| //===-- RISCVInstrInfoZfh.td - RISC-V 'Zfh' instructions ---*- tablegen -*-===// |
| // |
| // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| // See https://llvm.org/LICENSE.txt for license information. |
| // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| // |
| //===----------------------------------------------------------------------===// |
| // |
| // This file describes the RISC-V instructions from the standard 'Zfh' |
| // half-precision floating-point extension, version 1.0. |
| // |
| //===----------------------------------------------------------------------===// |
| |
| //===----------------------------------------------------------------------===// |
| // RISC-V specific DAG Nodes. |
| //===----------------------------------------------------------------------===// |
| |
| def SDT_RISCVFMV_H_X |
| : SDTypeProfile<1, 1, [SDTCisVT<0, f16>, SDTCisVT<1, XLenVT>]>; |
| def SDT_RISCVFMV_X_EXTH |
| : SDTypeProfile<1, 1, [SDTCisVT<0, XLenVT>, SDTCisVT<1, f16>]>; |
| |
| def riscv_fmv_h_x |
| : SDNode<"RISCVISD::FMV_H_X", SDT_RISCVFMV_H_X>; |
| def riscv_fmv_x_anyexth |
| : SDNode<"RISCVISD::FMV_X_ANYEXTH", SDT_RISCVFMV_X_EXTH>; |
| def riscv_fmv_x_signexth |
| : SDNode<"RISCVISD::FMV_X_SIGNEXTH", SDT_RISCVFMV_X_EXTH>; |
| |
| //===----------------------------------------------------------------------===// |
| // Operand and SDNode transformation definitions. |
| //===----------------------------------------------------------------------===// |
| |
| // Zhinxmin and Zhinx |
| |
| def FPR16INX : RegisterOperand<GPRF16> { |
| let ParserMatchClass = GPRAsFPR; |
| let DecoderMethod = "DecodeGPRRegisterClass"; |
| } |
| |
| def ZfhExt : ExtInfo<0, [HasStdExtZfh]>; |
| def Zfh64Ext : ExtInfo<0, [HasStdExtZfh, IsRV64]>; |
| def ZfhminExt : ExtInfo<0, [HasStdExtZfhOrZfhmin]>; |
| def ZhinxExt : ExtInfo<1, [HasStdExtZhinx]>; |
| def ZhinxminExt : ExtInfo<1, [HasStdExtZhinxOrZhinxmin]>; |
| def Zhinx64Ext : ExtInfo<1, [HasStdExtZhinx, IsRV64]>; |
| |
| def ZfhminDExt : ExtInfo<0, [HasStdExtZfhOrZfhmin, HasStdExtD]>; |
| def ZhinxminZdinxExt : ExtInfo<1, [HasStdExtZhinxOrZhinxmin, HasStdExtZdinx]>; |
| |
| def H : ExtInfo_r<ZfhExt, FPR16>; |
| def H_INX : ExtInfo_r<ZhinxExt, FPR16INX>; |
| |
| def HH : ExtInfo_rr<ZfhExt, FPR16, FPR16>; |
| def HH_INX : ExtInfo_rr<ZhinxExt, FPR16INX, FPR16INX>; |
| def XH : ExtInfo_rr<ZfhExt, GPR, FPR16>; |
| def XH_INX : ExtInfo_rr<ZhinxExt, GPR, FPR16INX>; |
| def HX : ExtInfo_rr<ZfhExt, FPR16, GPR>; |
| def HX_INX : ExtInfo_rr<ZhinxExt, FPR16INX, GPR>; |
| def XH_64 : ExtInfo_rr<Zfh64Ext, GPR, FPR16>; |
| def HX_64 : ExtInfo_rr<Zfh64Ext, FPR16, GPR>; |
| def XH_INX_64 : ExtInfo_rr<Zhinx64Ext, GPR, FPR16INX>; |
