mirror of
https://github.com/RPCS3/llvm.git
synced 2024-12-02 00:37:09 +00:00
7a15273892
JP Bonn! git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59876 91177308-0d34-0410-b5e6-96231b3b80d8
2163 lines
89 KiB
HTML
2163 lines
89 KiB
HTML
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
|
|
"http://www.w3.org/TR/html4/strict.dtd">
|
|
<html>
|
|
<head>
|
|
<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="#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 IR (intermediate representation) 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">
|
|
These essential documents must be read before reading this document:
|
|
<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 to the 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 (tblgen) application that manages domain-specific
|
|
information to support LLVM code generation. TableGen processes input from a
|
|
target description file (.td 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 FunctionPass, as are several SelectionDAG processing steps.
|
|
</li>
|
|
</ul>
|
|
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 (GAS), 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 lists of target machine dependent features.
|
|
</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 (intermediate representation)
|
|
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 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 DAG (directed
|
|
acyclic graph) 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 (that is, 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 .cpp and .h 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 LLVMDummy (for example, see the MIPS target). Alternatively, you can
|
|
split the library into LLVMDummyCodeGen and LLVMDummyAsmPrinter, 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 lots 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 TargetMachine. 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 LLVMTargetMachine. (To create a target from scratch, create a subclass of
|
|
TargetMachine.)</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>LLVMTargetMachine is designed as a base class for targets
|
|
implemented with the LLVM target-independent code generator. The
|
|
LLVMTargetMachine class should be specialized by a concrete target class that
|
|
implements the various virtual methods. LLVMTargetMachine is defined as a
|
|
subclass of TargetMachine in <tt>include/llvm/Target/TargetMachine.h</tt>. The
|
|
TargetMachine class implementation (<tt>TargetMachine.cpp</tt>) also processes numerous
|
|
command-line options. </p>
|
|
|
|
<p>To create a concrete target-specific subclass of
|
|
LLVMTargetMachine, start by copying an existing TargetMachine 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 XXX, the implementation of XXXTargetMachine
|
|
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. XXXTargetMachine 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>
|
|
|
|
<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);
|
|
virtual bool addAssemblyEmitter(PassManagerBase &PM, bool Fast,
|
|
std::ostream &Out);
|
|
};
|
|
|
|
} // end namespace llvm
|
|
</pre>
|
|
</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 XXXTargetMachine constructor should specify a
|
|
TargetDescription 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>
|
|
|
|
<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 class="doc_text">
|
|
<p>Hyphens separate portions of the TargetDescription string. </p>
|
|
<ul>
|
|
<li>The "E" in the string indicates a big-endian target data model; a
|
|
lower-case "e" would indicate little-endian. </li>
|
|
<li>"p:" is followed by pointer information: size, ABI alignment, and
|
|
preferred alignment. If only two figures follow "p:", 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: "i", "f", "v", or "a"
|
|
(corresponding to integer, floating point, vector, or aggregate). "i", "v", or
|
|
"a" are followed by ABI alignment and preferred alignment. "f" is followed by
|
|
three values, the first indicates the size of a long double, then ABI alignment
|
|
and preferred alignment.</li>
|
|
</ul>
|
|
<p>You must also register your target using the RegisterTarget
|
|
template. (See the TargetMachineRegistry class.) For example, in <tt>SparcTargetMachine.cpp</tt>,
|
|
the target is registered with:</p>
|
|
</div>
|
|
|
|
<div class="doc_code">
|
|
<pre>
|
|
namespace {
|
|
// Register the target.
