/usr/share/doc/llvm-6.0-doc/html/GarbageCollection.html is in llvm-6.0-doc 1:6.0-1ubuntu2.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 | <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
<title>Garbage Collection with LLVM — LLVM 6 documentation</title>
<link rel="stylesheet" href="_static/llvm-theme.css" type="text/css" />
<link rel="stylesheet" href="_static/pygments.css" type="text/css" />
<script type="text/javascript">
var DOCUMENTATION_OPTIONS = {
URL_ROOT: './',
VERSION: '6',
COLLAPSE_INDEX: false,
FILE_SUFFIX: '.html',
HAS_SOURCE: true,
SOURCELINK_SUFFIX: '.txt'
};
</script>
<script type="text/javascript" src="_static/jquery.js"></script>
<script type="text/javascript" src="_static/underscore.js"></script>
<script type="text/javascript" src="_static/doctools.js"></script>
<link rel="index" title="Index" href="genindex.html" />
<link rel="search" title="Search" href="search.html" />
<link rel="next" title="Writing an LLVM Pass" href="WritingAnLLVMPass.html" />
<link rel="prev" title="How To Use Instruction Mappings" href="HowToUseInstrMappings.html" />
<style type="text/css">
table.right { float: right; margin-left: 20px; }
table.right td { border: 1px solid #ccc; }
</style>
</head>
<body>
<div class="logo">
<a href="index.html">
<img src="_static/logo.png"
alt="LLVM Logo" width="250" height="88"/></a>
</div>
<div class="related" role="navigation" aria-label="related navigation">
<h3>Navigation</h3>
<ul>
<li class="right" style="margin-right: 10px">
<a href="genindex.html" title="General Index"
accesskey="I">index</a></li>
<li class="right" >
<a href="WritingAnLLVMPass.html" title="Writing an LLVM Pass"
accesskey="N">next</a> |</li>
<li class="right" >
<a href="HowToUseInstrMappings.html" title="How To Use Instruction Mappings"
accesskey="P">previous</a> |</li>
<li><a href="http://llvm.org/">LLVM Home</a> | </li>
<li><a href="index.html">Documentation</a>»</li>
</ul>
</div>
<div class="document">
<div class="documentwrapper">
<div class="body" role="main">
<div class="section" id="garbage-collection-with-llvm">
<h1>Garbage Collection with LLVM<a class="headerlink" href="#garbage-collection-with-llvm" title="Permalink to this headline">¶</a></h1>
<div class="contents local topic" id="contents">
<ul class="simple">
<li><a class="reference internal" href="#abstract" id="id1">Abstract</a></li>
<li><a class="reference internal" href="#quick-start" id="id2">Quick Start</a></li>
<li><a class="reference internal" href="#introduction" id="id3">Introduction</a><ul>
<li><a class="reference internal" href="#what-is-garbage-collection" id="id4">What is Garbage Collection?</a></li>
<li><a class="reference internal" href="#goals-and-non-goals" id="id5">Goals and non-goals</a></li>
</ul>
</li>
<li><a class="reference internal" href="#llvm-ir-features" id="id6">LLVM IR Features</a><ul>
<li><a class="reference internal" href="#specifying-gc-code-generation-gc" id="id7">Specifying GC code generation: <code class="docutils literal"><span class="pre">gc</span> <span class="pre">"..."</span></code></a></li>
<li><a class="reference internal" href="#gcroot" id="id8">Identifying GC roots on the stack</a><ul>
<li><a class="reference internal" href="#using-gc-statepoint" id="id9">Using <code class="docutils literal"><span class="pre">gc.statepoint</span></code></a></li>
<li><a class="reference internal" href="#using-llvm-gcwrite" id="id10">Using <code class="docutils literal"><span class="pre">llvm.gcwrite</span></code></a></li>
</ul>
</li>
<li><a class="reference internal" href="#reading-and-writing-references-in-the-heap" id="id11">Reading and writing references in the heap</a><ul>
<li><a class="reference internal" href="#write-barrier-llvm-gcwrite" id="id12">Write barrier: <code class="docutils literal"><span class="pre">llvm.gcwrite</span></code></a></li>
<li><a class="reference internal" href="#read-barrier-llvm-gcread" id="id13">Read barrier: <code class="docutils literal"><span class="pre">llvm.gcread</span></code></a></li>
</ul>
</li>
</ul>
</li>
<li><a class="reference internal" href="#built-in-gc-strategies" id="id14">Built In GC Strategies</a><ul>
<li><a class="reference internal" href="#the-shadow-stack-gc" id="id15">The Shadow Stack GC</a></li>
<li><a class="reference internal" href="#the-erlang-and-ocaml-gcs" id="id16">The ‘Erlang’ and ‘Ocaml’ GCs</a></li>
<li><a class="reference internal" href="#the-statepoint-example-gc" id="id17">The Statepoint Example GC</a></li>
<li><a class="reference internal" href="#the-coreclr-gc" id="id18">The CoreCLR GC</a></li>
</ul>
</li>
<li><a class="reference internal" href="#custom-gc-strategies" id="id19">Custom GC Strategies</a><ul>
<li><a class="reference internal" href="#collector-requirements" id="id20">Collector Requirements</a></li>
<li><a class="reference internal" href="#implementing-a-collector-plugin" id="id21">Implementing a collector plugin</a></li>
<li><a class="reference internal" href="#overview-of-available-features" id="id22">Overview of available features</a></li>
<li><a class="reference internal" href="#computing-stack-maps" id="id23">Computing stack maps</a></li>
<li><a class="reference internal" href="#initializing-roots-to-null-initroots" id="id24">Initializing roots to null: <code class="docutils literal"><span class="pre">InitRoots</span></code></a></li>
<li><a class="reference internal" href="#custom-lowering-of-intrinsics-customroots-customreadbarriers-and-customwritebarriers" id="id25">Custom lowering of intrinsics: <code class="docutils literal"><span class="pre">CustomRoots</span></code>, <code class="docutils literal"><span class="pre">CustomReadBarriers</span></code>, and <code class="docutils literal"><span class="pre">CustomWriteBarriers</span></code></a></li>
<li><a class="reference internal" href="#generating-safe-points-neededsafepoints" id="id26">Generating safe points: <code class="docutils literal"><span class="pre">NeededSafePoints</span></code></a></li>
<li><a class="reference internal" href="#emitting-assembly-code-gcmetadataprinter" id="id27">Emitting assembly code: <code class="docutils literal"><span class="pre">GCMetadataPrinter</span></code></a></li>
</ul>
</li>
<li><a class="reference internal" href="#references" id="id28">References</a></li>
</ul>
</div>
<div class="section" id="abstract">
<h2><a class="toc-backref" href="#id1">Abstract</a><a class="headerlink" href="#abstract" title="Permalink to this headline">¶</a></h2>
<p>This document covers how to integrate LLVM into a compiler for a language which
supports garbage collection. <strong>Note that LLVM itself does not provide a
garbage collector.</strong> You must provide your own.</p>
</div>
<div class="section" id="quick-start">
<h2><a class="toc-backref" href="#id2">Quick Start</a><a class="headerlink" href="#quick-start" title="Permalink to this headline">¶</a></h2>
<p>First, you should pick a collector strategy. LLVM includes a number of built
in ones, but you can also implement a loadable plugin with a custom definition.
Note that the collector strategy is a description of how LLVM should generate
code such that it interacts with your collector and runtime, not a description
of the collector itself.</p>
<p>Next, mark your generated functions as using your chosen collector strategy.
From c++, you can call:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="n">F</span><span class="p">.</span><span class="n">setGC</span><span class="p">(</span><span class="o"><</span><span class="n">collector</span> <span class="n">description</span> <span class="n">name</span><span class="o">></span><span class="p">);</span>
</pre></div>
</div>
<p>This will produce IR like the following fragment:</p>
<div class="highlight-llvm"><div class="highlight"><pre><span></span><span class="k">define</span> <span class="kt">void</span> <span class="vg">@foo</span><span class="p">()</span> <span class="k">gc</span> <span class="s">"<collector description name>"</span> <span class="p">{</span> <span class="p">...</span> <span class="p">}</span>
</pre></div>
</div>
<p>When generating LLVM IR for your functions, you will need to:</p>
<ul class="simple">
<li>Use <code class="docutils literal"><span class="pre">@llvm.gcread</span></code> and/or <code class="docutils literal"><span class="pre">@llvm.gcwrite</span></code> in place of standard load and
store instructions. These intrinsics are used to represent load and store
barriers. If you collector does not require such barriers, you can skip
this step.</li>
<li>Use the memory allocation routines provided by your garbage collector’s
runtime library.</li>
<li>If your collector requires them, generate type maps according to your
runtime’s binary interface. LLVM is not involved in the process. In
particular, the LLVM type system is not suitable for conveying such
information though the compiler.</li>
<li>Insert any coordination code required for interacting with your collector.
