1 ==============================
2 LLVM Language Reference Manual
3 ==============================
12 This document is a reference manual for the LLVM assembly language. LLVM
13 is a Static Single Assignment (SSA) based representation that provides
14 type safety, low-level operations, flexibility, and the capability of
15 representing 'all' high-level languages cleanly. It is the common code
16 representation used throughout all phases of the LLVM compilation
22 The LLVM code representation is designed to be used in three different
23 forms: as an in-memory compiler IR, as an on-disk bitcode representation
24 (suitable for fast loading by a Just-In-Time compiler), and as a human
25 readable assembly language representation. This allows LLVM to provide a
26 powerful intermediate representation for efficient compiler
27 transformations and analysis, while providing a natural means to debug
28 and visualize the transformations. The three different forms of LLVM are
29 all equivalent. This document describes the human readable
30 representation and notation.
32 The LLVM representation aims to be light-weight and low-level while
33 being expressive, typed, and extensible at the same time. It aims to be
34 a "universal IR" of sorts, by being at a low enough level that
35 high-level ideas may be cleanly mapped to it (similar to how
36 microprocessors are "universal IR's", allowing many source languages to
37 be mapped to them). By providing type information, LLVM can be used as
38 the target of optimizations: for example, through pointer analysis, it
39 can be proven that a C automatic variable is never accessed outside of
40 the current function, allowing it to be promoted to a simple SSA value
41 instead of a memory location.
48 It is important to note that this document describes 'well formed' LLVM
49 assembly language. There is a difference between what the parser accepts
50 and what is considered 'well formed'. For example, the following
51 instruction is syntactically okay, but not well formed:
57 because the definition of ``%x`` does not dominate all of its uses. The
58 LLVM infrastructure provides a verification pass that may be used to
59 verify that an LLVM module is well formed. This pass is automatically
60 run by the parser after parsing input assembly and by the optimizer
61 before it outputs bitcode. The violations pointed out by the verifier
62 pass indicate bugs in transformation passes or input to the parser.
69 LLVM identifiers come in two basic types: global and local. Global
70 identifiers (functions, global variables) begin with the ``'@'``
71 character. Local identifiers (register names, types) begin with the
72 ``'%'`` character. Additionally, there are three different formats for
73 identifiers, for different purposes:
75 #. Named values are represented as a string of characters with their
76 prefix. For example, ``%foo``, ``@DivisionByZero``,
77 ``%a.really.long.identifier``. The actual regular expression used is
78 '``[%@][a-zA-Z$._][a-zA-Z$._0-9]*``'. Identifiers which 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.
83 #. Unnamed values are represented as an unsigned numeric value with
84 their prefix. For example, ``%12``, ``@2``, ``%44``.
85 #. Constants, which are described in the section Constants_ below.
87 LLVM requires that values start with a prefix for two reasons: Compilers
88 don't need to worry about name clashes with reserved words, and the set
89 of reserved words may be expanded in the future without penalty.
90 Additionally, unnamed identifiers allow a compiler to quickly come up
91 with a temporary variable without having to avoid symbol table
94 Reserved words in LLVM are very similar to reserved words in other
95 languages. There are keywords for different opcodes ('``add``',
96 '``bitcast``', '``ret``', etc...), for primitive type names ('``void``',
97 '``i32``', etc...), and others. These reserved words cannot conflict
98 with variable names, because none of them start with a prefix character
101 Here is an example of LLVM code to multiply the integer variable
108 %result = mul i32 %X, 8
110 After strength reduction:
114 %result = shl i32 %X, 3
120 %0 = add i32 %X, %X ; yields {i32}:%0
121 %1 = add i32 %0, %0 ; yields {i32}:%1
122 %result = add i32 %1, %1
124 This last way of multiplying ``%X`` by 8 illustrates several important
125 lexical features of LLVM:
127 #. Comments are delimited with a '``;``' and go until the end of line.
128 #. Unnamed temporaries are created when the result of a computation is
129 not assigned to a named value.
130 #. Unnamed temporaries are numbered sequentially (using a per-function
131 incrementing counter, starting with 0).
133 It also shows a convention that we follow in this document. When
134 demonstrating instructions, we will follow an instruction with a comment
135 that defines the type and name of value produced.
143 LLVM programs are composed of ``Module``'s, each of which is a
144 translation unit of the input programs. Each module consists of
145 functions, global variables, and symbol table entries. Modules may be
146 combined together with the LLVM linker, which merges function (and
147 global variable) definitions, resolves forward declarations, and merges
148 symbol table entries. Here is an example of the "hello world" module:
152 ; Declare the string constant as a global constant.
153 @.str = private unnamed_addr constant [13 x i8] c"hello world\0A\00"
155 ; External declaration of the puts function
156 declare i32 @puts(i8* nocapture) nounwind
158 ; Definition of main function
159 define i32 @main() { ; i32()*
160 ; Convert [13 x i8]* to i8 *...
161 %cast210 = getelementptr [13 x i8]* @.str, i64 0, i64 0
163 ; Call puts function to write out the string to stdout.
164 call i32 @puts(i8* %cast210)
169 !1 = metadata !{i32 42}
172 This example is made up of a :ref:`global variable <globalvars>` named
173 "``.str``", an external declaration of the "``puts``" function, a
174 :ref:`function definition <functionstructure>` for "``main``" and
175 :ref:`named metadata <namedmetadatastructure>` "``foo``".
177 In general, a module is made up of a list of global values (where both
178 functions and global variables are global values). Global values are
179 represented by a pointer to a memory location (in this case, a pointer
180 to an array of char, and a pointer to a function), and have one of the
181 following :ref:`linkage types <linkage>`.
188 All Global Variables and Functions have one of the following types of
192 Global values with "``private``" linkage are only directly
193 accessible by objects in the current module. In particular, linking
194 code into a module with an private global value may cause the
195 private to be renamed as necessary to avoid collisions. Because the
196 symbol is private to the module, all references can be updated. This
197 doesn't show up in any symbol table in the object file.
199 Similar to ``private``, but the symbol is passed through the
200 assembler and evaluated by the linker. Unlike normal strong symbols,
201 they are removed by the linker from the final linked image
202 (executable or dynamic library).
203 ``linker_private_weak``
204 Similar to "``linker_private``", but the symbol is weak. Note that
205 ``linker_private_weak`` symbols are subject to coalescing by the
206 linker. The symbols are removed by the linker from the final linked
207 image (executable or dynamic library).
209 Similar to private, but the value shows as a local symbol
210 (``STB_LOCAL`` in the case of ELF) in the object file. This
211 corresponds to the notion of the '``static``' keyword in C.
212 ``available_externally``
213 Globals with "``available_externally``" linkage are never emitted
214 into the object file corresponding to the LLVM module. They exist to
215 allow inlining and other optimizations to take place given knowledge
216 of the definition of the global, which is known to be somewhere
217 outside the module. Globals with ``available_externally`` linkage
218 are allowed to be discarded at will, and are otherwise the same as
219 ``linkonce_odr``. This linkage type is only allowed on definitions,
222 Globals with "``linkonce``" linkage are merged with other globals of
223 the same name when linkage occurs. This can be used to implement
224 some forms of inline functions, templates, or other code which must
225 be generated in each translation unit that uses it, but where the
226 body may be overridden with a more definitive definition later.
227 Unreferenced ``linkonce`` globals are allowed to be discarded. Note
228 that ``linkonce`` linkage does not actually allow the optimizer to
229 inline the body of this function into callers because it doesn't
230 know if this definition of the function is the definitive definition
231 within the program or whether it will be overridden by a stronger
232 definition. To enable inlining and other optimizations, use
233 "``linkonce_odr``" linkage.
235 "``weak``" linkage has the same merging semantics as ``linkonce``
236 linkage, except that unreferenced globals with ``weak`` linkage may
237 not be discarded. This is used for globals that are declared "weak"
240 "``common``" linkage is most similar to "``weak``" linkage, but they
241 are used for tentative definitions in C, such as "``int X;``" at
242 global scope. Symbols with "``common``" linkage are merged in the
243 same way as ``weak symbols``, and they may not be deleted if
244 unreferenced. ``common`` symbols may not have an explicit section,
245 must have a zero initializer, and may not be marked
246 ':ref:`constant <globalvars>`'. Functions and aliases may not have
249 .. _linkage_appending:
252 "``appending``" linkage may only be applied to global variables of
253 pointer to array type. When two global variables with appending
254 linkage are linked together, the two global arrays are appended
255 together. This is the LLVM, typesafe, equivalent of having the
256 system linker append together "sections" with identical names when
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 ``linkonce_odr_auto_hide``
271 Similar to "``linkonce_odr``", but nothing in the translation unit
272 takes the address of this definition. For instance, functions that
273 had an inline definition, but the compiler decided not to inline it.
274 ``linkonce_odr_auto_hide`` may have only ``default`` visibility. The
275 symbols are removed by the linker from the final linked image
276 (executable or dynamic library).
278 If none of the above identifiers are used, the global is externally
279 visible, meaning that it participates in linkage and can be used to
280 resolve external symbol references.
282 The next two types of linkage are targeted for Microsoft Windows
283 platform only. They are designed to support importing (exporting)
284 symbols from (to) DLLs (Dynamic Link Libraries).
287 "``dllimport``" linkage causes the compiler to reference a function
288 or variable via a global pointer to a pointer that is set up by the
289 DLL exporting the symbol. On Microsoft Windows targets, the pointer
290 name is formed by combining ``__imp_`` and the function or variable
293 "``dllexport``" linkage causes the compiler to provide a global
294 pointer to a pointer in a DLL, so that it can be referenced with the
295 ``dllimport`` attribute. On Microsoft Windows targets, the pointer
296 name is formed by combining ``__imp_`` and the function or variable
299 For example, since the "``.LC0``" variable is defined to be internal, if
300 another module defined a "``.LC0``" variable and was linked with this
301 one, one of the two would be renamed, preventing a collision. Since
302 "``main``" and "``puts``" are external (i.e., lacking any linkage
303 declarations), they are accessible outside of the current module.
305 It is illegal for a function *declaration* to have any linkage type
306 other than ``external``, ``dllimport`` or ``extern_weak``.
313 LLVM :ref:`functions <functionstructure>`, :ref:`calls <i_call>` and
314 :ref:`invokes <i_invoke>` can all have an optional calling convention
315 specified for the call. The calling convention of any pair of dynamic
316 caller/callee must match, or the behavior of the program is undefined.
317 The following calling conventions are supported by LLVM, and more may be
320 "``ccc``" - The C calling convention
321 This calling convention (the default if no other calling convention
322 is specified) matches the target C calling conventions. This calling
323 convention supports varargs function calls and tolerates some
324 mismatch in the declared prototype and implemented declaration of
325 the function (as does normal C).
326 "``fastcc``" - The fast calling convention
327 This calling convention attempts to make calls as fast as possible
328 (e.g. by passing things in registers). This calling convention
329 allows the target to use whatever tricks it wants to produce fast
330 code for the target, without having to conform to an externally
331 specified ABI (Application Binary Interface). `Tail calls can only
332 be optimized when this, the GHC or the HiPE convention is
333 used. <CodeGenerator.html#id80>`_ This calling convention does not
334 support varargs and requires the prototype of all callees to exactly
335 match the prototype of the function definition.
336 "``coldcc``" - The cold calling convention
337 This calling convention attempts to make code in the caller as
338 efficient as possible under the assumption that the call is not
339 commonly executed. As such, these calls often preserve all registers
340 so that the call does not break any live ranges in the caller side.
341 This calling convention does not support varargs and requires the
342 prototype of all callees to exactly match the prototype of the
344 "``cc 10``" - GHC convention
345 This calling convention has been implemented specifically for use by
346 the `Glasgow Haskell Compiler (GHC) <http://www.haskell.org/ghc>`_.
347 It passes everything in registers, going to extremes to achieve this
348 by disabling callee save registers. This calling convention should
349 not be used lightly but only for specific situations such as an
350 alternative to the *register pinning* performance technique often
351 used when implementing functional programming languages. At the
352 moment only X86 supports this convention and it has the following
355 - On *X86-32* only supports up to 4 bit type parameters. No
356 floating point types are supported.
357 - On *X86-64* only supports up to 10 bit type parameters and 6
358 floating point parameters.
360 This calling convention supports `tail call
361 optimization <CodeGenerator.html#id80>`_ but requires both the
362 caller and callee are using it.
363 "``cc 11``" - The HiPE calling convention
364 This calling convention has been implemented specifically for use by
365 the `High-Performance Erlang
366 (HiPE) <http://www.it.uu.se/research/group/hipe/>`_ compiler, *the*
367 native code compiler of the `Ericsson's Open Source Erlang/OTP
368 system <http://www.erlang.org/download.shtml>`_. It uses more
369 registers for argument passing than the ordinary C calling
370 convention and defines no callee-saved registers. The calling
371 convention properly supports `tail call
372 optimization <CodeGenerator.html#id80>`_ but requires that both the
373 caller and the callee use it. It uses a *register pinning*
374 mechanism, similar to GHC's convention, for keeping frequently
375 accessed runtime components pinned to specific hardware registers.
376 At the moment only X86 supports this convention (both 32 and 64
378 "``cc <n>``" - Numbered convention
379 Any calling convention may be specified by number, allowing
380 target-specific calling conventions to be used. Target specific
381 calling conventions start at 64.
383 More calling conventions can be added/defined on an as-needed basis, to
384 support Pascal conventions or any other well-known target-independent
387 .. _visibilitystyles:
392 All Global Variables and Functions have one of the following visibility
395 "``default``" - Default style
396 On targets that use the ELF object file format, default visibility
397 means that the declaration is visible to other modules and, in
398 shared libraries, means that the declared entity may be overridden.
399 On Darwin, default visibility means that the declaration is visible
400 to other modules. Default visibility corresponds to "external
401 linkage" in the language.
402 "``hidden``" - Hidden style
403 Two declarations of an object with hidden visibility refer to the
404 same object if they are in the same shared object. Usually, hidden
405 visibility indicates that the symbol will not be placed into the
406 dynamic symbol table, so no other module (executable or shared
407 library) can reference it directly.
408 "``protected``" - Protected style
409 On ELF, protected visibility indicates that the symbol will be
410 placed in the dynamic symbol table, but that references within the
411 defining module will bind to the local symbol. That is, the symbol
412 cannot be overridden by another module.
419 LLVM IR allows you to specify name aliases for certain types. This can
420 make it easier to read the IR and make the IR more condensed
421 (particularly when recursive types are involved). An example of a name
426 %mytype = type { %mytype*, i32 }
428 You may give a name to any :ref:`type <typesystem>` except
429 ":ref:`void <t_void>`". Type name aliases may be used anywhere a type is
430 expected with the syntax "%mytype".
432 Note that type names are aliases for the structural type that they
433 indicate, and that you can therefore specify multiple names for the same
434 type. This often leads to confusing behavior when dumping out a .ll
435 file. Since LLVM IR uses structural typing, the name is not part of the
436 type. When printing out LLVM IR, the printer will pick *one name* to
437 render all types of a particular shape. This means that if you have code
438 where two different source types end up having the same LLVM type, that
439 the dumper will sometimes print the "wrong" or unexpected type. This is
440 an important design point and isn't going to change.
447 Global variables define regions of memory allocated at compilation time
448 instead of run-time. Global variables may optionally be initialized, may
449 have an explicit section to be placed in, and may have an optional
450 explicit alignment specified.
452 A variable may be defined as ``thread_local``, which means that it will
453 not be shared by threads (each thread will have a separated copy of the
454 variable). Not all targets support thread-local variables. Optionally, a
455 TLS model may be specified:
458 For variables that are only used within the current shared library.
460 For variables in modules that will not be loaded dynamically.
462 For variables defined in the executable and only used within it.
464 The models correspond to the ELF TLS models; see `ELF Handling For
465 Thread-Local Storage <http://people.redhat.com/drepper/tls.pdf>`_ for
466 more information on under which circumstances the different models may
467 be used. The target may choose a different TLS model if the specified
468 model is not supported, or if a better choice of model can be made.
470 A variable may be defined as a global ``constant``, which indicates that
471 the contents of the variable will **never** be modified (enabling better
472 optimization, allowing the global data to be placed in the read-only
473 section of an executable, etc). Note that variables that need runtime
474 initialization cannot be marked ``constant`` as there is a store to the
477 LLVM explicitly allows *declarations* of global variables to be marked
478 constant, even if the final definition of the global is not. This
479 capability can be used to enable slightly better optimization of the
480 program, but requires the language definition to guarantee that
481 optimizations based on the 'constantness' are valid for the translation
482 units that do not include the definition.
484 As SSA values, global variables define pointer values that are in scope
485 (i.e. they dominate) all basic blocks in the program. Global variables
486 always define a pointer to their "content" type because they describe a
487 region of memory, and all memory objects in LLVM are accessed through
490 Global variables can be marked with ``unnamed_addr`` which indicates
491 that the address is not significant, only the content. Constants marked
492 like this can be merged with other constants if they have the same
493 initializer. Note that a constant with significant address *can* be
494 merged with a ``unnamed_addr`` constant, the result being a constant
495 whose address is significant.
497 A global variable may be declared to reside in a target-specific
498 numbered address space. For targets that support them, address spaces
499 may affect how optimizations are performed and/or what target
500 instructions are used to access the variable. The default address space
501 is zero. The address space qualifier must precede any other attributes.
503 LLVM allows an explicit section to be specified for globals. If the
504 target supports it, it will emit globals to the section specified.
506 By default, global initializers are optimized by assuming that global
507 variables defined within the module are not modified from their
508 initial values before the start of the global initializer. This is
509 true even for variables potentially accessible from outside the
510 module, including those with external linkage or appearing in
511 ``@llvm.used``. This assumption may be suppressed by marking the
512 variable with ``externally_initialized``.
514 An explicit alignment may be specified for a global, which must be a
515 power of 2. If not present, or if the alignment is set to zero, the
516 alignment of the global is set by the target to whatever it feels
517 convenient. If an explicit alignment is specified, the global is forced
518 to have exactly that alignment. Targets and optimizers are not allowed
519 to over-align the global if the global has an assigned section. In this
520 case, the extra alignment could be observable: for example, code could
521 assume that the globals are densely packed in their section and try to
522 iterate over them as an array, alignment padding would break this
525 For example, the following defines a global in a numbered address space
526 with an initializer, section, and alignment:
530 @G = addrspace(5) constant float 1.0, section "foo", align 4
532 The following example defines a thread-local global with the
533 ``initialexec`` TLS model:
537 @G = thread_local(initialexec) global i32 0, align 4
539 .. _functionstructure:
544 LLVM function definitions consist of the "``define``" keyword, an
545 optional :ref:`linkage type <linkage>`, an optional :ref:`visibility
546 style <visibility>`, an optional :ref:`calling convention <callingconv>`,
547 an optional ``unnamed_addr`` attribute, a return type, an optional
548 :ref:`parameter attribute <paramattrs>` for the return type, a function
549 name, a (possibly empty) argument list (each with optional :ref:`parameter
550 attributes <paramattrs>`), optional :ref:`function attributes <fnattrs>`,
551 an optional section, an optional alignment, an optional :ref:`garbage
552 collector name <gc>`, an optional :ref:`prefix <prefixdata>`, an opening
553 curly brace, a list of basic blocks, and a closing curly brace.
555 LLVM function declarations consist of the "``declare``" keyword, an
556 optional :ref:`linkage type <linkage>`, an optional :ref:`visibility
557 style <visibility>`, an optional :ref:`calling convention <callingconv>`,
558 an optional ``unnamed_addr`` attribute, a return type, an optional
559 :ref:`parameter attribute <paramattrs>` for the return type, a function
560 name, a possibly empty list of arguments, an optional alignment, an optional
561 :ref:`garbage collector name <gc>` and an optional :ref:`prefix <prefixdata>`.
563 A function definition contains a list of basic blocks, forming the CFG
564 (Control Flow Graph) for the function. Each basic block may optionally
565 start with a label (giving the basic block a symbol table entry),
566 contains a list of instructions, and ends with a
567 :ref:`terminator <terminators>` instruction (such as a branch or function
568 return). If explicit label is not provided, a block is assigned an
569 implicit numbered label, using a next value from the same counter as used
570 for unnamed temporaries (:ref:`see above<identifiers>`). For example, if a
571 function entry block does not have explicit label, it will be assigned
572 label "%0", then first unnamed temporary in that block will be "%1", etc.
574 The first basic block in a function is special in two ways: it is
575 immediately executed on entrance to the function, and it is not allowed
576 to have predecessor basic blocks (i.e. there can not be any branches to
577 the entry block of a function). Because the block can have no
578 predecessors, it also cannot have any :ref:`PHI nodes <i_phi>`.
580 LLVM allows an explicit section to be specified for functions. If the
581 target supports it, it will emit functions to the section specified.
