1============================================== 2LLVM Atomic Instructions and Concurrency Guide 3============================================== 4 5.. contents:: 6 :local: 7 8Introduction 9============ 10 11LLVM supports instructions which are well-defined in the presence of threads and 12asynchronous signals. 13 14The atomic instructions are designed specifically to provide readable IR and 15optimized code generation for the following: 16 17* The C++11 ``<atomic>`` header. (`C++11 draft available here 18 <http://www.open-std.org/jtc1/sc22/wg21/>`_.) (`C11 draft available here 19 <http://www.open-std.org/jtc1/sc22/wg14/>`_.) 20 21* Proper semantics for Java-style memory, for both ``volatile`` and regular 22 shared variables. (`Java Specification 23 <http://docs.oracle.com/javase/specs/jls/se8/html/jls-17.html>`_) 24 25* gcc-compatible ``__sync_*`` builtins. (`Description 26 <https://gcc.gnu.org/onlinedocs/gcc/_005f_005fsync-Builtins.html>`_) 27 28* Other scenarios with atomic semantics, including ``static`` variables with 29 non-trivial constructors in C++. 30 31Atomic and volatile in the IR are orthogonal; "volatile" is the C/C++ volatile, 32which ensures that every volatile load and store happens and is performed in the 33stated order. A couple examples: if a SequentiallyConsistent store is 34immediately followed by another SequentiallyConsistent store to the same 35address, the first store can be erased. This transformation is not allowed for a 36pair of volatile stores. On the other hand, a non-volatile non-atomic load can 37be moved across a volatile load freely, but not an Acquire load. 38 39This document is intended to provide a guide to anyone either writing a frontend 40for LLVM or working on optimization passes for LLVM with a guide for how to deal 41with instructions with special semantics in the presence of concurrency. This 42is not intended to be a precise guide to the semantics; the details can get 43extremely complicated and unreadable, and are not usually necessary. 44 45.. _Optimization outside atomic: 46 47Optimization outside atomic 48=========================== 49 50The basic ``'load'`` and ``'store'`` allow a variety of optimizations, but can 51lead to undefined results in a concurrent environment; see `NotAtomic`_. This 52section specifically goes into the one optimizer restriction which applies in 53concurrent environments, which gets a bit more of an extended description 54because any optimization dealing with stores needs to be aware of it. 55 56From the optimizer's point of view, the rule is that if there are not any 57instructions with atomic ordering involved, concurrency does not matter, with 58one exception: if a variable might be visible to another thread or signal 59handler, a store cannot be inserted along a path where it might not execute 60otherwise. Take the following example: 61 62.. code-block:: c 63 64 /* C code, for readability; run through clang -O2 -S -emit-llvm to get 65 equivalent IR */ 66 int x; 67 void f(int* a) { 68 for (int i = 0; i < 100; i++) { 69 if (a[i]) 70 x += 1; 71 } 72 } 73 74The following is equivalent in non-concurrent situations: 75 76.. code-block:: c 77 78 int x; 79 void f(int* a) { 80 int xtemp = x; 81 for (int i = 0; i < 100; i++) { 82 if (a[i]) 83 xtemp += 1; 84 } 85 x = xtemp; 86 } 87 88However, LLVM is not allowed to transform the former to the latter: it could 89indirectly introduce undefined behavior if another thread can access ``x`` at 90the same time. (This example is particularly of interest because before the 91concurrency model was implemented, LLVM would perform this transformation.) 92 93Note that speculative loads are allowed; a load which is part of a race returns 94``undef``, but does not have undefined behavior. 95 96Atomic instructions 97=================== 98 99For cases where simple loads and stores are not sufficient, LLVM provides 100various atomic instructions. The exact guarantees provided depend on the 101ordering; see `Atomic orderings`_. 102 103``load atomic`` and ``store atomic`` provide the same basic functionality as 104non-atomic loads and stores, but provide additional guarantees in situations 105where threads and signals are involved. 