| def HX_INX_64 : ExtInfo_rr<Zhinx64Ext, FPR16INX, GPR>; |
| def HFmin : ExtInfo_rr<ZfhminExt, FPR16, FPR32>; |
| def HF_INXmin : ExtInfo_rr<ZhinxminExt, FPR16INX, FPR32INX>; |
| def HF_INX : ExtInfo_rr<ZhinxExt, FPR16INX, FPR32INX>; |
| def FHmin : ExtInfo_rr<ZfhminExt, FPR32, FPR16>; |
| def FH_INXmin : ExtInfo_rr<ZhinxminExt, FPR32INX, FPR16INX>; |
| def FH_INX : ExtInfo_rr<ZhinxExt, FPR32INX, FPR16INX>; |
| def DHmin : ExtInfo_rr<ZfhminDExt, FPR64, FPR16>; |
| def DH_INXmin : ExtInfo_rr<ZhinxminZdinxExt, FPR64INX, FPR16INX>; |
| def HDmin : ExtInfo_rr<ZfhminDExt, FPR16, FPR64>; |
| def HD_INXmin : ExtInfo_rr<ZhinxminZdinxExt, FPR16INX, FPR64INX>; |
| |
| defvar HINX = [H, H_INX]; |
| defvar HHINX = [HH, HH_INX]; |
| defvar XHINX = [XH, XH_INX]; |
| defvar HXINX = [HX, HX_INX]; |
| defvar XHIN64X = [XH_64, XH_INX_64]; |
| defvar HXIN64X = [HX_64, HX_INX_64]; |
| defvar HFINXmin = [HFmin, HF_INXmin]; |
| defvar FHINXmin = [FHmin, FH_INXmin]; |
| defvar DHINXmin = [DHmin, DH_INXmin]; |
| defvar HDINXmin = [HDmin, HD_INXmin]; |
| |
| //===----------------------------------------------------------------------===// |
| // Instructions |
| //===----------------------------------------------------------------------===// |
| |
| let Predicates = [HasStdExtZfhOrZfhmin] in { |
| def FLH : FPLoad_r<0b001, "flh", FPR16, WriteFLD16>; |
| |
| // Operands for stores are in the order srcreg, base, offset rather than |
| // reflecting the order these fields are specified in the instruction |
| // encoding. |
| def FSH : FPStore_r<0b001, "fsh", FPR16, WriteFST16>; |
| } // Predicates = [HasStdExtZfhOrZfhmin] |
| |
| let SchedRW = [WriteFMA16, ReadFMA16, ReadFMA16, ReadFMA16] in { |
| defm FMADD_H : FPFMA_rrr_frm_m<OPC_MADD, 0b10, "fmadd.h", HINX>; |
| defm FMSUB_H : FPFMA_rrr_frm_m<OPC_MSUB, 0b10, "fmsub.h", HINX>; |
| defm FNMSUB_H : FPFMA_rrr_frm_m<OPC_NMSUB, 0b10, "fnmsub.h", HINX>; |
| defm FNMADD_H : FPFMA_rrr_frm_m<OPC_NMADD, 0b10, "fnmadd.h", HINX>; |
| } |
| |
| defm : FPFMADynFrmAlias_m<FMADD_H, "fmadd.h", HINX>; |
| defm : FPFMADynFrmAlias_m<FMSUB_H, "fmsub.h", HINX>; |
| defm : FPFMADynFrmAlias_m<FNMSUB_H, "fnmsub.h", HINX>; |
| defm : FPFMADynFrmAlias_m<FNMADD_H, "fnmadd.h", HINX>; |
| |
| let SchedRW = [WriteFAdd16, ReadFAdd16, ReadFAdd16] in { |
| defm FADD_H : FPALU_rr_frm_m<0b0000010, "fadd.h", HINX, /*Commutable*/1>; |
| defm FSUB_H : FPALU_rr_frm_m<0b0000110, "fsub.h", HINX>; |
| } |
| let SchedRW = [WriteFMul16, ReadFMul16, ReadFMul16] in |
| defm FMUL_H : FPALU_rr_frm_m<0b0001010, "fmul.h", HINX, /*Commutable*/1>; |
| |
| let SchedRW = [WriteFDiv16, ReadFDiv16, ReadFDiv16] in |
| defm FDIV_H : FPALU_rr_frm_m<0b0001110, "fdiv.h", HINX>; |
| |
| defm : FPALUDynFrmAlias_m<FADD_H, "fadd.h", HINX>; |