|
|
RegisterTarget<SparcTargetMachine>X("sparc", "SPARC");
|
|
}
|
|
</pre>
|
|
</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 XXXRegisterInfo (where XXX identifies the target) and represents the
|
|
class register file data that is used for register allocation and 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 that include 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
|
|
XXXRegisterInfo 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 Register class (specified in <tt>Target.td</tt>) is used to
|
|
define an object for each register. The specified string n becomes the Name of
|
|
the register. The basic Register object does not have any subregisters and does
|
|
not specify any aliases.</p>
|
|
</div>
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<p>For example, in the <tt>X86RegisterInfo.td</tt> file, there are register
|
|
definitions that utilize the Register class, such as:</p>
|
|
</div>
|
|
<div class="doc_code">
|
|
<pre>
|
|
def AL : Register<"AL">,
|
|
DwarfRegNum<[0, 0, 0]>;
|
|
</pre>
|
|
</div>
|
|
|
|
<div class="doc_text">
|
|
<p>This defines the register AL and assigns it values (with
|
|
DwarfRegNum) that are used by <tt>gcc</tt>, <tt>gdb</tt>, or a debug information writer (such as
|
|
DwarfWriter in <tt>llvm/lib/CodeGen</tt>) to identify a register. For register AL,
|
|
DwarfRegNum takes an array of 3 values, representing 3 different modes: the
|
|
first element is for X86-64, the second for EH (exception handling) 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>
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<p>From the register info file, TableGen generates a
|
|
TargetRegisterDesc object for each register. TargetRegisterDesc is defined in
|
|
<tt>include/llvm/Target/TargetRegisterInfo.h</tt> with the following fields:</p>
|
|
</div>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<p>TableGen uses the entire target description file (<tt>.td</tt>) to
|
|
determine text names for the register (in the AsmName and Name fields of
|
|
TargetRegisterDesc) and the relationships of other registers to the defined
|
|
register (in the other TargetRegisterDesc fields). In this example, other
|
|
definitions establish the registers "AX", "EAX", and "RAX" as aliases for one
|
|
another, so TableGen generates a null-terminated array (AL_AliasSet) for this
|
|
register alias set. </p>
|
|
|
|
<p>The Register class is commonly used as a base class for more
|
|
complex classes. In <tt>Target.td</tt>, the Register class is the base for the
|
|
RegisterWithSubRegs class that is used to define registers that need to specify
|
|
subregisters in the SubRegs list, as shown here:</p>
|
|
</div>
|
|
<div class="doc_code">
|
|
<pre>
|
|
class RegisterWithSubRegs<string n,
|
|
list<Register> subregs> : Register<n> {
|
|
let SubRegs = subregs;
|
|
}</pre>
|
|
</div>
|
|
|
|
<div class="doc_text">
|
|
<p>In <tt>SparcRegisterInfo.td</tt>, additional register classes are defined
|
|
for SPARC: a Register subclass, SparcReg, and further subclasses: Ri, Rf, and
|
|
Rd. SPARC registers are identified by 5-bit ID numbers, which is a feature
|
|
common to these subclasses. Note the use of ‘let’ expressions to override values
|
|
that are initially defined in a superclass (such as SubRegs field in the Rd
|
|
class). </p>
|
|
</div>
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>In the <tt>SparcRegisterInfo.td</tt> file, there are register definitions
|
|
that utilize these subclasses of Register, such as:</p>
|
|
</div>
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>The last two registers shown above (D0 and D1) 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 RegisterClass 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>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<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 ValueType register type values
|
|
that are defined in <tt>include/llvm/CodeGen/ValueTypes.td</tt>. Defined values include
|
|
integer types (such as i16, i32, and i1 for Boolean), floating-point types
|
|
(f32, f64), and vector types (for example, v8i16 for an 8 x i16 vector). All
|
|
registers in a RegisterClass must have the same ValueType, 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 RegisterClass 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:
|
|
FPRegs, DFPRegs, and IntRegs. For all three register classes, the first
|
|
argument defines the namespace with the string “SP”. FPRegs defines a group of 32
|
|
single-precision floating-point registers (F0 to F31); DFPRegs defines a group
|
|
of 16 double-precision registers (D0-D15). For IntRegs, the MethodProtos and
