1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" 2 "http://www.w3.org/TR/html4/strict.dtd"> 3<html> 4<head> 5 <meta http-equiv="content-type" content="text/html; charset=utf-8"> 6 <title>The LLVM Target-Independent Code Generator</title> 7 <link rel="stylesheet" href="llvm.css" type="text/css"> 8 9 <style type="text/css"> 10 .unknown { background-color: #C0C0C0; text-align: center; } 11 .unknown:before { content: "?" } 12 .no { background-color: #C11B17 } 13 .no:before { content: "N" } 14 .partial { background-color: #F88017 } 15 .yes { background-color: #0F0; } 16 .yes:before { content: "Y" } 17 </style> 18 19</head> 20<body> 21 22<h1> 23 The LLVM Target-Independent Code Generator 24</h1> 25 26<ol> 27 <li><a href="#introduction">Introduction</a> 28 <ul> 29 <li><a href="#required">Required components in the code generator</a></li> 30 <li><a href="#high-level-design">The high-level design of the code 31 generator</a></li> 32 <li><a href="#tablegen">Using TableGen for target description</a></li> 33 </ul> 34 </li> 35 <li><a href="#targetdesc">Target description classes</a> 36 <ul> 37 <li><a href="#targetmachine">The <tt>TargetMachine</tt> class</a></li> 38 <li><a href="#targetdata">The <tt>TargetData</tt> class</a></li> 39 <li><a href="#targetlowering">The <tt>TargetLowering</tt> class</a></li> 40 <li><a href="#targetregisterinfo">The <tt>TargetRegisterInfo</tt> class</a></li> 41 <li><a href="#targetinstrinfo">The <tt>TargetInstrInfo</tt> class</a></li> 42 <li><a href="#targetframeinfo">The <tt>TargetFrameInfo</tt> class</a></li> 43 <li><a href="#targetsubtarget">The <tt>TargetSubtarget</tt> class</a></li> 44 <li><a href="#targetjitinfo">The <tt>TargetJITInfo</tt> class</a></li> 45 </ul> 46 </li> 47 <li><a href="#codegendesc">The "Machine" Code Generator classes</a> 48 <ul> 49 <li><a href="#machineinstr">The <tt>MachineInstr</tt> class</a></li> 50 <li><a href="#machinebasicblock">The <tt>MachineBasicBlock</tt> 51 class</a></li> 52 <li><a href="#machinefunction">The <tt>MachineFunction</tt> class</a></li> 53 </ul> 54 </li> 55 <li><a href="#mc">The "MC" Layer</a> 56 <ul> 57 <li><a href="#mcstreamer">The <tt>MCStreamer</tt> API</a></li> 58 <li><a href="#mccontext">The <tt>MCContext</tt> class</a> 59 <li><a href="#mcsymbol">The <tt>MCSymbol</tt> class</a></li> 60 <li><a href="#mcsection">The <tt>MCSection</tt> class</a></li> 61 <li><a href="#mcinst">The <tt>MCInst</tt> class</a></li> 62 </ul> 63 </li> 64 <li><a href="#codegenalgs">Target-independent code generation algorithms</a> 65 <ul> 66 <li><a href="#instselect">Instruction Selection</a> 67 <ul> 68 <li><a href="#selectiondag_intro">Introduction to SelectionDAGs</a></li> 69 <li><a href="#selectiondag_process">SelectionDAG Code Generation 70 Process</a></li> 71 <li><a href="#selectiondag_build">Initial SelectionDAG 72 Construction</a></li> 73 <li><a href="#selectiondag_legalize_types">SelectionDAG LegalizeTypes Phase</a></li> 74 <li><a href="#selectiondag_legalize">SelectionDAG Legalize Phase</a></li> 75 <li><a href="#selectiondag_optimize">SelectionDAG Optimization 76 Phase: the DAG Combiner</a></li> 77 <li><a href="#selectiondag_select">SelectionDAG Select Phase</a></li> 78 <li><a href="#selectiondag_sched">SelectionDAG Scheduling and Formation 79 Phase</a></li> 80 <li><a href="#selectiondag_future">Future directions for the 81 SelectionDAG</a></li> 82 </ul></li> 83 <li><a href="#liveintervals">Live Intervals</a> 84 <ul> 85 <li><a href="#livevariable_analysis">Live Variable Analysis</a></li> 86 <li><a href="#liveintervals_analysis">Live Intervals Analysis</a></li> 87 </ul></li> 88 <li><a href="#regalloc">Register Allocation</a> 89 <ul> 90 <li><a href="#regAlloc_represent">How registers are represented in 91 LLVM</a></li> 92 <li><a href="#regAlloc_howTo">Mapping virtual registers to physical 93 registers</a></li> 94 <li><a href="#regAlloc_twoAddr">Handling two address instructions</a></li> 95 <li><a href="#regAlloc_ssaDecon">The SSA deconstruction phase</a></li> 96 <li><a href="#regAlloc_fold">Instruction folding</a></li> 97 <li><a href="#regAlloc_builtIn">Built in register allocators</a></li> 98 </ul></li> 99 <li><a href="#codeemit">Code Emission</a></li> 100 </ul> 101 </li> 102 <li><a href="#nativeassembler">Implementing a Native Assembler</a></li> 103 104 <li><a href="#targetimpls">Target-specific Implementation Notes</a> 105 <ul> 106 <li><a href="#targetfeatures">Target Feature Matrix</a></li> 107 <li><a href="#tailcallopt">Tail call optimization</a></li> 108 <li><a href="#sibcallopt">Sibling call optimization</a></li> 109 <li><a href="#x86">The X86 backend</a></li> 110 <li><a href="#ppc">The PowerPC backend</a> 111 <ul> 112 <li><a href="#ppc_abi">LLVM PowerPC ABI</a></li> 113 <li><a href="#ppc_frame">Frame Layout</a></li> 114 <li><a href="#ppc_prolog">Prolog/Epilog</a></li> 115 <li><a href="#ppc_dynamic">Dynamic Allocation</a></li> 116 </ul></li> 117 <li><a href="#ptx">The PTX backend</a></li> 118 </ul></li> 119 120</ol> 121 122<div class="doc_author"> 123 <p>Written by the LLVM Team.</p> 124</div> 125 126<div class="doc_warning"> 127 <p>Warning: This is a work in progress.</p> 128</div> 129 130<!-- *********************************************************************** --> 131<h2> 132 <a name="introduction">Introduction</a> 133</h2> 134<!-- *********************************************************************** --> 135 136<div> 137 138<p>The LLVM target-independent code generator is a framework that provides a 139 suite of reusable components for translating the LLVM internal representation 140 to the machine code for a specified target—either in assembly form 141 (suitable for a static compiler) or in binary machine code format (usable for 142 a JIT compiler). The LLVM target-independent code generator consists of six 143 main components:</p> 144 145<ol> 146 <li><a href="#targetdesc">Abstract target description</a> interfaces which 147 capture important properties about various aspects of the machine, 148 independently of how they will be used. These interfaces are defined in 149 <tt>include/llvm/Target/</tt>.</li> 150 151 <li>Classes used to represent the <a href="#codegendesc">code being 152 generated</a> for a target. These classes are intended to be abstract 153 enough to represent the machine code for <i>any</i> target machine. These 154 classes are defined in <tt>include/llvm/CodeGen/</tt>. At this level, 155 concepts like "constant pool entries" and "jump tables" are explicitly 156 exposed.</li> 157 158 <li>Classes and algorithms used to represent code as the object file level, 159 the <a href="#mc">MC Layer</a>. These classes represent assembly level 160 constructs like labels, sections, and instructions. At this level, 161 concepts like "constant pool entries" and "jump tables" don't exist.</li> 162 163 <li><a href="#codegenalgs">Target-independent algorithms</a> used to implement 164 various phases of native code generation (register allocation, scheduling, 165 stack frame representation, etc). This code lives 166 in <tt>lib/CodeGen/</tt>.</li> 167 168 <li><a href="#targetimpls">Implementations of the abstract target description 169 interfaces</a> for particular targets. These machine descriptions make 170 use of the components provided by LLVM, and can optionally provide custom 171 target-specific passes, to build complete code generators for a specific 172 target. Target descriptions live in <tt>lib/Target/</tt>.</li> 173 174 <li><a href="#jit">The target-independent JIT components</a>. The LLVM JIT is 175 completely target independent (it uses the <tt>TargetJITInfo</tt> 176 structure to interface for target-specific issues. The code for the 177 target-independent JIT lives in <tt>lib/ExecutionEngine/JIT</tt>.</li> 178</ol> 179 180<p>Depending on which part of the code generator you are interested in working 181 on, different pieces of this will be useful to you. In any case, you should 182 be familiar with the <a href="#targetdesc">target description</a> 183 and <a href="#codegendesc">machine code representation</a> classes. If you 184 want to add a backend for a new target, you will need 185 to <a href="#targetimpls">implement the target description</a> classes for 186 your new target and understand the <a href="LangRef.html">LLVM code 187 representation</a>. If you are interested in implementing a 188 new <a href="#codegenalgs">code generation algorithm</a>, it should only 189 depend on the target-description and machine code representation classes, 190 ensuring that it is portable.</p> 191 192<!-- ======================================================================= --> 193<h3> 194 <a name="required">Required components in the code generator</a> 195</h3> 196 197<div> 198 199<p>The two pieces of the LLVM code generator are the high-level interface to the 200 code generator and the set of reusable components that can be used to build 201 target-specific backends. The two most important interfaces 202 (<a href="#targetmachine"><tt>TargetMachine</tt></a> 203 and <a href="#targetdata"><tt>TargetData</tt></a>) are the only ones that are 204 required to be defined for a backend to fit into the LLVM system, but the 205 others must be defined if the reusable code generator components are going to 206 be used.</p> 207 208<p>This design has two important implications. The first is that LLVM can 209 support completely non-traditional code generation targets. For example, the 210 C backend does not require register allocation, instruction selection, or any 211 of the other standard components provided by the system. As such, it only 212 implements these two interfaces, and does its own thing. Another example of 213 a code generator like this is a (purely hypothetical) backend that converts 214 LLVM to the GCC RTL form and uses GCC to emit machine code for a target.</p> 215 216<p>This design also implies that it is possible to design and implement 217 radically different code generators in the LLVM system that do not make use 218 of any of the built-in components. Doing so is not recommended at all, but 219 could be required for radically different targets that do not fit into the 220 LLVM machine description model: FPGAs for example.</p> 221 222</div> 223 224<!-- ======================================================================= --> 225<h3> 226 <a name="high-level-design">The high-level design of the code generator</a> 227</h3> 228 229<div> 230 231<p>The LLVM target-independent code generator is designed to support efficient 232 and quality code generation for standard register-based microprocessors. 233 Code generation in this model is divided into the following stages:</p> 234 235<ol> 236 <li><b><a href="#instselect">Instruction Selection</a></b> — This phase 237 determines an efficient way to express the input LLVM code in the target 238 instruction set. This stage produces the initial code for the program in 239 the target instruction set, then makes use of virtual registers in SSA 240 form and physical registers that represent any required register 241 assignments due to target constraints or calling conventions. This step 242 turns the LLVM code into a DAG of target instructions.</li> 243 244 <li><b><a href="#selectiondag_sched">Scheduling and Formation</a></b> — 245 This phase takes the DAG of target instructions produced by the 246 instruction selection phase, determines an ordering of the instructions, 247 then emits the instructions 248 as <tt><a href="#machineinstr">MachineInstr</a></tt>s with that ordering. 249 Note that we describe this in the <a href="#instselect">instruction 250 selection section</a> because it operates on 251 a <a href="#selectiondag_intro">SelectionDAG</a>.</li> 252 253 <li><b><a href="#ssamco">SSA-based Machine Code Optimizations</a></b> — 254 This optional stage consists of a series of machine-code optimizations 255 that operate on the SSA-form produced by the instruction selector. 256 Optimizations like modulo-scheduling or peephole optimization work 257 here.</li> 258 259 <li><b><a href="#regalloc">Register Allocation</a></b> — The target code 260 is transformed from an infinite virtual register file in SSA form to the 261 concrete register file used by the target. This phase introduces spill 262 code and eliminates all virtual register references from the program.</li> 263 264 <li><b><a href="#proepicode">Prolog/Epilog Code Insertion</a></b> — Once 265 the machine code has been generated for the function and the amount of 266 stack space required is known (used for LLVM alloca's and spill slots), 267 the prolog and epilog code for the function can be inserted and "abstract 268 stack location references" can be eliminated. This stage is responsible 269 for implementing optimizations like frame-pointer elimination and stack 270 packing.</li> 271 272 <li><b><a href="#latemco">Late Machine Code Optimizations</a></b> — 273 Optimizations that operate on "final" machine code can go here, such as 274 spill code scheduling and peephole optimizations.</li> 275 276 <li><b><a href="#codeemit">Code Emission</a></b> — The final stage 277 actually puts out the code for the current function, either in the target 278 assembler format or in machine code.</li> 279</ol> 280 281<p>The code generator is based on the assumption that the instruction selector 282 will use an optimal pattern matching selector to create high-quality 283 sequences of native instructions. Alternative code generator designs based 284 on pattern expansion and aggressive iterative peephole optimization are much 285 slower. This design permits efficient compilation (important for JIT 286 environments) and aggressive optimization (used when generating code offline) 287 by allowing components of varying levels of sophistication to be used for any 288 step of compilation.</p> 289 290<p>In addition to these stages, target implementations can insert arbitrary 291 target-specific passes into the flow. For example, the X86 target uses a 292 special pass to handle the 80x87 floating point stack architecture. Other 293 targets with unusual requirements can be supported with custom passes as 294 needed.</p> 295 296</div> 297 298<!-- ======================================================================= --> 299<h3> 300 <a name="tablegen">Using TableGen for target description</a> 301</h3> 302 303<div> 304 305<p>The target description classes require a detailed description of the target 306 architecture. These target descriptions often have a large amount of common 307 information (e.g., an <tt>add</tt> instruction is almost identical to a 308 <tt>sub</tt> instruction). In order to allow the maximum amount of 309 commonality to be factored out, the LLVM code generator uses 310 the <a href="TableGenFundamentals.html">TableGen</a> tool to describe big 311 chunks of the target machine, which allows the use of domain-specific and 312 target-specific abstractions to reduce the amount of repetition.</p> 313 314<p>As LLVM continues to be developed and refined, we plan to move more and more 315 of the target description to the <tt>.td</tt> form. Doing so gives us a 316 number of advantages. The most important is that it makes it easier to port 317 LLVM because it reduces the amount of C++ code that has to be written, and 318 the surface area of the code generator that needs to be understood before 319 someone can get something working. Second, it makes it easier to change 320 things. In particular, if tables and other things are all emitted 321 by <tt>tblgen</tt>, we only need a change in one place (<tt>tblgen</tt>) to 322 update all of the targets to a new interface.</p> 323 324</div> 325 326</div> 327 328<!-- *********************************************************************** --> 329<h2> 330 <a name="targetdesc">Target description classes</a> 331</h2> 332<!