1 ==============================
2 LLVM Language Reference Manual
3 ==============================
12 This document is a reference manual for the LLVM assembly language. LLVM
13 is a Static Single Assignment (SSA) based representation that provides
14 type safety, low-level operations, flexibility, and the capability of
15 representing 'all' high-level languages cleanly. It is the common code
16 representation used throughout all phases of the LLVM compilation
22 The LLVM code representation is designed to be used in three different
23 forms: as an in-memory compiler IR, as an on-disk bitcode representation
24 (suitable for fast loading by a Just-In-Time compiler), and as a human
25 readable assembly language representation. This allows LLVM to provide a
26 powerful intermediate representation for efficient compiler
27 transformations and analysis, while providing a natural means to debug
28 and visualize the transformations. The three different forms of LLVM are
29 all equivalent. This document describes the human readable
30 representation and notation.
32 The LLVM representation aims to be light-weight and low-level while
33 being expressive, typed, and extensible at the same time. It aims to be
34 a "universal IR" of sorts, by being at a low enough level that
35 high-level ideas may be cleanly mapped to it (similar to how
36 microprocessors are "universal IR's", allowing many source languages to
37 be mapped to them). By providing type information, LLVM can be used as
38 the target of optimizations: for example, through pointer analysis, it
39 can be proven that a C automatic variable is never accessed outside of
40 the current function, allowing it to be promoted to a simple SSA value
41 instead of a memory location.
48 It is important to note that this document describes 'well formed' LLVM
49 assembly language. There is a difference between what the parser accepts
50 and what is considered 'well formed'. For example, the following
51 instruction is syntactically okay, but not well formed:
57 because the definition of ``%x`` does not dominate all of its uses. The
58 LLVM infrastructure provides a verification pass that may be used to
59 verify that an LLVM module is well formed. This pass is automatically
60 run by the parser after parsing input assembly and by the optimizer
61 before it outputs bitcode. The violations pointed out by the verifier
62 pass indicate bugs in transformation passes or input to the parser.
69 LLVM identifiers come in two basic types: global and local. Global
70 identifiers (functions, global variables) begin with the ``'@'``
71 character. Local identifiers (register names, types) begin with the
72 ``'%'`` character. Additionally, there are three different formats for
73 identifiers, for different purposes:
75 #. Named values are represented as a string of characters with their
76 prefix. For example, ``%foo``, ``@DivisionByZero``,
77 ``%a.really.long.identifier``. The actual regular expression used is
78 '``[%@][a-zA-Z$._][a-zA-Z$._0-9]*``'. Identifiers that require other
79 characters in their names can be surrounded with quotes. Special
80 characters may be escaped using ``"\xx"`` where ``xx`` is the ASCII
81 code for the character in hexadecimal. In this way, any character can
82 be used in a name value, even quotes themselves. The ``"\01"`` prefix
83 can be used on global variables to suppress mangling.
84 #. Unnamed values are represented as an unsigned numeric value with
85 their prefix. For example, ``%12``, ``@2``, ``%44``.
86 #. Constants, which are described in the section Constants_ below.
88 LLVM requires that values start with a prefix for two reasons: Compilers
89 don't need to worry about name clashes with reserved words, and the set
90 of reserved words may be expanded in the future without penalty.
91 Additionally, unnamed identifiers allow a compiler to quickly come up
92 with a temporary variable without having to avoid symbol table
95 Reserved words in LLVM are very similar to reserved words in other
96 languages. There are keywords for different opcodes ('``add``',
97 '``bitcast``', '``ret``', etc...), for primitive type names ('``void``',
98 '``i32``', etc...), and others. These reserved words cannot conflict
99 with variable names, because none of them start with a prefix character
100 (``'%'`` or ``'@'``).
102 Here is an example of LLVM code to multiply the integer variable
109 %result = mul i32 %X, 8
111 After strength reduction:
115 %result = shl i32 %X, 3
121 %0 = add i32 %X, %X ; yields i32:%0
122 %1 = add i32 %0, %0 ; yields i32:%1
123 %result = add i32 %1, %1
125 This last way of multiplying ``%X`` by 8 illustrates several important
126 lexical features of LLVM:
128 #. Comments are delimited with a '``;``' and go until the end of line.
129 #. Unnamed temporaries are created when the result of a computation is
130 not assigned to a named value.
131 #. Unnamed temporaries are numbered sequentially (using a per-function
132 incrementing counter, starting with 0). Note that basic blocks and unnamed
133 function parameters are included in this numbering. For example, if the
134 entry basic block is not given a label name and all function parameters are
135 named, then it will get number 0.
137 It also shows a convention that we follow in this document. When
138 demonstrating instructions, we will follow an instruction with a comment
139 that defines the type and name of value produced.
147 LLVM programs are composed of ``Module``'s, each of which is a
148 translation unit of the input programs. Each module consists of
149 functions, global variables, and symbol table entries. Modules may be
150 combined together with the LLVM linker, which merges function (and
151 global variable) definitions, resolves forward declarations, and merges
152 symbol table entries. Here is an example of the "hello world" module:
156 ; Declare the string constant as a global constant.
157 @.str = private unnamed_addr constant [13 x i8] c"hello world\0A\00"
159 ; External declaration of the puts function
160 declare i32 @puts(i8* nocapture) nounwind
162 ; Definition of main function
163 define i32 @main() { ; i32()*
164 ; Convert [13 x i8]* to i8 *...
165 %cast210 = getelementptr [13 x i8]* @.str, i64 0, i64 0
167 ; Call puts function to write out the string to stdout.
168 call i32 @puts(i8* %cast210)
173 !0 = metadata !{i32 42, null, metadata !"string"}
176 This example is made up of a :ref:`global variable <globalvars>` named
177 "``.str``", an external declaration of the "``puts``" function, a
178 :ref:`function definition <functionstructure>` for "``main``" and
179 :ref:`named metadata <namedmetadatastructure>` "``foo``".
181 In general, a module is made up of a list of global values (where both
182 functions and global variables are global values). Global values are
183 represented by a pointer to a memory location (in this case, a pointer
184 to an array of char, and a pointer to a function), and have one of the
185 following :ref:`linkage types <linkage>`.
192 All Global Variables and Functions have one of the following types of
196 Global values with "``private``" linkage are only directly
197 accessible by objects in the current module. In particular, linking
198 code into a module with an private global value may cause the
199 private to be renamed as necessary to avoid collisions. Because the
200 symbol is private to the module, all references can be updated. This
201 doesn't show up in any symbol table in the object file.
203 Similar to private, but the value shows as a local symbol
204 (``STB_LOCAL`` in the case of ELF) in the object file. This
205 corresponds to the notion of the '``static``' keyword in C.
206 ``available_externally``
207 Globals with "``available_externally``" linkage are never emitted
208 into the object file corresponding to the LLVM module. They exist to
209 allow inlining and other optimizations to take place given knowledge
210 of the definition of the global, which is known to be somewhere
211 outside the module. Globals with ``available_externally`` linkage
212 are allowed to be discarded at will, and are otherwise the same as
213 ``linkonce_odr``. This linkage type is only allowed on definitions,
216 Globals with "``linkonce``" linkage are merged with other globals of
217 the same name when linkage occurs. This can be used to implement
218 some forms of inline functions, templates, or other code which must
219 be generated in each translation unit that uses it, but where the
220 body may be overridden with a more definitive definition later.
221 Unreferenced ``linkonce`` globals are allowed to be discarded. Note
222 that ``linkonce`` linkage does not actually allow the optimizer to
223 inline the body of this function into callers because it doesn't
224 know if this definition of the function is the definitive definition
225 within the program or whether it will be overridden by a stronger
226 definition. To enable inlining and other optimizations, use
227 "``linkonce_odr``" linkage.
229 "``weak``" linkage has the same merging semantics as ``linkonce``
230 linkage, except that unreferenced globals with ``weak`` linkage may
231 not be discarded. This is used for globals that are declared "weak"
234 "``common``" linkage is most similar to "``weak``" linkage, but they
235 are used for tentative definitions in C, such as "``int X;``" at
236 global scope. Symbols with "``common``" linkage are merged in the
237 same way as ``weak symbols``, and they may not be deleted if
238 unreferenced. ``common`` symbols may not have an explicit section,
239 must have a zero initializer, and may not be marked
240 ':ref:`constant <globalvars>`'. Functions and aliases may not have
243 .. _linkage_appending:
246 "``appending``" linkage may only be applied to global variables of
247 pointer to array type. When two global variables with appending
248 linkage are linked together, the two global arrays are appended
249 together. This is the LLVM, typesafe, equivalent of having the
250 system linker append together "sections" with identical names when
253 The semantics of this linkage follow the ELF object file model: the
254 symbol is weak until linked, if not linked, the symbol becomes null
255 instead of being an undefined reference.
256 ``linkonce_odr``, ``weak_odr``
257 Some languages allow differing globals to be merged, such as two
258 functions with different semantics. Other languages, such as
259 ``C++``, ensure that only equivalent globals are ever merged (the
260 "one definition rule" --- "ODR"). Such languages can use the
261 ``linkonce_odr`` and ``weak_odr`` linkage types to indicate that the
262 global will only be merged with equivalent globals. These linkage
263 types are otherwise the same as their non-``odr`` versions.
265 If none of the above identifiers are used, the global is externally
266 visible, meaning that it participates in linkage and can be used to
267 resolve external symbol references.
269 It is illegal for a function *declaration* to have any linkage type
270 other than ``external`` or ``extern_weak``.
277 LLVM :ref:`functions <functionstructure>`, :ref:`calls <i_call>` and
278 :ref:`invokes <i_invoke>` can all have an optional calling convention
279 specified for the call. The calling convention of any pair of dynamic
280 caller/callee must match, or the behavior of the program is undefined.
281 The following calling conventions are supported by LLVM, and more may be
284 "``ccc``" - The C calling convention
285 This calling convention (the default if no other calling convention
286 is specified) matches the target C calling conventions. This calling
287 convention supports varargs function calls and tolerates some
288 mismatch in the declared prototype and implemented declaration of
289 the function (as does normal C).
290 "``fastcc``" - The fast calling convention
291 This calling convention attempts to make calls as fast as possible
292 (e.g. by passing things in registers). This calling convention
293 allows the target to use whatever tricks it wants to produce fast
294 code for the target, without having to conform to an externally
295 specified ABI (Application Binary Interface). `Tail calls can only
296 be optimized when this, the GHC or the HiPE convention is
297 used. <CodeGenerator.html#id80>`_ This calling convention does not
298 support varargs and requires the prototype of all callees to exactly
299 match the prototype of the function definition.
300 "``coldcc``" - The cold calling convention
301 This calling convention attempts to make code in the caller as
302 efficient as possible under the assumption that the call is not
303 commonly executed. As such, these calls often preserve all registers
304 so that the call does not break any live ranges in the caller side.
305 This calling convention does not support varargs and requires the
306 prototype of all callees to exactly match the prototype of the
307 function definition. Furthermore the inliner doesn't consider such function
309 "``cc 10``" - GHC convention
310 This calling convention has been implemented specifically for use by
311 the `Glasgow Haskell Compiler (GHC) <http://www.haskell.org/ghc>`_.
312 It passes everything in registers, going to extremes to achieve this
313 by disabling callee save registers. This calling convention should
314 not be used lightly but only for specific situations such as an
315 alternative to the *register pinning* performance technique often
316 used when implementing functional programming languages. At the
317 moment only X86 supports this convention and it has the following
320 - On *X86-32* only supports up to 4 bit type parameters. No
321 floating point types are supported.
322 - On *X86-64* only supports up to 10 bit type parameters and 6
323 floating point parameters.
325 This calling convention supports `tail call
326 optimization <CodeGenerator.html#id80>`_ but requires both the
327 caller and callee are using it.
328 "``cc 11``" - The HiPE calling convention
329 This calling convention has been implemented specifically for use by
330 the `High-Performance Erlang
331 (HiPE) <http://www.it.uu.se/research/group/hipe/>`_ compiler, *the*
332 native code compiler of the `Ericsson's Open Source Erlang/OTP
333 system <http://www.erlang.org/download.shtml>`_. It uses more
334 registers for argument passing than the ordinary C calling
335 convention and defines no callee-saved registers. The calling
336 convention properly supports `tail call
337 optimization <CodeGenerator.html#id80>`_ but requires that both the
338 caller and the callee use it. It uses a *register pinning*
339 mechanism, similar to GHC's convention, for keeping frequently
340 accessed runtime components pinned to specific hardware registers.
341 At the moment only X86 supports this convention (both 32 and 64
343 "``webkit_jscc``" - WebKit's JavaScript calling convention
344 This calling convention has been implemented for `WebKit FTL JIT
345 <https://trac.webkit.org/wiki/FTLJIT>`_. It passes arguments on the
346 stack right to left (as cdecl does), and returns a value in the
347 platform's customary return register.
348 "``anyregcc``" - Dynamic calling convention for code patching
349 This is a special convention that supports patching an arbitrary code
350 sequence in place of a call site. This convention forces the call
351 arguments into registers but allows them to be dynamcially
352 allocated. This can currently only be used with calls to
353 llvm.experimental.patchpoint because only this intrinsic records
354 the location of its arguments in a side table. See :doc:`StackMaps`.
355 "``preserve_mostcc``" - The `PreserveMost` calling convention
356 This calling convention attempts to make the code in the caller as little
357 intrusive as possible. This calling convention behaves identical to the `C`
358 calling convention on how arguments and return values are passed, but it
359 uses a different set of caller/callee-saved registers. This alleviates the
360 burden of saving and recovering a large register set before and after the
361 call in the caller. If the arguments are passed in callee-saved registers,
362 then they will be preserved by the callee across the call. This doesn't
363 apply for values returned in callee-saved registers.
365 - On X86-64 the callee preserves all general purpose registers, except for
366 R11. R11 can be used as a scratch register. Floating-point registers
367 (XMMs/YMMs) are not preserved and need to be saved by the caller.
369 The idea behind this convention is to support calls to runtime functions
370 that have a hot path and a cold path. The hot path is usually a small piece
371 of code that doesn't many registers. The cold path might need to call out to
372 another function and therefore only needs to preserve the caller-saved
373 registers, which haven't already been saved by the caller. The
374 `PreserveMost` calling convention is very similar to the `cold` calling
375 convention in terms of caller/callee-saved registers, but they are used for
376 different types of function calls. `coldcc` is for function calls that are
377 rarely executed, whereas `preserve_mostcc` function calls are intended to be
378 on the hot path and definitely executed a lot. Furthermore `preserve_mostcc`
379 doesn't prevent the inliner from inlining the function call.
381 This calling convention will be used by a future version of the ObjectiveC
382 runtime and should therefore still be considered experimental at this time.
383 Although this convention was created to optimize certain runtime calls to
384 the ObjectiveC runtime, it is not limited to this runtime and might be used
385 by other runtimes in the future too. The current implementation only
386 supports X86-64, but the intention is to support more architectures in the
388 "``preserve_allcc``" - The `PreserveAll` calling convention
389 This calling convention attempts to make the code in the caller even less
390 intrusive than the `PreserveMost` calling convention. This calling
391 convention also behaves identical to the `C` calling convention on how
392 arguments and return values are passed, but it uses a different set of
393 caller/callee-saved registers. This removes the burden of saving and
394 recovering a large register set before and after the call in the caller. If
395 the arguments are passed in callee-saved registers, then they will be
396 preserved by the callee across the call. This doesn't apply for values
397 returned in callee-saved registers.
399 - On X86-64 the callee preserves all general purpose registers, except for
400 R11. R11 can be used as a scratch register. Furthermore it also preserves
401 all floating-point registers (XMMs/YMMs).
403 The idea behind this convention is to support calls to runtime functions
404 that don't need to call out to any other functions.
406 This calling convention, like the `PreserveMost` calling convention, will be
407 used by a future version of the ObjectiveC runtime and should be considered
408 experimental at this time.
409 "``cc <n>``" - Numbered convention
410 Any calling convention may be specified by number, allowing
411 target-specific calling conventions to be used. Target specific
412 calling conventions start at 64.
414 More calling conventions can be added/defined on an as-needed basis, to
415 support Pascal conventions or any other well-known target-independent
418 .. _visibilitystyles:
423 All Global Variables and Functions have one of the following visibility
426 "``default``" - Default style
427 On targets that use the ELF object file format, default visibility
428 means that the declaration is visible to other modules and, in
429 shared libraries, means that the declared entity may be overridden.
430 On Darwin, default visibility means that the declaration is visible
431 to other modules. Default visibility corresponds to "external
432 linkage" in the language.
433 "``hidden``" - Hidden style
434 Two declarations of an object with hidden visibility refer to the
435 same object if they are in the same shared object. Usually, hidden
436 visibility indicates that the symbol will not be placed into the
437 dynamic symbol table, so no other module (executable or shared
438 library) can reference it directly.
439 "``protected``" - Protected style
440 On ELF, protected visibility indicates that the symbol will be
441 placed in the dynamic symbol table, but that references within the
442 defining module will bind to the local symbol. That is, the symbol
443 cannot be overridden by another module.
445 A symbol with ``internal`` or ``private`` linkage must have ``default``
453 All Global Variables, Functions and Aliases can have one of the following
457 "``dllimport``" causes the compiler to reference a function or variable via
458 a global pointer to a pointer that is set up by the DLL exporting the
459 symbol. On Microsoft Windows targets, the pointer name is formed by
460 combining ``__imp_`` and the function or variable name.
462 "``dllexport``" causes the compiler to provide a global pointer to a pointer
463 in a DLL, so that it can be referenced with the ``dllimport`` attribute. On
464 Microsoft Windows targets, the pointer name is formed by combining
465 ``__imp_`` and the function or variable name. Since this storage class
466 exists for defining a dll interface, the compiler, assembler and linker know
467 it is externally referenced and must refrain from deleting the symbol.
471 Thread Local Storage Models
472 ---------------------------
474 A variable may be defined as ``thread_local``, which means that it will
475 not be shared by threads (each thread will have a separated copy of the
476 variable). Not all targets support thread-local variables. Optionally, a
477 TLS model may be specified:
480 For variables that are only used within the current shared library.
482 For variables in modules that will not be loaded dynamically.
484 For variables defined in the executable and only used within it.
486 If no explicit model is given, the "general dynamic" model is used.
488 The models correspond to the ELF TLS models; see `ELF Handling For
489 Thread-Local Storage <http://people.redhat.com/drepper/tls.pdf>`_ for
490 more information on under which circumstances the different models may
491 be used. The target may choose a different TLS model if the specified
492 model is not supported, or if a better choice of model can be made.
494 A model can also be specified in a alias, but then it only governs how
495 the alias is accessed. It will not have any effect in the aliasee.
502 LLVM IR allows you to specify both "identified" and "literal" :ref:`structure
503 types <t_struct>`. Literal types are uniqued structurally, but identified types
504 are never uniqued. An :ref:`opaque structural type <t_opaque>` can also be used
505 to forward declare a type that is not yet available.
507 An example of a identified structure specification is:
511 %mytype = type { %mytype*, i32 }
513 Prior to the LLVM 3.0 release, identified types were structurally uniqued. Only
514 literal types are uniqued in recent versions of LLVM.
521 Global variables define regions of memory allocated at compilation time
524 Global variables definitions must be initialized.
526 Global variables in other translation units can also be declared, in which
527 case they don't have an initializer.
529 Either global variable definitions or declarations may have an explicit section
530 to be placed in and may have an optional explicit alignment specified.
532 A variable may be defined as a global ``constant``, which indicates that
533 the contents of the variable will **never** be modified (enabling better
534 optimization, allowing the global data to be placed in the read-only
535 section of an executable, etc). Note that variables that need runtime
536 initialization cannot be marked ``constant`` as there is a store to the
539 LLVM explicitly allows *declarations* of global variables to be marked
540 constant, even if the final definition of the global is not. This
541 capability can be used to enable slightly better optimization of the
542 program, but requires the language definition to guarantee that
543 optimizations based on the 'constantness' are valid for the translation
544 units that do not include the definition.
546 As SSA values, global variables define pointer values that are in scope
547 (i.e. they dominate) all basic blocks in the program. Global variables
548 always define a pointer to their "content" type because they describe a
549 region of memory, and all memory objects in LLVM are accessed through
552 Global variables can be marked with ``unnamed_addr`` which indicates
553 that the address is not significant, only the content. Constants marked
554 like this can be merged with other constants if they have the same
555 initializer. Note that a constant with significant address *can* be
556 merged with a ``unnamed_addr`` constant, the result being a constant
557 whose address is significant.
559 A global variable may be declared to reside in a target-specific
560 numbered address space. For targets that support them, address spaces
561 may affect how optimizations are performed and/or what target
562 instructions are used to access the variable. The default address space
563 is zero. The address space qualifier must precede any other attributes.
565 LLVM allows an explicit section to be specified for globals. If the
566 target supports it, it will emit globals to the section specified.
567 Additionally, the global can placed in a comdat if the target has the necessary
570 By default, global initializers are optimized by assuming that global
571 variables defined within the module are not modified from their
572 initial values before the start of the global initializer. This is
573 true even for variables potentially accessible from outside the
574 module, including those with external linkage or appearing in
575 ``@llvm.used`` or dllexported variables. This assumption may be suppressed
576 by marking the variable with ``externally_initialized``.
578 An explicit alignment may be specified for a global, which must be a
579 power of 2. If not present, or if the alignment is set to zero, the
580 alignment of the global is set by the target to whatever it feels
581 convenient. If an explicit alignment is specified, the global is forced
582 to have exactly that alignment. Targets and optimizers are not allowed
583 to over-align the global if the global has an assigned section. In this
584 case, the extra alignment could be observable: for example, code could
585 assume that the globals are densely packed in their section and try to
586 iterate over them as an array, alignment padding would break this
587 iteration. The maximum alignment is ``1 << 29``.
589 Globals can also have a :ref:`DLL storage class <dllstorageclass>`.
591 Variables and aliasaes can have a
592 :ref:`Thread Local Storage Model <tls_model>`.
596 [@<GlobalVarName> =] [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal]
597 [unnamed_addr] [AddrSpace] [ExternallyInitialized]
598 <global | constant> <Type> [<InitializerConstant>]
599 [, section "name"] [, align <Alignment>]
601 For example, the following defines a global in a numbered address space
602 with an initializer, section, and alignment:
606 @G = addrspace(5) constant float 1.0, section "foo", align 4
608 The following example just declares a global variable
612 @G = external global i32
614 The following example defines a thread-local global with the
615 ``initialexec`` TLS model:
619 @G = thread_local(initialexec) global i32 0, align 4
621 .. _functionstructure:
626 LLVM function definitions consist of the "``define``" keyword, an
627 optional :ref:`linkage type <linkage>`, an optional :ref:`visibility
628 style <visibility>`, an optional :ref:`DLL storage class <dllstorageclass>`,
629 an optional :ref:`calling convention <callingconv>`,
630 an optional ``unnamed_addr`` attribute, a return type, an optional
631 :ref:`parameter attribute <paramattrs>` for the return type, a function
632 name, a (possibly empty) argument list (each with optional :ref:`parameter
633 attributes <paramattrs>`), optional :ref:`function attributes <fnattrs>`,
634 an optional section, an optional alignment,
635 an optional :ref:`comdat <langref_comdats>`,
636 an optional :ref:`garbage collector name <gc>`, an optional :ref:`prefix <prefixdata>`, an opening
637 curly brace, a list of basic blocks, and a closing curly brace.
639 LLVM function declarations consist of the "``declare``" keyword, an
640 optional :ref:`linkage type <linkage>`, an optional :ref:`visibility
641 style <visibility>`, an optional :ref:`DLL storage class <dllstorageclass>`,
642 an optional :ref:`calling convention <callingconv>`,
643 an optional ``unnamed_addr`` attribute, a return type, an optional
644 :ref:`parameter attribute <paramattrs>` for the return type, a function
645 name, a possibly empty list of arguments, an optional alignment, an optional
646 :ref:`garbage collector name <gc>` and an optional :ref:`prefix <prefixdata>`.
648 A function definition contains a list of basic blocks, forming the CFG (Control
649 Flow Graph) for the function. Each basic block may optionally start with a label
650 (giving the basic block a symbol table entry), contains a list of instructions,
651 and ends with a :ref:`terminator <terminators>` instruction (such as a branch or
652 function return). If an explicit label is not provided, a block is assigned an
653 implicit numbered label, using the next value from the same counter as used for
654 unnamed temporaries (:ref:`see above<identifiers>`). For example, if a function
655 entry block does not have an explicit label, it will be assigned label "%0",
656 then the first unnamed temporary in that block will be "%1", etc.
658 The first basic block in a function is special in two ways: it is
659 immediately executed on entrance to the function, and it is not allowed
660 to have predecessor basic blocks (i.e. there can not be any branches to
661 the entry block of a function). Because the block can have no
662 predecessors, it also cannot have any :ref:`PHI nodes <i_phi>`.
664 LLVM allows an explicit section to be specified for functions. If the
665 target supports it, it will emit functions to the section specified.
666 Additionally, the function can placed in a COMDAT.
668 An explicit alignment may be specified for a function. If not present,
669 or if the alignment is set to zero, the alignment of the function is set
670 by the target to whatever it feels convenient. If an explicit alignment
671 is specified, the function is forced to have at least that much
672 alignment. All alignments must be a power of 2.
674 If the ``unnamed_addr`` attribute is given, the address is know to not
675 be significant and two identical functions can be merged.
679 define [linkage] [visibility] [DLLStorageClass]
681 <ResultType> @<FunctionName> ([argument list])
682 [unnamed_addr] [fn Attrs] [section "name"] [comdat $<ComdatName>]
683 [align N] [gc] [prefix Constant] { ... }
685 The argument list is a comma seperated sequence of arguments where each
686 argument is of the following form
690 <type> [parameter Attrs] [name]
698 Aliases, unlike function or variables, don't create any new data. They
699 are just a new symbol and metadata for an existing position.
701 Aliases have a name and an aliasee that is either a global value or a
704 Aliases may have an optional :ref:`linkage type <linkage>`, an optional
705 :ref:`visibility style <visibility>`, an optional :ref:`DLL storage class
706 <dllstorageclass>` and an optional :ref:`tls model <tls_model>`.
710 @<Name> = [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal] [unnamed_addr] alias <AliaseeTy> @<Aliasee>
712 The linkage must be one of ``private``, ``internal``, ``linkonce``, ``weak``,
713 ``linkonce_odr``, ``weak_odr``, ``external``. Note that some system linkers
714 might not correctly handle dropping a weak symbol that is aliased.
716 Alias that are not ``unnamed_addr`` are guaranteed to have the same address as
717 the aliasee expression. ``unnamed_addr`` ones are only guaranteed to point
720 Since aliases are only a second name, some restrictions apply, of which
721 some can only be checked when producing an object file:
723 * The expression defining the aliasee must be computable at assembly
724 time. Since it is just a name, no relocations can be used.
726 * No alias in the expression can be weak as the possibility of the
727 intermediate alias being overridden cannot be represented in an
730 * No global value in the expression can be a declaration, since that
731 would require a relocation, which is not possible.
738 Comdat IR provides access to COFF and ELF object file COMDAT functionality.
740 Comdats have a name which represents the COMDAT key. All global objects that
741 specify this key will only end up in the final object file if the linker chooses
742 that key over some other key. Aliases are placed in the same COMDAT that their
743 aliasee computes to, if any.
745 Comdats have a selection kind to provide input on how the linker should
746 choose between keys in two different object files.
750 $<Name> = comdat SelectionKind
752 The selection kind must be one of the following:
755 The linker may choose any COMDAT key, the choice is arbitrary.
757 The linker may choose any COMDAT key but the sections must contain the
760 The linker will choose the section containing the largest COMDAT key.
762 The linker requires that only section with this COMDAT key exist.
764 The linker may choose any COMDAT key but the sections must contain the
767 Note that the Mach-O platform doesn't support COMDATs and ELF only supports
768 ``any`` as a selection kind.
770 Here is an example of a COMDAT group where a function will only be selected if
771 the COMDAT key's section is the largest:
775 $foo = comdat largest
776 @foo = global i32 2, comdat $foo
778 define void @bar() comdat $foo {
782 In a COFF object file, this will create a COMDAT section with selection kind
783 ``IMAGE_COMDAT_SELECT_LARGEST`` containing the contents of the ``@foo`` symbol
784 and another COMDAT section with selection kind
785 ``IMAGE_COMDAT_SELECT_ASSOCIATIVE`` which is associated with the first COMDAT
786 section and contains the contents of the ``@baz`` symbol.
788 There are some restrictions on the properties of the global object.
789 It, or an alias to it, must have the same name as the COMDAT group when
791 The contents and size of this object may be used during link-time to determine
792 which COMDAT groups get selected depending on the selection kind.
793 Because the name of the object must match the name of the COMDAT group, the
794 linkage of the global object must not be local; local symbols can get renamed
795 if a collision occurs in the symbol table.
797 The combined use of COMDATS and section attributes may yield surprising results.
804 @g1 = global i32 42, section "sec", comdat $foo
805 @g2 = global i32 42, section "sec", comdat $bar
807 From the object file perspective, this requires the creation of two sections
808 with the same name. This is necessary because both globals belong to different
809 COMDAT groups and COMDATs, at the object file level, are represented by
812 Note that certain IR constructs like global variables and functions may create
813 COMDATs in the object file in addition to any which are specified using COMDAT
814 IR. This arises, for example, when a global variable has linkonce_odr linkage.
816 .. _namedmetadatastructure:
821 Named metadata is a collection of metadata. :ref:`Metadata
822 nodes <metadata>` (but not metadata strings) are the only valid
823 operands for a named metadata.
827 ; Some unnamed metadata nodes, which are referenced by the named metadata.
828 !0 = metadata !{metadata !"zero"}
829 !1 = metadata !{metadata !"one"}
830 !2 = metadata !{metadata !"two"}
832 !name = !{!0, !1, !2}
839 The return type and each parameter of a function type may have a set of
840 *parameter attributes* associated with them. Parameter attributes are
841 used to communicate additional information about the result or
842 parameters of a function. Parameter attributes are considered to be part
843 of the function, not of the function type, so functions with different
844 parameter attributes can have the same function type.
846 Parameter attributes are simple keywords that follow the type specified.
847 If multiple parameter attributes are needed, they are space separated.
852 declare i32 @printf(i8* noalias nocapture, ...)
853 declare i32 @atoi(i8 zeroext)
854 declare signext i8 @returns_signed_char()
856 Note that any attributes for the function result (``nounwind``,
857 ``readonly``) come immediately after the argument list.
859 Currently, only the following parameter attributes are defined:
862 This indicates to the code generator that the parameter or return
863 value should be zero-extended to the extent required by the target's
864 ABI (which is usually 32-bits, but is 8-bits for a i1 on x86-64) by
865 the caller (for a parameter) or the callee (for a return value).
867 This indicates to the code generator that the parameter or return
868 value should be sign-extended to the extent required by the target's
869 ABI (which is usually 32-bits) by the caller (for a parameter) or
870 the callee (for a return value).
872 This indicates that this parameter or return value should be treated
873 in a special target-dependent fashion during while emitting code for
874 a function call or return (usually, by putting it in a register as
875 opposed to memory, though some targets use it to distinguish between
876 two different kinds of registers). Use of this attribute is
879 This indicates that the pointer parameter should really be passed by
880 value to the function. The attribute implies that a hidden copy of
881 the pointee is made between the caller and the callee, so the callee
882 is unable to modify the value in the caller. This attribute is only
883 valid on LLVM pointer arguments. It is generally used to pass
884 structs and arrays by value, but is also valid on pointers to
885 scalars. The copy is considered to belong to the caller not the
886 callee (for example, ``readonly`` functions should not write to
887 ``byval`` parameters). This is not a valid attribute for return
890 The byval attribute also supports specifying an alignment with the
891 align attribute. It indicates the alignment of the stack slot to
892 form and the known alignment of the pointer specified to the call
893 site. If the alignment is not specified, then the code generator
894 makes a target-specific assumption.
900 The ``inalloca`` argument attribute allows the caller to take the
901 address of outgoing stack arguments. An ``inalloca`` argument must
902 be a pointer to stack memory produced by an ``alloca`` instruction.
903 The alloca, or argument allocation, must also be tagged with the
904 inalloca keyword. Only the last argument may have the ``inalloca``
905 attribute, and that argument is guaranteed to be passed in memory.
907 An argument allocation may be used by a call at most once because
908 the call may deallocate it. The ``inalloca`` attribute cannot be
909 used in conjunction with other attributes that affect argument
910 storage, like ``inreg``, ``nest``, ``sret``, or ``byval``. The
911 ``inalloca`` attribute also disables LLVM's implicit lowering of
912 large aggregate return values, which means that frontend authors
913 must lower them with ``sret`` pointers.
915 When the call site is reached, the argument allocation must have
916 been the most recent stack allocation that is still live, or the
917 results are undefined. It is possible to allocate additional stack
918 space after an argument allocation and before its call site, but it
919 must be cleared off with :ref:`llvm.stackrestore
922 See :doc:`InAlloca` for more information on how to use this
926 This indicates that the pointer parameter specifies the address of a
927 structure that is the return value of the function in the source
928 program. This pointer must be guaranteed by the caller to be valid:
929 loads and stores to the structure may be assumed by the callee
930 not to trap and to be properly aligned. This may only be applied to
931 the first parameter. This is not a valid attribute for return
935 This indicates that the pointer value may be assumed by the optimizer to
936 have the specified alignment.
938 Note that this attribute has additional semantics when combined with the
944 This indicates that pointer values :ref:`based <pointeraliasing>` on
945 the argument or return value do not alias pointer values that are
946 not *based* on it, ignoring certain "irrelevant" dependencies. For a
947 call to the parent function, dependencies between memory references
948 from before or after the call and from those during the call are
949 "irrelevant" to the ``noalias`` keyword for the arguments and return
950 value used in that call. The caller shares the responsibility with
951 the callee for ensuring that these requirements are met. For further
952 details, please see the discussion of the NoAlias response in :ref:`alias
953 analysis <Must, May, or No>`.
955 Note that this definition of ``noalias`` is intentionally similar
956 to the definition of ``restrict`` in C99 for function arguments,
957 though it is slightly weaker.
959 For function return values, C99's ``restrict`` is not meaningful,
960 while LLVM's ``noalias`` is.
962 This indicates that the callee does not make any copies of the
963 pointer that outlive the callee itself. This is not a valid
964 attribute for return values.
969 This indicates that the pointer parameter can be excised using the
970 :ref:`trampoline intrinsics <int_trampoline>`. This is not a valid
971 attribute for return values and can only be applied to one parameter.
974 This indicates that the function always returns the argument as its return
975 value. This is an optimization hint to the code generator when generating
976 the caller, allowing tail call optimization and omission of register saves
977 and restores in some cases; it is not checked or enforced when generating
978 the callee. The parameter and the function return type must be valid
979 operands for the :ref:`bitcast instruction <i_bitcast>`. This is not a
980 valid attribute for return values and can only be applied to one parameter.