583 An explicit alignment may be specified for a function. If not present,
584 or if the alignment is set to zero, the alignment of the function is set
585 by the target to whatever it feels convenient. If an explicit alignment
586 is specified, the function is forced to have at least that much
587 alignment. All alignments must be a power of 2.
589 If the ``unnamed_addr`` attribute is given, the address is know to not
590 be significant and two identical functions can be merged.
594 define [linkage] [visibility]
596 <ResultType> @<FunctionName> ([argument list])
597 [fn Attrs] [section "name"] [align N]
598 [gc] [prefix Constant] { ... }
605 Aliases act as "second name" for the aliasee value (which can be either
606 function, global variable, another alias or bitcast of global value).
607 Aliases may have an optional :ref:`linkage type <linkage>`, and an optional
608 :ref:`visibility style <visibility>`.
612 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
614 The linkage must be one of ``private``, ``linker_private``,
615 ``linker_private_weak``, ``internal``, ``linkonce``, ``weak``,
616 ``linkonce_odr``, ``weak_odr``, ``linkonce_odr_auto_hide``, ``external``. Note
617 that some system linkers might not correctly handle dropping a weak symbol that
618 is aliased by a non weak alias.
620 .. _namedmetadatastructure:
625 Named metadata is a collection of metadata. :ref:`Metadata
626 nodes <metadata>` (but not metadata strings) are the only valid
627 operands for a named metadata.
631 ; Some unnamed metadata nodes, which are referenced by the named metadata.
632 !0 = metadata !{metadata !"zero"}
633 !1 = metadata !{metadata !"one"}
634 !2 = metadata !{metadata !"two"}
636 !name = !{!0, !1, !2}
643 The return type and each parameter of a function type may have a set of
644 *parameter attributes* associated with them. Parameter attributes are
645 used to communicate additional information about the result or
646 parameters of a function. Parameter attributes are considered to be part
647 of the function, not of the function type, so functions with different
648 parameter attributes can have the same function type.
650 Parameter attributes are simple keywords that follow the type specified.
651 If multiple parameter attributes are needed, they are space separated.
656 declare i32 @printf(i8* noalias nocapture, ...)
657 declare i32 @atoi(i8 zeroext)
658 declare signext i8 @returns_signed_char()
660 Note that any attributes for the function result (``nounwind``,
661 ``readonly``) come immediately after the argument list.
663 Currently, only the following parameter attributes are defined:
666 This indicates to the code generator that the parameter or return
667 value should be zero-extended to the extent required by the target's
668 ABI (which is usually 32-bits, but is 8-bits for a i1 on x86-64) by
669 the caller (for a parameter) or the callee (for a return value).
671 This indicates to the code generator that the parameter or return
672 value should be sign-extended to the extent required by the target's
673 ABI (which is usually 32-bits) by the caller (for a parameter) or
674 the callee (for a return value).
676 This indicates that this parameter or return value should be treated
677 in a special target-dependent fashion during while emitting code for
678 a function call or return (usually, by putting it in a register as
679 opposed to memory, though some targets use it to distinguish between
680 two different kinds of registers). Use of this attribute is
683 This indicates that the pointer parameter should really be passed by
684 value to the function. The attribute implies that a hidden copy of
685 the pointee is made between the caller and the callee, so the callee
686 is unable to modify the value in the caller. This attribute is only
687 valid on LLVM pointer arguments. It is generally used to pass
688 structs and arrays by value, but is also valid on pointers to
689 scalars. The copy is considered to belong to the caller not the
690 callee (for example, ``readonly`` functions should not write to
691 ``byval`` parameters). This is not a valid attribute for return
694 The byval attribute also supports specifying an alignment with the
695 align attribute. It indicates the alignment of the stack slot to
696 form and the known alignment of the pointer specified to the call
697 site. If the alignment is not specified, then the code generator
698 makes a target-specific assumption.
701 This indicates that the pointer parameter specifies the address of a
702 structure that is the return value of the function in the source
703 program. This pointer must be guaranteed by the caller to be valid:
704 loads and stores to the structure may be assumed by the callee
705 not to trap and to be properly aligned. This may only be applied to
706 the first parameter. This is not a valid attribute for return
709 This indicates that pointer values :ref:`based <pointeraliasing>` on
710 the argument or return value do not alias pointer values which are
711 not *based* on it, ignoring certain "irrelevant" dependencies. For a
712 call to the parent function, dependencies between memory references
713 from before or after the call and from those during the call are
714 "irrelevant" to the ``noalias`` keyword for the arguments and return
715 value used in that call. The caller shares the responsibility with
716 the callee for ensuring that these requirements are met. For further
717 details, please see the discussion of the NoAlias response in `alias
718 analysis <AliasAnalysis.html#MustMayNo>`_.
720 Note that this definition of ``noalias`` is intentionally similar
721 to the definition of ``restrict`` in C99 for function arguments,
722 though it is slightly weaker.
724 For function return values, C99's ``restrict`` is not meaningful,
725 while LLVM's ``noalias`` is.
727 This indicates that the callee does not make any copies of the
728 pointer that outlive the callee itself. This is not a valid
729 attribute for return values.
734 This indicates that the pointer parameter can be excised using the
735 :ref:`trampoline intrinsics <int_trampoline>`. This is not a valid
736 attribute for return values and can only be applied to one parameter.
739 This indicates that the function always returns the argument as its return
740 value. This is an optimization hint to the code generator when generating
741 the caller, allowing tail call optimization and omission of register saves
742 and restores in some cases; it is not checked or enforced when generating
743 the callee. The parameter and the function return type must be valid
744 operands for the :ref:`bitcast instruction <i_bitcast>`. This is not a
745 valid attribute for return values and can only be applied to one parameter.
749 Garbage Collector Names
750 -----------------------
752 Each function may specify a garbage collector name, which is simply a
757 define void @f() gc "name" { ... }
759 The compiler declares the supported values of *name*. Specifying a
760 collector which will cause the compiler to alter its output in order to
761 support the named garbage collection algorithm.
768 Prefix data is data associated with a function which the code generator
769 will emit immediately before the function body. The purpose of this feature
770 is to allow frontends to associate language-specific runtime metadata with
771 specific functions and make it available through the function pointer while
772 still allowing the function pointer to be called. To access the data for a
773 given function, a program may bitcast the function pointer to a pointer to
774 the constant's type. This implies that the IR symbol points to the start
777 To maintain the semantics of ordinary function calls, the prefix data must
778 have a particular format. Specifically, it must begin with a sequence of
779 bytes which decode to a sequence of machine instructions, valid for the
780 module's target, which transfer control to the point immediately succeeding
781 the prefix data, without performing any other visible action. This allows
782 the inliner and other passes to reason about the semantics of the function
783 definition without needing to reason about the prefix data. Obviously this
784 makes the format of the prefix data highly target dependent.
786 Prefix data is laid out as if it were an initializer for a global variable
787 of the prefix data's type. No padding is automatically placed between the
788 prefix data and the function body. If padding is required, it must be part
791 A trivial example of valid prefix data for the x86 architecture is ``i8 144``,
792 which encodes the ``nop`` instruction:
796 define void @f() prefix i8 144 { ... }
798 Generally prefix data can be formed by encoding a relative branch instruction
799 which skips the metadata, as in this example of valid prefix data for the
800 x86_64 architecture, where the first two bytes encode ``jmp .+10``:
804 %0 = type <{ i8, i8, i8* }>
806 define void @f() prefix %0 <{ i8 235, i8 8, i8* @md}> { ... }
808 A function may have prefix data but no body. This has similar semantics
809 to the ``available_externally`` linkage in that the data may be used by the
810 optimizers but will not be emitted in the object file.
817 Attribute groups are groups of attributes that are referenced by objects within
818 the IR. They are important for keeping ``.ll`` files readable, because a lot of
819 functions will use the same set of attributes. In the degenerative case of a
820 ``.ll`` file that corresponds to a single ``.c`` file, the single attribute
821 group will capture the important command line flags used to build that file.
823 An attribute group is a module-level object. To use an attribute group, an
824 object references the attribute group's ID (e.g. ``#37``). An object may refer
825 to more than one attribute group. In that situation, the attributes from the
826 different groups are merged.
828 Here is an example of attribute groups for a function that should always be
829 inlined, has a stack alignment of 4, and which shouldn't use SSE instructions:
833 ; Target-independent attributes:
834 attributes #0 = { alwaysinline alignstack=4 }
836 ; Target-dependent attributes:
837 attributes #1 = { "no-sse" }
839 ; Function @f has attributes: alwaysinline, alignstack=4, and "no-sse".
840 define void @f() #0 #1 { ... }
847 Function attributes are set to communicate additional information about
848 a function. Function attributes are considered to be part of the
849 function, not of the function type, so functions with different function
850 attributes can have the same function type.
852 Function attributes are simple keywords that follow the type specified.
853 If multiple attributes are needed, they are space separated. For
858 define void @f() noinline { ... }
859 define void @f() alwaysinline { ... }
860 define void @f() alwaysinline optsize { ... }
861 define void @f() optsize { ... }
864 This attribute indicates that, when emitting the prologue and
865 epilogue, the backend should forcibly align the stack pointer.
866 Specify the desired alignment, which must be a power of two, in
869 This attribute indicates that the inliner should attempt to inline
870 this function into callers whenever possible, ignoring any active
871 inlining size threshold for this caller.
873 This indicates that the callee function at a call site should be
874 recognized as a built-in function, even though the function's declaration
875 uses the ``nobuiltin`` attribute. This is only valid at call sites for
876 direct calls to functions which are declared with the ``nobuiltin``
879 This attribute indicates that this function is rarely called. When
880 computing edge weights, basic blocks post-dominated by a cold
881 function call are also considered to be cold; and, thus, given low
884 This attribute indicates that the source code contained a hint that
885 inlining this function is desirable (such as the "inline" keyword in
886 C/C++). It is just a hint; it imposes no requirements on the
889 This attribute suggests that optimization passes and code generator
890 passes make choices that keep the code size of this function as small
891 as possible and perform optimizations that may sacrifice runtime
892 performance in order to minimize the size of the generated code.
894 This attribute disables prologue / epilogue emission for the
895 function. This can have very system-specific consequences.
897 This indicates that the callee function at a call site is not recognized as
898 a built-in function. LLVM will retain the original call and not replace it
899 with equivalent code based on the semantics of the built-in function, unless
900 the call site uses the ``builtin`` attribute. This is valid at call sites
901 and on function declarations and definitions.
903 This attribute indicates that calls to the function cannot be
904 duplicated. A call to a ``noduplicate`` function may be moved
905 within its parent function, but may not be duplicated within
908 A function containing a ``noduplicate`` call may still
909 be an inlining candidate, provided that the call is not
910 duplicated by inlining. That implies that the function has
911 internal linkage and only has one call site, so the original
912 call is dead after inlining.
914 This attributes disables implicit floating point instructions.
916 This attribute indicates that the inliner should never inline this
917 function in any situation. This attribute may not be used together
918 with the ``alwaysinline`` attribute.
920 This attribute suppresses lazy symbol binding for the function. This
921 may make calls to the function faster, at the cost of extra program
922 startup time if the function is not called during program startup.
924 This attribute indicates that the code generator should not use a
925 red zone, even if the target-specific ABI normally permits it.
927 This function attribute indicates that the function never returns
928 normally. This produces undefined behavior at runtime if the
929 function ever does dynamically return.
931 This function attribute indicates that the function never returns
932 with an unwind or exceptional control flow. If the function does
933 unwind, its runtime behavior is undefined.
935 This function attribute indicates that the function is not optimized
936 by any optimization or code generator passes with the
937 exception of interprocedural optimization passes.
938 This attribute cannot be used together with the ``alwaysinline``
939 attribute; this attribute is also incompatible
940 with the ``minsize`` attribute and the ``optsize`` attribute.
942 The inliner should never inline this function in any situation.
943 Only functions with the ``alwaysinline`` attribute are valid
944 candidates for inlining inside the body of this function.
946 This attribute suggests that optimization passes and code generator
947 passes make choices that keep the code size of this function low,
948 and otherwise do optimizations specifically to reduce code size as
949 long as they do not significantly impact runtime performance.
951 On a function, this attribute indicates that the function computes its
952 result (or decides to unwind an exception) based strictly on its arguments,
953 without dereferencing any pointer arguments or otherwise accessing
954 any mutable state (e.g. memory, control registers, etc) visible to
955 caller functions. It does not write through any pointer arguments
956 (including ``byval`` arguments) and never changes any state visible
957 to callers. This means that it cannot unwind exceptions by calling
958 the ``C++`` exception throwing methods.
960 On an argument, this attribute indicates that the function does not
961 dereference that pointer argument, even though it may read or write the
962 memory that the pointer points to if accessed through other pointers.
964 On a function, this attribute indicates that the function does not write
965 through any pointer arguments (including ``byval`` arguments) or otherwise
966 modify any state (e.g. memory, control registers, etc) visible to
967 caller functions. It may dereference pointer arguments and read
968 state that may be set in the caller. A readonly function always
969 returns the same value (or unwinds an exception identically) when
970 called with the same set of arguments and global state. It cannot
971 unwind an exception by calling the ``C++`` exception throwing
974 On an argument, this attribute indicates that the function does not write
975 through this pointer argument, even though it may write to the memory that
976 the pointer points to.
978 This attribute indicates that this function can return twice. The C
979 ``setjmp`` is an example of such a function. The compiler disables
980 some optimizations (like tail calls) in the caller of these
983 This attribute indicates that AddressSanitizer checks
984 (dynamic address safety analysis) are enabled for this function.
986 This attribute indicates that MemorySanitizer checks (dynamic detection
987 of accesses to uninitialized memory) are enabled for this function.
989 This attribute indicates that ThreadSanitizer checks
990 (dynamic thread safety analysis) are enabled for this function.
992 This attribute indicates that the function should emit a stack
993 smashing protector. It is in the form of a "canary" --- a random value
994 placed on the stack before the local variables that's checked upon
995 return from the function to see if it has been overwritten. A
996 heuristic is used to determine if a function needs stack protectors
997 or not. The heuristic used will enable protectors for functions with:
999 - Character arrays larger than ``ssp-buffer-size`` (default 8).
1000 - Aggregates containing character arrays larger than ``ssp-buffer-size``.
1001 - Calls to alloca() with variable sizes or constant sizes greater than
1002 ``ssp-buffer-size``.
1004 If a function that has an ``ssp`` attribute is inlined into a
1005 function that doesn't have an ``ssp`` attribute, then the resulting
1006 function will have an ``ssp`` attribute.
1008 This attribute indicates that the function should *always* emit a
1009 stack smashing protector. This overrides the ``ssp`` function
1012 If a function that has an ``sspreq`` attribute is inlined into a
1013 function that doesn't have an ``sspreq`` attribute or which has an
1014 ``ssp`` or ``sspstrong`` attribute, then the resulting function will have
1015 an ``sspreq`` attribute.
1017 This attribute indicates that the function should emit a stack smashing
1018 protector. This attribute causes a strong heuristic to be used when
1019 determining if a function needs stack protectors. The strong heuristic
1020 will enable protectors for functions with:
1022 - Arrays of any size and type
1023 - Aggregates containing an array of any size and type.
1024 - Calls to alloca().
1025 - Local variables that have had their address taken.
1027 This overrides the ``ssp`` function attribute.
1029 If a function that has an ``sspstrong`` attribute is inlined into a
1030 function that doesn't have an ``sspstrong`` attribute, then the
1031 resulting function will have an ``sspstrong`` attribute.
1033 This attribute indicates that the ABI being targeted requires that
1034 an unwind table entry be produce for this function even if we can
1035 show that no exceptions passes by it. This is normally the case for
1036 the ELF x86-64 abi, but it can be disabled for some compilation
1041 Module-Level Inline Assembly
1042 ----------------------------
1044 Modules may contain "module-level inline asm" blocks, which corresponds
1045 to the GCC "file scope inline asm" blocks. These blocks are internally
1046 concatenated by LLVM and treated as a single unit, but may be separated
1047 in the ``.ll`` file if desired. The syntax is very simple:
1049 .. code-block:: llvm
1051 module asm "inline asm code goes here"
1052 module asm "more can go here"
1054 The strings can contain any character by escaping non-printable
1055 characters. The escape sequence used is simply "\\xx" where "xx" is the
1056 two digit hex code for the number.
1058 The inline asm code is simply printed to the machine code .s file when
1059 assembly code is generated.
1061 .. _langref_datalayout:
1066 A module may specify a target specific data layout string that specifies
1067 how data is to be laid out in memory. The syntax for the data layout is
1070 .. code-block:: llvm
1072 target datalayout = "layout specification"
1074 The *layout specification* consists of a list of specifications
1075 separated by the minus sign character ('-'). Each specification starts
1076 with a letter and may include other information after the letter to
1077 define some aspect of the data layout. The specifications accepted are
1081 Specifies that the target lays out data in big-endian form. That is,
1082 the bits with the most significance have the lowest address
1085 Specifies that the target lays out data in little-endian form. That
1086 is, the bits with the least significance have the lowest address
1089 Specifies the natural alignment of the stack in bits. Alignment
1090 promotion of stack variables is limited to the natural stack
1091 alignment to avoid dynamic stack realignment. The stack alignment
1092 must be a multiple of 8-bits. If omitted, the natural stack
1093 alignment defaults to "unspecified", which does not prevent any
1094 alignment promotions.
1095 ``p[n]:<size>:<abi>:<pref>``
1096 This specifies the *size* of a pointer and its ``<abi>`` and
1097 ``<pref>``\erred alignments for address space ``n``. All sizes are in
1098 bits. Specifying the ``<pref>`` alignment is optional. If omitted, the
1099 preceding ``:`` should be omitted too. The address space, ``n`` is
1100 optional, and if not specified, denotes the default address space 0.
1101 The value of ``n`` must be in the range [1,2^23).
1102 ``i<size>:<abi>:<pref>``
1103 This specifies the alignment for an integer type of a given bit
1104 ``<size>``. The value of ``<size>`` must be in the range [1,2^23).
1105 ``v<size>:<abi>:<pref>``
1106 This specifies the alignment for a vector type of a given bit
1108 ``f<size>:<abi>:<pref>``
1109 This specifies the alignment for a floating point type of a given bit
1110 ``<size>``. Only values of ``<size>`` that are supported by the target
1111 will work. 32 (float) and 64 (double) are supported on all targets; 80
1112 or 128 (different flavors of long double) are also supported on some
1114 ``a<size>:<abi>:<pref>``
1115 This specifies the alignment for an aggregate type of a given bit
1117 ``s<size>:<abi>:<pref>``
1118 This specifies the alignment for a stack object of a given bit
1120 ``n<size1>:<size2>:<size3>...``
1121 This specifies a set of native integer widths for the target CPU in
1122 bits. For example, it might contain ``n32`` for 32-bit PowerPC,
1123 ``n32:64`` for PowerPC 64, or ``n8:16:32:64`` for X86-64. Elements of
1124 this set are considered to support most general arithmetic operations
1127 When constructing the data layout for a given target, LLVM starts with a
1128 default set of specifications which are then (possibly) overridden by
1129 the specifications in the ``datalayout`` keyword. The default
1130 specifications are given in this list:
1132 - ``E`` - big endian
1133 - ``p:64:64:64`` - 64-bit pointers with 64-bit alignment.
1134 - ``p[n]:64:64:64`` - Other address spaces are assumed to be the
1135 same as the default address space.
1136 - ``S0`` - natural stack alignment is unspecified
1137 - ``i1:8:8`` - i1 is 8-bit (byte) aligned
1138 - ``i8:8:8`` - i8 is 8-bit (byte) aligned
1139 - ``i16:16:16`` - i16 is 16-bit aligned
1140 - ``i32:32:32`` - i32 is 32-bit aligned
1141 - ``i64:32:64`` - i64 has ABI alignment of 32-bits but preferred
1142 alignment of 64-bits
1143 - ``f16:16:16`` - half is 16-bit aligned
1144 - ``f32:32:32`` - float is 32-bit aligned
1145 - ``f64:64:64`` - double is 64-bit aligned
1146 - ``f128:128:128`` - quad is 128-bit aligned
1147 - ``v64:64:64`` - 64-bit vector is 64-bit aligned
1148 - ``v128:128:128`` - 128-bit vector is 128-bit aligned
1149 - ``a0:0:64`` - aggregates are 64-bit aligned
1151 When LLVM is determining the alignment for a given type, it uses the
1154 #. If the type sought is an exact match for one of the specifications,
1155 that specification is used.
1156 #. If no match is found, and the type sought is an integer type, then
1157 the smallest integer type that is larger than the bitwidth of the
1158 sought type is used. If none of the specifications are larger than
1159 the bitwidth then the largest integer type is used. For example,
1160 given the default specifications above, the i7 type will use the
1161 alignment of i8 (next largest) while both i65 and i256 will use the
1162 alignment of i64 (largest specified).
