llvm-6502/lib/Target/ARM/ARM.td

429 lines
22 KiB
TableGen
Raw Normal View History

//===-- ARM.td - Describe the ARM Target Machine -----------*- tablegen -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
//
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
// Target-independent interfaces which we are implementing
//===----------------------------------------------------------------------===//
include "llvm/Target/Target.td"
//===----------------------------------------------------------------------===//
// ARM Subtarget state.
//
def ModeThumb : SubtargetFeature<"thumb-mode", "InThumbMode", "true",
"Thumb mode">;
//===----------------------------------------------------------------------===//
// ARM Subtarget features.
//
def FeatureVFP2 : SubtargetFeature<"vfp2", "HasVFPv2", "true",
"Enable VFP2 instructions">;
def FeatureVFP3 : SubtargetFeature<"vfp3", "HasVFPv3", "true",
"Enable VFP3 instructions",
[FeatureVFP2]>;
def FeatureNEON : SubtargetFeature<"neon", "HasNEON", "true",
"Enable NEON instructions",
[FeatureVFP3]>;
def FeatureThumb2 : SubtargetFeature<"thumb2", "HasThumb2", "true",
"Enable Thumb2 instructions">;
def FeatureNoARM : SubtargetFeature<"noarm", "NoARM", "true",
"Does not support ARM mode execution",
[ModeThumb]>;
def FeatureFP16 : SubtargetFeature<"fp16", "HasFP16", "true",
"Enable half-precision floating point">;
def FeatureVFP4 : SubtargetFeature<"vfp4", "HasVFPv4", "true",
"Enable VFP4 instructions",
[FeatureVFP3, FeatureFP16]>;
def FeatureFPARMv8 : SubtargetFeature<"fp-armv8", "HasFPARMv8",
"true", "Enable ARMv8 FP",
[FeatureVFP4]>;
def FeatureD16 : SubtargetFeature<"d16", "HasD16", "true",
"Restrict VFP3 to 16 double registers">;
def FeatureHWDiv : SubtargetFeature<"hwdiv", "HasHardwareDivide", "true",
"Enable divide instructions">;
def FeatureHWDivARM : SubtargetFeature<"hwdiv-arm",
"HasHardwareDivideInARM", "true",
"Enable divide instructions in ARM mode">;
def FeatureT2XtPk : SubtargetFeature<"t2xtpk", "HasT2ExtractPack", "true",
"Enable Thumb2 extract and pack instructions">;
def FeatureDB : SubtargetFeature<"db", "HasDataBarrier", "true",
"Has data barrier (dmb / dsb) instructions">;
def FeatureSlowFPBrcc : SubtargetFeature<"slow-fp-brcc", "SlowFPBrcc", "true",
"FP compare + branch is slow">;
def FeatureVFPOnlySP : SubtargetFeature<"fp-only-sp", "FPOnlySP", "true",
"Floating point unit supports single precision only">;
def FeaturePerfMon : SubtargetFeature<"perfmon", "HasPerfMon", "true",
"Enable support for Performance Monitor extensions">;
def FeatureTrustZone : SubtargetFeature<"trustzone", "HasTrustZone", "true",
"Enable support for TrustZone security extensions">;
def FeatureCrypto : SubtargetFeature<"crypto", "HasCrypto", "true",
"Enable support for Cryptography extensions",
[FeatureNEON]>;
def FeatureCRC : SubtargetFeature<"crc", "HasCRC", "true",
"Enable support for CRC instructions">;
Making use of VFP / NEON floating point multiply-accumulate / subtraction is difficult on current ARM implementations for a few reasons. 1. Even though a single vmla has latency that is one cycle shorter than a pair of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause additional pipeline stall. So it's frequently better to single codegen vmul + vadd. 2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to stall for 4 cycles. We need to schedule them apart. 3. A vmla followed vmla is a special case. Obvious issuing back to back RAW vmla + vmla is very bad. But this isn't ideal either: vmul vadd vmla Instead, we want to expand the second vmla: vmla vmul vadd Even with the 4 cycle vmul stall, the second sequence is still 2 cycles faster. Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough but it isn't the optimial solution. This patch attempts to make it possible to use vmla / vmls in cases where it is profitable. A. Add missing isel predicates which cause vmla to be codegen'ed. B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to compute a fmul and a fmla. C. Add additional isel checks for vmla, avoid cases where vmla is feeding into fp instructions (except for the #3 exceptional case). D. Add ARM hazard recognizer to model the vmla / vmls hazards. E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the vmla / vmls will trigger one of the special hazards. Work in progress, only A+B are enabled. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120960 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-05 22:04:16 +00:00
// Some processors have FP multiply-accumulate instructions that don't
// play nicely with other VFP / NEON instructions, and it's generally better
// to just not use them.