983 This indicates that the parameter or return pointer is not null. This
984 attribute may only be applied to pointer typed parameters. This is not
985 checked or enforced by LLVM, the caller must ensure that the pointer
986 passed in is non-null, or the callee must ensure that the returned pointer
989 ``dereferenceable(<n>)``
990 This indicates that the parameter or return pointer is dereferenceable. This
991 attribute may only be applied to pointer typed parameters. A pointer that
992 is dereferenceable can be loaded from speculatively without a risk of
993 trapping. The number of bytes known to be dereferenceable must be provided
994 in parentheses. It is legal for the number of bytes to be less than the
995 size of the pointee type. The ``nonnull`` attribute does not imply
996 dereferenceability (consider a pointer to one element past the end of an
997 array), however ``dereferenceable(<n>)`` does imply ``nonnull`` in
998 ``addrspace(0)`` (which is the default address space).
1002 Garbage Collector Names
1003 -----------------------
1005 Each function may specify a garbage collector name, which is simply a
1008 .. code-block:: llvm
1010 define void @f() gc "name" { ... }
1012 The compiler declares the supported values of *name*. Specifying a
1013 collector will cause the compiler to alter its output in order to
1014 support the named garbage collection algorithm.
1021 Prefix data is data associated with a function which the code generator
1022 will emit immediately before the function body. The purpose of this feature
1023 is to allow frontends to associate language-specific runtime metadata with
1024 specific functions and make it available through the function pointer while
1025 still allowing the function pointer to be called. To access the data for a
1026 given function, a program may bitcast the function pointer to a pointer to
1027 the constant's type. This implies that the IR symbol points to the start
1030 To maintain the semantics of ordinary function calls, the prefix data must
1031 have a particular format. Specifically, it must begin with a sequence of
1032 bytes which decode to a sequence of machine instructions, valid for the
1033 module's target, which transfer control to the point immediately succeeding
1034 the prefix data, without performing any other visible action. This allows
1035 the inliner and other passes to reason about the semantics of the function
1036 definition without needing to reason about the prefix data. Obviously this
1037 makes the format of the prefix data highly target dependent.
1039 Prefix data is laid out as if it were an initializer for a global variable
1040 of the prefix data's type. No padding is automatically placed between the
1041 prefix data and the function body. If padding is required, it must be part
1044 A trivial example of valid prefix data for the x86 architecture is ``i8 144``,
1045 which encodes the ``nop`` instruction:
1047 .. code-block:: llvm
1049 define void @f() prefix i8 144 { ... }
1051 Generally prefix data can be formed by encoding a relative branch instruction
1052 which skips the metadata, as in this example of valid prefix data for the
1053 x86_64 architecture, where the first two bytes encode ``jmp .+10``:
1055 .. code-block:: llvm
1057 %0 = type <{ i8, i8, i8* }>
1059 define void @f() prefix %0 <{ i8 235, i8 8, i8* @md}> { ... }
1061 A function may have prefix data but no body. This has similar semantics
1062 to the ``available_externally`` linkage in that the data may be used by the
1063 optimizers but will not be emitted in the object file.
1070 Attribute groups are groups of attributes that are referenced by objects within
1071 the IR. They are important for keeping ``.ll`` files readable, because a lot of
1072 functions will use the same set of attributes. In the degenerative case of a
1073 ``.ll`` file that corresponds to a single ``.c`` file, the single attribute
1074 group will capture the important command line flags used to build that file.
1076 An attribute group is a module-level object. To use an attribute group, an
1077 object references the attribute group's ID (e.g. ``#37``). An object may refer
1078 to more than one attribute group. In that situation, the attributes from the
1079 different groups are merged.
1081 Here is an example of attribute groups for a function that should always be
1082 inlined, has a stack alignment of 4, and which shouldn't use SSE instructions:
1084 .. code-block:: llvm
1086 ; Target-independent attributes:
1087 attributes #0 = { alwaysinline alignstack=4 }
1089 ; Target-dependent attributes:
1090 attributes #1 = { "no-sse" }
1092 ; Function @f has attributes: alwaysinline, alignstack=4, and "no-sse".
1093 define void @f() #0 #1 { ... }
1100 Function attributes are set to communicate additional information about
1101 a function. Function attributes are considered to be part of the
1102 function, not of the function type, so functions with different function
1103 attributes can have the same function type.
1105 Function attributes are simple keywords that follow the type specified.
1106 If multiple attributes are needed, they are space separated. For
1109 .. code-block:: llvm
1111 define void @f() noinline { ... }
1112 define void @f() alwaysinline { ... }
1113 define void @f() alwaysinline optsize { ... }
1114 define void @f() optsize { ... }
1117 This attribute indicates that, when emitting the prologue and
1118 epilogue, the backend should forcibly align the stack pointer.
1119 Specify the desired alignment, which must be a power of two, in
1122 This attribute indicates that the inliner should attempt to inline
1123 this function into callers whenever possible, ignoring any active
1124 inlining size threshold for this caller.
1126 This indicates that the callee function at a call site should be
1127 recognized as a built-in function, even though the function's declaration
1128 uses the ``nobuiltin`` attribute. This is only valid at call sites for
1129 direct calls to functions that are declared with the ``nobuiltin``
1132 This attribute indicates that this function is rarely called. When
1133 computing edge weights, basic blocks post-dominated by a cold
1134 function call are also considered to be cold; and, thus, given low
1137 This attribute indicates that the source code contained a hint that
1138 inlining this function is desirable (such as the "inline" keyword in
1139 C/C++). It is just a hint; it imposes no requirements on the
1142 This attribute indicates that the function should be added to a
1143 jump-instruction table at code-generation time, and that all address-taken
1144 references to this function should be replaced with a reference to the
1145 appropriate jump-instruction-table function pointer. Note that this creates
1146 a new pointer for the original function, which means that code that depends
1147 on function-pointer identity can break. So, any function annotated with
1148 ``jumptable`` must also be ``unnamed_addr``.
1150 This attribute suggests that optimization passes and code generator
1151 passes make choices that keep the code size of this function as small
1152 as possible and perform optimizations that may sacrifice runtime
1153 performance in order to minimize the size of the generated code.
1155 This attribute disables prologue / epilogue emission for the
1156 function. This can have very system-specific consequences.
1158 This indicates that the callee function at a call site is not recognized as
1159 a built-in function. LLVM will retain the original call and not replace it
1160 with equivalent code based on the semantics of the built-in function, unless
1161 the call site uses the ``builtin`` attribute. This is valid at call sites
1162 and on function declarations and definitions.
1164 This attribute indicates that calls to the function cannot be
1165 duplicated. A call to a ``noduplicate`` function may be moved
1166 within its parent function, but may not be duplicated within
1167 its parent function.
1169 A function containing a ``noduplicate`` call may still
1170 be an inlining candidate, provided that the call is not
1171 duplicated by inlining. That implies that the function has
1172 internal linkage and only has one call site, so the original
1173 call is dead after inlining.
1175 This attributes disables implicit floating point instructions.
1177 This attribute indicates that the inliner should never inline this
1178 function in any situation. This attribute may not be used together
1179 with the ``alwaysinline`` attribute.
1181 This attribute suppresses lazy symbol binding for the function. This
1182 may make calls to the function faster, at the cost of extra program
1183 startup time if the function is not called during program startup.
1185 This attribute indicates that the code generator should not use a
1186 red zone, even if the target-specific ABI normally permits it.
1188 This function attribute indicates that the function never returns
1189 normally. This produces undefined behavior at runtime if the
1190 function ever does dynamically return.
1192 This function attribute indicates that the function never returns
1193 with an unwind or exceptional control flow. If the function does
1194 unwind, its runtime behavior is undefined.
1196 This function attribute indicates that the function is not optimized
1197 by any optimization or code generator passes with the
1198 exception of interprocedural optimization passes.
1199 This attribute cannot be used together with the ``alwaysinline``
1200 attribute; this attribute is also incompatible
1201 with the ``minsize`` attribute and the ``optsize`` attribute.
1203 This attribute requires the ``noinline`` attribute to be specified on
1204 the function as well, so the function is never inlined into any caller.
1205 Only functions with the ``alwaysinline`` attribute are valid
1206 candidates for inlining into the body of this function.
1208 This attribute suggests that optimization passes and code generator
1209 passes make choices that keep the code size of this function low,
1210 and otherwise do optimizations specifically to reduce code size as
1211 long as they do not significantly impact runtime performance.
1213 On a function, this attribute indicates that the function computes its
1214 result (or decides to unwind an exception) based strictly on its arguments,
1215 without dereferencing any pointer arguments or otherwise accessing
1216 any mutable state (e.g. memory, control registers, etc) visible to
1217 caller functions. It does not write through any pointer arguments
1218 (including ``byval`` arguments) and never changes any state visible
1219 to callers. This means that it cannot unwind exceptions by calling
1220 the ``C++`` exception throwing methods.
1222 On an argument, this attribute indicates that the function does not
1223 dereference that pointer argument, even though it may read or write the
1224 memory that the pointer points to if accessed through other pointers.
1226 On a function, this attribute indicates that the function does not write
1227 through any pointer arguments (including ``byval`` arguments) or otherwise
1228 modify any state (e.g. memory, control registers, etc) visible to
1229 caller functions. It may dereference pointer arguments and read
1230 state that may be set in the caller. A readonly function always
1231 returns the same value (or unwinds an exception identically) when
1232 called with the same set of arguments and global state. It cannot
1233 unwind an exception by calling the ``C++`` exception throwing
1236 On an argument, this attribute indicates that the function does not write
1237 through this pointer argument, even though it may write to the memory that
1238 the pointer points to.
1240 This attribute indicates that this function can return twice. The C
1241 ``setjmp`` is an example of such a function. The compiler disables
1242 some optimizations (like tail calls) in the caller of these
1244 ``sanitize_address``
1245 This attribute indicates that AddressSanitizer checks
1246 (dynamic address safety analysis) are enabled for this function.
1248 This attribute indicates that MemorySanitizer checks (dynamic detection
1249 of accesses to uninitialized memory) are enabled for this function.
1251 This attribute indicates that ThreadSanitizer checks
1252 (dynamic thread safety analysis) are enabled for this function.
1254 This attribute indicates that the function should emit a stack
1255 smashing protector. It is in the form of a "canary" --- a random value
1256 placed on the stack before the local variables that's checked upon
1257 return from the function to see if it has been overwritten. A
1258 heuristic is used to determine if a function needs stack protectors
1259 or not. The heuristic used will enable protectors for functions with:
1261 - Character arrays larger than ``ssp-buffer-size`` (default 8).
1262 - Aggregates containing character arrays larger than ``ssp-buffer-size``.
1263 - Calls to alloca() with variable sizes or constant sizes greater than
1264 ``ssp-buffer-size``.
1266 Variables that are identified as requiring a protector will be arranged
1267 on the stack such that they are adjacent to the stack protector guard.
1269 If a function that has an ``ssp`` attribute is inlined into a
1270 function that doesn't have an ``ssp`` attribute, then the resulting
1271 function will have an ``ssp`` attribute.
1273 This attribute indicates that the function should *always* emit a
1274 stack smashing protector. This overrides the ``ssp`` function
1277 Variables that are identified as requiring a protector will be arranged
1278 on the stack such that they are adjacent to the stack protector guard.
1279 The specific layout rules are:
1281 #. Large arrays and structures containing large arrays
1282 (``>= ssp-buffer-size``) are closest to the stack protector.
1283 #. Small arrays and structures containing small arrays
1284 (``< ssp-buffer-size``) are 2nd closest to the protector.
1285 #. Variables that have had their address taken are 3rd closest to the
1288 If a function that has an ``sspreq`` attribute is inlined into a
1289 function that doesn't have an ``sspreq`` attribute or which has an
1290 ``ssp`` or ``sspstrong`` attribute, then the resulting function will have
1291 an ``sspreq`` attribute.
1293 This attribute indicates that the function should emit a stack smashing
1294 protector. This attribute causes a strong heuristic to be used when
1295 determining if a function needs stack protectors. The strong heuristic
1296 will enable protectors for functions with:
1298 - Arrays of any size and type
1299 - Aggregates containing an array of any size and type.
1300 - Calls to alloca().
1301 - Local variables that have had their address taken.
1303 Variables that are identified as requiring a protector will be arranged
1304 on the stack such that they are adjacent to the stack protector guard.
1305 The specific layout rules are:
1307 #. Large arrays and structures containing large arrays
1308 (``>= ssp-buffer-size``) are closest to the stack protector.
1309 #. Small arrays and structures containing small arrays
1310 (``< ssp-buffer-size``) are 2nd closest to the protector.
1311 #. Variables that have had their address taken are 3rd closest to the
1314 This overrides the ``ssp`` function attribute.
1316 If a function that has an ``sspstrong`` attribute is inlined into a
1317 function that doesn't have an ``sspstrong`` attribute, then the
1318 resulting function will have an ``sspstrong`` attribute.
1320 This attribute indicates that the ABI being targeted requires that
1321 an unwind table entry be produce for this function even if we can
1322 show that no exceptions passes by it. This is normally the case for
1323 the ELF x86-64 abi, but it can be disabled for some compilation
1328 Module-Level Inline Assembly
1329 ----------------------------
1331 Modules may contain "module-level inline asm" blocks, which corresponds
1332 to the GCC "file scope inline asm" blocks. These blocks are internally
1333 concatenated by LLVM and treated as a single unit, but may be separated
1334 in the ``.ll`` file if desired. The syntax is very simple:
1336 .. code-block:: llvm
1338 module asm "inline asm code goes here"
1339 module asm "more can go here"
1341 The strings can contain any character by escaping non-printable
1342 characters. The escape sequence used is simply "\\xx" where "xx" is the
1343 two digit hex code for the number.
1345 The inline asm code is simply printed to the machine code .s file when
1346 assembly code is generated.
1348 .. _langref_datalayout:
1353 A module may specify a target specific data layout string that specifies
1354 how data is to be laid out in memory. The syntax for the data layout is
1357 .. code-block:: llvm
1359 target datalayout = "layout specification"
1361 The *layout specification* consists of a list of specifications
1362 separated by the minus sign character ('-'). Each specification starts
1363 with a letter and may include other information after the letter to
1364 define some aspect of the data layout. The specifications accepted are
1368 Specifies that the target lays out data in big-endian form. That is,
1369 the bits with the most significance have the lowest address
1372 Specifies that the target lays out data in little-endian form. That
1373 is, the bits with the least significance have the lowest address
1376 Specifies the natural alignment of the stack in bits. Alignment
1377 promotion of stack variables is limited to the natural stack
1378 alignment to avoid dynamic stack realignment. The stack alignment
1379 must be a multiple of 8-bits. If omitted, the natural stack
1380 alignment defaults to "unspecified", which does not prevent any
1381 alignment promotions.
1382 ``p[n]:<size>:<abi>:<pref>``
1383 This specifies the *size* of a pointer and its ``<abi>`` and
1384 ``<pref>``\erred alignments for address space ``n``. All sizes are in
1385 bits. The address space, ``n`` is optional, and if not specified,
1386 denotes the default address space 0. The value of ``n`` must be
1387 in the range [1,2^23).
1388 ``i<size>:<abi>:<pref>``
1389 This specifies the alignment for an integer type of a given bit
1390 ``<size>``. The value of ``<size>`` must be in the range [1,2^23).
1391 ``v<size>:<abi>:<pref>``
1392 This specifies the alignment for a vector type of a given bit
1394 ``f<size>:<abi>:<pref>``
1395 This specifies the alignment for a floating point type of a given bit
1396 ``<size>``. Only values of ``<size>`` that are supported by the target
1397 will work. 32 (float) and 64 (double) are supported on all targets; 80
1398 or 128 (different flavors of long double) are also supported on some
1401 This specifies the alignment for an object of aggregate type.
1403 If present, specifies that llvm names are mangled in the output. The
1406 * ``e``: ELF mangling: Private symbols get a ``.L`` prefix.
1407 * ``m``: Mips mangling: Private symbols get a ``$`` prefix.
1408 * ``o``: Mach-O mangling: Private symbols get ``L`` prefix. Other
1409 symbols get a ``_`` prefix.
1410 * ``w``: Windows COFF prefix: Similar to Mach-O, but stdcall and fastcall
1411 functions also get a suffix based on the frame size.
1412 ``n<size1>:<size2>:<size3>...``
1413 This specifies a set of native integer widths for the target CPU in
1414 bits. For example, it might contain ``n32`` for 32-bit PowerPC,
1415 ``n32:64`` for PowerPC 64, or ``n8:16:32:64`` for X86-64. Elements of
1416 this set are considered to support most general arithmetic operations
1419 On every specification that takes a ``<abi>:<pref>``, specifying the
1420 ``<pref>`` alignment is optional. If omitted, the preceding ``:``
1421 should be omitted too and ``<pref>`` will be equal to ``<abi>``.
1423 When constructing the data layout for a given target, LLVM starts with a
1424 default set of specifications which are then (possibly) overridden by
1425 the specifications in the ``datalayout`` keyword. The default
1426 specifications are given in this list:
1428 - ``E`` - big endian
1429 - ``p:64:64:64`` - 64-bit pointers with 64-bit alignment.
1430 - ``p[n]:64:64:64`` - Other address spaces are assumed to be the
1431 same as the default address space.
1432 - ``S0`` - natural stack alignment is unspecified
1433 - ``i1:8:8`` - i1 is 8-bit (byte) aligned
1434 - ``i8:8:8`` - i8 is 8-bit (byte) aligned
1435 - ``i16:16:16`` - i16 is 16-bit aligned
1436 - ``i32:32:32`` - i32 is 32-bit aligned
1437 - ``i64:32:64`` - i64 has ABI alignment of 32-bits but preferred
1438 alignment of 64-bits
1439 - ``f16:16:16`` - half is 16-bit aligned
1440 - ``f32:32:32`` - float is 32-bit aligned
1441 - ``f64:64:64`` - double is 64-bit aligned
1442 - ``f128:128:128`` - quad is 128-bit aligned
1443 - ``v64:64:64`` - 64-bit vector is 64-bit aligned
1444 - ``v128:128:128`` - 128-bit vector is 128-bit aligned
1445 - ``a:0:64`` - aggregates are 64-bit aligned
1447 When LLVM is determining the alignment for a given type, it uses the
1450 #. If the type sought is an exact match for one of the specifications,
1451 that specification is used.
1452 #. If no match is found, and the type sought is an integer type, then
1453 the smallest integer type that is larger than the bitwidth of the
1454 sought type is used. If none of the specifications are larger than
1455 the bitwidth then the largest integer type is used. For example,
1456 given the default specifications above, the i7 type will use the
1457 alignment of i8 (next largest) while both i65 and i256 will use the
1458 alignment of i64 (largest specified).
1459 #. If no match is found, and the type sought is a vector type, then the
1460 largest vector type that is smaller than the sought vector type will
1461 be used as a fall back. This happens because <128 x double> can be
1462 implemented in terms of 64 <2 x double>, for example.
1464 The function of the data layout string may not be what you expect.
1465 Notably, this is not a specification from the frontend of what alignment
1466 the code generator should use.
1468 Instead, if specified, the target data layout is required to match what
1469 the ultimate *code generator* expects. This string is used by the
1470 mid-level optimizers to improve code, and this only works if it matches
1471 what the ultimate code generator uses. If you would like to generate IR
1472 that does not embed this target-specific detail into the IR, then you
1473 don't have to specify the string. This will disable some optimizations
1474 that require precise layout information, but this also prevents those
1475 optimizations from introducing target specificity into the IR.
1482 A module may specify a target triple string that describes the target
1483 host. The syntax for the target triple is simply:
1485 .. code-block:: llvm
1487 target triple = "x86_64-apple-macosx10.7.0"
1489 The *target triple* string consists of a series of identifiers delimited
1490 by the minus sign character ('-'). The canonical forms are:
1494 ARCHITECTURE-VENDOR-OPERATING_SYSTEM
1495 ARCHITECTURE-VENDOR-OPERATING_SYSTEM-ENVIRONMENT
1497 This information is passed along to the backend so that it generates
1498 code for the proper architecture. It's possible to override this on the
1499 command line with the ``-mtriple`` command line option.
1501 .. _pointeraliasing:
1503 Pointer Aliasing Rules
1504 ----------------------
1506 Any memory access must be done through a pointer value associated with
1507 an address range of the memory access, otherwise the behavior is
1508 undefined. Pointer values are associated with address ranges according
1509 to the following rules:
1511 - A pointer value is associated with the addresses associated with any
1512 value it is *based* on.
1513 - An address of a global variable is associated with the address range
1514 of the variable's storage.
1515 - The result value of an allocation instruction is associated with the
1516 address range of the allocated storage.
1517 - A null pointer in the default address-space is associated with no
1519 - An integer constant other than zero or a pointer value returned from
1520 a function not defined within LLVM may be associated with address
1521 ranges allocated through mechanisms other than those provided by
1522 LLVM. Such ranges shall not overlap with any ranges of addresses
1523 allocated by mechanisms provided by LLVM.
1525 A pointer value is *based* on another pointer value according to the
1528 - A pointer value formed from a ``getelementptr`` operation is *based*
1529 on the first operand of the ``getelementptr``.
1530 - The result value of a ``bitcast`` is *based* on the operand of the
1532 - A pointer value formed by an ``inttoptr`` is *based* on all pointer
1533 values that contribute (directly or indirectly) to the computation of
1534 the pointer's value.
1535 - The "*based* on" relationship is transitive.
1537 Note that this definition of *"based"* is intentionally similar to the
1538 definition of *"based"* in C99, though it is slightly weaker.
1540 LLVM IR does not associate types with memory. The result type of a
1541 ``load`` merely indicates the size and alignment of the memory from
1542 which to load, as well as the interpretation of the value. The first
1543 operand type of a ``store`` similarly only indicates the size and
1544 alignment of the store.
1546 Consequently, type-based alias analysis, aka TBAA, aka
1547 ``-fstrict-aliasing``, is not applicable to general unadorned LLVM IR.
1548 :ref:`Metadata <metadata>` may be used to encode additional information
1549 which specialized optimization passes may use to implement type-based
1554 Volatile Memory Accesses
1555 ------------------------
1557 Certain memory accesses, such as :ref:`load <i_load>`'s,
1558 :ref:`store <i_store>`'s, and :ref:`llvm.memcpy <int_memcpy>`'s may be
1559 marked ``volatile``. The optimizers must not change the number of
1560 volatile operations or change their order of execution relative to other
1561 volatile operations. The optimizers *may* change the order of volatile
1562 operations relative to non-volatile operations. This is not Java's
1563 "volatile" and has no cross-thread synchronization behavior.
1565 IR-level volatile loads and stores cannot safely be optimized into
1566 llvm.memcpy or llvm.memmove intrinsics even when those intrinsics are
1567 flagged volatile. Likewise, the backend should never split or merge
1568 target-legal volatile load/store instructions.
1570 .. admonition:: Rationale
1572 Platforms may rely on volatile loads and stores of natively supported
1573 data width to be executed as single instruction. For example, in C
1574 this holds for an l-value of volatile primitive type with native
1575 hardware support, but not necessarily for aggregate types. The
1576 frontend upholds these expectations, which are intentionally
1577 unspecified in the IR. The rules above ensure that IR transformation
1578 do not violate the frontend's contract with the language.
1582 Memory Model for Concurrent Operations
1583 --------------------------------------
1585 The LLVM IR does not define any way to start parallel threads of
1586 execution or to register signal handlers. Nonetheless, there are
1587 platform-specific ways to create them, and we define LLVM IR's behavior
1588 in their presence. This model is inspired by the C++0x memory model.
1590 For a more informal introduction to this model, see the :doc:`Atomics`.
1592 We define a *happens-before* partial order as the least partial order
1595 - Is a superset of single-thread program order, and
1596 - When a *synchronizes-with* ``b``, includes an edge from ``a`` to
1597 ``b``. *Synchronizes-with* pairs are introduced by platform-specific
1598 techniques, like pthread locks, thread creation, thread joining,
1599 etc., and by atomic instructions. (See also :ref:`Atomic Memory Ordering
1600 Constraints <ordering>`).
1602 Note that program order does not introduce *happens-before* edges
1603 between a thread and signals executing inside that thread.
1605 Every (defined) read operation (load instructions, memcpy, atomic
1606 loads/read-modify-writes, etc.) R reads a series of bytes written by
1607 (defined) write operations (store instructions, atomic
1608 stores/read-modify-writes, memcpy, etc.). For the purposes of this
1609 section, initialized globals are considered to have a write of the
1610 initializer which is atomic and happens before any other read or write
1611 of the memory in question. For each byte of a read R, R\ :sub:`byte`
1612 may see any write to the same byte, except:
1614 - If write\ :sub:`1` happens before write\ :sub:`2`, and
1615 write\ :sub:`2` happens before R\ :sub:`byte`, then
1616 R\ :sub:`byte` does not see write\ :sub:`1`.
1617 - If R\ :sub:`byte` happens before write\ :sub:`3`, then
1618 R\ :sub:`byte` does not see write\ :sub:`3`.
1620 Given that definition, R\ :sub:`byte` is defined as follows:
1622 - If R is volatile, the result is target-dependent. (Volatile is
1623 supposed to give guarantees which can support ``sig_atomic_t`` in
1624 C/C++, and may be used for accesses to addresses that do not behave
1625 like normal memory. It does not generally provide cross-thread
1627 - Otherwise, if there is no write to the same byte that happens before
1628 R\ :sub:`byte`, R\ :sub:`byte` returns ``undef`` for that byte.
1629 - Otherwise, if R\ :sub:`byte` may see exactly one write,
1630 R\ :sub:`byte` returns the value written by that write.
1631 - Otherwise, if R is atomic, and all the writes R\ :sub:`byte` may
1632 see are atomic, it chooses one of the values written. See the :ref:`Atomic
1633 Memory Ordering Constraints <ordering>` section for additional
1634 constraints on how the choice is made.
1635 - Otherwise R\ :sub:`byte` returns ``undef``.
1637 R returns the value composed of the series of bytes it read. This
1638 implies that some bytes within the value may be ``undef`` **without**
1639 the entire value being ``undef``. Note that this only defines the
1640 semantics of the operation; it doesn't mean that targets will emit more
1641 than one instruction to read the series of bytes.
1643 Note that in cases where none of the atomic intrinsics are used, this
1644 model places only one restriction on IR transformations on top of what
1645 is required for single-threaded execution: introducing a store to a byte
1646 which might not otherwise be stored is not allowed in general.
1647 (Specifically, in the case where another thread might write to and read
1648 from an address, introducing a store can change a load that may see
1649 exactly one write into a load that may see multiple writes.)
1653 Atomic Memory Ordering Constraints
1654 ----------------------------------
1656 Atomic instructions (:ref:`cmpxchg <i_cmpxchg>`,
1657 :ref:`atomicrmw <i_atomicrmw>`, :ref:`fence <i_fence>`,
1658 :ref:`atomic load <i_load>`, and :ref:`atomic store <i_store>`) take
1659 ordering parameters that determine which other atomic instructions on
1660 the same address they *synchronize with*. These semantics are borrowed
1661 from Java and C++0x, but are somewhat more colloquial. If these
1662 descriptions aren't precise enough, check those specs (see spec
1663 references in the :doc:`atomics guide <Atomics>`).
1664 :ref:`fence <i_fence>` instructions treat these orderings somewhat
1665 differently since they don't take an address. See that instruction's
1666 documentation for details.
1668 For a simpler introduction to the ordering constraints, see the
1672 The set of values that can be read is governed by the happens-before
1673 partial order. A value cannot be read unless some operation wrote
1674 it. This is intended to provide a guarantee strong enough to model
1675 Java's non-volatile shared variables. This ordering cannot be
1676 specified for read-modify-write operations; it is not strong enough
1677 to make them atomic in any interesting way.
1679 In addition to the guarantees of ``unordered``, there is a single
1680 total order for modifications by ``monotonic`` operations on each
1681 address. All modification orders must be compatible with the
1682 happens-before order. There is no guarantee that the modification
1683 orders can be combined to a global total order for the whole program
1684 (and this often will not be possible). The read in an atomic
1685 read-modify-write operation (:ref:`cmpxchg <i_cmpxchg>` and
1686 :ref:`atomicrmw <i_atomicrmw>`) reads the value in the modification
1687 order immediately before the value it writes. If one atomic read
1688 happens before another atomic read of the same address, the later
1689 read must see the same value or a later value in the address's
1690 modification order. This disallows reordering of ``monotonic`` (or
1691 stronger) operations on the same address. If an address is written
1692 ``monotonic``-ally by one thread, and other threads ``monotonic``-ally
1693 read that address repeatedly, the other threads must eventually see
1694 the write. This corresponds to the C++0x/C1x
1695 ``memory_order_relaxed``.
1697 In addition to the guarantees of ``monotonic``, a
1698 *synchronizes-with* edge may be formed with a ``release`` operation.
1699 This is intended to model C++'s ``memory_order_acquire``.
1701 In addition to the guarantees of ``monotonic``, if this operation
1702 writes a value which is subsequently read by an ``acquire``
1703 operation, it *synchronizes-with* that operation. (This isn't a
1704 complete description; see the C++0x definition of a release
1705 sequence.) This corresponds to the C++0x/C1x
1706 ``memory_order_release``.
1707 ``acq_rel`` (acquire+release)
1708 Acts as both an ``acquire`` and ``release`` operation on its
1709 address. This corresponds to the C++0x/C1x ``memory_order_acq_rel``.
1710 ``seq_cst`` (sequentially consistent)
1711 In addition to the guarantees of ``acq_rel`` (``acquire`` for an
1712 operation that only reads, ``release`` for an operation that only
1713 writes), there is a global total order on all
1714 sequentially-consistent operations on all addresses, which is
1715 consistent with the *happens-before* partial order and with the
1716 modification orders of all the affected addresses. Each
1717 sequentially-consistent read sees the last preceding write to the
1718 same address in this global order. This corresponds to the C++0x/C1x
1719 ``memory_order_seq_cst`` and Java volatile.
1723 If an atomic operation is marked ``singlethread``, it only *synchronizes
1724 with* or participates in modification and seq\_cst total orderings with
1725 other operations running in the same thread (for example, in signal
1733 LLVM IR floating-point binary ops (:ref:`fadd <i_fadd>`,
1734 :ref:`fsub <i_fsub>`, :ref:`fmul <i_fmul>`, :ref:`fdiv <i_fdiv>`,
1735 :ref:`frem <i_frem>`) have the following flags that can set to enable
1736 otherwise unsafe floating point operations
1739 No NaNs - Allow optimizations to assume the arguments and result are not
1740 NaN. Such optimizations are required to retain defined behavior over
1741 NaNs, but the value of the result is undefined.
1744 No Infs - Allow optimizations to assume the arguments and result are not
1745 +/-Inf. Such optimizations are required to retain defined behavior over
1746 +/-Inf, but the value of the result is undefined.
1749 No Signed Zeros - Allow optimizations to treat the sign of a zero
1750 argument or result as insignificant.
1753 Allow Reciprocal - Allow optimizations to use the reciprocal of an
1754 argument rather than perform division.
1757 Fast - Allow algebraically equivalent transformations that may
1758 dramatically change results in floating point (e.g. reassociate). This
1759 flag implies all the others.
1763 Use-list Order Directives
1764 -------------------------
1766 Use-list directives encode the in-memory order of each use-list, allowing the
1767 order to be recreated. ``<order-indexes>`` is a comma-separated list of
1768 indexes that are assigned to the referenced value's uses. The referenced
1769 value's use-list is immediately sorted by these indexes.
1771 Use-list directives may appear at function scope or global scope. They are not
1772 instructions, and have no effect on the semantics of the IR. When they're at
1773 function scope, they must appear after the terminator of the final basic block.
1775 If basic blocks have their address taken via ``blockaddress()`` expressions,
1776 ``uselistorder_bb`` can be used to reorder their use-lists from outside their
1783 uselistorder <ty> <value>, { <order-indexes> }
1784 uselistorder_bb @function, %block { <order-indexes> }
1790 define void @foo(i32 %arg1, i32 %arg2) {
1792 ; ... instructions ...
1794 ; ... instructions ...
1796 ; At function scope.
1797 uselistorder i32 %arg1, { 1, 0, 2 }
1798 uselistorder label %bb, { 1, 0 }
1802 uselistorder i32* @global, { 1, 2, 0 }
1803 uselistorder i32 7, { 1, 0 }
1804 uselistorder i32 (i32) @bar, { 1, 0 }
1805 uselistorder_bb @foo, %bb, { 5, 1, 3, 2, 0, 4 }
1812 The LLVM type system is one of the most important features of the
1813 intermediate representation. Being typed enables a number of
1814 optimizations to be performed on the intermediate representation
1815 directly, without having to do extra analyses on the side before the
1816 transformation. A strong type system makes it easier to read the
1817 generated code and enables novel analyses and transformations that are
1818 not feasible to perform on normal three address code representations.
1828 The void type does not represent any value and has no size.
1846 The function type can be thought of as a function signature. It consists of a
1847 return type and a list of formal parameter types. The return type of a function
1848 type is a void type or first class type --- except for :ref:`label <t_label>`
1849 and :ref:`metadata <t_metadata>` types.
1855 <returntype> (<parameter list>)
1857 ...where '``<parameter list>``' is a comma-separated list of type
1858 specifiers. Optionally, the parameter list may include a type ``...``, which
1859 indicates that the function takes a variable number of arguments. Variable
1860 argument functions can access their arguments with the :ref:`variable argument
1861 handling intrinsic <int_varargs>` functions. '``<returntype>``' is any type
1862 except :ref:`label <t_label>` and :ref:`metadata <t_metadata>`.
1866 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1867 | ``i32 (i32)`` | function taking an ``i32``, returning an ``i32`` |
1868 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1869 | ``float (i16, i32 *) *`` | :ref:`Pointer <t_pointer>` to a function that takes an ``i16`` and a :ref:`pointer <t_pointer>` to ``i32``, returning ``float``. |
1870 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1871 | ``i32 (i8*, ...)`` | A vararg function that takes at least one :ref:`pointer <t_pointer>` to ``i8`` (char in C), which returns an integer. This is the signature for ``printf`` in LLVM. |
1872 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1873 | ``{i32, i32} (i32)`` | A function taking an ``i32``, returning a :ref:`structure <t_struct>` containing two ``i32`` values |
1874 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1881 The :ref:`first class <t_firstclass>` types are perhaps the most important.
1882 Values of these types are the only ones which can be produced by
1890 These are the types that are valid in registers from CodeGen's perspective.
1899 The integer type is a very simple type that simply specifies an
1900 arbitrary bit width for the integer type desired. Any bit width from 1
1901 bit to 2\ :sup:`23`\ -1 (about 8 million) can be specified.
1909 The number of bits the integer will occupy is specified by the ``N``
1915 +----------------+------------------------------------------------+
1916 | ``i1`` | a single-bit integer. |
1917 +----------------+------------------------------------------------+
1918 | ``i32`` | a 32-bit integer. |
1919 +----------------+------------------------------------------------+
1920 | ``i1942652`` | a really big integer of over 1 million bits. |
1921 +----------------+------------------------------------------------+
1925 Floating Point Types
1926 """"""""""""""""""""
1935 - 16-bit floating point value
1938 - 32-bit floating point value
1941 - 64-bit floating point value
1944 - 128-bit floating point value (112-bit mantissa)
1947 - 80-bit floating point value (X87)
1950 - 128-bit floating point value (two 64-bits)
1957 The x86_mmx type represents a value held in an MMX register on an x86
1958 machine. The operations allowed on it are quite limited: parameters and
1959 return values, load and store, and bitcast. User-specified MMX
1960 instructions are represented as intrinsic or asm calls with arguments
1961 and/or results of this type. There are no arrays, vectors or constants
1978 The pointer type is used to specify memory locations. Pointers are
1979 commonly used to reference objects in memory.