| defm : FPALUDynFrmAlias_m<FSUB_H, "fsub.h", HINX>; |
| defm : FPALUDynFrmAlias_m<FMUL_H, "fmul.h", HINX>; |
| defm : FPALUDynFrmAlias_m<FDIV_H, "fdiv.h", HINX>; |
| |
| defm FSQRT_H : FPUnaryOp_r_frm_m<0b0101110, 0b00000, HHINX, "fsqrt.h">, |
| Sched<[WriteFSqrt16, ReadFSqrt16]>; |
| defm : FPUnaryOpDynFrmAlias_m<FSQRT_H, "fsqrt.h", HHINX>; |
| |
| let SchedRW = [WriteFSGNJ16, ReadFSGNJ16, ReadFSGNJ16], |
| mayRaiseFPException = 0 in { |
| defm FSGNJ_H : FPALU_rr_m<0b0010010, 0b000, "fsgnj.h", HINX>; |
| defm FSGNJN_H : FPALU_rr_m<0b0010010, 0b001, "fsgnjn.h", HINX>; |
| defm FSGNJX_H : FPALU_rr_m<0b0010010, 0b010, "fsgnjx.h", HINX>; |
| } |
| |
| let SchedRW = [WriteFMinMax16, ReadFMinMax16, ReadFMinMax16] in { |
| defm FMIN_H : FPALU_rr_m<0b0010110, 0b000, "fmin.h", HINX, /*Commutable*/1>; |
| defm FMAX_H : FPALU_rr_m<0b0010110, 0b001, "fmax.h", HINX, /*Commutable*/1>; |
| } |
| |
| let IsSignExtendingOpW = 1 in |
| defm FCVT_W_H : FPUnaryOp_r_frm_m<0b1100010, 0b00000, XHINX, "fcvt.w.h">, |
| Sched<[WriteFCvtF16ToI32, ReadFCvtF16ToI32]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_W_H, "fcvt.w.h", XHINX>; |
| |
| let IsSignExtendingOpW = 1 in |
| defm FCVT_WU_H : FPUnaryOp_r_frm_m<0b1100010, 0b00001, XHINX, "fcvt.wu.h">, |
| Sched<[WriteFCvtF16ToI32, ReadFCvtF16ToI32]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_WU_H, "fcvt.wu.h", XHINX>; |
| |
| defm FCVT_H_W : FPUnaryOp_r_frm_m<0b1101010, 0b00000, HXINX, "fcvt.h.w">, |
| Sched<[WriteFCvtI32ToF16, ReadFCvtI32ToF16]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_H_W, "fcvt.h.w", HXINX>; |
| |
| defm FCVT_H_WU : FPUnaryOp_r_frm_m<0b1101010, 0b00001, HXINX, "fcvt.h.wu">, |
| Sched<[WriteFCvtI32ToF16, ReadFCvtI32ToF16]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_H_WU, "fcvt.h.wu", HXINX>; |
| |
| defm FCVT_H_S : FPUnaryOp_r_frm_m<0b0100010, 0b00000, HFINXmin, "fcvt.h.s">, |
| Sched<[WriteFCvtF32ToF16, ReadFCvtF32ToF16]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_H_S, "fcvt.h.s", HFINXmin>; |
| |
| defm FCVT_S_H : FPUnaryOp_r_m<0b0100000, 0b00010, 0b000, FHINXmin, "fcvt.s.h">, |
| Sched<[WriteFCvtF16ToF32, ReadFCvtF16ToF32]>; |
| |
| let Predicates = [HasStdExtZfhOrZfhmin] in { |
| let mayRaiseFPException = 0, IsSignExtendingOpW = 1 in |
| def FMV_X_H : FPUnaryOp_r<0b1110010, 0b00000, 0b000, GPR, FPR16, "fmv.x.h">, |
| Sched<[WriteFMovF16ToI16, ReadFMovF16ToI16]>; |
| |
| let mayRaiseFPException = 0 in |
| def FMV_H_X : FPUnaryOp_r<0b1111010, 0b00000, 0b000, FPR16, GPR, "fmv.h.x">, |
| Sched<[WriteFMovI16ToF16, ReadFMovI16ToF16]>; |
| } // Predicates = [HasStdExtZfhOrZfhmin] |
| |
| let SchedRW = [WriteFCmp16, ReadFCmp16, ReadFCmp16] in { |
| defm FEQ_H : FPCmp_rr_m<0b1010010, 0b010, "feq.h", HINX, /*Commutable*/1>; |
