• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1==============================
2LLVM Language Reference Manual
3==============================
4
5.. contents::
6   :local:
7   :depth: 4
8
9Abstract
10========
11
12This document is a reference manual for the LLVM assembly language. LLVM
13is a Static Single Assignment (SSA) based representation that provides
14type safety, low-level operations, flexibility, and the capability of
15representing 'all' high-level languages cleanly. It is the common code
16representation used throughout all phases of the LLVM compilation
17strategy.
18
19Introduction
20============
21
22The LLVM code representation is designed to be used in three different
23forms: 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
25readable assembly language representation. This allows LLVM to provide a
26powerful intermediate representation for efficient compiler
27transformations and analysis, while providing a natural means to debug
28and visualize the transformations. The three different forms of LLVM are
29all equivalent. This document describes the human readable
30representation and notation.
31
32The LLVM representation aims to be light-weight and low-level while
33being expressive, typed, and extensible at the same time. It aims to be
34a "universal IR" of sorts, by being at a low enough level that
35high-level ideas may be cleanly mapped to it (similar to how
36microprocessors are "universal IR's", allowing many source languages to
37be mapped to them). By providing type information, LLVM can be used as
38the target of optimizations: for example, through pointer analysis, it
39can be proven that a C automatic variable is never accessed outside of
40the current function, allowing it to be promoted to a simple SSA value
41instead of a memory location.
42
43.. _wellformed:
44
45Well-Formedness
46---------------
47
48It is important to note that this document describes 'well formed' LLVM
49assembly language. There is a difference between what the parser accepts
50and what is considered 'well formed'. For example, the following
51instruction is syntactically okay, but not well formed:
52
53.. code-block:: llvm
54
55    %x = add i32 1, %x
56
57because the definition of ``%x`` does not dominate all of its uses. The
58LLVM infrastructure provides a verification pass that may be used to
59verify that an LLVM module is well formed. This pass is automatically
60run by the parser after parsing input assembly and by the optimizer
61before it outputs bitcode. The violations pointed out by the verifier
62pass indicate bugs in transformation passes or input to the parser.
63
64.. _identifiers:
65
66Identifiers
67===========
68
69LLVM identifiers come in two basic types: global and local. Global
70identifiers (functions, global variables) begin with the ``'@'``
71character. Local identifiers (register names, types) begin with the
72``'%'`` character. Additionally, there are three different formats for
73identifiers, for different purposes:
74
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.
87
88LLVM requires that values start with a prefix for two reasons: Compilers
89don't need to worry about name clashes with reserved words, and the set
90of reserved words may be expanded in the future without penalty.
91Additionally, unnamed identifiers allow a compiler to quickly come up
92with a temporary variable without having to avoid symbol table
93conflicts.
94
95Reserved words in LLVM are very similar to reserved words in other
96languages. 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
99with variable names, because none of them start with a prefix character
100(``'%'`` or ``'@'``).
101
102Here is an example of LLVM code to multiply the integer variable
103'``%X``' by 8:
104
105The easy way:
106
107.. code-block:: llvm
108
109    %result = mul i32 %X, 8
110
111After strength reduction:
112
113.. code-block:: llvm
114
115    %result = shl i32 %X, 3
116
117And the hard way:
118
119.. code-block:: llvm
120
121    %0 = add i32 %X, %X           ; yields i32:%0
122    %1 = add i32 %0, %0           ; yields i32:%1
123    %result = add i32 %1, %1
124
125This last way of multiplying ``%X`` by 8 illustrates several important
126lexical features of LLVM:
127
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.
136
137It also shows a convention that we follow in this document. When
138demonstrating instructions, we will follow an instruction with a comment
139that defines the type and name of value produced.
140
141High Level Structure
142====================
143
144Module Structure
145----------------
146
147LLVM programs are composed of ``Module``'s, each of which is a
148translation unit of the input programs. Each module consists of
149functions, global variables, and symbol table entries. Modules may be
150combined together with the LLVM linker, which merges function (and
151global variable) definitions, resolves forward declarations, and merges
152symbol table entries. Here is an example of the "hello world" module:
153
154.. code-block:: llvm
155
156    ; Declare the string constant as a global constant.
157    @.str = private unnamed_addr constant [13 x i8] c"hello world\0A\00"
158
159    ; External declaration of the puts function
160    declare i32 @puts(i8* nocapture) nounwind
161
162    ; Definition of main function
163    define i32 @main() {   ; i32()*
164      ; Convert [13 x i8]* to i8  *...
165      %cast210 = getelementptr [13 x i8], [13 x i8]* @.str, i64 0, i64 0
166
167      ; Call puts function to write out the string to stdout.
168      call i32 @puts(i8* %cast210)
169      ret i32 0
170    }
171
172    ; Named metadata
173    !0 = !{i32 42, null, !"string"}
174    !foo = !{!0}
175
176This 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``".
180
181In general, a module is made up of a list of global values (where both
182functions and global variables are global values). Global values are
183represented by a pointer to a memory location (in this case, a pointer
184to an array of char, and a pointer to a function), and have one of the
185following :ref:`linkage types <linkage>`.
186
187.. _linkage:
188
189Linkage Types
190-------------
191
192All Global Variables and Functions have one of the following types of
193linkage:
194
195``private``
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.
202``internal``
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 into
208    the object file corresponding to the LLVM module. From the linker's
209    perspective, an ``available_externally`` global is equivalent to
210    an external declaration. They exist to allow inlining and other
211    optimizations to take place given knowledge of the definition of the
212    global, which is known to be somewhere outside the module. Globals
213    with ``available_externally`` linkage are allowed to be discarded at
214    will, and allow inlining and other optimizations. This linkage type is
215    only allowed on definitions, not declarations.
216``linkonce``
217    Globals with "``linkonce``" linkage are merged with other globals of
218    the same name when linkage occurs. This can be used to implement
219    some forms of inline functions, templates, or other code which must
220    be generated in each translation unit that uses it, but where the
221    body may be overridden with a more definitive definition later.
222    Unreferenced ``linkonce`` globals are allowed to be discarded. Note
223    that ``linkonce`` linkage does not actually allow the optimizer to
224    inline the body of this function into callers because it doesn't
225    know if this definition of the function is the definitive definition
226    within the program or whether it will be overridden by a stronger
227    definition. To enable inlining and other optimizations, use
228    "``linkonce_odr``" linkage.
229``weak``
230    "``weak``" linkage has the same merging semantics as ``linkonce``
231    linkage, except that unreferenced globals with ``weak`` linkage may
232    not be discarded. This is used for globals that are declared "weak"
233    in C source code.
234``common``
235    "``common``" linkage is most similar to "``weak``" linkage, but they
236    are used for tentative definitions in C, such as "``int X;``" at
237    global scope. Symbols with "``common``" linkage are merged in the
238    same way as ``weak symbols``, and they may not be deleted if
239    unreferenced. ``common`` symbols may not have an explicit section,
240    must have a zero initializer, and may not be marked
241    ':ref:`constant <globalvars>`'. Functions and aliases may not have
242    common linkage.
243
244.. _linkage_appending:
245
246``appending``
247    "``appending``" linkage may only be applied to global variables of
248    pointer to array type. When two global variables with appending
249    linkage are linked together, the two global arrays are appended
250    together. This is the LLVM, typesafe, equivalent of having the
251    system linker append together "sections" with identical names when
252    .o files are linked.
253
254    Unfortunately this doesn't correspond to any feature in .o files, so it
255    can only be used for variables like ``llvm.global_ctors`` which llvm
256    interprets specially.
257
258``extern_weak``
259    The semantics of this linkage follow the ELF object file model: the
260    symbol is weak until linked, if not linked, the symbol becomes null
261    instead of being an undefined reference.
262``linkonce_odr``, ``weak_odr``
263    Some languages allow differing globals to be merged, such as two
264    functions with different semantics. Other languages, such as
265    ``C++``, ensure that only equivalent globals are ever merged (the
266    "one definition rule" --- "ODR"). Such languages can use the
267    ``linkonce_odr`` and ``weak_odr`` linkage types to indicate that the
268    global will only be merged with equivalent globals. These linkage
269    types are otherwise the same as their non-``odr`` versions.
270``external``
271    If none of the above identifiers are used, the global is externally
272    visible, meaning that it participates in linkage and can be used to
273    resolve external symbol references.
274
275It is illegal for a function *declaration* to have any linkage type
276other than ``external`` or ``extern_weak``.
277
278.. _callingconv:
279
280Calling Conventions
281-------------------
282
283LLVM :ref:`functions <functionstructure>`, :ref:`calls <i_call>` and
284:ref:`invokes <i_invoke>` can all have an optional calling convention
285specified for the call. The calling convention of any pair of dynamic
286caller/callee must match, or the behavior of the program is undefined.
287The following calling conventions are supported by LLVM, and more may be
288added in the future:
289
290"``ccc``" - The C calling convention
291    This calling convention (the default if no other calling convention
292    is specified) matches the target C calling conventions. This calling
293    convention supports varargs function calls and tolerates some
294    mismatch in the declared prototype and implemented declaration of
295    the function (as does normal C).
296"``fastcc``" - The fast calling convention
297    This calling convention attempts to make calls as fast as possible
298    (e.g. by passing things in registers). This calling convention
299    allows the target to use whatever tricks it wants to produce fast
300    code for the target, without having to conform to an externally
301    specified ABI (Application Binary Interface). `Tail calls can only
302    be optimized when this, the GHC or the HiPE convention is
303    used. <CodeGenerator.html#id80>`_ This calling convention does not
304    support varargs and requires the prototype of all callees to exactly
305    match the prototype of the function definition.
306"``coldcc``" - The cold calling convention
307    This calling convention attempts to make code in the caller as
308    efficient as possible under the assumption that the call is not
309    commonly executed. As such, these calls often preserve all registers
310    so that the call does not break any live ranges in the caller side.
311    This calling convention does not support varargs and requires the
312    prototype of all callees to exactly match the prototype of the
313    function definition. Furthermore the inliner doesn't consider such function
314    calls for inlining.
315"``cc 10``" - GHC convention
316    This calling convention has been implemented specifically for use by
317    the `Glasgow Haskell Compiler (GHC) <http://www.haskell.org/ghc>`_.
318    It passes everything in registers, going to extremes to achieve this
319    by disabling callee save registers. This calling convention should
320    not be used lightly but only for specific situations such as an
321    alternative to the *register pinning* performance technique often
322    used when implementing functional programming languages. At the
323    moment only X86 supports this convention and it has the following
324    limitations:
325
326    -  On *X86-32* only supports up to 4 bit type parameters. No
327       floating point types are supported.
328    -  On *X86-64* only supports up to 10 bit type parameters and 6
329       floating point parameters.
330
331    This calling convention supports `tail call
332    optimization <CodeGenerator.html#id80>`_ but requires both the
333    caller and callee are using it.
334"``cc 11``" - The HiPE calling convention
335    This calling convention has been implemented specifically for use by
336    the `High-Performance Erlang
337    (HiPE) <http://www.it.uu.se/research/group/hipe/>`_ compiler, *the*
338    native code compiler of the `Ericsson's Open Source Erlang/OTP
339    system <http://www.erlang.org/download.shtml>`_. It uses more
340    registers for argument passing than the ordinary C calling
341    convention and defines no callee-saved registers. The calling
342    convention properly supports `tail call
343    optimization <CodeGenerator.html#id80>`_ but requires that both the
344    caller and the callee use it. It uses a *register pinning*
345    mechanism, similar to GHC's convention, for keeping frequently
346    accessed runtime components pinned to specific hardware registers.
347    At the moment only X86 supports this convention (both 32 and 64
348    bit).
349"``webkit_jscc``" - WebKit's JavaScript calling convention
350    This calling convention has been implemented for `WebKit FTL JIT
351    <https://trac.webkit.org/wiki/FTLJIT>`_. It passes arguments on the
352    stack right to left (as cdecl does), and returns a value in the
353    platform's customary return register.
354"``anyregcc``" - Dynamic calling convention for code patching
355    This is a special convention that supports patching an arbitrary code
356    sequence in place of a call site. This convention forces the call
357    arguments into registers but allows them to be dynamically
358    allocated. This can currently only be used with calls to
359    llvm.experimental.patchpoint because only this intrinsic records
360    the location of its arguments in a side table. See :doc:`StackMaps`.
361"``preserve_mostcc``" - The `PreserveMost` calling convention
362    This calling convention attempts to make the code in the caller as
363    unintrusive as possible. This convention behaves identically to the `C`
364    calling convention on how arguments and return values are passed, but it
365    uses a different set of caller/callee-saved registers. This alleviates the
366    burden of saving and recovering a large register set before and after the
367    call in the caller. If the arguments are passed in callee-saved registers,
368    then they will be preserved by the callee across the call. This doesn't
369    apply for values returned in callee-saved registers.
370
371    - On X86-64 the callee preserves all general purpose registers, except for
372      R11. R11 can be used as a scratch register. Floating-point registers
373      (XMMs/YMMs) are not preserved and need to be saved by the caller.
374
375    The idea behind this convention is to support calls to runtime functions
376    that have a hot path and a cold path. The hot path is usually a small piece
377    of code that doesn't use many registers. The cold path might need to call out to
378    another function and therefore only needs to preserve the caller-saved
379    registers, which haven't already been saved by the caller. The
380    `PreserveMost` calling convention is very similar to the `cold` calling
381    convention in terms of caller/callee-saved registers, but they are used for
382    different types of function calls. `coldcc` is for function calls that are
383    rarely executed, whereas `preserve_mostcc` function calls are intended to be
384    on the hot path and definitely executed a lot. Furthermore `preserve_mostcc`
385    doesn't prevent the inliner from inlining the function call.
386
387    This calling convention will be used by a future version of the ObjectiveC
388    runtime and should therefore still be considered experimental at this time.
389    Although this convention was created to optimize certain runtime calls to
390    the ObjectiveC runtime, it is not limited to this runtime and might be used
391    by other runtimes in the future too. The current implementation only
392    supports X86-64, but the intention is to support more architectures in the
393    future.
394"``preserve_allcc``" - The `PreserveAll` calling convention
395    This calling convention attempts to make the code in the caller even less
396    intrusive than the `PreserveMost` calling convention. This calling
397    convention also behaves identical to the `C` calling convention on how
398    arguments and return values are passed, but it uses a different set of
399    caller/callee-saved registers. This removes the burden of saving and
400    recovering a large register set before and after the call in the caller. If
401    the arguments are passed in callee-saved registers, then they will be
402    preserved by the callee across the call. This doesn't apply for values
403    returned in callee-saved registers.
404
405    - On X86-64 the callee preserves all general purpose registers, except for
406      R11. R11 can be used as a scratch register. Furthermore it also preserves
407      all floating-point registers (XMMs/YMMs).
408
409    The idea behind this convention is to support calls to runtime functions
410    that don't need to call out to any other functions.
411
412    This calling convention, like the `PreserveMost` calling convention, will be
413    used by a future version of the ObjectiveC runtime and should be considered
414    experimental at this time.
415"``cxx_fast_tlscc``" - The `CXX_FAST_TLS` calling convention for access functions
416    Clang generates an access function to access C++-style TLS. The access
417    function generally has an entry block, an exit block and an initialization
418    block that is run at the first time. The entry and exit blocks can access
419    a few TLS IR variables, each access will be lowered to a platform-specific
420    sequence.
421
422    This calling convention aims to minimize overhead in the caller by
423    preserving as many registers as possible (all the registers that are
424    perserved on the fast path, composed of the entry and exit blocks).
425
426    This calling convention behaves identical to the `C` calling convention on
427    how arguments and return values are passed, but it uses a different set of
428    caller/callee-saved registers.
429
430    Given that each platform has its own lowering sequence, hence its own set
431    of preserved registers, we can't use the existing `PreserveMost`.
432
433    - On X86-64 the callee preserves all general purpose registers, except for
434      RDI and RAX.
435"``swiftcc``" - This calling convention is used for Swift language.
436    - On X86-64 RCX and R8 are available for additional integer returns, and
437      XMM2 and XMM3 are available for additional FP/vector returns.
438    - On iOS platforms, we use AAPCS-VFP calling convention.
439"``cc <n>``" - Numbered convention
440    Any calling convention may be specified by number, allowing
441    target-specific calling conventions to be used. Target specific
442    calling conventions start at 64.
443
444More calling conventions can be added/defined on an as-needed basis, to
445support Pascal conventions or any other well-known target-independent
446convention.
447
448.. _visibilitystyles:
449
450Visibility Styles
451-----------------
452
453All Global Variables and Functions have one of the following visibility
454styles:
455
456"``default``" - Default style
457    On targets that use the ELF object file format, default visibility
458    means that the declaration is visible to other modules and, in
459    shared libraries, means that the declared entity may be overridden.
460    On Darwin, default visibility means that the declaration is visible
461    to other modules. Default visibility corresponds to "external
462    linkage" in the language.
463"``hidden``" - Hidden style
464    Two declarations of an object with hidden visibility refer to the
465    same object if they are in the same shared object. Usually, hidden
466    visibility indicates that the symbol will not be placed into the
467    dynamic symbol table, so no other module (executable or shared
468    library) can reference it directly.
469"``protected``" - Protected style
470    On ELF, protected visibility indicates that the symbol will be
471    placed in the dynamic symbol table, but that references within the
472    defining module will bind to the local symbol. That is, the symbol
473    cannot be overridden by another module.
474
475A symbol with ``internal`` or ``private`` linkage must have ``default``
476visibility.
477
478.. _dllstorageclass:
479
480DLL Storage Classes
481-------------------
482
483All Global Variables, Functions and Aliases can have one of the following
484DLL storage class:
485
486``dllimport``
487    "``dllimport``" causes the compiler to reference a function or variable via
488    a global pointer to a pointer that is set up by the DLL exporting the
489    symbol. On Microsoft Windows targets, the pointer name is formed by
490    combining ``__imp_`` and the function or variable name.
491``dllexport``
492    "``dllexport``" causes the compiler to provide a global pointer to a pointer
493    in a DLL, so that it can be referenced with the ``dllimport`` attribute. On
494    Microsoft Windows targets, the pointer name is formed by combining
495    ``__imp_`` and the function or variable name. Since this storage class
496    exists for defining a dll interface, the compiler, assembler and linker know
497    it is externally referenced and must refrain from deleting the symbol.
498
499.. _tls_model:
500
501Thread Local Storage Models
502---------------------------
503
504A variable may be defined as ``thread_local``, which means that it will
505not be shared by threads (each thread will have a separated copy of the
506variable). Not all targets support thread-local variables. Optionally, a
507TLS model may be specified:
508
509``localdynamic``
510    For variables that are only used within the current shared library.
511``initialexec``
512    For variables in modules that will not be loaded dynamically.
513``localexec``
514    For variables defined in the executable and only used within it.
515
516If no explicit model is given, the "general dynamic" model is used.
517
518The models correspond to the ELF TLS models; see `ELF Handling For
519Thread-Local Storage <http://people.redhat.com/drepper/tls.pdf>`_ for
520more information on under which circumstances the different models may
521be used. The target may choose a different TLS model if the specified
522model is not supported, or if a better choice of model can be made.
523
524A model can also be specified in an alias, but then it only governs how
525the alias is accessed. It will not have any effect in the aliasee.
526
527For platforms without linker support of ELF TLS model, the -femulated-tls
528flag can be used to generate GCC compatible emulated TLS code.
529
530.. _namedtypes:
531
532Structure Types
533---------------
534
535LLVM IR allows you to specify both "identified" and "literal" :ref:`structure
536types <t_struct>`. Literal types are uniqued structurally, but identified types
537are never uniqued. An :ref:`opaque structural type <t_opaque>` can also be used
538to forward declare a type that is not yet available.
539
540An example of an identified structure specification is:
541
542.. code-block:: llvm
543
544    %mytype = type { %mytype*, i32 }
545
546Prior to the LLVM 3.0 release, identified types were structurally uniqued. Only
547literal types are uniqued in recent versions of LLVM.
548
549.. _globalvars:
550
551Global Variables
552----------------
553
554Global variables define regions of memory allocated at compilation time
555instead of run-time.
556
557Global variable definitions must be initialized.
558
559Global variables in other translation units can also be declared, in which
560case they don't have an initializer.
561
562Either global variable definitions or declarations may have an explicit section
563to be placed in and may have an optional explicit alignment specified.
564
565A variable may be defined as a global ``constant``, which indicates that
566the contents of the variable will **never** be modified (enabling better
567optimization, allowing the global data to be placed in the read-only
568section of an executable, etc). Note that variables that need runtime
569initialization cannot be marked ``constant`` as there is a store to the
570variable.
571
572LLVM explicitly allows *declarations* of global variables to be marked
573constant, even if the final definition of the global is not. This
574capability can be used to enable slightly better optimization of the
575program, but requires the language definition to guarantee that
576optimizations based on the 'constantness' are valid for the translation
577units that do not include the definition.
578
579As SSA values, global variables define pointer values that are in scope
580(i.e. they dominate) all basic blocks in the program. Global variables
581always define a pointer to their "content" type because they describe a
582region of memory, and all memory objects in LLVM are accessed through
583pointers.
584
585Global variables can be marked with ``unnamed_addr`` which indicates
586that the address is not significant, only the content. Constants marked
587like this can be merged with other constants if they have the same
588initializer. Note that a constant with significant address *can* be
589merged with a ``unnamed_addr`` constant, the result being a constant
590whose address is significant.
591
592If the ``local_unnamed_addr`` attribute is given, the address is known to
593not be significant within the module.
594
595A global variable may be declared to reside in a target-specific
596numbered address space. For targets that support them, address spaces
597may affect how optimizations are performed and/or what target
598instructions are used to access the variable. The default address space
599is zero. The address space qualifier must precede any other attributes.
600
601LLVM allows an explicit section to be specified for globals. If the
602target supports it, it will emit globals to the section specified.
603Additionally, the global can placed in a comdat if the target has the necessary
604support.
605
606By default, global initializers are optimized by assuming that global
607variables defined within the module are not modified from their
608initial values before the start of the global initializer. This is
609true even for variables potentially accessible from outside the
610module, including those with external linkage or appearing in
611``@llvm.used`` or dllexported variables. This assumption may be suppressed
612by marking the variable with ``externally_initialized``.
613
614An explicit alignment may be specified for a global, which must be a
615power of 2. If not present, or if the alignment is set to zero, the
616alignment of the global is set by the target to whatever it feels
617convenient. If an explicit alignment is specified, the global is forced
618to have exactly that alignment. Targets and optimizers are not allowed
619to over-align the global if the global has an assigned section. In this
620case, the extra alignment could be observable: for example, code could
621assume that the globals are densely packed in their section and try to
622iterate over them as an array, alignment padding would break this
623iteration. The maximum alignment is ``1 << 29``.
624
625Globals can also have a :ref:`DLL storage class <dllstorageclass>` and
626an optional list of attached :ref:`metadata <metadata>`,
627
628Variables and aliases can have a
629:ref:`Thread Local Storage Model <tls_model>`.
630
631Syntax::
632
633      @<GlobalVarName> = [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal]
634                         [(unnamed_addr|local_unnamed_addr)] [AddrSpace]
635                         [ExternallyInitialized]
636                         <global | constant> <Type> [<InitializerConstant>]
637                         [, section "name"] [, comdat [($name)]]
638                         [, align <Alignment>] (, !name !N)*
639
640For example, the following defines a global in a numbered address space
641with an initializer, section, and alignment:
642
643.. code-block:: llvm
644
645    @G = addrspace(5) constant float 1.0, section "foo", align 4
646
647The following example just declares a global variable
648
649.. code-block:: llvm
650
651   @G = external global i32
652
653The following example defines a thread-local global with the
654``initialexec`` TLS model:
655
656.. code-block:: llvm
657
658    @G = thread_local(initialexec) global i32 0, align 4
659
660.. _functionstructure:
661
662Functions
663---------
664
665LLVM function definitions consist of the "``define``" keyword, an
666optional :ref:`linkage type <linkage>`, an optional :ref:`visibility
667style <visibility>`, an optional :ref:`DLL storage class <dllstorageclass>`,
668an optional :ref:`calling convention <callingconv>`,
669an optional ``unnamed_addr`` attribute, a return type, an optional
670:ref:`parameter attribute <paramattrs>` for the return type, a function
671name, a (possibly empty) argument list (each with optional :ref:`parameter
672attributes <paramattrs>`), optional :ref:`function attributes <fnattrs>`,
673an optional section, an optional alignment,
674an optional :ref:`comdat <langref_comdats>`,
675an optional :ref:`garbage collector name <gc>`, an optional :ref:`prefix <prefixdata>`,
676an optional :ref:`prologue <prologuedata>`,
677an optional :ref:`personality <personalityfn>`,
678an optional list of attached :ref:`metadata <metadata>`,
679an opening curly brace, a list of basic blocks, and a closing curly brace.
680
681LLVM function declarations consist of the "``declare``" keyword, an
682optional :ref:`linkage type <linkage>`, an optional :ref:`visibility style
683<visibility>`, an optional :ref:`DLL storage class <dllstorageclass>`, an
684optional :ref:`calling convention <callingconv>`, an optional ``unnamed_addr``
685or ``local_unnamed_addr`` attribute, a return type, an optional :ref:`parameter
686attribute <paramattrs>` for the return type, a function name, a possibly
687empty list of arguments, an optional alignment, an optional :ref:`garbage
688collector name <gc>`, an optional :ref:`prefix <prefixdata>`, and an optional
689:ref:`prologue <prologuedata>`.
690
691A function definition contains a list of basic blocks, forming the CFG (Control
692Flow Graph) for the function. Each basic block may optionally start with a label
693(giving the basic block a symbol table entry), contains a list of instructions,
694and ends with a :ref:`terminator <terminators>` instruction (such as a branch or
695function return). If an explicit label is not provided, a block is assigned an
696implicit numbered label, using the next value from the same counter as used for
697unnamed temporaries (:ref:`see above<identifiers>`). For example, if a function
698entry block does not have an explicit label, it will be assigned label "%0",
699then the first unnamed temporary in that block will be "%1", etc.
700
701The first basic block in a function is special in two ways: it is
702immediately executed on entrance to the function, and it is not allowed
703to have predecessor basic blocks (i.e. there can not be any branches to
704the entry block of a function). Because the block can have no
705predecessors, it also cannot have any :ref:`PHI nodes <i_phi>`.
706
707LLVM allows an explicit section to be specified for functions. If the
708target supports it, it will emit functions to the section specified.
709Additionally, the function can be placed in a COMDAT.
710
711An explicit alignment may be specified for a function. If not present,
712or if the alignment is set to zero, the alignment of the function is set
713by the target to whatever it feels convenient. If an explicit alignment
714is specified, the function is forced to have at least that much
715alignment. All alignments must be a power of 2.
716
717If the ``unnamed_addr`` attribute is given, the address is known to not
718be significant and two identical functions can be merged.
719
720If the ``local_unnamed_addr`` attribute is given, the address is known to
721not be significant within the module.
722
723Syntax::
724
725    define [linkage] [visibility] [DLLStorageClass]
726           [cconv] [ret attrs]
727           <ResultType> @<FunctionName> ([argument list])
728           [(unnamed_addr|local_unnamed_addr)] [fn Attrs] [section "name"]
729           [comdat [($name)]] [align N] [gc] [prefix Constant]
730           [prologue Constant] [personality Constant] (!name !N)* { ... }
731
732The argument list is a comma separated sequence of arguments where each
733argument is of the following form:
734
735Syntax::
736
737   <type> [parameter Attrs] [name]
738
739
740.. _langref_aliases:
741
742Aliases
743-------
744
745Aliases, unlike function or variables, don't create any new data. They
746are just a new symbol and metadata for an existing position.
747
748Aliases have a name and an aliasee that is either a global value or a
749constant expression.
750
751Aliases may have an optional :ref:`linkage type <linkage>`, an optional
752:ref:`visibility style <visibility>`, an optional :ref:`DLL storage class
753<dllstorageclass>` and an optional :ref:`tls model <tls_model>`.
754
755Syntax::
756
757    @<Name> = [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal] [(unnamed_addr|local_unnamed_addr)] alias <AliaseeTy>, <AliaseeTy>* @<Aliasee>
758
759The linkage must be one of ``private``, ``internal``, ``linkonce``, ``weak``,
760``linkonce_odr``, ``weak_odr``, ``external``. Note that some system linkers
761might not correctly handle dropping a weak symbol that is aliased.
762
763Aliases that are not ``unnamed_addr`` are guaranteed to have the same address as
764the aliasee expression. ``unnamed_addr`` ones are only guaranteed to point
765to the same content.
766
767If the ``local_unnamed_addr`` attribute is given, the address is known to
768not be significant within the module.
769
770Since aliases are only a second name, some restrictions apply, of which
771some can only be checked when producing an object file:
772
773* The expression defining the aliasee must be computable at assembly
774  time. Since it is just a name, no relocations can be used.
775
776* No alias in the expression can be weak as the possibility of the
777  intermediate alias being overridden cannot be represented in an
778  object file.
779
780* No global value in the expression can be a declaration, since that
781  would require a relocation, which is not possible.
782
783.. _langref_ifunc:
784
785IFuncs
786-------
787
788IFuncs, like as aliases, don't create any new data or func. They are just a new
789symbol that dynamic linker resolves at runtime by calling a resolver function.
790
791IFuncs have a name and a resolver that is a function called by dynamic linker
792that returns address of another function associated with the name.
793
794IFunc may have an optional :ref:`linkage type <linkage>` and an optional
795:ref:`visibility style <visibility>`.
796
797Syntax::
798
799    @<Name> = [Linkage] [Visibility] ifunc <IFuncTy>, <ResolverTy>* @<Resolver>
800
801
802.. _langref_comdats:
803
804Comdats
805-------
806
807Comdat IR provides access to COFF and ELF object file COMDAT functionality.
808
809Comdats have a name which represents the COMDAT key. All global objects that
810specify this key will only end up in the final object file if the linker chooses
811that key over some other key. Aliases are placed in the same COMDAT that their
812aliasee computes to, if any.
813
814Comdats have a selection kind to provide input on how the linker should
815choose between keys in two different object files.
816
817Syntax::
818
819    $<Name> = comdat SelectionKind
820
821The selection kind must be one of the following:
822
823``any``
824    The linker may choose any COMDAT key, the choice is arbitrary.
825``exactmatch``
826    The linker may choose any COMDAT key but the sections must contain the
827    same data.
828``largest``
829    The linker will choose the section containing the largest COMDAT key.
830``noduplicates``
831    The linker requires that only section with this COMDAT key exist.
832``samesize``
833    The linker may choose any COMDAT key but the sections must contain the
834    same amount of data.
835
836Note that the Mach-O platform doesn't support COMDATs and ELF only supports
837``any`` as a selection kind.
838
839Here is an example of a COMDAT group where a function will only be selected if
840the COMDAT key's section is the largest:
841
842.. code-block:: llvm
843
844   $foo = comdat largest
845   @foo = global i32 2, comdat($foo)
846
847   define void @bar() comdat($foo) {
848     ret void
849   }
850
851As a syntactic sugar the ``$name`` can be omitted if the name is the same as
852the global name:
853
854.. code-block:: llvm
855
856  $foo = comdat any
857  @foo = global i32 2, comdat
858
859
860In a COFF object file, this will create a COMDAT section with selection kind
861``IMAGE_COMDAT_SELECT_LARGEST`` containing the contents of the ``@foo`` symbol
862and another COMDAT section with selection kind
863``IMAGE_COMDAT_SELECT_ASSOCIATIVE`` which is associated with the first COMDAT
864section and contains the contents of the ``@bar`` symbol.
865
866There are some restrictions on the properties of the global object.
867It, or an alias to it, must have the same name as the COMDAT group when
868targeting COFF.
869The contents and size of this object may be used during link-time to determine
870which COMDAT groups get selected depending on the selection kind.
871Because the name of the object must match the name of the COMDAT group, the
872linkage of the global object must not be local; local symbols can get renamed
873if a collision occurs in the symbol table.
874
875The combined use of COMDATS and section attributes may yield surprising results.
876For example:
877
878.. code-block:: llvm
879
880   $foo = comdat any
881   $bar = comdat any
882   @g1 = global i32 42, section "sec", comdat($foo)
883   @g2 = global i32 42, section "sec", comdat($bar)
884
885From the object file perspective, this requires the creation of two sections
886with the same name. This is necessary because both globals belong to different
887COMDAT groups and COMDATs, at the object file level, are represented by
888sections.
889
890Note that certain IR constructs like global variables and functions may
891create COMDATs in the object file in addition to any which are specified using
892COMDAT IR. This arises when the code generator is configured to emit globals
893in individual sections (e.g. when `-data-sections` or `-function-sections`
894is supplied to `llc`).
895
896.. _namedmetadatastructure:
897
898Named Metadata
899--------------
900
901Named metadata is a collection of metadata. :ref:`Metadata
902nodes <metadata>` (but not metadata strings) are the only valid
903operands for a named metadata.
904
905#. Named metadata are represented as a string of characters with the
906   metadata prefix. The rules for metadata names are the same as for
907   identifiers, but quoted names are not allowed. ``"\xx"`` type escapes
908   are still valid, which allows any character to be part of a name.
909
910Syntax::
911
912    ; Some unnamed metadata nodes, which are referenced by the named metadata.
913    !0 = !{!"zero"}
914    !1 = !{!"one"}
915    !2 = !{!"two"}
916    ; A named metadata.
917    !name = !{!0, !1, !2}
918
919.. _paramattrs:
920
921Parameter Attributes
922--------------------
923
924The return type and each parameter of a function type may have a set of
925*parameter attributes* associated with them. Parameter attributes are
926used to communicate additional information about the result or
927parameters of a function. Parameter attributes are considered to be part
928of the function, not of the function type, so functions with different
929parameter attributes can have the same function type.
930
931Parameter attributes are simple keywords that follow the type specified.
932If multiple parameter attributes are needed, they are space separated.
933For example:
934
935.. code-block:: llvm
936
937    declare i32 @printf(i8* noalias nocapture, ...)
938    declare i32 @atoi(i8 zeroext)
939    declare signext i8 @returns_signed_char()
940
941Note that any attributes for the function result (``nounwind``,
942``readonly``) come immediately after the argument list.
943
944Currently, only the following parameter attributes are defined:
945
946``zeroext``
947    This indicates to the code generator that the parameter or return
948    value should be zero-extended to the extent required by the target's
949    ABI by the caller (for a parameter) or the callee (for a return value).
950``signext``
951    This indicates to the code generator that the parameter or return
952    value should be sign-extended to the extent required by the target's
953    ABI (which is usually 32-bits) by the caller (for a parameter) or
954    the callee (for a return value).
955``inreg``
956    This indicates that this parameter or return value should be treated
957    in a special target-dependent fashion while emitting code for
958    a function call or return (usually, by putting it in a register as
959    opposed to memory, though some targets use it to distinguish between
960    two different kinds of registers). Use of this attribute is
961    target-specific.
962``byval``
963    This indicates that the pointer parameter should really be passed by
964    value to the function. The attribute implies that a hidden copy of
965    the pointee is made between the caller and the callee, so the callee
966    is unable to modify the value in the caller. This attribute is only
967    valid on LLVM pointer arguments. It is generally used to pass
968    structs and arrays by value, but is also valid on pointers to
969    scalars. The copy is considered to belong to the caller not the
970    callee (for example, ``readonly`` functions should not write to
971    ``byval`` parameters). This is not a valid attribute for return
972    values.
973
974    The byval attribute also supports specifying an alignment with the
975    align attribute. It indicates the alignment of the stack slot to
976    form and the known alignment of the pointer specified to the call
977    site. If the alignment is not specified, then the code generator
978    makes a target-specific assumption.
979
980.. _attr_inalloca:
981
982``inalloca``
983
984    The ``inalloca`` argument attribute allows the caller to take the
985    address of outgoing stack arguments. An ``inalloca`` argument must
986    be a pointer to stack memory produced by an ``alloca`` instruction.
987    The alloca, or argument allocation, must also be tagged with the
988    inalloca keyword. Only the last argument may have the ``inalloca``
989    attribute, and that argument is guaranteed to be passed in memory.
990
991    An argument allocation may be used by a call at most once because
992    the call may deallocate it. The ``inalloca`` attribute cannot be
993    used in conjunction with other attributes that affect argument
994    storage, like ``inreg``, ``nest``, ``sret``, or ``byval``. The
995    ``inalloca`` attribute also disables LLVM's implicit lowering of
996    large aggregate return values, which means that frontend authors
997    must lower them with ``sret`` pointers.
998
999    When the call site is reached, the argument allocation must have
1000    been the most recent stack allocation that is still live, or the
1001    results are undefined. It is possible to allocate additional stack
1002    space after an argument allocation and before its call site, but it
1003    must be cleared off with :ref:`llvm.stackrestore
1004    <int_stackrestore>`.
1005
1006    See :doc:`InAlloca` for more information on how to use this
1007    attribute.
1008
1009``sret``
1010    This indicates that the pointer parameter specifies the address of a
1011    structure that is the return value of the function in the source
1012    program. This pointer must be guaranteed by the caller to be valid:
1013    loads and stores to the structure may be assumed by the callee
1014    not to trap and to be properly aligned. This may only be applied to
1015    the first parameter. This is not a valid attribute for return
1016    values.
1017
1018``align <n>``
1019    This indicates that the pointer value may be assumed by the optimizer to
1020    have the specified alignment.
1021
1022    Note that this attribute has additional semantics when combined with the
1023    ``byval`` attribute.
1024
1025.. _noalias:
1026
1027``noalias``
1028    This indicates that objects accessed via pointer values
1029    :ref:`based <pointeraliasing>` on the argument or return value are not also
1030    accessed, during the execution of the function, via pointer values not
1031    *based* on the argument or return value. The attribute on a return value
1032    also has additional semantics described below. The caller shares the
1033    responsibility with the callee for ensuring that these requirements are met.
1034    For further details, please see the discussion of the NoAlias response in
1035    :ref:`alias analysis <Must, May, or No>`.
1036
1037    Note that this definition of ``noalias`` is intentionally similar
1038    to the definition of ``restrict`` in C99 for function arguments.
1039
1040    For function return values, C99's ``restrict`` is not meaningful,
1041    while LLVM's ``noalias`` is. Furthermore, the semantics of the ``noalias``
1042    attribute on return values are stronger than the semantics of the attribute
1043    when used on function arguments. On function return values, the ``noalias``
1044    attribute indicates that the function acts like a system memory allocation
1045    function, returning a pointer to allocated storage disjoint from the
1046    storage for any other object accessible to the caller.
1047
1048``nocapture``
1049    This indicates that the callee does not make any copies of the
1050    pointer that outlive the callee itself. This is not a valid
1051    attribute for return values.  Addresses used in volatile operations
1052    are considered to be captured.
1053
1054.. _nest:
1055
1056``nest``
1057    This indicates that the pointer parameter can be excised using the
1058    :ref:`trampoline intrinsics <int_trampoline>`. This is not a valid
1059    attribute for return values and can only be applied to one parameter.
1060
1061``returned``
1062    This indicates that the function always returns the argument as its return
1063    value. This is a hint to the optimizer and code generator used when
1064    generating the caller, allowing value propagation, tail call optimization,
1065    and omission of register saves and restores in some cases; it is not
1066    checked or enforced when generating the callee. The parameter and the
1067    function return type must be valid operands for the
1068    :ref:`bitcast instruction <i_bitcast>`. This is not a valid attribute for
1069    return values and can only be applied to one parameter.
1070
1071``nonnull``
1072    This indicates that the parameter or return pointer is not null. This
1073    attribute may only be applied to pointer typed parameters. This is not
1074    checked or enforced by LLVM, the caller must ensure that the pointer
1075    passed in is non-null, or the callee must ensure that the returned pointer
1076    is non-null.
1077
1078``dereferenceable(<n>)``
1079    This indicates that the parameter or return pointer is dereferenceable. This
1080    attribute may only be applied to pointer typed parameters. A pointer that
1081    is dereferenceable can be loaded from speculatively without a risk of
1082    trapping. The number of bytes known to be dereferenceable must be provided
1083    in parentheses. It is legal for the number of bytes to be less than the
1084    size of the pointee type. The ``nonnull`` attribute does not imply
1085    dereferenceability (consider a pointer to one element past the end of an
1086    array), however ``dereferenceable(<n>)`` does imply ``nonnull`` in
1087    ``addrspace(0)`` (which is the default address space).
1088
1089``dereferenceable_or_null(<n>)``
1090    This indicates that the parameter or return value isn't both
1091    non-null and non-dereferenceable (up to ``<n>`` bytes) at the same
1092    time. All non-null pointers tagged with
1093    ``dereferenceable_or_null(<n>)`` are ``dereferenceable(<n>)``.
1094    For address space 0 ``dereferenceable_or_null(<n>)`` implies that
1095    a pointer is exactly one of ``dereferenceable(<n>)`` or ``null``,
1096    and in other address spaces ``dereferenceable_or_null(<n>)``
1097    implies that a pointer is at least one of ``dereferenceable(<n>)``
1098    or ``null`` (i.e. it may be both ``null`` and
1099    ``dereferenceable(<n>)``). This attribute may only be applied to
1100    pointer typed parameters.
1101
1102``swiftself``
1103    This indicates that the parameter is the self/context parameter. This is not
1104    a valid attribute for return values and can only be applied to one
1105    parameter.
1106
1107``swifterror``
1108    This attribute is motivated to model and optimize Swift error handling. It
1109    can be applied to a parameter with pointer to pointer type or a
1110    pointer-sized alloca. At the call site, the actual argument that corresponds
1111    to a ``swifterror`` parameter has to come from a ``swifterror`` alloca. A
1112    ``swifterror`` value (either the parameter or the alloca) can only be loaded
1113    and stored from, or used as a ``swifterror`` argument. This is not a valid
1114    attribute for return values and can only be applied to one parameter.
1115
1116    These constraints allow the calling convention to optimize access to
1117    ``swifterror`` variables by associating them with a specific register at
1118    call boundaries rather than placing them in memory. Since this does change
1119    the calling convention, a function which uses the ``swifterror`` attribute
1120    on a parameter is not ABI-compatible with one which does not.
1121
1122    These constraints also allow LLVM to assume that a ``swifterror`` argument
1123    does not alias any other memory visible within a function and that a
1124    ``swifterror`` alloca passed as an argument does not escape.
1125
1126.. _gc:
1127
1128Garbage Collector Strategy Names
1129--------------------------------
1130
1131Each function may specify a garbage collector strategy name, which is simply a
1132string:
1133
1134.. code-block:: llvm
1135
1136    define void @f() gc "name" { ... }
1137
1138The supported values of *name* includes those :ref:`built in to LLVM
1139<builtin-gc-strategies>` and any provided by loaded plugins. Specifying a GC
1140strategy will cause the compiler to alter its output in order to support the
1141named garbage collection algorithm. Note that LLVM itself does not contain a
1142garbage collector, this functionality is restricted to generating machine code
1143which can interoperate with a collector provided externally.
1144
1145.. _prefixdata:
1146
1147Prefix Data
1148-----------
1149
1150Prefix data is data associated with a function which the code
1151generator will emit immediately before the function's entrypoint.
1152The purpose of this feature is to allow frontends to associate
1153language-specific runtime metadata with specific functions and make it
1154available through the function pointer while still allowing the
1155function pointer to be called.
1156
1157To access the data for a given function, a program may bitcast the
1158function pointer to a pointer to the constant's type and dereference
1159index -1. This implies that the IR symbol points just past the end of
1160the prefix data. For instance, take the example of a function annotated
1161with a single ``i32``,
1162
1163.. code-block:: llvm
1164
1165    define void @f() prefix i32 123 { ... }
1166
1167The prefix data can be referenced as,
1168
1169.. code-block:: llvm
1170
1171    %0 = bitcast void* () @f to i32*
1172    %a = getelementptr inbounds i32, i32* %0, i32 -1
1173    %b = load i32, i32* %a
1174
1175Prefix data is laid out as if it were an initializer for a global variable
1176of the prefix data's type. The function will be placed such that the
1177beginning of the prefix data is aligned. This means that if the size
1178of the prefix data is not a multiple of the alignment size, the
1179function's entrypoint will not be aligned. If alignment of the
1180function's entrypoint is desired, padding must be added to the prefix
1181data.
1182
1183A function may have prefix data but no body. This has similar semantics
1184to the ``available_externally`` linkage in that the data may be used by the
1185optimizers but will not be emitted in the object file.
1186
1187.. _prologuedata:
1188
1189Prologue Data
1190-------------
1191
1192The ``prologue`` attribute allows arbitrary code (encoded as bytes) to
1193be inserted prior to the function body. This can be used for enabling
1194function hot-patching and instrumentation.
1195
1196To maintain the semantics of ordinary function calls, the prologue data must
1197have a particular format. Specifically, it must begin with a sequence of
1198bytes which decode to a sequence of machine instructions, valid for the
1199module's target, which transfer control to the point immediately succeeding
1200the prologue data, without performing any other visible action. This allows
1201the inliner and other passes to reason about the semantics of the function
1202definition without needing to reason about the prologue data. Obviously this
1203makes the format of the prologue data highly target dependent.
1204
1205A trivial example of valid prologue data for the x86 architecture is ``i8 144``,
1206which encodes the ``nop`` instruction:
1207
1208.. code-block:: llvm
1209
1210    define void @f() prologue i8 144 { ... }
1211
1212Generally prologue data can be formed by encoding a relative branch instruction
1213which skips the metadata, as in this example of valid prologue data for the
1214x86_64 architecture, where the first two bytes encode ``jmp .+10``:
1215
1216.. code-block:: llvm
1217
1218    %0 = type <{ i8, i8, i8* }>
1219
1220    define void @f() prologue %0 <{ i8 235, i8 8, i8* @md}> { ... }
1221
1222A function may have prologue data but no body. This has similar semantics
1223to the ``available_externally`` linkage in that the data may be used by the
1224optimizers but will not be emitted in the object file.
1225
1226.. _personalityfn:
1227
1228Personality Function
1229--------------------
1230
1231The ``personality`` attribute permits functions to specify what function
1232to use for exception handling.
1233
1234.. _attrgrp:
1235
1236Attribute Groups
1237----------------
1238
1239Attribute groups are groups of attributes that are referenced by objects within
1240the IR. They are important for keeping ``.ll`` files readable, because a lot of
1241functions will use the same set of attributes. In the degenerative case of a
1242``.ll`` file that corresponds to a single ``.c`` file, the single attribute
1243group will capture the important command line flags used to build that file.
1244
1245An attribute group is a module-level object. To use an attribute group, an
1246object references the attribute group's ID (e.g. ``#37``). An object may refer
1247to more than one attribute group. In that situation, the attributes from the
1248different groups are merged.
1249
1250Here is an example of attribute groups for a function that should always be
1251inlined, has a stack alignment of 4, and which shouldn't use SSE instructions:
1252
1253.. code-block:: llvm
1254
1255   ; Target-independent attributes:
1256   attributes #0 = { alwaysinline alignstack=4 }
1257
1258   ; Target-dependent attributes:
1259   attributes #1 = { "no-sse" }
1260
1261   ; Function @f has attributes: alwaysinline, alignstack=4, and "no-sse".
1262   define void @f() #0 #1 { ... }
1263
1264.. _fnattrs:
1265
1266Function Attributes
1267-------------------
1268
1269Function attributes are set to communicate additional information about
1270a function. Function attributes are considered to be part of the
1271function, not of the function type, so functions with different function
1272attributes can have the same function type.
1273
1274Function attributes are simple keywords that follow the type specified.
1275If multiple attributes are needed, they are space separated. For
1276example:
1277
1278.. code-block:: llvm
1279
1280    define void @f() noinline { ... }
1281    define void @f() alwaysinline { ... }
1282    define void @f() alwaysinline optsize { ... }
1283    define void @f() optsize { ... }
1284
1285``alignstack(<n>)``
1286    This attribute indicates that, when emitting the prologue and
1287    epilogue, the backend should forcibly align the stack pointer.
1288    Specify the desired alignment, which must be a power of two, in
1289    parentheses.
1290``allocsize(<EltSizeParam>[, <NumEltsParam>])``
1291    This attribute indicates that the annotated function will always return at
1292    least a given number of bytes (or null). Its arguments are zero-indexed
1293    parameter numbers; if one argument is provided, then it's assumed that at
1294    least ``CallSite.Args[EltSizeParam]`` bytes will be available at the
1295    returned pointer. If two are provided, then it's assumed that
1296    ``CallSite.Args[EltSizeParam] * CallSite.Args[NumEltsParam]`` bytes are
1297    available. The referenced parameters must be integer types. No assumptions
1298    are made about the contents of the returned block of memory.
1299``alwaysinline``
1300    This attribute indicates that the inliner should attempt to inline
1301    this function into callers whenever possible, ignoring any active
1302    inlining size threshold for this caller.
1303``builtin``
1304    This indicates that the callee function at a call site should be
1305    recognized as a built-in function, even though the function's declaration
1306    uses the ``nobuiltin`` attribute. This is only valid at call sites for
1307    direct calls to functions that are declared with the ``nobuiltin``
1308    attribute.
1309``cold``
1310    This attribute indicates that this function is rarely called. When
1311    computing edge weights, basic blocks post-dominated by a cold
1312    function call are also considered to be cold; and, thus, given low
1313    weight.
1314``convergent``
1315    In some parallel execution models, there exist operations that cannot be
1316    made control-dependent on any additional values.  We call such operations
1317    ``convergent``, and mark them with this attribute.
1318
1319    The ``convergent`` attribute may appear on functions or call/invoke
1320    instructions.  When it appears on a function, it indicates that calls to
1321    this function should not be made control-dependent on additional values.
1322    For example, the intrinsic ``llvm.nvvm.barrier0`` is ``convergent``, so
1323    calls to this intrinsic cannot be made control-dependent on additional
1324    values.
1325
1326    When it appears on a call/invoke, the ``convergent`` attribute indicates
1327    that we should treat the call as though we're calling a convergent
1328    function.  This is particularly useful on indirect calls; without this we
1329    may treat such calls as though the target is non-convergent.
1330
1331    The optimizer may remove the ``convergent`` attribute on functions when it
1332    can prove that the function does not execute any convergent operations.
1333    Similarly, the optimizer may remove ``convergent`` on calls/invokes when it
1334    can prove that the call/invoke cannot call a convergent function.
1335``inaccessiblememonly``
1336    This attribute indicates that the function may only access memory that
1337    is not accessible by the module being compiled. This is a weaker form
1338    of ``readnone``.
1339``inaccessiblemem_or_argmemonly``
1340    This attribute indicates that the function may only access memory that is
1341    either not accessible by the module being compiled, or is pointed to
1342    by its pointer arguments. This is a weaker form of  ``argmemonly``
1343``inlinehint``
1344    This attribute indicates that the source code contained a hint that
1345    inlining this function is desirable (such as the "inline" keyword in
1346    C/C++). It is just a hint; it imposes no requirements on the
1347    inliner.
1348``jumptable``
1349    This attribute indicates that the function should be added to a
1350    jump-instruction table at code-generation time, and that all address-taken
1351    references to this function should be replaced with a reference to the
1352    appropriate jump-instruction-table function pointer. Note that this creates
1353    a new pointer for the original function, which means that code that depends
1354    on function-pointer identity can break. So, any function annotated with
1355    ``jumptable`` must also be ``unnamed_addr``.
1356``minsize``
1357    This attribute suggests that optimization passes and code generator
1358    passes make choices that keep the code size of this function as small
1359    as possible and perform optimizations that may sacrifice runtime
1360    performance in order to minimize the size of the generated code.
1361``naked``
1362    This attribute disables prologue / epilogue emission for the
1363    function. This can have very system-specific consequences.
1364``nobuiltin``
1365    This indicates that the callee function at a call site is not recognized as
1366    a built-in function. LLVM will retain the original call and not replace it
1367    with equivalent code based on the semantics of the built-in function, unless
1368    the call site uses the ``builtin`` attribute. This is valid at call sites
1369    and on function declarations and definitions.
1370``noduplicate``
1371    This attribute indicates that calls to the function cannot be
1372    duplicated. A call to a ``noduplicate`` function may be moved
1373    within its parent function, but may not be duplicated within
1374    its parent function.
1375
1376    A function containing a ``noduplicate`` call may still
1377    be an inlining candidate, provided that the call is not
1378    duplicated by inlining. That implies that the function has
1379    internal linkage and only has one call site, so the original
1380    call is dead after inlining.
1381``noimplicitfloat``
1382    This attributes disables implicit floating point instructions.
1383``noinline``
1384    This attribute indicates that the inliner should never inline this
1385    function in any situation. This attribute may not be used together
1386    with the ``alwaysinline`` attribute.
1387``nonlazybind``
1388    This attribute suppresses lazy symbol binding for the function. This
1389    may make calls to the function faster, at the cost of extra program
1390    startup time if the function is not called during program startup.
1391``noredzone``
1392    This attribute indicates that the code generator should not use a
1393    red zone, even if the target-specific ABI normally permits it.
1394``noreturn``
1395    This function attribute indicates that the function never returns
1396    normally. This produces undefined behavior at runtime if the
1397    function ever does dynamically return.
1398``norecurse``
1399    This function attribute indicates that the function does not call itself
1400    either directly or indirectly down any possible call path. This produces
1401    undefined behavior at runtime if the function ever does recurse.
1402``nounwind``
1403    This function attribute indicates that the function never raises an
1404    exception. If the function does raise an exception, its runtime
1405    behavior is undefined. However, functions marked nounwind may still
1406    trap or generate asynchronous exceptions. Exception handling schemes
1407    that are recognized by LLVM to handle asynchronous exceptions, such
1408    as SEH, will still provide their implementation defined semantics.
1409``optnone``
1410    This function attribute indicates that most optimization passes will skip
1411    this function, with the exception of interprocedural optimization passes.
1412    Code generation defaults to the "fast" instruction selector.
1413    This attribute cannot be used together with the ``alwaysinline``
1414    attribute; this attribute is also incompatible
1415    with the ``minsize`` attribute and the ``optsize`` attribute.
1416
1417    This attribute requires the ``noinline`` attribute to be specified on
1418    the function as well, so the function is never inlined into any caller.
1419    Only functions with the ``alwaysinline`` attribute are valid
1420    candidates for inlining into the body of this function.
1421``optsize``
1422    This attribute suggests that optimization passes and code generator
1423    passes make choices that keep the code size of this function low,
1424    and otherwise do optimizations specifically to reduce code size as
1425    long as they do not significantly impact runtime performance.
1426``"patchable-function"``
1427    This attribute tells the code generator that the code
1428    generated for this function needs to follow certain conventions that
1429    make it possible for a runtime function to patch over it later.
1430    The exact effect of this attribute depends on its string value,
1431    for which there currently is one legal possibility:
1432
1433     * ``"prologue-short-redirect"`` - This style of patchable
1434       function is intended to support patching a function prologue to
1435       redirect control away from the function in a thread safe
1436       manner.  It guarantees that the first instruction of the
1437       function will be large enough to accommodate a short jump
1438       instruction, and will be sufficiently aligned to allow being
1439       fully changed via an atomic compare-and-swap instruction.
1440       While the first requirement can be satisfied by inserting large
1441       enough NOP, LLVM can and will try to re-purpose an existing
1442       instruction (i.e. one that would have to be emitted anyway) as
1443       the patchable instruction larger than a short jump.
1444
1445       ``"prologue-short-redirect"`` is currently only supported on
1446       x86-64.
1447
1448    This attribute by itself does not imply restrictions on
1449    inter-procedural optimizations.  All of the semantic effects the
1450    patching may have to be separately conveyed via the linkage type.
1451``readnone``
1452    On a function, this attribute indicates that the function computes its
1453    result (or decides to unwind an exception) based strictly on its arguments,
1454    without dereferencing any pointer arguments or otherwise accessing
1455    any mutable state (e.g. memory, control registers, etc) visible to
1456    caller functions. It does not write through any pointer arguments
1457    (including ``byval`` arguments) and never changes any state visible
1458    to callers. This means that it cannot unwind exceptions by calling
1459    the ``C++`` exception throwing methods.
1460
1461    On an argument, this attribute indicates that the function does not
1462    dereference that pointer argument, even though it may read or write the
1463    memory that the pointer points to if accessed through other pointers.
1464``readonly``
1465    On a function, this attribute indicates that the function does not write
1466    through any pointer arguments (including ``byval`` arguments) or otherwise
1467    modify any state (e.g. memory, control registers, etc) visible to
1468    caller functions. It may dereference pointer arguments and read
1469    state that may be set in the caller. A readonly function always
1470    returns the same value (or unwinds an exception identically) when
1471    called with the same set of arguments and global state. It cannot
1472    unwind an exception by calling the ``C++`` exception throwing
1473    methods.
1474
1475    On an argument, this attribute indicates that the function does not write
1476    through this pointer argument, even though it may write to the memory that
1477    the pointer points to.
1478``writeonly``
1479    On a function, this attribute indicates that the function may write to but
1480    does not read from memory.
1481
1482    On an argument, this attribute indicates that the function may write to but
1483    does not read through this pointer argument (even though it may read from
1484    the memory that the pointer points to).
1485``argmemonly``
1486    This attribute indicates that the only memory accesses inside function are
1487    loads and stores from objects pointed to by its pointer-typed arguments,
1488    with arbitrary offsets. Or in other words, all memory operations in the
1489    function can refer to memory only using pointers based on its function
1490    arguments.
1491    Note that ``argmemonly`` can be used together with ``readonly`` attribute
1492    in order to specify that function reads only from its arguments.