1981 Pointer types may have an optional address space attribute defining the
1982 numbered address space where the pointed-to object resides. The default
1983 address space is number zero. The semantics of non-zero address spaces
1984 are target-specific.
1986 Note that LLVM does not permit pointers to void (``void*``) nor does it
1987 permit pointers to labels (``label*``). Use ``i8*`` instead.
1997 +-------------------------+--------------------------------------------------------------------------------------------------------------+
1998 | ``[4 x i32]*`` | A :ref:`pointer <t_pointer>` to :ref:`array <t_array>` of four ``i32`` values. |
1999 +-------------------------+--------------------------------------------------------------------------------------------------------------+
2000 | ``i32 (i32*) *`` | A :ref:`pointer <t_pointer>` to a :ref:`function <t_function>` that takes an ``i32*``, returning an ``i32``. |
2001 +-------------------------+--------------------------------------------------------------------------------------------------------------+
2002 | ``i32 addrspace(5)*`` | A :ref:`pointer <t_pointer>` to an ``i32`` value that resides in address space #5. |
2003 +-------------------------+--------------------------------------------------------------------------------------------------------------+
2012 A vector type is a simple derived type that represents a vector of
2013 elements. Vector types are used when multiple primitive data are
2014 operated in parallel using a single instruction (SIMD). A vector type
2015 requires a size (number of elements) and an underlying primitive data
2016 type. Vector types are considered :ref:`first class <t_firstclass>`.
2022 < <# elements> x <elementtype> >
2024 The number of elements is a constant integer value larger than 0;
2025 elementtype may be any integer, floating point or pointer type. Vectors
2026 of size zero are not allowed.
2030 +-------------------+--------------------------------------------------+
2031 | ``<4 x i32>`` | Vector of 4 32-bit integer values. |
2032 +-------------------+--------------------------------------------------+
2033 | ``<8 x float>`` | Vector of 8 32-bit floating-point values. |
2034 +-------------------+--------------------------------------------------+
2035 | ``<2 x i64>`` | Vector of 2 64-bit integer values. |
2036 +-------------------+--------------------------------------------------+
2037 | ``<4 x i64*>`` | Vector of 4 pointers to 64-bit integer values. |
2038 +-------------------+--------------------------------------------------+
2047 The label type represents code labels.
2062 The metadata type represents embedded metadata. No derived types may be
2063 created from metadata except for :ref:`function <t_function>` arguments.
2076 Aggregate Types are a subset of derived types that can contain multiple
2077 member types. :ref:`Arrays <t_array>` and :ref:`structs <t_struct>` are
2078 aggregate types. :ref:`Vectors <t_vector>` are not considered to be
2088 The array type is a very simple derived type that arranges elements
2089 sequentially in memory. The array type requires a size (number of
2090 elements) and an underlying data type.
2096 [<# elements> x <elementtype>]
2098 The number of elements is a constant integer value; ``elementtype`` may
2099 be any type with a size.
2103 +------------------+--------------------------------------+
2104 | ``[40 x i32]`` | Array of 40 32-bit integer values. |
2105 +------------------+--------------------------------------+
2106 | ``[41 x i32]`` | Array of 41 32-bit integer values. |
2107 +------------------+--------------------------------------+
2108 | ``[4 x i8]`` | Array of 4 8-bit integer values. |
2109 +------------------+--------------------------------------+
2111 Here are some examples of multidimensional arrays:
2113 +-----------------------------+----------------------------------------------------------+
2114 | ``[3 x [4 x i32]]`` | 3x4 array of 32-bit integer values. |
2115 +-----------------------------+----------------------------------------------------------+
2116 | ``[12 x [10 x float]]`` | 12x10 array of single precision floating point values. |
2117 +-----------------------------+----------------------------------------------------------+
2118 | ``[2 x [3 x [4 x i16]]]`` | 2x3x4 array of 16-bit integer values. |
2119 +-----------------------------+----------------------------------------------------------+
2121 There is no restriction on indexing beyond the end of the array implied
2122 by a static type (though there are restrictions on indexing beyond the
2123 bounds of an allocated object in some cases). This means that
2124 single-dimension 'variable sized array' addressing can be implemented in
2125 LLVM with a zero length array type. An implementation of 'pascal style
2126 arrays' in LLVM could use the type "``{ i32, [0 x float]}``", for
2136 The structure type is used to represent a collection of data members
2137 together in memory. The elements of a structure may be any type that has
2140 Structures in memory are accessed using '``load``' and '``store``' by
2141 getting a pointer to a field with the '``getelementptr``' instruction.
2142 Structures in registers are accessed using the '``extractvalue``' and
2143 '``insertvalue``' instructions.
2145 Structures may optionally be "packed" structures, which indicate that
2146 the alignment of the struct is one byte, and that there is no padding
2147 between the elements. In non-packed structs, padding between field types
2148 is inserted as defined by the DataLayout string in the module, which is
2149 required to match what the underlying code generator expects.
2151 Structures can either be "literal" or "identified". A literal structure
2152 is defined inline with other types (e.g. ``{i32, i32}*``) whereas
2153 identified types are always defined at the top level with a name.
2154 Literal types are uniqued by their contents and can never be recursive
2155 or opaque since there is no way to write one. Identified types can be
2156 recursive, can be opaqued, and are never uniqued.
2162 %T1 = type { <type list> } ; Identified normal struct type
2163 %T2 = type <{ <type list> }> ; Identified packed struct type
2167 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2168 | ``{ i32, i32, i32 }`` | A triple of three ``i32`` values |
2169 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2170 | ``{ float, i32 (i32) * }`` | A pair, where the first element is a ``float`` and the second element is a :ref:`pointer <t_pointer>` to a :ref:`function <t_function>` that takes an ``i32``, returning an ``i32``. |
2171 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2172 | ``<{ i8, i32 }>`` | A packed struct known to be 5 bytes in size. |
2173 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2177 Opaque Structure Types
2178 """"""""""""""""""""""
2182 Opaque structure types are used to represent named structure types that
2183 do not have a body specified. This corresponds (for example) to the C
2184 notion of a forward declared structure.
2195 +--------------+-------------------+
2196 | ``opaque`` | An opaque type. |
2197 +--------------+-------------------+
2204 LLVM has several different basic types of constants. This section
2205 describes them all and their syntax.
2210 **Boolean constants**
2211 The two strings '``true``' and '``false``' are both valid constants
2213 **Integer constants**
2214 Standard integers (such as '4') are constants of the
2215 :ref:`integer <t_integer>` type. Negative numbers may be used with
2217 **Floating point constants**
2218 Floating point constants use standard decimal notation (e.g.
2219 123.421), exponential notation (e.g. 1.23421e+2), or a more precise
2220 hexadecimal notation (see below). The assembler requires the exact
2221 decimal value of a floating-point constant. For example, the
2222 assembler accepts 1.25 but rejects 1.3 because 1.3 is a repeating
2223 decimal in binary. Floating point constants must have a :ref:`floating
2224 point <t_floating>` type.
2225 **Null pointer constants**
2226 The identifier '``null``' is recognized as a null pointer constant
2227 and must be of :ref:`pointer type <t_pointer>`.
2229 The one non-intuitive notation for constants is the hexadecimal form of
2230 floating point constants. For example, the form
2231 '``double 0x432ff973cafa8000``' is equivalent to (but harder to read
2232 than) '``double 4.5e+15``'. The only time hexadecimal floating point
2233 constants are required (and the only time that they are generated by the
2234 disassembler) is when a floating point constant must be emitted but it
2235 cannot be represented as a decimal floating point number in a reasonable
2236 number of digits. For example, NaN's, infinities, and other special
2237 values are represented in their IEEE hexadecimal format so that assembly
2238 and disassembly do not cause any bits to change in the constants.
2240 When using the hexadecimal form, constants of types half, float, and
2241 double are represented using the 16-digit form shown above (which
2242 matches the IEEE754 representation for double); half and float values
2243 must, however, be exactly representable as IEEE 754 half and single
2244 precision, respectively. Hexadecimal format is always used for long
2245 double, and there are three forms of long double. The 80-bit format used
2246 by x86 is represented as ``0xK`` followed by 20 hexadecimal digits. The
2247 128-bit format used by PowerPC (two adjacent doubles) is represented by
2248 ``0xM`` followed by 32 hexadecimal digits. The IEEE 128-bit format is
2249 represented by ``0xL`` followed by 32 hexadecimal digits. Long doubles
2250 will only work if they match the long double format on your target.
2251 The IEEE 16-bit format (half precision) is represented by ``0xH``
2252 followed by 4 hexadecimal digits. All hexadecimal formats are big-endian
2253 (sign bit at the left).
2255 There are no constants of type x86_mmx.
2257 .. _complexconstants:
2262 Complex constants are a (potentially recursive) combination of simple
2263 constants and smaller complex constants.
2265 **Structure constants**
2266 Structure constants are represented with notation similar to
2267 structure type definitions (a comma separated list of elements,
2268 surrounded by braces (``{}``)). For example:
2269 "``{ i32 4, float 17.0, i32* @G }``", where "``@G``" is declared as
2270 "``@G = external global i32``". Structure constants must have
2271 :ref:`structure type <t_struct>`, and the number and types of elements
2272 must match those specified by the type.
2274 Array constants are represented with notation similar to array type
2275 definitions (a comma separated list of elements, surrounded by
2276 square brackets (``[]``)). For example:
2277 "``[ i32 42, i32 11, i32 74 ]``". Array constants must have
2278 :ref:`array type <t_array>`, and the number and types of elements must
2279 match those specified by the type.
2280 **Vector constants**
2281 Vector constants are represented with notation similar to vector
2282 type definitions (a comma separated list of elements, surrounded by
2283 less-than/greater-than's (``<>``)). For example:
2284 "``< i32 42, i32 11, i32 74, i32 100 >``". Vector constants
2285 must have :ref:`vector type <t_vector>`, and the number and types of
2286 elements must match those specified by the type.
2287 **Zero initialization**
2288 The string '``zeroinitializer``' can be used to zero initialize a
2289 value to zero of *any* type, including scalar and
2290 :ref:`aggregate <t_aggregate>` types. This is often used to avoid
2291 having to print large zero initializers (e.g. for large arrays) and
2292 is always exactly equivalent to using explicit zero initializers.
2294 A metadata node is a structure-like constant with :ref:`metadata
2295 type <t_metadata>`. For example:
2296 "``metadata !{ i32 0, metadata !"test" }``". Unlike other
2297 constants that are meant to be interpreted as part of the
2298 instruction stream, metadata is a place to attach additional
2299 information such as debug info.
2301 Global Variable and Function Addresses
2302 --------------------------------------
2304 The addresses of :ref:`global variables <globalvars>` and
2305 :ref:`functions <functionstructure>` are always implicitly valid
2306 (link-time) constants. These constants are explicitly referenced when
2307 the :ref:`identifier for the global <identifiers>` is used and always have
2308 :ref:`pointer <t_pointer>` type. For example, the following is a legal LLVM
2311 .. code-block:: llvm
2315 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2322 The string '``undef``' can be used anywhere a constant is expected, and
2323 indicates that the user of the value may receive an unspecified
2324 bit-pattern. Undefined values may be of any type (other than '``label``'
2325 or '``void``') and be used anywhere a constant is permitted.
2327 Undefined values are useful because they indicate to the compiler that
2328 the program is well defined no matter what value is used. This gives the
2329 compiler more freedom to optimize. Here are some examples of
2330 (potentially surprising) transformations that are valid (in pseudo IR):
2332 .. code-block:: llvm
2342 This is safe because all of the output bits are affected by the undef
2343 bits. Any output bit can have a zero or one depending on the input bits.
2345 .. code-block:: llvm
2356 These logical operations have bits that are not always affected by the
2357 input. For example, if ``%X`` has a zero bit, then the output of the
2358 '``and``' operation will always be a zero for that bit, no matter what
2359 the corresponding bit from the '``undef``' is. As such, it is unsafe to
2360 optimize or assume that the result of the '``and``' is '``undef``'.
2361 However, it is safe to assume that all bits of the '``undef``' could be
2362 0, and optimize the '``and``' to 0. Likewise, it is safe to assume that
2363 all the bits of the '``undef``' operand to the '``or``' could be set,
2364 allowing the '``or``' to be folded to -1.
2366 .. code-block:: llvm
2368 %A = select undef, %X, %Y
2369 %B = select undef, 42, %Y
2370 %C = select %X, %Y, undef
2380 This set of examples shows that undefined '``select``' (and conditional
2381 branch) conditions can go *either way*, but they have to come from one
2382 of the two operands. In the ``%A`` example, if ``%X`` and ``%Y`` were
2383 both known to have a clear low bit, then ``%A`` would have to have a
2384 cleared low bit. However, in the ``%C`` example, the optimizer is
2385 allowed to assume that the '``undef``' operand could be the same as
2386 ``%Y``, allowing the whole '``select``' to be eliminated.
2388 .. code-block:: llvm
2390 %A = xor undef, undef
2407 This example points out that two '``undef``' operands are not
2408 necessarily the same. This can be surprising to people (and also matches
2409 C semantics) where they assume that "``X^X``" is always zero, even if
2410 ``X`` is undefined. This isn't true for a number of reasons, but the
2411 short answer is that an '``undef``' "variable" can arbitrarily change
2412 its value over its "live range". This is true because the variable
2413 doesn't actually *have a live range*. Instead, the value is logically
2414 read from arbitrary registers that happen to be around when needed, so
2415 the value is not necessarily consistent over time. In fact, ``%A`` and
2416 ``%C`` need to have the same semantics or the core LLVM "replace all
2417 uses with" concept would not hold.
2419 .. code-block:: llvm
2427 These examples show the crucial difference between an *undefined value*
2428 and *undefined behavior*. An undefined value (like '``undef``') is
2429 allowed to have an arbitrary bit-pattern. This means that the ``%A``
2430 operation can be constant folded to '``undef``', because the '``undef``'
2431 could be an SNaN, and ``fdiv`` is not (currently) defined on SNaN's.
2432 However, in the second example, we can make a more aggressive
2433 assumption: because the ``undef`` is allowed to be an arbitrary value,
2434 we are allowed to assume that it could be zero. Since a divide by zero
2435 has *undefined behavior*, we are allowed to assume that the operation
2436 does not execute at all. This allows us to delete the divide and all
2437 code after it. Because the undefined operation "can't happen", the
2438 optimizer can assume that it occurs in dead code.
2440 .. code-block:: llvm
2442 a: store undef -> %X
2443 b: store %X -> undef
2448 These examples reiterate the ``fdiv`` example: a store *of* an undefined
2449 value can be assumed to not have any effect; we can assume that the
2450 value is overwritten with bits that happen to match what was already
2451 there. However, a store *to* an undefined location could clobber
2452 arbitrary memory, therefore, it has undefined behavior.
2459 Poison values are similar to :ref:`undef values <undefvalues>`, however
2460 they also represent the fact that an instruction or constant expression
2461 that cannot evoke side effects has nevertheless detected a condition
2462 that results in undefined behavior.
2464 There is currently no way of representing a poison value in the IR; they
2465 only exist when produced by operations such as :ref:`add <i_add>` with
2468 Poison value behavior is defined in terms of value *dependence*:
2470 - Values other than :ref:`phi <i_phi>` nodes depend on their operands.
2471 - :ref:`Phi <i_phi>` nodes depend on the operand corresponding to
2472 their dynamic predecessor basic block.
2473 - Function arguments depend on the corresponding actual argument values
2474 in the dynamic callers of their functions.
2475 - :ref:`Call <i_call>` instructions depend on the :ref:`ret <i_ret>`
2476 instructions that dynamically transfer control back to them.
2477 - :ref:`Invoke <i_invoke>` instructions depend on the
2478 :ref:`ret <i_ret>`, :ref:`resume <i_resume>`, or exception-throwing
2479 call instructions that dynamically transfer control back to them.
2480 - Non-volatile loads and stores depend on the most recent stores to all
2481 of the referenced memory addresses, following the order in the IR
2482 (including loads and stores implied by intrinsics such as
2483 :ref:`@llvm.memcpy <int_memcpy>`.)
2484 - An instruction with externally visible side effects depends on the
2485 most recent preceding instruction with externally visible side
2486 effects, following the order in the IR. (This includes :ref:`volatile
2487 operations <volatile>`.)
2488 - An instruction *control-depends* on a :ref:`terminator
2489 instruction <terminators>` if the terminator instruction has
2490 multiple successors and the instruction is always executed when
2491 control transfers to one of the successors, and may not be executed
2492 when control is transferred to another.
2493 - Additionally, an instruction also *control-depends* on a terminator
2494 instruction if the set of instructions it otherwise depends on would
2495 be different if the terminator had transferred control to a different
2497 - Dependence is transitive.
2499 Poison values have the same behavior as :ref:`undef values <undefvalues>`,
2500 with the additional effect that any instruction that has a *dependence*
2501 on a poison value has undefined behavior.
2503 Here are some examples:
2505 .. code-block:: llvm
2508 %poison = sub nuw i32 0, 1 ; Results in a poison value.
2509 %still_poison = and i32 %poison, 0 ; 0, but also poison.
2510 %poison_yet_again = getelementptr i32* @h, i32 %still_poison
2511 store i32 0, i32* %poison_yet_again ; memory at @h[0] is poisoned
2513 store i32 %poison, i32* @g ; Poison value stored to memory.
2514 %poison2 = load i32* @g ; Poison value loaded back from memory.
2516 store volatile i32 %poison, i32* @g ; External observation; undefined behavior.
2518 %narrowaddr = bitcast i32* @g to i16*
2519 %wideaddr = bitcast i32* @g to i64*
2520 %poison3 = load i16* %narrowaddr ; Returns a poison value.
2521 %poison4 = load i64* %wideaddr ; Returns a poison value.
2523 %cmp = icmp slt i32 %poison, 0 ; Returns a poison value.
2524 br i1 %cmp, label %true, label %end ; Branch to either destination.
2527 store volatile i32 0, i32* @g ; This is control-dependent on %cmp, so
2528 ; it has undefined behavior.
2532 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2533 ; Both edges into this PHI are
2534 ; control-dependent on %cmp, so this
2535 ; always results in a poison value.
2537 store volatile i32 0, i32* @g ; This would depend on the store in %true
2538 ; if %cmp is true, or the store in %entry
2539 ; otherwise, so this is undefined behavior.
2541 br i1 %cmp, label %second_true, label %second_end
2542 ; The same branch again, but this time the
2543 ; true block doesn't have side effects.
2550 store volatile i32 0, i32* @g ; This time, the instruction always depends
2551 ; on the store in %end. Also, it is
2552 ; control-equivalent to %end, so this is
2553 ; well-defined (ignoring earlier undefined
2554 ; behavior in this example).
2558 Addresses of Basic Blocks
2559 -------------------------
2561 ``blockaddress(@function, %block)``
2563 The '``blockaddress``' constant computes the address of the specified
2564 basic block in the specified function, and always has an ``i8*`` type.
2565 Taking the address of the entry block is illegal.
2567 This value only has defined behavior when used as an operand to the
2568 ':ref:`indirectbr <i_indirectbr>`' instruction, or for comparisons
2569 against null. Pointer equality tests between labels addresses results in
2570 undefined behavior --- though, again, comparison against null is ok, and
2571 no label is equal to the null pointer. This may be passed around as an
2572 opaque pointer sized value as long as the bits are not inspected. This
2573 allows ``ptrtoint`` and arithmetic to be performed on these values so
2574 long as the original value is reconstituted before the ``indirectbr``
2577 Finally, some targets may provide defined semantics when using the value
2578 as the operand to an inline assembly, but that is target specific.
2582 Constant Expressions
2583 --------------------
2585 Constant expressions are used to allow expressions involving other
2586 constants to be used as constants. Constant expressions may be of any
2587 :ref:`first class <t_firstclass>` type and may involve any LLVM operation
2588 that does not have side effects (e.g. load and call are not supported).
2589 The following is the syntax for constant expressions:
2591 ``trunc (CST to TYPE)``
2592 Truncate a constant to another type. The bit size of CST must be
2593 larger than the bit size of TYPE. Both types must be integers.
2594 ``zext (CST to TYPE)``
2595 Zero extend a constant to another type. The bit size of CST must be
2596 smaller than the bit size of TYPE. Both types must be integers.
2597 ``sext (CST to TYPE)``
2598 Sign extend a constant to another type. The bit size of CST must be
2599 smaller than the bit size of TYPE. Both types must be integers.
2600 ``fptrunc (CST to TYPE)``
2601 Truncate a floating point constant to another floating point type.
2602 The size of CST must be larger than the size of TYPE. Both types
2603 must be floating point.
2604 ``fpext (CST to TYPE)``
2605 Floating point extend a constant to another type. The size of CST
2606 must be smaller or equal to the size of TYPE. Both types must be
2608 ``fptoui (CST to TYPE)``
2609 Convert a floating point constant to the corresponding unsigned
2610 integer constant. TYPE must be a scalar or vector integer type. CST
2611 must be of scalar or vector floating point type. Both CST and TYPE
2612 must be scalars, or vectors of the same number of elements. If the
2613 value won't fit in the integer type, the results are undefined.
2614 ``fptosi (CST to TYPE)``
2615 Convert a floating point constant to the corresponding signed
2616 integer constant. TYPE must be a scalar or vector integer type. CST
2617 must be of scalar or vector floating point type. Both CST and TYPE
2618 must be scalars, or vectors of the same number of elements. If the
2619 value won't fit in the integer type, the results are undefined.
2620 ``uitofp (CST to TYPE)``
2621 Convert an unsigned integer constant to the corresponding floating
2622 point constant. TYPE must be a scalar or vector floating point type.
2623 CST must be of scalar or vector integer type. Both CST and TYPE must
2624 be scalars, or vectors of the same number of elements. If the value
2625 won't fit in the floating point type, the results are undefined.
2626 ``sitofp (CST to TYPE)``
2627 Convert a signed integer constant to the corresponding floating
2628 point constant. TYPE must be a scalar or vector floating point type.
2629 CST must be of scalar or vector integer type. Both CST and TYPE must
2630 be scalars, or vectors of the same number of elements. If the value
2631 won't fit in the floating point type, the results are undefined.
2632 ``ptrtoint (CST to TYPE)``
2633 Convert a pointer typed constant to the corresponding integer
2634 constant. ``TYPE`` must be an integer type. ``CST`` must be of
2635 pointer type. The ``CST`` value is zero extended, truncated, or
2636 unchanged to make it fit in ``TYPE``.
2637 ``inttoptr (CST to TYPE)``
2638 Convert an integer constant to a pointer constant. TYPE must be a
2639 pointer type. CST must be of integer type. The CST value is zero
2640 extended, truncated, or unchanged to make it fit in a pointer size.
2641 This one is *really* dangerous!
2642 ``bitcast (CST to TYPE)``
2643 Convert a constant, CST, to another TYPE. The constraints of the
2644 operands are the same as those for the :ref:`bitcast
2645 instruction <i_bitcast>`.
2646 ``addrspacecast (CST to TYPE)``
2647 Convert a constant pointer or constant vector of pointer, CST, to another
2648 TYPE in a different address space. The constraints of the operands are the
2649 same as those for the :ref:`addrspacecast instruction <i_addrspacecast>`.
2650 ``getelementptr (CSTPTR, IDX0, IDX1, ...)``, ``getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)``
2651 Perform the :ref:`getelementptr operation <i_getelementptr>` on
2652 constants. As with the :ref:`getelementptr <i_getelementptr>`
2653 instruction, the index list may have zero or more indexes, which are
2654 required to make sense for the type of "CSTPTR".
2655 ``select (COND, VAL1, VAL2)``
2656 Perform the :ref:`select operation <i_select>` on constants.
2657 ``icmp COND (VAL1, VAL2)``
2658 Performs the :ref:`icmp operation <i_icmp>` on constants.
2659 ``fcmp COND (VAL1, VAL2)``
2660 Performs the :ref:`fcmp operation <i_fcmp>` on constants.
2661 ``extractelement (VAL, IDX)``
2662 Perform the :ref:`extractelement operation <i_extractelement>` on
2664 ``insertelement (VAL, ELT, IDX)``
2665 Perform the :ref:`insertelement operation <i_insertelement>` on
2667 ``shufflevector (VEC1, VEC2, IDXMASK)``
2668 Perform the :ref:`shufflevector operation <i_shufflevector>` on
2670 ``extractvalue (VAL, IDX0, IDX1, ...)``
2671 Perform the :ref:`extractvalue operation <i_extractvalue>` on
2672 constants. The index list is interpreted in a similar manner as
2673 indices in a ':ref:`getelementptr <i_getelementptr>`' operation. At
2674 least one index value must be specified.
2675 ``insertvalue (VAL, ELT, IDX0, IDX1, ...)``
2676 Perform the :ref:`insertvalue operation <i_insertvalue>` on constants.
2677 The index list is interpreted in a similar manner as indices in a
2678 ':ref:`getelementptr <i_getelementptr>`' operation. At least one index
2679 value must be specified.
2680 ``OPCODE (LHS, RHS)``
2681 Perform the specified operation of the LHS and RHS constants. OPCODE
2682 may be any of the :ref:`binary <binaryops>` or :ref:`bitwise
2683 binary <bitwiseops>` operations. The constraints on operands are
2684 the same as those for the corresponding instruction (e.g. no bitwise
2685 operations on floating point values are allowed).
2692 Inline Assembler Expressions
2693 ----------------------------
2695 LLVM supports inline assembler expressions (as opposed to :ref:`Module-Level
2696 Inline Assembly <moduleasm>`) through the use of a special value. This
2697 value represents the inline assembler as a string (containing the
2698 instructions to emit), a list of operand constraints (stored as a
2699 string), a flag that indicates whether or not the inline asm expression
2700 has side effects, and a flag indicating whether the function containing
2701 the asm needs to align its stack conservatively. An example inline
2702 assembler expression is:
2704 .. code-block:: llvm
2706 i32 (i32) asm "bswap $0", "=r,r"
2708 Inline assembler expressions may **only** be used as the callee operand
2709 of a :ref:`call <i_call>` or an :ref:`invoke <i_invoke>` instruction.
2710 Thus, typically we have:
2712 .. code-block:: llvm
2714 %X = call i32 asm "bswap $0", "=r,r"(i32 %Y)
2716 Inline asms with side effects not visible in the constraint list must be
2717 marked as having side effects. This is done through the use of the
2718 '``sideeffect``' keyword, like so:
2720 .. code-block:: llvm
2722 call void asm sideeffect "eieio", ""()
2724 In some cases inline asms will contain code that will not work unless
2725 the stack is aligned in some way, such as calls or SSE instructions on
2726 x86, yet will not contain code that does that alignment within the asm.
2727 The compiler should make conservative assumptions about what the asm
2728 might contain and should generate its usual stack alignment code in the
2729 prologue if the '``alignstack``' keyword is present:
2731 .. code-block:: llvm
2733 call void asm alignstack "eieio", ""()
2735 Inline asms also support using non-standard assembly dialects. The
2736 assumed dialect is ATT. When the '``inteldialect``' keyword is present,
2737 the inline asm is using the Intel dialect. Currently, ATT and Intel are
2738 the only supported dialects. An example is:
2740 .. code-block:: llvm
2742 call void asm inteldialect "eieio", ""()
2744 If multiple keywords appear the '``sideeffect``' keyword must come
2745 first, the '``alignstack``' keyword second and the '``inteldialect``'
2751 The call instructions that wrap inline asm nodes may have a
2752 "``!srcloc``" MDNode attached to it that contains a list of constant
2753 integers. If present, the code generator will use the integer as the
2754 location cookie value when report errors through the ``LLVMContext``
2755 error reporting mechanisms. This allows a front-end to correlate backend
2756 errors that occur with inline asm back to the source code that produced
2759 .. code-block:: llvm
2761 call void asm sideeffect "something bad", ""(), !srcloc !42
2763 !42 = !{ i32 1234567 }
2765 It is up to the front-end to make sense of the magic numbers it places
2766 in the IR. If the MDNode contains multiple constants, the code generator
2767 will use the one that corresponds to the line of the asm that the error
2772 Metadata Nodes and Metadata Strings
2773 -----------------------------------
2775 LLVM IR allows metadata to be attached to instructions in the program
2776 that can convey extra information about the code to the optimizers and
2777 code generator. One example application of metadata is source-level
2778 debug information. There are two metadata primitives: strings and nodes.
2779 All metadata has the ``metadata`` type and is identified in syntax by a
2780 preceding exclamation point ('``!``').
2782 A metadata string is a string surrounded by double quotes. It can
2783 contain any character by escaping non-printable characters with
2784 "``\xx``" where "``xx``" is the two digit hex code. For example:
2787 Metadata nodes are represented with notation similar to structure
2788 constants (a comma separated list of elements, surrounded by braces and
2789 preceded by an exclamation point). Metadata nodes can have any values as
2790 their operand. For example:
2792 .. code-block:: llvm
2794 !{ metadata !"test\00", i32 10}
2796 A :ref:`named metadata <namedmetadatastructure>` is a collection of
2797 metadata nodes, which can be looked up in the module symbol table. For
2800 .. code-block:: llvm
2802 !foo = metadata !{!4, !3}
2804 Metadata can be used as function arguments. Here ``llvm.dbg.value``
2805 function is using two metadata arguments:
2807 .. code-block:: llvm
2809 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2811 Metadata can be attached with an instruction. Here metadata ``!21`` is
2812 attached to the ``add`` instruction using the ``!dbg`` identifier:
2814 .. code-block:: llvm
2816 %indvar.next = add i64 %indvar, 1, !dbg !21
2818 More information about specific metadata nodes recognized by the
2819 optimizers and code generator is found below.
2824 In LLVM IR, memory does not have types, so LLVM's own type system is not
2825 suitable for doing TBAA. Instead, metadata is added to the IR to
2826 describe a type system of a higher level language. This can be used to
2827 implement typical C/C++ TBAA, but it can also be used to implement
2828 custom alias analysis behavior for other languages.
2830 The current metadata format is very simple. TBAA metadata nodes have up
2831 to three fields, e.g.:
2833 .. code-block:: llvm
2835 !0 = metadata !{ metadata !"an example type tree" }
2836 !1 = metadata !{ metadata !"int", metadata !0 }
2837 !2 = metadata !{ metadata !"float", metadata !0 }
2838 !3 = metadata !{ metadata !"const float", metadata !2, i64 1 }
2840 The first field is an identity field. It can be any value, usually a
2841 metadata string, which uniquely identifies the type. The most important
2842 name in the tree is the name of the root node. Two trees with different
2843 root node names are entirely disjoint, even if they have leaves with
2846 The second field identifies the type's parent node in the tree, or is
2847 null or omitted for a root node. A type is considered to alias all of
2848 its descendants and all of its ancestors in the tree. Also, a type is
2849 considered to alias all types in other trees, so that bitcode produced
2850 from multiple front-ends is handled conservatively.
2852 If the third field is present, it's an integer which if equal to 1
2853 indicates that the type is "constant" (meaning
2854 ``pointsToConstantMemory`` should return true; see `other useful
2855 AliasAnalysis methods <AliasAnalysis.html#OtherItfs>`_).
2857 '``tbaa.struct``' Metadata
2858 ^^^^^^^^^^^^^^^^^^^^^^^^^^
2860 The :ref:`llvm.memcpy <int_memcpy>` is often used to implement
2861 aggregate assignment operations in C and similar languages, however it
2862 is defined to copy a contiguous region of memory, which is more than
2863 strictly necessary for aggregate types which contain holes due to
2864 padding. Also, it doesn't contain any TBAA information about the fields
2867 ``!tbaa.struct`` metadata can describe which memory subregions in a
2868 memcpy are padding and what the TBAA tags of the struct are.
2870 The current metadata format is very simple. ``!tbaa.struct`` metadata
2871 nodes are a list of operands which are in conceptual groups of three.
2872 For each group of three, the first operand gives the byte offset of a
2873 field in bytes, the second gives its size in bytes, and the third gives
2876 .. code-block:: llvm
2878 !4 = metadata !{ i64 0, i64 4, metadata !1, i64 8, i64 4, metadata !2 }
2880 This describes a struct with two fields. The first is at offset 0 bytes
2881 with size 4 bytes, and has tbaa tag !1. The second is at offset 8 bytes
2882 and has size 4 bytes and has tbaa tag !2.
2884 Note that the fields need not be contiguous. In this example, there is a
2885 4 byte gap between the two fields. This gap represents padding which
2886 does not carry useful data and need not be preserved.
2888 '``noalias``' and '``alias.scope``' Metadata
2889 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2891 ``noalias`` and ``alias.scope`` metadata provide the ability to specify generic
2892 noalias memory-access sets. This means that some collection of memory access
2893 instructions (loads, stores, memory-accessing calls, etc.) that carry
2894 ``noalias`` metadata can specifically be specified not to alias with some other
2895 collection of memory access instructions that carry ``alias.scope`` metadata.
2896 Each type of metadata specifies a list of scopes where each scope has an id and
2897 a domain. When evaluating an aliasing query, if for some some domain, the set
2898 of scopes with that domain in one instruction's ``alias.scope`` list is a
2899 subset of (or qual to) the set of scopes for that domain in another
2900 instruction's ``noalias`` list, then the two memory accesses are assumed not to
2903 The metadata identifying each domain is itself a list containing one or two
2904 entries. The first entry is the name of the domain. Note that if the name is a
2905 string then it can be combined accross functions and translation units. A
2906 self-reference can be used to create globally unique domain names. A
2907 descriptive string may optionally be provided as a second list entry.
2909 The metadata identifying each scope is also itself a list containing two or
2910 three entries. The first entry is the name of the scope. Note that if the name
2911 is a string then it can be combined accross functions and translation units. A
2912 self-reference can be used to create globally unique scope names. A metadata
2913 reference to the scope's domain is the second entry. A descriptive string may
2914 optionally be provided as a third list entry.
2918 .. code-block:: llvm
2920 ; Two scope domains:
2921 !0 = metadata !{metadata !0}
2922 !1 = metadata !{metadata !1}
2924 ; Some scopes in these domains:
2925 !2 = metadata !{metadata !2, metadata !0}
2926 !3 = metadata !{metadata !3, metadata !0}
2927 !4 = metadata !{metadata !4, metadata !1}
2930 !5 = metadata !{metadata !4} ; A list containing only scope !4
2931 !6 = metadata !{metadata !4, metadata !3, metadata !2}
2932 !7 = metadata !{metadata !3}
2934 ; These two instructions don't alias:
2935 %0 = load float* %c, align 4, !alias.scope !5
2936 store float %0, float* %arrayidx.i, align 4, !noalias !5
2938 ; These two instructions also don't alias (for domain !1, the set of scopes
2939 ; in the !alias.scope equals that in the !noalias list):
2940 %2 = load float* %c, align 4, !alias.scope !5
2941 store float %2, float* %arrayidx.i2, align 4, !noalias !6
2943 ; These two instructions don't alias (for domain !0, the set of scopes in
2944 ; the !noalias list is not a superset of, or equal to, the scopes in the
2945 ; !alias.scope list):
2946 %2 = load float* %c, align 4, !alias.scope !6
2947 store float %0, float* %arrayidx.i, align 4, !noalias !7
2949 '``fpmath``' Metadata
2950 ^^^^^^^^^^^^^^^^^^^^^
2952 ``fpmath`` metadata may be attached to any instruction of floating point
2953 type. It can be used to express the maximum acceptable error in the
2954 result of that instruction, in ULPs, thus potentially allowing the
2955 compiler to use a more efficient but less accurate method of computing
2956 it. ULP is defined as follows:
2958 If ``x`` is a real number that lies between two finite consecutive
2959 floating-point numbers ``a`` and ``b``, without being equal to one
2960 of them, then ``ulp(x) = |b - a|``, otherwise ``ulp(x)`` is the
2961 distance between the two non-equal finite floating-point numbers
2962 nearest ``x``. Moreover, ``ulp(NaN)`` is ``NaN``.