106 107``cmpxchg`` and ``atomicrmw`` are essentially like an atomic load followed by an 108atomic store (where the store is conditional for ``cmpxchg``), but no other 109memory operation can happen on any thread between the load and store. 110 111A ``fence`` provides Acquire and/or Release ordering which is not part of 112another operation; it is normally used along with Monotonic memory operations. 113A Monotonic load followed by an Acquire fence is roughly equivalent to an 114Acquire load, and a Monotonic store following a Release fence is roughly 115equivalent to a Release store. SequentiallyConsistent fences behave as both 116an Acquire and a Release fence, and offer some additional complicated 117guarantees, see the C++11 standard for details. 118 119Frontends generating atomic instructions generally need to be aware of the 120target to some degree; atomic instructions are guaranteed to be lock-free, and 121therefore an instruction which is wider than the target natively supports can be 122impossible to generate. 123 124.. _Atomic orderings: 125 126Atomic orderings 127================ 128 129In order to achieve a balance between performance and necessary guarantees, 130there are six levels of atomicity. They are listed in order of strength; each 131level includes all the guarantees of the previous level except for 132Acquire/Release. (See also `LangRef Ordering <LangRef.html#ordering>`_.) 133 134.. _NotAtomic: 135 136NotAtomic 137--------- 138 139NotAtomic is the obvious, a load or store which is not atomic. (This isn't 140really a level of atomicity, but is listed here for comparison.) This is 141essentially a regular load or store. If there is a race on a given memory 142location, loads from that location return undef. 143 144Relevant standard 145 This is intended to match shared variables in C/C++, and to be used in any 146 other context where memory access is necessary, and a race is impossible. (The 147 precise definition is in `LangRef Memory Model <LangRef.html#memmodel>`_.) 148 149Notes for frontends 150 The rule is essentially that all memory accessed with basic loads and stores 151 by multiple threads should be protected by a lock or other synchronization; 152 otherwise, you are likely to run into undefined behavior. If your frontend is 153 for a "safe" language like Java, use Unordered to load and store any shared 154 variable. Note that NotAtomic volatile loads and stores are not properly 155 atomic; do not try to use them as a substitute. (Per the C/C++ standards, 156 volatile does provide some limited guarantees around asynchronous signals, but 157 atomics are generally a better solution.) 158 159Notes for optimizers 160 Introducing loads to shared variables along a codepath where they would not 161 otherwise exist is allowed; introducing stores to shared variables is not. See 162 `Optimization outside atomic`_. 163 164Notes for code generation 165 The one interesting restriction here is that it is not allowed to write to 166 bytes outside of the bytes relevant to a store. This is mostly relevant to 167 unaligned stores: it is not allowed in general to convert an unaligned store 168 into two aligned stores of the same width as the unaligned store. Backends are 169 also expected to generate an i8 store as an i8 store, and not an instruction 170 which writes to surrounding bytes. (If you are writing a backend for an 171 architecture which cannot satisfy these restrictions and cares about 172 concurrency, please send an email to llvm-dev.) 173 174Unordered 175--------- 176 177Unordered is the lowest level of atomicity. It essentially guarantees that races 178produce somewhat sane results instead of having undefined behavior. It also 179guarantees the operation to be lock-free, so it does not depend on the data 180being part of a special atomic structure or depend on a separate per-process 181global lock. Note that code generation will fail for unsupported atomic 182operations; if you need such an operation, use explicit locking. 183 184Relevant standard 185 This is intended to match the Java memory model for shared variables. 