1493``returns_twice``
1494    This attribute indicates that this function can return twice. The C
1495    ``setjmp`` is an example of such a function. The compiler disables
1496    some optimizations (like tail calls) in the caller of these
1497    functions.
1498``safestack``
1499    This attribute indicates that
1500    `SafeStack <http://clang.llvm.org/docs/SafeStack.html>`_
1501    protection is enabled for this function.
1502
1503    If a function that has a ``safestack`` attribute is inlined into a
1504    function that doesn't have a ``safestack`` attribute or which has an
1505    ``ssp``, ``sspstrong`` or ``sspreq`` attribute, then the resulting
1506    function will have a ``safestack`` attribute.
1507``sanitize_address``
1508    This attribute indicates that AddressSanitizer checks
1509    (dynamic address safety analysis) are enabled for this function.
1510``sanitize_memory``
1511    This attribute indicates that MemorySanitizer checks (dynamic detection
1512    of accesses to uninitialized memory) are enabled for this function.
1513``sanitize_thread``
1514    This attribute indicates that ThreadSanitizer checks
1515    (dynamic thread safety analysis) are enabled for this function.
1516``ssp``
1517    This attribute indicates that the function should emit a stack
1518    smashing protector. It is in the form of a "canary" --- a random value
1519    placed on the stack before the local variables that's checked upon
1520    return from the function to see if it has been overwritten. A
1521    heuristic is used to determine if a function needs stack protectors
1522    or not. The heuristic used will enable protectors for functions with:
1523
1524    - Character arrays larger than ``ssp-buffer-size`` (default 8).
1525    - Aggregates containing character arrays larger than ``ssp-buffer-size``.
1526    - Calls to alloca() with variable sizes or constant sizes greater than
1527      ``ssp-buffer-size``.
1528
1529    Variables that are identified as requiring a protector will be arranged
1530    on the stack such that they are adjacent to the stack protector guard.
1531
1532    If a function that has an ``ssp`` attribute is inlined into a
1533    function that doesn't have an ``ssp`` attribute, then the resulting
1534    function will have an ``ssp`` attribute.
1535``sspreq``
1536    This attribute indicates that the function should *always* emit a
1537    stack smashing protector. This overrides the ``ssp`` function
1538    attribute.
1539
1540    Variables that are identified as requiring a protector will be arranged
1541    on the stack such that they are adjacent to the stack protector guard.
1542    The specific layout rules are:
1543
1544    #. Large arrays and structures containing large arrays
1545       (``>= ssp-buffer-size``) are closest to the stack protector.
1546    #. Small arrays and structures containing small arrays
1547       (``< ssp-buffer-size``) are 2nd closest to the protector.
1548    #. Variables that have had their address taken are 3rd closest to the
1549       protector.
1550
1551    If a function that has an ``sspreq`` attribute is inlined into a
1552    function that doesn't have an ``sspreq`` attribute or which has an
1553    ``ssp`` or ``sspstrong`` attribute, then the resulting function will have
1554    an ``sspreq`` attribute.
1555``sspstrong``
1556    This attribute indicates that the function should emit a stack smashing
1557    protector. This attribute causes a strong heuristic to be used when
1558    determining if a function needs stack protectors. The strong heuristic
1559    will enable protectors for functions with:
1560
1561    - Arrays of any size and type
1562    - Aggregates containing an array of any size and type.
1563    - Calls to alloca().
1564    - Local variables that have had their address taken.
1565
1566    Variables that are identified as requiring a protector will be arranged
1567    on the stack such that they are adjacent to the stack protector guard.
1568    The specific layout rules are:
1569
1570    #. Large arrays and structures containing large arrays
1571       (``>= ssp-buffer-size``) are closest to the stack protector.
1572    #. Small arrays and structures containing small arrays
1573       (``< ssp-buffer-size``) are 2nd closest to the protector.
1574    #. Variables that have had their address taken are 3rd closest to the
1575       protector.
1576
1577    This overrides the ``ssp`` function attribute.
1578
1579    If a function that has an ``sspstrong`` attribute is inlined into a
1580    function that doesn't have an ``sspstrong`` attribute, then the
1581    resulting function will have an ``sspstrong`` attribute.
1582``"thunk"``
1583    This attribute indicates that the function will delegate to some other
1584    function with a tail call. The prototype of a thunk should not be used for
1585    optimization purposes. The caller is expected to cast the thunk prototype to
1586    match the thunk target prototype.
1587``uwtable``
1588    This attribute indicates that the ABI being targeted requires that
1589    an unwind table entry be produced for this function even if we can
1590    show that no exceptions passes by it. This is normally the case for
1591    the ELF x86-64 abi, but it can be disabled for some compilation
1592    units.
1593
1594
1595.. _opbundles:
1596
1597Operand Bundles
1598---------------
1599
1600Note: operand bundles are a work in progress, and they should be
1601considered experimental at this time.
1602
1603Operand bundles are tagged sets of SSA values that can be associated
1604with certain LLVM instructions (currently only ``call`` s and
1605``invoke`` s).  In a way they are like metadata, but dropping them is
1606incorrect and will change program semantics.
1607
1608Syntax::
1609
1610    operand bundle set ::= '[' operand bundle (, operand bundle )* ']'
1611    operand bundle ::= tag '(' [ bundle operand ] (, bundle operand )* ')'
1612    bundle operand ::= SSA value
1613    tag ::= string constant
1614
1615Operand bundles are **not** part of a function's signature, and a
1616given function may be called from multiple places with different kinds
1617of operand bundles.  This reflects the fact that the operand bundles
1618are conceptually a part of the ``call`` (or ``invoke``), not the
1619callee being dispatched to.
1620
1621Operand bundles are a generic mechanism intended to support
1622runtime-introspection-like functionality for managed languages.  While
1623the exact semantics of an operand bundle depend on the bundle tag,
1624there are certain limitations to how much the presence of an operand
1625bundle can influence the semantics of a program.  These restrictions
1626are described as the semantics of an "unknown" operand bundle.  As
1627long as the behavior of an operand bundle is describable within these
1628restrictions, LLVM does not need to have special knowledge of the
1629operand bundle to not miscompile programs containing it.
1630
1631- The bundle operands for an unknown operand bundle escape in unknown
1632  ways before control is transferred to the callee or invokee.
1633- Calls and invokes with operand bundles have unknown read / write
1634  effect on the heap on entry and exit (even if the call target is
1635  ``readnone`` or ``readonly``), unless they're overridden with
1636  callsite specific attributes.
1637- An operand bundle at a call site cannot change the implementation
1638  of the called function.  Inter-procedural optimizations work as
1639  usual as long as they take into account the first two properties.
1640
1641More specific types of operand bundles are described below.
1642
1643.. _deopt_opbundles:
1644
1645Deoptimization Operand Bundles
1646^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1647
1648Deoptimization operand bundles are characterized by the ``"deopt"``
1649operand bundle tag.  These operand bundles represent an alternate
1650"safe" continuation for the call site they're attached to, and can be
1651used by a suitable runtime to deoptimize the compiled frame at the
1652specified call site.  There can be at most one ``"deopt"`` operand
1653bundle attached to a call site.  Exact details of deoptimization is
1654out of scope for the language reference, but it usually involves
1655rewriting a compiled frame into a set of interpreted frames.
1656
1657From the compiler's perspective, deoptimization operand bundles make
1658the call sites they're attached to at least ``readonly``.  They read
1659through all of their pointer typed operands (even if they're not
1660otherwise escaped) and the entire visible heap.  Deoptimization
1661operand bundles do not capture their operands except during
1662deoptimization, in which case control will not be returned to the
1663compiled frame.
1664
1665The inliner knows how to inline through calls that have deoptimization
1666operand bundles.  Just like inlining through a normal call site
1667involves composing the normal and exceptional continuations, inlining
1668through a call site with a deoptimization operand bundle needs to
1669appropriately compose the "safe" deoptimization continuation.  The
1670inliner does this by prepending the parent's deoptimization
1671continuation to every deoptimization continuation in the inlined body.
1672E.g. inlining ``@f`` into ``@g`` in the following example
1673
1674.. code-block:: llvm
1675
1676    define void @f() {
1677      call void @x()  ;; no deopt state
1678      call void @y() [ "deopt"(i32 10) ]
1679      call void @y() [ "deopt"(i32 10), "unknown"(i8* null) ]
1680      ret void
1681    }
1682
1683    define void @g() {
1684      call void @f() [ "deopt"(i32 20) ]
1685      ret void
1686    }
1687
1688will result in
1689
1690.. code-block:: llvm
1691
1692    define void @g() {
1693      call void @x()  ;; still no deopt state
1694      call void @y() [ "deopt"(i32 20, i32 10) ]
1695      call void @y() [ "deopt"(i32 20, i32 10), "unknown"(i8* null) ]
1696      ret void
1697    }
1698
1699It is the frontend's responsibility to structure or encode the
1700deoptimization state in a way that syntactically prepending the
1701caller's deoptimization state to the callee's deoptimization state is
1702semantically equivalent to composing the caller's deoptimization
1703continuation after the callee's deoptimization continuation.
1704
1705.. _ob_funclet:
1706
1707Funclet Operand Bundles
1708^^^^^^^^^^^^^^^^^^^^^^^
1709
1710Funclet operand bundles are characterized by the ``"funclet"``
1711operand bundle tag.  These operand bundles indicate that a call site
1712is within a particular funclet.  There can be at most one
1713``"funclet"`` operand bundle attached to a call site and it must have
1714exactly one bundle operand.
1715
1716If any funclet EH pads have been "entered" but not "exited" (per the
1717`description in the EH doc\ <ExceptionHandling.html#wineh-constraints>`_),
1718it is undefined behavior to execute a ``call`` or ``invoke`` which:
1719
1720* does not have a ``"funclet"`` bundle and is not a ``call`` to a nounwind
1721  intrinsic, or
1722* has a ``"funclet"`` bundle whose operand is not the most-recently-entered
1723  not-yet-exited funclet EH pad.
1724
1725Similarly, if no funclet EH pads have been entered-but-not-yet-exited,
1726executing a ``call`` or ``invoke`` with a ``"funclet"`` bundle is undefined behavior.
1727
1728GC Transition Operand Bundles
1729^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1730
1731GC transition operand bundles are characterized by the
1732``"gc-transition"`` operand bundle tag. These operand bundles mark a
1733call as a transition between a function with one GC strategy to a
1734function with a different GC strategy. If coordinating the transition
1735between GC strategies requires additional code generation at the call
1736site, these bundles may contain any values that are needed by the
1737generated code.  For more details, see :ref:`GC Transitions
1738<gc_transition_args>`.
1739
1740.. _moduleasm:
1741
1742Module-Level Inline Assembly
1743----------------------------
1744
1745Modules may contain "module-level inline asm" blocks, which corresponds
1746to the GCC "file scope inline asm" blocks. These blocks are internally
1747concatenated by LLVM and treated as a single unit, but may be separated
1748in the ``.ll`` file if desired. The syntax is very simple:
1749
1750.. code-block:: llvm
1751
1752    module asm "inline asm code goes here"
1753    module asm "more can go here"
1754
1755The strings can contain any character by escaping non-printable
1756characters. The escape sequence used is simply "\\xx" where "xx" is the
1757two digit hex code for the number.
1758
1759Note that the assembly string *must* be parseable by LLVM's integrated assembler
1760(unless it is disabled), even when emitting a ``.s`` file.
1761
1762.. _langref_datalayout:
1763
1764Data Layout
1765-----------
1766
1767A module may specify a target specific data layout string that specifies
1768how data is to be laid out in memory. The syntax for the data layout is
1769simply:
1770
1771.. code-block:: llvm
1772
1773    target datalayout = "layout specification"
1774
1775The *layout specification* consists of a list of specifications
1776separated by the minus sign character ('-'). Each specification starts
1777with a letter and may include other information after the letter to
1778define some aspect of the data layout. The specifications accepted are
1779as follows:
1780
1781``E``
1782    Specifies that the target lays out data in big-endian form. That is,
1783    the bits with the most significance have the lowest address
1784    location.
1785``e``
1786    Specifies that the target lays out data in little-endian form. That
1787    is, the bits with the least significance have the lowest address
1788    location.
1789``S<size>``
1790    Specifies the natural alignment of the stack in bits. Alignment
1791    promotion of stack variables is limited to the natural stack
1792    alignment to avoid dynamic stack realignment. The stack alignment
1793    must be a multiple of 8-bits. If omitted, the natural stack
1794    alignment defaults to "unspecified", which does not prevent any
1795    alignment promotions.
1796``p[n]:<size>:<abi>:<pref>``
1797    This specifies the *size* of a pointer and its ``<abi>`` and
1798    ``<pref>``\erred alignments for address space ``n``. All sizes are in
1799    bits. The address space, ``n``, is optional, and if not specified,
1800    denotes the default address space 0. The value of ``n`` must be
1801    in the range [1,2^23).
1802``i<size>:<abi>:<pref>``
1803    This specifies the alignment for an integer type of a given bit
1804    ``<size>``. The value of ``<size>`` must be in the range [1,2^23).
1805``v<size>:<abi>:<pref>``
1806    This specifies the alignment for a vector type of a given bit
1807    ``<size>``.
1808``f<size>:<abi>:<pref>``
1809    This specifies the alignment for a floating point type of a given bit
1810    ``<size>``. Only values of ``<size>`` that are supported by the target
1811    will work. 32 (float) and 64 (double) are supported on all targets; 80
1812    or 128 (different flavors of long double) are also supported on some
1813    targets.
1814``a:<abi>:<pref>``
1815    This specifies the alignment for an object of aggregate type.
1816``m:<mangling>``
1817    If present, specifies that llvm names are mangled in the output. The
1818    options are
1819
1820    * ``e``: ELF mangling: Private symbols get a ``.L`` prefix.
1821    * ``m``: Mips mangling: Private symbols get a ``$`` prefix.
1822    * ``o``: Mach-O mangling: Private symbols get ``L`` prefix. Other
1823      symbols get a ``_`` prefix.
1824    * ``w``: Windows COFF prefix:  Similar to Mach-O, but stdcall and fastcall
1825      functions also get a suffix based on the frame size.
1826    * ``x``: Windows x86 COFF prefix:  Similar to Windows COFF, but use a ``_``
1827      prefix for ``__cdecl`` functions.
1828``n<size1>:<size2>:<size3>...``
1829    This specifies a set of native integer widths for the target CPU in
1830    bits. For example, it might contain ``n32`` for 32-bit PowerPC,
1831    ``n32:64`` for PowerPC 64, or ``n8:16:32:64`` for X86-64. Elements of
1832    this set are considered to support most general arithmetic operations
1833    efficiently.
1834
1835On every specification that takes a ``<abi>:<pref>``, specifying the
1836``<pref>`` alignment is optional. If omitted, the preceding ``:``
1837should be omitted too and ``<pref>`` will be equal to ``<abi>``.
1838
1839When constructing the data layout for a given target, LLVM starts with a
1840default set of specifications which are then (possibly) overridden by
1841the specifications in the ``datalayout`` keyword. The default
1842specifications are given in this list:
1843
1844-  ``E`` - big endian
1845-  ``p:64:64:64`` - 64-bit pointers with 64-bit alignment.
1846-  ``p[n]:64:64:64`` - Other address spaces are assumed to be the
1847   same as the default address space.
1848-  ``S0`` - natural stack alignment is unspecified
1849-  ``i1:8:8`` - i1 is 8-bit (byte) aligned
1850-  ``i8:8:8`` - i8 is 8-bit (byte) aligned
1851-  ``i16:16:16`` - i16 is 16-bit aligned
1852-  ``i32:32:32`` - i32 is 32-bit aligned
1853-  ``i64:32:64`` - i64 has ABI alignment of 32-bits but preferred
1854   alignment of 64-bits
1855-  ``f16:16:16`` - half is 16-bit aligned
1856-  ``f32:32:32`` - float is 32-bit aligned
1857-  ``f64:64:64`` - double is 64-bit aligned
1858-  ``f128:128:128`` - quad is 128-bit aligned
1859-  ``v64:64:64`` - 64-bit vector is 64-bit aligned
1860-  ``v128:128:128`` - 128-bit vector is 128-bit aligned
1861-  ``a:0:64`` - aggregates are 64-bit aligned
1862
1863When LLVM is determining the alignment for a given type, it uses the
1864following rules:
1865
1866#. If the type sought is an exact match for one of the specifications,
1867   that specification is used.
1868#. If no match is found, and the type sought is an integer type, then
1869   the smallest integer type that is larger than the bitwidth of the
1870   sought type is used. If none of the specifications are larger than
1871   the bitwidth then the largest integer type is used. For example,
1872   given the default specifications above, the i7 type will use the
1873   alignment of i8 (next largest) while both i65 and i256 will use the
1874   alignment of i64 (largest specified).
1875#. If no match is found, and the type sought is a vector type, then the
1876   largest vector type that is smaller than the sought vector type will
1877   be used as a fall back. This happens because <128 x double> can be
1878   implemented in terms of 64 <2 x double>, for example.
1879
1880The function of the data layout string may not be what you expect.
1881Notably, this is not a specification from the frontend of what alignment
1882the code generator should use.
1883
1884Instead, if specified, the target data layout is required to match what
1885the ultimate *code generator* expects. This string is used by the
1886mid-level optimizers to improve code, and this only works if it matches
1887what the ultimate code generator uses. There is no way to generate IR
1888that does not embed this target-specific detail into the IR. If you
1889don't specify the string, the default specifications will be used to
1890generate a Data Layout and the optimization phases will operate
1891accordingly and introduce target specificity into the IR with respect to
1892these default specifications.
1893
1894.. _langref_triple:
1895
1896Target Triple
1897-------------
1898
1899A module may specify a target triple string that describes the target
1900host. The syntax for the target triple is simply:
1901
1902.. code-block:: llvm
1903
1904    target triple = "x86_64-apple-macosx10.7.0"
1905
1906The *target triple* string consists of a series of identifiers delimited
1907by the minus sign character ('-'). The canonical forms are:
1908
1909::
1910
1911    ARCHITECTURE-VENDOR-OPERATING_SYSTEM
1912    ARCHITECTURE-VENDOR-OPERATING_SYSTEM-ENVIRONMENT
1913
1914This information is passed along to the backend so that it generates
1915code for the proper architecture. It's possible to override this on the
1916command line with the ``-mtriple`` command line option.
1917
1918.. _pointeraliasing:
1919
1920Pointer Aliasing Rules
1921----------------------
1922
1923Any memory access must be done through a pointer value associated with
1924an address range of the memory access, otherwise the behavior is
1925undefined. Pointer values are associated with address ranges according
1926to the following rules:
1927
1928-  A pointer value is associated with the addresses associated with any
1929   value it is *based* on.
1930-  An address of a global variable is associated with the address range
1931   of the variable's storage.
1932-  The result value of an allocation instruction is associated with the
1933   address range of the allocated storage.
1934-  A null pointer in the default address-space is associated with no
1935   address.
1936-  An integer constant other than zero or a pointer value returned from
1937   a function not defined within LLVM may be associated with address
1938   ranges allocated through mechanisms other than those provided by
1939   LLVM. Such ranges shall not overlap with any ranges of addresses
1940   allocated by mechanisms provided by LLVM.
1941
1942A pointer value is *based* on another pointer value according to the
1943following rules:
1944
1945-  A pointer value formed from a ``getelementptr`` operation is *based*
1946   on the first value operand of the ``getelementptr``.
1947-  The result value of a ``bitcast`` is *based* on the operand of the
1948   ``bitcast``.
1949-  A pointer value formed by an ``inttoptr`` is *based* on all pointer
1950   values that contribute (directly or indirectly) to the computation of
1951   the pointer's value.
1952-  The "*based* on" relationship is transitive.
1953
1954Note that this definition of *"based"* is intentionally similar to the
1955definition of *"based"* in C99, though it is slightly weaker.
1956
1957LLVM IR does not associate types with memory. The result type of a
1958``load`` merely indicates the size and alignment of the memory from
1959which to load, as well as the interpretation of the value. The first
1960operand type of a ``store`` similarly only indicates the size and
1961alignment of the store.
1962
1963Consequently, type-based alias analysis, aka TBAA, aka
1964``-fstrict-aliasing``, is not applicable to general unadorned LLVM IR.
1965:ref:`Metadata <metadata>` may be used to encode additional information
1966which specialized optimization passes may use to implement type-based
1967alias analysis.
1968
1969.. _volatile:
1970
1971Volatile Memory Accesses
1972------------------------
1973
1974Certain memory accesses, such as :ref:`load <i_load>`'s,
1975:ref:`store <i_store>`'s, and :ref:`llvm.memcpy <int_memcpy>`'s may be
1976marked ``volatile``. The optimizers must not change the number of
1977volatile operations or change their order of execution relative to other
1978volatile operations. The optimizers *may* change the order of volatile
1979operations relative to non-volatile operations. This is not Java's
1980"volatile" and has no cross-thread synchronization behavior.
1981
1982IR-level volatile loads and stores cannot safely be optimized into
1983llvm.memcpy or llvm.memmove intrinsics even when those intrinsics are
1984flagged volatile. Likewise, the backend should never split or merge
1985target-legal volatile load/store instructions.
1986
1987.. admonition:: Rationale
1988
1989 Platforms may rely on volatile loads and stores of natively supported
1990 data width to be executed as single instruction. For example, in C
1991 this holds for an l-value of volatile primitive type with native
1992 hardware support, but not necessarily for aggregate types. The
1993 frontend upholds these expectations, which are intentionally
1994 unspecified in the IR. The rules above ensure that IR transformations
1995 do not violate the frontend's contract with the language.
1996
1997.. _memmodel:
1998
1999Memory Model for Concurrent Operations
2000--------------------------------------
2001
2002The LLVM IR does not define any way to start parallel threads of
2003execution or to register signal handlers. Nonetheless, there are
2004platform-specific ways to create them, and we define LLVM IR's behavior
2005in their presence. This model is inspired by the C++0x memory model.
2006
2007For a more informal introduction to this model, see the :doc:`Atomics`.
2008
2009We define a *happens-before* partial order as the least partial order
2010that
2011
2012-  Is a superset of single-thread program order, and
2013-  When a *synchronizes-with* ``b``, includes an edge from ``a`` to
2014   ``b``. *Synchronizes-with* pairs are introduced by platform-specific
2015   techniques, like pthread locks, thread creation, thread joining,
2016   etc., and by atomic instructions. (See also :ref:`Atomic Memory Ordering
2017   Constraints <ordering>`).
2018
2019Note that program order does not introduce *happens-before* edges
2020between a thread and signals executing inside that thread.
2021
2022Every (defined) read operation (load instructions, memcpy, atomic
2023loads/read-modify-writes, etc.) R reads a series of bytes written by
2024(defined) write operations (store instructions, atomic
2025stores/read-modify-writes, memcpy, etc.). For the purposes of this
2026section, initialized globals are considered to have a write of the
2027initializer which is atomic and happens before any other read or write
2028of the memory in question. For each byte of a read R, R\ :sub:`byte`
2029may see any write to the same byte, except:
2030
2031-  If write\ :sub:`1`  happens before write\ :sub:`2`, and
2032   write\ :sub:`2` happens before R\ :sub:`byte`, then
2033   R\ :sub:`byte` does not see write\ :sub:`1`.
2034-  If R\ :sub:`byte` happens before write\ :sub:`3`, then
2035   R\ :sub:`byte` does not see write\ :sub:`3`.
2036
2037Given that definition, R\ :sub:`byte` is defined as follows:
2038
2039-  If R is volatile, the result is target-dependent. (Volatile is
2040   supposed to give guarantees which can support ``sig_atomic_t`` in
2041   C/C++, and may be used for accesses to addresses that do not behave
2042   like normal memory. It does not generally provide cross-thread
2043   synchronization.)
2044-  Otherwise, if there is no write to the same byte that happens before
2045   R\ :sub:`byte`, R\ :sub:`byte` returns ``undef`` for that byte.
2046-  Otherwise, if R\ :sub:`byte` may see exactly one write,
2047   R\ :sub:`byte` returns the value written by that write.
2048-  Otherwise, if R is atomic, and all the writes R\ :sub:`byte` may
2049   see are atomic, it chooses one of the values written. See the :ref:`Atomic
2050   Memory Ordering Constraints <ordering>` section for additional
2051   constraints on how the choice is made.
2052-  Otherwise R\ :sub:`byte` returns ``undef``.
2053
2054R returns the value composed of the series of bytes it read. This
2055implies that some bytes within the value may be ``undef`` **without**
2056the entire value being ``undef``. Note that this only defines the
2057semantics of the operation; it doesn't mean that targets will emit more
2058than one instruction to read the series of bytes.
2059
2060Note that in cases where none of the atomic intrinsics are used, this
2061model places only one restriction on IR transformations on top of what
2062is required for single-threaded execution: introducing a store to a byte
2063which might not otherwise be stored is not allowed in general.
2064(Specifically, in the case where another thread might write to and read
2065from an address, introducing a store can change a load that may see
2066exactly one write into a load that may see multiple writes.)
2067
2068.. _ordering:
2069
2070Atomic Memory Ordering Constraints
2071----------------------------------
2072
2073Atomic instructions (:ref:`cmpxchg <i_cmpxchg>`,
2074:ref:`atomicrmw <i_atomicrmw>`, :ref:`fence <i_fence>`,
2075:ref:`atomic load <i_load>`, and :ref:`atomic store <i_store>`) take
2076ordering parameters that determine which other atomic instructions on
2077the same address they *synchronize with*. These semantics are borrowed
2078from Java and C++0x, but are somewhat more colloquial. If these
2079descriptions aren't precise enough, check those specs (see spec
2080references in the :doc:`atomics guide <Atomics>`).
2081:ref:`fence <i_fence>` instructions treat these orderings somewhat
2082differently since they don't take an address. See that instruction's
2083documentation for details.
2084
2085For a simpler introduction to the ordering constraints, see the
2086:doc:`Atomics`.
2087
2088``unordered``
2089    The set of values that can be read is governed by the happens-before
2090    partial order. A value cannot be read unless some operation wrote
2091    it. This is intended to provide a guarantee strong enough to model
2092    Java's non-volatile shared variables. This ordering cannot be
2093    specified for read-modify-write operations; it is not strong enough
2094    to make them atomic in any interesting way.
2095``monotonic``
2096    In addition to the guarantees of ``unordered``, there is a single
2097    total order for modifications by ``monotonic`` operations on each
2098    address. All modification orders must be compatible with the
2099    happens-before order. There is no guarantee that the modification
2100    orders can be combined to a global total order for the whole program
2101    (and this often will not be possible). The read in an atomic
2102    read-modify-write operation (:ref:`cmpxchg <i_cmpxchg>` and
2103    :ref:`atomicrmw <i_atomicrmw>`) reads the value in the modification
2104    order immediately before the value it writes. If one atomic read
2105    happens before another atomic read of the same address, the later
2106    read must see the same value or a later value in the address's
2107    modification order. This disallows reordering of ``monotonic`` (or
2108    stronger) operations on the same address. If an address is written
2109    ``monotonic``-ally by one thread, and other threads ``monotonic``-ally
2110    read that address repeatedly, the other threads must eventually see
2111    the write. This corresponds to the C++0x/C1x
2112    ``memory_order_relaxed``.
2113``acquire``
2114    In addition to the guarantees of ``monotonic``, a
2115    *synchronizes-with* edge may be formed with a ``release`` operation.
2116    This is intended to model C++'s ``memory_order_acquire``.
2117``release``
2118    In addition to the guarantees of ``monotonic``, if this operation
2119    writes a value which is subsequently read by an ``acquire``
2120    operation, it *synchronizes-with* that operation. (This isn't a
2121    complete description; see the C++0x definition of a release
2122    sequence.) This corresponds to the C++0x/C1x
2123    ``memory_order_release``.
2124``acq_rel`` (acquire+release)
2125    Acts as both an ``acquire`` and ``release`` operation on its
2126    address. This corresponds to the C++0x/C1x ``memory_order_acq_rel``.
2127``seq_cst`` (sequentially consistent)
2128    In addition to the guarantees of ``acq_rel`` (``acquire`` for an
2129    operation that only reads, ``release`` for an operation that only
2130    writes), there is a global total order on all
2131    sequentially-consistent operations on all addresses, which is
2132    consistent with the *happens-before* partial order and with the
2133    modification orders of all the affected addresses. Each
2134    sequentially-consistent read sees the last preceding write to the
2135    same address in this global order. This corresponds to the C++0x/C1x
2136    ``memory_order_seq_cst`` and Java volatile.
2137
2138.. _singlethread:
2139
2140If an atomic operation is marked ``singlethread``, it only *synchronizes
2141with* or participates in modification and seq\_cst total orderings with
2142other operations running in the same thread (for example, in signal
2143handlers).
2144
2145.. _fastmath:
2146
2147Fast-Math Flags
2148---------------
2149
2150LLVM IR floating-point binary ops (:ref:`fadd <i_fadd>`,
2151:ref:`fsub <i_fsub>`, :ref:`fmul <i_fmul>`, :ref:`fdiv <i_fdiv>`,
2152:ref:`frem <i_frem>`, :ref:`fcmp <i_fcmp>`) have the following flags that can
2153be set to enable otherwise unsafe floating point operations
2154
2155``nnan``
2156   No NaNs - Allow optimizations to assume the arguments and result are not
2157   NaN. Such optimizations are required to retain defined behavior over
2158   NaNs, but the value of the result is undefined.
2159
2160``ninf``
2161   No Infs - Allow optimizations to assume the arguments and result are not
2162   +/-Inf. Such optimizations are required to retain defined behavior over
2163   +/-Inf, but the value of the result is undefined.
2164
2165``nsz``
2166   No Signed Zeros - Allow optimizations to treat the sign of a zero
2167   argument or result as insignificant.
2168
2169``arcp``
2170   Allow Reciprocal - Allow optimizations to use the reciprocal of an
2171   argument rather than perform division.
2172
2173``fast``
2174   Fast - Allow algebraically equivalent transformations that may
2175   dramatically change results in floating point (e.g. reassociate). This
2176   flag implies all the others.
2177
2178.. _uselistorder:
2179
2180Use-list Order Directives
2181-------------------------
2182
2183Use-list directives encode the in-memory order of each use-list, allowing the
2184order to be recreated. ``<order-indexes>`` is a comma-separated list of
2185indexes that are assigned to the referenced value's uses. The referenced
2186value's use-list is immediately sorted by these indexes.
2187
2188Use-list directives may appear at function scope or global scope. They are not
2189instructions, and have no effect on the semantics of the IR. When they're at
2190function scope, they must appear after the terminator of the final basic block.
2191
2192If basic blocks have their address taken via ``blockaddress()`` expressions,
2193``uselistorder_bb`` can be used to reorder their use-lists from outside their
2194function's scope.
2195
2196:Syntax:
2197
2198::
2199
2200    uselistorder <ty> <value>, { <order-indexes> }
2201    uselistorder_bb @function, %block { <order-indexes> }
2202
2203:Examples:
2204
2205::
2206
2207    define void @foo(i32 %arg1, i32 %arg2) {
2208    entry:
2209      ; ... instructions ...
2210    bb:
2211      ; ... instructions ...
2212
2213      ; At function scope.
2214      uselistorder i32 %arg1, { 1, 0, 2 }
2215      uselistorder label %bb, { 1, 0 }
2216    }
2217
2218    ; At global scope.
2219    uselistorder i32* @global, { 1, 2, 0 }
2220    uselistorder i32 7, { 1, 0 }
2221    uselistorder i32 (i32) @bar, { 1, 0 }
2222    uselistorder_bb @foo, %bb, { 5, 1, 3, 2, 0, 4 }
2223
2224.. _source_filename:
2225
2226Source Filename
2227---------------
2228
2229The *source filename* string is set to the original module identifier,
2230which will be the name of the compiled source file when compiling from
2231source through the clang front end, for example. It is then preserved through
2232the IR and bitcode.
2233
2234This is currently necessary to generate a consistent unique global
2235identifier for local functions used in profile data, which prepends the
2236source file name to the local function name.
2237
2238The syntax for the source file name is simply:
2239
2240.. code-block:: llvm
2241
2242    source_filename = "/path/to/source.c"
2243
2244.. _typesystem:
2245
2246Type System
2247===========
2248
2249The LLVM type system is one of the most important features of the
2250intermediate representation. Being typed enables a number of
2251optimizations to be performed on the intermediate representation
2252directly, without having to do extra analyses on the side before the
2253transformation. A strong type system makes it easier to read the
2254generated code and enables novel analyses and transformations that are
2255not feasible to perform on normal three address code representations.
2256
2257.. _t_void:
2258
2259Void Type
2260---------
2261
2262:Overview:
2263
2264
2265The void type does not represent any value and has no size.
2266
2267:Syntax:
2268
2269
2270::
2271
2272      void
2273
2274
2275.. _t_function:
2276
2277Function Type
2278-------------
2279
2280:Overview:
2281
2282
2283The function type can be thought of as a function signature. It consists of a
2284return type and a list of formal parameter types. The return type of a function
2285type is a void type or first class type --- except for :ref:`label <t_label>`
2286and :ref:`metadata <t_metadata>` types.
2287
2288:Syntax:
2289
2290::
2291
2292      <returntype> (<parameter list>)
2293
2294...where '``<parameter list>``' is a comma-separated list of type
2295specifiers. Optionally, the parameter list may include a type ``...``, which
2296indicates that the function takes a variable number of arguments. Variable
2297argument functions can access their arguments with the :ref:`variable argument
2298handling intrinsic <int_varargs>` functions. '``<returntype>``' is any type
2299except :ref:`label <t_label>` and :ref:`metadata <t_metadata>`.
2300
2301:Examples:
2302
2303+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2304| ``i32 (i32)``                   | function taking an ``i32``, returning an ``i32``                                                                                                                    |
2305+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2306| ``float (i16, i32 *) *``        | :ref:`Pointer <t_pointer>` to a function that takes an ``i16`` and a :ref:`pointer <t_pointer>` to ``i32``, returning ``float``.                                    |
2307+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2308| ``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. |
2309+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2310| ``{i32, i32} (i32)``            | A function taking an ``i32``, returning a :ref:`structure <t_struct>` containing two ``i32`` values                                                                 |
2311+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2312
2313.. _t_firstclass:
2314
2315First Class Types
2316-----------------
2317
2318The :ref:`first class <t_firstclass>` types are perhaps the most important.
2319Values of these types are the only ones which can be produced by
2320instructions.
2321
2322.. _t_single_value:
2323
2324Single Value Types
2325^^^^^^^^^^^^^^^^^^
2326
2327These are the types that are valid in registers from CodeGen's perspective.
2328
2329.. _t_integer:
2330
2331Integer Type
2332""""""""""""
2333
2334:Overview:
2335
2336The integer type is a very simple type that simply specifies an
2337arbitrary bit width for the integer type desired. Any bit width from 1
2338bit to 2\ :sup:`23`\ -1 (about 8 million) can be specified.
2339
2340:Syntax:
2341
2342::
2343
2344      iN
2345
2346The number of bits the integer will occupy is specified by the ``N``
2347value.
2348
2349Examples:
2350*********
2351
2352+----------------+------------------------------------------------+
2353| ``i1``         | a single-bit integer.                          |
2354+----------------+------------------------------------------------+
2355| ``i32``        | a 32-bit integer.                              |
2356+----------------+------------------------------------------------+
2357| ``i1942652``   | a really big integer of over 1 million bits.   |
2358+----------------+------------------------------------------------+
2359
2360.. _t_floating:
2361
2362Floating Point Types
2363""""""""""""""""""""
2364
2365.. list-table::
2366   :header-rows: 1
2367
2368   * - Type
2369     - Description
2370
2371   * - ``half``
2372     - 16-bit floating point value
2373
2374   * - ``float``
2375     - 32-bit floating point value
2376
2377   * - ``double``
2378     - 64-bit floating point value
2379
2380   * - ``fp128``
2381     - 128-bit floating point value (112-bit mantissa)
2382
2383   * - ``x86_fp80``
2384     -  80-bit floating point value (X87)
2385
2386   * - ``ppc_fp128``
2387     - 128-bit floating point value (two 64-bits)
2388
2389X86_mmx Type
2390""""""""""""
2391
2392:Overview:
2393
2394The x86_mmx type represents a value held in an MMX register on an x86
2395machine. The operations allowed on it are quite limited: parameters and
2396return values, load and store, and bitcast. User-specified MMX
2397instructions are represented as intrinsic or asm calls with arguments
2398and/or results of this type. There are no arrays, vectors or constants
2399of this type.
2400
2401:Syntax:
2402
2403::
2404
2405      x86_mmx
2406
2407
2408.. _t_pointer:
2409
2410Pointer Type
2411""""""""""""
2412
2413:Overview:
2414
2415The pointer type is used to specify memory locations. Pointers are
2416commonly used to reference objects in memory.
2417
2418Pointer types may have an optional address space attribute defining the
2419numbered address space where the pointed-to object resides. The default
2420address space is number zero. The semantics of non-zero address spaces
2421are target-specific.
2422
2423Note that LLVM does not permit pointers to void (``void*``) nor does it
2424permit pointers to labels (``label*``). Use ``i8*`` instead.
2425
2426:Syntax:
2427
2428::
2429
2430      <type> *
2431
2432:Examples:
2433
2434+-------------------------+--------------------------------------------------------------------------------------------------------------+
2435| ``[4 x i32]*``          | A :ref:`pointer <t_pointer>` to :ref:`array <t_array>` of four ``i32`` values.                               |
2436+-------------------------+--------------------------------------------------------------------------------------------------------------+
2437| ``i32 (i32*) *``        | A :ref:`pointer <t_pointer>` to a :ref:`function <t_function>` that takes an ``i32*``, returning an ``i32``. |
2438+-------------------------+--------------------------------------------------------------------------------------------------------------+
2439| ``i32 addrspace(5)*``   | A :ref:`pointer <t_pointer>` to an ``i32`` value that resides in address space #5.                           |
2440+-------------------------+--------------------------------------------------------------------------------------------------------------+
2441
2442.. _t_vector:
2443
2444Vector Type
2445"""""""""""
2446
2447:Overview:
2448
2449A vector type is a simple derived type that represents a vector of
2450elements. Vector types are used when multiple primitive data are
2451operated in parallel using a single instruction (SIMD). A vector type
2452requires a size (number of elements) and an underlying primitive data
2453type. Vector types are considered :ref:`first class <t_firstclass>`.
2454
2455:Syntax:
2456
2457::
2458
2459      < <# elements> x <elementtype> >
2460
2461The number of elements is a constant integer value larger than 0;
2462elementtype may be any integer, floating point or pointer type. Vectors
2463of size zero are not allowed.
2464
2465:Examples:
2466
2467+-------------------+--------------------------------------------------+
2468| ``<4 x i32>``     | Vector of 4 32-bit integer values.               |
2469+-------------------+--------------------------------------------------+
2470| ``<8 x float>``   | Vector of 8 32-bit floating-point values.        |
2471+-------------------+--------------------------------------------------+
2472| ``<2 x i64>``     | Vector of 2 64-bit integer values.               |
2473+-------------------+--------------------------------------------------+
2474| ``<4 x i64*>``    | Vector of 4 pointers to 64-bit integer values.   |
2475+-------------------+--------------------------------------------------+
2476
2477.. _t_label:
2478
2479Label Type
2480^^^^^^^^^^
2481
2482:Overview:
2483
2484The label type represents code labels.
2485
2486:Syntax:
2487
2488::
2489
2490      label
2491
2492.. _t_token:
2493
2494Token Type
2495^^^^^^^^^^
2496
2497:Overview:
2498
2499The token type is used when a value is associated with an instruction
2500but all uses of the value must not attempt to introspect or obscure it.
2501As such, it is not appropriate to have a :ref:`phi <i_phi>` or
2502:ref:`select <i_select>` of type token.
2503
2504:Syntax:
2505
2506::
2507
2508      token
2509
2510
2511
2512.. _t_metadata:
2513
2514Metadata Type
2515^^^^^^^^^^^^^
2516
2517:Overview:
2518
2519The metadata type represents embedded metadata. No derived types may be
2520created from metadata except for :ref:`function <t_function>` arguments.
2521
2522:Syntax:
2523
2524::
2525
2526      metadata
2527
2528.. _t_aggregate:
2529
2530Aggregate Types
2531^^^^^^^^^^^^^^^
2532
2533Aggregate Types are a subset of derived types that can contain multiple
2534member types. :ref:`Arrays <t_array>` and :ref:`structs <t_struct>` are
2535aggregate types. :ref:`Vectors <t_vector>` are not considered to be
2536aggregate types.
2537
2538.. _t_array:
2539
2540Array Type
2541""""""""""
2542
2543:Overview:
2544
2545The array type is a very simple derived type that arranges elements
2546sequentially in memory. The array type requires a size (number of
2547elements) and an underlying data type.
2548
2549:Syntax:
2550
2551::
2552
2553      [<# elements> x <elementtype>]
2554
2555The number of elements is a constant integer value; ``elementtype`` may
2556be any type with a size.
2557
2558:Examples:
2559
2560+------------------+--------------------------------------+
2561| ``[40 x i32]``   | Array of 40 32-bit integer values.   |
2562+------------------+--------------------------------------+
2563| ``[41 x i32]``   | Array of 41 32-bit integer values.   |
2564+------------------+--------------------------------------+
2565| ``[4 x i8]``     | Array of 4 8-bit integer values.     |
2566+------------------+--------------------------------------+
2567
2568Here are some examples of multidimensional arrays:
2569
2570+-----------------------------+----------------------------------------------------------+
2571| ``[3 x [4 x i32]]``         | 3x4 array of 32-bit integer values.                      |
2572+-----------------------------+----------------------------------------------------------+
2573| ``[12 x [10 x float]]``     | 12x10 array of single precision floating point values.   |
2574+-----------------------------+----------------------------------------------------------+
2575| ``[2 x [3 x [4 x i16]]]``   | 2x3x4 array of 16-bit integer values.                    |
2576+-----------------------------+----------------------------------------------------------+
2577
2578There is no restriction on indexing beyond the end of the array implied
2579by a static type (though there are restrictions on indexing beyond the
2580bounds of an allocated object in some cases). This means that
2581single-dimension 'variable sized array' addressing can be implemented in
2582LLVM with a zero length array type. An implementation of 'pascal style
2583arrays' in LLVM could use the type "``{ i32, [0 x float]}``", for
2584example.
2585
2586.. _t_struct:
2587
2588Structure Type
2589""""""""""""""
2590
2591:Overview:
2592
2593The structure type is used to represent a collection of data members
2594together in memory. The elements of a structure may be any type that has
2595a size.
2596
2597Structures in memory are accessed using '``load``' and '``store``' by
2598getting a pointer to a field with the '``getelementptr``' instruction.
2599Structures in registers are accessed using the '``extractvalue``' and
2600'``insertvalue``' instructions.
2601
2602Structures may optionally be "packed" structures, which indicate that
2603the alignment of the struct is one byte, and that there is no padding
2604between the elements. In non-packed structs, padding between field types
2605is inserted as defined by the DataLayout string in the module, which is
2606required to match what the underlying code generator expects.
2607
2608Structures can either be "literal" or "identified". A literal structure
2609is defined inline with other types (e.g. ``{i32, i32}*``) whereas
2610identified types are always defined at the top level with a name.
2611Literal types are uniqued by their contents and can never be recursive
2612or opaque since there is no way to write one. Identified types can be
2613recursive, can be opaqued, and are never uniqued.
2614
2615:Syntax:
2616
2617::
2618
2619      %T1 = type { <type list> }     ; Identified normal struct type
2620      %T2 = type <{ <type list> }>   ; Identified packed struct type
2621
2622:Examples:
2623
2624+------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2625| ``{ i32, i32, i32 }``        | A triple of three ``i32`` values                                                                                                                                                      |
2626+------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2627| ``{ 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``.  |
2628+------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2629| ``<{ i8, i32 }>``            | A packed struct known to be 5 bytes in size.                                                                                                                                          |
2630+------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2631
2632.. _t_opaque:
2633
2634Opaque Structure Types
2635""""""""""""""""""""""
2636
2637:Overview:
2638
2639Opaque structure types are used to represent named structure types that
2640do not have a body specified. This corresponds (for example) to the C
2641notion of a forward declared structure.
2642
2643:Syntax:
2644
2645::
2646
2647      %X = type opaque
2648      %52 = type opaque
2649
2650:Examples:
2651
2652+--------------+-------------------+
2653| ``opaque``   | An opaque type.   |
2654+--------------+-------------------+
2655
2656.. _constants:
2657
2658Constants
2659=========
2660
2661LLVM has several different basic types of constants. This section
2662describes them all and their syntax.
2663
2664Simple Constants
2665----------------
2666
2667**Boolean constants**
2668    The two strings '``true``' and '``false``' are both valid constants
2669    of the ``i1`` type.
2670**Integer constants**
2671    Standard integers (such as '4') are constants of the
2672    :ref:`integer <t_integer>` type. Negative numbers may be used with
2673    integer types.
2674**Floating point constants**
2675    Floating point constants use standard decimal notation (e.g.
2676    123.421), exponential notation (e.g. 1.23421e+2), or a more precise
2677    hexadecimal notation (see below). The assembler requires the exact
2678    decimal value of a floating-point constant. For example, the
2679    assembler accepts 1.25 but rejects 1.3 because 1.3 is a repeating
2680    decimal in binary. Floating point constants must have a :ref:`floating
2681    point <t_floating>` type.
2682**Null pointer constants**
2683    The identifier '``null``' is recognized as a null pointer constant
2684    and must be of :ref:`pointer type <t_pointer>`.
2685**Token constants**
2686    The identifier '``none``' is recognized as an empty token constant
2687    and must be of :ref:`token type <t_token>`.
2688
2689The one non-intuitive notation for constants is the hexadecimal form of
2690floating point constants. For example, the form
2691'``double    0x432ff973cafa8000``' is equivalent to (but harder to read
2692than) '``double 4.5e+15``'. The only time hexadecimal floating point
2693constants are required (and the only time that they are generated by the
2694disassembler) is when a floating point constant must be emitted but it
2695cannot be represented as a decimal floating point number in a reasonable
2696number of digits. For example, NaN's, infinities, and other special
2697values are represented in their IEEE hexadecimal format so that assembly
2698and disassembly do not cause any bits to change in the constants.
2699
2700When using the hexadecimal form, constants of types half, float, and
2701double are represented using the 16-digit form shown above (which
2702matches the IEEE754 representation for double); half and float values
2703must, however, be exactly representable as IEEE 754 half and single
2704precision, respectively. Hexadecimal format is always used for long
2705double, and there are three forms of long double. The 80-bit format used
2706by x86 is represented as ``0xK`` followed by 20 hexadecimal digits. The
2707128-bit format used by PowerPC (two adjacent doubles) is represented by
2708``0xM`` followed by 32 hexadecimal digits. The IEEE 128-bit format is
2709represented by ``0xL`` followed by 32 hexadecimal digits. Long doubles
2710will only work if they match the long double format on your target.
2711The IEEE 16-bit format (half precision) is represented by ``0xH``
2712followed by 4 hexadecimal digits. All hexadecimal formats are big-endian
2713(sign bit at the left).
2714
2715There are no constants of type x86_mmx.
2716
2717.. _complexconstants:
2718
2719Complex Constants
2720-----------------
2721
2722Complex constants are a (potentially recursive) combination of simple
2723constants and smaller complex constants.
2724
2725**Structure constants**
2726    Structure constants are represented with notation similar to
2727    structure type definitions (a comma separated list of elements,
2728    surrounded by braces (``{}``)). For example:
2729    "``{ i32 4, float 17.0, i32* @G }``", where "``@G``" is declared as
2730    "``@G = external global i32``". Structure constants must have
2731    :ref:`structure type <t_struct>`, and the number and types of elements
2732    must match those specified by the type.
2733**Array constants**
2734    Array constants are represented with notation similar to array type
2735    definitions (a comma separated list of elements, surrounded by
2736    square brackets (``[]``)). For example:
2737    "``[ i32 42, i32 11, i32 74 ]``". Array constants must have
2738    :ref:`array type <t_array>`, and the number and types of elements must
2739    match those specified by the type. As a special case, character array
2740    constants may also be represented as a double-quoted string using the ``c``
2741    prefix. For example: "``c"Hello World\0A\00"``".
2742**Vector constants**
2743    Vector constants are represented with notation similar to vector
2744    type definitions (a comma separated list of elements, surrounded by
2745    less-than/greater-than's (``<>``)). For example:
2746    "``< i32 42, i32 11, i32 74, i32 100 >``". Vector constants
2747    must have :ref:`vector type <t_vector>`, and the number and types of
2748    elements must match those specified by the type.
2749**Zero initialization**
2750    The string '``zeroinitializer``' can be used to zero initialize a
2751    value to zero of *any* type, including scalar and
2752    :ref:`aggregate <t_aggregate>` types. This is often used to avoid
2753    having to print large zero initializers (e.g. for large arrays) and
2754    is always exactly equivalent to using explicit zero initializers.
2755**Metadata node**
2756    A metadata node is a constant tuple without types. For example:
2757    "``!{!0, !{!2, !0}, !"test"}``". Metadata can reference constant values,
2758    for example: "``!{!0, i32 0, i8* @global, i64 (i64)* @function, !"str"}``".
2759    Unlike other typed constants that are meant to be interpreted as part of
2760    the instruction stream, metadata is a place to attach additional
2761    information such as debug info.
2762
2763Global Variable and Function Addresses
2764--------------------------------------
2765
2766The addresses of :ref:`global variables <globalvars>` and
2767:ref:`functions <functionstructure>` are always implicitly valid
2768(link-time) constants. These constants are explicitly referenced when
2769the :ref:`identifier for the global <identifiers>` is used and always have
2770:ref:`pointer <t_pointer>` type. For example, the following is a legal LLVM
2771file:
2772
2773.. code-block:: llvm
2774
2775    @X = global i32 17
2776    @Y = global i32 42
2777    @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2778
2779.. _undefvalues:
2780
2781Undefined Values
2782----------------
2783
2784The string '``undef``' can be used anywhere a constant is expected, and
2785indicates that the user of the value may receive an unspecified
2786bit-pattern. Undefined values may be of any type (other than '``label``'
2787or '``void``') and be used anywhere a constant is permitted.
2788
2789Undefined values are useful because they indicate to the compiler that
2790the program is well defined no matter what value is used. This gives the
2791compiler more freedom to optimize. Here are some examples of
2792(potentially surprising) transformations that are valid (in pseudo IR):
2793
2794.. code-block:: llvm
2795
2796      %A = add %X, undef
2797      %B = sub %X, undef
2798      %C = xor %X, undef
2799    Safe:
2800      %A = undef
2801      %B = undef
2802      %C = undef
2803
2804This is safe because all of the output bits are affected by the undef
2805bits. Any output bit can have a zero or one depending on the input bits.
2806
2807.. code-block:: llvm
2808
2809      %A = or %X, undef
2810      %B = and %X, undef
2811    Safe:
2812      %A = -1
2813      %B = 0
2814    Unsafe:
2815      %A = undef
2816      %B = undef
2817
2818These logical operations have bits that are not always affected by the
2819input. For example, if ``%X`` has a zero bit, then the output of the
2820'``and``' operation will always be a zero for that bit, no matter what
2821the corresponding bit from the '``undef``' is. As such, it is unsafe to
2822optimize or assume that the result of the '``and``' is '``undef``'.
2823However, it is safe to assume that all bits of the '``undef``' could be
28240, and optimize the '``and``' to 0. Likewise, it is safe to assume that
2825all the bits of the '``undef``' operand to the '``or``' could be set,
2826allowing the '``or``' to be folded to -1.
2827
2828.. code-block:: llvm
2829
2830      %A = select undef, %X, %Y
2831      %B = select undef, 42, %Y
2832      %C = select %X, %Y, undef
2833    Safe:
2834      %A = %X     (or %Y)
2835      %B = 42     (or %Y)
2836      %C = %Y
2837    Unsafe:
2838      %A = undef
2839      %B = undef
2840      %C = undef
2841
2842This set of examples shows that undefined '``select``' (and conditional
2843branch) conditions can go *either way*, but they have to come from one
2844of the two operands. In the ``%A`` example, if ``%X`` and ``%Y`` were
2845both known to have a clear low bit, then ``%A`` would have to have a
2846cleared low bit. However, in the ``%C`` example, the optimizer is
2847allowed to assume that the '``undef``' operand could be the same as
2848``%Y``, allowing the whole '``select``' to be eliminated.
2849
2850.. code-block:: llvm
2851
2852      %A = xor undef, undef
2853
2854      %B = undef
2855      %C = xor %B, %B
2856
2857      %D = undef
2858      %E = icmp slt %D, 4
2859      %F = icmp gte %D, 4
2860
2861    Safe:
2862      %A = undef
2863      %B = undef
2864      %C = undef
2865      %D = undef
2866      %E = undef
2867      %F = undef
2868
2869This example points out that two '``undef``' operands are not
2870necessarily the same. This can be surprising to people (and also matches
2871C semantics) where they assume that "``X^X``" is always zero, even if
2872``X`` is undefined. This isn't true for a number of reasons, but the
2873short answer is that an '``undef``' "variable" can arbitrarily change
2874its value over its "live range". This is true because the variable
2875doesn't actually *have a live range*. Instead, the value is logically
2876read from arbitrary registers that happen to be around when needed, so
2877the value is not necessarily consistent over time. In fact, ``%A`` and
2878``%C`` need to have the same semantics or the core LLVM "replace all
2879uses with" concept would not hold.
2880
2881.. code-block:: llvm
2882
2883      %A = fdiv undef, %X
2884      %B = fdiv %X, undef
2885    Safe:
2886      %A = undef
2887    b: unreachable
2888
2889These examples show the crucial difference between an *undefined value*
2890and *undefined behavior*. An undefined value (like '``undef``') is
2891allowed to have an arbitrary bit-pattern. This means that the ``%A``
2892operation can be constant folded to '``undef``', because the '``undef``'
2893could be an SNaN, and ``fdiv`` is not (currently) defined on SNaN's.
2894However, in the second example, we can make a more aggressive
2895assumption: because the ``undef`` is allowed to be an arbitrary value,
2896we are allowed to assume that it could be zero. Since a divide by zero
2897has *undefined behavior*, we are allowed to assume that the operation
2898does not execute at all. This allows us to delete the divide and all
2899code after it. Because the undefined operation "can't happen", the
2900optimizer can assume that it occurs in dead code.
2901
2902.. code-block:: llvm
2903
2904    a:  store undef -> %X
2905    b:  store %X -> undef
2906    Safe:
2907    a: <deleted>
2908    b: unreachable
2909
2910These examples reiterate the ``fdiv`` example: a store *of* an undefined
2911value can be assumed to not have any effect; we can assume that the
2912value is overwritten with bits that happen to match what was already
2913there. However, a store *to* an undefined location could clobber
2914arbitrary memory, therefore, it has undefined behavior.
2915
2916.. _poisonvalues:
2917
2918Poison Values
2919-------------
2920
2921Poison values are similar to :ref:`undef values <undefvalues>`, however
2922they also represent the fact that an instruction or constant expression
2923that cannot evoke side effects has nevertheless detected a condition
2924that results in undefined behavior.
2925
2926There is currently no way of representing a poison value in the IR; they
2927only exist when produced by operations such as :ref:`add <i_add>` with
2928the ``nsw`` flag.
2929
2930Poison value behavior is defined in terms of value *dependence*:
2931
2932-  Values other than :ref:`phi <i_phi>` nodes depend on their operands.
2933-  :ref:`Phi <i_phi>` nodes depend on the operand corresponding to
2934   their dynamic predecessor basic block.
2935-  Function arguments depend on the corresponding actual argument values
2936   in the dynamic callers of their functions.
2937-  :ref:`Call <i_call>` instructions depend on the :ref:`ret <i_ret>`
2938   instructions that dynamically transfer control back to them.
2939-  :ref:`Invoke <i_invoke>` instructions depend on the
2940   :ref:`ret <i_ret>`, :ref:`resume <i_resume>`, or exception-throwing
2941   call instructions that dynamically transfer control back to them.
2942-  Non-volatile loads and stores depend on the most recent stores to all
2943   of the referenced memory addresses, following the order in the IR
2944   (including loads and stores implied by intrinsics such as
2945   :ref:`@llvm.memcpy <int_memcpy>`.)
2946-  An instruction with externally visible side effects depends on the
2947   most recent preceding instruction with externally visible side
2948   effects, following the order in the IR. (This includes :ref:`volatile
2949   operations <volatile>`.)
2950-  An instruction *control-depends* on a :ref:`terminator
2951   instruction <terminators>` if the terminator instruction has
2952   multiple successors and the instruction is always executed when
2953   control transfers to one of the successors, and may not be executed
2954   when control is transferred to another.
2955-  Additionally, an instruction also *control-depends* on a terminator
2956   instruction if the set of instructions it otherwise depends on would
2957   be different if the terminator had transferred control to a different
2958   successor.
2959-  Dependence is transitive.
2960
2961Poison values have the same behavior as :ref:`undef values <undefvalues>`,
2962with the additional effect that any instruction that has a *dependence*
2963on a poison value has undefined behavior.
2964
2965Here are some examples:
2966
2967.. code-block:: llvm
2968
2969    entry:
2970      %poison = sub nuw i32 0, 1           ; Results in a poison value.
2971      %still_poison = and i32 %poison, 0   ; 0, but also poison.
2972      %poison_yet_again = getelementptr i32, i32* @h, i32 %still_poison
2973      store i32 0, i32* %poison_yet_again  ; memory at @h[0] is poisoned
2974
2975      store i32 %poison, i32* @g           ; Poison value stored to memory.