2964 The metadata node shall consist of a single positive floating point
2965 number representing the maximum relative error, for example:
2967 .. code-block:: llvm
2969 !0 = metadata !{ float 2.5 } ; maximum acceptable inaccuracy is 2.5 ULPs
2971 '``range``' Metadata
2972 ^^^^^^^^^^^^^^^^^^^^
2974 ``range`` metadata may be attached only to ``load``, ``call`` and ``invoke`` of
2975 integer types. It expresses the possible ranges the loaded value or the value
2976 returned by the called function at this call site is in. The ranges are
2977 represented with a flattened list of integers. The loaded value or the value
2978 returned is known to be in the union of the ranges defined by each consecutive
2979 pair. Each pair has the following properties:
2981 - The type must match the type loaded by the instruction.
2982 - The pair ``a,b`` represents the range ``[a,b)``.
2983 - Both ``a`` and ``b`` are constants.
2984 - The range is allowed to wrap.
2985 - The range should not represent the full or empty set. That is,
2988 In addition, the pairs must be in signed order of the lower bound and
2989 they must be non-contiguous.
2993 .. code-block:: llvm
2995 %a = load i8* %x, align 1, !range !0 ; Can only be 0 or 1
2996 %b = load i8* %y, align 1, !range !1 ; Can only be 255 (-1), 0 or 1
2997 %c = call i8 @foo(), !range !2 ; Can only be 0, 1, 3, 4 or 5
2998 %d = invoke i8 @bar() to label %cont
2999 unwind label %lpad, !range !3 ; Can only be -2, -1, 3, 4 or 5
3001 !0 = metadata !{ i8 0, i8 2 }
3002 !1 = metadata !{ i8 255, i8 2 }
3003 !2 = metadata !{ i8 0, i8 2, i8 3, i8 6 }
3004 !3 = metadata !{ i8 -2, i8 0, i8 3, i8 6 }
3009 It is sometimes useful to attach information to loop constructs. Currently,
3010 loop metadata is implemented as metadata attached to the branch instruction
3011 in the loop latch block. This type of metadata refer to a metadata node that is
3012 guaranteed to be separate for each loop. The loop identifier metadata is
3013 specified with the name ``llvm.loop``.
3015 The loop identifier metadata is implemented using a metadata that refers to
3016 itself to avoid merging it with any other identifier metadata, e.g.,
3017 during module linkage or function inlining. That is, each loop should refer
3018 to their own identification metadata even if they reside in separate functions.
3019 The following example contains loop identifier metadata for two separate loop
3022 .. code-block:: llvm
3024 !0 = metadata !{ metadata !0 }
3025 !1 = metadata !{ metadata !1 }
3027 The loop identifier metadata can be used to specify additional
3028 per-loop metadata. Any operands after the first operand can be treated
3029 as user-defined metadata. For example the ``llvm.loop.unroll.count``
3030 suggests an unroll factor to the loop unroller:
3032 .. code-block:: llvm
3034 br i1 %exitcond, label %._crit_edge, label %.lr.ph, !llvm.loop !0
3036 !0 = metadata !{ metadata !0, metadata !1 }
3037 !1 = metadata !{ metadata !"llvm.loop.unroll.count", i32 4 }
3039 '``llvm.loop.vectorize``' and '``llvm.loop.interleave``'
3040 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3042 Metadata prefixed with ``llvm.loop.vectorize`` or ``llvm.loop.interleave`` are
3043 used to control per-loop vectorization and interleaving parameters such as
3044 vectorization width and interleave count. These metadata should be used in
3045 conjunction with ``llvm.loop`` loop identification metadata. The
3046 ``llvm.loop.vectorize`` and ``llvm.loop.interleave`` metadata are only
3047 optimization hints and the optimizer will only interleave and vectorize loops if
3048 it believes it is safe to do so. The ``llvm.mem.parallel_loop_access`` metadata
3049 which contains information about loop-carried memory dependencies can be helpful
3050 in determining the safety of these transformations.
3052 '``llvm.loop.interleave.count``' Metadata
3053 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3055 This metadata suggests an interleave count to the loop interleaver.
3056 The first operand is the string ``llvm.loop.interleave.count`` and the
3057 second operand is an integer specifying the interleave count. For
3060 .. code-block:: llvm
3062 !0 = metadata !{ metadata !"llvm.loop.interleave.count", i32 4 }
3064 Note that setting ``llvm.loop.interleave.count`` to 1 disables interleaving
3065 multiple iterations of the loop. If ``llvm.loop.interleave.count`` is set to 0
3066 then the interleave count will be determined automatically.
3068 '``llvm.loop.vectorize.enable``' Metadata
3069 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3071 This metadata selectively enables or disables vectorization for the loop. The
3072 first operand is the string ``llvm.loop.vectorize.enable`` and the second operand
3073 is a bit. If the bit operand value is 1 vectorization is enabled. A value of
3074 0 disables vectorization:
3076 .. code-block:: llvm
3078 !0 = metadata !{ metadata !"llvm.loop.vectorize.enable", i1 0 }
3079 !1 = metadata !{ metadata !"llvm.loop.vectorize.enable", i1 1 }
3081 '``llvm.loop.vectorize.width``' Metadata
3082 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3084 This metadata sets the target width of the vectorizer. The first
3085 operand is the string ``llvm.loop.vectorize.width`` and the second
3086 operand is an integer specifying the width. For example:
3088 .. code-block:: llvm
3090 !0 = metadata !{ metadata !"llvm.loop.vectorize.width", i32 4 }
3092 Note that setting ``llvm.loop.vectorize.width`` to 1 disables
3093 vectorization of the loop. If ``llvm.loop.vectorize.width`` is set to
3094 0 or if the loop does not have this metadata the width will be
3095 determined automatically.
3097 '``llvm.loop.unroll``'
3098 ^^^^^^^^^^^^^^^^^^^^^^
3100 Metadata prefixed with ``llvm.loop.unroll`` are loop unrolling
3101 optimization hints such as the unroll factor. ``llvm.loop.unroll``
3102 metadata should be used in conjunction with ``llvm.loop`` loop
3103 identification metadata. The ``llvm.loop.unroll`` metadata are only
3104 optimization hints and the unrolling will only be performed if the
3105 optimizer believes it is safe to do so.
3107 '``llvm.loop.unroll.count``' Metadata
3108 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3110 This metadata suggests an unroll factor to the loop unroller. The
3111 first operand is the string ``llvm.loop.unroll.count`` and the second
3112 operand is a positive integer specifying the unroll factor. For
3115 .. code-block:: llvm
3117 !0 = metadata !{ metadata !"llvm.loop.unroll.count", i32 4 }
3119 If the trip count of the loop is less than the unroll count the loop
3120 will be partially unrolled.
3122 '``llvm.loop.unroll.disable``' Metadata
3123 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3125 This metadata either disables loop unrolling. The metadata has a single operand
3126 which is the string ``llvm.loop.unroll.disable``. For example:
3128 .. code-block:: llvm
3130 !0 = metadata !{ metadata !"llvm.loop.unroll.disable" }
3132 '``llvm.loop.unroll.full``' Metadata
3133 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3135 This metadata either suggests that the loop should be unrolled fully. The
3136 metadata has a single operand which is the string ``llvm.loop.unroll.disable``.
3139 .. code-block:: llvm
3141 !0 = metadata !{ metadata !"llvm.loop.unroll.full" }
3146 Metadata types used to annotate memory accesses with information helpful
3147 for optimizations are prefixed with ``llvm.mem``.
3149 '``llvm.mem.parallel_loop_access``' Metadata
3150 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3152 The ``llvm.mem.parallel_loop_access`` metadata refers to a loop identifier,
3153 or metadata containing a list of loop identifiers for nested loops.
3154 The metadata is attached to memory accessing instructions and denotes that
3155 no loop carried memory dependence exist between it and other instructions denoted
3156 with the same loop identifier.
3158 Precisely, given two instructions ``m1`` and ``m2`` that both have the
3159 ``llvm.mem.parallel_loop_access`` metadata, with ``L1`` and ``L2`` being the
3160 set of loops associated with that metadata, respectively, then there is no loop
3161 carried dependence between ``m1`` and ``m2`` for loops in both ``L1`` and
3164 As a special case, if all memory accessing instructions in a loop have
3165 ``llvm.mem.parallel_loop_access`` metadata that refers to that loop, then the
3166 loop has no loop carried memory dependences and is considered to be a parallel
3169 Note that if not all memory access instructions have such metadata referring to
3170 the loop, then the loop is considered not being trivially parallel. Additional
3171 memory dependence analysis is required to make that determination. As a fail
3172 safe mechanism, this causes loops that were originally parallel to be considered
3173 sequential (if optimization passes that are unaware of the parallel semantics
3174 insert new memory instructions into the loop body).
3176 Example of a loop that is considered parallel due to its correct use of
3177 both ``llvm.loop`` and ``llvm.mem.parallel_loop_access``
3178 metadata types that refer to the same loop identifier metadata.
3180 .. code-block:: llvm
3184 %val0 = load i32* %arrayidx, !llvm.mem.parallel_loop_access !0
3186 store i32 %val0, i32* %arrayidx1, !llvm.mem.parallel_loop_access !0
3188 br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0
3192 !0 = metadata !{ metadata !0 }
3194 It is also possible to have nested parallel loops. In that case the
3195 memory accesses refer to a list of loop identifier metadata nodes instead of
3196 the loop identifier metadata node directly:
3198 .. code-block:: llvm
3202 %val1 = load i32* %arrayidx3, !llvm.mem.parallel_loop_access !2
3204 br label %inner.for.body
3208 %val0 = load i32* %arrayidx1, !llvm.mem.parallel_loop_access !0
3210 store i32 %val0, i32* %arrayidx2, !llvm.mem.parallel_loop_access !0
3212 br i1 %exitcond, label %inner.for.end, label %inner.for.body, !llvm.loop !1
3216 store i32 %val1, i32* %arrayidx4, !llvm.mem.parallel_loop_access !2
3218 br i1 %exitcond, label %outer.for.end, label %outer.for.body, !llvm.loop !2
3220 outer.for.end: ; preds = %for.body
3222 !0 = metadata !{ metadata !1, metadata !2 } ; a list of loop identifiers
3223 !1 = metadata !{ metadata !1 } ; an identifier for the inner loop
3224 !2 = metadata !{ metadata !2 } ; an identifier for the outer loop
3226 Module Flags Metadata
3227 =====================
3229 Information about the module as a whole is difficult to convey to LLVM's
3230 subsystems. The LLVM IR isn't sufficient to transmit this information.
3231 The ``llvm.module.flags`` named metadata exists in order to facilitate
3232 this. These flags are in the form of key / value pairs --- much like a
3233 dictionary --- making it easy for any subsystem who cares about a flag to
3236 The ``llvm.module.flags`` metadata contains a list of metadata triplets.
3237 Each triplet has the following form:
3239 - The first element is a *behavior* flag, which specifies the behavior
3240 when two (or more) modules are merged together, and it encounters two
3241 (or more) metadata with the same ID. The supported behaviors are
3243 - The second element is a metadata string that is a unique ID for the
3244 metadata. Each module may only have one flag entry for each unique ID (not
3245 including entries with the **Require** behavior).
3246 - The third element is the value of the flag.
3248 When two (or more) modules are merged together, the resulting
3249 ``llvm.module.flags`` metadata is the union of the modules' flags. That is, for
3250 each unique metadata ID string, there will be exactly one entry in the merged
3251 modules ``llvm.module.flags`` metadata table, and the value for that entry will
3252 be determined by the merge behavior flag, as described below. The only exception
3253 is that entries with the *Require* behavior are always preserved.
3255 The following behaviors are supported:
3266 Emits an error if two values disagree, otherwise the resulting value
3267 is that of the operands.
3271 Emits a warning if two values disagree. The result value will be the
3272 operand for the flag from the first module being linked.
3276 Adds a requirement that another module flag be present and have a
3277 specified value after linking is performed. The value must be a
3278 metadata pair, where the first element of the pair is the ID of the
3279 module flag to be restricted, and the second element of the pair is
3280 the value the module flag should be restricted to. This behavior can
3281 be used to restrict the allowable results (via triggering of an
3282 error) of linking IDs with the **Override** behavior.
3286 Uses the specified value, regardless of the behavior or value of the
3287 other module. If both modules specify **Override**, but the values
3288 differ, an error will be emitted.
3292 Appends the two values, which are required to be metadata nodes.
3296 Appends the two values, which are required to be metadata
3297 nodes. However, duplicate entries in the second list are dropped
3298 during the append operation.
3300 It is an error for a particular unique flag ID to have multiple behaviors,
3301 except in the case of **Require** (which adds restrictions on another metadata
3302 value) or **Override**.
3304 An example of module flags:
3306 .. code-block:: llvm
3308 !0 = metadata !{ i32 1, metadata !"foo", i32 1 }
3309 !1 = metadata !{ i32 4, metadata !"bar", i32 37 }
3310 !2 = metadata !{ i32 2, metadata !"qux", i32 42 }
3311 !3 = metadata !{ i32 3, metadata !"qux",
3313 metadata !"foo", i32 1
3316 !llvm.module.flags = !{ !0, !1, !2, !3 }
3318 - Metadata ``!0`` has the ID ``!"foo"`` and the value '1'. The behavior
3319 if two or more ``!"foo"`` flags are seen is to emit an error if their
3320 values are not equal.
3322 - Metadata ``!1`` has the ID ``!"bar"`` and the value '37'. The
3323 behavior if two or more ``!"bar"`` flags are seen is to use the value
3326 - Metadata ``!2`` has the ID ``!"qux"`` and the value '42'. The
3327 behavior if two or more ``!"qux"`` flags are seen is to emit a
3328 warning if their values are not equal.
3330 - Metadata ``!3`` has the ID ``!"qux"`` and the value:
3334 metadata !{ metadata !"foo", i32 1 }
3336 The behavior is to emit an error if the ``llvm.module.flags`` does not
3337 contain a flag with the ID ``!"foo"`` that has the value '1' after linking is
3340 Objective-C Garbage Collection Module Flags Metadata
3341 ----------------------------------------------------
3343 On the Mach-O platform, Objective-C stores metadata about garbage
3344 collection in a special section called "image info". The metadata
3345 consists of a version number and a bitmask specifying what types of
3346 garbage collection are supported (if any) by the file. If two or more
3347 modules are linked together their garbage collection metadata needs to
3348 be merged rather than appended together.
3350 The Objective-C garbage collection module flags metadata consists of the
3351 following key-value pairs:
3360 * - ``Objective-C Version``
3361 - **[Required]** --- The Objective-C ABI version. Valid values are 1 and 2.
3363 * - ``Objective-C Image Info Version``
3364 - **[Required]** --- The version of the image info section. Currently
3367 * - ``Objective-C Image Info Section``
3368 - **[Required]** --- The section to place the metadata. Valid values are
3369 ``"__OBJC, __image_info, regular"`` for Objective-C ABI version 1, and
3370 ``"__DATA,__objc_imageinfo, regular, no_dead_strip"`` for
3371 Objective-C ABI version 2.
3373 * - ``Objective-C Garbage Collection``
3374 - **[Required]** --- Specifies whether garbage collection is supported or
3375 not. Valid values are 0, for no garbage collection, and 2, for garbage
3376 collection supported.
3378 * - ``Objective-C GC Only``
3379 - **[Optional]** --- Specifies that only garbage collection is supported.
3380 If present, its value must be 6. This flag requires that the
3381 ``Objective-C Garbage Collection`` flag have the value 2.
3383 Some important flag interactions:
3385 - If a module with ``Objective-C Garbage Collection`` set to 0 is
3386 merged with a module with ``Objective-C Garbage Collection`` set to
3387 2, then the resulting module has the
3388 ``Objective-C Garbage Collection`` flag set to 0.
3389 - A module with ``Objective-C Garbage Collection`` set to 0 cannot be
3390 merged with a module with ``Objective-C GC Only`` set to 6.
3392 Automatic Linker Flags Module Flags Metadata
3393 --------------------------------------------
3395 Some targets support embedding flags to the linker inside individual object
3396 files. Typically this is used in conjunction with language extensions which
3397 allow source files to explicitly declare the libraries they depend on, and have
3398 these automatically be transmitted to the linker via object files.
3400 These flags are encoded in the IR using metadata in the module flags section,
3401 using the ``Linker Options`` key. The merge behavior for this flag is required
3402 to be ``AppendUnique``, and the value for the key is expected to be a metadata
3403 node which should be a list of other metadata nodes, each of which should be a
3404 list of metadata strings defining linker options.
3406 For example, the following metadata section specifies two separate sets of
3407 linker options, presumably to link against ``libz`` and the ``Cocoa``
3410 !0 = metadata !{ i32 6, metadata !"Linker Options",
3412 metadata !{ metadata !"-lz" },
3413 metadata !{ metadata !"-framework", metadata !"Cocoa" } } }
3414 !llvm.module.flags = !{ !0 }
3416 The metadata encoding as lists of lists of options, as opposed to a collapsed
3417 list of options, is chosen so that the IR encoding can use multiple option
3418 strings to specify e.g., a single library, while still having that specifier be
3419 preserved as an atomic element that can be recognized by a target specific
3420 assembly writer or object file emitter.
3422 Each individual option is required to be either a valid option for the target's
3423 linker, or an option that is reserved by the target specific assembly writer or
3424 object file emitter. No other aspect of these options is defined by the IR.
3426 C type width Module Flags Metadata
3427 ----------------------------------
3429 The ARM backend emits a section into each generated object file describing the
3430 options that it was compiled with (in a compiler-independent way) to prevent
3431 linking incompatible objects, and to allow automatic library selection. Some
3432 of these options are not visible at the IR level, namely wchar_t width and enum
3435 To pass this information to the backend, these options are encoded in module
3436 flags metadata, using the following key-value pairs:
3446 - * 0 --- sizeof(wchar_t) == 4
3447 * 1 --- sizeof(wchar_t) == 2
3450 - * 0 --- Enums are at least as large as an ``int``.
3451 * 1 --- Enums are stored in the smallest integer type which can
3452 represent all of its values.
3454 For example, the following metadata section specifies that the module was
3455 compiled with a ``wchar_t`` width of 4 bytes, and the underlying type of an
3456 enum is the smallest type which can represent all of its values::
3458 !llvm.module.flags = !{!0, !1}
3459 !0 = metadata !{i32 1, metadata !"short_wchar", i32 1}
3460 !1 = metadata !{i32 1, metadata !"short_enum", i32 0}
3462 .. _intrinsicglobalvariables:
3464 Intrinsic Global Variables
3465 ==========================
3467 LLVM has a number of "magic" global variables that contain data that
3468 affect code generation or other IR semantics. These are documented here.
3469 All globals of this sort should have a section specified as
3470 "``llvm.metadata``". This section and all globals that start with
3471 "``llvm.``" are reserved for use by LLVM.
3475 The '``llvm.used``' Global Variable
3476 -----------------------------------
3478 The ``@llvm.used`` global is an array which has
3479 :ref:`appending linkage <linkage_appending>`. This array contains a list of
3480 pointers to named global variables, functions and aliases which may optionally
3481 have a pointer cast formed of bitcast or getelementptr. For example, a legal
3484 .. code-block:: llvm
3489 @llvm.used = appending global [2 x i8*] [
3491 i8* bitcast (i32* @Y to i8*)
3492 ], section "llvm.metadata"
3494 If a symbol appears in the ``@llvm.used`` list, then the compiler, assembler,
3495 and linker are required to treat the symbol as if there is a reference to the
3496 symbol that it cannot see (which is why they have to be named). For example, if
3497 a variable has internal linkage and no references other than that from the
3498 ``@llvm.used`` list, it cannot be deleted. This is commonly used to represent
3499 references from inline asms and other things the compiler cannot "see", and
3500 corresponds to "``attribute((used))``" in GNU C.
3502 On some targets, the code generator must emit a directive to the
3503 assembler or object file to prevent the assembler and linker from
3504 molesting the symbol.
3506 .. _gv_llvmcompilerused:
3508 The '``llvm.compiler.used``' Global Variable
3509 --------------------------------------------
3511 The ``@llvm.compiler.used`` directive is the same as the ``@llvm.used``
3512 directive, except that it only prevents the compiler from touching the
3513 symbol. On targets that support it, this allows an intelligent linker to
3514 optimize references to the symbol without being impeded as it would be
3517 This is a rare construct that should only be used in rare circumstances,
3518 and should not be exposed to source languages.
3520 .. _gv_llvmglobalctors:
3522 The '``llvm.global_ctors``' Global Variable
3523 -------------------------------------------
3525 .. code-block:: llvm
3527 %0 = type { i32, void ()*, i8* }
3528 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor, i8* @data }]
3530 The ``@llvm.global_ctors`` array contains a list of constructor
3531 functions, priorities, and an optional associated global or function.
3532 The functions referenced by this array will be called in ascending order
3533 of priority (i.e. lowest first) when the module is loaded. The order of
3534 functions with the same priority is not defined.
3536 If the third field is present, non-null, and points to a global variable
3537 or function, the initializer function will only run if the associated
3538 data from the current module is not discarded.
3540 .. _llvmglobaldtors:
3542 The '``llvm.global_dtors``' Global Variable
3543 -------------------------------------------
3545 .. code-block:: llvm
3547 %0 = type { i32, void ()*, i8* }
3548 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor, i8* @data }]
3550 The ``@llvm.global_dtors`` array contains a list of destructor
3551 functions, priorities, and an optional associated global or function.
3552 The functions referenced by this array will be called in descending
3553 order of priority (i.e. highest first) when the module is unloaded. The
3554 order of functions with the same priority is not defined.
3556 If the third field is present, non-null, and points to a global variable
3557 or function, the destructor function will only run if the associated
3558 data from the current module is not discarded.
3560 Instruction Reference
3561 =====================
3563 The LLVM instruction set consists of several different classifications
3564 of instructions: :ref:`terminator instructions <terminators>`, :ref:`binary
3565 instructions <binaryops>`, :ref:`bitwise binary
3566 instructions <bitwiseops>`, :ref:`memory instructions <memoryops>`, and
3567 :ref:`other instructions <otherops>`.
3571 Terminator Instructions
3572 -----------------------
3574 As mentioned :ref:`previously <functionstructure>`, every basic block in a
3575 program ends with a "Terminator" instruction, which indicates which
3576 block should be executed after the current block is finished. These
3577 terminator instructions typically yield a '``void``' value: they produce
3578 control flow, not values (the one exception being the
3579 ':ref:`invoke <i_invoke>`' instruction).
3581 The terminator instructions are: ':ref:`ret <i_ret>`',
3582 ':ref:`br <i_br>`', ':ref:`switch <i_switch>`',
3583 ':ref:`indirectbr <i_indirectbr>`', ':ref:`invoke <i_invoke>`',
3584 ':ref:`resume <i_resume>`', and ':ref:`unreachable <i_unreachable>`'.
3588 '``ret``' Instruction
3589 ^^^^^^^^^^^^^^^^^^^^^
3596 ret <type> <value> ; Return a value from a non-void function
3597 ret void ; Return from void function
3602 The '``ret``' instruction is used to return control flow (and optionally
3603 a value) from a function back to the caller.
3605 There are two forms of the '``ret``' instruction: one that returns a
3606 value and then causes control flow, and one that just causes control
3612 The '``ret``' instruction optionally accepts a single argument, the
3613 return value. The type of the return value must be a ':ref:`first
3614 class <t_firstclass>`' type.
3616 A function is not :ref:`well formed <wellformed>` if it it has a non-void
3617 return type and contains a '``ret``' instruction with no return value or
3618 a return value with a type that does not match its type, or if it has a
3619 void return type and contains a '``ret``' instruction with a return
3625 When the '``ret``' instruction is executed, control flow returns back to
3626 the calling function's context. If the caller is a
3627 ":ref:`call <i_call>`" instruction, execution continues at the
3628 instruction after the call. If the caller was an
3629 ":ref:`invoke <i_invoke>`" instruction, execution continues at the
3630 beginning of the "normal" destination block. If the instruction returns
3631 a value, that value shall set the call or invoke instruction's return
3637 .. code-block:: llvm
3639 ret i32 5 ; Return an integer value of 5
3640 ret void ; Return from a void function
3641 ret { i32, i8 } { i32 4, i8 2 } ; Return a struct of values 4 and 2
3645 '``br``' Instruction
3646 ^^^^^^^^^^^^^^^^^^^^
3653 br i1 <cond>, label <iftrue>, label <iffalse>
3654 br label <dest> ; Unconditional branch
3659 The '``br``' instruction is used to cause control flow to transfer to a
3660 different basic block in the current function. There are two forms of
3661 this instruction, corresponding to a conditional branch and an
3662 unconditional branch.
3667 The conditional branch form of the '``br``' instruction takes a single
3668 '``i1``' value and two '``label``' values. The unconditional form of the
3669 '``br``' instruction takes a single '``label``' value as a target.
3674 Upon execution of a conditional '``br``' instruction, the '``i1``'
3675 argument is evaluated. If the value is ``true``, control flows to the
3676 '``iftrue``' ``label`` argument. If "cond" is ``false``, control flows
3677 to the '``iffalse``' ``label`` argument.
3682 .. code-block:: llvm
3685 %cond = icmp eq i32 %a, %b
3686 br i1 %cond, label %IfEqual, label %IfUnequal
3694 '``switch``' Instruction
3695 ^^^^^^^^^^^^^^^^^^^^^^^^
3702 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
3707 The '``switch``' instruction is used to transfer control flow to one of
3708 several different places. It is a generalization of the '``br``'
3709 instruction, allowing a branch to occur to one of many possible
3715 The '``switch``' instruction uses three parameters: an integer
3716 comparison value '``value``', a default '``label``' destination, and an
3717 array of pairs of comparison value constants and '``label``'s. The table
3718 is not allowed to contain duplicate constant entries.
3723 The ``switch`` instruction specifies a table of values and destinations.
3724 When the '``switch``' instruction is executed, this table is searched
3725 for the given value. If the value is found, control flow is transferred
3726 to the corresponding destination; otherwise, control flow is transferred
3727 to the default destination.
3732 Depending on properties of the target machine and the particular
3733 ``switch`` instruction, this instruction may be code generated in
3734 different ways. For example, it could be generated as a series of
3735 chained conditional branches or with a lookup table.
3740 .. code-block:: llvm
3742 ; Emulate a conditional br instruction
3743 %Val = zext i1 %value to i32
3744 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
3746 ; Emulate an unconditional br instruction
3747 switch i32 0, label %dest [ ]
3749 ; Implement a jump table:
3750 switch i32 %val, label %otherwise [ i32 0, label %onzero
3752 i32 2, label %ontwo ]
3756 '``indirectbr``' Instruction
3757 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3764 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
3769 The '``indirectbr``' instruction implements an indirect branch to a
3770 label within the current function, whose address is specified by
3771 "``address``". Address must be derived from a
3772 :ref:`blockaddress <blockaddress>` constant.
3777 The '``address``' argument is the address of the label to jump to. The
3778 rest of the arguments indicate the full set of possible destinations
3779 that the address may point to. Blocks are allowed to occur multiple
3780 times in the destination list, though this isn't particularly useful.
3782 This destination list is required so that dataflow analysis has an
3783 accurate understanding of the CFG.
3788 Control transfers to the block specified in the address argument. All
3789 possible destination blocks must be listed in the label list, otherwise
3790 this instruction has undefined behavior. This implies that jumps to
3791 labels defined in other functions have undefined behavior as well.
3796 This is typically implemented with a jump through a register.
3801 .. code-block:: llvm
3803 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3807 '``invoke``' Instruction
3808 ^^^^^^^^^^^^^^^^^^^^^^^^
3815 <result> = invoke [cconv] [ret attrs] <ptr to function ty> <function ptr val>(<function args>) [fn attrs]
3816 to label <normal label> unwind label <exception label>
3821 The '``invoke``' instruction causes control to transfer to a specified
3822 function, with the possibility of control flow transfer to either the
3823 '``normal``' label or the '``exception``' label. If the callee function
3824 returns with the "``ret``" instruction, control flow will return to the
3825 "normal" label. If the callee (or any indirect callees) returns via the
3826 ":ref:`resume <i_resume>`" instruction or other exception handling
3827 mechanism, control is interrupted and continued at the dynamically
3828 nearest "exception" label.
3830 The '``exception``' label is a `landing
3831 pad <ExceptionHandling.html#overview>`_ for the exception. As such,
3832 '``exception``' label is required to have the
3833 ":ref:`landingpad <i_landingpad>`" instruction, which contains the
3834 information about the behavior of the program after unwinding happens,
3835 as its first non-PHI instruction. The restrictions on the
3836 "``landingpad``" instruction's tightly couples it to the "``invoke``"
3837 instruction, so that the important information contained within the
3838 "``landingpad``" instruction can't be lost through normal code motion.
3843 This instruction requires several arguments:
3845 #. The optional "cconv" marker indicates which :ref:`calling
3846 convention <callingconv>` the call should use. If none is
3847 specified, the call defaults to using C calling conventions.
3848 #. The optional :ref:`Parameter Attributes <paramattrs>` list for return
3849 values. Only '``zeroext``', '``signext``', and '``inreg``' attributes
3851 #. '``ptr to function ty``': shall be the signature of the pointer to
3852 function value being invoked. In most cases, this is a direct
3853 function invocation, but indirect ``invoke``'s are just as possible,
3854 branching off an arbitrary pointer to function value.
3855 #. '``function ptr val``': An LLVM value containing a pointer to a
3856 function to be invoked.
3857 #. '``function args``': argument list whose types match the function
3858 signature argument types and parameter attributes. All arguments must
3859 be of :ref:`first class <t_firstclass>` type. If the function signature
3860 indicates the function accepts a variable number of arguments, the
3861 extra arguments can be specified.
3862 #. '``normal label``': the label reached when the called function
3863 executes a '``ret``' instruction.
3864 #. '``exception label``': the label reached when a callee returns via
3865 the :ref:`resume <i_resume>` instruction or other exception handling
3867 #. The optional :ref:`function attributes <fnattrs>` list. Only
3868 '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``'
3869 attributes are valid here.
3874 This instruction is designed to operate as a standard '``call``'
3875 instruction in most regards. The primary difference is that it
3876 establishes an association with a label, which is used by the runtime
3877 library to unwind the stack.
3879 This instruction is used in languages with destructors to ensure that
3880 proper cleanup is performed in the case of either a ``longjmp`` or a
3881 thrown exception. Additionally, this is important for implementation of
3882 '``catch``' clauses in high-level languages that support them.
3884 For the purposes of the SSA form, the definition of the value returned
3885 by the '``invoke``' instruction is deemed to occur on the edge from the
3886 current block to the "normal" label. If the callee unwinds then no
3887 return value is available.
3892 .. code-block:: llvm
3894 %retval = invoke i32 @Test(i32 15) to label %Continue
3895 unwind label %TestCleanup ; i32:retval set
3896 %retval = invoke coldcc i32 %Testfnptr(i32 15) to label %Continue
3897 unwind label %TestCleanup ; i32:retval set
3901 '``resume``' Instruction
3902 ^^^^^^^^^^^^^^^^^^^^^^^^
3909 resume <type> <value>
3914 The '``resume``' instruction is a terminator instruction that has no
3920 The '``resume``' instruction requires one argument, which must have the
3921 same type as the result of any '``landingpad``' instruction in the same
3927 The '``resume``' instruction resumes propagation of an existing
3928 (in-flight) exception whose unwinding was interrupted with a
3929 :ref:`landingpad <i_landingpad>` instruction.
3934 .. code-block:: llvm
3936 resume { i8*, i32 } %exn
3940 '``unreachable``' Instruction
3941 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3953 The '``unreachable``' instruction has no defined semantics. This
3954 instruction is used to inform the optimizer that a particular portion of
3955 the code is not reachable. This can be used to indicate that the code
3956 after a no-return function cannot be reached, and other facts.
3961 The '``unreachable``' instruction has no defined semantics.
3968 Binary operators are used to do most of the computation in a program.
3969 They require two operands of the same type, execute an operation on
3970 them, and produce a single value. The operands might represent multiple
3971 data, as is the case with the :ref:`vector <t_vector>` data type. The
3972 result value has the same type as its operands.
3974 There are several different binary operators:
3978 '``add``' Instruction
3979 ^^^^^^^^^^^^^^^^^^^^^
3986 <result> = add <ty> <op1>, <op2> ; yields ty:result
3987 <result> = add nuw <ty> <op1>, <op2> ; yields ty:result
3988 <result> = add nsw <ty> <op1>, <op2> ; yields ty:result
3989 <result> = add nuw nsw <ty> <op1>, <op2> ; yields ty:result
3994 The '``add``' instruction returns the sum of its two operands.
3999 The two arguments to the '``add``' instruction must be
4000 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4001 arguments must have identical types.
4006 The value produced is the integer sum of the two operands.
4008 If the sum has unsigned overflow, the result returned is the
4009 mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of
4012 Because LLVM integers use a two's complement representation, this
4013 instruction is appropriate for both signed and unsigned integers.
4015 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
4016 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
4017 result value of the ``add`` is a :ref:`poison value <poisonvalues>` if
4018 unsigned and/or signed overflow, respectively, occurs.
4023 .. code-block:: llvm
4025 <result> = add i32 4, %var ; yields i32:result = 4 + %var
4029 '``fadd``' Instruction
4030 ^^^^^^^^^^^^^^^^^^^^^^
4037 <result> = fadd [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
4042 The '``fadd``' instruction returns the sum of its two operands.
4047 The two arguments to the '``fadd``' instruction must be :ref:`floating
4048 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
4049 Both arguments must have identical types.
4054 The value produced is the floating point sum of the two operands. This
4055 instruction can also take any number of :ref:`fast-math flags <fastmath>`,
4056 which are optimization hints to enable otherwise unsafe floating point
4062 .. code-block:: llvm
4064 <result> = fadd float 4.0, %var ; yields float:result = 4.0 + %var
4066 '``sub``' Instruction
4067 ^^^^^^^^^^^^^^^^^^^^^
4074 <result> = sub <ty> <op1>, <op2> ; yields ty:result
4075 <result> = sub nuw <ty> <op1>, <op2> ; yields ty:result
4076 <result> = sub nsw <ty> <op1>, <op2> ; yields ty:result
4077 <result> = sub nuw nsw <ty> <op1>, <op2> ; yields ty:result
4082 The '``sub``' instruction returns the difference of its two operands.
4084 Note that the '``sub``' instruction is used to represent the '``neg``'
4085 instruction present in most other intermediate representations.