186 187Notes for frontends 188 This cannot be used for synchronization, but is useful for Java and other 189 "safe" languages which need to guarantee that the generated code never 190 exhibits undefined behavior. Note that this guarantee is cheap on common 191 platforms for loads of a native width, but can be expensive or unavailable for 192 wider loads, like a 64-bit store on ARM. (A frontend for Java or other "safe" 193 languages would normally split a 64-bit store on ARM into two 32-bit unordered 194 stores.) 195 196Notes for optimizers 197 In terms of the optimizer, this prohibits any transformation that transforms a 198 single load into multiple loads, transforms a store into multiple stores, 199 narrows a store, or stores a value which would not be stored otherwise. Some 200 examples of unsafe optimizations are narrowing an assignment into a bitfield, 201 rematerializing a load, and turning loads and stores into a memcpy 202 call. Reordering unordered operations is safe, though, and optimizers should 203 take advantage of that because unordered operations are common in languages 204 that need them. 205 206Notes for code generation 207 These operations are required to be atomic in the sense that if you use 208 unordered loads and unordered stores, a load cannot see a value which was 209 never stored. A normal load or store instruction is usually sufficient, but 210 note that an unordered load or store cannot be split into multiple 211 instructions (or an instruction which does multiple memory operations, like 212 ``LDRD`` on ARM without LPAE, or not naturally-aligned ``LDRD`` on LPAE ARM). 213 214Monotonic 215--------- 216 217Monotonic is the weakest level of atomicity that can be used in synchronization 218primitives, although it does not provide any general synchronization. It 219essentially guarantees that if you take all the operations affecting a specific 220address, a consistent ordering exists. 221 222Relevant standard 223 This corresponds to the C++11/C11 ``memory_order_relaxed``; see those 224 standards for the exact definition. 225 226Notes for frontends 227 If you are writing a frontend which uses this directly, use with caution. The 228 guarantees in terms of synchronization are very weak, so make sure these are 229 only used in a pattern which you know is correct. Generally, these would 230 either be used for atomic operations which do not protect other memory (like 231 an atomic counter), or along with a ``fence``. 232 233Notes for optimizers 234 In terms of the optimizer, this can be treated as a read+write on the relevant 235 memory location (and alias analysis will take advantage of that). In addition, 236 it is legal to reorder non-atomic and Unordered loads around Monotonic 237 loads. CSE/DSE and a few other optimizations are allowed, but Monotonic 238 operations are unlikely to be used in ways which would make those 239 optimizations useful. 240 241Notes for code generation 242 Code generation is essentially the same as that for unordered for loads and 243 stores. No fences are required. ``cmpxchg`` and ``atomicrmw`` are required 244 to appear as a single operation. 245 246Acquire 247------- 248 249Acquire provides a barrier of the sort necessary to acquire a lock to access 250other memory with normal loads and stores. 251 252Relevant standard 253 This corresponds to the C++11/C11 ``memory_order_acquire``. It should also be 254 used for C++11/C11 ``memory_order_consume``. 255 256Notes for frontends 257 If you are writing a frontend which uses this directly, use with caution. 258 Acquire only provides a semantic guarantee when paired with a Release 259 operation. 260 261Notes for optimizers 262 Optimizers not aware of atomics can treat this like a nothrow call. It is 263 also possible to move stores from before an Acquire load or read-modify-write 264 operation to after it, and move non-Acquire loads from before an Acquire 265 operation to after it. 266 267Notes for code generation 268 Architectures with weak memory ordering (essentially everything relevant today 269 except x86 and SPARC) require some sort of fence to maintain the Acquire 270 semantics. The precise fences required varies widely by architecture, but for 271 a simple implementation, most architectures provide a barrier which is strong 272 enough for everything (``dmb`` on ARM, ``sync`` on PowerPC, etc.). Putting 273 such a fence after the equivalent Monotonic operation is sufficient to 274 maintain Acquire semantics for a memory operation. 