2976      %poison2 = load i32, i32* @g         ; Poison value loaded back from memory.
2977
2978      store volatile i32 %poison, i32* @g  ; External observation; undefined behavior.
2979
2980      %narrowaddr = bitcast i32* @g to i16*
2981      %wideaddr = bitcast i32* @g to i64*
2982      %poison3 = load i16, i16* %narrowaddr ; Returns a poison value.
2983      %poison4 = load i64, i64* %wideaddr  ; Returns a poison value.
2984
2985      %cmp = icmp slt i32 %poison, 0       ; Returns a poison value.
2986      br i1 %cmp, label %true, label %end  ; Branch to either destination.
2987
2988    true:
2989      store volatile i32 0, i32* @g        ; This is control-dependent on %cmp, so
2990                                           ; it has undefined behavior.
2991      br label %end
2992
2993    end:
2994      %p = phi i32 [ 0, %entry ], [ 1, %true ]
2995                                           ; Both edges into this PHI are
2996                                           ; control-dependent on %cmp, so this
2997                                           ; always results in a poison value.
2998
2999      store volatile i32 0, i32* @g        ; This would depend on the store in %true
3000                                           ; if %cmp is true, or the store in %entry
3001                                           ; otherwise, so this is undefined behavior.
3002
3003      br i1 %cmp, label %second_true, label %second_end
3004                                           ; The same branch again, but this time the
3005                                           ; true block doesn't have side effects.
3006
3007    second_true:
3008      ; No side effects!
3009      ret void
3010
3011    second_end:
3012      store volatile i32 0, i32* @g        ; This time, the instruction always depends
3013                                           ; on the store in %end. Also, it is
3014                                           ; control-equivalent to %end, so this is
3015                                           ; well-defined (ignoring earlier undefined
3016                                           ; behavior in this example).
3017
3018.. _blockaddress:
3019
3020Addresses of Basic Blocks
3021-------------------------
3022
3023``blockaddress(@function, %block)``
3024
3025The '``blockaddress``' constant computes the address of the specified
3026basic block in the specified function, and always has an ``i8*`` type.
3027Taking the address of the entry block is illegal.
3028
3029This value only has defined behavior when used as an operand to the
3030':ref:`indirectbr <i_indirectbr>`' instruction, or for comparisons
3031against null. Pointer equality tests between labels addresses results in
3032undefined behavior --- though, again, comparison against null is ok, and
3033no label is equal to the null pointer. This may be passed around as an
3034opaque pointer sized value as long as the bits are not inspected. This
3035allows ``ptrtoint`` and arithmetic to be performed on these values so
3036long as the original value is reconstituted before the ``indirectbr``
3037instruction.
3038
3039Finally, some targets may provide defined semantics when using the value
3040as the operand to an inline assembly, but that is target specific.
3041
3042.. _constantexprs:
3043
3044Constant Expressions
3045--------------------
3046
3047Constant expressions are used to allow expressions involving other
3048constants to be used as constants. Constant expressions may be of any
3049:ref:`first class <t_firstclass>` type and may involve any LLVM operation
3050that does not have side effects (e.g. load and call are not supported).
3051The following is the syntax for constant expressions:
3052
3053``trunc (CST to TYPE)``
3054    Truncate a constant to another type. The bit size of CST must be
3055    larger than the bit size of TYPE. Both types must be integers.
3056``zext (CST to TYPE)``
3057    Zero extend a constant to another type. The bit size of CST must be
3058    smaller than the bit size of TYPE. Both types must be integers.
3059``sext (CST to TYPE)``
3060    Sign extend a constant to another type. The bit size of CST must be
3061    smaller than the bit size of TYPE. Both types must be integers.
3062``fptrunc (CST to TYPE)``
3063    Truncate a floating point constant to another floating point type.
3064    The size of CST must be larger than the size of TYPE. Both types
3065    must be floating point.
3066``fpext (CST to TYPE)``
3067    Floating point extend a constant to another type. The size of CST
3068    must be smaller or equal to the size of TYPE. Both types must be
3069    floating point.
3070``fptoui (CST to TYPE)``
3071    Convert a floating point constant to the corresponding unsigned
3072    integer constant. TYPE must be a scalar or vector integer type. CST
3073    must be of scalar or vector floating point type. Both CST and TYPE
3074    must be scalars, or vectors of the same number of elements. If the
3075    value won't fit in the integer type, the results are undefined.
3076``fptosi (CST to TYPE)``
3077    Convert a floating point constant to the corresponding signed
3078    integer constant. TYPE must be a scalar or vector integer type. CST
3079    must be of scalar or vector floating point type. Both CST and TYPE
3080    must be scalars, or vectors of the same number of elements. If the
3081    value won't fit in the integer type, the results are undefined.
3082``uitofp (CST to TYPE)``
3083    Convert an unsigned integer constant to the corresponding floating
3084    point constant. TYPE must be a scalar or vector floating point type.
3085    CST must be of scalar or vector integer type. Both CST and TYPE must
3086    be scalars, or vectors of the same number of elements. If the value
3087    won't fit in the floating point type, the results are undefined.
3088``sitofp (CST to TYPE)``
3089    Convert a signed integer constant to the corresponding floating
3090    point constant. TYPE must be a scalar or vector floating point type.
3091    CST must be of scalar or vector integer type. Both CST and TYPE must
3092    be scalars, or vectors of the same number of elements. If the value
3093    won't fit in the floating point type, the results are undefined.
3094``ptrtoint (CST to TYPE)``
3095    Convert a pointer typed constant to the corresponding integer
3096    constant. ``TYPE`` must be an integer type. ``CST`` must be of
3097    pointer type. The ``CST`` value is zero extended, truncated, or
3098    unchanged to make it fit in ``TYPE``.
3099``inttoptr (CST to TYPE)``
3100    Convert an integer constant to a pointer constant. TYPE must be a
3101    pointer type. CST must be of integer type. The CST value is zero
3102    extended, truncated, or unchanged to make it fit in a pointer size.
3103    This one is *really* dangerous!
3104``bitcast (CST to TYPE)``
3105    Convert a constant, CST, to another TYPE. The constraints of the
3106    operands are the same as those for the :ref:`bitcast
3107    instruction <i_bitcast>`.
3108``addrspacecast (CST to TYPE)``
3109    Convert a constant pointer or constant vector of pointer, CST, to another
3110    TYPE in a different address space. The constraints of the operands are the
3111    same as those for the :ref:`addrspacecast instruction <i_addrspacecast>`.
3112``getelementptr (TY, CSTPTR, IDX0, IDX1, ...)``, ``getelementptr inbounds (TY, CSTPTR, IDX0, IDX1, ...)``
3113    Perform the :ref:`getelementptr operation <i_getelementptr>` on
3114    constants. As with the :ref:`getelementptr <i_getelementptr>`
3115    instruction, the index list may have zero or more indexes, which are
3116    required to make sense for the type of "pointer to TY".
3117``select (COND, VAL1, VAL2)``
3118    Perform the :ref:`select operation <i_select>` on constants.
3119``icmp COND (VAL1, VAL2)``
3120    Performs the :ref:`icmp operation <i_icmp>` on constants.
3121``fcmp COND (VAL1, VAL2)``
3122    Performs the :ref:`fcmp operation <i_fcmp>` on constants.
3123``extractelement (VAL, IDX)``
3124    Perform the :ref:`extractelement operation <i_extractelement>` on
3125    constants.
3126``insertelement (VAL, ELT, IDX)``
3127    Perform the :ref:`insertelement operation <i_insertelement>` on
3128    constants.
3129``shufflevector (VEC1, VEC2, IDXMASK)``
3130    Perform the :ref:`shufflevector operation <i_shufflevector>` on
3131    constants.
3132``extractvalue (VAL, IDX0, IDX1, ...)``
3133    Perform the :ref:`extractvalue operation <i_extractvalue>` on
3134    constants. The index list is interpreted in a similar manner as
3135    indices in a ':ref:`getelementptr <i_getelementptr>`' operation. At
3136    least one index value must be specified.
3137``insertvalue (VAL, ELT, IDX0, IDX1, ...)``
3138    Perform the :ref:`insertvalue operation <i_insertvalue>` on constants.
3139    The index list is interpreted in a similar manner as indices in a
3140    ':ref:`getelementptr <i_getelementptr>`' operation. At least one index
3141    value must be specified.
3142``OPCODE (LHS, RHS)``
3143    Perform the specified operation of the LHS and RHS constants. OPCODE
3144    may be any of the :ref:`binary <binaryops>` or :ref:`bitwise
3145    binary <bitwiseops>` operations. The constraints on operands are
3146    the same as those for the corresponding instruction (e.g. no bitwise
3147    operations on floating point values are allowed).
3148
3149Other Values
3150============
3151
3152.. _inlineasmexprs:
3153
3154Inline Assembler Expressions
3155----------------------------
3156
3157LLVM supports inline assembler expressions (as opposed to :ref:`Module-Level
3158Inline Assembly <moduleasm>`) through the use of a special value. This value
3159represents the inline assembler as a template string (containing the
3160instructions to emit), a list of operand constraints (stored as a string), a
3161flag that indicates whether or not the inline asm expression has side effects,
3162and a flag indicating whether the function containing the asm needs to align its
3163stack conservatively.
3164
3165The template string supports argument substitution of the operands using "``$``"
3166followed by a number, to indicate substitution of the given register/memory
3167location, as specified by the constraint string. "``${NUM:MODIFIER}``" may also
3168be used, where ``MODIFIER`` is a target-specific annotation for how to print the
3169operand (See :ref:`inline-asm-modifiers`).
3170
3171A literal "``$``" may be included by using "``$$``" in the template. To include
3172other special characters into the output, the usual "``\XX``" escapes may be
3173used, just as in other strings. Note that after template substitution, the
3174resulting assembly string is parsed by LLVM's integrated assembler unless it is
3175disabled -- even when emitting a ``.s`` file -- and thus must contain assembly
3176syntax known to LLVM.
3177
3178LLVM's support for inline asm is modeled closely on the requirements of Clang's
3179GCC-compatible inline-asm support. Thus, the feature-set and the constraint and
3180modifier codes listed here are similar or identical to those in GCC's inline asm
3181support. However, to be clear, the syntax of the template and constraint strings
3182described here is *not* the same as the syntax accepted by GCC and Clang, and,
3183while most constraint letters are passed through as-is by Clang, some get
3184translated to other codes when converting from the C source to the LLVM
3185assembly.
3186
3187An example inline assembler expression is:
3188
3189.. code-block:: llvm
3190
3191    i32 (i32) asm "bswap $0", "=r,r"
3192
3193Inline assembler expressions may **only** be used as the callee operand
3194of a :ref:`call <i_call>` or an :ref:`invoke <i_invoke>` instruction.
3195Thus, typically we have:
3196
3197.. code-block:: llvm
3198
3199    %X = call i32 asm "bswap $0", "=r,r"(i32 %Y)
3200
3201Inline asms with side effects not visible in the constraint list must be
3202marked as having side effects. This is done through the use of the
3203'``sideeffect``' keyword, like so:
3204
3205.. code-block:: llvm
3206
3207    call void asm sideeffect "eieio", ""()
3208
3209In some cases inline asms will contain code that will not work unless
3210the stack is aligned in some way, such as calls or SSE instructions on
3211x86, yet will not contain code that does that alignment within the asm.
3212The compiler should make conservative assumptions about what the asm
3213might contain and should generate its usual stack alignment code in the
3214prologue if the '``alignstack``' keyword is present:
3215
3216.. code-block:: llvm
3217
3218    call void asm alignstack "eieio", ""()
3219
3220Inline asms also support using non-standard assembly dialects. The
3221assumed dialect is ATT. When the '``inteldialect``' keyword is present,
3222the inline asm is using the Intel dialect. Currently, ATT and Intel are
3223the only supported dialects. An example is:
3224
3225.. code-block:: llvm
3226
3227    call void asm inteldialect "eieio", ""()
3228
3229If multiple keywords appear the '``sideeffect``' keyword must come
3230first, the '``alignstack``' keyword second and the '``inteldialect``'
3231keyword last.
3232
3233Inline Asm Constraint String
3234^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3235
3236The constraint list is a comma-separated string, each element containing one or
3237more constraint codes.
3238
3239For each element in the constraint list an appropriate register or memory
3240operand will be chosen, and it will be made available to assembly template
3241string expansion as ``$0`` for the first constraint in the list, ``$1`` for the
3242second, etc.
3243
3244There are three different types of constraints, which are distinguished by a
3245prefix symbol in front of the constraint code: Output, Input, and Clobber. The
3246constraints must always be given in that order: outputs first, then inputs, then
3247clobbers. They cannot be intermingled.
3248
3249There are also three different categories of constraint codes:
3250
3251- Register constraint. This is either a register class, or a fixed physical
3252  register. This kind of constraint will allocate a register, and if necessary,
3253  bitcast the argument or result to the appropriate type.
3254- Memory constraint. This kind of constraint is for use with an instruction
3255  taking a memory operand. Different constraints allow for different addressing
3256  modes used by the target.
3257- Immediate value constraint. This kind of constraint is for an integer or other
3258  immediate value which can be rendered directly into an instruction. The
3259  various target-specific constraints allow the selection of a value in the
3260  proper range for the instruction you wish to use it with.
3261
3262Output constraints
3263""""""""""""""""""
3264
3265Output constraints are specified by an "``=``" prefix (e.g. "``=r``"). This
3266indicates that the assembly will write to this operand, and the operand will
3267then be made available as a return value of the ``asm`` expression. Output
3268constraints do not consume an argument from the call instruction. (Except, see
3269below about indirect outputs).
3270
3271Normally, it is expected that no output locations are written to by the assembly
3272expression until *all* of the inputs have been read. As such, LLVM may assign
3273the same register to an output and an input. If this is not safe (e.g. if the
3274assembly contains two instructions, where the first writes to one output, and
3275the second reads an input and writes to a second output), then the "``&``"
3276modifier must be used (e.g. "``=&r``") to specify that the output is an
3277"early-clobber" output. Marking an output as "early-clobber" ensures that LLVM
3278will not use the same register for any inputs (other than an input tied to this
3279output).
3280
3281Input constraints
3282"""""""""""""""""
3283
3284Input constraints do not have a prefix -- just the constraint codes. Each input
3285constraint will consume one argument from the call instruction. It is not
3286permitted for the asm to write to any input register or memory location (unless
3287that input is tied to an output). Note also that multiple inputs may all be
3288assigned to the same register, if LLVM can determine that they necessarily all
3289contain the same value.
3290
3291Instead of providing a Constraint Code, input constraints may also "tie"
3292themselves to an output constraint, by providing an integer as the constraint
3293string. Tied inputs still consume an argument from the call instruction, and
3294take up a position in the asm template numbering as is usual -- they will simply
3295be constrained to always use the same register as the output they've been tied
3296to. For example, a constraint string of "``=r,0``" says to assign a register for
3297output, and use that register as an input as well (it being the 0'th
3298constraint).
3299
3300It is permitted to tie an input to an "early-clobber" output. In that case, no
3301*other* input may share the same register as the input tied to the early-clobber
3302(even when the other input has the same value).
3303
3304You may only tie an input to an output which has a register constraint, not a
3305memory constraint. Only a single input may be tied to an output.
3306
3307There is also an "interesting" feature which deserves a bit of explanation: if a
3308register class constraint allocates a register which is too small for the value
3309type operand provided as input, the input value will be split into multiple
3310registers, and all of them passed to the inline asm.
3311
3312However, this feature is often not as useful as you might think.
3313
3314Firstly, the registers are *not* guaranteed to be consecutive. So, on those
3315architectures that have instructions which operate on multiple consecutive
3316instructions, this is not an appropriate way to support them. (e.g. the 32-bit
3317SparcV8 has a 64-bit load, which instruction takes a single 32-bit register. The
3318hardware then loads into both the named register, and the next register. This
3319feature of inline asm would not be useful to support that.)
3320
3321A few of the targets provide a template string modifier allowing explicit access
3322to the second register of a two-register operand (e.g. MIPS ``L``, ``M``, and
3323``D``). On such an architecture, you can actually access the second allocated
3324register (yet, still, not any subsequent ones). But, in that case, you're still
3325probably better off simply splitting the value into two separate operands, for
3326clarity. (e.g. see the description of the ``A`` constraint on X86, which,
3327despite existing only for use with this feature, is not really a good idea to
3328use)
3329
3330Indirect inputs and outputs
3331"""""""""""""""""""""""""""
3332
3333Indirect output or input constraints can be specified by the "``*``" modifier
3334(which goes after the "``=``" in case of an output). This indicates that the asm
3335will write to or read from the contents of an *address* provided as an input
3336argument. (Note that in this way, indirect outputs act more like an *input* than
3337an output: just like an input, they consume an argument of the call expression,
3338rather than producing a return value. An indirect output constraint is an
3339"output" only in that the asm is expected to write to the contents of the input
3340memory location, instead of just read from it).
3341
3342This is most typically used for memory constraint, e.g. "``=*m``", to pass the
3343address of a variable as a value.
3344
3345It is also possible to use an indirect *register* constraint, but only on output
3346(e.g. "``=*r``"). This will cause LLVM to allocate a register for an output
3347value normally, and then, separately emit a store to the address provided as
3348input, after the provided inline asm. (It's not clear what value this
3349functionality provides, compared to writing the store explicitly after the asm
3350statement, and it can only produce worse code, since it bypasses many
3351optimization passes. I would recommend not using it.)
3352
3353
3354Clobber constraints
3355"""""""""""""""""""
3356
3357A clobber constraint is indicated by a "``~``" prefix. A clobber does not
3358consume an input operand, nor generate an output. Clobbers cannot use any of the
3359general constraint code letters -- they may use only explicit register
3360constraints, e.g. "``~{eax}``". The one exception is that a clobber string of
3361"``~{memory}``" indicates that the assembly writes to arbitrary undeclared
3362memory locations -- not only the memory pointed to by a declared indirect
3363output.
3364
3365
3366Constraint Codes
3367""""""""""""""""
3368After a potential prefix comes constraint code, or codes.
3369
3370A Constraint Code is either a single letter (e.g. "``r``"), a "``^``" character
3371followed by two letters (e.g. "``^wc``"), or "``{``" register-name "``}``"
3372(e.g. "``{eax}``").
3373
3374The one and two letter constraint codes are typically chosen to be the same as
3375GCC's constraint codes.
3376
3377A single constraint may include one or more than constraint code in it, leaving
3378it up to LLVM to choose which one to use. This is included mainly for
3379compatibility with the translation of GCC inline asm coming from clang.
3380
3381There are two ways to specify alternatives, and either or both may be used in an
3382inline asm constraint list:
3383
33841) Append the codes to each other, making a constraint code set. E.g. "``im``"
3385   or "``{eax}m``". This means "choose any of the options in the set". The
3386   choice of constraint is made independently for each constraint in the
3387   constraint list.
3388
33892) Use "``|``" between constraint code sets, creating alternatives. Every
3390   constraint in the constraint list must have the same number of alternative
3391   sets. With this syntax, the same alternative in *all* of the items in the
3392   constraint list will be chosen together.
3393
3394Putting those together, you might have a two operand constraint string like
3395``"rm|r,ri|rm"``. This indicates that if operand 0 is ``r`` or ``m``, then
3396operand 1 may be one of ``r`` or ``i``. If operand 0 is ``r``, then operand 1
3397may be one of ``r`` or ``m``. But, operand 0 and 1 cannot both be of type m.
3398
3399However, the use of either of the alternatives features is *NOT* recommended, as
3400LLVM is not able to make an intelligent choice about which one to use. (At the
3401point it currently needs to choose, not enough information is available to do so
3402in a smart way.) Thus, it simply tries to make a choice that's most likely to
3403compile, not one that will be optimal performance. (e.g., given "``rm``", it'll
3404always choose to use memory, not registers). And, if given multiple registers,
3405or multiple register classes, it will simply choose the first one. (In fact, it
3406doesn't currently even ensure explicitly specified physical registers are
3407unique, so specifying multiple physical registers as alternatives, like
3408``{r11}{r12},{r11}{r12}``, will assign r11 to both operands, not at all what was
3409intended.)
3410
3411Supported Constraint Code List
3412""""""""""""""""""""""""""""""
3413
3414The constraint codes are, in general, expected to behave the same way they do in
3415GCC. LLVM's support is often implemented on an 'as-needed' basis, to support C
3416inline asm code which was supported by GCC. A mismatch in behavior between LLVM
3417and GCC likely indicates a bug in LLVM.
3418
3419Some constraint codes are typically supported by all targets:
3420
3421- ``r``: A register in the target's general purpose register class.
3422- ``m``: A memory address operand. It is target-specific what addressing modes
3423  are supported, typical examples are register, or register + register offset,
3424  or register + immediate offset (of some target-specific size).
3425- ``i``: An integer constant (of target-specific width). Allows either a simple
3426  immediate, or a relocatable value.
3427- ``n``: An integer constant -- *not* including relocatable values.
3428- ``s``: An integer constant, but allowing *only* relocatable values.
3429- ``X``: Allows an operand of any kind, no constraint whatsoever. Typically
3430  useful to pass a label for an asm branch or call.
3431
3432  .. FIXME: but that surely isn't actually okay to jump out of an asm
3433     block without telling llvm about the control transfer???)
3434
3435- ``{register-name}``: Requires exactly the named physical register.
3436
3437Other constraints are target-specific:
3438
3439AArch64:
3440
3441- ``z``: An immediate integer 0. Outputs ``WZR`` or ``XZR``, as appropriate.
3442- ``I``: An immediate integer valid for an ``ADD`` or ``SUB`` instruction,
3443  i.e. 0 to 4095 with optional shift by 12.
3444- ``J``: An immediate integer that, when negated, is valid for an ``ADD`` or
3445  ``SUB`` instruction, i.e. -1 to -4095 with optional left shift by 12.
3446- ``K``: An immediate integer that is valid for the 'bitmask immediate 32' of a
3447  logical instruction like ``AND``, ``EOR``, or ``ORR`` with a 32-bit register.
3448- ``L``: An immediate integer that is valid for the 'bitmask immediate 64' of a
3449  logical instruction like ``AND``, ``EOR``, or ``ORR`` with a 64-bit register.
3450- ``M``: An immediate integer for use with the ``MOV`` assembly alias on a
3451  32-bit register. This is a superset of ``K``: in addition to the bitmask
3452  immediate, also allows immediate integers which can be loaded with a single
3453  ``MOVZ`` or ``MOVL`` instruction.
3454- ``N``: An immediate integer for use with the ``MOV`` assembly alias on a
3455  64-bit register. This is a superset of ``L``.
3456- ``Q``: Memory address operand must be in a single register (no
3457  offsets). (However, LLVM currently does this for the ``m`` constraint as
3458  well.)
3459- ``r``: A 32 or 64-bit integer register (W* or X*).
3460- ``w``: A 32, 64, or 128-bit floating-point/SIMD register.
3461- ``x``: A lower 128-bit floating-point/SIMD register (``V0`` to ``V15``).
3462
3463AMDGPU:
3464
3465- ``r``: A 32 or 64-bit integer register.
3466- ``[0-9]v``: The 32-bit VGPR register, number 0-9.
3467- ``[0-9]s``: The 32-bit SGPR register, number 0-9.
3468
3469
3470All ARM modes:
3471
3472- ``Q``, ``Um``, ``Un``, ``Uq``, ``Us``, ``Ut``, ``Uv``, ``Uy``: Memory address
3473  operand. Treated the same as operand ``m``, at the moment.
3474
3475ARM and ARM's Thumb2 mode:
3476
3477- ``j``: An immediate integer between 0 and 65535 (valid for ``MOVW``)
3478- ``I``: An immediate integer valid for a data-processing instruction.
3479- ``J``: An immediate integer between -4095 and 4095.
3480- ``K``: An immediate integer whose bitwise inverse is valid for a
3481  data-processing instruction. (Can be used with template modifier "``B``" to
3482  print the inverted value).
3483- ``L``: An immediate integer whose negation is valid for a data-processing
3484  instruction. (Can be used with template modifier "``n``" to print the negated
3485  value).
3486- ``M``: A power of two or a integer between 0 and 32.
3487- ``N``: Invalid immediate constraint.
3488- ``O``: Invalid immediate constraint.
3489- ``r``: A general-purpose 32-bit integer register (``r0-r15``).
3490- ``l``: In Thumb2 mode, low 32-bit GPR registers (``r0-r7``). In ARM mode, same
3491  as ``r``.
3492- ``h``: In Thumb2 mode, a high 32-bit GPR register (``r8-r15``). In ARM mode,
3493  invalid.
3494- ``w``: A 32, 64, or 128-bit floating-point/SIMD register: ``s0-s31``,
3495  ``d0-d31``, or ``q0-q15``.
3496- ``x``: A 32, 64, or 128-bit floating-point/SIMD register: ``s0-s15``,
3497  ``d0-d7``, or ``q0-q3``.
3498- ``t``: A floating-point/SIMD register, only supports 32-bit values:
3499  ``s0-s31``.
3500
3501ARM's Thumb1 mode:
3502
3503- ``I``: An immediate integer between 0 and 255.
3504- ``J``: An immediate integer between -255 and -1.
3505- ``K``: An immediate integer between 0 and 255, with optional left-shift by
3506  some amount.
3507- ``L``: An immediate integer between -7 and 7.
3508- ``M``: An immediate integer which is a multiple of 4 between 0 and 1020.
3509- ``N``: An immediate integer between 0 and 31.
3510- ``O``: An immediate integer which is a multiple of 4 between -508 and 508.
3511- ``r``: A low 32-bit GPR register (``r0-r7``).
3512- ``l``: A low 32-bit GPR register (``r0-r7``).
3513- ``h``: A high GPR register (``r0-r7``).
3514- ``w``: A 32, 64, or 128-bit floating-point/SIMD register: ``s0-s31``,
3515  ``d0-d31``, or ``q0-q15``.
3516- ``x``: A 32, 64, or 128-bit floating-point/SIMD register: ``s0-s15``,
3517  ``d0-d7``, or ``q0-q3``.
3518- ``t``: A floating-point/SIMD register, only supports 32-bit values:
3519  ``s0-s31``.
3520
3521
3522Hexagon:
3523
3524- ``o``, ``v``: A memory address operand, treated the same as constraint ``m``,
3525  at the moment.
3526- ``r``: A 32 or 64-bit register.
3527
3528MSP430:
3529
3530- ``r``: An 8 or 16-bit register.
3531
3532MIPS:
3533
3534- ``I``: An immediate signed 16-bit integer.
3535- ``J``: An immediate integer zero.
3536- ``K``: An immediate unsigned 16-bit integer.
3537- ``L``: An immediate 32-bit integer, where the lower 16 bits are 0.
3538- ``N``: An immediate integer between -65535 and -1.
3539- ``O``: An immediate signed 15-bit integer.
3540- ``P``: An immediate integer between 1 and 65535.
3541- ``m``: A memory address operand. In MIPS-SE mode, allows a base address
3542  register plus 16-bit immediate offset. In MIPS mode, just a base register.
3543- ``R``: A memory address operand. In MIPS-SE mode, allows a base address
3544  register plus a 9-bit signed offset. In MIPS mode, the same as constraint
3545  ``m``.
3546- ``ZC``: A memory address operand, suitable for use in a ``pref``, ``ll``, or
3547  ``sc`` instruction on the given subtarget (details vary).
3548- ``r``, ``d``,  ``y``: A 32 or 64-bit GPR register.
3549- ``f``: A 32 or 64-bit FPU register (``F0-F31``), or a 128-bit MSA register
3550  (``W0-W31``). In the case of MSA registers, it is recommended to use the ``w``
3551  argument modifier for compatibility with GCC.
3552- ``c``: A 32-bit or 64-bit GPR register suitable for indirect jump (always
3553  ``25``).
3554- ``l``: The ``lo`` register, 32 or 64-bit.
3555- ``x``: Invalid.
3556
3557NVPTX:
3558
3559- ``b``: A 1-bit integer register.
3560- ``c`` or ``h``: A 16-bit integer register.
3561- ``r``: A 32-bit integer register.
3562- ``l`` or ``N``: A 64-bit integer register.
3563- ``f``: A 32-bit float register.
3564- ``d``: A 64-bit float register.
3565
3566
3567PowerPC:
3568
3569- ``I``: An immediate signed 16-bit integer.
3570- ``J``: An immediate unsigned 16-bit integer, shifted left 16 bits.
3571- ``K``: An immediate unsigned 16-bit integer.
3572- ``L``: An immediate signed 16-bit integer, shifted left 16 bits.
3573- ``M``: An immediate integer greater than 31.
3574- ``N``: An immediate integer that is an exact power of 2.
3575- ``O``: The immediate integer constant 0.
3576- ``P``: An immediate integer constant whose negation is a signed 16-bit
3577  constant.
3578- ``es``, ``o``, ``Q``, ``Z``, ``Zy``: A memory address operand, currently
3579  treated the same as ``m``.
3580- ``r``: A 32 or 64-bit integer register.
3581- ``b``: A 32 or 64-bit integer register, excluding ``R0`` (that is:
3582  ``R1-R31``).
3583- ``f``: A 32 or 64-bit float register (``F0-F31``), or when QPX is enabled, a
3584  128 or 256-bit QPX register (``Q0-Q31``; aliases the ``F`` registers).
3585- ``v``: For ``4 x f32`` or ``4 x f64`` types, when QPX is enabled, a
3586  128 or 256-bit QPX register (``Q0-Q31``), otherwise a 128-bit
3587  altivec vector register (``V0-V31``).
3588
3589  .. FIXME: is this a bug that v accepts QPX registers? I think this
3590     is supposed to only use the altivec vector registers?
3591
3592- ``y``: Condition register (``CR0-CR7``).
3593- ``wc``: An individual CR bit in a CR register.
3594- ``wa``, ``wd``, ``wf``: Any 128-bit VSX vector register, from the full VSX
3595  register set (overlapping both the floating-point and vector register files).
3596- ``ws``: A 32 or 64-bit floating point register, from the full VSX register
3597  set.
3598
3599Sparc:
3600
3601- ``I``: An immediate 13-bit signed integer.
3602- ``r``: A 32-bit integer register.
3603
3604SystemZ:
3605
3606- ``I``: An immediate unsigned 8-bit integer.
3607- ``J``: An immediate unsigned 12-bit integer.
3608- ``K``: An immediate signed 16-bit integer.
3609- ``L``: An immediate signed 20-bit integer.
3610- ``M``: An immediate integer 0x7fffffff.
3611- ``Q``: A memory address operand with a base address and a 12-bit immediate
3612  unsigned displacement.
3613- ``R``: A memory address operand with a base address, a 12-bit immediate
3614  unsigned displacement, and an index register.
3615- ``S``: A memory address operand with a base address and a 20-bit immediate
3616  signed displacement.
3617- ``T``: A memory address operand with a base address, a 20-bit immediate
3618  signed displacement, and an index register.
3619- ``r`` or ``d``: A 32, 64, or 128-bit integer register.
3620- ``a``: A 32, 64, or 128-bit integer address register (excludes R0, which in an
3621  address context evaluates as zero).
3622- ``h``: A 32-bit value in the high part of a 64bit data register
3623  (LLVM-specific)
3624- ``f``: A 32, 64, or 128-bit floating point register.
3625
3626X86:
3627
3628- ``I``: An immediate integer between 0 and 31.
3629- ``J``: An immediate integer between 0 and 64.
3630- ``K``: An immediate signed 8-bit integer.
3631- ``L``: An immediate integer, 0xff or 0xffff or (in 64-bit mode only)
3632  0xffffffff.
3633- ``M``: An immediate integer between 0 and 3.
3634- ``N``: An immediate unsigned 8-bit integer.
3635- ``O``: An immediate integer between 0 and 127.
3636- ``e``: An immediate 32-bit signed integer.
3637- ``Z``: An immediate 32-bit unsigned integer.
3638- ``o``, ``v``: Treated the same as ``m``, at the moment.
3639- ``q``: An 8, 16, 32, or 64-bit register which can be accessed as an 8-bit
3640  ``l`` integer register. On X86-32, this is the ``a``, ``b``, ``c``, and ``d``
3641  registers, and on X86-64, it is all of the integer registers.
3642- ``Q``: An 8, 16, 32, or 64-bit register which can be accessed as an 8-bit
3643  ``h`` integer register. This is the ``a``, ``b``, ``c``, and ``d`` registers.
3644- ``r`` or ``l``: An 8, 16, 32, or 64-bit integer register.
3645- ``R``: An 8, 16, 32, or 64-bit "legacy" integer register -- one which has
3646  existed since i386, and can be accessed without the REX prefix.
3647- ``f``: A 32, 64, or 80-bit '387 FPU stack pseudo-register.
3648- ``y``: A 64-bit MMX register, if MMX is enabled.
3649- ``x``: If SSE is enabled: a 32 or 64-bit scalar operand, or 128-bit vector
3650  operand in a SSE register. If AVX is also enabled, can also be a 256-bit
3651  vector operand in an AVX register. If AVX-512 is also enabled, can also be a
3652  512-bit vector operand in an AVX512 register, Otherwise, an error.
3653- ``Y``: The same as ``x``, if *SSE2* is enabled, otherwise an error.
3654- ``A``: Special case: allocates EAX first, then EDX, for a single operand (in
3655  32-bit mode, a 64-bit integer operand will get split into two registers). It
3656  is not recommended to use this constraint, as in 64-bit mode, the 64-bit
3657  operand will get allocated only to RAX -- if two 32-bit operands are needed,
3658  you're better off splitting it yourself, before passing it to the asm
3659  statement.
3660
3661XCore:
3662
3663- ``r``: A 32-bit integer register.
3664
3665
3666.. _inline-asm-modifiers:
3667
3668Asm template argument modifiers
3669^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3670
3671In the asm template string, modifiers can be used on the operand reference, like
3672"``${0:n}``".
3673
3674The modifiers are, in general, expected to behave the same way they do in
3675GCC. LLVM's support is often implemented on an 'as-needed' basis, to support C
3676inline asm code which was supported by GCC. A mismatch in behavior between LLVM
3677and GCC likely indicates a bug in LLVM.
3678
3679Target-independent:
3680
3681- ``c``: Print an immediate integer constant unadorned, without
3682  the target-specific immediate punctuation (e.g. no ``$`` prefix).
3683- ``n``: Negate and print immediate integer constant unadorned, without the
3684  target-specific immediate punctuation (e.g. no ``$`` prefix).
3685- ``l``: Print as an unadorned label, without the target-specific label
3686  punctuation (e.g. no ``$`` prefix).
3687
3688AArch64:
3689
3690- ``w``: Print a GPR register with a ``w*`` name instead of ``x*`` name. E.g.,
3691  instead of ``x30``, print ``w30``.
3692- ``x``: Print a GPR register with a ``x*`` name. (this is the default, anyhow).
3693- ``b``, ``h``, ``s``, ``d``, ``q``: Print a floating-point/SIMD register with a
3694  ``b*``, ``h*``, ``s*``, ``d*``, or ``q*`` name, rather than the default of
3695  ``v*``.
3696
3697AMDGPU:
3698
3699- ``r``: No effect.
3700
3701ARM:
3702
3703- ``a``: Print an operand as an address (with ``[`` and ``]`` surrounding a
3704  register).
3705- ``P``: No effect.
3706- ``q``: No effect.
3707- ``y``: Print a VFP single-precision register as an indexed double (e.g. print
3708  as ``d4[1]`` instead of ``s9``)
3709- ``B``: Bitwise invert and print an immediate integer constant without ``#``
3710  prefix.
3711- ``L``: Print the low 16-bits of an immediate integer constant.
3712- ``M``: Print as a register set suitable for ldm/stm. Also prints *all*
3713  register operands subsequent to the specified one (!), so use carefully.
3714- ``Q``: Print the low-order register of a register-pair, or the low-order
3715  register of a two-register operand.
3716- ``R``: Print the high-order register of a register-pair, or the high-order
3717  register of a two-register operand.
3718- ``H``: Print the second register of a register-pair. (On a big-endian system,
3719  ``H`` is equivalent to ``Q``, and on little-endian system, ``H`` is equivalent
3720  to ``R``.)
3721
3722  .. FIXME: H doesn't currently support printing the second register
3723     of a two-register operand.
3724
3725- ``e``: Print the low doubleword register of a NEON quad register.
3726- ``f``: Print the high doubleword register of a NEON quad register.
3727- ``m``: Print the base register of a memory operand without the ``[`` and ``]``
3728  adornment.
3729
3730Hexagon:
3731
3732- ``L``: Print the second register of a two-register operand. Requires that it
3733  has been allocated consecutively to the first.
3734
3735  .. FIXME: why is it restricted to consecutive ones? And there's
3736     nothing that ensures that happens, is there?
3737
3738- ``I``: Print the letter 'i' if the operand is an integer constant, otherwise
3739  nothing. Used to print 'addi' vs 'add' instructions.
3740
3741MSP430:
3742
3743No additional modifiers.
3744
3745MIPS:
3746
3747- ``X``: Print an immediate integer as hexadecimal
3748- ``x``: Print the low 16 bits of an immediate integer as hexadecimal.
3749- ``d``: Print an immediate integer as decimal.
3750- ``m``: Subtract one and print an immediate integer as decimal.
3751- ``z``: Print $0 if an immediate zero, otherwise print normally.
3752- ``L``: Print the low-order register of a two-register operand, or prints the
3753  address of the low-order word of a double-word memory operand.
3754
3755  .. FIXME: L seems to be missing memory operand support.
3756
3757- ``M``: Print the high-order register of a two-register operand, or prints the
3758  address of the high-order word of a double-word memory operand.
3759
3760  .. FIXME: M seems to be missing memory operand support.
3761
3762- ``D``: Print the second register of a two-register operand, or prints the
3763  second word of a double-word memory operand. (On a big-endian system, ``D`` is
3764  equivalent to ``L``, and on little-endian system, ``D`` is equivalent to
3765  ``M``.)
3766- ``w``: No effect. Provided for compatibility with GCC which requires this
3767  modifier in order to print MSA registers (``W0-W31``) with the ``f``
3768  constraint.
3769
3770NVPTX:
3771
3772- ``r``: No effect.
3773
3774PowerPC:
3775
3776- ``L``: Print the second register of a two-register operand. Requires that it
3777  has been allocated consecutively to the first.
3778
3779  .. FIXME: why is it restricted to consecutive ones? And there's
3780     nothing that ensures that happens, is there?
3781
3782- ``I``: Print the letter 'i' if the operand is an integer constant, otherwise
3783  nothing. Used to print 'addi' vs 'add' instructions.
3784- ``y``: For a memory operand, prints formatter for a two-register X-form
3785  instruction. (Currently always prints ``r0,OPERAND``).
3786- ``U``: Prints 'u' if the memory operand is an update form, and nothing
3787  otherwise. (NOTE: LLVM does not support update form, so this will currently
3788  always print nothing)
3789- ``X``: Prints 'x' if the memory operand is an indexed form. (NOTE: LLVM does
3790  not support indexed form, so this will currently always print nothing)
3791
3792Sparc:
3793
3794- ``r``: No effect.
3795
3796SystemZ:
3797
3798SystemZ implements only ``n``, and does *not* support any of the other
3799target-independent modifiers.
3800
3801X86:
3802
3803- ``c``: Print an unadorned integer or symbol name. (The latter is
3804  target-specific behavior for this typically target-independent modifier).
3805- ``A``: Print a register name with a '``*``' before it.
3806- ``b``: Print an 8-bit register name (e.g. ``al``); do nothing on a memory
3807  operand.
3808- ``h``: Print the upper 8-bit register name (e.g. ``ah``); do nothing on a
3809  memory operand.
3810- ``w``: Print the 16-bit register name (e.g. ``ax``); do nothing on a memory
3811  operand.
3812- ``k``: Print the 32-bit register name (e.g. ``eax``); do nothing on a memory
3813  operand.
3814- ``q``: Print the 64-bit register name (e.g. ``rax``), if 64-bit registers are
3815  available, otherwise the 32-bit register name; do nothing on a memory operand.
3816- ``n``: Negate and print an unadorned integer, or, for operands other than an
3817  immediate integer (e.g. a relocatable symbol expression), print a '-' before
3818  the operand. (The behavior for relocatable symbol expressions is a
3819  target-specific behavior for this typically target-independent modifier)
3820- ``H``: Print a memory reference with additional offset +8.
3821- ``P``: Print a memory reference or operand for use as the argument of a call
3822  instruction. (E.g. omit ``(rip)``, even though it's PC-relative.)
3823
3824XCore:
3825
3826No additional modifiers.
3827
3828
3829Inline Asm Metadata
3830^^^^^^^^^^^^^^^^^^^
3831
3832The call instructions that wrap inline asm nodes may have a
3833"``!srcloc``" MDNode attached to it that contains a list of constant
3834integers. If present, the code generator will use the integer as the
3835location cookie value when report errors through the ``LLVMContext``
3836error reporting mechanisms. This allows a front-end to correlate backend
3837errors that occur with inline asm back to the source code that produced
3838it. For example:
3839
3840.. code-block:: llvm
3841
3842    call void asm sideeffect "something bad", ""(), !srcloc !42
3843    ...
3844    !42 = !{ i32 1234567 }
3845
3846It is up to the front-end to make sense of the magic numbers it places
3847in the IR. If the MDNode contains multiple constants, the code generator
3848will use the one that corresponds to the line of the asm that the error
3849occurs on.
3850
3851.. _metadata:
3852
3853Metadata
3854========
3855
3856LLVM IR allows metadata to be attached to instructions in the program
3857that can convey extra information about the code to the optimizers and
3858code generator. One example application of metadata is source-level
3859debug information. There are two metadata primitives: strings and nodes.
3860
3861Metadata does not have a type, and is not a value. If referenced from a
3862``call`` instruction, it uses the ``metadata`` type.
3863
3864All metadata are identified in syntax by a exclamation point ('``!``').
3865
3866.. _metadata-string:
3867
3868Metadata Nodes and Metadata Strings
3869-----------------------------------
3870
3871A metadata string is a string surrounded by double quotes. It can
3872contain any character by escaping non-printable characters with
3873"``\xx``" where "``xx``" is the two digit hex code. For example:
3874"``!"test\00"``".
3875
3876Metadata nodes are represented with notation similar to structure
3877constants (a comma separated list of elements, surrounded by braces and
3878preceded by an exclamation point). Metadata nodes can have any values as
3879their operand. For example:
3880
3881.. code-block:: llvm
3882
3883    !{ !"test\00", i32 10}
3884
3885Metadata nodes that aren't uniqued use the ``distinct`` keyword. For example:
3886
3887.. code-block:: llvm
3888
3889    !0 = distinct !{!"test\00", i32 10}
3890
3891``distinct`` nodes are useful when nodes shouldn't be merged based on their
3892content. They can also occur when transformations cause uniquing collisions
3893when metadata operands change.
3894
3895A :ref:`named metadata <namedmetadatastructure>` is a collection of
3896metadata nodes, which can be looked up in the module symbol table. For
3897example:
3898
3899.. code-block:: llvm
3900
3901    !foo = !{!4, !3}
3902
3903Metadata can be used as function arguments. Here ``llvm.dbg.value``
3904function is using two metadata arguments:
3905
3906.. code-block:: llvm
3907
3908    call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
3909
3910Metadata can be attached to an instruction. Here metadata ``!21`` is attached
3911to the ``add`` instruction using the ``!dbg`` identifier:
3912
3913.. code-block:: llvm
3914
3915    %indvar.next = add i64 %indvar, 1, !dbg !21
3916
3917Metadata can also be attached to a function definition. Here metadata ``!22``
3918is attached to the ``foo`` function using the ``!dbg`` identifier:
3919
3920.. code-block:: llvm
3921
3922    define void @foo() !dbg !22 {
3923      ret void
3924    }
3925
3926More information about specific metadata nodes recognized by the
3927optimizers and code generator is found below.
3928
3929.. _specialized-metadata:
3930
3931Specialized Metadata Nodes
3932^^^^^^^^^^^^^^^^^^^^^^^^^^
3933
3934Specialized metadata nodes are custom data structures in metadata (as opposed
3935to generic tuples). Their fields are labelled, and can be specified in any
3936order.
3937
3938These aren't inherently debug info centric, but currently all the specialized
3939metadata nodes are related to debug info.
3940
3941.. _DICompileUnit:
3942
3943DICompileUnit
3944"""""""""""""
3945
3946``DICompileUnit`` nodes represent a compile unit. The ``enums:``,
3947``retainedTypes:``, ``subprograms:``, ``globals:``, ``imports:`` and ``macros:``
3948fields are tuples containing the debug info to be emitted along with the compile
3949unit, regardless of code optimizations (some nodes are only emitted if there are
3950references to them from instructions).
3951
3952.. code-block:: llvm
3953
3954    !0 = !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang",
3955                        isOptimized: true, flags: "-O2", runtimeVersion: 2,
3956                        splitDebugFilename: "abc.debug", emissionKind: FullDebug,
3957                        enums: !2, retainedTypes: !3, subprograms: !4,
3958                        globals: !5, imports: !6, macros: !7, dwoId: 0x0abcd)
3959
3960Compile unit descriptors provide the root scope for objects declared in a
3961specific compilation unit. File descriptors are defined using this scope.
3962These descriptors are collected by a named metadata ``!llvm.dbg.cu``. They
3963keep track of subprograms, global variables, type information, and imported
3964entities (declarations and namespaces).
3965
3966.. _DIFile:
3967
3968DIFile
3969""""""
3970
3971``DIFile`` nodes represent files. The ``filename:`` can include slashes.
3972
3973.. code-block:: llvm
3974
3975    !0 = !DIFile(filename: "path/to/file", directory: "/path/to/dir")
3976
3977Files are sometimes used in ``scope:`` fields, and are the only valid target
3978for ``file:`` fields.
3979
3980.. _DIBasicType:
3981
3982DIBasicType
3983"""""""""""
3984
3985``DIBasicType`` nodes represent primitive types, such as ``int``, ``bool`` and
3986``float``. ``tag:`` defaults to ``DW_TAG_base_type``.
3987
3988.. code-block:: llvm
3989
3990    !0 = !DIBasicType(name: "unsigned char", size: 8, align: 8,
3991                      encoding: DW_ATE_unsigned_char)
3992    !1 = !DIBasicType(tag: DW_TAG_unspecified_type, name: "decltype(nullptr)")
3993
3994The ``encoding:`` describes the details of the type. Usually it's one of the
3995following:
3996
3997.. code-block:: llvm
3998
3999  DW_ATE_address       = 1
4000  DW_ATE_boolean       = 2
4001  DW_ATE_float         = 4
4002  DW_ATE_signed        = 5
4003  DW_ATE_signed_char   = 6
4004  DW_ATE_unsigned      = 7
4005  DW_ATE_unsigned_char = 8
4006
4007.. _DISubroutineType:
4008
4009DISubroutineType
4010""""""""""""""""
4011
4012``DISubroutineType`` nodes represent subroutine types. Their ``types:`` field
4013refers to a tuple; the first operand is the return type, while the rest are the
4014types of the formal arguments in order. If the first operand is ``null``, that
4015represents a function with no return value (such as ``void foo() {}`` in C++).
4016
4017.. code-block:: llvm
4018
4019    !0 = !BasicType(name: "int", size: 32, align: 32, DW_ATE_signed)
4020    !1 = !BasicType(name: "char", size: 8, align: 8, DW_ATE_signed_char)
4021    !2 = !DISubroutineType(types: !{null, !0, !1}) ; void (int, char)
4022
4023.. _DIDerivedType:
4024
4025DIDerivedType
4026"""""""""""""
4027
4028``DIDerivedType`` nodes represent types derived from other types, such as
4029qualified types.
4030
4031.. code-block:: llvm
4032
4033    !0 = !DIBasicType(name: "unsigned char", size: 8, align: 8,
4034                      encoding: DW_ATE_unsigned_char)
4035    !1 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !0, size: 32,
4036                        align: 32)
4037
4038The following ``tag:`` values are valid:
4039
4040.. code-block:: llvm
4041
4042  DW_TAG_member             = 13
4043  DW_TAG_pointer_type       = 15
4044  DW_TAG_reference_type     = 16
4045  DW_TAG_typedef            = 22
4046  DW_TAG_inheritance        = 28
4047  DW_TAG_ptr_to_member_type = 31
4048  DW_TAG_const_type         = 38
4049  DW_TAG_friend             = 42
4050  DW_TAG_volatile_type      = 53
4051  DW_TAG_restrict_type      = 55
4052
4053.. _DIDerivedTypeMember:
4054
4055``DW_TAG_member`` is used to define a member of a :ref:`composite type
4056<DICompositeType>`. The type of the member is the ``baseType:``. The
4057``offset:`` is the member's bit offset.  If the composite type has an ODR
4058``identifier:`` and does not set ``flags: DIFwdDecl``, then the member is
4059uniqued based only on its ``name:`` and ``scope:``.
4060
4061``DW_TAG_inheritance`` and ``DW_TAG_friend`` are used in the ``elements:``
4062field of :ref:`composite types <DICompositeType>` to describe parents and
4063friends.
4064
4065``DW_TAG_typedef`` is used to provide a name for the ``baseType:``.
4066
4067``DW_TAG_pointer_type``, ``DW_TAG_reference_type``, ``DW_TAG_const_type``,
4068``DW_TAG_volatile_type`` and ``DW_TAG_restrict_type`` are used to qualify the
4069``baseType:``.
4070
4071Note that the ``void *`` type is expressed as a type derived from NULL.
4072
4073.. _DICompositeType:
4074
4075DICompositeType
4076"""""""""""""""
4077
4078``DICompositeType`` nodes represent types composed of other types, like
4079structures and unions. ``elements:`` points to a tuple of the composed types.
4080
4081If the source language supports ODR, the ``identifier:`` field gives the unique
4082identifier used for type merging between modules.  When specified,
4083:ref:`subprogram declarations <DISubprogramDeclaration>` and :ref:`member
4084derived types <DIDerivedTypeMember>` that reference the ODR-type in their
4085``scope:`` change uniquing rules.
4086
4087For a given ``identifier:``, there should only be a single composite type that
4088does not have  ``flags: DIFlagFwdDecl`` set.  LLVM tools that link modules
4089together will unique such definitions at parse time via the ``identifier:``
4090field, even if the nodes are ``distinct``.
4091
4092.. code-block:: llvm
4093
4094    !0 = !DIEnumerator(name: "SixKind", value: 7)
4095    !1 = !DIEnumerator(name: "SevenKind", value: 7)
4096    !2 = !DIEnumerator(name: "NegEightKind", value: -8)
4097    !3 = !DICompositeType(tag: DW_TAG_enumeration_type, name: "Enum", file: !12,
4098                          line: 2, size: 32, align: 32, identifier: "_M4Enum",
4099                          elements: !{!0, !1, !2})
4100
4101The following ``tag:`` values are valid:
4102
4103.. code-block:: llvm
4104
4105  DW_TAG_array_type       = 1
4106  DW_TAG_class_type       = 2
4107  DW_TAG_enumeration_type = 4
4108  DW_TAG_structure_type   = 19
4109  DW_TAG_union_type       = 23
4110
4111For ``DW_TAG_array_type``, the ``elements:`` should be :ref:`subrange
4112descriptors <DISubrange>`, each representing the range of subscripts at that
4113level of indexing. The ``DIFlagVector`` flag to ``flags:`` indicates that an
4114array type is a native packed vector.
4115
4116For ``DW_TAG_enumeration_type``, the ``elements:`` should be :ref:`enumerator
4117descriptors <DIEnumerator>`, each representing the definition of an enumeration
4118value for the set. All enumeration type descriptors are collected in the
4119``enums:`` field of the :ref:`compile unit <DICompileUnit>`.
4120
4121For ``DW_TAG_structure_type``, ``DW_TAG_class_type``, and
4122``DW_TAG_union_type``, the ``elements:`` should be :ref:`derived types
4123<DIDerivedType>` with ``tag: DW_TAG_member``, ``tag: DW_TAG_inheritance``, or
4124``tag: DW_TAG_friend``; or :ref:`subprograms <DISubprogram>` with
4125``isDefinition: false``.
4126
4127.. _DISubrange:
4128
4129DISubrange
4130""""""""""
4131
4132``DISubrange`` nodes are the elements for ``DW_TAG_array_type`` variants of
4133:ref:`DICompositeType`. ``count: -1`` indicates an empty array.
4134
4135.. code-block:: llvm
4136
4137    !0 = !DISubrange(count: 5, lowerBound: 0) ; array counting from 0
4138    !1 = !DISubrange(count: 5, lowerBound: 1) ; array counting from 1
4139    !2 = !DISubrange(count: -1) ; empty array.
4140
4141.. _DIEnumerator:
4142
4143DIEnumerator
4144""""""""""""
4145
4146``DIEnumerator`` nodes are the elements for ``DW_TAG_enumeration_type``
4147variants of :ref:`DICompositeType`.
4148
4149.. code-block:: llvm
4150
4151    !0 = !DIEnumerator(name: "SixKind", value: 7)
4152    !1 = !DIEnumerator(name: "SevenKind", value: 7)
4153    !2 = !DIEnumerator(name: "NegEightKind", value: -8)
4154
4155DITemplateTypeParameter
4156"""""""""""""""""""""""
4157
4158``DITemplateTypeParameter`` nodes represent type parameters to generic source
4159language constructs. They are used (optionally) in :ref:`DICompositeType` and
4160:ref:`DISubprogram` ``templateParams:`` fields.
4161
4162.. code-block:: llvm
4163
4164    !0 = !DITemplateTypeParameter(name: "Ty", type: !1)
4165
4166DITemplateValueParameter
4167""""""""""""""""""""""""
4168
4169``DITemplateValueParameter`` nodes represent value parameters to generic source
4170language constructs. ``tag:`` defaults to ``DW_TAG_template_value_parameter``,
4171but if specified can also be set to ``DW_TAG_GNU_template_template_param`` or
4172``DW_TAG_GNU_template_param_pack``. They are used (optionally) in
4173:ref:`DICompositeType` and :ref:`DISubprogram` ``templateParams:`` fields.
4174
4175.. code-block:: llvm
4176
4177    !0 = !DITemplateValueParameter(name: "Ty", type: !1, value: i32 7)
4178
4179DINamespace
4180"""""""""""
4181
4182``DINamespace`` nodes represent namespaces in the source language.
4183
4184.. code-block:: llvm
4185
4186    !0 = !DINamespace(name: "myawesomeproject", scope: !1, file: !2, line: 7)
4187
4188DIGlobalVariable
4189""""""""""""""""
4190
4191``DIGlobalVariable`` nodes represent global variables in the source language.
4192
4193.. code-block:: llvm
4194
4195    !0 = !DIGlobalVariable(name: "foo", linkageName: "foo", scope: !1,
4196                           file: !2, line: 7, type: !3, isLocal: true,
4197                           isDefinition: false, variable: i32* @foo,
4198                           declaration: !4)
4199
4200All global variables should be referenced by the `globals:` field of a
4201:ref:`compile unit <DICompileUnit>`.
4202
4203.. _DISubprogram:
4204
4205DISubprogram
4206""""""""""""
4207
4208``DISubprogram`` nodes represent functions from the source language. A
4209``DISubprogram`` may be attached to a function definition using ``!dbg``
4210metadata. The ``variables:`` field points at :ref:`variables <DILocalVariable>`
4211that must be retained, even if their IR counterparts are optimized out of
4212the IR. The ``type:`` field must point at an :ref:`DISubroutineType`.
4213
4214.. _DISubprogramDeclaration:
4215
4216When ``isDefinition: false``, subprograms describe a declaration in the type
4217tree as opposed to a definition of a function.  If the scope is a composite
4218type with an ODR ``identifier:`` and that does not set ``flags: DIFwdDecl``,
4219then the subprogram declaration is uniqued based only on its ``linkageName:``
4220and ``scope:``.
4221
4222.. code-block:: llvm
4223
4224    define void @_Z3foov() !dbg !0 {
4225      ...
4226    }
4227
4228    !0 = distinct !DISubprogram(name: "foo", linkageName: "_Zfoov", scope: !1,
4229                                file: !2, line: 7, type: !3, isLocal: true,
4230                                isDefinition: true, scopeLine: 8,
4231                                containingType: !4,
4232                                virtuality: DW_VIRTUALITY_pure_virtual,
4233                                virtualIndex: 10, flags: DIFlagPrototyped,
4234                                isOptimized: true, templateParams: !5,
4235                                declaration: !6, variables: !7)
4236
4237.. _DILexicalBlock:
4238
4239DILexicalBlock
4240""""""""""""""
4241
4242``DILexicalBlock`` nodes describe nested blocks within a :ref:`subprogram
4243<DISubprogram>`. The line number and column numbers are used to distinguish
4244two lexical blocks at same depth. They are valid targets for ``scope:``
4245fields.
4246
4247.. code-block:: llvm
4248
4249    !0 = distinct !DILexicalBlock(scope: !1, file: !2, line: 7, column: 35)
4250
4251Usually lexical blocks are ``distinct`` to prevent node merging based on
4252operands.
4253
4254.. _DILexicalBlockFile:
4255
4256DILexicalBlockFile
4257""""""""""""""""""
4258
4259``DILexicalBlockFile`` nodes are used to discriminate between sections of a
4260:ref:`lexical block <DILexicalBlock>`. The ``file:`` field can be changed to
4261indicate textual inclusion, or the ``discriminator:`` field can be used to
4262discriminate between control flow within a single block in the source language.
