diff options
| author | mike-m <mikem.llvm@gmail.com> | 2010-05-06 23:45:43 +0000 |
|---|---|---|
| committer | mike-m <mikem.llvm@gmail.com> | 2010-05-06 23:45:43 +0000 |
| commit | 68cb31901c590cabceee6e6356d62c84142114cb (patch) | |
| tree | 6444bddc975b662fbe47d63cd98a7b776a407c1a /docs/WritingAnLLVMBackend.html | |
| parent | c26ae5ab7e2d65b67c97524e66f50ce86445dec7 (diff) | |
| download | llvm-68cb31901c590cabceee6e6356d62c84142114cb.tar.gz llvm-68cb31901c590cabceee6e6356d62c84142114cb.tar.bz2 llvm-68cb31901c590cabceee6e6356d62c84142114cb.tar.xz | |
Overhauled llvm/clang docs builds. Closes PR6613.
NOTE: 2nd part changeset for cfe trunk to follow.
*** PRE-PATCH ISSUES ADDRESSED
- clang api docs fail build from objdir
- clang/llvm api docs collide in install PREFIX/
- clang/llvm main docs collide in install
- clang/llvm main docs have full of hard coded destination
assumptions and make use of absolute root in static html files;
namely CommandGuide tools hard codes a website destination
for cross references and some html cross references assume
website root paths
*** IMPROVEMENTS
- bumped Doxygen from 1.4.x -> 1.6.3
- splits llvm/clang docs into 'main' and 'api' (doxygen) build trees
- provide consistent, reliable doc builds for both main+api docs
- support buid vs. install vs. website intentions
- support objdir builds
- document targets with 'make help'
- correct clean and uninstall operations
- use recursive dir delete only where absolutely necessary
- added call function fn.RMRF which safeguards against botched 'rm -rf';
if any target (or any variable is evaluated) which attempts
to remove any dirs which match a hard-coded 'safelist', a verbose
error will be printed and make will error-stop.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103213 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'docs/WritingAnLLVMBackend.html')
| -rw-r--r-- | docs/WritingAnLLVMBackend.html | 2561 |
1 files changed, 0 insertions, 2561 deletions
diff --git a/docs/WritingAnLLVMBackend.html b/docs/WritingAnLLVMBackend.html deleted file mode 100644 index f49875c204..0000000000 --- a/docs/WritingAnLLVMBackend.html +++ /dev/null @@ -1,2561 +0,0 @@ -<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" - "http://www.w3.org/TR/html4/strict.dtd"> -<html> -<head> - <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> - <title>Writing an LLVM Compiler Backend</title> - <link rel="stylesheet" href="llvm.css" type="text/css"> -</head> - -<body> - -<div class="doc_title"> - Writing an LLVM Compiler Backend -</div> - -<ol> - <li><a href="#intro">Introduction</a> - <ul> - <li><a href="#Audience">Audience</a></li> - <li><a href="#Prerequisite">Prerequisite Reading</a></li> - <li><a href="#Basic">Basic Steps</a></li> - <li><a href="#Preliminaries">Preliminaries</a></li> - </ul> - <li><a href="#TargetMachine">Target Machine</a></li> - <li><a href="#TargetRegistration">Target Registration</a></li> - <li><a href="#RegisterSet">Register Set and Register Classes</a> - <ul> - <li><a href="#RegisterDef">Defining a Register</a></li> - <li><a href="#RegisterClassDef">Defining a Register Class</a></li> - <li><a href="#implementRegister">Implement a subclass of TargetRegisterInfo</a></li> - </ul></li> - <li><a href="#InstructionSet">Instruction Set</a> - <ul> - <li><a href="#operandMapping">Instruction Operand Mapping</a></li> - <li><a href="#implementInstr">Implement a subclass of TargetInstrInfo</a></li> - <li><a href="#branchFolding">Branch Folding and If Conversion</a></li> - </ul></li> - <li><a href="#InstructionSelector">Instruction Selector</a> - <ul> - <li><a href="#LegalizePhase">The SelectionDAG Legalize Phase</a> - <ul> - <li><a href="#promote">Promote</a></li> - <li><a href="#expand">Expand</a></li> - <li><a href="#custom">Custom</a></li> - <li><a href="#legal">Legal</a></li> - </ul></li> - <li><a href="#callingConventions">Calling Conventions</a></li> - </ul></li> - <li><a href="#assemblyPrinter">Assembly Printer</a></li> - <li><a href="#subtargetSupport">Subtarget Support</a></li> - <li><a href="#jitSupport">JIT Support</a> - <ul> - <li><a href="#mce">Machine Code Emitter</a></li> - <li><a href="#targetJITInfo">Target JIT Info</a></li> - </ul></li> -</ol> - -<div class="doc_author"> - <p>Written by <a href="http://www.woo.com">Mason Woo</a> and - <a href="http://misha.brukman.net">Misha Brukman</a></p> -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="intro">Introduction</a> -</div> -<!-- *********************************************************************** --> - -<div class="doc_text"> - -<p> -This document describes techniques for writing compiler backends that convert -the LLVM Intermediate Representation (IR) to code for a specified machine or -other languages. Code intended for a specific machine can take the form of -either assembly code or binary code (usable for a JIT compiler). -</p> - -<p> -The backend of LLVM features a target-independent code generator that may create -output for several types of target CPUs — including X86, PowerPC, Alpha, -and SPARC. The backend may also be used to generate code targeted at SPUs of the -Cell processor or GPUs to support the execution of compute kernels. -</p> - -<p> -The document focuses on existing examples found in subdirectories -of <tt>llvm/lib/Target</tt> in a downloaded LLVM release. In particular, this -document focuses on the example of creating a static compiler (one that emits -text assembly) for a SPARC target, because SPARC has fairly standard -characteristics, such as a RISC instruction set and straightforward calling -conventions. -</p> - -</div> - -<div class="doc_subsection"> - <a name="Audience">Audience</a> -</div> - -<div class="doc_text"> - -<p> -The audience for this document is anyone who needs to write an LLVM backend to -generate code for a specific hardware or software target. -</p> - -</div> - -<div class="doc_subsection"> - <a name="Prerequisite">Prerequisite Reading</a> -</div> - -<div class="doc_text"> - -<p> -These essential documents must be read before reading this document: -</p> - -<ul> -<li><i><a href="http://www.llvm.org/docs/LangRef.html">LLVM Language Reference - Manual</a></i> — a reference manual for the LLVM assembly language.</li> - -<li><i><a href="http://www.llvm.org/docs/CodeGenerator.html">The LLVM - Target-Independent Code Generator</a></i> — a guide to the components - (classes and code generation algorithms) for translating the LLVM internal - representation into machine code for a specified target. Pay particular - attention to the descriptions of code generation stages: Instruction - Selection, Scheduling and Formation, SSA-based Optimization, Register - Allocation, Prolog/Epilog Code Insertion, Late Machine Code Optimizations, - and Code Emission.</li> - -<li><i><a href="http://www.llvm.org/docs/TableGenFundamentals.html">TableGen - Fundamentals</a></i> —a document that describes the TableGen - (<tt>tblgen</tt>) application that manages domain-specific information to - support LLVM code generation. TableGen processes input from a target - description file (<tt>.td</tt> suffix) and generates C++ code that can be - used for code generation.</li> - -<li><i><a href="http://www.llvm.org/docs/WritingAnLLVMPass.html">Writing an LLVM - Pass</a></i> — The assembly printer is a <tt>FunctionPass</tt>, as are - several SelectionDAG processing steps.</li> -</ul> - -<p> -To follow the SPARC examples in this document, have a copy of -<i><a href="http://www.sparc.org/standards/V8.pdf">The SPARC Architecture -Manual, Version 8</a></i> for reference. For details about the ARM instruction -set, refer to the <i><a href="http://infocenter.arm.com/">ARM Architecture -Reference Manual</a></i>. For more about the GNU Assembler format -(<tt>GAS</tt>), see -<i><a href="http://sourceware.org/binutils/docs/as/index.html">Using As</a></i>, -especially for the assembly printer. <i>Using As</i> contains a list of target -machine dependent features. -</p> - -</div> - -<div class="doc_subsection"> - <a name="Basic">Basic Steps</a> -</div> - -<div class="doc_text"> - -<p> -To write a compiler backend for LLVM that converts the LLVM IR to code for a -specified target (machine or other language), follow these steps: -</p> - -<ul> -<li>Create a subclass of the TargetMachine class that describes characteristics - of your target machine. Copy existing examples of specific TargetMachine - class and header files; for example, start with - <tt>SparcTargetMachine.cpp</tt> and <tt>SparcTargetMachine.h</tt>, but - change the file names for your target. Similarly, change code that - references "Sparc" to reference your target. </li> - -<li>Describe the register set of the target. Use TableGen to generate code for - register definition, register aliases, and register classes from a - target-specific <tt>RegisterInfo.td</tt> input file. You should also write - additional code for a subclass of the TargetRegisterInfo class that - represents the class register file data used for register allocation and - also describes the interactions between registers.</li> - -<li>Describe the instruction set of the target. Use TableGen to generate code - for target-specific instructions from target-specific versions of - <tt>TargetInstrFormats.td</tt> and <tt>TargetInstrInfo.td</tt>. You should - write additional code for a subclass of the TargetInstrInfo class to - represent machine instructions supported by the target machine. </li> - -<li>Describe the selection and conversion of the LLVM IR from a Directed Acyclic - Graph (DAG) representation of instructions to native target-specific - instructions. Use TableGen to generate code that matches patterns and - selects instructions based on additional information in a target-specific - version of <tt>TargetInstrInfo.td</tt>. Write code - for <tt>XXXISelDAGToDAG.cpp</tt>, where XXX identifies the specific target, - to perform pattern matching and DAG-to-DAG instruction selection. Also write - code in <tt>XXXISelLowering.cpp</tt> to replace or remove operations and - data types that are not supported natively in a SelectionDAG. </li> - -<li>Write code for an assembly printer that converts LLVM IR to a GAS format for - your target machine. You should add assembly strings to the instructions - defined in your target-specific version of <tt>TargetInstrInfo.td</tt>. You - should also write code for a subclass of AsmPrinter that performs the - LLVM-to-assembly conversion and a trivial subclass of TargetAsmInfo.</li> - -<li>Optionally, add support for subtargets (i.e., variants with different - capabilities). You should also write code for a subclass of the - TargetSubtarget class, which allows you to use the <tt>-mcpu=</tt> - and <tt>-mattr=</tt> command-line options.