-- *********************************************************************** --> 333 334<div> 335 336<p>The LLVM target description classes (located in the 337 <tt>include/llvm/Target</tt> directory) provide an abstract description of 338 the target machine independent of any particular client. These classes are 339 designed to capture the <i>abstract</i> properties of the target (such as the 340 instructions and registers it has), and do not incorporate any particular 341 pieces of code generation algorithms.</p> 342 343<p>All of the target description classes (except the 344 <tt><a href="#targetdata">TargetData</a></tt> class) are designed to be 345 subclassed by the concrete target implementation, and have virtual methods 346 implemented. To get to these implementations, the 347 <tt><a href="#targetmachine">TargetMachine</a></tt> class provides accessors 348 that should be implemented by the target.</p> 349 350<!-- ======================================================================= --> 351<h3> 352 <a name="targetmachine">The <tt>TargetMachine</tt> class</a> 353</h3> 354 355<div> 356 357<p>The <tt>TargetMachine</tt> class provides virtual methods that are used to 358 access the target-specific implementations of the various target description 359 classes via the <tt>get*Info</tt> methods (<tt>getInstrInfo</tt>, 360 <tt>getRegisterInfo</tt>, <tt>getFrameInfo</tt>, etc.). This class is 361 designed to be specialized by a concrete target implementation 362 (e.g., <tt>X86TargetMachine</tt>) which implements the various virtual 363 methods. The only required target description class is 364 the <a href="#targetdata"><tt>TargetData</tt></a> class, but if the code 365 generator components are to be used, the other interfaces should be 366 implemented as well.</p> 367 368</div> 369 370<!-- ======================================================================= --> 371<h3> 372 <a name="targetdata">The <tt>TargetData</tt> class</a> 373</h3> 374 375<div> 376 377<p>The <tt>TargetData</tt> class is the only required target description class, 378 and it is the only class that is not extensible (you cannot derived a new 379 class from it). <tt>TargetData</tt> specifies information about how the 380 target lays out memory for structures, the alignment requirements for various 381 data types, the size of pointers in the target, and whether the target is 382 little-endian or big-endian.</p> 383 384</div> 385 386<!-- ======================================================================= --> 387<h3> 388 <a name="targetlowering">The <tt>TargetLowering</tt> class</a> 389</h3> 390 391<div> 392 393<p>The <tt>TargetLowering</tt> class is used by SelectionDAG based instruction 394 selectors primarily to describe how LLVM code should be lowered to 395 SelectionDAG operations. Among other things, this class indicates:</p> 396 397<ul> 398 <li>an initial register class to use for various <tt>ValueType</tt>s,</li> 399 400 <li>which operations are natively supported by the target machine,</li> 401 402 <li>the return type of <tt>setcc</tt> operations,</li> 403 404 <li>the type to use for shift amounts, and</li> 405 406 <li>various high-level characteristics, like whether it is profitable to turn 407 division by a constant into a multiplication sequence</li> 408</ul> 409 410</div> 411 412<!-- ======================================================================= --> 413<h3> 414 <a name="targetregisterinfo">The <tt>TargetRegisterInfo</tt> class</a> 415</h3> 416 417<div> 418 419<p>The <tt>TargetRegisterInfo</tt> class is used to describe the register file 420 of the target and any interactions between the registers.</p> 421 422<p>Registers in the code generator are represented in the code generator by 423 unsigned integers. Physical registers (those that actually exist in the 424 target description) are unique small numbers, and virtual registers are 425 generally large. Note that register #0 is reserved as a flag value.</p> 426 427<p>Each register in the processor description has an associated 428 <tt>TargetRegisterDesc</tt> entry, which provides a textual name for the 429 register (used for assembly output and debugging dumps) and a set of aliases 430 (used to indicate whether one register overlaps with another).</p> 431 432<p>In addition to the per-register description, the <tt>TargetRegisterInfo</tt> 433 class exposes a set of processor specific register classes (instances of the 434 <tt>TargetRegisterClass</tt> class). Each register class contains sets of 435 registers that have the same properties (for example, they are all 32-bit 436 integer registers). Each SSA virtual register created by the instruction 437 selector has an associated register class. When the register allocator runs, 438 it replaces virtual registers with a physical register in the set.</p> 439 440<p>The target-specific implementations of these classes is auto-generated from 441 a <a href="TableGenFundamentals.html">TableGen</a> description of the 442 register file.</p> 443 444</div> 445 446<!-- ======================================================================= --> 447<h3> 448 <a name="targetinstrinfo">The <tt>TargetInstrInfo</tt> class</a> 449</h3> 450 451<div> 452 453<p>The <tt>TargetInstrInfo</tt> class is used to describe the machine 454 instructions supported by the target. It is essentially an array of 455 <tt>TargetInstrDescriptor</tt> objects, each of which describes one 456 instruction the target supports. Descriptors define things like the mnemonic 457 for the opcode, the number of operands, the list of implicit register uses 458 and defs, whether the instruction has certain target-independent properties 459 (accesses memory, is commutable, etc), and holds any target-specific 460 flags.</p> 461 462</div> 463 464<!-- ======================================================================= --> 465<h3> 466 <a name="targetframeinfo">The <tt>TargetFrameInfo</tt> class</a> 467</h3> 468 469<div> 470 471<p>The <tt>TargetFrameInfo</tt> class is used to provide information about the 472 stack frame layout of the target. It holds the direction of stack growth, the 473 known stack alignment on entry to each function, and the offset to the local 474 area. The offset to the local area is the offset from the stack pointer on 475 function entry to the first location where function data (local variables, 476 spill locations) can be stored.</p> 477 478</div> 479 480<!-- ======================================================================= --> 481<h3> 482 <a name="targetsubtarget">The <tt>TargetSubtarget</tt> class</a> 483</h3> 484 485<div> 486 487<p>The <tt>TargetSubtarget</tt> class is used to provide information about the 488 specific chip set being targeted. A sub-target informs code generation of 489 which instructions are supported, instruction latencies and instruction 490 execution itinerary; i.e., which processing units are used, in what order, 491 and for how long.</p> 492 493</div> 494 495 496<!-- ======================================================================= --> 497<h3> 498 <a name="targetjitinfo">The <tt>TargetJITInfo</tt> class</a> 499</h3> 500 501<div> 502 503<p>The <tt>TargetJITInfo</tt> class exposes an abstract interface used by the 504 Just-In-Time code generator to perform target-specific activities, such as 505 emitting stubs. If a <tt>TargetMachine</tt> supports JIT code generation, it 506 should provide one of these objects through the <tt>getJITInfo</tt> 507 method.</p> 508 509</div> 510 511</div> 512 513<!-- *********************************************************************** --> 514<h2> 515 <a name="codegendesc">Machine code description classes</a> 516</h2> 517<!-- *********************************************************************** --> 518 519<div> 520 521<p>At the high-level, LLVM code is translated to a machine specific 522 representation formed out of 523 <a href="#machinefunction"><tt>MachineFunction</tt></a>, 524 <a href="#machinebasicblock"><tt>MachineBasicBlock</tt></a>, 525 and <a href="#machineinstr"><tt>MachineInstr</tt></a> instances (defined 526 in <tt>include/llvm/CodeGen</tt>). This representation is completely target 527 agnostic, representing instructions in their most abstract form: an opcode 528 and a series of operands. This representation is designed to support both an 529 SSA representation for machine code, as well as a register allocated, non-SSA 530 form.</p> 531 532<!-- ======================================================================= --> 533<h3> 534 <a name="machineinstr">The <tt>MachineInstr</tt> class</a> 535</h3> 536 537<div> 538 539<p>Target machine instructions are represented as instances of the 540 <tt>MachineInstr</tt> class. This class is an extremely abstract way of 541 representing machine instructions. In particular, it only keeps track of an 542 opcode number and a set of operands.</p> 543 544<p>The opcode number is a simple unsigned integer that only has meaning to a 545 specific backend. All of the instructions for a target should be defined in 546 the <tt>*InstrInfo.td</tt> file for the target. The opcode enum values are 547 auto-generated from this description. The <tt>MachineInstr</tt> class does 548 not have any information about how to interpret the instruction (i.e., what 549 the semantics of the instruction are); for that you must refer to the 550 <tt><a href="#targetinstrinfo">TargetInstrInfo</a></tt> class.</p> 551 552<p>The operands of a machine instruction can be of several different types: a 553 register reference, a constant integer, a basic block reference, etc. In 554 addition, a machine operand should be marked as a def or a use of the value 555 (though only registers are allowed to be defs).</p> 556 557<p>By convention, the LLVM code generator orders instruction operands so that 558 all register definitions come before the register uses, even on architectures 559 that are normally printed in other orders. For example, the SPARC add 560 instruction: "<tt>add %i1, %i2, %i3</tt>" adds the "%i1", and "%i2" registers 561 and stores the result into the "%i3" register. In the LLVM code generator, 562 the operands should be stored as "<tt>%i3, %i1, %i2</tt>": with the 563 destination first.</p> 564 565<p>Keeping destination (definition) operands at the beginning of the operand 566 list has several advantages. In particular, the debugging printer will print 567 the instruction like this:</p> 568 569<div class="doc_code"> 570<pre> 571%r3 = add %i1, %i2 572</pre> 573</div> 574 575<p>Also if the first operand is a def, it is easier to <a href="#buildmi">create 576 instructions</a> whose only def is the first operand.</p> 577 578<!-- _______________________________________________________________________ --> 579<h4> 580 <a name="buildmi">Using the <tt>MachineInstrBuilder.h</tt> functions</a> 581</h4> 582 583<div> 584 585<p>Machine instructions are created by using the <tt>BuildMI</tt> functions, 586 located in the <tt>include/llvm/CodeGen/MachineInstrBuilder.h</tt> file. The 587 <tt>BuildMI</tt> functions make it easy to build arbitrary machine 588 instructions. Usage of the <tt>BuildMI</tt> functions look like this:</p> 589 590<div class="doc_code"> 591<pre> 592// Create a 'DestReg = mov 42' (rendered in X86 assembly as 'mov DestReg, 42') 593// instruction. The '1' specifies how many operands will be added. 594MachineInstr *MI = BuildMI(X86::MOV32ri, 1, DestReg).addImm(42); 595 596// Create the same instr, but insert it at the end of a basic block. 597MachineBasicBlock &MBB = ... 598BuildMI(MBB, X86::MOV32ri, 1, DestReg).addImm(42); 599 600// Create the same instr, but insert it before a specified iterator point. 601MachineBasicBlock::iterator MBBI = ... 602BuildMI(MBB, MBBI, X86::MOV32ri, 1, DestReg).addImm(42); 603 604// Create a 'cmp Reg, 0' instruction, no destination reg. 605MI = BuildMI(X86::CMP32ri, 2).addReg(Reg).addImm(0); 606// Create an 'sahf' instruction which takes no operands and stores nothing. 607MI = BuildMI(X86::SAHF, 0); 608 609// Create a self looping branch instruction. 610BuildMI(MBB, X86::JNE, 1).addMBB(&MBB); 611</pre> 612</div> 613 614<p>The key thing to remember with the <tt>BuildMI</tt> functions is that you 615 have to specify the number of operands that the machine instruction will 616 take. This allows for efficient memory allocation. You also need to specify 617 if operands default to be uses of values, not definitions. If you need to 618 add a definition operand (other than the optional destination register), you 619 must explicitly mark it as such:</p> 620 621<div class="doc_code"> 622<pre> 623MI.addReg(Reg, RegState::Define); 624</pre> 625</div> 626 627</div> 628 629<!-- _______________________________________________________________________ --> 630<h4> 631 <a name="fixedregs">Fixed (preassigned) registers</a> 632</h4> 633 634<div> 635 636<p>One important issue that the code generator needs to be aware of is the 637 presence of fixed registers. In particular, there are often places in the 638 instruction stream where the register allocator <em>must</em> arrange for a 639 particular value to be in a particular register. This can occur due to 640 limitations of the instruction set (e.g., the X86 can only do a 32-bit divide 641 with the <tt>EAX</tt>/<tt>EDX</tt> registers), or external factors like 642 calling conventions. In any case, the instruction selector should emit code 643 that copies a virtual register into or out of a physical register when 644 needed.</p> 645 646<p>For example, consider this simple LLVM example:</p> 647 648<div class="doc_code"> 649<pre> 650define i32 @test(i32 %X, i32 %Y) { 651 %Z = udiv i32 %X, %Y 652 ret i32 %Z 653} 654</pre> 655</div> 656 657<p>The X86 instruction selector produces this machine code for the <tt>div</tt> 658 and <tt>ret</tt> (use "<tt>llc X.bc -march=x86 -print-machineinstrs</tt>" to 659 get this):</p> 660 661<div class="doc_code"> 662<pre> 663;; Start of div 664%EAX = mov %reg1024 ;; Copy X (in reg1024) into EAX 665%reg1027 = sar %reg1024, 31 666%EDX = mov %reg1027 ;; Sign extend X into EDX 667idiv %reg1025 ;; Divide by Y (in reg1025) 668%reg1026 = mov %EAX ;; Read the result (Z) out of EAX 669 670;; Start of ret 671%EAX = mov %reg1026 ;; 32-bit return value goes in EAX 672ret 673</pre> 674</div> 675 676<p>By the end of code generation, the register allocator has coalesced the 677 registers and deleted the resultant identity moves producing the following 678 code:</p> 679 680<div class="doc_code"> 681<pre> 682;; X is in EAX, Y is in ECX 683mov %EAX, %EDX 684sar %EDX, 31 685idiv %ECX 686ret 687</pre> 688</div> 689 690<p>This approach is extremely general (if it can handle the X86 architecture, it 691 can handle anything!) and allows all of the target specific knowledge about 692 the instruction stream to be isolated in the instruction selector. Note that 693 physical registers should have a short lifetime for good code generation, and 694 all physical registers are assumed dead on entry to and exit from basic 695 blocks (before register allocation). Thus, if you need a value to be live 696 across basic block boundaries, it <em>must</em> live in a virtual 697 register.</p> 698 699</div> 700 701<!-- _______________________________________________________________________ --> 702<h4> 703 <a name="ssa">Machine code in SSA form</a> 704</h4> 705 706<div> 707 708<p><tt>MachineInstr</tt>'s are initially selected in SSA-form, and are 709 maintained in SSA-form until register allocation happens. For the most part, 710 this is trivially simple since LLVM is already in SSA form; LLVM PHI nodes 711 become machine code PHI nodes, and virtual registers are only allowed to have 712 a single definition.</p> 713 714<p>After register allocation, machine code is no longer in SSA-form because 715 there are no virtual registers left in the code.</p> 716 717</div> 718 719</div> 720 721<!