4263
4264.. code-block:: llvm
4265
4266    !0 = !DILexicalBlock(scope: !3, file: !4, line: 7, column: 35)
4267    !1 = !DILexicalBlockFile(scope: !0, file: !4, discriminator: 0)
4268    !2 = !DILexicalBlockFile(scope: !0, file: !4, discriminator: 1)
4269
4270.. _DILocation:
4271
4272DILocation
4273""""""""""
4274
4275``DILocation`` nodes represent source debug locations. The ``scope:`` field is
4276mandatory, and points at an :ref:`DILexicalBlockFile`, an
4277:ref:`DILexicalBlock`, or an :ref:`DISubprogram`.
4278
4279.. code-block:: llvm
4280
4281    !0 = !DILocation(line: 2900, column: 42, scope: !1, inlinedAt: !2)
4282
4283.. _DILocalVariable:
4284
4285DILocalVariable
4286"""""""""""""""
4287
4288``DILocalVariable`` nodes represent local variables in the source language. If
4289the ``arg:`` field is set to non-zero, then this variable is a subprogram
4290parameter, and it will be included in the ``variables:`` field of its
4291:ref:`DISubprogram`.
4292
4293.. code-block:: llvm
4294
4295    !0 = !DILocalVariable(name: "this", arg: 1, scope: !3, file: !2, line: 7,
4296                          type: !3, flags: DIFlagArtificial)
4297    !1 = !DILocalVariable(name: "x", arg: 2, scope: !4, file: !2, line: 7,
4298                          type: !3)
4299    !2 = !DILocalVariable(name: "y", scope: !5, file: !2, line: 7, type: !3)
4300
4301DIExpression
4302""""""""""""
4303
4304``DIExpression`` nodes represent DWARF expression sequences. They are used in
4305:ref:`debug intrinsics<dbg_intrinsics>` (such as ``llvm.dbg.declare``) to
4306describe how the referenced LLVM variable relates to the source language
4307variable.
4308
4309The current supported vocabulary is limited:
4310
4311- ``DW_OP_deref`` dereferences the working expression.
4312- ``DW_OP_plus, 93`` adds ``93`` to the working expression.
4313- ``DW_OP_bit_piece, 16, 8`` specifies the offset and size (``16`` and ``8``
4314  here, respectively) of the variable piece from the working expression.
4315
4316.. code-block:: llvm
4317
4318    !0 = !DIExpression(DW_OP_deref)
4319    !1 = !DIExpression(DW_OP_plus, 3)
4320    !2 = !DIExpression(DW_OP_bit_piece, 3, 7)
4321    !3 = !DIExpression(DW_OP_deref, DW_OP_plus, 3, DW_OP_bit_piece, 3, 7)
4322
4323DIObjCProperty
4324""""""""""""""
4325
4326``DIObjCProperty`` nodes represent Objective-C property nodes.
4327
4328.. code-block:: llvm
4329
4330    !3 = !DIObjCProperty(name: "foo", file: !1, line: 7, setter: "setFoo",
4331                         getter: "getFoo", attributes: 7, type: !2)
4332
4333DIImportedEntity
4334""""""""""""""""
4335
4336``DIImportedEntity`` nodes represent entities (such as modules) imported into a
4337compile unit.
4338
4339.. code-block:: llvm
4340
4341   !2 = !DIImportedEntity(tag: DW_TAG_imported_module, name: "foo", scope: !0,
4342                          entity: !1, line: 7)
4343
4344DIMacro
4345"""""""
4346
4347``DIMacro`` nodes represent definition or undefinition of a macro identifiers.
4348The ``name:`` field is the macro identifier, followed by macro parameters when
4349defining a function-like macro, and the ``value`` field is the token-string
4350used to expand the macro identifier.
4351
4352.. code-block:: llvm
4353
4354   !2 = !DIMacro(macinfo: DW_MACINFO_define, line: 7, name: "foo(x)",
4355                 value: "((x) + 1)")
4356   !3 = !DIMacro(macinfo: DW_MACINFO_undef, line: 30, name: "foo")
4357
4358DIMacroFile
4359"""""""""""
4360
4361``DIMacroFile`` nodes represent inclusion of source files.
4362The ``nodes:`` field is a list of ``DIMacro`` and ``DIMacroFile`` nodes that
4363appear in the included source file.
4364
4365.. code-block:: llvm
4366
4367   !2 = !DIMacroFile(macinfo: DW_MACINFO_start_file, line: 7, file: !2,
4368                     nodes: !3)
4369
4370'``tbaa``' Metadata
4371^^^^^^^^^^^^^^^^^^^
4372
4373In LLVM IR, memory does not have types, so LLVM's own type system is not
4374suitable for doing TBAA. Instead, metadata is added to the IR to
4375describe a type system of a higher level language. This can be used to
4376implement typical C/C++ TBAA, but it can also be used to implement
4377custom alias analysis behavior for other languages.
4378
4379The current metadata format is very simple. TBAA metadata nodes have up
4380to three fields, e.g.:
4381
4382.. code-block:: llvm
4383
4384    !0 = !{ !"an example type tree" }
4385    !1 = !{ !"int", !0 }
4386    !2 = !{ !"float", !0 }
4387    !3 = !{ !"const float", !2, i64 1 }
4388
4389The first field is an identity field. It can be any value, usually a
4390metadata string, which uniquely identifies the type. The most important
4391name in the tree is the name of the root node. Two trees with different
4392root node names are entirely disjoint, even if they have leaves with
4393common names.
4394
4395The second field identifies the type's parent node in the tree, or is
4396null or omitted for a root node. A type is considered to alias all of
4397its descendants and all of its ancestors in the tree. Also, a type is
4398considered to alias all types in other trees, so that bitcode produced
4399from multiple front-ends is handled conservatively.
4400
4401If the third field is present, it's an integer which if equal to 1
4402indicates that the type is "constant" (meaning
4403``pointsToConstantMemory`` should return true; see `other useful
4404AliasAnalysis methods <AliasAnalysis.html#OtherItfs>`_).
4405
4406'``tbaa.struct``' Metadata
4407^^^^^^^^^^^^^^^^^^^^^^^^^^
4408
4409The :ref:`llvm.memcpy <int_memcpy>` is often used to implement
4410aggregate assignment operations in C and similar languages, however it
4411is defined to copy a contiguous region of memory, which is more than
4412strictly necessary for aggregate types which contain holes due to
4413padding. Also, it doesn't contain any TBAA information about the fields
4414of the aggregate.
4415
4416``!tbaa.struct`` metadata can describe which memory subregions in a
4417memcpy are padding and what the TBAA tags of the struct are.
4418
4419The current metadata format is very simple. ``!tbaa.struct`` metadata
4420nodes are a list of operands which are in conceptual groups of three.
4421For each group of three, the first operand gives the byte offset of a
4422field in bytes, the second gives its size in bytes, and the third gives
4423its tbaa tag. e.g.:
4424
4425.. code-block:: llvm
4426
4427    !4 = !{ i64 0, i64 4, !1, i64 8, i64 4, !2 }
4428
4429This describes a struct with two fields. The first is at offset 0 bytes
4430with size 4 bytes, and has tbaa tag !1. The second is at offset 8 bytes
4431and has size 4 bytes and has tbaa tag !2.
4432
4433Note that the fields need not be contiguous. In this example, there is a
44344 byte gap between the two fields. This gap represents padding which
4435does not carry useful data and need not be preserved.
4436
4437'``noalias``' and '``alias.scope``' Metadata
4438^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4439
4440``noalias`` and ``alias.scope`` metadata provide the ability to specify generic
4441noalias memory-access sets. This means that some collection of memory access
4442instructions (loads, stores, memory-accessing calls, etc.) that carry
4443``noalias`` metadata can specifically be specified not to alias with some other
4444collection of memory access instructions that carry ``alias.scope`` metadata.
4445Each type of metadata specifies a list of scopes where each scope has an id and
4446a domain.
4447
4448When evaluating an aliasing query, if for some domain, the set
4449of scopes with that domain in one instruction's ``alias.scope`` list is a
4450subset of (or equal to) the set of scopes for that domain in another
4451instruction's ``noalias`` list, then the two memory accesses are assumed not to
4452alias.
4453
4454Because scopes in one domain don't affect scopes in other domains, separate
4455domains can be used to compose multiple independent noalias sets.  This is
4456used for example during inlining.  As the noalias function parameters are
4457turned into noalias scope metadata, a new domain is used every time the
4458function is inlined.
4459
4460The metadata identifying each domain is itself a list containing one or two
4461entries. The first entry is the name of the domain. Note that if the name is a
4462string then it can be combined across functions and translation units. A
4463self-reference can be used to create globally unique domain names. A
4464descriptive string may optionally be provided as a second list entry.
4465
4466The metadata identifying each scope is also itself a list containing two or
4467three entries. The first entry is the name of the scope. Note that if the name
4468is a string then it can be combined across functions and translation units. A
4469self-reference can be used to create globally unique scope names. A metadata
4470reference to the scope's domain is the second entry. A descriptive string may
4471optionally be provided as a third list entry.
4472
4473For example,
4474
4475.. code-block:: llvm
4476
4477    ; Two scope domains:
4478    !0 = !{!0}
4479    !1 = !{!1}
4480
4481    ; Some scopes in these domains:
4482    !2 = !{!2, !0}
4483    !3 = !{!3, !0}
4484    !4 = !{!4, !1}
4485
4486    ; Some scope lists:
4487    !5 = !{!4} ; A list containing only scope !4
4488    !6 = !{!4, !3, !2}
4489    !7 = !{!3}
4490
4491    ; These two instructions don't alias:
4492    %0 = load float, float* %c, align 4, !alias.scope !5
4493    store float %0, float* %arrayidx.i, align 4, !noalias !5
4494
4495    ; These two instructions also don't alias (for domain !1, the set of scopes
4496    ; in the !alias.scope equals that in the !noalias list):
4497    %2 = load float, float* %c, align 4, !alias.scope !5
4498    store float %2, float* %arrayidx.i2, align 4, !noalias !6
4499
4500    ; These two instructions may alias (for domain !0, the set of scopes in
4501    ; the !noalias list is not a superset of, or equal to, the scopes in the
4502    ; !alias.scope list):
4503    %2 = load float, float* %c, align 4, !alias.scope !6
4504    store float %0, float* %arrayidx.i, align 4, !noalias !7
4505
4506'``fpmath``' Metadata
4507^^^^^^^^^^^^^^^^^^^^^
4508
4509``fpmath`` metadata may be attached to any instruction of floating point
4510type. It can be used to express the maximum acceptable error in the
4511result of that instruction, in ULPs, thus potentially allowing the
4512compiler to use a more efficient but less accurate method of computing
4513it. ULP is defined as follows:
4514
4515    If ``x`` is a real number that lies between two finite consecutive
4516    floating-point numbers ``a`` and ``b``, without being equal to one
4517    of them, then ``ulp(x) = |b - a|``, otherwise ``ulp(x)`` is the
4518    distance between the two non-equal finite floating-point numbers
4519    nearest ``x``. Moreover, ``ulp(NaN)`` is ``NaN``.
4520
4521The metadata node shall consist of a single positive float type number
4522representing the maximum relative error, for example:
4523
4524.. code-block:: llvm
4525
4526    !0 = !{ float 2.5 } ; maximum acceptable inaccuracy is 2.5 ULPs
4527
4528.. _range-metadata:
4529
4530'``range``' Metadata
4531^^^^^^^^^^^^^^^^^^^^
4532
4533``range`` metadata may be attached only to ``load``, ``call`` and ``invoke`` of
4534integer types. It expresses the possible ranges the loaded value or the value
4535returned by the called function at this call site is in. The ranges are
4536represented with a flattened list of integers. The loaded value or the value
4537returned is known to be in the union of the ranges defined by each consecutive
4538pair. Each pair has the following properties:
4539
4540-  The type must match the type loaded by the instruction.
4541-  The pair ``a,b`` represents the range ``[a,b)``.
4542-  Both ``a`` and ``b`` are constants.
4543-  The range is allowed to wrap.
4544-  The range should not represent the full or empty set. That is,
4545   ``a!=b``.
4546
4547In addition, the pairs must be in signed order of the lower bound and
4548they must be non-contiguous.
4549
4550Examples:
4551
4552.. code-block:: llvm
4553
4554      %a = load i8, i8* %x, align 1, !range !0 ; Can only be 0 or 1
4555      %b = load i8, i8* %y, align 1, !range !1 ; Can only be 255 (-1), 0 or 1
4556      %c = call i8 @foo(),       !range !2 ; Can only be 0, 1, 3, 4 or 5
4557      %d = invoke i8 @bar() to label %cont
4558             unwind label %lpad, !range !3 ; Can only be -2, -1, 3, 4 or 5
4559    ...
4560    !0 = !{ i8 0, i8 2 }
4561    !1 = !{ i8 255, i8 2 }
4562    !2 = !{ i8 0, i8 2, i8 3, i8 6 }
4563    !3 = !{ i8 -2, i8 0, i8 3, i8 6 }
4564
4565'``unpredictable``' Metadata
4566^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4567
4568``unpredictable`` metadata may be attached to any branch or switch
4569instruction. It can be used to express the unpredictability of control
4570flow. Similar to the llvm.expect intrinsic, it may be used to alter
4571optimizations related to compare and branch instructions. The metadata
4572is treated as a boolean value; if it exists, it signals that the branch
4573or switch that it is attached to is completely unpredictable.
4574
4575'``llvm.loop``'
4576^^^^^^^^^^^^^^^
4577
4578It is sometimes useful to attach information to loop constructs. Currently,
4579loop metadata is implemented as metadata attached to the branch instruction
4580in the loop latch block. This type of metadata refer to a metadata node that is
4581guaranteed to be separate for each loop. The loop identifier metadata is
4582specified with the name ``llvm.loop``.
4583
4584The loop identifier metadata is implemented using a metadata that refers to
4585itself to avoid merging it with any other identifier metadata, e.g.,
4586during module linkage or function inlining. That is, each loop should refer
4587to their own identification metadata even if they reside in separate functions.
4588The following example contains loop identifier metadata for two separate loop
4589constructs:
4590
4591.. code-block:: llvm
4592
4593    !0 = !{!0}
4594    !1 = !{!1}
4595
4596The loop identifier metadata can be used to specify additional
4597per-loop metadata. Any operands after the first operand can be treated
4598as user-defined metadata. For example the ``llvm.loop.unroll.count``
4599suggests an unroll factor to the loop unroller:
4600
4601.. code-block:: llvm
4602
4603      br i1 %exitcond, label %._crit_edge, label %.lr.ph, !llvm.loop !0
4604    ...
4605    !0 = !{!0, !1}
4606    !1 = !{!"llvm.loop.unroll.count", i32 4}
4607
4608'``llvm.loop.vectorize``' and '``llvm.loop.interleave``'
4609^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4610
4611Metadata prefixed with ``llvm.loop.vectorize`` or ``llvm.loop.interleave`` are
4612used to control per-loop vectorization and interleaving parameters such as
4613vectorization width and interleave count. These metadata should be used in
4614conjunction with ``llvm.loop`` loop identification metadata. The
4615``llvm.loop.vectorize`` and ``llvm.loop.interleave`` metadata are only
4616optimization hints and the optimizer will only interleave and vectorize loops if
4617it believes it is safe to do so. The ``llvm.mem.parallel_loop_access`` metadata
4618which contains information about loop-carried memory dependencies can be helpful
4619in determining the safety of these transformations.
4620
4621'``llvm.loop.interleave.count``' Metadata
4622^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4623
4624This metadata suggests an interleave count to the loop interleaver.
4625The first operand is the string ``llvm.loop.interleave.count`` and the
4626second operand is an integer specifying the interleave count. For
4627example:
4628
4629.. code-block:: llvm
4630
4631   !0 = !{!"llvm.loop.interleave.count", i32 4}
4632
4633Note that setting ``llvm.loop.interleave.count`` to 1 disables interleaving
4634multiple iterations of the loop. If ``llvm.loop.interleave.count`` is set to 0
4635then the interleave count will be determined automatically.
4636
4637'``llvm.loop.vectorize.enable``' Metadata
4638^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4639
4640This metadata selectively enables or disables vectorization for the loop. The
4641first operand is the string ``llvm.loop.vectorize.enable`` and the second operand
4642is a bit. If the bit operand value is 1 vectorization is enabled. A value of
46430 disables vectorization:
4644
4645.. code-block:: llvm
4646
4647   !0 = !{!"llvm.loop.vectorize.enable", i1 0}
4648   !1 = !{!"llvm.loop.vectorize.enable", i1 1}
4649
4650'``llvm.loop.vectorize.width``' Metadata
4651^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4652
4653This metadata sets the target width of the vectorizer. The first
4654operand is the string ``llvm.loop.vectorize.width`` and the second
4655operand is an integer specifying the width. For example:
4656
4657.. code-block:: llvm
4658
4659   !0 = !{!"llvm.loop.vectorize.width", i32 4}
4660
4661Note that setting ``llvm.loop.vectorize.width`` to 1 disables
4662vectorization of the loop. If ``llvm.loop.vectorize.width`` is set to
46630 or if the loop does not have this metadata the width will be
4664determined automatically.
4665
4666'``llvm.loop.unroll``'
4667^^^^^^^^^^^^^^^^^^^^^^
4668
4669Metadata prefixed with ``llvm.loop.unroll`` are loop unrolling
4670optimization hints such as the unroll factor. ``llvm.loop.unroll``
4671metadata should be used in conjunction with ``llvm.loop`` loop
4672identification metadata. The ``llvm.loop.unroll`` metadata are only
4673optimization hints and the unrolling will only be performed if the
4674optimizer believes it is safe to do so.
4675
4676'``llvm.loop.unroll.count``' Metadata
4677^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4678
4679This metadata suggests an unroll factor to the loop unroller. The
4680first operand is the string ``llvm.loop.unroll.count`` and the second
4681operand is a positive integer specifying the unroll factor. For
4682example:
4683
4684.. code-block:: llvm
4685
4686   !0 = !{!"llvm.loop.unroll.count", i32 4}
4687
4688If the trip count of the loop is less than the unroll count the loop
4689will be partially unrolled.
4690
4691'``llvm.loop.unroll.disable``' Metadata
4692^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4693
4694This metadata disables loop unrolling. The metadata has a single operand
4695which is the string ``llvm.loop.unroll.disable``. For example:
4696
4697.. code-block:: llvm
4698
4699   !0 = !{!"llvm.loop.unroll.disable"}
4700
4701'``llvm.loop.unroll.runtime.disable``' Metadata
4702^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4703
4704This metadata disables runtime loop unrolling. The metadata has a single
4705operand which is the string ``llvm.loop.unroll.runtime.disable``. For example:
4706
4707.. code-block:: llvm
4708
4709   !0 = !{!"llvm.loop.unroll.runtime.disable"}
4710
4711'``llvm.loop.unroll.enable``' Metadata
4712^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4713
4714This metadata suggests that the loop should be fully unrolled if the trip count
4715is known at compile time and partially unrolled if the trip count is not known
4716at compile time. The metadata has a single operand which is the string
4717``llvm.loop.unroll.enable``.  For example:
4718
4719.. code-block:: llvm
4720
4721   !0 = !{!"llvm.loop.unroll.enable"}
4722
4723'``llvm.loop.unroll.full``' Metadata
4724^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4725
4726This metadata suggests that the loop should be unrolled fully. The
4727metadata has a single operand which is the string ``llvm.loop.unroll.full``.
4728For example:
4729
4730.. code-block:: llvm
4731
4732   !0 = !{!"llvm.loop.unroll.full"}
4733
4734'``llvm.loop.licm_versioning.disable``' Metadata
4735^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4736
4737This metadata indicates that the loop should not be versioned for the purpose
4738of enabling loop-invariant code motion (LICM). The metadata has a single operand
4739which is the string ``llvm.loop.licm_versioning.disable``. For example:
4740
4741.. code-block:: llvm
4742
4743   !0 = !{!"llvm.loop.licm_versioning.disable"}
4744
4745'``llvm.loop.distribute.enable``' Metadata
4746^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4747
4748Loop distribution allows splitting a loop into multiple loops.  Currently,
4749this is only performed if the entire loop cannot be vectorized due to unsafe
4750memory dependencies.  The transformation will atempt to isolate the unsafe
4751dependencies into their own loop.
4752
4753This metadata can be used to selectively enable or disable distribution of the
4754loop.  The first operand is the string ``llvm.loop.distribute.enable`` and the
4755second operand is a bit. If the bit operand value is 1 distribution is
4756enabled. A value of 0 disables distribution:
4757
4758.. code-block:: llvm
4759
4760   !0 = !{!"llvm.loop.distribute.enable", i1 0}
4761   !1 = !{!"llvm.loop.distribute.enable", i1 1}
4762
4763This metadata should be used in conjunction with ``llvm.loop`` loop
4764identification metadata.
4765
4766'``llvm.mem``'
4767^^^^^^^^^^^^^^^
4768
4769Metadata types used to annotate memory accesses with information helpful
4770for optimizations are prefixed with ``llvm.mem``.
4771
4772'``llvm.mem.parallel_loop_access``' Metadata
4773^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4774
4775The ``llvm.mem.parallel_loop_access`` metadata refers to a loop identifier,
4776or metadata containing a list of loop identifiers for nested loops.
4777The metadata is attached to memory accessing instructions and denotes that
4778no loop carried memory dependence exist between it and other instructions denoted
4779with the same loop identifier. The metadata on memory reads also implies that
4780if conversion (i.e. speculative execution within a loop iteration) is safe.
4781
4782Precisely, given two instructions ``m1`` and ``m2`` that both have the
4783``llvm.mem.parallel_loop_access`` metadata, with ``L1`` and ``L2`` being the
4784set of loops associated with that metadata, respectively, then there is no loop
4785carried dependence between ``m1`` and ``m2`` for loops in both ``L1`` and
4786``L2``.
4787
4788As a special case, if all memory accessing instructions in a loop have
4789``llvm.mem.parallel_loop_access`` metadata that refers to that loop, then the
4790loop has no loop carried memory dependences and is considered to be a parallel
4791loop.
4792
4793Note that if not all memory access instructions have such metadata referring to
4794the loop, then the loop is considered not being trivially parallel. Additional
4795memory dependence analysis is required to make that determination. As a fail
4796safe mechanism, this causes loops that were originally parallel to be considered
4797sequential (if optimization passes that are unaware of the parallel semantics
4798insert new memory instructions into the loop body).
4799
4800Example of a loop that is considered parallel due to its correct use of
4801both ``llvm.loop`` and ``llvm.mem.parallel_loop_access``
4802metadata types that refer to the same loop identifier metadata.
4803
4804.. code-block:: llvm
4805
4806   for.body:
4807     ...
4808     %val0 = load i32, i32* %arrayidx, !llvm.mem.parallel_loop_access !0
4809     ...
4810     store i32 %val0, i32* %arrayidx1, !llvm.mem.parallel_loop_access !0
4811     ...
4812     br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0
4813
4814   for.end:
4815   ...
4816   !0 = !{!0}
4817
4818It is also possible to have nested parallel loops. In that case the
4819memory accesses refer to a list of loop identifier metadata nodes instead of
4820the loop identifier metadata node directly:
4821
4822.. code-block:: llvm
4823
4824   outer.for.body:
4825     ...
4826     %val1 = load i32, i32* %arrayidx3, !llvm.mem.parallel_loop_access !2
4827     ...
4828     br label %inner.for.body
4829
4830   inner.for.body:
4831     ...
4832     %val0 = load i32, i32* %arrayidx1, !llvm.mem.parallel_loop_access !0
4833     ...
4834     store i32 %val0, i32* %arrayidx2, !llvm.mem.parallel_loop_access !0
4835     ...
4836     br i1 %exitcond, label %inner.for.end, label %inner.for.body, !llvm.loop !1
4837
4838   inner.for.end:
4839     ...
4840     store i32 %val1, i32* %arrayidx4, !llvm.mem.parallel_loop_access !2
4841     ...
4842     br i1 %exitcond, label %outer.for.end, label %outer.for.body, !llvm.loop !2
4843
4844   outer.for.end:                                          ; preds = %for.body
4845   ...
4846   !0 = !{!1, !2} ; a list of loop identifiers
4847   !1 = !{!1} ; an identifier for the inner loop
4848   !2 = !{!2} ; an identifier for the outer loop
4849
4850'``invariant.group``' Metadata
4851^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4852
4853The ``invariant.group`` metadata may be attached to ``load``/``store`` instructions.
4854The existence of the ``invariant.group`` metadata on the instruction tells
4855the optimizer that every ``load`` and ``store`` to the same pointer operand
4856within the same invariant group can be assumed to load or store the same
4857value (but see the ``llvm.invariant.group.barrier`` intrinsic which affects
4858when two pointers are considered the same).
4859
4860Examples:
4861
4862.. code-block:: llvm
4863
4864   @unknownPtr = external global i8
4865   ...
4866   %ptr = alloca i8
4867   store i8 42, i8* %ptr, !invariant.group !0
4868   call void @foo(i8* %ptr)
4869
4870   %a = load i8, i8* %ptr, !invariant.group !0 ; Can assume that value under %ptr didn't change
4871   call void @foo(i8* %ptr)
4872   %b = load i8, i8* %ptr, !invariant.group !1 ; Can't assume anything, because group changed
4873
4874   %newPtr = call i8* @getPointer(i8* %ptr)
4875   %c = load i8, i8* %newPtr, !invariant.group !0 ; Can't assume anything, because we only have information about %ptr
4876
4877   %unknownValue = load i8, i8* @unknownPtr
4878   store i8 %unknownValue, i8* %ptr, !invariant.group !0 ; Can assume that %unknownValue == 42
4879
4880   call void @foo(i8* %ptr)
4881   %newPtr2 = call i8* @llvm.invariant.group.barrier(i8* %ptr)
4882   %d = load i8, i8* %newPtr2, !invariant.group !0  ; Can't step through invariant.group.barrier to get value of %ptr
4883
4884   ...
4885   declare void @foo(i8*)
4886   declare i8* @getPointer(i8*)
4887   declare i8* @llvm.invariant.group.barrier(i8*)
4888
4889   !0 = !{!"magic ptr"}
4890   !1 = !{!"other ptr"}
4891
4892
4893
4894Module Flags Metadata
4895=====================
4896
4897Information about the module as a whole is difficult to convey to LLVM's
4898subsystems. The LLVM IR isn't sufficient to transmit this information.
4899The ``llvm.module.flags`` named metadata exists in order to facilitate
4900this. These flags are in the form of key / value pairs --- much like a
4901dictionary --- making it easy for any subsystem who cares about a flag to
4902look it up.
4903
4904The ``llvm.module.flags`` metadata contains a list of metadata triplets.
4905Each triplet has the following form:
4906
4907-  The first element is a *behavior* flag, which specifies the behavior
4908   when two (or more) modules are merged together, and it encounters two
4909   (or more) metadata with the same ID. The supported behaviors are
4910   described below.
4911-  The second element is a metadata string that is a unique ID for the
4912   metadata. Each module may only have one flag entry for each unique ID (not
4913   including entries with the **Require** behavior).
4914-  The third element is the value of the flag.
4915
4916When two (or more) modules are merged together, the resulting
4917``llvm.module.flags`` metadata is the union of the modules' flags. That is, for
4918each unique metadata ID string, there will be exactly one entry in the merged
4919modules ``llvm.module.flags`` metadata table, and the value for that entry will
4920be determined by the merge behavior flag, as described below. The only exception
4921is that entries with the *Require* behavior are always preserved.
4922
4923The following behaviors are supported:
4924
4925.. list-table::
4926   :header-rows: 1
4927   :widths: 10 90
4928
4929   * - Value
4930     - Behavior
4931
4932   * - 1
4933     - **Error**
4934           Emits an error if two values disagree, otherwise the resulting value
4935           is that of the operands.
4936
4937   * - 2
4938     - **Warning**
4939           Emits a warning if two values disagree. The result value will be the
4940           operand for the flag from the first module being linked.
4941
4942   * - 3
4943     - **Require**
4944           Adds a requirement that another module flag be present and have a
4945           specified value after linking is performed. The value must be a
4946           metadata pair, where the first element of the pair is the ID of the
4947           module flag to be restricted, and the second element of the pair is
4948           the value the module flag should be restricted to. This behavior can
4949           be used to restrict the allowable results (via triggering of an
4950           error) of linking IDs with the **Override** behavior.
4951
4952   * - 4
4953     - **Override**
4954           Uses the specified value, regardless of the behavior or value of the
4955           other module. If both modules specify **Override**, but the values
4956           differ, an error will be emitted.
4957
4958   * - 5
4959     - **Append**
4960           Appends the two values, which are required to be metadata nodes.
4961
4962   * - 6
4963     - **AppendUnique**
4964           Appends the two values, which are required to be metadata
4965           nodes. However, duplicate entries in the second list are dropped
4966           during the append operation.
4967
4968It is an error for a particular unique flag ID to have multiple behaviors,
4969except in the case of **Require** (which adds restrictions on another metadata
4970value) or **Override**.
4971
4972An example of module flags:
4973
4974.. code-block:: llvm
4975
4976    !0 = !{ i32 1, !"foo", i32 1 }
4977    !1 = !{ i32 4, !"bar", i32 37 }
4978    !2 = !{ i32 2, !"qux", i32 42 }
4979    !3 = !{ i32 3, !"qux",
4980      !{
4981        !"foo", i32 1
4982      }
4983    }
4984    !llvm.module.flags = !{ !0, !1, !2, !3 }
4985
4986-  Metadata ``!0`` has the ID ``!"foo"`` and the value '1'. The behavior
4987   if two or more ``!"foo"`` flags are seen is to emit an error if their
4988   values are not equal.
4989
4990-  Metadata ``!1`` has the ID ``!"bar"`` and the value '37'. The
4991   behavior if two or more ``!"bar"`` flags are seen is to use the value
4992   '37'.
4993
4994-  Metadata ``!2`` has the ID ``!"qux"`` and the value '42'. The
4995   behavior if two or more ``!"qux"`` flags are seen is to emit a
4996   warning if their values are not equal.
4997
4998-  Metadata ``!3`` has the ID ``!"qux"`` and the value:
4999
5000   ::
5001
5002       !{ !"foo", i32 1 }
5003
5004   The behavior is to emit an error if the ``llvm.module.flags`` does not
5005   contain a flag with the ID ``!"foo"`` that has the value '1' after linking is
5006   performed.
5007
5008Objective-C Garbage Collection Module Flags Metadata
5009----------------------------------------------------
5010
5011On the Mach-O platform, Objective-C stores metadata about garbage
5012collection in a special section called "image info". The metadata
5013consists of a version number and a bitmask specifying what types of
5014garbage collection are supported (if any) by the file. If two or more
5015modules are linked together their garbage collection metadata needs to
5016be merged rather than appended together.
5017
5018The Objective-C garbage collection module flags metadata consists of the
5019following key-value pairs:
5020
5021.. list-table::
5022   :header-rows: 1
5023   :widths: 30 70
5024
5025   * - Key
5026     - Value
5027
5028   * - ``Objective-C Version``
5029     - **[Required]** --- The Objective-C ABI version. Valid values are 1 and 2.
5030
5031   * - ``Objective-C Image Info Version``
5032     - **[Required]** --- The version of the image info section. Currently
5033       always 0.
5034
5035   * - ``Objective-C Image Info Section``
5036     - **[Required]** --- The section to place the metadata. Valid values are
5037       ``"__OBJC, __image_info, regular"`` for Objective-C ABI version 1, and
5038       ``"__DATA,__objc_imageinfo, regular, no_dead_strip"`` for
5039       Objective-C ABI version 2.
5040
5041   * - ``Objective-C Garbage Collection``
5042     - **[Required]** --- Specifies whether garbage collection is supported or
5043       not. Valid values are 0, for no garbage collection, and 2, for garbage
5044       collection supported.
5045
5046   * - ``Objective-C GC Only``
5047     - **[Optional]** --- Specifies that only garbage collection is supported.
5048       If present, its value must be 6. This flag requires that the
5049       ``Objective-C Garbage Collection`` flag have the value 2.
5050
5051Some important flag interactions:
5052
5053-  If a module with ``Objective-C Garbage Collection`` set to 0 is
5054   merged with a module with ``Objective-C Garbage Collection`` set to
5055   2, then the resulting module has the
5056   ``Objective-C Garbage Collection`` flag set to 0.
5057-  A module with ``Objective-C Garbage Collection`` set to 0 cannot be
5058   merged with a module with ``Objective-C GC Only`` set to 6.
5059
5060Automatic Linker Flags Module Flags Metadata
5061--------------------------------------------
5062
5063Some targets support embedding flags to the linker inside individual object
5064files. Typically this is used in conjunction with language extensions which
5065allow source files to explicitly declare the libraries they depend on, and have
5066these automatically be transmitted to the linker via object files.
5067
5068These flags are encoded in the IR using metadata in the module flags section,
5069using the ``Linker Options`` key. The merge behavior for this flag is required
5070to be ``AppendUnique``, and the value for the key is expected to be a metadata
5071node which should be a list of other metadata nodes, each of which should be a
5072list of metadata strings defining linker options.
5073
5074For example, the following metadata section specifies two separate sets of
5075linker options, presumably to link against ``libz`` and the ``Cocoa``
5076framework::
5077
5078    !0 = !{ i32 6, !"Linker Options",
5079       !{
5080          !{ !"-lz" },
5081          !{ !"-framework", !"Cocoa" } } }
5082    !llvm.module.flags = !{ !0 }
5083
5084The metadata encoding as lists of lists of options, as opposed to a collapsed
5085list of options, is chosen so that the IR encoding can use multiple option
5086strings to specify e.g., a single library, while still having that specifier be
5087preserved as an atomic element that can be recognized by a target specific
5088assembly writer or object file emitter.
5089
5090Each individual option is required to be either a valid option for the target's
5091linker, or an option that is reserved by the target specific assembly writer or
5092object file emitter. No other aspect of these options is defined by the IR.
5093
5094C type width Module Flags Metadata
5095----------------------------------
5096
5097The ARM backend emits a section into each generated object file describing the
5098options that it was compiled with (in a compiler-independent way) to prevent
5099linking incompatible objects, and to allow automatic library selection. Some
5100of these options are not visible at the IR level, namely wchar_t width and enum
5101width.
5102
5103To pass this information to the backend, these options are encoded in module
5104flags metadata, using the following key-value pairs:
5105
5106.. list-table::
5107   :header-rows: 1
5108   :widths: 30 70
5109
5110   * - Key
5111     - Value
5112
5113   * - short_wchar
5114     - * 0 --- sizeof(wchar_t) == 4
5115       * 1 --- sizeof(wchar_t) == 2
5116
5117   * - short_enum
5118     - * 0 --- Enums are at least as large as an ``int``.
5119       * 1 --- Enums are stored in the smallest integer type which can
5120         represent all of its values.
5121
5122For example, the following metadata section specifies that the module was
5123compiled with a ``wchar_t`` width of 4 bytes, and the underlying type of an
5124enum is the smallest type which can represent all of its values::
5125
5126    !llvm.module.flags = !{!0, !1}
5127    !0 = !{i32 1, !"short_wchar", i32 1}
5128    !1 = !{i32 1, !"short_enum", i32 0}
5129
5130.. _intrinsicglobalvariables:
5131
5132Intrinsic Global Variables
5133==========================
5134
5135LLVM has a number of "magic" global variables that contain data that
5136affect code generation or other IR semantics. These are documented here.
5137All globals of this sort should have a section specified as
5138"``llvm.metadata``". This section and all globals that start with
5139"``llvm.``" are reserved for use by LLVM.
5140
5141.. _gv_llvmused:
5142
5143The '``llvm.used``' Global Variable
5144-----------------------------------
5145
5146The ``@llvm.used`` global is an array which has
5147:ref:`appending linkage <linkage_appending>`. This array contains a list of
5148pointers to named global variables, functions and aliases which may optionally
5149have a pointer cast formed of bitcast or getelementptr. For example, a legal
5150use of it is:
5151
5152.. code-block:: llvm
5153
5154    @X = global i8 4
5155    @Y = global i32 123
5156
5157    @llvm.used = appending global [2 x i8*] [
5158       i8* @X,
5159       i8* bitcast (i32* @Y to i8*)
5160    ], section "llvm.metadata"
5161
5162If a symbol appears in the ``@llvm.used`` list, then the compiler, assembler,
5163and linker are required to treat the symbol as if there is a reference to the
5164symbol that it cannot see (which is why they have to be named). For example, if
5165a variable has internal linkage and no references other than that from the
5166``@llvm.used`` list, it cannot be deleted. This is commonly used to represent
5167references from inline asms and other things the compiler cannot "see", and
5168corresponds to "``attribute((used))``" in GNU C.
5169
5170On some targets, the code generator must emit a directive to the
5171assembler or object file to prevent the assembler and linker from
5172molesting the symbol.
5173
5174.. _gv_llvmcompilerused:
5175
5176The '``llvm.compiler.used``' Global Variable
5177--------------------------------------------
5178
5179The ``@llvm.compiler.used`` directive is the same as the ``@llvm.used``
5180directive, except that it only prevents the compiler from touching the
5181symbol. On targets that support it, this allows an intelligent linker to
5182optimize references to the symbol without being impeded as it would be
5183by ``@llvm.used``.
5184
5185This is a rare construct that should only be used in rare circumstances,
5186and should not be exposed to source languages.
5187
5188.. _gv_llvmglobalctors:
5189
5190The '``llvm.global_ctors``' Global Variable
5191-------------------------------------------
5192
5193.. code-block:: llvm
5194
5195    %0 = type { i32, void ()*, i8* }
5196    @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor, i8* @data }]
5197
5198The ``@llvm.global_ctors`` array contains a list of constructor
5199functions, priorities, and an optional associated global or function.
5200The functions referenced by this array will be called in ascending order
5201of priority (i.e. lowest first) when the module is loaded. The order of
5202functions with the same priority is not defined.
5203
5204If the third field is present, non-null, and points to a global variable
5205or function, the initializer function will only run if the associated
5206data from the current module is not discarded.
5207
5208.. _llvmglobaldtors:
5209
5210The '``llvm.global_dtors``' Global Variable
5211-------------------------------------------
5212
5213.. code-block:: llvm
5214
5215    %0 = type { i32, void ()*, i8* }
5216    @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor, i8* @data }]
5217
5218The ``@llvm.global_dtors`` array contains a list of destructor
5219functions, priorities, and an optional associated global or function.
5220The functions referenced by this array will be called in descending
5221order of priority (i.e. highest first) when the module is unloaded. The
5222order of functions with the same priority is not defined.
5223
5224If the third field is present, non-null, and points to a global variable
5225or function, the destructor function will only run if the associated
5226data from the current module is not discarded.
5227
5228Instruction Reference
5229=====================
5230
5231The LLVM instruction set consists of several different classifications
5232of instructions: :ref:`terminator instructions <terminators>`, :ref:`binary
5233instructions <binaryops>`, :ref:`bitwise binary
5234instructions <bitwiseops>`, :ref:`memory instructions <memoryops>`, and
5235:ref:`other instructions <otherops>`.
5236
5237.. _terminators:
5238
5239Terminator Instructions
5240-----------------------
5241
5242As mentioned :ref:`previously <functionstructure>`, every basic block in a
5243program ends with a "Terminator" instruction, which indicates which
5244block should be executed after the current block is finished. These
5245terminator instructions typically yield a '``void``' value: they produce
5246control flow, not values (the one exception being the
5247':ref:`invoke <i_invoke>`' instruction).
5248
5249The terminator instructions are: ':ref:`ret <i_ret>`',
5250':ref:`br <i_br>`', ':ref:`switch <i_switch>`',
5251':ref:`indirectbr <i_indirectbr>`', ':ref:`invoke <i_invoke>`',
5252':ref:`resume <i_resume>`', ':ref:`catchswitch <i_catchswitch>`',
5253':ref:`catchret <i_catchret>`',
5254':ref:`cleanupret <i_cleanupret>`',
5255and ':ref:`unreachable <i_unreachable>`'.
5256
5257.. _i_ret:
5258
5259'``ret``' Instruction
5260^^^^^^^^^^^^^^^^^^^^^
5261
5262Syntax:
5263"""""""
5264
5265::
5266
5267      ret <type> <value>       ; Return a value from a non-void function
5268      ret void                 ; Return from void function
5269
5270Overview:
5271"""""""""
5272
5273The '``ret``' instruction is used to return control flow (and optionally
5274a value) from a function back to the caller.
5275
5276There are two forms of the '``ret``' instruction: one that returns a
5277value and then causes control flow, and one that just causes control
5278flow to occur.
5279
5280Arguments:
5281""""""""""
5282
5283The '``ret``' instruction optionally accepts a single argument, the
5284return value. The type of the return value must be a ':ref:`first
5285class <t_firstclass>`' type.
5286
5287A function is not :ref:`well formed <wellformed>` if it it has a non-void
5288return type and contains a '``ret``' instruction with no return value or
5289a return value with a type that does not match its type, or if it has a
5290void return type and contains a '``ret``' instruction with a return
5291value.
5292
5293Semantics:
5294""""""""""
5295
5296When the '``ret``' instruction is executed, control flow returns back to
5297the calling function's context. If the caller is a
5298":ref:`call <i_call>`" instruction, execution continues at the
5299instruction after the call. If the caller was an
5300":ref:`invoke <i_invoke>`" instruction, execution continues at the
5301beginning of the "normal" destination block. If the instruction returns
5302a value, that value shall set the call or invoke instruction's return
5303value.
5304
5305Example:
5306""""""""
5307
5308.. code-block:: llvm
5309
5310      ret i32 5                       ; Return an integer value of 5
5311      ret void                        ; Return from a void function
5312      ret { i32, i8 } { i32 4, i8 2 } ; Return a struct of values 4 and 2
5313
5314.. _i_br:
5315
5316'``br``' Instruction
5317^^^^^^^^^^^^^^^^^^^^
5318
5319Syntax:
5320"""""""
5321
5322::
5323
5324      br i1 <cond>, label <iftrue>, label <iffalse>
5325      br label <dest>          ; Unconditional branch
5326
5327Overview:
5328"""""""""
5329
5330The '``br``' instruction is used to cause control flow to transfer to a
5331different basic block in the current function. There are two forms of
5332this instruction, corresponding to a conditional branch and an
5333unconditional branch.
5334
5335Arguments:
5336""""""""""
5337
5338The conditional branch form of the '``br``' instruction takes a single
5339'``i1``' value and two '``label``' values. The unconditional form of the
5340'``br``' instruction takes a single '``label``' value as a target.
5341
5342Semantics:
5343""""""""""
5344
5345Upon execution of a conditional '``br``' instruction, the '``i1``'
5346argument is evaluated. If the value is ``true``, control flows to the
5347'``iftrue``' ``label`` argument. If "cond" is ``false``, control flows
5348to the '``iffalse``' ``label`` argument.
5349
5350Example:
5351""""""""
5352
5353.. code-block:: llvm
5354
5355    Test:
5356      %cond = icmp eq i32 %a, %b
5357      br i1 %cond, label %IfEqual, label %IfUnequal
5358    IfEqual:
5359      ret i32 1
5360    IfUnequal:
5361      ret i32 0
5362
5363.. _i_switch:
5364
5365'``switch``' Instruction
5366^^^^^^^^^^^^^^^^^^^^^^^^
5367
5368Syntax:
5369"""""""
5370
5371::
5372
5373      switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
5374
5375Overview:
5376"""""""""
5377
5378The '``switch``' instruction is used to transfer control flow to one of
5379several different places. It is a generalization of the '``br``'
5380instruction, allowing a branch to occur to one of many possible
5381destinations.
5382
5383Arguments:
5384""""""""""
5385
5386The '``switch``' instruction uses three parameters: an integer
5387comparison value '``value``', a default '``label``' destination, and an
5388array of pairs of comparison value constants and '``label``'s. The table
5389is not allowed to contain duplicate constant entries.
5390
5391Semantics:
5392""""""""""
5393
5394The ``switch`` instruction specifies a table of values and destinations.
5395When the '``switch``' instruction is executed, this table is searched
5396for the given value. If the value is found, control flow is transferred
5397to the corresponding destination; otherwise, control flow is transferred
5398to the default destination.
5399
5400Implementation:
5401"""""""""""""""
5402
5403Depending on properties of the target machine and the particular
5404``switch`` instruction, this instruction may be code generated in
5405different ways. For example, it could be generated as a series of
5406chained conditional branches or with a lookup table.
5407
5408Example:
5409""""""""
5410
5411.. code-block:: llvm
5412
5413     ; Emulate a conditional br instruction
5414     %Val = zext i1 %value to i32
5415     switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
5416
5417     ; Emulate an unconditional br instruction
5418     switch i32 0, label %dest [ ]
5419
5420     ; Implement a jump table:
5421     switch i32 %val, label %otherwise [ i32 0, label %onzero
5422                                         i32 1, label %onone
5423                                         i32 2, label %ontwo ]
5424
5425.. _i_indirectbr:
5426
5427'``indirectbr``' Instruction
5428^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5429
5430Syntax:
5431"""""""
5432
5433::
5434
5435      indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
5436
5437Overview:
5438"""""""""
5439
5440The '``indirectbr``' instruction implements an indirect branch to a
5441label within the current function, whose address is specified by
5442"``address``". Address must be derived from a
5443:ref:`blockaddress <blockaddress>` constant.
5444
5445Arguments:
5446""""""""""
5447
5448The '``address``' argument is the address of the label to jump to. The
5449rest of the arguments indicate the full set of possible destinations
5450that the address may point to. Blocks are allowed to occur multiple
5451times in the destination list, though this isn't particularly useful.
5452
5453This destination list is required so that dataflow analysis has an
5454accurate understanding of the CFG.
5455
5456Semantics:
5457""""""""""
5458
5459Control transfers to the block specified in the address argument. All
5460possible destination blocks must be listed in the label list, otherwise
5461this instruction has undefined behavior. This implies that jumps to
5462labels defined in other functions have undefined behavior as well.
5463
5464Implementation:
5465"""""""""""""""
5466
5467This is typically implemented with a jump through a register.
5468
5469Example:
5470""""""""
5471
5472.. code-block:: llvm
5473
5474     indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
5475
5476.. _i_invoke:
5477
5478'``invoke``' Instruction
5479^^^^^^^^^^^^^^^^^^^^^^^^
5480
5481Syntax:
5482"""""""
5483
5484::
5485
5486      <result> = invoke [cconv] [ret attrs] <ty>|<fnty> <fnptrval>(<function args>) [fn attrs]
5487                    [operand bundles] to label <normal label> unwind label <exception label>
5488
5489Overview:
5490"""""""""
5491
5492The '``invoke``' instruction causes control to transfer to a specified
5493function, with the possibility of control flow transfer to either the
5494'``normal``' label or the '``exception``' label. If the callee function
5495returns with the "``ret``" instruction, control flow will return to the
5496"normal" label. If the callee (or any indirect callees) returns via the
5497":ref:`resume <i_resume>`" instruction or other exception handling
5498mechanism, control is interrupted and continued at the dynamically
5499nearest "exception" label.
5500
5501The '``exception``' label is a `landing
5502pad <ExceptionHandling.html#overview>`_ for the exception. As such,
5503'``exception``' label is required to have the
5504":ref:`landingpad <i_landingpad>`" instruction, which contains the
5505information about the behavior of the program after unwinding happens,
5506as its first non-PHI instruction. The restrictions on the
5507"``landingpad``" instruction's tightly couples it to the "``invoke``"
5508instruction, so that the important information contained within the
5509"``landingpad``" instruction can't be lost through normal code motion.
5510
5511Arguments:
5512""""""""""
5513
5514This instruction requires several arguments:
5515
5516#. The optional "cconv" marker indicates which :ref:`calling
5517   convention <callingconv>` the call should use. If none is
5518   specified, the call defaults to using C calling conventions.
5519#. The optional :ref:`Parameter Attributes <paramattrs>` list for return
5520   values. Only '``zeroext``', '``signext``', and '``inreg``' attributes
5521   are valid here.
5522#. '``ty``': the type of the call instruction itself which is also the
5523   type of the return value. Functions that return no value are marked
5524   ``void``.
5525#. '``fnty``': shall be the signature of the function being invoked. The
5526   argument types must match the types implied by this signature. This
5527   type can be omitted if the function is not varargs.
5528#. '``fnptrval``': An LLVM value containing a pointer to a function to
5529   be invoked. In most cases, this is a direct function invocation, but
5530   indirect ``invoke``'s are just as possible, calling an arbitrary pointer
5531   to function value.
5532#. '``function args``': argument list whose types match the function
5533   signature argument types and parameter attributes. All arguments must
5534   be of :ref:`first class <t_firstclass>` type. If the function signature
5535   indicates the function accepts a variable number of arguments, the
5536   extra arguments can be specified.
5537#. '``normal label``': the label reached when the called function
5538   executes a '``ret``' instruction.
5539#. '``exception label``': the label reached when a callee returns via
5540   the :ref:`resume <i_resume>` instruction or other exception handling
5541   mechanism.
5542#. The optional :ref:`function attributes <fnattrs>` list. Only
5543   '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``'
5544   attributes are valid here.
5545#. The optional :ref:`operand bundles <opbundles>` list.
5546
5547Semantics:
5548""""""""""
5549
5550This instruction is designed to operate as a standard '``call``'
5551instruction in most regards. The primary difference is that it
5552establishes an association with a label, which is used by the runtime
5553library to unwind the stack.
5554
5555This instruction is used in languages with destructors to ensure that
5556proper cleanup is performed in the case of either a ``longjmp`` or a
5557thrown exception. Additionally, this is important for implementation of
5558'``catch``' clauses in high-level languages that support them.
5559
5560For the purposes of the SSA form, the definition of the value returned
5561by the '``invoke``' instruction is deemed to occur on the edge from the
5562current block to the "normal" label. If the callee unwinds then no
5563return value is available.
5564
5565Example:
5566""""""""
5567
5568.. code-block:: llvm
5569
5570      %retval = invoke i32 @Test(i32 15) to label %Continue
5571                  unwind label %TestCleanup              ; i32:retval set
5572      %retval = invoke coldcc i32 %Testfnptr(i32 15) to label %Continue
5573                  unwind label %TestCleanup              ; i32:retval set
5574
5575.. _i_resume:
5576
5577'``resume``' Instruction
5578^^^^^^^^^^^^^^^^^^^^^^^^
5579
5580Syntax:
5581"""""""
5582
5583::
5584
5585      resume <type> <value>
5586
5587Overview:
5588"""""""""
5589
5590The '``resume``' instruction is a terminator instruction that has no
5591successors.
5592
5593Arguments:
5594""""""""""
5595
5596The '``resume``' instruction requires one argument, which must have the
5597same type as the result of any '``landingpad``' instruction in the same
5598function.
5599
5600Semantics:
5601""""""""""
5602
5603The '``resume``' instruction resumes propagation of an existing
5604(in-flight) exception whose unwinding was interrupted with a
5605:ref:`landingpad <i_landingpad>` instruction.
5606
5607Example:
5608""""""""
5609
5610.. code-block:: llvm
5611
5612      resume { i8*, i32 } %exn
5613
5614.. _i_catchswitch:
5615
5616'``catchswitch``' Instruction
5617^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5618
5619Syntax:
5620"""""""
5621
5622::
5623
5624      <resultval> = catchswitch within <parent> [ label <handler1>, label <handler2>, ... ] unwind to caller
5625      <resultval> = catchswitch within <parent> [ label <handler1>, label <handler2>, ... ] unwind label <default>
5626
5627Overview:
5628"""""""""
5629
5630The '``catchswitch``' instruction is used by `LLVM's exception handling system
5631<ExceptionHandling.html#overview>`_ to describe the set of possible catch handlers
5632that may be executed by the :ref:`EH personality routine <personalityfn>`.
5633
5634Arguments:
5635""""""""""
5636
5637The ``parent`` argument is the token of the funclet that contains the
5638``catchswitch`` instruction. If the ``catchswitch`` is not inside a funclet,
5639this operand may be the token ``none``.
5640
5641The ``default`` argument is the label of another basic block beginning with
5642either a ``cleanuppad`` or ``catchswitch`` instruction.  This unwind destination
5643must be a legal target with respect to the ``parent`` links, as described in
5644the `exception handling documentation\ <ExceptionHandling.html#wineh-constraints>`_.
5645
5646The ``handlers`` are a nonempty list of successor blocks that each begin with a
5647:ref:`catchpad <i_catchpad>` instruction.
5648
5649Semantics:
5650""""""""""
5651
5652Executing this instruction transfers control to one of the successors in
5653``handlers``, if appropriate, or continues to unwind via the unwind label if
5654present.
5655
5656The ``catchswitch`` is both a terminator and a "pad" instruction, meaning that
5657it must be both the first non-phi instruction and last instruction in the basic
5658block. Therefore, it must be the only non-phi instruction in the block.
5659
5660Example:
5661""""""""
5662
5663.. code-block:: llvm
5664
5665    dispatch1:
5666      %cs1 = catchswitch within none [label %handler0, label %handler1] unwind to caller
5667    dispatch2:
5668      %cs2 = catchswitch within %parenthandler [label %handler0] unwind label %cleanup
5669
5670.. _i_catchret:
5671
5672'``catchret``' Instruction
5673^^^^^^^^^^^^^^^^^^^^^^^^^^
5674
5675Syntax:
5676"""""""
5677
5678::
5679
5680      catchret from <token> to label <normal>
5681
5682Overview:
5683"""""""""
5684
5685The '``catchret``' instruction is a terminator instruction that has a
5686single successor.
5687
5688
5689Arguments:
5690""""""""""
5691
5692The first argument to a '``catchret``' indicates which ``catchpad`` it
5693exits.  It must be a :ref:`catchpad <i_catchpad>`.
5694The second argument to a '``catchret``' specifies where control will
5695transfer to next.
5696
5697Semantics:
5698""""""""""
5699
5700The '``catchret``' instruction ends an existing (in-flight) exception whose
5701unwinding was interrupted with a :ref:`catchpad <i_catchpad>` instruction.  The
5702:ref:`personality function <personalityfn>` gets a chance to execute arbitrary
5703code to, for example, destroy the active exception.  Control then transfers to
5704``normal``.
5705
5706The ``token`` argument must be a token produced by a ``catchpad`` instruction.
5707If the specified ``catchpad`` is not the most-recently-entered not-yet-exited
5708funclet pad (as described in the `EH documentation\ <ExceptionHandling.html#wineh-constraints>`_),
5709the ``catchret``'s behavior is undefined.
5710
5711Example:
5712""""""""
5713
5714.. code-block:: llvm
5715
5716      catchret from %catch label %continue
5717
5718.. _i_cleanupret:
5719
5720'``cleanupret``' Instruction
5721^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5722
5723Syntax:
5724"""""""
5725
5726::
5727
5728      cleanupret from <value> unwind label <continue>
5729      cleanupret from <value> unwind to caller
5730
5731Overview:
5732"""""""""
5733
5734The '``cleanupret``' instruction is a terminator instruction that has
5735an optional successor.
5736
5737
5738Arguments:
5739""""""""""
5740
5741The '``cleanupret``' instruction requires one argument, which indicates
5742which ``cleanuppad`` it exits, and must be a :ref:`cleanuppad <i_cleanuppad>`.
5743If the specified ``cleanuppad`` is not the most-recently-entered not-yet-exited
5744funclet pad (as described in the `EH documentation\ <ExceptionHandling.html#wineh-constraints>`_),
5745the ``cleanupret``'s behavior is undefined.
5746
5747The '``cleanupret``' instruction also has an optional successor, ``continue``,
5748which must be the label of another basic block beginning with either a
5749``cleanuppad`` or ``catchswitch`` instruction.  This unwind destination must
5750be a legal target with respect to the ``parent`` links, as described in the
5751`exception handling documentation\ <ExceptionHandling.html#wineh-constraints>`_.
5752
5753Semantics:
5754""""""""""
5755
5756The '``cleanupret``' instruction indicates to the
5757:ref:`personality function <personalityfn>` that one
5758:ref:`cleanuppad <i_cleanuppad>` it transferred control to has ended.
5759It transfers control to ``continue`` or unwinds out of the function.
5760
5761Example:
5762""""""""
5763
5764.. code-block:: llvm
5765
5766      cleanupret from %cleanup unwind to caller
5767      cleanupret from %cleanup unwind label %continue
5768
5769.. _i_unreachable:
5770
5771'``unreachable``' Instruction
5772^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5773
5774Syntax:
5775"""""""
5776
5777::
5778
5779      unreachable
5780
5781Overview:
5782"""""""""
5783
5784The '``unreachable``' instruction has no defined semantics. This
5785instruction is used to inform the optimizer that a particular portion of
5786the code is not reachable. This can be used to indicate that the code
5787after a no-return function cannot be reached, and other facts.
5788
5789Semantics:
5790""""""""""
5791
5792The '``unreachable``' instruction has no defined semantics.
5793
5794.. _binaryops:
5795
5796Binary Operations
5797-----------------
5798
5799Binary operators are used to do most of the computation in a program.
5800They require two operands of the same type, execute an operation on
5801them, and produce a single value. The operands might represent multiple
5802data, as is the case with the :ref:`vector <t_vector>` data type. The
5803result value has the same type as its operands.
5804
5805There are several different binary operators:
5806
5807.. _i_add:
5808
5809'``add``' Instruction
5810^^^^^^^^^^^^^^^^^^^^^
5811
5812Syntax:
5813"""""""
5814
5815::
5816
5817      <result> = add <ty> <op1>, <op2>          ; yields ty:result
5818      <result> = add nuw <ty> <op1>, <op2>      ; yields ty:result
5819      <result> = add nsw <ty> <op1>, <op2>      ; yields ty:result
5820      <result> = add nuw nsw <ty> <op1>, <op2>  ; yields ty:result
5821
5822Overview:
5823"""""""""
5824
5825The '``add``' instruction returns the sum of its two operands.