4090 The two arguments to the '``sub``' instruction must be
4091 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4092 arguments must have identical types.
4097 The value produced is the integer difference of the two operands.
4099 If the difference has unsigned overflow, the result returned is the
4100 mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of
4103 Because LLVM integers use a two's complement representation, this
4104 instruction is appropriate for both signed and unsigned integers.
4106 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
4107 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
4108 result value of the ``sub`` is a :ref:`poison value <poisonvalues>` if
4109 unsigned and/or signed overflow, respectively, occurs.
4114 .. code-block:: llvm
4116 <result> = sub i32 4, %var ; yields i32:result = 4 - %var
4117 <result> = sub i32 0, %val ; yields i32:result = -%var
4121 '``fsub``' Instruction
4122 ^^^^^^^^^^^^^^^^^^^^^^
4129 <result> = fsub [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
4134 The '``fsub``' instruction returns the difference of its two operands.
4136 Note that the '``fsub``' instruction is used to represent the '``fneg``'
4137 instruction present in most other intermediate representations.
4142 The two arguments to the '``fsub``' instruction must be :ref:`floating
4143 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
4144 Both arguments must have identical types.
4149 The value produced is the floating point difference of the two operands.
4150 This instruction can also take any number of :ref:`fast-math
4151 flags <fastmath>`, which are optimization hints to enable otherwise
4152 unsafe floating point optimizations:
4157 .. code-block:: llvm
4159 <result> = fsub float 4.0, %var ; yields float:result = 4.0 - %var
4160 <result> = fsub float -0.0, %val ; yields float:result = -%var
4162 '``mul``' Instruction
4163 ^^^^^^^^^^^^^^^^^^^^^
4170 <result> = mul <ty> <op1>, <op2> ; yields ty:result
4171 <result> = mul nuw <ty> <op1>, <op2> ; yields ty:result
4172 <result> = mul nsw <ty> <op1>, <op2> ; yields ty:result
4173 <result> = mul nuw nsw <ty> <op1>, <op2> ; yields ty:result
4178 The '``mul``' instruction returns the product of its two operands.
4183 The two arguments to the '``mul``' instruction must be
4184 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4185 arguments must have identical types.
4190 The value produced is the integer product of the two operands.
4192 If the result of the multiplication has unsigned overflow, the result
4193 returned is the mathematical result modulo 2\ :sup:`n`\ , where n is the
4194 bit width of the result.
4196 Because LLVM integers use a two's complement representation, and the
4197 result is the same width as the operands, this instruction returns the
4198 correct result for both signed and unsigned integers. If a full product
4199 (e.g. ``i32`` * ``i32`` -> ``i64``) is needed, the operands should be
4200 sign-extended or zero-extended as appropriate to the width of the full
4203 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
4204 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
4205 result value of the ``mul`` is a :ref:`poison value <poisonvalues>` if
4206 unsigned and/or signed overflow, respectively, occurs.
4211 .. code-block:: llvm
4213 <result> = mul i32 4, %var ; yields i32:result = 4 * %var
4217 '``fmul``' Instruction
4218 ^^^^^^^^^^^^^^^^^^^^^^
4225 <result> = fmul [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
4230 The '``fmul``' instruction returns the product of its two operands.
4235 The two arguments to the '``fmul``' instruction must be :ref:`floating
4236 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
4237 Both arguments must have identical types.
4242 The value produced is the floating point product of the two operands.
4243 This instruction can also take any number of :ref:`fast-math
4244 flags <fastmath>`, which are optimization hints to enable otherwise
4245 unsafe floating point optimizations:
4250 .. code-block:: llvm
4252 <result> = fmul float 4.0, %var ; yields float:result = 4.0 * %var
4254 '``udiv``' Instruction
4255 ^^^^^^^^^^^^^^^^^^^^^^
4262 <result> = udiv <ty> <op1>, <op2> ; yields ty:result
4263 <result> = udiv exact <ty> <op1>, <op2> ; yields ty:result
4268 The '``udiv``' instruction returns the quotient of its two operands.
4273 The two arguments to the '``udiv``' instruction must be
4274 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4275 arguments must have identical types.
4280 The value produced is the unsigned integer quotient of the two operands.
4282 Note that unsigned integer division and signed integer division are
4283 distinct operations; for signed integer division, use '``sdiv``'.
4285 Division by zero leads to undefined behavior.
4287 If the ``exact`` keyword is present, the result value of the ``udiv`` is
4288 a :ref:`poison value <poisonvalues>` if %op1 is not a multiple of %op2 (as
4289 such, "((a udiv exact b) mul b) == a").
4294 .. code-block:: llvm
4296 <result> = udiv i32 4, %var ; yields i32:result = 4 / %var
4298 '``sdiv``' Instruction
4299 ^^^^^^^^^^^^^^^^^^^^^^
4306 <result> = sdiv <ty> <op1>, <op2> ; yields ty:result
4307 <result> = sdiv exact <ty> <op1>, <op2> ; yields ty:result
4312 The '``sdiv``' instruction returns the quotient of its two operands.
4317 The two arguments to the '``sdiv``' instruction must be
4318 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4319 arguments must have identical types.
4324 The value produced is the signed integer quotient of the two operands
4325 rounded towards zero.
4327 Note that signed integer division and unsigned integer division are
4328 distinct operations; for unsigned integer division, use '``udiv``'.
4330 Division by zero leads to undefined behavior. Overflow also leads to
4331 undefined behavior; this is a rare case, but can occur, for example, by
4332 doing a 32-bit division of -2147483648 by -1.
4334 If the ``exact`` keyword is present, the result value of the ``sdiv`` is
4335 a :ref:`poison value <poisonvalues>` if the result would be rounded.
4340 .. code-block:: llvm
4342 <result> = sdiv i32 4, %var ; yields i32:result = 4 / %var
4346 '``fdiv``' Instruction
4347 ^^^^^^^^^^^^^^^^^^^^^^
4354 <result> = fdiv [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
4359 The '``fdiv``' instruction returns the quotient of its two operands.
4364 The two arguments to the '``fdiv``' instruction must be :ref:`floating
4365 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
4366 Both arguments must have identical types.
4371 The value produced is the floating point quotient of the two operands.
4372 This instruction can also take any number of :ref:`fast-math
4373 flags <fastmath>`, which are optimization hints to enable otherwise
4374 unsafe floating point optimizations:
4379 .. code-block:: llvm
4381 <result> = fdiv float 4.0, %var ; yields float:result = 4.0 / %var
4383 '``urem``' Instruction
4384 ^^^^^^^^^^^^^^^^^^^^^^
4391 <result> = urem <ty> <op1>, <op2> ; yields ty:result
4396 The '``urem``' instruction returns the remainder from the unsigned
4397 division of its two arguments.
4402 The two arguments to the '``urem``' instruction must be
4403 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4404 arguments must have identical types.
4409 This instruction returns the unsigned integer *remainder* of a division.
4410 This instruction always performs an unsigned division to get the
4413 Note that unsigned integer remainder and signed integer remainder are
4414 distinct operations; for signed integer remainder, use '``srem``'.
4416 Taking the remainder of a division by zero leads to undefined behavior.
4421 .. code-block:: llvm
4423 <result> = urem i32 4, %var ; yields i32:result = 4 % %var
4425 '``srem``' Instruction
4426 ^^^^^^^^^^^^^^^^^^^^^^
4433 <result> = srem <ty> <op1>, <op2> ; yields ty:result
4438 The '``srem``' instruction returns the remainder from the signed
4439 division of its two operands. This instruction can also take
4440 :ref:`vector <t_vector>` versions of the values in which case the elements
4446 The two arguments to the '``srem``' instruction must be
4447 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4448 arguments must have identical types.
4453 This instruction returns the *remainder* of a division (where the result
4454 is either zero or has the same sign as the dividend, ``op1``), not the
4455 *modulo* operator (where the result is either zero or has the same sign
4456 as the divisor, ``op2``) of a value. For more information about the
4457 difference, see `The Math
4458 Forum <http://mathforum.org/dr.math/problems/anne.4.28.99.html>`_. For a
4459 table of how this is implemented in various languages, please see
4461 operation <http://en.wikipedia.org/wiki/Modulo_operation>`_.
4463 Note that signed integer remainder and unsigned integer remainder are
4464 distinct operations; for unsigned integer remainder, use '``urem``'.
4466 Taking the remainder of a division by zero leads to undefined behavior.
4467 Overflow also leads to undefined behavior; this is a rare case, but can
4468 occur, for example, by taking the remainder of a 32-bit division of
4469 -2147483648 by -1. (The remainder doesn't actually overflow, but this
4470 rule lets srem be implemented using instructions that return both the
4471 result of the division and the remainder.)
4476 .. code-block:: llvm
4478 <result> = srem i32 4, %var ; yields i32:result = 4 % %var
4482 '``frem``' Instruction
4483 ^^^^^^^^^^^^^^^^^^^^^^
4490 <result> = frem [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
4495 The '``frem``' instruction returns the remainder from the division of
4501 The two arguments to the '``frem``' instruction must be :ref:`floating
4502 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
4503 Both arguments must have identical types.
4508 This instruction returns the *remainder* of a division. The remainder
4509 has the same sign as the dividend. This instruction can also take any
4510 number of :ref:`fast-math flags <fastmath>`, which are optimization hints
4511 to enable otherwise unsafe floating point optimizations:
4516 .. code-block:: llvm
4518 <result> = frem float 4.0, %var ; yields float:result = 4.0 % %var
4522 Bitwise Binary Operations
4523 -------------------------
4525 Bitwise binary operators are used to do various forms of bit-twiddling
4526 in a program. They are generally very efficient instructions and can
4527 commonly be strength reduced from other instructions. They require two
4528 operands of the same type, execute an operation on them, and produce a
4529 single value. The resulting value is the same type as its operands.
4531 '``shl``' Instruction
4532 ^^^^^^^^^^^^^^^^^^^^^
4539 <result> = shl <ty> <op1>, <op2> ; yields ty:result
4540 <result> = shl nuw <ty> <op1>, <op2> ; yields ty:result
4541 <result> = shl nsw <ty> <op1>, <op2> ; yields ty:result
4542 <result> = shl nuw nsw <ty> <op1>, <op2> ; yields ty:result
4547 The '``shl``' instruction returns the first operand shifted to the left
4548 a specified number of bits.
4553 Both arguments to the '``shl``' instruction must be the same
4554 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
4555 '``op2``' is treated as an unsigned value.
4560 The value produced is ``op1`` \* 2\ :sup:`op2` mod 2\ :sup:`n`,
4561 where ``n`` is the width of the result. If ``op2`` is (statically or
4562 dynamically) negative or equal to or larger than the number of bits in
4563 ``op1``, the result is undefined. If the arguments are vectors, each
4564 vector element of ``op1`` is shifted by the corresponding shift amount
4567 If the ``nuw`` keyword is present, then the shift produces a :ref:`poison
4568 value <poisonvalues>` if it shifts out any non-zero bits. If the
4569 ``nsw`` keyword is present, then the shift produces a :ref:`poison
4570 value <poisonvalues>` if it shifts out any bits that disagree with the
4571 resultant sign bit. As such, NUW/NSW have the same semantics as they
4572 would if the shift were expressed as a mul instruction with the same
4573 nsw/nuw bits in (mul %op1, (shl 1, %op2)).
4578 .. code-block:: llvm
4580 <result> = shl i32 4, %var ; yields i32: 4 << %var
4581 <result> = shl i32 4, 2 ; yields i32: 16
4582 <result> = shl i32 1, 10 ; yields i32: 1024
4583 <result> = shl i32 1, 32 ; undefined
4584 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 2, i32 4>
4586 '``lshr``' Instruction
4587 ^^^^^^^^^^^^^^^^^^^^^^
4594 <result> = lshr <ty> <op1>, <op2> ; yields ty:result
4595 <result> = lshr exact <ty> <op1>, <op2> ; yields ty:result
4600 The '``lshr``' instruction (logical shift right) returns the first
4601 operand shifted to the right a specified number of bits with zero fill.
4606 Both arguments to the '``lshr``' instruction must be the same
4607 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
4608 '``op2``' is treated as an unsigned value.
4613 This instruction always performs a logical shift right operation. The
4614 most significant bits of the result will be filled with zero bits after
4615 the shift. If ``op2`` is (statically or dynamically) equal to or larger
4616 than the number of bits in ``op1``, the result is undefined. If the
4617 arguments are vectors, each vector element of ``op1`` is shifted by the
4618 corresponding shift amount in ``op2``.
4620 If the ``exact`` keyword is present, the result value of the ``lshr`` is
4621 a :ref:`poison value <poisonvalues>` if any of the bits shifted out are
4627 .. code-block:: llvm
4629 <result> = lshr i32 4, 1 ; yields i32:result = 2
4630 <result> = lshr i32 4, 2 ; yields i32:result = 1
4631 <result> = lshr i8 4, 3 ; yields i8:result = 0
4632 <result> = lshr i8 -2, 1 ; yields i8:result = 0x7F
4633 <result> = lshr i32 1, 32 ; undefined
4634 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1>
4636 '``ashr``' Instruction
4637 ^^^^^^^^^^^^^^^^^^^^^^
4644 <result> = ashr <ty> <op1>, <op2> ; yields ty:result
4645 <result> = ashr exact <ty> <op1>, <op2> ; yields ty:result
4650 The '``ashr``' instruction (arithmetic shift right) returns the first
4651 operand shifted to the right a specified number of bits with sign
4657 Both arguments to the '``ashr``' instruction must be the same
4658 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
4659 '``op2``' is treated as an unsigned value.
4664 This instruction always performs an arithmetic shift right operation,
4665 The most significant bits of the result will be filled with the sign bit
4666 of ``op1``. If ``op2`` is (statically or dynamically) equal to or larger
4667 than the number of bits in ``op1``, the result is undefined. If the
4668 arguments are vectors, each vector element of ``op1`` is shifted by the
4669 corresponding shift amount in ``op2``.
4671 If the ``exact`` keyword is present, the result value of the ``ashr`` is
4672 a :ref:`poison value <poisonvalues>` if any of the bits shifted out are
4678 .. code-block:: llvm
4680 <result> = ashr i32 4, 1 ; yields i32:result = 2
4681 <result> = ashr i32 4, 2 ; yields i32:result = 1
4682 <result> = ashr i8 4, 3 ; yields i8:result = 0
4683 <result> = ashr i8 -2, 1 ; yields i8:result = -1
4684 <result> = ashr i32 1, 32 ; undefined
4685 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> ; yields: result=<2 x i32> < i32 -1, i32 0>
4687 '``and``' Instruction
4688 ^^^^^^^^^^^^^^^^^^^^^
4695 <result> = and <ty> <op1>, <op2> ; yields ty:result
4700 The '``and``' instruction returns the bitwise logical and of its two
4706 The two arguments to the '``and``' instruction must be
4707 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4708 arguments must have identical types.
4713 The truth table used for the '``and``' instruction is:
4730 .. code-block:: llvm
4732 <result> = and i32 4, %var ; yields i32:result = 4 & %var
4733 <result> = and i32 15, 40 ; yields i32:result = 8
4734 <result> = and i32 4, 8 ; yields i32:result = 0
4736 '``or``' Instruction
4737 ^^^^^^^^^^^^^^^^^^^^
4744 <result> = or <ty> <op1>, <op2> ; yields ty:result
4749 The '``or``' instruction returns the bitwise logical inclusive or of its
4755 The two arguments to the '``or``' instruction must be
4756 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4757 arguments must have identical types.
4762 The truth table used for the '``or``' instruction is:
4781 <result> = or i32 4, %var ; yields i32:result = 4 | %var
4782 <result> = or i32 15, 40 ; yields i32:result = 47
4783 <result> = or i32 4, 8 ; yields i32:result = 12
4785 '``xor``' Instruction
4786 ^^^^^^^^^^^^^^^^^^^^^
4793 <result> = xor <ty> <op1>, <op2> ; yields ty:result
4798 The '``xor``' instruction returns the bitwise logical exclusive or of
4799 its two operands. The ``xor`` is used to implement the "one's
4800 complement" operation, which is the "~" operator in C.
4805 The two arguments to the '``xor``' instruction must be
4806 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4807 arguments must have identical types.
4812 The truth table used for the '``xor``' instruction is:
4829 .. code-block:: llvm
4831 <result> = xor i32 4, %var ; yields i32:result = 4 ^ %var
4832 <result> = xor i32 15, 40 ; yields i32:result = 39
4833 <result> = xor i32 4, 8 ; yields i32:result = 12
4834 <result> = xor i32 %V, -1 ; yields i32:result = ~%V
4839 LLVM supports several instructions to represent vector operations in a
4840 target-independent manner. These instructions cover the element-access
4841 and vector-specific operations needed to process vectors effectively.
4842 While LLVM does directly support these vector operations, many
4843 sophisticated algorithms will want to use target-specific intrinsics to
4844 take full advantage of a specific target.
4846 .. _i_extractelement:
4848 '``extractelement``' Instruction
4849 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4856 <result> = extractelement <n x <ty>> <val>, <ty2> <idx> ; yields <ty>
4861 The '``extractelement``' instruction extracts a single scalar element
4862 from a vector at a specified index.
4867 The first operand of an '``extractelement``' instruction is a value of
4868 :ref:`vector <t_vector>` type. The second operand is an index indicating
4869 the position from which to extract the element. The index may be a
4870 variable of any integer type.
4875 The result is a scalar of the same type as the element type of ``val``.
4876 Its value is the value at position ``idx`` of ``val``. If ``idx``
4877 exceeds the length of ``val``, the results are undefined.
4882 .. code-block:: llvm
4884 <result> = extractelement <4 x i32> %vec, i32 0 ; yields i32
4886 .. _i_insertelement:
4888 '``insertelement``' Instruction
4889 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4896 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, <ty2> <idx> ; yields <n x <ty>>
4901 The '``insertelement``' instruction inserts a scalar element into a
4902 vector at a specified index.
4907 The first operand of an '``insertelement``' instruction is a value of
4908 :ref:`vector <t_vector>` type. The second operand is a scalar value whose
4909 type must equal the element type of the first operand. The third operand
4910 is an index indicating the position at which to insert the value. The
4911 index may be a variable of any integer type.
4916 The result is a vector of the same type as ``val``. Its element values
4917 are those of ``val`` except at position ``idx``, where it gets the value
4918 ``elt``. If ``idx`` exceeds the length of ``val``, the results are
4924 .. code-block:: llvm
4926 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 ; yields <4 x i32>
4928 .. _i_shufflevector:
4930 '``shufflevector``' Instruction
4931 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4938 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> ; yields <m x <ty>>
4943 The '``shufflevector``' instruction constructs a permutation of elements
4944 from two input vectors, returning a vector with the same element type as
4945 the input and length that is the same as the shuffle mask.
4950 The first two operands of a '``shufflevector``' instruction are vectors
4951 with the same type. The third argument is a shuffle mask whose element
4952 type is always 'i32'. The result of the instruction is a vector whose
4953 length is the same as the shuffle mask and whose element type is the
4954 same as the element type of the first two operands.
4956 The shuffle mask operand is required to be a constant vector with either
4957 constant integer or undef values.
4962 The elements of the two input vectors are numbered from left to right
4963 across both of the vectors. The shuffle mask operand specifies, for each
4964 element of the result vector, which element of the two input vectors the
4965 result element gets. The element selector may be undef (meaning "don't
4966 care") and the second operand may be undef if performing a shuffle from
4972 .. code-block:: llvm
4974 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4975 <4 x i32> <i32 0, i32 4, i32 1, i32 5> ; yields <4 x i32>
4976 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4977 <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32> - Identity shuffle.
4978 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4979 <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32>
4980 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4981 <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 > ; yields <8 x i32>
4983 Aggregate Operations
4984 --------------------
4986 LLVM supports several instructions for working with
4987 :ref:`aggregate <t_aggregate>` values.
4991 '``extractvalue``' Instruction
4992 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4999 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
5004 The '``extractvalue``' instruction extracts the value of a member field
5005 from an :ref:`aggregate <t_aggregate>` value.
5010 The first operand of an '``extractvalue``' instruction is a value of
5011 :ref:`struct <t_struct>` or :ref:`array <t_array>` type. The operands are
5012 constant indices to specify which value to extract in a similar manner
5013 as indices in a '``getelementptr``' instruction.
5015 The major differences to ``getelementptr`` indexing are:
5017 - Since the value being indexed is not a pointer, the first index is
5018 omitted and assumed to be zero.
5019 - At least one index must be specified.
5020 - Not only struct indices but also array indices must be in bounds.
5025 The result is the value at the position in the aggregate specified by
5031 .. code-block:: llvm
5033 <result> = extractvalue {i32, float} %agg, 0 ; yields i32
5037 '``insertvalue``' Instruction
5038 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5045 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* ; yields <aggregate type>
5050 The '``insertvalue``' instruction inserts a value into a member field in
5051 an :ref:`aggregate <t_aggregate>` value.
5056 The first operand of an '``insertvalue``' instruction is a value of
5057 :ref:`struct <t_struct>` or :ref:`array <t_array>` type. The second operand is
5058 a first-class value to insert. The following operands are constant
5059 indices indicating the position at which to insert the value in a
5060 similar manner as indices in a '``extractvalue``' instruction. The value
5061 to insert must have the same type as the value identified by the
5067 The result is an aggregate of the same type as ``val``. Its value is
5068 that of ``val`` except that the value at the position specified by the
5069 indices is that of ``elt``.
5074 .. code-block:: llvm
5076 %agg1 = insertvalue {i32, float} undef, i32 1, 0 ; yields {i32 1, float undef}
5077 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 ; yields {i32 1, float %val}
5078 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 ; yields {i32 1, float %val}
5082 Memory Access and Addressing Operations
5083 ---------------------------------------
5085 A key design point of an SSA-based representation is how it represents
5086 memory. In LLVM, no memory locations are in SSA form, which makes things
5087 very simple. This section describes how to read, write, and allocate
5092 '``alloca``' Instruction
5093 ^^^^^^^^^^^^^^^^^^^^^^^^
5100 <result> = alloca [inalloca] <type> [, <ty> <NumElements>] [, align <alignment>] ; yields type*:result
5105 The '``alloca``' instruction allocates memory on the stack frame of the
5106 currently executing function, to be automatically released when this
5107 function returns to its caller. The object is always allocated in the
5108 generic address space (address space zero).
5113 The '``alloca``' instruction allocates ``sizeof(<type>)*NumElements``
5114 bytes of memory on the runtime stack, returning a pointer of the
5115 appropriate type to the program. If "NumElements" is specified, it is
5116 the number of elements allocated, otherwise "NumElements" is defaulted
5117 to be one. If a constant alignment is specified, the value result of the
5118 allocation is guaranteed to be aligned to at least that boundary. The
5119 alignment may not be greater than ``1 << 29``. If not specified, or if
5120 zero, the target can choose to align the allocation on any convenient
5121 boundary compatible with the type.
5123 '``type``' may be any sized type.
5128 Memory is allocated; a pointer is returned. The operation is undefined
5129 if there is insufficient stack space for the allocation. '``alloca``'d
5130 memory is automatically released when the function returns. The
5131 '``alloca``' instruction is commonly used to represent automatic
5132 variables that must have an address available. When the function returns
5133 (either with the ``ret`` or ``resume`` instructions), the memory is
5134 reclaimed. Allocating zero bytes is legal, but the result is undefined.
5135 The order in which memory is allocated (ie., which way the stack grows)
5141 .. code-block:: llvm
5143 %ptr = alloca i32 ; yields i32*:ptr
5144 %ptr = alloca i32, i32 4 ; yields i32*:ptr
5145 %ptr = alloca i32, i32 4, align 1024 ; yields i32*:ptr
5146 %ptr = alloca i32, align 1024 ; yields i32*:ptr
5150 '``load``' Instruction
5151 ^^^^^^^^^^^^^^^^^^^^^^
5158 <result> = load [volatile] <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>][, !invariant.load !<index>]
5159 <result> = load atomic [volatile] <ty>* <pointer> [singlethread] <ordering>, align <alignment>
5160 !<index> = !{ i32 1 }
5165 The '``load``' instruction is used to read from memory.
5170 The argument to the ``load`` instruction specifies the memory address
5171 from which to load. The pointer must point to a :ref:`first
5172 class <t_firstclass>` type. If the ``load`` is marked as ``volatile``,
5173 then the optimizer is not allowed to modify the number or order of
5174 execution of this ``load`` with other :ref:`volatile
5175 operations <volatile>`.
5177 If the ``load`` is marked as ``atomic``, it takes an extra
5178 :ref:`ordering <ordering>` and optional ``singlethread`` argument. The
5179 ``release`` and ``acq_rel`` orderings are not valid on ``load``
5180 instructions. Atomic loads produce :ref:`defined <memmodel>` results
5181 when they may see multiple atomic stores. The type of the pointee must
5182 be an integer type whose bit width is a power of two greater than or
5183 equal to eight and less than or equal to a target-specific size limit.
5184 ``align`` must be explicitly specified on atomic loads, and the load has
5185 undefined behavior if the alignment is not set to a value which is at
5186 least the size in bytes of the pointee. ``!nontemporal`` does not have
5187 any defined semantics for atomic loads.
5189 The optional constant ``align`` argument specifies the alignment of the
5190 operation (that is, the alignment of the memory address). A value of 0
5191 or an omitted ``align`` argument means that the operation has the ABI
5192 alignment for the target. It is the responsibility of the code emitter
5193 to ensure that the alignment information is correct. Overestimating the
5194 alignment results in undefined behavior. Underestimating the alignment
5195 may produce less efficient code. An alignment of 1 is always safe. The
5196 maximum possible alignment is ``1 << 29``.
5198 The optional ``!nontemporal`` metadata must reference a single
5199 metadata name ``<index>`` corresponding to a metadata node with one
5200 ``i32`` entry of value 1. The existence of the ``!nontemporal``
5201 metadata on the instruction tells the optimizer and code generator
5202 that this load is not expected to be reused in the cache. The code
5203 generator may select special instructions to save cache bandwidth, such
5204 as the ``MOVNT`` instruction on x86.
5206 The optional ``!invariant.load`` metadata must reference a single
5207 metadata name ``<index>`` corresponding to a metadata node with no
5208 entries. The existence of the ``!invariant.load`` metadata on the
5209 instruction tells the optimizer and code generator that this load
5210 address points to memory which does not change value during program
5211 execution. The optimizer may then move this load around, for example, by
5212 hoisting it out of loops using loop invariant code motion.
5217 The location of memory pointed to is loaded. If the value being loaded
5218 is of scalar type then the number of bytes read does not exceed the
5219 minimum number of bytes needed to hold all bits of the type. For
5220 example, loading an ``i24`` reads at most three bytes. When loading a
5221 value of a type like ``i20`` with a size that is not an integral number
5222 of bytes, the result is undefined if the value was not originally
5223 written using a store of the same type.
5228 .. code-block:: llvm
5230 %ptr = alloca i32 ; yields i32*:ptr
5231 store i32 3, i32* %ptr ; yields void
5232 %val = load i32* %ptr ; yields i32:val = i32 3
5236 '``store``' Instruction
5237 ^^^^^^^^^^^^^^^^^^^^^^^
5244 store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] ; yields void
5245 store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> ; yields void
5250 The '``store``' instruction is used to write to memory.
5255 There are two arguments to the ``store`` instruction: a value to store
5256 and an address at which to store it. The type of the ``<pointer>``
5257 operand must be a pointer to the :ref:`first class <t_firstclass>` type of
5258 the ``<value>`` operand. If the ``store`` is marked as ``volatile``,
5259 then the optimizer is not allowed to modify the number or order of
5260 execution of this ``store`` with other :ref:`volatile
5261 operations <volatile>`.
5263 If the ``store`` is marked as ``atomic``, it takes an extra
5264 :ref:`ordering <ordering>` and optional ``singlethread`` argument. The
5265 ``acquire`` and ``acq_rel`` orderings aren't valid on ``store``
5266 instructions. Atomic loads produce :ref:`defined <memmodel>` results
5267 when they may see multiple atomic stores. The type of the pointee must
5268 be an integer type whose bit width is a power of two greater than or
5269 equal to eight and less than or equal to a target-specific size limit.
5270 ``align`` must be explicitly specified on atomic stores, and the store
5271 has undefined behavior if the alignment is not set to a value which is
5272 at least the size in bytes of the pointee. ``!nontemporal`` does not
5273 have any defined semantics for atomic stores.
5275 The optional constant ``align`` argument specifies the alignment of the
5276 operation (that is, the alignment of the memory address). A value of 0
5277 or an omitted ``align`` argument means that the operation has the ABI
5278 alignment for the target. It is the responsibility of the code emitter
5279 to ensure that the alignment information is correct. Overestimating the
5280 alignment results in undefined behavior. Underestimating the
5281 alignment may produce less efficient code. An alignment of 1 is always
5282 safe. The maximum possible alignment is ``1 << 29``.
5284 The optional ``!nontemporal`` metadata must reference a single metadata
5285 name ``<index>`` corresponding to a metadata node with one ``i32`` entry of
5286 value 1. The existence of the ``!nontemporal`` metadata on the instruction
5287 tells the optimizer and code generator that this load is not expected to
5288 be reused in the cache. The code generator may select special
5289 instructions to save cache bandwidth, such as the MOVNT instruction on
5295 The contents of memory are updated to contain ``<value>`` at the
5296 location specified by the ``<pointer>`` operand. If ``<value>`` is
5297 of scalar type then the number of bytes written does not exceed the
5298 minimum number of bytes needed to hold all bits of the type. For
5299 example, storing an ``i24`` writes at most three bytes. When writing a
5300 value of a type like ``i20`` with a size that is not an integral number
5301 of bytes, it is unspecified what happens to the extra bits that do not
5302 belong to the type, but they will typically be overwritten.
5307 .. code-block:: llvm
5309 %ptr = alloca i32 ; yields i32*:ptr
5310 store i32 3, i32* %ptr ; yields void
5311 %val = load i32* %ptr ; yields i32:val = i32 3
5315 '``fence``' Instruction
5316 ^^^^^^^^^^^^^^^^^^^^^^^
5323 fence [singlethread] <ordering> ; yields void
5328 The '``fence``' instruction is used to introduce happens-before edges
5334 '``fence``' instructions take an :ref:`ordering <ordering>` argument which
5335 defines what *synchronizes-with* edges they add. They can only be given
5336 ``acquire``, ``release``, ``acq_rel``, and ``seq_cst`` orderings.
5341 A fence A which has (at least) ``release`` ordering semantics
5342 *synchronizes with* a fence B with (at least) ``acquire`` ordering
5343 semantics if and only if there exist atomic operations X and Y, both
5344 operating on some atomic object M, such that A is sequenced before X, X
5345 modifies M (either directly or through some side effect of a sequence
5346 headed by X), Y is sequenced before B, and Y observes M. This provides a
5347 *happens-before* dependency between A and B. Rather than an explicit
5348 ``fence``, one (but not both) of the atomic operations X or Y might
5349 provide a ``release`` or ``acquire`` (resp.) ordering constraint and
5350 still *synchronize-with* the explicit ``fence`` and establish the
5351 *happens-before* edge.
5353 A ``fence`` which has ``seq_cst`` ordering, in addition to having both
5354 ``acquire`` and ``release`` semantics specified above, participates in
5355 the global program order of other ``seq_cst`` operations and/or fences.
5357 The optional ":ref:`singlethread <singlethread>`" argument specifies
5358 that the fence only synchronizes with other fences in the same thread.
5359 (This is useful for interacting with signal handlers.)
5364 .. code-block:: llvm
5366 fence acquire ; yields void
5367 fence singlethread seq_cst ; yields void
5371 '``cmpxchg``' Instruction
5372 ^^^^^^^^^^^^^^^^^^^^^^^^^
5379 cmpxchg [weak] [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <success ordering> <failure ordering> ; yields { ty, i1 }
5384 The '``cmpxchg``' instruction is used to atomically modify memory. It
5385 loads a value in memory and compares it to a given value. If they are
5386 equal, it tries to store a new value into the memory.
5391 There are three arguments to the '``cmpxchg``' instruction: an address
5392 to operate on, a value to compare to the value currently be at that
5393 address, and a new value to place at that address if the compared values
5394 are equal. The type of '<cmp>' must be an integer type whose bit width
5395 is a power of two greater than or equal to eight and less than or equal
5396 to a target-specific size limit. '<cmp>' and '<new>' must have the same
5397 type, and the type of '<pointer>' must be a pointer to that type. If the
5398 ``cmpxchg`` is marked as ``volatile``, then the optimizer is not allowed
5399 to modify the number or order of execution of this ``cmpxchg`` with
5400 other :ref:`volatile operations <volatile>`.
5402 The success and failure :ref:`ordering <ordering>` arguments specify how this
5403 ``cmpxchg`` synchronizes with other atomic operations. Both ordering parameters
5404 must be at least ``monotonic``, the ordering constraint on failure must be no
5405 stronger than that on success, and the failure ordering cannot be either
5406 ``release`` or ``acq_rel``.
5408 The optional "``singlethread``" argument declares that the ``cmpxchg``
5409 is only atomic with respect to code (usually signal handlers) running in
5410 the same thread as the ``cmpxchg``. Otherwise the cmpxchg is atomic with
5411 respect to all other code in the system.
5413 The pointer passed into cmpxchg must have alignment greater than or
5414 equal to the size in memory of the operand.
5419 The contents of memory at the location specified by the '``<pointer>``' operand
5420 is read and compared to '``<cmp>``'; if the read value is the equal, the
5421 '``<new>``' is written. The original value at the location is returned, together
5422 with a flag indicating success (true) or failure (false).
5424 If the cmpxchg operation is marked as ``weak`` then a spurious failure is
5425 permitted: the operation may not write ``<new>`` even if the comparison
5428 If the cmpxchg operation is strong (the default), the i1 value is 1 if and only
5429 if the value loaded equals ``cmp``.
5431 A successful ``cmpxchg`` is a read-modify-write instruction for the purpose of
5432 identifying release sequences. A failed ``cmpxchg`` is equivalent to an atomic
5433 load with an ordering parameter determined the second ordering parameter.
5438 .. code-block:: llvm
5441 %orig = atomic load i32* %ptr unordered ; yields i32
5445 %cmp = phi i32 [ %orig, %entry ], [%old, %loop]
5446 %squared = mul i32 %cmp, %cmp
5447 %val_success = cmpxchg i32* %ptr, i32 %cmp, i32 %squared acq_rel monotonic ; yields { i32, i1 }
5448 %value_loaded = extractvalue { i32, i1 } %val_success, 0
5449 %success = extractvalue { i32, i1 } %val_success, 1
5450 br i1 %success, label %done, label %loop
5457 '``atomicrmw``' Instruction
5458 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
5465 atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering> ; yields ty
5470 The '``atomicrmw``' instruction is used to atomically modify memory.
5475 There are three arguments to the '``atomicrmw``' instruction: an
5476 operation to apply, an address whose value to modify, an argument to the
5477 operation. The operation must be one of the following keywords:
5491 The type of '<value>' must be an integer type whose bit width is a power
5492 of two greater than or equal to eight and less than or equal to a
5493 target-specific size limit. The type of the '``<pointer>``' operand must
5494 be a pointer to that type. If the ``atomicrmw`` is marked as
5495 ``volatile``, then the optimizer is not allowed to modify the number or
5496 order of execution of this ``atomicrmw`` with other :ref:`volatile
5497 operations <volatile>`.
