1 =====================================
2 Garbage Collection with LLVM
3 =====================================
11 This document covers how to integrate LLVM into a compiler for a language which
12 supports garbage collection. **Note that LLVM itself does not provide a
13 garbage collector.** You must provide your own.
18 First, you should pick a collector strategy. LLVM includes a number of built
19 in ones, but you can also implement a loadable plugin with a custom definition.
20 Note that the collector strategy is a description of how LLVM should generate
21 code such that it interacts with your collector and runtime, not a description
22 of the collector itself.
24 Next, mark your generated functions as using your chosen collector strategy.
25 From c++, you can call:
29 F.setGC(<collector description name>);
32 This will produce IR like the following fragment:
36 define void @foo() gc "<collector description name>" { ... }
39 When generating LLVM IR for your functions, you will need to:
41 * Use ``@llvm.gcread`` and/or ``@llvm.gcwrite`` in place of standard load and
42 store instructions. These intrinsics are used to represent load and store
43 barriers. If you collector does not require such barriers, you can skip
46 * Use the memory allocation routines provided by your garbage collector's
49 * If your collector requires them, generate type maps according to your
50 runtime's binary interface. LLVM is not involved in the process. In
51 particular, the LLVM type system is not suitable for conveying such
52 information though the compiler.
54 * Insert any coordination code required for interacting with your collector.
55 Many collectors require running application code to periodically check a
56 flag and conditionally call a runtime function. This is often referred to
59 You will need to identify roots (i.e. references to heap objects your collector
60 needs to know about) in your generated IR, so that LLVM can encode them into
61 your final stack maps. Depending on the collector strategy chosen, this is
62 accomplished by using either the ``@llvm.gcroot`` intrinsics or an
63 ``gc.statepoint`` relocation sequence.
65 Don't forget to create a root for each intermediate value that is generated when
66 evaluating an expression. In ``h(f(), g())``, the result of ``f()`` could
67 easily be collected if evaluating ``g()`` triggers a collection.
69 Finally, you need to link your runtime library with the generated program
70 executable (for a static compiler) or ensure the appropriate symbols are
71 available for the runtime linker (for a JIT compiler).
77 What is Garbage Collection?
78 ---------------------------
80 Garbage collection is a widely used technique that frees the programmer from
81 having to know the lifetimes of heap objects, making software easier to produce
82 and maintain. Many programming languages rely on garbage collection for
83 automatic memory management. There are two primary forms of garbage collection:
84 conservative and accurate.
86 Conservative garbage collection often does not require any special support from
87 either the language or the compiler: it can handle non-type-safe programming
88 languages (such as C/C++) and does not require any special information from the
89 compiler. The `Boehm collector
90 <http://www.hpl.hp.com/personal/Hans_Boehm/gc/>`__ is an example of a
91 state-of-the-art conservative collector.
93 Accurate garbage collection requires the ability to identify all pointers in the
94 program at run-time (which requires that the source-language be type-safe in
95 most cases). Identifying pointers at run-time requires compiler support to
96 locate all places that hold live pointer variables at run-time, including the
97 :ref:`processor stack and registers <gcroot>`.
99 Conservative garbage collection is attractive because it does not require any
100 special compiler support, but it does have problems. In particular, because the
101 conservative garbage collector cannot *know* that a particular word in the
102 machine is a pointer, it cannot move live objects in the heap (preventing the
103 use of compacting and generational GC algorithms) and it can occasionally suffer
104 from memory leaks due to integer values that happen to point to objects in the
105 program. In addition, some aggressive compiler transformations can break
106 conservative garbage collectors (though these seem rare in practice).
108 Accurate garbage collectors do not suffer from any of these problems, but they
109 can suffer from degraded scalar optimization of the program. In particular,
110 because the runtime must be able to identify and update all pointers active in
111 the program, some optimizations are less effective. In practice, however, the
112 locality and performance benefits of using aggressive garbage collection
113 techniques dominates any low-level losses.
115 This document describes the mechanisms and interfaces provided by LLVM to
116 support accurate garbage collection.