Making use of VFP / NEON floating point multiply-accumulate / subtraction is difficult on current ARM implementations for a few reasons. 1. Even though a single vmla has latency that is one cycle shorter than a pair of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause additional pipeline stall. So it's frequently better to single codegen vmul + vadd. 2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to stall for 4 cycles. We need to schedule them apart. 3. A vmla followed vmla is a special case. Obvious issuing back to back RAW vmla + vmla is very bad. But this isn't ideal either: vmul vadd vmla Instead, we want to expand the second vmla: vmla vmul vadd Even with the 4 cycle vmul stall, the second sequence is still 2 cycles faster. Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough but it isn't the optimial solution. This patch attempts to make it possible to use vmla / vmls in cases where it is profitable. A. Add missing isel predicates which cause vmla to be codegen'ed. B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to compute a fmul and a fmla. C. Add additional isel checks for vmla, avoid cases where vmla is feeding into fp instructions (except for the #3 exceptional case). D. Add ARM hazard recognizer to model the vmla / vmls hazards. E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the vmla / vmls will trigger one of the special hazards. Work in progress, only A+B are enabled. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120960 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-05 22:04:16 +00:00
def FeatureHasSlowFPVMLx : SubtargetFeature<"slowfpvmlx", "SlowFPVMLx", "true",
"Disable VFP / NEON MAC instructions">;
// Cortex-A8 / A9 Advanced SIMD has multiplier accumulator forwarding.
def FeatureVMLxForwarding : SubtargetFeature<"vmlx-forwarding",
"HasVMLxForwarding", "true",
"Has multiplier accumulator forwarding">;
// Some processors benefit from using NEON instructions for scalar
// single-precision FP operations.
def FeatureNEONForFP : SubtargetFeature<"neonfp", "UseNEONForSinglePrecisionFP",
"true",
"Use NEON for single precision FP">;
// Disable 32-bit to 16-bit narrowing for experimentation.
def FeaturePref32BitThumb : SubtargetFeature<"32bit", "Pref32BitThumb", "true",
"Prefer 32-bit Thumb instrs">;
/// Some instructions update CPSR partially, which can add false dependency for
/// out-of-order implementation, e.g. Cortex-A9, unless each individual bit is
/// mapped to a separate physical register. Avoid partial CPSR update for these
/// processors.
def FeatureAvoidPartialCPSR : SubtargetFeature<"avoid-partial-cpsr",
"AvoidCPSRPartialUpdate", "true",
"Avoid CPSR partial update for OOO execution">;
def FeatureAvoidMOVsShOp : SubtargetFeature<"avoid-movs-shop",
"AvoidMOVsShifterOperand", "true",
"Avoid movs instructions with shifter operand">;
// Some processors perform return stack prediction. CodeGen should avoid issue
// "normal" call instructions to callees which do not return.
def FeatureHasRAS : SubtargetFeature<"ras", "HasRAS", "true",
"Has return address stack">;
/// Some M architectures don't have the DSP extension (v7E-M vs. v7M)
def FeatureDSPThumb2 : SubtargetFeature<"t2dsp", "Thumb2DSP", "true",
"Supports v7 DSP instructions in Thumb2">;
// Multiprocessing extension.