5502 The contents of memory at the location specified by the '``<pointer>``'
5503 operand are atomically read, modified, and written back. The original
5504 value at the location is returned. The modification is specified by the
5507 - xchg: ``*ptr = val``
5508 - add: ``*ptr = *ptr + val``
5509 - sub: ``*ptr = *ptr - val``
5510 - and: ``*ptr = *ptr & val``
5511 - nand: ``*ptr = ~(*ptr & val)``
5512 - or: ``*ptr = *ptr | val``
5513 - xor: ``*ptr = *ptr ^ val``
5514 - max: ``*ptr = *ptr > val ? *ptr : val`` (using a signed comparison)
5515 - min: ``*ptr = *ptr < val ? *ptr : val`` (using a signed comparison)
5516 - umax: ``*ptr = *ptr > val ? *ptr : val`` (using an unsigned
5518 - umin: ``*ptr = *ptr < val ? *ptr : val`` (using an unsigned
5524 .. code-block:: llvm
5526 %old = atomicrmw add i32* %ptr, i32 1 acquire ; yields i32
5528 .. _i_getelementptr:
5530 '``getelementptr``' Instruction
5531 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5538 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
5539 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
5540 <result> = getelementptr <ptr vector> ptrval, <vector index type> idx
5545 The '``getelementptr``' instruction is used to get the address of a
5546 subelement of an :ref:`aggregate <t_aggregate>` data structure. It performs
5547 address calculation only and does not access memory.
5552 The first argument is always a pointer or a vector of pointers, and
5553 forms the basis of the calculation. The remaining arguments are indices
5554 that indicate which of the elements of the aggregate object are indexed.
5555 The interpretation of each index is dependent on the type being indexed
5556 into. The first index always indexes the pointer value given as the
5557 first argument, the second index indexes a value of the type pointed to
5558 (not necessarily the value directly pointed to, since the first index
5559 can be non-zero), etc. The first type indexed into must be a pointer
5560 value, subsequent types can be arrays, vectors, and structs. Note that
5561 subsequent types being indexed into can never be pointers, since that
5562 would require loading the pointer before continuing calculation.
5564 The type of each index argument depends on the type it is indexing into.
5565 When indexing into a (optionally packed) structure, only ``i32`` integer
5566 **constants** are allowed (when using a vector of indices they must all
5567 be the **same** ``i32`` integer constant). When indexing into an array,
5568 pointer or vector, integers of any width are allowed, and they are not
5569 required to be constant. These integers are treated as signed values
5572 For example, let's consider a C code fragment and how it gets compiled
5588 int *foo(struct ST *s) {
5589 return &s[1].Z.B[5][13];
5592 The LLVM code generated by Clang is:
5594 .. code-block:: llvm
5596 %struct.RT = type { i8, [10 x [20 x i32]], i8 }
5597 %struct.ST = type { i32, double, %struct.RT }
5599 define i32* @foo(%struct.ST* %s) nounwind uwtable readnone optsize ssp {
5601 %arrayidx = getelementptr inbounds %struct.ST* %s, i64 1, i32 2, i32 1, i64 5, i64 13
5608 In the example above, the first index is indexing into the
5609 '``%struct.ST*``' type, which is a pointer, yielding a '``%struct.ST``'
5610 = '``{ i32, double, %struct.RT }``' type, a structure. The second index
5611 indexes into the third element of the structure, yielding a
5612 '``%struct.RT``' = '``{ i8 , [10 x [20 x i32]], i8 }``' type, another
5613 structure. The third index indexes into the second element of the
5614 structure, yielding a '``[10 x [20 x i32]]``' type, an array. The two
5615 dimensions of the array are subscripted into, yielding an '``i32``'
5616 type. The '``getelementptr``' instruction returns a pointer to this
5617 element, thus computing a value of '``i32*``' type.
5619 Note that it is perfectly legal to index partially through a structure,
5620 returning a pointer to an inner element. Because of this, the LLVM code
5621 for the given testcase is equivalent to:
5623 .. code-block:: llvm
5625 define i32* @foo(%struct.ST* %s) {
5626 %t1 = getelementptr %struct.ST* %s, i32 1 ; yields %struct.ST*:%t1
5627 %t2 = getelementptr %struct.ST* %t1, i32 0, i32 2 ; yields %struct.RT*:%t2
5628 %t3 = getelementptr %struct.RT* %t2, i32 0, i32 1 ; yields [10 x [20 x i32]]*:%t3
5629 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 ; yields [20 x i32]*:%t4
5630 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 ; yields i32*:%t5
5634 If the ``inbounds`` keyword is present, the result value of the
5635 ``getelementptr`` is a :ref:`poison value <poisonvalues>` if the base
5636 pointer is not an *in bounds* address of an allocated object, or if any
5637 of the addresses that would be formed by successive addition of the
5638 offsets implied by the indices to the base address with infinitely
5639 precise signed arithmetic are not an *in bounds* address of that
5640 allocated object. The *in bounds* addresses for an allocated object are
5641 all the addresses that point into the object, plus the address one byte
5642 past the end. In cases where the base is a vector of pointers the
5643 ``inbounds`` keyword applies to each of the computations element-wise.
5645 If the ``inbounds`` keyword is not present, the offsets are added to the
5646 base address with silently-wrapping two's complement arithmetic. If the
5647 offsets have a different width from the pointer, they are sign-extended
5648 or truncated to the width of the pointer. The result value of the
5649 ``getelementptr`` may be outside the object pointed to by the base
5650 pointer. The result value may not necessarily be used to access memory
5651 though, even if it happens to point into allocated storage. See the
5652 :ref:`Pointer Aliasing Rules <pointeraliasing>` section for more
5655 The getelementptr instruction is often confusing. For some more insight
5656 into how it works, see :doc:`the getelementptr FAQ <GetElementPtr>`.
5661 .. code-block:: llvm
5663 ; yields [12 x i8]*:aptr
5664 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
5666 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
5668 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
5670 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
5672 In cases where the pointer argument is a vector of pointers, each index
5673 must be a vector with the same number of elements. For example:
5675 .. code-block:: llvm
5677 %A = getelementptr <4 x i8*> %ptrs, <4 x i64> %offsets,
5679 Conversion Operations
5680 ---------------------
5682 The instructions in this category are the conversion instructions
5683 (casting) which all take a single operand and a type. They perform
5684 various bit conversions on the operand.
5686 '``trunc .. to``' Instruction
5687 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5694 <result> = trunc <ty> <value> to <ty2> ; yields ty2
5699 The '``trunc``' instruction truncates its operand to the type ``ty2``.
5704 The '``trunc``' instruction takes a value to trunc, and a type to trunc
5705 it to. Both types must be of :ref:`integer <t_integer>` types, or vectors
5706 of the same number of integers. The bit size of the ``value`` must be
5707 larger than the bit size of the destination type, ``ty2``. Equal sized
5708 types are not allowed.
5713 The '``trunc``' instruction truncates the high order bits in ``value``
5714 and converts the remaining bits to ``ty2``. Since the source size must
5715 be larger than the destination size, ``trunc`` cannot be a *no-op cast*.
5716 It will always truncate bits.
5721 .. code-block:: llvm
5723 %X = trunc i32 257 to i8 ; yields i8:1
5724 %Y = trunc i32 123 to i1 ; yields i1:true
5725 %Z = trunc i32 122 to i1 ; yields i1:false
5726 %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> ; yields <i8 8, i8 7>
5728 '``zext .. to``' Instruction
5729 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5736 <result> = zext <ty> <value> to <ty2> ; yields ty2
5741 The '``zext``' instruction zero extends its operand to type ``ty2``.
5746 The '``zext``' instruction takes a value to cast, and a type to cast it
5747 to. Both types must be of :ref:`integer <t_integer>` types, or vectors of
5748 the same number of integers. The bit size of the ``value`` must be
5749 smaller than the bit size of the destination type, ``ty2``.
5754 The ``zext`` fills the high order bits of the ``value`` with zero bits
5755 until it reaches the size of the destination type, ``ty2``.
5757 When zero extending from i1, the result will always be either 0 or 1.
5762 .. code-block:: llvm
5764 %X = zext i32 257 to i64 ; yields i64:257
5765 %Y = zext i1 true to i32 ; yields i32:1
5766 %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7>
5768 '``sext .. to``' Instruction
5769 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5776 <result> = sext <ty> <value> to <ty2> ; yields ty2
5781 The '``sext``' sign extends ``value`` to the type ``ty2``.
5786 The '``sext``' instruction takes a value to cast, and a type to cast it
5787 to. Both types must be of :ref:`integer <t_integer>` types, or vectors of
5788 the same number of integers. The bit size of the ``value`` must be
5789 smaller than the bit size of the destination type, ``ty2``.
5794 The '``sext``' instruction performs a sign extension by copying the sign
5795 bit (highest order bit) of the ``value`` until it reaches the bit size
5796 of the type ``ty2``.
5798 When sign extending from i1, the extension always results in -1 or 0.
5803 .. code-block:: llvm
5805 %X = sext i8 -1 to i16 ; yields i16 :65535
5806 %Y = sext i1 true to i32 ; yields i32:-1
5807 %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7>
5809 '``fptrunc .. to``' Instruction
5810 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5817 <result> = fptrunc <ty> <value> to <ty2> ; yields ty2
5822 The '``fptrunc``' instruction truncates ``value`` to type ``ty2``.
5827 The '``fptrunc``' instruction takes a :ref:`floating point <t_floating>`
5828 value to cast and a :ref:`floating point <t_floating>` type to cast it to.
5829 The size of ``value`` must be larger than the size of ``ty2``. This
5830 implies that ``fptrunc`` cannot be used to make a *no-op cast*.
5835 The '``fptrunc``' instruction truncates a ``value`` from a larger
5836 :ref:`floating point <t_floating>` type to a smaller :ref:`floating
5837 point <t_floating>` type. If the value cannot fit within the
5838 destination type, ``ty2``, then the results are undefined.
5843 .. code-block:: llvm
5845 %X = fptrunc double 123.0 to float ; yields float:123.0
5846 %Y = fptrunc double 1.0E+300 to float ; yields undefined
5848 '``fpext .. to``' Instruction
5849 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5856 <result> = fpext <ty> <value> to <ty2> ; yields ty2
5861 The '``fpext``' extends a floating point ``value`` to a larger floating
5867 The '``fpext``' instruction takes a :ref:`floating point <t_floating>`
5868 ``value`` to cast, and a :ref:`floating point <t_floating>` type to cast it
5869 to. The source type must be smaller than the destination type.
5874 The '``fpext``' instruction extends the ``value`` from a smaller
5875 :ref:`floating point <t_floating>` type to a larger :ref:`floating
5876 point <t_floating>` type. The ``fpext`` cannot be used to make a
5877 *no-op cast* because it always changes bits. Use ``bitcast`` to make a
5878 *no-op cast* for a floating point cast.
5883 .. code-block:: llvm
5885 %X = fpext float 3.125 to double ; yields double:3.125000e+00
5886 %Y = fpext double %X to fp128 ; yields fp128:0xL00000000000000004000900000000000
5888 '``fptoui .. to``' Instruction
5889 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5896 <result> = fptoui <ty> <value> to <ty2> ; yields ty2
5901 The '``fptoui``' converts a floating point ``value`` to its unsigned
5902 integer equivalent of type ``ty2``.
5907 The '``fptoui``' instruction takes a value to cast, which must be a
5908 scalar or vector :ref:`floating point <t_floating>` value, and a type to
5909 cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If
5910 ``ty`` is a vector floating point type, ``ty2`` must be a vector integer
5911 type with the same number of elements as ``ty``
5916 The '``fptoui``' instruction converts its :ref:`floating
5917 point <t_floating>` operand into the nearest (rounding towards zero)
5918 unsigned integer value. If the value cannot fit in ``ty2``, the results
5924 .. code-block:: llvm
5926 %X = fptoui double 123.0 to i32 ; yields i32:123
5927 %Y = fptoui float 1.0E+300 to i1 ; yields undefined:1
5928 %Z = fptoui float 1.04E+17 to i8 ; yields undefined:1
5930 '``fptosi .. to``' Instruction
5931 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5938 <result> = fptosi <ty> <value> to <ty2> ; yields ty2
5943 The '``fptosi``' instruction converts :ref:`floating point <t_floating>`
5944 ``value`` to type ``ty2``.
5949 The '``fptosi``' instruction takes a value to cast, which must be a
5950 scalar or vector :ref:`floating point <t_floating>` value, and a type to
5951 cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If
5952 ``ty`` is a vector floating point type, ``ty2`` must be a vector integer
5953 type with the same number of elements as ``ty``
5958 The '``fptosi``' instruction converts its :ref:`floating
5959 point <t_floating>` operand into the nearest (rounding towards zero)
5960 signed integer value. If the value cannot fit in ``ty2``, the results
5966 .. code-block:: llvm
5968 %X = fptosi double -123.0 to i32 ; yields i32:-123
5969 %Y = fptosi float 1.0E-247 to i1 ; yields undefined:1
5970 %Z = fptosi float 1.04E+17 to i8 ; yields undefined:1
5972 '``uitofp .. to``' Instruction
5973 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5980 <result> = uitofp <ty> <value> to <ty2> ; yields ty2
5985 The '``uitofp``' instruction regards ``value`` as an unsigned integer
5986 and converts that value to the ``ty2`` type.
5991 The '``uitofp``' instruction takes a value to cast, which must be a
5992 scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to
5993 ``ty2``, which must be an :ref:`floating point <t_floating>` type. If
5994 ``ty`` is a vector integer type, ``ty2`` must be a vector floating point
5995 type with the same number of elements as ``ty``
6000 The '``uitofp``' instruction interprets its operand as an unsigned
6001 integer quantity and converts it to the corresponding floating point
6002 value. If the value cannot fit in the floating point value, the results
6008 .. code-block:: llvm
6010 %X = uitofp i32 257 to float ; yields float:257.0
6011 %Y = uitofp i8 -1 to double ; yields double:255.0
6013 '``sitofp .. to``' Instruction
6014 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6021 <result> = sitofp <ty> <value> to <ty2> ; yields ty2
6026 The '``sitofp``' instruction regards ``value`` as a signed integer and
6027 converts that value to the ``ty2`` type.
6032 The '``sitofp``' instruction takes a value to cast, which must be a
6033 scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to
6034 ``ty2``, which must be an :ref:`floating point <t_floating>` type. If
6035 ``ty`` is a vector integer type, ``ty2`` must be a vector floating point
6036 type with the same number of elements as ``ty``
6041 The '``sitofp``' instruction interprets its operand as a signed integer
6042 quantity and converts it to the corresponding floating point value. If
6043 the value cannot fit in the floating point value, the results are
6049 .. code-block:: llvm
6051 %X = sitofp i32 257 to float ; yields float:257.0
6052 %Y = sitofp i8 -1 to double ; yields double:-1.0
6056 '``ptrtoint .. to``' Instruction
6057 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6064 <result> = ptrtoint <ty> <value> to <ty2> ; yields ty2
6069 The '``ptrtoint``' instruction converts the pointer or a vector of
6070 pointers ``value`` to the integer (or vector of integers) type ``ty2``.
6075 The '``ptrtoint``' instruction takes a ``value`` to cast, which must be
6076 a a value of type :ref:`pointer <t_pointer>` or a vector of pointers, and a
6077 type to cast it to ``ty2``, which must be an :ref:`integer <t_integer>` or
6078 a vector of integers type.
6083 The '``ptrtoint``' instruction converts ``value`` to integer type
6084 ``ty2`` by interpreting the pointer value as an integer and either
6085 truncating or zero extending that value to the size of the integer type.
6086 If ``value`` is smaller than ``ty2`` then a zero extension is done. If
6087 ``value`` is larger than ``ty2`` then a truncation is done. If they are
6088 the same size, then nothing is done (*no-op cast*) other than a type
6094 .. code-block:: llvm
6096 %X = ptrtoint i32* %P to i8 ; yields truncation on 32-bit architecture
6097 %Y = ptrtoint i32* %P to i64 ; yields zero extension on 32-bit architecture
6098 %Z = ptrtoint <4 x i32*> %P to <4 x i64>; yields vector zero extension for a vector of addresses on 32-bit architecture
6102 '``inttoptr .. to``' Instruction
6103 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6110 <result> = inttoptr <ty> <value> to <ty2> ; yields ty2
6115 The '``inttoptr``' instruction converts an integer ``value`` to a
6116 pointer type, ``ty2``.
6121 The '``inttoptr``' instruction takes an :ref:`integer <t_integer>` value to
6122 cast, and a type to cast it to, which must be a :ref:`pointer <t_pointer>`
6128 The '``inttoptr``' instruction converts ``value`` to type ``ty2`` by
6129 applying either a zero extension or a truncation depending on the size
6130 of the integer ``value``. If ``value`` is larger than the size of a
6131 pointer then a truncation is done. If ``value`` is smaller than the size
6132 of a pointer then a zero extension is done. If they are the same size,
6133 nothing is done (*no-op cast*).
6138 .. code-block:: llvm
6140 %X = inttoptr i32 255 to i32* ; yields zero extension on 64-bit architecture
6141 %Y = inttoptr i32 255 to i32* ; yields no-op on 32-bit architecture
6142 %Z = inttoptr i64 0 to i32* ; yields truncation on 32-bit architecture
6143 %Z = inttoptr <4 x i32> %G to <4 x i8*>; yields truncation of vector G to four pointers
6147 '``bitcast .. to``' Instruction
6148 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6155 <result> = bitcast <ty> <value> to <ty2> ; yields ty2
6160 The '``bitcast``' instruction converts ``value`` to type ``ty2`` without
6166 The '``bitcast``' instruction takes a value to cast, which must be a
6167 non-aggregate first class value, and a type to cast it to, which must
6168 also be a non-aggregate :ref:`first class <t_firstclass>` type. The
6169 bit sizes of ``value`` and the destination type, ``ty2``, must be
6170 identical. If the source type is a pointer, the destination type must
6171 also be a pointer of the same size. This instruction supports bitwise
6172 conversion of vectors to integers and to vectors of other types (as
6173 long as they have the same size).
6178 The '``bitcast``' instruction converts ``value`` to type ``ty2``. It
6179 is always a *no-op cast* because no bits change with this
6180 conversion. The conversion is done as if the ``value`` had been stored
6181 to memory and read back as type ``ty2``. Pointer (or vector of
6182 pointers) types may only be converted to other pointer (or vector of
6183 pointers) types with the same address space through this instruction.
6184 To convert pointers to other types, use the :ref:`inttoptr <i_inttoptr>`
6185 or :ref:`ptrtoint <i_ptrtoint>` instructions first.
6190 .. code-block:: llvm
6192 %X = bitcast i8 255 to i8 ; yields i8 :-1
6193 %Y = bitcast i32* %x to sint* ; yields sint*:%x
6194 %Z = bitcast <2 x int> %V to i64; ; yields i64: %V
6195 %Z = bitcast <2 x i32*> %V to <2 x i64*> ; yields <2 x i64*>
6197 .. _i_addrspacecast:
6199 '``addrspacecast .. to``' Instruction
6200 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6207 <result> = addrspacecast <pty> <ptrval> to <pty2> ; yields pty2
6212 The '``addrspacecast``' instruction converts ``ptrval`` from ``pty`` in
6213 address space ``n`` to type ``pty2`` in address space ``m``.
6218 The '``addrspacecast``' instruction takes a pointer or vector of pointer value
6219 to cast and a pointer type to cast it to, which must have a different
6225 The '``addrspacecast``' instruction converts the pointer value
6226 ``ptrval`` to type ``pty2``. It can be a *no-op cast* or a complex
6227 value modification, depending on the target and the address space
6228 pair. Pointer conversions within the same address space must be
6229 performed with the ``bitcast`` instruction. Note that if the address space
6230 conversion is legal then both result and operand refer to the same memory
6236 .. code-block:: llvm
6238 %X = addrspacecast i32* %x to i32 addrspace(1)* ; yields i32 addrspace(1)*:%x
6239 %Y = addrspacecast i32 addrspace(1)* %y to i64 addrspace(2)* ; yields i64 addrspace(2)*:%y
6240 %Z = addrspacecast <4 x i32*> %z to <4 x float addrspace(3)*> ; yields <4 x float addrspace(3)*>:%z
6247 The instructions in this category are the "miscellaneous" instructions,
6248 which defy better classification.
6252 '``icmp``' Instruction
6253 ^^^^^^^^^^^^^^^^^^^^^^
6260 <result> = icmp <cond> <ty> <op1>, <op2> ; yields i1 or <N x i1>:result
6265 The '``icmp``' instruction returns a boolean value or a vector of
6266 boolean values based on comparison of its two integer, integer vector,
6267 pointer, or pointer vector operands.
6272 The '``icmp``' instruction takes three operands. The first operand is
6273 the condition code indicating the kind of comparison to perform. It is
6274 not a value, just a keyword. The possible condition code are:
6277 #. ``ne``: not equal
6278 #. ``ugt``: unsigned greater than
6279 #. ``uge``: unsigned greater or equal
6280 #. ``ult``: unsigned less than
6281 #. ``ule``: unsigned less or equal
6282 #. ``sgt``: signed greater than
6283 #. ``sge``: signed greater or equal
6284 #. ``slt``: signed less than
6285 #. ``sle``: signed less or equal
6287 The remaining two arguments must be :ref:`integer <t_integer>` or
6288 :ref:`pointer <t_pointer>` or integer :ref:`vector <t_vector>` typed. They
6289 must also be identical types.
6294 The '``icmp``' compares ``op1`` and ``op2`` according to the condition
6295 code given as ``cond``. The comparison performed always yields either an
6296 :ref:`i1 <t_integer>` or vector of ``i1`` result, as follows:
6298 #. ``eq``: yields ``true`` if the operands are equal, ``false``
6299 otherwise. No sign interpretation is necessary or performed.
6300 #. ``ne``: yields ``true`` if the operands are unequal, ``false``
6301 otherwise. No sign interpretation is necessary or performed.
6302 #. ``ugt``: interprets the operands as unsigned values and yields
6303 ``true`` if ``op1`` is greater than ``op2``.
6304 #. ``uge``: interprets the operands as unsigned values and yields
6305 ``true`` if ``op1`` is greater than or equal to ``op2``.
6306 #. ``ult``: interprets the operands as unsigned values and yields
6307 ``true`` if ``op1`` is less than ``op2``.
6308 #. ``ule``: interprets the operands as unsigned values and yields
6309 ``true`` if ``op1`` is less than or equal to ``op2``.
6310 #. ``sgt``: interprets the operands as signed values and yields ``true``
6311 if ``op1`` is greater than ``op2``.
6312 #. ``sge``: interprets the operands as signed values and yields ``true``
6313 if ``op1`` is greater than or equal to ``op2``.
6314 #. ``slt``: interprets the operands as signed values and yields ``true``
6315 if ``op1`` is less than ``op2``.
6316 #. ``sle``: interprets the operands as signed values and yields ``true``
6317 if ``op1`` is less than or equal to ``op2``.
6319 If the operands are :ref:`pointer <t_pointer>` typed, the pointer values
6320 are compared as if they were integers.
6322 If the operands are integer vectors, then they are compared element by
6323 element. The result is an ``i1`` vector with the same number of elements
6324 as the values being compared. Otherwise, the result is an ``i1``.
6329 .. code-block:: llvm
6331 <result> = icmp eq i32 4, 5 ; yields: result=false
6332 <result> = icmp ne float* %X, %X ; yields: result=false
6333 <result> = icmp ult i16 4, 5 ; yields: result=true
6334 <result> = icmp sgt i16 4, 5 ; yields: result=false
6335 <result> = icmp ule i16 -4, 5 ; yields: result=false
6336 <result> = icmp sge i16 4, 5 ; yields: result=false
6338 Note that the code generator does not yet support vector types with the
6339 ``icmp`` instruction.
6343 '``fcmp``' Instruction
6344 ^^^^^^^^^^^^^^^^^^^^^^
6351 <result> = fcmp <cond> <ty> <op1>, <op2> ; yields i1 or <N x i1>:result
6356 The '``fcmp``' instruction returns a boolean value or vector of boolean
6357 values based on comparison of its operands.
6359 If the operands are floating point scalars, then the result type is a
6360 boolean (:ref:`i1 <t_integer>`).
6362 If the operands are floating point vectors, then the result type is a
6363 vector of boolean with the same number of elements as the operands being
6369 The '``fcmp``' instruction takes three operands. The first operand is
6370 the condition code indicating the kind of comparison to perform. It is
6371 not a value, just a keyword. The possible condition code are:
6373 #. ``false``: no comparison, always returns false
6374 #. ``oeq``: ordered and equal
6375 #. ``ogt``: ordered and greater than
6376 #. ``oge``: ordered and greater than or equal
6377 #. ``olt``: ordered and less than
6378 #. ``ole``: ordered and less than or equal
6379 #. ``one``: ordered and not equal
6380 #. ``ord``: ordered (no nans)
6381 #. ``ueq``: unordered or equal
6382 #. ``ugt``: unordered or greater than
6383 #. ``uge``: unordered or greater than or equal
6384 #. ``ult``: unordered or less than
6385 #. ``ule``: unordered or less than or equal
6386 #. ``une``: unordered or not equal
6387 #. ``uno``: unordered (either nans)
6388 #. ``true``: no comparison, always returns true
6390 *Ordered* means that neither operand is a QNAN while *unordered* means
6391 that either operand may be a QNAN.
6393 Each of ``val1`` and ``val2`` arguments must be either a :ref:`floating
6394 point <t_floating>` type or a :ref:`vector <t_vector>` of floating point
6395 type. They must have identical types.
6400 The '``fcmp``' instruction compares ``op1`` and ``op2`` according to the
6401 condition code given as ``cond``. If the operands are vectors, then the
6402 vectors are compared element by element. Each comparison performed
6403 always yields an :ref:`i1 <t_integer>` result, as follows:
6405 #. ``false``: always yields ``false``, regardless of operands.
6406 #. ``oeq``: yields ``true`` if both operands are not a QNAN and ``op1``
6407 is equal to ``op2``.
6408 #. ``ogt``: yields ``true`` if both operands are not a QNAN and ``op1``
6409 is greater than ``op2``.
6410 #. ``oge``: yields ``true`` if both operands are not a QNAN and ``op1``
6411 is greater than or equal to ``op2``.
6412 #. ``olt``: yields ``true`` if both operands are not a QNAN and ``op1``
6413 is less than ``op2``.
6414 #. ``ole``: yields ``true`` if both operands are not a QNAN and ``op1``
6415 is less than or equal to ``op2``.
6416 #. ``one``: yields ``true`` if both operands are not a QNAN and ``op1``
6417 is not equal to ``op2``.
6418 #. ``ord``: yields ``true`` if both operands are not a QNAN.
6419 #. ``ueq``: yields ``true`` if either operand is a QNAN or ``op1`` is
6421 #. ``ugt``: yields ``true`` if either operand is a QNAN or ``op1`` is
6422 greater than ``op2``.
6423 #. ``uge``: yields ``true`` if either operand is a QNAN or ``op1`` is
6424 greater than or equal to ``op2``.
6425 #. ``ult``: yields ``true`` if either operand is a QNAN or ``op1`` is
6427 #. ``ule``: yields ``true`` if either operand is a QNAN or ``op1`` is
6428 less than or equal to ``op2``.
6429 #. ``une``: yields ``true`` if either operand is a QNAN or ``op1`` is
6430 not equal to ``op2``.
6431 #. ``uno``: yields ``true`` if either operand is a QNAN.
6432 #. ``true``: always yields ``true``, regardless of operands.
6437 .. code-block:: llvm
6439 <result> = fcmp oeq float 4.0, 5.0 ; yields: result=false
6440 <result> = fcmp one float 4.0, 5.0 ; yields: result=true
6441 <result> = fcmp olt float 4.0, 5.0 ; yields: result=true
6442 <result> = fcmp ueq double 1.0, 2.0 ; yields: result=false
6444 Note that the code generator does not yet support vector types with the
6445 ``fcmp`` instruction.
6449 '``phi``' Instruction
6450 ^^^^^^^^^^^^^^^^^^^^^
6457 <result> = phi <ty> [ <val0>, <label0>], ...
6462 The '``phi``' instruction is used to implement the φ node in the SSA
6463 graph representing the function.
6468 The type of the incoming values is specified with the first type field.
6469 After this, the '``phi``' instruction takes a list of pairs as
6470 arguments, with one pair for each predecessor basic block of the current
6471 block. Only values of :ref:`first class <t_firstclass>` type may be used as
6472 the value arguments to the PHI node. Only labels may be used as the
6475 There must be no non-phi instructions between the start of a basic block
6476 and the PHI instructions: i.e. PHI instructions must be first in a basic
6479 For the purposes of the SSA form, the use of each incoming value is
6480 deemed to occur on the edge from the corresponding predecessor block to
6481 the current block (but after any definition of an '``invoke``'
6482 instruction's return value on the same edge).
6487 At runtime, the '``phi``' instruction logically takes on the value
6488 specified by the pair corresponding to the predecessor basic block that
6489 executed just prior to the current block.
6494 .. code-block:: llvm
6496 Loop: ; Infinite loop that counts from 0 on up...
6497 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
6498 %nextindvar = add i32 %indvar, 1
6503 '``select``' Instruction
6504 ^^^^^^^^^^^^^^^^^^^^^^^^
6511 <result> = select selty <cond>, <ty> <val1>, <ty> <val2> ; yields ty
6513 selty is either i1 or {<N x i1>}
6518 The '``select``' instruction is used to choose one value based on a
6519 condition, without IR-level branching.
6524 The '``select``' instruction requires an 'i1' value or a vector of 'i1'
6525 values indicating the condition, and two values of the same :ref:`first
6526 class <t_firstclass>` type. If the val1/val2 are vectors and the
6527 condition is a scalar, then entire vectors are selected, not individual
6533 If the condition is an i1 and it evaluates to 1, the instruction returns
6534 the first value argument; otherwise, it returns the second value
6537 If the condition is a vector of i1, then the value arguments must be
6538 vectors of the same size, and the selection is done element by element.
6543 .. code-block:: llvm
6545 %X = select i1 true, i8 17, i8 42 ; yields i8:17
6549 '``call``' Instruction
6550 ^^^^^^^^^^^^^^^^^^^^^^
6557 <result> = [tail | musttail] call [cconv] [ret attrs] <ty> [<fnty>*] <fnptrval>(<function args>) [fn attrs]
6562 The '``call``' instruction represents a simple function call.
6567 This instruction requires several arguments:
6569 #. The optional ``tail`` and ``musttail`` markers indicate that the optimizers
6570 should perform tail call optimization. The ``tail`` marker is a hint that
6571 `can be ignored <CodeGenerator.html#sibcallopt>`_. The ``musttail`` marker
6572 means that the call must be tail call optimized in order for the program to
6573 be correct. The ``musttail`` marker provides these guarantees:
6575 #. The call will not cause unbounded stack growth if it is part of a
6576 recursive cycle in the call graph.
6577 #. Arguments with the :ref:`inalloca <attr_inalloca>` attribute are
6580 Both markers imply that the callee does not access allocas or varargs from
6581 the caller. Calls marked ``musttail`` must obey the following additional
6584 - The call must immediately precede a :ref:`ret <i_ret>` instruction,
6585 or a pointer bitcast followed by a ret instruction.
6586 - The ret instruction must return the (possibly bitcasted) value
6587 produced by the call or void.
6588 - The caller and callee prototypes must match. Pointer types of
6589 parameters or return types may differ in pointee type, but not
6591 - The calling conventions of the caller and callee must match.
6592 - All ABI-impacting function attributes, such as sret, byval, inreg,
6593 returned, and inalloca, must match.
6594 - The callee must be varargs iff the caller is varargs. Bitcasting a
6595 non-varargs function to the appropriate varargs type is legal so
6596 long as the non-varargs prefixes obey the other rules.
6598 Tail call optimization for calls marked ``tail`` is guaranteed to occur if
6599 the following conditions are met:
6601 - Caller and callee both have the calling convention ``fastcc``.
6602 - The call is in tail position (ret immediately follows call and ret
6603 uses value of call or is void).
6604 - Option ``-tailcallopt`` is enabled, or
6605 ``llvm::GuaranteedTailCallOpt`` is ``true``.
6606 - `Platform-specific constraints are
6607 met. <CodeGenerator.html#tailcallopt>`_
6609 #. The optional "cconv" marker indicates which :ref:`calling
6610 convention <callingconv>` the call should use. If none is
6611 specified, the call defaults to using C calling conventions. The
6612 calling convention of the call must match the calling convention of
6613 the target function, or else the behavior is undefined.
6614 #. The optional :ref:`Parameter Attributes <paramattrs>` list for return
6615 values. Only '``zeroext``', '``signext``', and '``inreg``' attributes
6617 #. '``ty``': the type of the call instruction itself which is also the
6618 type of the return value. Functions that return no value are marked
6620 #. '``fnty``': shall be the signature of the pointer to function value
6621 being invoked. The argument types must match the types implied by
6622 this signature. This type can be omitted if the function is not
6623 varargs and if the function type does not return a pointer to a
6625 #. '``fnptrval``': An LLVM value containing a pointer to a function to
6626 be invoked. In most cases, this is a direct function invocation, but
6627 indirect ``call``'s are just as possible, calling an arbitrary pointer
6629 #. '``function args``': argument list whose types match the function
6630 signature argument types and parameter attributes. All arguments must
6631 be of :ref:`first class <t_firstclass>` type. If the function signature
6632 indicates the function accepts a variable number of arguments, the
6633 extra arguments can be specified.
6634 #. The optional :ref:`function attributes <fnattrs>` list. Only
6635 '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``'
6636 attributes are valid here.
6641 The '``call``' instruction is used to cause control flow to transfer to
6642 a specified function, with its incoming arguments bound to the specified
6643 values. Upon a '``ret``' instruction in the called function, control
6644 flow continues with the instruction after the function call, and the
6645 return value of the function is bound to the result argument.
6650 .. code-block:: llvm
6652 %retval = call i32 @test(i32 %argc)
6653 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) ; yields i32
6654 %X = tail call i32 @foo() ; yields i32
6655 %Y = tail call fastcc i32 @foo() ; yields i32
6656 call void %foo(i8 97 signext)
6658 %struct.A = type { i32, i8 }
6659 %r = call %struct.A @foo() ; yields { i32, i8 }
6660 %gr = extractvalue %struct.A %r, 0 ; yields i32
6661 %gr1 = extractvalue %struct.A %r, 1 ; yields i8
6662 %Z = call void @foo() noreturn ; indicates that %foo never returns normally
6663 %ZZ = call zeroext i32 @bar() ; Return value is %zero extended
6665 llvm treats calls to some functions with names and arguments that match
6666 the standard C99 library as being the C99 library functions, and may
6667 perform optimizations or generate code for them under that assumption.
6668 This is something we'd like to change in the future to provide better
6669 support for freestanding environments and non-C-based languages.
