsOverloadWhileRW, IsStreamingCompatible]>; 1896def SVWHILERW_D : SInst<"svwhilerw[_{1}]", "Pcc", "lUld", MergeNone, "aarch64_sve_whilerw_d", [IsOverloadWhileRW, IsStreamingCompatible]>; 1897 1898def SVWHILEWR_B : SInst<"svwhilewr[_{1}]", "Pcc", "cUc", MergeNone, "aarch64_sve_whilewr_b", [IsOverloadWhileRW, IsStreamingCompatible]>; 1899def SVWHILEWR_H : SInst<"svwhilewr[_{1}]", "Pcc", "sUsh", MergeNone, "aarch64_sve_whilewr_h", [IsOverloadWhileRW, IsStreamingCompatible]>; 1900def SVWHILEWR_S : SInst<"svwhilewr[_{1}]", "Pcc", "iUif", MergeNone, "aarch64_sve_whilewr_s", [IsOverloadWhileRW, IsStreamingCompatible]>; 1901def SVWHILEWR_D : SInst<"svwhilewr[_{1}]", "Pcc", "lUld", MergeNone, "aarch64_sve_whilewr_d", [IsOverloadWhileRW, IsStreamingCompatible]>; 1902} 1903 1904let TargetGuard = "sve2,bf16" in { 1905def SVWHILERW_H_BF16 : SInst<"svwhilerw[_{1}]", "Pcc", "b", MergeNone, "aarch64_sve_whilerw_h", [IsOverloadWhileRW, IsStreamingCompatible]>; 1906def SVWHILEWR_H_BF16 : SInst<"svwhilewr[_{1}]", "Pcc", "b", MergeNone, "aarch64_sve_whilewr_h", [IsOverloadWhileRW, IsStreamingCompatible]>; 1907} 1908 1909//////////////////////////////////////////////////////////////////////////////// 1910// SVE2 - Extended table lookup/permute 1911let TargetGuard = "sve2" in { 1912def SVTBL2 : SInst<"svtbl2[_{d}]", "d2u", "csilUcUsUiUlhfd", MergeNone, "", [IsStreamingCompatible]>; 1913def SVTBX : SInst<"svtbx[_{d}]", "dddu", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tbx", [IsStreamingCompatible]>; 1914} 1915 1916let TargetGuard = "sve2,bf16" in { 1917def SVTBL2_BF16 : SInst<"svtbl2[_{d}]", "d2u", "b", MergeNone, "", [IsStreamingCompatible]>; 1918def SVTBX_BF16 : SInst<"svtbx[_{d}]", "dddu", "b", MergeNone, "aarch64_sve_tbx", [IsStreamingCompatible]>; 1919} 1920 1921//////////////////////////////////////////////////////////////////////////////// 1922// SVE2 - Optional 1923 1924let TargetGuard = "sve2-aes" in { 1925def SVAESD : SInst<"svaesd[_{d}]", "ddd", "Uc", MergeNone, "aarch64_sve_aesd", [IsOverloadNone]>; 1926def SVAESIMC : SInst<"svaesimc[_{d}]", "dd", "Uc", MergeNone, "aarch64_sve_aesimc", [IsOverloadNone]>; 1927def SVAESE : SInst<"svaese[_{d}]", "ddd", "Uc", MergeNone, "aarch64_sve_aese", [IsOverloadNone]>; 1928def SVAESMC : SInst<"svaesmc[_{d}]", "dd", "Uc", MergeNone, "aarch64_sve_aesmc", [IsOverloadNone]>; 1929 1930def SVPMULLB_PAIR_U64 : SInst<"svpmullb_pair[_{d}]", "ddd", "Ul", MergeNone, "aarch64_sve_pmullb_pair">; 1931def SVPMULLB_PAIR_N_U64 : SInst<"svpmullb_pair[_n_{d}]", "dda", "Ul", MergeNone, "aarch64_sve_pmullb_pair">; 1932 1933def SVPMULLT_PAIR_U64 : SInst<"svpmullt_pair[_{d}]", "ddd", "Ul", MergeNone, "aarch64_sve_pmullt_pair">; 1934def SVPMULLT_PAIR_N_U64 : SInst<"svpmullt_pair[_n_{d}]", "dda", "Ul", MergeNone, "aarch64_sve_pmullt_pair">; 1935} 1936 1937let TargetGuard = "sve2-sha3" in { 1938def SVRAX1 : SInst<"svrax1[_{d}]", "ddd", "lUl", MergeNone, "aarch64_sve_rax1", [IsOverloadNone]>; 1939} 1940 1941let TargetGuard = "sve2-sm4" in { 1942def SVSM4E : SInst<"svsm4e[_{d}]", "ddd", "Ui", MergeNone, "aarch64_sve_sm4e", [IsOverloadNone]>; 1943def SVSM4EKEY : SInst<"svsm4ekey[_{d}]", "ddd", "Ui", MergeNone, "aarch64_sve_sm4ekey", [IsOverloadNone]>; 1944} 1945 1946let TargetGuard = "sve2-bitperm" in { 1947def SVBDEP : SInst<"svbdep[_{d}]", "ddd", "UcUsUiUl", MergeNone, "aarch64_sve_bdep_x">; 1948def SVBDEP_N : SInst<"svbdep[_n_{d}]", "dda", "UcUsUiUl", MergeNone, "aarch64_sve_bdep_x">; 1949def SVBEXT : SInst<"svbext[_{d}]", "ddd", "UcUsUiUl", MergeNone, "aarch64_sve_bext_x">; 1950def SVBEXT_N : SInst<"svbext[_n_{d}]", "dda", "UcUsUiUl", MergeNone, "aarch64_sve_bext_x">; 1951def SVBGRP : SInst<"svbgrp[_{d}]", "ddd", "UcUsUiUl", MergeNone, "aarch64_sve_bgrp_x">; 1952def SVBGRP_N : SInst<"svbgrp[_n_{d}]", "dda", "UcUsUiUl", MergeNone, "aarch64_sve_bgrp_x">; 1953} 1954 1955let TargetGuard = "sve2p1|sme" in { 1956def SVPSEL_B : SInst<"svpsel_lane_b8", "PPPm", "Pc", MergeNone, "", [IsStreamingOrSVE2p1], []>; 1957def SVPSEL_H : SInst<"svpsel_lane_b16", "PPPm", "Ps", MergeNone, "", [IsStreamingOrSVE2p1], []>; 1958def SVPSEL_S : SInst<"svpsel_lane_b32", "PPPm", "Pi", MergeNone, "", [IsStreamingOrSVE2p1], []>; 1959def SVPSEL_D : SInst<"svpsel_lane_b64", "PPPm", "Pl", MergeNone, "", [IsStreamingOrSVE2p1], []>; 1960} 1961 1962// Standalone sve2.1 builtins 1963let