llvm-6502/include/llvm/Target/TargetItinerary.td
Andrew Trick 2661b411cc I'm introducing a new machine model to simultaneously allow simple
subtarget CPU descriptions and support new features of
MachineScheduler.

MachineModel has three categories of data:
1) Basic properties for coarse grained instruction cost model.
2) Scheduler Read/Write resources for simple per-opcode and operand cost model (TBD).
3) Instruction itineraties for detailed per-cycle reservation tables.

These will all live side-by-side. Any subtarget can use any
combination of them. Instruction itineraries will not change in the
near term. In the long run, I expect them to only be relevant for
in-order VLIW machines that have complex contraints and require a
precise scheduling/bundling model. Once itineraries are only actively
used by VLIW-ish targets, they could be replaced by something more
appropriate for those targets.

This tablegen backend rewrite sets things up for introducing
MachineModel type #2: per opcode/operand cost model.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159891 91177308-0d34-0410-b5e6-96231b3b80d8
2012-07-07 04:00:00 +00:00

137 lines
6.2 KiB
TableGen

//===- TargetItinerary.td - Target Itinierary Description --*- tablegen -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file defines the target-independent scheduling interfaces
// which should be implemented by each target that uses instruction
// itineraries for scheduling. Itineraries are details reservation
// tables for each instruction class. They are most appropriate for
// in-order machine with complicated scheduling or bundling constraints.
//
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
// Processor functional unit - These values represent the function units
// available across all chip sets for the target. Eg., IntUnit, FPUnit, ...
// These may be independent values for each chip set or may be shared across
// all chip sets of the target. Each functional unit is treated as a resource
// during scheduling and has an affect instruction order based on availability
// during a time interval.
//
class FuncUnit;
//===----------------------------------------------------------------------===//
// Pipeline bypass / forwarding - These values specifies the symbolic names of
// pipeline bypasses which can be used to forward results of instructions
// that are forwarded to uses.
class Bypass;
def NoBypass : Bypass;
class ReservationKind<bits<1> val> {
int Value = val;
}
def Required : ReservationKind<0>;
def Reserved : ReservationKind<1>;
//===----------------------------------------------------------------------===//
// Instruction stage - These values represent a non-pipelined step in
// the execution of an instruction. Cycles represents the number of
// discrete time slots needed to complete the stage. Units represent
// the choice of functional units that can be used to complete the
// stage. Eg. IntUnit1, IntUnit2. NextCycles indicates how many
// cycles should elapse from the start of this stage to the start of
// the next stage in the itinerary. For example:
//
// A stage is specified in one of two ways:
//
// InstrStage<1, [FU_x, FU_y]> - TimeInc defaults to Cycles
// InstrStage<1, [FU_x, FU_y], 0> - TimeInc explicit
//
class InstrStage<int cycles, list<FuncUnit> units,
int timeinc = -1,
ReservationKind kind = Required> {
int Cycles = cycles; // length of stage in machine cycles
list<FuncUnit> Units = units; // choice of functional units
int TimeInc = timeinc; // cycles till start of next stage
int Kind = kind.Value; // kind of FU reservation
}
//===----------------------------------------------------------------------===//
// Instruction itinerary - An itinerary represents a sequential series of steps
// required to complete an instruction. Itineraries are represented as lists of
// instruction stages.
//
//===----------------------------------------------------------------------===//
// Instruction itinerary classes - These values represent 'named' instruction
// itinerary. Using named itineraries simplifies managing groups of
// instructions across chip sets. An instruction uses the same itinerary class
// across all chip sets. Thus a new chip set can be added without modifying
// instruction information.
//
class InstrItinClass;
def NoItinerary : InstrItinClass;
//===----------------------------------------------------------------------===//
// Instruction itinerary data - These values provide a runtime map of an
// instruction itinerary class (name) to its itinerary data.
//
// NumMicroOps represents the number of micro-operations that each instruction
// in the class are decoded to. If the number is zero, then it means the
// instruction can decode into variable number of micro-ops and it must be
// determined dynamically. This directly relates to the itineraries
// global IssueWidth property, which constrains the number of microops
// that can issue per cycle.
//
// OperandCycles are optional "cycle counts". They specify the cycle after
// instruction issue the values which correspond to specific operand indices
// are defined or read. Bypasses are optional "pipeline forwarding pathes", if
// a def by an instruction is available on a specific bypass and the use can
// read from the same bypass, then the operand use latency is reduced by one.
//
// InstrItinData<IIC_iLoad_i , [InstrStage<1, [A9_Pipe1]>,
// InstrStage<1, [A9_AGU]>],
// [3, 1], [A9_LdBypass]>,
// InstrItinData<IIC_iMVNr , [InstrStage<1, [A9_Pipe0, A9_Pipe1]>],
// [1, 1], [NoBypass, A9_LdBypass]>,
//
// In this example, the instruction of IIC_iLoadi reads its input on cycle 1
// (after issue) and the result of the load is available on cycle 3. The result
// is available via forwarding path A9_LdBypass. If it's used by the first
// source operand of instructions of IIC_iMVNr class, then the operand latency
// is reduced by 1.
class InstrItinData<InstrItinClass Class, list<InstrStage> stages,
list<int> operandcycles = [],
list<Bypass> bypasses = [], int uops = 1> {
InstrItinClass TheClass = Class;
int NumMicroOps = uops;
list<InstrStage> Stages = stages;
list<int> OperandCycles = operandcycles;
list<Bypass> Bypasses = bypasses;
}
//===----------------------------------------------------------------------===//
// Processor itineraries - These values represent the set of all itinerary
// classes for a given chip set.
//
// Set property values to -1 to use the default.
// See InstrItineraryProps for comments and defaults.
class ProcessorItineraries<list<FuncUnit> fu, list<Bypass> bp,
list<InstrItinData> iid> {
list<FuncUnit> FU = fu;
list<Bypass> BP = bp;
list<InstrItinData> IID = iid;
}
// NoItineraries - A marker that can be used by processors without schedule
// info. Subtargets using NoItineraries can bypass the scheduler's
// expensive HazardRecognizer because no reservation table is needed.
def NoItineraries : ProcessorItineraries<[], [], []>;