| defm FLT_H : FPCmp_rr_m<0b1010010, 0b001, "flt.h", HINX>; |
| defm FLE_H : FPCmp_rr_m<0b1010010, 0b000, "fle.h", HINX>; |
| } |
| |
| let mayRaiseFPException = 0 in |
| defm FCLASS_H : FPUnaryOp_r_m<0b1110010, 0b00000, 0b001, XHINX, "fclass.h">, |
| Sched<[WriteFClass16, ReadFClass16]>; |
| |
| defm FCVT_L_H : FPUnaryOp_r_frm_m<0b1100010, 0b00010, XHIN64X, "fcvt.l.h">, |
| Sched<[WriteFCvtF16ToI64, ReadFCvtF16ToI64]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_L_H, "fcvt.l.h", XHIN64X>; |
| |
| defm FCVT_LU_H : FPUnaryOp_r_frm_m<0b1100010, 0b00011, XHIN64X, "fcvt.lu.h">, |
| Sched<[WriteFCvtF16ToI64, ReadFCvtF16ToI64]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_LU_H, "fcvt.lu.h", XHIN64X>; |
| |
| defm FCVT_H_L : FPUnaryOp_r_frm_m<0b1101010, 0b00010, HXIN64X, "fcvt.h.l">, |
| Sched<[WriteFCvtI64ToF16, ReadFCvtI64ToF16]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_H_L, "fcvt.h.l", HXIN64X>; |
| |
| defm FCVT_H_LU : FPUnaryOp_r_frm_m<0b1101010, 0b00011, HXIN64X, "fcvt.h.lu">, |
| Sched<[WriteFCvtI64ToF16, ReadFCvtI64ToF16]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_H_LU, "fcvt.h.lu", HXIN64X>; |
| |
| defm FCVT_H_D : FPUnaryOp_r_frm_m<0b0100010, 0b00001, HDINXmin, "fcvt.h.d">, |
| Sched<[WriteFCvtF64ToF16, ReadFCvtF64ToF16]>; |
| defm : FPUnaryOpDynFrmAlias_m<FCVT_H_D, "fcvt.h.d", HDINXmin>; |
| |
| defm FCVT_D_H : FPUnaryOp_r_m<0b0100001, 0b00010, 0b000, DHINXmin, "fcvt.d.h">, |
| Sched<[WriteFCvtF16ToF64, ReadFCvtF16ToF64]>; |
| |
| //===----------------------------------------------------------------------===// |
| // Assembler Pseudo Instructions (User-Level ISA, Version 2.2, Chapter 20) |
| //===----------------------------------------------------------------------===// |
| |
| let Predicates = [HasStdExtZfhOrZfhmin] in { |
| def : InstAlias<"flh $rd, (${rs1})", (FLH FPR16:$rd, GPR:$rs1, 0), 0>; |
| def : InstAlias<"fsh $rs2, (${rs1})", (FSH FPR16:$rs2, GPR:$rs1, 0), 0>; |
| } // Predicates = [HasStdExtZfhOrZfhmin] |
| |
| let Predicates = [HasStdExtZfh] in { |
| def : InstAlias<"fmv.h $rd, $rs", (FSGNJ_H FPR16:$rd, FPR16:$rs, FPR16:$rs)>; |
| def : InstAlias<"fabs.h $rd, $rs", (FSGNJX_H FPR16:$rd, FPR16:$rs, FPR16:$rs)>; |
| def : InstAlias<"fneg.h $rd, $rs", (FSGNJN_H FPR16:$rd, FPR16:$rs, FPR16:$rs)>; |
| |
| // fgt.h/fge.h are recognised by the GNU assembler but the canonical |
| // flt.h/fle.h forms will always be printed. Therefore, set a zero weight. |
| def : InstAlias<"fgt.h $rd, $rs, $rt", |
| (FLT_H GPR:$rd, FPR16:$rt, FPR16:$rs), 0>; |
| def : InstAlias<"fge.h $rd, $rs, $rt", |
| (FLE_H GPR:$rd, FPR16:$rt, FPR16:$rs), 0>; |
| } // Predicates = [HasStdExtZfh] |
| |
| let Predicates = [HasStdExtZfhOrZfhmin] in { |
