maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/lib/Target/RISCV/RISCVInstrInfoA.td

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

392 lines
17 KiB
TableGen
Raw Normal View History

[RISCV] MC layer support for the standard RV32A instruction set extension llvm-svn: 317791
2017-11-09 23:00:03 +08:00
//===-- RISCVInstrInfoA.td - RISC-V 'A' instructions -------*- tablegen -*-===//
//
Update the file headers across all of the LLVM projects in the monorepo to reflect the new license. We understand that people may be surprised that we're moving the header entirely to discuss the new license. We checked this carefully with the Foundation's lawyer and we believe this is the correct approach. Essentially, all code in the project is now made available by the LLVM project under our new license, so you will see that the license headers include that license only. Some of our contributors have contributed code under our old license, and accordingly, we have retained a copy of our old license notice in the top-level files in each project and repository. llvm-svn: 351636
2019-01-19 16:50:56 +08:00
// 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
[RISCV] MC layer support for the standard RV32A instruction set extension llvm-svn: 317791
2017-11-09 23:00:03 +08:00
//
//===----------------------------------------------------------------------===//
//
// This file describes the RISC-V instructions from the standard 'A', Atomic
// Instructions extension.
//
//===----------------------------------------------------------------------===//
[RISCV] Add Custom Parser for Atomic Memory Operands Summary: GCC Accepts both (reg) and 0(reg) for atomic instruction memory operands. These instructions do not allow for an offset in their encoding, so in the latter case, the 0 is silently dropped. Due to how we have structured the RISCVAsmParser, the easiest way to add support for parsing this offset is to add a custom AsmOperand and parser. This parser drops all the parens, and just keeps the register. This commit also adds a custom printer for these operands, which matches the GCC canonical printer, printing both `(a0)` and `0(a0)` as `(a0)`. Reviewers: asb, lewis-revill Reviewed By: asb Subscribers: s.egerton, hiraditya, rbar, johnrusso, simoncook, apazos, sabuasal, niosHD, kito-cheng, shiva0217, jrtc27, MaskRay, zzheng, edward-jones, rogfer01, MartinMosbeck, brucehoult, the_o, rkruppe, jfb, PkmX, jocewei, psnobl, benna, Jim, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65205 llvm-svn: 367553
2019-08-01 20:42:31 +08:00
//===----------------------------------------------------------------------===//
// Operand and SDNode transformation definitions.
//===----------------------------------------------------------------------===//
// A parse method for (${gpr}) or 0(${gpr}), where the 0 is be silently ignored.
// Used for GNU as Compatibility.
def AtomicMemOpOperand : AsmOperandClass {
let Name = "AtomicMemOpOperand";
let RenderMethod = "addRegOperands";
[RISCV] Check register class for AMO memory operands Summary: AMO memory operands use a custom parser in order to accept both (reg) and 0(reg). However, the validation predicate used for these operands was only checking that they were registers, and not the register class, so non-GPRs (such as FPRs) were also accepted. Thus, fix this by making the predicate check that they are GPRs. Reviewers: asb, lenary Reviewed By: asb, lenary Subscribers: hiraditya, rbar, johnrusso, simoncook, sabuasal, niosHD, kito-cheng, shiva0217, MaskRay, zzheng, edward-jones, rogfer01, MartinMosbeck, brucehoult, the_o, rkruppe, PkmX, jocewei, psnobl, benna, Jim, s.egerton, pzheng, sameer.abuasal, apazos, luismarques, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D72471
2020-01-13 08:50:37 +08:00
let PredicateMethod = "isGPR";
[RISCV] Add Custom Parser for Atomic Memory Operands Summary: GCC Accepts both (reg) and 0(reg) for atomic instruction memory operands. These instructions do not allow for an offset in their encoding, so in the latter case, the 0 is silently dropped. Due to how we have structured the RISCVAsmParser, the easiest way to add support for parsing this offset is to add a custom AsmOperand and parser. This parser drops all the parens, and just keeps the register. This commit also adds a custom printer for these operands, which matches the GCC canonical printer, printing both `(a0)` and `0(a0)` as `(a0)`. Reviewers: asb, lewis-revill Reviewed By: asb Subscribers: s.egerton, hiraditya, rbar, johnrusso, simoncook, apazos, sabuasal, niosHD, kito-cheng, shiva0217, jrtc27, MaskRay, zzheng, edward-jones, rogfer01, MartinMosbeck, brucehoult, the_o, rkruppe, jfb, PkmX, jocewei, psnobl, benna, Jim, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65205 llvm-svn: 367553
2019-08-01 20:42:31 +08:00
let ParserMethod = "parseAtomicMemOp";
}
def GPRMemAtomic : RegisterOperand<GPR> {
let ParserMatchClass = AtomicMemOpOperand;
let PrintMethod = "printAtomicMemOp";
}
[RISCV] MC layer support for the standard RV32A instruction set extension llvm-svn: 317791