275 276Release 277------- 278 279Release is similar to Acquire, but with a barrier of the sort necessary to 280release a lock. 281 282Relevant standard 283 This corresponds to the C++11/C11 ``memory_order_release``. 284 285Notes for frontends 286 If you are writing a frontend which uses this directly, use with caution. 287 Release only provides a semantic guarantee when paired with a Acquire 288 operation. 289 290Notes for optimizers 291 Optimizers not aware of atomics can treat this like a nothrow call. It is 292 also possible to move loads from after a Release store or read-modify-write 293 operation to before it, and move non-Release stores from after an Release 294 operation to before it. 295 296Notes for code generation 297 See the section on Acquire; a fence before the relevant operation is usually 298 sufficient for Release. Note that a store-store fence is not sufficient to 299 implement Release semantics; store-store fences are generally not exposed to 300 IR because they are extremely difficult to use correctly. 301 302AcquireRelease 303-------------- 304 305AcquireRelease (``acq_rel`` in IR) provides both an Acquire and a Release 306barrier (for fences and operations which both read and write memory). 307 308Relevant standard 309 This corresponds to the C++11/C11 ``memory_order_acq_rel``. 310 311Notes for frontends 312 If you are writing a frontend which uses this directly, use with caution. 313 Acquire only provides a semantic guarantee when paired with a Release 314 operation, and vice versa. 315 316Notes for optimizers 317 In general, optimizers should treat this like a nothrow call; the possible 318 optimizations are usually not interesting. 319 320Notes for code generation 321 This operation has Acquire and Release semantics; see the sections on Acquire 322 and Release. 323 324SequentiallyConsistent 325---------------------- 326 327SequentiallyConsistent (``seq_cst`` in IR) provides Acquire semantics for loads 328and Release semantics for stores. Additionally, it guarantees that a total 329ordering exists between all SequentiallyConsistent operations. 330 331Relevant standard 332 This corresponds to the C++11/C11 ``memory_order_seq_cst``, Java volatile, and 333 the gcc-compatible ``__sync_*`` builtins which do not specify otherwise. 334 335Notes for frontends 336 If a frontend is exposing atomic operations, these are much easier to reason 337 about for the programmer than other kinds of operations, and using them is 338 generally a practical performance tradeoff. 339 340Notes for optimizers 341 Optimizers not aware of atomics can treat this like a nothrow call. For 342 SequentiallyConsistent loads and stores, the same reorderings are allowed as 343 for Acquire loads and Release stores, except that SequentiallyConsistent 344 operations may not be reordered. 345 346Notes for code generation 347 SequentiallyConsistent loads minimally require the same barriers as Acquire 348 operations and SequentiallyConsistent stores require Release 349 barriers. Additionally, the code generator must enforce ordering between 350 SequentiallyConsistent stores followed by SequentiallyConsistent loads. This 351 is usually done by emitting either a full fence before the loads or a full 352 fence after the stores; which is preferred varies by architecture. 353 354Atomics and IR optimization 355=========================== 356 357Predicates for optimizer writers to query: 358 359* ``isSimple()``: A load or store which is not volatile or atomic. This is 360 what, for example, memcpyopt would check for operations it might transform. 361 362* ``isUnordered()``: A load or store which is not volatile and at most 363 Unordered. This would be checked, for example, by LICM before hoisting an 364 operation. 365 366* ``mayReadFromMemory()``/``mayWriteToMemory()``: Existing predicate, but note 367 that they return true for any operation which is volatile or at least 368 Monotonic. 369 370* ``isStrongerThan`` / ``isAtLeastOrStrongerThan``: These are predicates on 371 orderings. They can be useful for passes that are aware of atomics, for 372 example to do DSE across a single atomic access, but not across a 373 release-acquire pair (see MemoryDependencyAnalysis for an example of this) 374 375* Alias analysis: Note that AA will return ModRef for anything Acquire or 376 Release, and for the address accessed by any Monotonic operation. 