-- ======================================================================= --> 722<h3> 723 <a name="machinebasicblock">The <tt>MachineBasicBlock</tt> class</a> 724</h3> 725 726<div> 727 728<p>The <tt>MachineBasicBlock</tt> class contains a list of machine instructions 729 (<tt><a href="#machineinstr">MachineInstr</a></tt> instances). It roughly 730 corresponds to the LLVM code input to the instruction selector, but there can 731 be a one-to-many mapping (i.e. one LLVM basic block can map to multiple 732 machine basic blocks). The <tt>MachineBasicBlock</tt> class has a 733 "<tt>getBasicBlock</tt>" method, which returns the LLVM basic block that it 734 comes from.</p> 735 736</div> 737 738<!-- ======================================================================= --> 739<h3> 740 <a name="machinefunction">The <tt>MachineFunction</tt> class</a> 741</h3> 742 743<div> 744 745<p>The <tt>MachineFunction</tt> class contains a list of machine basic blocks 746 (<tt><a href="#machinebasicblock">MachineBasicBlock</a></tt> instances). It 747 corresponds one-to-one with the LLVM function input to the instruction 748 selector. In addition to a list of basic blocks, 749 the <tt>MachineFunction</tt> contains a a <tt>MachineConstantPool</tt>, 750 a <tt>MachineFrameInfo</tt>, a <tt>MachineFunctionInfo</tt>, and a 751 <tt>MachineRegisterInfo</tt>. See 752 <tt>include/llvm/CodeGen/MachineFunction.h</tt> for more information.</p> 753 754</div> 755 756</div> 757 758<!-- *********************************************************************** --> 759<h2> 760 <a name="mc">The "MC" Layer</a> 761</h2> 762<!-- *********************************************************************** --> 763 764<div> 765 766<p> 767The MC Layer is used to represent and process code at the raw machine code 768level, devoid of "high level" information like "constant pools", "jump tables", 769"global variables" or anything like that. At this level, LLVM handles things 770like label names, machine instructions, and sections in the object file. The 771code in this layer is used for a number of important purposes: the tail end of 772the code generator uses it to write a .s or .o file, and it is also used by the 773llvm-mc tool to implement standalone machine code assemblers and disassemblers. 774</p> 775 776<p> 777This section describes some of the important classes. There are also a number 778of important subsystems that interact at this layer, they are described later 779in this manual. 780</p> 781 782<!-- ======================================================================= --> 783<h3> 784 <a name="mcstreamer">The <tt>MCStreamer</tt> API</a> 785</h3> 786 787<div> 788 789<p> 790MCStreamer is best thought of as an assembler API. It is an abstract API which 791is <em>implemented</em> in different ways (e.g. to output a .s file, output an 792ELF .o file, etc) but whose API correspond directly to what you see in a .s 793file. MCStreamer has one method per directive, such as EmitLabel, 794EmitSymbolAttribute, SwitchSection, EmitValue (for .byte, .word), etc, which 795directly correspond to assembly level directives. It also has an 796EmitInstruction method, which is used to output an MCInst to the streamer. 797</p> 798 799<p> 800This API is most important for two clients: the llvm-mc stand-alone assembler is 801effectively a parser that parses a line, then invokes a method on MCStreamer. In 802the code generator, the <a href="#codeemit">Code Emission</a> phase of the code 803generator lowers higher level LLVM IR and Machine* constructs down to the MC 804layer, emitting directives through MCStreamer.</p> 805 806<p> 807On the implementation side of MCStreamer, there are two major implementations: 808one for writing out a .s file (MCAsmStreamer), and one for writing out a .o 809file (MCObjectStreamer). MCAsmStreamer is a straight-forward implementation 810that prints out a directive for each method (e.g. EmitValue -> .byte), but 811MCObjectStreamer implements a full assembler. 812</p> 813 814</div> 815 816<!-- ======================================================================= --> 817<h3> 818 <a name="mccontext">The <tt>MCContext</tt> class</a> 819</h3> 820 821<div> 822 823<p> 824The MCContext class is the owner of a variety of uniqued data structures at the 825MC layer, including symbols, sections, etc. As such, this is the class that you 826interact with to create symbols and sections. This class can not be subclassed. 827</p> 828 829</div> 830 831<!-- ======================================================================= --> 832<h3> 833 <a name="mcsymbol">The <tt>MCSymbol</tt> class</a> 834</h3> 835 836<div> 837 838<p> 839The MCSymbol class represents a symbol (aka label) in the assembly file. There 840are two interesting kinds of symbols: assembler temporary symbols, and normal 841symbols. Assembler temporary symbols are used and processed by the assembler 842but are discarded when the object file is produced. The distinction is usually 843represented by adding a prefix to the label, for example "L" labels are 844assembler temporary labels in MachO. 845</p> 846 847<p>MCSymbols are created by MCContext and uniqued there. This means that 848MCSymbols can be compared for pointer equivalence to find out if they are the 849same symbol. Note that pointer inequality does not guarantee the labels will 850end up at different addresses though. It's perfectly legal to output something 851like this to the .s file:<p> 852 853<pre> 854 foo: 855 bar: 856 .byte 4 857</pre> 858 859<p>In this case, both the foo and bar symbols will have the same address.</p> 860 861</div> 862 863<!-- ======================================================================= --> 864<h3> 865 <a name="mcsection">The <tt>MCSection</tt> class</a> 866</h3> 867 868<div> 869 870<p> 871The MCSection class represents an object-file specific section. It is subclassed 872by object file specific implementations (e.g. <tt>MCSectionMachO</tt>, 873<tt>MCSectionCOFF</tt>, <tt>MCSectionELF</tt>) and these are created and uniqued 874by MCContext. The MCStreamer has a notion of the current section, which can be 875changed with the SwitchToSection method (which corresponds to a ".section" 876directive in a .s file). 877</p> 878 879</div> 880 881<!-- ======================================================================= --> 882<h3> 883 <a name="mcinst">The <tt>MCInst</tt> class</a> 884</h3> 885 886<div> 887 888<p> 889The MCInst class is a target-independent representation of an instruction. It 890is a simple class (much more so than <a href="#machineinstr">MachineInstr</a>) 891that holds a target-specific opcode and a vector of MCOperands. MCOperand, in 892turn, is a simple discriminated union of three cases: 1) a simple immediate, 8932) a target register ID, 3) a symbolic expression (e.g. "Lfoo-Lbar+42") as an 894MCExpr. 895</p> 896 897<p>MCInst is the common currency used to represent machine instructions at the 898MC layer. It is the type used by the instruction encoder, the instruction 899printer, and the type generated by the assembly parser and disassembler. 900</p> 901 902</div> 903 904</div> 905 906<!-- *********************************************************************** --> 907<h2> 908 <a name="codegenalgs">Target-independent code generation algorithms</a> 909</h2> 910<!-- *********************************************************************** --> 911 912<div> 913 914<p>This section documents the phases described in the 915 <a href="#high-level-design">high-level design of the code generator</a>. 916 It explains how they work and some of the rationale behind their design.</p> 917 918<!-- ======================================================================= --> 919<h3> 920 <a name="instselect">Instruction Selection</a> 921</h3> 922 923<div> 924 925<p>Instruction Selection is the process of translating LLVM code presented to 926 the code generator into target-specific machine instructions. There are 927 several well-known ways to do this in the literature. LLVM uses a 928 SelectionDAG based instruction selector.</p> 929 930<p>Portions of the DAG instruction selector are generated from the target 931 description (<tt>*.td</tt>) files. Our goal is for the entire instruction 932 selector to be generated from these <tt>.td</tt> files, though currently 933 there are still things that require custom C++ code.</p> 934 935<!-- _______________________________________________________________________ --> 936<h4> 937 <a name="selectiondag_intro">Introduction to SelectionDAGs</a> 938</h4> 939 940<div> 941 942<p>The SelectionDAG provides an abstraction for code representation in a way 943 that is amenable to instruction selection using automatic techniques 944 (e.g. dynamic-programming based optimal pattern matching selectors). It is 945 also well-suited to other phases of code generation; in particular, 946 instruction scheduling (SelectionDAG's are very close to scheduling DAGs 947 post-selection). Additionally, the SelectionDAG provides a host 948 representation where a large variety of very-low-level (but 949 target-independent) <a href="#selectiondag_optimize">optimizations</a> may be 950 performed; ones which require extensive information about the instructions 951 efficiently supported by the target.</p> 952 953<p>The SelectionDAG is a Directed-Acyclic-Graph whose nodes are instances of the 954 <tt>SDNode</tt> class. The primary payload of the <tt>SDNode</tt> is its 955 operation code (Opcode) that indicates what operation the node performs and 956 the operands to the operation. The various operation node types are 957 described at the top of the <tt>include/llvm/CodeGen/SelectionDAGNodes.h</tt> 958 file.</p> 959 960<p>Although most operations define a single value, each node in the graph may 961 define multiple values. For example, a combined div/rem operation will 962 define both the dividend and the remainder. Many other situations require 963 multiple values as well. Each node also has some number of operands, which 964 are edges to the node defining the used value. Because nodes may define 965 multiple values, edges are represented by instances of the <tt>SDValue</tt> 966 class, which is a <tt><SDNode, unsigned></tt> pair, indicating the node 967 and result value being used, respectively. Each value produced by 968 an <tt>SDNode</tt> has an associated <tt>MVT</tt> (Machine Value Type) 969 indicating what the type of the value is.</p> 970 971<p>SelectionDAGs contain two different kinds of values: those that represent 972 data flow and those that represent control flow dependencies. Data values 973 are simple edges with an integer or floating point value type. Control edges 974 are represented as "chain" edges which are of type <tt>MVT::Other</tt>. 975 These edges provide an ordering between nodes that have side effects (such as 976 loads, stores, calls, returns, etc). All nodes that have side effects should 977 take a token chain as input and produce a new one as output. By convention, 978 token chain inputs are always operand #0, and chain results are always the 979 last value produced by an operation.</p> 980 981<p>A SelectionDAG has designated "Entry" and "Root" nodes. The Entry node is 982 always a marker node with an Opcode of <tt>ISD::EntryToken</tt>. The Root 983 node is the final side-effecting node in the token chain. For example, in a 984 single basic block function it would be the return node.</p> 985 986<p>One important concept for SelectionDAGs is the notion of a "legal" vs. 987 "illegal" DAG. A legal DAG for a target is one that only uses supported 988 operations and supported types. On a 32-bit PowerPC, for example, a DAG with 989 a value of type i1, i8, i16, or i64 would be illegal, as would a DAG that 990 uses a SREM or UREM operation. The 991 <a href="#selectinodag_legalize_types">legalize types</a> and 992 <a href="#selectiondag_legalize">legalize operations</a> phases are 993 responsible for turning an illegal DAG into a legal DAG.</p> 994 995</div> 996 997<!-- _______________________________________________________________________ --> 998<h4> 999 <a name="selectiondag_process">SelectionDAG Instruction Selection Process</a> 1000</h4> 1001 1002<div> 1003 1004<p>SelectionDAG-based instruction selection consists of the following steps:</p> 1005 1006<ol> 1007 <li><a href="#selectiondag_build">Build initial DAG</a> — This stage 1008 performs a simple translation from the input LLVM code to an illegal 1009 SelectionDAG.</li> 1010 1011 <li><a href="#selectiondag_optimize">Optimize SelectionDAG</a> — This 1012 stage performs simple optimizations on the SelectionDAG to simplify it, 1013 and recognize meta instructions (like rotates 1014 and <tt>div</tt>/<tt>rem</tt> pairs) for targets that support these meta 1015 operations. This makes the resultant code more efficient and 1016 the <a href="#selectiondag_select">select instructions from DAG</a> phase 1017 (below) simpler.</li> 1018 1019 <li><a href="#selectiondag_legalize_types">Legalize SelectionDAG Types</a> 1020 — This stage transforms SelectionDAG nodes to eliminate any types 1021 that are unsupported on the target.</li> 1022 1023 <li><a href="#selectiondag_optimize">Optimize SelectionDAG</a> — The 1024 SelectionDAG optimizer is run to clean up redundancies exposed by type 1025 legalization.</li> 1026 1027 <li><a href="#selectiondag_legalize">Legalize SelectionDAG Ops</a> — 1028 This stage transforms SelectionDAG nodes to eliminate any operations 1029 that are unsupported on the target.</li> 1030 1031 <li><a href="#selectiondag_optimize">Optimize SelectionDAG</a> — The 1032 SelectionDAG optimizer is run to eliminate inefficiencies introduced by 1033 operation legalization.</li> 1034 1035 <li><a href="#selectiondag_select">Select instructions from DAG</a> — 1036 Finally, the target instruction selector matches the DAG operations to 1037 target instructions. This process translates the target-independent input 1038 DAG into another DAG of target instructions.</li> 1039 1040 <li><a href="#selectiondag_sched">SelectionDAG Scheduling and Formation</a> 1041 — The last phase assigns a linear order to the instructions in the 1042 target-instruction DAG and emits them into the MachineFunction being 1043 compiled. This step uses traditional prepass scheduling techniques.</li> 1044</ol> 1045 1046<p>After all of these steps are complete, the SelectionDAG is destroyed and the 1047 rest of the code generation passes are run.</p> 1048 1049<p>One great way to visualize what is going on here is to take advantage of a 1050 few LLC command line options. The following options pop up a window 1051 displaying the SelectionDAG at specific times (if you only get errors printed 1052 to the console while using this, you probably 1053 <a href="ProgrammersManual.html#ViewGraph">need to configure your system</a> 1054 to add support for it).</p> 1055 1056<ul> 1057 <li><tt>-view-dag-combine1-dags</tt> displays the DAG after being built, 1058 before the first optimization pass.</li> 1059 1060 <li><tt>-view-legalize-dags</tt> displays the DAG before Legalization.</li> 1061 1062 <li><tt>-view-dag-combine2-dags</tt> displays the DAG before the second 1063 optimization pass.</li> 1064 1065 <li><tt>-view-isel-dags</tt> displays the DAG before the Select phase.</li> 1066 1067 <li><tt>-view-sched-dags</tt> displays the DAG before Scheduling.</li> 1068</ul> 1069 1070<p>The <tt>-view-sunit-dags</tt> displays the Scheduler's dependency graph. 1071 This graph is based on the final SelectionDAG, with nodes that must be 1072 scheduled together bundled into a single scheduling-unit node, and with 1073 immediate operands and other nodes that aren't relevant for scheduling 1074 omitted.</p> 1075 1076</div> 1077 1078<!