121 LLVM's intermediate representation provides :ref:`garbage collection intrinsics
122 <gc_intrinsics>` that offer support for a broad class of collector models. For
123 instance, the intrinsics permit:
125 * semi-space collectors
127 * mark-sweep collectors
129 * generational collectors
131 * incremental collectors
133 * concurrent collectors
135 * cooperative collectors
139 We hope that the support built into the LLVM IR is sufficient to support a
140 broad class of garbage collected languages including Scheme, ML, Java, C#,
141 Perl, Python, Lua, Ruby, other scripting languages, and more.
143 Note that LLVM **does not itself provide a garbage collector** --- this should
144 be part of your language's runtime library. LLVM provides a framework for
145 describing the garbage collectors requirements to the compiler. In particular,
146 LLVM provides support for generating stack maps at call sites, polling for a
147 safepoint, and emitting load and store barriers. You can also extend LLVM -
148 possibly through a loadable :ref:`code generation plugins <plugin>` - to
149 generate code and data structures which conforms to the *binary interface*
150 specified by the *runtime library*. This is similar to the relationship between
151 LLVM and DWARF debugging info, for example. The difference primarily lies in
152 the lack of an established standard in the domain of garbage collection --- thus
153 the need for a flexible extension mechanism.
155 The aspects of the binary interface with which LLVM's GC support is
158 * Creation of GC safepoints within code where collection is allowed to execute
161 * Computation of the stack map. For each safe point in the code, object
162 references within the stack frame must be identified so that the collector may
163 traverse and perhaps update them.
165 * Write barriers when storing object references to the heap. These are commonly
166 used to optimize incremental scans in generational collectors.
168 * Emission of read barriers when loading object references. These are useful
169 for interoperating with concurrent collectors.
171 There are additional areas that LLVM does not directly address:
173 * Registration of global roots with the runtime.
175 * Registration of stack map entries with the runtime.
177 * The functions used by the program to allocate memory, trigger a collection,
180 * Computation or compilation of type maps, or registration of them with the
181 runtime. These are used to crawl the heap for object references.
183 In general, LLVM's support for GC does not include features which can be
184 adequately addressed with other features of the IR and does not specify a
185 particular binary interface. On the plus side, this means that you should be
186 able to integrate LLVM with an existing runtime. On the other hand, it can
187 have the effect of leaving a lot of work for the developer of a novel
188 language. We try to mitigate this by providing built in collector strategy
189 descriptions that can work with many common collector designs and easy
190 extension points. If you don't already have a specific binary interface
191 you need to support, we recommend trying to use one of these built in collector
199 This section describes the garbage collection facilities provided by the
200 :doc:`LLVM intermediate representation <LangRef>`. The exact behavior of these
201 IR features is specified by the selected :ref:`GC strategy description
204 Specifying GC code generation: ``gc "..."``
205 -------------------------------------------
209 define <returntype> @name(...) gc "name" { ... }
211 The ``gc`` function attribute is used to specify the desired GC strategy to the
212 compiler. Its programmatic equivalent is the ``setGC`` method of ``Function``.
214 Setting ``gc "name"`` on a function triggers a search for a matching subclass
215 of GCStrategy. Some collector strategies are built in. You can add others
216 using either the loadable plugin mechanism, or by patching your copy of LLVM.
217 It is the selected GC strategy which defines the exact nature of the code
218 generated to support GC. If none is found, the compiler will raise an error.
220 Specifying the GC style on a per-function basis allows LLVM to link together
221 programs that use different garbage collection algorithms (or none at all).
225 Identifying GC roots on the stack
226 ----------------------------------
228 LLVM currently supports two different mechanisms for describing references in
229 compiled code at safepoints. ``llvm.gcroot`` is the older mechanism;
230 ``gc.statepoint`` has been added more recently. At the moment, you can choose
231 either implementation (on a per :ref:`GC strategy <plugin>` basis). Longer
232 term, we will probably either migrate away from ``llvm.gcroot`` entirely, or
233 substantially merge their implementations. Note that most new development
234 work is focused on ``gc.statepoint``.
236 Using ``gc.statepoint``
237 ^^^^^^^^^^^^^^^^^^^^^^^^
238 :doc:`This page <Statepoints>` contains detailed documentation for
241 Using ``llvm.gcwrite``
242 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
246 void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
248 The ``llvm.gcroot`` intrinsic is used to inform LLVM that a stack variable
249 references an object on the heap and is to be tracked for garbage collection.
250 The exact impact on generated code is specified by the Function's selected
251 :ref:`GC strategy <plugin>`. All calls to ``llvm.gcroot`` **must** reside
252 inside the first basic block.
254 The first argument **must** be a value referring to an alloca instruction or a
255 bitcast of an alloca. The second contains a pointer to metadata that should be
256 associated with the pointer, and **must** be a constant or global value
257 address. If your target collector uses tags, use a null pointer for metadata.
259 A compiler which performs manual SSA construction **must** ensure that SSA
260 values representing GC references are stored in to the alloca passed to the
261 respective ``gcroot`` before every call site and reloaded after every call.