def FeatureMP : SubtargetFeature<"mp", "HasMPExtension", "true",
"Supports Multiprocessing extension">;
// Virtualization extension - requires HW divide (ARMv7-AR ARMARM - 4.4.8).
def FeatureVirtualization : SubtargetFeature<"virtualization",
"HasVirtualization", "true",
"Supports Virtualization extension",
[FeatureHWDiv, FeatureHWDivARM]>;
// M-series ISA
def FeatureMClass : SubtargetFeature<"mclass", "ARMProcClass", "MClass",
"Is microcontroller profile ('M' series)">;
// R-series ISA
def FeatureRClass : SubtargetFeature<"rclass", "ARMProcClass", "RClass",
"Is realtime profile ('R' series)">;
// A-series ISA
def FeatureAClass : SubtargetFeature<"aclass", "ARMProcClass", "AClass",
"Is application profile ('A' series)">;
// Special TRAP encoding for NaCl, which looks like a TRAP in Thumb too.
// See ARMInstrInfo.td for details.
def FeatureNaClTrap : SubtargetFeature<"nacl-trap", "UseNaClTrap", "true",
"NaCl trap">;
// ARM ISAs.
def HasV4TOps : SubtargetFeature<"v4t", "HasV4TOps", "true",
"Support ARM v4T instructions">;
def HasV5TOps : SubtargetFeature<"v5t", "HasV5TOps", "true",
"Support ARM v5T instructions",
[HasV4TOps]>;
def HasV5TEOps : SubtargetFeature<"v5te", "HasV5TEOps", "true",
"Support ARM v5TE, v5TEj, and v5TExp instructions",
[HasV5TOps]>;
def HasV6Ops : SubtargetFeature<"v6", "HasV6Ops", "true",
"Support ARM v6 instructions",
[HasV5TEOps]>;
def HasV6MOps : SubtargetFeature<"v6m", "HasV6MOps", "true",
"Support ARM v6M instructions",
[HasV6Ops]>;
def HasV6T2Ops : SubtargetFeature<"v6t2", "HasV6T2Ops", "true",
"Support ARM v6t2 instructions",
[HasV6MOps, FeatureThumb2]>;
def HasV7Ops : SubtargetFeature<"v7", "HasV7Ops", "true",
"Support ARM v7 instructions",
[HasV6T2Ops, FeaturePerfMon]>;
def HasV8Ops : SubtargetFeature<"v8", "HasV8Ops", "true",
"Support ARM v8 instructions",
[HasV7Ops, FeatureVirtualization,
FeatureMP]>;
//===----------------------------------------------------------------------===//
// ARM Processors supported.
//
include "ARMSchedule.td"
// ARM processor families.
def ProcA5 : SubtargetFeature<"a5", "ARMProcFamily", "CortexA5",
"Cortex-A5 ARM processors",
[FeatureSlowFPBrcc, FeatureHasSlowFPVMLx,
FeatureVMLxForwarding, FeatureT2XtPk,
FeatureTrustZone, FeatureMP]>;
def ProcA7 : SubtargetFeature<"a7", "ARMProcFamily", "CortexA7",
"Cortex-A7 ARM processors",
[FeatureSlowFPBrcc, FeatureHasSlowFPVMLx,
FeatureVMLxForwarding, FeatureT2XtPk,
FeatureVFP4, FeatureMP,
FeatureHWDiv, FeatureHWDivARM,
FeatureTrustZone, FeatureVirtualization]>;
def ProcA8 : SubtargetFeature<"a8", "ARMProcFamily", "CortexA8",
"Cortex-A8 ARM processors",
[FeatureSlowFPBrcc, FeatureHasSlowFPVMLx,
FeatureVMLxForwarding, FeatureT2XtPk,
FeatureTrustZone]>;
def ProcA9 : SubtargetFeature<"a9", "ARMProcFamily", "CortexA9",
"Cortex-A9 ARM processors",