5826
5827Arguments:
5828""""""""""
5829
5830The two arguments to the '``add``' instruction must be
5831:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5832arguments must have identical types.
5833
5834Semantics:
5835""""""""""
5836
5837The value produced is the integer sum of the two operands.
5838
5839If the sum has unsigned overflow, the result returned is the
5840mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of
5841the result.
5842
5843Because LLVM integers use a two's complement representation, this
5844instruction is appropriate for both signed and unsigned integers.
5845
5846``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
5847respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
5848result value of the ``add`` is a :ref:`poison value <poisonvalues>` if
5849unsigned and/or signed overflow, respectively, occurs.
5850
5851Example:
5852""""""""
5853
5854.. code-block:: llvm
5855
5856      <result> = add i32 4, %var          ; yields i32:result = 4 + %var
5857
5858.. _i_fadd:
5859
5860'``fadd``' Instruction
5861^^^^^^^^^^^^^^^^^^^^^^
5862
5863Syntax:
5864"""""""
5865
5866::
5867
5868      <result> = fadd [fast-math flags]* <ty> <op1>, <op2>   ; yields ty:result
5869
5870Overview:
5871"""""""""
5872
5873The '``fadd``' instruction returns the sum of its two operands.
5874
5875Arguments:
5876""""""""""
5877
5878The two arguments to the '``fadd``' instruction must be :ref:`floating
5879point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
5880Both arguments must have identical types.
5881
5882Semantics:
5883""""""""""
5884
5885The value produced is the floating point sum of the two operands. This
5886instruction can also take any number of :ref:`fast-math flags <fastmath>`,
5887which are optimization hints to enable otherwise unsafe floating point
5888optimizations:
5889
5890Example:
5891""""""""
5892
5893.. code-block:: llvm
5894
5895      <result> = fadd float 4.0, %var          ; yields float:result = 4.0 + %var
5896
5897'``sub``' Instruction
5898^^^^^^^^^^^^^^^^^^^^^
5899
5900Syntax:
5901"""""""
5902
5903::
5904
5905      <result> = sub <ty> <op1>, <op2>          ; yields ty:result
5906      <result> = sub nuw <ty> <op1>, <op2>      ; yields ty:result
5907      <result> = sub nsw <ty> <op1>, <op2>      ; yields ty:result
5908      <result> = sub nuw nsw <ty> <op1>, <op2>  ; yields ty:result
5909
5910Overview:
5911"""""""""
5912
5913The '``sub``' instruction returns the difference of its two operands.
5914
5915Note that the '``sub``' instruction is used to represent the '``neg``'
5916instruction present in most other intermediate representations.
5917
5918Arguments:
5919""""""""""
5920
5921The two arguments to the '``sub``' instruction must be
5922:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5923arguments must have identical types.
5924
5925Semantics:
5926""""""""""
5927
5928The value produced is the integer difference of the two operands.
5929
5930If the difference has unsigned overflow, the result returned is the
5931mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of
5932the result.
5933
5934Because LLVM integers use a two's complement representation, this
5935instruction is appropriate for both signed and unsigned integers.
5936
5937``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
5938respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
5939result value of the ``sub`` is a :ref:`poison value <poisonvalues>` if
5940unsigned and/or signed overflow, respectively, occurs.
5941
5942Example:
5943""""""""
5944
5945.. code-block:: llvm
5946
5947      <result> = sub i32 4, %var          ; yields i32:result = 4 - %var
5948      <result> = sub i32 0, %val          ; yields i32:result = -%var
5949
5950.. _i_fsub:
5951
5952'``fsub``' Instruction
5953^^^^^^^^^^^^^^^^^^^^^^
5954
5955Syntax:
5956"""""""
5957
5958::
5959
5960      <result> = fsub [fast-math flags]* <ty> <op1>, <op2>   ; yields ty:result
5961
5962Overview:
5963"""""""""
5964
5965The '``fsub``' instruction returns the difference of its two operands.
5966
5967Note that the '``fsub``' instruction is used to represent the '``fneg``'
5968instruction present in most other intermediate representations.
5969
5970Arguments:
5971""""""""""
5972
5973The two arguments to the '``fsub``' instruction must be :ref:`floating
5974point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
5975Both arguments must have identical types.
5976
5977Semantics:
5978""""""""""
5979
5980The value produced is the floating point difference of the two operands.
5981This instruction can also take any number of :ref:`fast-math
5982flags <fastmath>`, which are optimization hints to enable otherwise
5983unsafe floating point optimizations:
5984
5985Example:
5986""""""""
5987
5988.. code-block:: llvm
5989
5990      <result> = fsub float 4.0, %var           ; yields float:result = 4.0 - %var
5991      <result> = fsub float -0.0, %val          ; yields float:result = -%var
5992
5993'``mul``' Instruction
5994^^^^^^^^^^^^^^^^^^^^^
5995
5996Syntax:
5997"""""""
5998
5999::
6000
6001      <result> = mul <ty> <op1>, <op2>          ; yields ty:result
6002      <result> = mul nuw <ty> <op1>, <op2>      ; yields ty:result
6003      <result> = mul nsw <ty> <op1>, <op2>      ; yields ty:result
6004      <result> = mul nuw nsw <ty> <op1>, <op2>  ; yields ty:result
6005
6006Overview:
6007"""""""""
6008
6009The '``mul``' instruction returns the product of its two operands.
6010
6011Arguments:
6012""""""""""
6013
6014The two arguments to the '``mul``' instruction must be
6015:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6016arguments must have identical types.
6017
6018Semantics:
6019""""""""""
6020
6021The value produced is the integer product of the two operands.
6022
6023If the result of the multiplication has unsigned overflow, the result
6024returned is the mathematical result modulo 2\ :sup:`n`\ , where n is the
6025bit width of the result.
6026
6027Because LLVM integers use a two's complement representation, and the
6028result is the same width as the operands, this instruction returns the
6029correct result for both signed and unsigned integers. If a full product
6030(e.g. ``i32`` * ``i32`` -> ``i64``) is needed, the operands should be
6031sign-extended or zero-extended as appropriate to the width of the full
6032product.
6033
6034``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
6035respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
6036result value of the ``mul`` is a :ref:`poison value <poisonvalues>` if
6037unsigned and/or signed overflow, respectively, occurs.
6038
6039Example:
6040""""""""
6041
6042.. code-block:: llvm
6043
6044      <result> = mul i32 4, %var          ; yields i32:result = 4 * %var
6045
6046.. _i_fmul:
6047
6048'``fmul``' Instruction
6049^^^^^^^^^^^^^^^^^^^^^^
6050
6051Syntax:
6052"""""""
6053
6054::
6055
6056      <result> = fmul [fast-math flags]* <ty> <op1>, <op2>   ; yields ty:result
6057
6058Overview:
6059"""""""""
6060
6061The '``fmul``' instruction returns the product of its two operands.
6062
6063Arguments:
6064""""""""""
6065
6066The two arguments to the '``fmul``' instruction must be :ref:`floating
6067point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
6068Both arguments must have identical types.
6069
6070Semantics:
6071""""""""""
6072
6073The value produced is the floating point product of the two operands.
6074This instruction can also take any number of :ref:`fast-math
6075flags <fastmath>`, which are optimization hints to enable otherwise
6076unsafe floating point optimizations:
6077
6078Example:
6079""""""""
6080
6081.. code-block:: llvm
6082
6083      <result> = fmul float 4.0, %var          ; yields float:result = 4.0 * %var
6084
6085'``udiv``' Instruction
6086^^^^^^^^^^^^^^^^^^^^^^
6087
6088Syntax:
6089"""""""
6090
6091::
6092
6093      <result> = udiv <ty> <op1>, <op2>         ; yields ty:result
6094      <result> = udiv exact <ty> <op1>, <op2>   ; yields ty:result
6095
6096Overview:
6097"""""""""
6098
6099The '``udiv``' instruction returns the quotient of its two operands.
6100
6101Arguments:
6102""""""""""
6103
6104The two arguments to the '``udiv``' instruction must be
6105:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6106arguments must have identical types.
6107
6108Semantics:
6109""""""""""
6110
6111The value produced is the unsigned integer quotient of the two operands.
6112
6113Note that unsigned integer division and signed integer division are
6114distinct operations; for signed integer division, use '``sdiv``'.
6115
6116Division by zero leads to undefined behavior.
6117
6118If the ``exact`` keyword is present, the result value of the ``udiv`` is
6119a :ref:`poison value <poisonvalues>` if %op1 is not a multiple of %op2 (as
6120such, "((a udiv exact b) mul b) == a").
6121
6122Example:
6123""""""""
6124
6125.. code-block:: llvm
6126
6127      <result> = udiv i32 4, %var          ; yields i32:result = 4 / %var
6128
6129'``sdiv``' Instruction
6130^^^^^^^^^^^^^^^^^^^^^^
6131
6132Syntax:
6133"""""""
6134
6135::
6136
6137      <result> = sdiv <ty> <op1>, <op2>         ; yields ty:result
6138      <result> = sdiv exact <ty> <op1>, <op2>   ; yields ty:result
6139
6140Overview:
6141"""""""""
6142
6143The '``sdiv``' instruction returns the quotient of its two operands.
6144
6145Arguments:
6146""""""""""
6147
6148The two arguments to the '``sdiv``' instruction must be
6149:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6150arguments must have identical types.
6151
6152Semantics:
6153""""""""""
6154
6155The value produced is the signed integer quotient of the two operands
6156rounded towards zero.
6157
6158Note that signed integer division and unsigned integer division are
6159distinct operations; for unsigned integer division, use '``udiv``'.
6160
6161Division by zero leads to undefined behavior. Overflow also leads to
6162undefined behavior; this is a rare case, but can occur, for example, by
6163doing a 32-bit division of -2147483648 by -1.
6164
6165If the ``exact`` keyword is present, the result value of the ``sdiv`` is
6166a :ref:`poison value <poisonvalues>` if the result would be rounded.
6167
6168Example:
6169""""""""
6170
6171.. code-block:: llvm
6172
6173      <result> = sdiv i32 4, %var          ; yields i32:result = 4 / %var
6174
6175.. _i_fdiv:
6176
6177'``fdiv``' Instruction
6178^^^^^^^^^^^^^^^^^^^^^^
6179
6180Syntax:
6181"""""""
6182
6183::
6184
6185      <result> = fdiv [fast-math flags]* <ty> <op1>, <op2>   ; yields ty:result
6186
6187Overview:
6188"""""""""
6189
6190The '``fdiv``' instruction returns the quotient of its two operands.
6191
6192Arguments:
6193""""""""""
6194
6195The two arguments to the '``fdiv``' instruction must be :ref:`floating
6196point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
6197Both arguments must have identical types.
6198
6199Semantics:
6200""""""""""
6201
6202The value produced is the floating point quotient of the two operands.
6203This instruction can also take any number of :ref:`fast-math
6204flags <fastmath>`, which are optimization hints to enable otherwise
6205unsafe floating point optimizations:
6206
6207Example:
6208""""""""
6209
6210.. code-block:: llvm
6211
6212      <result> = fdiv float 4.0, %var          ; yields float:result = 4.0 / %var
6213
6214'``urem``' Instruction
6215^^^^^^^^^^^^^^^^^^^^^^
6216
6217Syntax:
6218"""""""
6219
6220::
6221
6222      <result> = urem <ty> <op1>, <op2>   ; yields ty:result
6223
6224Overview:
6225"""""""""
6226
6227The '``urem``' instruction returns the remainder from the unsigned
6228division of its two arguments.
6229
6230Arguments:
6231""""""""""
6232
6233The two arguments to the '``urem``' instruction must be
6234:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6235arguments must have identical types.
6236
6237Semantics:
6238""""""""""
6239
6240This instruction returns the unsigned integer *remainder* of a division.
6241This instruction always performs an unsigned division to get the
6242remainder.
6243
6244Note that unsigned integer remainder and signed integer remainder are
6245distinct operations; for signed integer remainder, use '``srem``'.
6246
6247Taking the remainder of a division by zero leads to undefined behavior.
6248
6249Example:
6250""""""""
6251
6252.. code-block:: llvm
6253
6254      <result> = urem i32 4, %var          ; yields i32:result = 4 % %var
6255
6256'``srem``' Instruction
6257^^^^^^^^^^^^^^^^^^^^^^
6258
6259Syntax:
6260"""""""
6261
6262::
6263
6264      <result> = srem <ty> <op1>, <op2>   ; yields ty:result
6265
6266Overview:
6267"""""""""
6268
6269The '``srem``' instruction returns the remainder from the signed
6270division of its two operands. This instruction can also take
6271:ref:`vector <t_vector>` versions of the values in which case the elements
6272must be integers.
6273
6274Arguments:
6275""""""""""
6276
6277The two arguments to the '``srem``' instruction must be
6278:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6279arguments must have identical types.
6280
6281Semantics:
6282""""""""""
6283
6284This instruction returns the *remainder* of a division (where the result
6285is either zero or has the same sign as the dividend, ``op1``), not the
6286*modulo* operator (where the result is either zero or has the same sign
6287as the divisor, ``op2``) of a value. For more information about the
6288difference, see `The Math
6289Forum <http://mathforum.org/dr.math/problems/anne.4.28.99.html>`_. For a
6290table of how this is implemented in various languages, please see
6291`Wikipedia: modulo
6292operation <http://en.wikipedia.org/wiki/Modulo_operation>`_.
6293
6294Note that signed integer remainder and unsigned integer remainder are
6295distinct operations; for unsigned integer remainder, use '``urem``'.
6296
6297Taking the remainder of a division by zero leads to undefined behavior.
6298Overflow also leads to undefined behavior; this is a rare case, but can
6299occur, for example, by taking the remainder of a 32-bit division of
6300-2147483648 by -1. (The remainder doesn't actually overflow, but this
6301rule lets srem be implemented using instructions that return both the
6302result of the division and the remainder.)
6303
6304Example:
6305""""""""
6306
6307.. code-block:: llvm
6308
6309      <result> = srem i32 4, %var          ; yields i32:result = 4 % %var
6310
6311.. _i_frem:
6312
6313'``frem``' Instruction
6314^^^^^^^^^^^^^^^^^^^^^^
6315
6316Syntax:
6317"""""""
6318
6319::
6320
6321      <result> = frem [fast-math flags]* <ty> <op1>, <op2>   ; yields ty:result
6322
6323Overview:
6324"""""""""
6325
6326The '``frem``' instruction returns the remainder from the division of
6327its two operands.
6328
6329Arguments:
6330""""""""""
6331
6332The two arguments to the '``frem``' instruction must be :ref:`floating
6333point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
6334Both arguments must have identical types.
6335
6336Semantics:
6337""""""""""
6338
6339This instruction returns the *remainder* of a division. The remainder
6340has the same sign as the dividend. This instruction can also take any
6341number of :ref:`fast-math flags <fastmath>`, which are optimization hints
6342to enable otherwise unsafe floating point optimizations:
6343
6344Example:
6345""""""""
6346
6347.. code-block:: llvm
6348
6349      <result> = frem float 4.0, %var          ; yields float:result = 4.0 % %var
6350
6351.. _bitwiseops:
6352
6353Bitwise Binary Operations
6354-------------------------
6355
6356Bitwise binary operators are used to do various forms of bit-twiddling
6357in a program. They are generally very efficient instructions and can
6358commonly be strength reduced from other instructions. They require two
6359operands of the same type, execute an operation on them, and produce a
6360single value. The resulting value is the same type as its operands.
6361
6362'``shl``' Instruction
6363^^^^^^^^^^^^^^^^^^^^^
6364
6365Syntax:
6366"""""""
6367
6368::
6369
6370      <result> = shl <ty> <op1>, <op2>           ; yields ty:result
6371      <result> = shl nuw <ty> <op1>, <op2>       ; yields ty:result
6372      <result> = shl nsw <ty> <op1>, <op2>       ; yields ty:result
6373      <result> = shl nuw nsw <ty> <op1>, <op2>   ; yields ty:result
6374
6375Overview:
6376"""""""""
6377
6378The '``shl``' instruction returns the first operand shifted to the left
6379a specified number of bits.
6380
6381Arguments:
6382""""""""""
6383
6384Both arguments to the '``shl``' instruction must be the same
6385:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
6386'``op2``' is treated as an unsigned value.
6387
6388Semantics:
6389""""""""""
6390
6391The value produced is ``op1`` \* 2\ :sup:`op2` mod 2\ :sup:`n`,
6392where ``n`` is the width of the result. If ``op2`` is (statically or
6393dynamically) equal to or larger than the number of bits in
6394``op1``, the result is undefined. If the arguments are vectors, each
6395vector element of ``op1`` is shifted by the corresponding shift amount
6396in ``op2``.
6397
6398If the ``nuw`` keyword is present, then the shift produces a :ref:`poison
6399value <poisonvalues>` if it shifts out any non-zero bits. If the
6400``nsw`` keyword is present, then the shift produces a :ref:`poison
6401value <poisonvalues>` if it shifts out any bits that disagree with the
6402resultant sign bit. As such, NUW/NSW have the same semantics as they
6403would if the shift were expressed as a mul instruction with the same
6404nsw/nuw bits in (mul %op1, (shl 1, %op2)).
6405
6406Example:
6407""""""""
6408
6409.. code-block:: llvm
6410
6411      <result> = shl i32 4, %var   ; yields i32: 4 << %var
6412      <result> = shl i32 4, 2      ; yields i32: 16
6413      <result> = shl i32 1, 10     ; yields i32: 1024
6414      <result> = shl i32 1, 32     ; undefined
6415      <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2>   ; yields: result=<2 x i32> < i32 2, i32 4>
6416
6417'``lshr``' Instruction
6418^^^^^^^^^^^^^^^^^^^^^^
6419
6420Syntax:
6421"""""""
6422
6423::
6424
6425      <result> = lshr <ty> <op1>, <op2>         ; yields ty:result
6426      <result> = lshr exact <ty> <op1>, <op2>   ; yields ty:result
6427
6428Overview:
6429"""""""""
6430
6431The '``lshr``' instruction (logical shift right) returns the first
6432operand shifted to the right a specified number of bits with zero fill.
6433
6434Arguments:
6435""""""""""
6436
6437Both arguments to the '``lshr``' instruction must be the same
6438:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
6439'``op2``' is treated as an unsigned value.
6440
6441Semantics:
6442""""""""""
6443
6444This instruction always performs a logical shift right operation. The
6445most significant bits of the result will be filled with zero bits after
6446the shift. If ``op2`` is (statically or dynamically) equal to or larger
6447than the number of bits in ``op1``, the result is undefined. If the
6448arguments are vectors, each vector element of ``op1`` is shifted by the
6449corresponding shift amount in ``op2``.
6450
6451If the ``exact`` keyword is present, the result value of the ``lshr`` is
6452a :ref:`poison value <poisonvalues>` if any of the bits shifted out are
6453non-zero.
6454
6455Example:
6456""""""""
6457
6458.. code-block:: llvm
6459
6460      <result> = lshr i32 4, 1   ; yields i32:result = 2
6461      <result> = lshr i32 4, 2   ; yields i32:result = 1
6462      <result> = lshr i8  4, 3   ; yields i8:result = 0
6463      <result> = lshr i8 -2, 1   ; yields i8:result = 0x7F
6464      <result> = lshr i32 1, 32  ; undefined
6465      <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2>   ; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1>
6466
6467'``ashr``' Instruction
6468^^^^^^^^^^^^^^^^^^^^^^
6469
6470Syntax:
6471"""""""
6472
6473::
6474
6475      <result> = ashr <ty> <op1>, <op2>         ; yields ty:result
6476      <result> = ashr exact <ty> <op1>, <op2>   ; yields ty:result
6477
6478Overview:
6479"""""""""
6480
6481The '``ashr``' instruction (arithmetic shift right) returns the first
6482operand shifted to the right a specified number of bits with sign
6483extension.
6484
6485Arguments:
6486""""""""""
6487
6488Both arguments to the '``ashr``' instruction must be the same
6489:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
6490'``op2``' is treated as an unsigned value.
6491
6492Semantics:
6493""""""""""
6494
6495This instruction always performs an arithmetic shift right operation,
6496The most significant bits of the result will be filled with the sign bit
6497of ``op1``. If ``op2`` is (statically or dynamically) equal to or larger
6498than the number of bits in ``op1``, the result is undefined. If the
6499arguments are vectors, each vector element of ``op1`` is shifted by the
6500corresponding shift amount in ``op2``.
6501
6502If the ``exact`` keyword is present, the result value of the ``ashr`` is
6503a :ref:`poison value <poisonvalues>` if any of the bits shifted out are
6504non-zero.
6505
6506Example:
6507""""""""
6508
6509.. code-block:: llvm
6510
6511      <result> = ashr i32 4, 1   ; yields i32:result = 2
6512      <result> = ashr i32 4, 2   ; yields i32:result = 1
6513      <result> = ashr i8  4, 3   ; yields i8:result = 0
6514      <result> = ashr i8 -2, 1   ; yields i8:result = -1
6515      <result> = ashr i32 1, 32  ; undefined
6516      <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3>   ; yields: result=<2 x i32> < i32 -1, i32 0>
6517
6518'``and``' Instruction
6519^^^^^^^^^^^^^^^^^^^^^
6520
6521Syntax:
6522"""""""
6523
6524::
6525
6526      <result> = and <ty> <op1>, <op2>   ; yields ty:result
6527
6528Overview:
6529"""""""""
6530
6531The '``and``' instruction returns the bitwise logical and of its two
6532operands.
6533
6534Arguments:
6535""""""""""
6536
6537The two arguments to the '``and``' instruction must be
6538:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6539arguments must have identical types.
6540
6541Semantics:
6542""""""""""
6543
6544The truth table used for the '``and``' instruction is:
6545
6546+-----+-----+-----+
6547| In0 | In1 | Out |
6548+-----+-----+-----+
6549|   0 |   0 |   0 |
6550+-----+-----+-----+
6551|   0 |   1 |   0 |
6552+-----+-----+-----+
6553|   1 |   0 |   0 |
6554+-----+-----+-----+
6555|   1 |   1 |   1 |
6556+-----+-----+-----+
6557
6558Example:
6559""""""""
6560
6561.. code-block:: llvm
6562
6563      <result> = and i32 4, %var         ; yields i32:result = 4 & %var
6564      <result> = and i32 15, 40          ; yields i32:result = 8
6565      <result> = and i32 4, 8            ; yields i32:result = 0
6566
6567'``or``' Instruction
6568^^^^^^^^^^^^^^^^^^^^
6569
6570Syntax:
6571"""""""
6572
6573::
6574
6575      <result> = or <ty> <op1>, <op2>   ; yields ty:result
6576
6577Overview:
6578"""""""""
6579
6580The '``or``' instruction returns the bitwise logical inclusive or of its
6581two operands.
6582
6583Arguments:
6584""""""""""
6585
6586The two arguments to the '``or``' instruction must be
6587:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6588arguments must have identical types.
6589
6590Semantics:
6591""""""""""
6592
6593The truth table used for the '``or``' instruction is:
6594
6595+-----+-----+-----+
6596| In0 | In1 | Out |
6597+-----+-----+-----+
6598|   0 |   0 |   0 |
6599+-----+-----+-----+
6600|   0 |   1 |   1 |
6601+-----+-----+-----+
6602|   1 |   0 |   1 |
6603+-----+-----+-----+
6604|   1 |   1 |   1 |
6605+-----+-----+-----+
6606
6607Example:
6608""""""""
6609
6610::
6611
6612      <result> = or i32 4, %var         ; yields i32:result = 4 | %var
6613      <result> = or i32 15, 40          ; yields i32:result = 47
6614      <result> = or i32 4, 8            ; yields i32:result = 12
6615
6616'``xor``' Instruction
6617^^^^^^^^^^^^^^^^^^^^^
6618
6619Syntax:
6620"""""""
6621
6622::
6623
6624      <result> = xor <ty> <op1>, <op2>   ; yields ty:result
6625
6626Overview:
6627"""""""""
6628
6629The '``xor``' instruction returns the bitwise logical exclusive or of
6630its two operands. The ``xor`` is used to implement the "one's
6631complement" operation, which is the "~" operator in C.
6632
6633Arguments:
6634""""""""""
6635
6636The two arguments to the '``xor``' instruction must be
6637:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6638arguments must have identical types.
6639
6640Semantics:
6641""""""""""
6642
6643The truth table used for the '``xor``' instruction is:
6644
6645+-----+-----+-----+
6646| In0 | In1 | Out |
6647+-----+-----+-----+
6648|   0 |   0 |   0 |
6649+-----+-----+-----+
6650|   0 |   1 |   1 |
6651+-----+-----+-----+
6652|   1 |   0 |   1 |
6653+-----+-----+-----+
6654|   1 |   1 |   0 |
6655+-----+-----+-----+
6656
6657Example:
6658""""""""
6659
6660.. code-block:: llvm
6661
6662      <result> = xor i32 4, %var         ; yields i32:result = 4 ^ %var
6663      <result> = xor i32 15, 40          ; yields i32:result = 39
6664      <result> = xor i32 4, 8            ; yields i32:result = 12
6665      <result> = xor i32 %V, -1          ; yields i32:result = ~%V
6666
6667Vector Operations
6668-----------------
6669
6670LLVM supports several instructions to represent vector operations in a
6671target-independent manner. These instructions cover the element-access
6672and vector-specific operations needed to process vectors effectively.
6673While LLVM does directly support these vector operations, many
6674sophisticated algorithms will want to use target-specific intrinsics to
6675take full advantage of a specific target.
6676
6677.. _i_extractelement:
6678
6679'``extractelement``' Instruction
6680^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6681
6682Syntax:
6683"""""""
6684
6685::
6686
6687      <result> = extractelement <n x <ty>> <val>, <ty2> <idx>  ; yields <ty>
6688
6689Overview:
6690"""""""""
6691
6692The '``extractelement``' instruction extracts a single scalar element
6693from a vector at a specified index.
6694
6695Arguments:
6696""""""""""
6697
6698The first operand of an '``extractelement``' instruction is a value of
6699:ref:`vector <t_vector>` type. The second operand is an index indicating
6700the position from which to extract the element. The index may be a
6701variable of any integer type.
6702
6703Semantics:
6704""""""""""
6705
6706The result is a scalar of the same type as the element type of ``val``.
6707Its value is the value at position ``idx`` of ``val``. If ``idx``
6708exceeds the length of ``val``, the results are undefined.
6709
6710Example:
6711""""""""
6712
6713.. code-block:: llvm
6714
6715      <result> = extractelement <4 x i32> %vec, i32 0    ; yields i32
6716
6717.. _i_insertelement:
6718
6719'``insertelement``' Instruction
6720^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6721
6722Syntax:
6723"""""""
6724
6725::
6726
6727      <result> = insertelement <n x <ty>> <val>, <ty> <elt>, <ty2> <idx>    ; yields <n x <ty>>
6728
6729Overview:
6730"""""""""
6731
6732The '``insertelement``' instruction inserts a scalar element into a
6733vector at a specified index.
6734
6735Arguments:
6736""""""""""
6737
6738The first operand of an '``insertelement``' instruction is a value of
6739:ref:`vector <t_vector>` type. The second operand is a scalar value whose
6740type must equal the element type of the first operand. The third operand
6741is an index indicating the position at which to insert the value. The
6742index may be a variable of any integer type.
6743
6744Semantics:
6745""""""""""
6746
6747The result is a vector of the same type as ``val``. Its element values
6748are those of ``val`` except at position ``idx``, where it gets the value
6749``elt``. If ``idx`` exceeds the length of ``val``, the results are
6750undefined.
6751
6752Example:
6753""""""""
6754
6755.. code-block:: llvm
6756
6757      <result> = insertelement <4 x i32> %vec, i32 1, i32 0    ; yields <4 x i32>
6758
6759.. _i_shufflevector:
6760
6761'``shufflevector``' Instruction
6762^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6763
6764Syntax:
6765"""""""
6766
6767::
6768
6769      <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask>    ; yields <m x <ty>>
6770
6771Overview:
6772"""""""""
6773
6774The '``shufflevector``' instruction constructs a permutation of elements
6775from two input vectors, returning a vector with the same element type as
6776the input and length that is the same as the shuffle mask.
6777
6778Arguments:
6779""""""""""
6780
6781The first two operands of a '``shufflevector``' instruction are vectors
6782with the same type. The third argument is a shuffle mask whose element
6783type is always 'i32'. The result of the instruction is a vector whose
6784length is the same as the shuffle mask and whose element type is the
6785same as the element type of the first two operands.
6786
6787The shuffle mask operand is required to be a constant vector with either
6788constant integer or undef values.
6789
6790Semantics:
6791""""""""""
6792
6793The elements of the two input vectors are numbered from left to right
6794across both of the vectors. The shuffle mask operand specifies, for each
6795element of the result vector, which element of the two input vectors the
6796result element gets. The element selector may be undef (meaning "don't
6797care") and the second operand may be undef if performing a shuffle from
6798only one vector.
6799
6800Example:
6801""""""""
6802
6803.. code-block:: llvm
6804
6805      <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
6806                              <4 x i32> <i32 0, i32 4, i32 1, i32 5>  ; yields <4 x i32>
6807      <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
6808                              <4 x i32> <i32 0, i32 1, i32 2, i32 3>  ; yields <4 x i32> - Identity shuffle.
6809      <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
6810                              <4 x i32> <i32 0, i32 1, i32 2, i32 3>  ; yields <4 x i32>
6811      <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
6812                              <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 >  ; yields <8 x i32>
6813
6814Aggregate Operations
6815--------------------
6816
6817LLVM supports several instructions for working with
6818:ref:`aggregate <t_aggregate>` values.
6819
6820.. _i_extractvalue:
6821
6822'``extractvalue``' Instruction
6823^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6824
6825Syntax:
6826"""""""
6827
6828::
6829
6830      <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
6831
6832Overview:
6833"""""""""
6834
6835The '``extractvalue``' instruction extracts the value of a member field
6836from an :ref:`aggregate <t_aggregate>` value.
6837
6838Arguments:
6839""""""""""
6840
6841The first operand of an '``extractvalue``' instruction is a value of
6842:ref:`struct <t_struct>` or :ref:`array <t_array>` type. The other operands are
6843constant indices to specify which value to extract in a similar manner
6844as indices in a '``getelementptr``' instruction.
6845
6846The major differences to ``getelementptr`` indexing are:
6847
6848-  Since the value being indexed is not a pointer, the first index is
6849   omitted and assumed to be zero.
6850-  At least one index must be specified.
6851-  Not only struct indices but also array indices must be in bounds.
6852
6853Semantics:
6854""""""""""
6855
6856The result is the value at the position in the aggregate specified by
6857the index operands.
6858
6859Example:
6860""""""""
6861
6862.. code-block:: llvm
6863
6864      <result> = extractvalue {i32, float} %agg, 0    ; yields i32
6865
6866.. _i_insertvalue:
6867
6868'``insertvalue``' Instruction
6869^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6870
6871Syntax:
6872"""""""
6873
6874::
6875
6876      <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}*    ; yields <aggregate type>
6877
6878Overview:
6879"""""""""
6880
6881The '``insertvalue``' instruction inserts a value into a member field in
6882an :ref:`aggregate <t_aggregate>` value.
6883
6884Arguments:
6885""""""""""
6886
6887The first operand of an '``insertvalue``' instruction is a value of
6888:ref:`struct <t_struct>` or :ref:`array <t_array>` type. The second operand is
6889a first-class value to insert. The following operands are constant
6890indices indicating the position at which to insert the value in a
6891similar manner as indices in a '``extractvalue``' instruction. The value
6892to insert must have the same type as the value identified by the
6893indices.
6894
6895Semantics:
6896""""""""""
6897
6898The result is an aggregate of the same type as ``val``. Its value is
6899that of ``val`` except that the value at the position specified by the
6900indices is that of ``elt``.
6901
6902Example:
6903""""""""
6904
6905.. code-block:: llvm
6906
6907      %agg1 = insertvalue {i32, float} undef, i32 1, 0              ; yields {i32 1, float undef}
6908      %agg2 = insertvalue {i32, float} %agg1, float %val, 1         ; yields {i32 1, float %val}
6909      %agg3 = insertvalue {i32, {float}} undef, float %val, 1, 0    ; yields {i32 undef, {float %val}}
6910
6911.. _memoryops:
6912
6913Memory Access and Addressing Operations
6914---------------------------------------
6915
6916A key design point of an SSA-based representation is how it represents
6917memory. In LLVM, no memory locations are in SSA form, which makes things
6918very simple. This section describes how to read, write, and allocate
6919memory in LLVM.
6920
6921.. _i_alloca:
6922
6923'``alloca``' Instruction
6924^^^^^^^^^^^^^^^^^^^^^^^^
6925
6926Syntax:
6927"""""""
6928
6929::
6930
6931      <result> = alloca [inalloca] <type> [, <ty> <NumElements>] [, align <alignment>]     ; yields type*:result
6932
6933Overview:
6934"""""""""
6935
6936The '``alloca``' instruction allocates memory on the stack frame of the
6937currently executing function, to be automatically released when this
6938function returns to its caller. The object is always allocated in the
6939generic address space (address space zero).
6940
6941Arguments:
6942""""""""""
6943
6944The '``alloca``' instruction allocates ``sizeof(<type>)*NumElements``
6945bytes of memory on the runtime stack, returning a pointer of the
6946appropriate type to the program. If "NumElements" is specified, it is
6947the number of elements allocated, otherwise "NumElements" is defaulted
6948to be one. If a constant alignment is specified, the value result of the
6949allocation is guaranteed to be aligned to at least that boundary. The
6950alignment may not be greater than ``1 << 29``. If not specified, or if
6951zero, the target can choose to align the allocation on any convenient
6952boundary compatible with the type.
6953
6954'``type``' may be any sized type.
6955
6956Semantics:
6957""""""""""
6958
6959Memory is allocated; a pointer is returned. The operation is undefined
6960if there is insufficient stack space for the allocation. '``alloca``'d
6961memory is automatically released when the function returns. The
6962'``alloca``' instruction is commonly used to represent automatic
6963variables that must have an address available. When the function returns
6964(either with the ``ret`` or ``resume`` instructions), the memory is
6965reclaimed. Allocating zero bytes is legal, but the result is undefined.
6966The order in which memory is allocated (ie., which way the stack grows)
6967is not specified.
6968
6969Example:
6970""""""""
6971
6972.. code-block:: llvm
6973
6974      %ptr = alloca i32                             ; yields i32*:ptr
6975      %ptr = alloca i32, i32 4                      ; yields i32*:ptr
6976      %ptr = alloca i32, i32 4, align 1024          ; yields i32*:ptr
6977      %ptr = alloca i32, align 1024                 ; yields i32*:ptr
6978
6979.. _i_load:
6980
6981'``load``' Instruction
6982^^^^^^^^^^^^^^^^^^^^^^
6983
6984Syntax:
6985"""""""
6986
6987::
6988
6989      <result> = load [volatile] <ty>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>][, !invariant.load !<index>][, !invariant.group !<index>][, !nonnull !<index>][, !dereferenceable !<deref_bytes_node>][, !dereferenceable_or_null !<deref_bytes_node>][, !align !<align_node>]
6990      <result> = load atomic [volatile] <ty>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> [, !invariant.group !<index>]
6991      !<index> = !{ i32 1 }
6992      !<deref_bytes_node> = !{i64 <dereferenceable_bytes>}
6993      !<align_node> = !{ i64 <value_alignment> }
6994
6995Overview:
6996"""""""""
6997
6998The '``load``' instruction is used to read from memory.
6999
7000Arguments:
7001""""""""""
7002
7003The argument to the ``load`` instruction specifies the memory address from which
7004to load. The type specified must be a :ref:`first class <t_firstclass>` type of
7005known size (i.e. not containing an :ref:`opaque structural type <t_opaque>`). If
7006the ``load`` is marked as ``volatile``, then the optimizer is not allowed to
7007modify the number or order of execution of this ``load`` with other
7008:ref:`volatile operations <volatile>`.
7009
7010If the ``load`` is marked as ``atomic``, it takes an extra :ref:`ordering
7011<ordering>` and optional ``singlethread`` argument. The ``release`` and
7012``acq_rel`` orderings are not valid on ``load`` instructions. Atomic loads
7013produce :ref:`defined <memmodel>` results when they may see multiple atomic
7014stores. The type of the pointee must be an integer, pointer, or floating-point
7015type whose bit width is a power of two greater than or equal to eight and less
7016than or equal to a target-specific size limit.  ``align`` must be explicitly
7017specified on atomic loads, and the load has undefined behavior if the alignment
7018is not set to a value which is at least the size in bytes of the
7019pointee. ``!nontemporal`` does not have any defined semantics for atomic loads.
7020
7021The optional constant ``align`` argument specifies the alignment of the
7022operation (that is, the alignment of the memory address). A value of 0
7023or an omitted ``align`` argument means that the operation has the ABI
7024alignment for the target. It is the responsibility of the code emitter
7025to ensure that the alignment information is correct. Overestimating the
7026alignment results in undefined behavior. Underestimating the alignment
7027may produce less efficient code. An alignment of 1 is always safe. The
7028maximum possible alignment is ``1 << 29``. An alignment value higher
7029than the size of the loaded type implies memory up to the alignment
7030value bytes can be safely loaded without trapping in the default
7031address space. Access of the high bytes can interfere with debugging
7032tools, so should not be accessed if the function has the
7033``sanitize_thread`` or ``sanitize_address`` attributes.
7034
7035The optional ``!nontemporal`` metadata must reference a single
7036metadata name ``<index>`` corresponding to a metadata node with one
7037``i32`` entry of value 1. The existence of the ``!nontemporal``
7038metadata on the instruction tells the optimizer and code generator
7039that this load is not expected to be reused in the cache. The code
7040generator may select special instructions to save cache bandwidth, such
7041as the ``MOVNT`` instruction on x86.
7042
7043The optional ``!invariant.load`` metadata must reference a single
7044metadata name ``<index>`` corresponding to a metadata node with no
7045entries. The existence of the ``!invariant.load`` metadata on the
7046instruction tells the optimizer and code generator that the address
7047operand to this load points to memory which can be assumed unchanged.
7048Being invariant does not imply that a location is dereferenceable,
7049but it does imply that once the location is known dereferenceable
7050its value is henceforth unchanging.
7051
7052The optional ``!invariant.group`` metadata must reference a single metadata name
7053 ``<index>`` corresponding to a metadata node. See ``invariant.group`` metadata.
7054
7055The optional ``!nonnull`` metadata must reference a single
7056metadata name ``<index>`` corresponding to a metadata node with no
7057entries. The existence of the ``!nonnull`` metadata on the
7058instruction tells the optimizer that the value loaded is known to
7059never be null. This is analogous to the ``nonnull`` attribute
7060on parameters and return values. This metadata can only be applied
7061to loads of a pointer type.
7062
7063The optional ``!dereferenceable`` metadata must reference a single metadata
7064name ``<deref_bytes_node>`` corresponding to a metadata node with one ``i64``
7065entry. The existence of the ``!dereferenceable`` metadata on the instruction
7066tells the optimizer that the value loaded is known to be dereferenceable.
7067The number of bytes known to be dereferenceable is specified by the integer
7068value in the metadata node. This is analogous to the ''dereferenceable''
7069attribute on parameters and return values. This metadata can only be applied
7070to loads of a pointer type.
7071
7072The optional ``!dereferenceable_or_null`` metadata must reference a single
7073metadata name ``<deref_bytes_node>`` corresponding to a metadata node with one
7074``i64`` entry. The existence of the ``!dereferenceable_or_null`` metadata on the
7075instruction tells the optimizer that the value loaded is known to be either
7076dereferenceable or null.
7077The number of bytes known to be dereferenceable is specified by the integer
7078value in the metadata node. This is analogous to the ''dereferenceable_or_null''
7079attribute on parameters and return values. This metadata can only be applied
7080to loads of a pointer type.
7081
7082The optional ``!align`` metadata must reference a single metadata name
7083``<align_node>`` corresponding to a metadata node with one ``i64`` entry.
7084The existence of the ``!align`` metadata on the instruction tells the
7085optimizer that the value loaded is known to be aligned to a boundary specified
7086by the integer value in the metadata node. The alignment must be a power of 2.
7087This is analogous to the ''align'' attribute on parameters and return values.
7088This metadata can only be applied to loads of a pointer type.
7089
7090Semantics:
7091""""""""""
7092
7093The location of memory pointed to is loaded. If the value being loaded
7094is of scalar type then the number of bytes read does not exceed the
7095minimum number of bytes needed to hold all bits of the type. For
7096example, loading an ``i24`` reads at most three bytes. When loading a
7097value of a type like ``i20`` with a size that is not an integral number
7098of bytes, the result is undefined if the value was not originally
7099written using a store of the same type.
7100
7101Examples:
7102"""""""""
7103
7104.. code-block:: llvm
7105
7106      %ptr = alloca i32                               ; yields i32*:ptr
7107      store i32 3, i32* %ptr                          ; yields void
7108      %val = load i32, i32* %ptr                      ; yields i32:val = i32 3
7109
7110.. _i_store:
7111
7112'``store``' Instruction
7113^^^^^^^^^^^^^^^^^^^^^^^
7114
7115Syntax:
7116"""""""
7117
7118::
7119
7120      store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>][, !invariant.group !<index>]        ; yields void
7121      store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> [, !invariant.group !<index>] ; yields void
7122
7123Overview:
7124"""""""""
7125
7126The '``store``' instruction is used to write to memory.
7127
7128Arguments:
7129""""""""""
7130
7131There are two arguments to the ``store`` instruction: a value to store and an
7132address at which to store it. The type of the ``<pointer>`` operand must be a
7133pointer to the :ref:`first class <t_firstclass>` type of the ``<value>``
7134operand. If the ``store`` is marked as ``volatile``, then the optimizer is not
7135allowed to modify the number or order of execution of this ``store`` with other
7136:ref:`volatile operations <volatile>`.  Only values of :ref:`first class
7137<t_firstclass>` types of known size (i.e. not containing an :ref:`opaque
7138structural type <t_opaque>`) can be stored.
7139
7140If the ``store`` is marked as ``atomic``, it takes an extra :ref:`ordering
7141<ordering>` and optional ``singlethread`` argument. The ``acquire`` and
7142``acq_rel`` orderings aren't valid on ``store`` instructions. Atomic loads
7143produce :ref:`defined <memmodel>` results when they may see multiple atomic
7144stores. The type of the pointee must be an integer, pointer, or floating-point
7145type whose bit width is a power of two greater than or equal to eight and less
7146than or equal to a target-specific size limit.  ``align`` must be explicitly
7147specified on atomic stores, and the store has undefined behavior if the
7148alignment is not set to a value which is at least the size in bytes of the
7149pointee. ``!nontemporal`` does not have any defined semantics for atomic stores.
7150
7151The optional constant ``align`` argument specifies the alignment of the
7152operation (that is, the alignment of the memory address). A value of 0
7153or an omitted ``align`` argument means that the operation has the ABI
7154alignment for the target. It is the responsibility of the code emitter
7155to ensure that the alignment information is correct. Overestimating the
7156alignment results in undefined behavior. Underestimating the
7157alignment may produce less efficient code. An alignment of 1 is always
7158safe. The maximum possible alignment is ``1 << 29``. An alignment
7159value higher than the size of the stored type implies memory up to the
7160alignment value bytes can be stored to without trapping in the default
7161address space. Storing to the higher bytes however may result in data
7162races if another thread can access the same address. Introducing a
7163data race is not allowed. Storing to the extra bytes is not allowed
7164even in situations where a data race is known to not exist if the
7165function has the ``sanitize_address`` attribute.
7166
7167The optional ``!nontemporal`` metadata must reference a single metadata
7168name ``<index>`` corresponding to a metadata node with one ``i32`` entry of
7169value 1. The existence of the ``!nontemporal`` metadata on the instruction
7170tells the optimizer and code generator that this load is not expected to
7171be reused in the cache. The code generator may select special
7172instructions to save cache bandwidth, such as the ``MOVNT`` instruction on
7173x86.
7174
7175The optional ``!invariant.group`` metadata must reference a
7176single metadata name ``<index>``. See ``invariant.group`` metadata.
7177
7178Semantics:
7179""""""""""
7180
7181The contents of memory are updated to contain ``<value>`` at the
7182location specified by the ``<pointer>`` operand. If ``<value>`` is
7183of scalar type then the number of bytes written does not exceed the
7184minimum number of bytes needed to hold all bits of the type. For
7185example, storing an ``i24`` writes at most three bytes. When writing a
7186value of a type like ``i20`` with a size that is not an integral number
7187of bytes, it is unspecified what happens to the extra bits that do not
7188belong to the type, but they will typically be overwritten.
7189
7190Example:
7191""""""""
7192
7193.. code-block:: llvm
7194
7195      %ptr = alloca i32                               ; yields i32*:ptr
7196      store i32 3, i32* %ptr                          ; yields void
7197      %val = load i32, i32* %ptr                      ; yields i32:val = i32 3
7198
7199.. _i_fence:
7200
7201'``fence``' Instruction
7202^^^^^^^^^^^^^^^^^^^^^^^
7203
7204Syntax:
7205"""""""
7206
7207::
7208
7209      fence [singlethread] <ordering>                   ; yields void
7210
7211Overview:
7212"""""""""
7213
7214The '``fence``' instruction is used to introduce happens-before edges
7215between operations.
7216
7217Arguments:
7218""""""""""
7219
7220'``fence``' instructions take an :ref:`ordering <ordering>` argument which
7221defines what *synchronizes-with* edges they add. They can only be given
7222``acquire``, ``release``, ``acq_rel``, and ``seq_cst`` orderings.
7223
7224Semantics:
7225""""""""""
7226
7227A fence A which has (at least) ``release`` ordering semantics
7228*synchronizes with* a fence B with (at least) ``acquire`` ordering
7229semantics if and only if there exist atomic operations X and Y, both
7230operating on some atomic object M, such that A is sequenced before X, X
7231modifies M (either directly or through some side effect of a sequence
7232headed by X), Y is sequenced before B, and Y observes M. This provides a
7233*happens-before* dependency between A and B. Rather than an explicit
7234``fence``, one (but not both) of the atomic operations X or Y might
7235provide a ``release`` or ``acquire`` (resp.) ordering constraint and
7236still *synchronize-with* the explicit ``fence`` and establish the
7237*happens-before* edge.
7238
7239A ``fence`` which has ``seq_cst`` ordering, in addition to having both
7240``acquire`` and ``release`` semantics specified above, participates in
7241the global program order of other ``seq_cst`` operations and/or fences.
7242
7243The optional ":ref:`singlethread <singlethread>`" argument specifies
7244that the fence only synchronizes with other fences in the same thread.
7245(This is useful for interacting with signal handlers.)
7246
7247Example:
7248""""""""
7249
7250.. code-block:: llvm
7251
7252      fence acquire                          ; yields void
7253      fence singlethread seq_cst             ; yields void
7254
7255.. _i_cmpxchg:
7256
7257'``cmpxchg``' Instruction
7258^^^^^^^^^^^^^^^^^^^^^^^^^
7259
7260Syntax:
7261"""""""
7262
7263::
7264
7265      cmpxchg [weak] [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <success ordering> <failure ordering> ; yields  { ty, i1 }
7266
7267Overview:
7268"""""""""
7269
7270The '``cmpxchg``' instruction is used to atomically modify memory. It
7271loads a value in memory and compares it to a given value. If they are
7272equal, it tries to store a new value into the memory.
7273
7274Arguments:
7275""""""""""
7276
7277There are three arguments to the '``cmpxchg``' instruction: an address
7278to operate on, a value to compare to the value currently be at that
7279address, and a new value to place at that address if the compared values
7280are equal. The type of '<cmp>' must be an integer or pointer type whose
7281bit width is a power of two greater than or equal to eight and less
7282than or equal to a target-specific size limit. '<cmp>' and '<new>' must
7283have the same type, and the type of '<pointer>' must be a pointer to
7284that type. If the ``cmpxchg`` is marked as ``volatile``, then the
7285optimizer is not allowed to modify the number or order of execution of
7286this ``cmpxchg`` with other :ref:`volatile operations <volatile>`.
7287
7288The success and failure :ref:`ordering <ordering>` arguments specify how this
7289``cmpxchg`` synchronizes with other atomic operations. Both ordering parameters
7290must be at least ``monotonic``, the ordering constraint on failure must be no
7291stronger than that on success, and the failure ordering cannot be either
7292``release`` or ``acq_rel``.
7293
7294The optional "``singlethread``" argument declares that the ``cmpxchg``
7295is only atomic with respect to code (usually signal handlers) running in
7296the same thread as the ``cmpxchg``. Otherwise the cmpxchg is atomic with
7297respect to all other code in the system.
7298
7299The pointer passed into cmpxchg must have alignment greater than or
7300equal to the size in memory of the operand.
7301
7302Semantics:
7303""""""""""
7304
7305The contents of memory at the location specified by the '``<pointer>``' operand
7306is read and compared to '``<cmp>``'; if the read value is the equal, the
7307'``<new>``' is written. The original value at the location is returned, together
7308with a flag indicating success (true) or failure (false).
7309
7310If the cmpxchg operation is marked as ``weak`` then a spurious failure is
7311permitted: the operation may not write ``<new>`` even if the comparison
7312matched.
7313
7314If the cmpxchg operation is strong (the default), the i1 value is 1 if and only
7315if the value loaded equals ``cmp``.
7316
7317A successful ``cmpxchg`` is a read-modify-write instruction for the purpose of
7318identifying release sequences. A failed ``cmpxchg`` is equivalent to an atomic
7319load with an ordering parameter determined the second ordering parameter.
7320
7321Example:
7322""""""""
7323
7324.. code-block:: llvm
7325
7326    entry:
7327      %orig = load atomic i32, i32* %ptr unordered, align 4                      ; yields i32
7328      br label %loop
7329
7330    loop:
7331      %cmp = phi i32 [ %orig, %entry ], [%value_loaded, %loop]
7332      %squared = mul i32 %cmp, %cmp
7333      %val_success = cmpxchg i32* %ptr, i32 %cmp, i32 %squared acq_rel monotonic ; yields  { i32, i1 }
7334      %value_loaded = extractvalue { i32, i1 } %val_success, 0
7335      %success = extractvalue { i32, i1 } %val_success, 1
7336      br i1 %success, label %done, label %loop
7337
7338    done:
7339      ...
7340
7341.. _i_atomicrmw:
7342
7343'``atomicrmw``' Instruction
7344^^^^^^^^^^^^^^^^^^^^^^^^^^^
7345
7346Syntax:
7347"""""""
7348
7349::
7350
7351      atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering>                   ; yields ty
7352
7353Overview:
7354"""""""""
7355
7356The '``atomicrmw``' instruction is used to atomically modify memory.
7357
7358Arguments:
7359""""""""""
7360
7361There are three arguments to the '``atomicrmw``' instruction: an
7362operation to apply, an address whose value to modify, an argument to the
7363operation. The operation must be one of the following keywords:
7364
7365-  xchg
7366-  add
7367-  sub
7368-  and
7369-  nand
7370-  or
7371-  xor
7372-  max
7373-  min
7374-  umax
7375-  umin
7376
7377The type of '<value>' must be an integer type whose bit width is a power
7378of two greater than or equal to eight and less than or equal to a
7379target-specific size limit. The type of the '``<pointer>``' operand must
7380be a pointer to that type. If the ``atomicrmw`` is marked as
7381``volatile``, then the optimizer is not allowed to modify the number or
7382order of execution of this ``atomicrmw`` with other :ref:`volatile
7383operations <volatile>`.
7384
7385Semantics:
7386""""""""""
7387
7388The contents of memory at the location specified by the '``<pointer>``'
7389operand are atomically read, modified, and written back. The original
7390value at the location is returned. The modification is specified by the
7391operation argument:
7392
7393-  xchg: ``*ptr = val``
7394-  add: ``*ptr = *ptr + val``
7395-  sub: ``*ptr = *ptr - val``
7396-  and: ``*ptr = *ptr & val``
7397-  nand: ``*ptr = ~(*ptr & val)``
7398-  or: ``*ptr = *ptr | val``
7399-  xor: ``*ptr = *ptr ^ val``
7400-  max: ``*ptr = *ptr > val ? *ptr : val`` (using a signed comparison)
7401-  min: ``*ptr = *ptr < val ? *ptr : val`` (using a signed comparison)
7402-  umax: ``*ptr = *ptr > val ? *ptr : val`` (using an unsigned
7403   comparison)
7404-  umin: ``*ptr = *ptr < val ? *ptr : val`` (using an unsigned
7405   comparison)
7406
7407Example:
7408""""""""
7409
7410.. code-block:: llvm
7411
7412      %old = atomicrmw add i32* %ptr, i32 1 acquire                        ; yields i32
7413
7414.. _i_getelementptr:
7415
7416'``getelementptr``' Instruction
7417^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7418
7419Syntax:
7420"""""""
7421
7422::
7423
7424      <result> = getelementptr <ty>, <ty>* <ptrval>{, <ty> <idx>}*
7425      <result> = getelementptr inbounds <ty>, <ty>* <ptrval>{, <ty> <idx>}*
7426      <result> = getelementptr <ty>, <ptr vector> <ptrval>, <vector index type> <idx>
7427
7428Overview:
7429"""""""""
7430
7431The '``getelementptr``' instruction is used to get the address of a
7432subelement of an :ref:`aggregate <t_aggregate>` data structure. It performs
7433address calculation only and does not access memory. The instruction can also
7434be used to calculate a vector of such addresses.
7435
7436Arguments:
7437""""""""""
7438
7439The first argument is always a type used as the basis for the calculations.
7440The second argument is always a pointer or a vector of pointers, and is the
7441base address to start from. The remaining arguments are indices
7442that indicate which of the elements of the aggregate object are indexed.
7443The interpretation of each index is dependent on the type being indexed
7444into. The first index always indexes the pointer value given as the
7445first argument, the second index indexes a value of the type pointed to
7446(not necessarily the value directly pointed to, since the first index
7447can be non-zero), etc. The first type indexed into must be a pointer
7448value, subsequent types can be arrays, vectors, and structs. Note that
7449subsequent types being indexed into can never be pointers, since that
7450would require loading the pointer before continuing calculation.
7451
7452The type of each index argument depends on the type it is indexing into.
7453When indexing into a (optionally packed) structure, only ``i32`` integer
7454**constants** are allowed (when using a vector of indices they must all
7455be the **same** ``i32`` integer constant). When indexing into an array,
7456pointer or vector, integers of any width are allowed, and they are not
7457required to be constant. These integers are treated as signed values
7458where relevant.
7459
7460For example, let's consider a C code fragment and how it gets compiled
7461to LLVM:
7462
7463.. code-block:: c
7464
7465    struct RT {
7466      char A;
7467      int B[10][20];
7468      char C;
7469    };
7470    struct ST {
7471      int X;
7472      double Y;
7473      struct RT Z;
7474    };
7475
7476    int *foo(struct ST *s) {
7477      return &s[1].Z.B[5][13];
7478    }
7479
7480The LLVM code generated by Clang is:
7481
7482.. code-block:: llvm
7483
7484    %struct.RT = type { i8, [10 x [20 x i32]], i8 }
7485    %struct.ST = type { i32, double, %struct.RT }
7486
7487    define i32* @foo(%struct.ST* %s) nounwind uwtable readnone optsize ssp {
7488    entry:
7489      %arrayidx = getelementptr inbounds %struct.ST, %struct.ST* %s, i64 1, i32 2, i32 1, i64 5, i64 13
7490      ret i32* %arrayidx
7491    }
7492
7493Semantics:
7494""""""""""
7495
7496In the example above, the first index is indexing into the
7497'``%struct.ST*``' type, which is a pointer, yielding a '``%struct.ST``'
7498= '``{ i32, double, %struct.RT }``' type, a structure. The second index
7499indexes into the third element of the structure, yielding a
7500'``%struct.RT``' = '``{ i8 , [10 x [20 x i32]], i8 }``' type, another
7501structure. The third index indexes into the second element of the
7502structure, yielding a '``[10 x [20 x i32]]``' type, an array. The two
7503dimensions of the array are subscripted into, yielding an '``i32``'
7504type. The '``getelementptr``' instruction returns a pointer to this
7505element, thus computing a value of '``i32*``' type.
7506
7507Note that it is perfectly legal to index partially through a structure,
7508returning a pointer to an inner element. Because of this, the LLVM code
7509for the given testcase is equivalent to:
7510
7511.. code-block:: llvm
7512
7513    define i32* @foo(%struct.ST* %s) {
7514      %t1 = getelementptr %struct.ST, %struct.ST* %s, i32 1                        ; yields %struct.ST*:%t1
7515      %t2 = getelementptr %struct.ST, %struct.ST* %t1, i32 0, i32 2                ; yields %struct.RT*:%t2
7516      %t3 = getelementptr %struct.RT, %struct.RT* %t2, i32 0, i32 1                ; yields [10 x [20 x i32]]*:%t3
7517      %t4 = getelementptr [10 x [20 x i32]], [10 x [20 x i32]]* %t3, i32 0, i32 5  ; yields [20 x i32]*:%t4
7518      %t5 = getelementptr [20 x i32], [20 x i32]* %t4, i32 0, i32 13               ; yields i32*:%t5
7519      ret i32* %t5
7520    }
7521
7522If the ``inbounds`` keyword is present, the result value of the
7523``getelementptr`` is a :ref:`poison value <poisonvalues>` if the base
7524pointer is not an *in bounds* address of an allocated object, or if any
7525of the addresses that would be formed by successive addition of the
7526offsets implied by the indices to the base address with infinitely
7527precise signed arithmetic are not an *in bounds* address of that
7528allocated object. The *in bounds* addresses for an allocated object are
7529all the addresses that point into the object, plus the address one byte
7530past the end. In cases where the base is a vector of pointers the
7531``inbounds`` keyword applies to each of the computations element-wise.