TargetGuard = "sve2p1" in { 1964def SVORQV : SInst<"svorqv[_{d}]", "{Pd", "csilUcUsUiUl", MergeNone, "aarch64_sve_orqv", [IsReductionQV]>; 1965def SVEORQV : SInst<"sveorqv[_{d}]", "{Pd", "csilUcUsUiUl", MergeNone, "aarch64_sve_eorqv", [IsReductionQV]>; 1966def SVADDQV : SInst<"svaddqv[_{d}]", "{Pd", "hfdcsilUcUsUiUl", MergeNone, "aarch64_sve_addqv", [IsReductionQV]>; 1967def SVANDQV : SInst<"svandqv[_{d}]", "{Pd", "csilUcUsUiUl", MergeNone, "aarch64_sve_andqv", [IsReductionQV]>; 1968def SVSMAXQV : SInst<"svmaxqv[_{d}]", "{Pd", "csil", MergeNone, "aarch64_sve_smaxqv", [IsReductionQV]>; 1969def SVUMAXQV : SInst<"svmaxqv[_{d}]", "{Pd", "UcUsUiUl", MergeNone, "aarch64_sve_umaxqv", [IsReductionQV]>; 1970def SVSMINQV : SInst<"svminqv[_{d}]", "{Pd", "csil", MergeNone, "aarch64_sve_sminqv", [IsReductionQV]>; 1971def SVUMINQV : SInst<"svminqv[_{d}]", "{Pd", "UcUsUiUl", MergeNone, "aarch64_sve_uminqv", [IsReductionQV]>; 1972 1973def SVFMAXNMQV: SInst<"svmaxnmqv[_{d}]", "{Pd", "hfd", MergeNone, "aarch64_sve_fmaxnmqv", [IsReductionQV]>; 1974def SVFMINNMQV: SInst<"svminnmqv[_{d}]", "{Pd", "hfd", MergeNone, "aarch64_sve_fminnmqv", [IsReductionQV]>; 1975def SVFMAXQV: SInst<"svmaxqv[_{d}]", "{Pd", "hfd", MergeNone, "aarch64_sve_fmaxqv", [IsReductionQV]>; 1976def SVFMINQV: SInst<"svminqv[_{d}]", "{Pd", "hfd", MergeNone, "aarch64_sve_fminqv", [IsReductionQV]>; 1977} 1978 1979let TargetGuard = "sve2p1|sme2" in { 1980def SVPEXT_SINGLE : SInst<"svpext_lane_{d}", "P}i", "QcQsQiQl", MergeNone, "aarch64_sve_pext", [IsStreamingOrSVE2p1], [ImmCheck<1, ImmCheck0_3>]>; 1981def SVPEXT_X2 : SInst<"svpext_lane_{d}_x2", "2.P}i", "QcQsQiQl", MergeNone, "aarch64_sve_pext_x2", [IsStreamingOrSVE2p1], [ImmCheck<1, ImmCheck0_1>]>; 1982 1983def SVPSEL_COUNT_ALIAS_B : SInst<"svpsel_lane_c8", "}}Pm", "Pc", MergeNone, "", [IsStreamingOrSVE2p1], []>; 1984def SVPSEL_COUNT_ALIAS_H : SInst<"svpsel_lane_c16", "}}Pm", "Ps", MergeNone, "", [IsStreamingOrSVE2p1], []>; 1985def SVPSEL_COUNT_ALIAS_S : SInst<"svpsel_lane_c32", "}}Pm", "Pi", MergeNone, "", [IsStreamingOrSVE2p1], []>; 1986def SVPSEL_COUNT_ALIAS_D : SInst<"svpsel_lane_c64", "}}Pm", "Pl", MergeNone, "", [IsStreamingOrSVE2p1], []>; 1987 1988def SVWHILEGE_COUNT : SInst<"svwhilege_{d}[_{1}]", "}lli", "QcQsQiQl", MergeNone, "aarch64_sve_whilege_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<2, ImmCheck2_4_Mul2>]>; 1989def SVWHILEGT_COUNT : SInst<"svwhilegt_{d}[_{1}]", "}lli", "QcQsQiQl", MergeNone, "aarch64_sve_whilegt_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<2, ImmCheck2_4_Mul2>]>; 1990def SVWHILELE_COUNT : SInst<"svwhilele_{d}[_{1}]", "}lli", "QcQsQiQl", MergeNone, "aarch64_sve_whilele_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<2, ImmCheck2_4_Mul2>]>; 1991def SVWHILELT_COUNT : SInst<"svwhilelt_{d}[_{1}]", "}lli", "QcQsQiQl", MergeNone, "aarch64_sve_whilelt_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<2, ImmCheck2_4_Mul2>]>; 1992def SVWHILELO_COUNT : SInst<"svwhilelt_{d}[_{1}]", "}nni", "QcQsQiQl", MergeNone, "aarch64_sve_whilelo_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<2, ImmCheck2_4_Mul2>]>; 1993def SVWHILELS_COUNT : SInst<"svwhilele_{d}[_{1}]", "}nni", "QcQsQiQl", MergeNone, "aarch64_sve_whilels_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<2, ImmCheck2_4_Mul2>]>; 1994def SVWHILEHI_COUNT : SInst<"svwhilegt_{d}[_{1}]", "}nni", "QcQsQiQl", MergeNone, "aarch64_sve_whilehi_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<2, ImmCheck2_4_Mul2>]>; 1995def SVWHILEHS_COUNT : SInst<"svwhilege_{d}[_{1}]", "}nni", "QcQsQiQl", MergeNone, "aarch64_sve_whilehs_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<2, ImmCheck2_4_Mul2>]>; 1996} 1997 1998multiclass MultiVecLoad<string i> { 1999 def SV # NAME # B_X2 : MInst<"sv" # i # "[_{2}]_x2", "2}c", "cUc", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2000 def SV # NAME # H_X2 : MInst<"sv" # i # "[_{2}]_x2", "2}c", "sUshb", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2001 def SV # NAME # W_X2 : MInst<"sv" # i # "[_{2}]_x2", "2}c", "iUif", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2002 def SV # NAME # D_X2 : MInst<"sv" # i # "[_{2}]_x2", "2}c", "lUld", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2003 def SV # NAME # B_X4 : MInst<"sv" # i # "[_{2}]_x4", "4}c", "cUc", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2004 