</li> - -<li>Optionally, add JIT support and create a machine code emitter (subclass of - TargetJITInfo) that is used to emit binary code directly into memory. </li> -</ul> - -<p> -In the <tt>.cpp</tt> and <tt>.h</tt>. files, initially stub up these methods and -then implement them later. Initially, you may not know which private members -that the class will need and which components will need to be subclassed. -</p> - -</div> - -<div class="doc_subsection"> - <a name="Preliminaries">Preliminaries</a> -</div> - -<div class="doc_text"> - -<p> -To actually create your compiler backend, you need to create and modify a few -files. The absolute minimum is discussed here. But to actually use the LLVM -target-independent code generator, you must perform the steps described in -the <a href="http://www.llvm.org/docs/CodeGenerator.html">LLVM -Target-Independent Code Generator</a> document. -</p> - -<p> -First, you should create a subdirectory under <tt>lib/Target</tt> to hold all -the files related to your target. If your target is called "Dummy," create the -directory <tt>lib/Target/Dummy</tt>. -</p> - -<p> -In this new -directory, create a <tt>Makefile</tt>. It is easiest to copy a -<tt>Makefile</tt> of another target and modify it. It should at least contain -the <tt>LEVEL</tt>, <tt>LIBRARYNAME</tt> and <tt>TARGET</tt> variables, and then -include <tt>$(LEVEL)/Makefile.common</tt>. The library can be -named <tt>LLVMDummy</tt> (for example, see the MIPS target). Alternatively, you -can split the library into <tt>LLVMDummyCodeGen</tt> -and <tt>LLVMDummyAsmPrinter</tt>, the latter of which should be implemented in a -subdirectory below <tt>lib/Target/Dummy</tt> (for example, see the PowerPC -target). -</p> - -<p> -Note that these two naming schemes are hardcoded into <tt>llvm-config</tt>. -Using any other naming scheme will confuse <tt>llvm-config</tt> and produce a -lot of (seemingly unrelated) linker errors when linking <tt>llc</tt>. -</p> - -<p> -To make your target actually do something, you need to implement a subclass of -<tt>TargetMachine</tt>. This implementation should typically be in the file -<tt>lib/Target/DummyTargetMachine.cpp</tt>, but any file in -the <tt>lib/Target</tt> directory will be built and should work. To use LLVM's -target independent code generator, you should do what all current machine -backends do: create a subclass of <tt>LLVMTargetMachine</tt>. (To create a -target from scratch, create a subclass of <tt>TargetMachine</tt>.) -</p> - -<p> -To get LLVM to actually build and link your target, you need to add it to -the <tt>TARGETS_TO_BUILD</tt> variable. To do this, you modify the configure -script to know about your target when parsing the <tt>--enable-targets</tt> -option. Search the configure script for <tt>TARGETS_TO_BUILD</tt>, add your -target to the lists there (some creativity required), and then -reconfigure. Alternatively, you can change <tt>autotools/configure.ac</tt> and -regenerate configure by running <tt>./autoconf/AutoRegen.sh</tt>. -</p> - -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="TargetMachine">Target Machine</a> -</div> -<!-- *********************************************************************** --> - -<div class="doc_text"> - -<p> -<tt>LLVMTargetMachine</tt> is designed as a base class for targets implemented -with the LLVM target-independent code generator. The <tt>LLVMTargetMachine</tt> -class should be specialized by a concrete target class that implements the -various virtual methods. <tt>LLVMTargetMachine</tt> is defined as a subclass of -<tt>TargetMachine</tt> in <tt>include/llvm/Target/TargetMachine.h</tt>. The -<tt>TargetMachine</tt> class implementation (<tt>TargetMachine.cpp</tt>) also -processes numerous command-line options. -</p> - -<p> -To create a concrete target-specific subclass of <tt>LLVMTargetMachine</tt>, -start by copying an existing <tt>TargetMachine</tt> class and header. You -should name the files that you create to reflect your specific target. For -instance, for the SPARC target, name the files <tt>SparcTargetMachine.h</tt> and -<tt>SparcTargetMachine.cpp</tt>. -</p> - -<p> -For a target machine <tt>XXX</tt>, the implementation of -<tt>XXXTargetMachine</tt> must have access methods to obtain objects that -represent target components. These methods are named <tt>get*Info</tt>, and are -intended to obtain the instruction set (<tt>getInstrInfo</tt>), register set -(<tt>getRegisterInfo</tt>), stack frame layout (<tt>getFrameInfo</tt>), and -similar information. <tt>XXXTargetMachine</tt> must also implement the -<tt>getTargetData</tt> method to access an object with target-specific data -characteristics, such as data type size and alignment requirements. -</p> - -<p> -For instance, for the SPARC target, the header file -<tt>SparcTargetMachine.h</tt> declares prototypes for several <tt>get*Info</tt> -and <tt>getTargetData</tt> methods that simply return a class member. -</p> - -<div class="doc_code"> -<pre> -namespace llvm { - -class Module; - -class SparcTargetMachine : public LLVMTargetMachine { - const TargetData DataLayout; // Calculates type size & alignment - SparcSubtarget Subtarget; - SparcInstrInfo InstrInfo; - TargetFrameInfo FrameInfo; - -protected: - virtual const TargetAsmInfo *createTargetAsmInfo() const; - -public: - SparcTargetMachine(const Module &M, const std::string &FS); - - virtual const SparcInstrInfo *getInstrInfo() const {return &InstrInfo; } - virtual const TargetFrameInfo *getFrameInfo() const {return &FrameInfo; } - virtual const TargetSubtarget *getSubtargetImpl() const{return &Subtarget; } - virtual const TargetRegisterInfo *getRegisterInfo() const { - return &InstrInfo.getRegisterInfo(); - } - virtual const TargetData *getTargetData() const { return &DataLayout; } - static unsigned getModuleMatchQuality(const Module &M); - - // Pass Pipeline Configuration - virtual bool addInstSelector(PassManagerBase &PM, bool Fast); - virtual bool addPreEmitPass(PassManagerBase &PM, bool Fast); -}; - -} // end namespace llvm -</pre> -</div> - -</div> - - -<div class="doc_text"> - -<ul> -<li><tt>getInstrInfo()</tt></li> -<li><tt>getRegisterInfo()</tt></li> -<li><tt>getFrameInfo()</tt></li> -<li><tt>getTargetData()</tt></li> -<li><tt>getSubtargetImpl()</tt></li> -</ul> - -<p>For some targets, you also need to support the following methods:</p> - -<ul> -<li><tt>getTargetLowering()</tt></li> -<li><tt>getJITInfo()</tt></li> -</ul> - -<p> -In addition, the <tt>XXXTargetMachine</tt> constructor should specify a -<tt>TargetDescription</tt> string that determines the data layout for the target -machine, including characteristics such as pointer size, alignment, and -endianness. For example, the constructor for SparcTargetMachine contains the -following: -</p> - -<div class="doc_code"> -<pre> -SparcTargetMachine::SparcTargetMachine(const Module &M, const std::string &FS) - : DataLayout("E-p:32:32-f128:128:128"), - Subtarget(M, FS), InstrInfo(Subtarget), - FrameInfo(TargetFrameInfo::StackGrowsDown, 8, 0) { -} -</pre> -</div> - -</div> - -<div class="doc_text"> - -<p>Hyphens separate portions of the <tt>TargetDescription</tt> string.</p> - -<ul> -<li>An upper-case "<tt>E</tt>" in the string indicates a big-endian target data - model. a lower-case "<tt>e</tt>" indicates little-endian.</li> - -<li>"<tt>p:</tt>" is followed by pointer information: size, ABI alignment, and - preferred alignment. If only two figures follow "<tt>p:</tt>", then the - first value is pointer size, and the second value is both ABI and preferred - alignment.</li> - -<li>Then a letter for numeric type alignment: "<tt>i</tt>", "<tt>f</tt>", - "<tt>v</tt>", or "<tt>a</tt>" (corresponding to integer, floating point, - vector, or aggregate). "<tt>i</tt>", "<tt>v</tt>", or "<tt>a</tt>" are - followed by ABI alignment and preferred alignment. "<tt>f</tt>" is followed - by three values: the first indicates the size of a long double, then ABI - alignment, and then ABI preferred alignment.</li> -</ul> - -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="TargetRegistration">Target Registration</a> -</div> -<!-- *********************************************************************** --> - -<div class="doc_text"> - -<p> -You must also register your target with the <tt>TargetRegistry</tt>, which is -what other LLVM tools use to be able to lookup and use your target at -runtime. The <tt>TargetRegistry</tt> can be used directly, but for most targets -there are helper templates which should take care of the work for you.</p> - -<p> -All targets should declare a global <tt>Target</tt> object which is used to -represent the target during registration. Then, in the target's TargetInfo -library, the target should define that object and use -the <tt>RegisterTarget</tt> template to register the target. For example, the Sparc registration code looks like this: -</p> - -<div class="doc_code"> -<pre> -Target llvm::TheSparcTarget; - -extern "C" void LLVMInitializeSparcTargetInfo() { - RegisterTarget<Triple::sparc, /*HasJIT=*/false> - X(TheSparcTarget, "sparc", "Sparc"); -} -</pre> -</div> - -<p> -This allows the <tt>TargetRegistry</tt> to look up the target by name or by -target triple. In addition, most targets will also register additional features -which are available in separate libraries. These registration steps are -separate, because some clients may wish to only link in some parts of the target --- the JIT code generator does not require the use of the assembler printer, for -example. Here is an example of registering the Sparc assembly printer: -</p> - -<div class="doc_code"> -<pre> -extern "C" void LLVMInitializeSparcAsmPrinter() { - RegisterAsmPrinter<SparcAsmPrinter> X(TheSparcTarget); -} -</pre> -</div> - -<p> -For more information, see -"<a href="/doxygen/TargetRegistry_8h-source.html">llvm/Target/TargetRegistry.h</a>". -</p> - -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="RegisterSet">Register Set and Register Classes</a> -</div> -<!-- *********************************************************************** --> - -<div class="doc_text"> - -<p> -You should describe a concrete target-specific class that represents the -register file of a target machine. This class is called <tt>XXXRegisterInfo</tt> -(where <tt>XXX</tt> identifies the target) and represents the class register -file data that is used for register allocation. It also describes the -interactions between registers. -</p> - -<p> -You also need to define register classes to categorize related registers. A -register class should be added for groups of registers that are all treated the -same way for some instruction. Typical examples are register classes for -integer, floating-point, or vector registers. A register allocator allows an -instruction to use any register in a specified register class to perform the -instruction in a similar manner. Register classes allocate virtual registers to -instructions from these sets, and register classes let the target-independent -register allocator automatically choose the actual registers. -</p> - -<p> -Much of the code for registers, including register definition, register aliases, -and register classes, is generated by TableGen from <tt>XXXRegisterInfo.td</tt> -input files and placed in <tt>XXXGenRegisterInfo.h.inc</tt> and -<tt>XXXGenRegisterInfo.inc</tt> output files. Some of the code in the -implementation of <tt>XXXRegisterInfo</tt> requires hand-coding. -</p> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="RegisterDef">Defining a Register</a> -</div> - -<div class="doc_text"> - -<p> -The <tt>XXXRegisterInfo.td</tt> file typically starts with register definitions -for a target machine. The <tt>Register</tt> class (specified -in <tt>Target.td</tt>) is used to define an object for each register. The -specified string <tt>n</tt> becomes the <tt>Name</tt> of the register. The -basic <tt>Register</tt> object does not have any subregisters and does not -specify any aliases. -</p> - -<div class="doc_code"> -<pre> -class Register<string n> { - string Namespace = ""; - string AsmName = n; - string Name = n; - int SpillSize = 0; - int SpillAlignment = 0; - list<Register> Aliases = []; - list<Register> SubRegs = []; - list<int> DwarfNumbers = []; -} -</pre> -</div> - -<p> -For example, in the <tt>X86RegisterInfo.td</tt> file, there are register -definitions that utilize the Register class, such as: -</p> - -<div class="doc_code"> -<pre> -def AL : Register<"AL">, DwarfRegNum<[0, 0, 0]>; -</pre> -</div> - -<p> -This defines the register <tt>AL</tt> and assigns it values (with -<tt>DwarfRegNum</tt>) that are used by <tt>gcc</tt>, <tt>gdb</tt>, or a debug -information writer to identify a register. For register -<tt>AL</tt>, <tt>DwarfRegNum</tt> takes an array of 3 values representing 3 -different modes: the first element is for X86-64, the second for exception -handling (EH) on X86-32, and the third is generic. -1 is a special Dwarf number -that indicates the gcc number is undefined, and -2 indicates the register number -is invalid for this mode. -</p> - -<p> -From the previously described line in the <tt>X86RegisterInfo.td</tt> file, -TableGen generates this code in the <tt>X86GenRegisterInfo.inc</tt> file: -</p> - -<div class="doc_code"> -<pre> -static const unsigned GR8[] = { X86::AL, ... }; - -const unsigned AL_AliasSet[] = { X86::AX, X86::EAX, X86::RAX, 0 }; - -const TargetRegisterDesc RegisterDescriptors[] = { - ... -{ "AL", "AL", AL_AliasSet, Empty_SubRegsSet, Empty_SubRegsSet, AL_SuperRegsSet }, ... -</pre> -</div> - -<p> -From the register info file, TableGen generates a <tt>TargetRegisterDesc</tt> -object for each register. <tt>TargetRegisterDesc</tt> is defined in -<tt>include/llvm/Target/TargetRegisterInfo.h</tt> with the following fields: -</p> - -<div class="doc_code"> -<pre> -struct TargetRegisterDesc { - const char *AsmName; // Assembly language name for the register - const char *Name; // Printable name for the reg (for debugging) - const unsigned *AliasSet; // Register Alias Set - const unsigned *SubRegs; // Sub-register set - const unsigned *ImmSubRegs; // Immediate sub-register set - const unsigned *SuperRegs; // Super-register set -};</pre> -</div> - -<p> -TableGen uses the entire target description file (<tt>.td</tt>) to determine -text names for the register (in the <tt>AsmName</tt> and <tt>Name</tt> fields of -<tt>TargetRegisterDesc</tt>) and the relationships of other registers to the -defined register (in the other <tt>TargetRegisterDesc</tt> fields). In this -example, other definitions establish the registers "<tt>AX</tt>", -"<tt>EAX</tt>", and "<tt>RAX</tt>" as aliases for one another, so TableGen -generates a null-terminated array (<tt>AL_AliasSet</tt>) for this register alias -set. -</p> - -<p> -The <tt>Register</tt> class is commonly used as a base class for more complex -classes. In <tt>Target.td</tt>, the <tt>Register</tt> class is the base for the -<tt>RegisterWithSubRegs</tt> class that is used to define registers that need to -specify subregisters in the <tt>SubRegs</tt> list, as shown here: -</p> - -<div class="doc_code"> -<pre> -class RegisterWithSubRegs<string n, -list<Register> subregs> : Register<n> { - let SubRegs = subregs; -} -</pre> -</div> - -<p> -In <tt>SparcRegisterInfo.td</tt>, additional register classes are defined for -SPARC: a Register subclass, SparcReg, and further subclasses: <tt>Ri</tt>, -<tt>Rf</tt>, and <tt>Rd</tt>. SPARC registers are identified by 5-bit ID -numbers, which is a feature common to these subclasses. Note the use of -'<tt>let</tt>' expressions to override values that are initially defined in a -superclass (such as <tt>SubRegs</tt> field in the <tt>Rd</tt> class). -</p> - -<div class="doc_code"> -<pre> -class SparcReg<string n> : Register<n> { - field bits<5> Num; - let Namespace = "SP"; -} -// Ri - 32-bit integer registers -class Ri<bits<5> num, string n> : -SparcReg<n> { - let Num = num; -} -// Rf - 32-bit floating-point registers -class Rf<bits<5> num, string n> : -SparcReg<n> { - let Num = num; -} -// Rd - Slots in the FP register file for 64-bit -floating-point values. -class Rd<bits<5> num, string n, -list<Register> subregs> : SparcReg<n> { - let Num = num; - let SubRegs = subregs; -} -</pre> -</div> - -<p> -In the <tt>SparcRegisterInfo.td</tt> file, there are register definitions that -utilize these subclasses of <tt>Register</tt>, such as: -</p> - -<div class="doc_code"> -<pre> -def G0 : Ri< 0, "G0">, -DwarfRegNum<[0]>; -def G1 : Ri< 1, "G1">, DwarfRegNum<[1]>; -... -def F0 : Rf< 0, "F0">, -DwarfRegNum<[32]>; -def F1 : Rf< 1, "F1">, -DwarfRegNum<[33]>; -... -def D0 : Rd< 0, "F0", [F0, F1]>, -DwarfRegNum<[32]>; -def D1 : Rd< 2, "F2", [F2, F3]>, -DwarfRegNum<[34]>; -</pre> -</div> - -<p> -The last two registers shown above (<tt>D0</tt> and <tt>D1</tt>) are -double-precision floating-point registers that are aliases for pairs of -single-precision floating-point sub-registers. In addition to aliases, the -sub-register and super-register relationships of the defined register are in -fields of a register's TargetRegisterDesc. -</p> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="RegisterClassDef">Defining a Register Class</a> -</div> - -<div class="doc_text"> - -<p> -The <tt>RegisterClass</tt> class (specified in <tt>Target.td</tt>) is used to -define an object that represents a group of related registers and also defines -the default allocation order of the registers. A target description file -<tt>XXXRegisterInfo.td</tt> that uses <tt>Target.td</tt> can construct register -classes using the following class: -</p> - -<div class="doc_code"> -<pre> -class RegisterClass<string namespace, -list<ValueType> regTypes, int alignment, - list<Register> regList> { - string Namespace = namespace; - list<ValueType> RegTypes = regTypes; - int Size = 0; // spill size, in bits; zero lets tblgen pick the size - int Alignment = alignment; - - // CopyCost is the cost of copying a value between two registers - // default value 1 means a single instruction - // A negative value means copying is extremely expensive or impossible - int CopyCost = 1; - list<Register> MemberList = regList; - - // for register classes that are subregisters of this class - list<RegisterClass> SubRegClassList = []; - - code MethodProtos = [{}]; // to insert arbitrary code - code MethodBodies = [{}]; -} -</pre> -</div> - -<p>To define a RegisterClass, use the following 4 arguments:</p> - -<ul> -<li>The first argument of the definition is the name of the namespace.</li> - -<li>The second argument is a list of <tt>ValueType</tt> register type values - that are defined in <tt>include/llvm/CodeGen/ValueTypes.td</tt>. Defined - values include integer types (such as <tt>i16</tt>, <tt>i32</tt>, - and <tt>i1</tt> for Boolean), floating-point types - (<tt>f32</tt>, <tt>f64</tt>), and vector types (for example, <tt>v8i16</tt> - for an <tt>8 x i16</tt> vector). All registers in a <tt>RegisterClass</tt> - must have the same <tt>ValueType</tt>, but some registers may store vector - data in different configurations. For example a register that can process a - 128-bit vector may be able to handle 16 8-bit integer elements, 8 16-bit - integers, 4 32-bit integers, and so on. </li> - -<li>The third argument of the <tt>RegisterClass</tt> definition specifies the - alignment required of the registers when they are stored or loaded to - memory.</li> - -<li>The final argument, <tt>regList</tt>, specifies which registers are in this - class. If an <tt>allocation_order_*</tt> method is not specified, - then <tt>regList</tt> also defines the order of allocation used by the - register allocator.</li> -</ul> - -<p> -In <tt>SparcRegisterInfo.td</tt>, three RegisterClass objects are defined: -<tt>FPRegs</tt>, <tt>DFPRegs</tt>, and <tt>IntRegs</tt>. For all three register -classes, the first argument defines the namespace with the string -'<tt>SP</tt>'. <tt>FPRegs</tt> defines a group of 32 single-precision -floating-point registers (<tt>F0</tt> to <tt>F31</tt>); <tt>DFPRegs</tt> defines -a group of 16 double-precision registers -(<tt>D0-D15</tt>). For <tt>IntRegs</tt>, the <tt>MethodProtos</tt> -and <tt>MethodBodies</tt> methods are used by TableGen to insert the specified -code into generated output. -</p> - -<div class="doc_code"> -<pre> -def FPRegs : RegisterClass<"SP", [f32], 32, - [F0, F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F12, F13, F14, F15, - F16, F17, F18, F19, F20, F21, F22, F23, F24, F25, F26, F27, F28, F29, F30, F31]>; - -def DFPRegs : RegisterClass<"SP", [f64], 64, - [D0, D1, D2, D3, D4, D5, D6, D7, D8, D9, D10, D11, D12, D13, D14, D15]>; - -def IntRegs : RegisterClass<"SP", [i32], 32, - [L0, L1, L2, L3, L4, L5, L6, L7, - I0, I1, I2, I3, I4, I5, - O0, O1, O2, O3, O4, O5, O7, - G1, - // Non-allocatable regs: - G2, G3, G4, - O6, // stack ptr - I6, // frame ptr - I7, // return address - G0, // constant zero - G5, G6, G7 // reserved for kernel - ]> { - let MethodProtos = [{ - iterator allocation_order_end(const MachineFunction &MF) const; - }]; - let MethodBodies = [{ - IntRegsClass::iterator - IntRegsClass::allocation_order_end(const MachineFunction &MF) const { - return end() - 10 // Don't allocate special registers - -1; - } - }]; -} -</pre> -</div> - -<p> -Using <tt>SparcRegisterInfo.td</tt> with TableGen generates several output files -that are intended for inclusion in other source code that you write. -<tt>SparcRegisterInfo.td</tt> generates <tt>SparcGenRegisterInfo.h.inc</tt>, -which should be included in the header file for the implementation of the SPARC -register implementation that you write (<tt>SparcRegisterInfo.h</tt>). In -<tt>SparcGenRegisterInfo.h.inc</tt> a new structure is defined called -<tt>SparcGenRegisterInfo</tt> that uses <tt>TargetRegisterInfo</tt> as its -base. It also specifies types, based upon the defined register -classes: <tt>DFPRegsClass</tt>, <tt>FPRegsClass</tt>, and <tt>IntRegsClass</tt>. -</p> - -<p> -<tt>SparcRegisterInfo.td</tt> also generates <tt>SparcGenRegisterInfo.inc</tt>, -which is included at the bottom of <tt>SparcRegisterInfo.cpp</tt>, the SPARC -register implementation. The code below shows only the generated integer -registers and associated register classes. The order of registers -in <tt>IntRegs</tt> reflects the order in the definition of <tt>IntRegs</tt> in -the target description file. Take special note of the use -of <tt>MethodBodies</tt> in <tt>SparcRegisterInfo.td</tt> to create code in -<tt>SparcGenRegisterInfo.inc</tt>. <tt>MethodProtos</tt> generates similar code -in <tt>SparcGenRegisterInfo.h.inc</tt>. -</p> - -<div class="doc_code"> -<pre> // IntRegs Register Class... - static const unsigned IntRegs[] = { - SP::L0, SP::L1, SP::L2, SP::L3, SP::L4, SP::L5, - SP::L6, SP::L7, SP::I0, SP::I1, SP::I2, SP::I3, - SP::I4, SP::I5, SP::O0, SP::O1, SP::O2, SP::O3, - SP::O4, SP::O5, SP::O7, SP::G1, SP::G2, SP::G3, - SP::G4, SP::O6, SP::I6, SP::I7, SP::G0, SP::G5, - SP::G6, SP::G7, - }; - - // IntRegsVTs Register Class Value Types... - static const MVT::ValueType IntRegsVTs[] = { - MVT::i32, MVT::Other - }; - -namespace SP { // Register class instances - DFPRegsClass DFPRegsRegClass; - FPRegsClass FPRegsRegClass; - IntRegsClass IntRegsRegClass; -... - // IntRegs Sub-register Classess... - static const TargetRegisterClass* const IntRegsSubRegClasses [] = { - NULL - }; -... - // IntRegs Super-register Classess... - static const TargetRegisterClass* const IntRegsSuperRegClasses [] = { - NULL - }; -... - // IntRegs Register Class sub-classes... - static const TargetRegisterClass* const IntRegsSubclasses [] = { - NULL - }; -... - // IntRegs Register Class super-classes... - static const TargetRegisterClass* const IntRegsSuperclasses [] = { - NULL - }; -... - IntRegsClass::iterator - IntRegsClass::allocation_order_end(const MachineFunction &MF) const { - return end()-10 // Don't allocate special registers - -1; - } - - IntRegsClass::IntRegsClass() : TargetRegisterClass(IntRegsRegClassID, - IntRegsVTs, IntRegsSubclasses, IntRegsSuperclasses, IntRegsSubRegClasses, - IntRegsSuperRegClasses, 4, 4, 1, IntRegs, IntRegs + 32) {} -} -</pre> -</div> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="implementRegister">Implement a subclass of</a> - <a href="http://www.llvm.org/docs/CodeGenerator.html#targetregisterinfo">TargetRegisterInfo</a> -</div> - -<div class="doc_text"> - -<p> -The final step is to hand code portions of <tt>XXXRegisterInfo</tt>, which -implements the interface described in <tt>TargetRegisterInfo.h</tt>. These -functions return <tt>0</tt>, <tt>NULL</tt>, or <tt>false</tt>, unless -overridden. Here is a list of functions that are overridden for the SPARC -implementation in <tt>SparcRegisterInfo.cpp</tt>: -</p> - -<ul> -<li><tt>getCalleeSavedRegs</tt> — Returns a list of callee-saved registers - in the order of the desired callee-save stack frame offset.