377 378To support optimizing around atomic operations, make sure you are using the 379right predicates; everything should work if that is done. If your pass should 380optimize some atomic operations (Unordered operations in particular), make sure 381it doesn't replace an atomic load or store with a non-atomic operation. 382 383Some examples of how optimizations interact with various kinds of atomic 384operations: 385 386* ``memcpyopt``: An atomic operation cannot be optimized into part of a 387 memcpy/memset, including unordered loads/stores. It can pull operations 388 across some atomic operations. 389 390* LICM: Unordered loads/stores can be moved out of a loop. It just treats 391 monotonic operations like a read+write to a memory location, and anything 392 stricter than that like a nothrow call. 393 394* DSE: Unordered stores can be DSE'ed like normal stores. Monotonic stores can 395 be DSE'ed in some cases, but it's tricky to reason about, and not especially 396 important. It is possible in some case for DSE to operate across a stronger 397 atomic operation, but it is fairly tricky. DSE delegates this reasoning to 398 MemoryDependencyAnalysis (which is also used by other passes like GVN). 399 400* Folding a load: Any atomic load from a constant global can be constant-folded, 401 because it cannot be observed. Similar reasoning allows sroa with 402 atomic loads and stores. 403 404Atomics and Codegen 405=================== 406 407Atomic operations are represented in the SelectionDAG with ``ATOMIC_*`` opcodes. 408On architectures which use barrier instructions for all atomic ordering (like 409ARM), appropriate fences can be emitted by the AtomicExpand Codegen pass if 410``setInsertFencesForAtomic()`` was used. 411 412The MachineMemOperand for all atomic operations is currently marked as volatile; 413this is not correct in the IR sense of volatile, but CodeGen handles anything 414marked volatile very conservatively. This should get fixed at some point. 415 416One very important property of the atomic operations is that if your backend 417supports any inline lock-free atomic operations of a given size, you should 418support *ALL* operations of that size in a lock-free manner. 419 420When the target implements atomic ``cmpxchg`` or LL/SC instructions (as most do) 421this is trivial: all the other operations can be implemented on top of those 422primitives. However, on many older CPUs (e.g. ARMv5, SparcV8, Intel 80386) there 423are atomic load and store instructions, but no ``cmpxchg`` or LL/SC. As it is 424invalid to implement ``atomic load`` using the native instruction, but 425``cmpxchg`` using a library call to a function that uses a mutex, ``atomic 426load`` must *also* expand to a library call on such architectures, so that it 427can remain atomic with regards to a simultaneous ``cmpxchg``, by using the same 428mutex. 429 430AtomicExpandPass can help with that: it will expand all atomic operations to the 431proper ``__atomic_*`` libcalls for any size above the maximum set by 432``setMaxAtomicSizeInBitsSupported`` (which defaults to 0). 433 434On x86, all atomic loads generate a ``MOV``. SequentiallyConsistent stores 435generate an ``XCHG``, other stores generate a ``MOV``. SequentiallyConsistent 436fences generate an ``MFENCE``, other fences do not cause any code to be 437generated. ``cmpxchg`` uses the ``LOCK CMPXCHG`` instruction. ``atomicrmw xchg`` 438uses ``XCHG``, ``atomicrmw add`` and ``atomicrmw sub`` use ``XADD``, and all 439other ``atomicrmw`` operations generate a loop with ``LOCK CMPXCHG``. Depending 440on the users of the result, some ``atomicrmw`` operations can be translated into 441operations like ``LOCK AND``, but that does not work in general. 442 443On ARM (before v8), MIPS, and many other RISC architectures, Acquire, Release, 444and SequentiallyConsistent semantics require barrier instructions for every such 445operation. Loads and stores generate normal instructions. ``cmpxchg`` and 446``atomicrmw`` can be represented using a loop with LL/SC-style instructions 447which take some sort of exclusive lock on a cache line (``LDREX`` and ``STREX`` 448on ARM, etc.). 