262 A compiler which uses mem2reg to raise imperative code using ``alloca`` into
263 SSA form need only add a call to ``@llvm.gcroot`` for those variables which
264 are pointers into the GC heap.
266 It is also important to mark intermediate values with ``llvm.gcroot``. For
267 example, consider ``h(f(), g())``. Beware leaking the result of ``f()`` in the
268 case that ``g()`` triggers a collection. Note, that stack variables must be
269 initialized and marked with ``llvm.gcroot`` in function's prologue.
271 The ``%metadata`` argument can be used to avoid requiring heap objects to have
272 'isa' pointers or tag bits. [Appel89_, Goldberg91_, Tolmach94_] If specified,
273 its value will be tracked along with the location of the pointer in the stack
276 Consider the following fragment of Java code:
281 Object X; // A null-initialized reference to an object
285 This block (which may be located in the middle of a function or in a loop nest),
286 could be compiled to this LLVM code:
291 ;; In the entry block for the function, allocate the
292 ;; stack space for X, which is an LLVM pointer.
295 ;; Tell LLVM that the stack space is a stack root.
296 ;; Java has type-tags on objects, so we pass null as metadata.
297 %tmp = bitcast %Object** %X to i8**
298 call void @llvm.gcroot(i8** %tmp, i8* null)
301 ;; "CodeBlock" is the block corresponding to the start
302 ;; of the scope above.
304 ;; Java null-initializes pointers.
305 store %Object* null, %Object** %X
309 ;; As the pointer goes out of scope, store a null value into
310 ;; it, to indicate that the value is no longer live.
311 store %Object* null, %Object** %X
314 Reading and writing references in the heap
315 ------------------------------------------
317 Some collectors need to be informed when the mutator (the program that needs
318 garbage collection) either reads a pointer from or writes a pointer to a field
319 of a heap object. The code fragments inserted at these points are called *read
320 barriers* and *write barriers*, respectively. The amount of code that needs to
321 be executed is usually quite small and not on the critical path of any
322 computation, so the overall performance impact of the barrier is tolerable.
324 Barriers often require access to the *object pointer* rather than the *derived
325 pointer* (which is a pointer to the field within the object). Accordingly,
326 these intrinsics take both pointers as separate arguments for completeness. In
327 this snippet, ``%object`` is the object pointer, and ``%derived`` is the derived
333 %class.Array = type { %class.Object, i32, [0 x %class.Object*] }
336 ;; Load the object pointer from a gcroot.
337 %object = load %class.Array** %object_addr
339 ;; Compute the derived pointer.
340 %derived = getelementptr %object, i32 0, i32 2, i32 %n
342 LLVM does not enforce this relationship between the object and derived pointer
343 (although a particular :ref:`collector strategy <plugin>` might). However, it
344 would be an unusual collector that violated it.
346 The use of these intrinsics is naturally optional if the target GC does not
347 require the corresponding barrier. The GC strategy used with such a collector
348 should replace the intrinsic calls with the corresponding ``load`` or
349 ``store`` instruction if they are used.
351 One known deficiency with the current design is that the barrier intrinsics do
352 not include the size or alignment of the underlying operation performed. It is
353 currently assumed that the operation is of pointer size and the alignment is
354 assumed to be the target machine's default alignment.
356 Write barrier: ``llvm.gcwrite``
357 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
361 void @llvm.gcwrite(i8* %value, i8* %object, i8** %derived)
363 For write barriers, LLVM provides the ``llvm.gcwrite`` intrinsic function. It
364 has exactly the same semantics as a non-volatile ``store`` to the derived
365 pointer (the third argument). The exact code generated is specified by the
366 Function's selected :ref:`GC strategy <plugin>`.
368 Many important algorithms require write barriers, including generational and
369 concurrent collectors. Additionally, write barriers could be used to implement
372 Read barrier: ``llvm.gcread``
373 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
377 i8* @llvm.gcread(i8* %object, i8** %derived)
379 For read barriers, LLVM provides the ``llvm.gcread`` intrinsic function. It has
380 exactly the same semantics as a non-volatile ``load`` from the derived pointer
381 (the second argument). The exact code generated is specified by the Function's
382 selected :ref:`GC strategy <plugin>`.
384 Read barriers are needed by fewer algorithms than write barriers, and may have a
385 greater performance impact since pointer reads are more frequent than writes.