6673 '``va_arg``' Instruction
6674 ^^^^^^^^^^^^^^^^^^^^^^^^
6681 <resultval> = va_arg <va_list*> <arglist>, <argty>
6686 The '``va_arg``' instruction is used to access arguments passed through
6687 the "variable argument" area of a function call. It is used to implement
6688 the ``va_arg`` macro in C.
6693 This instruction takes a ``va_list*`` value and the type of the
6694 argument. It returns a value of the specified argument type and
6695 increments the ``va_list`` to point to the next argument. The actual
6696 type of ``va_list`` is target specific.
6701 The '``va_arg``' instruction loads an argument of the specified type
6702 from the specified ``va_list`` and causes the ``va_list`` to point to
6703 the next argument. For more information, see the variable argument
6704 handling :ref:`Intrinsic Functions <int_varargs>`.
6706 It is legal for this instruction to be called in a function which does
6707 not take a variable number of arguments, for example, the ``vfprintf``
6710 ``va_arg`` is an LLVM instruction instead of an :ref:`intrinsic
6711 function <intrinsics>` because it takes a type as an argument.
6716 See the :ref:`variable argument processing <int_varargs>` section.
6718 Note that the code generator does not yet fully support va\_arg on many
6719 targets. Also, it does not currently support va\_arg with aggregate
6720 types on any target.
6724 '``landingpad``' Instruction
6725 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6732 <resultval> = landingpad <resultty> personality <type> <pers_fn> <clause>+
6733 <resultval> = landingpad <resultty> personality <type> <pers_fn> cleanup <clause>*
6735 <clause> := catch <type> <value>
6736 <clause> := filter <array constant type> <array constant>
6741 The '``landingpad``' instruction is used by `LLVM's exception handling
6742 system <ExceptionHandling.html#overview>`_ to specify that a basic block
6743 is a landing pad --- one where the exception lands, and corresponds to the
6744 code found in the ``catch`` portion of a ``try``/``catch`` sequence. It
6745 defines values supplied by the personality function (``pers_fn``) upon
6746 re-entry to the function. The ``resultval`` has the type ``resultty``.
6751 This instruction takes a ``pers_fn`` value. This is the personality
6752 function associated with the unwinding mechanism. The optional
6753 ``cleanup`` flag indicates that the landing pad block is a cleanup.
6755 A ``clause`` begins with the clause type --- ``catch`` or ``filter`` --- and
6756 contains the global variable representing the "type" that may be caught
6757 or filtered respectively. Unlike the ``catch`` clause, the ``filter``
6758 clause takes an array constant as its argument. Use
6759 "``[0 x i8**] undef``" for a filter which cannot throw. The
6760 '``landingpad``' instruction must contain *at least* one ``clause`` or
6761 the ``cleanup`` flag.
6766 The '``landingpad``' instruction defines the values which are set by the
6767 personality function (``pers_fn``) upon re-entry to the function, and
6768 therefore the "result type" of the ``landingpad`` instruction. As with
6769 calling conventions, how the personality function results are
6770 represented in LLVM IR is target specific.
6772 The clauses are applied in order from top to bottom. If two
6773 ``landingpad`` instructions are merged together through inlining, the
6774 clauses from the calling function are appended to the list of clauses.
6775 When the call stack is being unwound due to an exception being thrown,
6776 the exception is compared against each ``clause`` in turn. If it doesn't
6777 match any of the clauses, and the ``cleanup`` flag is not set, then
6778 unwinding continues further up the call stack.
6780 The ``landingpad`` instruction has several restrictions:
6782 - A landing pad block is a basic block which is the unwind destination
6783 of an '``invoke``' instruction.
6784 - A landing pad block must have a '``landingpad``' instruction as its
6785 first non-PHI instruction.
6786 - There can be only one '``landingpad``' instruction within the landing
6788 - A basic block that is not a landing pad block may not include a
6789 '``landingpad``' instruction.
6790 - All '``landingpad``' instructions in a function must have the same
6791 personality function.
6796 .. code-block:: llvm
6798 ;; A landing pad which can catch an integer.
6799 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6801 ;; A landing pad that is a cleanup.
6802 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6804 ;; A landing pad which can catch an integer and can only throw a double.
6805 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6807 filter [1 x i8**] [@_ZTId]
6814 LLVM supports the notion of an "intrinsic function". These functions
6815 have well known names and semantics and are required to follow certain
6816 restrictions. Overall, these intrinsics represent an extension mechanism
6817 for the LLVM language that does not require changing all of the
6818 transformations in LLVM when adding to the language (or the bitcode
6819 reader/writer, the parser, etc...).
6821 Intrinsic function names must all start with an "``llvm.``" prefix. This
6822 prefix is reserved in LLVM for intrinsic names; thus, function names may
6823 not begin with this prefix. Intrinsic functions must always be external
6824 functions: you cannot define the body of intrinsic functions. Intrinsic
6825 functions may only be used in call or invoke instructions: it is illegal
6826 to take the address of an intrinsic function. Additionally, because
6827 intrinsic functions are part of the LLVM language, it is required if any
6828 are added that they be documented here.
6830 Some intrinsic functions can be overloaded, i.e., the intrinsic
6831 represents a family of functions that perform the same operation but on
6832 different data types. Because LLVM can represent over 8 million
6833 different integer types, overloading is used commonly to allow an
6834 intrinsic function to operate on any integer type. One or more of the
6835 argument types or the result type can be overloaded to accept any
6836 integer type. Argument types may also be defined as exactly matching a
6837 previous argument's type or the result type. This allows an intrinsic
6838 function which accepts multiple arguments, but needs all of them to be
6839 of the same type, to only be overloaded with respect to a single
6840 argument or the result.
6842 Overloaded intrinsics will have the names of its overloaded argument
6843 types encoded into its function name, each preceded by a period. Only
6844 those types which are overloaded result in a name suffix. Arguments
6845 whose type is matched against another type do not. For example, the
6846 ``llvm.ctpop`` function can take an integer of any width and returns an
6847 integer of exactly the same integer width. This leads to a family of
6848 functions such as ``i8 @llvm.ctpop.i8(i8 %val)`` and
6849 ``i29 @llvm.ctpop.i29(i29 %val)``. Only one type, the return type, is
6850 overloaded, and only one type suffix is required. Because the argument's
6851 type is matched against the return type, it does not require its own
6854 To learn how to add an intrinsic function, please see the `Extending
6855 LLVM Guide <ExtendingLLVM.html>`_.
6859 Variable Argument Handling Intrinsics
6860 -------------------------------------
6862 Variable argument support is defined in LLVM with the
6863 :ref:`va_arg <i_va_arg>` instruction and these three intrinsic
6864 functions. These functions are related to the similarly named macros
6865 defined in the ``<stdarg.h>`` header file.
6867 All of these functions operate on arguments that use a target-specific
6868 value type "``va_list``". The LLVM assembly language reference manual
6869 does not define what this type is, so all transformations should be
6870 prepared to handle these functions regardless of the type used.
6872 This example shows how the :ref:`va_arg <i_va_arg>` instruction and the
6873 variable argument handling intrinsic functions are used.
6875 .. code-block:: llvm
6877 define i32 @test(i32 %X, ...) {
6878 ; Initialize variable argument processing
6880 %ap2 = bitcast i8** %ap to i8*
6881 call void @llvm.va_start(i8* %ap2)
6883 ; Read a single integer argument
6884 %tmp = va_arg i8** %ap, i32
6886 ; Demonstrate usage of llvm.va_copy and llvm.va_end
6888 %aq2 = bitcast i8** %aq to i8*
6889 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
6890 call void @llvm.va_end(i8* %aq2)
6892 ; Stop processing of arguments.
6893 call void @llvm.va_end(i8* %ap2)
6897 declare void @llvm.va_start(i8*)
6898 declare void @llvm.va_copy(i8*, i8*)
6899 declare void @llvm.va_end(i8*)
6903 '``llvm.va_start``' Intrinsic
6904 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6911 declare void @llvm.va_start(i8* <arglist>)
6916 The '``llvm.va_start``' intrinsic initializes ``*<arglist>`` for
6917 subsequent use by ``va_arg``.
6922 The argument is a pointer to a ``va_list`` element to initialize.
6927 The '``llvm.va_start``' intrinsic works just like the ``va_start`` macro
6928 available in C. In a target-dependent way, it initializes the
6929 ``va_list`` element to which the argument points, so that the next call
6930 to ``va_arg`` will produce the first variable argument passed to the
6931 function. Unlike the C ``va_start`` macro, this intrinsic does not need
6932 to know the last argument of the function as the compiler can figure
6935 '``llvm.va_end``' Intrinsic
6936 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
6943 declare void @llvm.va_end(i8* <arglist>)
6948 The '``llvm.va_end``' intrinsic destroys ``*<arglist>``, which has been
6949 initialized previously with ``llvm.va_start`` or ``llvm.va_copy``.
6954 The argument is a pointer to a ``va_list`` to destroy.
6959 The '``llvm.va_end``' intrinsic works just like the ``va_end`` macro
6960 available in C. In a target-dependent way, it destroys the ``va_list``
6961 element to which the argument points. Calls to
6962 :ref:`llvm.va_start <int_va_start>` and
6963 :ref:`llvm.va_copy <int_va_copy>` must be matched exactly with calls to
6968 '``llvm.va_copy``' Intrinsic
6969 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6976 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
6981 The '``llvm.va_copy``' intrinsic copies the current argument position
6982 from the source argument list to the destination argument list.
6987 The first argument is a pointer to a ``va_list`` element to initialize.
6988 The second argument is a pointer to a ``va_list`` element to copy from.
6993 The '``llvm.va_copy``' intrinsic works just like the ``va_copy`` macro
6994 available in C. In a target-dependent way, it copies the source
6995 ``va_list`` element into the destination ``va_list`` element. This
6996 intrinsic is necessary because the `` llvm.va_start`` intrinsic may be
6997 arbitrarily complex and require, for example, memory allocation.
6999 Accurate Garbage Collection Intrinsics
7000 --------------------------------------
7002 LLVM support for `Accurate Garbage Collection <GarbageCollection.html>`_
7003 (GC) requires the implementation and generation of these intrinsics.
7004 These intrinsics allow identification of :ref:`GC roots on the
7005 stack <int_gcroot>`, as well as garbage collector implementations that
7006 require :ref:`read <int_gcread>` and :ref:`write <int_gcwrite>` barriers.
7007 Front-ends for type-safe garbage collected languages should generate
7008 these intrinsics to make use of the LLVM garbage collectors. For more
7009 details, see `Accurate Garbage Collection with
7010 LLVM <GarbageCollection.html>`_.
7012 The garbage collection intrinsics only operate on objects in the generic
7013 address space (address space zero).
7017 '``llvm.gcroot``' Intrinsic
7018 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7025 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
7030 The '``llvm.gcroot``' intrinsic declares the existence of a GC root to
7031 the code generator, and allows some metadata to be associated with it.
7036 The first argument specifies the address of a stack object that contains
7037 the root pointer. The second pointer (which must be either a constant or
7038 a global value address) contains the meta-data to be associated with the
7044 At runtime, a call to this intrinsic stores a null pointer into the
7045 "ptrloc" location. At compile-time, the code generator generates
7046 information to allow the runtime to find the pointer at GC safe points.
7047 The '``llvm.gcroot``' intrinsic may only be used in a function which
7048 :ref:`specifies a GC algorithm <gc>`.
7052 '``llvm.gcread``' Intrinsic
7053 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7060 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
7065 The '``llvm.gcread``' intrinsic identifies reads of references from heap
7066 locations, allowing garbage collector implementations that require read
7072 The second argument is the address to read from, which should be an
7073 address allocated from the garbage collector. The first object is a
7074 pointer to the start of the referenced object, if needed by the language
7075 runtime (otherwise null).
7080 The '``llvm.gcread``' intrinsic has the same semantics as a load
7081 instruction, but may be replaced with substantially more complex code by
7082 the garbage collector runtime, as needed. The '``llvm.gcread``'
7083 intrinsic may only be used in a function which :ref:`specifies a GC
7088 '``llvm.gcwrite``' Intrinsic
7089 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7096 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
7101 The '``llvm.gcwrite``' intrinsic identifies writes of references to heap
7102 locations, allowing garbage collector implementations that require write
7103 barriers (such as generational or reference counting collectors).
7108 The first argument is the reference to store, the second is the start of
7109 the object to store it to, and the third is the address of the field of
7110 Obj to store to. If the runtime does not require a pointer to the
7111 object, Obj may be null.
7116 The '``llvm.gcwrite``' intrinsic has the same semantics as a store
7117 instruction, but may be replaced with substantially more complex code by
7118 the garbage collector runtime, as needed. The '``llvm.gcwrite``'
7119 intrinsic may only be used in a function which :ref:`specifies a GC
7122 Code Generator Intrinsics
7123 -------------------------
7125 These intrinsics are provided by LLVM to expose special features that
7126 may only be implemented with code generator support.
7128 '``llvm.returnaddress``' Intrinsic
7129 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7136 declare i8 *@llvm.returnaddress(i32 <level>)
7141 The '``llvm.returnaddress``' intrinsic attempts to compute a
7142 target-specific value indicating the return address of the current
7143 function or one of its callers.
7148 The argument to this intrinsic indicates which function to return the
7149 address for. Zero indicates the calling function, one indicates its
7150 caller, etc. The argument is **required** to be a constant integer
7156 The '``llvm.returnaddress``' intrinsic either returns a pointer
7157 indicating the return address of the specified call frame, or zero if it
7158 cannot be identified. The value returned by this intrinsic is likely to
7159 be incorrect or 0 for arguments other than zero, so it should only be
7160 used for debugging purposes.
7162 Note that calling this intrinsic does not prevent function inlining or
7163 other aggressive transformations, so the value returned may not be that
7164 of the obvious source-language caller.
7166 '``llvm.frameaddress``' Intrinsic
7167 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7174 declare i8* @llvm.frameaddress(i32 <level>)
7179 The '``llvm.frameaddress``' intrinsic attempts to return the
7180 target-specific frame pointer value for the specified stack frame.
7185 The argument to this intrinsic indicates which function to return the
7186 frame pointer for. Zero indicates the calling function, one indicates
7187 its caller, etc. The argument is **required** to be a constant integer
7193 The '``llvm.frameaddress``' intrinsic either returns a pointer
7194 indicating the frame address of the specified call frame, or zero if it
7195 cannot be identified. The value returned by this intrinsic is likely to
7196 be incorrect or 0 for arguments other than zero, so it should only be
7197 used for debugging purposes.
7199 Note that calling this intrinsic does not prevent function inlining or
7200 other aggressive transformations, so the value returned may not be that
7201 of the obvious source-language caller.
7203 .. _int_read_register:
7204 .. _int_write_register:
7206 '``llvm.read_register``' and '``llvm.write_register``' Intrinsics
7207 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7214 declare i32 @llvm.read_register.i32(metadata)
7215 declare i64 @llvm.read_register.i64(metadata)
7216 declare void @llvm.write_register.i32(metadata, i32 @value)
7217 declare void @llvm.write_register.i64(metadata, i64 @value)
7218 !0 = metadata !{metadata !"sp\00"}
7223 The '``llvm.read_register``' and '``llvm.write_register``' intrinsics
7224 provides access to the named register. The register must be valid on
7225 the architecture being compiled to. The type needs to be compatible
7226 with the register being read.
7231 The '``llvm.read_register``' intrinsic returns the current value of the
7232 register, where possible. The '``llvm.write_register``' intrinsic sets
7233 the current value of the register, where possible.
7235 This is useful to implement named register global variables that need
7236 to always be mapped to a specific register, as is common practice on
7237 bare-metal programs including OS kernels.
7239 The compiler doesn't check for register availability or use of the used
7240 register in surrounding code, including inline assembly. Because of that,
7241 allocatable registers are not supported.
7243 Warning: So far it only works with the stack pointer on selected
7244 architectures (ARM, AArch64, PowerPC and x86_64). Significant amount of
7245 work is needed to support other registers and even more so, allocatable
7250 '``llvm.stacksave``' Intrinsic
7251 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7258 declare i8* @llvm.stacksave()
7263 The '``llvm.stacksave``' intrinsic is used to remember the current state
7264 of the function stack, for use with
7265 :ref:`llvm.stackrestore <int_stackrestore>`. This is useful for
7266 implementing language features like scoped automatic variable sized
7272 This intrinsic returns a opaque pointer value that can be passed to
7273 :ref:`llvm.stackrestore <int_stackrestore>`. When an
7274 ``llvm.stackrestore`` intrinsic is executed with a value saved from
7275 ``llvm.stacksave``, it effectively restores the state of the stack to
7276 the state it was in when the ``llvm.stacksave`` intrinsic executed. In
7277 practice, this pops any :ref:`alloca <i_alloca>` blocks from the stack that
7278 were allocated after the ``llvm.stacksave`` was executed.
7280 .. _int_stackrestore:
7282 '``llvm.stackrestore``' Intrinsic
7283 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7290 declare void @llvm.stackrestore(i8* %ptr)
7295 The '``llvm.stackrestore``' intrinsic is used to restore the state of
7296 the function stack to the state it was in when the corresponding
7297 :ref:`llvm.stacksave <int_stacksave>` intrinsic executed. This is
7298 useful for implementing language features like scoped automatic variable
7299 sized arrays in C99.
7304 See the description for :ref:`llvm.stacksave <int_stacksave>`.
7306 '``llvm.prefetch``' Intrinsic
7307 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7314 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
7319 The '``llvm.prefetch``' intrinsic is a hint to the code generator to
7320 insert a prefetch instruction if supported; otherwise, it is a noop.
7321 Prefetches have no effect on the behavior of the program but can change
7322 its performance characteristics.
7327 ``address`` is the address to be prefetched, ``rw`` is the specifier
7328 determining if the fetch should be for a read (0) or write (1), and
7329 ``locality`` is a temporal locality specifier ranging from (0) - no
7330 locality, to (3) - extremely local keep in cache. The ``cache type``
7331 specifies whether the prefetch is performed on the data (1) or
7332 instruction (0) cache. The ``rw``, ``locality`` and ``cache type``
7333 arguments must be constant integers.
7338 This intrinsic does not modify the behavior of the program. In
7339 particular, prefetches cannot trap and do not produce a value. On
7340 targets that support this intrinsic, the prefetch can provide hints to
7341 the processor cache for better performance.
7343 '``llvm.pcmarker``' Intrinsic
7344 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7351 declare void @llvm.pcmarker(i32 <id>)
7356 The '``llvm.pcmarker``' intrinsic is a method to export a Program
7357 Counter (PC) in a region of code to simulators and other tools. The
7358 method is target specific, but it is expected that the marker will use
7359 exported symbols to transmit the PC of the marker. The marker makes no
7360 guarantees that it will remain with any specific instruction after
7361 optimizations. It is possible that the presence of a marker will inhibit
7362 optimizations. The intended use is to be inserted after optimizations to
7363 allow correlations of simulation runs.
7368 ``id`` is a numerical id identifying the marker.
7373 This intrinsic does not modify the behavior of the program. Backends
7374 that do not support this intrinsic may ignore it.
7376 '``llvm.readcyclecounter``' Intrinsic
7377 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7384 declare i64 @llvm.readcyclecounter()
7389 The '``llvm.readcyclecounter``' intrinsic provides access to the cycle
7390 counter register (or similar low latency, high accuracy clocks) on those
7391 targets that support it. On X86, it should map to RDTSC. On Alpha, it
7392 should map to RPCC. As the backing counters overflow quickly (on the
7393 order of 9 seconds on alpha), this should only be used for small
7399 When directly supported, reading the cycle counter should not modify any
7400 memory. Implementations are allowed to either return a application
7401 specific value or a system wide value. On backends without support, this
7402 is lowered to a constant 0.
7404 Note that runtime support may be conditional on the privilege-level code is
7405 running at and the host platform.
7407 '``llvm.clear_cache``' Intrinsic
7408 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7415 declare void @llvm.clear_cache(i8*, i8*)
7420 The '``llvm.clear_cache``' intrinsic ensures visibility of modifications
7421 in the specified range to the execution unit of the processor. On
7422 targets with non-unified instruction and data cache, the implementation
7423 flushes the instruction cache.
7428 On platforms with coherent instruction and data caches (e.g. x86), this
7429 intrinsic is a nop. On platforms with non-coherent instruction and data
7430 cache (e.g. ARM, MIPS), the intrinsic is lowered either to appropriate
7431 instructions or a system call, if cache flushing requires special
7434 The default behavior is to emit a call to ``__clear_cache`` from the run
7437 This instrinsic does *not* empty the instruction pipeline. Modifications
7438 of the current function are outside the scope of the intrinsic.
7440 Standard C Library Intrinsics
7441 -----------------------------
7443 LLVM provides intrinsics for a few important standard C library
7444 functions. These intrinsics allow source-language front-ends to pass
7445 information about the alignment of the pointer arguments to the code
7446 generator, providing opportunity for more efficient code generation.
7450 '``llvm.memcpy``' Intrinsic
7451 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7456 This is an overloaded intrinsic. You can use ``llvm.memcpy`` on any
7457 integer bit width and for different address spaces. Not all targets
7458 support all bit widths however.
7462 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
7463 i32 <len>, i32 <align>, i1 <isvolatile>)
7464 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
7465 i64 <len>, i32 <align>, i1 <isvolatile>)
7470 The '``llvm.memcpy.*``' intrinsics copy a block of memory from the
7471 source location to the destination location.
7473 Note that, unlike the standard libc function, the ``llvm.memcpy.*``
7474 intrinsics do not return a value, takes extra alignment/isvolatile
7475 arguments and the pointers can be in specified address spaces.
7480 The first argument is a pointer to the destination, the second is a
7481 pointer to the source. The third argument is an integer argument
7482 specifying the number of bytes to copy, the fourth argument is the
7483 alignment of the source and destination locations, and the fifth is a
7484 boolean indicating a volatile access.
7486 If the call to this intrinsic has an alignment value that is not 0 or 1,
7487 then the caller guarantees that both the source and destination pointers
7488 are aligned to that boundary.
7490 If the ``isvolatile`` parameter is ``true``, the ``llvm.memcpy`` call is
7491 a :ref:`volatile operation <volatile>`. The detailed access behavior is not
7492 very cleanly specified and it is unwise to depend on it.
7497 The '``llvm.memcpy.*``' intrinsics copy a block of memory from the
7498 source location to the destination location, which are not allowed to
7499 overlap. It copies "len" bytes of memory over. If the argument is known
7500 to be aligned to some boundary, this can be specified as the fourth
7501 argument, otherwise it should be set to 0 or 1 (both meaning no alignment).
7503 '``llvm.memmove``' Intrinsic
7504 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7509 This is an overloaded intrinsic. You can use llvm.memmove on any integer
7510 bit width and for different address space. Not all targets support all
7515 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
7516 i32 <len>, i32 <align>, i1 <isvolatile>)
7517 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
7518 i64 <len>, i32 <align>, i1 <isvolatile>)
7523 The '``llvm.memmove.*``' intrinsics move a block of memory from the
7524 source location to the destination location. It is similar to the
7525 '``llvm.memcpy``' intrinsic but allows the two memory locations to
7528 Note that, unlike the standard libc function, the ``llvm.memmove.*``
7529 intrinsics do not return a value, takes extra alignment/isvolatile
7530 arguments and the pointers can be in specified address spaces.
7535 The first argument is a pointer to the destination, the second is a
7536 pointer to the source. The third argument is an integer argument
7537 specifying the number of bytes to copy, the fourth argument is the
7538 alignment of the source and destination locations, and the fifth is a
7539 boolean indicating a volatile access.
7541 If the call to this intrinsic has an alignment value that is not 0 or 1,
7542 then the caller guarantees that the source and destination pointers are
7543 aligned to that boundary.
7545 If the ``isvolatile`` parameter is ``true``, the ``llvm.memmove`` call
7546 is a :ref:`volatile operation <volatile>`. The detailed access behavior is
7547 not very cleanly specified and it is unwise to depend on it.
7552 The '``llvm.memmove.*``' intrinsics copy a block of memory from the
7553 source location to the destination location, which may overlap. It
7554 copies "len" bytes of memory over. If the argument is known to be
7555 aligned to some boundary, this can be specified as the fourth argument,
7556 otherwise it should be set to 0 or 1 (both meaning no alignment).
7558 '``llvm.memset.*``' Intrinsics
7559 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7564 This is an overloaded intrinsic. You can use llvm.memset on any integer
7565 bit width and for different address spaces. However, not all targets
7566 support all bit widths.
7570 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
7571 i32 <len>, i32 <align>, i1 <isvolatile>)
7572 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
7573 i64 <len>, i32 <align>, i1 <isvolatile>)
7578 The '``llvm.memset.*``' intrinsics fill a block of memory with a
7579 particular byte value.
7581 Note that, unlike the standard libc function, the ``llvm.memset``
7582 intrinsic does not return a value and takes extra alignment/volatile
7583 arguments. Also, the destination can be in an arbitrary address space.
7588 The first argument is a pointer to the destination to fill, the second
7589 is the byte value with which to fill it, the third argument is an
7590 integer argument specifying the number of bytes to fill, and the fourth
7591 argument is the known alignment of the destination location.
7593 If the call to this intrinsic has an alignment value that is not 0 or 1,
7594 then the caller guarantees that the destination pointer is aligned to
7597 If the ``isvolatile`` parameter is ``true``, the ``llvm.memset`` call is
7598 a :ref:`volatile operation <volatile>`. The detailed access behavior is not
7599 very cleanly specified and it is unwise to depend on it.
7604 The '``llvm.memset.*``' intrinsics fill "len" bytes of memory starting
7605 at the destination location. If the argument is known to be aligned to
7606 some boundary, this can be specified as the fourth argument, otherwise
7607 it should be set to 0 or 1 (both meaning no alignment).
7609 '``llvm.sqrt.*``' Intrinsic
7610 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7615 This is an overloaded intrinsic. You can use ``llvm.sqrt`` on any
7616 floating point or vector of floating point type. Not all targets support
7621 declare float @llvm.sqrt.f32(float %Val)
7622 declare double @llvm.sqrt.f64(double %Val)
7623 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
7624 declare fp128 @llvm.sqrt.f128(fp128 %Val)
7625 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
7630 The '``llvm.sqrt``' intrinsics return the sqrt of the specified operand,
7631 returning the same value as the libm '``sqrt``' functions would. Unlike
7632 ``sqrt`` in libm, however, ``llvm.sqrt`` has undefined behavior for
7633 negative numbers other than -0.0 (which allows for better optimization,
7634 because there is no need to worry about errno being set).
7635 ``llvm.sqrt(-0.0)`` is defined to return -0.0 like IEEE sqrt.
7640 The argument and return value are floating point numbers of the same
7646 This function returns the sqrt of the specified operand if it is a
7647 nonnegative floating point number.
7649 '``llvm.powi.*``' Intrinsic
7650 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7655 This is an overloaded intrinsic. You can use ``llvm.powi`` on any
7656 floating point or vector of floating point type. Not all targets support
7661 declare float @llvm.powi.f32(float %Val, i32 %power)
7662 declare double @llvm.powi.f64(double %Val, i32 %power)
7663 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
7664 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
7665 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
7670 The '``llvm.powi.*``' intrinsics return the first operand raised to the
7671 specified (positive or negative) power. The order of evaluation of
7672 multiplications is not defined. When a vector of floating point type is
7673 used, the second argument remains a scalar integer value.
7678 The second argument is an integer power, and the first is a value to
7679 raise to that power.
7684 This function returns the first value raised to the second power with an
7685 unspecified sequence of rounding operations.
7687 '``llvm.sin.*``' Intrinsic
7688 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7693 This is an overloaded intrinsic. You can use ``llvm.sin`` on any
7694 floating point or vector of floating point type. Not all targets support
7699 declare float @llvm.sin.f32(float %Val)
7700 declare double @llvm.sin.f64(double %Val)
7701 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
7702 declare fp128 @llvm.sin.f128(fp128 %Val)
7703 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
7708 The '``llvm.sin.*``' intrinsics return the sine of the operand.
7713 The argument and return value are floating point numbers of the same
7719 This function returns the sine of the specified operand, returning the
7720 same values as the libm ``sin`` functions would, and handles error
7721 conditions in the same way.
7723 '``llvm.cos.*``' Intrinsic
7724 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7729 This is an overloaded intrinsic. You can use ``llvm.cos`` on any
7730 floating point or vector of floating point type. Not all targets support
7735 declare float @llvm.cos.f32(float %Val)
7736 declare double @llvm.cos.f64(double %Val)
7737 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
7738 declare fp128 @llvm.cos.f128(fp128 %Val)
7739 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
7744 The '``llvm.cos.*``' intrinsics return the cosine of the operand.
7749 The argument and return value are floating point numbers of the same
7755 This function returns the cosine of the specified operand, returning the
7756 same values as the libm ``cos`` functions would, and handles error
7757 conditions in the same way.
7759 '``llvm.pow.*``' Intrinsic
7760 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7765 This is an overloaded intrinsic. You can use ``llvm.pow`` on any
7766 floating point or vector of floating point type. Not all targets support
7771 declare float @llvm.pow.f32(float %Val, float %Power)
7772 declare double @llvm.pow.f64(double %Val, double %Power)
7773 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
7774 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
7775 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
7780 The '``llvm.pow.*``' intrinsics return the first operand raised to the
7781 specified (positive or negative) power.
7786 The second argument is a floating point power, and the first is a value
7787 to raise to that power.
7792 This function returns the first value raised to the second power,
7793 returning the same values as the libm ``pow`` functions would, and
7794 handles error conditions in the same way.
7796 '``llvm.exp.*``' Intrinsic
7797 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7802 This is an overloaded intrinsic. You can use ``llvm.exp`` on any
7803 floating point or vector of floating point type. Not all targets support
7808 declare float @llvm.exp.f32(float %Val)
7809 declare double @llvm.exp.f64(double %Val)
7810 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
7811 declare fp128 @llvm.exp.f128(fp128 %Val)
7812 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
7817 The '``llvm.exp.*``' intrinsics perform the exp function.
7822 The argument and return value are floating point numbers of the same
7828 This function returns the same values as the libm ``exp`` functions
7829 would, and handles error conditions in the same way.
7831 '``llvm.exp2.*``' Intrinsic
7832 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7837 This is an overloaded intrinsic. You can use ``llvm.exp2`` on any
7838 floating point or vector of floating point type. Not all targets support
7843 declare float @llvm.exp2.f32(float %Val)
7844 declare double @llvm.exp2.f64(double %Val)
7845 declare x86_fp80 @llvm.exp2.f80(x86_fp80 %Val)
7846 declare fp128 @llvm.exp2.f128(fp128 %Val)
7847 declare ppc_fp128 @llvm.exp2.ppcf128(ppc_fp128 %Val)
7852 The '``llvm.exp2.*``' intrinsics perform the exp2 function.
7857 The argument and return value are floating point numbers of the same
7863 This function returns the same values as the libm ``exp2`` functions
7864 would, and handles error conditions in the same way.
7866 '``llvm.log.*``' Intrinsic
7867 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7872 This is an overloaded intrinsic. You can use ``llvm.log`` on any
7873 floating point or vector of floating point type. Not all targets support
7878 declare float @llvm.log.f32(float %Val)
7879 declare double @llvm.log.f64(double %Val)
7880 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
7881 declare fp128 @llvm.log.f128(fp128 %Val)
7882 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
7887 The '``llvm.log.*``' intrinsics perform the log function.
7892 The argument and return value are floating point numbers of the same
7898 This function returns the same values as the libm ``log`` functions
7899 would, and handles error conditions in the same way.
7901 '``llvm.log10.*``' Intrinsic
7902 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7907 This is an overloaded intrinsic. You can use ``llvm.log10`` on any
7908 floating point or vector of floating point type. Not all targets support
7913 declare float @llvm.log10.f32(float %Val)
7914 declare double @llvm.log10.f64(double %Val)
7915 declare x86_fp80 @llvm.log10.f80(x86_fp80 %Val)
7916 declare fp128 @llvm.log10.f128(fp128 %Val)
7917 declare ppc_fp128 @llvm.log10.ppcf128(ppc_fp128 %Val)
7922 The '``llvm.log10.*``' intrinsics perform the log10 function.
7927 The argument and return value are floating point numbers of the same
7933 This function returns the same values as the libm ``log10`` functions
7934 would, and handles error conditions in the same way.
7936 '``llvm.log2.*``' Intrinsic
7937 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7942 This is an overloaded intrinsic. You can use ``llvm.log2`` on any
7943 floating point or vector of floating point type. Not all targets support
7948 declare float @llvm.log2.f32(float %Val)
7949 declare double @llvm.log2.f64(double %Val)
7950 declare x86_fp80 @llvm.log2.f80(x86_fp80 %Val)
7951 declare fp128 @llvm.log2.f128(fp128 %Val)
7952 declare ppc_fp128 @llvm.log2.ppcf128(ppc_fp128 %Val)
7957 The '``llvm.log2.*``' intrinsics perform the log2 function.
7962 The argument and return value are floating point numbers of the same
7968 This function returns the same values as the libm ``log2`` functions
7969 would, and handles error conditions in the same way.
7971 '``llvm.fma.*``' Intrinsic
7972 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7977 This is an overloaded intrinsic. You can use ``llvm.fma`` on any
7978 floating point or vector of floating point type. Not all targets support
7983 declare float @llvm.fma.f32(float %a, float %b, float %c)
7984 declare double @llvm.fma.f64(double %a, double %b, double %c)
7985 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
7986 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
7987 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
7992 The '``llvm.fma.*``' intrinsics perform the fused multiply-add
7998 The argument and return value are floating point numbers of the same
8004 This function returns the same values as the libm ``fma`` functions
8005 would, and does not set errno.
8007 '``llvm.fabs.*``' Intrinsic
8008 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8013 This is an overloaded intrinsic. You can use ``llvm.fabs`` on any
8014 floating point or vector of floating point type. Not all targets support
8019 declare float @llvm.fabs.f32(float %Val)
8020 declare double @llvm.fabs.f64(double %Val)
8021 declare x86_fp80 @llvm.fabs.f80(x86_fp80 %Val)
8022 declare fp128 @llvm.fabs.f128(fp128 %Val)
8023 declare ppc_fp128 @llvm.fabs.ppcf128(ppc_fp128 %Val)
8028 The '``llvm.fabs.*``' intrinsics return the absolute value of the
8034 The argument and return value are floating point numbers of the same
8040 This function returns the same values as the libm ``fabs`` functions
8041 would, and handles error conditions in the same way.
8043 '``llvm.copysign.*``' Intrinsic
8044 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8049 This is an overloaded intrinsic. You can use ``llvm.copysign`` on any
8050 floating point or vector of floating point type. Not all targets support
8055 declare float @llvm.copysign.f32(float %Mag, float %Sgn)
8056 declare double @llvm.copysign.f64(double %Mag, double %Sgn)
8057 declare x86_fp80 @llvm.copysign.f80(x86_fp80 %Mag, x86_fp80 %Sgn)
8058 declare fp128 @llvm.copysign.f128(fp128 %Mag, fp128 %Sgn)
8059 declare ppc_fp128 @llvm.copysign.ppcf128(ppc_fp128 %Mag, ppc_fp128 %Sgn)
8064 The '``llvm.copysign.*``' intrinsics return a value with the magnitude of the
8065 first operand and the sign of the second operand.
8070 The arguments and return value are floating point numbers of the same
8076 This function returns the same values as the libm ``copysign``
8077 functions would, and handles error conditions in the same way.