| def PseudoFLH : PseudoFloatLoad<"flh", FPR16>; |
| def PseudoFSH : PseudoStore<"fsh", FPR16>; |
| let usesCustomInserter = 1 in { |
| def PseudoQuietFLE_H : PseudoQuietFCMP<FPR16>; |
| def PseudoQuietFLT_H : PseudoQuietFCMP<FPR16>; |
| } |
| } // Predicates = [HasStdExtZfhOrZfhmin] |
| |
| let Predicates = [HasStdExtZhinx] in { |
| def : InstAlias<"fmv.h $rd, $rs", (FSGNJ_H_INX FPR16INX:$rd, FPR16INX:$rs, FPR16INX:$rs)>; |
| def : InstAlias<"fabs.h $rd, $rs", (FSGNJX_H_INX FPR16INX:$rd, FPR16INX:$rs, FPR16INX:$rs)>; |
| def : InstAlias<"fneg.h $rd, $rs", (FSGNJN_H_INX FPR16INX:$rd, FPR16INX:$rs, FPR16INX:$rs)>; |
| |
| def : InstAlias<"fgt.h $rd, $rs, $rt", |
| (FLT_H_INX GPR:$rd, FPR16INX:$rt, FPR16INX:$rs), 0>; |
| def : InstAlias<"fge.h $rd, $rs, $rt", |
| (FLE_H_INX GPR:$rd, FPR16INX:$rt, FPR16INX:$rs), 0>; |
| } // Predicates = [HasStdExtZhinx] |
| |
| //===----------------------------------------------------------------------===// |
| // Pseudo-instructions and codegen patterns |
| //===----------------------------------------------------------------------===// |
| |
| let Predicates = [HasStdExtZfh] in { |
| |
| // Floating point constant -0.0 |
| def : Pat<(f16 (fpimmneg0)), (FSGNJN_H (FMV_H_X X0), (FMV_H_X X0))>; |
| |
| /// Float conversion operations |
| |
| // [u]int32<->float conversion patterns must be gated on IsRV32 or IsRV64, so |
| // are defined later. |
| |
| /// Float arithmetic operations |
| |
| def : PatFprFprDynFrm<any_fadd, FADD_H, FPR16>; |
| def : PatFprFprDynFrm<any_fsub, FSUB_H, FPR16>; |
| def : PatFprFprDynFrm<any_fmul, FMUL_H, FPR16>; |
| def : PatFprFprDynFrm<any_fdiv, FDIV_H, FPR16>; |
| |
| def : Pat<(any_fsqrt FPR16:$rs1), (FSQRT_H FPR16:$rs1, 0b111)>; |
| |
| def : Pat<(fneg FPR16:$rs1), (FSGNJN_H $rs1, $rs1)>; |
| def : Pat<(fabs FPR16:$rs1), (FSGNJX_H $rs1, $rs1)>; |
| |
| def : PatFprFpr<fcopysign, FSGNJ_H, FPR16>; |
| def : Pat<(fcopysign FPR16:$rs1, (fneg FPR16:$rs2)), (FSGNJN_H $rs1, $rs2)>; |
| def : Pat<(fcopysign FPR16:$rs1, FPR32:$rs2), |
| (FSGNJ_H $rs1, (FCVT_H_S $rs2, 0b111))>; |
| |
| // fmadd: rs1 * rs2 + rs3 |
| def : Pat<(any_fma FPR16:$rs1, FPR16:$rs2, FPR16:$rs3), |
| (FMADD_H $rs1, $rs2, $rs3, 0b111)>; |
| |
| // fmsub: rs1 * rs2 - rs3 |
| def : Pat<(any_fma FPR16:$rs1, FPR16:$rs2, (fneg FPR16:$rs3)), |
| (FMSUB_H FPR16:$rs1, FPR16:$rs2, FPR16:$rs3, 0b111)>; |
| |
| // fnmsub: -rs1 * rs2 + rs3 |
| def : Pat<(any_fma (fneg FPR16:$rs1), FPR16:$rs2, FPR16:$rs3), |
| (FNMSUB_H FPR16:$rs1, FPR16:$rs2, FPR16:$rs3, 0b111)>; |
| |
| // fnmadd: -rs1 * rs2 - rs3 |
| def : Pat<(any_fma (fneg FPR16:$rs1), FPR16:$rs2, (fneg FPR16:$rs3)), |
| (FNMADD_H FPR16:$rs1, FPR16:$rs2, FPR16:$rs3, 0b111)>; |
| |
| // fnmadd: -(rs1 * rs2 + rs3) (the nsz flag on the FMA) |