-- _______________________________________________________________________ --> 1079<h4> 1080 <a name="selectiondag_build">Initial SelectionDAG Construction</a> 1081</h4> 1082 1083<div> 1084 1085<p>The initial SelectionDAG is naïvely peephole expanded from the LLVM 1086 input by the <tt>SelectionDAGLowering</tt> class in the 1087 <tt>lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp</tt> file. The intent of 1088 this pass is to expose as much low-level, target-specific details to the 1089 SelectionDAG as possible. This pass is mostly hard-coded (e.g. an 1090 LLVM <tt>add</tt> turns into an <tt>SDNode add</tt> while a 1091 <tt>getelementptr</tt> is expanded into the obvious arithmetic). This pass 1092 requires target-specific hooks to lower calls, returns, varargs, etc. For 1093 these features, the <tt><a href="#targetlowering">TargetLowering</a></tt> 1094 interface is used.</p> 1095 1096</div> 1097 1098<!-- _______________________________________________________________________ --> 1099<h4> 1100 <a name="selectiondag_legalize_types">SelectionDAG LegalizeTypes Phase</a> 1101</h4> 1102 1103<div> 1104 1105<p>The Legalize phase is in charge of converting a DAG to only use the types 1106 that are natively supported by the target.</p> 1107 1108<p>There are two main ways of converting values of unsupported scalar types to 1109 values of supported types: converting small types to larger types 1110 ("promoting"), and breaking up large integer types into smaller ones 1111 ("expanding"). For example, a target might require that all f32 values are 1112 promoted to f64 and that all i1/i8/i16 values are promoted to i32. The same 1113 target might require that all i64 values be expanded into pairs of i32 1114 values. These changes can insert sign and zero extensions as needed to make 1115 sure that the final code has the same behavior as the input.</p> 1116 1117<p>There are two main ways of converting values of unsupported vector types to 1118 value of supported types: splitting vector types, multiple times if 1119 necessary, until a legal type is found, and extending vector types by adding 1120 elements to the end to round them out to legal types ("widening"). If a 1121 vector gets split all the way down to single-element parts with no supported 1122 vector type being found, the elements are converted to scalars 1123 ("scalarizing").</p> 1124 1125<p>A target implementation tells the legalizer which types are supported (and 1126 which register class to use for them) by calling the 1127 <tt>addRegisterClass</tt> method in its TargetLowering constructor.</p> 1128 1129</div> 1130 1131<!-- _______________________________________________________________________ --> 1132<h4> 1133 <a name="selectiondag_legalize">SelectionDAG Legalize Phase</a> 1134</h4> 1135 1136<div> 1137 1138<p>The Legalize phase is in charge of converting a DAG to only use the 1139 operations that are natively supported by the target.</p> 1140 1141<p>Targets often have weird constraints, such as not supporting every operation 1142 on every supported datatype (e.g. X86 does not support byte conditional moves 1143 and PowerPC does not support sign-extending loads from a 16-bit memory 1144 location). Legalize takes care of this by open-coding another sequence of 1145 operations to emulate the operation ("expansion"), by promoting one type to a 1146 larger type that supports the operation ("promotion"), or by using a 1147 target-specific hook to implement the legalization ("custom").</p> 1148 1149<p>A target implementation tells the legalizer which operations are not 1150 supported (and which of the above three actions to take) by calling the 1151 <tt>setOperationAction</tt> method in its <tt>TargetLowering</tt> 1152 constructor.</p> 1153 1154<p>Prior to the existence of the Legalize passes, we required that every target 1155 <a href="#selectiondag_optimize">selector</a> supported and handled every 1156 operator and type even if they are not natively supported. The introduction 1157 of the Legalize phases allows all of the canonicalization patterns to be 1158 shared across targets, and makes it very easy to optimize the canonicalized 1159 code because it is still in the form of a DAG.</p> 1160 1161</div> 1162 1163<!-- _______________________________________________________________________ --> 1164<h4> 1165 <a name="selectiondag_optimize"> 1166 SelectionDAG Optimization Phase: the DAG Combiner 1167 </a> 1168</h4> 1169 1170<div> 1171 1172<p>The SelectionDAG optimization phase is run multiple times for code 1173 generation, immediately after the DAG is built and once after each 1174 legalization. The first run of the pass allows the initial code to be 1175 cleaned up (e.g. performing optimizations that depend on knowing that the 1176 operators have restricted type inputs). Subsequent runs of the pass clean up 1177 the messy code generated by the Legalize passes, which allows Legalize to be 1178 very simple (it can focus on making code legal instead of focusing on 1179 generating <em>good</em> and legal code).</p> 1180 1181<p>One important class of optimizations performed is optimizing inserted sign 1182 and zero extension instructions. We currently use ad-hoc techniques, but 1183 could move to more rigorous techniques in the future. Here are some good 1184 papers on the subject:</p> 1185 1186<p>"<a href="http://www.eecs.harvard.edu/~nr/pubs/widen-abstract.html">Widening 1187 integer arithmetic</a>"<br> 1188 Kevin Redwine and Norman Ramsey<br> 1189 International Conference on Compiler Construction (CC) 2004</p> 1190 1191<p>"<a href="http://portal.acm.org/citation.cfm?doid=512529.512552">Effective 1192 sign extension elimination</a>"<br> 1193 Motohiro Kawahito, Hideaki Komatsu, and Toshio Nakatani<br> 1194 Proceedings of the ACM SIGPLAN 2002 Conference on Programming Language Design 1195 and Implementation.</p> 1196 1197</div> 1198 1199<!-- _______________________________________________________________________ --> 1200<h4> 1201 <a name="selectiondag_select">SelectionDAG Select Phase</a> 1202</h4> 1203 1204<div> 1205 1206<p>The Select phase is the bulk of the target-specific code for instruction 1207 selection. This phase takes a legal SelectionDAG as input, pattern matches 1208 the instructions supported by the target to this DAG, and produces a new DAG 1209 of target code. For example, consider the following LLVM fragment:</p> 1210 1211<div class="doc_code"> 1212<pre> 1213%t1 = fadd float %W, %X 1214%t2 = fmul float %t1, %Y 1215%t3 = fadd float %t2, %Z 1216</pre> 1217</div> 1218 1219<p>This LLVM code corresponds to a SelectionDAG that looks basically like 1220 this:</p> 1221 1222<div class="doc_code"> 1223<pre> 1224(fadd:f32 (fmul:f32 (fadd:f32 W, X), Y), Z) 1225</pre> 1226</div> 1227 1228<p>If a target supports floating point multiply-and-add (FMA) operations, one of 1229 the adds can be merged with the multiply. On the PowerPC, for example, the 1230 output of the instruction selector might look like this DAG:</p> 1231 1232<div class="doc_code"> 1233<pre> 1234(FMADDS (FADDS W, X), Y, Z) 1235</pre> 1236</div> 1237 1238<p>The <tt>FMADDS</tt> instruction is a ternary instruction that multiplies its 1239first two operands and adds the third (as single-precision floating-point 1240numbers). The <tt>FADDS</tt> instruction is a simple binary single-precision 1241add instruction. To perform this pattern match, the PowerPC backend includes 1242the following instruction definitions:</p> 1243 1244<div class="doc_code"> 1245<pre> 1246def FMADDS : AForm_1<59, 29, 1247 (ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRC, F4RC:$FRB), 1248 "fmadds $FRT, $FRA, $FRC, $FRB", 1249 [<b>(set F4RC:$FRT, (fadd (fmul F4RC:$FRA, F4RC:$FRC), 1250 F4RC:$FRB))</b>]>; 1251def FADDS : AForm_2<59, 21, 1252 (ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRB), 1253 "fadds $FRT, $FRA, $FRB", 1254 [<b>(set F4RC:$FRT, (fadd F4RC:$FRA, F4RC:$FRB))</b>]>; 1255</pre> 1256</div> 1257 1258<p>The portion of the instruction definition in bold indicates the pattern used 1259 to match the instruction. The DAG operators 1260 (like <tt>fmul</tt>/<tt>fadd</tt>) are defined in 1261 the <tt>include/llvm/Target/TargetSelectionDAG.td</tt> file. " 1262 <tt>F4RC</tt>" is the register class of the input and result values.</p> 1263 1264<p>The TableGen DAG instruction selector generator reads the instruction 1265 patterns in the <tt>.td</tt> file and automatically builds parts of the 1266 pattern matching code for your target. It has the following strengths:</p> 1267 1268<ul> 1269 <li>At compiler-compiler time, it analyzes your instruction patterns and tells 1270 you if your patterns make sense or not.</li> 1271 1272 <li>It can handle arbitrary constraints on operands for the pattern match. In 1273 particular, it is straight-forward to say things like "match any immediate 1274 that is a 13-bit sign-extended value". For examples, see the 1275 <tt>immSExt16</tt> and related <tt>tblgen</tt> classes in the PowerPC 1276 backend.</li> 1277 1278 <li>It knows several important identities for the patterns defined. For 1279 example, it knows that addition is commutative, so it allows the 1280 <tt>FMADDS</tt> pattern above to match "<tt>(fadd X, (fmul Y, Z))</tt>" as 1281 well as "<tt>(fadd (fmul X, Y), Z)</tt>", without the target author having 1282 to specially handle this case.</li> 1283 1284 <li>It has a full-featured type-inferencing system. In particular, you should 1285 rarely have to explicitly tell the system what type parts of your patterns 1286 are. In the <tt>FMADDS</tt> case above, we didn't have to tell 1287 <tt>tblgen</tt> that all of the nodes in the pattern are of type 'f32'. 1288 It was able to infer and propagate this knowledge from the fact that 1289 <tt>F4RC</tt> has type 'f32'.</li> 1290 1291 <li>Targets can define their own (and rely on built-in) "pattern fragments". 1292 Pattern fragments are chunks of reusable patterns that get inlined into 1293 your patterns during compiler-compiler time. For example, the integer 1294 "<tt>(not x)</tt>" operation is actually defined as a pattern fragment 1295 that expands as "<tt>(xor x, -1)</tt>", since the SelectionDAG does not 1296 have a native '<tt>not</tt>' operation. Targets can define their own 1297 short-hand fragments as they see fit. See the definition of 1298 '<tt>not</tt>' and '<tt>ineg</tt>' for examples.</li> 1299 1300 <li>In addition to instructions, targets can specify arbitrary patterns that 1301 map to one or more instructions using the 'Pat' class. For example, the 1302 PowerPC has no way to load an arbitrary integer immediate into a register 1303 in one instruction. To tell tblgen how to do this, it defines: 1304 <br> 1305 <br> 1306<div class="doc_code"> 1307<pre> 1308// Arbitrary immediate support. Implement in terms of LIS/ORI. 1309def : Pat<(i32 imm:$imm), 1310 (ORI (LIS (HI16 imm:$imm)), (LO16 imm:$imm))>; 1311</pre> 1312</div> 1313 <br> 1314 If none of the single-instruction patterns for loading an immediate into a 1315 register match, this will be used. This rule says "match an arbitrary i32 1316 immediate, turning it into an <tt>ORI</tt> ('or a 16-bit immediate') and 1317 an <tt>LIS</tt> ('load 16-bit immediate, where the immediate is shifted to 1318 the left 16 bits') instruction". To make this work, the 1319 <tt>LO16</tt>/<tt>HI16</tt> node transformations are used to manipulate 1320 the input immediate (in this case, take the high or low 16-bits of the 1321 immediate).</li> 1322 1323 <li>While the system does automate a lot, it still allows you to write custom 1324 C++ code to match special cases if there is something that is hard to 1325 express.</li> 1326</ul> 1327 1328<p>While it has many strengths, the system currently has some limitations, 1329 primarily because it is a work in progress and is not yet finished:</p> 1330 1331<ul> 1332 <li>Overall, there is no way to define or match SelectionDAG nodes that define 1333 multiple values (e.g. <tt>SMUL_LOHI</tt>, <tt>LOAD</tt>, <tt>CALL</tt>, 1334 etc). This is the biggest reason that you currently still <em>have 1335 to</em> write custom C++ code for your instruction selector.</li> 1336 1337 <li>There is no great way to support matching complex addressing modes yet. 1338 In the future, we will extend pattern fragments to allow them to define 1339 multiple values (e.g. the four operands of the <a href="#x86_memory">X86 1340 addressing mode</a>, which are currently matched with custom C++ code). 1341 In addition, we'll extend fragments so that a fragment can match multiple 1342 different patterns.</li> 1343 1344 <li>We don't automatically infer flags like isStore/isLoad yet.</li> 1345 1346 <li>We don't automatically generate the set of supported registers and 1347 operations for the <a href="#selectiondag_legalize">Legalizer</a> 1348 yet.</li> 1349 1350 <li>We don't have a way of tying in custom legalized nodes yet.</li> 1351</ul> 1352 1353<p>Despite these limitations, the instruction selector generator is still quite 1354 useful for most of the binary and logical operations in typical instruction 1355 sets. If you run into any problems or can't figure out how to do something, 1356 please let Chris know!</p> 1357 1358</div> 1359 1360<!-- _______________________________________________________________________ --> 1361<h4> 1362 <a name="selectiondag_sched">SelectionDAG Scheduling and Formation Phase</a> 1363</h4> 1364 1365<div> 1366 1367<p>The scheduling phase takes the DAG of target instructions from the selection 1368 phase and assigns an order. The scheduler can pick an order depending on 1369 various constraints of the machines (i.e. order for minimal register pressure 1370 or try to cover instruction latencies). Once an order is established, the 1371 DAG is converted to a list 1372 of <tt><a href="#machineinstr">MachineInstr</a></tt>s and the SelectionDAG is 1373 destroyed.</p> 1374 1375<p>Note that this phase is logically separate from the instruction selection 1376 phase, but is tied to it closely in the code because it operates on 1377 SelectionDAGs.</p> 1378 1379</div> 1380 1381<!-- _______________________________________________________________________ --> 1382<h4> 1383 <a name="selectiondag_future">Future directions for the SelectionDAG</a> 1384</h4> 1385 1386<div> 1387 1388<ol> 1389 <li>Optional function-at-a-time selection.</li> 1390 1391 <li>Auto-generate entire selector from <tt>.td</tt> file.</li> 1392</ol> 1393 1394</div> 1395 1396</div> 1397 1398<!-- ======================================================================= --> 1399<h3> 1400 <a name="ssamco">SSA-based Machine Code Optimizations</a> 1401</h3> 1402<div><p>To Be Written</p></div> 1403 1404<!-- ======================================================================= --> 1405<h3> 1406 <a name="liveintervals">Live Intervals</a> 1407</h3> 1408 1409<div> 1410 1411<p>Live Intervals are the ranges (intervals) where a variable is <i>live</i>. 1412 They are used by some <a href="#regalloc">register allocator</a> passes to 1413 determine if two or more virtual registers which require the same physical 1414 register are live at the same point in the program (i.e., they conflict). 1415 When this situation occurs, one virtual register must be <i>spilled</i>.</p> 1416 1417<!-- _______________________________________________________________________ --> 1418<h4> 1419 <a name="livevariable_analysis">Live Variable Analysis</a> 1420</h4> 1421 1422<div> 1423 1424<p>The first step in determining the live intervals of variables is to calculate 1425 the set of registers that are immediately dead after the instruction (i.e., 1426 the instruction calculates the value, but it is never used) and the set of 1427 registers that are used by the instruction, but are never used after the 1428 instruction (i.e., they are killed). Live variable information is computed 1429 for each <i>virtual</i> register and <i>register allocatable</i> physical 1430 register in the function. This is done in a very efficient manner because it 1431 uses SSA to sparsely compute lifetime information for virtual registers 1432 (which are in SSA form) and only has to track physical registers within a 1433 block. Before register allocation, LLVM can assume that physical registers 1434 are only live within a single basic block. This allows it to do a single, 1435 local analysis to resolve physical register lifetimes within each basic 1436 block. If a physical register is not register allocatable (e.g., a stack 1437 pointer or condition codes), it is not tracked.