Many collectors require running application code to periodically check a
flag and conditionally call a runtime function. This is often referred to
as a safepoint poll.</li>
</ul>
<p>You will need to identify roots (i.e. references to heap objects your collector
needs to know about) in your generated IR, so that LLVM can encode them into
your final stack maps. Depending on the collector strategy chosen, this is
accomplished by using either the <code class="docutils literal"><span class="pre">@llvm.gcroot</span></code> intrinsics or an
<code class="docutils literal"><span class="pre">gc.statepoint</span></code> relocation sequence.</p>
<p>Don’t forget to create a root for each intermediate value that is generated when
evaluating an expression. In <code class="docutils literal"><span class="pre">h(f(),</span> <span class="pre">g())</span></code>, the result of <code class="docutils literal"><span class="pre">f()</span></code> could
easily be collected if evaluating <code class="docutils literal"><span class="pre">g()</span></code> triggers a collection.</p>
<p>Finally, you need to link your runtime library with the generated program
executable (for a static compiler) or ensure the appropriate symbols are
available for the runtime linker (for a JIT compiler).</p>
</div>
<div class="section" id="introduction">
<h2><a class="toc-backref" href="#id3">Introduction</a><a class="headerlink" href="#introduction" title="Permalink to this headline">¶</a></h2>
<div class="section" id="what-is-garbage-collection">
<h3><a class="toc-backref" href="#id4">What is Garbage Collection?</a><a class="headerlink" href="#what-is-garbage-collection" title="Permalink to this headline">¶</a></h3>
<p>Garbage collection is a widely used technique that frees the programmer from
having to know the lifetimes of heap objects, making software easier to produce
and maintain. Many programming languages rely on garbage collection for
automatic memory management. There are two primary forms of garbage collection:
conservative and accurate.</p>
<p>Conservative garbage collection often does not require any special support from
either the language or the compiler: it can handle non-type-safe programming
languages (such as C/C++) and does not require any special information from the
compiler. The <a class="reference external" href="http://www.hpl.hp.com/personal/Hans_Boehm/gc/">Boehm collector</a> is an example of a
state-of-the-art conservative collector.</p>
<p>Accurate garbage collection requires the ability to identify all pointers in the
program at run-time (which requires that the source-language be type-safe in
most cases). Identifying pointers at run-time requires compiler support to
locate all places that hold live pointer variables at run-time, including the
<a class="reference internal" href="#gcroot"><span class="std std-ref">processor stack and registers</span></a>.</p>
<p>Conservative garbage collection is attractive because it does not require any
special compiler support, but it does have problems. In particular, because the
conservative garbage collector cannot <em>know</em> that a particular word in the
machine is a pointer, it cannot move live objects in the heap (preventing the
use of compacting and generational GC algorithms) and it can occasionally suffer
from memory leaks due to integer values that happen to point to objects in the
program. In addition, some aggressive compiler transformations can break
conservative garbage collectors (though these seem rare in practice).</p>
<p>Accurate garbage collectors do not suffer from any of these problems, but they
can suffer from degraded scalar optimization of the program. In particular,
because the runtime must be able to identify and update all pointers active in
the program, some optimizations are less effective. In practice, however, the
locality and performance benefits of using aggressive garbage collection
techniques dominates any low-level losses.</p>
<p>This document describes the mechanisms and interfaces provided by LLVM to
support accurate garbage collection.</p>
</div>
<div class="section" id="goals-and-non-goals">
<h3><a class="toc-backref" href="#id5">Goals and non-goals</a><a class="headerlink" href="#goals-and-non-goals" title="Permalink to this headline">¶</a></h3>
<p>LLVM’s intermediate representation provides <a class="reference internal" href="#gc-intrinsics"><span class="std std-ref">garbage collection intrinsics</span></a> that offer support for a broad class of collector models. For
instance, the intrinsics permit:</p>
<ul class="simple">
<li>semi-space collectors</li>
<li>mark-sweep collectors</li>
<li>generational collectors</li>
<li>incremental collectors</li>
<li>concurrent collectors</li>
<li>cooperative collectors</li>
<li>reference counting</li>
</ul>
<p>We hope that the support built into the LLVM IR is sufficient to support a
broad class of garbage collected languages including Scheme, ML, Java, C#,
Perl, Python, Lua, Ruby, other scripting languages, and more.</p>
<p>Note that LLVM <strong>does not itself provide a garbage collector</strong> — this should
be part of your language’s runtime library. LLVM provides a framework for
describing the garbage collectors requirements to the compiler. In particular,
LLVM provides support for generating stack maps at call sites, polling for a
safepoint, and emitting load and store barriers. You can also extend LLVM -
possibly through a loadable <a class="reference internal" href="#plugin"><span class="std std-ref">code generation plugins</span></a> - to
generate code and data structures which conforms to the <em>binary interface</em>
specified by the <em>runtime library</em>. This is similar to the relationship between
LLVM and DWARF debugging info, for example. The difference primarily lies in
the lack of an established standard in the domain of garbage collection — thus
the need for a flexible extension mechanism.</p>
<p>The aspects of the binary interface with which LLVM’s GC support is
concerned are:</p>
<ul class="simple">
<li>Creation of GC safepoints within code where collection is allowed to execute
safely.</li>
<li>Computation of the stack map. For each safe point in the code, object
references within the stack frame must be identified so that the collector may
traverse and perhaps update them.</li>
<li>Write barriers when storing object references to the heap. These are commonly
used to optimize incremental scans in generational collectors.</li>
<li>Emission of read barriers when loading object references. These are useful
for interoperating with concurrent collectors.</li>
</ul>
<p>There are additional areas that LLVM does not directly address:</p>
<ul class="simple">
<li>Registration of global roots with the runtime.</li>
<li>Registration of stack map entries with the runtime.</li>
<li>The functions used by the program to allocate memory, trigger a collection,
etc.</li>
<li>Computation or compilation of type maps, or registration of them with the
runtime. These are used to crawl the heap for object references.</li>
</ul>
<p>In general, LLVM’s support for GC does not include features which can be
adequately addressed with other features of the IR and does not specify a
particular binary interface. On the plus side, this means that you should be
able to integrate LLVM with an existing runtime. On the other hand, it can
have the effect of leaving a lot of work for the developer of a novel
language. We try to mitigate this by providing built in collector strategy
descriptions that can work with many common collector designs and easy
extension points. If you don’t already have a specific binary interface
you need to support, we recommend trying to use one of these built in collector
strategies.</p>
</div>
</div>
<div class="section" id="llvm-ir-features">
<span id="gc-intrinsics"></span><h2><a class="toc-backref" href="#id6">LLVM IR Features</a><a class="headerlink" href="#llvm-ir-features" title="Permalink to this headline">¶</a></h2>
<p>This section describes the garbage collection facilities provided by the
<a class="reference internal" href="LangRef.html"><span class="doc">LLVM intermediate representation</span></a>. The exact behavior of these
IR features is specified by the selected <a class="reference internal" href="#plugin"><span class="std std-ref">GC strategy description</span></a>.</p>
<div class="section" id="specifying-gc-code-generation-gc">
<h3><a class="toc-backref" href="#id7">Specifying GC code generation: <code class="docutils literal"><span class="pre">gc</span> <span class="pre">"..."</span></code></a><a class="headerlink" href="#specifying-gc-code-generation-gc" title="Permalink to this headline">¶</a></h3>
<div class="highlight-text"><div class="highlight"><pre><span></span>define <returntype> @name(...) gc "name" { ... }
</pre></div>
</div>
<p>The <code class="docutils literal"><span class="pre">gc</span></code> function attribute is used to specify the desired GC strategy to the
compiler. Its programmatic equivalent is the <code class="docutils literal"><span class="pre">setGC</span></code> method of <code class="docutils literal"><span class="pre">Function</span></code>.</p>
<p>Setting <code class="docutils literal"><span class="pre">gc</span> <span class="pre">"name"</span></code> on a function triggers a search for a matching subclass
of GCStrategy. Some collector strategies are built in. You can add others
using either the loadable plugin mechanism, or by patching your copy of LLVM.
It is the selected GC strategy which defines the exact nature of the code
generated to support GC. If none is found, the compiler will raise an error.</p>
<p>Specifying the GC style on a per-function basis allows LLVM to link together
programs that use different garbage collection algorithms (or none at all).</p>
</div>
<div class="section" id="gcroot">
<span id="identifying-gc-roots-on-the-stack"></span><h3><a class="toc-backref" href="#id8">Identifying GC roots on the stack</a><a class="headerlink" href="#gcroot" title="Permalink to this headline">¶</a></h3>
<p>LLVM currently supports two different mechanisms for describing references in
compiled code at safepoints. <code class="docutils literal"><span class="pre">llvm.gcroot</span></code> is the older mechanism;
<code class="docutils literal"><span class="pre">gc.statepoint</span></code> has been added more recently. At the moment, you can choose
either implementation (on a per <a class="reference internal" href="#plugin"><span class="std std-ref">GC strategy</span></a> basis). Longer
term, we will probably either migrate away from <code class="docutils literal"><span class="pre">llvm.gcroot</span></code> entirely, or
substantially merge their implementations. Note that most new development
work is focused on <code class="docutils literal"><span class="pre">gc.statepoint</span></code>.</p>
<div class="section" id="using-gc-statepoint">
<h4><a class="toc-backref" href="#id9">Using <code class="docutils literal"><span class="pre">gc.statepoint</span></code></a><a class="headerlink" href="#using-gc-statepoint" title="Permalink to this headline">¶</a></h4>
<p><a class="reference internal" href="Statepoints.html"><span class="doc">This page</span></a> contains detailed documentation for
<code class="docutils literal"><span class="pre">gc.statepoint</span></code>.</p>
</div>
<div class="section" id="using-llvm-gcwrite">
<h4><a class="toc-backref" href="#id10">Using <code class="docutils literal"><span class="pre">llvm.gcwrite</span></code></a><a class="headerlink" href="#using-llvm-gcwrite" title="Permalink to this headline">¶</a></h4>
<div class="highlight-llvm"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="vg">@llvm.gcroot</span><span class="p">(</span><span class="k">i8</span><span class="p">**</span> <span class="nv">%ptrloc</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%metadata</span><span class="p">)</span>
</pre></div>
</div>
<p>The <code class="docutils literal"><span class="pre">llvm.gcroot</span></code> intrinsic is used to inform LLVM that a stack variable
references an object on the heap and is to be tracked for garbage collection.
The exact impact on generated code is specified by the Function’s selected
<a class="reference internal" href="#plugin"><span class="std std-ref">GC strategy</span></a>. All calls to <code class="docutils literal"><span class="pre">llvm.gcroot</span></code> <strong>must</strong> reside
inside the first basic block.</p>
<p>The first argument <strong>must</strong> be a value referring to an alloca instruction or a
bitcast of an alloca. The second contains a pointer to metadata that should be
associated with the pointer, and <strong>must</strong> be a constant or global value
address. If your target collector uses tags, use a null pointer for metadata.</p>
<p>A compiler which performs manual SSA construction <strong>must</strong> ensure that SSA
values representing GC references are stored in to the alloca passed to the
respective <code class="docutils literal"><span class="pre">gcroot</span></code> before every call site and reloaded after every call.