2017-11-09 23:00:03 +08:00
//===----------------------------------------------------------------------===//
// Instruction class templates
//===----------------------------------------------------------------------===//
let hasSideEffects = 0, mayLoad = 1, mayStore = 0 in
class LR_r<bit aq, bit rl, bits<3> funct3, string opcodestr>
: RVInstRAtomic<0b00010, aq, rl, funct3, OPC_AMO,
[RISCV] Add Custom Parser for Atomic Memory Operands Summary: GCC Accepts both (reg) and 0(reg) for atomic instruction memory operands. These instructions do not allow for an offset in their encoding, so in the latter case, the 0 is silently dropped. Due to how we have structured the RISCVAsmParser, the easiest way to add support for parsing this offset is to add a custom AsmOperand and parser. This parser drops all the parens, and just keeps the register. This commit also adds a custom printer for these operands, which matches the GCC canonical printer, printing both `(a0)` and `0(a0)` as `(a0)`. Reviewers: asb, lewis-revill Reviewed By: asb Subscribers: s.egerton, hiraditya, rbar, johnrusso, simoncook, apazos, sabuasal, niosHD, kito-cheng, shiva0217, jrtc27, MaskRay, zzheng, edward-jones, rogfer01, MartinMosbeck, brucehoult, the_o, rkruppe, jfb, PkmX, jocewei, psnobl, benna, Jim, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65205 llvm-svn: 367553
2019-08-01 20:42:31 +08:00
(outs GPR:$rd), (ins GPRMemAtomic:$rs1),
opcodestr, "$rd, $rs1"> {
[RISCV] MC layer support for the standard RV32A instruction set extension llvm-svn: 317791
2017-11-09 23:00:03 +08:00
let rs2 = 0;
}
multiclass LR_r_aq_rl<bits<3> funct3, string opcodestr> {
def "" : LR_r<0, 0, funct3, opcodestr>;
def _AQ : LR_r<1, 0, funct3, opcodestr # ".aq">;
def _RL : LR_r<0, 1, funct3, opcodestr # ".rl">;
def _AQ_RL : LR_r<1, 1, funct3, opcodestr # ".aqrl">;
}
let hasSideEffects = 0, mayLoad = 1, mayStore = 1 in
class AMO_rr<bits<5> funct5, bit aq, bit rl, bits<3> funct3, string opcodestr>
: RVInstRAtomic<funct5, aq, rl, funct3, OPC_AMO,
[RISCV] Add Custom Parser for Atomic Memory Operands Summary: GCC Accepts both (reg) and 0(reg) for atomic instruction memory operands. These instructions do not allow for an offset in their encoding, so in the latter case, the 0 is silently dropped. Due to how we have structured the RISCVAsmParser, the easiest way to add support for parsing this offset is to add a custom AsmOperand and parser. This parser drops all the parens, and just keeps the register. This commit also adds a custom printer for these operands, which matches the GCC canonical printer, printing both `(a0)` and `0(a0)` as `(a0)`. Reviewers: asb, lewis-revill Reviewed By: asb Subscribers: s.egerton, hiraditya, rbar, johnrusso, simoncook, apazos, sabuasal, niosHD, kito-cheng, shiva0217, jrtc27, MaskRay, zzheng, edward-jones, rogfer01, MartinMosbeck, brucehoult, the_o, rkruppe, jfb, PkmX, jocewei, psnobl, benna, Jim, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65205 llvm-svn: 367553
2019-08-01 20:42:31 +08:00
(outs GPR:$rd), (ins GPRMemAtomic:$rs1, GPR:$rs2),
opcodestr, "$rd, $rs2, $rs1">;
[RISCV] MC layer support for the standard RV32A instruction set extension llvm-svn: 317791
2017-11-09 23:00:03 +08:00
multiclass AMO_rr_aq_rl<bits<5> funct5, bits<3> funct3, string opcodestr> {
def "" : AMO_rr<funct5, 0, 0, funct3, opcodestr>;
def _AQ : AMO_rr<funct5, 1, 0, funct3, opcodestr # ".aq">;
def _RL : AMO_rr<funct5, 0, 1, funct3, opcodestr # ".rl">;
def _AQ_RL : AMO_rr<funct5, 1, 1, funct3, opcodestr # ".aqrl">;
}
[RISCV] atomic_store_nn have a different layout to regular store We cannot directy reuse the patterns of StPat because for some reason the store DAG node and the atomic_store_nn DAG nodes put the ptr and the value in different positions. Currently we attempt to store the address to an address formed by the value. Differential Revision: https://reviews.llvm.org/D51217 llvm-svn: 340722
2018-08-27 15:08:18 +08:00
multiclass AtomicStPat<PatFrag StoreOp, RVInst Inst, RegisterClass StTy> {
def : Pat<(StoreOp GPR:$rs1, StTy:$rs2), (Inst StTy:$rs2, GPR:$rs1, 0)>;
def : Pat<(StoreOp AddrFI:$rs1, StTy:$rs2), (Inst StTy:$rs2, AddrFI:$rs1, 0)>;
def : Pat<(StoreOp (add GPR:$rs1, simm12:$imm12), StTy:$rs2),
(Inst StTy:$rs2, GPR:$rs1, simm12:$imm12)>;
def : Pat<(StoreOp (add AddrFI:$rs1, simm12:$imm12), StTy:$rs2),
(Inst StTy:$rs2, AddrFI:$rs1, simm12:$imm12)>;
def : Pat<(StoreOp (IsOrAdd AddrFI:$rs1, simm12:$imm12), StTy:$rs2),
(Inst StTy:$rs2, AddrFI:$rs1, simm12:$imm12)>;
}
[RISCV] MC layer support for the standard RV32A instruction set extension llvm-svn: 317791
2017-11-09 23:00:03 +08:00
//===----------------------------------------------------------------------===//
// Instructions
//===----------------------------------------------------------------------===//
let Predicates = [HasStdExtA] in {
[RISCV] Scheduler description for the Rocket core Pipeline scheduler model for the RISC-V Rocket micro-architecture using the MIScheduler interface. Support for both 32 and 64-bit Rocket cores is implemented. Differential revision: https://reviews.llvm.org/D68685