This patch combines several changes from Evan Cheng for rdar://8659675. Making use of VFP / NEON floating point multiply-accumulate / subtraction is difficult on current ARM implementations for a few reasons. 1. Even though a single vmla has latency that is one cycle shorter than a pair of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause additional pipeline stall. So it's frequently better to single codegen vmul + vadd. 2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to stall for 4 cycles. We need to schedule them apart. 3. A vmla followed vmla is a special case. Obvious issuing back to back RAW vmla + vmla is very bad. But this isn't ideal either: vmul vadd vmla Instead, we want to expand the second vmla: vmla vmul vadd Even with the 4 cycle vmul stall, the second sequence is still 2 cycles faster. Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough but it isn't the optimial solution. This patch attempts to make it possible to use vmla / vmls in cases where it is profitable. A. Add missing isel predicates which cause vmla to be codegen'ed. B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to compute a fmul and a fmla. C. Add additional isel checks for vmla, avoid cases where vmla is feeding into fp instructions (except for the #3 exceptional case). D. Add ARM hazard recognizer to model the vmla / vmls hazards. E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the vmla / vmls will trigger one of the special hazards. Enable these fp vmlx codegen changes for Cortex-A9. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@129775 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-19 18:11:57 +00:00
[FeatureVMLxForwarding,
FeatureT2XtPk, FeatureFP16,
FeatureAvoidPartialCPSR,
FeatureTrustZone]>;
def ProcSwift : SubtargetFeature<"swift", "ARMProcFamily", "Swift",
"Swift ARM processors",
[FeatureNEONForFP, FeatureT2XtPk,
FeatureVFP4, FeatureMP, FeatureHWDiv,
FeatureHWDivARM, FeatureAvoidPartialCPSR,
FeatureAvoidMOVsShOp,
FeatureHasSlowFPVMLx, FeatureTrustZone]>;
def ProcA12 : SubtargetFeature<"a12", "ARMProcFamily", "CortexA12",
"Cortex-A12 ARM processors",
[FeatureVMLxForwarding,
FeatureT2XtPk, FeatureVFP4,
FeatureHWDiv, FeatureHWDivARM,
FeatureAvoidPartialCPSR,
FeatureVirtualization,
FeatureTrustZone]>;
// FIXME: It has not been determined if A15 has these features.
def ProcA15 : SubtargetFeature<"a15", "ARMProcFamily", "CortexA15",
"Cortex-A15 ARM processors",
[FeatureT2XtPk, FeatureVFP4,
FeatureMP, FeatureHWDiv, FeatureHWDivARM,
FeatureAvoidPartialCPSR,
FeatureTrustZone, FeatureVirtualization]>;
def ProcA53 : SubtargetFeature<"a53", "ARMProcFamily", "CortexA53",
"Cortex-A53 ARM processors",
[FeatureHWDiv, FeatureHWDivARM,
FeatureTrustZone, FeatureT2XtPk,
FeatureCrypto, FeatureCRC]>;
def ProcA57 : SubtargetFeature<"a57", "ARMProcFamily", "CortexA57",
"Cortex-A57 ARM processors",
[FeatureHWDiv, FeatureHWDivARM,
FeatureTrustZone, FeatureT2XtPk,
FeatureCrypto, FeatureCRC]>;
def ProcR5 : SubtargetFeature<"r5", "ARMProcFamily", "CortexR5",
"Cortex-R5 ARM processors",
[FeatureSlowFPBrcc,
FeatureHWDiv, FeatureHWDivARM,
FeatureHasSlowFPVMLx,
FeatureAvoidPartialCPSR,
FeatureT2XtPk]>;
// FIXME: krait has currently the same features as A9
// plus VFP4 and hardware division features.