389 .. _builtin-gc-strategies:
391 Built In GC Strategies
392 ======================
394 LLVM includes built in support for several varieties of garbage collectors.
397 ----------------------
399 To use this collector strategy, mark your functions with:
403 F.setGC("shadow-stack");
405 Unlike many GC algorithms which rely on a cooperative code generator to compile
406 stack maps, this algorithm carefully maintains a linked list of stack roots
407 [:ref:`Henderson2002 <henderson02>`]. This so-called "shadow stack" mirrors the
408 machine stack. Maintaining this data structure is slower than using a stack map
409 compiled into the executable as constant data, but has a significant portability
410 advantage because it requires no special support from the target code generator,
411 and does not require tricky platform-specific code to crawl the machine stack.
413 The tradeoff for this simplicity and portability is:
415 * High overhead per function call.
419 Still, it's an easy way to get started. After your compiler and runtime are up
420 and running, writing a :ref:`plugin <plugin>` will allow you to take advantage
421 of :ref:`more advanced GC features <collector-algos>` of LLVM in order to
425 The shadow stack doesn't imply a memory allocation algorithm. A semispace
426 collector or building atop ``malloc`` are great places to start, and can be
427 implemented with very little code.
429 When it comes time to collect, however, your runtime needs to traverse the stack
430 roots, and for this it needs to integrate with the shadow stack. Luckily, doing
431 so is very simple. (This code is heavily commented to help you understand the
432 data structure, but there are only 20 lines of meaningful code.)
436 /// @brief The map for a single function's stack frame. One of these is
437 /// compiled as constant data into the executable for each function.
439 /// Storage of metadata values is elided if the %metadata parameter to
440 /// @llvm.gcroot is null.
442 int32_t NumRoots; //< Number of roots in stack frame.
443 int32_t NumMeta; //< Number of metadata entries. May be < NumRoots.
444 const void *Meta[0]; //< Metadata for each root.
447 /// @brief A link in the dynamic shadow stack. One of these is embedded in
448 /// the stack frame of each function on the call stack.
450 StackEntry *Next; //< Link to next stack entry (the caller's).
451 const FrameMap *Map; //< Pointer to constant FrameMap.
452 void *Roots[0]; //< Stack roots (in-place array).
455 /// @brief The head of the singly-linked list of StackEntries. Functions push
456 /// and pop onto this in their prologue and epilogue.
458 /// Since there is only a global list, this technique is not threadsafe.
459 StackEntry *llvm_gc_root_chain;
461 /// @brief Calls Visitor(root, meta) for each GC root on the stack.
462 /// root and meta are exactly the values passed to
465 /// Visitor could be a function to recursively mark live objects. Or it
466 /// might copy them to another heap or generation.
468 /// @param Visitor A function to invoke for every GC root on the stack.
469 void visitGCRoots(void (*Visitor)(void **Root, const void *Meta)) {
470 for (StackEntry *R = llvm_gc_root_chain; R; R = R->Next) {
473 // For roots [0, NumMeta), the metadata pointer is in the FrameMap.
474 for (unsigned e = R->Map->NumMeta; i != e; ++i)
475 Visitor(&R->Roots[i], R->Map->Meta[i]);
477 // For roots [NumMeta, NumRoots), the metadata pointer is null.
478 for (unsigned e = R->Map->NumRoots; i != e; ++i)
479 Visitor(&R->Roots[i], NULL);
484 The 'Erlang' and 'Ocaml' GCs
485 -----------------------------
487 LLVM ships with two example collectors which leverage the ``gcroot``
488 mechanisms. To our knowledge, these are not actually used by any language
489 runtime, but they do provide a reasonable starting point for someone interested
490 in writing an ``gcroot`` compatible GC plugin. In particular, these are the
491 only in tree examples of how to produce a custom binary stack map format using
492 a ``gcroot`` strategy.
494 As there names imply, the binary format produced is intended to model that
495 used by the Erlang and OCaml compilers respectively.
497 .. _statepoint_example_gc:
499 The Statepoint Example GC
500 -------------------------
504 F.setGC("statepoint-example");
506 This GC provides an example of how one might use the infrastructure provided
507 by ``gc.statepoint``. This example GC is compatible with the
508 :ref:`PlaceSafepoints` and :ref:`RewriteStatepointsForGC` utility passes
509 which simplify ``gc.statepoint`` sequence insertion. If you need to build a
510 custom GC strategy around the ``gc.statepoints`` mechanisms, it is recommended
511 that you use this one as a starting point.
513 This GC strategy does not support read or write barriers. As a result, these
514 intrinsics are lowered to normal loads and stores.
516 The stack map format generated by this GC strategy can be found in the
517 :ref:`stackmap-section` using a format documented :ref:`here
518 <statepoint-stackmap-format>`. This format is intended to be the standard
519 format supported by LLVM going forward.
522 -------------------------
528 This GC leverages the ``gc.statepoint`` mechanism to support the
529 `CoreCLR <https://github.com/dotnet/coreclr>`__ runtime.