</li> - -<li><tt>getCalleeSavedRegClasses</tt> — Returns a list of preferred - register classes with which to spill each callee saved register.</li> - -<li><tt>getReservedRegs</tt> — Returns a bitset indexed by physical - register numbers, indicating if a particular register is unavailable.</li> - -<li><tt>hasFP</tt> — Return a Boolean indicating if a function should have - a dedicated frame pointer register.</li> - -<li><tt>eliminateCallFramePseudoInstr</tt> — If call frame setup or - destroy pseudo instructions are used, this can be called to eliminate - them.</li> - -<li><tt>eliminateFrameIndex</tt> — Eliminate abstract frame indices from - instructions that may use them.</li> - -<li><tt>emitPrologue</tt> — Insert prologue code into the function.</li> - -<li><tt>emitEpilogue</tt> — Insert epilogue code into the function.</li> -</ul> - -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="InstructionSet">Instruction Set</a> -</div> - -<!-- *********************************************************************** --> -<div class="doc_text"> - -<p> -During the early stages of code generation, the LLVM IR code is converted to a -<tt>SelectionDAG</tt> with nodes that are instances of the <tt>SDNode</tt> class -containing target instructions. An <tt>SDNode</tt> has an opcode, operands, type -requirements, and operation properties. For example, is an operation -commutative, does an operation load from memory. The various operation node -types are described in the <tt>include/llvm/CodeGen/SelectionDAGNodes.h</tt> -file (values of the <tt>NodeType</tt> enum in the <tt>ISD</tt> namespace). -</p> - -<p> -TableGen uses the following target description (<tt>.td</tt>) input files to -generate much of the code for instruction definition: -</p> - -<ul> -<li><tt>Target.td</tt> — Where the <tt>Instruction</tt>, <tt>Operand</tt>, - <tt>InstrInfo</tt>, and other fundamental classes are defined.</li> - -<li><tt>TargetSelectionDAG.td</tt>— Used by <tt>SelectionDAG</tt> - instruction selection generators, contains <tt>SDTC*</tt> classes (selection - DAG type constraint), definitions of <tt>SelectionDAG</tt> nodes (such as - <tt>imm</tt>, <tt>cond</tt>, <tt>bb</tt>, <tt>add</tt>, <tt>fadd</tt>, - <tt>sub</tt>), and pattern support (<tt>Pattern</tt>, <tt>Pat</tt>, - <tt>PatFrag</tt>, <tt>PatLeaf</tt>, <tt>ComplexPattern</tt>.</li> - -<li><tt>XXXInstrFormats.td</tt> — Patterns for definitions of - target-specific instructions.</li> - -<li><tt>XXXInstrInfo.td</tt> — Target-specific definitions of instruction - templates, condition codes, and instructions of an instruction set. For - architecture modifications, a different file name may be used. For example, - for Pentium with SSE instruction, this file is <tt>X86InstrSSE.td</tt>, and - for Pentium with MMX, this file is <tt>X86InstrMMX.td</tt>.</li> -</ul> - -<p> -There is also a target-specific <tt>XXX.td</tt> file, where <tt>XXX</tt> is the -name of the target. The <tt>XXX.td</tt> file includes the other <tt>.td</tt> -input files, but its contents are only directly important for subtargets. -</p> - -<p> -You should describe a concrete target-specific class <tt>XXXInstrInfo</tt> that -represents machine instructions supported by a target machine. -<tt>XXXInstrInfo</tt> contains an array of <tt>XXXInstrDescriptor</tt> objects, -each of which describes one instruction. An instruction descriptor defines:</p> - -<ul> -<li>Opcode mnemonic</li> - -<li>Number of operands</li> - -<li>List of implicit register definitions and uses</li> - -<li>Target-independent properties (such as memory access, is commutable)</li> - -<li>Target-specific flags </li> -</ul> - -<p> -The Instruction class (defined in <tt>Target.td</tt>) is mostly used as a base -for more complex instruction classes. -</p> - -<div class="doc_code"> -<pre>class Instruction { - string Namespace = ""; - dag OutOperandList; // An dag containing the MI def operand list. - dag InOperandList; // An dag containing the MI use operand list. - string AsmString = ""; // The .s format to print the instruction with. - list<dag> Pattern; // Set to the DAG pattern for this instruction - list<Register> Uses = []; - list<Register> Defs = []; - list<Predicate> Predicates = []; // predicates turned into isel match code - ... remainder not shown for space ... -} -</pre> -</div> - -<p> -A <tt>SelectionDAG</tt> node (<tt>SDNode</tt>) should contain an object -representing a target-specific instruction that is defined -in <tt>XXXInstrInfo.td</tt>. The instruction objects should represent -instructions from the architecture manual of the target machine (such as the -SPARC Architecture Manual for the SPARC target). -</p> - -<p> -A single instruction from the architecture manual is often modeled as multiple -target instructions, depending upon its operands. For example, a manual might -describe an add instruction that takes a register or an immediate operand. An -LLVM target could model this with two instructions named <tt>ADDri</tt> and -<tt>ADDrr</tt>. -</p> - -<p> -You should define a class for each instruction category and define each opcode -as a subclass of the category with appropriate parameters such as the fixed -binary encoding of opcodes and extended opcodes. You should map the register -bits to the bits of the instruction in which they are encoded (for the -JIT). Also you should specify how the instruction should be printed when the -automatic assembly printer is used. -</p> - -<p> -As is described in the SPARC Architecture Manual, Version 8, there are three -major 32-bit formats for instructions. Format 1 is only for the <tt>CALL</tt> -instruction. Format 2 is for branch on condition codes and <tt>SETHI</tt> (set -high bits of a register) instructions. Format 3 is for other instructions. -</p> - -<p> -Each of these formats has corresponding classes in <tt>SparcInstrFormat.td</tt>. -<tt>InstSP</tt> is a base class for other instruction classes. Additional base -classes are specified for more precise formats: for example -in <tt>SparcInstrFormat.td</tt>, <tt>F2_1</tt> is for <tt>SETHI</tt>, -and <tt>F2_2</tt> is for branches. There are three other base -classes: <tt>F3_1</tt> for register/register operations, <tt>F3_2</tt> for -register/immediate operations, and <tt>F3_3</tt> for floating-point -operations. <tt>SparcInstrInfo.td</tt> also adds the base class Pseudo for -synthetic SPARC instructions. -</p> - -<p> -<tt>SparcInstrInfo.td</tt> largely consists of operand and instruction -definitions for the SPARC target. In <tt>SparcInstrInfo.td</tt>, the following -target description file entry, <tt>LDrr</tt>, defines the Load Integer -instruction for a Word (the <tt>LD</tt> SPARC opcode) from a memory address to a -register. The first parameter, the value 3 (<tt>11<sub>2</sub></tt>), is the -operation value for this category of operation. The second parameter -(<tt>000000<sub>2</sub></tt>) is the specific operation value -for <tt>LD</tt>/Load Word. The third parameter is the output destination, which -is a register operand and defined in the <tt>Register</tt> target description -file (<tt>IntRegs</tt>). -</p> - -<div class="doc_code"> -<pre>def LDrr : F3_1 <3, 0b000000, (outs IntRegs:$dst), (ins MEMrr:$addr), - "ld [$addr], $dst", - [(set IntRegs:$dst, (load ADDRrr:$addr))]>; -</pre> -</div> - -<p> -The fourth parameter is the input source, which uses the address -operand <tt>MEMrr</tt> that is defined earlier in <tt>SparcInstrInfo.td</tt>: -</p> - -<div class="doc_code"> -<pre>def MEMrr : Operand<i32> { - let PrintMethod = "printMemOperand"; - let MIOperandInfo = (ops IntRegs, IntRegs); -} -</pre> -</div> - -<p> -The fifth parameter is a string that is used by the assembly printer and can be -left as an empty string until the assembly printer interface is implemented. The -sixth and final parameter is the pattern used to match the instruction during -the SelectionDAG Select Phase described in -(<a href="http://www.llvm.org/docs/CodeGenerator.html">The LLVM -Target-Independent Code Generator</a>). This parameter is detailed in the next -section, <a href="#InstructionSelector">Instruction Selector</a>. -</p> - -<p> -Instruction class definitions are not overloaded for different operand types, so -separate versions of instructions are needed for register, memory, or immediate -value operands. For example, to perform a Load Integer instruction for a Word -from an immediate operand to a register, the following instruction class is -defined: -</p> - -<div class="doc_code"> -<pre>def LDri : F3_2 <3, 0b000000, (outs IntRegs:$dst), (ins MEMri:$addr), - "ld [$addr], $dst", - [(set IntRegs:$dst, (load ADDRri:$addr))]>; -</pre> -</div> - -<p> -Writing these definitions for so many similar instructions can involve a lot of -cut and paste. In td files, the <tt>multiclass</tt> directive enables the -creation of templates to define several instruction classes at once (using -the <tt>defm</tt> directive). For example in <tt>SparcInstrInfo.td</tt>, the -<tt>multiclass</tt> pattern <tt>F3_12</tt> is defined to create 2 instruction -classes each time <tt>F3_12</tt> is invoked: -</p> - -<div class="doc_code"> -<pre>multiclass F3_12 <string OpcStr, bits<6> Op3Val, SDNode OpNode> { - def rr : F3_1 <2, Op3Val, - (outs IntRegs:$dst), (ins IntRegs:$b, IntRegs:$c), - !strconcat(OpcStr, " $b, $c, $dst"), - [(set IntRegs:$dst, (OpNode IntRegs:$b, IntRegs:$c))]>; - def ri : F3_2 <2, Op3Val, - (outs IntRegs:$dst), (ins IntRegs:$b, i32imm:$c), - !strconcat(OpcStr, " $b, $c, $dst"), - [(set IntRegs:$dst, (OpNode IntRegs:$b, simm13:$c))]>; -} -</pre> -</div> - -<p> -So when the <tt>defm</tt> directive is used for the <tt>XOR</tt> -and <tt>ADD</tt> instructions, as seen below, it creates four instruction -objects: <tt>XORrr</tt>, <tt>XORri</tt>, <tt>ADDrr</tt>, and <tt>ADDri</tt>. -</p> - -<div class="doc_code"> -<pre> -defm XOR : F3_12<"xor", 0b000011, xor>; -defm ADD : F3_12<"add", 0b000000, add>; -</pre> -</div> - -<p> -<tt>SparcInstrInfo.td</tt> also includes definitions for condition codes that -are referenced by branch instructions. The following definitions -in <tt>SparcInstrInfo.td</tt> indicate the bit location of the SPARC condition -code. For example, the 10<sup>th</sup> bit represents the 'greater than' -condition for integers, and the 22<sup>nd</sup> bit represents the 'greater -than' condition for floats. -</p> - -<div class="doc_code"> -<pre> -def ICC_NE : ICC_VAL< 9>; // Not Equal -def ICC_E : ICC_VAL< 1>; // Equal -def ICC_G : ICC_VAL<10>; // Greater -... -def FCC_U : FCC_VAL<23>; // Unordered -def FCC_G : FCC_VAL<22>; // Greater -def FCC_UG : FCC_VAL<21>; // Unordered or Greater -... -</pre> -</div> - -<p> -(Note that <tt>Sparc.h</tt> also defines enums that correspond to the same SPARC -condition codes. Care must be taken to ensure the values in <tt>Sparc.h</tt> -correspond to the values in <tt>SparcInstrInfo.td</tt>. I.e., -<tt>SPCC::ICC_NE = 9</tt>, <tt>SPCC::FCC_U = 23</tt> and so on.) -</p> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="operandMapping">Instruction Operand Mapping</a> -</div> - -<div class="doc_text"> - -<p> -The code generator backend maps instruction operands to fields in the -instruction. Operands are assigned to unbound fields in the instruction in the -order they are defined. Fields are bound when they are assigned a value. For -example, the Sparc target defines the <tt>XNORrr</tt> instruction as -a <tt>F3_1</tt> format instruction having three operands. -</p> - -<div class="doc_code"> -<pre> -def XNORrr : F3_1<2, 0b000111, - (outs IntRegs:$dst), (ins IntRegs:$b, IntRegs:$c), - "xnor $b, $c, $dst", - [(set IntRegs:$dst, (not (xor IntRegs:$b, IntRegs:$c)))]>; -</pre> -</div> - -<p> -The instruction templates in <tt>SparcInstrFormats.td</tt> show the base class -for <tt>F3_1</tt> is <tt>InstSP</tt>. -</p> - -<div class="doc_code"> -<pre> -class InstSP<dag outs, dag ins, string asmstr, list<dag> pattern> : Instruction { - field bits<32> Inst; - let Namespace = "SP"; - bits<2> op; - let Inst{31-30} = op; - dag OutOperandList = outs; - dag InOperandList = ins; - let AsmString = asmstr; - let Pattern = pattern; -} -</pre> -</div> - -<p><tt>InstSP</tt> leaves the <tt>op</tt> field unbound.