1163 #. If no match is found, and the type sought is a vector type, then the
1164 largest vector type that is smaller than the sought vector type will
1165 be used as a fall back. This happens because <128 x double> can be
1166 implemented in terms of 64 <2 x double>, for example.
1168 The function of the data layout string may not be what you expect.
1169 Notably, this is not a specification from the frontend of what alignment
1170 the code generator should use.
1172 Instead, if specified, the target data layout is required to match what
1173 the ultimate *code generator* expects. This string is used by the
1174 mid-level optimizers to improve code, and this only works if it matches
1175 what the ultimate code generator uses. If you would like to generate IR
1176 that does not embed this target-specific detail into the IR, then you
1177 don't have to specify the string. This will disable some optimizations
1178 that require precise layout information, but this also prevents those
1179 optimizations from introducing target specificity into the IR.
1186 A module may specify a target triple string that describes the target
1187 host. The syntax for the target triple is simply:
1189 .. code-block:: llvm
1191 target triple = "x86_64-apple-macosx10.7.0"
1193 The *target triple* string consists of a series of identifiers delimited
1194 by the minus sign character ('-'). The canonical forms are:
1198 ARCHITECTURE-VENDOR-OPERATING_SYSTEM
1199 ARCHITECTURE-VENDOR-OPERATING_SYSTEM-ENVIRONMENT
1201 This information is passed along to the backend so that it generates
1202 code for the proper architecture. It's possible to override this on the
1203 command line with the ``-mtriple`` command line option.
1205 .. _pointeraliasing:
1207 Pointer Aliasing Rules
1208 ----------------------
1210 Any memory access must be done through a pointer value associated with
1211 an address range of the memory access, otherwise the behavior is
1212 undefined. Pointer values are associated with address ranges according
1213 to the following rules:
1215 - A pointer value is associated with the addresses associated with any
1216 value it is *based* on.
1217 - An address of a global variable is associated with the address range
1218 of the variable's storage.
1219 - The result value of an allocation instruction is associated with the
1220 address range of the allocated storage.
1221 - A null pointer in the default address-space is associated with no
1223 - An integer constant other than zero or a pointer value returned from
1224 a function not defined within LLVM may be associated with address
1225 ranges allocated through mechanisms other than those provided by
1226 LLVM. Such ranges shall not overlap with any ranges of addresses
1227 allocated by mechanisms provided by LLVM.
1229 A pointer value is *based* on another pointer value according to the
1232 - A pointer value formed from a ``getelementptr`` operation is *based*
1233 on the first operand of the ``getelementptr``.
1234 - The result value of a ``bitcast`` is *based* on the operand of the
1236 - A pointer value formed by an ``inttoptr`` is *based* on all pointer
1237 values that contribute (directly or indirectly) to the computation of
1238 the pointer's value.
1239 - The "*based* on" relationship is transitive.
1241 Note that this definition of *"based"* is intentionally similar to the
1242 definition of *"based"* in C99, though it is slightly weaker.
1244 LLVM IR does not associate types with memory. The result type of a
1245 ``load`` merely indicates the size and alignment of the memory from
1246 which to load, as well as the interpretation of the value. The first
1247 operand type of a ``store`` similarly only indicates the size and
1248 alignment of the store.
1250 Consequently, type-based alias analysis, aka TBAA, aka
1251 ``-fstrict-aliasing``, is not applicable to general unadorned LLVM IR.
1252 :ref:`Metadata <metadata>` may be used to encode additional information
1253 which specialized optimization passes may use to implement type-based
1258 Volatile Memory Accesses
1259 ------------------------
1261 Certain memory accesses, such as :ref:`load <i_load>`'s,
1262 :ref:`store <i_store>`'s, and :ref:`llvm.memcpy <int_memcpy>`'s may be
1263 marked ``volatile``. The optimizers must not change the number of
1264 volatile operations or change their order of execution relative to other
1265 volatile operations. The optimizers *may* change the order of volatile
1266 operations relative to non-volatile operations. This is not Java's
1267 "volatile" and has no cross-thread synchronization behavior.
1269 IR-level volatile loads and stores cannot safely be optimized into
1270 llvm.memcpy or llvm.memmove intrinsics even when those intrinsics are
1271 flagged volatile. Likewise, the backend should never split or merge
1272 target-legal volatile load/store instructions.
1274 .. admonition:: Rationale
1276 Platforms may rely on volatile loads and stores of natively supported
1277 data width to be executed as single instruction. For example, in C
1278 this holds for an l-value of volatile primitive type with native
1279 hardware support, but not necessarily for aggregate types. The
1280 frontend upholds these expectations, which are intentionally
1281 unspecified in the IR. The rules above ensure that IR transformation
1282 do not violate the frontend's contract with the language.
1286 Memory Model for Concurrent Operations
1287 --------------------------------------
1289 The LLVM IR does not define any way to start parallel threads of
1290 execution or to register signal handlers. Nonetheless, there are
1291 platform-specific ways to create them, and we define LLVM IR's behavior
1292 in their presence. This model is inspired by the C++0x memory model.
1294 For a more informal introduction to this model, see the :doc:`Atomics`.
1296 We define a *happens-before* partial order as the least partial order
1299 - Is a superset of single-thread program order, and
1300 - When a *synchronizes-with* ``b``, includes an edge from ``a`` to
1301 ``b``. *Synchronizes-with* pairs are introduced by platform-specific
1302 techniques, like pthread locks, thread creation, thread joining,
1303 etc., and by atomic instructions. (See also :ref:`Atomic Memory Ordering
1304 Constraints <ordering>`).
1306 Note that program order does not introduce *happens-before* edges
1307 between a thread and signals executing inside that thread.
1309 Every (defined) read operation (load instructions, memcpy, atomic
1310 loads/read-modify-writes, etc.) R reads a series of bytes written by
1311 (defined) write operations (store instructions, atomic
1312 stores/read-modify-writes, memcpy, etc.). For the purposes of this
1313 section, initialized globals are considered to have a write of the
1314 initializer which is atomic and happens before any other read or write
1315 of the memory in question. For each byte of a read R, R\ :sub:`byte`
1316 may see any write to the same byte, except:
1318 - If write\ :sub:`1` happens before write\ :sub:`2`, and
1319 write\ :sub:`2` happens before R\ :sub:`byte`, then
1320 R\ :sub:`byte` does not see write\ :sub:`1`.
1321 - If R\ :sub:`byte` happens before write\ :sub:`3`, then
1322 R\ :sub:`byte` does not see write\ :sub:`3`.
1324 Given that definition, R\ :sub:`byte` is defined as follows:
1326 - If R is volatile, the result is target-dependent. (Volatile is
1327 supposed to give guarantees which can support ``sig_atomic_t`` in
1328 C/C++, and may be used for accesses to addresses which do not behave
1329 like normal memory. It does not generally provide cross-thread
1331 - Otherwise, if there is no write to the same byte that happens before
1332 R\ :sub:`byte`, R\ :sub:`byte` returns ``undef`` for that byte.
1333 - Otherwise, if R\ :sub:`byte` may see exactly one write,
1334 R\ :sub:`byte` returns the value written by that write.
1335 - Otherwise, if R is atomic, and all the writes R\ :sub:`byte` may
1336 see are atomic, it chooses one of the values written. See the :ref:`Atomic
1337 Memory Ordering Constraints <ordering>` section for additional
1338 constraints on how the choice is made.
1339 - Otherwise R\ :sub:`byte` returns ``undef``.
1341 R returns the value composed of the series of bytes it read. This
1342 implies that some bytes within the value may be ``undef`` **without**
1343 the entire value being ``undef``. Note that this only defines the
1344 semantics of the operation; it doesn't mean that targets will emit more
1345 than one instruction to read the series of bytes.
1347 Note that in cases where none of the atomic intrinsics are used, this
1348 model places only one restriction on IR transformations on top of what
1349 is required for single-threaded execution: introducing a store to a byte
1350 which might not otherwise be stored is not allowed in general.
1351 (Specifically, in the case where another thread might write to and read
1352 from an address, introducing a store can change a load that may see
1353 exactly one write into a load that may see multiple writes.)
1357 Atomic Memory Ordering Constraints
1358 ----------------------------------
1360 Atomic instructions (:ref:`cmpxchg <i_cmpxchg>`,
1361 :ref:`atomicrmw <i_atomicrmw>`, :ref:`fence <i_fence>`,
1362 :ref:`atomic load <i_load>`, and :ref:`atomic store <i_store>`) take
1363 an ordering parameter that determines which other atomic instructions on
1364 the same address they *synchronize with*. These semantics are borrowed
1365 from Java and C++0x, but are somewhat more colloquial. If these
1366 descriptions aren't precise enough, check those specs (see spec
1367 references in the :doc:`atomics guide <Atomics>`).
1368 :ref:`fence <i_fence>` instructions treat these orderings somewhat
1369 differently since they don't take an address. See that instruction's
1370 documentation for details.
1372 For a simpler introduction to the ordering constraints, see the
1376 The set of values that can be read is governed by the happens-before
1377 partial order. A value cannot be read unless some operation wrote
1378 it. This is intended to provide a guarantee strong enough to model
1379 Java's non-volatile shared variables. This ordering cannot be
1380 specified for read-modify-write operations; it is not strong enough
1381 to make them atomic in any interesting way.
1383 In addition to the guarantees of ``unordered``, there is a single
1384 total order for modifications by ``monotonic`` operations on each
1385 address. All modification orders must be compatible with the
1386 happens-before order. There is no guarantee that the modification
1387 orders can be combined to a global total order for the whole program
1388 (and this often will not be possible). The read in an atomic
1389 read-modify-write operation (:ref:`cmpxchg <i_cmpxchg>` and
1390 :ref:`atomicrmw <i_atomicrmw>`) reads the value in the modification
1391 order immediately before the value it writes. If one atomic read
1392 happens before another atomic read of the same address, the later
1393 read must see the same value or a later value in the address's
1394 modification order. This disallows reordering of ``monotonic`` (or
1395 stronger) operations on the same address. If an address is written
1396 ``monotonic``-ally by one thread, and other threads ``monotonic``-ally
1397 read that address repeatedly, the other threads must eventually see
1398 the write. This corresponds to the C++0x/C1x
1399 ``memory_order_relaxed``.
1401 In addition to the guarantees of ``monotonic``, a
1402 *synchronizes-with* edge may be formed with a ``release`` operation.
1403 This is intended to model C++'s ``memory_order_acquire``.
1405 In addition to the guarantees of ``monotonic``, if this operation
1406 writes a value which is subsequently read by an ``acquire``
1407 operation, it *synchronizes-with* that operation. (This isn't a
1408 complete description; see the C++0x definition of a release
1409 sequence.) This corresponds to the C++0x/C1x
1410 ``memory_order_release``.
1411 ``acq_rel`` (acquire+release)
1412 Acts as both an ``acquire`` and ``release`` operation on its
1413 address. This corresponds to the C++0x/C1x ``memory_order_acq_rel``.
1414 ``seq_cst`` (sequentially consistent)
1415 In addition to the guarantees of ``acq_rel`` (``acquire`` for an
1416 operation which only reads, ``release`` for an operation which only
1417 writes), there is a global total order on all
1418 sequentially-consistent operations on all addresses, which is
1419 consistent with the *happens-before* partial order and with the
1420 modification orders of all the affected addresses. Each
1421 sequentially-consistent read sees the last preceding write to the
1422 same address in this global order. This corresponds to the C++0x/C1x
1423 ``memory_order_seq_cst`` and Java volatile.
1427 If an atomic operation is marked ``singlethread``, it only *synchronizes
1428 with* or participates in modification and seq\_cst total orderings with
1429 other operations running in the same thread (for example, in signal
1437 LLVM IR floating-point binary ops (:ref:`fadd <i_fadd>`,
1438 :ref:`fsub <i_fsub>`, :ref:`fmul <i_fmul>`, :ref:`fdiv <i_fdiv>`,
1439 :ref:`frem <i_frem>`) have the following flags that can set to enable
1440 otherwise unsafe floating point operations
1443 No NaNs - Allow optimizations to assume the arguments and result are not
1444 NaN. Such optimizations are required to retain defined behavior over
1445 NaNs, but the value of the result is undefined.
1448 No Infs - Allow optimizations to assume the arguments and result are not
1449 +/-Inf. Such optimizations are required to retain defined behavior over
1450 +/-Inf, but the value of the result is undefined.
1453 No Signed Zeros - Allow optimizations to treat the sign of a zero
1454 argument or result as insignificant.
1457 Allow Reciprocal - Allow optimizations to use the reciprocal of an
1458 argument rather than perform division.
1461 Fast - Allow algebraically equivalent transformations that may
1462 dramatically change results in floating point (e.g. reassociate). This
1463 flag implies all the others.
1470 The LLVM type system is one of the most important features of the
1471 intermediate representation. Being typed enables a number of
1472 optimizations to be performed on the intermediate representation
1473 directly, without having to do extra analyses on the side before the
1474 transformation. A strong type system makes it easier to read the
1475 generated code and enables novel analyses and transformations that are
1476 not feasible to perform on normal three address code representations.
1478 .. _typeclassifications:
1480 Type Classifications
1481 --------------------
1483 The types fall into a few useful classifications:
1492 * - :ref:`integer <t_integer>`
1493 - ``i1``, ``i2``, ``i3``, ... ``i8``, ... ``i16``, ... ``i32``, ...
1496 * - :ref:`floating point <t_floating>`
1497 - ``half``, ``float``, ``double``, ``x86_fp80``, ``fp128``,
1505 - :ref:`integer <t_integer>`, :ref:`floating point <t_floating>`,
1506 :ref:`pointer <t_pointer>`, :ref:`vector <t_vector>`,
1507 :ref:`structure <t_struct>`, :ref:`array <t_array>`,
1508 :ref:`label <t_label>`, :ref:`metadata <t_metadata>`.
1510 * - :ref:`primitive <t_primitive>`
1511 - :ref:`label <t_label>`,
1512 :ref:`void <t_void>`,
1513 :ref:`integer <t_integer>`,
1514 :ref:`floating point <t_floating>`,
1515 :ref:`x86mmx <t_x86mmx>`,
1516 :ref:`metadata <t_metadata>`.
1518 * - :ref:`derived <t_derived>`
1519 - :ref:`array <t_array>`,
1520 :ref:`function <t_function>`,
1521 :ref:`pointer <t_pointer>`,
1522 :ref:`structure <t_struct>`,
1523 :ref:`vector <t_vector>`,
1524 :ref:`opaque <t_opaque>`.
1526 The :ref:`first class <t_firstclass>` types are perhaps the most important.
1527 Values of these types are the only ones which can be produced by
1535 The primitive types are the fundamental building blocks of the LLVM
1546 The integer type is a very simple type that simply specifies an
1547 arbitrary bit width for the integer type desired. Any bit width from 1
1548 bit to 2\ :sup:`23`\ -1 (about 8 million) can be specified.
1557 The number of bits the integer will occupy is specified by the ``N``
1563 +----------------+------------------------------------------------+
1564 | ``i1`` | a single-bit integer. |
1565 +----------------+------------------------------------------------+
1566 | ``i32`` | a 32-bit integer. |
1567 +----------------+------------------------------------------------+
1568 | ``i1942652`` | a really big integer of over 1 million bits. |
1569 +----------------+------------------------------------------------+
1573 Floating Point Types
1574 ^^^^^^^^^^^^^^^^^^^^
1583 - 16-bit floating point value
1586 - 32-bit floating point value
1589 - 64-bit floating point value
1592 - 128-bit floating point value (112-bit mantissa)
1595 - 80-bit floating point value (X87)
1598 - 128-bit floating point value (two 64-bits)
1608 The x86mmx type represents a value held in an MMX register on an x86
1609 machine. The operations allowed on it are quite limited: parameters and
1610 return values, load and store, and bitcast. User-specified MMX
1611 instructions are represented as intrinsic or asm calls with arguments
1612 and/or results of this type. There are no arrays, vectors or constants
1630 The void type does not represent any value and has no size.
1647 The label type represents code labels.
1664 The metadata type represents embedded metadata. No derived types may be
1665 created from metadata except for :ref:`function <t_function>` arguments.
1679 The real power in LLVM comes from the derived types in the system. This
1680 is what allows a programmer to represent arrays, functions, pointers,
1681 and other useful types. Each of these types contain one or more element
1682 types which may be a primitive type, or another derived type. For
1683 example, it is possible to have a two dimensional array, using an array
1684 as the element type of another array.
1691 Aggregate Types are a subset of derived types that can contain multiple
1692 member types. :ref:`Arrays <t_array>` and :ref:`structs <t_struct>` are
1693 aggregate types. :ref:`Vectors <t_vector>` are not considered to be
1704 The array type is a very simple derived type that arranges elements
1705 sequentially in memory. The array type requires a size (number of
1706 elements) and an underlying data type.
1713 [<# elements> x <elementtype>]
1715 The number of elements is a constant integer value; ``elementtype`` may
1716 be any type with a size.
1721 +------------------+--------------------------------------+
1722 | ``[40 x i32]`` | Array of 40 32-bit integer values. |
1723 +------------------+--------------------------------------+
1724 | ``[41 x i32]`` | Array of 41 32-bit integer values. |
1725 +------------------+--------------------------------------+
1726 | ``[4 x i8]`` | Array of 4 8-bit integer values. |
1727 +------------------+--------------------------------------+
1729 Here are some examples of multidimensional arrays:
1731 +-----------------------------+----------------------------------------------------------+
1732 | ``[3 x [4 x i32]]`` | 3x4 array of 32-bit integer values. |
1733 +-----------------------------+----------------------------------------------------------+
1734 | ``[12 x [10 x float]]`` | 12x10 array of single precision floating point values. |
1735 +-----------------------------+----------------------------------------------------------+
1736 | ``[2 x [3 x [4 x i16]]]`` | 2x3x4 array of 16-bit integer values. |
1737 +-----------------------------+----------------------------------------------------------+
1739 There is no restriction on indexing beyond the end of the array implied
1740 by a static type (though there are restrictions on indexing beyond the
1741 bounds of an allocated object in some cases). This means that
1742 single-dimension 'variable sized array' addressing can be implemented in
1743 LLVM with a zero length array type. An implementation of 'pascal style
1744 arrays' in LLVM could use the type "``{ i32, [0 x float]}``", for
1755 The function type can be thought of as a function signature. It consists of a
1756 return type and a list of formal parameter types. The return type of a function
1757 type is a void type or first class type --- except for :ref:`label <t_label>`
1758 and :ref:`metadata <t_metadata>` types.
1765 <returntype> (<parameter list>)
1767 ...where '``<parameter list>``' is a comma-separated list of type
1768 specifiers. Optionally, the parameter list may include a type ``...``, which
1769 indicates that the function takes a variable number of arguments. Variable
1770 argument functions can access their arguments with the :ref:`variable argument
1771 handling intrinsic <int_varargs>` functions. '``<returntype>``' is any type
1772 except :ref:`label <t_label>` and :ref:`metadata <t_metadata>`.
1777 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1778 | ``i32 (i32)`` | function taking an ``i32``, returning an ``i32`` |
1779 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1780 | ``float (i16, i32 *) *`` | :ref:`Pointer <t_pointer>` to a function that takes an ``i16`` and a :ref:`pointer <t_pointer>` to ``i32``, returning ``float``. |
1781 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1782 | ``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. |
1783 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1784 | ``{i32, i32} (i32)`` | A function taking an ``i32``, returning a :ref:`structure <t_struct>` containing two ``i32`` values |
1785 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1795 The structure type is used to represent a collection of data members
1796 together in memory. The elements of a structure may be any type that has
1799 Structures in memory are accessed using '``load``' and '``store``' by
1800 getting a pointer to a field with the '``getelementptr``' instruction.
1801 Structures in registers are accessed using the '``extractvalue``' and
1802 '``insertvalue``' instructions.
1804 Structures may optionally be "packed" structures, which indicate that
1805 the alignment of the struct is one byte, and that there is no padding
1806 between the elements. In non-packed structs, padding between field types
1807 is inserted as defined by the DataLayout string in the module, which is
1808 required to match what the underlying code generator expects.
1810 Structures can either be "literal" or "identified". A literal structure
1811 is defined inline with other types (e.g. ``{i32, i32}*``) whereas
1812 identified types are always defined at the top level with a name.
1813 Literal types are uniqued by their contents and can never be recursive
1814 or opaque since there is no way to write one. Identified types can be
1815 recursive, can be opaqued, and are never uniqued.