531 Support for this GC strategy is a work in progress. This strategy will
533 :ref:`statepoint-example GC<statepoint_example_gc>` strategy in
534 certain aspects like:
536 * Base-pointers of interior pointers are not explicitly
537 tracked and reported.
539 * A different format is used for encoding stack maps.
541 * Safe-point polls are only needed before loop-back edges
542 and before tail-calls (not needed at function-entry).
547 If none of the built in GC strategy descriptions met your needs above, you will
548 need to define a custom GCStrategy and possibly, a custom LLVM pass to perform
549 lowering. Your best example of where to start defining a custom GCStrategy
550 would be to look at one of the built in strategies.
552 You may be able to structure this additional code as a loadable plugin library.
553 Loadable plugins are sufficient if all you need is to enable a different
554 combination of built in functionality, but if you need to provide a custom
555 lowering pass, you will need to build a patched version of LLVM. If you think
556 you need a patched build, please ask for advice on llvm-dev. There may be an
557 easy way we can extend the support to make it work for your use case without
558 requiring a custom build.
560 Collector Requirements
561 ----------------------
563 You should be able to leverage any existing collector library that includes the following elements:
565 #. A memory allocator which exposes an allocation function your compiled
568 #. A binary format for the stack map. A stack map describes the location
569 of references at a safepoint and is used by precise collectors to identify
570 references within a stack frame on the machine stack. Note that collectors
571 which conservatively scan the stack don't require such a structure.
573 #. A stack crawler to discover functions on the call stack, and enumerate the
574 references listed in the stack map for each call site.
576 #. A mechanism for identifying references in global locations (e.g. global
579 #. If you collector requires them, an LLVM IR implementation of your collectors
580 load and store barriers. Note that since many collectors don't require
581 barriers at all, LLVM defaults to lowering such barriers to normal loads
582 and stores unless you arrange otherwise.
585 Implementing a collector plugin
586 -------------------------------
588 User code specifies which GC code generation to use with the ``gc`` function
589 attribute or, equivalently, with the ``setGC`` method of ``Function``.
591 To implement a GC plugin, it is necessary to subclass ``llvm::GCStrategy``,
592 which can be accomplished in a few lines of boilerplate code. LLVM's
593 infrastructure provides access to several important algorithms. For an
594 uncontroversial collector, all that remains may be to compile LLVM's computed
595 stack map to assembly code (using the binary representation expected by the
596 runtime library). This can be accomplished in about 100 lines of code.
598 This is not the appropriate place to implement a garbage collected heap or a
599 garbage collector itself. That code should exist in the language's runtime
600 library. The compiler plugin is responsible for generating code which conforms
601 to the binary interface defined by library, most essentially the :ref:`stack map
604 To subclass ``llvm::GCStrategy`` and register it with the compiler:
608 // lib/MyGC/MyGC.cpp - Example LLVM GC plugin
610 #include "llvm/CodeGen/GCStrategy.h"
611 #include "llvm/CodeGen/GCMetadata.h"
612 #include "llvm/Support/Compiler.h"
614 using namespace llvm;
617 class LLVM_LIBRARY_VISIBILITY MyGC : public GCStrategy {
622 GCRegistry::Add<MyGC>
623 X("mygc", "My bespoke garbage collector.");
626 This boilerplate collector does nothing. More specifically:
628 * ``llvm.gcread`` calls are replaced with the corresponding ``load``
631 * ``llvm.gcwrite`` calls are replaced with the corresponding ``store``
634 * No safe points are added to the code.
636 * The stack map is not compiled into the executable.
638 Using the LLVM makefiles, this code
639 can be compiled as a plugin using a simple makefile:
649 include $(LEVEL)/Makefile.common
651 Once the plugin is compiled, code using it may be compiled using ``llc
652 -load=MyGC.so`` (though MyGC.so may have some other platform-specific
658 define void @f() gc "mygc" {
662 $ llvm-as < sample.ll | llc -load=MyGC.so
664 It is also possible to statically link the collector plugin into tools, such as
665 a language-specific compiler front-end.