</p> - -<div class="doc_code"> -<pre> -class F3<dag outs, dag ins, string asmstr, list<dag> pattern> - : InstSP<outs, ins, asmstr, pattern> { - bits<5> rd; - bits<6> op3; - bits<5> rs1; - let op{1} = 1; // Op = 2 or 3 - let Inst{29-25} = rd; - let Inst{24-19} = op3; - let Inst{18-14} = rs1; -} -</pre> -</div> - -<p> -<tt>F3</tt> binds the <tt>op</tt> field and defines the <tt>rd</tt>, -<tt>op3</tt>, and <tt>rs1</tt> fields. <tt>F3</tt> format instructions will -bind the operands <tt>rd</tt>, <tt>op3</tt>, and <tt>rs1</tt> fields. -</p> - -<div class="doc_code"> -<pre> -class F3_1<bits<2> opVal, bits<6> op3val, dag outs, dag ins, - string asmstr, list<dag> pattern> : F3<outs, ins, asmstr, pattern> { - bits<8> asi = 0; // asi not currently used - bits<5> rs2; - let op = opVal; - let op3 = op3val; - let Inst{13} = 0; // i field = 0 - let Inst{12-5} = asi; // address space identifier - let Inst{4-0} = rs2; -} -</pre> -</div> - -<p> -<tt>F3_1</tt> binds the <tt>op3</tt> field and defines the <tt>rs2</tt> -fields. <tt>F3_1</tt> format instructions will bind the operands to the <tt>rd</tt>, -<tt>rs1</tt>, and <tt>rs2</tt> fields. This results in the <tt>XNORrr</tt> -instruction binding <tt>$dst</tt>, <tt>$b</tt>, and <tt>$c</tt> operands to -the <tt>rd</tt>, <tt>rs1</tt>, and <tt>rs2</tt> fields respectively. -</p> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="implementInstr">Implement a subclass of </a> - <a href="http://www.llvm.org/docs/CodeGenerator.html#targetinstrinfo">TargetInstrInfo</a> -</div> - -<div class="doc_text"> - -<p> -The final step is to hand code portions of <tt>XXXInstrInfo</tt>, which -implements the interface described in <tt>TargetInstrInfo.h</tt>. These -functions return <tt>0</tt> or a Boolean or they assert, unless -overridden. Here's a list of functions that are overridden for the SPARC -implementation in <tt>SparcInstrInfo.cpp</tt>: -</p> - -<ul> -<li><tt>isMoveInstr</tt> — Return true if the instruction is a register to - register move; false, otherwise.</li> - -<li><tt>isLoadFromStackSlot</tt> — If the specified machine instruction is - a direct load from a stack slot, return the register number of the - destination and the <tt>FrameIndex</tt> of the stack slot.</li> - -<li><tt>isStoreToStackSlot</tt> — If the specified machine instruction is - a direct store to a stack slot, return the register number of the - destination and the <tt>FrameIndex</tt> of the stack slot.</li> - -<li><tt>copyRegToReg</tt> — Copy values between a pair of registers.</li> - -<li><tt>storeRegToStackSlot</tt> — Store a register value to a stack - slot.</li> - -<li><tt>loadRegFromStackSlot</tt> — Load a register value from a stack - slot.</li> - -<li><tt>storeRegToAddr</tt> — Store a register value to memory.</li> - -<li><tt>loadRegFromAddr</tt> — Load a register value from memory.</li> - -<li><tt>foldMemoryOperand</tt> — Attempt to combine instructions of any - load or store instruction for the specified operand(s).</li> -</ul> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="branchFolding">Branch Folding and If Conversion</a> -</div> -<div class="doc_text"> - -<p> -Performance can be improved by combining instructions or by eliminating -instructions that are never reached. The <tt>AnalyzeBranch</tt> method -in <tt>XXXInstrInfo</tt> may be implemented to examine conditional instructions -and remove unnecessary instructions. <tt>AnalyzeBranch</tt> looks at the end of -a machine basic block (MBB) for opportunities for improvement, such as branch -folding and if conversion. The <tt>BranchFolder</tt> and <tt>IfConverter</tt> -machine function passes (see the source files <tt>BranchFolding.cpp</tt> and -<tt>IfConversion.cpp</tt> in the <tt>lib/CodeGen</tt> directory) call -<tt>AnalyzeBranch</tt> to improve the control flow graph that represents the -instructions. -</p> - -<p> -Several implementations of <tt>AnalyzeBranch</tt> (for ARM, Alpha, and X86) can -be examined as models for your own <tt>AnalyzeBranch</tt> implementation. Since -SPARC does not implement a useful <tt>AnalyzeBranch</tt>, the ARM target -implementation is shown below. -</p> - -<p><tt>AnalyzeBranch</tt> returns a Boolean value and takes four parameters:</p> - -<ul> -<li><tt>MachineBasicBlock &MBB</tt> — The incoming block to be - examined.</li> - -<li><tt>MachineBasicBlock *&TBB</tt> — A destination block that is - returned. For a conditional branch that evaluates to true, <tt>TBB</tt> is - the destination.</li> - -<li><tt>MachineBasicBlock *&FBB</tt> — For a conditional branch that - evaluates to false, <tt>FBB</tt> is returned as the destination.</li> - -<li><tt>std::vector<MachineOperand> &Cond</tt> — List of - operands to evaluate a condition for a conditional branch.</li> -</ul> - -<p> -In the simplest case, if a block ends without a branch, then it falls through to -the successor block. No destination blocks are specified for either <tt>TBB</tt> -or <tt>FBB</tt>, so both parameters return <tt>NULL</tt>. The start of -the <tt>AnalyzeBranch</tt> (see code below for the ARM target) shows the -function parameters and the code for the simplest case. -</p> - -<div class="doc_code"> -<pre>bool ARMInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB, - MachineBasicBlock *&TBB, MachineBasicBlock *&FBB, - std::vector<MachineOperand> &Cond) const -{ - MachineBasicBlock::iterator I = MBB.end(); - if (I == MBB.begin() || !isUnpredicatedTerminator(--I)) - return false; -</pre> -</div> - -<p> -If a block ends with a single unconditional branch instruction, then -<tt>AnalyzeBranch</tt> (shown below) should return the destination of that -branch in the <tt>TBB</tt> parameter. -</p> - -<div class="doc_code"> -<pre> - if (LastOpc == ARM::B || LastOpc == ARM::tB) { - TBB = LastInst->getOperand(0).getMBB(); - return false; - } -</pre> -</div> - -<p> -If a block ends with two unconditional branches, then the second branch is never -reached. In that situation, as shown below, remove the last branch instruction -and return the penultimate branch in the <tt>TBB</tt> parameter. -</p> - -<div class="doc_code"> -<pre> - if ((SecondLastOpc == ARM::B || SecondLastOpc==ARM::tB) && - (LastOpc == ARM::B || LastOpc == ARM::tB)) { - TBB = SecondLastInst->getOperand(0).getMBB(); - I = LastInst; - I->eraseFromParent(); - return false; - } -</pre> -</div> - -<p> -A block may end with a single conditional branch instruction that falls through -to successor block if the condition evaluates to false. In that case, -<tt>AnalyzeBranch</tt> (shown below) should return the destination of that -conditional branch in the <tt>TBB</tt> parameter and a list of operands in -the <tt>Cond</tt> parameter to evaluate the condition. -</p> - -<div class="doc_code"> -<pre> - if (LastOpc == ARM::Bcc || LastOpc == ARM::tBcc) { - // Block ends with fall-through condbranch. - TBB = LastInst->getOperand(0).getMBB(); - Cond.push_back(LastInst->getOperand(1)); - Cond.push_back(LastInst->getOperand(2)); - return false; - } -</pre> -</div> - -<p> -If a block ends with both a conditional branch and an ensuing unconditional -branch, then <tt>AnalyzeBranch</tt> (shown below) should return the conditional -branch destination (assuming it corresponds to a conditional evaluation of -'<tt>true</tt>') in the <tt>TBB</tt> parameter and the unconditional branch -destination in the <tt>FBB</tt> (corresponding to a conditional evaluation of -'<tt>false</tt>'). A list of operands to evaluate the condition should be -returned in the <tt>Cond</tt> parameter. -</p> - -<div class="doc_code"> -<pre> - unsigned SecondLastOpc = SecondLastInst->getOpcode(); - - if ((SecondLastOpc == ARM::Bcc && LastOpc == ARM::B) || - (SecondLastOpc == ARM::tBcc && LastOpc == ARM::tB)) { - TBB = SecondLastInst->getOperand(0).getMBB(); - Cond.push_back(SecondLastInst->getOperand(1)); - Cond.push_back(SecondLastInst->getOperand(2)); - FBB = LastInst->getOperand(0).getMBB(); - return false; - } -</pre> -</div> - -<p> -For the last two cases (ending with a single conditional branch or ending with -one conditional and one unconditional branch), the operands returned in -the <tt>Cond</tt> parameter can be passed to methods of other instructions to -create new branches or perform other operations. An implementation -of <tt>AnalyzeBranch</tt> requires the helper methods <tt>RemoveBranch</tt> -and <tt>InsertBranch</tt> to manage subsequent operations. -</p> - -<p> -<tt>AnalyzeBranch</tt> should return false indicating success in most circumstances. -<tt>AnalyzeBranch</tt> should only return true when the method is stumped about what to -do, for example, if a block has three terminating branches. <tt>AnalyzeBranch</tt> may -return true if it encounters a terminator it cannot handle, such as an indirect -branch. -</p> - -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="InstructionSelector">Instruction Selector</a> -</div> -<!-- *********************************************************************** --> - -<div class="doc_text"> - -<p> -LLVM uses a <tt>SelectionDAG</tt> to represent LLVM IR instructions, and nodes -of the <tt>SelectionDAG</tt> ideally represent native target -instructions. During code generation, instruction selection passes are performed -to convert non-native DAG instructions into native target-specific -instructions. The pass described in <tt>XXXISelDAGToDAG.cpp</tt> is used to -match patterns and perform DAG-to-DAG instruction selection. Optionally, a pass -may be defined (in <tt>XXXBranchSelector.cpp</tt>) to perform similar DAG-to-DAG -operations for branch instructions. Later, the code in -<tt>XXXISelLowering.cpp</tt> replaces or removes operations and data types not -supported natively (legalizes) in a <tt>SelectionDAG</tt>. -</p> - -<p> -TableGen generates code for instruction selection using the following target -description input files: -</p> - -<ul> -<li><tt>XXXInstrInfo.td</tt> — Contains definitions of instructions in a - target-specific instruction set, generates <tt>XXXGenDAGISel.inc</tt>, which - is included in <tt>XXXISelDAGToDAG.cpp</tt>.</li> - -<li><tt>XXXCallingConv.td</tt> — Contains the calling and return value - conventions for the target architecture, and it generates - <tt>XXXGenCallingConv.inc</tt>, which is included in - <tt>XXXISelLowering.cpp</tt>.