1822 %T1 = type { <type list> } ; Identified normal struct type
1823 %T2 = type <{ <type list> }> ; Identified packed struct type
1828 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1829 | ``{ i32, i32, i32 }`` | A triple of three ``i32`` values |
1830 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1831 | ``{ 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``. |
1832 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1833 | ``<{ i8, i32 }>`` | A packed struct known to be 5 bytes in size. |
1834 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1838 Opaque Structure Types
1839 ^^^^^^^^^^^^^^^^^^^^^^
1844 Opaque structure types are used to represent named structure types that
1845 do not have a body specified. This corresponds (for example) to the C
1846 notion of a forward declared structure.
1859 +--------------+-------------------+
1860 | ``opaque`` | An opaque type. |
1861 +--------------+-------------------+
1871 The pointer type is used to specify memory locations. Pointers are
1872 commonly used to reference objects in memory.
1874 Pointer types may have an optional address space attribute defining the
1875 numbered address space where the pointed-to object resides. The default
1876 address space is number zero. The semantics of non-zero address spaces
1877 are target-specific.
1879 Note that LLVM does not permit pointers to void (``void*``) nor does it
1880 permit pointers to labels (``label*``). Use ``i8*`` instead.
1892 +-------------------------+--------------------------------------------------------------------------------------------------------------+
1893 | ``[4 x i32]*`` | A :ref:`pointer <t_pointer>` to :ref:`array <t_array>` of four ``i32`` values. |
1894 +-------------------------+--------------------------------------------------------------------------------------------------------------+
1895 | ``i32 (i32*) *`` | A :ref:`pointer <t_pointer>` to a :ref:`function <t_function>` that takes an ``i32*``, returning an ``i32``. |
1896 +-------------------------+--------------------------------------------------------------------------------------------------------------+
1897 | ``i32 addrspace(5)*`` | A :ref:`pointer <t_pointer>` to an ``i32`` value that resides in address space #5. |
1898 +-------------------------+--------------------------------------------------------------------------------------------------------------+
1908 A vector type is a simple derived type that represents a vector of
1909 elements. Vector types are used when multiple primitive data are
1910 operated in parallel using a single instruction (SIMD). A vector type
1911 requires a size (number of elements) and an underlying primitive data
1912 type. Vector types are considered :ref:`first class <t_firstclass>`.
1919 < <# elements> x <elementtype> >
1921 The number of elements is a constant integer value larger than 0;
1922 elementtype may be any integer or floating point type, or a pointer to
1923 these types. Vectors of size zero are not allowed.
1928 +-------------------+--------------------------------------------------+
1929 | ``<4 x i32>`` | Vector of 4 32-bit integer values. |
1930 +-------------------+--------------------------------------------------+
1931 | ``<8 x float>`` | Vector of 8 32-bit floating-point values. |
1932 +-------------------+--------------------------------------------------+
1933 | ``<2 x i64>`` | Vector of 2 64-bit integer values. |
1934 +-------------------+--------------------------------------------------+
1935 | ``<4 x i64*>`` | Vector of 4 pointers to 64-bit integer values. |
1936 +-------------------+--------------------------------------------------+
1941 LLVM has several different basic types of constants. This section
1942 describes them all and their syntax.
1947 **Boolean constants**
1948 The two strings '``true``' and '``false``' are both valid constants
1950 **Integer constants**
1951 Standard integers (such as '4') are constants of the
1952 :ref:`integer <t_integer>` type. Negative numbers may be used with
1954 **Floating point constants**
1955 Floating point constants use standard decimal notation (e.g.
1956 123.421), exponential notation (e.g. 1.23421e+2), or a more precise
1957 hexadecimal notation (see below). The assembler requires the exact
1958 decimal value of a floating-point constant. For example, the
1959 assembler accepts 1.25 but rejects 1.3 because 1.3 is a repeating
1960 decimal in binary. Floating point constants must have a :ref:`floating
1961 point <t_floating>` type.
1962 **Null pointer constants**
1963 The identifier '``null``' is recognized as a null pointer constant
1964 and must be of :ref:`pointer type <t_pointer>`.
1966 The one non-intuitive notation for constants is the hexadecimal form of
1967 floating point constants. For example, the form
1968 '``double 0x432ff973cafa8000``' is equivalent to (but harder to read
1969 than) '``double 4.5e+15``'. The only time hexadecimal floating point
1970 constants are required (and the only time that they are generated by the
1971 disassembler) is when a floating point constant must be emitted but it
1972 cannot be represented as a decimal floating point number in a reasonable
1973 number of digits. For example, NaN's, infinities, and other special
1974 values are represented in their IEEE hexadecimal format so that assembly
1975 and disassembly do not cause any bits to change in the constants.
1977 When using the hexadecimal form, constants of types half, float, and
1978 double are represented using the 16-digit form shown above (which
1979 matches the IEEE754 representation for double); half and float values
1980 must, however, be exactly representable as IEEE 754 half and single
1981 precision, respectively. Hexadecimal format is always used for long
1982 double, and there are three forms of long double. The 80-bit format used
1983 by x86 is represented as ``0xK`` followed by 20 hexadecimal digits. The
1984 128-bit format used by PowerPC (two adjacent doubles) is represented by
1985 ``0xM`` followed by 32 hexadecimal digits. The IEEE 128-bit format is
1986 represented by ``0xL`` followed by 32 hexadecimal digits. Long doubles
1987 will only work if they match the long double format on your target.
1988 The IEEE 16-bit format (half precision) is represented by ``0xH``
1989 followed by 4 hexadecimal digits. All hexadecimal formats are big-endian
1990 (sign bit at the left).
1992 There are no constants of type x86mmx.
1994 .. _complexconstants:
1999 Complex constants are a (potentially recursive) combination of simple
2000 constants and smaller complex constants.
2002 **Structure constants**
2003 Structure constants are represented with notation similar to
2004 structure type definitions (a comma separated list of elements,
2005 surrounded by braces (``{}``)). For example:
2006 "``{ i32 4, float 17.0, i32* @G }``", where "``@G``" is declared as
2007 "``@G = external global i32``". Structure constants must have
2008 :ref:`structure type <t_struct>`, and the number and types of elements
2009 must match those specified by the type.
2011 Array constants are represented with notation similar to array type
2012 definitions (a comma separated list of elements, surrounded by
2013 square brackets (``[]``)). For example:
2014 "``[ i32 42, i32 11, i32 74 ]``". Array constants must have
2015 :ref:`array type <t_array>`, and the number and types of elements must
2016 match those specified by the type.
2017 **Vector constants**
2018 Vector constants are represented with notation similar to vector
2019 type definitions (a comma separated list of elements, surrounded by
2020 less-than/greater-than's (``<>``)). For example:
2021 "``< i32 42, i32 11, i32 74, i32 100 >``". Vector constants
2022 must have :ref:`vector type <t_vector>`, and the number and types of
2023 elements must match those specified by the type.
2024 **Zero initialization**
2025 The string '``zeroinitializer``' can be used to zero initialize a
2026 value to zero of *any* type, including scalar and
2027 :ref:`aggregate <t_aggregate>` types. This is often used to avoid
2028 having to print large zero initializers (e.g. for large arrays) and
2029 is always exactly equivalent to using explicit zero initializers.
2031 A metadata node is a structure-like constant with :ref:`metadata
2032 type <t_metadata>`. For example:
2033 "``metadata !{ i32 0, metadata !"test" }``". Unlike other
2034 constants that are meant to be interpreted as part of the
2035 instruction stream, metadata is a place to attach additional
2036 information such as debug info.
2038 Global Variable and Function Addresses
2039 --------------------------------------
2041 The addresses of :ref:`global variables <globalvars>` and
2042 :ref:`functions <functionstructure>` are always implicitly valid
2043 (link-time) constants. These constants are explicitly referenced when
2044 the :ref:`identifier for the global <identifiers>` is used and always have
2045 :ref:`pointer <t_pointer>` type. For example, the following is a legal LLVM
2048 .. code-block:: llvm
2052 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2059 The string '``undef``' can be used anywhere a constant is expected, and
2060 indicates that the user of the value may receive an unspecified
2061 bit-pattern. Undefined values may be of any type (other than '``label``'
2062 or '``void``') and be used anywhere a constant is permitted.
2064 Undefined values are useful because they indicate to the compiler that
2065 the program is well defined no matter what value is used. This gives the
2066 compiler more freedom to optimize. Here are some examples of
2067 (potentially surprising) transformations that are valid (in pseudo IR):
2069 .. code-block:: llvm
2079 This is safe because all of the output bits are affected by the undef
2080 bits. Any output bit can have a zero or one depending on the input bits.
2082 .. code-block:: llvm
2093 These logical operations have bits that are not always affected by the
2094 input. For example, if ``%X`` has a zero bit, then the output of the
2095 '``and``' operation will always be a zero for that bit, no matter what
2096 the corresponding bit from the '``undef``' is. As such, it is unsafe to
2097 optimize or assume that the result of the '``and``' is '``undef``'.
2098 However, it is safe to assume that all bits of the '``undef``' could be
2099 0, and optimize the '``and``' to 0. Likewise, it is safe to assume that
2100 all the bits of the '``undef``' operand to the '``or``' could be set,
2101 allowing the '``or``' to be folded to -1.
2103 .. code-block:: llvm
2105 %A = select undef, %X, %Y
2106 %B = select undef, 42, %Y
2107 %C = select %X, %Y, undef
2117 This set of examples shows that undefined '``select``' (and conditional
2118 branch) conditions can go *either way*, but they have to come from one
2119 of the two operands. In the ``%A`` example, if ``%X`` and ``%Y`` were
2120 both known to have a clear low bit, then ``%A`` would have to have a
2121 cleared low bit. However, in the ``%C`` example, the optimizer is
2122 allowed to assume that the '``undef``' operand could be the same as
2123 ``%Y``, allowing the whole '``select``' to be eliminated.
2125 .. code-block:: llvm
2127 %A = xor undef, undef
2144 This example points out that two '``undef``' operands are not
2145 necessarily the same. This can be surprising to people (and also matches
2146 C semantics) where they assume that "``X^X``" is always zero, even if
2147 ``X`` is undefined. This isn't true for a number of reasons, but the
2148 short answer is that an '``undef``' "variable" can arbitrarily change
2149 its value over its "live range". This is true because the variable
2150 doesn't actually *have a live range*. Instead, the value is logically
2151 read from arbitrary registers that happen to be around when needed, so
2152 the value is not necessarily consistent over time. In fact, ``%A`` and
2153 ``%C`` need to have the same semantics or the core LLVM "replace all
2154 uses with" concept would not hold.
2156 .. code-block:: llvm
2164 These examples show the crucial difference between an *undefined value*
2165 and *undefined behavior*. An undefined value (like '``undef``') is
2166 allowed to have an arbitrary bit-pattern. This means that the ``%A``
2167 operation can be constant folded to '``undef``', because the '``undef``'
2168 could be an SNaN, and ``fdiv`` is not (currently) defined on SNaN's.
2169 However, in the second example, we can make a more aggressive
2170 assumption: because the ``undef`` is allowed to be an arbitrary value,
2171 we are allowed to assume that it could be zero. Since a divide by zero
2172 has *undefined behavior*, we are allowed to assume that the operation
2173 does not execute at all. This allows us to delete the divide and all
2174 code after it. Because the undefined operation "can't happen", the
2175 optimizer can assume that it occurs in dead code.
2177 .. code-block:: llvm
2179 a: store undef -> %X
2180 b: store %X -> undef
2185 These examples reiterate the ``fdiv`` example: a store *of* an undefined
2186 value can be assumed to not have any effect; we can assume that the
2187 value is overwritten with bits that happen to match what was already
2188 there. However, a store *to* an undefined location could clobber
2189 arbitrary memory, therefore, it has undefined behavior.
2196 Poison values are similar to :ref:`undef values <undefvalues>`, however
2197 they also represent the fact that an instruction or constant expression
2198 which cannot evoke side effects has nevertheless detected a condition
2199 which results in undefined behavior.
2201 There is currently no way of representing a poison value in the IR; they
2202 only exist when produced by operations such as :ref:`add <i_add>` with
2205 Poison value behavior is defined in terms of value *dependence*:
2207 - Values other than :ref:`phi <i_phi>` nodes depend on their operands.
2208 - :ref:`Phi <i_phi>` nodes depend on the operand corresponding to
2209 their dynamic predecessor basic block.
2210 - Function arguments depend on the corresponding actual argument values
2211 in the dynamic callers of their functions.
2212 - :ref:`Call <i_call>` instructions depend on the :ref:`ret <i_ret>`
2213 instructions that dynamically transfer control back to them.
2214 - :ref:`Invoke <i_invoke>` instructions depend on the
2215 :ref:`ret <i_ret>`, :ref:`resume <i_resume>`, or exception-throwing
2216 call instructions that dynamically transfer control back to them.
2217 - Non-volatile loads and stores depend on the most recent stores to all
2218 of the referenced memory addresses, following the order in the IR
2219 (including loads and stores implied by intrinsics such as
2220 :ref:`@llvm.memcpy <int_memcpy>`.)
2221 - An instruction with externally visible side effects depends on the
2222 most recent preceding instruction with externally visible side
2223 effects, following the order in the IR. (This includes :ref:`volatile
2224 operations <volatile>`.)
2225 - An instruction *control-depends* on a :ref:`terminator
2226 instruction <terminators>` if the terminator instruction has
2227 multiple successors and the instruction is always executed when
2228 control transfers to one of the successors, and may not be executed
2229 when control is transferred to another.
2230 - Additionally, an instruction also *control-depends* on a terminator
2231 instruction if the set of instructions it otherwise depends on would
2232 be different if the terminator had transferred control to a different
2234 - Dependence is transitive.
2236 Poison Values have the same behavior as :ref:`undef values <undefvalues>`,
2237 with the additional affect that any instruction which has a *dependence*
2238 on a poison value has undefined behavior.
2240 Here are some examples:
2242 .. code-block:: llvm
2245 %poison = sub nuw i32 0, 1 ; Results in a poison value.
2246 %still_poison = and i32 %poison, 0 ; 0, but also poison.
2247 %poison_yet_again = getelementptr i32* @h, i32 %still_poison
2248 store i32 0, i32* %poison_yet_again ; memory at @h[0] is poisoned
2250 store i32 %poison, i32* @g ; Poison value stored to memory.
2251 %poison2 = load i32* @g ; Poison value loaded back from memory.
2253 store volatile i32 %poison, i32* @g ; External observation; undefined behavior.
2255 %narrowaddr = bitcast i32* @g to i16*
2256 %wideaddr = bitcast i32* @g to i64*
2257 %poison3 = load i16* %narrowaddr ; Returns a poison value.
2258 %poison4 = load i64* %wideaddr ; Returns a poison value.
2260 %cmp = icmp slt i32 %poison, 0 ; Returns a poison value.
2261 br i1 %cmp, label %true, label %end ; Branch to either destination.
2264 store volatile i32 0, i32* @g ; This is control-dependent on %cmp, so
2265 ; it has undefined behavior.
2269 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2270 ; Both edges into this PHI are
2271 ; control-dependent on %cmp, so this
2272 ; always results in a poison value.
2274 store volatile i32 0, i32* @g ; This would depend on the store in %true
2275 ; if %cmp is true, or the store in %entry
2276 ; otherwise, so this is undefined behavior.
2278 br i1 %cmp, label %second_true, label %second_end
2279 ; The same branch again, but this time the
2280 ; true block doesn't have side effects.
2287 store volatile i32 0, i32* @g ; This time, the instruction always depends
2288 ; on the store in %end. Also, it is
2289 ; control-equivalent to %end, so this is
2290 ; well-defined (ignoring earlier undefined
2291 ; behavior in this example).
2295 Addresses of Basic Blocks
2296 -------------------------
2298 ``blockaddress(@function, %block)``
2300 The '``blockaddress``' constant computes the address of the specified
2301 basic block in the specified function, and always has an ``i8*`` type.
2302 Taking the address of the entry block is illegal.
2304 This value only has defined behavior when used as an operand to the
2305 ':ref:`indirectbr <i_indirectbr>`' instruction, or for comparisons
2306 against null. Pointer equality tests between labels addresses results in
2307 undefined behavior --- though, again, comparison against null is ok, and
2308 no label is equal to the null pointer. This may be passed around as an
2309 opaque pointer sized value as long as the bits are not inspected. This
2310 allows ``ptrtoint`` and arithmetic to be performed on these values so
2311 long as the original value is reconstituted before the ``indirectbr``
2314 Finally, some targets may provide defined semantics when using the value
2315 as the operand to an inline assembly, but that is target specific.
2319 Constant Expressions
2320 --------------------
2322 Constant expressions are used to allow expressions involving other
2323 constants to be used as constants. Constant expressions may be of any
2324 :ref:`first class <t_firstclass>` type and may involve any LLVM operation
2325 that does not have side effects (e.g. load and call are not supported).
2326 The following is the syntax for constant expressions:
2328 ``trunc (CST to TYPE)``
2329 Truncate a constant to another type. The bit size of CST must be
2330 larger than the bit size of TYPE. Both types must be integers.
2331 ``zext (CST to TYPE)``
2332 Zero extend a constant to another type. The bit size of CST must be
2333 smaller than the bit size of TYPE. Both types must be integers.
2334 ``sext (CST to TYPE)``
2335 Sign extend a constant to another type. The bit size of CST must be
2336 smaller than the bit size of TYPE. Both types must be integers.
2337 ``fptrunc (CST to TYPE)``
2338 Truncate a floating point constant to another floating point type.
2339 The size of CST must be larger than the size of TYPE. Both types
2340 must be floating point.
2341 ``fpext (CST to TYPE)``
2342 Floating point extend a constant to another type. The size of CST
2343 must be smaller or equal to the size of TYPE. Both types must be
2345 ``fptoui (CST to TYPE)``
2346 Convert a floating point constant to the corresponding unsigned
2347 integer constant. TYPE must be a scalar or vector integer type. CST
2348 must be of scalar or vector floating point type. Both CST and TYPE
2349 must be scalars, or vectors of the same number of elements. If the
2350 value won't fit in the integer type, the results are undefined.
2351 ``fptosi (CST to TYPE)``
2352 Convert a floating point constant to the corresponding signed
2353 integer constant. TYPE must be a scalar or vector integer type. CST
2354 must be of scalar or vector floating point type. Both CST and TYPE
2355 must be scalars, or vectors of the same number of elements. If the
2356 value won't fit in the integer type, the results are undefined.
2357 ``uitofp (CST to TYPE)``
2358 Convert an unsigned integer constant to the corresponding floating
2359 point constant. TYPE must be a scalar or vector floating point type.
2360 CST must be of scalar or vector integer type. Both CST and TYPE must
2361 be scalars, or vectors of the same number of elements. If the value
2362 won't fit in the floating point type, the results are undefined.
2363 ``sitofp (CST to TYPE)``
2364 Convert a signed integer constant to the corresponding floating
2365 point constant. TYPE must be a scalar or vector floating point type.
2366 CST must be of scalar or vector integer type. Both CST and TYPE must
2367 be scalars, or vectors of the same number of elements. If the value
2368 won't fit in the floating point type, the results are undefined.
2369 ``ptrtoint (CST to TYPE)``
2370 Convert a pointer typed constant to the corresponding integer
2371 constant. ``TYPE`` must be an integer type. ``CST`` must be of
2372 pointer type. The ``CST`` value is zero extended, truncated, or
2373 unchanged to make it fit in ``TYPE``.
2374 ``inttoptr (CST to TYPE)``
2375 Convert an integer constant to a pointer constant. TYPE must be a
2376 pointer type. CST must be of integer type. The CST value is zero
2377 extended, truncated, or unchanged to make it fit in a pointer size.
2378 This one is *really* dangerous!
2379 ``bitcast (CST to TYPE)``
2380 Convert a constant, CST, to another TYPE. The constraints of the
2381 operands are the same as those for the :ref:`bitcast
2382 instruction <i_bitcast>`.
2383 ``getelementptr (CSTPTR, IDX0, IDX1, ...)``, ``getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)``
2384 Perform the :ref:`getelementptr operation <i_getelementptr>` on
2385 constants. As with the :ref:`getelementptr <i_getelementptr>`
2386 instruction, the index list may have zero or more indexes, which are
2387 required to make sense for the type of "CSTPTR".
2388 ``select (COND, VAL1, VAL2)``
2389 Perform the :ref:`select operation <i_select>` on constants.
2390 ``icmp COND (VAL1, VAL2)``
2391 Performs the :ref:`icmp operation <i_icmp>` on constants.
2392 ``fcmp COND (VAL1, VAL2)``
2393 Performs the :ref:`fcmp operation <i_fcmp>` on constants.
2394 ``extractelement (VAL, IDX)``
2395 Perform the :ref:`extractelement operation <i_extractelement>` on
2397 ``insertelement (VAL, ELT, IDX)``
2398 Perform the :ref:`insertelement operation <i_insertelement>` on
2400 ``shufflevector (VEC1, VEC2, IDXMASK)``
2401 Perform the :ref:`shufflevector operation <i_shufflevector>` on
2403 ``extractvalue (VAL, IDX0, IDX1, ...)``
2404 Perform the :ref:`extractvalue operation <i_extractvalue>` on
2405 constants. The index list is interpreted in a similar manner as
2406 indices in a ':ref:`getelementptr <i_getelementptr>`' operation. At
2407 least one index value must be specified.