449 450It is often easiest for backends to use AtomicExpandPass to lower some of the 451atomic constructs. Here are some lowerings it can do: 452 453* cmpxchg -> loop with load-linked/store-conditional 454 by overriding ``shouldExpandAtomicCmpXchgInIR()``, ``emitLoadLinked()``, 455 ``emitStoreConditional()`` 456* large loads/stores -> ll-sc/cmpxchg 457 by overriding ``shouldExpandAtomicStoreInIR()``/``shouldExpandAtomicLoadInIR()`` 458* strong atomic accesses -> monotonic accesses + fences by overriding 459 ``shouldInsertFencesForAtomic()``, ``emitLeadingFence()``, and 460 ``emitTrailingFence()`` 461* atomic rmw -> loop with cmpxchg or load-linked/store-conditional 462 by overriding ``expandAtomicRMWInIR()`` 463* expansion to __atomic_* libcalls for unsupported sizes. 464 465For an example of all of these, look at the ARM backend. 466 467Libcalls: __atomic_* 468==================== 469 470There are two kinds of atomic library calls that are generated by LLVM. Please 471note that both sets of library functions somewhat confusingly share the names of 472builtin functions defined by clang. Despite this, the library functions are 473not directly related to the builtins: it is *not* the case that ``__atomic_*`` 474builtins lower to ``__atomic_*`` library calls and ``__sync_*`` builtins lower 475to ``__sync_*`` library calls. 476 477The first set of library functions are named ``__atomic_*``. This set has been 478"standardized" by GCC, and is described below. (See also `GCC's documentation 479<https://gcc.gnu.org/wiki/Atomic/GCCMM/LIbrary>`_) 480 481LLVM's AtomicExpandPass will translate atomic operations on data sizes above 482``MaxAtomicSizeInBitsSupported`` into calls to these functions. 483 484There are four generic functions, which can be called with data of any size or 485alignment:: 486 487 void __atomic_load(size_t size, void *ptr, void *ret, int ordering) 488 void __atomic_store(size_t size, void *ptr, void *val, int ordering) 489 void __atomic_exchange(size_t size, void *ptr, void *val, void *ret, int ordering) 490 bool __atomic_compare_exchange(size_t size, void *ptr, void *expected, void *desired, int success_order, int failure_order) 491 492There are also size-specialized versions of the above functions, which can only 493be used with *naturally-aligned* pointers of the appropriate size. In the 494signatures below, "N" is one of 1, 2, 4, 8, and 16, and "iN" is the appropriate 495integer type of that size; if no such integer type exists, the specialization 496cannot be used:: 497 498 iN __atomic_load_N(iN *ptr, iN val, int ordering) 499 void __atomic_store_N(iN *ptr, iN val, int ordering) 500 iN __atomic_exchange_N(iN *ptr, iN val, int ordering) 501 bool __atomic_compare_exchange_N(iN *ptr, iN *expected, iN desired, int success_order, int failure_order) 502 503Finally there are some read-modify-write functions, which are only available in 504the size-specific variants (any other sizes use a ``__atomic_compare_exchange`` 505loop):: 506 507 iN __atomic_fetch_add_N(iN *ptr, iN val, int ordering) 508 iN __atomic_fetch_sub_N(iN *ptr, iN val, int ordering) 509 iN __atomic_fetch_and_N(iN *ptr, iN val, int ordering) 510 iN __atomic_fetch_or_N(iN *ptr, iN val, int ordering) 511 iN __atomic_fetch_xor_N(iN *ptr, iN val, int ordering) 512 iN __atomic_fetch_nand_N(iN *ptr, iN val, int ordering) 513 514This set of library functions have some interesting implementation requirements 515to take note of: 516 517- They support all sizes and alignments -- including those which cannot be 518 implemented natively on any existing hardware. Therefore, they will certainly 519 use mutexes in for some sizes/alignments. 520 521- As a consequence, they cannot be shipped in a statically linked 522 compiler-support library, as they have state which must be shared amongst all 523 DSOs loaded in the program. They must be provided in a shared library used by 524 all objects. 