</p> 1438 1439<p>Physical registers may be live in to or out of a function. Live in values are 1440 typically arguments in registers. Live out values are typically return values 1441 in registers. Live in values are marked as such, and are given a dummy 1442 "defining" instruction during live intervals analysis. If the last basic 1443 block of a function is a <tt>return</tt>, then it's marked as using all live 1444 out values in the function.</p> 1445 1446<p><tt>PHI</tt> nodes need to be handled specially, because the calculation of 1447 the live variable information from a depth first traversal of the CFG of the 1448 function won't guarantee that a virtual register used by the <tt>PHI</tt> 1449 node is defined before it's used. When a <tt>PHI</tt> node is encountered, 1450 only the definition is handled, because the uses will be handled in other 1451 basic blocks.</p> 1452 1453<p>For each <tt>PHI</tt> node of the current basic block, we simulate an 1454 assignment at the end of the current basic block and traverse the successor 1455 basic blocks. If a successor basic block has a <tt>PHI</tt> node and one of 1456 the <tt>PHI</tt> node's operands is coming from the current basic block, then 1457 the variable is marked as <i>alive</i> within the current basic block and all 1458 of its predecessor basic blocks, until the basic block with the defining 1459 instruction is encountered.</p> 1460 1461</div> 1462 1463<!-- _______________________________________________________________________ --> 1464<h4> 1465 <a name="liveintervals_analysis">Live Intervals Analysis</a> 1466</h4> 1467 1468<div> 1469 1470<p>We now have the information available to perform the live intervals analysis 1471 and build the live intervals themselves. We start off by numbering the basic 1472 blocks and machine instructions. We then handle the "live-in" values. These 1473 are in physical registers, so the physical register is assumed to be killed 1474 by the end of the basic block. Live intervals for virtual registers are 1475 computed for some ordering of the machine instructions <tt>[1, N]</tt>. A 1476 live interval is an interval <tt>[i, j)</tt>, where <tt>1 <= i <= j 1477 < N</tt>, for which a variable is live.</p> 1478 1479<p><i><b>More to come...</b></i></p> 1480 1481</div> 1482 1483</div> 1484 1485<!-- ======================================================================= --> 1486<h3> 1487 <a name="regalloc">Register Allocation</a> 1488</h3> 1489 1490<div> 1491 1492<p>The <i>Register Allocation problem</i> consists in mapping a program 1493 <i>P<sub>v</sub></i>, that can use an unbounded number of virtual registers, 1494 to a program <i>P<sub>p</sub></i> that contains a finite (possibly small) 1495 number of physical registers. Each target architecture has a different number 1496 of physical registers. If the number of physical registers is not enough to 1497 accommodate all the virtual registers, some of them will have to be mapped 1498 into memory. These virtuals are called <i>spilled virtuals</i>.</p> 1499 1500<!-- _______________________________________________________________________ --> 1501 1502<h4> 1503 <a name="regAlloc_represent">How registers are represented in LLVM</a> 1504</h4> 1505 1506<div> 1507 1508<p>In LLVM, physical registers are denoted by integer numbers that normally 1509 range from 1 to 1023. To see how this numbering is defined for a particular 1510 architecture, you can read the <tt>GenRegisterNames.inc</tt> file for that 1511 architecture. For instance, by 1512 inspecting <tt>lib/Target/X86/X86GenRegisterNames.inc</tt> we see that the 1513 32-bit register <tt>EAX</tt> is denoted by 15, and the MMX register 1514 <tt>MM0</tt> is mapped to 48.</p> 1515 1516<p>Some architectures contain registers that share the same physical location. A 1517 notable example is the X86 platform. For instance, in the X86 architecture, 1518 the registers <tt>EAX</tt>, <tt>AX</tt> and <tt>AL</tt> share the first eight 1519 bits. These physical registers are marked as <i>aliased</i> in LLVM. Given a 1520 particular architecture, you can check which registers are aliased by 1521 inspecting its <tt>RegisterInfo.td</tt> file. Moreover, the method 1522 <tt>TargetRegisterInfo::getAliasSet(p_reg)</tt> returns an array containing 1523 all the physical registers aliased to the register <tt>p_reg</tt>.</p> 1524 1525<p>Physical registers, in LLVM, are grouped in <i>Register Classes</i>. 1526 Elements in the same register class are functionally equivalent, and can be 1527 interchangeably used. Each virtual register can only be mapped to physical 1528 registers of a particular class. For instance, in the X86 architecture, some 1529 virtuals can only be allocated to 8 bit registers. A register class is 1530 described by <tt>TargetRegisterClass</tt> objects. To discover if a virtual 1531 register is compatible with a given physical, this code can be used:</p> 1532 1533<div class="doc_code"> 1534<pre> 1535bool RegMapping_Fer::compatible_class(MachineFunction &mf, 1536 unsigned v_reg, 1537 unsigned p_reg) { 1538 assert(TargetRegisterInfo::isPhysicalRegister(p_reg) && 1539 "Target register must be physical"); 1540 const TargetRegisterClass *trc = mf.getRegInfo().getRegClass(v_reg); 1541 return trc->contains(p_reg); 1542} 1543</pre> 1544</div> 1545 1546<p>Sometimes, mostly for debugging purposes, it is useful to change the number 1547 of physical registers available in the target architecture. This must be done 1548 statically, inside the <tt>TargetRegsterInfo.td</tt> file. Just <tt>grep</tt> 1549 for <tt>RegisterClass</tt>, the last parameter of which is a list of 1550 registers. Just commenting some out is one simple way to avoid them being 1551 used. A more polite way is to explicitly exclude some registers from 1552 the <i>allocation order</i>. See the definition of the <tt>GR8</tt> register 1553 class in <tt>lib/Target/X86/X86RegisterInfo.td</tt> for an example of this. 1554 </p> 1555 1556<p>Virtual registers are also denoted by integer numbers. Contrary to physical 1557 registers, different virtual registers never share the same number. Whereas 1558 physical registers are statically defined in a <tt>TargetRegisterInfo.td</tt> 1559 file and cannot be created by the application developer, that is not the case 1560 with virtual registers. In order to create new virtual registers, use the 1561 method <tt>MachineRegisterInfo::createVirtualRegister()</tt>. This method 1562 will return a new virtual register. Use an <tt>IndexedMap<Foo, 1563 VirtReg2IndexFunctor></tt> to hold information per virtual register. If you 1564 need to enumerate all virtual registers, use the function 1565 <tt>TargetRegisterInfo::index2VirtReg()</tt> to find the virtual register 1566 numbers:</p> 1567 1568<div class="doc_code"> 1569<pre> 1570 for (unsigned i = 0, e = MRI->getNumVirtRegs(); i != e; ++i) { 1571 unsigned VirtReg = TargetRegisterInfo::index2VirtReg(i); 1572 stuff(VirtReg); 1573 } 1574</pre> 1575</div> 1576 1577<p>Before register allocation, the operands of an instruction are mostly virtual 1578 registers, although physical registers may also be used. In order to check if 1579 a given machine operand is a register, use the boolean 1580 function <tt>MachineOperand::isRegister()</tt>. To obtain the integer code of 1581 a register, use <tt>MachineOperand::getReg()</tt>. An instruction may define 1582 or use a register. For instance, <tt>ADD reg:1026 := reg:1025 reg:1024</tt> 1583 defines the registers 1024, and uses registers 1025 and 1026. Given a 1584 register operand, the method <tt>MachineOperand::isUse()</tt> informs if that 1585 register is being used by the instruction. The 1586 method <tt>MachineOperand::isDef()</tt> informs if that registers is being 1587 defined.</p> 1588 1589<p>We will call physical registers present in the LLVM bitcode before register 1590 allocation <i>pre-colored registers</i>. Pre-colored registers are used in 1591 many different situations, for instance, to pass parameters of functions 1592 calls, and to store results of particular instructions. There are two types 1593 of pre-colored registers: the ones <i>implicitly</i> defined, and 1594 those <i>explicitly</i> defined. Explicitly defined registers are normal 1595 operands, and can be accessed 1596 with <tt>MachineInstr::getOperand(int)::getReg()</tt>. In order to check 1597 which registers are implicitly defined by an instruction, use 1598 the <tt>TargetInstrInfo::get(opcode)::ImplicitDefs</tt>, 1599 where <tt>opcode</tt> is the opcode of the target instruction. One important 1600 difference between explicit and implicit physical registers is that the 1601 latter are defined statically for each instruction, whereas the former may 1602 vary depending on the program being compiled. For example, an instruction 1603 that represents a function call will always implicitly define or use the same 1604 set of physical registers. To read the registers implicitly used by an 1605 instruction, 1606 use <tt>TargetInstrInfo::get(opcode)::ImplicitUses</tt>. Pre-colored 1607 registers impose constraints on any register allocation algorithm. The 1608 register allocator must make sure that none of them are overwritten by 1609 the values of virtual registers while still alive.</p> 1610 1611</div> 1612 1613<!-- _______________________________________________________________________ --> 1614 1615<h4> 1616 <a name="regAlloc_howTo">Mapping virtual registers to physical registers</a> 1617</h4> 1618 1619<div> 1620 1621<p>There are two ways to map virtual registers to physical registers (or to 1622 memory slots). The first way, that we will call <i>direct mapping</i>, is 1623 based on the use of methods of the classes <tt>TargetRegisterInfo</tt>, 1624 and <tt>MachineOperand</tt>. The second way, that we will call <i>indirect 1625 mapping</i>, relies on the <tt>VirtRegMap</tt> class in order to insert loads 1626 and stores sending and getting values to and from memory.</p> 1627 1628<p>The direct mapping provides more flexibility to the developer of the register 1629 allocator; however, it is more error prone, and demands more implementation 1630 work. Basically, the programmer will have to specify where load and store 1631 instructions should be inserted in the target function being compiled in 1632 order to get and store values in memory. To assign a physical register to a 1633 virtual register present in a given operand, 1634 use <tt>MachineOperand::setReg(p_reg)</tt>. To insert a store instruction, 1635 use <tt>TargetInstrInfo::storeRegToStackSlot(...)</tt>, and to insert a 1636 load instruction, use <tt>TargetInstrInfo::loadRegFromStackSlot</tt>.</p> 1637 1638<p>The indirect mapping shields the application developer from the complexities 1639 of inserting load and store instructions. In order to map a virtual register 1640 to a physical one, use <tt>VirtRegMap::assignVirt2Phys(vreg, preg)</tt>. In 1641 order to map a certain virtual register to memory, 1642 use <tt>VirtRegMap::assignVirt2StackSlot(vreg)</tt>. This method will return 1643 the stack slot where <tt>vreg</tt>'s value will be located. If it is 1644 necessary to map another virtual register to the same stack slot, 1645 use <tt>VirtRegMap::assignVirt2StackSlot(vreg, stack_location)</tt>. One 1646 important point to consider when using the indirect mapping, is that even if 1647 a virtual register is mapped to memory, it still needs to be mapped to a 1648 physical register. This physical register is the location where the virtual 1649 register is supposed to be found before being stored or after being 1650 reloaded.</p> 1651 1652<p>If the indirect strategy is used, after all the virtual registers have been 1653 mapped to physical registers or stack slots, it is necessary to use a spiller 1654 object to place load and store instructions in the code. Every virtual that 1655 has been mapped to a stack slot will be stored to memory after been defined 1656 and will be loaded before being used. The implementation of the spiller tries 1657 to recycle load/store instructions, avoiding unnecessary instructions. For an 1658 example of how to invoke the spiller, 1659 see <tt>RegAllocLinearScan::runOnMachineFunction</tt> 1660 in <tt>lib/CodeGen/RegAllocLinearScan.cpp</tt>.</p> 1661 1662</div> 1663 1664<!-- _______________________________________________________________________ --> 1665<h4> 1666 <a name="regAlloc_twoAddr">Handling two address instructions</a> 1667</h4> 1668 1669<div> 1670 1671<p>With very rare exceptions (e.g., function calls), the LLVM machine code 1672 instructions are three address instructions. That is, each instruction is 1673 expected to define at most one register, and to use at most two registers. 1674 However, some architectures use two address instructions. In this case, the 1675 defined register is also one of the used register. For instance, an 1676 instruction such as <tt>ADD %EAX, %EBX</tt>, in X86 is actually equivalent 1677 to <tt>%EAX = %EAX + %EBX</tt>.</p> 1678 1679<p>In order to produce correct code, LLVM must convert three address 1680 instructions that represent two address instructions into true two address 1681 instructions. LLVM provides the pass <tt>TwoAddressInstructionPass</tt> for 1682 this specific purpose. It must be run before register allocation takes 1683 place. After its execution, the resulting code may no longer be in SSA 1684 form. This happens, for instance, in situations where an instruction such 1685 as <tt>%a = ADD %b %c</tt> is converted to two instructions such as:</p> 1686 1687<div class="doc_code"> 1688<pre> 1689%a = MOVE %b 1690%a = ADD %a %c 1691</pre> 1692</div> 1693 1694<p>Notice that, internally, the second instruction is represented as 1695 <tt>ADD %a[def/use] %c</tt>. I.e., the register operand <tt>%a</tt> is both 1696 used and defined by the instruction.</p> 1697 1698</div> 1699 1700<!-- _______________________________________________________________________ --> 1701<h4> 1702 <a name="regAlloc_ssaDecon">The SSA deconstruction phase</a> 1703</h4> 1704 1705<div> 1706 1707<p>An important transformation that happens during register allocation is called 1708 the <i>SSA Deconstruction Phase</i>. The SSA form simplifies many analyses 1709 that are performed on the control flow graph of programs. However, 1710 traditional instruction sets do not implement PHI instructions. Thus, in 1711 order to generate executable code, compilers must replace PHI instructions 1712 with other instructions that preserve their semantics.</p> 1713 1714<p>There are many ways in which PHI instructions can safely be removed from the 1715 target code. The most traditional PHI deconstruction algorithm replaces PHI 1716 instructions with copy instructions. That is the strategy adopted by 1717 LLVM. The SSA deconstruction algorithm is implemented 1718 in <tt>lib/CodeGen/PHIElimination.cpp</tt>. In order to invoke this pass, the 1719 identifier <tt>PHIEliminationID</tt> must be marked as required in the code 1720 of the register allocator.</p> 1721 1722</div> 1723 1724<!-- _______________________________________________________________________ --> 1725<h4> 1726 <a name="regAlloc_fold">Instruction folding</a> 1727</h4> 1728 1729<div> 1730 1731<p><i>Instruction folding</i> is an optimization performed during register 1732 allocation that removes unnecessary copy instructions. For instance, a 1733 sequence of instructions such as:</p> 1734 1735<div class="doc_code"> 1736<pre> 1737%EBX = LOAD %mem_address 1738%EAX = COPY %EBX 1739</pre> 1740</div> 1741 1742<p>can be safely substituted by the single instruction:</p> 1743 1744<div class="doc_code"> 1745<pre> 1746%EAX = LOAD %mem_address 1747</pre> 1748</div> 1749 1750<p>Instructions can be folded with 1751 the <tt>TargetRegisterInfo::foldMemoryOperand(...)