A compiler which uses mem2reg to raise imperative code using <code class="docutils literal"><span class="pre">alloca</span></code> into
SSA form need only add a call to <code class="docutils literal"><span class="pre">@llvm.gcroot</span></code> for those variables which
are pointers into the GC heap.</p>
<p>It is also important to mark intermediate values with <code class="docutils literal"><span class="pre">llvm.gcroot</span></code>. For
example, consider <code class="docutils literal"><span class="pre">h(f(),</span> <span class="pre">g())</span></code>. Beware leaking the result of <code class="docutils literal"><span class="pre">f()</span></code> in the
case that <code class="docutils literal"><span class="pre">g()</span></code> triggers a collection. Note, that stack variables must be
initialized and marked with <code class="docutils literal"><span class="pre">llvm.gcroot</span></code> in function’s prologue.</p>
<p>The <code class="docutils literal"><span class="pre">%metadata</span></code> argument can be used to avoid requiring heap objects to have
‘isa’ pointers or tag bits. [<a class="reference internal" href="#appel89">Appel89</a>, <a class="reference internal" href="#goldberg91">Goldberg91</a>, <a class="reference internal" href="#tolmach94">Tolmach94</a>] If specified,
its value will be tracked along with the location of the pointer in the stack
frame.</p>
<p>Consider the following fragment of Java code:</p>
<div class="highlight-java"><div class="highlight"><pre><span></span><span class="o">{</span>
<span class="n">Object</span> <span class="n">X</span><span class="o">;</span> <span class="c1">// A null-initialized reference to an object</span>
<span class="o">...</span>
<span class="o">}</span>
</pre></div>
</div>
<p>This block (which may be located in the middle of a function or in a loop nest),
could be compiled to this LLVM code:</p>
<div class="highlight-llvm"><div class="highlight"><pre><span></span><span class="nl">Entry:</span>
<span class="c">;; In the entry block for the function, allocate the</span>
<span class="c">;; stack space for X, which is an LLVM pointer.</span>
<span class="nv">%X</span> <span class="p">=</span> <span class="k">alloca</span> <span class="nv">%Object</span><span class="p">*</span>
<span class="c">;; Tell LLVM that the stack space is a stack root.</span>
<span class="c">;; Java has type-tags on objects, so we pass null as metadata.</span>
<span class="nv">%tmp</span> <span class="p">=</span> <span class="k">bitcast</span> <span class="nv">%Object</span><span class="p">**</span> <span class="nv">%X</span> <span class="k">to</span> <span class="k">i8</span><span class="p">**</span>
<span class="k">call</span> <span class="kt">void</span> <span class="vg">@llvm.gcroot</span><span class="p">(</span><span class="k">i8</span><span class="p">**</span> <span class="nv">%tmp</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="k">null</span><span class="p">)</span>
<span class="p">...</span>
<span class="c">;; "CodeBlock" is the block corresponding to the start</span>
<span class="c">;; of the scope above.</span>
<span class="nl">CodeBlock:</span>
<span class="c">;; Java null-initializes pointers.</span>
<span class="k">store</span> <span class="nv">%Object</span><span class="p">*</span> <span class="k">null</span><span class="p">,</span> <span class="nv">%Object</span><span class="p">**</span> <span class="nv">%X</span>
<span class="p">...</span>
<span class="c">;; As the pointer goes out of scope, store a null value into</span>
<span class="c">;; it, to indicate that the value is no longer live.</span>
<span class="k">store</span> <span class="nv">%Object</span><span class="p">*</span> <span class="k">null</span><span class="p">,</span> <span class="nv">%Object</span><span class="p">**</span> <span class="nv">%X</span>
<span class="p">...</span>
</pre></div>
</div>
</div>
</div>
<div class="section" id="reading-and-writing-references-in-the-heap">
<h3><a class="toc-backref" href="#id11">Reading and writing references in the heap</a><a class="headerlink" href="#reading-and-writing-references-in-the-heap" title="Permalink to this headline">¶</a></h3>
<p>Some collectors need to be informed when the mutator (the program that needs
garbage collection) either reads a pointer from or writes a pointer to a field
of a heap object. The code fragments inserted at these points are called <em>read
barriers</em> and <em>write barriers</em>, respectively. The amount of code that needs to
be executed is usually quite small and not on the critical path of any
computation, so the overall performance impact of the barrier is tolerable.</p>
<p>Barriers often require access to the <em>object pointer</em> rather than the <em>derived
pointer</em> (which is a pointer to the field within the object). Accordingly,
these intrinsics take both pointers as separate arguments for completeness. In
this snippet, <code class="docutils literal"><span class="pre">%object</span></code> is the object pointer, and <code class="docutils literal"><span class="pre">%derived</span></code> is the derived
pointer:</p>
<div class="highlight-llvm"><div class="highlight"><pre><span></span><span class="c">;; An array type.</span>
<span class="nv">%class.Array</span> <span class="p">=</span> <span class="k">type</span> <span class="p">{</span> <span class="nv">%class.Object</span><span class="p">,</span> <span class="k">i32</span><span class="p">,</span> <span class="p">[</span><span class="m">0</span> <span class="k">x</span> <span class="nv">%class.Object</span><span class="p">*]</span> <span class="p">}</span>
<span class="p">...</span>
<span class="c">;; Load the object pointer from a gcroot.</span>
<span class="nv">%object</span> <span class="p">=</span> <span class="k">load</span> <span class="nv">%class.Array</span><span class="p">**</span> <span class="nv">%object_addr</span>
<span class="c">;; Compute the derived pointer.</span>
<span class="nv">%derived</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="nv">%object</span><span class="p">,</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">2</span><span class="p">,</span> <span class="k">i32</span> <span class="nv">%n</span>
</pre></div>
</div>
<p>LLVM does not enforce this relationship between the object and derived pointer
(although a particular <a class="reference internal" href="#plugin"><span class="std std-ref">collector strategy</span></a> might). However, it
would be an unusual collector that violated it.</p>
<p>The use of these intrinsics is naturally optional if the target GC does not
require the corresponding barrier. The GC strategy used with such a collector
should replace the intrinsic calls with the corresponding <code class="docutils literal"><span class="pre">load</span></code> or
<code class="docutils literal"><span class="pre">store</span></code> instruction if they are used.</p>
<p>One known deficiency with the current design is that the barrier intrinsics do
not include the size or alignment of the underlying operation performed. It is
currently assumed that the operation is of pointer size and the alignment is
assumed to be the target machine’s default alignment.</p>
<div class="section" id="write-barrier-llvm-gcwrite">
<h4><a class="toc-backref" href="#id12">Write barrier: <code class="docutils literal"><span class="pre">llvm.gcwrite</span></code></a><a class="headerlink" href="#write-barrier-llvm-gcwrite" title="Permalink to this headline">¶</a></h4>
<div class="highlight-llvm"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="vg">@llvm.gcwrite</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%value</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%object</span><span class="p">,</span> <span class="k">i8</span><span class="p">**</span> <span class="nv">%derived</span><span class="p">)</span>
</pre></div>
</div>
<p>For write barriers, LLVM provides the <code class="docutils literal"><span class="pre">llvm.gcwrite</span></code> intrinsic function. It
has exactly the same semantics as a non-volatile <code class="docutils literal"><span class="pre">store</span></code> to the derived
pointer (the third argument). The exact code generated is specified by the
Function’s selected <a class="reference internal" href="#plugin"><span class="std std-ref">GC strategy</span></a>.</p>
<p>Many important algorithms require write barriers, including generational and
concurrent collectors. Additionally, write barriers could be used to implement
reference counting.</p>
</div>
<div class="section" id="read-barrier-llvm-gcread">
<h4><a class="toc-backref" href="#id13">Read barrier: <code class="docutils literal"><span class="pre">llvm.gcread</span></code></a><a class="headerlink" href="#read-barrier-llvm-gcread" title="Permalink to this headline">¶</a></h4>
<div class="highlight-llvm"><div class="highlight"><pre><span></span><span class="k">i8</span><span class="p">*</span> <span class="vg">@llvm.gcread</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%object</span><span class="p">,</span> <span class="k">i8</span><span class="p">**</span> <span class="nv">%derived</span><span class="p">)</span>
</pre></div>
</div>
<p>For read barriers, LLVM provides the <code class="docutils literal"><span class="pre">llvm.gcread</span></code> intrinsic function. It has
exactly the same semantics as a non-volatile <code class="docutils literal"><span class="pre">load</span></code> from the derived pointer
(the second argument). The exact code generated is specified by the Function’s
selected <a class="reference internal" href="#plugin"><span class="std std-ref">GC strategy</span></a>.</p>
<p>Read barriers are needed by fewer algorithms than write barriers, and may have a
greater performance impact since pointer reads are more frequent than writes.</p>
</div>
</div>
</div>
<div class="section" id="built-in-gc-strategies">
<span id="builtin-gc-strategies"></span><span id="plugin"></span><h2><a class="toc-backref" href="#id14">Built In GC Strategies</a><a class="headerlink" href="#built-in-gc-strategies" title="Permalink to this headline">¶</a></h2>
<p>LLVM includes built in support for several varieties of garbage collectors.</p>
<div class="section" id="the-shadow-stack-gc">
<h3><a class="toc-backref" href="#id15">The Shadow Stack GC</a><a class="headerlink" href="#the-shadow-stack-gc" title="Permalink to this headline">¶</a></h3>
<p>To use this collector strategy, mark your functions with:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="n">F</span><span class="p">.</span><span class="n">setGC</span><span class="p">(</span><span class="s">"shadow-stack"</span><span class="p">);</span>
</pre></div>
</div>
<p>Unlike many GC algorithms which rely on a cooperative code generator to compile
stack maps, this algorithm carefully maintains a linked list of stack roots
[<a class="reference internal" href="#henderson02"><span class="std std-ref">Henderson2002</span></a>]. This so-called “shadow stack” mirrors the
machine stack. Maintaining this data structure is slower than using a stack map
compiled into the executable as constant data, but has a significant portability
advantage because it requires no special support from the target code generator,
and does not require tricky platform-specific code to crawl the machine stack.</p>
<p>The tradeoff for this simplicity and portability is:</p>
<ul class="simple">
<li>High overhead per function call.</li>