525 526- The set of atomic sizes supported lock-free must be a superset of the sizes 527 any compiler can emit. That is: if a new compiler introduces support for 528 inline-lock-free atomics of size N, the ``__atomic_*`` functions must also have a 529 lock-free implementation for size N. This is a requirement so that code 530 produced by an old compiler (which will have called the ``__atomic_*`` function) 531 interoperates with code produced by the new compiler (which will use native 532 the atomic instruction). 533 534Note that it's possible to write an entirely target-independent implementation 535of these library functions by using the compiler atomic builtins themselves to 536implement the operations on naturally-aligned pointers of supported sizes, and a 537generic mutex implementation otherwise. 538 539Libcalls: __sync_* 540================== 541 542Some targets or OS/target combinations can support lock-free atomics, but for 543various reasons, it is not practical to emit the instructions inline. 544 545There's two typical examples of this. 546 547Some CPUs support multiple instruction sets which can be swiched back and forth 548on function-call boundaries. For example, MIPS supports the MIPS16 ISA, which 549has a smaller instruction encoding than the usual MIPS32 ISA. ARM, similarly, 550has the Thumb ISA. In MIPS16 and earlier versions of Thumb, the atomic 551instructions are not encodable. However, those instructions are available via a 552function call to a function with the longer encoding. 553 554Additionally, a few OS/target pairs provide kernel-supported lock-free 555atomics. ARM/Linux is an example of this: the kernel `provides 556<https://www.kernel.org/doc/Documentation/arm/kernel_user_helpers.txt>`_ a 557function which on older CPUs contains a "magically-restartable" atomic sequence 558(which looks atomic so long as there's only one CPU), and contains actual atomic 559instructions on newer multicore models. This sort of functionality can typically 560be provided on any architecture, if all CPUs which are missing atomic 561compare-and-swap support are uniprocessor (no SMP). This is almost always the 562case. The only common architecture without that property is SPARC -- SPARCV8 SMP 563systems were common, yet it doesn't support any sort of compare-and-swap 564operation. 565 566In either of these cases, the Target in LLVM can claim support for atomics of an 567appropriate size, and then implement some subset of the operations via libcalls 568to a ``__sync_*`` function. Such functions *must* not use locks in their 569implementation, because unlike the ``__atomic_*`` routines used by 570AtomicExpandPass, these may be mixed-and-matched with native instructions by the 571target lowering. 572 573Further, these routines do not need to be shared, as they are stateless. So, 574there is no issue with having multiple copies included in one binary. Thus, 575typically these routines are implemented by the statically-linked compiler 576runtime support library. 577 578LLVM will emit a call to an appropriate ``__sync_*`` routine if the target 579ISelLowering code has set the corresponding ``ATOMIC_CMPXCHG``, ``ATOMIC_SWAP``, 580or ``ATOMIC_LOAD_*`` operation to "Expand", and if it has opted-into the 581availability of those library functions via a call to ``initSyncLibcalls()``. 582 583The full set of functions that may be called by LLVM is (for ``N`` being 1, 2, 5844, 8, or 16):: 585 586 iN __sync_val_compare_and_swap_N(iN *ptr, iN expected, iN desired) 587 iN __sync_lock_test_and_set_N(iN *ptr, iN val) 588 iN __sync_fetch_and_add_N(iN *ptr, iN val) 589 iN __sync_fetch_and_sub_N(iN *ptr, iN val) 590 iN __sync_fetch_and_and_N(iN *ptr, iN val) 591 iN __sync_fetch_and_or_N(iN *ptr, iN val) 592 iN __sync_fetch_and_xor_N(iN *ptr, iN val) 593 iN __sync_fetch_and_nand_N(iN *ptr, iN val) 594 iN __sync_fetch_and_max_N(iN *ptr, iN val) 595 iN __sync_fetch_and_umax_N(iN *ptr, iN val) 596 iN __sync_fetch_and_min_N(iN *ptr, iN val) 597 iN __sync_fetch_and_umin_N(iN *ptr, iN val) 598 599This list doesn't include any function for atomic load or store; all known 600architectures support atomic loads and stores directly (possibly by emitting a 601fence on either side of a normal load or store.) 602 603There's also, somewhat separately, the possibility to lower ``ATOMIC_FENCE`` to 604``__sync_synchronize()``. This may happen or not happen independent of all the 605above, controlled purely by ``setOperationAction(ISD::ATOMIC_FENCE, ...)``. 606