| def : Pat<(fneg (any_fma_nsz FPR16:$rs1, FPR16:$rs2, FPR16:$rs3)), |
| (FNMADD_H FPR16:$rs1, FPR16:$rs2, FPR16:$rs3, 0b111)>; |
| |
| // The ratified 20191213 ISA spec defines fmin and fmax in a way that matches |
| // LLVM's fminnum and fmaxnum |
| // <https://github.com/riscv/riscv-isa-manual/commit/cd20cee7efd9bac7c5aa127ec3b451749d2b3cce>. |
| def : PatFprFpr<fminnum, FMIN_H, FPR16>; |
| def : PatFprFpr<fmaxnum, FMAX_H, FPR16>; |
| |
| /// Setcc |
| // FIXME: SETEQ/SETLT/SETLE imply nonans, can we pick better instructions for |
| // strict versions of those. |
| |
| // Match non-signaling FEQ_D |
| def : PatSetCC<FPR16, any_fsetcc, SETEQ, FEQ_H>; |
| def : PatSetCC<FPR16, any_fsetcc, SETOEQ, FEQ_H>; |
| def : PatSetCC<FPR16, strict_fsetcc, SETLT, PseudoQuietFLT_H>; |
| def : PatSetCC<FPR16, strict_fsetcc, SETOLT, PseudoQuietFLT_H>; |
| def : PatSetCC<FPR16, strict_fsetcc, SETLE, PseudoQuietFLE_H>; |
| def : PatSetCC<FPR16, strict_fsetcc, SETOLE, PseudoQuietFLE_H>; |
| |
| // Match signaling FEQ_H |
| def : Pat<(strict_fsetccs FPR16:$rs1, FPR16:$rs2, SETEQ), |
| (AND (FLE_H $rs1, $rs2), |
| (FLE_H $rs2, $rs1))>; |
| def : Pat<(strict_fsetccs FPR16:$rs1, FPR16:$rs2, SETOEQ), |
| (AND (FLE_H $rs1, $rs2), |
| (FLE_H $rs2, $rs1))>; |
| // If both operands are the same, use a single FLE. |
| def : Pat<(strict_fsetccs FPR16:$rs1, FPR16:$rs1, SETEQ), |
| (FLE_H $rs1, $rs1)>; |
| def : Pat<(strict_fsetccs FPR16:$rs1, FPR16:$rs1, SETOEQ), |
| (FLE_H $rs1, $rs1)>; |
| |
| def : PatSetCC<FPR16, any_fsetccs, SETLT, FLT_H>; |
| def : PatSetCC<FPR16, any_fsetccs, SETOLT, FLT_H>; |
| def : PatSetCC<FPR16, any_fsetccs, SETLE, FLE_H>; |
| def : PatSetCC<FPR16, any_fsetccs, SETOLE, FLE_H>; |
| |
| defm Select_FPR16 : SelectCC_GPR_rrirr<FPR16>; |
| |
| def PseudoFROUND_H : PseudoFROUND<FPR16>; |
| } // Predicates = [HasStdExtZfh] |
| |
| let Predicates = [HasStdExtZfhOrZfhmin] in { |
| /// Loads |
| |
| defm : LdPat<load, FLH, f16>; |
| |
| /// Stores |
| |
| defm : StPat<store, FSH, FPR16, f16>; |
| |
| /// Floating point constant +0.0 |
| def : Pat<(f16 (fpimm0)), (FMV_H_X X0)>; |
| |
| /// Float conversion operations |
| |
| // f32 -> f16, f16 -> f32 |
| def : Pat<(any_fpround FPR32:$rs1), (FCVT_H_S FPR32:$rs1, 0b111)>; |
| def : Pat<(any_fpextend FPR16:$rs1), (FCVT_S_H FPR16:$rs1)>; |
| |
| // Moves (no conversion) |
| def : Pat<(riscv_fmv_h_x GPR:$src), (FMV_H_X GPR:$src)>; |
| def : Pat<(riscv_fmv_x_anyexth FPR16:$src), (FMV_X_H FPR16:$src)>; |
| def : Pat<(riscv_fmv_x_signexth FPR16:$src), (FMV_X_H FPR16:$src)>; |
| |
| def : Pat<(fcopysign FPR32:$rs1, FPR16:$rs2), (FSGNJ_S $rs1, (FCVT_S_H $rs2))>; |
| } // Predicates = [HasStdExtZfhOrZfhmin] |
| |