669 Overview of available features
670 ------------------------------
672 ``GCStrategy`` provides a range of features through which a plugin may do useful
673 work. Some of these are callbacks, some are algorithms that can be enabled,
674 disabled, or customized. This matrix summarizes the supported (and planned)
675 features and correlates them with the collection techniques which typically
678 .. |v| unicode:: 0x2714
681 .. |x| unicode:: 0x2718
684 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
685 | Algorithm | Done | Shadow | refcount | mark- | copying | incremental | threaded | concurrent |
686 | | | stack | | sweep | | | | |
687 +============+======+========+==========+=======+=========+=============+==========+============+
688 | stack map | |v| | | | |x| | |x| | |x| | |x| | |x| |
689 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
690 | initialize | |v| | |x| | |x| | |x| | |x| | |x| | |x| | |x| |
691 | roots | | | | | | | | |
692 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
693 | derived | NO | | | | | | **N**\* | **N**\* |
694 | pointers | | | | | | | | |
695 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
696 | **custom | |v| | | | | | | | |
697 | lowering** | | | | | | | | |
698 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
699 | *gcroot* | |v| | |x| | |x| | | | | | |
700 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
701 | *gcwrite* | |v| | | |x| | | | |x| | | |x| |
702 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
703 | *gcread* | |v| | | | | | | | |x| |
704 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
705 | **safe | | | | | | | | |
706 | points** | | | | | | | | |
707 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
708 | *in | |v| | | | |x| | |x| | |x| | |x| | |x| |
709 | calls* | | | | | | | | |
710 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
711 | *before | |v| | | | | | | |x| | |x| |
712 | calls* | | | | | | | | |
713 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
714 | *for | NO | | | | | | **N** | **N** |
715 | loops* | | | | | | | | |
716 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
717 | *before | |v| | | | | | | |x| | |x| |
718 | escape* | | | | | | | | |
719 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
720 | emit code | NO | | | | | | **N** | **N** |
721 | at safe | | | | | | | | |
722 | points | | | | | | | | |
723 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
724 | **output** | | | | | | | | |
725 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
726 | *assembly* | |v| | | | |x| | |x| | |x| | |x| | |x| |
727 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
728 | *JIT* | NO | | | **?** | **?** | **?** | **?** | **?** |
729 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
730 | *obj* | NO | | | **?** | **?** | **?** | **?** | **?** |
731 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
732 | live | NO | | | **?** | **?** | **?** | **?** | **?** |
733 | analysis | | | | | | | | |
734 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
735 | register | NO | | | **?** | **?** | **?** | **?** | **?** |
736 | map | | | | | | | | |
737 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
738 | \* Derived pointers only pose a hasard to copying collections. |
739 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
740 | **?** denotes a feature which could be utilized if available. |
741 +------------+------+--------+----------+-------+---------+-------------+----------+------------+
743 To be clear, the collection techniques above are defined as:
746 The mutator carefully maintains a linked list of stack roots.
749 The mutator maintains a reference count for each object and frees an object
750 when its count falls to zero.
753 When the heap is exhausted, the collector marks reachable objects starting
754 from the roots, then deallocates unreachable objects in a sweep phase.
757 As reachability analysis proceeds, the collector copies objects from one heap
758 area to another, compacting them in the process. Copying collectors enable
759 highly efficient "bump pointer" allocation and can improve locality of
763 (Including generational collectors.) Incremental collectors generally have all
764 the properties of a copying collector (regardless of whether the mature heap
765 is compacting), but bring the added complexity of requiring write barriers.
768 Denotes a multithreaded mutator; the collector must still stop the mutator
769 ("stop the world") before beginning reachability analysis. Stopping a
770 multithreaded mutator is a complicated problem. It generally requires highly
771 platform-specific code in the runtime, and the production of carefully
772 designed machine code at safe points.
775 In this technique, the mutator and the collector run concurrently, with the
776 goal of eliminating pause times. In a *cooperative* collector, the mutator
777 further aids with collection should a pause occur, allowing collection to take
778 advantage of multiprocessor hosts. The "stop the world" problem of threaded
779 collectors is generally still present to a limited extent. Sophisticated
780 marking algorithms are necessary. Read barriers may be necessary.
782 As the matrix indicates, LLVM's garbage collection infrastructure is already
783 suitable for a wide variety of collectors, but does not currently extend to
784 multithreaded programs. This will be added in the future as there is
792 LLVM automatically computes a stack map. One of the most important features
793 of a ``GCStrategy`` is to compile this information into the executable in
794 the binary representation expected by the runtime library.
796 The stack map consists of the location and identity of each GC root in the
797 each function in the module. For each root:
799 * ``RootNum``: The index of the root.
801 * ``StackOffset``: The offset of the object relative to the frame pointer.
803 * ``RootMetadata``: The value passed as the ``%metadata`` parameter to the
804 ``@llvm.gcroot`` intrinsic.
806 Also, for the function as a whole:
808 * ``getFrameSize()``: The overall size of the function's initial stack frame,
809 not accounting for any dynamic allocation.
811 * ``roots_size()``: The count of roots in the function.