def SV # NAME # H_X4 : MInst<"sv" # i # "[_{2}]_x4", "4}c", "sUshb", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2005 def SV # NAME # W_X4 : MInst<"sv" # i # "[_{2}]_x4", "4}c", "iUif", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2006 def SV # NAME # D_X4 : MInst<"sv" # i # "[_{2}]_x4", "4}c", "lUld", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2007 2008 def SV # NAME # B_VNUM_X2 : MInst<"sv" # i # "_vnum" # "[_{2}]_x2", "2}cl", "cUc", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2009 def SV # NAME # H_VNUM_X2 : MInst<"sv" # i # "_vnum" # "[_{2}]_x2", "2}cl", "sUshb", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2010 def SV # NAME # W_VNUM_X2 : MInst<"sv" # i # "_vnum" # "[_{2}]_x2", "2}cl", "iUif", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2011 def SV # NAME # D_VNUM_X2 : MInst<"sv" # i # "_vnum" # "[_{2}]_x2", "2}cl", "lUld", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2012 def SV # NAME # B_VNUM_X4 : MInst<"sv" # i # "_vnum" # "[_{2}]_x4", "4}cl", "cUc", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2013 def SV # NAME # H_VNUM_X4 : MInst<"sv" # i # "_vnum" # "[_{2}]_x4", "4}cl", "sUshb", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2014 def SV # NAME # W_VNUM_X4 : MInst<"sv" # i # "_vnum" # "[_{2}]_x4", "4}cl", "iUif", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2015 def SV # NAME # D_VNUM_X4 : MInst<"sv" # i # "_vnum" # "[_{2}]_x4", "4}cl", "lUld", [IsStructLoad, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2016} 2017 2018let TargetGuard = "sve2p1|sme2" in { 2019 defm LD1 : MultiVecLoad<"ld1">; 2020 defm LDNT1 : MultiVecLoad<"ldnt1">; 2021} 2022 2023multiclass MultiVecStore<string i> { 2024 def SV # NAME # B_X2 : MInst<"sv" # i # "[_{2}_x2]", "v}p2", "cUc", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2025 def SV # NAME # H_X2 : MInst<"sv" # i # "[_{2}_x2]", "v}p2", "sUshb", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2026 def SV # NAME # W_X2 : MInst<"sv" # i # "[_{2}_x2]", "v}p2", "iUif", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2027 def SV # NAME # D_X2 : MInst<"sv" # i # "[_{2}_x2]", "v}p2", "lUld", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2028 def SV # NAME # B_X4 : MInst<"sv" # i # "[_{2}_x4]", "v}p4", "cUc", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2029 def SV # NAME # H_X4 : MInst<"sv" # i # "[_{2}_x4]", "v}p4", "sUshb", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2030 def SV # NAME # W_X4 : MInst<"sv" # i # "[_{2}_x4]", "v}p4", "iUif", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2031 def SV # NAME # D_X4 : MInst<"sv" # i # "[_{2}_x4]", "v}p4", "lUld", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2032 2033 def SV # NAME # B_VNUM_X2 : MInst<"sv" # i # "_vnum" # "[_{2}_x2]", "v}pl2", "cUc", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2034 def SV # NAME # H_VNUM_X2 : MInst<"sv" # i # "_vnum" # "[_{2}_x2]", "v}pl2", "sUshb", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2035 def SV # NAME # W_VNUM_X2 : MInst<"sv" # i # "_vnum" # "[_{2}_x2]", "v}pl2", "iUif", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2036 def SV # NAME # D_VNUM_X2 : MInst<"sv" # i # "_vnum" # "[_{2}_x2]", "v}pl2", "lUld", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x2">; 2037 def SV # NAME # B_VNUM_X4 : MInst<"sv" # i # "_vnum" # "[_{2}_x4]", "v}pl4", "cUc", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2038 def SV # NAME # H_VNUM_X4 : MInst<"sv" # i # "_vnum" # "[_{2}_x4]", "v}pl4", "sUshb", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2039 def SV # NAME # W_VNUM_X4 : MInst<"sv" # i # "_vnum" # "[_{2}_x4]", "v}pl4", "iUif", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2040 def SV # NAME # D_VNUM_X4 : MInst<"sv" # i # "_vnum" # "[_{2}_x4]", "v}pl4", "lUld", [IsStructStore, IsStreamingOrSVE2p1], MemEltTyDefault, "aarch64_sve_" # i # "_pn_x4">; 2041} 2042 2043let TargetGuard = "sve2p1|sme2" in { 2044 defm ST1 : MultiVecStore<"st1">; 2045 defm STNT1 : MultiVecStore<"stnt1">; 2046} 2047 2048let TargetGuard = "sve2p1|sme2" in { 2049def SVDOT_X2_S : SInst<"svdot[_{d}_{2}]", "ddhh", "i", MergeNone, "aarch64_sve_sdot_x2", [IsStreamingOrSVE2p1], []>; 2050def SVDOT_X2_U : SInst<"svdot[_{d}_{2}]", "ddhh", "Ui", MergeNone, "aarch64_sve_udot_x2", [IsStreamingOrSVE2p1], []>; 2051def SVDOT_X2_F : SInst<"svdot[_{d}_{2}]", "ddhh", "f", MergeNone, "aarch64_sve_fdot_x2", [IsStreamingOrSVE2p1], []>; 2052def SVDOT_LANE_X2_S : SInst<"svdot_lane[_{d}_{2}]", "ddhhi", "i", MergeNone, "aarch64_sve_sdot_lane_x2", [IsStreamingOrSVE2p1], [ImmCheck<3, ImmCheck0_3>]>; 2053def SVDOT_LANE_X2_U : SInst<"svdot_lane[_{d}_{2}]", "ddhhi", "Ui", MergeNone, "aarch64_sve_udot_lane_x2", [IsStreamingOrSVE2p1], [ImmCheck<3, ImmCheck0_3>]>; 2054def SVDOT_LANE_X2_F : SInst<"svdot_lane[_{d}_{2}]", "ddhhi", "f", MergeNone, "aarch64_sve_fdot_lane_x2", [IsStreamingOrSVE2p1], [ImmCheck<3, ImmCheck0_3>]>; 2055} 2056 2057let TargetGuard = "sve2p1|sme2" in { 2058def SVSCLAMP : SInst<"svclamp[_{d}]", "dddd", "csil", MergeNone, "aarch64_sve_sclamp", [IsStreamingOrSVE2p1], []>; 2059def SVUCLAMP : SInst<"svclamp[_{d}]", "dddd", "UcUsUiUl", MergeNone, "aarch64_sve_uclamp", [IsStreamingOrSVE2p1], []>; 2060 2061defm SVREVD : SInstZPZ<"svrevd", "csilUcUsUiUlbhfd", "aarch64_sve_revd">; 2062} 2063 2064let TargetGuard = "sve2p1|sme2" in { 2065 def SVPTRUE_COUNT : SInst<"svptrue_{d}", "}v", "QcQsQiQl", MergeNone, "aarch64_sve_ptrue_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], []>; 2066 2067 def SVPFALSE_COUNT_ALIAS : SInst<"svpfalse_c", "}v", "", MergeNone, "", [IsOverloadNone, IsStreamingOrSVE2p1]>; 2068 2069 def SVFCLAMP : SInst<"svclamp[_{d}]", "dddd", "hfd", MergeNone, "aarch64_sve_fclamp", [IsStreamingOrSVE2p1], []>; 2070 def SVCNTP_COUNT : SInst<"svcntp_{d}", "n}i", "QcQsQiQl", MergeNone, "aarch64_sve_cntp_{d}", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<1, ImmCheck2_4_Mul2>]>; 2071} 2072 2073let TargetGuard = "(sve2|sme2),b16b16" in { 2074defm SVMUL_BF : SInstZPZZ<"svmul", "b", "aarch64_sve_fmul", "aarch64_sve_fmul_u", [IsStreamingCompatible]>; 2075defm SVADD_BF : SInstZPZZ<"svadd", "b", "aarch64_sve_fadd", "aarch64_sve_fadd_u", [IsStreamingCompatible]>; 2076defm SVSUB_BF : SInstZPZZ<"svsub", "b", "aarch64_sve_fsub", "aarch64_sve_fsub_u", [IsStreamingCompatible]>; 2077defm SVMAXNM_BF : SInstZPZZ<"svmaxnm","b", "aarch64_sve_fmaxnm", "aarch64_sve_fmaxnm_u", [IsStreamingCompatible]>; 2078defm SVMINNM_BF : SInstZPZZ<"svminnm","b", "aarch64_sve_fminnm", "aarch64_sve_fminnm_u", [IsStreamingCompatible]>; 2079defm SVMAX_BF : SInstZPZZ<"svmax", "b", "aarch64_sve_fmax", "aarch64_sve_fmax_u", [IsStreamingCompatible]>; 2080defm SVMIN_BF : SInstZPZZ<"svmin", "b", "aarch64_sve_fmin", "aarch64_sve_fmin_u", [IsStreamingCompatible]>; 2081defm SVMLA_BF : SInstZPZZZ<"svmla", "b", "aarch64_sve_fmla", "aarch64_sve_fmla_u", [IsStreamingCompatible]>; 2082defm SVMLS_BF : SInstZPZZZ<"svmls", "b", "aarch64_sve_fmls", "aarch64_sve_fmls_u", [IsStreamingCompatible]>; 2083def SVMLA_LANE_BF : SInst<"svmla_lane[_{d}]", "ddddi", "b", MergeNone, "aarch64_sve_fmla_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 2084def SVMLS_LANE_BF : SInst<"svmls_lane[_{d}]", "ddddi", "b", MergeNone, "aarch64_sve_fmls_lane", [IsStreamingCompatible], [ImmCheck<3, ImmCheckLaneIndex, 2>]>; 2085def SVMUL_LANE_BF : SInst<"svmul_lane[_{d}]", "dddi", "b", MergeNone, "aarch64_sve_fmul_lane", [IsStreamingCompatible], [ImmCheck<2, ImmCheckLaneIndex, 1>]>; 2086def SVFCLAMP_BF : SInst<"svclamp[_{d}]", "dddd", "b", MergeNone, "aarch64_sve_fclamp", [IsStreamingCompatible], []>; 2087} 2088 2089// SME2 2090 2091// SME intrinsics which operate only on vectors and do not require ZA should be added here, 2092// as they could possibly become SVE instructions in the future. 2093 2094multiclass MinMaxIntr<string i, string zm, string mul, string t> { 2095 def SVS # NAME : SInst<"sv" # i # "[" # zm # "_{d}_" # mul # "]", t, "csil", MergeNone, "aarch64_sve_s" # i # zm # "_" # mul, [IsStreaming], []>; 2096 def SVU # NAME : SInst<"sv" # i # "[" # zm # "_{d}_" # mul # "]", t, "UcUsUiUl", MergeNone, "aarch64_sve_u" # i # zm # "_" # mul, [IsStreaming], []>; 2097 def SVF # NAME : SInst<"sv" # i # "[" # zm # "_{d}_" # mul # "]", t, "hfd", MergeNone, "aarch64_sve_f" # i # zm # "_" # mul, [IsStreaming], []>; 2098} 2099 2100let TargetGuard = "sme2" in { 2101// == SMAX / UMAX / FMAX == 2102 defm MAX_SINGLE_X2 : MinMaxIntr<"max", "_single", "x2", "22d">; 2103 defm MAX_MULTI_X2 : MinMaxIntr<"max", "", "x2", "222">; 2104 defm MAX_SINGLE_X4 : MinMaxIntr<"max", "_single", "x4", "44d">; 2105 defm MAX_MULTI_X4 : MinMaxIntr<"max", "", "x4", "444">; 2106 2107// == SMIN / UMIN / FMIN == 2108 defm MIN_SINGLE_X2 : MinMaxIntr<"min", "_single", "x2", "22d">; 2109 defm MIN_MULTI_X2 : MinMaxIntr<"min", "", "x2", "222">; 2110 defm MIN_SINGLE_X4 : MinMaxIntr<"min", "_single", "x4", "44d">; 2111 defm MIN_MULTI_X4 : MinMaxIntr<"min", "", "x4", "444">; 2112} 2113 2114multiclass SInstMinMaxByVector<string name> { 2115 def NAME # _SINGLE_X2 : SInst<"sv" # name # "nm[_single_{d}_x2]", "22d", "hfd", MergeNone, "aarch64_sve_f" # name # "nm_single_x2", [IsStreaming], []>; 2116 def NAME # _SINGLE_X4 : SInst<"sv" # name # "nm[_single_{d}_x4]", "44d", "hfd", MergeNone, "aarch64_sve_f" # name # "nm_single_x4", [IsStreaming], []>; 2117 2118 def NAME # _X2 : SInst<"sv" # name # "nm[_{d}_x2]", "222", "hfd", MergeNone, "aarch64_sve_f" # name # "nm_x2", [IsStreaming], []>; 2119 def NAME # _X4 : SInst<"sv" # name # "nm[_{d}_x4]", "444", "hfd", MergeNone, "aarch64_sve_f" # name # "nm_x4", [IsStreaming], []>; 2120} 2121 2122let TargetGuard = "sme2" in { 2123// == FMINNM / FMAXNM == 2124 defm SVMINNM : SInstMinMaxByVector<"min">; 2125 defm SVMAXNM : SInstMinMaxByVector<"max">; 2126} 2127 2128let TargetGuard = "sme2" in { 2129 // FRINTA / FRINTM / FRINTN / FRINTP 2130 def SVRINTA_X2 : SInst<"svrinta[_{d}_x2]", "22", "f", MergeNone, "aarch64_sve_frinta_x2", [IsStreaming], []>; 2131 def SVRINTA_X4 : SInst<"svrinta[_{d}_x4]", "44", "f", MergeNone, "aarch64_sve_frinta_x4", [IsStreaming], []>; 2132 2133 def SVRINTM_X2 : SInst<"svrintm[_{d}_x2]", "22", "f", MergeNone, "aarch64_sve_frintm_x2", [IsStreaming], []>; 2134 def SVRINTM_X4 : SInst<"svrintm[_{d}_x4]", "44", "f", MergeNone, "aarch64_sve_frintm_x4", [IsStreaming], []>; 2135 2136 def SVRINTN_X2 : SInst<"svrintn[_{d}_x2]", "22", "f", MergeNone, "aarch64_sve_frintn_x2", [IsStreaming], []>; 2137 def SVRINTN_X4 : SInst<"svrintn[_{d}_x4]", "44", "f", MergeNone, "aarch64_sve_frintn_x4", [IsStreaming], []>; 2138 2139 def SVRINTP_X2 : SInst<"svrintp[_{d}_x2]", "22", "f", MergeNone, "aarch64_sve_frintp_x2", [IsStreaming], []>; 2140 def SVRINTP_X4 : SInst<"svrintp[_{d}_x4]", "44", "f", MergeNone, "aarch64_sve_frintp_x4", [IsStreaming], []>; 2141} 2142 2143let TargetGuard = "sme2" in { 2144 def SVSCLAMP_X2 : SInst<"svclamp[_single_{d}_x2]", "22dd", "csil", MergeNone, "aarch64_sve_sclamp_single_x2", [IsStreaming], []>; 2145 def SVUCLAMP_X2 : SInst<"svclamp[_single_{d}_x2]", "22dd", "UcUsUiUl", MergeNone, "aarch64_sve_uclamp_single_x2", [IsStreaming], []>; 2146 def SVFCLAMP_X2 : SInst<"svclamp[_single_{d}_x2]", "22dd", "hfd", MergeNone, "aarch64_sve_fclamp_single_x2", [IsStreaming], []>; 2147 2148 def SVSCLAMP_X4 : SInst<"svclamp[_single_{d}_x4]", "44dd", "csil", MergeNone, "aarch64_sve_sclamp_single_x4", [IsStreaming], []>; 2149 def SVUCLAMP_X4 : SInst<"svclamp[_single_{d}_x4]", "44dd", "UcUsUiUl", MergeNone, "aarch64_sve_uclamp_single_x4", [IsStreaming], []>; 2150 def SVFCLAMP_X4 : SInst<"svclamp[_single_{d}_x4]", "44dd", "hfd", MergeNone, "aarch64_sve_fclamp_single_x4", [IsStreaming], []>; 2151} 2152 2153let TargetGuard = "sme2" in { 2154// == ADD (vectors) == 2155 def SVADD_SINGLE_X2 : SInst<"svadd[_single_{d}_x2]", "22d", "cUcsUsiUilUl", MergeNone, "aarch64_sve_add_single_x2", [IsStreaming], []>; 2156 def SVADD_SINGLE_X4 : SInst<"svadd[_single_{d}_x4]", "44d", "cUcsUsiUilUl", MergeNone, "aarch64_sve_add_single_x4", [IsStreaming], []>; 2157 2158 // 2-way and 4-way selects 2159 def SVSEL_X2 : SInst<"svsel[_{d}_x2]", "2}22", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_sel_x2", [IsStreaming], []>; 2160 def SVSEL_X4 : SInst<"svsel[_{d}_x4]", "4}44", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_sel_x4", [IsStreaming], []>; 2161 2162 // SRSHL / URSHL 2163 def SVSRSHL_SINGLE_X2 : SInst<"svrshl[_single_{d}_x2]", "22d", "csil", MergeNone, "aarch64_sve_srshl_single_x2", [IsStreaming], []>; 2164 def SVURSHL_SINGLE_X2 : SInst<"svrshl[_single_{d}_x2]", "22d", "UcUsUiUl", MergeNone, "aarch64_sve_urshl_single_x2", [IsStreaming], []>; 2165 def SVSRSHL_SINGLE_X4 : SInst<"svrshl[_single_{d}_x4]", "44d", "csil", MergeNone, "aarch64_sve_srshl_single_x4", [IsStreaming], []>; 2166 def SVURSHL_SINGLE_X4 : SInst<"svrshl[_single_{d}_x4]", "44d", "UcUsUiUl", MergeNone, "aarch64_sve_urshl_single_x4", [IsStreaming], []>; 2167 2168 def SVSRSHL_X2 : SInst<"svrshl[_{d}_x2]", "222", "csil", MergeNone, "aarch64_sve_srshl_x2", [IsStreaming], []>; 