7532
7533If the ``inbounds`` keyword is not present, the offsets are added to the
7534base address with silently-wrapping two's complement arithmetic. If the
7535offsets have a different width from the pointer, they are sign-extended
7536or truncated to the width of the pointer. The result value of the
7537``getelementptr`` may be outside the object pointed to by the base
7538pointer. The result value may not necessarily be used to access memory
7539though, even if it happens to point into allocated storage. See the
7540:ref:`Pointer Aliasing Rules <pointeraliasing>` section for more
7541information.
7542
7543The getelementptr instruction is often confusing. For some more insight
7544into how it works, see :doc:`the getelementptr FAQ <GetElementPtr>`.
7545
7546Example:
7547""""""""
7548
7549.. code-block:: llvm
7550
7551        ; yields [12 x i8]*:aptr
7552        %aptr = getelementptr {i32, [12 x i8]}, {i32, [12 x i8]}* %saptr, i64 0, i32 1
7553        ; yields i8*:vptr
7554        %vptr = getelementptr {i32, <2 x i8>}, {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
7555        ; yields i8*:eptr
7556        %eptr = getelementptr [12 x i8], [12 x i8]* %aptr, i64 0, i32 1
7557        ; yields i32*:iptr
7558        %iptr = getelementptr [10 x i32], [10 x i32]* @arr, i16 0, i16 0
7559
7560Vector of pointers:
7561"""""""""""""""""""
7562
7563The ``getelementptr`` returns a vector of pointers, instead of a single address,
7564when one or more of its arguments is a vector. In such cases, all vector
7565arguments should have the same number of elements, and every scalar argument
7566will be effectively broadcast into a vector during address calculation.
7567
7568.. code-block:: llvm
7569
7570     ; All arguments are vectors:
7571     ;   A[i] = ptrs[i] + offsets[i]*sizeof(i8)
7572     %A = getelementptr i8, <4 x i8*> %ptrs, <4 x i64> %offsets
7573
7574     ; Add the same scalar offset to each pointer of a vector:
7575     ;   A[i] = ptrs[i] + offset*sizeof(i8)
7576     %A = getelementptr i8, <4 x i8*> %ptrs, i64 %offset
7577
7578     ; Add distinct offsets to the same pointer:
7579     ;   A[i] = ptr + offsets[i]*sizeof(i8)
7580     %A = getelementptr i8, i8* %ptr, <4 x i64> %offsets
7581
7582     ; In all cases described above the type of the result is <4 x i8*>
7583
7584The two following instructions are equivalent:
7585
7586.. code-block:: llvm
7587
7588     getelementptr  %struct.ST, <4 x %struct.ST*> %s, <4 x i64> %ind1,
7589       <4 x i32> <i32 2, i32 2, i32 2, i32 2>,
7590       <4 x i32> <i32 1, i32 1, i32 1, i32 1>,
7591       <4 x i32> %ind4,
7592       <4 x i64> <i64 13, i64 13, i64 13, i64 13>
7593
7594     getelementptr  %struct.ST, <4 x %struct.ST*> %s, <4 x i64> %ind1,
7595       i32 2, i32 1, <4 x i32> %ind4, i64 13
7596
7597Let's look at the C code, where the vector version of ``getelementptr``
7598makes sense:
7599
7600.. code-block:: c
7601
7602    // Let's assume that we vectorize the following loop:
7603    double *A, B; int *C;
7604    for (int i = 0; i < size; ++i) {
7605      A[i] = B[C[i]];
7606    }
7607
7608.. code-block:: llvm
7609
7610    ; get pointers for 8 elements from array B
7611    %ptrs = getelementptr double, double* %B, <8 x i32> %C
7612    ; load 8 elements from array B into A
7613    %A = call <8 x double> @llvm.masked.gather.v8f64(<8 x double*> %ptrs,
7614         i32 8, <8 x i1> %mask, <8 x double> %passthru)
7615
7616Conversion Operations
7617---------------------
7618
7619The instructions in this category are the conversion instructions
7620(casting) which all take a single operand and a type. They perform
7621various bit conversions on the operand.
7622
7623'``trunc .. to``' Instruction
7624^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7625
7626Syntax:
7627"""""""
7628
7629::
7630
7631      <result> = trunc <ty> <value> to <ty2>             ; yields ty2
7632
7633Overview:
7634"""""""""
7635
7636The '``trunc``' instruction truncates its operand to the type ``ty2``.
7637
7638Arguments:
7639""""""""""
7640
7641The '``trunc``' instruction takes a value to trunc, and a type to trunc
7642it to. Both types must be of :ref:`integer <t_integer>` types, or vectors
7643of the same number of integers. The bit size of the ``value`` must be
7644larger than the bit size of the destination type, ``ty2``. Equal sized
7645types are not allowed.
7646
7647Semantics:
7648""""""""""
7649
7650The '``trunc``' instruction truncates the high order bits in ``value``
7651and converts the remaining bits to ``ty2``. Since the source size must
7652be larger than the destination size, ``trunc`` cannot be a *no-op cast*.
7653It will always truncate bits.
7654
7655Example:
7656""""""""
7657
7658.. code-block:: llvm
7659
7660      %X = trunc i32 257 to i8                        ; yields i8:1
7661      %Y = trunc i32 123 to i1                        ; yields i1:true
7662      %Z = trunc i32 122 to i1                        ; yields i1:false
7663      %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> ; yields <i8 8, i8 7>
7664
7665'``zext .. to``' Instruction
7666^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7667
7668Syntax:
7669"""""""
7670
7671::
7672
7673      <result> = zext <ty> <value> to <ty2>             ; yields ty2
7674
7675Overview:
7676"""""""""
7677
7678The '``zext``' instruction zero extends its operand to type ``ty2``.
7679
7680Arguments:
7681""""""""""
7682
7683The '``zext``' instruction takes a value to cast, and a type to cast it
7684to. Both types must be of :ref:`integer <t_integer>` types, or vectors of
7685the same number of integers. The bit size of the ``value`` must be
7686smaller than the bit size of the destination type, ``ty2``.
7687
7688Semantics:
7689""""""""""
7690
7691The ``zext`` fills the high order bits of the ``value`` with zero bits
7692until it reaches the size of the destination type, ``ty2``.
7693
7694When zero extending from i1, the result will always be either 0 or 1.
7695
7696Example:
7697""""""""
7698
7699.. code-block:: llvm
7700
7701      %X = zext i32 257 to i64              ; yields i64:257
7702      %Y = zext i1 true to i32              ; yields i32:1
7703      %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7>
7704
7705'``sext .. to``' Instruction
7706^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7707
7708Syntax:
7709"""""""
7710
7711::
7712
7713      <result> = sext <ty> <value> to <ty2>             ; yields ty2
7714
7715Overview:
7716"""""""""
7717
7718The '``sext``' sign extends ``value`` to the type ``ty2``.
7719
7720Arguments:
7721""""""""""
7722
7723The '``sext``' instruction takes a value to cast, and a type to cast it
7724to. Both types must be of :ref:`integer <t_integer>` types, or vectors of
7725the same number of integers. The bit size of the ``value`` must be
7726smaller than the bit size of the destination type, ``ty2``.
7727
7728Semantics:
7729""""""""""
7730
7731The '``sext``' instruction performs a sign extension by copying the sign
7732bit (highest order bit) of the ``value`` until it reaches the bit size
7733of the type ``ty2``.
7734
7735When sign extending from i1, the extension always results in -1 or 0.
7736
7737Example:
7738""""""""
7739
7740.. code-block:: llvm
7741
7742      %X = sext i8  -1 to i16              ; yields i16   :65535
7743      %Y = sext i1 true to i32             ; yields i32:-1
7744      %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7>
7745
7746'``fptrunc .. to``' Instruction
7747^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7748
7749Syntax:
7750"""""""
7751
7752::
7753
7754      <result> = fptrunc <ty> <value> to <ty2>             ; yields ty2
7755
7756Overview:
7757"""""""""
7758
7759The '``fptrunc``' instruction truncates ``value`` to type ``ty2``.
7760
7761Arguments:
7762""""""""""
7763
7764The '``fptrunc``' instruction takes a :ref:`floating point <t_floating>`
7765value to cast and a :ref:`floating point <t_floating>` type to cast it to.
7766The size of ``value`` must be larger than the size of ``ty2``. This
7767implies that ``fptrunc`` cannot be used to make a *no-op cast*.
7768
7769Semantics:
7770""""""""""
7771
7772The '``fptrunc``' instruction casts a ``value`` from a larger
7773:ref:`floating point <t_floating>` type to a smaller :ref:`floating
7774point <t_floating>` type. If the value cannot fit (i.e. overflows) within the
7775destination type, ``ty2``, then the results are undefined. If the cast produces
7776an inexact result, how rounding is performed (e.g. truncation, also known as
7777round to zero) is undefined.
7778
7779Example:
7780""""""""
7781
7782.. code-block:: llvm
7783
7784      %X = fptrunc double 123.0 to float         ; yields float:123.0
7785      %Y = fptrunc double 1.0E+300 to float      ; yields undefined
7786
7787'``fpext .. to``' Instruction
7788^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7789
7790Syntax:
7791"""""""
7792
7793::
7794
7795      <result> = fpext <ty> <value> to <ty2>             ; yields ty2
7796
7797Overview:
7798"""""""""
7799
7800The '``fpext``' extends a floating point ``value`` to a larger floating
7801point value.
7802
7803Arguments:
7804""""""""""
7805
7806The '``fpext``' instruction takes a :ref:`floating point <t_floating>`
7807``value`` to cast, and a :ref:`floating point <t_floating>` type to cast it
7808to. The source type must be smaller than the destination type.
7809
7810Semantics:
7811""""""""""
7812
7813The '``fpext``' instruction extends the ``value`` from a smaller
7814:ref:`floating point <t_floating>` type to a larger :ref:`floating
7815point <t_floating>` type. The ``fpext`` cannot be used to make a
7816*no-op cast* because it always changes bits. Use ``bitcast`` to make a
7817*no-op cast* for a floating point cast.
7818
7819Example:
7820""""""""
7821
7822.. code-block:: llvm
7823
7824      %X = fpext float 3.125 to double         ; yields double:3.125000e+00
7825      %Y = fpext double %X to fp128            ; yields fp128:0xL00000000000000004000900000000000
7826
7827'``fptoui .. to``' Instruction
7828^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7829
7830Syntax:
7831"""""""
7832
7833::
7834
7835      <result> = fptoui <ty> <value> to <ty2>             ; yields ty2
7836
7837Overview:
7838"""""""""
7839
7840The '``fptoui``' converts a floating point ``value`` to its unsigned
7841integer equivalent of type ``ty2``.
7842
7843Arguments:
7844""""""""""
7845
7846The '``fptoui``' instruction takes a value to cast, which must be a
7847scalar or vector :ref:`floating point <t_floating>` value, and a type to
7848cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If
7849``ty`` is a vector floating point type, ``ty2`` must be a vector integer
7850type with the same number of elements as ``ty``
7851
7852Semantics:
7853""""""""""
7854
7855The '``fptoui``' instruction converts its :ref:`floating
7856point <t_floating>` operand into the nearest (rounding towards zero)
7857unsigned integer value. If the value cannot fit in ``ty2``, the results
7858are undefined.
7859
7860Example:
7861""""""""
7862
7863.. code-block:: llvm
7864
7865      %X = fptoui double 123.0 to i32      ; yields i32:123
7866      %Y = fptoui float 1.0E+300 to i1     ; yields undefined:1
7867      %Z = fptoui float 1.04E+17 to i8     ; yields undefined:1
7868
7869'``fptosi .. to``' Instruction
7870^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7871
7872Syntax:
7873"""""""
7874
7875::
7876
7877      <result> = fptosi <ty> <value> to <ty2>             ; yields ty2
7878
7879Overview:
7880"""""""""
7881
7882The '``fptosi``' instruction converts :ref:`floating point <t_floating>`
7883``value`` to type ``ty2``.
7884
7885Arguments:
7886""""""""""
7887
7888The '``fptosi``' instruction takes a value to cast, which must be a
7889scalar or vector :ref:`floating point <t_floating>` value, and a type to
7890cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If
7891``ty`` is a vector floating point type, ``ty2`` must be a vector integer
7892type with the same number of elements as ``ty``
7893
7894Semantics:
7895""""""""""
7896
7897The '``fptosi``' instruction converts its :ref:`floating
7898point <t_floating>` operand into the nearest (rounding towards zero)
7899signed integer value. If the value cannot fit in ``ty2``, the results
7900are undefined.
7901
7902Example:
7903""""""""
7904
7905.. code-block:: llvm
7906
7907      %X = fptosi double -123.0 to i32      ; yields i32:-123
7908      %Y = fptosi float 1.0E-247 to i1      ; yields undefined:1
7909      %Z = fptosi float 1.04E+17 to i8      ; yields undefined:1
7910
7911'``uitofp .. to``' Instruction
7912^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7913
7914Syntax:
7915"""""""
7916
7917::
7918
7919      <result> = uitofp <ty> <value> to <ty2>             ; yields ty2
7920
7921Overview:
7922"""""""""
7923
7924The '``uitofp``' instruction regards ``value`` as an unsigned integer
7925and converts that value to the ``ty2`` type.
7926
7927Arguments:
7928""""""""""
7929
7930The '``uitofp``' instruction takes a value to cast, which must be a
7931scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to
7932``ty2``, which must be an :ref:`floating point <t_floating>` type. If
7933``ty`` is a vector integer type, ``ty2`` must be a vector floating point
7934type with the same number of elements as ``ty``
7935
7936Semantics:
7937""""""""""
7938
7939The '``uitofp``' instruction interprets its operand as an unsigned
7940integer quantity and converts it to the corresponding floating point
7941value. If the value cannot fit in the floating point value, the results
7942are undefined.
7943
7944Example:
7945""""""""
7946
7947.. code-block:: llvm
7948
7949      %X = uitofp i32 257 to float         ; yields float:257.0
7950      %Y = uitofp i8 -1 to double          ; yields double:255.0
7951
7952'``sitofp .. to``' Instruction
7953^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7954
7955Syntax:
7956"""""""
7957
7958::
7959
7960      <result> = sitofp <ty> <value> to <ty2>             ; yields ty2
7961
7962Overview:
7963"""""""""
7964
7965The '``sitofp``' instruction regards ``value`` as a signed integer and
7966converts that value to the ``ty2`` type.
7967
7968Arguments:
7969""""""""""
7970
7971The '``sitofp``' instruction takes a value to cast, which must be a
7972scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to
7973``ty2``, which must be an :ref:`floating point <t_floating>` type. If
7974``ty`` is a vector integer type, ``ty2`` must be a vector floating point
7975type with the same number of elements as ``ty``
7976
7977Semantics:
7978""""""""""
7979
7980The '``sitofp``' instruction interprets its operand as a signed integer
7981quantity and converts it to the corresponding floating point value. If
7982the value cannot fit in the floating point value, the results are
7983undefined.
7984
7985Example:
7986""""""""
7987
7988.. code-block:: llvm
7989
7990      %X = sitofp i32 257 to float         ; yields float:257.0
7991      %Y = sitofp i8 -1 to double          ; yields double:-1.0
7992
7993.. _i_ptrtoint:
7994
7995'``ptrtoint .. to``' Instruction
7996^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7997
7998Syntax:
7999"""""""
8000
8001::
8002
8003      <result> = ptrtoint <ty> <value> to <ty2>             ; yields ty2
8004
8005Overview:
8006"""""""""
8007
8008The '``ptrtoint``' instruction converts the pointer or a vector of
8009pointers ``value`` to the integer (or vector of integers) type ``ty2``.
8010
8011Arguments:
8012""""""""""
8013
8014The '``ptrtoint``' instruction takes a ``value`` to cast, which must be
8015a value of type :ref:`pointer <t_pointer>` or a vector of pointers, and a
8016type to cast it to ``ty2``, which must be an :ref:`integer <t_integer>` or
8017a vector of integers type.
8018
8019Semantics:
8020""""""""""
8021
8022The '``ptrtoint``' instruction converts ``value`` to integer type
8023``ty2`` by interpreting the pointer value as an integer and either
8024truncating or zero extending that value to the size of the integer type.
8025If ``value`` is smaller than ``ty2`` then a zero extension is done. If
8026``value`` is larger than ``ty2`` then a truncation is done. If they are
8027the same size, then nothing is done (*no-op cast*) other than a type
8028change.
8029
8030Example:
8031""""""""
8032
8033.. code-block:: llvm
8034
8035      %X = ptrtoint i32* %P to i8                         ; yields truncation on 32-bit architecture
8036      %Y = ptrtoint i32* %P to i64                        ; yields zero extension on 32-bit architecture
8037      %Z = ptrtoint <4 x i32*> %P to <4 x i64>; yields vector zero extension for a vector of addresses on 32-bit architecture
8038
8039.. _i_inttoptr:
8040
8041'``inttoptr .. to``' Instruction
8042^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8043
8044Syntax:
8045"""""""
8046
8047::
8048
8049      <result> = inttoptr <ty> <value> to <ty2>             ; yields ty2
8050
8051Overview:
8052"""""""""
8053
8054The '``inttoptr``' instruction converts an integer ``value`` to a
8055pointer type, ``ty2``.
8056
8057Arguments:
8058""""""""""
8059
8060The '``inttoptr``' instruction takes an :ref:`integer <t_integer>` value to
8061cast, and a type to cast it to, which must be a :ref:`pointer <t_pointer>`
8062type.
8063
8064Semantics:
8065""""""""""
8066
8067The '``inttoptr``' instruction converts ``value`` to type ``ty2`` by
8068applying either a zero extension or a truncation depending on the size
8069of the integer ``value``. If ``value`` is larger than the size of a
8070pointer then a truncation is done. If ``value`` is smaller than the size
8071of a pointer then a zero extension is done. If they are the same size,
8072nothing is done (*no-op cast*).
8073
8074Example:
8075""""""""
8076
8077.. code-block:: llvm
8078
8079      %X = inttoptr i32 255 to i32*          ; yields zero extension on 64-bit architecture
8080      %Y = inttoptr i32 255 to i32*          ; yields no-op on 32-bit architecture
8081      %Z = inttoptr i64 0 to i32*            ; yields truncation on 32-bit architecture
8082      %Z = inttoptr <4 x i32> %G to <4 x i8*>; yields truncation of vector G to four pointers
8083
8084.. _i_bitcast:
8085
8086'``bitcast .. to``' Instruction
8087^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8088
8089Syntax:
8090"""""""
8091
8092::
8093
8094      <result> = bitcast <ty> <value> to <ty2>             ; yields ty2
8095
8096Overview:
8097"""""""""
8098
8099The '``bitcast``' instruction converts ``value`` to type ``ty2`` without
8100changing any bits.
8101
8102Arguments:
8103""""""""""
8104
8105The '``bitcast``' instruction takes a value to cast, which must be a
8106non-aggregate first class value, and a type to cast it to, which must
8107also be a non-aggregate :ref:`first class <t_firstclass>` type. The
8108bit sizes of ``value`` and the destination type, ``ty2``, must be
8109identical. If the source type is a pointer, the destination type must
8110also be a pointer of the same size. This instruction supports bitwise
8111conversion of vectors to integers and to vectors of other types (as
8112long as they have the same size).
8113
8114Semantics:
8115""""""""""
8116
8117The '``bitcast``' instruction converts ``value`` to type ``ty2``. It
8118is always a *no-op cast* because no bits change with this
8119conversion. The conversion is done as if the ``value`` had been stored
8120to memory and read back as type ``ty2``. Pointer (or vector of
8121pointers) types may only be converted to other pointer (or vector of
8122pointers) types with the same address space through this instruction.
8123To convert pointers to other types, use the :ref:`inttoptr <i_inttoptr>`
8124or :ref:`ptrtoint <i_ptrtoint>` instructions first.
8125
8126Example:
8127""""""""
8128
8129.. code-block:: llvm
8130
8131      %X = bitcast i8 255 to i8              ; yields i8 :-1
8132      %Y = bitcast i32* %x to sint*          ; yields sint*:%x
8133      %Z = bitcast <2 x int> %V to i64;        ; yields i64: %V
8134      %Z = bitcast <2 x i32*> %V to <2 x i64*> ; yields <2 x i64*>
8135
8136.. _i_addrspacecast:
8137
8138'``addrspacecast .. to``' Instruction
8139^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8140
8141Syntax:
8142"""""""
8143
8144::
8145
8146      <result> = addrspacecast <pty> <ptrval> to <pty2>       ; yields pty2
8147
8148Overview:
8149"""""""""
8150
8151The '``addrspacecast``' instruction converts ``ptrval`` from ``pty`` in
8152address space ``n`` to type ``pty2`` in address space ``m``.
8153
8154Arguments:
8155""""""""""
8156
8157The '``addrspacecast``' instruction takes a pointer or vector of pointer value
8158to cast and a pointer type to cast it to, which must have a different
8159address space.
8160
8161Semantics:
8162""""""""""
8163
8164The '``addrspacecast``' instruction converts the pointer value
8165``ptrval`` to type ``pty2``. It can be a *no-op cast* or a complex
8166value modification, depending on the target and the address space
8167pair. Pointer conversions within the same address space must be
8168performed with the ``bitcast`` instruction. Note that if the address space
8169conversion is legal then both result and operand refer to the same memory
8170location.
8171
8172Example:
8173""""""""
8174
8175.. code-block:: llvm
8176
8177      %X = addrspacecast i32* %x to i32 addrspace(1)*    ; yields i32 addrspace(1)*:%x
8178      %Y = addrspacecast i32 addrspace(1)* %y to i64 addrspace(2)*    ; yields i64 addrspace(2)*:%y
8179      %Z = addrspacecast <4 x i32*> %z to <4 x float addrspace(3)*>   ; yields <4 x float addrspace(3)*>:%z
8180
8181.. _otherops:
8182
8183Other Operations
8184----------------
8185
8186The instructions in this category are the "miscellaneous" instructions,
8187which defy better classification.
8188
8189.. _i_icmp:
8190
8191'``icmp``' Instruction
8192^^^^^^^^^^^^^^^^^^^^^^
8193
8194Syntax:
8195"""""""
8196
8197::
8198
8199      <result> = icmp <cond> <ty> <op1>, <op2>   ; yields i1 or <N x i1>:result
8200
8201Overview:
8202"""""""""
8203
8204The '``icmp``' instruction returns a boolean value or a vector of
8205boolean values based on comparison of its two integer, integer vector,
8206pointer, or pointer vector operands.
8207
8208Arguments:
8209""""""""""
8210
8211The '``icmp``' instruction takes three operands. The first operand is
8212the condition code indicating the kind of comparison to perform. It is
8213not a value, just a keyword. The possible condition codes are:
8214
8215#. ``eq``: equal
8216#. ``ne``: not equal
8217#. ``ugt``: unsigned greater than
8218#. ``uge``: unsigned greater or equal
8219#. ``ult``: unsigned less than
8220#. ``ule``: unsigned less or equal
8221#. ``sgt``: signed greater than
8222#. ``sge``: signed greater or equal
8223#. ``slt``: signed less than
8224#. ``sle``: signed less or equal
8225
8226The remaining two arguments must be :ref:`integer <t_integer>` or
8227:ref:`pointer <t_pointer>` or integer :ref:`vector <t_vector>` typed. They
8228must also be identical types.
8229
8230Semantics:
8231""""""""""
8232
8233The '``icmp``' compares ``op1`` and ``op2`` according to the condition
8234code given as ``cond``. The comparison performed always yields either an
8235:ref:`i1 <t_integer>` or vector of ``i1`` result, as follows:
8236
8237#. ``eq``: yields ``true`` if the operands are equal, ``false``
8238   otherwise. No sign interpretation is necessary or performed.
8239#. ``ne``: yields ``true`` if the operands are unequal, ``false``
8240   otherwise. No sign interpretation is necessary or performed.
8241#. ``ugt``: interprets the operands as unsigned values and yields
8242   ``true`` if ``op1`` is greater than ``op2``.
8243#. ``uge``: interprets the operands as unsigned values and yields
8244   ``true`` if ``op1`` is greater than or equal to ``op2``.
8245#. ``ult``: interprets the operands as unsigned values and yields
8246   ``true`` if ``op1`` is less than ``op2``.
8247#. ``ule``: interprets the operands as unsigned values and yields
8248   ``true`` if ``op1`` is less than or equal to ``op2``.
8249#. ``sgt``: interprets the operands as signed values and yields ``true``
8250   if ``op1`` is greater than ``op2``.
8251#. ``sge``: interprets the operands as signed values and yields ``true``
8252   if ``op1`` is greater than or equal to ``op2``.
8253#. ``slt``: interprets the operands as signed values and yields ``true``
8254   if ``op1`` is less than ``op2``.
8255#. ``sle``: interprets the operands as signed values and yields ``true``
8256   if ``op1`` is less than or equal to ``op2``.
8257
8258If the operands are :ref:`pointer <t_pointer>` typed, the pointer values
8259are compared as if they were integers.
8260
8261If the operands are integer vectors, then they are compared element by
8262element. The result is an ``i1`` vector with the same number of elements
8263as the values being compared. Otherwise, the result is an ``i1``.
8264
8265Example:
8266""""""""
8267
8268.. code-block:: llvm
8269
8270      <result> = icmp eq i32 4, 5          ; yields: result=false
8271      <result> = icmp ne float* %X, %X     ; yields: result=false
8272      <result> = icmp ult i16  4, 5        ; yields: result=true
8273      <result> = icmp sgt i16  4, 5        ; yields: result=false
8274      <result> = icmp ule i16 -4, 5        ; yields: result=false
8275      <result> = icmp sge i16  4, 5        ; yields: result=false
8276
8277.. _i_fcmp:
8278
8279'``fcmp``' Instruction
8280^^^^^^^^^^^^^^^^^^^^^^
8281
8282Syntax:
8283"""""""
8284
8285::
8286
8287      <result> = fcmp [fast-math flags]* <cond> <ty> <op1>, <op2>     ; yields i1 or <N x i1>:result
8288
8289Overview:
8290"""""""""
8291
8292The '``fcmp``' instruction returns a boolean value or vector of boolean
8293values based on comparison of its operands.
8294
8295If the operands are floating point scalars, then the result type is a
8296boolean (:ref:`i1 <t_integer>`).
8297
8298If the operands are floating point vectors, then the result type is a
8299vector of boolean with the same number of elements as the operands being
8300compared.
8301
8302Arguments:
8303""""""""""
8304
8305The '``fcmp``' instruction takes three operands. The first operand is
8306the condition code indicating the kind of comparison to perform. It is
8307not a value, just a keyword. The possible condition codes are:
8308
8309#. ``false``: no comparison, always returns false
8310#. ``oeq``: ordered and equal
8311#. ``ogt``: ordered and greater than
8312#. ``oge``: ordered and greater than or equal
8313#. ``olt``: ordered and less than
8314#. ``ole``: ordered and less than or equal
8315#. ``one``: ordered and not equal
8316#. ``ord``: ordered (no nans)
8317#. ``ueq``: unordered or equal
8318#. ``ugt``: unordered or greater than
8319#. ``uge``: unordered or greater than or equal
8320#. ``ult``: unordered or less than
8321#. ``ule``: unordered or less than or equal
8322#. ``une``: unordered or not equal
8323#. ``uno``: unordered (either nans)
8324#. ``true``: no comparison, always returns true
8325
8326*Ordered* means that neither operand is a QNAN while *unordered* means
8327that either operand may be a QNAN.
8328
8329Each of ``val1`` and ``val2`` arguments must be either a :ref:`floating
8330point <t_floating>` type or a :ref:`vector <t_vector>` of floating point
8331type. They must have identical types.
8332
8333Semantics:
8334""""""""""
8335
8336The '``fcmp``' instruction compares ``op1`` and ``op2`` according to the
8337condition code given as ``cond``. If the operands are vectors, then the
8338vectors are compared element by element. Each comparison performed
8339always yields an :ref:`i1 <t_integer>` result, as follows:
8340
8341#. ``false``: always yields ``false``, regardless of operands.
8342#. ``oeq``: yields ``true`` if both operands are not a QNAN and ``op1``
8343   is equal to ``op2``.
8344#. ``ogt``: yields ``true`` if both operands are not a QNAN and ``op1``
8345   is greater than ``op2``.
8346#. ``oge``: yields ``true`` if both operands are not a QNAN and ``op1``
8347   is greater than or equal to ``op2``.
8348#. ``olt``: yields ``true`` if both operands are not a QNAN and ``op1``
8349   is less than ``op2``.
8350#. ``ole``: yields ``true`` if both operands are not a QNAN and ``op1``
8351   is less than or equal to ``op2``.
8352#. ``one``: yields ``true`` if both operands are not a QNAN and ``op1``
8353   is not equal to ``op2``.
8354#. ``ord``: yields ``true`` if both operands are not a QNAN.
8355#. ``ueq``: yields ``true`` if either operand is a QNAN or ``op1`` is
8356   equal to ``op2``.
8357#. ``ugt``: yields ``true`` if either operand is a QNAN or ``op1`` is
8358   greater than ``op2``.
8359#. ``uge``: yields ``true`` if either operand is a QNAN or ``op1`` is
8360   greater than or equal to ``op2``.
8361#. ``ult``: yields ``true`` if either operand is a QNAN or ``op1`` is
8362   less than ``op2``.
8363#. ``ule``: yields ``true`` if either operand is a QNAN or ``op1`` is
8364   less than or equal to ``op2``.
8365#. ``une``: yields ``true`` if either operand is a QNAN or ``op1`` is
8366   not equal to ``op2``.
8367#. ``uno``: yields ``true`` if either operand is a QNAN.
8368#. ``true``: always yields ``true``, regardless of operands.
8369
8370The ``fcmp`` instruction can also optionally take any number of
8371:ref:`fast-math flags <fastmath>`, which are optimization hints to enable
8372otherwise unsafe floating point optimizations.
8373
8374Any set of fast-math flags are legal on an ``fcmp`` instruction, but the
8375only flags that have any effect on its semantics are those that allow
8376assumptions to be made about the values of input arguments; namely
8377``nnan``, ``ninf``, and ``nsz``. See :ref:`fastmath` for more information.
8378
8379Example:
8380""""""""
8381
8382.. code-block:: llvm
8383
8384      <result> = fcmp oeq float 4.0, 5.0    ; yields: result=false
8385      <result> = fcmp one float 4.0, 5.0    ; yields: result=true
8386      <result> = fcmp olt float 4.0, 5.0    ; yields: result=true
8387      <result> = fcmp ueq double 1.0, 2.0   ; yields: result=false
8388
8389.. _i_phi:
8390
8391'``phi``' Instruction
8392^^^^^^^^^^^^^^^^^^^^^
8393
8394Syntax:
8395"""""""
8396
8397::
8398
8399      <result> = phi <ty> [ <val0>, <label0>], ...
8400
8401Overview:
8402"""""""""
8403
8404The '``phi``' instruction is used to implement the φ node in the SSA
8405graph representing the function.
8406
8407Arguments:
8408""""""""""
8409
8410The type of the incoming values is specified with the first type field.
8411After this, the '``phi``' instruction takes a list of pairs as
8412arguments, with one pair for each predecessor basic block of the current
8413block. Only values of :ref:`first class <t_firstclass>` type may be used as
8414the value arguments to the PHI node. Only labels may be used as the
8415label arguments.
8416
8417There must be no non-phi instructions between the start of a basic block
8418and the PHI instructions: i.e. PHI instructions must be first in a basic
8419block.
8420
8421For the purposes of the SSA form, the use of each incoming value is
8422deemed to occur on the edge from the corresponding predecessor block to
8423the current block (but after any definition of an '``invoke``'
8424instruction's return value on the same edge).
8425
8426Semantics:
8427""""""""""
8428
8429At runtime, the '``phi``' instruction logically takes on the value
8430specified by the pair corresponding to the predecessor basic block that
8431executed just prior to the current block.
8432
8433Example:
8434""""""""
8435
8436.. code-block:: llvm
8437
8438    Loop:       ; Infinite loop that counts from 0 on up...
8439      %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
8440      %nextindvar = add i32 %indvar, 1
8441      br label %Loop
8442
8443.. _i_select:
8444
8445'``select``' Instruction
8446^^^^^^^^^^^^^^^^^^^^^^^^
8447
8448Syntax:
8449"""""""
8450
8451::
8452
8453      <result> = select selty <cond>, <ty> <val1>, <ty> <val2>             ; yields ty
8454
8455      selty is either i1 or {<N x i1>}
8456
8457Overview:
8458"""""""""
8459
8460The '``select``' instruction is used to choose one value based on a
8461condition, without IR-level branching.
8462
8463Arguments:
8464""""""""""
8465
8466The '``select``' instruction requires an 'i1' value or a vector of 'i1'
8467values indicating the condition, and two values of the same :ref:`first
8468class <t_firstclass>` type.
8469
8470Semantics:
8471""""""""""
8472
8473If the condition is an i1 and it evaluates to 1, the instruction returns
8474the first value argument; otherwise, it returns the second value
8475argument.
8476
8477If the condition is a vector of i1, then the value arguments must be
8478vectors of the same size, and the selection is done element by element.
8479
8480If the condition is an i1 and the value arguments are vectors of the
8481same size, then an entire vector is selected.
8482
8483Example:
8484""""""""
8485
8486.. code-block:: llvm
8487
8488      %X = select i1 true, i8 17, i8 42          ; yields i8:17
8489
8490.. _i_call:
8491
8492'``call``' Instruction
8493^^^^^^^^^^^^^^^^^^^^^^
8494
8495Syntax:
8496"""""""
8497
8498::
8499
8500      <result> = [tail | musttail | notail ] call [fast-math flags] [cconv] [ret attrs] <ty>|<fnty> <fnptrval>(<function args>) [fn attrs]
8501                   [ operand bundles ]
8502
8503Overview:
8504"""""""""
8505
8506The '``call``' instruction represents a simple function call.
8507
8508Arguments:
8509""""""""""
8510
8511This instruction requires several arguments:
8512
8513#. The optional ``tail`` and ``musttail`` markers indicate that the optimizers
8514   should perform tail call optimization. The ``tail`` marker is a hint that
8515   `can be ignored <CodeGenerator.html#sibcallopt>`_. The ``musttail`` marker
8516   means that the call must be tail call optimized in order for the program to
8517   be correct. The ``musttail`` marker provides these guarantees:
8518
8519   #. The call will not cause unbounded stack growth if it is part of a
8520      recursive cycle in the call graph.
8521   #. Arguments with the :ref:`inalloca <attr_inalloca>` attribute are
8522      forwarded in place.
8523
8524   Both markers imply that the callee does not access allocas or varargs from
8525   the caller. Calls marked ``musttail`` must obey the following additional
8526   rules:
8527
8528   - The call must immediately precede a :ref:`ret <i_ret>` instruction,
8529     or a pointer bitcast followed by a ret instruction.
8530   - The ret instruction must return the (possibly bitcasted) value
8531     produced by the call or void.
8532   - The caller and callee prototypes must match. Pointer types of
8533     parameters or return types may differ in pointee type, but not
8534     in address space.
8535   - The calling conventions of the caller and callee must match.
8536   - All ABI-impacting function attributes, such as sret, byval, inreg,
8537     returned, and inalloca, must match.
8538   - The callee must be varargs iff the caller is varargs. Bitcasting a
8539     non-varargs function to the appropriate varargs type is legal so
8540     long as the non-varargs prefixes obey the other rules.
8541
8542   Tail call optimization for calls marked ``tail`` is guaranteed to occur if
8543   the following conditions are met:
8544
8545   -  Caller and callee both have the calling convention ``fastcc``.
8546   -  The call is in tail position (ret immediately follows call and ret
8547      uses value of call or is void).
8548   -  Option ``-tailcallopt`` is enabled, or
8549      ``llvm::GuaranteedTailCallOpt`` is ``true``.
8550   -  `Platform-specific constraints are
8551      met. <CodeGenerator.html#tailcallopt>`_
8552
8553#. The optional ``notail`` marker indicates that the optimizers should not add
8554   ``tail`` or ``musttail`` markers to the call. It is used to prevent tail
8555   call optimization from being performed on the call.
8556
8557#. The optional ``fast-math flags`` marker indicates that the call has one or more
8558   :ref:`fast-math flags <fastmath>`, which are optimization hints to enable
8559   otherwise unsafe floating-point optimizations. Fast-math flags are only valid
8560   for calls that return a floating-point scalar or vector type.
8561
8562#. The optional "cconv" marker indicates which :ref:`calling
8563   convention <callingconv>` the call should use. If none is
8564   specified, the call defaults to using C calling conventions. The
8565   calling convention of the call must match the calling convention of
8566   the target function, or else the behavior is undefined.
8567#. The optional :ref:`Parameter Attributes <paramattrs>` list for return
8568   values. Only '``zeroext``', '``signext``', and '``inreg``' attributes
8569   are valid here.
8570#. '``ty``': the type of the call instruction itself which is also the
8571   type of the return value. Functions that return no value are marked
8572   ``void``.
8573#. '``fnty``': shall be the signature of the function being called. The
8574   argument types must match the types implied by this signature. This
8575   type can be omitted if the function is not varargs.
8576#. '``fnptrval``': An LLVM value containing a pointer to a function to
8577   be called. In most cases, this is a direct function call, but
8578   indirect ``call``'s are just as possible, calling an arbitrary pointer
8579   to function value.
8580#. '``function args``': argument list whose types match the function
8581   signature argument types and parameter attributes. All arguments must
8582   be of :ref:`first class <t_firstclass>` type. If the function signature
8583   indicates the function accepts a variable number of arguments, the
8584   extra arguments can be specified.
8585#. The optional :ref:`function attributes <fnattrs>` list. Only
8586   '``noreturn``', '``nounwind``', '``readonly``' , '``readnone``',
8587   and '``convergent``' attributes are valid here.
8588#. The optional :ref:`operand bundles <opbundles>` list.
8589
8590Semantics:
8591""""""""""
8592
8593The '``call``' instruction is used to cause control flow to transfer to
8594a specified function, with its incoming arguments bound to the specified
8595values. Upon a '``ret``' instruction in the called function, control
8596flow continues with the instruction after the function call, and the
8597return value of the function is bound to the result argument.
8598
8599Example:
8600""""""""
8601
8602.. code-block:: llvm
8603
8604      %retval = call i32 @test(i32 %argc)
8605      call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42)        ; yields i32
8606      %X = tail call i32 @foo()                                    ; yields i32
8607      %Y = tail call fastcc i32 @foo()  ; yields i32
8608      call void %foo(i8 97 signext)
8609
8610      %struct.A = type { i32, i8 }
8611      %r = call %struct.A @foo()                        ; yields { i32, i8 }
8612      %gr = extractvalue %struct.A %r, 0                ; yields i32
8613      %gr1 = extractvalue %struct.A %r, 1               ; yields i8
8614      %Z = call void @foo() noreturn                    ; indicates that %foo never returns normally
8615      %ZZ = call zeroext i32 @bar()                     ; Return value is %zero extended
8616
8617llvm treats calls to some functions with names and arguments that match
8618the standard C99 library as being the C99 library functions, and may
8619perform optimizations or generate code for them under that assumption.
8620This is something we'd like to change in the future to provide better
8621support for freestanding environments and non-C-based languages.
8622
8623.. _i_va_arg:
8624
8625'``va_arg``' Instruction
8626^^^^^^^^^^^^^^^^^^^^^^^^
8627
8628Syntax:
8629"""""""
8630
8631::
8632
8633      <resultval> = va_arg <va_list*> <arglist>, <argty>
8634
8635Overview:
8636"""""""""
8637
8638The '``va_arg``' instruction is used to access arguments passed through
8639the "variable argument" area of a function call. It is used to implement
8640the ``va_arg`` macro in C.
8641
8642Arguments:
8643""""""""""
8644
8645This instruction takes a ``va_list*`` value and the type of the
8646argument. It returns a value of the specified argument type and
8647increments the ``va_list`` to point to the next argument. The actual
8648type of ``va_list`` is target specific.
8649
8650Semantics:
8651""""""""""
8652
8653The '``va_arg``' instruction loads an argument of the specified type
8654from the specified ``va_list`` and causes the ``va_list`` to point to
8655the next argument. For more information, see the variable argument
8656handling :ref:`Intrinsic Functions <int_varargs>`.
8657
8658It is legal for this instruction to be called in a function which does
8659not take a variable number of arguments, for example, the ``vfprintf``
8660function.
8661
8662``va_arg`` is an LLVM instruction instead of an :ref:`intrinsic
8663function <intrinsics>` because it takes a type as an argument.
8664
8665Example:
8666""""""""
8667
8668See the :ref:`variable argument processing <int_varargs>` section.
8669
8670Note that the code generator does not yet fully support va\_arg on many
8671targets. Also, it does not currently support va\_arg with aggregate
8672types on any target.
8673
8674.. _i_landingpad:
8675
8676'``landingpad``' Instruction
8677^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8678
8679Syntax:
8680"""""""
8681
8682::
8683
8684      <resultval> = landingpad <resultty> <clause>+
8685      <resultval> = landingpad <resultty> cleanup <clause>*
8686
8687      <clause> := catch <type> <value>
8688      <clause> := filter <array constant type> <array constant>
8689
8690Overview:
8691"""""""""
8692
8693The '``landingpad``' instruction is used by `LLVM's exception handling
8694system <ExceptionHandling.html#overview>`_ to specify that a basic block
8695is a landing pad --- one where the exception lands, and corresponds to the
8696code found in the ``catch`` portion of a ``try``/``catch`` sequence. It
8697defines values supplied by the :ref:`personality function <personalityfn>` upon
8698re-entry to the function. The ``resultval`` has the type ``resultty``.
8699
8700Arguments:
8701""""""""""
8702
8703The optional
8704``cleanup`` flag indicates that the landing pad block is a cleanup.
8705
8706A ``clause`` begins with the clause type --- ``catch`` or ``filter`` --- and
8707contains the global variable representing the "type" that may be caught
8708or filtered respectively. Unlike the ``catch`` clause, the ``filter``
8709clause takes an array constant as its argument. Use
8710"``[0 x i8**] undef``" for a filter which cannot throw. The
8711'``landingpad``' instruction must contain *at least* one ``clause`` or
8712the ``cleanup`` flag.
8713
8714Semantics:
8715""""""""""
8716
8717The '``landingpad``' instruction defines the values which are set by the
8718:ref:`personality function <personalityfn>` upon re-entry to the function, and
8719therefore the "result type" of the ``landingpad`` instruction. As with
8720calling conventions, how the personality function results are
8721represented in LLVM IR is target specific.
8722
8723The clauses are applied in order from top to bottom. If two
8724``landingpad`` instructions are merged together through inlining, the
8725clauses from the calling function are appended to the list of clauses.
8726When the call stack is being unwound due to an exception being thrown,
8727the exception is compared against each ``clause`` in turn. If it doesn't
8728match any of the clauses, and the ``cleanup`` flag is not set, then
8729unwinding continues further up the call stack.
8730
8731The ``landingpad`` instruction has several restrictions:
8732
8733-  A landing pad block is a basic block which is the unwind destination
8734   of an '``invoke``' instruction.
8735-  A landing pad block must have a '``landingpad``' instruction as its
8736   first non-PHI instruction.
8737-  There can be only one '``landingpad``' instruction within the landing
8738   pad block.
8739-  A basic block that is not a landing pad block may not include a
8740   '``landingpad``' instruction.
8741
8742Example:
8743""""""""
8744
8745.. code-block:: llvm
8746
8747      ;; A landing pad which can catch an integer.
8748      %res = landingpad { i8*, i32 }
8749               catch i8** @_ZTIi
8750      ;; A landing pad that is a cleanup.
8751      %res = landingpad { i8*, i32 }
8752               cleanup
8753      ;; A landing pad which can catch an integer and can only throw a double.
8754      %res = landingpad { i8*, i32 }
8755               catch i8** @_ZTIi
8756               filter [1 x i8**] [@_ZTId]
8757
8758.. _i_catchpad:
8759
8760'``catchpad``' Instruction
8761^^^^^^^^^^^^^^^^^^^^^^^^^^
8762
8763Syntax:
8764"""""""
8765
8766::
8767
8768      <resultval> = catchpad within <catchswitch> [<args>*]
8769
8770Overview:
8771"""""""""
8772
8773The '``catchpad``' instruction is used by `LLVM's exception handling
8774system <ExceptionHandling.html#overview>`_ to specify that a basic block
8775begins a catch handler --- one where a personality routine attempts to transfer
8776control to catch an exception.
8777
8778Arguments:
8779""""""""""
8780
8781The ``catchswitch`` operand must always be a token produced by a
8782:ref:`catchswitch <i_catchswitch>` instruction in a predecessor block. This
8783ensures that each ``catchpad`` has exactly one predecessor block, and it always
8784terminates in a ``catchswitch``.
8785
8786The ``args`` correspond to whatever information the personality routine
8787requires to know if this is an appropriate handler for the exception. Control
8788will transfer to the ``catchpad`` if this is the first appropriate handler for
8789the exception.
8790
8791The ``resultval`` has the type :ref:`token <t_token>` and is used to match the
8792``catchpad`` to corresponding :ref:`catchrets <i_catchret>` and other nested EH
8793pads.
8794
8795Semantics:
8796""""""""""
8797
8798When the call stack is being unwound due to an exception being thrown, the
8799exception is compared against the ``args``. If it doesn't match, control will
8800not reach the ``catchpad`` instruction.  The representation of ``args`` is
8801entirely target and personality function-specific.
8802
8803Like the :ref:`landingpad <i_landingpad>` instruction, the ``catchpad``
8804instruction must be the first non-phi of its parent basic block.
8805
8806The meaning of the tokens produced and consumed by ``catchpad`` and other "pad"
8807instructions is described in the
8808`Windows exception handling documentation\ <ExceptionHandling.html#wineh>`_.
8809
8810When a ``catchpad`` has been "entered" but not yet "exited" (as
8811described in the `EH documentation\ <ExceptionHandling.html#wineh-constraints>`_),
8812it is undefined behavior to execute a :ref:`call <i_call>` or :ref:`invoke <i_invoke>`
8813that does not carry an appropriate :ref:`"funclet" bundle <ob_funclet>`.
8814
8815Example:
8816""""""""
8817
8818.. code-block:: llvm
8819
8820    dispatch:
8821      %cs = catchswitch within none [label %handler0] unwind to caller
8822      ;; A catch block which can catch an integer.
8823    handler0:
8824      %tok = catchpad within %cs [i8** @_ZTIi]
8825
8826.. _i_cleanuppad:
8827
8828'``cleanuppad``' Instruction
8829^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8830
8831Syntax:
8832"""""""
8833
8834::
8835
8836      <resultval> = cleanuppad within <parent> [<args>*]
8837
8838Overview:
8839"""""""""
8840
8841The '``cleanuppad``' instruction is used by `LLVM's exception handling
8842system <ExceptionHandling.html#overview>`_ to specify that a basic block
8843is a cleanup block --- one where a personality routine attempts to
8844transfer control to run cleanup actions.
8845The ``args`` correspond to whatever additional
8846information the :ref:`personality function <personalityfn>` requires to
8847execute the cleanup.
8848The ``resultval`` has the type :ref:`token <t_token>` and is used to
8849match the ``cleanuppad`` to corresponding :ref:`cleanuprets <i_cleanupret>`.
8850The ``parent`` argument is the token of the funclet that contains the
8851``cleanuppad`` instruction. If the ``cleanuppad`` is not inside a funclet,
8852this operand may be the token ``none``.
8853
8854Arguments:
8855""""""""""
8856
8857The instruction takes a list of arbitrary values which are interpreted
8858by the :ref:`personality function <personalityfn>`.
8859
8860Semantics:
8861""""""""""
8862
8863When the call stack is being unwound due to an exception being thrown,
8864the :ref:`personality function <personalityfn>` transfers control to the
8865``cleanuppad`` with the aid of the personality-specific arguments.
8866As with calling conventions, how the personality function results are
8867represented in LLVM IR is target specific.
8868
8869The ``cleanuppad`` instruction has several restrictions:
8870
8871-  A cleanup block is a basic block which is the unwind destination of
8872   an exceptional instruction.
8873-  A cleanup block must have a '``cleanuppad``' instruction as its
8874   first non-PHI instruction.
8875-  There can be only one '``cleanuppad``' instruction within the
8876   cleanup block.
8877-  A basic block that is not a cleanup block may not include a
8878   '``cleanuppad``' instruction.
8879
8880When a ``cleanuppad`` has been "entered" but not yet "exited" (as
8881described in the `EH documentation\ <ExceptionHandling.html#wineh-constraints>`_),
8882it is undefined behavior to execute a :ref:`call <i_call>` or :ref:`invoke <i_invoke>`
8883that does not carry an appropriate :ref:`"funclet" bundle <ob_funclet>`.
8884
8885Example:
8886""""""""
8887
8888.. code-block:: llvm
8889
8890      %tok = cleanuppad within %cs []
8891
8892.. _intrinsics:
8893
8894Intrinsic Functions
8895===================
8896
8897LLVM supports the notion of an "intrinsic function". These functions
8898have well known names and semantics and are required to follow certain
8899restrictions. Overall, these intrinsics represent an extension mechanism
8900for the LLVM language that does not require changing all of the
8901transformations in LLVM when adding to the language (or the bitcode
8902reader/writer, the parser, etc...).
8903
8904Intrinsic function names must all start with an "``llvm.``" prefix. This
8905prefix is reserved in LLVM for intrinsic names; thus, function names may
8906not begin with this prefix. Intrinsic functions must always be external
8907functions: you cannot define the body of intrinsic functions. Intrinsic
8908functions may only be used in call or invoke instructions: it is illegal
8909to take the address of an intrinsic function. Additionally, because
8910intrinsic functions are part of the LLVM language, it is required if any
8911are added that they be documented here.
8912
8913Some intrinsic functions can be overloaded, i.e., the intrinsic
8914represents a family of functions that perform the same operation but on
8915different data types. Because LLVM can represent over 8 million
8916different integer types, overloading is used commonly to allow an
8917intrinsic function to operate on any integer type. One or more of the
8918argument types or the result type can be overloaded to accept any
8919integer type. Argument types may also be defined as exactly matching a
8920previous argument's type or the result type. This allows an intrinsic
8921function which accepts multiple arguments, but needs all of them to be
8922of the same type, to only be overloaded with respect to a single
8923argument or the result.
8924
8925Overloaded intrinsics will have the names of its overloaded argument
8926types encoded into its function name, each preceded by a period. Only
8927those types which are overloaded result in a name suffix. Arguments
8928whose type is matched against another type do not. For example, the
8929``llvm.ctpop`` function can take an integer of any width and returns an
8930integer of exactly the same integer width. This leads to a family of
8931functions such as ``i8 @llvm.ctpop.i8(i8 %val)`` and
8932``i29 @llvm.ctpop.i29(i29 %val)``. Only one type, the return type, is
8933overloaded, and only one type suffix is required. Because the argument's
8934type is matched against the return type, it does not require its own
8935name suffix.
8936
8937To learn how to add an intrinsic function, please see the `Extending
8938LLVM Guide <ExtendingLLVM.html>`_.
8939
8940.. _int_varargs:
8941
8942Variable Argument Handling Intrinsics
8943-------------------------------------
8944
8945Variable argument support is defined in LLVM with the
8946:ref:`va_arg <i_va_arg>` instruction and these three intrinsic
8947functions. These functions are related to the similarly named macros
8948defined in the ``<stdarg.h>`` header file.
8949
8950All of these functions operate on arguments that use a target-specific
8951value type "``va_list``". The LLVM assembly language reference manual
8952does not define what this type is, so all transformations should be
8953prepared to handle these functions regardless of the type used.
8954
8955This example shows how the :ref:`va_arg <i_va_arg>` instruction and the
8956variable argument handling intrinsic functions are used.
8957
8958.. code-block:: llvm
8959
8960    ; This struct is different for every platform. For most platforms,
8961    ; it is merely an i8*.
8962    %struct.va_list = type { i8* }
8963
8964    ; For Unix x86_64 platforms, va_list is the following struct:
8965    ; %struct.va_list = type { i32, i32, i8*, i8* }
8966
8967    define i32 @test(i32 %X, ...) {
8968      ; Initialize variable argument processing
8969      %ap = alloca %struct.va_list
8970      %ap2 = bitcast %struct.va_list* %ap to i8*
8971      call void @llvm.va_start(i8* %ap2)
8972
8973      ; Read a single integer argument
8974      %tmp = va_arg i8* %ap2, i32
8975
8976      ; Demonstrate usage of llvm.va_copy and llvm.va_end
8977      %aq = alloca i8*
8978      %aq2 = bitcast i8** %aq to i8*
8979      call void @llvm.va_copy(i8* %aq2, i8* %ap2)
8980      call void @llvm.va_end(i8* %aq2)
8981
8982      ; Stop processing of arguments.
8983      call void @llvm.va_end(i8* %ap2)
8984      ret i32 %tmp
8985    }
8986
8987    declare void @llvm.va_start(i8*)
8988    declare void @llvm.va_copy(i8*, i8*)
8989    declare void @llvm.va_end(i8*)
8990
8991.. _int_va_start:
8992
8993'``llvm.va_start``' Intrinsic
8994^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8995
8996Syntax:
8997"""""""
8998
8999::
9000
9001      declare void @llvm.va_start(i8* <arglist>)
9002
9003Overview:
9004"""""""""
9005
9006The '``llvm.va_start``' intrinsic initializes ``*<arglist>`` for
9007subsequent use by ``va_arg``.
9008
9009Arguments:
9010""""""""""
9011
9012The argument is a pointer to a ``va_list`` element to initialize.
9013
9014Semantics:
9015""""""""""
9016
9017The '``llvm.va_start``' intrinsic works just like the ``va_start`` macro
9018available in C. In a target-dependent way, it initializes the
9019``va_list`` element to which the argument points, so that the next call
9020to ``va_arg`` will produce the first variable argument passed to the
9021function. Unlike the C ``va_start`` macro, this intrinsic does not need
9022to know the last argument of the function as the compiler can figure
9023that out.
9024
9025'``llvm.va_end``' Intrinsic
9026^^^^^^^^^^^^^^^^^^^^^^^^^^^
9027
9028Syntax:
9029"""""""
9030
9031::
9032
9033      declare void @llvm.va_end(i8* <arglist>)
9034
9035Overview:
9036"""""""""
9037
9038The '``llvm.va_end``' intrinsic destroys ``*<arglist>``, which has been
9039initialized previously with ``llvm.va_start`` or ``llvm.va_copy``.
9040
9041Arguments:
9042""""""""""
9043
9044The argument is a pointer to a ``va_list`` to destroy.
9045
9046Semantics:
9047""""""""""
9048
9049The '``llvm.va_end``' intrinsic works just like the ``va_end`` macro
9050available in C. In a target-dependent way, it destroys the ``va_list``
9051element to which the argument points. Calls to
9052:ref:`llvm.va_start <int_va_start>` and
9053:ref:`llvm.va_copy <int_va_copy>` must be matched exactly with calls to
9054``llvm.va_end``.
9055
9056.. _int_va_copy:
9057
9058'``llvm.va_copy``' Intrinsic
9059^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9060
9061Syntax:
9062"""""""
9063
9064::
9065
9066      declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
9067
9068Overview:
9069"""""""""
9070
9071The '``llvm.va_copy``' intrinsic copies the current argument position
9072from the source argument list to the destination argument list.
9073
9074Arguments:
9075""""""""""
9076
9077The first argument is a pointer to a ``va_list`` element to initialize.
9078The second argument is a pointer to a ``va_list`` element to copy from.
9079
9080Semantics:
9081""""""""""
9082
9083The '``llvm.va_copy``' intrinsic works just like the ``va_copy`` macro
9084available in C. In a target-dependent way, it copies the source
9085``va_list`` element into the destination ``va_list`` element. This
9086intrinsic is necessary because the `` llvm.va_start`` intrinsic may be
9087arbitrarily complex and require, for example, memory allocation.
9088
9089Accurate Garbage Collection Intrinsics
9090--------------------------------------
9091
9092LLVM's support for `Accurate Garbage Collection <GarbageCollection.html>`_
9093(GC) requires the frontend to generate code containing appropriate intrinsic
9094calls and select an appropriate GC strategy which knows how to lower these
9095intrinsics in a manner which is appropriate for the target collector.
9096
9097These intrinsics allow identification of :ref:`GC roots on the
9098stack <int_gcroot>`, as well as garbage collector implementations that
9099require :ref:`read <int_gcread>` and :ref:`write <int_gcwrite>` barriers.
9100Frontends for type-safe garbage collected languages should generate
9101these intrinsics to make use of the LLVM garbage collectors. For more
9102details, see `Garbage Collection with LLVM <GarbageCollection.html>`_.
9103
9104Experimental Statepoint Intrinsics
9105^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9106
9107LLVM provides an second experimental set of intrinsics for describing garbage
9108collection safepoints in compiled code. These intrinsics are an alternative
9109to the ``llvm.gcroot`` intrinsics, but are compatible with the ones for
9110:ref:`read <int_gcread>` and :ref:`write <int_gcwrite>` barriers. The
9111differences in approach are covered in the `Garbage Collection with LLVM
9112<GarbageCollection.html>`_ documentation. The intrinsics themselves are
9113described in :doc:`Statepoints`.