813 To access the stack map, use ``GCFunctionMetadata::roots_begin()`` and
814 -``end()`` from the :ref:`GCMetadataPrinter <assembly>`:
818 for (iterator I = begin(), E = end(); I != E; ++I) {
819 GCFunctionInfo *FI = *I;
820 unsigned FrameSize = FI->getFrameSize();
821 size_t RootCount = FI->roots_size();
823 for (GCFunctionInfo::roots_iterator RI = FI->roots_begin(),
824 RE = FI->roots_end();
826 int RootNum = RI->Num;
827 int RootStackOffset = RI->StackOffset;
828 Constant *RootMetadata = RI->Metadata;
832 If the ``llvm.gcroot`` intrinsic is eliminated before code generation by a
833 custom lowering pass, LLVM will compute an empty stack map. This may be useful
834 for collector plugins which implement reference counting or a shadow stack.
838 Initializing roots to null: ``InitRoots``
839 -----------------------------------------
847 When set, LLVM will automatically initialize each root to ``null`` upon entry to
848 the function. This prevents the GC's sweep phase from visiting uninitialized
849 pointers, which will almost certainly cause it to crash. This initialization
850 occurs before custom lowering, so the two may be used together.
852 Since LLVM does not yet compute liveness information, there is no means of
853 distinguishing an uninitialized stack root from an initialized one. Therefore,
854 this feature should be used by all GC plugins. It is enabled by default.
856 Custom lowering of intrinsics: ``CustomRoots``, ``CustomReadBarriers``, and ``CustomWriteBarriers``
857 ---------------------------------------------------------------------------------------------------
859 For GCs which use barriers or unusual treatment of stack roots, these
860 flags allow the collector to perform arbitrary transformations of the
865 class MyGC : public GCStrategy {
869 CustomReadBarriers = true;
870 CustomWriteBarriers = true;
874 If any of these flags are set, LLVM suppresses its default lowering for
875 the corresponding intrinsics. Instead, you must provide a custom Pass
876 which lowers the intrinsics as desired. If you have opted in to custom
877 lowering of a particular intrinsic your pass **must** eliminate all
878 instances of the corresponding intrinsic in functions which opt in to
879 your GC. The best example of such a pass is the ShadowStackGC and it's
880 ShadowStackGCLowering pass.
882 There is currently no way to register such a custom lowering pass
883 without building a custom copy of LLVM.
887 Generating safe points: ``NeededSafePoints``
888 --------------------------------------------
890 LLVM can compute four kinds of safe points:
895 /// PointKind - The type of a collector-safe point.
898 Loop, //< Instr is a loop (backwards branch).
899 Return, //< Instr is a return instruction.
900 PreCall, //< Instr is a call instruction.
901 PostCall //< Instr is the return address of a call.
905 A collector can request any combination of the four by setting the
906 ``NeededSafePoints`` mask:
911 NeededSafePoints = 1 << GC::Loop
917 It can then use the following routines to access safe points.
921 for (iterator I = begin(), E = end(); I != E; ++I) {
922 GCFunctionInfo *MD = *I;
923 size_t PointCount = MD->size();
925 for (GCFunctionInfo::iterator PI = MD->begin(),
926 PE = MD->end(); PI != PE; ++PI) {
927 GC::PointKind PointKind = PI->Kind;
928 unsigned PointNum = PI->Num;
932 Almost every collector requires ``PostCall`` safe points, since these correspond
933 to the moments when the function is suspended during a call to a subroutine.
935 Threaded programs generally require ``Loop`` safe points to guarantee that the
936 application will reach a safe point within a bounded amount of time, even if it
937 is executing a long-running loop which contains no function calls.
939 Threaded collectors may also require ``Return`` and ``PreCall`` safe points to
940 implement "stop the world" techniques using self-modifying code, where it is
941 important that the program not exit the function without reaching a safe point
942 (because only the topmost function has been patched).
946 Emitting assembly code: ``GCMetadataPrinter``
947 ---------------------------------------------
949 LLVM allows a plugin to print arbitrary assembly code before and after the rest
950 of a module's assembly code. At the end of the module, the GC can compile the
951 LLVM stack map into assembly code. (At the beginning, this information is not
954 Since AsmWriter and CodeGen are separate components of LLVM, a separate abstract
955 base class and registry is provided for printing assembly code, the
956 ``GCMetadaPrinter`` and ``GCMetadataPrinterRegistry``. The AsmWriter will look
957 for such a subclass if the ``GCStrategy`` sets ``UsesMetadata``:
965 This separation allows JIT-only clients to be smaller.
967 Note that LLVM does not currently have analogous APIs to support code generation
968 in the JIT, nor using the object writers.