2020-01-24 04:51:04 +08:00
defm LR_W : LR_r_aq_rl<0b010, "lr.w">, Sched<[WriteAtomicLDW, ReadAtomicLDW]>;
defm SC_W : AMO_rr_aq_rl<0b00011, 0b010, "sc.w">,
Sched<[WriteAtomicSTW, ReadAtomicSTW, ReadAtomicSTW]>;
defm AMOSWAP_W : AMO_rr_aq_rl<0b00001, 0b010, "amoswap.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
defm AMOADD_W : AMO_rr_aq_rl<0b00000, 0b010, "amoadd.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
defm AMOXOR_W : AMO_rr_aq_rl<0b00100, 0b010, "amoxor.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
defm AMOAND_W : AMO_rr_aq_rl<0b01100, 0b010, "amoand.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
defm AMOOR_W : AMO_rr_aq_rl<0b01000, 0b010, "amoor.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
defm AMOMIN_W : AMO_rr_aq_rl<0b10000, 0b010, "amomin.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
defm AMOMAX_W : AMO_rr_aq_rl<0b10100, 0b010, "amomax.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
defm AMOMINU_W : AMO_rr_aq_rl<0b11000, 0b010, "amominu.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
defm AMOMAXU_W : AMO_rr_aq_rl<0b11100, 0b010, "amomaxu.w">,
Sched<[WriteAtomicW, ReadAtomicWA, ReadAtomicWD]>;
[RISCV] MC layer support for the standard RV32A instruction set extension llvm-svn: 317791
2017-11-09 23:00:03 +08:00
} // Predicates = [HasStdExtA]
[RISCV] MC layer support for the standard RV64A instruction set extension llvm-svn: 320027
2017-12-07 18:59:12 +08:00
let Predicates = [HasStdExtA, IsRV64] in {
[RISCV] Scheduler description for the Rocket core Pipeline scheduler model for the RISC-V Rocket micro-architecture using the MIScheduler interface. Support for both 32 and 64-bit Rocket cores is implemented. Differential revision: https://reviews.llvm.org/D68685
2020-01-24 04:51:04 +08:00
defm LR_D : LR_r_aq_rl<0b011, "lr.d">, Sched<[WriteAtomicLDD, ReadAtomicLDD]>;
defm SC_D : AMO_rr_aq_rl<0b00011, 0b011, "sc.d">,
Sched<[WriteAtomicSTD, ReadAtomicSTD, ReadAtomicSTD]>;
defm AMOSWAP_D : AMO_rr_aq_rl<0b00001, 0b011, "amoswap.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
defm AMOADD_D : AMO_rr_aq_rl<0b00000, 0b011, "amoadd.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
defm AMOXOR_D : AMO_rr_aq_rl<0b00100, 0b011, "amoxor.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
defm AMOAND_D : AMO_rr_aq_rl<0b01100, 0b011, "amoand.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
defm AMOOR_D : AMO_rr_aq_rl<0b01000, 0b011, "amoor.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
defm AMOMIN_D : AMO_rr_aq_rl<0b10000, 0b011, "amomin.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
defm AMOMAX_D : AMO_rr_aq_rl<0b10100, 0b011, "amomax.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
defm AMOMINU_D : AMO_rr_aq_rl<0b11000, 0b011, "amominu.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
defm AMOMAXU_D : AMO_rr_aq_rl<0b11100, 0b011, "amomaxu.d">,
Sched<[WriteAtomicD, ReadAtomicDA, ReadAtomicDD]>;
[RISCV][NFC] Fix HasStedExtA -> HasStdExtA typo in comment Differential Revision: https://reviews.llvm.org/D64011 Patch by James Clarke. llvm-svn: 366173
2019-07-16 11:54:08 +08:00
} // Predicates = [HasStdExtA, IsRV64]
[RISCV] Add codegen support for atomic load/stores with RV32A Fences are inserted according to table A.6 in the current draft of version 2.3 of the RISC-V Instruction Set Manual, which incorporates the memory model changes and definitions contributed by the RISC-V Memory Consistency Model task group. Instruction selection failures will now occur for 8/16/32-bit atomicrmw and cmpxchg operations when targeting RV32IA until lowering for these operations is added in a follow-on patch. Differential Revision: https://reviews.llvm.org/D47589 llvm-svn: 334591
2018-06-13 20:04:51 +08:00
//===----------------------------------------------------------------------===//
// Pseudo-instructions and codegen patterns
//===----------------------------------------------------------------------===//
let Predicates = [HasStdExtA] in {
/// Atomic loads and stores
// Fences will be inserted for atomic load/stores according to the logic in
// RISCVTargetLowering::{emitLeadingFence,emitTrailingFence}.
defm : LdPat<atomic_load_8, LB>;
defm : LdPat<atomic_load_16, LH>;
defm : LdPat<atomic_load_32, LW>;
[RISCV] atomic_store_nn have a different layout to regular store We cannot directy reuse the patterns of StPat because for some reason the store DAG node and the atomic_store_nn DAG nodes put the ptr and the value in different positions. Currently we attempt to store the address to an address formed by the value. Differential Revision: https://reviews.llvm.org/D51217 llvm-svn: 340722
2018-08-27 15:08:18 +08:00
defm : AtomicStPat<atomic_store_8, SB, GPR>;
defm : AtomicStPat<atomic_store_16, SH, GPR>;
defm : AtomicStPat<atomic_store_32, SW, GPR>;