<li>Not thread-safe.</li>
</ul>
<p>Still, it’s an easy way to get started. After your compiler and runtime are up
and running, writing a <a class="reference internal" href="#plugin"><span class="std std-ref">plugin</span></a> will allow you to take advantage
of <a class="reference internal" href="#collector-algos"><span class="std std-ref">more advanced GC features</span></a> of LLVM in order to
improve performance.</p>
<p>The shadow stack doesn’t imply a memory allocation algorithm. A semispace
collector or building atop <code class="docutils literal"><span class="pre">malloc</span></code> are great places to start, and can be
implemented with very little code.</p>
<p>When it comes time to collect, however, your runtime needs to traverse the stack
roots, and for this it needs to integrate with the shadow stack. Luckily, doing
so is very simple. (This code is heavily commented to help you understand the
data structure, but there are only 20 lines of meaningful code.)</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="c1">/// @brief The map for a single function's stack frame. One of these is</span>
<span class="c1">/// compiled as constant data into the executable for each function.</span>
<span class="c1">///</span>
<span class="c1">/// Storage of metadata values is elided if the %metadata parameter to</span>
<span class="c1">/// @llvm.gcroot is null.</span>
<span class="k">struct</span> <span class="n">FrameMap</span> <span class="p">{</span>
<span class="kt">int32_t</span> <span class="n">NumRoots</span><span class="p">;</span> <span class="c1">//< Number of roots in stack frame.</span>
<span class="kt">int32_t</span> <span class="n">NumMeta</span><span class="p">;</span> <span class="c1">//< Number of metadata entries. May be < NumRoots.</span>
<span class="k">const</span> <span class="kt">void</span> <span class="o">*</span><span class="n">Meta</span><span class="p">[</span><span class="mi">0</span><span class="p">];</span> <span class="c1">//< Metadata for each root.</span>
<span class="p">};</span>
<span class="c1">/// @brief A link in the dynamic shadow stack. One of these is embedded in</span>
<span class="c1">/// the stack frame of each function on the call stack.</span>
<span class="k">struct</span> <span class="n">StackEntry</span> <span class="p">{</span>
<span class="n">StackEntry</span> <span class="o">*</span><span class="n">Next</span><span class="p">;</span> <span class="c1">//< Link to next stack entry (the caller's).</span>
<span class="k">const</span> <span class="n">FrameMap</span> <span class="o">*</span><span class="n">Map</span><span class="p">;</span> <span class="c1">//< Pointer to constant FrameMap.</span>
<span class="kt">void</span> <span class="o">*</span><span class="n">Roots</span><span class="p">[</span><span class="mi">0</span><span class="p">];</span> <span class="c1">//< Stack roots (in-place array).</span>
<span class="p">};</span>
<span class="c1">/// @brief The head of the singly-linked list of StackEntries. Functions push</span>
<span class="c1">/// and pop onto this in their prologue and epilogue.</span>
<span class="c1">///</span>
<span class="c1">/// Since there is only a global list, this technique is not threadsafe.</span>
<span class="n">StackEntry</span> <span class="o">*</span><span class="n">llvm_gc_root_chain</span><span class="p">;</span>
<span class="c1">/// @brief Calls Visitor(root, meta) for each GC root on the stack.</span>
<span class="c1">/// root and meta are exactly the values passed to</span>
<span class="c1">/// @llvm.gcroot.</span>
<span class="c1">///</span>
<span class="c1">/// Visitor could be a function to recursively mark live objects. Or it</span>
<span class="c1">/// might copy them to another heap or generation.</span>
<span class="c1">///</span>
<span class="c1">/// @param Visitor A function to invoke for every GC root on the stack.</span>
<span class="kt">void</span> <span class="nf">visitGCRoots</span><span class="p">(</span><span class="kt">void</span> <span class="p">(</span><span class="o">*</span><span class="n">Visitor</span><span class="p">)(</span><span class="kt">void</span> <span class="o">**</span><span class="n">Root</span><span class="p">,</span> <span class="k">const</span> <span class="kt">void</span> <span class="o">*</span><span class="n">Meta</span><span class="p">))</span> <span class="p">{</span>
<span class="k">for</span> <span class="p">(</span><span class="n">StackEntry</span> <span class="o">*</span><span class="n">R</span> <span class="o">=</span> <span class="n">llvm_gc_root_chain</span><span class="p">;</span> <span class="n">R</span><span class="p">;</span> <span class="n">R</span> <span class="o">=</span> <span class="n">R</span><span class="o">-></span><span class="n">Next</span><span class="p">)</span> <span class="p">{</span>
<span class="kt">unsigned</span> <span class="n">i</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span>
<span class="c1">// For roots [0, NumMeta), the metadata pointer is in the FrameMap.</span>
<span class="k">for</span> <span class="p">(</span><span class="kt">unsigned</span> <span class="n">e</span> <span class="o">=</span> <span class="n">R</span><span class="o">-></span><span class="n">Map</span><span class="o">-></span><span class="n">NumMeta</span><span class="p">;</span> <span class="n">i</span> <span class="o">!=</span> <span class="n">e</span><span class="p">;</span> <span class="o">++</span><span class="n">i</span><span class="p">)</span>
<span class="n">Visitor</span><span class="p">(</span><span class="o">&</span><span class="n">R</span><span class="o">-></span><span class="n">Roots</span><span class="p">[</span><span class="n">i</span><span class="p">],</span> <span class="n">R</span><span class="o">-></span><span class="n">Map</span><span class="o">-></span><span class="n">Meta</span><span class="p">[</span><span class="n">i</span><span class="p">]);</span>
<span class="c1">// For roots [NumMeta, NumRoots), the metadata pointer is null.</span>
<span class="k">for</span> <span class="p">(</span><span class="kt">unsigned</span> <span class="n">e</span> <span class="o">=</span> <span class="n">R</span><span class="o">-></span><span class="n">Map</span><span class="o">-></span><span class="n">NumRoots</span><span class="p">;</span> <span class="n">i</span> <span class="o">!=</span> <span class="n">e</span><span class="p">;</span> <span class="o">++</span><span class="n">i</span><span class="p">)</span>
<span class="n">Visitor</span><span class="p">(</span><span class="o">&</span><span class="n">R</span><span class="o">-></span><span class="n">Roots</span><span class="p">[</span><span class="n">i</span><span class="p">],</span> <span class="nb">NULL</span><span class="p">);</span>
<span class="p">}</span>
<span class="p">}</span>
</pre></div>
</div>
</div>
<div class="section" id="the-erlang-and-ocaml-gcs">
<h3><a class="toc-backref" href="#id16">The ‘Erlang’ and ‘Ocaml’ GCs</a><a class="headerlink" href="#the-erlang-and-ocaml-gcs" title="Permalink to this headline">¶</a></h3>
<p>LLVM ships with two example collectors which leverage the <code class="docutils literal"><span class="pre">gcroot</span></code>
mechanisms. To our knowledge, these are not actually used by any language
runtime, but they do provide a reasonable starting point for someone interested
in writing an <code class="docutils literal"><span class="pre">gcroot</span></code> compatible GC plugin. In particular, these are the
only in tree examples of how to produce a custom binary stack map format using
a <code class="docutils literal"><span class="pre">gcroot</span></code> strategy.</p>
<p>As there names imply, the binary format produced is intended to model that
used by the Erlang and OCaml compilers respectively.</p>
</div>
<div class="section" id="the-statepoint-example-gc">
<span id="statepoint-example-gc"></span><h3><a class="toc-backref" href="#id17">The Statepoint Example GC</a><a class="headerlink" href="#the-statepoint-example-gc" title="Permalink to this headline">¶</a></h3>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="n">F</span><span class="p">.</span><span class="n">setGC</span><span class="p">(</span><span class="s">"statepoint-example"</span><span class="p">);</span>
</pre></div>
</div>
<p>This GC provides an example of how one might use the infrastructure provided
by <code class="docutils literal"><span class="pre">gc.statepoint</span></code>. This example GC is compatible with the
<a class="reference internal" href="Statepoints.html#placesafepoints"><span class="std std-ref">PlaceSafepoints</span></a> and <a class="reference internal" href="Statepoints.html#rewritestatepointsforgc"><span class="std std-ref">RewriteStatepointsForGC</span></a> utility passes
which simplify <code class="docutils literal"><span class="pre">gc.statepoint</span></code> sequence insertion. If you need to build a
custom GC strategy around the <code class="docutils literal"><span class="pre">gc.statepoints</span></code> mechanisms, it is recommended
that you use this one as a starting point.</p>
<p>This GC strategy does not support read or write barriers. As a result, these
intrinsics are lowered to normal loads and stores.</p>
<p>The stack map format generated by this GC strategy can be found in the
<a class="reference internal" href="StackMaps.html#stackmap-section"><span class="std std-ref">Stack Map Section</span></a> using a format documented <a class="reference internal" href="Statepoints.html#statepoint-stackmap-format"><span class="std std-ref">here</span></a>. This format is intended to be the standard
format supported by LLVM going forward.</p>
</div>
<div class="section" id="the-coreclr-gc">
<h3><a class="toc-backref" href="#id18">The CoreCLR GC</a><a class="headerlink" href="#the-coreclr-gc" title="Permalink to this headline">¶</a></h3>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="n">F</span><span class="p">.</span><span class="n">setGC</span><span class="p">(</span><span class="s">"coreclr"</span><span class="p">);</span>
</pre></div>
</div>
<p>This GC leverages the <code class="docutils literal"><span class="pre">gc.statepoint</span></code> mechanism to support the
<a class="reference external" href="https://github.com/dotnet/coreclr">CoreCLR</a> runtime.</p>
<p>Support for this GC strategy is a work in progress. This strategy will
differ from
<a class="reference internal" href="#statepoint-example-gc"><span class="std std-ref">statepoint-example GC</span></a> strategy in
certain aspects like:</p>
<ul class="simple">
<li>Base-pointers of interior pointers are not explicitly
tracked and reported.</li>
<li>A different format is used for encoding stack maps.</li>
<li>Safe-point polls are only needed before loop-back edges
and before tail-calls (not needed at function-entry).</li>
</ul>
</div>
</div>
<div class="section" id="custom-gc-strategies">
<h2><a class="toc-backref" href="#id19">Custom GC Strategies</a><a class="headerlink" href="#custom-gc-strategies" title="Permalink to this headline">¶</a></h2>
<p>If none of the built in GC strategy descriptions met your needs above, you will
need to define a custom GCStrategy and possibly, a custom LLVM pass to perform
lowering. Your best example of where to start defining a custom GCStrategy
would be to look at one of the built in strategies.</p>
<p>You may be able to structure this additional code as a loadable plugin library.