8079 '``llvm.floor.*``' Intrinsic
8080 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8085 This is an overloaded intrinsic. You can use ``llvm.floor`` on any
8086 floating point or vector of floating point type. Not all targets support
8091 declare float @llvm.floor.f32(float %Val)
8092 declare double @llvm.floor.f64(double %Val)
8093 declare x86_fp80 @llvm.floor.f80(x86_fp80 %Val)
8094 declare fp128 @llvm.floor.f128(fp128 %Val)
8095 declare ppc_fp128 @llvm.floor.ppcf128(ppc_fp128 %Val)
8100 The '``llvm.floor.*``' intrinsics return the floor of the operand.
8105 The argument and return value are floating point numbers of the same
8111 This function returns the same values as the libm ``floor`` functions
8112 would, and handles error conditions in the same way.
8114 '``llvm.ceil.*``' Intrinsic
8115 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8120 This is an overloaded intrinsic. You can use ``llvm.ceil`` on any
8121 floating point or vector of floating point type. Not all targets support
8126 declare float @llvm.ceil.f32(float %Val)
8127 declare double @llvm.ceil.f64(double %Val)
8128 declare x86_fp80 @llvm.ceil.f80(x86_fp80 %Val)
8129 declare fp128 @llvm.ceil.f128(fp128 %Val)
8130 declare ppc_fp128 @llvm.ceil.ppcf128(ppc_fp128 %Val)
8135 The '``llvm.ceil.*``' intrinsics return the ceiling of the operand.
8140 The argument and return value are floating point numbers of the same
8146 This function returns the same values as the libm ``ceil`` functions
8147 would, and handles error conditions in the same way.
8149 '``llvm.trunc.*``' Intrinsic
8150 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8155 This is an overloaded intrinsic. You can use ``llvm.trunc`` on any
8156 floating point or vector of floating point type. Not all targets support
8161 declare float @llvm.trunc.f32(float %Val)
8162 declare double @llvm.trunc.f64(double %Val)
8163 declare x86_fp80 @llvm.trunc.f80(x86_fp80 %Val)
8164 declare fp128 @llvm.trunc.f128(fp128 %Val)
8165 declare ppc_fp128 @llvm.trunc.ppcf128(ppc_fp128 %Val)
8170 The '``llvm.trunc.*``' intrinsics returns the operand rounded to the
8171 nearest integer not larger in magnitude than the operand.
8176 The argument and return value are floating point numbers of the same
8182 This function returns the same values as the libm ``trunc`` functions
8183 would, and handles error conditions in the same way.
8185 '``llvm.rint.*``' Intrinsic
8186 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8191 This is an overloaded intrinsic. You can use ``llvm.rint`` on any
8192 floating point or vector of floating point type. Not all targets support
8197 declare float @llvm.rint.f32(float %Val)
8198 declare double @llvm.rint.f64(double %Val)
8199 declare x86_fp80 @llvm.rint.f80(x86_fp80 %Val)
8200 declare fp128 @llvm.rint.f128(fp128 %Val)
8201 declare ppc_fp128 @llvm.rint.ppcf128(ppc_fp128 %Val)
8206 The '``llvm.rint.*``' intrinsics returns the operand rounded to the
8207 nearest integer. It may raise an inexact floating-point exception if the
8208 operand isn't an integer.
8213 The argument and return value are floating point numbers of the same
8219 This function returns the same values as the libm ``rint`` functions
8220 would, and handles error conditions in the same way.
8222 '``llvm.nearbyint.*``' Intrinsic
8223 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8228 This is an overloaded intrinsic. You can use ``llvm.nearbyint`` on any
8229 floating point or vector of floating point type. Not all targets support
8234 declare float @llvm.nearbyint.f32(float %Val)
8235 declare double @llvm.nearbyint.f64(double %Val)
8236 declare x86_fp80 @llvm.nearbyint.f80(x86_fp80 %Val)
8237 declare fp128 @llvm.nearbyint.f128(fp128 %Val)
8238 declare ppc_fp128 @llvm.nearbyint.ppcf128(ppc_fp128 %Val)
8243 The '``llvm.nearbyint.*``' intrinsics returns the operand rounded to the
8249 The argument and return value are floating point numbers of the same
8255 This function returns the same values as the libm ``nearbyint``
8256 functions would, and handles error conditions in the same way.
8258 '``llvm.round.*``' Intrinsic
8259 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8264 This is an overloaded intrinsic. You can use ``llvm.round`` on any
8265 floating point or vector of floating point type. Not all targets support
8270 declare float @llvm.round.f32(float %Val)
8271 declare double @llvm.round.f64(double %Val)
8272 declare x86_fp80 @llvm.round.f80(x86_fp80 %Val)
8273 declare fp128 @llvm.round.f128(fp128 %Val)
8274 declare ppc_fp128 @llvm.round.ppcf128(ppc_fp128 %Val)
8279 The '``llvm.round.*``' intrinsics returns the operand rounded to the
8285 The argument and return value are floating point numbers of the same
8291 This function returns the same values as the libm ``round``
8292 functions would, and handles error conditions in the same way.
8294 Bit Manipulation Intrinsics
8295 ---------------------------
8297 LLVM provides intrinsics for a few important bit manipulation
8298 operations. These allow efficient code generation for some algorithms.
8300 '``llvm.bswap.*``' Intrinsics
8301 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8306 This is an overloaded intrinsic function. You can use bswap on any
8307 integer type that is an even number of bytes (i.e. BitWidth % 16 == 0).
8311 declare i16 @llvm.bswap.i16(i16 <id>)
8312 declare i32 @llvm.bswap.i32(i32 <id>)
8313 declare i64 @llvm.bswap.i64(i64 <id>)
8318 The '``llvm.bswap``' family of intrinsics is used to byte swap integer
8319 values with an even number of bytes (positive multiple of 16 bits).
8320 These are useful for performing operations on data that is not in the
8321 target's native byte order.
8326 The ``llvm.bswap.i16`` intrinsic returns an i16 value that has the high
8327 and low byte of the input i16 swapped. Similarly, the ``llvm.bswap.i32``
8328 intrinsic returns an i32 value that has the four bytes of the input i32
8329 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the
8330 returned i32 will have its bytes in 3, 2, 1, 0 order. The
8331 ``llvm.bswap.i48``, ``llvm.bswap.i64`` and other intrinsics extend this
8332 concept to additional even-byte lengths (6 bytes, 8 bytes and more,
8335 '``llvm.ctpop.*``' Intrinsic
8336 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8341 This is an overloaded intrinsic. You can use llvm.ctpop on any integer
8342 bit width, or on any vector with integer elements. Not all targets
8343 support all bit widths or vector types, however.
8347 declare i8 @llvm.ctpop.i8(i8 <src>)
8348 declare i16 @llvm.ctpop.i16(i16 <src>)
8349 declare i32 @llvm.ctpop.i32(i32 <src>)
8350 declare i64 @llvm.ctpop.i64(i64 <src>)
8351 declare i256 @llvm.ctpop.i256(i256 <src>)
8352 declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
8357 The '``llvm.ctpop``' family of intrinsics counts the number of bits set
8363 The only argument is the value to be counted. The argument may be of any
8364 integer type, or a vector with integer elements. The return type must
8365 match the argument type.
8370 The '``llvm.ctpop``' intrinsic counts the 1's in a variable, or within
8371 each element of a vector.
8373 '``llvm.ctlz.*``' Intrinsic
8374 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8379 This is an overloaded intrinsic. You can use ``llvm.ctlz`` on any
8380 integer bit width, or any vector whose elements are integers. Not all
8381 targets support all bit widths or vector types, however.
8385 declare i8 @llvm.ctlz.i8 (i8 <src>, i1 <is_zero_undef>)
8386 declare i16 @llvm.ctlz.i16 (i16 <src>, i1 <is_zero_undef>)
8387 declare i32 @llvm.ctlz.i32 (i32 <src>, i1 <is_zero_undef>)
8388 declare i64 @llvm.ctlz.i64 (i64 <src>, i1 <is_zero_undef>)
8389 declare i256 @llvm.ctlz.i256(i256 <src>, i1 <is_zero_undef>)
8390 declase <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
8395 The '``llvm.ctlz``' family of intrinsic functions counts the number of
8396 leading zeros in a variable.
8401 The first argument is the value to be counted. This argument may be of
8402 any integer type, or a vectory with integer element type. The return
8403 type must match the first argument type.
8405 The second argument must be a constant and is a flag to indicate whether
8406 the intrinsic should ensure that a zero as the first argument produces a
8407 defined result. Historically some architectures did not provide a
8408 defined result for zero values as efficiently, and many algorithms are
8409 now predicated on avoiding zero-value inputs.
8414 The '``llvm.ctlz``' intrinsic counts the leading (most significant)
8415 zeros in a variable, or within each element of the vector. If
8416 ``src == 0`` then the result is the size in bits of the type of ``src``
8417 if ``is_zero_undef == 0`` and ``undef`` otherwise. For example,
8418 ``llvm.ctlz(i32 2) = 30``.
8420 '``llvm.cttz.*``' Intrinsic
8421 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8426 This is an overloaded intrinsic. You can use ``llvm.cttz`` on any
8427 integer bit width, or any vector of integer elements. Not all targets
8428 support all bit widths or vector types, however.
8432 declare i8 @llvm.cttz.i8 (i8 <src>, i1 <is_zero_undef>)
8433 declare i16 @llvm.cttz.i16 (i16 <src>, i1 <is_zero_undef>)
8434 declare i32 @llvm.cttz.i32 (i32 <src>, i1 <is_zero_undef>)
8435 declare i64 @llvm.cttz.i64 (i64 <src>, i1 <is_zero_undef>)
8436 declare i256 @llvm.cttz.i256(i256 <src>, i1 <is_zero_undef>)
8437 declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
8442 The '``llvm.cttz``' family of intrinsic functions counts the number of
8448 The first argument is the value to be counted. This argument may be of
8449 any integer type, or a vectory with integer element type. The return
8450 type must match the first argument type.
8452 The second argument must be a constant and is a flag to indicate whether
8453 the intrinsic should ensure that a zero as the first argument produces a
8454 defined result. Historically some architectures did not provide a
8455 defined result for zero values as efficiently, and many algorithms are
8456 now predicated on avoiding zero-value inputs.
8461 The '``llvm.cttz``' intrinsic counts the trailing (least significant)
8462 zeros in a variable, or within each element of a vector. If ``src == 0``
8463 then the result is the size in bits of the type of ``src`` if
8464 ``is_zero_undef == 0`` and ``undef`` otherwise. For example,
8465 ``llvm.cttz(2) = 1``.
8467 Arithmetic with Overflow Intrinsics
8468 -----------------------------------
8470 LLVM provides intrinsics for some arithmetic with overflow operations.
8472 '``llvm.sadd.with.overflow.*``' Intrinsics
8473 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8478 This is an overloaded intrinsic. You can use ``llvm.sadd.with.overflow``
8479 on any integer bit width.
8483 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
8484 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
8485 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
8490 The '``llvm.sadd.with.overflow``' family of intrinsic functions perform
8491 a signed addition of the two arguments, and indicate whether an overflow
8492 occurred during the signed summation.
8497 The arguments (%a and %b) and the first element of the result structure
8498 may be of integer types of any bit width, but they must have the same
8499 bit width. The second element of the result structure must be of type
8500 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
8506 The '``llvm.sadd.with.overflow``' family of intrinsic functions perform
8507 a signed addition of the two variables. They return a structure --- the
8508 first element of which is the signed summation, and the second element
8509 of which is a bit specifying if the signed summation resulted in an
8515 .. code-block:: llvm
8517 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
8518 %sum = extractvalue {i32, i1} %res, 0
8519 %obit = extractvalue {i32, i1} %res, 1
8520 br i1 %obit, label %overflow, label %normal
8522 '``llvm.uadd.with.overflow.*``' Intrinsics
8523 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8528 This is an overloaded intrinsic. You can use ``llvm.uadd.with.overflow``
8529 on any integer bit width.
8533 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
8534 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
8535 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
8540 The '``llvm.uadd.with.overflow``' family of intrinsic functions perform
8541 an unsigned addition of the two arguments, and indicate whether a carry
8542 occurred during the unsigned summation.
8547 The arguments (%a and %b) and the first element of the result structure
8548 may be of integer types of any bit width, but they must have the same
8549 bit width. The second element of the result structure must be of type
8550 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
8556 The '``llvm.uadd.with.overflow``' family of intrinsic functions perform
8557 an unsigned addition of the two arguments. They return a structure --- the
8558 first element of which is the sum, and the second element of which is a
8559 bit specifying if the unsigned summation resulted in a carry.
8564 .. code-block:: llvm
8566 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
8567 %sum = extractvalue {i32, i1} %res, 0
8568 %obit = extractvalue {i32, i1} %res, 1
8569 br i1 %obit, label %carry, label %normal
8571 '``llvm.ssub.with.overflow.*``' Intrinsics
8572 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8577 This is an overloaded intrinsic. You can use ``llvm.ssub.with.overflow``
8578 on any integer bit width.
8582 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
8583 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
8584 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
8589 The '``llvm.ssub.with.overflow``' family of intrinsic functions perform
8590 a signed subtraction of the two arguments, and indicate whether an
8591 overflow occurred during the signed subtraction.
8596 The arguments (%a and %b) and the first element of the result structure
8597 may be of integer types of any bit width, but they must have the same
8598 bit width. The second element of the result structure must be of type
8599 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
8605 The '``llvm.ssub.with.overflow``' family of intrinsic functions perform
8606 a signed subtraction of the two arguments. They return a structure --- the
8607 first element of which is the subtraction, and the second element of
8608 which is a bit specifying if the signed subtraction resulted in an
8614 .. code-block:: llvm
8616 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
8617 %sum = extractvalue {i32, i1} %res, 0
8618 %obit = extractvalue {i32, i1} %res, 1
8619 br i1 %obit, label %overflow, label %normal
8621 '``llvm.usub.with.overflow.*``' Intrinsics
8622 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8627 This is an overloaded intrinsic. You can use ``llvm.usub.with.overflow``
8628 on any integer bit width.
8632 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
8633 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
8634 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
8639 The '``llvm.usub.with.overflow``' family of intrinsic functions perform
8640 an unsigned subtraction of the two arguments, and indicate whether an
8641 overflow occurred during the unsigned subtraction.
8646 The arguments (%a and %b) and the first element of the result structure
8647 may be of integer types of any bit width, but they must have the same
8648 bit width. The second element of the result structure must be of type
8649 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
8655 The '``llvm.usub.with.overflow``' family of intrinsic functions perform
8656 an unsigned subtraction of the two arguments. They return a structure ---
8657 the first element of which is the subtraction, and the second element of
8658 which is a bit specifying if the unsigned subtraction resulted in an
8664 .. code-block:: llvm
8666 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
8667 %sum = extractvalue {i32, i1} %res, 0
8668 %obit = extractvalue {i32, i1} %res, 1
8669 br i1 %obit, label %overflow, label %normal
8671 '``llvm.smul.with.overflow.*``' Intrinsics
8672 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8677 This is an overloaded intrinsic. You can use ``llvm.smul.with.overflow``
8678 on any integer bit width.
8682 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
8683 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
8684 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
8689 The '``llvm.smul.with.overflow``' family of intrinsic functions perform
8690 a signed multiplication of the two arguments, and indicate whether an
8691 overflow occurred during the signed multiplication.
8696 The arguments (%a and %b) and the first element of the result structure
8697 may be of integer types of any bit width, but they must have the same
8698 bit width. The second element of the result structure must be of type
8699 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
8705 The '``llvm.smul.with.overflow``' family of intrinsic functions perform
8706 a signed multiplication of the two arguments. They return a structure ---
8707 the first element of which is the multiplication, and the second element
8708 of which is a bit specifying if the signed multiplication resulted in an
8714 .. code-block:: llvm
8716 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
8717 %sum = extractvalue {i32, i1} %res, 0
8718 %obit = extractvalue {i32, i1} %res, 1
8719 br i1 %obit, label %overflow, label %normal
8721 '``llvm.umul.with.overflow.*``' Intrinsics
8722 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8727 This is an overloaded intrinsic. You can use ``llvm.umul.with.overflow``
8728 on any integer bit width.
8732 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
8733 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
8734 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
8739 The '``llvm.umul.with.overflow``' family of intrinsic functions perform
8740 a unsigned multiplication of the two arguments, and indicate whether an
8741 overflow occurred during the unsigned multiplication.
8746 The arguments (%a and %b) and the first element of the result structure
8747 may be of integer types of any bit width, but they must have the same
8748 bit width. The second element of the result structure must be of type
8749 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
8755 The '``llvm.umul.with.overflow``' family of intrinsic functions perform
8756 an unsigned multiplication of the two arguments. They return a structure ---
8757 the first element of which is the multiplication, and the second
8758 element of which is a bit specifying if the unsigned multiplication
8759 resulted in an overflow.
8764 .. code-block:: llvm
8766 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
8767 %sum = extractvalue {i32, i1} %res, 0
8768 %obit = extractvalue {i32, i1} %res, 1
8769 br i1 %obit, label %overflow, label %normal
8771 Specialised Arithmetic Intrinsics
8772 ---------------------------------
8774 '``llvm.fmuladd.*``' Intrinsic
8775 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8782 declare float @llvm.fmuladd.f32(float %a, float %b, float %c)
8783 declare double @llvm.fmuladd.f64(double %a, double %b, double %c)
8788 The '``llvm.fmuladd.*``' intrinsic functions represent multiply-add
8789 expressions that can be fused if the code generator determines that (a) the
8790 target instruction set has support for a fused operation, and (b) that the
8791 fused operation is more efficient than the equivalent, separate pair of mul
8792 and add instructions.
8797 The '``llvm.fmuladd.*``' intrinsics each take three arguments: two
8798 multiplicands, a and b, and an addend c.
8807 %0 = call float @llvm.fmuladd.f32(%a, %b, %c)
8809 is equivalent to the expression a \* b + c, except that rounding will
8810 not be performed between the multiplication and addition steps if the
8811 code generator fuses the operations. Fusion is not guaranteed, even if
8812 the target platform supports it. If a fused multiply-add is required the
8813 corresponding llvm.fma.\* intrinsic function should be used
8814 instead. This never sets errno, just as '``llvm.fma.*``'.
8819 .. code-block:: llvm
8821 %r2 = call float @llvm.fmuladd.f32(float %a, float %b, float %c) ; yields float:r2 = (a * b) + c
8823 Half Precision Floating Point Intrinsics
8824 ----------------------------------------
8826 For most target platforms, half precision floating point is a
8827 storage-only format. This means that it is a dense encoding (in memory)
8828 but does not support computation in the format.
8830 This means that code must first load the half-precision floating point
8831 value as an i16, then convert it to float with
8832 :ref:`llvm.convert.from.fp16 <int_convert_from_fp16>`. Computation can
8833 then be performed on the float value (including extending to double
8834 etc). To store the value back to memory, it is first converted to float
8835 if needed, then converted to i16 with
8836 :ref:`llvm.convert.to.fp16 <int_convert_to_fp16>`, then storing as an
8839 .. _int_convert_to_fp16:
8841 '``llvm.convert.to.fp16``' Intrinsic
8842 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8849 declare i16 @llvm.convert.to.fp16.f32(float %a)
8850 declare i16 @llvm.convert.to.fp16.f64(double %a)
8855 The '``llvm.convert.to.fp16``' intrinsic function performs a conversion from a
8856 conventional floating point type to half precision floating point format.
8861 The intrinsic function contains single argument - the value to be
8867 The '``llvm.convert.to.fp16``' intrinsic function performs a conversion from a
8868 conventional floating point format to half precision floating point format. The
8869 return value is an ``i16`` which contains the converted number.
8874 .. code-block:: llvm
8876 %res = call i16 @llvm.convert.to.fp16.f32(float %a)
8877 store i16 %res, i16* @x, align 2
8879 .. _int_convert_from_fp16:
8881 '``llvm.convert.from.fp16``' Intrinsic
8882 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8889 declare float @llvm.convert.from.fp16.f32(i16 %a)
8890 declare double @llvm.convert.from.fp16.f64(i16 %a)
8895 The '``llvm.convert.from.fp16``' intrinsic function performs a
8896 conversion from half precision floating point format to single precision
8897 floating point format.
8902 The intrinsic function contains single argument - the value to be
8908 The '``llvm.convert.from.fp16``' intrinsic function performs a
8909 conversion from half single precision floating point format to single
8910 precision floating point format. The input half-float value is
8911 represented by an ``i16`` value.
8916 .. code-block:: llvm
8918 %a = load i16* @x, align 2
8919 %res = call float @llvm.convert.from.fp16(i16 %a)
8924 The LLVM debugger intrinsics (which all start with ``llvm.dbg.``
8925 prefix), are described in the `LLVM Source Level
8926 Debugging <SourceLevelDebugging.html#format_common_intrinsics>`_
8929 Exception Handling Intrinsics
8930 -----------------------------
8932 The LLVM exception handling intrinsics (which all start with
8933 ``llvm.eh.`` prefix), are described in the `LLVM Exception
8934 Handling <ExceptionHandling.html#format_common_intrinsics>`_ document.
8938 Trampoline Intrinsics
8939 ---------------------
8941 These intrinsics make it possible to excise one parameter, marked with
8942 the :ref:`nest <nest>` attribute, from a function. The result is a
8943 callable function pointer lacking the nest parameter - the caller does
8944 not need to provide a value for it. Instead, the value to use is stored
8945 in advance in a "trampoline", a block of memory usually allocated on the
8946 stack, which also contains code to splice the nest value into the
8947 argument list. This is used to implement the GCC nested function address
8950 For example, if the function is ``i32 f(i8* nest %c, i32 %x, i32 %y)``
8951 then the resulting function pointer has signature ``i32 (i32, i32)*``.
8952 It can be created as follows:
8954 .. code-block:: llvm
8956 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
8957 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
8958 call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval)
8959 %p = call i8* @llvm.adjust.trampoline(i8* %tramp1)
8960 %fp = bitcast i8* %p to i32 (i32, i32)*
8962 The call ``%val = call i32 %fp(i32 %x, i32 %y)`` is then equivalent to
8963 ``%val = call i32 %f(i8* %nval, i32 %x, i32 %y)``.
8967 '``llvm.init.trampoline``' Intrinsic
8968 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8975 declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
8980 This fills the memory pointed to by ``tramp`` with executable code,
8981 turning it into a trampoline.
8986 The ``llvm.init.trampoline`` intrinsic takes three arguments, all
8987 pointers. The ``tramp`` argument must point to a sufficiently large and
8988 sufficiently aligned block of memory; this memory is written to by the
8989 intrinsic. Note that the size and the alignment are target-specific -
8990 LLVM currently provides no portable way of determining them, so a
8991 front-end that generates this intrinsic needs to have some
8992 target-specific knowledge. The ``func`` argument must hold a function
8993 bitcast to an ``i8*``.
8998 The block of memory pointed to by ``tramp`` is filled with target
8999 dependent code, turning it into a function. Then ``tramp`` needs to be
9000 passed to :ref:`llvm.adjust.trampoline <int_at>` to get a pointer which can
9001 be :ref:`bitcast (to a new function) and called <int_trampoline>`. The new
9002 function's signature is the same as that of ``func`` with any arguments
9003 marked with the ``nest`` attribute removed. At most one such ``nest``
9004 argument is allowed, and it must be of pointer type. Calling the new
9005 function is equivalent to calling ``func`` with the same argument list,
9006 but with ``nval`` used for the missing ``nest`` argument. If, after
9007 calling ``llvm.init.trampoline``, the memory pointed to by ``tramp`` is
9008 modified, then the effect of any later call to the returned function
9009 pointer is undefined.
9013 '``llvm.adjust.trampoline``' Intrinsic
9014 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9021 declare i8* @llvm.adjust.trampoline(i8* <tramp>)
9026 This performs any required machine-specific adjustment to the address of
9027 a trampoline (passed as ``tramp``).
9032 ``tramp`` must point to a block of memory which already has trampoline
9033 code filled in by a previous call to
9034 :ref:`llvm.init.trampoline <int_it>`.
9039 On some architectures the address of the code to be executed needs to be
9040 different than the address where the trampoline is actually stored. This
9041 intrinsic returns the executable address corresponding to ``tramp``
9042 after performing the required machine specific adjustments. The pointer
9043 returned can then be :ref:`bitcast and executed <int_trampoline>`.
9048 This class of intrinsics provides information about the lifetime of
9049 memory objects and ranges where variables are immutable.
9053 '``llvm.lifetime.start``' Intrinsic
9054 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9061 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
9066 The '``llvm.lifetime.start``' intrinsic specifies the start of a memory
9072 The first argument is a constant integer representing the size of the
9073 object, or -1 if it is variable sized. The second argument is a pointer
9079 This intrinsic indicates that before this point in the code, the value
9080 of the memory pointed to by ``ptr`` is dead. This means that it is known
9081 to never be used and has an undefined value. A load from the pointer
9082 that precedes this intrinsic can be replaced with ``'undef'``.
9086 '``llvm.lifetime.end``' Intrinsic
9087 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9094 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
9099 The '``llvm.lifetime.end``' intrinsic specifies the end of a memory
9105 The first argument is a constant integer representing the size of the
9106 object, or -1 if it is variable sized. The second argument is a pointer
9112 This intrinsic indicates that after this point in the code, the value of
9113 the memory pointed to by ``ptr`` is dead. This means that it is known to
9114 never be used and has an undefined value. Any stores into the memory
9115 object following this intrinsic may be removed as dead.
9117 '``llvm.invariant.start``' Intrinsic
9118 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9125 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
9130 The '``llvm.invariant.start``' intrinsic specifies that the contents of
9131 a memory object will not change.
9136 The first argument is a constant integer representing the size of the
9137 object, or -1 if it is variable sized. The second argument is a pointer
9143 This intrinsic indicates that until an ``llvm.invariant.end`` that uses
9144 the return value, the referenced memory location is constant and
9147 '``llvm.invariant.end``' Intrinsic
9148 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9155 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
9160 The '``llvm.invariant.end``' intrinsic specifies that the contents of a
9161 memory object are mutable.
9166 The first argument is the matching ``llvm.invariant.start`` intrinsic.
9167 The second argument is a constant integer representing the size of the
9168 object, or -1 if it is variable sized and the third argument is a
9169 pointer to the object.
9174 This intrinsic indicates that the memory is mutable again.
9179 This class of intrinsics is designed to be generic and has no specific
9182 '``llvm.var.annotation``' Intrinsic
9183 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9190 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
9195 The '``llvm.var.annotation``' intrinsic.
9200 The first argument is a pointer to a value, the second is a pointer to a
9201 global string, the third is a pointer to a global string which is the
9202 source file name, and the last argument is the line number.
9207 This intrinsic allows annotation of local variables with arbitrary
9208 strings. This can be useful for special purpose optimizations that want
9209 to look for these annotations. These have no other defined use; they are
9210 ignored by code generation and optimization.
9212 '``llvm.ptr.annotation.*``' Intrinsic
9213 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9218 This is an overloaded intrinsic. You can use '``llvm.ptr.annotation``' on a
9219 pointer to an integer of any width. *NOTE* you must specify an address space for
9220 the pointer. The identifier for the default address space is the integer
9225 declare i8* @llvm.ptr.annotation.p<address space>i8(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
9226 declare i16* @llvm.ptr.annotation.p<address space>i16(i16* <val>, i8* <str>, i8* <str>, i32 <int>)
9227 declare i32* @llvm.ptr.annotation.p<address space>i32(i32* <val>, i8* <str>, i8* <str>, i32 <int>)
9228 declare i64* @llvm.ptr.annotation.p<address space>i64(i64* <val>, i8* <str>, i8* <str>, i32 <int>)
9229 declare i256* @llvm.ptr.annotation.p<address space>i256(i256* <val>, i8* <str>, i8* <str>, i32 <int>)
9234 The '``llvm.ptr.annotation``' intrinsic.
9239 The first argument is a pointer to an integer value of arbitrary bitwidth
9240 (result of some expression), the second is a pointer to a global string, the
9241 third is a pointer to a global string which is the source file name, and the
9242 last argument is the line number. It returns the value of the first argument.
9247 This intrinsic allows annotation of a pointer to an integer with arbitrary
9248 strings. This can be useful for special purpose optimizations that want to look
9249 for these annotations. These have no other defined use; they are ignored by code
9250 generation and optimization.
9252 '``llvm.annotation.*``' Intrinsic
9253 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9258 This is an overloaded intrinsic. You can use '``llvm.annotation``' on
9259 any integer bit width.
9263 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
9264 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
9265 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
9266 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
9267 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
9272 The '``llvm.annotation``' intrinsic.
9277 The first argument is an integer value (result of some expression), the
9278 second is a pointer to a global string, the third is a pointer to a
9279 global string which is the source file name, and the last argument is
9280 the line number. It returns the value of the first argument.
9285 This intrinsic allows annotations to be put on arbitrary expressions
9286 with arbitrary strings. This can be useful for special purpose
9287 optimizations that want to look for these annotations. These have no
9288 other defined use; they are ignored by code generation and optimization.
9290 '``llvm.trap``' Intrinsic
9291 ^^^^^^^^^^^^^^^^^^^^^^^^^
9298 declare void @llvm.trap() noreturn nounwind
9303 The '``llvm.trap``' intrinsic.
9313 This intrinsic is lowered to the target dependent trap instruction. If
9314 the target does not have a trap instruction, this intrinsic will be
9315 lowered to a call of the ``abort()`` function.
9317 '``llvm.debugtrap``' Intrinsic
9318 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9325 declare void @llvm.debugtrap() nounwind
9330 The '``llvm.debugtrap``' intrinsic.
9340 This intrinsic is lowered to code which is intended to cause an
9341 execution trap with the intention of requesting the attention of a
9344 '``llvm.stackprotector``' Intrinsic
9345 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9352 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
9357 The ``llvm.stackprotector`` intrinsic takes the ``guard`` and stores it
9358 onto the stack at ``slot``. The stack slot is adjusted to ensure that it
9359 is placed on the stack before local variables.
9364 The ``llvm.stackprotector`` intrinsic requires two pointer arguments.
9365 The first argument is the value loaded from the stack guard
9366 ``@__stack_chk_guard``. The second variable is an ``alloca`` that has
9367 enough space to hold the value of the guard.
9372 This intrinsic causes the prologue/epilogue inserter to force the position of
9373 the ``AllocaInst`` stack slot to be before local variables on the stack. This is
9374 to ensure that if a local variable on the stack is overwritten, it will destroy
9375 the value of the guard. When the function exits, the guard on the stack is
9376 checked against the original guard by ``llvm.stackprotectorcheck``. If they are
9377 different, then ``llvm.stackprotectorcheck`` causes the program to abort by
9378 calling the ``__stack_chk_fail()`` function.
9380 '``llvm.stackprotectorcheck``' Intrinsic
9381 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9388 declare void @llvm.stackprotectorcheck(i8** <guard>)
9393 The ``llvm.stackprotectorcheck`` intrinsic compares ``guard`` against an already
9394 created stack protector and if they are not equal calls the
9395 ``__stack_chk_fail()`` function.
9400 The ``llvm.stackprotectorcheck`` intrinsic requires one pointer argument, the
9401 the variable ``@__stack_chk_guard``.
9406 This intrinsic is provided to perform the stack protector check by comparing
9407 ``guard`` with the stack slot created by ``llvm.stackprotector`` and if the
9408 values do not match call the ``__stack_chk_fail()`` function.
9410 The reason to provide this as an IR level intrinsic instead of implementing it
9411 via other IR operations is that in order to perform this operation at the IR
9412 level without an intrinsic, one would need to create additional basic blocks to
9413 handle the success/failure cases. This makes it difficult to stop the stack
9414 protector check from disrupting sibling tail calls in Codegen. With this
9415 intrinsic, we are able to generate the stack protector basic blocks late in
9416 codegen after the tail call decision has occurred.
9418 '``llvm.objectsize``' Intrinsic
9419 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9426 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <min>)
9427 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <min>)
9432 The ``llvm.objectsize`` intrinsic is designed to provide information to
9433 the optimizers to determine at compile time whether a) an operation
9434 (like memcpy) will overflow a buffer that corresponds to an object, or
9435 b) that a runtime check for overflow isn't necessary. An object in this
9436 context means an allocation of a specific class, structure, array, or
9442 The ``llvm.objectsize`` intrinsic takes two arguments. The first
9443 argument is a pointer to or into the ``object``. The second argument is
9444 a boolean and determines whether ``llvm.objectsize`` returns 0 (if true)
9445 or -1 (if false) when the object size is unknown. The second argument
9446 only accepts constants.
9451 The ``llvm.objectsize`` intrinsic is lowered to a constant representing
9452 the size of the object concerned. If the size cannot be determined at
9453 compile time, ``llvm.objectsize`` returns ``i32/i64 -1 or 0`` (depending
9454 on the ``min`` argument).
9456 '``llvm.expect``' Intrinsic
9457 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9462 This is an overloaded intrinsic. You can use ``llvm.expect`` on any
9467 declare i1 @llvm.expect.i1(i1 <val>, i1 <expected_val>)
9468 declare i32 @llvm.expect.i32(i32 <val>, i32 <expected_val>)
9469 declare i64 @llvm.expect.i64(i64 <val>, i64 <expected_val>)
9474 The ``llvm.expect`` intrinsic provides information about expected (the
9475 most probable) value of ``val``, which can be used by optimizers.
9480 The ``llvm.expect`` intrinsic takes two arguments. The first argument is
9481 a value. The second argument is an expected value, this needs to be a
9482 constant value, variables are not allowed.
9487 This intrinsic is lowered to the ``val``.
9489 '``llvm.assume``' Intrinsic
9490 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9497 declare void @llvm.assume(i1 %cond)
9502 The ``llvm.assume`` allows the optimizer to assume that the provided
9503 condition is true. This information can then be used in simplifying other parts
9509 The condition which the optimizer may assume is always true.
9514 The intrinsic allows the optimizer to assume that the provided condition is
9515 always true whenever the control flow reaches the intrinsic call. No code is
9516 generated for this intrinsic, and instructions that contribute only to the
9517 provided condition are not used for code generation. If the condition is
9518 violated during execution, the behavior is undefined.
9520 Please note that optimizer might limit the transformations performed on values
9521 used by the ``llvm.assume`` intrinsic in order to preserve the instructions
9522 only used to form the intrinsic's input argument. This might prove undesirable
9523 if the extra information provided by the ``llvm.assume`` intrinsic does cause
9524 sufficient overall improvement in code quality. For this reason,
9525 ``llvm.assume`` should not be used to document basic mathematical invariants
9526 that the optimizer can otherwise deduce or facts that are of little use to the
9529 '``llvm.donothing``' Intrinsic
9530 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9537 declare void @llvm.donothing() nounwind readnone
9542 The ``llvm.donothing`` intrinsic doesn't perform any operation. It's the
9543 only intrinsic that can be called with an invoke instruction.
9553 This intrinsic does nothing, and it's removed by optimizers and ignored
9556 Stack Map Intrinsics
9557 --------------------
9559 LLVM provides experimental intrinsics to support runtime patching
9560 mechanisms commonly desired in dynamic language JITs. These intrinsics
9561 are described in :doc:`StackMaps`.