[RISCV] Codegen for i8, i16, and i32 atomicrmw with RV32A Introduce a new RISCVExpandPseudoInsts pass to expand atomic pseudo-instructions after register allocation. This is necessary in order to ensure that register spills aren't introduced between LL and SC, thus breaking the forward progress guarantee for the operation. AArch64 does something similar for CmpXchg (though only at O0), and Mips is moving towards this approach (see D31287). See also [this mailing list post](http://lists.llvm.org/pipermail/llvm-dev/2016-May/099490.html) from James Knight, which summarises the issues with lowering to ll/sc in IR or pre-RA. See the [accompanying RFC thread](http://lists.llvm.org/pipermail/llvm-dev/2018-June/123993.html) for an overview of the lowering strategy. Differential Revision: https://reviews.llvm.org/D47882 llvm-svn: 342534
2018-09-19 18:54:22 +08:00
/// AMOs
multiclass AMOPat<string AtomicOp, string BaseInst> {
def : PatGprGpr<!cast<PatFrag>(AtomicOp#"_monotonic"),
!cast<RVInst>(BaseInst)>;
def : PatGprGpr<!cast<PatFrag>(AtomicOp#"_acquire"),
!cast<RVInst>(BaseInst#"_AQ")>;
def : PatGprGpr<!cast<PatFrag>(AtomicOp#"_release"),
!cast<RVInst>(BaseInst#"_RL")>;
def : PatGprGpr<!cast<PatFrag>(AtomicOp#"_acq_rel"),
!cast<RVInst>(BaseInst#"_AQ_RL")>;
def : PatGprGpr<!cast<PatFrag>(AtomicOp#"_seq_cst"),
!cast<RVInst>(BaseInst#"_AQ_RL")>;
}
defm : AMOPat<"atomic_swap_32", "AMOSWAP_W">;
defm : AMOPat<"atomic_load_add_32", "AMOADD_W">;
defm : AMOPat<"atomic_load_and_32", "AMOAND_W">;
defm : AMOPat<"atomic_load_or_32", "AMOOR_W">;
defm : AMOPat<"atomic_load_xor_32", "AMOXOR_W">;
defm : AMOPat<"atomic_load_max_32", "AMOMAX_W">;
defm : AMOPat<"atomic_load_min_32", "AMOMIN_W">;
defm : AMOPat<"atomic_load_umax_32", "AMOMAXU_W">;
defm : AMOPat<"atomic_load_umin_32", "AMOMINU_W">;
def : Pat<(atomic_load_sub_32_monotonic GPR:$addr, GPR:$incr),
(AMOADD_W GPR:$addr, (SUB X0, GPR:$incr))>;
def : Pat<(atomic_load_sub_32_acquire GPR:$addr, GPR:$incr),
(AMOADD_W_AQ GPR:$addr, (SUB X0, GPR:$incr))>;
def : Pat<(atomic_load_sub_32_release GPR:$addr, GPR:$incr),
(AMOADD_W_RL GPR:$addr, (SUB X0, GPR:$incr))>;
def : Pat<(atomic_load_sub_32_acq_rel GPR:$addr, GPR:$incr),
(AMOADD_W_AQ_RL GPR:$addr, (SUB X0, GPR:$incr))>;
def : Pat<(atomic_load_sub_32_seq_cst GPR:$addr, GPR:$incr),
(AMOADD_W_AQ_RL GPR:$addr, (SUB X0, GPR:$incr))>;
/// Pseudo AMOs
class PseudoAMO : Pseudo<(outs GPR:$res, GPR:$scratch),
[RISCV] Remove XLenVT==i32 assumptions from RISCVInstrInfo td 1. brcond operates on an condition. 2. atomic_fence and the pseudo AMO instructions should all take xlen immediates This allows the same definitions and patterns to work for RV64 (XLenVT==i64). llvm-svn: 343678
2018-10-03 19:14:26 +08:00
(ins GPR:$addr, GPR:$incr, ixlenimm:$ordering), []> {
[RISCV] Codegen for i8, i16, and i32 atomicrmw with RV32A Introduce a new RISCVExpandPseudoInsts pass to expand atomic pseudo-instructions after register allocation. This is necessary in order to ensure that register spills aren't introduced between LL and SC, thus breaking the forward progress guarantee for the operation. AArch64 does something similar for CmpXchg (though only at O0), and Mips is moving towards this approach (see D31287). See also [this mailing list post](http://lists.llvm.org/pipermail/llvm-dev/2016-May/099490.html) from James Knight, which summarises the issues with lowering to ll/sc in IR or pre-RA. See the [accompanying RFC thread](http://lists.llvm.org/pipermail/llvm-dev/2018-June/123993.html) for an overview of the lowering strategy. Differential Revision: https://reviews.llvm.org/D47882 llvm-svn: 342534
2018-09-19 18:54:22 +08:00
let Constraints = "@earlyclobber $res,@earlyclobber $scratch";
let mayLoad = 1;
let mayStore = 1;
let hasSideEffects = 0;
}
def PseudoAtomicLoadNand32 : PseudoAMO;
[RISCV] Implement codegen for cmpxchg on RV32IA Utilise a similar ('late') lowering strategy to D47882. The changes to AtomicExpandPass allow this strategy to be utilised by other targets which implement shouldExpandAtomicCmpXchgInIR. All cmpxchg are lowered as 'strong' currently and failure ordering is ignored. This is conservative but correct. Differential Revision: https://reviews.llvm.org/D48131 llvm-svn: 347914
2018-11-30 04:43:42 +08:00
// Ordering constants must be kept in sync with the AtomicOrdering enum in
[RISCV] Codegen for i8, i16, and i32 atomicrmw with RV32A Introduce a new RISCVExpandPseudoInsts pass to expand atomic pseudo-instructions after register allocation. This is necessary in order to ensure that register spills aren't introduced between LL and SC, thus breaking the forward progress guarantee for the operation. AArch64 does something similar for CmpXchg (though only at O0), and Mips is moving towards this approach (see D31287). See also [this mailing list post](http://lists.llvm.org/pipermail/llvm-dev/2016-May/099490.html) from James Knight, which summarises the issues with lowering to ll/sc in IR or pre-RA. See the [accompanying RFC thread](http://lists.llvm.org/pipermail/llvm-dev/2018-June/123993.html) for an overview of the lowering strategy. Differential Revision: https://reviews.llvm.org/D47882 llvm-svn: 342534
2018-09-19 18:54:22 +08:00
// AtomicOrdering.h.