Loadable plugins are sufficient if all you need is to enable a different
combination of built in functionality, but if you need to provide a custom
lowering pass, you will need to build a patched version of LLVM. If you think
you need a patched build, please ask for advice on llvm-dev. There may be an
easy way we can extend the support to make it work for your use case without
requiring a custom build.</p>
<div class="section" id="collector-requirements">
<h3><a class="toc-backref" href="#id20">Collector Requirements</a><a class="headerlink" href="#collector-requirements" title="Permalink to this headline">¶</a></h3>
<p>You should be able to leverage any existing collector library that includes the following elements:</p>
<ol class="arabic simple">
<li>A memory allocator which exposes an allocation function your compiled
code can call.</li>
<li>A binary format for the stack map. A stack map describes the location
of references at a safepoint and is used by precise collectors to identify
references within a stack frame on the machine stack. Note that collectors
which conservatively scan the stack don’t require such a structure.</li>
<li>A stack crawler to discover functions on the call stack, and enumerate the
references listed in the stack map for each call site.</li>
<li>A mechanism for identifying references in global locations (e.g. global
variables).</li>
<li>If you collector requires them, an LLVM IR implementation of your collectors
load and store barriers. Note that since many collectors don’t require
barriers at all, LLVM defaults to lowering such barriers to normal loads
and stores unless you arrange otherwise.</li>
</ol>
</div>
<div class="section" id="implementing-a-collector-plugin">
<h3><a class="toc-backref" href="#id21">Implementing a collector plugin</a><a class="headerlink" href="#implementing-a-collector-plugin" title="Permalink to this headline">¶</a></h3>
<p>User code specifies which GC code generation to use with the <code class="docutils literal"><span class="pre">gc</span></code> function
attribute or, equivalently, with the <code class="docutils literal"><span class="pre">setGC</span></code> method of <code class="docutils literal"><span class="pre">Function</span></code>.</p>
<p>To implement a GC plugin, it is necessary to subclass <code class="docutils literal"><span class="pre">llvm::GCStrategy</span></code>,
which can be accomplished in a few lines of boilerplate code. LLVM’s
infrastructure provides access to several important algorithms. For an
uncontroversial collector, all that remains may be to compile LLVM’s computed
stack map to assembly code (using the binary representation expected by the
runtime library). This can be accomplished in about 100 lines of code.</p>
<p>This is not the appropriate place to implement a garbage collected heap or a
garbage collector itself. That code should exist in the language’s runtime
library. The compiler plugin is responsible for generating code which conforms
to the binary interface defined by library, most essentially the <a class="reference internal" href="#stack-map"><span class="std std-ref">stack map</span></a>.</p>
<p>To subclass <code class="docutils literal"><span class="pre">llvm::GCStrategy</span></code> and register it with the compiler:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="c1">// lib/MyGC/MyGC.cpp - Example LLVM GC plugin</span>
<span class="cp">#include</span> <span class="cpf">"llvm/CodeGen/GCStrategy.h"</span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf">"llvm/CodeGen/GCMetadata.h"</span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf">"llvm/Support/Compiler.h"</span><span class="cp"></span>
<span class="k">using</span> <span class="k">namespace</span> <span class="n">llvm</span><span class="p">;</span>
<span class="k">namespace</span> <span class="p">{</span>
<span class="k">class</span> <span class="nc">LLVM_LIBRARY_VISIBILITY</span> <span class="nl">MyGC</span> <span class="p">:</span> <span class="k">public</span> <span class="n">GCStrategy</span> <span class="p">{</span>
<span class="k">public</span><span class="o">:</span>
<span class="n">MyGC</span><span class="p">()</span> <span class="p">{}</span>
<span class="p">};</span>
<span class="n">GCRegistry</span><span class="o">::</span><span class="n">Add</span><span class="o"><</span><span class="n">MyGC</span><span class="o">></span>
<span class="n">X</span><span class="p">(</span><span class="s">"mygc"</span><span class="p">,</span> <span class="s">"My bespoke garbage collector."</span><span class="p">);</span>
<span class="p">}</span>
</pre></div>
</div>
<p>This boilerplate collector does nothing. More specifically:</p>
<ul class="simple">
<li><code class="docutils literal"><span class="pre">llvm.gcread</span></code> calls are replaced with the corresponding <code class="docutils literal"><span class="pre">load</span></code>
instruction.</li>
<li><code class="docutils literal"><span class="pre">llvm.gcwrite</span></code> calls are replaced with the corresponding <code class="docutils literal"><span class="pre">store</span></code>
instruction.</li>
<li>No safe points are added to the code.</li>
<li>The stack map is not compiled into the executable.</li>
</ul>
<p>Using the LLVM makefiles, this code
can be compiled as a plugin using a simple makefile:</p>
<div class="highlight-make"><div class="highlight"><pre><span></span><span class="c"># lib/MyGC/Makefile</span>
<span class="nv">LEVEL</span> <span class="o">:=</span> ../..
<span class="nv">LIBRARYNAME</span> <span class="o">=</span> MyGC
<span class="nv">LOADABLE_MODULE</span> <span class="o">=</span> <span class="m">1</span>
<span class="cp">include $(LEVEL)/Makefile.common</span>
</pre></div>
</div>
<p>Once the plugin is compiled, code using it may be compiled using <code class="docutils literal"><span class="pre">llc</span>
<span class="pre">-load=MyGC.so</span></code> (though MyGC.so may have some other platform-specific
extension):</p>
<div class="highlight-default"><div class="highlight"><pre><span></span>$ cat sample.ll
define void @f() gc "mygc" {
entry:
ret void
}
$ llvm-as < sample.ll | llc -load=MyGC.so
</pre></div>
</div>
<p>It is also possible to statically link the collector plugin into tools, such as
a language-specific compiler front-end.</p>
</div>
<div class="section" id="overview-of-available-features">
<span id="collector-algos"></span><h3><a class="toc-backref" href="#id22">Overview of available features</a><a class="headerlink" href="#overview-of-available-features" title="Permalink to this headline">¶</a></h3>
<p><code class="docutils literal"><span class="pre">GCStrategy</span></code> provides a range of features through which a plugin may do useful
work. Some of these are callbacks, some are algorithms that can be enabled,
disabled, or customized. This matrix summarizes the supported (and planned)
features and correlates them with the collection techniques which typically
require them.</p>
<table border="1" class="docutils">
<colgroup>
<col width="14%" />
<col width="7%" />
<col width="9%" />
<col width="11%" />
<col width="8%" />
<col width="10%" />
<col width="15%" />
<col width="11%" />
<col width="14%" />
</colgroup>
<thead valign="bottom">
<tr class="row-odd"><th class="head">Algorithm</th>
<th class="head">Done</th>
<th class="head">Shadow
stack</th>
<th class="head">refcount</th>
<th class="head">mark-
sweep</th>
<th class="head">copying</th>
<th class="head">incremental</th>
<th class="head">threaded</th>
<th class="head">concurrent</th>
</tr>
</thead>
<tbody valign="top">
<tr class="row-even"><td>stack map</td>
<td>✔</td>
<td> </td>
<td> </td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
</tr>
<tr class="row-odd"><td>initialize
roots</td>
<td>✔</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
</tr>
<tr class="row-even"><td>derived
pointers</td>
<td>NO</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td><strong>N</strong>*</td>
<td><strong>N</strong>*</td>
</tr>
<tr class="row-odd"><td><strong>custom
lowering</strong></td>
<td>✔</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr class="row-even"><td><em>gcroot</em></td>
<td>✔</td>
<td>✘</td>
<td>✘</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr class="row-odd"><td><em>gcwrite</em></td>
<td>✔</td>
<td> </td>
<td>✘</td>
<td> </td>
<td> </td>
<td>✘</td>
<td> </td>
<td>✘</td>
</tr>
<tr class="row-even"><td><em>gcread</em></td>
<td>✔</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td>✘</td>
</tr>
<tr class="row-odd"><td><strong>safe
points</strong></td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr class="row-even"><td><em>in
calls</em></td>
<td>✔</td>
<td> </td>
<td> </td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
</tr>
<tr class="row-odd"><td><em>before
calls</em></td>
<td>✔</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td>✘</td>
<td>✘</td>
</tr>
<tr class="row-even"><td><em>for
loops</em></td>
<td>NO</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td><strong>N</strong></td>
<td><strong>N</strong></td>
</tr>
<tr class="row-odd"><td><em>before
escape</em></td>
<td>✔</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td>✘</td>
<td>✘</td>
</tr>
<tr class="row-even"><td>emit code
at safe
points</td>
<td>NO</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td><strong>N</strong></td>
<td><strong>N</strong></td>
</tr>
<tr class="row-odd"><td><strong>output</strong></td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr class="row-even"><td><em>assembly</em></td>
<td>✔</td>
<td> </td>
<td> </td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
<td>✘</td>
</tr>
<tr class="row-odd"><td><em>JIT</em></td>
<td>NO</td>
<td> </td>
<td> </td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
</tr>
<tr class="row-even"><td><em>obj</em></td>
<td>NO</td>
<td> </td>
<td> </td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
</tr>
<tr class="row-odd"><td>live
analysis</td>
<td>NO</td>
<td> </td>
<td> </td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
</tr>
<tr class="row-even"><td>register
map</td>
<td>NO</td>
<td> </td>
<td> </td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
<td><strong>?</strong></td>
</tr>
<tr class="row-odd"><td colspan="9">* Derived pointers only pose a hasard to copying collections.</td>
</tr>
