-//===- ARMSchedule.td - ARM Scheduling Definitions ---------*- tablegen -*-===//
-//
+//===-- ARMSchedule.td - ARM Scheduling Definitions --------*- tablegen -*-===//
+//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
-//
+//
//===----------------------------------------------------------------------===//
+//===----------------------------------------------------------------------===//
+// Instruction scheduling annotations for out-of-order CPUs.
+// These annotations are independent of the itinerary class defined below.
+// Here we define the subtarget independent read/write per-operand resources.
+// The subtarget schedule definitions will then map these to the subtarget's
+// resource usages.
+// For example:
+// The instruction cycle timings table might contain an entry for an operation
+// like the following:
+// Rd <- ADD Rn, Rm, <shift> Rs
+// Uops | Latency from register | Uops - resource requirements - latency
+// 2 | Rn: 1 Rm: 4 Rs: 4 | uop T0, Rm, Rs - P01 - 3
+// | | uopc Rd, Rn, T0 - P01 - 1
+// This is telling us that the result will be available in destination register
+// Rd after a minimum of three cycles after the result in Rm and Rs is available
+// and one cycle after the result in Rn is available. The micro-ops can execute
+// on resource P01.
+// To model this, we need to express that we need to dispatch two micro-ops,
+// that the resource P01 is needed and that the latency to Rn is different than
+// the latency to Rm and Rs. The scheduler can decrease Rn's producer latency by
+// two.
+// We will do this by assigning (abstract) resources to register defs/uses.
+// ARMSchedule.td:
+// def WriteALUsr : SchedWrite;
+// def ReadAdvanceALUsr : ScheRead;
+//
+// ARMInstrInfo.td:
+// def ADDrs : I<>, Sched<[WriteALUsr, ReadAdvanceALUsr, ReadDefault,
+// ReadDefault]> { ...}
+// ReadAdvance read resources allow us to define "pipeline by-passes" or
+// shorter latencies to certain registers as needed in the example above.
+// The "ReadDefault" can be omitted.
+// Next, the subtarget td file assigns resources to the abstract resources
+// defined here.
+// ARMScheduleSubtarget.td:
+// // Resources.
+// def P01 : ProcResource<3>; // ALU unit (3 of it).
+// ...
+// // Resource usages.
+// def : WriteRes<WriteALUsr, [P01, P01]> {
+// Latency = 4; // Latency of 4.
+// NumMicroOps = 2; // Dispatch 2 micro-ops.
+// // The two instances of resource P01 are occupied for one cycle. It is one
+// // cycle because these resources happen to be pipelined.
+// ResourceCycles = [1, 1];
+// }
+// def : ReadAdvance<ReadAdvanceALUsr, 3>;
+
+// Basic ALU operation.
+def WriteALU : SchedWrite;
+def ReadALU : SchedRead;
+
+// Basic ALU with shifts.
+def WriteALUsi : SchedWrite; // Shift by immediate.
+def WriteALUsr : SchedWrite; // Shift by register.
+def WriteALUSsr : SchedWrite; // Shift by register (flag setting).
+def ReadALUsr : SchedRead; // Some operands are read later.
+
+// Compares.
+def WriteCMP : SchedWrite;
+def WriteCMPsi : SchedWrite;
+def WriteCMPsr : SchedWrite;
+
+// Division.
+def WriteDiv : SchedWrite;
+
+// Loads.
+def WriteLd : SchedWrite;
+def WritePreLd : SchedWrite;
+
+// Branches.
+def WriteBr : SchedWrite;
+def WriteBrL : SchedWrite;
+def WriteBrTbl : SchedWrite;
+
+// Fixpoint conversions.
+def WriteCvtFP : SchedWrite;
+
+// Noop.
+def WriteNoop : SchedWrite;
+
+// Define TII for use in SchedVariant Predicates.