</tt> method. Care must be 1752 taken when folding instructions; a folded instruction can be quite different 1753 from the original 1754 instruction. See <tt>LiveIntervals::addIntervalsForSpills</tt> 1755 in <tt>lib/CodeGen/LiveIntervalAnalysis.cpp</tt> for an example of its 1756 use.</p> 1757 1758</div> 1759 1760<!-- _______________________________________________________________________ --> 1761 1762<h4> 1763 <a name="regAlloc_builtIn">Built in register allocators</a> 1764</h4> 1765 1766<div> 1767 1768<p>The LLVM infrastructure provides the application developer with three 1769 different register allocators:</p> 1770 1771<ul> 1772 <li><i>Fast</i> — This register allocator is the default for debug 1773 builds. It allocates registers on a basic block level, attempting to keep 1774 values in registers and reusing registers as appropriate.</li> 1775 1776 <li><i>Basic</i> — This is an incremental approach to register 1777 allocation. Live ranges are assigned to registers one at a time in 1778 an order that is driven by heuristics. Since code can be rewritten 1779 on-the-fly during allocation, this framework allows interesting 1780 allocators to be developed as extensions. It is not itself a 1781 production register allocator but is a potentially useful 1782 stand-alone mode for triaging bugs and as a performance baseline. 1783 1784 <li><i>Greedy</i> — <i>The default allocator</i>. This is a 1785 highly tuned implementation of the <i>Basic</i> allocator that 1786 incorporates global live range splitting. This allocator works hard 1787 to minimize the cost of spill code. 1788 1789 <li><i>PBQP</i> — A Partitioned Boolean Quadratic Programming (PBQP) 1790 based register allocator. This allocator works by constructing a PBQP 1791 problem representing the register allocation problem under consideration, 1792 solving this using a PBQP solver, and mapping the solution back to a 1793 register assignment.</li> 1794</ul> 1795 1796<p>The type of register allocator used in <tt>llc</tt> can be chosen with the 1797 command line option <tt>-regalloc=...</tt>:</p> 1798 1799<div class="doc_code"> 1800<pre> 1801$ llc -regalloc=linearscan file.bc -o ln.s; 1802$ llc -regalloc=fast file.bc -o fa.s; 1803$ llc -regalloc=pbqp file.bc -o pbqp.s; 1804</pre> 1805</div> 1806 1807</div> 1808 1809</div> 1810 1811<!-- ======================================================================= --> 1812<h3> 1813 <a name="proepicode">Prolog/Epilog Code Insertion</a> 1814</h3> 1815 1816<!-- _______________________________________________________________________ --> 1817<h4> 1818 <a name="compact_unwind">Compact Unwind</a> 1819</h4> 1820 1821<div> 1822 1823<p>Throwing an exception requires <em>unwinding</em> out of a function. The 1824 information on how to unwind a given function is traditionally expressed in 1825 DWARF unwind (a.k.a. frame) info. But that format was originally developed 1826 for debuggers to backtrace, and each Frame Description Entry (FDE) requires 1827 ~20-30 bytes per function. There is also the cost of mapping from an address 1828 in a function to the corresponding FDE at runtime. An alternative unwind 1829 encoding is called <em>compact unwind</em> and requires just 4-bytes per 1830 function.</p> 1831 1832<p>The compact unwind encoding is a 32-bit value, which is encoded in an 1833 architecture-specific way. It specifies which registers to restore and from 1834 where, and how to unwind out of the function. When the linker creates a final 1835 linked image, it will create a <code>__TEXT,__unwind_info</code> 1836 section. This section is a small and fast way for the runtime to access 1837 unwind info for any given function. If we emit compact unwind info for the 1838 function, that compact unwind info will be encoded in 1839 the <code>__TEXT,__unwind_info</code> section. If we emit DWARF unwind info, 1840 the <code>__TEXT,__unwind_info</code> section will contain the offset of the 1841 FDE in the <code>__TEXT,__eh_frame</code> section in the final linked 1842 image.</p> 1843 1844<p>For X86, there are three modes for the compact unwind encoding:</p> 1845 1846<dl> 1847 <dt><i>Function with a Frame Pointer (<code>EBP</code> or <code>RBP</code>)</i></dt> 1848 <dd><p><code>EBP/RBP</code>-based frame, where <code>EBP/RBP</code> is pushed 1849 onto the stack immediately after the return address, 1850 then <code>ESP/RSP</code> is moved to <code>EBP/RBP</code>. Thus to 1851 unwind, <code>ESP/RSP</code> is restored with the 1852 current <code>EBP/RBP</code> value, then <code>EBP/RBP</code> is restored 1853 by popping the stack, and the return is done by popping the stack once 1854 more into the PC. All non-volatile registers that need to be restored must 1855 have been saved in a small range on the stack that 1856 starts <code>EBP-4</code> to <code>EBP-1020</code> (<code>RBP-8</code> 1857 to <code>RBP-1020</code>). The offset (divided by 4 in 32-bit mode and 8 1858 in 64-bit mode) is encoded in bits 16-23 (mask: <code>0x00FF0000</code>). 1859 The registers saved are encoded in bits 0-14 1860 (mask: <code>0x00007FFF</code>) as five 3-bit entries from the following 1861 table:</p> 1862<table border="1" cellspacing="0"> 1863 <tr> 1864 <th>Compact Number</th> 1865 <th>i386 Register</th> 1866 <th>x86-64 Regiser</th> 1867 </tr> 1868 <tr> 1869 <td>1</td> 1870 <td><code>EBX</code></td> 1871 <td><code>RBX</code></td> 1872 </tr> 1873 <tr> 1874 <td>2</td> 1875 <td><code>ECX</code></td> 1876 <td><code>R12</code></td> 1877 </tr> 1878 <tr> 1879 <td>3</td> 1880 <td><code>EDX</code></td> 1881 <td><code>R13</code></td> 1882 </tr> 1883 <tr> 1884 <td>4</td> 1885 <td><code>EDI</code></td> 1886 <td><code>R14</code></td> 1887 </tr> 1888 <tr> 1889 <td>5</td> 1890 <td><code>ESI</code></td> 1891 <td><code>R15</code></td> 1892 </tr> 1893 <tr> 1894 <td>6</td> 1895 <td><code>EBP</code></td> 1896 <td><code>RBP</code></td> 1897 </tr> 1898</table> 1899 1900</dd> 1901 1902 <dt><i>Frameless with a Small Constant Stack Size (<code>EBP</code> 1903 or <code>RBP</code> is not used as a frame pointer)</i></dt> 1904 <dd><p>To return, a constant (encoded in the compact unwind encoding) is added 1905 to the <code>ESP/RSP</code>. Then the return is done by popping the stack 1906 into the PC. All non-volatile registers that need to be restored must have 1907 been saved on the stack immediately after the return address. The stack 1908 size (divided by 4 in 32-bit mode and 8 in 64-bit mode) is encoded in bits 1909 16-23 (mask: <code>0x00FF0000</code>). There is a maximum stack size of 1910 1024 bytes in 32-bit mode and 2048 in 64-bit mode. The number of registers 1911 saved is encoded in bits 9-12 (mask: <code>0x00001C00</code>). Bits 0-9 1912 (mask: <code>0x000003FF</code>) contain which registers were saved and 1913 their order. (See 1914 the <code>encodeCompactUnwindRegistersWithoutFrame()</code> function 1915 in <code>lib/Target/X86FrameLowering.cpp</code> for the encoding 1916 algorithm.)</p></dd> 1917 1918 <dt><i>Frameless with a Large Constant Stack Size (<code>EBP</code> 1919 or <code>RBP</code> is not used as a frame pointer)</i></dt> 1920 <dd><p>This case is like the "Frameless with a Small Constant Stack Size" 1921 case, but the stack size is too large to encode in the compact unwind 1922 encoding. Instead it requires that the function contains "<code>subl 1923 $nnnnnn, %esp</code>" in its prolog. The compact encoding contains the 1924 offset to the <code>$nnnnnn</code> value in the function in bits 9-12 1925 (mask: <code>0x00001C00</code>).</p></dd> 1926</dl> 1927 1928</div> 1929 1930<!-- ======================================================================= --> 1931<h3> 1932 <a name="latemco">Late Machine Code Optimizations</a> 1933</h3> 1934<div><p>To Be Written</p></div> 1935 1936<!-- ======================================================================= --> 1937<h3> 1938 <a name="codeemit">Code Emission</a> 1939</h3> 1940 1941<div> 1942 1943<p>The code emission step of code generation is responsible for lowering from 1944the code generator abstractions (like <a 1945href="#machinefunction">MachineFunction</a>, <a 1946href="#machineinstr">MachineInstr</a>, etc) down 1947to the abstractions used by the MC layer (<a href="#mcinst">MCInst</a>, 1948<a href="#mcstreamer">MCStreamer</a>, etc). This is 1949done with a combination of several different classes: the (misnamed) 1950target-independent AsmPrinter class, target-specific subclasses of AsmPrinter 1951(such as SparcAsmPrinter), and the TargetLoweringObjectFile class.</p> 1952 1953<p>Since the MC layer works at the level of abstraction of object files, it 1954doesn't have a notion of functions, global variables etc. Instead, it thinks 1955about labels, directives, and instructions. A key class used at this time is 1956the MCStreamer class. This is an abstract API that is implemented in different 1957ways (e.g. to output a .s file, output an ELF .o file, etc) that is effectively 1958an "assembler API". MCStreamer has one method per directive, such as EmitLabel, 1959EmitSymbolAttribute, SwitchSection, etc, which directly correspond to assembly 1960level directives. 1961</p> 1962 1963<p>If you are interested in implementing a code generator for a target, there 1964are three important things that you have to implement for your target:</p> 1965 1966<ol> 1967<li>First, you need a subclass of AsmPrinter for your target. This class 1968implements the general lowering process converting MachineFunction's into MC 1969label constructs. The AsmPrinter base class provides a number of useful methods 1970and routines, and also allows you to override the lowering process in some 1971important ways. You should get much of the lowering for free if you are 1972implementing an ELF, COFF, or MachO target, because the TargetLoweringObjectFile 1973class implements much of the common logic.</li> 1974 1975<li>Second, you need to implement an instruction printer for your target. The 1976instruction printer takes an <a href="#mcinst">MCInst</a> and renders it to a 1977raw_ostream as text. Most of this is automatically generated from the .td file 1978(when you specify something like "<tt>add $dst, $src1, $src2</tt>" in the 1979instructions), but you need to implement routines to print operands.</li> 1980 1981<li>Third, you need to implement code that lowers a <a 1982href="#machineinstr">MachineInstr</a> to an MCInst, usually implemented in 1983"<target>MCInstLower.cpp". This lowering process is often target 1984specific, and is responsible for turning jump table entries, constant pool 1985indices, global variable addresses, etc into MCLabels as appropriate. This 1986translation layer is also responsible for expanding pseudo ops used by the code 1987generator into the actual machine instructions they correspond to. The MCInsts 1988that are generated by this are fed into the instruction printer or the encoder. 1989</li> 1990 1991</ol> 1992 1993<p>Finally, at your choosing, you can also implement an subclass of 1994MCCodeEmitter which lowers MCInst's into machine code bytes and relocations. 1995This is important if you want to support direct .o file emission, or would like 1996to implement an assembler for your target.</p> 1997 1998</div> 1999 2000</div> 2001 2002<!-- *********************************************************************** --> 2003<h2> 2004 <a name="nativeassembler">Implementing a Native Assembler</a> 2005</h2> 2006<!-- *********************************************************************** --> 2007 2008<div> 2009 2010<p>Though you're probably reading this because you want to write or maintain a 2011compiler backend, LLVM also fully supports building a native assemblers too. 2012We've tried hard to automate the generation of the assembler from the .td files 2013(in particular the instruction syntax and encodings), which means that a large 2014part of the manual and repetitive data entry can be factored and shared with the 2015compiler.</p> 2016 2017<!-- ======================================================================= --> 2018<h3 id="na_instparsing">Instruction Parsing</h3> 2019 2020<div><p>To Be Written</p></div> 2021 2022 2023<!-- ======================================================================= --> 2024<h3 id="na_instaliases"> 2025 Instruction Alias Processing 2026</h3> 2027 2028<div> 2029<p>Once the instruction is parsed, it enters the MatchInstructionImpl function. 2030The MatchInstructionImpl function performs alias processing and then does 2031actual matching.</p> 2032 2033<p>Alias processing is the phase that canonicalizes different lexical forms of 2034the same instructions down to one representation. There are several different 2035kinds of alias that are possible to implement and they are listed below in the 2036order that they are processed (which is in order from simplest/weakest to most 2037complex/powerful). Generally you want to use the first alias mechanism that 2038meets the needs of your instruction, because it will allow a more concise 2039description.</p> 2040 2041<!-- _______________________________________________________________________ --> 2042<h4>Mnemonic Aliases</h4> 2043 2044<div> 2045 2046<p>The first phase of alias processing is simple instruction mnemonic 2047remapping for classes of instructions which are allowed with two different 2048mnemonics. This phase is a simple and unconditionally remapping from one input 2049mnemonic to one output mnemonic. It isn't possible for this form of alias to 2050look at the operands at all, so the remapping must apply for all forms of a 2051given mnemonic. Mnemonic aliases are defined simply, for example X86 has: 2052</p> 2053 2054<div class="doc_code"> 2055<pre> 2056def : MnemonicAlias<"cbw", "cbtw">; 2057def : MnemonicAlias<"smovq", "movsq">; 2058def : MnemonicAlias<"fldcww", "fldcw">; 2059def : MnemonicAlias<"fucompi", "fucomip">; 2060def : MnemonicAlias<"ud2a", "ud2">; 2061</pre> 2062</div> 2063 2064<p>... and many others. With a MnemonicAlias definition, the mnemonic is 2065remapped simply and directly. Though MnemonicAlias's can't look at any aspect 2066of the instruction (such as the operands) they can depend on global modes (the 2067same ones supported by the matcher), through a Requires clause:</p> 2068 2069<div class="doc_code"> 2070<pre> 2071def : MnemonicAlias<"pushf", "pushfq">, Requires<[In64BitMode]>; 2072def : MnemonicAlias<"pushf", "pushfl">, Requires<[In32BitMode]>; 2073</pre> 2074</div> 2075 2076<p>In this example, the mnemonic gets mapped into different a new one depending 2077on the current instruction set.</p> 2078 2079</div> 2080 2081<!-- _______________________________________________________________________ --> 2082<h4>Instruction Aliases</h4> 2083 2084<div> 2085 2086<p>The most general phase of alias processing occurs while matching is 2087happening: it provides new forms for the matcher to match along with a specific 2088instruction to generate. An instruction alias has two parts: the string to 2089match and the instruction to generate. For example: 2090</p> 2091 2092<div class="doc_code"> 2093<pre> 2094def : InstAlias<"movsx $src, $dst", (MOVSX16rr8W GR16:$dst, GR8 :$src)>; 2095def : InstAlias<"movsx $src, $dst", (MOVSX16rm8W GR16:$dst, i8mem:$src)>; 2096def : InstAlias<"movsx $src, $dst", (MOVSX32rr8 GR32:$dst, GR8 :$src)>; 2097def : InstAlias<"movsx $src, $dst", (MOVSX32rr16 GR32:$dst, GR16 :$src)>; 2098def : InstAlias<"movsx $src, $dst", (MOVSX64rr8 GR64:$dst, GR8 :$src)>; 2099def : InstAlias<"movsx $src, $dst", (MOVSX64rr16 GR64:$dst, GR16 :$src)>; 2100def : InstAlias<"movsx $src, $dst", (MOVSX64rr32 GR64:$dst, GR32 :$src)>; 2101</pre> 2102</div> 2103 2104<p>This shows a powerful example of the instruction aliases, matching the 2105same mnemonic in multiple different ways depending on what operands are present 2106in the assembly. The result of instruction aliases can include operands in a 2107different order than the destination instruction, and can use an input 2108multiple times, for example:</p> 2109 2110<div class="doc_code"> 2111<pre> 2112def : InstAlias<"clrb $reg", (XOR8rr GR8 :$reg, GR8 :$reg)>; 2113def : InstAlias<"clrw $reg", (XOR16rr GR16:$reg, GR16:$reg)>; 2114def : InstAlias<"clrl $reg", (XOR32rr GR32:$reg, GR32:$reg)>; 2115def : InstAlias<"clrq $reg", (XOR64rr GR64:$reg, GR64:$reg)>; 2116</pre> 2117</div> 2118 2119<p>This example also shows that tied operands are only listed once. In the X86 2120backend, XOR8rr has two input GR8's and one output GR8 (where an input is tied 2121to the output). InstAliases take a flattened operand list without duplicates 2122for tied operands. The result of an instruction alias can also use immediates 2123and fixed physical registers which are added as simple immediate operands in the 2124result, for example:</p> 2125 2126<div class="doc_code"> 2127<pre> 2128// Fixed Immediate operand. 