<tr class="row-even"><td colspan="9"><strong>?</strong> denotes a feature which could be utilized if available.</td>
</tr>
</tbody>
</table>
<p>To be clear, the collection techniques above are defined as:</p>
<dl class="docutils">
<dt>Shadow Stack</dt>
<dd>The mutator carefully maintains a linked list of stack roots.</dd>
<dt>Reference Counting</dt>
<dd>The mutator maintains a reference count for each object and frees an object
when its count falls to zero.</dd>
<dt>Mark-Sweep</dt>
<dd>When the heap is exhausted, the collector marks reachable objects starting
from the roots, then deallocates unreachable objects in a sweep phase.</dd>
<dt>Copying</dt>
<dd>As reachability analysis proceeds, the collector copies objects from one heap
area to another, compacting them in the process. Copying collectors enable
highly efficient “bump pointer” allocation and can improve locality of
reference.</dd>
<dt>Incremental</dt>
<dd>(Including generational collectors.) Incremental collectors generally have all
the properties of a copying collector (regardless of whether the mature heap
is compacting), but bring the added complexity of requiring write barriers.</dd>
<dt>Threaded</dt>
<dd>Denotes a multithreaded mutator; the collector must still stop the mutator
(“stop the world”) before beginning reachability analysis. Stopping a
multithreaded mutator is a complicated problem. It generally requires highly
platform-specific code in the runtime, and the production of carefully
designed machine code at safe points.</dd>
<dt>Concurrent</dt>
<dd>In this technique, the mutator and the collector run concurrently, with the
goal of eliminating pause times. In a <em>cooperative</em> collector, the mutator
further aids with collection should a pause occur, allowing collection to take
advantage of multiprocessor hosts. The “stop the world” problem of threaded
collectors is generally still present to a limited extent. Sophisticated
marking algorithms are necessary. Read barriers may be necessary.</dd>
</dl>
<p>As the matrix indicates, LLVM’s garbage collection infrastructure is already
suitable for a wide variety of collectors, but does not currently extend to
multithreaded programs. This will be added in the future as there is
interest.</p>
</div>
<div class="section" id="computing-stack-maps">
<span id="stack-map"></span><h3><a class="toc-backref" href="#id23">Computing stack maps</a><a class="headerlink" href="#computing-stack-maps" title="Permalink to this headline">¶</a></h3>
<p>LLVM automatically computes a stack map. One of the most important features
of a <code class="docutils literal"><span class="pre">GCStrategy</span></code> is to compile this information into the executable in
the binary representation expected by the runtime library.</p>
<p>The stack map consists of the location and identity of each GC root in the
each function in the module. For each root:</p>
<ul class="simple">
<li><code class="docutils literal"><span class="pre">RootNum</span></code>: The index of the root.</li>
<li><code class="docutils literal"><span class="pre">StackOffset</span></code>: The offset of the object relative to the frame pointer.</li>
<li><code class="docutils literal"><span class="pre">RootMetadata</span></code>: The value passed as the <code class="docutils literal"><span class="pre">%metadata</span></code> parameter to the
<code class="docutils literal"><span class="pre">@llvm.gcroot</span></code> intrinsic.</li>
</ul>
<p>Also, for the function as a whole:</p>
<ul class="simple">
<li><dl class="first docutils">
<dt><code class="docutils literal"><span class="pre">getFrameSize()</span></code>: The overall size of the function’s initial stack frame,</dt>
<dd>not accounting for any dynamic allocation.</dd>
</dl>
</li>
<li><code class="docutils literal"><span class="pre">roots_size()</span></code>: The count of roots in the function.</li>
</ul>
<p>To access the stack map, use <code class="docutils literal"><span class="pre">GCFunctionMetadata::roots_begin()</span></code> and
-<code class="docutils literal"><span class="pre">end()</span></code> from the <a class="reference internal" href="#assembly"><span class="std std-ref">GCMetadataPrinter</span></a>:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="k">for</span> <span class="p">(</span><span class="n">iterator</span> <span class="n">I</span> <span class="o">=</span> <span class="n">begin</span><span class="p">(),</span> <span class="n">E</span> <span class="o">=</span> <span class="n">end</span><span class="p">();</span> <span class="n">I</span> <span class="o">!=</span> <span class="n">E</span><span class="p">;</span> <span class="o">++</span><span class="n">I</span><span class="p">)</span> <span class="p">{</span>
<span class="n">GCFunctionInfo</span> <span class="o">*</span><span class="n">FI</span> <span class="o">=</span> <span class="o">*</span><span class="n">I</span><span class="p">;</span>
<span class="kt">unsigned</span> <span class="n">FrameSize</span> <span class="o">=</span> <span class="n">FI</span><span class="o">-></span><span class="n">getFrameSize</span><span class="p">();</span>
<span class="kt">size_t</span> <span class="n">RootCount</span> <span class="o">=</span> <span class="n">FI</span><span class="o">-></span><span class="n">roots_size</span><span class="p">();</span>
<span class="k">for</span> <span class="p">(</span><span class="n">GCFunctionInfo</span><span class="o">::</span><span class="n">roots_iterator</span> <span class="n">RI</span> <span class="o">=</span> <span class="n">FI</span><span class="o">-></span><span class="n">roots_begin</span><span class="p">(),</span>
<span class="n">RE</span> <span class="o">=</span> <span class="n">FI</span><span class="o">-></span><span class="n">roots_end</span><span class="p">();</span>
<span class="n">RI</span> <span class="o">!=</span> <span class="n">RE</span><span class="p">;</span> <span class="o">++</span><span class="n">RI</span><span class="p">)</span> <span class="p">{</span>
<span class="kt">int</span> <span class="n">RootNum</span> <span class="o">=</span> <span class="n">RI</span><span class="o">-></span><span class="n">Num</span><span class="p">;</span>
<span class="kt">int</span> <span class="n">RootStackOffset</span> <span class="o">=</span> <span class="n">RI</span><span class="o">-></span><span class="n">StackOffset</span><span class="p">;</span>
<span class="n">Constant</span> <span class="o">*</span><span class="n">RootMetadata</span> <span class="o">=</span> <span class="n">RI</span><span class="o">-></span><span class="n">Metadata</span><span class="p">;</span>
<span class="p">}</span>
<span class="p">}</span>
</pre></div>
</div>
<p>If the <code class="docutils literal"><span class="pre">llvm.gcroot</span></code> intrinsic is eliminated before code generation by a
custom lowering pass, LLVM will compute an empty stack map. This may be useful
for collector plugins which implement reference counting or a shadow stack.</p>
</div>
<div class="section" id="initializing-roots-to-null-initroots">
<span id="init-roots"></span><h3><a class="toc-backref" href="#id24">Initializing roots to null: <code class="docutils literal"><span class="pre">InitRoots</span></code></a><a class="headerlink" href="#initializing-roots-to-null-initroots" title="Permalink to this headline">¶</a></h3>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="n">MyGC</span><span class="o">::</span><span class="n">MyGC</span><span class="p">()</span> <span class="p">{</span>
<span class="n">InitRoots</span> <span class="o">=</span> <span class="nb">true</span><span class="p">;</span>
<span class="p">}</span>
</pre></div>
</div>
<p>When set, LLVM will automatically initialize each root to <code class="docutils literal"><span class="pre">null</span></code> upon entry to
the function. This prevents the GC’s sweep phase from visiting uninitialized
pointers, which will almost certainly cause it to crash. This initialization
occurs before custom lowering, so the two may be used together.</p>
<p>Since LLVM does not yet compute liveness information, there is no means of
distinguishing an uninitialized stack root from an initialized one. Therefore,
this feature should be used by all GC plugins. It is enabled by default.</p>
</div>
<div class="section" id="custom-lowering-of-intrinsics-customroots-customreadbarriers-and-customwritebarriers">
<h3><a class="toc-backref" href="#id25">Custom lowering of intrinsics: <code class="docutils literal"><span class="pre">CustomRoots</span></code>, <code class="docutils literal"><span class="pre">CustomReadBarriers</span></code>, and <code class="docutils literal"><span class="pre">CustomWriteBarriers</span></code></a><a class="headerlink" href="#custom-lowering-of-intrinsics-customroots-customreadbarriers-and-customwritebarriers" title="Permalink to this headline">¶</a></h3>
<p>For GCs which use barriers or unusual treatment of stack roots, these
flags allow the collector to perform arbitrary transformations of the
LLVM IR:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">MyGC</span> <span class="o">:</span> <span class="k">public</span> <span class="n">GCStrategy</span> <span class="p">{</span>
<span class="k">public</span><span class="o">:</span>
<span class="n">MyGC</span><span class="p">()</span> <span class="p">{</span>
<span class="n">CustomRoots</span> <span class="o">=</span> <span class="nb">true</span><span class="p">;</span>
<span class="n">CustomReadBarriers</span> <span class="o">=</span> <span class="nb">true</span><span class="p">;</span>
<span class="n">CustomWriteBarriers</span> <span class="o">=</span> <span class="nb">true</span><span class="p">;</span>
<span class="p">}</span>
<span class="p">};</span>
</pre></div>
</div>
<p>If any of these flags are set, LLVM suppresses its default lowering for
the corresponding intrinsics. Instead, you must provide a custom Pass
which lowers the intrinsics as desired. If you have opted in to custom
lowering of a particular intrinsic your pass <strong>must</strong> eliminate all
instances of the corresponding intrinsic in functions which opt in to
your GC. The best example of such a pass is the ShadowStackGC and it’s
ShadowStackGCLowering pass.</p>
<p>There is currently no way to register such a custom lowering pass