+def : PredicateProlog<[{
+ const ARMBaseInstrInfo *TII =
+ static_cast<const ARMBaseInstrInfo*>(SchedModel->getInstrInfo());
+ (void)TII;
+}]>;
+
+def IsPredicatedPred : SchedPredicate<[{TII->isPredicated(MI)}]>;
//===----------------------------------------------------------------------===//
// Instruction Itinerary classes used for ARM
def IIC_iMOVsi : InstrItinClass;
def IIC_iMOVsr : InstrItinClass;
def IIC_iMOVix2 : InstrItinClass;
+def IIC_iMOVix2addpc : InstrItinClass;
+def IIC_iMOVix2ld : InstrItinClass;
def IIC_iMVNi : InstrItinClass;
def IIC_iMVNr : InstrItinClass;
def IIC_iMVNsi : InstrItinClass;
def IIC_iMAC32 : InstrItinClass;
def IIC_iMUL64 : InstrItinClass;
def IIC_iMAC64 : InstrItinClass;
+def IIC_iDIV : InstrItinClass;
def IIC_iLoad_i : InstrItinClass;
def IIC_iLoad_r : InstrItinClass;
def IIC_iLoad_si : InstrItinClass;
def IIC_iLoad_d_i : InstrItinClass;
def IIC_iLoad_d_r : InstrItinClass;
def IIC_iLoad_d_ru : InstrItinClass;
-def IIC_iLoad_m : InstrItinClass<0>; // micro-coded
-def IIC_iLoad_mu : InstrItinClass<0>; // micro-coded
-def IIC_iLoad_mBr : InstrItinClass<0>; // micro-coded
-def IIC_iPop : InstrItinClass<0>; // micro-coded
-def IIC_iPop_Br : InstrItinClass<0>; // micro-coded
+def IIC_iLoad_m : InstrItinClass;
+def IIC_iLoad_mu : InstrItinClass;
+def IIC_iLoad_mBr : InstrItinClass;
+def IIC_iPop : InstrItinClass;
+def IIC_iPop_Br : InstrItinClass;
def IIC_iLoadiALU : InstrItinClass;
def IIC_iStore_i : InstrItinClass;
def IIC_iStore_r : InstrItinClass;
def IIC_iStore_d_i : InstrItinClass;
def IIC_iStore_d_r : InstrItinClass;
def IIC_iStore_d_ru : InstrItinClass;
-def IIC_iStore_m : InstrItinClass<0>; // micro-coded
-def IIC_iStore_mu : InstrItinClass<0>; // micro-coded
+def IIC_iStore_m : InstrItinClass;
+def IIC_iStore_mu : InstrItinClass;
def IIC_Preload : InstrItinClass;
def IIC_Br : InstrItinClass;
def IIC_fpSTAT : InstrItinClass;
def IIC_fpMUL64 : InstrItinClass;
def IIC_fpMAC32 : InstrItinClass;
def IIC_fpMAC64 : InstrItinClass;
+def IIC_fpFMAC32 : InstrItinClass;
+def IIC_fpFMAC64 : InstrItinClass;
def IIC_fpDIV32 : InstrItinClass;
def IIC_fpDIV64 : InstrItinClass;
def IIC_fpSQRT32 : InstrItinClass;
def IIC_fpSQRT64 : InstrItinClass;
def IIC_fpLoad32 : InstrItinClass;
def IIC_fpLoad64 : InstrItinClass;
-def IIC_fpLoad_m : InstrItinClass<0>; // micro-coded
-def IIC_fpLoad_mu : InstrItinClass<0>; // micro-coded
+def IIC_fpLoad_m : InstrItinClass;
+def IIC_fpLoad_mu : InstrItinClass;
def IIC_fpStore32 : InstrItinClass;
def IIC_fpStore64 : InstrItinClass;
-def IIC_fpStore_m : InstrItinClass<0>; // micro-coded
-def IIC_fpStore_mu : InstrItinClass<0>; // micro-coded
+def IIC_fpStore_m : InstrItinClass;
+def IIC_fpStore_mu : InstrItinClass;
def IIC_VLD1 : InstrItinClass;
def IIC_VLD1x2 : InstrItinClass;
def IIC_VLD1x3 : InstrItinClass;
def IIC_VLD3ln : InstrItinClass;
def IIC_VLD3u : InstrItinClass;
def IIC_VLD3lnu : InstrItinClass;
+def IIC_VLD3dup : InstrItinClass;
+def IIC_VLD3dupu : InstrItinClass;
def IIC_VLD4 : InstrItinClass;
def IIC_VLD4ln : InstrItinClass;
def IIC_VLD4u : InstrItinClass;
def IIC_VLD4lnu : InstrItinClass;
+def IIC_VLD4dup : InstrItinClass;
+def IIC_VLD4dupu : InstrItinClass;
def IIC_VST1 : InstrItinClass;
def IIC_VST1x2 : InstrItinClass;
def IIC_VST1x3 : InstrItinClass;
def IIC_VPERMQ3 : InstrItinClass;
def IIC_VMACD : InstrItinClass;
def IIC_VMACQ : InstrItinClass;
+def IIC_VFMACD : InstrItinClass;
+def IIC_VFMACQ : InstrItinClass;
def IIC_VRECSD : InstrItinClass;
def IIC_VRECSQ : InstrItinClass;
def IIC_VCNTiD : InstrItinClass;
//===----------------------------------------------------------------------===//
// Processor instruction itineraries.
-def GenericItineraries : ProcessorItineraries<[], [], []>;
-
include "ARMScheduleV6.td"
include "ARMScheduleA8.td"
include "ARMScheduleA9.td"
+include "ARMScheduleSwift.td"
projects / oota-llvm.git / blobdiff