2408 ``insertvalue (VAL, ELT, IDX0, IDX1, ...)``
2409 Perform the :ref:`insertvalue operation <i_insertvalue>` on constants.
2410 The index list is interpreted in a similar manner as indices in a
2411 ':ref:`getelementptr <i_getelementptr>`' operation. At least one index
2412 value must be specified.
2413 ``OPCODE (LHS, RHS)``
2414 Perform the specified operation of the LHS and RHS constants. OPCODE
2415 may be any of the :ref:`binary <binaryops>` or :ref:`bitwise
2416 binary <bitwiseops>` operations. The constraints on operands are
2417 the same as those for the corresponding instruction (e.g. no bitwise
2418 operations on floating point values are allowed).
2425 Inline Assembler Expressions
2426 ----------------------------
2428 LLVM supports inline assembler expressions (as opposed to :ref:`Module-Level
2429 Inline Assembly <moduleasm>`) through the use of a special value. This
2430 value represents the inline assembler as a string (containing the
2431 instructions to emit), a list of operand constraints (stored as a
2432 string), a flag that indicates whether or not the inline asm expression
2433 has side effects, and a flag indicating whether the function containing
2434 the asm needs to align its stack conservatively. An example inline
2435 assembler expression is:
2437 .. code-block:: llvm
2439 i32 (i32) asm "bswap $0", "=r,r"
2441 Inline assembler expressions may **only** be used as the callee operand
2442 of a :ref:`call <i_call>` or an :ref:`invoke <i_invoke>` instruction.
2443 Thus, typically we have:
2445 .. code-block:: llvm
2447 %X = call i32 asm "bswap $0", "=r,r"(i32 %Y)
2449 Inline asms with side effects not visible in the constraint list must be
2450 marked as having side effects. This is done through the use of the
2451 '``sideeffect``' keyword, like so:
2453 .. code-block:: llvm
2455 call void asm sideeffect "eieio", ""()
2457 In some cases inline asms will contain code that will not work unless
2458 the stack is aligned in some way, such as calls or SSE instructions on
2459 x86, yet will not contain code that does that alignment within the asm.
2460 The compiler should make conservative assumptions about what the asm
2461 might contain and should generate its usual stack alignment code in the
2462 prologue if the '``alignstack``' keyword is present:
2464 .. code-block:: llvm
2466 call void asm alignstack "eieio", ""()
2468 Inline asms also support using non-standard assembly dialects. The
2469 assumed dialect is ATT. When the '``inteldialect``' keyword is present,
2470 the inline asm is using the Intel dialect. Currently, ATT and Intel are
2471 the only supported dialects. An example is:
2473 .. code-block:: llvm
2475 call void asm inteldialect "eieio", ""()
2477 If multiple keywords appear the '``sideeffect``' keyword must come
2478 first, the '``alignstack``' keyword second and the '``inteldialect``'
2484 The call instructions that wrap inline asm nodes may have a
2485 "``!srcloc``" MDNode attached to it that contains a list of constant
2486 integers. If present, the code generator will use the integer as the
2487 location cookie value when report errors through the ``LLVMContext``
2488 error reporting mechanisms. This allows a front-end to correlate backend
2489 errors that occur with inline asm back to the source code that produced
2492 .. code-block:: llvm
2494 call void asm sideeffect "something bad", ""(), !srcloc !42
2496 !42 = !{ i32 1234567 }
2498 It is up to the front-end to make sense of the magic numbers it places
2499 in the IR. If the MDNode contains multiple constants, the code generator
2500 will use the one that corresponds to the line of the asm that the error
2505 Metadata Nodes and Metadata Strings
2506 -----------------------------------
2508 LLVM IR allows metadata to be attached to instructions in the program
2509 that can convey extra information about the code to the optimizers and
2510 code generator. One example application of metadata is source-level
2511 debug information. There are two metadata primitives: strings and nodes.
2512 All metadata has the ``metadata`` type and is identified in syntax by a
2513 preceding exclamation point ('``!``').
2515 A metadata string is a string surrounded by double quotes. It can
2516 contain any character by escaping non-printable characters with
2517 "``\xx``" where "``xx``" is the two digit hex code. For example:
2520 Metadata nodes are represented with notation similar to structure
2521 constants (a comma separated list of elements, surrounded by braces and
2522 preceded by an exclamation point). Metadata nodes can have any values as
2523 their operand. For example:
2525 .. code-block:: llvm
2527 !{ metadata !"test\00", i32 10}
2529 A :ref:`named metadata <namedmetadatastructure>` is a collection of
2530 metadata nodes, which can be looked up in the module symbol table. For
2533 .. code-block:: llvm
2535 !foo = metadata !{!4, !3}
2537 Metadata can be used as function arguments. Here ``llvm.dbg.value``
2538 function is using two metadata arguments:
2540 .. code-block:: llvm
2542 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2544 Metadata can be attached with an instruction. Here metadata ``!21`` is
2545 attached to the ``add`` instruction using the ``!dbg`` identifier:
2547 .. code-block:: llvm
2549 %indvar.next = add i64 %indvar, 1, !dbg !21
2551 More information about specific metadata nodes recognized by the
2552 optimizers and code generator is found below.
2557 In LLVM IR, memory does not have types, so LLVM's own type system is not
2558 suitable for doing TBAA. Instead, metadata is added to the IR to
2559 describe a type system of a higher level language. This can be used to
2560 implement typical C/C++ TBAA, but it can also be used to implement
2561 custom alias analysis behavior for other languages.
2563 The current metadata format is very simple. TBAA metadata nodes have up
2564 to three fields, e.g.:
2566 .. code-block:: llvm
2568 !0 = metadata !{ metadata !"an example type tree" }
2569 !1 = metadata !{ metadata !"int", metadata !0 }
2570 !2 = metadata !{ metadata !"float", metadata !0 }
2571 !3 = metadata !{ metadata !"const float", metadata !2, i64 1 }
2573 The first field is an identity field. It can be any value, usually a
2574 metadata string, which uniquely identifies the type. The most important
2575 name in the tree is the name of the root node. Two trees with different
2576 root node names are entirely disjoint, even if they have leaves with
2579 The second field identifies the type's parent node in the tree, or is
2580 null or omitted for a root node. A type is considered to alias all of
2581 its descendants and all of its ancestors in the tree. Also, a type is
2582 considered to alias all types in other trees, so that bitcode produced
2583 from multiple front-ends is handled conservatively.
2585 If the third field is present, it's an integer which if equal to 1
2586 indicates that the type is "constant" (meaning
2587 ``pointsToConstantMemory`` should return true; see `other useful
2588 AliasAnalysis methods <AliasAnalysis.html#OtherItfs>`_).
2590 '``tbaa.struct``' Metadata
2591 ^^^^^^^^^^^^^^^^^^^^^^^^^^
2593 The :ref:`llvm.memcpy <int_memcpy>` is often used to implement
2594 aggregate assignment operations in C and similar languages, however it
2595 is defined to copy a contiguous region of memory, which is more than
2596 strictly necessary for aggregate types which contain holes due to
2597 padding. Also, it doesn't contain any TBAA information about the fields
2600 ``!tbaa.struct`` metadata can describe which memory subregions in a
2601 memcpy are padding and what the TBAA tags of the struct are.
2603 The current metadata format is very simple. ``!tbaa.struct`` metadata
2604 nodes are a list of operands which are in conceptual groups of three.
2605 For each group of three, the first operand gives the byte offset of a
2606 field in bytes, the second gives its size in bytes, and the third gives
2609 .. code-block:: llvm
2611 !4 = metadata !{ i64 0, i64 4, metadata !1, i64 8, i64 4, metadata !2 }
2613 This describes a struct with two fields. The first is at offset 0 bytes
2614 with size 4 bytes, and has tbaa tag !1. The second is at offset 8 bytes
2615 and has size 4 bytes and has tbaa tag !2.
2617 Note that the fields need not be contiguous. In this example, there is a
2618 4 byte gap between the two fields. This gap represents padding which
2619 does not carry useful data and need not be preserved.
2621 '``fpmath``' Metadata
2622 ^^^^^^^^^^^^^^^^^^^^^
2624 ``fpmath`` metadata may be attached to any instruction of floating point
2625 type. It can be used to express the maximum acceptable error in the
2626 result of that instruction, in ULPs, thus potentially allowing the
2627 compiler to use a more efficient but less accurate method of computing
2628 it. ULP is defined as follows:
2630 If ``x`` is a real number that lies between two finite consecutive
2631 floating-point numbers ``a`` and ``b``, without being equal to one
2632 of them, then ``ulp(x) = |b - a|``, otherwise ``ulp(x)`` is the
2633 distance between the two non-equal finite floating-point numbers
2634 nearest ``x``. Moreover, ``ulp(NaN)`` is ``NaN``.
2636 The metadata node shall consist of a single positive floating point
2637 number representing the maximum relative error, for example:
2639 .. code-block:: llvm
2641 !0 = metadata !{ float 2.5 } ; maximum acceptable inaccuracy is 2.5 ULPs
2643 '``range``' Metadata
2644 ^^^^^^^^^^^^^^^^^^^^
2646 ``range`` metadata may be attached only to loads of integer types. It
2647 expresses the possible ranges the loaded value is in. The ranges are
2648 represented with a flattened list of integers. The loaded value is known
2649 to be in the union of the ranges defined by each consecutive pair. Each
2650 pair has the following properties:
2652 - The type must match the type loaded by the instruction.
2653 - The pair ``a,b`` represents the range ``[a,b)``.
2654 - Both ``a`` and ``b`` are constants.
2655 - The range is allowed to wrap.
2656 - The range should not represent the full or empty set. That is,
2659 In addition, the pairs must be in signed order of the lower bound and
2660 they must be non-contiguous.
2664 .. code-block:: llvm
2666 %a = load i8* %x, align 1, !range !0 ; Can only be 0 or 1
2667 %b = load i8* %y, align 1, !range !1 ; Can only be 255 (-1), 0 or 1
2668 %c = load i8* %z, align 1, !range !2 ; Can only be 0, 1, 3, 4 or 5
2669 %d = load i8* %z, align 1, !range !3 ; Can only be -2, -1, 3, 4 or 5
2671 !0 = metadata !{ i8 0, i8 2 }
2672 !1 = metadata !{ i8 255, i8 2 }
2673 !2 = metadata !{ i8 0, i8 2, i8 3, i8 6 }
2674 !3 = metadata !{ i8 -2, i8 0, i8 3, i8 6 }
2679 It is sometimes useful to attach information to loop constructs. Currently,
2680 loop metadata is implemented as metadata attached to the branch instruction
2681 in the loop latch block. This type of metadata refer to a metadata node that is
2682 guaranteed to be separate for each loop. The loop identifier metadata is
2683 specified with the name ``llvm.loop``.
2685 The loop identifier metadata is implemented using a metadata that refers to
2686 itself to avoid merging it with any other identifier metadata, e.g.,
2687 during module linkage or function inlining. That is, each loop should refer
2688 to their own identification metadata even if they reside in separate functions.
2689 The following example contains loop identifier metadata for two separate loop
2692 .. code-block:: llvm
2694 !0 = metadata !{ metadata !0 }
2695 !1 = metadata !{ metadata !1 }
2697 The loop identifier metadata can be used to specify additional per-loop
2698 metadata. Any operands after the first operand can be treated as user-defined
2699 metadata. For example the ``llvm.vectorizer.unroll`` metadata is understood
2700 by the loop vectorizer to indicate how many times to unroll the loop:
2702 .. code-block:: llvm
2704 br i1 %exitcond, label %._crit_edge, label %.lr.ph, !llvm.loop !0
2706 !0 = metadata !{ metadata !0, metadata !1 }
2707 !1 = metadata !{ metadata !"llvm.vectorizer.unroll", i32 2 }
2712 Metadata types used to annotate memory accesses with information helpful
2713 for optimizations are prefixed with ``llvm.mem``.
2715 '``llvm.mem.parallel_loop_access``' Metadata
2716 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2718 For a loop to be parallel, in addition to using
2719 the ``llvm.loop`` metadata to mark the loop latch branch instruction,
2720 also all of the memory accessing instructions in the loop body need to be
2721 marked with the ``llvm.mem.parallel_loop_access`` metadata. If there
2722 is at least one memory accessing instruction not marked with the metadata,
2723 the loop must be considered a sequential loop. This causes parallel loops to be
2724 converted to sequential loops due to optimization passes that are unaware of
2725 the parallel semantics and that insert new memory instructions to the loop
2728 Example of a loop that is considered parallel due to its correct use of
2729 both ``llvm.loop`` and ``llvm.mem.parallel_loop_access``
2730 metadata types that refer to the same loop identifier metadata.
2732 .. code-block:: llvm
2736 %0 = load i32* %arrayidx, align 4, !llvm.mem.parallel_loop_access !0
2738 store i32 %0, i32* %arrayidx4, align 4, !llvm.mem.parallel_loop_access !0
2740 br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0
2744 !0 = metadata !{ metadata !0 }
2746 It is also possible to have nested parallel loops. In that case the
2747 memory accesses refer to a list of loop identifier metadata nodes instead of
2748 the loop identifier metadata node directly:
2750 .. code-block:: llvm
2757 %0 = load i32* %arrayidx, align 4, !llvm.mem.parallel_loop_access !0
2759 store i32 %0, i32* %arrayidx4, align 4, !llvm.mem.parallel_loop_access !0
2761 br i1 %exitcond, label %inner.for.end, label %inner.for.body, !llvm.loop !1
2765 %0 = load i32* %arrayidx, align 4, !llvm.mem.parallel_loop_access !0
2767 store i32 %0, i32* %arrayidx4, align 4, !llvm.mem.parallel_loop_access !0
2769 br i1 %exitcond, label %outer.for.end, label %outer.for.body, !llvm.loop !2
2771 outer.for.end: ; preds = %for.body
2773 !0 = metadata !{ metadata !1, metadata !2 } ; a list of loop identifiers
2774 !1 = metadata !{ metadata !1 } ; an identifier for the inner loop
2775 !2 = metadata !{ metadata !2 } ; an identifier for the outer loop
2777 '``llvm.vectorizer``'
2778 ^^^^^^^^^^^^^^^^^^^^^
2780 Metadata prefixed with ``llvm.vectorizer`` is used to control per-loop
2781 vectorization parameters such as vectorization factor and unroll factor.
2783 ``llvm.vectorizer`` metadata should be used in conjunction with ``llvm.loop``
2784 loop identification metadata.
2786 '``llvm.vectorizer.unroll``' Metadata
2787 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2789 This metadata instructs the loop vectorizer to unroll the specified
2790 loop exactly ``N`` times.
2792 The first operand is the string ``llvm.vectorizer.unroll`` and the second
2793 operand is an integer specifying the unroll factor. For example:
2795 .. code-block:: llvm
2797 !0 = metadata !{ metadata !"llvm.vectorizer.unroll", i32 4 }
2799 Note that setting ``llvm.vectorizer.unroll`` to 1 disables unrolling of the
2802 If ``llvm.vectorizer.unroll`` is set to 0 then the amount of unrolling will be
2803 determined automatically.
2805 '``llvm.vectorizer.width``' Metadata
2806 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2808 This metadata sets the target width of the vectorizer to ``N``. Without
2809 this metadata, the vectorizer will choose a width automatically.
2810 Regardless of this metadata, the vectorizer will only vectorize loops if
2811 it believes it is valid to do so.
2813 The first operand is the string ``llvm.vectorizer.width`` and the second
2814 operand is an integer specifying the width. For example:
2816 .. code-block:: llvm
2818 !0 = metadata !{ metadata !"llvm.vectorizer.width", i32 4 }
2820 Note that setting ``llvm.vectorizer.width`` to 1 disables vectorization of the
2823 If ``llvm.vectorizer.width`` is set to 0 then the width will be determined
2826 Module Flags Metadata
2827 =====================
2829 Information about the module as a whole is difficult to convey to LLVM's
2830 subsystems. The LLVM IR isn't sufficient to transmit this information.
2831 The ``llvm.module.flags`` named metadata exists in order to facilitate
2832 this. These flags are in the form of key / value pairs --- much like a
2833 dictionary --- making it easy for any subsystem who cares about a flag to
2836 The ``llvm.module.flags`` metadata contains a list of metadata triplets.
2837 Each triplet has the following form:
2839 - The first element is a *behavior* flag, which specifies the behavior
2840 when two (or more) modules are merged together, and it encounters two
2841 (or more) metadata with the same ID. The supported behaviors are
2843 - The second element is a metadata string that is a unique ID for the
2844 metadata. Each module may only have one flag entry for each unique ID (not
2845 including entries with the **Require** behavior).
2846 - The third element is the value of the flag.
2848 When two (or more) modules are merged together, the resulting
2849 ``llvm.module.flags`` metadata is the union of the modules' flags. That is, for
2850 each unique metadata ID string, there will be exactly one entry in the merged
2851 modules ``llvm.module.flags`` metadata table, and the value for that entry will
2852 be determined by the merge behavior flag, as described below. The only exception
2853 is that entries with the *Require* behavior are always preserved.
2855 The following behaviors are supported:
2866 Emits an error if two values disagree, otherwise the resulting value
2867 is that of the operands.
2871 Emits a warning if two values disagree. The result value will be the
2872 operand for the flag from the first module being linked.
2876 Adds a requirement that another module flag be present and have a
2877 specified value after linking is performed. The value must be a
2878 metadata pair, where the first element of the pair is the ID of the
2879 module flag to be restricted, and the second element of the pair is
2880 the value the module flag should be restricted to. This behavior can
2881 be used to restrict the allowable results (via triggering of an
2882 error) of linking IDs with the **Override** behavior.
2886 Uses the specified value, regardless of the behavior or value of the
2887 other module. If both modules specify **Override**, but the values
2888 differ, an error will be emitted.
2892 Appends the two values, which are required to be metadata nodes.
2896 Appends the two values, which are required to be metadata
2897 nodes. However, duplicate entries in the second list are dropped
2898 during the append operation.
2900 It is an error for a particular unique flag ID to have multiple behaviors,
2901 except in the case of **Require** (which adds restrictions on another metadata
2902 value) or **Override**.
2904 An example of module flags:
2906 .. code-block:: llvm
2908 !0 = metadata !{ i32 1, metadata !"foo", i32 1 }
2909 !1 = metadata !{ i32 4, metadata !"bar", i32 37 }
2910 !2 = metadata !{ i32 2, metadata !"qux", i32 42 }
2911 !3 = metadata !{ i32 3, metadata !"qux",
2913 metadata !"foo", i32 1
2916 !llvm.module.flags = !{ !0, !1, !2, !3 }
2918 - Metadata ``!0`` has the ID ``!"foo"`` and the value '1'. The behavior
2919 if two or more ``!"foo"`` flags are seen is to emit an error if their
2920 values are not equal.
2922 - Metadata ``!1`` has the ID ``!"bar"`` and the value '37'. The
2923 behavior if two or more ``!"bar"`` flags are seen is to use the value
2926 - Metadata ``!2`` has the ID ``!"qux"`` and the value '42'. The
2927 behavior if two or more ``!"qux"`` flags are seen is to emit a
2928 warning if their values are not equal.
2930 - Metadata ``!3`` has the ID ``!"qux"`` and the value:
2934 metadata !{ metadata !"foo", i32 1 }
2936 The behavior is to emit an error if the ``llvm.module.flags`` does not
2937 contain a flag with the ID ``!"foo"`` that has the value '1' after linking is
2940 Objective-C Garbage Collection Module Flags Metadata
2941 ----------------------------------------------------
2943 On the Mach-O platform, Objective-C stores metadata about garbage
2944 collection in a special section called "image info". The metadata
2945 consists of a version number and a bitmask specifying what types of
2946 garbage collection are supported (if any) by the file. If two or more
2947 modules are linked together their garbage collection metadata needs to
2948 be merged rather than appended together.
2950 The Objective-C garbage collection module flags metadata consists of the
2951 following key-value pairs:
2960 * - ``Objective-C Version``
2961 - **[Required]** --- The Objective-C ABI version. Valid values are 1 and 2.
2963 * - ``Objective-C Image Info Version``
2964 - **[Required]** --- The version of the image info section. Currently
2967 * - ``Objective-C Image Info Section``
2968 - **[Required]** --- The section to place the metadata. Valid values are
2969 ``"__OBJC, __image_info, regular"`` for Objective-C ABI version 1, and
2970 ``"__DATA,__objc_imageinfo, regular, no_dead_strip"`` for
2971 Objective-C ABI version 2.
2973 * - ``Objective-C Garbage Collection``
2974 - **[Required]** --- Specifies whether garbage collection is supported or
2975 not. Valid values are 0, for no garbage collection, and 2, for garbage
2976 collection supported.