2129def : InstAlias<"aad", (AAD8i8 10)>; 2130 2131// Fixed register operand. 2132def : InstAlias<"fcomi", (COM_FIr ST1)>; 2133 2134// Simple alias. 2135def : InstAlias<"fcomi $reg", (COM_FIr RST:$reg)>; 2136</pre> 2137</div> 2138 2139 2140<p>Instruction aliases can also have a Requires clause to make them 2141subtarget specific.</p> 2142 2143<p>If the back-end supports it, the instruction printer can automatically emit 2144 the alias rather than what's being aliased. It typically leads to better, 2145 more readable code. If it's better to print out what's being aliased, then 2146 pass a '0' as the third parameter to the InstAlias definition.</p> 2147 2148</div> 2149 2150</div> 2151 2152<!-- ======================================================================= --> 2153<h3 id="na_matching">Instruction Matching</h3> 2154 2155<div><p>To Be Written</p></div> 2156 2157</div> 2158 2159<!-- *********************************************************************** --> 2160<h2> 2161 <a name="targetimpls">Target-specific Implementation Notes</a> 2162</h2> 2163<!-- *********************************************************************** --> 2164 2165<div> 2166 2167<p>This section of the document explains features or design decisions that are 2168 specific to the code generator for a particular target. First we start 2169 with a table that summarizes what features are supported by each target.</p> 2170 2171<!-- ======================================================================= --> 2172<h3> 2173 <a name="targetfeatures">Target Feature Matrix</a> 2174</h3> 2175 2176<div> 2177 2178<p>Note that this table does not include the C backend or Cpp backends, since 2179they do not use the target independent code generator infrastructure. It also 2180doesn't list features that are not supported fully by any target yet. It 2181considers a feature to be supported if at least one subtarget supports it. A 2182feature being supported means that it is useful and works for most cases, it 2183does not indicate that there are zero known bugs in the implementation. Here 2184is the key:</p> 2185 2186 2187<table border="1" cellspacing="0"> 2188 <tr> 2189 <th>Unknown</th> 2190 <th>No support</th> 2191 <th>Partial Support</th> 2192 <th>Complete Support</th> 2193 </tr> 2194 <tr> 2195 <td class="unknown"></td> 2196 <td class="no"></td> 2197 <td class="partial"></td> 2198 <td class="yes"></td> 2199 </tr> 2200</table> 2201 2202<p>Here is the table:</p> 2203 2204<table width="689" border="1" cellspacing="0"> 2205<tr><td></td> 2206<td colspan="13" align="center" style="background-color:#ffc">Target</td> 2207</tr> 2208 <tr> 2209 <th>Feature</th> 2210 <th>ARM</th> 2211 <th>Alpha</th> 2212 <th>Blackfin</th> 2213 <th>CellSPU</th> 2214 <th>MBlaze</th> 2215 <th>MSP430</th> 2216 <th>Mips</th> 2217 <th>PTX</th> 2218 <th>PowerPC</th> 2219 <th>Sparc</th> 2220 <th>SystemZ</th> 2221 <th>X86</th> 2222 <th>XCore</th> 2223 </tr> 2224 2225<tr> 2226 <td><a href="#feat_reliable">is generally reliable</a></td> 2227 <td class="yes"></td> <!-- ARM --> 2228 <td class="unknown"></td> <!-- Alpha --> 2229 <td class="no"></td> <!-- Blackfin --> 2230 <td class="no"></td> <!-- CellSPU --> 2231 <td class="no"></td> <!-- MBlaze --> 2232 <td class="unknown"></td> <!-- MSP430 --> 2233 <td class="no"></td> <!-- Mips --> 2234 <td class="no"></td> <!-- PTX --> 2235 <td class="yes"></td> <!-- PowerPC --> 2236 <td class="yes"></td> <!-- Sparc --> 2237 <td class="unknown"></td> <!-- SystemZ --> 2238 <td class="yes"></td> <!-- X86 --> 2239 <td class="unknown"></td> <!-- XCore --> 2240</tr> 2241 2242<tr> 2243 <td><a href="#feat_asmparser">assembly parser</a></td> 2244 <td class="no"></td> <!-- ARM --> 2245 <td class="no"></td> <!-- Alpha --> 2246 <td class="no"></td> <!-- Blackfin --> 2247 <td class="no"></td> <!-- CellSPU --> 2248 <td class="yes"></td> <!-- MBlaze --> 2249 <td class="no"></td> <!-- MSP430 --> 2250 <td class="no"></td> <!-- Mips --> 2251 <td class="no"></td> <!-- PTX --> 2252 <td class="no"></td> <!-- PowerPC --> 2253 <td class="no"></td> <!-- Sparc --> 2254 <td class="no"></td> <!-- SystemZ --> 2255 <td class="yes"></td> <!-- X86 --> 2256 <td class="no"></td> <!-- XCore --> 2257</tr> 2258 2259<tr> 2260 <td><a href="#feat_disassembler">disassembler</a></td> 2261 <td class="yes"></td> <!-- ARM --> 2262 <td class="no"></td> <!-- Alpha --> 2263 <td class="no"></td> <!-- Blackfin --> 2264 <td class="no"></td> <!-- CellSPU --> 2265 <td class="yes"></td> <!-- MBlaze --> 2266 <td class="no"></td> <!-- MSP430 --> 2267 <td class="no"></td> <!-- Mips --> 2268 <td class="no"></td> <!-- PTX --> 2269 <td class="no"></td> <!-- PowerPC --> 2270 <td class="no"></td> <!-- Sparc --> 2271 <td class="no"></td> <!-- SystemZ --> 2272 <td class="yes"></td> <!-- X86 --> 2273 <td class="no"></td> <!-- XCore --> 2274</tr> 2275 2276<tr> 2277 <td><a href="#feat_inlineasm">inline asm</a></td> 2278 <td class="yes"></td> <!-- ARM --> 2279 <td class="unknown"></td> <!-- Alpha --> 2280 <td class="yes"></td> <!-- Blackfin --> 2281 <td class="no"></td> <!-- CellSPU --> 2282 <td class="yes"></td> <!-- MBlaze --> 2283 <td class="unknown"></td> <!-- MSP430 --> 2284 <td class="no"></td> <!-- Mips --> 2285 <td class="unknown"></td> <!-- PTX --> 2286 <td class="yes"></td> <!-- PowerPC --> 2287 <td class="unknown"></td> <!-- Sparc --> 2288 <td class="unknown"></td> <!-- SystemZ --> 2289 <td class="yes"></td> <!-- X86 --> 2290 <td class="unknown"></td> <!-- XCore --> 2291</tr> 2292 2293<tr> 2294 <td><a href="#feat_jit">jit</a></td> 2295 <td class="partial"><a href="#feat_jit_arm">*</a></td> <!-- ARM --> 2296 <td class="no"></td> <!-- Alpha --> 2297 <td class="no"></td> <!-- Blackfin --> 2298 <td class="no"></td> <!-- CellSPU --> 2299 <td class="no"></td> <!-- MBlaze --> 2300 <td class="unknown"></td> <!-- MSP430 --> 2301 <td class="no"></td> <!-- Mips --> 2302 <td class="unknown"></td> <!-- PTX --> 2303 <td class="yes"></td> <!-- PowerPC --> 2304 <td class="unknown"></td> <!-- Sparc --> 2305 <td class="unknown"></td> <!-- SystemZ --> 2306 <td class="yes"></td> <!-- X86 --> 2307 <td class="unknown"></td> <!-- XCore --> 2308</tr> 2309 2310<tr> 2311 <td><a href="#feat_objectwrite">.o file writing</a></td> 2312 <td class="no"></td> <!-- ARM --> 2313 <td class="no"></td> <!-- Alpha --> 2314 <td class="no"></td> <!-- Blackfin --> 2315 <td class="no"></td> <!-- CellSPU --> 2316 <td class="yes"></td> <!-- MBlaze --> 2317 <td class="no"></td> <!-- MSP430 --> 2318 <td class="no"></td> <!-- Mips --> 2319 <td class="no"></td> <!-- PTX --> 2320 <td class="no"></td> <!-- PowerPC --> 2321 <td class="no"></td> <!-- Sparc --> 2322 <td class="no"></td> <!-- SystemZ --> 2323 <td class="yes"></td> <!-- X86 --> 2324 <td class="no"></td> <!-- XCore --> 2325</tr> 2326 2327<tr> 2328 <td><a href="#feat_tailcall">tail calls</a></td> 2329 <td class="yes"></td> <!-- ARM --> 2330 <td class="unknown"></td> <!-- Alpha --> 2331 <td class="no"></td> <!-- Blackfin --> 2332 <td class="no"></td> <!-- CellSPU --> 2333 <td class="no"></td> <!-- MBlaze --> 2334 <td class="unknown"></td> <!-- MSP430 --> 2335 <td class="no"></td> <!-- Mips --> 2336 <td class="unknown"></td> <!-- PTX --> 2337 <td class="yes"></td> <!-- PowerPC --> 2338 <td class="unknown"></td> <!-- Sparc --> 2339 <td class="unknown"></td> <!-- SystemZ --> 2340 <td class="yes"></td> <!-- X86 --> 2341 <td class="unknown"></td> <!-- XCore --> 2342</tr> 2343 2344 2345</table> 2346 2347<!-- _______________________________________________________________________ --> 2348<h4 id="feat_reliable">Is Generally Reliable</h4> 2349 2350<div> 2351<p>This box indicates whether the target is considered to be production quality. 2352This indicates that the target has been used as a static compiler to 2353compile large amounts of code by a variety of different people and is in 2354continuous use.</p> 2355</div> 2356 2357<!-- _______________________________________________________________________ --> 2358<h4 id="feat_asmparser">Assembly Parser</h4> 2359 2360<div> 2361<p>This box indicates whether the target supports parsing target specific .s 2362files by implementing the MCAsmParser interface. This is required for llvm-mc 2363to be able to act as a native assembler and is required for inline assembly 2364support in the native .o file writer.</p> 2365 2366</div> 2367 2368 2369<!-- _______________________________________________________________________ --> 2370<h4 id="feat_disassembler">Disassembler</h4> 2371 2372<div> 2373<p>This box indicates whether the target supports the MCDisassembler API for 2374disassembling machine opcode bytes into MCInst's.</p> 2375 2376</div> 2377 2378<!-- _______________________________________________________________________ --> 2379<h4 id="feat_inlineasm">Inline Asm</h4> 2380 2381<div> 2382<p>This box indicates whether the target supports most popular inline assembly 2383constraints and modifiers.</p> 2384 2385</div> 2386 2387<!-- _______________________________________________________________________ --> 2388<h4 id="feat_jit">JIT Support</h4> 2389 2390<div> 2391<p>This box indicates whether the target supports the JIT compiler through 2392the ExecutionEngine interface.</p> 2393 2394<p id="feat_jit_arm">The ARM backend has basic support for integer code 2395in ARM codegen mode, but lacks NEON and full Thumb support.</p> 2396 2397</div> 2398 2399<!-- _______________________________________________________________________ --> 2400<h4 id="feat_objectwrite">.o File Writing</h4> 2401 2402<div> 2403 2404<p>This box indicates whether the target supports writing .o files (e.g. MachO, 2405ELF, and/or COFF) files directly from the target. Note that the target also 2406must include an assembly parser and general inline assembly support for full 2407inline assembly support in the .o writer.</p> 2408 2409<p>Targets that don't support this feature can obviously still write out .o 2410files, they just rely on having an external assembler to translate from a .s 2411file to a .o file (as is the case for many C compilers).</p> 2412 2413</div> 2414 2415<!-- _______________________________________________________________________ --> 2416<h4 id="feat_tailcall">Tail Calls</h4> 2417 2418<div> 2419 2420<p>This box indicates whether the target supports guaranteed tail calls. These 2421are calls marked "<a href="LangRef.html#i_call">tail</a>" and use the fastcc 2422calling convention. Please see the <a href="#tailcallopt">tail call section 2423more more details</a>.</p> 2424 2425</div> 2426 2427</div> 2428 2429<!-- ======================================================================= --> 2430<h3> 2431 <a name="tailcallopt">Tail call optimization</a> 2432</h3> 2433 2434<div> 2435 2436<p>Tail call optimization, callee reusing the stack of the caller, is currently 2437 supported on x86/x86-64 and PowerPC. It is performed if:</p> 2438 2439<ul> 2440 <li>Caller and callee have the calling convention <tt>fastcc</tt> or 2441 <tt>cc 10</tt> (GHC call convention).</li> 2442 2443 <li>The call is a tail call - in tail position (ret immediately follows call 2444 and ret uses value of call or is void).</li> 2445 2446 <li>Option <tt>-tailcallopt</tt> is enabled.</li> 2447 2448 <li>Platform specific constraints are met.</li> 2449</ul> 2450 2451<p>x86/x86-64 constraints:</p> 2452 2453<ul> 2454 <li>No variable argument lists are used.</li> 2455 2456 <li>On x86-64 when generating GOT/PIC code only module-local calls (visibility 2457 = hidden or protected) are supported.</li> 2458</ul> 2459 2460<p>PowerPC constraints:</p> 2461 2462<ul> 2463 <li>No variable argument lists are used.</li> 2464 2465 <li>No byval parameters are used.</li> 2466 2467 <li>On ppc32/64 GOT/PIC only module-local calls (visibility = hidden or protected) are supported.</li> 2468</ul> 2469 2470<p>Example:</p> 2471 2472<p>Call as <tt>llc -tailcallopt test.ll</tt>.</p> 2473 2474<div class="doc_code"> 2475<pre> 2476declare fastcc i32 @tailcallee(i32 inreg %a1, i32 inreg %a2, i32 %a3, i32 %a4) 2477 2478define fastcc i32 @tailcaller(i32 %in1, i32 %in2) { 2479 %l1 = add i32 %in1, %in2 2480 %tmp = tail call fastcc i32 @tailcallee(i32 %in1 inreg, i32 %in2 inreg, i32 %in1, i32 %l1) 2481 ret i32 %tmp 2482} 2483</pre> 2484</div> 2485 2486<p>Implications of <tt>-tailcallopt</tt>:</p> 2487 2488<p>To support tail call optimization in situations where the callee has more 2489 arguments than the caller a 'callee pops arguments' convention is used. This 2490 currently causes each <tt>fastcc</tt> call that is not tail call optimized 2491 (because one or more of above constraints are not met) to be followed by a 2492 readjustment of the stack. So performance might be worse in such cases.</p> 2493 2494</div> 2495<!-- ======================================================================= --> 2496<h3> 2497 <a name="sibcallopt">Sibling call optimization</a> 2498</h3> 2499 2500<div> 2501 2502<p>Sibling call optimization is a restricted form of tail call optimization. 2503 Unlike tail call optimization described in the previous section, it can be 2504 performed automatically on any tail calls when <tt>-tailcallopt</tt> option 2505 is not specified.</p> 2506 2507<p>Sibling call optimization is currently performed on x86/x86-64 when the 2508 following constraints are met:</p> 2509 2510<ul> 2511 <li>Caller and callee have the same calling convention. It can be either 2512 <tt>c</tt> or <tt>fastcc</tt>. 2513 2514 <li>The call is a tail call - in tail position (ret immediately follows call 2515 and ret uses value of call or is void).</li> 2516 2517 <li>Caller and callee have matching return type or the callee result is not 2518 used. 2519 2520 <li>If any of the callee arguments are being passed in stack, they must be 2521 available in caller's own incoming argument stack and the frame offsets 2522 must be the same. 2523</ul> 2524 2525<p>Example:</p> 2526<div class="doc_code"> 2527<pre> 2528declare i32 @bar(i32, i32) 2529 2530define i32 @foo(i32 %a, i32 %b, i32 %c) { 2531entry: 2532 %0 = tail call i32 @bar(i32 %a, i32 %b) 2533 ret i32 %0 2534} 2535</pre> 2536</div> 2537 2538</div> 2539<!-- ======================================================================= --> 2540<h3> 2541 <a name="x86">The X86 backend</a> 2542</h3> 2543 2544<div> 2545 2546<p>The X86 code generator lives in the <tt>lib/Target/X86</tt> directory. This 2547 code generator is capable of targeting a variety of x86-32 and x86-64 2548 processors, and includes support for ISA extensions such as MMX and SSE.</p> 2549 2550<!-- _______________________________________________________________________ --> 2551<h4> 2552 <a name="x86_tt">X86 Target Triples supported</a> 2553</h4> 2554 2555<div> 2556 2557<p>The following are the known target triples that are supported by the X86 2558 backend. This is not an exhaustive list, and it would be useful to add those 2559 that people test.