def : Pat<(atomic_load_nand_32_monotonic GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand32 GPR:$addr, GPR:$incr, 2)>;
def : Pat<(atomic_load_nand_32_acquire GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand32 GPR:$addr, GPR:$incr, 4)>;
def : Pat<(atomic_load_nand_32_release GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand32 GPR:$addr, GPR:$incr, 5)>;
def : Pat<(atomic_load_nand_32_acq_rel GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand32 GPR:$addr, GPR:$incr, 6)>;
def : Pat<(atomic_load_nand_32_seq_cst GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand32 GPR:$addr, GPR:$incr, 7)>;
class PseudoMaskedAMO
: Pseudo<(outs GPR:$res, GPR:$scratch),
[RISCV] Remove XLenVT==i32 assumptions from RISCVInstrInfo td 1. brcond operates on an condition. 2. atomic_fence and the pseudo AMO instructions should all take xlen immediates This allows the same definitions and patterns to work for RV64 (XLenVT==i64). llvm-svn: 343678
2018-10-03 19:14:26 +08:00
(ins GPR:$addr, GPR:$incr, GPR:$mask, ixlenimm:$ordering), []> {
[RISCV] Codegen for i8, i16, and i32 atomicrmw with RV32A Introduce a new RISCVExpandPseudoInsts pass to expand atomic pseudo-instructions after register allocation. This is necessary in order to ensure that register spills aren't introduced between LL and SC, thus breaking the forward progress guarantee for the operation. AArch64 does something similar for CmpXchg (though only at O0), and Mips is moving towards this approach (see D31287). See also [this mailing list post](http://lists.llvm.org/pipermail/llvm-dev/2016-May/099490.html) from James Knight, which summarises the issues with lowering to ll/sc in IR or pre-RA. See the [accompanying RFC thread](http://lists.llvm.org/pipermail/llvm-dev/2018-June/123993.html) for an overview of the lowering strategy. Differential Revision: https://reviews.llvm.org/D47882 llvm-svn: 342534
2018-09-19 18:54:22 +08:00
let Constraints = "@earlyclobber $res,@earlyclobber $scratch";
let mayLoad = 1;
let mayStore = 1;
let hasSideEffects = 0;
}
class PseudoMaskedAMOMinMax
: Pseudo<(outs GPR:$res, GPR:$scratch1, GPR:$scratch2),
[RISCV] Remove XLenVT==i32 assumptions from RISCVInstrInfo td 1. brcond operates on an condition. 2. atomic_fence and the pseudo AMO instructions should all take xlen immediates This allows the same definitions and patterns to work for RV64 (XLenVT==i64). llvm-svn: 343678
2018-10-03 19:14:26 +08:00
(ins GPR:$addr, GPR:$incr, GPR:$mask, ixlenimm:$sextshamt,
ixlenimm:$ordering), []> {
[RISCV] Codegen for i8, i16, and i32 atomicrmw with RV32A Introduce a new RISCVExpandPseudoInsts pass to expand atomic pseudo-instructions after register allocation. This is necessary in order to ensure that register spills aren't introduced between LL and SC, thus breaking the forward progress guarantee for the operation. AArch64 does something similar for CmpXchg (though only at O0), and Mips is moving towards this approach (see D31287). See also [this mailing list post](http://lists.llvm.org/pipermail/llvm-dev/2016-May/099490.html) from James Knight, which summarises the issues with lowering to ll/sc in IR or pre-RA. See the [accompanying RFC thread](http://lists.llvm.org/pipermail/llvm-dev/2018-June/123993.html) for an overview of the lowering strategy. Differential Revision: https://reviews.llvm.org/D47882 llvm-svn: 342534
2018-09-19 18:54:22 +08:00
let Constraints = "@earlyclobber $res,@earlyclobber $scratch1,"
"@earlyclobber $scratch2";
let mayLoad = 1;
let mayStore = 1;
let hasSideEffects = 0;
}
class PseudoMaskedAMOUMinUMax
: Pseudo<(outs GPR:$res, GPR:$scratch1, GPR:$scratch2),
[RISCV] Remove XLenVT==i32 assumptions from RISCVInstrInfo td 1. brcond operates on an condition. 2. atomic_fence and the pseudo AMO instructions should all take xlen immediates This allows the same definitions and patterns to work for RV64 (XLenVT==i64). llvm-svn: 343678
2018-10-03 19:14:26 +08:00
(ins GPR:$addr, GPR:$incr, GPR:$mask, ixlenimm:$ordering), []> {
[RISCV] Codegen for i8, i16, and i32 atomicrmw with RV32A Introduce a new RISCVExpandPseudoInsts pass to expand atomic pseudo-instructions after register allocation. This is necessary in order to ensure that register spills aren't introduced between LL and SC, thus breaking the forward progress guarantee for the operation. AArch64 does something similar for CmpXchg (though only at O0), and Mips is moving towards this approach (see D31287). See also [this mailing list post](http://lists.llvm.org/pipermail/llvm-dev/2016-May/099490.html) from James Knight, which summarises the issues with lowering to ll/sc in IR or pre-RA. See the [accompanying RFC thread](http://lists.llvm.org/pipermail/llvm-dev/2018-June/123993.html) for an overview of the lowering strategy. Differential Revision: https://reviews.llvm.org/D47882 llvm-svn: 342534
2018-09-19 18:54:22 +08:00
let Constraints = "@earlyclobber $res,@earlyclobber $scratch1,"
"@earlyclobber $scratch2";
let mayLoad = 1;
let mayStore = 1;
let hasSideEffects = 0;
}
class PseudoMaskedAMOPat<Intrinsic intrin, Pseudo AMOInst>
Reapply r372285 "GlobalISel: Don't materialize immarg arguments to intrinsics" This reverts r372314, reapplying r372285 and the commits which depend on it (r372286-r372293, and r372296-r372297) This was missing one switch to getTargetConstant in an untested case. llvm-svn: 372338
2019-09-20 00:26:14 +08:00
: Pat<(intrin GPR:$addr, GPR:$incr, GPR:$mask, timm:$ordering),