|
|
MethodBodies methods are used by TableGen to insert the specified code into generated
|
|
output.</p>
|
|
</div>
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<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
|
|
SparcGenRegisterInfo that uses TargetRegisterInfo as its base. It also
|
|
specifies types, based upon the defined register classes: DFPRegsClass, FPRegsClass,
|
|
and IntRegsClass. </p>
|
|
|
|
<p><tt>SparcRegisterInfo.td</tt> also generates SparcGenRegisterInfo.inc,
|
|
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 IntRegs reflects the
|
|
order in the definition of IntRegs in the target description file. Take special
|
|
note of the use of MethodBodies in <tt>SparcRegisterInfo.td</tt> to create code in
|
|
<tt>SparcGenRegisterInfo.inc</tt>. MethodProtos generates similar code in
|
|
<tt>SparcGenRegisterInfo.h.inc</tt>.</p>
|
|
</div>
|
|
|
|
<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 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 XXXRegisterInfo, which
|
|
implements the interface described in <tt>TargetRegisterInfo.h</tt>. These functions
|
|
return 0, NULL, or false, unless overridden. Here’s 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 SelectionDAG with nodes that are instances of the SDNode class
|
|
containing target instructions. An SDNode 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 NodeType enum in the ISD namespace).</p>
|
|
|
|
<p>TableGen uses the following target description (.td) input files
|
|
to generate much of the code for instruction definition:</p>
|
|
<ul>
|
|
<li><tt>Target.td</tt>, where the Instruction, Operand, InstrInfo, and other
|
|
fundamental classes are defined</li>
|
|
|
|
<li><tt>TargetSelectionDAG.td</tt>, used by SelectionDAG instruction selection
|
|
generators, contains SDTC* classes (selection DAG type constraint), definitions
|
|
of SelectionDAG nodes (such as imm, cond, bb, add, fadd, sub), and pattern
|
|
support (Pattern, Pat, PatFrag, PatLeaf, ComplexPattern)</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 XXX is the
|
|
name of the target. The <tt>XXX.td</tt> file includes the other .td input files, but its
|
|
contents are only directly important for subtargets.</p>
|
|
|
|
<p>You should describe
|
|
a concrete target-specific class
|
|
XXXInstrInfo that represents machine
|
|
instructions supported by a target machine. XXXInstrInfo contains an array of
|
|
XXXInstrDescriptor 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>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>A SelectionDAG node (SDNode) 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 ADDri and ADDrr.</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 CALL instruction. Format 2 is for
|
|
branch on condition codes and SETHI (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>. InstSP is a base
|
|
class for other instruction classes. Additional base classes are specified for
|
|
more precise formats: for example in <tt>SparcInstrFormat.td</tt>, F2_1 is for SETHI,
|
|
and F2_2 is for branches. There are three other base classes: F3_1 for
|
|
register/register operations, F3_2 for register/immediate operations, and F3_3 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, LDrr, defines
|
|
the Load Integer instruction for a Word (the LD SPARC opcode) from a memory
|
|
address to a register. The first parameter, the value 3 (11<sub>2</sub>), is
|
|
the operation value for this category of operation. The second parameter
|
|
(000000<sub>2</sub>) is the specific operation value for LD/Load Word. The
|
|
third parameter is the output destination, which is a register operand and
|
|
defined in the Register target description file (IntRegs). </p>
|
|
</div>
|
|
<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:$addr))]>;
|
|
</pre>
|
|
</div>
|
|
|
|
<div class="doc_text">
|
|
<p>The fourth
|
|
parameter is the input source, which uses the address operand MEMrr that is
|
|
defined earlier in <tt>SparcInstrInfo.td</tt>:</p>
|
|
</div>
|
|
<div class="doc_code">
|
|
<pre>def MEMrr : Operand<i32> {
|
|
let PrintMethod = "printMemOperand";
|
|
let MIOperandInfo = (ops IntRegs, IntRegs);
|
|
}
|
|
</pre>
|
|
</div>
|
|
<div class="doc_text">
|
|
<p>The fifth parameter is a string that is used by the assembly
|
|
printer and can be left as an empty string until the assembly printer interface
|
|
is implemented. The sixth and final parameter is the pattern used to match the
|
|
instruction during the SelectionDAG Select Phase described in
|
|
(<a href="http://www.llvm.org/docs/CodeGenerator.html">The LLVM Target-Independent Code Generator</a>).