</li> -</ul> - -<p> -The implementation of an instruction selection pass must include a header that -declares the <tt>FunctionPass</tt> class or a subclass of <tt>FunctionPass</tt>. In -<tt>XXXTargetMachine.cpp</tt>, a Pass Manager (PM) should add each instruction -selection pass into the queue of passes to run. -</p> - -<p> -The LLVM static compiler (<tt>llc</tt>) is an excellent tool for visualizing the -contents of DAGs. To display the <tt>SelectionDAG</tt> before or after specific -processing phases, use the command line options for <tt>llc</tt>, described -at <a href="http://llvm.org/docs/CodeGenerator.html#selectiondag_process"> -SelectionDAG Instruction Selection Process</a>. -</p> - -<p> -To describe instruction selector behavior, you should add patterns for lowering -LLVM code into a <tt>SelectionDAG</tt> as the last parameter of the instruction -definitions in <tt>XXXInstrInfo.td</tt>. For example, in -<tt>SparcInstrInfo.td</tt>, this entry defines a register store operation, and -the last parameter describes a pattern with the store DAG operator. -</p> - -<div class="doc_code"> -<pre> -def STrr : F3_1< 3, 0b000100, (outs), (ins MEMrr:$addr, IntRegs:$src), - "st $src, [$addr]", [(store IntRegs:$src, ADDRrr:$addr)]>; -</pre> -</div> - -<p> -<tt>ADDRrr</tt> is a memory mode that is also defined in -<tt>SparcInstrInfo.td</tt>: -</p> - -<div class="doc_code"> -<pre> -def ADDRrr : ComplexPattern<i32, 2, "SelectADDRrr", [], []>; -</pre> -</div> - -<p> -The definition of <tt>ADDRrr</tt> refers to <tt>SelectADDRrr</tt>, which is a -function defined in an implementation of the Instructor Selector (such -as <tt>SparcISelDAGToDAG.cpp</tt>). -</p> - -<p> -In <tt>lib/Target/TargetSelectionDAG.td</tt>, the DAG operator for store is -defined below: -</p> - -<div class="doc_code"> -<pre> -def store : PatFrag<(ops node:$val, node:$ptr), - (st node:$val, node:$ptr), [{ - if (StoreSDNode *ST = dyn_cast<StoreSDNode>(N)) - return !ST->isTruncatingStore() && - ST->getAddressingMode() == ISD::UNINDEXED; - return false; -}]>; -</pre> -</div> - -<p> -<tt>XXXInstrInfo.td</tt> also generates (in <tt>XXXGenDAGISel.inc</tt>) the -<tt>SelectCode</tt> method that is used to call the appropriate processing -method for an instruction. In this example, <tt>SelectCode</tt> -calls <tt>Select_ISD_STORE</tt> for the <tt>ISD::STORE</tt> opcode. -</p> - -<div class="doc_code"> -<pre> -SDNode *SelectCode(SDValue N) { - ... - MVT::ValueType NVT = N.getNode()->getValueType(0); - switch (N.getOpcode()) { - case ISD::STORE: { - switch (NVT) { - default: - return Select_ISD_STORE(N); - break; - } - break; - } - ... -</pre> -</div> - -<p> -The pattern for <tt>STrr</tt> is matched, so elsewhere in -<tt>XXXGenDAGISel.inc</tt>, code for <tt>STrr</tt> is created for -<tt>Select_ISD_STORE</tt>. The <tt>Emit_22</tt> method is also generated -in <tt>XXXGenDAGISel.inc</tt> to complete the processing of this -instruction. -</p> - -<div class="doc_code"> -<pre> -SDNode *Select_ISD_STORE(const SDValue &N) { - SDValue Chain = N.getOperand(0); - if (Predicate_store(N.getNode())) { - SDValue N1 = N.getOperand(1); - SDValue N2 = N.getOperand(2); - SDValue CPTmp0; - SDValue CPTmp1; - - // Pattern: (st:void IntRegs:i32:$src, - // ADDRrr:i32:$addr)<<P:Predicate_store>> - // Emits: (STrr:void ADDRrr:i32:$addr, IntRegs:i32:$src) - // Pattern complexity = 13 cost = 1 size = 0 - if (SelectADDRrr(N, N2, CPTmp0, CPTmp1) && - N1.getNode()->getValueType(0) == MVT::i32 && - N2.getNode()->getValueType(0) == MVT::i32) { - return Emit_22(N, SP::STrr, CPTmp0, CPTmp1); - } -... -</pre> -</div> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="LegalizePhase">The SelectionDAG Legalize Phase</a> -</div> - -<div class="doc_text"> - -<p> -The Legalize phase converts a DAG to use types and operations that are natively -supported by the target. For natively unsupported types and operations, you need -to add code to the target-specific XXXTargetLowering implementation to convert -unsupported types and operations to supported ones. -</p> - -<p> -In the constructor for the <tt>XXXTargetLowering</tt> class, first use the -<tt>addRegisterClass</tt> method to specify which types are supports and which -register classes are associated with them. The code for the register classes are -generated by TableGen from <tt>XXXRegisterInfo.td</tt> and placed -in <tt>XXXGenRegisterInfo.h.inc</tt>. For example, the implementation of the -constructor for the SparcTargetLowering class (in -<tt>SparcISelLowering.cpp</tt>) starts with the following code: -</p> - -<div class="doc_code"> -<pre> -addRegisterClass(MVT::i32, SP::IntRegsRegisterClass); -addRegisterClass(MVT::f32, SP::FPRegsRegisterClass); -addRegisterClass(MVT::f64, SP::DFPRegsRegisterClass); -</pre> -</div> - -<p> -You should examine the node types in the <tt>ISD</tt> namespace -(<tt>include/llvm/CodeGen/SelectionDAGNodes.h</tt>) and determine which -operations the target natively supports. For operations that do <b>not</b> have -native support, add a callback to the constructor for the XXXTargetLowering -class, so the instruction selection process knows what to do. The TargetLowering -class callback methods (declared in <tt>llvm/Target/TargetLowering.h</tt>) are: -</p> - -<ul> -<li><tt>setOperationAction</tt> — General operation.</li> - -<li><tt>setLoadExtAction</tt> — Load with extension.</li> - -<li><tt>setTruncStoreAction</tt> — Truncating store.</li> - -<li><tt>setIndexedLoadAction</tt> — Indexed load.</li> - -<li><tt>setIndexedStoreAction</tt> — Indexed store.</li> - -<li><tt>setConvertAction</tt> — Type conversion.</li> - -<li><tt>setCondCodeAction</tt> — Support for a given condition code.</li> -</ul> - -<p> -Note: on older releases, <tt>setLoadXAction</tt> is used instead -of <tt>setLoadExtAction</tt>. Also, on older releases, -<tt>setCondCodeAction</tt> may not be supported. Examine your release -to see what methods are specifically supported. -</p> - -<p> -These callbacks are used to determine that an operation does or does not work -with a specified type (or types). And in all cases, the third parameter is -a <tt>LegalAction</tt> type enum value: <tt>Promote</tt>, <tt>Expand</tt>, -<tt>Custom</tt>, or <tt>Legal</tt>. <tt>SparcISelLowering.cpp</tt> -contains examples of all four <tt>LegalAction</tt> values. -</p> - -</div> - -<!-- _______________________________________________________________________ --> -<div class="doc_subsubsection"> - <a name="promote">Promote</a> -</div> - -<div class="doc_text"> - -<p> -For an operation without native support for a given type, the specified type may -be promoted to a larger type that is supported. For example, SPARC does not -support a sign-extending load for Boolean values (<tt>i1</tt> type), so -in <tt>SparcISelLowering.cpp</tt> the third parameter below, <tt>Promote</tt>, -changes <tt>i1</tt> type values to a large type before loading. -</p> - -<div class="doc_code"> -<pre> -setLoadExtAction(ISD::SEXTLOAD, MVT::i1, Promote); -</pre> -</div> - -</div> - -<!-- _______________________________________________________________________ --> -<div class="doc_subsubsection"> - <a name="expand">Expand</a> -</div> - -<div class="doc_text"> - -<p> -For a type without native support, a value may need to be broken down further, -rather than promoted. For an operation without native support, a combination of -other operations may be used to similar effect. In SPARC, the floating-point -sine and cosine trig operations are supported by expansion to other operations, -as indicated by the third parameter, <tt>Expand</tt>, to -<tt>setOperationAction</tt>: -</p> - -<div class="doc_code"> -<pre> -setOperationAction(ISD::FSIN, MVT::f32, Expand); -setOperationAction(ISD::FCOS, MVT::f32, Expand); -</pre> -</div> - -</div> - -<!-- _______________________________________________________________________ --> -<div class="doc_subsubsection"> - <a name="custom">Custom</a> -</div> - -<div class="doc_text"> - -<p> -For some operations, simple type promotion or operation expansion may be -insufficient. In some cases, a special intrinsic function must be implemented. -</p> - -<p> -For example, a constant value may require special treatment, or an operation may -require spilling and restoring registers in the stack and working with register -allocators. -</p> - -<p> -As seen in <tt>SparcISelLowering.cpp</tt> code below, to perform a type -conversion from a floating point value to a signed integer, first the -<tt>setOperationAction</tt> should be called with <tt>Custom</tt> as the third -parameter: -</p> - -<div class="doc_code"> -<pre> -setOperationAction(ISD::FP_TO_SINT, MVT::i32, Custom); -</pre> -</div> - -<p> -In the <tt>LowerOperation</tt> method, for each <tt>Custom</tt> operation, a -case statement should be added to indicate what function to call. In the -following code, an <tt>FP_TO_SINT</tt> opcode will call -the <tt>LowerFP_TO_SINT</tt> method: -</p> - -<div class="doc_code"> -<pre> -SDValue SparcTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) { - switch (Op.getOpcode()) { - case ISD::FP_TO_SINT: return LowerFP_TO_SINT(Op, DAG); - ... - } -} -</pre> -</div> - -<p> -Finally, the <tt>LowerFP_TO_SINT</tt> method is implemented, using an FP -register to convert the floating-point value to an integer. -</p> - -<div class="doc_code"> -<pre> -static SDValue LowerFP_TO_SINT(SDValue Op, SelectionDAG &DAG) { - assert(Op.getValueType() == MVT::i32); - Op = DAG.getNode(SPISD::FTOI, MVT::f32, Op.getOperand(0)); - return DAG.getNode(ISD::BIT_CONVERT, MVT::i32, Op); -} -</pre> -</div> - -</div> - -<!-- _______________________________________________________________________ --> -<div class="doc_subsubsection"> - <a name="legal">Legal</a> -</div> - -<div class="doc_text"> - -<p> -The <tt>Legal</tt> LegalizeAction enum value simply indicates that an -operation <b>is</b> natively supported. <tt>Legal</tt> represents the default -condition, so it is rarely used. In <tt>SparcISelLowering.cpp</tt>, the action -for <tt>CTPOP</tt> (an operation to count the bits set in an integer) is -natively supported only for SPARC v9. The following code enables -the <tt>Expand</tt> conversion technique for non-v9 SPARC implementations. -</p> - -<div class="doc_code"> -<pre> -setOperationAction(ISD::CTPOP, MVT::i32, Expand); -... -if (TM.getSubtarget<SparcSubtarget>().isV9()) - setOperationAction(ISD::CTPOP, MVT::i32, Legal); - case ISD::SETULT: return SPCC::ICC_CS; - case ISD::SETULE: return SPCC::ICC_LEU; - case ISD::SETUGT: return SPCC::ICC_GU; - case ISD::SETUGE: return SPCC::ICC_CC; - } -} -</pre> -</div> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="callingConventions">Calling Conventions</a> -</div> - -<div class="doc_text"> - -<p> -To support target-specific calling conventions, <tt>XXXGenCallingConv.td</tt> -uses interfaces (such as CCIfType and CCAssignToReg) that are defined in -<tt>lib/Target/TargetCallingConv.td</tt>. TableGen can take the target -descriptor file <tt>XXXGenCallingConv.td</tt> and generate the header -file <tt>XXXGenCallingConv.inc</tt>, which is typically included -in <tt>XXXISelLowering.cpp</tt>. You can use the interfaces in -<tt>TargetCallingConv.td</tt> to specify: -</p> - -<ul> -<li>The order of parameter allocation.</li> - -<li>Where parameters and return values are placed (that is, on the stack or in - registers).</li> - -<li>Which registers may be used.</li> - -<li>Whether the caller or callee unwinds the stack.