9114
9115.. _int_gcroot:
9116
9117'``llvm.gcroot``' Intrinsic
9118^^^^^^^^^^^^^^^^^^^^^^^^^^^
9119
9120Syntax:
9121"""""""
9122
9123::
9124
9125      declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
9126
9127Overview:
9128"""""""""
9129
9130The '``llvm.gcroot``' intrinsic declares the existence of a GC root to
9131the code generator, and allows some metadata to be associated with it.
9132
9133Arguments:
9134""""""""""
9135
9136The first argument specifies the address of a stack object that contains
9137the root pointer. The second pointer (which must be either a constant or
9138a global value address) contains the meta-data to be associated with the
9139root.
9140
9141Semantics:
9142""""""""""
9143
9144At runtime, a call to this intrinsic stores a null pointer into the
9145"ptrloc" location. At compile-time, the code generator generates
9146information to allow the runtime to find the pointer at GC safe points.
9147The '``llvm.gcroot``' intrinsic may only be used in a function which
9148:ref:`specifies a GC algorithm <gc>`.
9149
9150.. _int_gcread:
9151
9152'``llvm.gcread``' Intrinsic
9153^^^^^^^^^^^^^^^^^^^^^^^^^^^
9154
9155Syntax:
9156"""""""
9157
9158::
9159
9160      declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
9161
9162Overview:
9163"""""""""
9164
9165The '``llvm.gcread``' intrinsic identifies reads of references from heap
9166locations, allowing garbage collector implementations that require read
9167barriers.
9168
9169Arguments:
9170""""""""""
9171
9172The second argument is the address to read from, which should be an
9173address allocated from the garbage collector. The first object is a
9174pointer to the start of the referenced object, if needed by the language
9175runtime (otherwise null).
9176
9177Semantics:
9178""""""""""
9179
9180The '``llvm.gcread``' intrinsic has the same semantics as a load
9181instruction, but may be replaced with substantially more complex code by
9182the garbage collector runtime, as needed. The '``llvm.gcread``'
9183intrinsic may only be used in a function which :ref:`specifies a GC
9184algorithm <gc>`.
9185
9186.. _int_gcwrite:
9187
9188'``llvm.gcwrite``' Intrinsic
9189^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9190
9191Syntax:
9192"""""""
9193
9194::
9195
9196      declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
9197
9198Overview:
9199"""""""""
9200
9201The '``llvm.gcwrite``' intrinsic identifies writes of references to heap
9202locations, allowing garbage collector implementations that require write
9203barriers (such as generational or reference counting collectors).
9204
9205Arguments:
9206""""""""""
9207
9208The first argument is the reference to store, the second is the start of
9209the object to store it to, and the third is the address of the field of
9210Obj to store to. If the runtime does not require a pointer to the
9211object, Obj may be null.
9212
9213Semantics:
9214""""""""""
9215
9216The '``llvm.gcwrite``' intrinsic has the same semantics as a store
9217instruction, but may be replaced with substantially more complex code by
9218the garbage collector runtime, as needed. The '``llvm.gcwrite``'
9219intrinsic may only be used in a function which :ref:`specifies a GC
9220algorithm <gc>`.
9221
9222Code Generator Intrinsics
9223-------------------------
9224
9225These intrinsics are provided by LLVM to expose special features that
9226may only be implemented with code generator support.
9227
9228'``llvm.returnaddress``' Intrinsic
9229^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9230
9231Syntax:
9232"""""""
9233
9234::
9235
9236      declare i8  *@llvm.returnaddress(i32 <level>)
9237
9238Overview:
9239"""""""""
9240
9241The '``llvm.returnaddress``' intrinsic attempts to compute a
9242target-specific value indicating the return address of the current
9243function or one of its callers.
9244
9245Arguments:
9246""""""""""
9247
9248The argument to this intrinsic indicates which function to return the
9249address for. Zero indicates the calling function, one indicates its
9250caller, etc. The argument is **required** to be a constant integer
9251value.
9252
9253Semantics:
9254""""""""""
9255
9256The '``llvm.returnaddress``' intrinsic either returns a pointer
9257indicating the return address of the specified call frame, or zero if it
9258cannot be identified. The value returned by this intrinsic is likely to
9259be incorrect or 0 for arguments other than zero, so it should only be
9260used for debugging purposes.
9261
9262Note that calling this intrinsic does not prevent function inlining or
9263other aggressive transformations, so the value returned may not be that
9264of the obvious source-language caller.
9265
9266'``llvm.frameaddress``' Intrinsic
9267^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9268
9269Syntax:
9270"""""""
9271
9272::
9273
9274      declare i8* @llvm.frameaddress(i32 <level>)
9275
9276Overview:
9277"""""""""
9278
9279The '``llvm.frameaddress``' intrinsic attempts to return the
9280target-specific frame pointer value for the specified stack frame.
9281
9282Arguments:
9283""""""""""
9284
9285The argument to this intrinsic indicates which function to return the
9286frame pointer for. Zero indicates the calling function, one indicates
9287its caller, etc. The argument is **required** to be a constant integer
9288value.
9289
9290Semantics:
9291""""""""""
9292
9293The '``llvm.frameaddress``' intrinsic either returns a pointer
9294indicating the frame address of the specified call frame, or zero if it
9295cannot be identified. The value returned by this intrinsic is likely to
9296be incorrect or 0 for arguments other than zero, so it should only be
9297used for debugging purposes.
9298
9299Note that calling this intrinsic does not prevent function inlining or
9300other aggressive transformations, so the value returned may not be that
9301of the obvious source-language caller.
9302
9303'``llvm.localescape``' and '``llvm.localrecover``' Intrinsics
9304^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9305
9306Syntax:
9307"""""""
9308
9309::
9310
9311      declare void @llvm.localescape(...)
9312      declare i8* @llvm.localrecover(i8* %func, i8* %fp, i32 %idx)
9313
9314Overview:
9315"""""""""
9316
9317The '``llvm.localescape``' intrinsic escapes offsets of a collection of static
9318allocas, and the '``llvm.localrecover``' intrinsic applies those offsets to a
9319live frame pointer to recover the address of the allocation. The offset is
9320computed during frame layout of the caller of ``llvm.localescape``.
9321
9322Arguments:
9323""""""""""
9324
9325All arguments to '``llvm.localescape``' must be pointers to static allocas or
9326casts of static allocas. Each function can only call '``llvm.localescape``'
9327once, and it can only do so from the entry block.
9328
9329The ``func`` argument to '``llvm.localrecover``' must be a constant
9330bitcasted pointer to a function defined in the current module. The code
9331generator cannot determine the frame allocation offset of functions defined in
9332other modules.
9333
9334The ``fp`` argument to '``llvm.localrecover``' must be a frame pointer of a
9335call frame that is currently live. The return value of '``llvm.localaddress``'
9336is one way to produce such a value, but various runtimes also expose a suitable
9337pointer in platform-specific ways.
9338
9339The ``idx`` argument to '``llvm.localrecover``' indicates which alloca passed to
9340'``llvm.localescape``' to recover. It is zero-indexed.
9341
9342Semantics:
9343""""""""""
9344
9345These intrinsics allow a group of functions to share access to a set of local
9346stack allocations of a one parent function. The parent function may call the
9347'``llvm.localescape``' intrinsic once from the function entry block, and the
9348child functions can use '``llvm.localrecover``' to access the escaped allocas.
9349The '``llvm.localescape``' intrinsic blocks inlining, as inlining changes where
9350the escaped allocas are allocated, which would break attempts to use
9351'``llvm.localrecover``'.
9352
9353.. _int_read_register:
9354.. _int_write_register:
9355
9356'``llvm.read_register``' and '``llvm.write_register``' Intrinsics
9357^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9358
9359Syntax:
9360"""""""
9361
9362::
9363
9364      declare i32 @llvm.read_register.i32(metadata)
9365      declare i64 @llvm.read_register.i64(metadata)
9366      declare void @llvm.write_register.i32(metadata, i32 @value)
9367      declare void @llvm.write_register.i64(metadata, i64 @value)
9368      !0 = !{!"sp\00"}
9369
9370Overview:
9371"""""""""
9372
9373The '``llvm.read_register``' and '``llvm.write_register``' intrinsics
9374provides access to the named register. The register must be valid on
9375the architecture being compiled to. The type needs to be compatible
9376with the register being read.
9377
9378Semantics:
9379""""""""""
9380
9381The '``llvm.read_register``' intrinsic returns the current value of the
9382register, where possible. The '``llvm.write_register``' intrinsic sets
9383the current value of the register, where possible.
9384
9385This is useful to implement named register global variables that need
9386to always be mapped to a specific register, as is common practice on
9387bare-metal programs including OS kernels.
9388
9389The compiler doesn't check for register availability or use of the used
9390register in surrounding code, including inline assembly. Because of that,
9391allocatable registers are not supported.
9392
9393Warning: So far it only works with the stack pointer on selected
9394architectures (ARM, AArch64, PowerPC and x86_64). Significant amount of
9395work is needed to support other registers and even more so, allocatable
9396registers.
9397
9398.. _int_stacksave:
9399
9400'``llvm.stacksave``' Intrinsic
9401^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9402
9403Syntax:
9404"""""""
9405
9406::
9407
9408      declare i8* @llvm.stacksave()
9409
9410Overview:
9411"""""""""
9412
9413The '``llvm.stacksave``' intrinsic is used to remember the current state
9414of the function stack, for use with
9415:ref:`llvm.stackrestore <int_stackrestore>`. This is useful for
9416implementing language features like scoped automatic variable sized
9417arrays in C99.
9418
9419Semantics:
9420""""""""""
9421
9422This intrinsic returns a opaque pointer value that can be passed to
9423:ref:`llvm.stackrestore <int_stackrestore>`. When an
9424``llvm.stackrestore`` intrinsic is executed with a value saved from
9425``llvm.stacksave``, it effectively restores the state of the stack to
9426the state it was in when the ``llvm.stacksave`` intrinsic executed. In
9427practice, this pops any :ref:`alloca <i_alloca>` blocks from the stack that
9428were allocated after the ``llvm.stacksave`` was executed.
9429
9430.. _int_stackrestore:
9431
9432'``llvm.stackrestore``' Intrinsic
9433^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9434
9435Syntax:
9436"""""""
9437
9438::
9439
9440      declare void @llvm.stackrestore(i8* %ptr)
9441
9442Overview:
9443"""""""""
9444
9445The '``llvm.stackrestore``' intrinsic is used to restore the state of
9446the function stack to the state it was in when the corresponding
9447:ref:`llvm.stacksave <int_stacksave>` intrinsic executed. This is
9448useful for implementing language features like scoped automatic variable
9449sized arrays in C99.
9450
9451Semantics:
9452""""""""""
9453
9454See the description for :ref:`llvm.stacksave <int_stacksave>`.
9455
9456.. _int_get_dynamic_area_offset:
9457
9458'``llvm.get.dynamic.area.offset``' Intrinsic
9459^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9460
9461Syntax:
9462"""""""
9463
9464::
9465
9466      declare i32 @llvm.get.dynamic.area.offset.i32()
9467      declare i64 @llvm.get.dynamic.area.offset.i64()
9468
9469      Overview:
9470      """""""""
9471
9472      The '``llvm.get.dynamic.area.offset.*``' intrinsic family is used to
9473      get the offset from native stack pointer to the address of the most
9474      recent dynamic alloca on the caller's stack. These intrinsics are
9475      intendend for use in combination with
9476      :ref:`llvm.stacksave <int_stacksave>` to get a
9477      pointer to the most recent dynamic alloca. This is useful, for example,
9478      for AddressSanitizer's stack unpoisoning routines.
9479
9480Semantics:
9481""""""""""
9482
9483      These intrinsics return a non-negative integer value that can be used to
9484      get the address of the most recent dynamic alloca, allocated by :ref:`alloca <i_alloca>`
9485      on the caller's stack. In particular, for targets where stack grows downwards,
9486      adding this offset to the native stack pointer would get the address of the most
9487      recent dynamic alloca. For targets where stack grows upwards, the situation is a bit more
9488      complicated, because substracting this value from stack pointer would get the address
9489      one past the end of the most recent dynamic alloca.
9490
9491      Although for most targets `llvm.get.dynamic.area.offset <int_get_dynamic_area_offset>`
9492      returns just a zero, for others, such as PowerPC and PowerPC64, it returns a
9493      compile-time-known constant value.
9494
9495      The return value type of :ref:`llvm.get.dynamic.area.offset <int_get_dynamic_area_offset>`
9496      must match the target's generic address space's (address space 0) pointer type.
9497
9498'``llvm.prefetch``' Intrinsic
9499^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9500
9501Syntax:
9502"""""""
9503
9504::
9505
9506      declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
9507
9508Overview:
9509"""""""""
9510
9511The '``llvm.prefetch``' intrinsic is a hint to the code generator to
9512insert a prefetch instruction if supported; otherwise, it is a noop.
9513Prefetches have no effect on the behavior of the program but can change
9514its performance characteristics.
9515
9516Arguments:
9517""""""""""
9518
9519``address`` is the address to be prefetched, ``rw`` is the specifier
9520determining if the fetch should be for a read (0) or write (1), and
9521``locality`` is a temporal locality specifier ranging from (0) - no
9522locality, to (3) - extremely local keep in cache. The ``cache type``
9523specifies whether the prefetch is performed on the data (1) or
9524instruction (0) cache. The ``rw``, ``locality`` and ``cache type``
9525arguments must be constant integers.
9526
9527Semantics:
9528""""""""""
9529
9530This intrinsic does not modify the behavior of the program. In
9531particular, prefetches cannot trap and do not produce a value. On
9532targets that support this intrinsic, the prefetch can provide hints to
9533the processor cache for better performance.
9534
9535'``llvm.pcmarker``' Intrinsic
9536^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9537
9538Syntax:
9539"""""""
9540
9541::
9542
9543      declare void @llvm.pcmarker(i32 <id>)
9544
9545Overview:
9546"""""""""
9547
9548The '``llvm.pcmarker``' intrinsic is a method to export a Program
9549Counter (PC) in a region of code to simulators and other tools. The
9550method is target specific, but it is expected that the marker will use
9551exported symbols to transmit the PC of the marker. The marker makes no
9552guarantees that it will remain with any specific instruction after
9553optimizations. It is possible that the presence of a marker will inhibit
9554optimizations. The intended use is to be inserted after optimizations to
9555allow correlations of simulation runs.
9556
9557Arguments:
9558""""""""""
9559
9560``id`` is a numerical id identifying the marker.
9561
9562Semantics:
9563""""""""""
9564
9565This intrinsic does not modify the behavior of the program. Backends
9566that do not support this intrinsic may ignore it.
9567
9568'``llvm.readcyclecounter``' Intrinsic
9569^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9570
9571Syntax:
9572"""""""
9573
9574::
9575
9576      declare i64 @llvm.readcyclecounter()
9577
9578Overview:
9579"""""""""
9580
9581The '``llvm.readcyclecounter``' intrinsic provides access to the cycle
9582counter register (or similar low latency, high accuracy clocks) on those
9583targets that support it. On X86, it should map to RDTSC. On Alpha, it
9584should map to RPCC. As the backing counters overflow quickly (on the
9585order of 9 seconds on alpha), this should only be used for small
9586timings.
9587
9588Semantics:
9589""""""""""
9590
9591When directly supported, reading the cycle counter should not modify any
9592memory. Implementations are allowed to either return a application
9593specific value or a system wide value. On backends without support, this
9594is lowered to a constant 0.
9595
9596Note that runtime support may be conditional on the privilege-level code is
9597running at and the host platform.
9598
9599'``llvm.clear_cache``' Intrinsic
9600^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9601
9602Syntax:
9603"""""""
9604
9605::
9606
9607      declare void @llvm.clear_cache(i8*, i8*)
9608
9609Overview:
9610"""""""""
9611
9612The '``llvm.clear_cache``' intrinsic ensures visibility of modifications
9613in the specified range to the execution unit of the processor. On
9614targets with non-unified instruction and data cache, the implementation
9615flushes the instruction cache.
9616
9617Semantics:
9618""""""""""
9619
9620On platforms with coherent instruction and data caches (e.g. x86), this
9621intrinsic is a nop. On platforms with non-coherent instruction and data
9622cache (e.g. ARM, MIPS), the intrinsic is lowered either to appropriate
9623instructions or a system call, if cache flushing requires special
9624privileges.
9625
9626The default behavior is to emit a call to ``__clear_cache`` from the run
9627time library.
9628
9629This instrinsic does *not* empty the instruction pipeline. Modifications
9630of the current function are outside the scope of the intrinsic.
9631
9632'``llvm.instrprof_increment``' Intrinsic
9633^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9634
9635Syntax:
9636"""""""
9637
9638::
9639
9640      declare void @llvm.instrprof_increment(i8* <name>, i64 <hash>,
9641                                             i32 <num-counters>, i32 <index>)
9642
9643Overview:
9644"""""""""
9645
9646The '``llvm.instrprof_increment``' intrinsic can be emitted by a
9647frontend for use with instrumentation based profiling. These will be
9648lowered by the ``-instrprof`` pass to generate execution counts of a
9649program at runtime.
9650
9651Arguments:
9652""""""""""
9653
9654The first argument is a pointer to a global variable containing the
9655name of the entity being instrumented. This should generally be the
9656(mangled) function name for a set of counters.
9657
9658The second argument is a hash value that can be used by the consumer
9659of the profile data to detect changes to the instrumented source, and
9660the third is the number of counters associated with ``name``. It is an
9661error if ``hash`` or ``num-counters`` differ between two instances of
9662``instrprof_increment`` that refer to the same name.
9663
9664The last argument refers to which of the counters for ``name`` should
9665be incremented. It should be a value between 0 and ``num-counters``.
9666
9667Semantics:
9668""""""""""
9669
9670This intrinsic represents an increment of a profiling counter. It will
9671cause the ``-instrprof`` pass to generate the appropriate data
9672structures and the code to increment the appropriate value, in a
9673format that can be written out by a compiler runtime and consumed via
9674the ``llvm-profdata`` tool.
9675
9676'``llvm.instrprof_value_profile``' Intrinsic
9677^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9678
9679Syntax:
9680"""""""
9681
9682::
9683
9684      declare void @llvm.instrprof_value_profile(i8* <name>, i64 <hash>,
9685                                                 i64 <value>, i32 <value_kind>,
9686                                                 i32 <index>)
9687
9688Overview:
9689"""""""""
9690
9691The '``llvm.instrprof_value_profile``' intrinsic can be emitted by a
9692frontend for use with instrumentation based profiling. This will be
9693lowered by the ``-instrprof`` pass to find out the target values,
9694instrumented expressions take in a program at runtime.
9695
9696Arguments:
9697""""""""""
9698
9699The first argument is a pointer to a global variable containing the
9700name of the entity being instrumented. ``name`` should generally be the
9701(mangled) function name for a set of counters.
9702
9703The second argument is a hash value that can be used by the consumer
9704of the profile data to detect changes to the instrumented source. It
9705is an error if ``hash`` differs between two instances of
9706``llvm.instrprof_*`` that refer to the same name.
9707
9708The third argument is the value of the expression being profiled. The profiled
9709expression's value should be representable as an unsigned 64-bit value. The
9710fourth argument represents the kind of value profiling that is being done. The
9711supported value profiling kinds are enumerated through the
9712``InstrProfValueKind`` type declared in the
9713``<include/llvm/ProfileData/InstrProf.h>`` header file. The last argument is the
9714index of the instrumented expression within ``name``. It should be >= 0.
9715
9716Semantics:
9717""""""""""
9718
9719This intrinsic represents the point where a call to a runtime routine
9720should be inserted for value profiling of target expressions. ``-instrprof``
9721pass will generate the appropriate data structures and replace the
9722``llvm.instrprof_value_profile`` intrinsic with the call to the profile
9723runtime library with proper arguments.
9724
9725'``llvm.thread.pointer``' Intrinsic
9726^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9727
9728Syntax:
9729"""""""
9730
9731::
9732
9733      declare i8* @llvm.thread.pointer()
9734
9735Overview:
9736"""""""""
9737
9738The '``llvm.thread.pointer``' intrinsic returns the value of the thread
9739pointer.
9740
9741Semantics:
9742""""""""""
9743
9744The '``llvm.thread.pointer``' intrinsic returns a pointer to the TLS area
9745for the current thread.  The exact semantics of this value are target
9746specific: it may point to the start of TLS area, to the end, or somewhere
9747in the middle.  Depending on the target, this intrinsic may read a register,
9748call a helper function, read from an alternate memory space, or perform
9749other operations necessary to locate the TLS area.  Not all targets support
9750this intrinsic.
9751
9752Standard C Library Intrinsics
9753-----------------------------
9754
9755LLVM provides intrinsics for a few important standard C library
9756functions. These intrinsics allow source-language front-ends to pass
9757information about the alignment of the pointer arguments to the code
9758generator, providing opportunity for more efficient code generation.
9759
9760.. _int_memcpy:
9761
9762'``llvm.memcpy``' Intrinsic
9763^^^^^^^^^^^^^^^^^^^^^^^^^^^
9764
9765Syntax:
9766"""""""
9767
9768This is an overloaded intrinsic. You can use ``llvm.memcpy`` on any
9769integer bit width and for different address spaces. Not all targets
9770support all bit widths however.
9771
9772::
9773
9774      declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
9775                                              i32 <len>, i32 <align>, i1 <isvolatile>)
9776      declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
9777                                              i64 <len>, i32 <align>, i1 <isvolatile>)
9778
9779Overview:
9780"""""""""
9781
9782The '``llvm.memcpy.*``' intrinsics copy a block of memory from the
9783source location to the destination location.
9784
9785Note that, unlike the standard libc function, the ``llvm.memcpy.*``
9786intrinsics do not return a value, takes extra alignment/isvolatile
9787arguments and the pointers can be in specified address spaces.
9788
9789Arguments:
9790""""""""""
9791
9792The first argument is a pointer to the destination, the second is a
9793pointer to the source. The third argument is an integer argument
9794specifying the number of bytes to copy, the fourth argument is the
9795alignment of the source and destination locations, and the fifth is a
9796boolean indicating a volatile access.
9797
9798If the call to this intrinsic has an alignment value that is not 0 or 1,
9799then the caller guarantees that both the source and destination pointers
9800are aligned to that boundary.
9801
9802If the ``isvolatile`` parameter is ``true``, the ``llvm.memcpy`` call is
9803a :ref:`volatile operation <volatile>`. The detailed access behavior is not
9804very cleanly specified and it is unwise to depend on it.
9805
9806Semantics:
9807""""""""""
9808
9809The '``llvm.memcpy.*``' intrinsics copy a block of memory from the
9810source location to the destination location, which are not allowed to
9811overlap. It copies "len" bytes of memory over. If the argument is known
9812to be aligned to some boundary, this can be specified as the fourth
9813argument, otherwise it should be set to 0 or 1 (both meaning no alignment).
9814
9815'``llvm.memmove``' Intrinsic
9816^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9817
9818Syntax:
9819"""""""
9820
9821This is an overloaded intrinsic. You can use llvm.memmove on any integer
9822bit width and for different address space. Not all targets support all
9823bit widths however.
9824
9825::
9826
9827      declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
9828                                               i32 <len>, i32 <align>, i1 <isvolatile>)
9829      declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
9830                                               i64 <len>, i32 <align>, i1 <isvolatile>)
9831
9832Overview:
9833"""""""""
9834
9835The '``llvm.memmove.*``' intrinsics move a block of memory from the
9836source location to the destination location. It is similar to the
9837'``llvm.memcpy``' intrinsic but allows the two memory locations to
9838overlap.
9839
9840Note that, unlike the standard libc function, the ``llvm.memmove.*``
9841intrinsics do not return a value, takes extra alignment/isvolatile
9842arguments and the pointers can be in specified address spaces.
9843
9844Arguments:
9845""""""""""
9846
9847The first argument is a pointer to the destination, the second is a
9848pointer to the source. The third argument is an integer argument
9849specifying the number of bytes to copy, the fourth argument is the
9850alignment of the source and destination locations, and the fifth is a
9851boolean indicating a volatile access.
9852
9853If the call to this intrinsic has an alignment value that is not 0 or 1,
9854then the caller guarantees that the source and destination pointers are
9855aligned to that boundary.
9856
9857If the ``isvolatile`` parameter is ``true``, the ``llvm.memmove`` call
9858is a :ref:`volatile operation <volatile>`. The detailed access behavior is
9859not very cleanly specified and it is unwise to depend on it.
9860
9861Semantics:
9862""""""""""
9863
9864The '``llvm.memmove.*``' intrinsics copy a block of memory from the
9865source location to the destination location, which may overlap. It
9866copies "len" bytes of memory over. If the argument is known to be
9867aligned to some boundary, this can be specified as the fourth argument,
9868otherwise it should be set to 0 or 1 (both meaning no alignment).
9869
9870'``llvm.memset.*``' Intrinsics
9871^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9872
9873Syntax:
9874"""""""
9875
9876This is an overloaded intrinsic. You can use llvm.memset on any integer
9877bit width and for different address spaces. However, not all targets
9878support all bit widths.
9879
9880::
9881
9882      declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
9883                                         i32 <len>, i32 <align>, i1 <isvolatile>)
9884      declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
9885                                         i64 <len>, i32 <align>, i1 <isvolatile>)
9886
9887Overview:
9888"""""""""
9889
9890The '``llvm.memset.*``' intrinsics fill a block of memory with a
9891particular byte value.
9892
9893Note that, unlike the standard libc function, the ``llvm.memset``
9894intrinsic does not return a value and takes extra alignment/volatile
9895arguments. Also, the destination can be in an arbitrary address space.
9896
9897Arguments:
9898""""""""""
9899
9900The first argument is a pointer to the destination to fill, the second
9901is the byte value with which to fill it, the third argument is an
9902integer argument specifying the number of bytes to fill, and the fourth
9903argument is the known alignment of the destination location.
9904
9905If the call to this intrinsic has an alignment value that is not 0 or 1,
9906then the caller guarantees that the destination pointer is aligned to
9907that boundary.
9908
9909If the ``isvolatile`` parameter is ``true``, the ``llvm.memset`` call is
9910a :ref:`volatile operation <volatile>`. The detailed access behavior is not
9911very cleanly specified and it is unwise to depend on it.
9912
9913Semantics:
9914""""""""""
9915
9916The '``llvm.memset.*``' intrinsics fill "len" bytes of memory starting
9917at the destination location. If the argument is known to be aligned to
9918some boundary, this can be specified as the fourth argument, otherwise
9919it should be set to 0 or 1 (both meaning no alignment).
9920
9921'``llvm.sqrt.*``' Intrinsic
9922^^^^^^^^^^^^^^^^^^^^^^^^^^^
9923
9924Syntax:
9925"""""""
9926
9927This is an overloaded intrinsic. You can use ``llvm.sqrt`` on any
9928floating point or vector of floating point type. Not all targets support
9929all types however.
9930
9931::
9932
9933      declare float     @llvm.sqrt.f32(float %Val)
9934      declare double    @llvm.sqrt.f64(double %Val)
9935      declare x86_fp80  @llvm.sqrt.f80(x86_fp80 %Val)
9936      declare fp128     @llvm.sqrt.f128(fp128 %Val)
9937      declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
9938
9939Overview:
9940"""""""""
9941
9942The '``llvm.sqrt``' intrinsics return the sqrt of the specified operand,
9943returning the same value as the libm '``sqrt``' functions would. Unlike
9944``sqrt`` in libm, however, ``llvm.sqrt`` has undefined behavior for
9945negative numbers other than -0.0 (which allows for better optimization,
9946because there is no need to worry about errno being set).
9947``llvm.sqrt(-0.0)`` is defined to return -0.0 like IEEE sqrt.
9948
9949Arguments:
9950""""""""""
9951
9952The argument and return value are floating point numbers of the same
9953type.
9954
9955Semantics:
9956""""""""""
9957
9958This function returns the sqrt of the specified operand if it is a
9959nonnegative floating point number.
9960
9961'``llvm.powi.*``' Intrinsic
9962^^^^^^^^^^^^^^^^^^^^^^^^^^^
9963
9964Syntax:
9965"""""""
9966
9967This is an overloaded intrinsic. You can use ``llvm.powi`` on any
9968floating point or vector of floating point type. Not all targets support
9969all types however.
9970
9971::
9972
9973      declare float     @llvm.powi.f32(float  %Val, i32 %power)
9974      declare double    @llvm.powi.f64(double %Val, i32 %power)
9975      declare x86_fp80  @llvm.powi.f80(x86_fp80  %Val, i32 %power)
9976      declare fp128     @llvm.powi.f128(fp128 %Val, i32 %power)
9977      declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128  %Val, i32 %power)
9978
9979Overview:
9980"""""""""
9981
9982The '``llvm.powi.*``' intrinsics return the first operand raised to the
9983specified (positive or negative) power. The order of evaluation of
9984multiplications is not defined. When a vector of floating point type is
9985used, the second argument remains a scalar integer value.
9986
9987Arguments:
9988""""""""""
9989
9990The second argument is an integer power, and the first is a value to
9991raise to that power.
9992
9993Semantics:
9994""""""""""
9995
9996This function returns the first value raised to the second power with an
9997unspecified sequence of rounding operations.
9998
9999'``llvm.sin.*``' Intrinsic
10000^^^^^^^^^^^^^^^^^^^^^^^^^^
10001
10002Syntax:
10003"""""""
10004
10005This is an overloaded intrinsic. You can use ``llvm.sin`` on any
10006floating point or vector of floating point type. Not all targets support
10007all types however.
10008
10009::
10010
10011      declare float     @llvm.sin.f32(float  %Val)
10012      declare double    @llvm.sin.f64(double %Val)
10013      declare x86_fp80  @llvm.sin.f80(x86_fp80  %Val)
10014      declare fp128     @llvm.sin.f128(fp128 %Val)
10015      declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128  %Val)
10016
10017Overview:
10018"""""""""
10019
10020The '``llvm.sin.*``' intrinsics return the sine of the operand.
10021
10022Arguments:
10023""""""""""
10024
10025The argument and return value are floating point numbers of the same
10026type.
10027
10028Semantics:
10029""""""""""
10030
10031This function returns the sine of the specified operand, returning the
10032same values as the libm ``sin`` functions would, and handles error
10033conditions in the same way.
10034
10035'``llvm.cos.*``' Intrinsic
10036^^^^^^^^^^^^^^^^^^^^^^^^^^
10037
10038Syntax:
10039"""""""
10040
10041This is an overloaded intrinsic. You can use ``llvm.cos`` on any
10042floating point or vector of floating point type. Not all targets support
10043all types however.
10044
10045::
10046
10047      declare float     @llvm.cos.f32(float  %Val)
10048      declare double    @llvm.cos.f64(double %Val)
10049      declare x86_fp80  @llvm.cos.f80(x86_fp80  %Val)
10050      declare fp128     @llvm.cos.f128(fp128 %Val)
10051      declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128  %Val)
10052
10053Overview:
10054"""""""""
10055
10056The '``llvm.cos.*``' intrinsics return the cosine of the operand.
10057
10058Arguments:
10059""""""""""
10060
10061The argument and return value are floating point numbers of the same
10062type.
10063
10064Semantics:
10065""""""""""
10066
10067This function returns the cosine of the specified operand, returning the
10068same values as the libm ``cos`` functions would, and handles error
10069conditions in the same way.
10070
10071'``llvm.pow.*``' Intrinsic
10072^^^^^^^^^^^^^^^^^^^^^^^^^^
10073
10074Syntax:
10075"""""""
10076
10077This is an overloaded intrinsic. You can use ``llvm.pow`` on any
10078floating point or vector of floating point type. Not all targets support
10079all types however.
10080
10081::
10082
10083      declare float     @llvm.pow.f32(float  %Val, float %Power)
10084      declare double    @llvm.pow.f64(double %Val, double %Power)
10085      declare x86_fp80  @llvm.pow.f80(x86_fp80  %Val, x86_fp80 %Power)
10086      declare fp128     @llvm.pow.f128(fp128 %Val, fp128 %Power)
10087      declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128  %Val, ppc_fp128 Power)
10088
10089Overview:
10090"""""""""
10091
10092The '``llvm.pow.*``' intrinsics return the first operand raised to the
10093specified (positive or negative) power.
10094
10095Arguments:
10096""""""""""
10097
10098The second argument is a floating point power, and the first is a value
10099to raise to that power.
10100
10101Semantics:
10102""""""""""
10103
10104This function returns the first value raised to the second power,
10105returning the same values as the libm ``pow`` functions would, and
10106handles error conditions in the same way.
10107
10108'``llvm.exp.*``' Intrinsic
10109^^^^^^^^^^^^^^^^^^^^^^^^^^
10110
10111Syntax:
10112"""""""
10113
10114This is an overloaded intrinsic. You can use ``llvm.exp`` on any
10115floating point or vector of floating point type. Not all targets support
10116all types however.
10117
10118::
10119
10120      declare float     @llvm.exp.f32(float  %Val)
10121      declare double    @llvm.exp.f64(double %Val)
10122      declare x86_fp80  @llvm.exp.f80(x86_fp80  %Val)
10123      declare fp128     @llvm.exp.f128(fp128 %Val)
10124      declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128  %Val)
10125
10126Overview:
10127"""""""""
10128
10129The '``llvm.exp.*``' intrinsics perform the exp function.
10130
10131Arguments:
10132""""""""""
10133
10134The argument and return value are floating point numbers of the same
10135type.
10136
10137Semantics:
10138""""""""""
10139
10140This function returns the same values as the libm ``exp`` functions
10141would, and handles error conditions in the same way.
10142
10143'``llvm.exp2.*``' Intrinsic
10144^^^^^^^^^^^^^^^^^^^^^^^^^^^
10145
10146Syntax:
10147"""""""
10148
10149This is an overloaded intrinsic. You can use ``llvm.exp2`` on any
10150floating point or vector of floating point type. Not all targets support
10151all types however.
10152
10153::
10154
10155      declare float     @llvm.exp2.f32(float  %Val)
10156      declare double    @llvm.exp2.f64(double %Val)
10157      declare x86_fp80  @llvm.exp2.f80(x86_fp80  %Val)
10158      declare fp128     @llvm.exp2.f128(fp128 %Val)
10159      declare ppc_fp128 @llvm.exp2.ppcf128(ppc_fp128  %Val)
10160
10161Overview:
10162"""""""""
10163
10164The '``llvm.exp2.*``' intrinsics perform the exp2 function.
10165
10166Arguments:
10167""""""""""
10168
10169The argument and return value are floating point numbers of the same
10170type.
10171
10172Semantics:
10173""""""""""
10174
10175This function returns the same values as the libm ``exp2`` functions
10176would, and handles error conditions in the same way.
10177
10178'``llvm.log.*``' Intrinsic
10179^^^^^^^^^^^^^^^^^^^^^^^^^^
10180
10181Syntax:
10182"""""""
10183
10184This is an overloaded intrinsic. You can use ``llvm.log`` on any
10185floating point or vector of floating point type. Not all targets support
10186all types however.
10187
10188::
10189
10190      declare float     @llvm.log.f32(float  %Val)
10191      declare double    @llvm.log.f64(double %Val)
10192      declare x86_fp80  @llvm.log.f80(x86_fp80  %Val)
10193      declare fp128     @llvm.log.f128(fp128 %Val)
10194      declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128  %Val)
10195
10196Overview:
10197"""""""""
10198
10199The '``llvm.log.*``' intrinsics perform the log function.
10200
10201Arguments:
10202""""""""""
10203
10204The argument and return value are floating point numbers of the same
10205type.
10206
10207Semantics:
10208""""""""""
10209
10210This function returns the same values as the libm ``log`` functions
10211would, and handles error conditions in the same way.
10212
10213'``llvm.log10.*``' Intrinsic
10214^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10215
10216Syntax:
10217"""""""
10218
10219This is an overloaded intrinsic. You can use ``llvm.log10`` on any
10220floating point or vector of floating point type. Not all targets support
10221all types however.
10222
10223::
10224
10225      declare float     @llvm.log10.f32(float  %Val)
10226      declare double    @llvm.log10.f64(double %Val)
10227      declare x86_fp80  @llvm.log10.f80(x86_fp80  %Val)
10228      declare fp128     @llvm.log10.f128(fp128 %Val)
10229      declare ppc_fp128 @llvm.log10.ppcf128(ppc_fp128  %Val)
10230
10231Overview:
10232"""""""""
10233
10234The '``llvm.log10.*``' intrinsics perform the log10 function.
10235
10236Arguments:
10237""""""""""
10238
10239The argument and return value are floating point numbers of the same
10240type.
10241
10242Semantics:
10243""""""""""
10244
10245This function returns the same values as the libm ``log10`` functions
10246would, and handles error conditions in the same way.
10247
10248'``llvm.log2.*``' Intrinsic
10249^^^^^^^^^^^^^^^^^^^^^^^^^^^
10250
10251Syntax:
10252"""""""
10253
10254This is an overloaded intrinsic. You can use ``llvm.log2`` on any
10255floating point or vector of floating point type. Not all targets support
10256all types however.
10257
10258::
10259
10260      declare float     @llvm.log2.f32(float  %Val)
10261      declare double    @llvm.log2.f64(double %Val)
10262      declare x86_fp80  @llvm.log2.f80(x86_fp80  %Val)
10263      declare fp128     @llvm.log2.f128(fp128 %Val)
10264      declare ppc_fp128 @llvm.log2.ppcf128(ppc_fp128  %Val)
10265
10266Overview:
10267"""""""""
10268
10269The '``llvm.log2.*``' intrinsics perform the log2 function.
10270
10271Arguments:
10272""""""""""
10273
10274The argument and return value are floating point numbers of the same
10275type.
10276
10277Semantics:
10278""""""""""
10279
10280This function returns the same values as the libm ``log2`` functions
10281would, and handles error conditions in the same way.
10282
10283'``llvm.fma.*``' Intrinsic
10284^^^^^^^^^^^^^^^^^^^^^^^^^^
10285
10286Syntax:
10287"""""""
10288
10289This is an overloaded intrinsic. You can use ``llvm.fma`` on any
10290floating point or vector of floating point type. Not all targets support
10291all types however.
10292
10293::
10294
10295      declare float     @llvm.fma.f32(float  %a, float  %b, float  %c)
10296      declare double    @llvm.fma.f64(double %a, double %b, double %c)
10297      declare x86_fp80  @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
10298      declare fp128     @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
10299      declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
10300
10301Overview:
10302"""""""""
10303
10304The '``llvm.fma.*``' intrinsics perform the fused multiply-add
10305operation.
10306
10307Arguments:
10308""""""""""
10309
10310The argument and return value are floating point numbers of the same
10311type.
10312
10313Semantics:
10314""""""""""
10315
10316This function returns the same values as the libm ``fma`` functions
10317would, and does not set errno.
10318
10319'``llvm.fabs.*``' Intrinsic
10320^^^^^^^^^^^^^^^^^^^^^^^^^^^
10321
10322Syntax:
10323"""""""
10324
10325This is an overloaded intrinsic. You can use ``llvm.fabs`` on any
10326floating point or vector of floating point type. Not all targets support
10327all types however.
10328
10329::
10330
10331      declare float     @llvm.fabs.f32(float  %Val)
10332      declare double    @llvm.fabs.f64(double %Val)
10333      declare x86_fp80  @llvm.fabs.f80(x86_fp80 %Val)
10334      declare fp128     @llvm.fabs.f128(fp128 %Val)
10335      declare ppc_fp128 @llvm.fabs.ppcf128(ppc_fp128 %Val)
10336
10337Overview:
10338"""""""""
10339
10340The '``llvm.fabs.*``' intrinsics return the absolute value of the
10341operand.
10342
10343Arguments:
10344""""""""""
10345
10346The argument and return value are floating point numbers of the same
10347type.
10348
10349Semantics:
10350""""""""""
10351
10352This function returns the same values as the libm ``fabs`` functions
10353would, and handles error conditions in the same way.
10354
10355'``llvm.minnum.*``' Intrinsic
10356^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10357
10358Syntax:
10359"""""""
10360
10361This is an overloaded intrinsic. You can use ``llvm.minnum`` on any
10362floating point or vector of floating point type. Not all targets support
10363all types however.
10364
10365::
10366
10367      declare float     @llvm.minnum.f32(float %Val0, float %Val1)
10368      declare double    @llvm.minnum.f64(double %Val0, double %Val1)
10369      declare x86_fp80  @llvm.minnum.f80(x86_fp80 %Val0, x86_fp80 %Val1)
10370      declare fp128     @llvm.minnum.f128(fp128 %Val0, fp128 %Val1)
10371      declare ppc_fp128 @llvm.minnum.ppcf128(ppc_fp128 %Val0, ppc_fp128 %Val1)
10372
10373Overview:
10374"""""""""
10375
10376The '``llvm.minnum.*``' intrinsics return the minimum of the two
10377arguments.
10378
10379
10380Arguments:
10381""""""""""
10382
10383The arguments and return value are floating point numbers of the same
10384type.
10385
10386Semantics:
10387""""""""""
10388
10389Follows the IEEE-754 semantics for minNum, which also match for libm's
10390fmin.
10391
10392If either operand is a NaN, returns the other non-NaN operand. Returns
10393NaN only if both operands are NaN. If the operands compare equal,
10394returns a value that compares equal to both operands. This means that
10395fmin(+/-0.0, +/-0.0) could return either -0.0 or 0.0.
10396
10397'``llvm.maxnum.*``' Intrinsic
10398^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10399
10400Syntax:
10401"""""""
10402
10403This is an overloaded intrinsic. You can use ``llvm.maxnum`` on any
10404floating point or vector of floating point type. Not all targets support
10405all types however.
10406
10407::
10408
10409      declare float     @llvm.maxnum.f32(float  %Val0, float  %Val1l)
10410      declare double    @llvm.maxnum.f64(double %Val0, double %Val1)
10411      declare x86_fp80  @llvm.maxnum.f80(x86_fp80  %Val0, x86_fp80  %Val1)
10412      declare fp128     @llvm.maxnum.f128(fp128 %Val0, fp128 %Val1)
10413      declare ppc_fp128 @llvm.maxnum.ppcf128(ppc_fp128  %Val0, ppc_fp128  %Val1)
10414
10415Overview:
10416"""""""""
10417
10418The '``llvm.maxnum.*``' intrinsics return the maximum of the two
10419arguments.
10420
10421
10422Arguments:
10423""""""""""
10424
10425The arguments and return value are floating point numbers of the same
10426type.
10427
10428Semantics:
10429""""""""""
10430Follows the IEEE-754 semantics for maxNum, which also match for libm's
10431fmax.
10432
10433If either operand is a NaN, returns the other non-NaN operand. Returns
10434NaN only if both operands are NaN. If the operands compare equal,
10435returns a value that compares equal to both operands. This means that
10436fmax(+/-0.0, +/-0.0) could return either -0.0 or 0.0.
10437
10438'``llvm.copysign.*``' Intrinsic
10439^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10440
10441Syntax:
10442"""""""
10443
10444This is an overloaded intrinsic. You can use ``llvm.copysign`` on any
10445floating point or vector of floating point type. Not all targets support
10446all types however.
10447
10448::
10449
10450      declare float     @llvm.copysign.f32(float  %Mag, float  %Sgn)
10451      declare double    @llvm.copysign.f64(double %Mag, double %Sgn)
10452      declare x86_fp80  @llvm.copysign.f80(x86_fp80  %Mag, x86_fp80  %Sgn)
10453      declare fp128     @llvm.copysign.f128(fp128 %Mag, fp128 %Sgn)
10454      declare ppc_fp128 @llvm.copysign.ppcf128(ppc_fp128  %Mag, ppc_fp128  %Sgn)
10455
10456Overview:
10457"""""""""
10458
10459The '``llvm.copysign.*``' intrinsics return a value with the magnitude of the
10460first operand and the sign of the second operand.
10461
10462Arguments:
10463""""""""""
10464
10465The arguments and return value are floating point numbers of the same
10466type.
10467
10468Semantics:
10469""""""""""
10470
10471This function returns the same values as the libm ``copysign``
10472functions would, and handles error conditions in the same way.
10473
10474'``llvm.floor.*``' Intrinsic
10475^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10476
10477Syntax:
10478"""""""
10479
10480This is an overloaded intrinsic. You can use ``llvm.floor`` on any
10481floating point or vector of floating point type. Not all targets support
10482all types however.
10483
10484::
10485
10486      declare float     @llvm.floor.f32(float  %Val)
10487      declare double    @llvm.floor.f64(double %Val)
10488      declare x86_fp80  @llvm.floor.f80(x86_fp80  %Val)
10489      declare fp128     @llvm.floor.f128(fp128 %Val)
10490      declare ppc_fp128 @llvm.floor.ppcf128(ppc_fp128  %Val)
10491
10492Overview:
10493"""""""""
10494
10495The '``llvm.floor.*``' intrinsics return the floor of the operand.
10496
10497Arguments:
10498""""""""""
10499
10500The argument and return value are floating point numbers of the same
10501type.
10502
10503Semantics:
10504""""""""""
10505
10506This function returns the same values as the libm ``floor`` functions
10507would, and handles error conditions in the same way.
10508
10509'``llvm.ceil.*``' Intrinsic
10510^^^^^^^^^^^^^^^^^^^^^^^^^^^
10511
10512Syntax:
10513"""""""
10514
10515This is an overloaded intrinsic. You can use ``llvm.ceil`` on any
10516floating point or vector of floating point type. Not all targets support
10517all types however.
10518
10519::
10520
10521      declare float     @llvm.ceil.f32(float  %Val)
10522      declare double    @llvm.ceil.f64(double %Val)
10523      declare x86_fp80  @llvm.ceil.f80(x86_fp80  %Val)
10524      declare fp128     @llvm.ceil.f128(fp128 %Val)
10525      declare ppc_fp128 @llvm.ceil.ppcf128(ppc_fp128  %Val)
10526
10527Overview:
10528"""""""""
10529
10530The '``llvm.ceil.*``' intrinsics return the ceiling of the operand.
10531
10532Arguments:
10533""""""""""
10534
10535The argument and return value are floating point numbers of the same
10536type.
10537
10538Semantics:
10539""""""""""
10540
10541This function returns the same values as the libm ``ceil`` functions
10542would, and handles error conditions in the same way.
10543
10544'``llvm.trunc.*``' Intrinsic
10545^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10546
10547Syntax:
10548"""""""
10549
10550This is an overloaded intrinsic. You can use ``llvm.trunc`` on any
10551floating point or vector of floating point type. Not all targets support
10552all types however.
10553
10554::
10555
10556      declare float     @llvm.trunc.f32(float  %Val)
10557      declare double    @llvm.trunc.f64(double %Val)
10558      declare x86_fp80  @llvm.trunc.f80(x86_fp80  %Val)
10559      declare fp128     @llvm.trunc.f128(fp128 %Val)
10560      declare ppc_fp128 @llvm.trunc.ppcf128(ppc_fp128  %Val)
10561
10562Overview:
10563"""""""""
10564
10565The '``llvm.trunc.*``' intrinsics returns the operand rounded to the
10566nearest integer not larger in magnitude than the operand.
10567
10568Arguments:
10569""""""""""
10570
10571The argument and return value are floating point numbers of the same
10572type.
10573
10574Semantics:
10575""""""""""
10576
10577This function returns the same values as the libm ``trunc`` functions
10578would, and handles error conditions in the same way.
10579
10580'``llvm.rint.*``' Intrinsic
10581^^^^^^^^^^^^^^^^^^^^^^^^^^^
10582
10583Syntax:
10584"""""""
10585
10586This is an overloaded intrinsic. You can use ``llvm.rint`` on any
10587floating point or vector of floating point type. Not all targets support
10588all types however.
10589
10590::
10591
10592      declare float     @llvm.rint.f32(float  %Val)
10593      declare double    @llvm.rint.f64(double %Val)
10594      declare x86_fp80  @llvm.rint.f80(x86_fp80  %Val)
10595      declare fp128     @llvm.rint.f128(fp128 %Val)
10596      declare ppc_fp128 @llvm.rint.ppcf128(ppc_fp128  %Val)
10597
10598Overview:
10599"""""""""
10600
10601The '``llvm.rint.*``' intrinsics returns the operand rounded to the
10602nearest integer. It may raise an inexact floating-point exception if the
10603operand isn't an integer.
10604
10605Arguments:
10606""""""""""
10607
10608The argument and return value are floating point numbers of the same
10609type.
10610
10611Semantics:
10612""""""""""
10613
10614This function returns the same values as the libm ``rint`` functions
10615would, and handles error conditions in the same way.
10616
10617'``llvm.nearbyint.*``' Intrinsic
10618^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10619
10620Syntax:
10621"""""""
10622
10623This is an overloaded intrinsic. You can use ``llvm.nearbyint`` on any
10624floating point or vector of floating point type. Not all targets support
10625all types however.
10626
10627::
10628
10629      declare float     @llvm.nearbyint.f32(float  %Val)
10630      declare double    @llvm.nearbyint.f64(double %Val)
10631      declare x86_fp80  @llvm.nearbyint.f80(x86_fp80  %Val)
10632      declare fp128     @llvm.nearbyint.f128(fp128 %Val)
10633      declare ppc_fp128 @llvm.nearbyint.ppcf128(ppc_fp128  %Val)
10634
10635Overview:
10636"""""""""
10637
10638The '``llvm.nearbyint.*``' intrinsics returns the operand rounded to the
10639nearest integer.
10640
10641Arguments:
10642""""""""""
10643
10644The argument and return value are floating point numbers of the same
10645type.
10646
10647Semantics:
10648""""""""""
10649
10650This function returns the same values as the libm ``nearbyint``
10651functions would, and handles error conditions in the same way.
10652
10653'``llvm.round.*``' Intrinsic
10654^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10655
10656Syntax:
10657"""""""
10658
10659This is an overloaded intrinsic. You can use ``llvm.round`` on any
10660floating point or vector of floating point type. Not all targets support
10661all types however.
10662
10663::
10664
10665      declare float     @llvm.round.f32(float  %Val)
10666      declare double    @llvm.round.f64(double %Val)
10667      declare x86_fp80  @llvm.round.f80(x86_fp80  %Val)
10668      declare fp128     @llvm.round.f128(fp128 %Val)
10669      declare ppc_fp128 @llvm.round.ppcf128(ppc_fp128  %Val)
10670
10671Overview:
10672"""""""""
10673
10674The '``llvm.round.*``' intrinsics returns the operand rounded to the
10675nearest integer.
10676
10677Arguments:
10678""""""""""
10679
10680The argument and return value are floating point numbers of the same
10681type.
10682
10683Semantics:
10684""""""""""
10685
10686This function returns the same values as the libm ``round``
10687functions would, and handles error conditions in the same way.
10688
10689Bit Manipulation Intrinsics
10690---------------------------
10691
10692LLVM provides intrinsics for a few important bit manipulation
10693operations. These allow efficient code generation for some algorithms.
10694
10695'``llvm.bitreverse.*``' Intrinsics
10696^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10697
10698Syntax:
10699"""""""
10700
10701This is an overloaded intrinsic function. You can use bitreverse on any
10702integer type.
10703
10704::
10705
10706      declare i16 @llvm.bitreverse.i16(i16 <id>)
10707      declare i32 @llvm.bitreverse.i32(i32 <id>)
10708      declare i64 @llvm.bitreverse.i64(i64 <id>)
10709
10710Overview:
10711"""""""""
10712
10713The '``llvm.bitreverse``' family of intrinsics is used to reverse the
10714bitpattern of an integer value; for example ``0b10110110`` becomes
10715``0b01101101``.
10716
10717Semantics:
10718""""""""""
10719
10720The ``llvm.bitreverse.iN`` intrinsic returns an i16 value that has bit
10721``M`` in the input moved to bit ``N-M`` in the output.
10722
10723'``llvm.bswap.*``' Intrinsics
10724^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10725
10726Syntax:
10727"""""""
10728
10729This is an overloaded intrinsic function. You can use bswap on any
10730integer type that is an even number of bytes (i.e. BitWidth % 16 == 0).
10731
10732::
10733
10734      declare i16 @llvm.bswap.i16(i16 <id>)
10735      declare i32 @llvm.bswap.i32(i32 <id>)
10736      declare i64 @llvm.bswap.i64(i64 <id>)
10737
10738Overview:
10739"""""""""
10740
10741The '``llvm.bswap``' family of intrinsics is used to byte swap integer
10742values with an even number of bytes (positive multiple of 16 bits).
10743These are useful for performing operations on data that is not in the
10744target's native byte order.
10745
10746Semantics:
10747""""""""""
10748
10749The ``llvm.bswap.i16`` intrinsic returns an i16 value that has the high
10750and low byte of the input i16 swapped. Similarly, the ``llvm.bswap.i32``
10751intrinsic returns an i32 value that has the four bytes of the input i32
10752swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the
10753returned i32 will have its bytes in 3, 2, 1, 0 order. The
10754``llvm.bswap.i48``, ``llvm.bswap.i64`` and other intrinsics extend this
10755concept to additional even-byte lengths (6 bytes, 8 bytes and more,
10756respectively).
10757
10758'``llvm.ctpop.*``' Intrinsic
10759^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10760
10761Syntax:
10762"""""""
10763
10764This is an overloaded intrinsic. You can use llvm.ctpop on any integer
10765bit width, or on any vector with integer elements. Not all targets
10766support all bit widths or vector types, however.
10767
10768::
10769
10770      declare i8 @llvm.ctpop.i8(i8  <src>)
10771      declare i16 @llvm.ctpop.i16(i16 <src>)
10772      declare i32 @llvm.ctpop.i32(i32 <src>)
10773      declare i64 @llvm.ctpop.i64(i64 <src>)
10774      declare i256 @llvm.ctpop.i256(i256 <src>)
10775      declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
10776
10777Overview:
10778"""""""""
10779
10780The '``llvm.ctpop``' family of intrinsics counts the number of bits set
10781in a value.
10782
10783Arguments:
10784""""""""""
10785
10786The only argument is the value to be counted. The argument may be of any
10787integer type, or a vector with integer elements. The return type must
10788match the argument type.
10789
10790Semantics:
10791""""""""""
10792
10793The '``llvm.ctpop``' intrinsic counts the 1's in a variable, or within
10794each element of a vector.
10795
10796'``llvm.ctlz.*``' Intrinsic
10797^^^^^^^^^^^^^^^^^^^^^^^^^^^
10798
10799Syntax:
10800"""""""
10801
10802This is an overloaded intrinsic. You can use ``llvm.ctlz`` on any
10803integer bit width, or any vector whose elements are integers. Not all
10804targets support all bit widths or vector types, however.
10805
10806::
10807
10808      declare i8   @llvm.ctlz.i8  (i8   <src>, i1 <is_zero_undef>)
10809      declare i16  @llvm.ctlz.i16 (i16  <src>, i1 <is_zero_undef>)
10810      declare i32  @llvm.ctlz.i32 (i32  <src>, i1 <is_zero_undef>)
10811      declare i64  @llvm.ctlz.i64 (i64  <src>, i1 <is_zero_undef>)
10812      declare i256 @llvm.ctlz.i256(i256 <src>, i1 <is_zero_undef>)
10813      declare <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
10814
10815Overview:
10816"""""""""
10817
10818The '``llvm.ctlz``' family of intrinsic functions counts the number of
10819leading zeros in a variable.
10820
10821Arguments:
10822""""""""""
10823
10824The first argument is the value to be counted. This argument may be of
10825any integer type, or a vector with integer element type. The return
10826type must match the first argument type.
10827
10828The second argument must be a constant and is a flag to indicate whether
10829the intrinsic should ensure that a zero as the first argument produces a
10830defined result. Historically some architectures did not provide a
10831defined result for zero values as efficiently, and many algorithms are
10832now predicated on avoiding zero-value inputs.
10833
10834Semantics:
10835""""""""""
10836
10837The '``llvm.ctlz``' intrinsic counts the leading (most significant)
10838zeros in a variable, or within each element of the vector. If
10839``src == 0`` then the result is the size in bits of the type of ``src``
10840if ``is_zero_undef == 0`` and ``undef`` otherwise. For example,
10841``llvm.ctlz(i32 2) = 30``.
10842
10843'``llvm.cttz.*``' Intrinsic
10844^^^^^^^^^^^^^^^^^^^^^^^^^^^
10845
10846Syntax:
10847"""""""
10848
10849This is an overloaded intrinsic. You can use ``llvm.cttz`` on any
10850integer bit width, or any vector of integer elements. Not all targets
10851support all bit widths or vector types, however.
10852
10853::
10854
10855      declare i8   @llvm.cttz.i8  (i8   <src>, i1 <is_zero_undef>)
10856      declare i16  @llvm.cttz.i16 (i16  <src>, i1 <is_zero_undef>)
10857      declare i32  @llvm.cttz.i32 (i32  <src>, i1 <is_zero_undef>)
10858      declare i64  @llvm.cttz.i64 (i64  <src>, i1 <is_zero_undef>)
10859      declare i256 @llvm.cttz.i256(i256 <src>, i1 <is_zero_undef>)
10860      declare <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
10861
10862Overview:
10863"""""""""
10864
10865The '``llvm.cttz``' family of intrinsic functions counts the number of
10866trailing zeros.
10867
10868Arguments:
10869""""""""""
10870
10871The first argument is the value to be counted. This argument may be of
10872any integer type, or a vector with integer element type. The return
10873type must match the first argument type.
10874
10875The second argument must be a constant and is a flag to indicate whether
10876the intrinsic should ensure that a zero as the first argument produces a
10877defined result. Historically some architectures did not provide a
10878defined result for zero values as efficiently, and many algorithms are
10879now predicated on avoiding zero-value inputs.