| let Predicates = [HasStdExtZfh, IsRV32] in { |
| // half->[u]int. Round-to-zero must be used. |
| def : Pat<(i32 (any_fp_to_sint FPR16:$rs1)), (FCVT_W_H $rs1, 0b001)>; |
| def : Pat<(i32 (any_fp_to_uint FPR16:$rs1)), (FCVT_WU_H $rs1, 0b001)>; |
| |
| // Saturating half->[u]int32. |
| def : Pat<(i32 (riscv_fcvt_x FPR16:$rs1, timm:$frm)), (FCVT_W_H $rs1, timm:$frm)>; |
| def : Pat<(i32 (riscv_fcvt_xu FPR16:$rs1, timm:$frm)), (FCVT_WU_H $rs1, timm:$frm)>; |
| |
| // half->int32 with current rounding mode. |
| def : Pat<(i32 (any_lrint FPR16:$rs1)), (FCVT_W_H $rs1, 0b111)>; |
| |
| // half->int32 rounded to nearest with ties rounded away from zero. |
| def : Pat<(i32 (any_lround FPR16:$rs1)), (FCVT_W_H $rs1, 0b100)>; |
| |
| // [u]int->half. Match GCC and default to using dynamic rounding mode. |
| def : Pat<(any_sint_to_fp (i32 GPR:$rs1)), (FCVT_H_W $rs1, 0b111)>; |
| def : Pat<(any_uint_to_fp (i32 GPR:$rs1)), (FCVT_H_WU $rs1, 0b111)>; |
| } // Predicates = [HasStdExtZfh, IsRV32] |
| |
| let Predicates = [HasStdExtZfh, IsRV64] in { |
| // Use target specific isd nodes to help us remember the result is sign |
| // extended. Matching sext_inreg+fptoui/fptosi may cause the conversion to be |
| // duplicated if it has another user that didn't need the sign_extend. |
| def : Pat<(riscv_any_fcvt_w_rv64 FPR16:$rs1, timm:$frm), (FCVT_W_H $rs1, timm:$frm)>; |
| def : Pat<(riscv_any_fcvt_wu_rv64 FPR16:$rs1, timm:$frm), (FCVT_WU_H $rs1, timm:$frm)>; |
| |
| // half->[u]int64. Round-to-zero must be used. |
| def : Pat<(i64 (any_fp_to_sint FPR16:$rs1)), (FCVT_L_H $rs1, 0b001)>; |
| def : Pat<(i64 (any_fp_to_uint FPR16:$rs1)), (FCVT_LU_H $rs1, 0b001)>; |
| |
| // Saturating half->[u]int64. |
| def : Pat<(i64 (riscv_fcvt_x FPR16:$rs1, timm:$frm)), (FCVT_L_H $rs1, timm:$frm)>; |
| def : Pat<(i64 (riscv_fcvt_xu FPR16:$rs1, timm:$frm)), (FCVT_LU_H $rs1, timm:$frm)>; |
| |
| // half->int64 with current rounding mode. |
| def : Pat<(i64 (any_lrint FPR16:$rs1)), (FCVT_L_H $rs1, 0b111)>; |
| def : Pat<(i64 (any_llrint FPR16:$rs1)), (FCVT_L_H $rs1, 0b111)>; |
| |
| // half->int64 rounded to nearest with ties rounded away from zero. |
| def : Pat<(i64 (any_lround FPR16:$rs1)), (FCVT_L_H $rs1, 0b100)>; |
| def : Pat<(i64 (any_llround FPR16:$rs1)), (FCVT_L_H $rs1, 0b100)>; |
| |
| // [u]int->fp. Match GCC and default to using dynamic rounding mode. |
| def : Pat<(any_sint_to_fp (i64 (sexti32 (i64 GPR:$rs1)))), (FCVT_H_W $rs1, 0b111)>; |
| def : Pat<(any_uint_to_fp (i64 (zexti32 (i64 GPR:$rs1)))), (FCVT_H_WU $rs1, 0b111)>; |
| def : Pat<(any_sint_to_fp (i64 GPR:$rs1)), (FCVT_H_L $rs1, 0b111)>; |
| def : Pat<(any_uint_to_fp (i64 GPR:$rs1)), (FCVT_H_LU $rs1, 0b111)>; |
| } // Predicates = [HasStdExtZfh, IsRV64] |
| |