</li> -</ul> - -<p> -The following example demonstrates the use of the <tt>CCIfType</tt> and -<tt>CCAssignToReg</tt> interfaces. If the <tt>CCIfType</tt> predicate is true -(that is, if the current argument is of type <tt>f32</tt> or <tt>f64</tt>), then -the action is performed. In this case, the <tt>CCAssignToReg</tt> action assigns -the argument value to the first available register: either <tt>R0</tt> -or <tt>R1</tt>. -</p> - -<div class="doc_code"> -<pre> -CCIfType<[f32,f64], CCAssignToReg<[R0, R1]>> -</pre> -</div> - -<p> -<tt>SparcCallingConv.td</tt> contains definitions for a target-specific -return-value calling convention (RetCC_Sparc32) and a basic 32-bit C calling -convention (<tt>CC_Sparc32</tt>). The definition of <tt>RetCC_Sparc32</tt> -(shown below) indicates which registers are used for specified scalar return -types. A single-precision float is returned to register <tt>F0</tt>, and a -double-precision float goes to register <tt>D0</tt>. A 32-bit integer is -returned in register <tt>I0</tt> or <tt>I1</tt>. -</p> - -<div class="doc_code"> -<pre> -def RetCC_Sparc32 : CallingConv<[ - CCIfType<[i32], CCAssignToReg<[I0, I1]>>, - CCIfType<[f32], CCAssignToReg<[F0]>>, - CCIfType<[f64], CCAssignToReg<[D0]>> -]>; -</pre> -</div> - -<p> -The definition of <tt>CC_Sparc32</tt> in <tt>SparcCallingConv.td</tt> introduces -<tt>CCAssignToStack</tt>, which assigns the value to a stack slot with the -specified size and alignment. In the example below, the first parameter, 4, -indicates the size of the slot, and the second parameter, also 4, indicates the -stack alignment along 4-byte units. (Special cases: if size is zero, then the -ABI size is used; if alignment is zero, then the ABI alignment is used.) -</p> - -<div class="doc_code"> -<pre> -def CC_Sparc32 : CallingConv<[ - // All arguments get passed in integer registers if there is space. - CCIfType<[i32, f32, f64], CCAssignToReg<[I0, I1, I2, I3, I4, I5]>>, - CCAssignToStack<4, 4> -]>; -</pre> -</div> - -<p> -<tt>CCDelegateTo</tt> is another commonly used interface, which tries to find a -specified sub-calling convention, and, if a match is found, it is invoked. In -the following example (in <tt>X86CallingConv.td</tt>), the definition of -<tt>RetCC_X86_32_C</tt> ends with <tt>CCDelegateTo</tt>. After the current value -is assigned to the register <tt>ST0</tt> or <tt>ST1</tt>, -the <tt>RetCC_X86Common</tt> is invoked. -</p> - -<div class="doc_code"> -<pre> -def RetCC_X86_32_C : CallingConv<[ - CCIfType<[f32], CCAssignToReg<[ST0, ST1]>>, - CCIfType<[f64], CCAssignToReg<[ST0, ST1]>>, - CCDelegateTo<RetCC_X86Common> -]>; -</pre> -</div> - -<p> -<tt>CCIfCC</tt> is an interface that attempts to match the given name to the -current calling convention. If the name identifies the current calling -convention, then a specified action is invoked. In the following example (in -<tt>X86CallingConv.td</tt>), if the <tt>Fast</tt> calling convention is in use, -then <tt>RetCC_X86_32_Fast</tt> is invoked. If the <tt>SSECall</tt> calling -convention is in use, then <tt>RetCC_X86_32_SSE</tt> is invoked. -</p> - -<div class="doc_code"> -<pre> -def RetCC_X86_32 : CallingConv<[ - CCIfCC<"CallingConv::Fast", CCDelegateTo<RetCC_X86_32_Fast>>, - CCIfCC<"CallingConv::X86_SSECall", CCDelegateTo<RetCC_X86_32_SSE>>, - CCDelegateTo<RetCC_X86_32_C> -]>; -</pre> -</div> - -<p>Other calling convention interfaces include:</p> - -<ul> -<li><tt>CCIf <predicate, action></tt> — If the predicate matches, - apply the action.</li> - -<li><tt>CCIfInReg <action></tt> — If the argument is marked with the - '<tt>inreg</tt>' attribute, then apply the action.</li> - -<li><tt>CCIfNest <action></tt> — Inf the argument is marked with the - '<tt>nest</tt>' attribute, then apply the action.</li> - -<li><tt>CCIfNotVarArg <action></tt> — If the current function does - not take a variable number of arguments, apply the action.</li> - -<li><tt>CCAssignToRegWithShadow <registerList, shadowList></tt> — - similar to <tt>CCAssignToReg</tt>, but with a shadow list of registers.</li> - -<li><tt>CCPassByVal <size, align></tt> — Assign value to a stack - slot with the minimum specified size and alignment.</li> - -<li><tt>CCPromoteToType <type></tt> — Promote the current value to - the specified type.</li> - -<li><tt>CallingConv <[actions]></tt> — Define each calling - convention that is supported.</li> -</ul> - -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="assemblyPrinter">Assembly Printer</a> -</div> -<!-- *********************************************************************** --> - -<div class="doc_text"> - -<p> -During the code emission stage, the code generator may utilize an LLVM pass to -produce assembly output. To do this, you want to implement the code for a -printer that converts LLVM IR to a GAS-format assembly language for your target -machine, using the following steps: -</p> - -<ul> -<li>Define all the assembly strings for your target, adding them to the - instructions defined in the <tt>XXXInstrInfo.td</tt> file. - (See <a href="#InstructionSet">Instruction Set</a>.) TableGen will produce - an output file (<tt>XXXGenAsmWriter.inc</tt>) with an implementation of - the <tt>printInstruction</tt> method for the XXXAsmPrinter class.</li> - -<li>Write <tt>XXXTargetAsmInfo.h</tt>, which contains the bare-bones declaration - of the <tt>XXXTargetAsmInfo</tt> class (a subclass - of <tt>TargetAsmInfo</tt>).</li> - -<li>Write <tt>XXXTargetAsmInfo.cpp</tt>, which contains target-specific values - for <tt>TargetAsmInfo</tt> properties and sometimes new implementations for - methods.</li> - -<li>Write <tt>XXXAsmPrinter.cpp</tt>, which implements the <tt>AsmPrinter</tt> - class that performs the LLVM-to-assembly conversion.</li> -</ul> - -<p> -The code in <tt>XXXTargetAsmInfo.h</tt> is usually a trivial declaration of the -<tt>XXXTargetAsmInfo</tt> class for use in <tt>XXXTargetAsmInfo.cpp</tt>. -Similarly, <tt>XXXTargetAsmInfo.cpp</tt> usually has a few declarations of -<tt>XXXTargetAsmInfo</tt> replacement values that override the default values -in <tt>TargetAsmInfo.cpp</tt>. For example in <tt>SparcTargetAsmInfo.cpp</tt>: -</p> - -<div class="doc_code"> -<pre> -SparcTargetAsmInfo::SparcTargetAsmInfo(const SparcTargetMachine &TM) { - Data16bitsDirective = "\t.half\t"; - Data32bitsDirective = "\t.word\t"; - Data64bitsDirective = 0; // .xword is only supported by V9. - ZeroDirective = "\t.skip\t"; - CommentString = "!"; - ConstantPoolSection = "\t.section \".rodata\",#alloc\n"; -} -</pre> -</div> - -<p> -The X86 assembly printer implementation (<tt>X86TargetAsmInfo</tt>) is an -example where the target specific <tt>TargetAsmInfo</tt> class uses an -overridden methods: <tt>ExpandInlineAsm</tt>. -</p> - -<p> -A target-specific implementation of AsmPrinter is written in -<tt>XXXAsmPrinter.cpp</tt>, which implements the <tt>AsmPrinter</tt> class that -converts the LLVM to printable assembly. The implementation must include the -following headers that have declarations for the <tt>AsmPrinter</tt> and -<tt>MachineFunctionPass</tt> classes. The <tt>MachineFunctionPass</tt> is a -subclass of <tt>FunctionPass</tt>. -</p> - -<div class="doc_code"> -<pre> -#include "llvm/CodeGen/AsmPrinter.h" -#include "llvm/CodeGen/MachineFunctionPass.h" -</pre> -</div> - -<p> -As a <tt>FunctionPass</tt>, <tt>AsmPrinter</tt> first -calls <tt>doInitialization</tt> to set up the <tt>AsmPrinter</tt>. In -<tt>SparcAsmPrinter</tt>, a <tt>Mangler</tt> object is instantiated to process -variable names. -</p> - -<p> -In <tt>XXXAsmPrinter.cpp</tt>, the <tt>runOnMachineFunction</tt> method -(declared in <tt>MachineFunctionPass</tt>) must be implemented -for <tt>XXXAsmPrinter</tt>. In <tt>MachineFunctionPass</tt>, -the <tt>runOnFunction</tt> method invokes <tt>runOnMachineFunction</tt>. -Target-specific implementations of <tt>runOnMachineFunction</tt> differ, but -generally do the following to process each machine function: -</p> - -<ul> -<li>Call <tt>SetupMachineFunction</tt> to perform initialization.</li> - -<li>Call <tt>EmitConstantPool</tt> to print out (to the output stream) constants - which have been spilled to memory.</li> - -<li>Call <tt>EmitJumpTableInfo</tt> to print out jump tables used by the current - function.</li> - -<li>Print out the label for the current function.</li> - -<li>Print out the code for the function, including basic block labels and the - assembly for the instruction (using <tt>printInstruction</tt>)</li> -</ul> - -<p> -The <tt>XXXAsmPrinter</tt> implementation must also include the code generated -by TableGen that is output in the <tt>XXXGenAsmWriter.inc</tt> file. The code -in <tt>XXXGenAsmWriter.inc</tt> contains an implementation of the -<tt>printInstruction</tt> method that may call these methods: -</p> - -<ul> -<li><tt>printOperand</tt></li> - -<li><tt>printMemOperand</tt></li> - -<li><tt>printCCOperand (for conditional statements)</tt></li> - -<li><tt>printDataDirective</tt></li> - -<li><tt>printDeclare</tt></li> - -<li><tt>printImplicitDef</tt></li> - -<li><tt>printInlineAsm</tt></li> -</ul> - -<p> -The implementations of <tt>printDeclare</tt>, <tt>printImplicitDef</tt>, -<tt>printInlineAsm</tt>, and <tt>printLabel</tt> in <tt>AsmPrinter.cpp</tt> are -generally adequate for printing assembly and do not need to be -overridden. -</p> - -<p> -The <tt>printOperand</tt> method is implemented with a long switch/case -statement for the type of operand: register, immediate, basic block, external -symbol, global address, constant pool index, or jump table index. For an -instruction with a memory address operand, the <tt>printMemOperand</tt> method -should be implemented to generate the proper output. Similarly, -<tt>printCCOperand</tt> should be used to print a conditional operand. -</p> - -<p><tt>doFinalization</tt> should be overridden in <tt>XXXAsmPrinter</tt>, and -it should be called to shut down the assembly printer. During -<tt>doFinalization</tt>, global variables and constants are printed to -output. -</p> - -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="subtargetSupport">Subtarget Support</a> -</div> -<!-- *********************************************************************** --> - -<div class="doc_text"> - -<p> -Subtarget support is used to inform the code generation process of instruction -set variations for a given chip set. For example, the LLVM SPARC implementation -provided covers three major versions of the SPARC microprocessor architecture: -Version 8 (V8, which is a 32-bit architecture), Version 9 (V9, a 64-bit -architecture), and the UltraSPARC architecture. V8 has 16 double-precision -floating-point registers that are also usable as either 32 single-precision or 8 -quad-precision registers. V8 is also purely big-endian. V9 has 32 -double-precision floating-point registers that are also usable as 16 -quad-precision registers, but cannot be used as single-precision registers. The -UltraSPARC architecture combines V9 with UltraSPARC Visual Instruction Set -extensions. -</p> - -<p> -If subtarget support is needed, you should implement a target-specific -XXXSubtarget class for your architecture. This class should process the -command-line options <tt>-mcpu=</tt> and <tt>-mattr=</tt>. -</p> - -<p> -TableGen uses definitions in the <tt>Target.td</tt> and <tt>Sparc.td</tt> files -to generate code in <tt>SparcGenSubtarget.inc</tt>. In <tt>Target.td</tt>, shown -below, the <tt>SubtargetFeature</tt> interface is defined. The first 4 string -parameters of the <tt>SubtargetFeature</tt> interface are a feature name, an -attribute set by the feature, the value of the attribute, and a description of -the feature. (The fifth parameter is a list of features whose presence is -implied, and its default value is an empty array.) -</p> - -<div class="doc_code"> -<pre> -class SubtargetFeature<string n, string a, string v, string d, - list<SubtargetFeature> i = []> { - string Name = n; - string Attribute = a; - string Value = v; - string Desc = d; - list<SubtargetFeature> Implies = i; -} -</pre> -</div> - -<p> -In the <tt>Sparc.td</tt> file, the SubtargetFeature is used to define the -following features. -</p> - -<div class="doc_code"> -<pre> -def FeatureV9 : SubtargetFeature<"v9", "IsV9", "true", - "Enable SPARC-V9 instructions">; -def FeatureV8Deprecated : SubtargetFeature<"deprecated-v8", - "V8DeprecatedInsts", "true", - "Enable deprecated V8 instructions in V9 mode">; -def FeatureVIS : SubtargetFeature<"vis", "IsVIS", "true", - "Enable UltraSPARC Visual Instruction Set extensions">; -</pre> -</div> - -<p> -Elsewhere in <tt>Sparc.td</tt>, the Proc class is defined and then is used to -define particular SPARC processor subtypes that may have the previously -described features. -</p> - -<div class="doc_code"> -<pre> -class Proc<string Name, list<SubtargetFeature> Features> - : Processor<Name, NoItineraries, Features>; - -def : Proc<"generic", []>; -def : Proc<"v8", []>; -def : Proc<"supersparc", []>; -def : Proc<"sparclite", []>; -def : Proc<"f934", []>; -def : Proc<"hypersparc", []>; -def : Proc<"sparclite86x", []>; -def : Proc<"sparclet", []>; -def : Proc<"tsc701", []>; -def : Proc<"v9", [FeatureV9]>; -def : Proc<"ultrasparc", [FeatureV9, FeatureV8Deprecated]>; -def : Proc<"ultrasparc3", [FeatureV9, FeatureV8Deprecated]>; -def : Proc<"ultrasparc3-vis", [FeatureV9, FeatureV8Deprecated, FeatureVIS]>; -</pre> -</div> - -<p> -From <tt>Target.td</tt> and <tt>Sparc.td</tt> files, the resulting -SparcGenSubtarget.inc specifies enum values to identify the features, arrays of -constants to represent the CPU features and CPU subtypes, and the -ParseSubtargetFeatures method that parses the features string that sets -specified subtarget options. The generated <tt>SparcGenSubtarget.inc</tt> file -should be included in the <tt>SparcSubtarget.cpp</tt>. The target-specific -implementation of the XXXSubtarget method should follow this pseudocode: -</p> - -<div class="doc_code"> -<pre> -XXXSubtarget::XXXSubtarget(const Module &M, const std::string &FS) { - // Set the default features - // Determine default and user specified characteristics of the CPU - // Call ParseSubtargetFeatures(FS, CPU) to parse the features string - // Perform any additional operations -} -</pre> -</div> - -</div> - -<!