llvm-project/llvm/utils/TableGen
Andrea Di Biagio 9da4d6db33 [MC][Tablegen] Allow the definition of processor register files in the scheduling model for llvm-mca
This patch allows the description of register files in processor scheduling
models. This addresses PR36662.

A new tablegen class named 'RegisterFile' has been added to TargetSchedule.td.
Targets can optionally describe register files for their processors using that
class. In particular, class RegisterFile allows to specify:
 - The total number of physical registers.
 - Which target registers are accessible through the register file.
 - The cost of allocating a register at register renaming stage.

Example (from this patch - see file X86/X86ScheduleBtVer2.td)

  def FpuPRF : RegisterFile<72, [VR64, VR128, VR256], [1, 1, 2]>

Here, FpuPRF describes a register file for MMX/XMM/YMM registers. On Jaguar
(btver2), a YMM register definition consumes 2 physical registers, while MMX/XMM
register definitions only cost 1 physical register.

The syntax allows to specify an empty set of register classes.  An empty set of
register classes means: this register file models all the registers specified by
the Target.  For each register class, users can specify an optional register
cost. By default, register costs default to 1.  A value of 0 for the number of
physical registers means: "this register file has an unbounded number of
physical registers".

This patch is structured in two parts.

* Part 1 - MC/Tablegen *

A first part adds the tablegen definition of RegisterFile, and teaches the
SubtargetEmitter how to emit information related to register files.

Information about register files is accessible through an instance of
MCExtraProcessorInfo.
The idea behind this design is to logically partition the processor description
which is only used by external tools (like llvm-mca) from the processor
information used by the llvm machine schedulers.
I think that this design would make easier for targets to get rid of the extra
processor information if they don't want it.

* Part 2 - llvm-mca related *

The second part of this patch is related to changes to llvm-mca.

The main differences are:
 1) class RegisterFile now needs to take into account the "cost of a register"
when allocating physical registers at register renaming stage.
 2) Point 1. triggered a minor refactoring which lef to the removal of the
"maximum 32 register files" restriction.
 3) The BackendStatistics view has been updated so that we can print out extra
details related to each register file implemented by the processor.

The effect of point 3. is also visible in tests register-files-[1..5].s.