[RISCV] Codegen for i8, i16, and i32 atomicrmw with RV32A Introduce a new RISCVExpandPseudoInsts pass to expand atomic pseudo-instructions after register allocation. This is necessary in order to ensure that register spills aren't introduced between LL and SC, thus breaking the forward progress guarantee for the operation. AArch64 does something similar for CmpXchg (though only at O0), and Mips is moving towards this approach (see D31287). See also [this mailing list post](http://lists.llvm.org/pipermail/llvm-dev/2016-May/099490.html) from James Knight, which summarises the issues with lowering to ll/sc in IR or pre-RA. See the [accompanying RFC thread](http://lists.llvm.org/pipermail/llvm-dev/2018-June/123993.html) for an overview of the lowering strategy. Differential Revision: https://reviews.llvm.org/D47882 llvm-svn: 342534
2018-09-19 18:54:22 +08:00
(AMOInst GPR:$addr, GPR:$incr, GPR:$mask, imm:$ordering)>;
class PseudoMaskedAMOMinMaxPat<Intrinsic intrin, Pseudo AMOInst>
: Pat<(intrin GPR:$addr, GPR:$incr, GPR:$mask, GPR:$shiftamt,
Reapply r372285 "GlobalISel: Don't materialize immarg arguments to intrinsics" This reverts r372314, reapplying r372285 and the commits which depend on it (r372286-r372293, and r372296-r372297) This was missing one switch to getTargetConstant in an untested case. llvm-svn: 372338
2019-09-20 00:26:14 +08:00
timm:$ordering),
[RISCV] Codegen for i8, i16, and i32 atomicrmw with RV32A Introduce a new RISCVExpandPseudoInsts pass to expand atomic pseudo-instructions after register allocation. This is necessary in order to ensure that register spills aren't introduced between LL and SC, thus breaking the forward progress guarantee for the operation. AArch64 does something similar for CmpXchg (though only at O0), and Mips is moving towards this approach (see D31287). See also [this mailing list post](http://lists.llvm.org/pipermail/llvm-dev/2016-May/099490.html) from James Knight, which summarises the issues with lowering to ll/sc in IR or pre-RA. See the [accompanying RFC thread](http://lists.llvm.org/pipermail/llvm-dev/2018-June/123993.html) for an overview of the lowering strategy. Differential Revision: https://reviews.llvm.org/D47882 llvm-svn: 342534
2018-09-19 18:54:22 +08:00
(AMOInst GPR:$addr, GPR:$incr, GPR:$mask, GPR:$shiftamt,
imm:$ordering)>;
def PseudoMaskedAtomicSwap32 : PseudoMaskedAMO;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_xchg_i32,
PseudoMaskedAtomicSwap32>;
def PseudoMaskedAtomicLoadAdd32 : PseudoMaskedAMO;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_add_i32,
PseudoMaskedAtomicLoadAdd32>;
def PseudoMaskedAtomicLoadSub32 : PseudoMaskedAMO;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_sub_i32,
PseudoMaskedAtomicLoadSub32>;
def PseudoMaskedAtomicLoadNand32 : PseudoMaskedAMO;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_nand_i32,
PseudoMaskedAtomicLoadNand32>;
def PseudoMaskedAtomicLoadMax32 : PseudoMaskedAMOMinMax;
def : PseudoMaskedAMOMinMaxPat<int_riscv_masked_atomicrmw_max_i32,
PseudoMaskedAtomicLoadMax32>;
def PseudoMaskedAtomicLoadMin32 : PseudoMaskedAMOMinMax;
def : PseudoMaskedAMOMinMaxPat<int_riscv_masked_atomicrmw_min_i32,
PseudoMaskedAtomicLoadMin32>;
def PseudoMaskedAtomicLoadUMax32 : PseudoMaskedAMOUMinUMax;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_umax_i32,
PseudoMaskedAtomicLoadUMax32>;
def PseudoMaskedAtomicLoadUMin32 : PseudoMaskedAMOUMinUMax;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_umin_i32,
PseudoMaskedAtomicLoadUMin32>;
[RISCV] Implement codegen for cmpxchg on RV32IA Utilise a similar ('late') lowering strategy to D47882. The changes to AtomicExpandPass allow this strategy to be utilised by other targets which implement shouldExpandAtomicCmpXchgInIR. All cmpxchg are lowered as 'strong' currently and failure ordering is ignored. This is conservative but correct. Differential Revision: https://reviews.llvm.org/D48131 llvm-svn: 347914
2018-11-30 04:43:42 +08:00
/// Compare and exchange
class PseudoCmpXchg
: Pseudo<(outs GPR:$res, GPR:$scratch),
[RISCV] Add codegen support for RV64A In order to support codegen RV64A, this patch: * Introduces masked atomics intrinsics for atomicrmw operations and cmpxchg that use the i64 type. These are ultimately lowered to masked operations using lr.w/sc.w, but we need to use these alternate intrinsics for RV64 because i32 is not legal * Modifies RISCVExpandPseudoInsts.cpp to handle PseudoAtomicLoadNand64 and PseudoCmpXchg64 * Modifies the AtomicExpandPass hooks in RISCVTargetLowering to sext/trunc as needed for RV64 and to select the i64 intrinsic IDs when necessary * Adds appropriate patterns to RISCVInstrInfoA.td * Updates test/CodeGen/RISCV/atomic-*.ll to show RV64A support This ends up being a fairly mechanical change, as the logic for RV32A is effectively reused. Differential Revision: https://reviews.llvm.org/D53233 llvm-svn: 351422
2019-01-17 18:04:39 +08:00
(ins GPR:$addr, GPR:$cmpval, GPR:$newval, ixlenimm:$ordering), []> {
[RISCV] Implement codegen for cmpxchg on RV32IA Utilise a similar ('late') lowering strategy to D47882. The changes to AtomicExpandPass allow this strategy to be utilised by other targets which implement shouldExpandAtomicCmpXchgInIR. All cmpxchg are lowered as 'strong' currently and failure ordering is ignored. This is conservative but correct. Differential Revision: https://reviews.llvm.org/D48131 llvm-svn: 347914
2018-11-30 04:43:42 +08:00
let Constraints = "@earlyclobber $res,@earlyclobber $scratch";
let mayLoad = 1;
let mayStore = 1;
let hasSideEffects = 0;
}
// Ordering constants must be kept in sync with the AtomicOrdering enum in
// AtomicOrdering.h.