2978 * - ``Objective-C GC Only``
2979 - **[Optional]** --- Specifies that only garbage collection is supported.
2980 If present, its value must be 6. This flag requires that the
2981 ``Objective-C Garbage Collection`` flag have the value 2.
2983 Some important flag interactions:
2985 - If a module with ``Objective-C Garbage Collection`` set to 0 is
2986 merged with a module with ``Objective-C Garbage Collection`` set to
2987 2, then the resulting module has the
2988 ``Objective-C Garbage Collection`` flag set to 0.
2989 - A module with ``Objective-C Garbage Collection`` set to 0 cannot be
2990 merged with a module with ``Objective-C GC Only`` set to 6.
2992 Automatic Linker Flags Module Flags Metadata
2993 --------------------------------------------
2995 Some targets support embedding flags to the linker inside individual object
2996 files. Typically this is used in conjunction with language extensions which
2997 allow source files to explicitly declare the libraries they depend on, and have
2998 these automatically be transmitted to the linker via object files.
3000 These flags are encoded in the IR using metadata in the module flags section,
3001 using the ``Linker Options`` key. The merge behavior for this flag is required
3002 to be ``AppendUnique``, and the value for the key is expected to be a metadata
3003 node which should be a list of other metadata nodes, each of which should be a
3004 list of metadata strings defining linker options.
3006 For example, the following metadata section specifies two separate sets of
3007 linker options, presumably to link against ``libz`` and the ``Cocoa``
3010 !0 = metadata !{ i32 6, metadata !"Linker Options",
3012 metadata !{ metadata !"-lz" },
3013 metadata !{ metadata !"-framework", metadata !"Cocoa" } } }
3014 !llvm.module.flags = !{ !0 }
3016 The metadata encoding as lists of lists of options, as opposed to a collapsed
3017 list of options, is chosen so that the IR encoding can use multiple option
3018 strings to specify e.g., a single library, while still having that specifier be
3019 preserved as an atomic element that can be recognized by a target specific
3020 assembly writer or object file emitter.
3022 Each individual option is required to be either a valid option for the target's
3023 linker, or an option that is reserved by the target specific assembly writer or
3024 object file emitter. No other aspect of these options is defined by the IR.
3026 .. _intrinsicglobalvariables:
3028 Intrinsic Global Variables
3029 ==========================
3031 LLVM has a number of "magic" global variables that contain data that
3032 affect code generation or other IR semantics. These are documented here.
3033 All globals of this sort should have a section specified as
3034 "``llvm.metadata``". This section and all globals that start with
3035 "``llvm.``" are reserved for use by LLVM.
3039 The '``llvm.used``' Global Variable
3040 -----------------------------------
3042 The ``@llvm.used`` global is an array which has
3043 :ref:`appending linkage <linkage_appending>`. This array contains a list of
3044 pointers to named global variables, functions and aliases which may optionally
3045 have a pointer cast formed of bitcast or getelementptr. For example, a legal
3048 .. code-block:: llvm
3053 @llvm.used = appending global [2 x i8*] [
3055 i8* bitcast (i32* @Y to i8*)
3056 ], section "llvm.metadata"
3058 If a symbol appears in the ``@llvm.used`` list, then the compiler, assembler,
3059 and linker are required to treat the symbol as if there is a reference to the
3060 symbol that it cannot see (which is why they have to be named). For example, if
3061 a variable has internal linkage and no references other than that from the
3062 ``@llvm.used`` list, it cannot be deleted. This is commonly used to represent
3063 references from inline asms and other things the compiler cannot "see", and
3064 corresponds to "``attribute((used))``" in GNU C.
3066 On some targets, the code generator must emit a directive to the
3067 assembler or object file to prevent the assembler and linker from
3068 molesting the symbol.
3070 .. _gv_llvmcompilerused:
3072 The '``llvm.compiler.used``' Global Variable
3073 --------------------------------------------
3075 The ``@llvm.compiler.used`` directive is the same as the ``@llvm.used``
3076 directive, except that it only prevents the compiler from touching the
3077 symbol. On targets that support it, this allows an intelligent linker to
3078 optimize references to the symbol without being impeded as it would be
3081 This is a rare construct that should only be used in rare circumstances,
3082 and should not be exposed to source languages.
3084 .. _gv_llvmglobalctors:
3086 The '``llvm.global_ctors``' Global Variable
3087 -------------------------------------------
3089 .. code-block:: llvm
3091 %0 = type { i32, void ()* }
3092 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
3094 The ``@llvm.global_ctors`` array contains a list of constructor
3095 functions and associated priorities. The functions referenced by this
3096 array will be called in ascending order of priority (i.e. lowest first)
3097 when the module is loaded. The order of functions with the same priority
3100 .. _llvmglobaldtors:
3102 The '``llvm.global_dtors``' Global Variable
3103 -------------------------------------------
3105 .. code-block:: llvm
3107 %0 = type { i32, void ()* }
3108 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
3110 The ``@llvm.global_dtors`` array contains a list of destructor functions
3111 and associated priorities. The functions referenced by this array will
3112 be called in descending order of priority (i.e. highest first) when the
3113 module is loaded. The order of functions with the same priority is not
3116 Instruction Reference
3117 =====================
3119 The LLVM instruction set consists of several different classifications
3120 of instructions: :ref:`terminator instructions <terminators>`, :ref:`binary
3121 instructions <binaryops>`, :ref:`bitwise binary
3122 instructions <bitwiseops>`, :ref:`memory instructions <memoryops>`, and
3123 :ref:`other instructions <otherops>`.
3127 Terminator Instructions
3128 -----------------------
3130 As mentioned :ref:`previously <functionstructure>`, every basic block in a
3131 program ends with a "Terminator" instruction, which indicates which
3132 block should be executed after the current block is finished. These
3133 terminator instructions typically yield a '``void``' value: they produce
3134 control flow, not values (the one exception being the
3135 ':ref:`invoke <i_invoke>`' instruction).
3137 The terminator instructions are: ':ref:`ret <i_ret>`',
3138 ':ref:`br <i_br>`', ':ref:`switch <i_switch>`',
3139 ':ref:`indirectbr <i_indirectbr>`', ':ref:`invoke <i_invoke>`',
3140 ':ref:`resume <i_resume>`', and ':ref:`unreachable <i_unreachable>`'.
3144 '``ret``' Instruction
3145 ^^^^^^^^^^^^^^^^^^^^^
3152 ret <type> <value> ; Return a value from a non-void function
3153 ret void ; Return from void function
3158 The '``ret``' instruction is used to return control flow (and optionally
3159 a value) from a function back to the caller.
3161 There are two forms of the '``ret``' instruction: one that returns a
3162 value and then causes control flow, and one that just causes control
3168 The '``ret``' instruction optionally accepts a single argument, the
3169 return value. The type of the return value must be a ':ref:`first
3170 class <t_firstclass>`' type.
3172 A function is not :ref:`well formed <wellformed>` if it it has a non-void
3173 return type and contains a '``ret``' instruction with no return value or
3174 a return value with a type that does not match its type, or if it has a
3175 void return type and contains a '``ret``' instruction with a return
3181 When the '``ret``' instruction is executed, control flow returns back to
3182 the calling function's context. If the caller is a
3183 ":ref:`call <i_call>`" instruction, execution continues at the
3184 instruction after the call. If the caller was an
3185 ":ref:`invoke <i_invoke>`" instruction, execution continues at the
3186 beginning of the "normal" destination block. If the instruction returns
3187 a value, that value shall set the call or invoke instruction's return
3193 .. code-block:: llvm
3195 ret i32 5 ; Return an integer value of 5
3196 ret void ; Return from a void function
3197 ret { i32, i8 } { i32 4, i8 2 } ; Return a struct of values 4 and 2
3201 '``br``' Instruction
3202 ^^^^^^^^^^^^^^^^^^^^
3209 br i1 <cond>, label <iftrue>, label <iffalse>
3210 br label <dest> ; Unconditional branch
3215 The '``br``' instruction is used to cause control flow to transfer to a
3216 different basic block in the current function. There are two forms of
3217 this instruction, corresponding to a conditional branch and an
3218 unconditional branch.
3223 The conditional branch form of the '``br``' instruction takes a single
3224 '``i1``' value and two '``label``' values. The unconditional form of the
3225 '``br``' instruction takes a single '``label``' value as a target.
3230 Upon execution of a conditional '``br``' instruction, the '``i1``'
3231 argument is evaluated. If the value is ``true``, control flows to the
3232 '``iftrue``' ``label`` argument. If "cond" is ``false``, control flows
3233 to the '``iffalse``' ``label`` argument.
3238 .. code-block:: llvm
3241 %cond = icmp eq i32 %a, %b
3242 br i1 %cond, label %IfEqual, label %IfUnequal
3250 '``switch``' Instruction
3251 ^^^^^^^^^^^^^^^^^^^^^^^^
3258 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
3263 The '``switch``' instruction is used to transfer control flow to one of
3264 several different places. It is a generalization of the '``br``'
3265 instruction, allowing a branch to occur to one of many possible
3271 The '``switch``' instruction uses three parameters: an integer
3272 comparison value '``value``', a default '``label``' destination, and an
3273 array of pairs of comparison value constants and '``label``'s. The table
3274 is not allowed to contain duplicate constant entries.
3279 The ``switch`` instruction specifies a table of values and destinations.
3280 When the '``switch``' instruction is executed, this table is searched
3281 for the given value. If the value is found, control flow is transferred
3282 to the corresponding destination; otherwise, control flow is transferred
3283 to the default destination.
3288 Depending on properties of the target machine and the particular
3289 ``switch`` instruction, this instruction may be code generated in
3290 different ways. For example, it could be generated as a series of
3291 chained conditional branches or with a lookup table.
3296 .. code-block:: llvm
3298 ; Emulate a conditional br instruction
3299 %Val = zext i1 %value to i32
3300 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
3302 ; Emulate an unconditional br instruction
3303 switch i32 0, label %dest [ ]
3305 ; Implement a jump table:
3306 switch i32 %val, label %otherwise [ i32 0, label %onzero
3308 i32 2, label %ontwo ]
3312 '``indirectbr``' Instruction
3313 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3320 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
3325 The '``indirectbr``' instruction implements an indirect branch to a
3326 label within the current function, whose address is specified by
3327 "``address``". Address must be derived from a
3328 :ref:`blockaddress <blockaddress>` constant.
3333 The '``address``' argument is the address of the label to jump to. The
3334 rest of the arguments indicate the full set of possible destinations
3335 that the address may point to. Blocks are allowed to occur multiple
3336 times in the destination list, though this isn't particularly useful.
3338 This destination list is required so that dataflow analysis has an
3339 accurate understanding of the CFG.
3344 Control transfers to the block specified in the address argument. All
3345 possible destination blocks must be listed in the label list, otherwise
3346 this instruction has undefined behavior. This implies that jumps to
3347 labels defined in other functions have undefined behavior as well.
3352 This is typically implemented with a jump through a register.
3357 .. code-block:: llvm
3359 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3363 '``invoke``' Instruction
3364 ^^^^^^^^^^^^^^^^^^^^^^^^
3371 <result> = invoke [cconv] [ret attrs] <ptr to function ty> <function ptr val>(<function args>) [fn attrs]
3372 to label <normal label> unwind label <exception label>
3377 The '``invoke``' instruction causes control to transfer to a specified
3378 function, with the possibility of control flow transfer to either the
3379 '``normal``' label or the '``exception``' label. If the callee function
3380 returns with the "``ret``" instruction, control flow will return to the
3381 "normal" label. If the callee (or any indirect callees) returns via the
3382 ":ref:`resume <i_resume>`" instruction or other exception handling
3383 mechanism, control is interrupted and continued at the dynamically
3384 nearest "exception" label.
3386 The '``exception``' label is a `landing
3387 pad <ExceptionHandling.html#overview>`_ for the exception. As such,
3388 '``exception``' label is required to have the
3389 ":ref:`landingpad <i_landingpad>`" instruction, which contains the
3390 information about the behavior of the program after unwinding happens,
3391 as its first non-PHI instruction. The restrictions on the
3392 "``landingpad``" instruction's tightly couples it to the "``invoke``"
3393 instruction, so that the important information contained within the
3394 "``landingpad``" instruction can't be lost through normal code motion.
3399 This instruction requires several arguments:
3401 #. The optional "cconv" marker indicates which :ref:`calling
3402 convention <callingconv>` the call should use. If none is
3403 specified, the call defaults to using C calling conventions.
3404 #. The optional :ref:`Parameter Attributes <paramattrs>` list for return
3405 values. Only '``zeroext``', '``signext``', and '``inreg``' attributes
3407 #. '``ptr to function ty``': shall be the signature of the pointer to
3408 function value being invoked. In most cases, this is a direct
3409 function invocation, but indirect ``invoke``'s are just as possible,
3410 branching off an arbitrary pointer to function value.
3411 #. '``function ptr val``': An LLVM value containing a pointer to a
3412 function to be invoked.
3413 #. '``function args``': argument list whose types match the function
3414 signature argument types and parameter attributes. All arguments must
3415 be of :ref:`first class <t_firstclass>` type. If the function signature
3416 indicates the function accepts a variable number of arguments, the
3417 extra arguments can be specified.
3418 #. '``normal label``': the label reached when the called function
3419 executes a '``ret``' instruction.
3420 #. '``exception label``': the label reached when a callee returns via
3421 the :ref:`resume <i_resume>` instruction or other exception handling
3423 #. The optional :ref:`function attributes <fnattrs>` list. Only
3424 '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``'
3425 attributes are valid here.
3430 This instruction is designed to operate as a standard '``call``'
3431 instruction in most regards. The primary difference is that it
3432 establishes an association with a label, which is used by the runtime
3433 library to unwind the stack.
3435 This instruction is used in languages with destructors to ensure that
3436 proper cleanup is performed in the case of either a ``longjmp`` or a
3437 thrown exception. Additionally, this is important for implementation of
3438 '``catch``' clauses in high-level languages that support them.
3440 For the purposes of the SSA form, the definition of the value returned
3441 by the '``invoke``' instruction is deemed to occur on the edge from the
3442 current block to the "normal" label. If the callee unwinds then no
3443 return value is available.
3448 .. code-block:: llvm
3450 %retval = invoke i32 @Test(i32 15) to label %Continue
3451 unwind label %TestCleanup ; {i32}:retval set
3452 %retval = invoke coldcc i32 %Testfnptr(i32 15) to label %Continue
3453 unwind label %TestCleanup ; {i32}:retval set
3457 '``resume``' Instruction
3458 ^^^^^^^^^^^^^^^^^^^^^^^^
3465 resume <type> <value>
3470 The '``resume``' instruction is a terminator instruction that has no
3476 The '``resume``' instruction requires one argument, which must have the
3477 same type as the result of any '``landingpad``' instruction in the same
3483 The '``resume``' instruction resumes propagation of an existing
3484 (in-flight) exception whose unwinding was interrupted with a
3485 :ref:`landingpad <i_landingpad>` instruction.
3490 .. code-block:: llvm
3492 resume { i8*, i32 } %exn
3496 '``unreachable``' Instruction
3497 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3509 The '``unreachable``' instruction has no defined semantics. This
3510 instruction is used to inform the optimizer that a particular portion of
3511 the code is not reachable. This can be used to indicate that the code
3512 after a no-return function cannot be reached, and other facts.
3517 The '``unreachable``' instruction has no defined semantics.
3524 Binary operators are used to do most of the computation in a program.
3525 They require two operands of the same type, execute an operation on
3526 them, and produce a single value. The operands might represent multiple
3527 data, as is the case with the :ref:`vector <t_vector>` data type. The
3528 result value has the same type as its operands.
3530 There are several different binary operators:
3534 '``add``' Instruction
3535 ^^^^^^^^^^^^^^^^^^^^^
3542 <result> = add <ty> <op1>, <op2> ; yields {ty}:result
3543 <result> = add nuw <ty> <op1>, <op2> ; yields {ty}:result
3544 <result> = add nsw <ty> <op1>, <op2> ; yields {ty}:result
3545 <result> = add nuw nsw <ty> <op1>, <op2> ; yields {ty}:result
3550 The '``add``' instruction returns the sum of its two operands.
3555 The two arguments to the '``add``' instruction must be
3556 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
3557 arguments must have identical types.
3562 The value produced is the integer sum of the two operands.
3564 If the sum has unsigned overflow, the result returned is the
3565 mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of
3568 Because LLVM integers use a two's complement representation, this
3569 instruction is appropriate for both signed and unsigned integers.
3571 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
3572 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
3573 result value of the ``add`` is a :ref:`poison value <poisonvalues>` if
3574 unsigned and/or signed overflow, respectively, occurs.
3579 .. code-block:: llvm
3581 <result> = add i32 4, %var ; yields {i32}:result = 4 + %var
3585 '``fadd``' Instruction
3586 ^^^^^^^^^^^^^^^^^^^^^^
3593 <result> = fadd [fast-math flags]* <ty> <op1>, <op2> ; yields {ty}:result
3598 The '``fadd``' instruction returns the sum of its two operands.
3603 The two arguments to the '``fadd``' instruction must be :ref:`floating
3604 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
3605 Both arguments must have identical types.
3610 The value produced is the floating point sum of the two operands. This
3611 instruction can also take any number of :ref:`fast-math flags <fastmath>`,
3612 which are optimization hints to enable otherwise unsafe floating point
3618 .. code-block:: llvm
3620 <result> = fadd float 4.0, %var ; yields {float}:result = 4.0 + %var
3622 '``sub``' Instruction
3623 ^^^^^^^^^^^^^^^^^^^^^
3630 <result> = sub <ty> <op1>, <op2> ; yields {ty}:result
3631 <result> = sub nuw <ty> <op1>, <op2> ; yields {ty}:result
3632 <result> = sub nsw <ty> <op1>, <op2> ; yields {ty}:result
3633 <result> = sub nuw nsw <ty> <op1>, <op2> ; yields {ty}:result
3638 The '``sub``' instruction returns the difference of its two operands.
3640 Note that the '``sub``' instruction is used to represent the '``neg``'
3641 instruction present in most other intermediate representations.
3646 The two arguments to the '``sub``' instruction must be
3647 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
3648 arguments must have identical types.
3653 The value produced is the integer difference of the two operands.
3655 If the difference has unsigned overflow, the result returned is the
3656 mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of
3659 Because LLVM integers use a two's complement representation, this
3660 instruction is appropriate for both signed and unsigned integers.
3662 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
3663 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
3664 result value of the ``sub`` is a :ref:`poison value <poisonvalues>` if
3665 unsigned and/or signed overflow, respectively, occurs.
3670 .. code-block:: llvm
3672 <result> = sub i32 4, %var ; yields {i32}:result = 4 - %var
3673 <result> = sub i32 0, %val ; yields {i32}:result = -%var
3677 '``fsub``' Instruction
3678 ^^^^^^^^^^^^^^^^^^^^^^
3685 <result> = fsub [fast-math flags]* <ty> <op1>, <op2> ; yields {ty}:result
3690 The '``fsub``' instruction returns the difference of its two operands.
3692 Note that the '``fsub``' instruction is used to represent the '``fneg``'
3693 instruction present in most other intermediate representations.
3698 The two arguments to the '``fsub``' instruction must be :ref:`floating
3699 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
3700 Both arguments must have identical types.
3705 The value produced is the floating point difference of the two operands.
3706 This instruction can also take any number of :ref:`fast-math
3707 flags <fastmath>`, which are optimization hints to enable otherwise
3708 unsafe floating point optimizations:
3713 .. code-block:: llvm
3715 <result> = fsub float 4.0, %var ; yields {float}:result = 4.0 - %var
3716 <result> = fsub float -0.0, %val ; yields {float}:result = -%var
3718 '``mul``' Instruction
3719 ^^^^^^^^^^^^^^^^^^^^^
3726 <result> = mul <ty> <op1>, <op2> ; yields {ty}:result
3727 <result> = mul nuw <ty> <op1>, <op2> ; yields {ty}:result
3728 <result> = mul nsw <ty> <op1>, <op2> ; yields {ty}:result
3729 <result> = mul nuw nsw <ty> <op1>, <op2> ; yields {ty}:result
3734 The '``mul``' instruction returns the product of its two operands.
3739 The two arguments to the '``mul``' instruction must be
3740 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
3741 arguments must have identical types.
3746 The value produced is the integer product of the two operands.
3748 If the result of the multiplication has unsigned overflow, the result
3749 returned is the mathematical result modulo 2\ :sup:`n`\ , where n is the
3750 bit width of the result.