|
|
This parameter is detailed in the next section, <a href="#InstructionSelector">Instruction Selector</a>.</p>
|
|
|
|
<p>Instruction class definitions are not overloaded for different
|
|
operand types, so separate versions of instructions are needed for register,
|
|
memory, or immediate value operands. For example, to perform a
|
|
Load Integer instruction for a Word
|
|
from an immediate operand to a register, the following instruction class is
|
|
defined: </p>
|
|
</div>
|
|
<div class="doc_code">
|
|
<pre>def LDri : F3_2 <3, 0b000000, (outs IntRegs:$dst), (ins MEMri:$addr),
|
|
"ld [$addr], $dst",
|
|
[(set IntRegs:$dst, (load ADDRri:$addr))]>;
|
|
</pre>
|
|
</div>
|
|
<div class="doc_text">
|
|
<p>Writing these definitions for so many similar instructions can
|
|
involve a lot of cut and paste. In td files, the <tt>multiclass</tt> directive enables
|
|
the creation of templates to define several instruction classes at once (using
|
|
the <tt>defm</tt> directive). For example in
|
|
<tt>SparcInstrInfo.td</tt>, the <tt>multiclass</tt> pattern F3_12 is defined to create 2
|
|
instruction classes each time F3_12 is invoked: </p>
|
|
</div>
|
|
<div class="doc_code">
|
|
<pre>multiclass F3_12 <string OpcStr, bits<6> Op3Val, SDNode OpNode> {
|
|
def rr : F3_1 <2, Op3Val,
|
|
(outs IntRegs:$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>
|
|
<div class="doc_text">
|
|
<p>So when the <tt>defm</tt> directive is used for the XOR and ADD
|
|
instructions, as seen below, it creates four instruction objects: XORrr, XORri,
|
|
ADDrr, and ADDri.</p>
|
|
</div>
|
|
<div class="doc_code">
|
|
<pre>defm XOR : F3_12<"xor", 0b000011, xor>;
|
|
defm ADD : F3_12<"add", 0b000000, add>;
|
|
</pre>
|
|
</div>
|
|
|
|
<div class="doc_text">
|
|
<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>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<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>; that is, <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 XNORrr instruction as a F3_1 format
|
|
instruction having three operands.</p>
|
|
</div>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<p>The instruction templates in <tt>SparcInstrFormats.td</tt> show the base class for F3_1 is InstSP.</p>
|
|
</div>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>
|
|
InstSP leaves the op field unbound.
|
|
</p>
|
|
</div>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>
|
|
F3 binds the op field and defines the rd, op3, and rs1 fields. F3 format instructions will
|
|
bind the operands rd, op3, and rs1 fields.
|
|
</p>
|
|
</div>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>
|
|
F3_1 binds the op3 field and defines the rs2 fields. F3_1 format instructions will
|
|
bind the operands to the rd, rs1, and rs2 fields. This results in the XNORrr instruction
|
|
binding $dst, $b, and $c operands to the rd, rs1, and rs2 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 XXXInstrInfo, which
|
|
implements the interface described in <tt>TargetInstrInfo.h</tt>. These functions return
|
|
0 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 FrameIndex 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 FrameIndex 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 XXXInstrInfo 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>MachineBasicBlock &MBB – the incoming block to be
|
|
examined</li>
|
|
|
|
<li>MachineBasicBlock *&TBB – a destination block that is
|
|
returned; for a conditional branch that evaluates to true, TBB is the
|
|
destination </li>
|
|
|
|
<li>MachineBasicBlock *&FBB – for a conditional branch that
|
|
evaluates to false, FBB is returned as the destination</li>
|
|
|
|
<li>std::vector<MachineOperand> &Cond – 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 TBB or FBB, so both parameters return NULL. 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>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<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 TBB parameter. </p>
|
|
</div>
|
|
|
|
<div class="doc_code">
|
|
<pre>if (LastOpc == ARM::B || LastOpc == ARM::tB) {
|
|
TBB = LastInst->getOperand(0).getMBB();
|
|
return false;
|
|
}
|
|
</pre>
|
|
</div>
|
|
|
|
<div class="doc_text">
|
|
<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 TBB parameter. </p>
|
|
</div>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<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 TBB parameter and a list of operands in the <tt>Cond</tt>
|
|
parameter to evaluate the condition. </p>
|
|
</div>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<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 ‘true’) in the TBB parameter and the unconditional branch
|
|
destination in the FBB (corresponding to a conditional evaluation of ‘false’).