def ProcKrait : SubtargetFeature<"krait", "ARMProcFamily", "Krait",
"Qualcomm ARM processors",
[FeatureVMLxForwarding,
FeatureT2XtPk, FeatureFP16,
FeatureAvoidPartialCPSR,
FeatureTrustZone,
FeatureVFP4,
FeatureHWDiv,
FeatureHWDivARM]>;
def FeatureAPCS : SubtargetFeature<"apcs", "TargetABI", "ARM_ABI_APCS",
"Use the APCS ABI">;
def FeatureAAPCS : SubtargetFeature<"aapcs", "TargetABI", "ARM_ABI_AAPCS",
"Use the AAPCS ABI">;
class ProcNoItin<string Name, list<SubtargetFeature> Features>
: Processor<Name, NoItineraries, Features>;
// V4 Processors.
def : ProcNoItin<"generic", []>;
def : ProcNoItin<"arm8", []>;
def : ProcNoItin<"arm810", []>;
def : ProcNoItin<"strongarm", []>;
def : ProcNoItin<"strongarm110", []>;
def : ProcNoItin<"strongarm1100", []>;
def : ProcNoItin<"strongarm1110", []>;
// V4T Processors.
def : ProcNoItin<"arm7tdmi", [HasV4TOps]>;
def : ProcNoItin<"arm7tdmi-s", [HasV4TOps]>;
def : ProcNoItin<"arm710t", [HasV4TOps]>;
def : ProcNoItin<"arm720t", [HasV4TOps]>;
def : ProcNoItin<"arm9", [HasV4TOps]>;
def : ProcNoItin<"arm9tdmi", [HasV4TOps]>;
def : ProcNoItin<"arm920", [HasV4TOps]>;
def : ProcNoItin<"arm920t", [HasV4TOps]>;
def : ProcNoItin<"arm922t", [HasV4TOps]>;
def : ProcNoItin<"arm940t", [HasV4TOps]>;
def : ProcNoItin<"ep9312", [HasV4TOps]>;
// V5T Processors.
def : ProcNoItin<"arm10tdmi", [HasV5TOps]>;
def : ProcNoItin<"arm1020t", [HasV5TOps]>;
// V5TE Processors.
def : ProcNoItin<"arm9e", [HasV5TEOps]>;
def : ProcNoItin<"arm926ej-s", [HasV5TEOps]>;
def : ProcNoItin<"arm946e-s", [HasV5TEOps]>;
def : ProcNoItin<"arm966e-s", [HasV5TEOps]>;
def : ProcNoItin<"arm968e-s", [HasV5TEOps]>;
def : ProcNoItin<"arm10e", [HasV5TEOps]>;
def : ProcNoItin<"arm1020e", [HasV5TEOps]>;
def : ProcNoItin<"arm1022e", [HasV5TEOps]>;
def : ProcNoItin<"xscale", [HasV5TEOps]>;
def : ProcNoItin<"iwmmxt", [HasV5TEOps]>;
// V6 Processors.
def : Processor<"arm1136j-s", ARMV6Itineraries, [HasV6Ops]>;
def : Processor<"arm1136jf-s", ARMV6Itineraries, [HasV6Ops, FeatureVFP2,
Making use of VFP / NEON floating point multiply-accumulate / subtraction is difficult on current ARM implementations for a few reasons. 1. Even though a single vmla has latency that is one cycle shorter than a pair of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause additional pipeline stall. So it's frequently better to single codegen vmul + vadd. 2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to stall for 4 cycles. We need to schedule them apart. 3. A vmla followed vmla is a special case. Obvious issuing back to back RAW vmla + vmla is very bad. But this isn't ideal either: vmul vadd vmla Instead, we want to expand the second vmla: vmla vmul vadd Even with the 4 cycle vmul stall, the second sequence is still 2 cycles faster. Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough but it isn't the optimial solution. This patch attempts to make it possible to use vmla / vmls in cases where it is profitable. A. Add missing isel predicates which cause vmla to be codegen'ed. B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to compute a fmul and a fmla. C. Add additional isel checks for vmla, avoid cases where vmla is feeding into fp instructions (except for the #3 exceptional case). D. Add ARM hazard recognizer to model the vmla / vmls hazards. E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the vmla / vmls will trigger one of the special hazards. Work in progress, only A+B are enabled. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120960 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-05 22:04:16 +00:00
FeatureHasSlowFPVMLx]>;
def : Processor<"arm1176jz-s", ARMV6Itineraries, [HasV6Ops]>;
def : Processor<"arm1176jzf-s", ARMV6Itineraries, [HasV6Ops, FeatureVFP2,