2169 def SVURSHL_X2 : SInst<"svrshl[_{d}_x2]", "222", "UcUsUiUl", MergeNone, "aarch64_sve_urshl_x2", [IsStreaming], []>; 2170 def SVSRSHL_X4 : SInst<"svrshl[_{d}_x4]", "444", "csil", MergeNone, "aarch64_sve_srshl_x4", [IsStreaming], []>; 2171 def SVURSHL_X4 : SInst<"svrshl[_{d}_x4]", "444", "UcUsUiUl", MergeNone, "aarch64_sve_urshl_x4", [IsStreaming], []>; 2172 2173 def SVQRSHRN_X4 : SInst<"svqrshrn[_n]_{0}[_{d}_x4]", "q4i", "il", MergeNone, "aarch64_sve_sqrshrn_x4", [IsStreaming], [ImmCheck<1, ImmCheckShiftRight, 0>]>; 2174 def SVUQRSHRN_X4 : SInst<"svqrshrn[_n]_{0}[_{d}_x4]", "b4i", "UiUl", MergeNone, "aarch64_sve_uqrshrn_x4", [IsStreaming], [ImmCheck<1, ImmCheckShiftRight, 0>]>; 2175 2176 // SQRSHR / UQRSHR 2177 def SVQRSHR_X2 : SInst<"svqrshr[_n]_{0}[_{d}_x2]", "h2i", "i", MergeNone, "aarch64_sve_sqrshr_x2", [IsStreaming], [ImmCheck<1, ImmCheck1_16>]>; 2178 def SVUQRSHR_X2 : SInst<"svqrshr[_n]_{0}[_{d}_x2]", "e2i", "Ui", MergeNone, "aarch64_sve_uqrshr_x2", [IsStreaming], [ImmCheck<1, ImmCheck1_16>]>; 2179 def SVQRSHR_X4 : SInst<"svqrshr[_n]_{0}[_{d}_x4]", "q4i", "il", MergeNone, "aarch64_sve_sqrshr_x4", [IsStreaming], [ImmCheck<1, ImmCheckShiftRight, 0>]>; 2180 def SVUQRSHR_X4 : SInst<"svqrshr[_n]_{0}[_{d}_x4]", "b4i", "UiUl", MergeNone, "aarch64_sve_uqrshr_x4", [IsStreaming], [ImmCheck<1, ImmCheckShiftRight, 0>]>; 2181 2182 // SQRSHRU 2183 def SVSQRSHRU_X2 : SInst<"svqrshru[_n]_{0}[_{d}_x2]", "e2i", "i", MergeNone, "aarch64_sve_sqrshru_x2", [IsStreaming], [ImmCheck<1, ImmCheck1_16>]>; 2184 def SVSQRSHRU_X4 : SInst<"svqrshru[_n]_{0}[_{d}_x4]", "b4i", "il", MergeNone, "aarch64_sve_sqrshru_x4", [IsStreaming], [ImmCheck<1, ImmCheckShiftRight, 0>]>; 2185 2186 def SVSQRSHRUN_X4 : SInst<"svqrshrun[_n]_{0}[_{d}_x4]", "b4i", "il", MergeNone, "aarch64_sve_sqrshrun_x4", [IsStreaming], [ImmCheck<1, ImmCheckShiftRight, 0>]>; 2187 2188 def REINTERPRET_SVBOOL_TO_SVCOUNT : Inst<"svreinterpret[_c]", "}P", "Pc", MergeNone, "", [IsStreamingCompatible], []>; 2189 def REINTERPRET_SVCOUNT_TO_SVBOOL : Inst<"svreinterpret[_b]", "P}", "Pc", MergeNone, "", [IsStreamingCompatible], []>; 2190 2191 // SQDMULH 2192 def SVSQDMULH_SINGLE_X2 : SInst<"svqdmulh[_single_{d}_x2]", "22d", "csil", MergeNone, "aarch64_sve_sqdmulh_single_vgx2", [IsStreaming], []>; 2193 def SVSQDMULH_SINGLE_X4 : SInst<"svqdmulh[_single_{d}_x4]", "44d", "csil", MergeNone, "aarch64_sve_sqdmulh_single_vgx4", [IsStreaming], []>; 2194 def SVSQDMULH_X2 : SInst<"svqdmulh[_{d}_x2]", "222", "csil", MergeNone, "aarch64_sve_sqdmulh_vgx2", [IsStreaming], []>; 2195 def SVSQDMULH_X4 : SInst<"svqdmulh[_{d}_x4]", "444", "csil", MergeNone, "aarch64_sve_sqdmulh_vgx4", [IsStreaming], []>; 2196} 2197 2198let TargetGuard = "sve2p1|sme2" in { 2199 // SQRSHRN / UQRSHRN 2200 def SVQRSHRN_X2 : SInst<"svqrshrn[_n]_{0}[_{d}_x2]", "h2i", "i", MergeNone, "aarch64_sve_sqrshrn_x2", [IsStreamingCompatible], [ImmCheck<1, ImmCheck1_16>]>; 2201 def SVUQRSHRN_X2 : SInst<"svqrshrn[_n]_{0}[_{d}_x2]", "e2i", "Ui", MergeNone, "aarch64_sve_uqrshrn_x2", [IsStreamingCompatible], [ImmCheck<1, ImmCheck1_16>]>; 2202 2203 // SQRSHRUN 2204 def SVSQRSHRUN_X2 : SInst<"svqrshrun[_n]_{0}[_{d}_x2]", "e2i", "i", MergeNone, "aarch64_sve_sqrshrun_x2", [IsStreamingCompatible], [ImmCheck<1, ImmCheck1_16>]>; 2205} 2206 2207let TargetGuard = "sve2p1" in { 2208 // ZIPQ1, ZIPQ2, UZPQ1, UZPQ2 2209 def SVZIPQ1 : SInst<"svzipq1[_{d}]", "ddd", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_zipq1", [], []>; 2210 def SVZIPQ2 : SInst<"svzipq2[_{d}]", "ddd", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_zipq2", [], []>; 2211 def SVUZPQ1 : SInst<"svuzpq1[_{d}]", "ddd", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_uzpq1", [], []>; 2212 def SVUZPQ2 : SInst<"svuzpq2[_{d}]", "ddd", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_uzpq2", [], []>; 2213 // TBLQ, TBXQ 2214 def SVTBLQ : SInst<"svtblq[_{d}]", "ddu", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_tblq">; 2215 def SVTBXQ : SInst<"svtbxq[_{d}]", "dddu", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_tbxq">; 2216 // EXTQ 2217 def EXTQ : SInst<"svextq[_{d}]", "dddk", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_extq", [], [ImmCheck<2, ImmCheck0_15>]>; 2218 // PMOV 2219 // Move to Pred 2220 multiclass PMOV_TO_PRED<string name, string types, string