972 // lib/MyGC/MyGCPrinter.cpp - Example LLVM GC printer
974 #include "llvm/CodeGen/GCMetadataPrinter.h"
975 #include "llvm/Support/Compiler.h"
977 using namespace llvm;
980 class LLVM_LIBRARY_VISIBILITY MyGCPrinter : public GCMetadataPrinter {
982 virtual void beginAssembly(AsmPrinter &AP);
984 virtual void finishAssembly(AsmPrinter &AP);
987 GCMetadataPrinterRegistry::Add<MyGCPrinter>
988 X("mygc", "My bespoke garbage collector.");
991 The collector should use ``AsmPrinter`` to print portable assembly code. The
992 collector itself contains the stack map for the entire module, and may access
993 the ``GCFunctionInfo`` using its own ``begin()`` and ``end()`` methods. Here's
998 #include "llvm/CodeGen/AsmPrinter.h"
999 #include "llvm/IR/Function.h"
1000 #include "llvm/IR/DataLayout.h"
1001 #include "llvm/Target/TargetAsmInfo.h"
1002 #include "llvm/Target/TargetMachine.h"
1004 void MyGCPrinter::beginAssembly(AsmPrinter &AP) {
1008 void MyGCPrinter::finishAssembly(AsmPrinter &AP) {
1009 MCStreamer &OS = AP.OutStreamer;
1010 unsigned IntPtrSize = AP.getPointerSize();
1012 // Put this in the data section.
1013 OS.SwitchSection(AP.getObjFileLowering().getDataSection());
1015 // For each function...
1016 for (iterator FI = begin(), FE = end(); FI != FE; ++FI) {
1017 GCFunctionInfo &MD = **FI;
1019 // A compact GC layout. Emit this data structure:
1022 // int32_t PointCount;
1023 // void *SafePointAddress[PointCount];
1024 // int32_t StackFrameSize; // in words
1025 // int32_t StackArity;
1026 // int32_t LiveCount;
1027 // int32_t LiveOffsets[LiveCount];
1028 // } __gcmap_<FUNCTIONNAME>;
1030 // Align to address width.
1031 AP.EmitAlignment(IntPtrSize == 4 ? 2 : 3);
1034 OS.AddComment("safe point count");
1035 AP.emitInt32(MD.size());
1037 // And each safe point...
1038 for (GCFunctionInfo::iterator PI = MD.begin(),
1039 PE = MD.end(); PI != PE; ++PI) {
1040 // Emit the address of the safe point.
1041 OS.AddComment("safe point address");
1042 MCSymbol *Label = PI->Label;
1043 AP.EmitLabelPlusOffset(Label/*Hi*/, 0/*Offset*/, 4/*Size*/);
1046 // Stack information never change in safe points! Only print info from the
1048 GCFunctionInfo::iterator PI = MD.begin();
1050 // Emit the stack frame size.
1051 OS.AddComment("stack frame size (in words)");
1052 AP.emitInt32(MD.getFrameSize() / IntPtrSize);
1054 // Emit stack arity, i.e. the number of stacked arguments.
1055 unsigned RegisteredArgs = IntPtrSize == 4 ? 5 : 6;
1056 unsigned StackArity = MD.getFunction().arg_size() > RegisteredArgs ?
1057 MD.getFunction().arg_size() - RegisteredArgs : 0;
1058 OS.AddComment("stack arity");
1059 AP.emitInt32(StackArity);
1061 // Emit the number of live roots in the function.
1062 OS.AddComment("live root count");
1063 AP.emitInt32(MD.live_size(PI));
1065 // And for each live root...
1066 for (GCFunctionInfo::live_iterator LI = MD.live_begin(PI),
1067 LE = MD.live_end(PI);
1069 // Emit live root's offset within the stack frame.
1070 OS.AddComment("stack index (offset / wordsize)");
1071 AP.emitInt32(LI->StackOffset);
1081 [Appel89] Runtime Tags Aren't Necessary. Andrew W. Appel. Lisp and Symbolic
1082 Computation 19(7):703-705, July 1989.
1086 [Goldberg91] Tag-free garbage collection for strongly typed programming
1087 languages. Benjamin Goldberg. ACM SIGPLAN PLDI'91.
1091 [Tolmach94] Tag-free garbage collection using explicit type parameters. Andrew
1092 Tolmach. Proceedings of the 1994 ACM conference on LISP and functional
1097 [Henderson2002] `Accurate Garbage Collection in an Uncooperative Environment
1098 <http://citeseer.ist.psu.edu/henderson02accurate.html>`__