multiclass PseudoCmpXchgPat<string Op, Pseudo CmpXchgInst> {
def : Pat<(!cast<PatFrag>(Op#"_monotonic") GPR:$addr, GPR:$cmp, GPR:$new),
(CmpXchgInst GPR:$addr, GPR:$cmp, GPR:$new, 2)>;
def : Pat<(!cast<PatFrag>(Op#"_acquire") GPR:$addr, GPR:$cmp, GPR:$new),
(CmpXchgInst GPR:$addr, GPR:$cmp, GPR:$new, 4)>;
def : Pat<(!cast<PatFrag>(Op#"_release") GPR:$addr, GPR:$cmp, GPR:$new),
(CmpXchgInst GPR:$addr, GPR:$cmp, GPR:$new, 5)>;
def : Pat<(!cast<PatFrag>(Op#"_acq_rel") GPR:$addr, GPR:$cmp, GPR:$new),
(CmpXchgInst GPR:$addr, GPR:$cmp, GPR:$new, 6)>;
def : Pat<(!cast<PatFrag>(Op#"_seq_cst") GPR:$addr, GPR:$cmp, GPR:$new),
(CmpXchgInst GPR:$addr, GPR:$cmp, GPR:$new, 7)>;
}
def PseudoCmpXchg32 : PseudoCmpXchg;
defm : PseudoCmpXchgPat<"atomic_cmp_swap_32", PseudoCmpXchg32>;
def PseudoMaskedCmpXchg32
: Pseudo<(outs GPR:$res, GPR:$scratch),
(ins GPR:$addr, GPR:$cmpval, GPR:$newval, GPR:$mask,
[RISCV] Add codegen support for RV64A In order to support codegen RV64A, this patch: * Introduces masked atomics intrinsics for atomicrmw operations and cmpxchg that use the i64 type. These are ultimately lowered to masked operations using lr.w/sc.w, but we need to use these alternate intrinsics for RV64 because i32 is not legal * Modifies RISCVExpandPseudoInsts.cpp to handle PseudoAtomicLoadNand64 and PseudoCmpXchg64 * Modifies the AtomicExpandPass hooks in RISCVTargetLowering to sext/trunc as needed for RV64 and to select the i64 intrinsic IDs when necessary * Adds appropriate patterns to RISCVInstrInfoA.td * Updates test/CodeGen/RISCV/atomic-*.ll to show RV64A support This ends up being a fairly mechanical change, as the logic for RV32A is effectively reused. Differential Revision: https://reviews.llvm.org/D53233 llvm-svn: 351422
2019-01-17 18:04:39 +08:00
ixlenimm:$ordering), []> {
[RISCV] Implement codegen for cmpxchg on RV32IA Utilise a similar ('late') lowering strategy to D47882. The changes to AtomicExpandPass allow this strategy to be utilised by other targets which implement shouldExpandAtomicCmpXchgInIR. All cmpxchg are lowered as 'strong' currently and failure ordering is ignored. This is conservative but correct. Differential Revision: https://reviews.llvm.org/D48131 llvm-svn: 347914
2018-11-30 04:43:42 +08:00
let Constraints = "@earlyclobber $res,@earlyclobber $scratch";
let mayLoad = 1;
let mayStore = 1;
let hasSideEffects = 0;
}
def : Pat<(int_riscv_masked_cmpxchg_i32
Reapply r372285 "GlobalISel: Don't materialize immarg arguments to intrinsics" This reverts r372314, reapplying r372285 and the commits which depend on it (r372286-r372293, and r372296-r372297) This was missing one switch to getTargetConstant in an untested case. llvm-svn: 372338
2019-09-20 00:26:14 +08:00
GPR:$addr, GPR:$cmpval, GPR:$newval, GPR:$mask, timm:$ordering),
[RISCV] Implement codegen for cmpxchg on RV32IA Utilise a similar ('late') lowering strategy to D47882. The changes to AtomicExpandPass allow this strategy to be utilised by other targets which implement shouldExpandAtomicCmpXchgInIR. All cmpxchg are lowered as 'strong' currently and failure ordering is ignored. This is conservative but correct. Differential Revision: https://reviews.llvm.org/D48131 llvm-svn: 347914
2018-11-30 04:43:42 +08:00
(PseudoMaskedCmpXchg32
GPR:$addr, GPR:$cmpval, GPR:$newval, GPR:$mask, imm:$ordering)>;
[RISCV] atomic_store_nn have a different layout to regular store We cannot directy reuse the patterns of StPat because for some reason the store DAG node and the atomic_store_nn DAG nodes put the ptr and the value in different positions. Currently we attempt to store the address to an address formed by the value. Differential Revision: https://reviews.llvm.org/D51217 llvm-svn: 340722
2018-08-27 15:08:18 +08:00
} // Predicates = [HasStdExtA]
[RISCV] Add codegen support for RV64A In order to support codegen RV64A, this patch: * Introduces masked atomics intrinsics for atomicrmw operations and cmpxchg that use the i64 type. These are ultimately lowered to masked operations using lr.w/sc.w, but we need to use these alternate intrinsics for RV64 because i32 is not legal * Modifies RISCVExpandPseudoInsts.cpp to handle PseudoAtomicLoadNand64 and PseudoCmpXchg64 * Modifies the AtomicExpandPass hooks in RISCVTargetLowering to sext/trunc as needed for RV64 and to select the i64 intrinsic IDs when necessary * Adds appropriate patterns to RISCVInstrInfoA.td * Updates test/CodeGen/RISCV/atomic-*.ll to show RV64A support This ends up being a fairly mechanical change, as the logic for RV32A is effectively reused. Differential Revision: https://reviews.llvm.org/D53233 llvm-svn: 351422
2019-01-17 18:04:39 +08:00
let Predicates = [HasStdExtA, IsRV64] in {
/// 64-bit atomic loads and stores
// Fences will be inserted for atomic load/stores according to the logic in
// RISCVTargetLowering::{emitLeadingFence,emitTrailingFence}.