|
|
A list of operands to evaluate the condition should be returned in the <tt>Cond</tt>
|
|
parameter.</p>
|
|
</div>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<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 SelectionDAG to represent LLVM IR instructions, and nodes
|
|
of the SelectionDAG 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 Selection DAG. </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 FunctionPass class or a subclass of FunctionPass. 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 SelectionDAG 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 SelectionDAG 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>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<p>ADDRrr is a memory mode that is also defined in <tt>SparcInstrInfo.td</tt>:</p>
|
|
</div>
|
|
|
|
<div class="doc_code">
|
|
<pre>def ADDRrr : ComplexPattern<i32, 2, "SelectADDRrr", [], []>;
|
|
</pre>
|
|
</div>
|
|
|
|
<div class="doc_text">
|
|
<p>The definition of ADDRrr refers to SelectADDRrr, 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>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<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
|
|
ISD::STORE opcode.</p>
|
|
</div>
|
|
|
|
<div class="doc_code">
|
|
<pre>SDNode *SelectCode(SDOperand N) {
|
|
...
|
|
MVT::ValueType NVT = N.Val->getValueType(0);
|
|
switch (N.getOpcode()) {
|
|
case ISD::STORE: {
|
|
switch (NVT) {
|
|
default:
|
|
return Select_ISD_STORE(N);
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
...
|
|
</pre>
|
|
</div>
|
|
<div class="doc_text">
|
|
<p>The pattern for STrr is matched, so elsewhere in
|
|
<tt>XXXGenDAGISel.inc</tt>, code for STrr 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>
|
|
|
|
<div class="doc_code">
|
|
<pre>SDNode *Select_ISD_STORE(const SDOperand &N) {
|
|
SDOperand Chain = N.getOperand(0);
|
|
if (Predicate_store(N.Val)) {
|
|
SDOperand N1 = N.getOperand(1);
|
|
SDOperand N2 = N.getOperand(2);
|
|
SDOperand CPTmp0;
|
|
SDOperand 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.Val->getValueType(0) == MVT::i32 &&
|
|
N2.Val->getValueType(0) == MVT::i32) {
|
|
return Emit_22(N, SP::STrr, CPTmp0, CPTmp1);
|
|
}
|
|
...
|
|
</pre>
|
|
</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 XXXTargetLowering 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>
|
|
|
|
<div class="doc_code">
|
|
<pre>addRegisterClass(MVT::i32, SP::IntRegsRegisterClass);
|
|
addRegisterClass(MVT::f32, SP::FPRegsRegisterClass);
|
|
addRegisterClass(MVT::f64, SP::DFPRegsRegisterClass);
|
|
</pre>
|
|
</div>
|
|
|
|
<div class="doc_text">
|
|
<p>You should examine the node types in the ISD 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 LegalAction 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 LegalAction 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>
|
|
|
|
<div class="doc_code">
|
|
<pre>setLoadExtAction(ISD::SEXTLOAD, MVT::i1, Promote);
|
|
</pre>
|
|
</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>
|
|
|
|
<div class="doc_code">
|
|
<pre>setOperationAction(ISD::FSIN, MVT::f32, Expand);
|
|
setOperationAction(ISD::FCOS, MVT::f32, Expand);
|
|
</pre>
|
|
</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>
|
|
|
|
<div class="doc_code">
|
|
<pre>setOperationAction(ISD::FP_TO_SINT, MVT::i32, Custom);
|
|
</pre>
|
|
</div>
|
|
<div class="doc_text">
|
|
<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 FP_TO_SINT opcode will call the <tt>LowerFP_TO_SINT</tt> method:</p>
|
|
</div>
|
|
|
|
<div class="doc_code">
|
|
<pre>SDOperand SparcTargetLowering::LowerOperation(
|
|
SDOperand Op, SelectionDAG &DAG) {
|
|
switch (Op.getOpcode()) {
|
|
case ISD::FP_TO_SINT: return LowerFP_TO_SINT(Op, DAG);
|
|
...