without building a custom copy of LLVM.</p>
</div>
<div class="section" id="generating-safe-points-neededsafepoints">
<span id="safe-points"></span><h3><a class="toc-backref" href="#id26">Generating safe points: <code class="docutils literal"><span class="pre">NeededSafePoints</span></code></a><a class="headerlink" href="#generating-safe-points-neededsafepoints" title="Permalink to this headline">¶</a></h3>
<p>LLVM can compute four kinds of safe points:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="k">namespace</span> <span class="n">GC</span> <span class="p">{</span>
<span class="c1">/// PointKind - The type of a collector-safe point.</span>
<span class="c1">///</span>
<span class="k">enum</span> <span class="n">PointKind</span> <span class="p">{</span>
<span class="n">Loop</span><span class="p">,</span> <span class="c1">//< Instr is a loop (backwards branch).</span>
<span class="n">Return</span><span class="p">,</span> <span class="c1">//< Instr is a return instruction.</span>
<span class="n">PreCall</span><span class="p">,</span> <span class="c1">//< Instr is a call instruction.</span>
<span class="n">PostCall</span> <span class="c1">//< Instr is the return address of a call.</span>
<span class="p">};</span>
<span class="p">}</span>
</pre></div>
</div>
<p>A collector can request any combination of the four by setting the
<code class="docutils literal"><span class="pre">NeededSafePoints</span></code> mask:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="n">MyGC</span><span class="o">::</span><span class="n">MyGC</span><span class="p">()</span> <span class="p">{</span>
<span class="n">NeededSafePoints</span> <span class="o">=</span> <span class="mi">1</span> <span class="o"><<</span> <span class="n">GC</span><span class="o">::</span><span class="n">Loop</span>
<span class="o">|</span> <span class="mi">1</span> <span class="o"><<</span> <span class="n">GC</span><span class="o">::</span><span class="n">Return</span>
<span class="o">|</span> <span class="mi">1</span> <span class="o"><<</span> <span class="n">GC</span><span class="o">::</span><span class="n">PreCall</span>
<span class="o">|</span> <span class="mi">1</span> <span class="o"><<</span> <span class="n">GC</span><span class="o">::</span><span class="n">PostCall</span><span class="p">;</span>
<span class="p">}</span>
</pre></div>
</div>
<p>It can then use the following routines to access safe points.</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="k">for</span> <span class="p">(</span><span class="n">iterator</span> <span class="n">I</span> <span class="o">=</span> <span class="n">begin</span><span class="p">(),</span> <span class="n">E</span> <span class="o">=</span> <span class="n">end</span><span class="p">();</span> <span class="n">I</span> <span class="o">!=</span> <span class="n">E</span><span class="p">;</span> <span class="o">++</span><span class="n">I</span><span class="p">)</span> <span class="p">{</span>
<span class="n">GCFunctionInfo</span> <span class="o">*</span><span class="n">MD</span> <span class="o">=</span> <span class="o">*</span><span class="n">I</span><span class="p">;</span>
<span class="kt">size_t</span> <span class="n">PointCount</span> <span class="o">=</span> <span class="n">MD</span><span class="o">-></span><span class="n">size</span><span class="p">();</span>
<span class="k">for</span> <span class="p">(</span><span class="n">GCFunctionInfo</span><span class="o">::</span><span class="n">iterator</span> <span class="n">PI</span> <span class="o">=</span> <span class="n">MD</span><span class="o">-></span><span class="n">begin</span><span class="p">(),</span>
<span class="n">PE</span> <span class="o">=</span> <span class="n">MD</span><span class="o">-></span><span class="n">end</span><span class="p">();</span> <span class="n">PI</span> <span class="o">!=</span> <span class="n">PE</span><span class="p">;</span> <span class="o">++</span><span class="n">PI</span><span class="p">)</span> <span class="p">{</span>
<span class="n">GC</span><span class="o">::</span><span class="n">PointKind</span> <span class="n">PointKind</span> <span class="o">=</span> <span class="n">PI</span><span class="o">-></span><span class="n">Kind</span><span class="p">;</span>
<span class="kt">unsigned</span> <span class="n">PointNum</span> <span class="o">=</span> <span class="n">PI</span><span class="o">-></span><span class="n">Num</span><span class="p">;</span>
<span class="p">}</span>
<span class="p">}</span>
</pre></div>
</div>
<p>Almost every collector requires <code class="docutils literal"><span class="pre">PostCall</span></code> safe points, since these correspond
to the moments when the function is suspended during a call to a subroutine.</p>
<p>Threaded programs generally require <code class="docutils literal"><span class="pre">Loop</span></code> safe points to guarantee that the
application will reach a safe point within a bounded amount of time, even if it
is executing a long-running loop which contains no function calls.</p>
<p>Threaded collectors may also require <code class="docutils literal"><span class="pre">Return</span></code> and <code class="docutils literal"><span class="pre">PreCall</span></code> safe points to
implement “stop the world” techniques using self-modifying code, where it is
important that the program not exit the function without reaching a safe point
(because only the topmost function has been patched).</p>
</div>
<div class="section" id="emitting-assembly-code-gcmetadataprinter">
<span id="assembly"></span><h3><a class="toc-backref" href="#id27">Emitting assembly code: <code class="docutils literal"><span class="pre">GCMetadataPrinter</span></code></a><a class="headerlink" href="#emitting-assembly-code-gcmetadataprinter" title="Permalink to this headline">¶</a></h3>
<p>LLVM allows a plugin to print arbitrary assembly code before and after the rest
of a module’s assembly code. At the end of the module, the GC can compile the
LLVM stack map into assembly code. (At the beginning, this information is not
yet computed.)</p>
<p>Since AsmWriter and CodeGen are separate components of LLVM, a separate abstract
base class and registry is provided for printing assembly code, the
<code class="docutils literal"><span class="pre">GCMetadaPrinter</span></code> and <code class="docutils literal"><span class="pre">GCMetadataPrinterRegistry</span></code>. The AsmWriter will look
for such a subclass if the <code class="docutils literal"><span class="pre">GCStrategy</span></code> sets <code class="docutils literal"><span class="pre">UsesMetadata</span></code>:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="n">MyGC</span><span class="o">::</span><span class="n">MyGC</span><span class="p">()</span> <span class="p">{</span>
<span class="n">UsesMetadata</span> <span class="o">=</span> <span class="nb">true</span><span class="p">;</span>
<span class="p">}</span>
</pre></div>
</div>
<p>This separation allows JIT-only clients to be smaller.</p>
<p>Note that LLVM does not currently have analogous APIs to support code generation
in the JIT, nor using the object writers.</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="c1">// lib/MyGC/MyGCPrinter.cpp - Example LLVM GC printer</span>
<span class="cp">#include</span> <span class="cpf">"llvm/CodeGen/GCMetadataPrinter.h"</span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf">"llvm/Support/Compiler.h"</span><span class="cp"></span>
<span class="k">using</span> <span class="k">namespace</span> <span class="n">llvm</span><span class="p">;</span>
<span class="k">namespace</span> <span class="p">{</span>
<span class="k">class</span> <span class="nc">LLVM_LIBRARY_VISIBILITY</span> <span class="nl">MyGCPrinter</span> <span class="p">:</span> <span class="k">public</span> <span class="n">GCMetadataPrinter</span> <span class="p">{</span>
<span class="k">public</span><span class="o">:</span>
<span class="k">virtual</span> <span class="kt">void</span> <span class="n">beginAssembly</span><span class="p">(</span><span class="n">AsmPrinter</span> <span class="o">&</span><span class="n">AP</span><span class="p">);</span>
<span class="k">virtual</span> <span class="kt">void</span> <span class="nf">finishAssembly</span><span class="p">(</span><span class="n">AsmPrinter</span> <span class="o">&</span><span class="n">AP</span><span class="p">);</span>
<span class="p">};</span>
<span class="n">GCMetadataPrinterRegistry</span><span class="o">::</span><span class="n">Add</span><span class="o"><</span><span class="n">MyGCPrinter</span><span class="o">></span>
<span class="n">X</span><span class="p">(</span><span class="s">"mygc"</span><span class="p">,</span> <span class="s">"My bespoke garbage collector."</span><span class="p">);</span>
<span class="p">}</span>
</pre></div>
</div>
<p>The collector should use <code class="docutils literal"><span class="pre">AsmPrinter</span></code> to print portable assembly code. The
collector itself contains the stack map for the entire module, and may access
the <code class="docutils literal"><span class="pre">GCFunctionInfo</span></code> using its own <code class="docutils literal"><span class="pre">begin()</span></code> and <code class="docutils literal"><span class="pre">end()</span></code> methods. Here’s
a realistic example:</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="cp">#include</span> <span class="cpf">"llvm/CodeGen/AsmPrinter.h"</span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf">"llvm/IR/Function.h"</span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf">"llvm/IR/DataLayout.h"</span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf">"llvm/Target/TargetAsmInfo.h"</span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf">"llvm/Target/TargetMachine.h"</span><span class="cp"></span>
<span class="kt">void</span> <span class="n">MyGCPrinter</span><span class="o">::</span><span class="n">beginAssembly</span><span class="p">(</span><span class="n">AsmPrinter</span> <span class="o">&</span><span class="n">AP</span><span class="p">)</span> <span class="p">{</span>
<span class="c1">// Nothing to do.</span>
<span class="p">}</span>
<span class="kt">void</span> <span class="n">MyGCPrinter</span><span class="o">::</span><span class="n">finishAssembly</span><span class="p">(</span><span class="n">AsmPrinter</span> <span class="o">&</span><span class="n">AP</span><span class="p">)</span> <span class="p">{</span>