</p> 2560 2561<ul> 2562 <li><b>i686-pc-linux-gnu</b> — Linux</li> 2563 2564 <li><b>i386-unknown-freebsd5.3</b> — FreeBSD 5.3</li> 2565 2566 <li><b>i686-pc-cygwin</b> — Cygwin on Win32</li> 2567 2568 <li><b>i686-pc-mingw32</b> — MingW on Win32</li> 2569 2570 <li><b>i386-pc-mingw32msvc</b> — MingW crosscompiler on Linux</li> 2571 2572 <li><b>i686-apple-darwin*</b> — Apple Darwin on X86</li> 2573 2574 <li><b>x86_64-unknown-linux-gnu</b> — Linux</li> 2575</ul> 2576 2577</div> 2578 2579<!-- _______________________________________________________________________ --> 2580<h4> 2581 <a name="x86_cc">X86 Calling Conventions supported</a> 2582</h4> 2583 2584 2585<div> 2586 2587<p>The following target-specific calling conventions are known to backend:</p> 2588 2589<ul> 2590<li><b>x86_StdCall</b> — stdcall calling convention seen on Microsoft 2591 Windows platform (CC ID = 64).</li> 2592<li><b>x86_FastCall</b> — fastcall calling convention seen on Microsoft 2593 Windows platform (CC ID = 65).</li> 2594<li><b>x86_ThisCall</b> — Similar to X86_StdCall. Passes first argument 2595 in ECX, others via stack. Callee is responsible for stack cleaning. This 2596 convention is used by MSVC by default for methods in its ABI 2597 (CC ID = 70).</li> 2598</ul> 2599 2600</div> 2601 2602<!-- _______________________________________________________________________ --> 2603<h4> 2604 <a name="x86_memory">Representing X86 addressing modes in MachineInstrs</a> 2605</h4> 2606 2607<div> 2608 2609<p>The x86 has a very flexible way of accessing memory. It is capable of 2610 forming memory addresses of the following expression directly in integer 2611 instructions (which use ModR/M addressing):</p> 2612 2613<div class="doc_code"> 2614<pre> 2615SegmentReg: Base + [1,2,4,8] * IndexReg + Disp32 2616</pre> 2617</div> 2618 2619<p>In order to represent this, LLVM tracks no less than 5 operands for each 2620 memory operand of this form. This means that the "load" form of 2621 '<tt>mov</tt>' has the following <tt>MachineOperand</tt>s in this order:</p> 2622 2623<div class="doc_code"> 2624<pre> 2625Index: 0 | 1 2 3 4 5 2626Meaning: DestReg, | BaseReg, Scale, IndexReg, Displacement Segment 2627OperandTy: VirtReg, | VirtReg, UnsImm, VirtReg, SignExtImm PhysReg 2628</pre> 2629</div> 2630 2631<p>Stores, and all other instructions, treat the four memory operands in the 2632 same way and in the same order. If the segment register is unspecified 2633 (regno = 0), then no segment override is generated. "Lea" operations do not 2634 have a segment register specified, so they only have 4 operands for their 2635 memory reference.</p> 2636 2637</div> 2638 2639<!-- _______________________________________________________________________ --> 2640<h4> 2641 <a name="x86_memory">X86 address spaces supported</a> 2642</h4> 2643 2644<div> 2645 2646<p>x86 has a feature which provides 2647 the ability to perform loads and stores to different address spaces 2648 via the x86 segment registers. A segment override prefix byte on an 2649 instruction causes the instruction's memory access to go to the specified 2650 segment. LLVM address space 0 is the default address space, which includes 2651 the stack, and any unqualified memory accesses in a program. Address spaces 2652 1-255 are currently reserved for user-defined code. The GS-segment is 2653 represented by address space 256, while the FS-segment is represented by 2654 address space 257. Other x86 segments have yet to be allocated address space 2655 numbers.</p> 2656 2657<p>While these address spaces may seem similar to TLS via the 2658 <tt>thread_local</tt> keyword, and often use the same underlying hardware, 2659 there are some fundamental differences.</p> 2660 2661<p>The <tt>thread_local</tt> keyword applies to global variables and 2662 specifies that they are to be allocated in thread-local memory. There are 2663 no type qualifiers involved, and these variables can be pointed to with 2664 normal pointers and accessed with normal loads and stores. 2665 The <tt>thread_local</tt> keyword is target-independent at the LLVM IR 2666 level (though LLVM doesn't yet have implementations of it for some 2667 configurations).<p> 2668 2669<p>Special address spaces, in contrast, apply to static types. Every 2670 load and store has a particular address space in its address operand type, 2671 and this is what determines which address space is accessed. 2672 LLVM ignores these special address space qualifiers on global variables, 2673 and does not provide a way to directly allocate storage in them. 2674 At the LLVM IR level, the behavior of these special address spaces depends 2675 in part on the underlying OS or runtime environment, and they are specific 2676 to x86 (and LLVM doesn't yet handle them correctly in some cases).</p> 2677 2678<p>Some operating systems and runtime environments use (or may in the future 2679 use) the FS/GS-segment registers for various low-level purposes, so care 2680 should be taken when considering them.</p> 2681 2682</div> 2683 2684<!-- _______________________________________________________________________ --> 2685<h4> 2686 <a name="x86_names">Instruction naming</a> 2687</h4> 2688 2689<div> 2690 2691<p>An instruction name consists of the base name, a default operand size, and a 2692 a character per operand with an optional special size. For example:</p> 2693 2694<div class="doc_code"> 2695<pre> 2696ADD8rr -> add, 8-bit register, 8-bit register 2697IMUL16rmi -> imul, 16-bit register, 16-bit memory, 16-bit immediate 2698IMUL16rmi8 -> imul, 16-bit register, 16-bit memory, 8-bit immediate 2699MOVSX32rm16 -> movsx, 32-bit register, 16-bit memory 2700</pre> 2701</div> 2702 2703</div> 2704 2705</div> 2706 2707<!-- ======================================================================= --> 2708<h3> 2709 <a name="ppc">The PowerPC backend</a> 2710</h3> 2711 2712<div> 2713 2714<p>The PowerPC code generator lives in the lib/Target/PowerPC directory. The 2715 code generation is retargetable to several variations or <i>subtargets</i> of 2716 the PowerPC ISA; including ppc32, ppc64 and altivec.</p> 2717 2718<!-- _______________________________________________________________________ --> 2719<h4> 2720 <a name="ppc_abi">LLVM PowerPC ABI</a> 2721</h4> 2722 2723<div> 2724 2725<p>LLVM follows the AIX PowerPC ABI, with two deviations. LLVM uses a PC 2726 relative (PIC) or static addressing for accessing global values, so no TOC 2727 (r2) is used. Second, r31 is used as a frame pointer to allow dynamic growth 2728 of a stack frame. LLVM takes advantage of having no TOC to provide space to 2729 save the frame pointer in the PowerPC linkage area of the caller frame. 2730 Other details of PowerPC ABI can be found at <a href= 2731 "http://developer.apple.com/documentation/DeveloperTools/Conceptual/LowLevelABI/Articles/32bitPowerPC.html" 2732 >PowerPC ABI.</a> Note: This link describes the 32 bit ABI. The 64 bit ABI 2733 is similar except space for GPRs are 8 bytes wide (not 4) and r13 is reserved 2734 for system use.</p> 2735 2736</div> 2737 2738<!-- _______________________________________________________________________ --> 2739<h4> 2740 <a name="ppc_frame">Frame Layout</a> 2741</h4> 2742 2743<div> 2744 2745<p>The size of a PowerPC frame is usually fixed for the duration of a 2746 function's invocation. Since the frame is fixed size, all references 2747 into the frame can be accessed via fixed offsets from the stack pointer. The 2748 exception to this is when dynamic alloca or variable sized arrays are 2749 present, then a base pointer (r31) is used as a proxy for the stack pointer 2750 and stack pointer is free to grow or shrink. A base pointer is also used if 2751 llvm-gcc is not passed the -fomit-frame-pointer flag. The stack pointer is 2752 always aligned to 16 bytes, so that space allocated for altivec vectors will 2753 be properly aligned.</p> 2754 2755<p>An invocation frame is laid out as follows (low memory at top);</p> 2756 2757<table class="layout"> 2758 <tr> 2759 <td>Linkage<br><br></td> 2760 </tr> 2761 <tr> 2762 <td>Parameter area<br><br></td> 2763 </tr> 2764 <tr> 2765 <td>Dynamic area<br><br></td> 2766 </tr> 2767 <tr> 2768 <td>Locals area<br><br></td> 2769 </tr> 2770 <tr> 2771 <td>Saved registers area<br><br></td> 2772 </tr> 2773 <tr style="border-style: none hidden none hidden;"> 2774 <td><br></td> 2775 </tr> 2776 <tr> 2777 <td>Previous Frame<br><br></td> 2778 </tr> 2779</table> 2780 2781<p>The <i>linkage</i> area is used by a callee to save special registers prior 2782 to allocating its own frame. Only three entries are relevant to LLVM. The 2783 first entry is the previous stack pointer (sp), aka link. This allows 2784 probing tools like gdb or exception handlers to quickly scan the frames in 2785 the stack. A function epilog can also use the link to pop the frame from the 2786 stack. The third entry in the linkage area is used to save the return 2787 address from the lr register. Finally, as mentioned above, the last entry is 2788 used to save the previous frame pointer (r31.) The entries in the linkage 2789 area are the size of a GPR, thus the linkage area is 24 bytes long in 32 bit 2790 mode and 48 bytes in 64 bit mode.</p> 2791 2792<p>32 bit linkage area</p> 2793 2794<table class="layout"> 2795 <tr> 2796 <td>0</td> 2797 <td>Saved SP (r1)</td> 2798 </tr> 2799 <tr> 2800 <td>4</td> 2801 <td>Saved CR</td> 2802 </tr> 2803 <tr> 2804 <td>8</td> 2805 <td>Saved LR</td> 2806 </tr> 2807 <tr> 2808 <td>12</td> 2809 <td>Reserved</td> 2810 </tr> 2811 <tr> 2812 <td>16</td> 2813 <td>Reserved</td> 2814 </tr> 2815 <tr> 2816 <td>20</td> 2817 <td>Saved FP (r31)</td> 2818 </tr> 2819</table> 2820 2821<p>64 bit linkage area</p> 2822 2823<table class="layout"> 2824 <tr> 2825 <td>0</td> 2826 <td>Saved SP (r1)</td> 2827 </tr> 2828 <tr> 2829 <td>8</td> 2830 <td>Saved CR</td> 2831 </tr> 2832 <tr> 2833 <td>16</td> 2834 <td>Saved LR</td> 2835 </tr> 2836 <tr> 2837 <td>24</td> 2838 <td>Reserved</td> 2839 </tr> 2840 <tr> 2841 <td>32</td> 2842 <td>Reserved</td> 2843 </tr> 2844 <tr> 2845 <td>40</td> 2846 <td>Saved FP (r31)</td> 2847 </tr> 2848</table> 2849 2850<p>The <i>parameter area</i> is used to store arguments being passed to a callee 2851 function. Following the PowerPC ABI, the first few arguments are actually 2852 passed in registers, with the space in the parameter area unused. However, 2853 if there are not enough registers or the callee is a thunk or vararg 2854 function, these register arguments can be spilled into the parameter area. 2855 Thus, the parameter area must be large enough to store all the parameters for 2856 the largest call sequence made by the caller. The size must also be 2857 minimally large enough to spill registers r3-r10. This allows callees blind 2858 to the call signature, such as thunks and vararg functions, enough space to 2859 cache the argument registers. Therefore, the parameter area is minimally 32 2860 bytes (64 bytes in 64 bit mode.) Also note that since the parameter area is 2861 a fixed offset from the top of the frame, that a callee can access its spilt 2862 arguments using fixed offsets from the stack pointer (or base pointer.)</p> 2863 2864<p>Combining the information about the linkage, parameter areas and alignment. A 2865 stack frame is minimally 64 bytes in 32 bit mode and 128 bytes in 64 bit 2866 mode.</p> 2867 2868<p>The <i>dynamic area</i> starts out as size zero. If a function uses dynamic 2869 alloca then space is added to the stack, the linkage and parameter areas are 2870 shifted to top of stack, and the new space is available immediately below the 2871 linkage and parameter areas. The cost of shifting the linkage and parameter 2872 areas is minor since only the link value needs to be copied. The link value 2873 can be easily fetched by adding the original frame size to the base pointer. 2874 Note that allocations in the dynamic space need to observe 16 byte 2875 alignment.</p> 2876 2877<p>The <i>locals area</i> is where the llvm compiler reserves space for local 2878 variables.</p> 2879 2880<p>The <i>saved registers area</i> is where the llvm compiler spills callee 2881 saved registers on entry to the callee.</p> 2882 2883</div> 2884 2885<!-- _______________________________________________________________________ --> 2886<h4> 2887 <a name="ppc_prolog">Prolog/Epilog</a> 2888</h4> 2889 2890<div> 2891 2892<p>The llvm prolog and epilog are the same as described in the PowerPC ABI, with 2893 the following exceptions. Callee saved registers are spilled after the frame 2894 is created. This allows the llvm epilog/prolog support to be common with 2895 other targets. The base pointer callee saved register r31 is saved in the 2896 TOC slot of linkage area. This simplifies allocation of space for the base 2897 pointer and makes it convenient to locate programatically and during 2898 debugging.</p> 2899 2900</div> 2901 2902<!-- _______________________________________________________________________ --> 2903<h4> 2904 <a name="ppc_dynamic">Dynamic Allocation</a> 2905</h4> 2906 2907<div> 2908 2909<p><i>TODO - More to come.</i></p> 2910 2911</div> 2912 2913</div> 2914 2915<!-- ======================================================================= --> 2916<h3> 2917 <a name="ptx">The PTX backend</a> 2918</h3> 2919 2920<div> 2921 2922<p>The PTX code generator lives in the lib/Target/PTX directory. It is 2923 currently a work-in-progress, but already supports most of the code 2924 generation functionality needed to generate correct PTX kernels for 2925 CUDA devices.</p> 2926 2927<p>The code generator can target PTX 2.0+, and shader model 1.0+. The 2928 PTX ISA Reference Manual is used as the primary source of ISA 2929 information, though an effort is made to make the output of the code 2930 generator match the output of the NVidia nvcc compiler, whenever 2931 possible.</p> 2932 2933<p>Code Generator Options:</p> 2934<table border="1" cellspacing="0"> 2935 <tr> 2936 <th>Option</th> 2937 <th>Description</th> 2938 </tr> 2939 <tr> 2940 <td><code>double</code></td> 2941 <td align="left">If enabled, the map_f64_to_f32 directive is 2942 disabled in the PTX output, allowing native double-precision 2943 arithmetic</td> 2944 </tr> 2945 <tr> 2946 <td><code>no-fma</code></td> 2947 <td align="left">Disable generation of Fused-Multiply Add 2948 instructions, which may be beneficial for some devices</td> 2949 </tr> 2950 <tr> 2951 <td><code>smxy / computexy</code></td> 2952 <td align="left">Set shader model/compute capability to x.y, 2953 e.g. sm20 or compute13</td> 2954 </tr> 2955</table> 2956 2957<p>Working:</p> 2958<ul> 2959 <li>Arithmetic instruction selection (including combo FMA)</li> 2960 <li>Bitwise instruction selection</li> 2961 <li>Control-flow instruction selection</li> 2962 <li>Function calls (only on SM 2.0+ and no return arguments)</li> 2963 <li>Addresses spaces (0 = global, 1 = constant, 2 = local, 4 = 2964 shared)</li> 2965 <li>Thread synchronization (bar.sync)</li> 2966 <li>Special register reads ([N]TID, [N]CTAID, PMx, CLOCK, etc.)</li> 2967</ul> 2968 2969<p>In Progress:</p> 2970<ul> 2971 <li>Robust call instruction selection</li> 2972 <li>Stack frame allocation</li> 2973 <li>Device-specific instruction scheduling optimizations</li> 2974</ul> 2975 2976 2977</div> 2978 2979</div> 2980 2981<!-- *********************************************************************** --> 2982<hr> 2983<address> 2984 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img 2985 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a> 2986 <a href="http://validator.w3.org/check/referer"><img 2987 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a> 2988 2989 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br> 2990 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br> 2991 Last modified: $Date: 2011-09-19 14:15:46 -0400 (Mon, 19 Sep 2011) $ 2992</address> 2993 2994</body> 2995</html> 2996