10880
10881Semantics:
10882""""""""""
10883
10884The '``llvm.cttz``' intrinsic counts the trailing (least significant)
10885zeros in a variable, or within each element of a vector. If ``src == 0``
10886then the result is the size in bits of the type of ``src`` if
10887``is_zero_undef == 0`` and ``undef`` otherwise. For example,
10888``llvm.cttz(2) = 1``.
10889
10890.. _int_overflow:
10891
10892Arithmetic with Overflow Intrinsics
10893-----------------------------------
10894
10895LLVM provides intrinsics for fast arithmetic overflow checking.
10896
10897Each of these intrinsics returns a two-element struct. The first
10898element of this struct contains the result of the corresponding
10899arithmetic operation modulo 2\ :sup:`n`\ , where n is the bit width of
10900the result. Therefore, for example, the first element of the struct
10901returned by ``llvm.sadd.with.overflow.i32`` is always the same as the
10902result of a 32-bit ``add`` instruction with the same operands, where
10903the ``add`` is *not* modified by an ``nsw`` or ``nuw`` flag.
10904
10905The second element of the result is an ``i1`` that is 1 if the
10906arithmetic operation overflowed and 0 otherwise. An operation
10907overflows if, for any values of its operands ``A`` and ``B`` and for
10908any ``N`` larger than the operands' width, ``ext(A op B) to iN`` is
10909not equal to ``(ext(A) to iN) op (ext(B) to iN)`` where ``ext`` is
10910``sext`` for signed overflow and ``zext`` for unsigned overflow, and
10911``op`` is the underlying arithmetic operation.
10912
10913The behavior of these intrinsics is well-defined for all argument
10914values.
10915
10916'``llvm.sadd.with.overflow.*``' Intrinsics
10917^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10918
10919Syntax:
10920"""""""
10921
10922This is an overloaded intrinsic. You can use ``llvm.sadd.with.overflow``
10923on any integer bit width.
10924
10925::
10926
10927      declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
10928      declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
10929      declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
10930
10931Overview:
10932"""""""""
10933
10934The '``llvm.sadd.with.overflow``' family of intrinsic functions perform
10935a signed addition of the two arguments, and indicate whether an overflow
10936occurred during the signed summation.
10937
10938Arguments:
10939""""""""""
10940
10941The arguments (%a and %b) and the first element of the result structure
10942may be of integer types of any bit width, but they must have the same
10943bit width. The second element of the result structure must be of type
10944``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
10945addition.
10946
10947Semantics:
10948""""""""""
10949
10950The '``llvm.sadd.with.overflow``' family of intrinsic functions perform
10951a signed addition of the two variables. They return a structure --- the
10952first element of which is the signed summation, and the second element
10953of which is a bit specifying if the signed summation resulted in an
10954overflow.
10955
10956Examples:
10957"""""""""
10958
10959.. code-block:: llvm
10960
10961      %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
10962      %sum = extractvalue {i32, i1} %res, 0
10963      %obit = extractvalue {i32, i1} %res, 1
10964      br i1 %obit, label %overflow, label %normal
10965
10966'``llvm.uadd.with.overflow.*``' Intrinsics
10967^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10968
10969Syntax:
10970"""""""
10971
10972This is an overloaded intrinsic. You can use ``llvm.uadd.with.overflow``
10973on any integer bit width.
10974
10975::
10976
10977      declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
10978      declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
10979      declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
10980
10981Overview:
10982"""""""""
10983
10984The '``llvm.uadd.with.overflow``' family of intrinsic functions perform
10985an unsigned addition of the two arguments, and indicate whether a carry
10986occurred during the unsigned summation.
10987
10988Arguments:
10989""""""""""
10990
10991The arguments (%a and %b) and the first element of the result structure
10992may be of integer types of any bit width, but they must have the same
10993bit width. The second element of the result structure must be of type
10994``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
10995addition.
10996
10997Semantics:
10998""""""""""
10999
11000The '``llvm.uadd.with.overflow``' family of intrinsic functions perform
11001an unsigned addition of the two arguments. They return a structure --- the
11002first element of which is the sum, and the second element of which is a
11003bit specifying if the unsigned summation resulted in a carry.
11004
11005Examples:
11006"""""""""
11007
11008.. code-block:: llvm
11009
11010      %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
11011      %sum = extractvalue {i32, i1} %res, 0
11012      %obit = extractvalue {i32, i1} %res, 1
11013      br i1 %obit, label %carry, label %normal
11014
11015'``llvm.ssub.with.overflow.*``' Intrinsics
11016^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11017
11018Syntax:
11019"""""""
11020
11021This is an overloaded intrinsic. You can use ``llvm.ssub.with.overflow``
11022on any integer bit width.
11023
11024::
11025
11026      declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
11027      declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
11028      declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
11029
11030Overview:
11031"""""""""
11032
11033The '``llvm.ssub.with.overflow``' family of intrinsic functions perform
11034a signed subtraction of the two arguments, and indicate whether an
11035overflow occurred during the signed subtraction.
11036
11037Arguments:
11038""""""""""
11039
11040The arguments (%a and %b) and the first element of the result structure
11041may be of integer types of any bit width, but they must have the same
11042bit width. The second element of the result structure must be of type
11043``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
11044subtraction.
11045
11046Semantics:
11047""""""""""
11048
11049The '``llvm.ssub.with.overflow``' family of intrinsic functions perform
11050a signed subtraction of the two arguments. They return a structure --- the
11051first element of which is the subtraction, and the second element of
11052which is a bit specifying if the signed subtraction resulted in an
11053overflow.
11054
11055Examples:
11056"""""""""
11057
11058.. code-block:: llvm
11059
11060      %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
11061      %sum = extractvalue {i32, i1} %res, 0
11062      %obit = extractvalue {i32, i1} %res, 1
11063      br i1 %obit, label %overflow, label %normal
11064
11065'``llvm.usub.with.overflow.*``' Intrinsics
11066^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11067
11068Syntax:
11069"""""""
11070
11071This is an overloaded intrinsic. You can use ``llvm.usub.with.overflow``
11072on any integer bit width.
11073
11074::
11075
11076      declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
11077      declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
11078      declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
11079
11080Overview:
11081"""""""""
11082
11083The '``llvm.usub.with.overflow``' family of intrinsic functions perform
11084an unsigned subtraction of the two arguments, and indicate whether an
11085overflow occurred during the unsigned subtraction.
11086
11087Arguments:
11088""""""""""
11089
11090The arguments (%a and %b) and the first element of the result structure
11091may be of integer types of any bit width, but they must have the same
11092bit width. The second element of the result structure must be of type
11093``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
11094subtraction.
11095
11096Semantics:
11097""""""""""
11098
11099The '``llvm.usub.with.overflow``' family of intrinsic functions perform
11100an unsigned subtraction of the two arguments. They return a structure ---
11101the first element of which is the subtraction, and the second element of
11102which is a bit specifying if the unsigned subtraction resulted in an
11103overflow.
11104
11105Examples:
11106"""""""""
11107
11108.. code-block:: llvm
11109
11110      %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
11111      %sum = extractvalue {i32, i1} %res, 0
11112      %obit = extractvalue {i32, i1} %res, 1
11113      br i1 %obit, label %overflow, label %normal
11114
11115'``llvm.smul.with.overflow.*``' Intrinsics
11116^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11117
11118Syntax:
11119"""""""
11120
11121This is an overloaded intrinsic. You can use ``llvm.smul.with.overflow``
11122on any integer bit width.
11123
11124::
11125
11126      declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
11127      declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
11128      declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
11129
11130Overview:
11131"""""""""
11132
11133The '``llvm.smul.with.overflow``' family of intrinsic functions perform
11134a signed multiplication of the two arguments, and indicate whether an
11135overflow occurred during the signed multiplication.
11136
11137Arguments:
11138""""""""""
11139
11140The arguments (%a and %b) and the first element of the result structure
11141may be of integer types of any bit width, but they must have the same
11142bit width. The second element of the result structure must be of type
11143``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
11144multiplication.
11145
11146Semantics:
11147""""""""""
11148
11149The '``llvm.smul.with.overflow``' family of intrinsic functions perform
11150a signed multiplication of the two arguments. They return a structure ---
11151the first element of which is the multiplication, and the second element
11152of which is a bit specifying if the signed multiplication resulted in an
11153overflow.
11154
11155Examples:
11156"""""""""
11157
11158.. code-block:: llvm
11159
11160      %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
11161      %sum = extractvalue {i32, i1} %res, 0
11162      %obit = extractvalue {i32, i1} %res, 1
11163      br i1 %obit, label %overflow, label %normal
11164
11165'``llvm.umul.with.overflow.*``' Intrinsics
11166^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11167
11168Syntax:
11169"""""""
11170
11171This is an overloaded intrinsic. You can use ``llvm.umul.with.overflow``
11172on any integer bit width.
11173
11174::
11175
11176      declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
11177      declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
11178      declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
11179
11180Overview:
11181"""""""""
11182
11183The '``llvm.umul.with.overflow``' family of intrinsic functions perform
11184a unsigned multiplication of the two arguments, and indicate whether an
11185overflow occurred during the unsigned multiplication.
11186
11187Arguments:
11188""""""""""
11189
11190The arguments (%a and %b) and the first element of the result structure
11191may be of integer types of any bit width, but they must have the same
11192bit width. The second element of the result structure must be of type
11193``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
11194multiplication.
11195
11196Semantics:
11197""""""""""
11198
11199The '``llvm.umul.with.overflow``' family of intrinsic functions perform
11200an unsigned multiplication of the two arguments. They return a structure ---
11201the first element of which is the multiplication, and the second
11202element of which is a bit specifying if the unsigned multiplication
11203resulted in an overflow.
11204
11205Examples:
11206"""""""""
11207
11208.. code-block:: llvm
11209
11210      %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
11211      %sum = extractvalue {i32, i1} %res, 0
11212      %obit = extractvalue {i32, i1} %res, 1
11213      br i1 %obit, label %overflow, label %normal
11214
11215Specialised Arithmetic Intrinsics
11216---------------------------------
11217
11218'``llvm.canonicalize.*``' Intrinsic
11219^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11220
11221Syntax:
11222"""""""
11223
11224::
11225
11226      declare float @llvm.canonicalize.f32(float %a)
11227      declare double @llvm.canonicalize.f64(double %b)
11228
11229Overview:
11230"""""""""
11231
11232The '``llvm.canonicalize.*``' intrinsic returns the platform specific canonical
11233encoding of a floating point number. This canonicalization is useful for
11234implementing certain numeric primitives such as frexp. The canonical encoding is
11235defined by IEEE-754-2008 to be:
11236
11237::
11238
11239      2.1.8 canonical encoding: The preferred encoding of a floating-point
11240      representation in a format. Applied to declets, significands of finite
11241      numbers, infinities, and NaNs, especially in decimal formats.
11242
11243This operation can also be considered equivalent to the IEEE-754-2008
11244conversion of a floating-point value to the same format. NaNs are handled
11245according to section 6.2.
11246
11247Examples of non-canonical encodings:
11248
11249- x87 pseudo denormals, pseudo NaNs, pseudo Infinity, Unnormals. These are
11250  converted to a canonical representation per hardware-specific protocol.
11251- Many normal decimal floating point numbers have non-canonical alternative
11252  encodings.
11253- Some machines, like GPUs or ARMv7 NEON, do not support subnormal values.
11254  These are treated as non-canonical encodings of zero and will be flushed to
11255  a zero of the same sign by this operation.
11256
11257Note that per IEEE-754-2008 6.2, systems that support signaling NaNs with
11258default exception handling must signal an invalid exception, and produce a
11259quiet NaN result.
11260
11261This function should always be implementable as multiplication by 1.0, provided
11262that the compiler does not constant fold the operation. Likewise, division by
112631.0 and ``llvm.minnum(x, x)`` are possible implementations. Addition with
11264-0.0 is also sufficient provided that the rounding mode is not -Infinity.
11265
11266``@llvm.canonicalize`` must preserve the equality relation. That is:
11267
11268- ``(@llvm.canonicalize(x) == x)`` is equivalent to ``(x == x)``
11269- ``(@llvm.canonicalize(x) == @llvm.canonicalize(y))`` is equivalent to
11270  to ``(x == y)``
11271
11272Additionally, the sign of zero must be conserved:
11273``@llvm.canonicalize(-0.0) = -0.0`` and ``@llvm.canonicalize(+0.0) = +0.0``
11274
11275The payload bits of a NaN must be conserved, with two exceptions.
11276First, environments which use only a single canonical representation of NaN
11277must perform said canonicalization. Second, SNaNs must be quieted per the
11278usual methods.
11279
11280The canonicalization operation may be optimized away if:
11281
11282- The input is known to be canonical. For example, it was produced by a
11283  floating-point operation that is required by the standard to be canonical.
11284- The result is consumed only by (or fused with) other floating-point
11285  operations. That is, the bits of the floating point value are not examined.
11286
11287'``llvm.fmuladd.*``' Intrinsic
11288^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11289
11290Syntax:
11291"""""""
11292
11293::
11294
11295      declare float @llvm.fmuladd.f32(float %a, float %b, float %c)
11296      declare double @llvm.fmuladd.f64(double %a, double %b, double %c)
11297
11298Overview:
11299"""""""""
11300
11301The '``llvm.fmuladd.*``' intrinsic functions represent multiply-add
11302expressions that can be fused if the code generator determines that (a) the
11303target instruction set has support for a fused operation, and (b) that the
11304fused operation is more efficient than the equivalent, separate pair of mul
11305and add instructions.
11306
11307Arguments:
11308""""""""""
11309
11310The '``llvm.fmuladd.*``' intrinsics each take three arguments: two
11311multiplicands, a and b, and an addend c.
11312
11313Semantics:
11314""""""""""
11315
11316The expression:
11317
11318::
11319
11320      %0 = call float @llvm.fmuladd.f32(%a, %b, %c)
11321
11322is equivalent to the expression a \* b + c, except that rounding will
11323not be performed between the multiplication and addition steps if the
11324code generator fuses the operations. Fusion is not guaranteed, even if
11325the target platform supports it. If a fused multiply-add is required the
11326corresponding llvm.fma.\* intrinsic function should be used
11327instead. This never sets errno, just as '``llvm.fma.*``'.
11328
11329Examples:
11330"""""""""
11331
11332.. code-block:: llvm
11333
11334      %r2 = call float @llvm.fmuladd.f32(float %a, float %b, float %c) ; yields float:r2 = (a * b) + c
11335
11336Half Precision Floating Point Intrinsics
11337----------------------------------------
11338
11339For most target platforms, half precision floating point is a
11340storage-only format. This means that it is a dense encoding (in memory)
11341but does not support computation in the format.
11342
11343This means that code must first load the half-precision floating point
11344value as an i16, then convert it to float with
11345:ref:`llvm.convert.from.fp16 <int_convert_from_fp16>`. Computation can
11346then be performed on the float value (including extending to double
11347etc). To store the value back to memory, it is first converted to float
11348if needed, then converted to i16 with
11349:ref:`llvm.convert.to.fp16 <int_convert_to_fp16>`, then storing as an
11350i16 value.
11351
11352.. _int_convert_to_fp16:
11353
11354'``llvm.convert.to.fp16``' Intrinsic
11355^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11356
11357Syntax:
11358"""""""
11359
11360::
11361
11362      declare i16 @llvm.convert.to.fp16.f32(float %a)
11363      declare i16 @llvm.convert.to.fp16.f64(double %a)
11364
11365Overview:
11366"""""""""
11367
11368The '``llvm.convert.to.fp16``' intrinsic function performs a conversion from a
11369conventional floating point type to half precision floating point format.
11370
11371Arguments:
11372""""""""""
11373
11374The intrinsic function contains single argument - the value to be
11375converted.
11376
11377Semantics:
11378""""""""""
11379
11380The '``llvm.convert.to.fp16``' intrinsic function performs a conversion from a
11381conventional floating point format to half precision floating point format. The
11382return value is an ``i16`` which contains the converted number.
11383
11384Examples:
11385"""""""""
11386
11387.. code-block:: llvm
11388
11389      %res = call i16 @llvm.convert.to.fp16.f32(float %a)
11390      store i16 %res, i16* @x, align 2
11391
11392.. _int_convert_from_fp16:
11393
11394'``llvm.convert.from.fp16``' Intrinsic
11395^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11396
11397Syntax:
11398"""""""
11399
11400::
11401
11402      declare float @llvm.convert.from.fp16.f32(i16 %a)
11403      declare double @llvm.convert.from.fp16.f64(i16 %a)
11404
11405Overview:
11406"""""""""
11407
11408The '``llvm.convert.from.fp16``' intrinsic function performs a
11409conversion from half precision floating point format to single precision
11410floating point format.
11411
11412Arguments:
11413""""""""""
11414
11415The intrinsic function contains single argument - the value to be
11416converted.
11417
11418Semantics:
11419""""""""""
11420
11421The '``llvm.convert.from.fp16``' intrinsic function performs a
11422conversion from half single precision floating point format to single
11423precision floating point format. The input half-float value is
11424represented by an ``i16`` value.
11425
11426Examples:
11427"""""""""
11428
11429.. code-block:: llvm
11430
11431      %a = load i16, i16* @x, align 2
11432      %res = call float @llvm.convert.from.fp16(i16 %a)
11433
11434.. _dbg_intrinsics:
11435
11436Debugger Intrinsics
11437-------------------
11438
11439The LLVM debugger intrinsics (which all start with ``llvm.dbg.``
11440prefix), are described in the `LLVM Source Level
11441Debugging <SourceLevelDebugging.html#format_common_intrinsics>`_
11442document.
11443
11444Exception Handling Intrinsics
11445-----------------------------
11446
11447The LLVM exception handling intrinsics (which all start with
11448``llvm.eh.`` prefix), are described in the `LLVM Exception
11449Handling <ExceptionHandling.html#format_common_intrinsics>`_ document.
11450
11451.. _int_trampoline:
11452
11453Trampoline Intrinsics
11454---------------------
11455
11456These intrinsics make it possible to excise one parameter, marked with
11457the :ref:`nest <nest>` attribute, from a function. The result is a
11458callable function pointer lacking the nest parameter - the caller does
11459not need to provide a value for it. Instead, the value to use is stored
11460in advance in a "trampoline", a block of memory usually allocated on the
11461stack, which also contains code to splice the nest value into the
11462argument list. This is used to implement the GCC nested function address
11463extension.
11464
11465For example, if the function is ``i32 f(i8* nest %c, i32 %x, i32 %y)``
11466then the resulting function pointer has signature ``i32 (i32, i32)*``.
11467It can be created as follows:
11468
11469.. code-block:: llvm
11470
11471      %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
11472      %tramp1 = getelementptr [10 x i8], [10 x i8]* %tramp, i32 0, i32 0
11473      call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval)
11474      %p = call i8* @llvm.adjust.trampoline(i8* %tramp1)
11475      %fp = bitcast i8* %p to i32 (i32, i32)*
11476
11477The call ``%val = call i32 %fp(i32 %x, i32 %y)`` is then equivalent to
11478``%val = call i32 %f(i8* %nval, i32 %x, i32 %y)``.
11479
11480.. _int_it:
11481
11482'``llvm.init.trampoline``' Intrinsic
11483^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11484
11485Syntax:
11486"""""""
11487
11488::
11489
11490      declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
11491
11492Overview:
11493"""""""""
11494
11495This fills the memory pointed to by ``tramp`` with executable code,
11496turning it into a trampoline.
11497
11498Arguments:
11499""""""""""
11500
11501The ``llvm.init.trampoline`` intrinsic takes three arguments, all
11502pointers. The ``tramp`` argument must point to a sufficiently large and
11503sufficiently aligned block of memory; this memory is written to by the
11504intrinsic. Note that the size and the alignment are target-specific -
11505LLVM currently provides no portable way of determining them, so a
11506front-end that generates this intrinsic needs to have some
11507target-specific knowledge. The ``func`` argument must hold a function
11508bitcast to an ``i8*``.
11509
11510Semantics:
11511""""""""""
11512
11513The block of memory pointed to by ``tramp`` is filled with target
11514dependent code, turning it into a function. Then ``tramp`` needs to be
11515passed to :ref:`llvm.adjust.trampoline <int_at>` to get a pointer which can
11516be :ref:`bitcast (to a new function) and called <int_trampoline>`. The new
11517function's signature is the same as that of ``func`` with any arguments
11518marked with the ``nest`` attribute removed. At most one such ``nest``
11519argument is allowed, and it must be of pointer type. Calling the new
11520function is equivalent to calling ``func`` with the same argument list,
11521but with ``nval`` used for the missing ``nest`` argument. If, after
11522calling ``llvm.init.trampoline``, the memory pointed to by ``tramp`` is
11523modified, then the effect of any later call to the returned function
11524pointer is undefined.
11525
11526.. _int_at:
11527
11528'``llvm.adjust.trampoline``' Intrinsic
11529^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11530
11531Syntax:
11532"""""""
11533
11534::
11535
11536      declare i8* @llvm.adjust.trampoline(i8* <tramp>)
11537
11538Overview:
11539"""""""""
11540
11541This performs any required machine-specific adjustment to the address of
11542a trampoline (passed as ``tramp``).
11543
11544Arguments:
11545""""""""""
11546
11547``tramp`` must point to a block of memory which already has trampoline
11548code filled in by a previous call to
11549:ref:`llvm.init.trampoline <int_it>`.
11550
11551Semantics:
11552""""""""""
11553
11554On some architectures the address of the code to be executed needs to be
11555different than the address where the trampoline is actually stored. This
11556intrinsic returns the executable address corresponding to ``tramp``
11557after performing the required machine specific adjustments. The pointer
11558returned can then be :ref:`bitcast and executed <int_trampoline>`.
11559
11560.. _int_mload_mstore:
11561
11562Masked Vector Load and Store Intrinsics
11563---------------------------------------
11564
11565LLVM provides intrinsics for predicated vector load and store operations. The predicate is specified by a mask operand, which holds one bit per vector element, switching the associated vector lane on or off. The memory addresses corresponding to the "off" lanes are not accessed. When all bits of the mask are on, the intrinsic is identical to a regular vector load or store. When all bits are off, no memory is accessed.
11566
11567.. _int_mload:
11568
11569'``llvm.masked.load.*``' Intrinsics
11570^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11571
11572Syntax:
11573"""""""
11574This is an overloaded intrinsic. The loaded data is a vector of any integer, floating point or pointer data type.
11575
11576::
11577
11578      declare <16 x float>  @llvm.masked.load.v16f32.p0v16f32 (<16 x float>* <ptr>, i32 <alignment>, <16 x i1> <mask>, <16 x float> <passthru>)
11579      declare <2 x double>  @llvm.masked.load.v2f64.p0v2f64  (<2 x double>* <ptr>, i32 <alignment>, <2 x i1>  <mask>, <2 x double> <passthru>)
11580      ;; The data is a vector of pointers to double
11581      declare <8 x double*> @llvm.masked.load.v8p0f64.p0v8p0f64    (<8 x double*>* <ptr>, i32 <alignment>, <8 x i1> <mask>, <8 x double*> <passthru>)
11582      ;; The data is a vector of function pointers
11583      declare <8 x i32 ()*> @llvm.masked.load.v8p0f_i32f.p0v8p0f_i32f (<8 x i32 ()*>* <ptr>, i32 <alignment>, <8 x i1> <mask>, <8 x i32 ()*> <passthru>)
11584
11585Overview:
11586"""""""""
11587
11588Reads a vector from memory according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes. The masked-off lanes in the result vector are taken from the corresponding lanes of the '``passthru``' operand.
11589
11590
11591Arguments:
11592""""""""""
11593
11594The first operand is the base pointer for the load. The second operand is the alignment of the source location. It must be a constant integer value. The third operand, mask, is a vector of boolean values with the same number of elements as the return type. The fourth is a pass-through value that is used to fill the masked-off lanes of the result. The return type, underlying type of the base pointer and the type of the '``passthru``' operand are the same vector types.
11595
11596
11597Semantics:
11598""""""""""
11599
11600The '``llvm.masked.load``' intrinsic is designed for conditional reading of selected vector elements in a single IR operation. It is useful for targets that support vector masked loads and allows vectorizing predicated basic blocks on these targets. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar load operations.
11601The result of this operation is equivalent to a regular vector load instruction followed by a 'select' between the loaded and the passthru values, predicated on the same mask. However, using this intrinsic prevents exceptions on memory access to masked-off lanes.
11602
11603
11604::
11605
11606       %res = call <16 x float> @llvm.masked.load.v16f32.p0v16f32 (<16 x float>* %ptr, i32 4, <16 x i1>%mask, <16 x float> %passthru)
11607
11608       ;; The result of the two following instructions is identical aside from potential memory access exception
11609       %loadlal = load <16 x float>, <16 x float>* %ptr, align 4
11610       %res = select <16 x i1> %mask, <16 x float> %loadlal, <16 x float> %passthru
11611
11612.. _int_mstore:
11613
11614'``llvm.masked.store.*``' Intrinsics
11615^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11616
11617Syntax:
11618"""""""
11619This is an overloaded intrinsic. The data stored in memory is a vector of any integer, floating point or pointer data type.
11620
11621::
11622
11623       declare void @llvm.masked.store.v8i32.p0v8i32  (<8  x i32>   <value>, <8  x i32>*   <ptr>, i32 <alignment>,  <8  x i1> <mask>)
11624       declare void @llvm.masked.store.v16f32.p0v16f32 (<16 x float> <value>, <16 x float>* <ptr>, i32 <alignment>,  <16 x i1> <mask>)
11625       ;; The data is a vector of pointers to double
11626       declare void @llvm.masked.store.v8p0f64.p0v8p0f64    (<8 x double*> <value>, <8 x double*>* <ptr>, i32 <alignment>, <8 x i1> <mask>)
11627       ;; The data is a vector of function pointers
11628       declare void @llvm.masked.store.v4p0f_i32f.p0v4p0f_i32f (<4 x i32 ()*> <value>, <4 x i32 ()*>* <ptr>, i32 <alignment>, <4 x i1> <mask>)
11629
11630Overview:
11631"""""""""
11632
11633Writes a vector to memory according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes.
11634
11635Arguments:
11636""""""""""
11637
11638The first operand is the vector value to be written to memory. The second operand is the base pointer for the store, it has the same underlying type as the value operand. The third operand is the alignment of the destination location. The fourth operand, mask, is a vector of boolean values. The types of the mask and the value operand must have the same number of vector elements.
11639
11640
11641Semantics:
11642""""""""""
11643
11644The '``llvm.masked.store``' intrinsics is designed for conditional writing of selected vector elements in a single IR operation. It is useful for targets that support vector masked store and allows vectorizing predicated basic blocks on these targets. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar store operations.
11645The result of this operation is equivalent to a load-modify-store sequence. However, using this intrinsic prevents exceptions and data races on memory access to masked-off lanes.
11646
11647::
11648
11649       call void @llvm.masked.store.v16f32.p0v16f32(<16 x float> %value, <16 x float>* %ptr, i32 4,  <16 x i1> %mask)
11650
11651       ;; The result of the following instructions is identical aside from potential data races and memory access exceptions
11652       %oldval = load <16 x float>, <16 x float>* %ptr, align 4
11653       %res = select <16 x i1> %mask, <16 x float> %value, <16 x float> %oldval
11654       store <16 x float> %res, <16 x float>* %ptr, align 4
11655
11656
11657Masked Vector Gather and Scatter Intrinsics
11658-------------------------------------------
11659
11660LLVM provides intrinsics for vector gather and scatter operations. They are similar to :ref:`Masked Vector Load and Store <int_mload_mstore>`, except they are designed for arbitrary memory accesses, rather than sequential memory accesses. Gather and scatter also employ a mask operand, which holds one bit per vector element, switching the associated vector lane on or off. The memory addresses corresponding to the "off" lanes are not accessed. When all bits are off, no memory is accessed.
11661
11662.. _int_mgather:
11663
11664'``llvm.masked.gather.*``' Intrinsics
11665^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11666
11667Syntax:
11668"""""""
11669This is an overloaded intrinsic. The loaded data are multiple scalar values of any integer, floating point or pointer data type gathered together into one vector.
11670
11671::
11672
11673      declare <16 x float> @llvm.masked.gather.v16f32   (<16 x float*> <ptrs>, i32 <alignment>, <16 x i1> <mask>, <16 x float> <passthru>)
11674      declare <2 x double> @llvm.masked.gather.v2f64    (<2 x double*> <ptrs>, i32 <alignment>, <2 x i1>  <mask>, <2 x double> <passthru>)
11675      declare <8 x float*> @llvm.masked.gather.v8p0f32  (<8 x float**> <ptrs>, i32 <alignment>, <8 x i1>  <mask>, <8 x float*> <passthru>)
11676
11677Overview:
11678"""""""""
11679
11680Reads scalar values from arbitrary memory locations and gathers them into one vector. The memory locations are provided in the vector of pointers '``ptrs``'. The memory is accessed according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes. The masked-off lanes in the result vector are taken from the corresponding lanes of the '``passthru``' operand.
11681
11682
11683Arguments:
11684""""""""""
11685
11686The first operand is a vector of pointers which holds all memory addresses to read. The second operand is an alignment of the source addresses. It must be a constant integer value. The third operand, mask, is a vector of boolean values with the same number of elements as the return type. The fourth is a pass-through value that is used to fill the masked-off lanes of the result. The return type, underlying type of the vector of pointers and the type of the '``passthru``' operand are the same vector types.
11687
11688
11689Semantics:
11690""""""""""
11691
11692The '``llvm.masked.gather``' intrinsic is designed for conditional reading of multiple scalar values from arbitrary memory locations in a single IR operation. It is useful for targets that support vector masked gathers and allows vectorizing basic blocks with data and control divergence. Other targets may support this intrinsic differently, for example by lowering it into a sequence of scalar load operations.
11693The semantics of this operation are equivalent to a sequence of conditional scalar loads with subsequent gathering all loaded values into a single vector. The mask restricts memory access to certain lanes and facilitates vectorization of predicated basic blocks.
11694
11695
11696::
11697
11698       %res = call <4 x double> @llvm.masked.gather.v4f64 (<4 x double*> %ptrs, i32 8, <4 x i1>%mask, <4 x double> <true, true, true, true>)
11699
11700       ;; The gather with all-true mask is equivalent to the following instruction sequence
11701       %ptr0 = extractelement <4 x double*> %ptrs, i32 0
11702       %ptr1 = extractelement <4 x double*> %ptrs, i32 1
11703       %ptr2 = extractelement <4 x double*> %ptrs, i32 2
11704       %ptr3 = extractelement <4 x double*> %ptrs, i32 3
11705
11706       %val0 = load double, double* %ptr0, align 8
11707       %val1 = load double, double* %ptr1, align 8
11708       %val2 = load double, double* %ptr2, align 8
11709       %val3 = load double, double* %ptr3, align 8
11710
11711       %vec0    = insertelement <4 x double>undef, %val0, 0
11712       %vec01   = insertelement <4 x double>%vec0, %val1, 1
11713       %vec012  = insertelement <4 x double>%vec01, %val2, 2
11714       %vec0123 = insertelement <4 x double>%vec012, %val3, 3
11715
11716.. _int_mscatter:
11717
11718'``llvm.masked.scatter.*``' Intrinsics
11719^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11720
11721Syntax:
11722"""""""
11723This is an overloaded intrinsic. The data stored in memory is a vector of any integer, floating point or pointer data type. Each vector element is stored in an arbitrary memory address. Scatter with overlapping addresses is guaranteed to be ordered from least-significant to most-significant element.
11724
11725::
11726
11727       declare void @llvm.masked.scatter.v8i32   (<8 x i32>     <value>, <8 x i32*>     <ptrs>, i32 <alignment>, <8 x i1>  <mask>)
11728       declare void @llvm.masked.scatter.v16f32  (<16 x float>  <value>, <16 x float*>  <ptrs>, i32 <alignment>, <16 x i1> <mask>)
11729       declare void @llvm.masked.scatter.v4p0f64 (<4 x double*> <value>, <4 x double**> <ptrs>, i32 <alignment>, <4 x i1>  <mask>)
11730
11731Overview:
11732"""""""""
11733
11734Writes each element from the value vector to the corresponding memory address. The memory addresses are represented as a vector of pointers. Writing is done according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes.
11735
11736Arguments:
11737""""""""""
11738
11739The first operand is a vector value to be written to memory. The second operand is a vector of pointers, pointing to where the value elements should be stored. It has the same underlying type as the value operand. The third operand is an alignment of the destination addresses. The fourth operand, mask, is a vector of boolean values. The types of the mask and the value operand must have the same number of vector elements.
11740
11741
11742Semantics:
11743""""""""""
11744
11745The '``llvm.masked.scatter``' intrinsics is designed for writing selected vector elements to arbitrary memory addresses in a single IR operation. The operation may be conditional, when not all bits in the mask are switched on. It is useful for targets that support vector masked scatter and allows vectorizing basic blocks with data and control divergence. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar store operations.
11746
11747::
11748
11749       ;; This instruction unconditionally stores data vector in multiple addresses
11750       call @llvm.masked.scatter.v8i32 (<8 x i32> %value, <8 x i32*> %ptrs, i32 4,  <8 x i1>  <true, true, .. true>)
11751
11752       ;; It is equivalent to a list of scalar stores
11753       %val0 = extractelement <8 x i32> %value, i32 0
11754       %val1 = extractelement <8 x i32> %value, i32 1
11755       ..
11756       %val7 = extractelement <8 x i32> %value, i32 7
11757       %ptr0 = extractelement <8 x i32*> %ptrs, i32 0
11758       %ptr1 = extractelement <8 x i32*> %ptrs, i32 1
11759       ..
11760       %ptr7 = extractelement <8 x i32*> %ptrs, i32 7
11761       ;; Note: the order of the following stores is important when they overlap:
11762       store i32 %val0, i32* %ptr0, align 4
11763       store i32 %val1, i32* %ptr1, align 4
11764       ..
11765       store i32 %val7, i32* %ptr7, align 4
11766
11767
11768Memory Use Markers
11769------------------
11770
11771This class of intrinsics provides information about the lifetime of
11772memory objects and ranges where variables are immutable.
11773
11774.. _int_lifestart:
11775
11776'``llvm.lifetime.start``' Intrinsic
11777^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11778
11779Syntax:
11780"""""""
11781
11782::
11783
11784      declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
11785
11786Overview:
11787"""""""""
11788
11789The '``llvm.lifetime.start``' intrinsic specifies the start of a memory
11790object's lifetime.
11791
11792Arguments:
11793""""""""""
11794
11795The first argument is a constant integer representing the size of the
11796object, or -1 if it is variable sized. The second argument is a pointer
11797to the object.
11798
11799Semantics:
11800""""""""""
11801
11802This intrinsic indicates that before this point in the code, the value
11803of the memory pointed to by ``ptr`` is dead. This means that it is known
11804to never be used and has an undefined value. A load from the pointer
11805that precedes this intrinsic can be replaced with ``'undef'``.
11806
11807.. _int_lifeend:
11808
11809'``llvm.lifetime.end``' Intrinsic
11810^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11811
11812Syntax:
11813"""""""
11814
11815::
11816
11817      declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
11818
11819Overview:
11820"""""""""
11821
11822The '``llvm.lifetime.end``' intrinsic specifies the end of a memory
11823object's lifetime.
11824
11825Arguments:
11826""""""""""
11827
11828The first argument is a constant integer representing the size of the
11829object, or -1 if it is variable sized. The second argument is a pointer
11830to the object.
11831
11832Semantics:
11833""""""""""
11834
11835This intrinsic indicates that after this point in the code, the value of
11836the memory pointed to by ``ptr`` is dead. This means that it is known to
11837never be used and has an undefined value. Any stores into the memory
11838object following this intrinsic may be removed as dead.
11839
11840'``llvm.invariant.start``' Intrinsic
11841^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11842
11843Syntax:
11844"""""""
11845
11846::
11847
11848      declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
11849
11850Overview:
11851"""""""""
11852
11853The '``llvm.invariant.start``' intrinsic specifies that the contents of
11854a memory object will not change.
11855
11856Arguments:
11857""""""""""
11858
11859The first argument is a constant integer representing the size of the
11860object, or -1 if it is variable sized. The second argument is a pointer
11861to the object.
11862
11863Semantics:
11864""""""""""
11865
11866This intrinsic indicates that until an ``llvm.invariant.end`` that uses
11867the return value, the referenced memory location is constant and
11868unchanging.
11869
11870'``llvm.invariant.end``' Intrinsic
11871^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11872
11873Syntax:
11874"""""""
11875
11876::
11877
11878      declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
11879
11880Overview:
11881"""""""""
11882
11883The '``llvm.invariant.end``' intrinsic specifies that the contents of a
11884memory object are mutable.
11885
11886Arguments:
11887""""""""""
11888
11889The first argument is the matching ``llvm.invariant.start`` intrinsic.
11890The second argument is a constant integer representing the size of the
11891object, or -1 if it is variable sized and the third argument is a
11892pointer to the object.
11893
11894Semantics:
11895""""""""""
11896
11897This intrinsic indicates that the memory is mutable again.
11898
11899'``llvm.invariant.group.barrier``' Intrinsic
11900^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11901
11902Syntax:
11903"""""""
11904
11905::
11906
11907      declare i8* @llvm.invariant.group.barrier(i8* <ptr>)
11908
11909Overview:
11910"""""""""
11911
11912The '``llvm.invariant.group.barrier``' intrinsic can be used when an invariant
11913established by invariant.group metadata no longer holds, to obtain a new pointer
11914value that does not carry the invariant information.
11915
11916
11917Arguments:
11918""""""""""
11919
11920The ``llvm.invariant.group.barrier`` takes only one argument, which is
11921the pointer to the memory for which the ``invariant.group`` no longer holds.
11922
11923Semantics:
11924""""""""""
11925
11926Returns another pointer that aliases its argument but which is considered different
11927for the purposes of ``load``/``store`` ``invariant.group`` metadata.
11928
11929General Intrinsics
11930------------------
11931
11932This class of intrinsics is designed to be generic and has no specific
11933purpose.
11934
11935'``llvm.var.annotation``' Intrinsic
11936^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11937
11938Syntax:
11939"""""""
11940
11941::
11942
11943      declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32  <int>)
11944
11945Overview:
11946"""""""""
11947
11948The '``llvm.var.annotation``' intrinsic.
11949
11950Arguments:
11951""""""""""
11952
11953The first argument is a pointer to a value, the second is a pointer to a
11954global string, the third is a pointer to a global string which is the
11955source file name, and the last argument is the line number.
11956
11957Semantics:
11958""""""""""
11959
11960This intrinsic allows annotation of local variables with arbitrary
11961strings. This can be useful for special purpose optimizations that want
11962to look for these annotations. These have no other defined use; they are
11963ignored by code generation and optimization.
11964
11965'``llvm.ptr.annotation.*``' Intrinsic
11966^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11967
11968Syntax:
11969"""""""
11970
11971This is an overloaded intrinsic. You can use '``llvm.ptr.annotation``' on a
11972pointer to an integer of any width. *NOTE* you must specify an address space for
11973the pointer. The identifier for the default address space is the integer
11974'``0``'.
11975
11976::
11977
11978      declare i8*   @llvm.ptr.annotation.p<address space>i8(i8* <val>, i8* <str>, i8* <str>, i32  <int>)
11979      declare i16*  @llvm.ptr.annotation.p<address space>i16(i16* <val>, i8* <str>, i8* <str>, i32  <int>)
11980      declare i32*  @llvm.ptr.annotation.p<address space>i32(i32* <val>, i8* <str>, i8* <str>, i32  <int>)
11981      declare i64*  @llvm.ptr.annotation.p<address space>i64(i64* <val>, i8* <str>, i8* <str>, i32  <int>)
11982      declare i256* @llvm.ptr.annotation.p<address space>i256(i256* <val>, i8* <str>, i8* <str>, i32  <int>)
11983
11984Overview:
11985"""""""""
11986
11987The '``llvm.ptr.annotation``' intrinsic.
11988
11989Arguments:
11990""""""""""
11991
11992The first argument is a pointer to an integer value of arbitrary bitwidth
11993(result of some expression), the second is a pointer to a global string, the
11994third is a pointer to a global string which is the source file name, and the
11995last argument is the line number. It returns the value of the first argument.
11996
11997Semantics:
11998""""""""""
11999
12000This intrinsic allows annotation of a pointer to an integer with arbitrary
12001strings. This can be useful for special purpose optimizations that want to look
12002for these annotations. These have no other defined use; they are ignored by code
12003generation and optimization.
12004
12005'``llvm.annotation.*``' Intrinsic
12006^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12007
12008Syntax:
12009"""""""
12010
12011This is an overloaded intrinsic. You can use '``llvm.annotation``' on
12012any integer bit width.
12013
12014::
12015
12016      declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32  <int>)
12017      declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32  <int>)
12018      declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32  <int>)
12019      declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32  <int>)
12020      declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32  <int>)
12021
12022Overview:
12023"""""""""
12024
12025The '``llvm.annotation``' intrinsic.
12026
12027Arguments:
12028""""""""""
12029
12030The first argument is an integer value (result of some expression), the
12031second is a pointer to a global string, the third is a pointer to a
12032global string which is the source file name, and the last argument is
12033the line number. It returns the value of the first argument.
12034
12035Semantics:
12036""""""""""
12037
12038This intrinsic allows annotations to be put on arbitrary expressions
12039with arbitrary strings. This can be useful for special purpose
12040optimizations that want to look for these annotations. These have no
12041other defined use; they are ignored by code generation and optimization.
12042
12043'``llvm.trap``' Intrinsic
12044^^^^^^^^^^^^^^^^^^^^^^^^^
12045
12046Syntax:
12047"""""""
12048
12049::
12050
12051      declare void @llvm.trap() noreturn nounwind
12052
12053Overview:
12054"""""""""
12055
12056The '``llvm.trap``' intrinsic.
12057
12058Arguments:
12059""""""""""
12060
12061None.
12062
12063Semantics:
12064""""""""""
12065
12066This intrinsic is lowered to the target dependent trap instruction. If
12067the target does not have a trap instruction, this intrinsic will be
12068lowered to a call of the ``abort()`` function.
12069
12070'``llvm.debugtrap``' Intrinsic
12071^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12072
12073Syntax:
12074"""""""
12075
12076::
12077
12078      declare void @llvm.debugtrap() nounwind
12079
12080Overview:
12081"""""""""
12082
12083The '``llvm.debugtrap``' intrinsic.
12084
12085Arguments:
12086""""""""""
12087
12088None.
12089
12090Semantics:
12091""""""""""
12092
12093This intrinsic is lowered to code which is intended to cause an
12094execution trap with the intention of requesting the attention of a
12095debugger.
12096
12097'``llvm.stackprotector``' Intrinsic
12098^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12099
12100Syntax:
12101"""""""
12102
12103::
12104
12105      declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
12106
12107Overview:
12108"""""""""
12109
12110The ``llvm.stackprotector`` intrinsic takes the ``guard`` and stores it
12111onto the stack at ``slot``. The stack slot is adjusted to ensure that it
12112is placed on the stack before local variables.
12113
12114Arguments:
12115""""""""""
12116
12117The ``llvm.stackprotector`` intrinsic requires two pointer arguments.
12118The first argument is the value loaded from the stack guard
12119``@__stack_chk_guard``. The second variable is an ``alloca`` that has
12120enough space to hold the value of the guard.
12121
12122Semantics:
12123""""""""""
12124
12125This intrinsic causes the prologue/epilogue inserter to force the position of
12126the ``AllocaInst`` stack slot to be before local variables on the stack. This is
12127to ensure that if a local variable on the stack is overwritten, it will destroy
12128the value of the guard. When the function exits, the guard on the stack is
12129checked against the original guard by ``llvm.stackprotectorcheck``. If they are
12130different, then ``llvm.stackprotectorcheck`` causes the program to abort by
12131calling the ``__stack_chk_fail()`` function.
12132
12133'``llvm.stackguard``' Intrinsic
12134^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12135
12136Syntax:
12137"""""""
12138
12139::
12140
12141      declare i8* @llvm.stackguard()
12142
12143Overview:
12144"""""""""
12145
12146The ``llvm.stackguard`` intrinsic returns the system stack guard value.
12147
12148It should not be generated by frontends, since it is only for internal usage.
12149The reason why we create this intrinsic is that we still support IR form Stack
12150Protector in FastISel.
12151
12152Arguments:
12153""""""""""
12154
12155None.
12156
12157Semantics:
12158""""""""""
12159
12160On some platforms, the value returned by this intrinsic remains unchanged
12161between loads in the same thread. On other platforms, it returns the same
12162global variable value, if any, e.g. ``@__stack_chk_guard``.
12163
12164Currently some platforms have IR-level customized stack guard loading (e.g.
12165X86 Linux) that is not handled by ``llvm.stackguard()``, while they should be
12166in the future.
12167
12168'``llvm.objectsize``' Intrinsic
12169^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12170
12171Syntax:
12172"""""""
12173
12174::
12175
12176      declare i32 @llvm.objectsize.i32(i8* <object>, i1 <min>)
12177      declare i64 @llvm.objectsize.i64(i8* <object>, i1 <min>)
12178
12179Overview:
12180"""""""""
12181
12182The ``llvm.objectsize`` intrinsic is designed to provide information to
12183the optimizers to determine at compile time whether a) an operation
12184(like memcpy) will overflow a buffer that corresponds to an object, or
12185b) that a runtime check for overflow isn't necessary. An object in this
12186context means an allocation of a specific class, structure, array, or
12187other object.
12188
12189Arguments:
12190""""""""""
12191
12192The ``llvm.objectsize`` intrinsic takes two arguments. The first
12193argument is a pointer to or into the ``object``. The second argument is
12194a boolean and determines whether ``llvm.objectsize`` returns 0 (if true)
12195or -1 (if false) when the object size is unknown. The second argument
12196only accepts constants.
12197
12198Semantics:
12199""""""""""
12200
12201The ``llvm.objectsize`` intrinsic is lowered to a constant representing
12202the size of the object concerned. If the size cannot be determined at
12203compile time, ``llvm.objectsize`` returns ``i32/i64 -1 or 0`` (depending
12204on the ``min`` argument).
12205
12206'``llvm.expect``' Intrinsic
12207^^^^^^^^^^^^^^^^^^^^^^^^^^^
12208
12209Syntax:
12210"""""""
12211
12212This is an overloaded intrinsic. You can use ``llvm.expect`` on any
12213integer bit width.
12214
12215::
12216
12217      declare i1 @llvm.expect.i1(i1 <val>, i1 <expected_val>)
12218      declare i32 @llvm.expect.i32(i32 <val>, i32 <expected_val>)
12219      declare i64 @llvm.expect.i64(i64 <val>, i64 <expected_val>)
12220
12221Overview:
12222"""""""""
12223
12224The ``llvm.expect`` intrinsic provides information about expected (the
12225most probable) value of ``val``, which can be used by optimizers.
12226
12227Arguments:
12228""""""""""
12229
12230The ``llvm.expect`` intrinsic takes two arguments. The first argument is
12231a value. The second argument is an expected value, this needs to be a
12232constant value, variables are not allowed.
12233
12234Semantics:
12235""""""""""
12236
12237This intrinsic is lowered to the ``val``.
12238
12239.. _int_assume:
12240
12241'``llvm.assume``' Intrinsic
12242^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12243
12244Syntax:
12245"""""""
12246
12247::
12248
12249      declare void @llvm.assume(i1 %cond)
12250
12251Overview:
12252"""""""""
12253
12254The ``llvm.assume`` allows the optimizer to assume that the provided
12255condition is true. This information can then be used in simplifying other parts
12256of the code.
12257
12258Arguments:
12259""""""""""
12260
12261The condition which the optimizer may assume is always true.
12262
12263Semantics:
12264""""""""""
12265
12266The intrinsic allows the optimizer to assume that the provided condition is
12267always true whenever the control flow reaches the intrinsic call. No code is
12268generated for this intrinsic, and instructions that contribute only to the
12269provided condition are not used for code generation. If the condition is
12270violated during execution, the behavior is undefined.
12271
12272Note that the optimizer might limit the transformations performed on values
12273used by the ``llvm.assume`` intrinsic in order to preserve the instructions
12274only used to form the intrinsic's input argument. This might prove undesirable
12275if the extra information provided by the ``llvm.assume`` intrinsic does not cause
12276sufficient overall improvement in code quality. For this reason,
12277``llvm.assume`` should not be used to document basic mathematical invariants
12278that the optimizer can otherwise deduce or facts that are of little use to the
12279optimizer.
12280
12281.. _type.test:
12282
12283'``llvm.type.test``' Intrinsic
12284^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12285
12286Syntax:
12287"""""""
12288
12289::
12290
12291      declare i1 @llvm.type.test(i8* %ptr, metadata %type) nounwind readnone
12292
12293
12294Arguments:
12295""""""""""
12296
12297The first argument is a pointer to be tested. The second argument is a
12298metadata object representing a :doc:`type identifier <TypeMetadata>`.
12299
12300Overview:
12301"""""""""
12302
12303The ``llvm.type.test`` intrinsic tests whether the given pointer is associated
12304with the given type identifier.
12305
12306'``llvm.type.checked.load``' Intrinsic
12307^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12308
12309Syntax:
12310"""""""
12311
12312::
12313
12314      declare {i8*, i1} @llvm.type.checked.load(i8* %ptr, i32 %offset, metadata %type) argmemonly nounwind readonly
12315
12316
12317Arguments:
12318""""""""""
12319
12320The first argument is a pointer from which to load a function pointer. The
12321second argument is the byte offset from which to load the function pointer. The
12322third argument is a metadata object representing a :doc:`type identifier
12323<TypeMetadata>`.
12324
12325Overview:
12326"""""""""
12327
12328The ``llvm.type.checked.load`` intrinsic safely loads a function pointer from a
12329virtual table pointer using type metadata. This intrinsic is used to implement
12330control flow integrity in conjunction with virtual call optimization. The
12331virtual call optimization pass will optimize away ``llvm.type.checked.load``
12332intrinsics associated with devirtualized calls, thereby removing the type
12333check in cases where it is not needed to enforce the control flow integrity
12334constraint.
12335
12336If the given pointer is associated with a type metadata identifier, this
12337function returns true as the second element of its return value. (Note that
12338the function may also return true if the given pointer is not associated
12339with a type metadata identifier.) If the function's return value's second
12340element is true, the following rules apply to the first element:
12341
12342- If the given pointer is associated with the given type metadata identifier,
12343  it is the function pointer loaded from the given byte offset from the given
12344  pointer.
12345
12346- If the given pointer is not associated with the given type metadata
12347  identifier, it is one of the following (the choice of which is unspecified):
12348
12349  1. The function pointer that would have been loaded from an arbitrarily chosen
12350     (through an unspecified mechanism) pointer associated with the type
12351     metadata.
12352
12353  2. If the function has a non-void return type, a pointer to a function that
12354     returns an unspecified value without causing side effects.
12355
12356If the function's return value's second element is false, the value of the
12357first element is undefined.
12358
12359
12360'``llvm.donothing``' Intrinsic
12361^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12362
12363Syntax:
12364"""""""
12365
12366::
12367
12368      declare void @llvm.donothing() nounwind readnone
12369
12370Overview:
12371"""""""""
12372
12373The ``llvm.donothing`` intrinsic doesn't perform any operation. It's one of only
12374three intrinsics (besides ``llvm.experimental.patchpoint`` and
12375``llvm.experimental.gc.statepoint``) that can be called with an invoke
12376instruction.
12377
12378Arguments:
12379""""""""""
12380
12381None.
12382
12383Semantics:
12384""""""""""
12385
12386This intrinsic does nothing, and it's removed by optimizers and ignored
12387by codegen.
12388
12389'``llvm.experimental.deoptimize``' Intrinsic
12390^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12391
12392Syntax:
12393"""""""
12394
12395::
12396
12397      declare type @llvm.experimental.deoptimize(...) [ "deopt"(...) ]
12398
12399Overview:
12400"""""""""
12401
12402This intrinsic, together with :ref:`deoptimization operand bundles
12403<deopt_opbundles>`, allow frontends to express transfer of control and
12404frame-local state from the currently executing (typically more specialized,
12405hence faster) version of a function into another (typically more generic, hence
12406slower) version.
12407
12408In languages with a fully integrated managed runtime like Java and JavaScript
12409this intrinsic can be used to implement "uncommon trap" or "side exit" like
12410functionality.  In unmanaged languages like C and C++, this intrinsic can be
12411used to represent the slow paths of specialized functions.
12412
12413
12414Arguments:
12415""""""""""
12416
12417The intrinsic takes an arbitrary number of arguments, whose meaning is
12418decided by the :ref:`lowering strategy<deoptimize_lowering>`.
12419
12420Semantics:
12421""""""""""
12422
12423The ``@llvm.experimental.deoptimize`` intrinsic executes an attached
12424deoptimization continuation (denoted using a :ref:`deoptimization
12425operand bundle <deopt_opbundles>`) and returns the value returned by
12426the deoptimization continuation.  Defining the semantic properties of
12427the continuation itself is out of scope of the language reference --
12428as far as LLVM is concerned, the deoptimization continuation can
12429invoke arbitrary side effects, including reading from and writing to
12430the entire heap.
12431
12432Deoptimization continuations expressed using ``"deopt"`` operand bundles always
12433continue execution to the end of the physical frame containing them, so all
12434calls to ``@llvm.experimental.deoptimize`` must be in "tail position":
12435
12436   - ``@llvm.experimental.deoptimize`` cannot be invoked.
12437   - The call must immediately precede a :ref:`ret <i_ret>` instruction.
12438   - The ``ret`` instruction must return the value produced by the
12439     ``@llvm.experimental.deoptimize`` call if there is one, or void.
12440
12441Note that the above restrictions imply that the return type for a call to
12442``@llvm.experimental.deoptimize`` will match the return type of its immediate
12443caller.
12444
12445The inliner composes the ``"deopt"`` continuations of the caller into the
12446``"deopt"`` continuations present in the inlinee, and also updates calls to this
12447intrinsic to return directly from the frame of the function it inlined into.
12448
12449All declarations of ``@llvm.experimental.deoptimize`` must share the
12450same calling convention.
12451
12452.. _deoptimize_lowering:
12453
12454Lowering:
12455"""""""""
12456
12457Calls to ``@llvm.experimental.deoptimize`` are lowered to calls to the
12458symbol ``__llvm_deoptimize`` (it is the frontend's responsibility to
12459ensure that this symbol is defined).  The call arguments to
12460``@llvm.experimental.deoptimize`` are lowered as if they were formal
12461arguments of the specified types, and not as varargs.
12462
12463
12464'``llvm.experimental.guard``' Intrinsic
12465^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12466
12467Syntax:
12468"""""""
12469
12470::
12471
12472      declare void @llvm.experimental.guard(i1, ...) [ "deopt"(...) ]
12473
12474Overview:
12475"""""""""
12476
12477This intrinsic, together with :ref:`deoptimization operand bundles
12478<deopt_opbundles>`, allows frontends to express guards or checks on
12479optimistic assumptions made during compilation.  The semantics of
12480``@llvm.experimental.guard`` is defined in terms of
12481``@llvm.experimental.deoptimize`` -- its body is defined to be
12482equivalent to:
12483
12484.. code-block:: llvm
12485
12486	define void @llvm.experimental.guard(i1 %pred, <args...>) {
12487	  %realPred = and i1 %pred, undef
12488	  br i1 %realPred, label %continue, label %leave [, !make.implicit !{}]
12489
12490	leave:
12491	  call void @llvm.experimental.deoptimize(<args...>) [ "deopt"() ]
12492	  ret void
12493
12494	continue:
12495	  ret void
12496	}
12497
12498
12499with the optional ``[, !make.implicit !{}]`` present if and only if it
12500is present on the call site.  For more details on ``!make.implicit``,
12501see :doc:`FaultMaps`.
12502
12503In words, ``@llvm.experimental.guard`` executes the attached
12504``"deopt"`` continuation if (but **not** only if) its first argument
12505is ``false``.  Since the optimizer is allowed to replace the ``undef``
12506with an arbitrary value, it can optimize guard to fail "spuriously",
12507i.e. without the original condition being false (hence the "not only
12508if"); and this allows for "check widening" type optimizations.
12509
12510``@llvm.experimental.guard`` cannot be invoked.
12511
12512
12513'``llvm.load.relative``' Intrinsic
12514^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12515
12516Syntax:
12517"""""""
12518
12519::
12520
12521      declare i8* @llvm.load.relative.iN(i8* %ptr, iN %offset) argmemonly nounwind readonly
12522
12523Overview:
12524"""""""""
12525
12526This intrinsic loads a 32-bit value from the address ``%ptr + %offset``,
12527adds ``%ptr`` to that value and returns it. The constant folder specifically
12528recognizes the form of this intrinsic and the constant initializers it may
12529load from; if a loaded constant initializer is known to have the form
12530``i32 trunc(x - %ptr)``, the intrinsic call is folded to ``x``.
12531
12532LLVM provides that the calculation of such a constant initializer will
12533not overflow at link time under the medium code model if ``x`` is an
12534``unnamed_addr`` function. However, it does not provide this guarantee for
12535a constant initializer folded into a function body. This intrinsic can be
12536used to avoid the possibility of overflows when loading from such a constant.
12537
12538Stack Map Intrinsics
12539--------------------
12540
12541LLVM provides experimental intrinsics to support runtime patching
12542mechanisms commonly desired in dynamic language JITs. These intrinsics
12543are described in :doc:`StackMaps`.
12544