3752 Because LLVM integers use a two's complement representation, and the
3753 result is the same width as the operands, this instruction returns the
3754 correct result for both signed and unsigned integers. If a full product
3755 (e.g. ``i32`` * ``i32`` -> ``i64``) is needed, the operands should be
3756 sign-extended or zero-extended as appropriate to the width of the full
3759 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
3760 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
3761 result value of the ``mul`` is a :ref:`poison value <poisonvalues>` if
3762 unsigned and/or signed overflow, respectively, occurs.
3767 .. code-block:: llvm
3769 <result> = mul i32 4, %var ; yields {i32}:result = 4 * %var
3773 '``fmul``' Instruction
3774 ^^^^^^^^^^^^^^^^^^^^^^
3781 <result> = fmul [fast-math flags]* <ty> <op1>, <op2> ; yields {ty}:result
3786 The '``fmul``' instruction returns the product of its two operands.
3791 The two arguments to the '``fmul``' instruction must be :ref:`floating
3792 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
3793 Both arguments must have identical types.
3798 The value produced is the floating point product of the two operands.
3799 This instruction can also take any number of :ref:`fast-math
3800 flags <fastmath>`, which are optimization hints to enable otherwise
3801 unsafe floating point optimizations:
3806 .. code-block:: llvm
3808 <result> = fmul float 4.0, %var ; yields {float}:result = 4.0 * %var
3810 '``udiv``' Instruction
3811 ^^^^^^^^^^^^^^^^^^^^^^
3818 <result> = udiv <ty> <op1>, <op2> ; yields {ty}:result
3819 <result> = udiv exact <ty> <op1>, <op2> ; yields {ty}:result
3824 The '``udiv``' instruction returns the quotient of its two operands.
3829 The two arguments to the '``udiv``' instruction must be
3830 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
3831 arguments must have identical types.
3836 The value produced is the unsigned integer quotient of the two operands.
3838 Note that unsigned integer division and signed integer division are
3839 distinct operations; for signed integer division, use '``sdiv``'.
3841 Division by zero leads to undefined behavior.
3843 If the ``exact`` keyword is present, the result value of the ``udiv`` is
3844 a :ref:`poison value <poisonvalues>` if %op1 is not a multiple of %op2 (as
3845 such, "((a udiv exact b) mul b) == a").
3850 .. code-block:: llvm
3852 <result> = udiv i32 4, %var ; yields {i32}:result = 4 / %var
3854 '``sdiv``' Instruction
3855 ^^^^^^^^^^^^^^^^^^^^^^
3862 <result> = sdiv <ty> <op1>, <op2> ; yields {ty}:result
3863 <result> = sdiv exact <ty> <op1>, <op2> ; yields {ty}:result
3868 The '``sdiv``' instruction returns the quotient of its two operands.
3873 The two arguments to the '``sdiv``' instruction must be
3874 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
3875 arguments must have identical types.
3880 The value produced is the signed integer quotient of the two operands
3881 rounded towards zero.
3883 Note that signed integer division and unsigned integer division are
3884 distinct operations; for unsigned integer division, use '``udiv``'.
3886 Division by zero leads to undefined behavior. Overflow also leads to
3887 undefined behavior; this is a rare case, but can occur, for example, by
3888 doing a 32-bit division of -2147483648 by -1.
3890 If the ``exact`` keyword is present, the result value of the ``sdiv`` is
3891 a :ref:`poison value <poisonvalues>` if the result would be rounded.
3896 .. code-block:: llvm
3898 <result> = sdiv i32 4, %var ; yields {i32}:result = 4 / %var
3902 '``fdiv``' Instruction
3903 ^^^^^^^^^^^^^^^^^^^^^^
3910 <result> = fdiv [fast-math flags]* <ty> <op1>, <op2> ; yields {ty}:result
3915 The '``fdiv``' instruction returns the quotient of its two operands.
3920 The two arguments to the '``fdiv``' instruction must be :ref:`floating
3921 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
3922 Both arguments must have identical types.
3927 The value produced is the floating point quotient of the two operands.
3928 This instruction can also take any number of :ref:`fast-math
3929 flags <fastmath>`, which are optimization hints to enable otherwise
3930 unsafe floating point optimizations:
3935 .. code-block:: llvm
3937 <result> = fdiv float 4.0, %var ; yields {float}:result = 4.0 / %var
3939 '``urem``' Instruction
3940 ^^^^^^^^^^^^^^^^^^^^^^
3947 <result> = urem <ty> <op1>, <op2> ; yields {ty}:result
3952 The '``urem``' instruction returns the remainder from the unsigned
3953 division of its two arguments.
3958 The two arguments to the '``urem``' instruction must be
3959 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
3960 arguments must have identical types.
3965 This instruction returns the unsigned integer *remainder* of a division.
3966 This instruction always performs an unsigned division to get the
3969 Note that unsigned integer remainder and signed integer remainder are
3970 distinct operations; for signed integer remainder, use '``srem``'.
3972 Taking the remainder of a division by zero leads to undefined behavior.
3977 .. code-block:: llvm
3979 <result> = urem i32 4, %var ; yields {i32}:result = 4 % %var
3981 '``srem``' Instruction
3982 ^^^^^^^^^^^^^^^^^^^^^^
3989 <result> = srem <ty> <op1>, <op2> ; yields {ty}:result
3994 The '``srem``' instruction returns the remainder from the signed
3995 division of its two operands. This instruction can also take
3996 :ref:`vector <t_vector>` versions of the values in which case the elements
4002 The two arguments to the '``srem``' instruction must be
4003 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4004 arguments must have identical types.
4009 This instruction returns the *remainder* of a division (where the result
4010 is either zero or has the same sign as the dividend, ``op1``), not the
4011 *modulo* operator (where the result is either zero or has the same sign
4012 as the divisor, ``op2``) of a value. For more information about the
4013 difference, see `The Math
4014 Forum <http://mathforum.org/dr.math/problems/anne.4.28.99.html>`_. For a
4015 table of how this is implemented in various languages, please see
4017 operation <http://en.wikipedia.org/wiki/Modulo_operation>`_.
4019 Note that signed integer remainder and unsigned integer remainder are
4020 distinct operations; for unsigned integer remainder, use '``urem``'.
4022 Taking the remainder of a division by zero leads to undefined behavior.
4023 Overflow also leads to undefined behavior; this is a rare case, but can
4024 occur, for example, by taking the remainder of a 32-bit division of
4025 -2147483648 by -1. (The remainder doesn't actually overflow, but this
4026 rule lets srem be implemented using instructions that return both the
4027 result of the division and the remainder.)
4032 .. code-block:: llvm
4034 <result> = srem i32 4, %var ; yields {i32}:result = 4 % %var
4038 '``frem``' Instruction
4039 ^^^^^^^^^^^^^^^^^^^^^^
4046 <result> = frem [fast-math flags]* <ty> <op1>, <op2> ; yields {ty}:result
4051 The '``frem``' instruction returns the remainder from the division of
4057 The two arguments to the '``frem``' instruction must be :ref:`floating
4058 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
4059 Both arguments must have identical types.
4064 This instruction returns the *remainder* of a division. The remainder
4065 has the same sign as the dividend. This instruction can also take any
4066 number of :ref:`fast-math flags <fastmath>`, which are optimization hints
4067 to enable otherwise unsafe floating point optimizations:
4072 .. code-block:: llvm
4074 <result> = frem float 4.0, %var ; yields {float}:result = 4.0 % %var
4078 Bitwise Binary Operations
4079 -------------------------
4081 Bitwise binary operators are used to do various forms of bit-twiddling
4082 in a program. They are generally very efficient instructions and can
4083 commonly be strength reduced from other instructions. They require two
4084 operands of the same type, execute an operation on them, and produce a
4085 single value. The resulting value is the same type as its operands.
4087 '``shl``' Instruction
4088 ^^^^^^^^^^^^^^^^^^^^^
4095 <result> = shl <ty> <op1>, <op2> ; yields {ty}:result
4096 <result> = shl nuw <ty> <op1>, <op2> ; yields {ty}:result
4097 <result> = shl nsw <ty> <op1>, <op2> ; yields {ty}:result
4098 <result> = shl nuw nsw <ty> <op1>, <op2> ; yields {ty}:result
4103 The '``shl``' instruction returns the first operand shifted to the left
4104 a specified number of bits.
4109 Both arguments to the '``shl``' instruction must be the same
4110 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
4111 '``op2``' is treated as an unsigned value.
4116 The value produced is ``op1`` \* 2\ :sup:`op2` mod 2\ :sup:`n`,
4117 where ``n`` is the width of the result. If ``op2`` is (statically or
4118 dynamically) negative or equal to or larger than the number of bits in
4119 ``op1``, the result is undefined. If the arguments are vectors, each
4120 vector element of ``op1`` is shifted by the corresponding shift amount
4123 If the ``nuw`` keyword is present, then the shift produces a :ref:`poison
4124 value <poisonvalues>` if it shifts out any non-zero bits. If the
4125 ``nsw`` keyword is present, then the shift produces a :ref:`poison
4126 value <poisonvalues>` if it shifts out any bits that disagree with the
4127 resultant sign bit. As such, NUW/NSW have the same semantics as they
4128 would if the shift were expressed as a mul instruction with the same
4129 nsw/nuw bits in (mul %op1, (shl 1, %op2)).
4134 .. code-block:: llvm
4136 <result> = shl i32 4, %var ; yields {i32}: 4 << %var
4137 <result> = shl i32 4, 2 ; yields {i32}: 16
4138 <result> = shl i32 1, 10 ; yields {i32}: 1024
4139 <result> = shl i32 1, 32 ; undefined
4140 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 2, i32 4>
4142 '``lshr``' Instruction
4143 ^^^^^^^^^^^^^^^^^^^^^^
4150 <result> = lshr <ty> <op1>, <op2> ; yields {ty}:result
4151 <result> = lshr exact <ty> <op1>, <op2> ; yields {ty}:result
4156 The '``lshr``' instruction (logical shift right) returns the first
4157 operand shifted to the right a specified number of bits with zero fill.
4162 Both arguments to the '``lshr``' instruction must be the same
4163 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
4164 '``op2``' is treated as an unsigned value.
4169 This instruction always performs a logical shift right operation. The
4170 most significant bits of the result will be filled with zero bits after
4171 the shift. If ``op2`` is (statically or dynamically) equal to or larger
4172 than the number of bits in ``op1``, the result is undefined. If the
4173 arguments are vectors, each vector element of ``op1`` is shifted by the
4174 corresponding shift amount in ``op2``.
4176 If the ``exact`` keyword is present, the result value of the ``lshr`` is
4177 a :ref:`poison value <poisonvalues>` if any of the bits shifted out are
4183 .. code-block:: llvm
4185 <result> = lshr i32 4, 1 ; yields {i32}:result = 2
4186 <result> = lshr i32 4, 2 ; yields {i32}:result = 1
4187 <result> = lshr i8 4, 3 ; yields {i8}:result = 0
4188 <result> = lshr i8 -2, 1 ; yields {i8}:result = 0x7F
4189 <result> = lshr i32 1, 32 ; undefined
4190 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1>
4192 '``ashr``' Instruction
4193 ^^^^^^^^^^^^^^^^^^^^^^
4200 <result> = ashr <ty> <op1>, <op2> ; yields {ty}:result
4201 <result> = ashr exact <ty> <op1>, <op2> ; yields {ty}:result
4206 The '``ashr``' instruction (arithmetic shift right) returns the first
4207 operand shifted to the right a specified number of bits with sign
4213 Both arguments to the '``ashr``' instruction must be the same
4214 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
4215 '``op2``' is treated as an unsigned value.
4220 This instruction always performs an arithmetic shift right operation,
4221 The most significant bits of the result will be filled with the sign bit
4222 of ``op1``. If ``op2`` is (statically or dynamically) equal to or larger
4223 than the number of bits in ``op1``, the result is undefined. If the
4224 arguments are vectors, each vector element of ``op1`` is shifted by the
4225 corresponding shift amount in ``op2``.
4227 If the ``exact`` keyword is present, the result value of the ``ashr`` is
4228 a :ref:`poison value <poisonvalues>` if any of the bits shifted out are
4234 .. code-block:: llvm
4236 <result> = ashr i32 4, 1 ; yields {i32}:result = 2
4237 <result> = ashr i32 4, 2 ; yields {i32}:result = 1
4238 <result> = ashr i8 4, 3 ; yields {i8}:result = 0
4239 <result> = ashr i8 -2, 1 ; yields {i8}:result = -1
4240 <result> = ashr i32 1, 32 ; undefined
4241 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> ; yields: result=<2 x i32> < i32 -1, i32 0>
4243 '``and``' Instruction
4244 ^^^^^^^^^^^^^^^^^^^^^
4251 <result> = and <ty> <op1>, <op2> ; yields {ty}:result
4256 The '``and``' instruction returns the bitwise logical and of its two
4262 The two arguments to the '``and``' instruction must be
4263 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4264 arguments must have identical types.
4269 The truth table used for the '``and``' instruction is:
4286 .. code-block:: llvm
4288 <result> = and i32 4, %var ; yields {i32}:result = 4 & %var
4289 <result> = and i32 15, 40 ; yields {i32}:result = 8
4290 <result> = and i32 4, 8 ; yields {i32}:result = 0
4292 '``or``' Instruction
4293 ^^^^^^^^^^^^^^^^^^^^
4300 <result> = or <ty> <op1>, <op2> ; yields {ty}:result
4305 The '``or``' instruction returns the bitwise logical inclusive or of its
4311 The two arguments to the '``or``' instruction must be
4312 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4313 arguments must have identical types.
4318 The truth table used for the '``or``' instruction is:
4337 <result> = or i32 4, %var ; yields {i32}:result = 4 | %var
4338 <result> = or i32 15, 40 ; yields {i32}:result = 47
4339 <result> = or i32 4, 8 ; yields {i32}:result = 12
4341 '``xor``' Instruction
4342 ^^^^^^^^^^^^^^^^^^^^^
4349 <result> = xor <ty> <op1>, <op2> ; yields {ty}:result
4354 The '``xor``' instruction returns the bitwise logical exclusive or of
4355 its two operands. The ``xor`` is used to implement the "one's
4356 complement" operation, which is the "~" operator in C.
4361 The two arguments to the '``xor``' instruction must be
4362 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
4363 arguments must have identical types.
4368 The truth table used for the '``xor``' instruction is:
4385 .. code-block:: llvm
4387 <result> = xor i32 4, %var ; yields {i32}:result = 4 ^ %var
4388 <result> = xor i32 15, 40 ; yields {i32}:result = 39
4389 <result> = xor i32 4, 8 ; yields {i32}:result = 12
4390 <result> = xor i32 %V, -1 ; yields {i32}:result = ~%V
4395 LLVM supports several instructions to represent vector operations in a
4396 target-independent manner. These instructions cover the element-access
4397 and vector-specific operations needed to process vectors effectively.
4398 While LLVM does directly support these vector operations, many
4399 sophisticated algorithms will want to use target-specific intrinsics to
4400 take full advantage of a specific target.
4402 .. _i_extractelement:
4404 '``extractelement``' Instruction
4405 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4412 <result> = extractelement <n x <ty>> <val>, i32 <idx> ; yields <ty>
4417 The '``extractelement``' instruction extracts a single scalar element
4418 from a vector at a specified index.
4423 The first operand of an '``extractelement``' instruction is a value of
4424 :ref:`vector <t_vector>` type. The second operand is an index indicating
4425 the position from which to extract the element. The index may be a
4431 The result is a scalar of the same type as the element type of ``val``.
4432 Its value is the value at position ``idx`` of ``val``. If ``idx``
4433 exceeds the length of ``val``, the results are undefined.
4438 .. code-block:: llvm
4440 <result> = extractelement <4 x i32> %vec, i32 0 ; yields i32
4442 .. _i_insertelement:
4444 '``insertelement``' Instruction
4445 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4452 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> ; yields <n x <ty>>
4457 The '``insertelement``' instruction inserts a scalar element into a
4458 vector at a specified index.
4463 The first operand of an '``insertelement``' instruction is a value of
4464 :ref:`vector <t_vector>` type. The second operand is a scalar value whose
4465 type must equal the element type of the first operand. The third operand
4466 is an index indicating the position at which to insert the value. The
4467 index may be a variable.
4472 The result is a vector of the same type as ``val``. Its element values
4473 are those of ``val`` except at position ``idx``, where it gets the value
4474 ``elt``. If ``idx`` exceeds the length of ``val``, the results are
4480 .. code-block:: llvm
4482 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 ; yields <4 x i32>
4484 .. _i_shufflevector:
4486 '``shufflevector``' Instruction
4487 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4494 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> ; yields <m x <ty>>
4499 The '``shufflevector``' instruction constructs a permutation of elements
4500 from two input vectors, returning a vector with the same element type as
4501 the input and length that is the same as the shuffle mask.
4506 The first two operands of a '``shufflevector``' instruction are vectors
4507 with the same type. The third argument is a shuffle mask whose element
4508 type is always 'i32'. The result of the instruction is a vector whose
4509 length is the same as the shuffle mask and whose element type is the
4510 same as the element type of the first two operands.
4512 The shuffle mask operand is required to be a constant vector with either
4513 constant integer or undef values.
4518 The elements of the two input vectors are numbered from left to right
4519 across both of the vectors. The shuffle mask operand specifies, for each
4520 element of the result vector, which element of the two input vectors the
4521 result element gets. The element selector may be undef (meaning "don't
4522 care") and the second operand may be undef if performing a shuffle from
4528 .. code-block:: llvm
4530 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4531 <4 x i32> <i32 0, i32 4, i32 1, i32 5> ; yields <4 x i32>
4532 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4533 <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32> - Identity shuffle.
4534 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4535 <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32>
4536 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4537 <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 > ; yields <8 x i32>
4539 Aggregate Operations
4540 --------------------
4542 LLVM supports several instructions for working with
4543 :ref:`aggregate <t_aggregate>` values.
4547 '``extractvalue``' Instruction
4548 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4555 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4560 The '``extractvalue``' instruction extracts the value of a member field
4561 from an :ref:`aggregate <t_aggregate>` value.
4566 The first operand of an '``extractvalue``' instruction is a value of
4567 :ref:`struct <t_struct>` or :ref:`array <t_array>` type. The operands are
4568 constant indices to specify which value to extract in a similar manner
4569 as indices in a '``getelementptr``' instruction.
4571 The major differences to ``getelementptr`` indexing are:
4573 - Since the value being indexed is not a pointer, the first index is
4574 omitted and assumed to be zero.
4575 - At least one index must be specified.
4576 - Not only struct indices but also array indices must be in bounds.
4581 The result is the value at the position in the aggregate specified by
4587 .. code-block:: llvm
4589 <result> = extractvalue {i32, float} %agg, 0 ; yields i32
4593 '``insertvalue``' Instruction
4594 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4601 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* ; yields <aggregate type>
4606 The '``insertvalue``' instruction inserts a value into a member field in
4607 an :ref:`aggregate <t_aggregate>` value.
4612 The first operand of an '``insertvalue``' instruction is a value of
4613 :ref:`struct <t_struct>` or :ref:`array <t_array>` type. The second operand is
4614 a first-class value to insert. The following operands are constant
4615 indices indicating the position at which to insert the value in a
4616 similar manner as indices in a '``extractvalue``' instruction. The value
4617 to insert must have the same type as the value identified by the
4623 The result is an aggregate of the same type as ``val``. Its value is
4624 that of ``val`` except that the value at the position specified by the
4625 indices is that of ``elt``.
4630 .. code-block:: llvm
4632 %agg1 = insertvalue {i32, float} undef, i32 1, 0 ; yields {i32 1, float undef}
4633 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 ; yields {i32 1, float %val}
4634 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 ; yields {i32 1, float %val}
4638 Memory Access and Addressing Operations
4639 ---------------------------------------
4641 A key design point of an SSA-based representation is how it represents
4642 memory. In LLVM, no memory locations are in SSA form, which makes things
4643 very simple. This section describes how to read, write, and allocate
4648 '``alloca``' Instruction
4649 ^^^^^^^^^^^^^^^^^^^^^^^^
4656 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] ; yields {type*}:result
4661 The '``alloca``' instruction allocates memory on the stack frame of the
4662 currently executing function, to be automatically released when this
4663 function returns to its caller. The object is always allocated in the
4664 generic address space (address space zero).
4669 The '``alloca``' instruction allocates ``sizeof(<type>)*NumElements``
4670 bytes of memory on the runtime stack, returning a pointer of the
4671 appropriate type to the program. If "NumElements" is specified, it is
4672 the number of elements allocated, otherwise "NumElements" is defaulted
4673 to be one. If a constant alignment is specified, the value result of the
4674 allocation is guaranteed to be aligned to at least that boundary. If not
4675 specified, or if zero, the target can choose to align the allocation on
4676 any convenient boundary compatible with the type.
4678 '``type``' may be any sized type.