-- *********************************************************************** --> -<div class="doc_section"> - <a name="jitSupport">JIT Support</a> -</div> -<!-- *********************************************************************** --> - -<div class="doc_text"> - -<p> -The implementation of a target machine optionally includes a Just-In-Time (JIT) -code generator that emits machine code and auxiliary structures as binary output -that can be written directly to memory. To do this, implement JIT code -generation by performing the following steps: -</p> - -<ul> -<li>Write an <tt>XXXCodeEmitter.cpp</tt> file that contains a machine function - pass that transforms target-machine instructions into relocatable machine - code.</li> - -<li>Write an <tt>XXXJITInfo.cpp</tt> file that implements the JIT interfaces for - target-specific code-generation activities, such as emitting machine code - and stubs.</li> - -<li>Modify <tt>XXXTargetMachine</tt> so that it provides a - <tt>TargetJITInfo</tt> object through its <tt>getJITInfo</tt> method.</li> -</ul> - -<p> -There are several different approaches to writing the JIT support code. For -instance, TableGen and target descriptor files may be used for creating a JIT -code generator, but are not mandatory. For the Alpha and PowerPC target -machines, TableGen is used to generate <tt>XXXGenCodeEmitter.inc</tt>, which -contains the binary coding of machine instructions and the -<tt>getBinaryCodeForInstr</tt> method to access those codes. Other JIT -implementations do not. -</p> - -<p> -Both <tt>XXXJITInfo.cpp</tt> and <tt>XXXCodeEmitter.cpp</tt> must include the -<tt>llvm/CodeGen/MachineCodeEmitter.h</tt> header file that defines the -<tt>MachineCodeEmitter</tt> class containing code for several callback functions -that write data (in bytes, words, strings, etc.) to the output stream. -</p> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="mce">Machine Code Emitter</a> -</div> - -<div class="doc_text"> - -<p> -In <tt>XXXCodeEmitter.cpp</tt>, a target-specific of the <tt>Emitter</tt> class -is implemented as a function pass (subclass -of <tt>MachineFunctionPass</tt>). The target-specific implementation -of <tt>runOnMachineFunction</tt> (invoked by -<tt>runOnFunction</tt> in <tt>MachineFunctionPass</tt>) iterates through the -<tt>MachineBasicBlock</tt> calls <tt>emitInstruction</tt> to process each -instruction and emit binary code. <tt>emitInstruction</tt> is largely -implemented with case statements on the instruction types defined in -<tt>XXXInstrInfo.h</tt>. For example, in <tt>X86CodeEmitter.cpp</tt>, -the <tt>emitInstruction</tt> method is built around the following switch/case -statements: -</p> - -<div class="doc_code"> -<pre> -switch (Desc->TSFlags & X86::FormMask) { -case X86II::Pseudo: // for not yet implemented instructions - ... // or pseudo-instructions - break; -case X86II::RawFrm: // for instructions with a fixed opcode value - ... - break; -case X86II::AddRegFrm: // for instructions that have one register operand - ... // added to their opcode - break; -case X86II::MRMDestReg:// for instructions that use the Mod/RM byte - ... // to specify a destination (register) - break; -case X86II::MRMDestMem:// for instructions that use the Mod/RM byte - ... // to specify a destination (memory) - break; -case X86II::MRMSrcReg: // for instructions that use the Mod/RM byte - ... // to specify a source (register) - break; -case X86II::MRMSrcMem: // for instructions that use the Mod/RM byte - ... // to specify a source (memory) - break; -case X86II::MRM0r: case X86II::MRM1r: // for instructions that operate on -case X86II::MRM2r: case X86II::MRM3r: // a REGISTER r/m operand and -case X86II::MRM4r: case X86II::MRM5r: // use the Mod/RM byte and a field -case X86II::MRM6r: case X86II::MRM7r: // to hold extended opcode data - ... - break; -case X86II::MRM0m: case X86II::MRM1m: // for instructions that operate on -case X86II::MRM2m: case X86II::MRM3m: // a MEMORY r/m operand and -case X86II::MRM4m: case X86II::MRM5m: // use the Mod/RM byte and a field -case X86II::MRM6m: case X86II::MRM7m: // to hold extended opcode data - ... - break; -case X86II::MRMInitReg: // for instructions whose source and - ... // destination are the same register - break; -} -</pre> -</div> - -<p> -The implementations of these case statements often first emit the opcode and -then get the operand(s). Then depending upon the operand, helper methods may be -called to process the operand(s). For example, in <tt>X86CodeEmitter.cpp</tt>, -for the <tt>X86II::AddRegFrm</tt> case, the first data emitted -(by <tt>emitByte</tt>) is the opcode added to the register operand. Then an -object representing the machine operand, <tt>MO1</tt>, is extracted. The helper -methods such as <tt>isImmediate</tt>, -<tt>isGlobalAddress</tt>, <tt>isExternalSymbol</tt>, <tt>isConstantPoolIndex</tt>, and -<tt>isJumpTableIndex</tt> determine the operand -type. (<tt>X86CodeEmitter.cpp</tt> also has private methods such -as <tt>emitConstant</tt>, <tt>emitGlobalAddress</tt>, -<tt>emitExternalSymbolAddress</tt>, <tt>emitConstPoolAddress</tt>, -and <tt>emitJumpTableAddress</tt> that emit the data into the output stream.) -</p> - -<div class="doc_code"> -<pre> -case X86II::AddRegFrm: - MCE.emitByte(BaseOpcode + getX86RegNum(MI.getOperand(CurOp++).getReg())); - - if (CurOp != NumOps) { - const MachineOperand &MO1 = MI.getOperand(CurOp++); - unsigned Size = X86InstrInfo::sizeOfImm(Desc); - if (MO1.isImmediate()) - emitConstant(MO1.getImm(), Size); - else { - unsigned rt = Is64BitMode ? X86::reloc_pcrel_word - : (IsPIC ? X86::reloc_picrel_word : X86::reloc_absolute_word); - if (Opcode == X86::MOV64ri) - rt = X86::reloc_absolute_dword; // FIXME: add X86II flag? - if (MO1.isGlobalAddress()) { - bool NeedStub = isa<Function>(MO1.getGlobal()); - bool isLazy = gvNeedsLazyPtr(MO1.getGlobal()); - emitGlobalAddress(MO1.getGlobal(), rt, MO1.getOffset(), 0, - NeedStub, isLazy); - } else if (MO1.isExternalSymbol()) - emitExternalSymbolAddress(MO1.getSymbolName(), rt); - else if (MO1.isConstantPoolIndex()) - emitConstPoolAddress(MO1.getIndex(), rt); - else if (MO1.isJumpTableIndex()) - emitJumpTableAddress(MO1.getIndex(), rt); - } - } - break; -</pre> -</div> - -<p> -In the previous example, <tt>XXXCodeEmitter.cpp</tt> uses the -variable <tt>rt</tt>, which is a RelocationType enum that may be used to -relocate addresses (for example, a global address with a PIC base offset). The -<tt>RelocationType</tt> enum for that target is defined in the short -target-specific <tt>XXXRelocations.h</tt> file. The <tt>RelocationType</tt> is used by -the <tt>relocate</tt> method defined in <tt>XXXJITInfo.cpp</tt> to rewrite -addresses for referenced global symbols. -</p> - -<p> -For example, <tt>X86Relocations.h</tt> specifies the following relocation types -for the X86 addresses. In all four cases, the relocated value is added to the -value already in memory. For <tt>reloc_pcrel_word</tt> -and <tt>reloc_picrel_word</tt>, there is an additional initial adjustment. -</p> - -<div class="doc_code"> -<pre> -enum RelocationType { - reloc_pcrel_word = 0, // add reloc value after adjusting for the PC loc - reloc_picrel_word = 1, // add reloc value after adjusting for the PIC base - reloc_absolute_word = 2, // absolute relocation; no additional adjustment - reloc_absolute_dword = 3 // absolute relocation; no additional adjustment -}; -</pre> -</div> - -</div> - -<!-- ======================================================================= --> -<div class="doc_subsection"> - <a name="targetJITInfo">Target JIT Info</a> -</div> - -<div class="doc_text"> - -<p> -<tt>XXXJITInfo.cpp</tt> implements the JIT interfaces for target-specific -code-generation activities, such as emitting machine code and stubs. At minimum, -a target-specific version of <tt>XXXJITInfo</tt> implements the following: -</p> - -<ul> -<li><tt>getLazyResolverFunction</tt> — Initializes the JIT, gives the - target a function that is used for compilation.</li> - -<li><tt>emitFunctionStub</tt> — Returns a native function with a specified - address for a callback function.</li> - -<li><tt>relocate</tt> — Changes the addresses of referenced globals, based - on relocation types.</li> - -<li>Callback function that are wrappers to a function stub that is used when the - real target is not initially known.</li> -</ul> - -<p> -<tt>getLazyResolverFunction</tt> is generally trivial to implement. It makes the -incoming parameter as the global <tt>JITCompilerFunction</tt> and returns the -callback function that will be used a function wrapper. For the Alpha target -(in <tt>AlphaJITInfo.cpp</tt>), the <tt>getLazyResolverFunction</tt> -implementation is simply: -</p> - -<div class="doc_code"> -<pre> -TargetJITInfo::LazyResolverFn AlphaJITInfo::getLazyResolverFunction( - JITCompilerFn F) { - JITCompilerFunction = F; - return AlphaCompilationCallback; -} -</pre> -</div> - -<p> -For the X86 target, the <tt>getLazyResolverFunction</tt> implementation is a -little more complication, because it returns a different callback function for -processors with SSE instructions and XMM registers. -</p> - -<p> -The callback function initially saves and later restores the callee register -values, incoming arguments, and frame and return address. The callback function -needs low-level access to the registers or stack, so it is typically implemented -with assembler. -</p> - -</div> - -<!-- *********************************************************************** --> - -<hr> -<address> - <a href="http://jigsaw.w3.org/css-validator/check/referer"><img - src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a> - <a href="http://validator.w3.org/check/referer"><img - src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a> - - <a href="http://www.woo.com">Mason Woo</a> and <a href="http://misha.brukman.net">Misha Brukman</a><br> - <a href="http://llvm.org">The LLVM Compiler Infrastructure</a> - <br> - Last modified: $Date$ -</address> - -</body> -</html> |