Differential Revision: https://reviews.llvm.org/D44980

llvm-svn: 329067
2018-04-03 13:36:24 +00:00
..
AsmMatcherEmitter.cpp Fix signed/unsigned comparison warning in AsmGenMatcher generated code. NFCI. 2018-02-17 12:29:47 +00:00
AsmWriterEmitter.cpp [AArch64][TableGen] Skip tied result operands for InstAlias 2017-11-20 14:36:40 +00:00
AsmWriterInst.cpp
AsmWriterInst.h
Attributes.cpp Remove redundant includes from utils/TableGen. 2017-12-13 21:31:13 +00:00
CMakeLists.txt TableGen: Allow setting SDNodeProperties on intrinsics 2017-12-20 19:36:28 +00:00
CTagsEmitter.cpp Revert r318822 "[llvm-tblgen] - Stop using std::string in RecordKeeper." 2017-11-23 06:52:44 +00:00
CallingConvEmitter.cpp [TableGen] Simplify CallingConvEmitter.cpp. NFC. 2017-10-16 14:52:26 +00:00
CodeEmitterGen.cpp [tablegen] Avoid creating a temporary vector in getInstructionCase 2017-07-04 06:16:53 +00:00
CodeGenDAGPatterns.cpp Fix a bunch of typoes. NFC 2018-03-30 22:22:31 +00:00
CodeGenDAGPatterns.h [TableGen] Print more helpful information in case of type contradiction 2017-12-21 17:12:43 +00:00
CodeGenHwModes.cpp TableGen support for parameterized register class information 2017-09-14 16:56:21 +00:00
CodeGenHwModes.h TableGen support for parameterized register class information 2017-09-14 16:56:21 +00:00
CodeGenInstruction.cpp [globalisel][tablegen] Generalize pointer-type inference by introducing ptypeN. NFC 2017-11-18 00:16:44 +00:00
CodeGenInstruction.h Fix layering of MachineValueType.h by moving it from CodeGen to Support 2018-03-23 23:58:25 +00:00
CodeGenIntrinsics.h Fix layering of MachineValueType.h by moving it from CodeGen to Support 2018-03-23 23:58:25 +00:00
CodeGenMapTable.cpp [mips] Improve diagnostics for instruction mapping 2018-01-08 16:25:40 +00:00
CodeGenRegisters.cpp [X86] Add phony registers for high halves of regs with low halves 2018-03-20 18:46:55 +00:00
CodeGenRegisters.h Fix layering of MachineValueType.h by moving it from CodeGen to Support 2018-03-23 23:58:25 +00:00
CodeGenSchedule.cpp [MC][Tablegen] Allow the definition of processor register files in the scheduling model for llvm-mca 2018-04-03 13:36:24 +00:00
CodeGenSchedule.h [MC][Tablegen] Allow the definition of processor register files in the scheduling model for llvm-mca 2018-04-03 13:36:24 +00:00
CodeGenTarget.cpp TableGen: More helpful error messages 2018-04-01 17:08:49 +00:00
CodeGenTarget.h [MachineOperand][Target] MachineOperand::isRenamable semantics changes 2018-02-23 18:25:08 +00:00
DAGISelEmitter.cpp Allow separation of declarations and definitions in <Target>ISelDAGToDAG.inc 2017-11-10 18:36:04 +00:00
DAGISelMatcher.cpp Remove redundant includes from utils/TableGen. 2017-12-13 21:31:13 +00:00
DAGISelMatcher.h Fix layering of MachineValueType.h by moving it from CodeGen to Support 2018-03-23 23:58:25 +00:00
DAGISelMatcherEmitter.cpp [SelectionDAG] Add a isel matcher op to check the type of node results other than result 0. 2017-11-22 07:11:01 +00:00
DAGISelMatcherGen.cpp TableGen support for parameterized register class information 2017-09-14 16:56:21 +00:00
DAGISelMatcherOpt.cpp [TableGen] Use less stack in DAGISelMatcherOpt 2017-02-06 19:41:44 +00:00
DFAPacketizerEmitter.cpp Avoid int to string conversion in Twine or raw_ostream contexts. 2017-12-28 16:58:54 +00:00
DisassemblerEmitter.cpp Fix layering by moving X86DisassemblerDecoderCommon to Support 2018-03-23 23:58:20 +00:00
FastISelEmitter.cpp Strip trailing whitespace 2017-10-06 15:33:55 +00:00
FixedLenDecoderEmitter.cpp TableGen: Use DefInit::getDef() instead of the type's getRecord() 2018-03-05 14:01:30 +00:00
GlobalISelEmitter.cpp Fix layering of MachineValueType.h by moving it from CodeGen to Support 2018-03-23 23:58:25 +00:00
InfoByHwMode.cpp Remove redundant includes from utils/TableGen. 2017-12-13 21:31:13 +00:00
InfoByHwMode.h Fix layering of MachineValueType.h by moving it from CodeGen to Support 2018-03-23 23:58:25 +00:00
InstrDocsEmitter.cpp [Docs] Add tablegen backend for target opcode documentation 2017-11-14 15:35:15 +00:00
InstrInfoEmitter.cpp [MachineOperand][Target] MachineOperand::isRenamable semantics changes 2018-02-23 18:25:08 +00:00
IntrinsicEmitter.cpp Avoid int to string conversion in Twine or raw_ostream contexts. 2017-12-28 16:58:54 +00:00
LLVMBuild.txt Add missing dependency (headers are included from MC, so a link dependency could exist easily enough) 2018-03-29 00:29:43 +00:00
OptParserEmitter.cpp [Bash-autocompletion] Add support for -std= 2017-08-29 02:01:56 +00:00
PseudoLoweringEmitter.cpp
RegisterBankEmitter.cpp [globalisel][regbank] Warn about MIR ambiguities when register bank/class names clash. 2017-11-01 22:13:05 +00:00
RegisterInfoEmitter.cpp Fix layering of MachineValueType.h by moving it from CodeGen to Support 2018-03-23 23:58:25 +00:00
SDNodeProperties.cpp TableGen: Allow setting SDNodeProperties on intrinsics 2017-12-20 19:36:28 +00:00
SDNodeProperties.h TableGen: Allow setting SDNodeProperties on intrinsics 2017-12-20 19:36:28 +00:00
SearchableTableEmitter.cpp TableGen: Support Intrinsic values in SearchableTable 2018-04-01 17:08:58 +00:00
SequenceToOffsetTable.h Remove usages of deprecated std::unary_function and std::binary_function. 2017-09-14 18:33:25 +00:00
SubtargetEmitter.cpp [MC][Tablegen] Allow the definition of processor register files in the scheduling model for llvm-mca 2018-04-03 13:36:24 +00:00
SubtargetFeatureInfo.cpp Reverting r315590; it did not include changes for llvm-tblgen, which is causing link errors for several people. 2017-10-15 14:32:27 +00:00
SubtargetFeatureInfo.h [globalisel][tablegen] Compute available feature bits correctly. 2017-04-29 17:30:09 +00:00
TableGen.cpp Remove redundant includes from utils/TableGen. 2017-12-13 21:31:13 +00:00
TableGenBackends.h [Docs] Add tablegen backend for target opcode documentation 2017-11-14 15:35:15 +00:00
Types.cpp [globalisel][tablegen] Import SelectionDAG's rule predicates and support the equivalent in GIRule. 2017-04-21 15:59:56 +00:00
Types.h
X86DisassemblerShared.h [X86] Use unique_ptr to simplify memory management. NFC 2018-03-24 07:15:47 +00:00
X86DisassemblerTables.cpp [X86] Add a new disassembler opcode map for 3DNow. Stop treating 3DNow as an attribute. 2018-03-24 07:48:54 +00:00
X86DisassemblerTables.h [X86] Add a new disassembler opcode map for 3DNow. Stop treating 3DNow as an attribute. 2018-03-24 07:48:54 +00:00
X86EVEX2VEXTablesEmitter.cpp [X86] Rename VROUNDYPS* and VROUNDYPD* instructions to VROUNDPSY* and VROUNDPDY*. Fix itinerary mistake on all memory forms of VROUNDPD 2018-03-22 21:55:20 +00:00
X86FoldTablesEmitter.cpp Fix a bunch of typoes. NFC 2018-03-30 22:22:31 +00:00
X86ModRMFilters.cpp
X86ModRMFilters.h fix trivial typos in comments; NFC 2017-07-04 13:09:29 +00:00
X86RecognizableInstr.cpp [X86][TableGen] Add a missing error check to make sure EVEX instructions use one PS/PD/XS/XD prefixes. 2018-04-03 06:37:01 +00:00
X86RecognizableInstr.h [X86] Reduce number of OpPrefix bits in TSFlags to 2. NFCI 2018-04-03 06:37:04 +00:00
tdtags