<span class="n">MCStreamer</span> <span class="o">&</span><span class="n">OS</span> <span class="o">=</span> <span class="n">AP</span><span class="p">.</span><span class="n">OutStreamer</span><span class="p">;</span>
<span class="kt">unsigned</span> <span class="n">IntPtrSize</span> <span class="o">=</span> <span class="n">AP</span><span class="p">.</span><span class="n">getPointerSize</span><span class="p">();</span>
<span class="c1">// Put this in the data section.</span>
<span class="n">OS</span><span class="p">.</span><span class="n">SwitchSection</span><span class="p">(</span><span class="n">AP</span><span class="p">.</span><span class="n">getObjFileLowering</span><span class="p">().</span><span class="n">getDataSection</span><span class="p">());</span>
<span class="c1">// For each function...</span>
<span class="k">for</span> <span class="p">(</span><span class="n">iterator</span> <span class="n">FI</span> <span class="o">=</span> <span class="n">begin</span><span class="p">(),</span> <span class="n">FE</span> <span class="o">=</span> <span class="n">end</span><span class="p">();</span> <span class="n">FI</span> <span class="o">!=</span> <span class="n">FE</span><span class="p">;</span> <span class="o">++</span><span class="n">FI</span><span class="p">)</span> <span class="p">{</span>
<span class="n">GCFunctionInfo</span> <span class="o">&</span><span class="n">MD</span> <span class="o">=</span> <span class="o">**</span><span class="n">FI</span><span class="p">;</span>
<span class="c1">// A compact GC layout. Emit this data structure:</span>
<span class="c1">//</span>
<span class="c1">// struct {</span>
<span class="c1">// int32_t PointCount;</span>
<span class="c1">// void *SafePointAddress[PointCount];</span>
<span class="c1">// int32_t StackFrameSize; // in words</span>
<span class="c1">// int32_t StackArity;</span>
<span class="c1">// int32_t LiveCount;</span>
<span class="c1">// int32_t LiveOffsets[LiveCount];</span>
<span class="c1">// } __gcmap_<FUNCTIONNAME>;</span>
<span class="c1">// Align to address width.</span>
<span class="n">AP</span><span class="p">.</span><span class="n">EmitAlignment</span><span class="p">(</span><span class="n">IntPtrSize</span> <span class="o">==</span> <span class="mi">4</span> <span class="o">?</span> <span class="mi">2</span> <span class="o">:</span> <span class="mi">3</span><span class="p">);</span>
<span class="c1">// Emit PointCount.</span>
<span class="n">OS</span><span class="p">.</span><span class="n">AddComment</span><span class="p">(</span><span class="s">"safe point count"</span><span class="p">);</span>
<span class="n">AP</span><span class="p">.</span><span class="n">EmitInt32</span><span class="p">(</span><span class="n">MD</span><span class="p">.</span><span class="n">size</span><span class="p">());</span>
<span class="c1">// And each safe point...</span>
<span class="k">for</span> <span class="p">(</span><span class="n">GCFunctionInfo</span><span class="o">::</span><span class="n">iterator</span> <span class="n">PI</span> <span class="o">=</span> <span class="n">MD</span><span class="p">.</span><span class="n">begin</span><span class="p">(),</span>
<span class="n">PE</span> <span class="o">=</span> <span class="n">MD</span><span class="p">.</span><span class="n">end</span><span class="p">();</span> <span class="n">PI</span> <span class="o">!=</span> <span class="n">PE</span><span class="p">;</span> <span class="o">++</span><span class="n">PI</span><span class="p">)</span> <span class="p">{</span>
<span class="c1">// Emit the address of the safe point.</span>
<span class="n">OS</span><span class="p">.</span><span class="n">AddComment</span><span class="p">(</span><span class="s">"safe point address"</span><span class="p">);</span>
<span class="n">MCSymbol</span> <span class="o">*</span><span class="n">Label</span> <span class="o">=</span> <span class="n">PI</span><span class="o">-></span><span class="n">Label</span><span class="p">;</span>
<span class="n">AP</span><span class="p">.</span><span class="n">EmitLabelPlusOffset</span><span class="p">(</span><span class="n">Label</span><span class="cm">/*Hi*/</span><span class="p">,</span> <span class="mi">0</span><span class="cm">/*Offset*/</span><span class="p">,</span> <span class="mi">4</span><span class="cm">/*Size*/</span><span class="p">);</span>
<span class="p">}</span>
<span class="c1">// Stack information never change in safe points! Only print info from the</span>
<span class="c1">// first call-site.</span>
<span class="n">GCFunctionInfo</span><span class="o">::</span><span class="n">iterator</span> <span class="n">PI</span> <span class="o">=</span> <span class="n">MD</span><span class="p">.</span><span class="n">begin</span><span class="p">();</span>
<span class="c1">// Emit the stack frame size.</span>
<span class="n">OS</span><span class="p">.</span><span class="n">AddComment</span><span class="p">(</span><span class="s">"stack frame size (in words)"</span><span class="p">);</span>
<span class="n">AP</span><span class="p">.</span><span class="n">EmitInt32</span><span class="p">(</span><span class="n">MD</span><span class="p">.</span><span class="n">getFrameSize</span><span class="p">()</span> <span class="o">/</span> <span class="n">IntPtrSize</span><span class="p">);</span>
<span class="c1">// Emit stack arity, i.e. the number of stacked arguments.</span>
<span class="kt">unsigned</span> <span class="n">RegisteredArgs</span> <span class="o">=</span> <span class="n">IntPtrSize</span> <span class="o">==</span> <span class="mi">4</span> <span class="o">?</span> <span class="mi">5</span> <span class="o">:</span> <span class="mi">6</span><span class="p">;</span>
<span class="kt">unsigned</span> <span class="n">StackArity</span> <span class="o">=</span> <span class="n">MD</span><span class="p">.</span><span class="n">getFunction</span><span class="p">().</span><span class="n">arg_size</span><span class="p">()</span> <span class="o">></span> <span class="n">RegisteredArgs</span> <span class="o">?</span>
<span class="n">MD</span><span class="p">.</span><span class="n">getFunction</span><span class="p">().</span><span class="n">arg_size</span><span class="p">()</span> <span class="o">-</span> <span class="nl">RegisteredArgs</span> <span class="p">:</span> <span class="mi">0</span><span class="p">;</span>
<span class="n">OS</span><span class="p">.</span><span class="n">AddComment</span><span class="p">(</span><span class="s">"stack arity"</span><span class="p">);</span>
<span class="n">AP</span><span class="p">.</span><span class="n">EmitInt32</span><span class="p">(</span><span class="n">StackArity</span><span class="p">);</span>
<span class="c1">// Emit the number of live roots in the function.</span>
<span class="n">OS</span><span class="p">.</span><span class="n">AddComment</span><span class="p">(</span><span class="s">"live root count"</span><span class="p">);</span>
<span class="n">AP</span><span class="p">.</span><span class="n">EmitInt32</span><span class="p">(</span><span class="n">MD</span><span class="p">.</span><span class="n">live_size</span><span class="p">(</span><span class="n">PI</span><span class="p">));</span>
<span class="c1">// And for each live root...</span>
<span class="k">for</span> <span class="p">(</span><span class="n">GCFunctionInfo</span><span class="o">::</span><span class="n">live_iterator</span> <span class="n">LI</span> <span class="o">=</span> <span class="n">MD</span><span class="p">.</span><span class="n">live_begin</span><span class="p">(</span><span class="n">PI</span><span class="p">),</span>
<span class="n">LE</span> <span class="o">=</span> <span class="n">MD</span><span class="p">.</span><span class="n">live_end</span><span class="p">(</span><span class="n">PI</span><span class="p">);</span>
<span class="n">LI</span> <span class="o">!=</span> <span class="n">LE</span><span class="p">;</span> <span class="o">++</span><span class="n">LI</span><span class="p">)</span> <span class="p">{</span>
<span class="c1">// Emit live root's offset within the stack frame.</span>
<span class="n">OS</span><span class="p">.</span><span class="n">AddComment</span><span class="p">(</span><span class="s">"stack index (offset / wordsize)"</span><span class="p">);</span>
<span class="n">AP</span><span class="p">.</span><span class="n">EmitInt32</span><span class="p">(</span><span class="n">LI</span><span class="o">-></span><span class="n">StackOffset</span><span class="p">);</span>
<span class="p">}</span>
<span class="p">}</span>
<span class="p">}</span>
</pre></div>
</div>
</div>
</div>
<div class="section" id="references">
<h2><a class="toc-backref" href="#id28">References</a><a class="headerlink" href="#references" title="Permalink to this headline">¶</a></h2>
<p id="appel89">[Appel89] Runtime Tags Aren’t Necessary. Andrew W. Appel. Lisp and Symbolic
Computation 19(7):703-705, July 1989.</p>
<p id="goldberg91">[Goldberg91] Tag-free garbage collection for strongly typed programming
languages. Benjamin Goldberg. ACM SIGPLAN PLDI‘91.</p>
<p id="tolmach94">[Tolmach94] Tag-free garbage collection using explicit type parameters. Andrew
Tolmach. Proceedings of the 1994 ACM conference on LISP and functional
programming.</p>
<p id="henderson02">[Henderson2002] <a class="reference external" href="http://citeseer.ist.psu.edu/henderson02accurate.html">Accurate Garbage Collection in an Uncooperative Environment</a></p>
</div>
</div>
</div>
</div>
<div class="clearer"></div>
</div>
<div class="related" role="navigation" aria-label="related navigation">
<h3>Navigation</h3>
<ul>
<li class="right" style="margin-right: 10px">
<a href="genindex.html" title="General Index"
>index</a></li>
<li class="right" >
<a href="WritingAnLLVMPass.html" title="Writing an LLVM Pass"
>next</a> |</li>
<li class="right" >
<a href="HowToUseInstrMappings.html" title="How To Use Instruction Mappings"
>previous</a> |</li>
<li><a href="http://llvm.org/">LLVM Home</a> | </li>
<li><a href="index.html">Documentation</a>»</li>
</ul>
</div>
<div class="footer" role="contentinfo">
© Copyright 2003-2018, LLVM Project.
Last updated on 2018-04-05.
Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.6.7.
</div>
</body>
</html>
|