defm : LdPat<atomic_load_64, LD>;
defm : AtomicStPat<atomic_store_64, SD, GPR>;
defm : AMOPat<"atomic_swap_64", "AMOSWAP_D">;
defm : AMOPat<"atomic_load_add_64", "AMOADD_D">;
defm : AMOPat<"atomic_load_and_64", "AMOAND_D">;
defm : AMOPat<"atomic_load_or_64", "AMOOR_D">;
defm : AMOPat<"atomic_load_xor_64", "AMOXOR_D">;
defm : AMOPat<"atomic_load_max_64", "AMOMAX_D">;
defm : AMOPat<"atomic_load_min_64", "AMOMIN_D">;
defm : AMOPat<"atomic_load_umax_64", "AMOMAXU_D">;
defm : AMOPat<"atomic_load_umin_64", "AMOMINU_D">;
/// 64-bit AMOs
def : Pat<(atomic_load_sub_64_monotonic GPR:$addr, GPR:$incr),
(AMOADD_D GPR:$addr, (SUB X0, GPR:$incr))>;
def : Pat<(atomic_load_sub_64_acquire GPR:$addr, GPR:$incr),
(AMOADD_D_AQ GPR:$addr, (SUB X0, GPR:$incr))>;
def : Pat<(atomic_load_sub_64_release GPR:$addr, GPR:$incr),
(AMOADD_D_RL GPR:$addr, (SUB X0, GPR:$incr))>;
def : Pat<(atomic_load_sub_64_acq_rel GPR:$addr, GPR:$incr),
(AMOADD_D_AQ_RL GPR:$addr, (SUB X0, GPR:$incr))>;
def : Pat<(atomic_load_sub_64_seq_cst GPR:$addr, GPR:$incr),
(AMOADD_D_AQ_RL GPR:$addr, (SUB X0, GPR:$incr))>;
/// 64-bit pseudo AMOs
def PseudoAtomicLoadNand64 : PseudoAMO;
// Ordering constants must be kept in sync with the AtomicOrdering enum in
// AtomicOrdering.h.
def : Pat<(atomic_load_nand_64_monotonic GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand64 GPR:$addr, GPR:$incr, 2)>;
def : Pat<(atomic_load_nand_64_acquire GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand64 GPR:$addr, GPR:$incr, 4)>;
def : Pat<(atomic_load_nand_64_release GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand64 GPR:$addr, GPR:$incr, 5)>;
def : Pat<(atomic_load_nand_64_acq_rel GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand64 GPR:$addr, GPR:$incr, 6)>;
def : Pat<(atomic_load_nand_64_seq_cst GPR:$addr, GPR:$incr),
(PseudoAtomicLoadNand64 GPR:$addr, GPR:$incr, 7)>;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_xchg_i64,
PseudoMaskedAtomicSwap32>;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_add_i64,
PseudoMaskedAtomicLoadAdd32>;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_sub_i64,
PseudoMaskedAtomicLoadSub32>;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_nand_i64,
PseudoMaskedAtomicLoadNand32>;
def : PseudoMaskedAMOMinMaxPat<int_riscv_masked_atomicrmw_max_i64,
PseudoMaskedAtomicLoadMax32>;
def : PseudoMaskedAMOMinMaxPat<int_riscv_masked_atomicrmw_min_i64,
PseudoMaskedAtomicLoadMin32>;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_umax_i64,
PseudoMaskedAtomicLoadUMax32>;
def : PseudoMaskedAMOPat<int_riscv_masked_atomicrmw_umin_i64,
PseudoMaskedAtomicLoadUMin32>;
/// 64-bit compare and exchange
def PseudoCmpXchg64 : PseudoCmpXchg;
defm : PseudoCmpXchgPat<"atomic_cmp_swap_64", PseudoCmpXchg64>;
def : Pat<(int_riscv_masked_cmpxchg_i64
Reapply r372285 "GlobalISel: Don't materialize immarg arguments to intrinsics" This reverts r372314, reapplying r372285 and the commits which depend on it (r372286-r372293, and r372296-r372297) This was missing one switch to getTargetConstant in an untested case. llvm-svn: 372338
2019-09-20 00:26:14 +08:00
GPR:$addr, GPR:$cmpval, GPR:$newval, GPR:$mask, timm:$ordering),
[RISCV] Add codegen support for RV64A In order to support codegen RV64A, this patch: * Introduces masked atomics intrinsics for atomicrmw operations and cmpxchg that use the i64 type. These are ultimately lowered to masked operations using lr.w/sc.w, but we need to use these alternate intrinsics for RV64 because i32 is not legal * Modifies RISCVExpandPseudoInsts.cpp to handle PseudoAtomicLoadNand64 and PseudoCmpXchg64 * Modifies the AtomicExpandPass hooks in RISCVTargetLowering to sext/trunc as needed for RV64 and to select the i64 intrinsic IDs when necessary * Adds appropriate patterns to RISCVInstrInfoA.td * Updates test/CodeGen/RISCV/atomic-*.ll to show RV64A support This ends up being a fairly mechanical change, as the logic for RV32A is effectively reused. Differential Revision: https://reviews.llvm.org/D53233 llvm-svn: 351422
2019-01-17 18:04:39 +08:00
(PseudoMaskedCmpXchg32
GPR:$addr, GPR:$cmpval, GPR:$newval, GPR:$mask, imm:$ordering)>;
} // Predicates = [HasStdExtA, IsRV64]