intrinsic, list<FlagType> flags=[], ImmCheckType immCh > { 2221 def _LANE : Inst<name # "_lane[_{d}]", "Pdi", types, MergeNone, intrinsic, flags, [ImmCheck<1, immCh>]>; 2222 def _LANE_ZERO : SInst<name # "[_{d}]", "Pd", types, MergeNone, intrinsic # "_zero", flags, []>; 2223 } 2224 defm SVPMOV_B_TO_PRED : PMOV_TO_PRED<"svpmov", "cUc", "aarch64_sve_pmov_to_pred_lane", [], ImmCheck0_0>; 2225 defm SVPMOV_H_TO_PRED : PMOV_TO_PRED<"svpmov", "sUs", "aarch64_sve_pmov_to_pred_lane", [], ImmCheck0_1>; 2226 defm SVPMOV_S_TO_PRED : PMOV_TO_PRED<"svpmov", "iUi", "aarch64_sve_pmov_to_pred_lane", [], ImmCheck0_3>; 2227 defm SVPMOV_D_TO_PRED : PMOV_TO_PRED<"svpmov", "lUl", "aarch64_sve_pmov_to_pred_lane", [], ImmCheck0_7>; 2228 2229 // Move to Vector 2230 multiclass PMOV_TO_VEC<string name, string types, string intrinsic, list<FlagType> flags=[], ImmCheckType immCh > { 2231 def _M : SInst<name # "_lane[_{d}]", "ddPi", types, MergeOp1, intrinsic # "_merging", flags, [ImmCheck<2, immCh>]>; 2232 def _Z : SInst<name # "_{d}_z", "dP", types, MergeNone, intrinsic # "_zeroing", flags, []>; 2233 } 2234 def SVPMOV_TO_VEC_LANE_B : SInst<"svpmov_{d}_z", "dP", "cUc", MergeNone, "aarch64_sve_pmov_to_vector_lane_zeroing", [], []>; 2235 defm SVPMOV_TO_VEC_LANE_H : PMOV_TO_VEC<"svpmov", "sUs", "aarch64_sve_pmov_to_vector_lane", [], ImmCheck1_1>; 2236 defm SVPMOV_TO_VEC_LANE_S : PMOV_TO_VEC<"svpmov", "iUi", "aarch64_sve_pmov_to_vector_lane", [], ImmCheck1_3>; 2237 defm SVPMOV_TO_VEC_LANE_D : PMOV_TO_VEC<"svpmov", "lUl", "aarch64_sve_pmov_to_vector_lane" ,[], ImmCheck1_7>; 2238} 2239 2240// 2241// Multi-vector convert to/from floating-point. 2242// 2243let TargetGuard = "sme2" in { 2244 def SVCVT_F16_X2 : SInst<"svcvt_f16[_f32_x2]", "e2", "f", MergeNone, "aarch64_sve_fcvt_x2", [IsStreaming],[]>; 2245 def SVCVT_BF16_X2 : SInst<"svcvt_bf16[_f32_x2]", "$2", "f", MergeNone, "aarch64_sve_bfcvt_x2", [IsOverloadNone, IsStreaming],[]>; 2246 2247 def SVCVT_F32_U32_X2 : SInst<"svcvt_{d}[_u32_x2]", "2.d2.u", "f", MergeNone, "aarch64_sve_ucvtf_x2", [IsStreaming, IsOverloadWhileOrMultiVecCvt], []>; 2248 def SVCVT_U32_F32_X2 : SInst<"svcvt_{d}[_f32_x2]", "2.d2.M", "Ui", MergeNone, "aarch64_sve_fcvtzu_x2", [IsStreaming, IsOverloadWhileOrMultiVecCvt], []>; 2249 def SVCVT_F32_S32_X2 : SInst<"svcvt_{d}[_s32_x2]", "2.d2.x", "f", MergeNone, "aarch64_sve_scvtf_x2", [IsStreaming, IsOverloadWhileOrMultiVecCvt], []>; 2250 def SVCVT_S32_F32_X2 : SInst<"svcvt_{d}[_f32_x2]", "2.d2.M", "i", MergeNone, "aarch64_sve_fcvtzs_x2", [IsStreaming, IsOverloadWhileOrMultiVecCvt], []>; 2251 2252 def SVCVT_F32_U32_X4 : SInst<"svcvt_{d}[_u32_x4]", "4.d4.u", "f", MergeNone, "aarch64_sve_ucvtf_x4", [IsStreaming, IsOverloadWhileOrMultiVecCvt], []>; 2253 def SVCVT_U32_F32_X4 : SInst<"svcvt_{d}[_f32_x4]", "4.d4.M", "Ui", MergeNone, "aarch64_sve_fcvtzu_x4", [IsStreaming, IsOverloadWhileOrMultiVecCvt], []>; 2254 def SVCVT_F32_S32_X4 : SInst<"svcvt_{d}[_s32_x4]", "4.d4.x", "f", MergeNone, "aarch64_sve_scvtf_x4", [IsStreaming, IsOverloadWhileOrMultiVecCvt], []>; 2255 def SVCVT_S32_F32_X4 : SInst<"svcvt_{d}[_f32_x4]", "4.d4.M", "i", MergeNone, "aarch64_sve_fcvtzs_x4", [IsStreaming, IsOverloadWhileOrMultiVecCvt], []>; 2256} 2257 2258// 2259// Multi-vector floating-point convert from single-precision to interleaved half-precision/BFloat16 2260// 2261let TargetGuard = "sme2" in { 2262 def SVCVTN_F16_X2 : SInst<"svcvtn_f16[_f32_x2]", "e2", "f", MergeNone, "aarch64_sve_fcvtn_x2", [IsStreaming],[]>; 2263 def SVCVTN_BF16_X2 : SInst<"svcvtn_bf16[_f32_x2]", "$2", "f", MergeNone, "aarch64_sve_bfcvtn_x2", [IsOverloadNone, IsStreaming],[]>; 2264} 2265 2266// 2267// Multi-vector saturating extract narrow 2268// 2269let TargetGuard = "sme2" in { 2270 def SVQCVT_S16_S32_X2 : SInst<"svqcvt_s16[_{d}_x2]", "h2.d", "i", MergeNone, "aarch64_sve_sqcvt_x2", [IsStreaming], []>; 2271 def SVQCVT_U16_U32_X2 : SInst<"svqcvt_u16[_{d}_x2]", "e2.d", "Ui", MergeNone, "aarch64_sve_uqcvt_x2", [IsStreaming], []>; 2272 def SVQCVT_U16_S32_X2 : SInst<"svqcvt_u16[_{d}_x2]", "e2.d", "i", MergeNone, "aarch64_sve_sqcvtu_x2", [IsStreaming], []>; 2273 2274 def SVQCVT_S8_S32_X4 : SInst<"svqcvt_s8[_{d}_x4]", "q4.d", "i", MergeNone, "aarch64_sve_sqcvt_x4", [IsStreaming], []>; 2275 def SVQCVT_U8_U32_X4 : SInst<"svqcvt_u8[_{d}_x4]", "b4.d", "Ui", MergeNone, "aarch64_sve_uqcvt_x4", [IsStreaming], []>; 2276 def