|
|
}
|
|
}
|
|
</pre>
|
|
</div>
|
|
<div class="doc_text">
|
|
<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>
|
|
|
|
<div class="doc_code">
|
|
<pre>static SDOperand LowerFP_TO_SINT(SDOperand 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 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 CTPOP (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>
|
|
|
|
<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 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 CCIfType and
|
|
CCAssignToReg interfaces. If the CCIfType predicate is true (that is, if the
|
|
current argument is of type f32 or f64), then the action is performed. In this
|
|
case, the CCAssignToReg action assigns the argument value to the first
|
|
available register: either R0 or R1. </p>
|
|
</div>
|
|
<div class="doc_code">
|
|
<pre>CCIfType<[f32,f64], CCAssignToReg<[R0, R1]>>
|
|
</pre>
|
|
</div>
|
|
<div class="doc_text">
|
|
<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
|
|
(CC_Sparc32). The definition of RetCC_Sparc32 (shown below) indicates which
|
|
registers are used for specified scalar return types. A single-precision float
|
|
is returned to register F0, and a double-precision float goes to register D0. A
|
|
32-bit integer is returned in register I0 or I1. </p>
|
|
</div>
|
|
|
|
<div class="doc_code">
|
|
<pre>def RetCC_Sparc32 : CallingConv<[
|
|
CCIfType<[i32], CCAssignToReg<[I0, I1]>>,
|
|
CCIfType<[f32], CCAssignToReg<[F0]>>,
|
|
CCIfType<[f64], CCAssignToReg<[D0]>>
|
|
]>;
|
|
</pre>
|
|
</div>
|
|
<div class="doc_text">
|
|
<p>The definition of CC_Sparc32 in <tt>SparcCallingConv.td</tt> introduces
|
|
CCAssignToStack, 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>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>CCDelegateTo 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 RetCC_X86_32_C
|
|
ends with CCDelegateTo. After the current value is assigned to the register ST0
|
|
or ST1, the RetCC_X86Common is invoked.</p>
|
|
</div>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>CCIfCC 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 Fast calling convention is in use, then RetCC_X86_32_Fast
|
|
is invoked. If the SSECall calling convention is in use, then RetCC_X86_32_SSE
|
|
is invoked. </p>
|
|
</div>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>Other calling convention interfaces include:</p>
|
|
<ul>
|
|
<li>CCIf <predicate, action> - if the predicate matches, apply
|
|
the action</li>
|
|
|
|
<li>CCIfInReg <action> - if the argument is marked with the
|
|
‘inreg’ attribute, then apply the action </li>
|
|
|
|
<li>CCIfNest <action> - if the argument is marked with the
|
|
‘nest’ attribute, then apply the action</li>
|
|
|
|
<li>CCIfNotVarArg <action> - if the current function does not
|
|
take a variable number of arguments, apply the action</li>
|
|
|
|
<li>CCAssignToRegWithShadow <registerList, shadowList> -
|
|
similar to CCAssignToReg, but with a shadow list of registers</li>
|
|
|
|
<li>CCPassByVal <size, align> - assign value to a stack slot
|
|
with the minimum specified size and alignment </li>
|
|
|
|
<li>CCPromoteToType <type> - promote the current value to the specified
|
|
type</li>
|
|
|
|
<li>CallingConv <[actions]> - 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 XXXTargetAsmInfo class (a subclass of TargetAsmInfo). </li>
|
|
|
|
<li>Write <tt>XXXTargetAsmInfo.cpp</tt>, which contains target-specific values
|
|
for TargetAsmInfo properties and sometimes new implementations for methods</li>
|
|
|
|
<li>Write <tt>XXXAsmPrinter.cpp</tt>, which implements the AsmPrinter 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 XXXTargetAsmInfo class for use in <tt>XXXTargetAsmInfo.cpp</tt>. Similarly,
|
|
<tt>XXXTargetAsmInfo.cpp</tt> usually has a few declarations of XXXTargetAsmInfo replacement
|
|
values that override the default values in <tt>TargetAsmInfo.cpp</tt>. For example in
|
|
<tt>SparcTargetAsmInfo.cpp</tt>, </p>
|
|
</div>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>The X86 assembly printer implementation (X86TargetAsmInfo) is an
|
|
example where the target specific TargetAsmInfo class uses overridden methods:
|
|
<tt>ExpandInlineAsm</tt> and <tt>PreferredEHDataFormat</tt>. </p>
|
|
|
|
<p>A target-specific implementation of AsmPrinter is written in
|
|
<tt>XXXAsmPrinter.cpp</tt>, which implements the AsmPrinter class that converts the LLVM
|
|
to printable assembly. The implementation must include the following headers
|
|
that have declarations for the AsmPrinter and MachineFunctionPass classes. The
|
|
MachineFunctionPass is a subclass of FunctionPass. </p>
|
|
</div>
|
|
|
|
<div class="doc_code">
|
|
<pre>#include "llvm/CodeGen/AsmPrinter.h"
|
|
#include "llvm/CodeGen/MachineFunctionPass.h"
|
|
</pre>
|
|
</div>
|
|
|
|
<div class="doc_text">
|
|
<p>As a FunctionPass, AsmPrinter first calls <tt>doInitialization</tt> to set
|
|
up the AsmPrinter. In SparcAsmPrinter, a Mangler object is instantiated to
|
|
process variable names.</p>
|
|
|
|
<p>In <tt>XXXAsmPrinter.cpp</tt>, the <tt>runOnMachineFunction</tt> method (declared
|
|
in MachineFunctionPass) must be implemented for XXXAsmPrinter. In
|
|
MachineFunctionPass, 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 XXXAsmPrinter 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>
|
|
|
|
<li><tt>printLabel</tt></li>
|
|
|
|
<li><tt>printPICJumpTableEntry</tt></li>
|
|
|
|
<li><tt>printPICJumpTableSetLabel</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. (<tt>printBasicBlockLabel</tt> is
|
|
another method that is implemented in <tt>AsmPrinter.cpp</tt> that may be directly used
|
|
in an implementation of XXXAsmPrinter.)</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 XXXAsmPrinter, 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
|
|
SubtargetFeature interface is defined. The first 4 string parameters of the
|
|
SubtargetFeature 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>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<p>In the <tt>Sparc.td</tt> file, the SubtargetFeature is used to define the
|
|
following features. </p>
|
|
</div>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<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>
|
|
|
|
<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>
|
|
|
|
<div class="doc_text">
|
|
<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>
|
|
|
|
<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 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 XXXTargetMachine so that it provides a TargetJITInfo
|
|
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 MachineCodeEmitter
|
|
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 Emitter class is
|
|
implemented as a function pass (subclass of MachineFunctionPass). The
|
|
target-specific implementation of <tt>runOnMachineFunction</tt> (invoked by
|
|
<tt>runOnFunction</tt> in MachineFunctionPass) iterates through the MachineBasicBlock
|
|
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>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<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, MO1, 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>
|
|
|
|
<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>
|
|
<div class="doc_text">
|
|
<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 RelocationType enum for
|
|
that target is defined in the short target-specific <tt>XXXRelocations.h</tt> file. The
|
|
RelocationType 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>
|
|
|
|
<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 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 XXXJITInfo 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 JITCompilerFunction 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>
|
|
|
|
<div class="doc_code">
|
|
<pre>TargetJITInfo::LazyResolverFn AlphaJITInfo::getLazyResolverFunction(
|
|
JITCompilerFn F)
|
|
{
|
|
JITCompilerFunction = F;
|
|
return AlphaCompilationCallback;
|
|
}
|
|
</pre>
|
|
</div>
|
|
<div class="doc_text">
|
|
<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" alt="Valid CSS!"></a>
|
|
<a href="http://validator.w3.org/check/referer"><img
|
|
src="http://www.w3.org/Icons/valid-html401" 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>
|