Making use of VFP / NEON floating point multiply-accumulate / subtraction is difficult on current ARM implementations for a few reasons. 1. Even though a single vmla has latency that is one cycle shorter than a pair of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause additional pipeline stall. So it's frequently better to single codegen vmul + vadd. 2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to stall for 4 cycles. We need to schedule them apart. 3. A vmla followed vmla is a special case. Obvious issuing back to back RAW vmla + vmla is very bad. But this isn't ideal either: vmul vadd vmla Instead, we want to expand the second vmla: vmla vmul vadd Even with the 4 cycle vmul stall, the second sequence is still 2 cycles faster. Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough but it isn't the optimial solution. This patch attempts to make it possible to use vmla / vmls in cases where it is profitable. A. Add missing isel predicates which cause vmla to be codegen'ed. B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to compute a fmul and a fmla. C. Add additional isel checks for vmla, avoid cases where vmla is feeding into fp instructions (except for the #3 exceptional case). D. Add ARM hazard recognizer to model the vmla / vmls hazards. E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the vmla / vmls will trigger one of the special hazards. Work in progress, only A+B are enabled. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120960 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-05 22:04:16 +00:00
FeatureHasSlowFPVMLx]>;
def : Processor<"mpcorenovfp", ARMV6Itineraries, [HasV6Ops]>;
def : Processor<"mpcore", ARMV6Itineraries, [HasV6Ops, FeatureVFP2,
Making use of VFP / NEON floating point multiply-accumulate / subtraction is difficult on current ARM implementations for a few reasons. 1. Even though a single vmla has latency that is one cycle shorter than a pair of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause additional pipeline stall. So it's frequently better to single codegen vmul + vadd. 2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to stall for 4 cycles. We need to schedule them apart. 3. A vmla followed vmla is a special case. Obvious issuing back to back RAW vmla + vmla is very bad. But this isn't ideal either: vmul vadd vmla Instead, we want to expand the second vmla: vmla vmul vadd Even with the 4 cycle vmul stall, the second sequence is still 2 cycles faster. Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough but it isn't the optimial solution. This patch attempts to make it possible to use vmla / vmls in cases where it is profitable. A. Add missing isel predicates which cause vmla to be codegen'ed. B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to compute a fmul and a fmla. C. Add additional isel checks for vmla, avoid cases where vmla is feeding into fp instructions (except for the #3 exceptional case). D. Add ARM hazard recognizer to model the vmla / vmls hazards. E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the vmla / vmls will trigger one of the special hazards. Work in progress, only A+B are enabled. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120960 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-05 22:04:16 +00:00
FeatureHasSlowFPVMLx]>;
// V6M Processors.
def : Processor<"cortex-m0", ARMV6Itineraries, [HasV6MOps, FeatureNoARM,
FeatureDB, FeatureMClass]>;
// V6T2 Processors.
def : Processor<"arm1156t2-s", ARMV6Itineraries, [HasV6T2Ops,
FeatureDSPThumb2]>;
def : Processor<"arm1156t2f-s", ARMV6Itineraries, [HasV6T2Ops, FeatureVFP2,
FeatureHasSlowFPVMLx,
FeatureDSPThumb2]>;
// V7a Processors.