SVQCVT_U8_S32_X4 : SInst<"svqcvt_u8[_{d}_x4]", "b4.d", "i", MergeNone, "aarch64_sve_sqcvtu_x4", [IsStreaming], []>; 2277 2278 def SVQCVT_S16_S64_X4 : SInst<"svqcvt_s16[_{d}_x4]", "q4.d", "l", MergeNone, "aarch64_sve_sqcvt_x4", [IsStreaming], []>; 2279 def SVQCVT_U16_U64_X4 : SInst<"svqcvt_u16[_{d}_x4]", "b4.d", "Ul", MergeNone, "aarch64_sve_uqcvt_x4", [IsStreaming], []>; 2280 def SVQCVT_U16_S64_X4 : SInst<"svqcvt_u16[_{d}_x4]", "b4.d", "l", MergeNone, "aarch64_sve_sqcvtu_x4", [IsStreaming], []>; 2281} 2282 2283// 2284// Multi-vector saturating extract narrow and interleave 2285// 2286let TargetGuard = "sme2|sve2p1" in { 2287 def SVQCVTN_S16_S32_X2 : SInst<"svqcvtn_s16[_{d}_x2]", "h2.d", "i", MergeNone, "aarch64_sve_sqcvtn_x2", [IsStreamingCompatible], []>; 2288 def SVQCVTN_U16_U32_X2 : SInst<"svqcvtn_u16[_{d}_x2]", "e2.d", "Ui", MergeNone, "aarch64_sve_uqcvtn_x2", [IsStreamingCompatible], []>; 2289 def SVQCVTN_U16_S32_X2 : SInst<"svqcvtn_u16[_{d}_x2]", "e2.d", "i", MergeNone, "aarch64_sve_sqcvtun_x2", [IsStreamingCompatible], []>; 2290} 2291 2292let TargetGuard = "sme2" in { 2293 def SVQCVTN_S8_S32_X4 : SInst<"svqcvtn_s8[_{d}_x4]", "q4.d", "i", MergeNone, "aarch64_sve_sqcvtn_x4", [IsStreaming], []>; 2294 def SVQCVTN_U8_U32_X4 : SInst<"svqcvtn_u8[_{d}_x4]", "b4.d", "Ui", MergeNone, "aarch64_sve_uqcvtn_x4", [IsStreaming], []>; 2295 def SVQCVTN_U8_S32_X4 : SInst<"svqcvtn_u8[_{d}_x4]", "b4.d", "i", MergeNone, "aarch64_sve_sqcvtun_x4", [IsStreaming], []>; 2296 2297 def SVQCVTN_S16_S64_X4 : SInst<"svqcvtn_s16[_{d}_x4]", "q4.d", "l", MergeNone, "aarch64_sve_sqcvtn_x4", [IsStreaming], []>; 2298 def SVQCVTN_U16_U64_X4 : SInst<"svqcvtn_u16[_{d}_x4]", "b4.d", "Ul", MergeNone, "aarch64_sve_uqcvtn_x4", [IsStreaming], []>; 2299 def SVQCVTN_U16_S64_X4 : SInst<"svqcvtn_u16[_{d}_x4]", "b4.d", "l", MergeNone, "aarch64_sve_sqcvtun_x4", [IsStreaming], []>; 2300} 2301 2302// 2303// Multi-vector zip/unzip 2304// 2305 2306let TargetGuard = "sme2" in { 2307 def SVZIP_X2 : SInst<"svzip[_{d}_x2]", "22", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_zip_x2", [IsStreaming], []>; 2308 def SVZIPQ_X2 : SInst<"svzipq[_{d}_x2]", "22", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_zipq_x2", [IsStreaming], []>; 2309 def SVZIP_X4 : SInst<"svzip[_{d}_x4]", "44", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_zip_x4", [IsStreaming], []>; 2310 def SVZIPQ_X4 : SInst<"svzipq[_{d}_x4]", "44", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_zipq_x4", [IsStreaming], []>; 2311 2312 def SVUZP_X2 : SInst<"svuzp[_{d}_x2]", "22", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_uzp_x2", [IsStreaming], []>; 2313 def SVUZPQ_X2 : SInst<"svuzpq[_{d}_x2]", "22", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_uzpq_x2", [IsStreaming], []>; 2314 def SVUZP_X4 : SInst<"svuzp[_{d}_x4]", "44", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_uzp_x4", [IsStreaming], []>; 2315 def SVUZPQ_X4 : SInst<"svuzpq[_{d}_x4]", "44", "cUcsUsiUilUlbhfd", MergeNone, "aarch64_sve_uzpq_x4", [IsStreaming], []>; 2316} 2317 2318// 2319// Multi-vector unpack 2320// 2321 2322let TargetGuard = "sme2" in { 2323 def SVSUNPK_X2 : SInst<"svunpk_{d}[_{1}_x2]", "2h", "sil", MergeNone, "aarch64_sve_sunpk_x2", [IsStreaming], []>; 2324 def SVUUNPK_X2 : SInst<"svunpk_{d}[_{1}_x2]", "2h", "UsUiUl", MergeNone, "aarch64_sve_uunpk_x2", [IsStreaming], []>; 2325 def SVSUNPK_X4 : SInst<"svunpk_{d}[_{3}_x4]", "42.h", "sil", MergeNone, "aarch64_sve_sunpk_x4", [IsStreaming], []>; 2326 def SVUUNPK_X4 : SInst<"svunpk_{d}[_{3}_x4]", "42.h", "UsUiUl", MergeNone, "aarch64_sve_uunpk_x4", [IsStreaming], []>; 2327} 2328 2329let TargetGuard = "sve2p1|sme2" in { 2330// == BFloat16 multiply-subtract == 2331 def SVBFMLSLB : SInst<"svbfmlslb[_{d}]", "dd$$", "f", MergeNone, "aarch64_sve_bfmlslb", [IsOverloadNone, IsStreamingOrSVE2p1], []>; 2332 def SVBFMLSLT : SInst<"svbfmlslt[_{d}]", "dd$$", "f", MergeNone, "aarch64_sve_bfmlslt", [IsOverloadNone, IsStreamingOrSVE2p1], []>; 2333 2334 def SVBFMLSLB_LANE : SInst<"svbfmlslb_lane[_{d}]", "dd$$i", "f", MergeNone, "aarch64_sve_bfmlslb_lane", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<3, ImmCheck0_7>]>; 2335 def SVBFMLSLT_LANE : SInst<"svbfmlslt_lane[_{d}]", "dd$$i", "f", MergeNone, "aarch64_sve_bfmlslt_lane", [IsOverloadNone, IsStreamingOrSVE2p1], [ImmCheck<3, ImmCheck0_7>]>; 2336} 2337