| let Predicates = [HasStdExtZfhOrZfhmin, HasStdExtD] in { |
| /// Float conversion operations |
| // f64 -> f16, f16 -> f64 |
| def : Pat<(any_fpround FPR64:$rs1), (FCVT_H_D FPR64:$rs1, 0b111)>; |
| def : Pat<(any_fpextend FPR16:$rs1), (FCVT_D_H FPR16:$rs1)>; |
| |
| /// Float arithmetic operations |
| def : Pat<(fcopysign FPR16:$rs1, FPR64:$rs2), |
| (FSGNJ_H $rs1, (FCVT_H_D $rs2, 0b111))>; |
| def : Pat<(fcopysign FPR64:$rs1, FPR16:$rs2), (FSGNJ_D $rs1, (FCVT_D_H $rs2))>; |
| } // Predicates = [HasStdExtZfhOrZfhmin, HasStdExtD] |
| |
| let Predicates = [HasStdExtZfhmin, NoStdExtZfh] in { |
| // Floating point constant -0.0 |
| def : Pat<(f16 (fpimmneg0)), (FCVT_H_S (FSGNJN_S (FMV_W_X X0), (FMV_W_X X0)), 0b111)>; |
| } // Predicates = [HasStdExtZfhmin, NoStdExtZfh] |
| |
| let Predicates = [HasStdExtZfhmin, NoStdExtZfh, IsRV32] in { |
| // half->[u]int. Round-to-zero must be used. |
| def : Pat<(i32 (any_fp_to_sint FPR16:$rs1)), (FCVT_W_S (FCVT_S_H $rs1), 0b001)>; |
| def : Pat<(i32 (any_fp_to_uint FPR16:$rs1)), (FCVT_WU_S (FCVT_S_H $rs1), 0b001)>; |
| |
| // half->int32 with current rounding mode. |
| def : Pat<(i32 (any_lrint FPR16:$rs1)), (FCVT_W_S (FCVT_S_H $rs1), 0b111)>; |
| |
| // half->int32 rounded to nearest with ties rounded away from zero. |
| def : Pat<(i32 (any_lround FPR16:$rs1)), (FCVT_W_S (FCVT_S_H $rs1), 0b100)>; |
| |
| // [u]int->half. Match GCC and default to using dynamic rounding mode. |
| def : Pat<(any_sint_to_fp (i32 GPR:$rs1)), (FCVT_H_S (FCVT_S_W $rs1, 0b111), 0b111)>; |
| def : Pat<(any_uint_to_fp (i32 GPR:$rs1)), (FCVT_H_S (FCVT_S_WU $rs1, 0b111), 0b111)>; |
| } // Predicates = [HasStdExtZfhmin, NoStdExtZfh, IsRV32] |
| |
| let Predicates = [HasStdExtZfhmin, NoStdExtZfh, IsRV64] in { |
| // half->[u]int64. Round-to-zero must be used. |
| def : Pat<(i64 (any_fp_to_sint FPR16:$rs1)), (FCVT_L_S (FCVT_S_H $rs1), 0b001)>; |
| def : Pat<(i64 (any_fp_to_uint FPR16:$rs1)), (FCVT_LU_S (FCVT_S_H $rs1), 0b001)>; |
| |
| // half->int64 with current rounding mode. |
| def : Pat<(i64 (any_lrint FPR16:$rs1)), (FCVT_L_S (FCVT_S_H $rs1), 0b111)>; |
| def : Pat<(i64 (any_llrint FPR16:$rs1)), (FCVT_L_S (FCVT_S_H $rs1), 0b111)>; |
| |
| // half->int64 rounded to nearest with ties rounded away from zero. |
| def : Pat<(i64 (any_lround FPR16:$rs1)), (FCVT_L_S (FCVT_S_H $rs1), 0b100)>; |
| def : Pat<(i64 (any_llround FPR16:$rs1)), (FCVT_L_S (FCVT_S_H $rs1), 0b100)>; |
| |
| // [u]int->fp. Match GCC and default to using dynamic rounding mode. |
| def : Pat<(any_sint_to_fp (i64 GPR:$rs1)), (FCVT_H_S (FCVT_S_L $rs1, 0b111), 0b111)>; |
| def : Pat<(any_uint_to_fp (i64 GPR:$rs1)), (FCVT_H_S (FCVT_S_LU $rs1, 0b111), 0b111)>; |
| } // Predicates = [HasStdExtZfhmin, NoStdExtZfh, IsRV64] |