// FIXME: A5 has currently the same Schedule model as A8
def : ProcessorModel<"cortex-a5", CortexA8Model,
[ProcA5, HasV7Ops, FeatureNEON, FeatureDB,
FeatureVFP4, FeatureDSPThumb2,
FeatureHasRAS, FeatureAClass]>;
def : ProcessorModel<"cortex-a7", CortexA8Model,
[ProcA7, HasV7Ops, FeatureNEON, FeatureDB,
FeatureDSPThumb2, FeatureHasRAS,
FeatureAClass]>;
def : ProcessorModel<"cortex-a8", CortexA8Model,
[ProcA8, HasV7Ops, FeatureNEON, FeatureDB,
FeatureDSPThumb2, FeatureHasRAS,
FeatureAClass]>;
def : ProcessorModel<"cortex-a9", CortexA9Model,
[ProcA9, HasV7Ops, FeatureNEON, FeatureDB,
FeatureDSPThumb2, FeatureHasRAS,
FeatureAClass]>;
def : ProcessorModel<"cortex-a9-mp", CortexA9Model,
[ProcA9, HasV7Ops, FeatureNEON, FeatureDB,
FeatureDSPThumb2, FeatureMP,
FeatureHasRAS, FeatureAClass]>;
// FIXME: A12 has currently the same Schedule model as A9
def : ProcessorModel<"cortex-a12", CortexA9Model,
[ProcA12, HasV7Ops, FeatureNEON, FeatureDB,
FeatureDSPThumb2, FeatureMP,
FeatureHasRAS, FeatureAClass]>;
// FIXME: A15 has currently the same ProcessorModel as A9.
def : ProcessorModel<"cortex-a15", CortexA9Model,
[ProcA15, HasV7Ops, FeatureNEON, FeatureDB,
FeatureDSPThumb2, FeatureHasRAS,
FeatureAClass]>;
// FIXME: R5 has currently the same ProcessorModel as A8.
def : ProcessorModel<"cortex-r5", CortexA8Model,
[ProcR5, HasV7Ops, FeatureDB,
FeatureVFP3, FeatureDSPThumb2,
FeatureHasRAS, FeatureVFPOnlySP,
FeatureD16, FeatureRClass]>;
// V7M Processors.
def : ProcNoItin<"cortex-m3", [HasV7Ops,
FeatureThumb2, FeatureNoARM, FeatureDB,
FeatureHWDiv, FeatureMClass]>;
// V7EM Processors.
def : ProcNoItin<"cortex-m4", [HasV7Ops,
FeatureThumb2, FeatureNoARM, FeatureDB,
FeatureHWDiv, FeatureDSPThumb2,
FeatureT2XtPk, FeatureVFP4,
FeatureVFPOnlySP, FeatureD16,
FeatureMClass]>;
// Swift uArch Processors.
def : ProcessorModel<"swift", SwiftModel,
[ProcSwift, HasV7Ops, FeatureNEON,
FeatureDB, FeatureDSPThumb2,
FeatureHasRAS, FeatureAClass]>;
// V8 Processors
def : ProcNoItin<"cortex-a53", [ProcA53, HasV8Ops, FeatureAClass,
FeatureDB, FeatureFPARMv8,
FeatureNEON, FeatureDSPThumb2]>;
def : ProcNoItin<"cortex-a57", [ProcA57, HasV8Ops, FeatureAClass,
FeatureDB, FeatureFPARMv8,
FeatureNEON, FeatureDSPThumb2]>;
// FIXME: krait has currently the same Schedule model as A9
def : ProcessorModel<"krait", CortexA9Model,
[ProcKrait, HasV7Ops,
FeatureNEON, FeatureDB,
FeatureDSPThumb2, FeatureHasRAS,
FeatureAClass]>;
//===----------------------------------------------------------------------===//
// Register File Description
//===----------------------------------------------------------------------===//
include "ARMRegisterInfo.td"
include "ARMCallingConv.td"
//===----------------------------------------------------------------------===//
// Instruction Descriptions
//===----------------------------------------------------------------------===//
include "ARMInstrInfo.td"
def ARMInstrInfo : InstrInfo;
//===----------------------------------------------------------------------===//
// Declare the target which we are implementing
//===----------------------------------------------------------------------===//
def ARM : Target {
// Pull in Instruction Info:
let InstructionSet = ARMInstrInfo;
}