4683 Memory is allocated; a pointer is returned. The operation is undefined
4684 if there is insufficient stack space for the allocation. '``alloca``'d
4685 memory is automatically released when the function returns. The
4686 '``alloca``' instruction is commonly used to represent automatic
4687 variables that must have an address available. When the function returns
4688 (either with the ``ret`` or ``resume`` instructions), the memory is
4689 reclaimed. Allocating zero bytes is legal, but the result is undefined.
4690 The order in which memory is allocated (ie., which way the stack grows)
4696 .. code-block:: llvm
4698 %ptr = alloca i32 ; yields {i32*}:ptr
4699 %ptr = alloca i32, i32 4 ; yields {i32*}:ptr
4700 %ptr = alloca i32, i32 4, align 1024 ; yields {i32*}:ptr
4701 %ptr = alloca i32, align 1024 ; yields {i32*}:ptr
4705 '``load``' Instruction
4706 ^^^^^^^^^^^^^^^^^^^^^^
4713 <result> = load [volatile] <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>][, !invariant.load !<index>]
4714 <result> = load atomic [volatile] <ty>* <pointer> [singlethread] <ordering>, align <alignment>
4715 !<index> = !{ i32 1 }
4720 The '``load``' instruction is used to read from memory.
4725 The argument to the ``load`` instruction specifies the memory address
4726 from which to load. The pointer must point to a :ref:`first
4727 class <t_firstclass>` type. If the ``load`` is marked as ``volatile``,
4728 then the optimizer is not allowed to modify the number or order of
4729 execution of this ``load`` with other :ref:`volatile
4730 operations <volatile>`.
4732 If the ``load`` is marked as ``atomic``, it takes an extra
4733 :ref:`ordering <ordering>` and optional ``singlethread`` argument. The
4734 ``release`` and ``acq_rel`` orderings are not valid on ``load``
4735 instructions. Atomic loads produce :ref:`defined <memmodel>` results
4736 when they may see multiple atomic stores. The type of the pointee must
4737 be an integer type whose bit width is a power of two greater than or
4738 equal to eight and less than or equal to a target-specific size limit.
4739 ``align`` must be explicitly specified on atomic loads, and the load has
4740 undefined behavior if the alignment is not set to a value which is at
4741 least the size in bytes of the pointee. ``!nontemporal`` does not have
4742 any defined semantics for atomic loads.
4744 The optional constant ``align`` argument specifies the alignment of the
4745 operation (that is, the alignment of the memory address). A value of 0
4746 or an omitted ``align`` argument means that the operation has the ABI
4747 alignment for the target. It is the responsibility of the code emitter
4748 to ensure that the alignment information is correct. Overestimating the
4749 alignment results in undefined behavior. Underestimating the alignment
4750 may produce less efficient code. An alignment of 1 is always safe.
4752 The optional ``!nontemporal`` metadata must reference a single
4753 metadata name ``<index>`` corresponding to a metadata node with one
4754 ``i32`` entry of value 1. The existence of the ``!nontemporal``
4755 metadata on the instruction tells the optimizer and code generator
4756 that this load is not expected to be reused in the cache. The code
4757 generator may select special instructions to save cache bandwidth, such
4758 as the ``MOVNT`` instruction on x86.
4760 The optional ``!invariant.load`` metadata must reference a single
4761 metadata name ``<index>`` corresponding to a metadata node with no
4762 entries. The existence of the ``!invariant.load`` metadata on the
4763 instruction tells the optimizer and code generator that this load
4764 address points to memory which does not change value during program
4765 execution. The optimizer may then move this load around, for example, by
4766 hoisting it out of loops using loop invariant code motion.
4771 The location of memory pointed to is loaded. If the value being loaded
4772 is of scalar type then the number of bytes read does not exceed the
4773 minimum number of bytes needed to hold all bits of the type. For
4774 example, loading an ``i24`` reads at most three bytes. When loading a
4775 value of a type like ``i20`` with a size that is not an integral number
4776 of bytes, the result is undefined if the value was not originally
4777 written using a store of the same type.
4782 .. code-block:: llvm
4784 %ptr = alloca i32 ; yields {i32*}:ptr
4785 store i32 3, i32* %ptr ; yields {void}
4786 %val = load i32* %ptr ; yields {i32}:val = i32 3
4790 '``store``' Instruction
4791 ^^^^^^^^^^^^^^^^^^^^^^^
4798 store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] ; yields {void}
4799 store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> ; yields {void}
4804 The '``store``' instruction is used to write to memory.
4809 There are two arguments to the ``store`` instruction: a value to store
4810 and an address at which to store it. The type of the ``<pointer>``
4811 operand must be a pointer to the :ref:`first class <t_firstclass>` type of
4812 the ``<value>`` operand. If the ``store`` is marked as ``volatile``,
4813 then the optimizer is not allowed to modify the number or order of
4814 execution of this ``store`` with other :ref:`volatile
4815 operations <volatile>`.
4817 If the ``store`` is marked as ``atomic``, it takes an extra
4818 :ref:`ordering <ordering>` and optional ``singlethread`` argument. The
4819 ``acquire`` and ``acq_rel`` orderings aren't valid on ``store``
4820 instructions. Atomic loads produce :ref:`defined <memmodel>` results
4821 when they may see multiple atomic stores. The type of the pointee must
4822 be an integer type whose bit width is a power of two greater than or
4823 equal to eight and less than or equal to a target-specific size limit.
4824 ``align`` must be explicitly specified on atomic stores, and the store
4825 has undefined behavior if the alignment is not set to a value which is
4826 at least the size in bytes of the pointee. ``!nontemporal`` does not
4827 have any defined semantics for atomic stores.
4829 The optional constant ``align`` argument specifies the alignment of the
4830 operation (that is, the alignment of the memory address). A value of 0
4831 or an omitted ``align`` argument means that the operation has the ABI
4832 alignment for the target. It is the responsibility of the code emitter
4833 to ensure that the alignment information is correct. Overestimating the
4834 alignment results in undefined behavior. Underestimating the
4835 alignment may produce less efficient code. An alignment of 1 is always
4838 The optional ``!nontemporal`` metadata must reference a single metadata
4839 name ``<index>`` corresponding to a metadata node with one ``i32`` entry of
4840 value 1. The existence of the ``!nontemporal`` metadata on the instruction
4841 tells the optimizer and code generator that this load is not expected to
4842 be reused in the cache. The code generator may select special
4843 instructions to save cache bandwidth, such as the MOVNT instruction on
4849 The contents of memory are updated to contain ``<value>`` at the
4850 location specified by the ``<pointer>`` operand. If ``<value>`` is
4851 of scalar type then the number of bytes written does not exceed the
4852 minimum number of bytes needed to hold all bits of the type. For
4853 example, storing an ``i24`` writes at most three bytes. When writing a
4854 value of a type like ``i20`` with a size that is not an integral number
4855 of bytes, it is unspecified what happens to the extra bits that do not
4856 belong to the type, but they will typically be overwritten.
4861 .. code-block:: llvm
4863 %ptr = alloca i32 ; yields {i32*}:ptr
4864 store i32 3, i32* %ptr ; yields {void}
4865 %val = load i32* %ptr ; yields {i32}:val = i32 3
4869 '``fence``' Instruction
4870 ^^^^^^^^^^^^^^^^^^^^^^^
4877 fence [singlethread] <ordering> ; yields {void}
4882 The '``fence``' instruction is used to introduce happens-before edges
4888 '``fence``' instructions take an :ref:`ordering <ordering>` argument which
4889 defines what *synchronizes-with* edges they add. They can only be given
4890 ``acquire``, ``release``, ``acq_rel``, and ``seq_cst`` orderings.
4895 A fence A which has (at least) ``release`` ordering semantics
4896 *synchronizes with* a fence B with (at least) ``acquire`` ordering
4897 semantics if and only if there exist atomic operations X and Y, both
4898 operating on some atomic object M, such that A is sequenced before X, X
4899 modifies M (either directly or through some side effect of a sequence
4900 headed by X), Y is sequenced before B, and Y observes M. This provides a
4901 *happens-before* dependency between A and B. Rather than an explicit
4902 ``fence``, one (but not both) of the atomic operations X or Y might
4903 provide a ``release`` or ``acquire`` (resp.) ordering constraint and
4904 still *synchronize-with* the explicit ``fence`` and establish the
4905 *happens-before* edge.
4907 A ``fence`` which has ``seq_cst`` ordering, in addition to having both
4908 ``acquire`` and ``release`` semantics specified above, participates in
4909 the global program order of other ``seq_cst`` operations and/or fences.
4911 The optional ":ref:`singlethread <singlethread>`" argument specifies
4912 that the fence only synchronizes with other fences in the same thread.
4913 (This is useful for interacting with signal handlers.)
4918 .. code-block:: llvm
4920 fence acquire ; yields {void}
4921 fence singlethread seq_cst ; yields {void}
4925 '``cmpxchg``' Instruction
4926 ^^^^^^^^^^^^^^^^^^^^^^^^^
4933 cmpxchg [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <ordering> ; yields {ty}
4938 The '``cmpxchg``' instruction is used to atomically modify memory. It
4939 loads a value in memory and compares it to a given value. If they are
4940 equal, it stores a new value into the memory.
4945 There are three arguments to the '``cmpxchg``' instruction: an address
4946 to operate on, a value to compare to the value currently be at that
4947 address, and a new value to place at that address if the compared values
4948 are equal. The type of '<cmp>' must be an integer type whose bit width
4949 is a power of two greater than or equal to eight and less than or equal
4950 to a target-specific size limit. '<cmp>' and '<new>' must have the same
4951 type, and the type of '<pointer>' must be a pointer to that type. If the
4952 ``cmpxchg`` is marked as ``volatile``, then the optimizer is not allowed
4953 to modify the number or order of execution of this ``cmpxchg`` with
4954 other :ref:`volatile operations <volatile>`.
4956 The :ref:`ordering <ordering>` argument specifies how this ``cmpxchg``
4957 synchronizes with other atomic operations.
4959 The optional "``singlethread``" argument declares that the ``cmpxchg``
4960 is only atomic with respect to code (usually signal handlers) running in
4961 the same thread as the ``cmpxchg``. Otherwise the cmpxchg is atomic with
4962 respect to all other code in the system.
4964 The pointer passed into cmpxchg must have alignment greater than or
4965 equal to the size in memory of the operand.
4970 The contents of memory at the location specified by the '``<pointer>``'
4971 operand is read and compared to '``<cmp>``'; if the read value is the
4972 equal, '``<new>``' is written. The original value at the location is
4975 A successful ``cmpxchg`` is a read-modify-write instruction for the purpose
4976 of identifying release sequences. A failed ``cmpxchg`` is equivalent to an
4977 atomic load with an ordering parameter determined by dropping any
4978 ``release`` part of the ``cmpxchg``'s ordering.
4983 .. code-block:: llvm
4986 %orig = atomic load i32* %ptr unordered ; yields {i32}
4990 %cmp = phi i32 [ %orig, %entry ], [%old, %loop]
4991 %squared = mul i32 %cmp, %cmp
4992 %old = cmpxchg i32* %ptr, i32 %cmp, i32 %squared ; yields {i32}
4993 %success = icmp eq i32 %cmp, %old
4994 br i1 %success, label %done, label %loop
5001 '``atomicrmw``' Instruction
5002 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
5009 atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering> ; yields {ty}
5014 The '``atomicrmw``' instruction is used to atomically modify memory.
5019 There are three arguments to the '``atomicrmw``' instruction: an
5020 operation to apply, an address whose value to modify, an argument to the
5021 operation. The operation must be one of the following keywords:
5035 The type of '<value>' must be an integer type whose bit width is a power
5036 of two greater than or equal to eight and less than or equal to a
5037 target-specific size limit. The type of the '``<pointer>``' operand must
5038 be a pointer to that type. If the ``atomicrmw`` is marked as
5039 ``volatile``, then the optimizer is not allowed to modify the number or
5040 order of execution of this ``atomicrmw`` with other :ref:`volatile
5041 operations <volatile>`.
5046 The contents of memory at the location specified by the '``<pointer>``'
5047 operand are atomically read, modified, and written back. The original
5048 value at the location is returned. The modification is specified by the
5051 - xchg: ``*ptr = val``
5052 - add: ``*ptr = *ptr + val``
5053 - sub: ``*ptr = *ptr - val``
5054 - and: ``*ptr = *ptr & val``
5055 - nand: ``*ptr = ~(*ptr & val)``
5056 - or: ``*ptr = *ptr | val``
5057 - xor: ``*ptr = *ptr ^ val``
5058 - max: ``*ptr = *ptr > val ? *ptr : val`` (using a signed comparison)
5059 - min: ``*ptr = *ptr < val ? *ptr : val`` (using a signed comparison)
5060 - umax: ``*ptr = *ptr > val ? *ptr : val`` (using an unsigned
5062 - umin: ``*ptr = *ptr < val ? *ptr : val`` (using an unsigned
5068 .. code-block:: llvm
5070 %old = atomicrmw add i32* %ptr, i32 1 acquire ; yields {i32}
5072 .. _i_getelementptr:
5074 '``getelementptr``' Instruction
5075 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5082 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
5083 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
5084 <result> = getelementptr <ptr vector> ptrval, <vector index type> idx
5089 The '``getelementptr``' instruction is used to get the address of a
5090 subelement of an :ref:`aggregate <t_aggregate>` data structure. It performs
5091 address calculation only and does not access memory.
5096 The first argument is always a pointer or a vector of pointers, and
5097 forms the basis of the calculation. The remaining arguments are indices
5098 that indicate which of the elements of the aggregate object are indexed.
5099 The interpretation of each index is dependent on the type being indexed
5100 into. The first index always indexes the pointer value given as the
5101 first argument, the second index indexes a value of the type pointed to
5102 (not necessarily the value directly pointed to, since the first index
5103 can be non-zero), etc. The first type indexed into must be a pointer
5104 value, subsequent types can be arrays, vectors, and structs. Note that
5105 subsequent types being indexed into can never be pointers, since that
5106 would require loading the pointer before continuing calculation.
5108 The type of each index argument depends on the type it is indexing into.
5109 When indexing into a (optionally packed) structure, only ``i32`` integer
5110 **constants** are allowed (when using a vector of indices they must all
5111 be the **same** ``i32`` integer constant). When indexing into an array,
5112 pointer or vector, integers of any width are allowed, and they are not
5113 required to be constant. These integers are treated as signed values
5116 For example, let's consider a C code fragment and how it gets compiled
5132 int *foo(struct ST *s) {
5133 return &s[1].Z.B[5][13];
5136 The LLVM code generated by Clang is:
5138 .. code-block:: llvm
5140 %struct.RT = type { i8, [10 x [20 x i32]], i8 }
5141 %struct.ST = type { i32, double, %struct.RT }
5143 define i32* @foo(%struct.ST* %s) nounwind uwtable readnone optsize ssp {
5145 %arrayidx = getelementptr inbounds %struct.ST* %s, i64 1, i32 2, i32 1, i64 5, i64 13
5152 In the example above, the first index is indexing into the
5153 '``%struct.ST*``' type, which is a pointer, yielding a '``%struct.ST``'
5154 = '``{ i32, double, %struct.RT }``' type, a structure. The second index
5155 indexes into the third element of the structure, yielding a
5156 '``%struct.RT``' = '``{ i8 , [10 x [20 x i32]], i8 }``' type, another
5157 structure. The third index indexes into the second element of the
5158 structure, yielding a '``[10 x [20 x i32]]``' type, an array. The two
5159 dimensions of the array are subscripted into, yielding an '``i32``'
5160 type. The '``getelementptr``' instruction returns a pointer to this
5161 element, thus computing a value of '``i32*``' type.
5163 Note that it is perfectly legal to index partially through a structure,
5164 returning a pointer to an inner element. Because of this, the LLVM code
5165 for the given testcase is equivalent to:
5167 .. code-block:: llvm
5169 define i32* @foo(%struct.ST* %s) {
5170 %t1 = getelementptr %struct.ST* %s, i32 1 ; yields %struct.ST*:%t1
5171 %t2 = getelementptr %struct.ST* %t1, i32 0, i32 2 ; yields %struct.RT*:%t2
5172 %t3 = getelementptr %struct.RT* %t2, i32 0, i32 1 ; yields [10 x [20 x i32]]*:%t3
5173 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 ; yields [20 x i32]*:%t4
5174 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 ; yields i32*:%t5
5178 If the ``inbounds`` keyword is present, the result value of the
5179 ``getelementptr`` is a :ref:`poison value <poisonvalues>` if the base
5180 pointer is not an *in bounds* address of an allocated object, or if any
5181 of the addresses that would be formed by successive addition of the
5182 offsets implied by the indices to the base address with infinitely
5183 precise signed arithmetic are not an *in bounds* address of that
5184 allocated object. The *in bounds* addresses for an allocated object are
5185 all the addresses that point into the object, plus the address one byte
5186 past the end. In cases where the base is a vector of pointers the
5187 ``inbounds`` keyword applies to each of the computations element-wise.
5189 If the ``inbounds`` keyword is not present, the offsets are added to the
5190 base address with silently-wrapping two's complement arithmetic. If the
5191 offsets have a different width from the pointer, they are sign-extended
5192 or truncated to the width of the pointer. The result value of the
5193 ``getelementptr`` may be outside the object pointed to by the base
5194 pointer. The result value may not necessarily be used to access memory
5195 though, even if it happens to point into allocated storage. See the
5196 :ref:`Pointer Aliasing Rules <pointeraliasing>` section for more
5199 The getelementptr instruction is often confusing. For some more insight
5200 into how it works, see :doc:`the getelementptr FAQ <GetElementPtr>`.
5205 .. code-block:: llvm
5207 ; yields [12 x i8]*:aptr
5208 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
5210 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
5212 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
5214 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
5216 In cases where the pointer argument is a vector of pointers, each index
5217 must be a vector with the same number of elements. For example:
5219 .. code-block:: llvm
5221 %A = getelementptr <4 x i8*> %ptrs, <4 x i64> %offsets,
5223 Conversion Operations
5224 ---------------------
5226 The instructions in this category are the conversion instructions
5227 (casting) which all take a single operand and a type. They perform
5228 various bit conversions on the operand.
5230 '``trunc .. to``' Instruction
5231 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5238 <result> = trunc <ty> <value> to <ty2> ; yields ty2
5243 The '``trunc``' instruction truncates its operand to the type ``ty2``.
5248 The '``trunc``' instruction takes a value to trunc, and a type to trunc
5249 it to. Both types must be of :ref:`integer <t_integer>` types, or vectors
5250 of the same number of integers. The bit size of the ``value`` must be
5251 larger than the bit size of the destination type, ``ty2``. Equal sized
5252 types are not allowed.
5257 The '``trunc``' instruction truncates the high order bits in ``value``
5258 and converts the remaining bits to ``ty2``. Since the source size must
5259 be larger than the destination size, ``trunc`` cannot be a *no-op cast*.
5260 It will always truncate bits.
5265 .. code-block:: llvm
5267 %X = trunc i32 257 to i8 ; yields i8:1
5268 %Y = trunc i32 123 to i1 ; yields i1:true
5269 %Z = trunc i32 122 to i1 ; yields i1:false
5270 %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> ; yields <i8 8, i8 7>
5272 '``zext .. to``' Instruction
5273 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5280 <result> = zext <ty> <value> to <ty2> ; yields ty2
5285 The '``zext``' instruction zero extends its operand to type ``ty2``.
5290 The '``zext``' instruction takes a value to cast, and a type to cast it
5291 to. Both types must be of :ref:`integer <t_integer>` types, or vectors of
5292 the same number of integers. The bit size of the ``value`` must be
5293 smaller than the bit size of the destination type, ``ty2``.
5298 The ``zext`` fills the high order bits of the ``value`` with zero bits
5299 until it reaches the size of the destination type, ``ty2``.
5301 When zero extending from i1, the result will always be either 0 or 1.
5306 .. code-block:: llvm
5308 %X = zext i32 257 to i64 ; yields i64:257
5309 %Y = zext i1 true to i32 ; yields i32:1
5310 %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7>
5312 '``sext .. to``' Instruction
5313 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5320 <result> = sext <ty> <value> to <ty2> ; yields ty2
5325 The '``sext``' sign extends ``value`` to the type ``ty2``.
5330 The '``sext``' instruction takes a value to cast, and a type to cast it
5331 to. Both types must be of :ref:`integer <t_integer>` types, or vectors of
5332 the same number of integers. The bit size of the ``value`` must be
5333 smaller than the bit size of the destination type, ``ty2``.
5338 The '``sext``' instruction performs a sign extension by copying the sign
5339 bit (highest order bit) of the ``value`` until it reaches the bit size
5340 of the type ``ty2``.
5342 When sign extending from i1, the extension always results in -1 or 0.
5347 .. code-block:: llvm
5349 %X = sext i8 -1 to i16 ; yields i16 :65535
5350 %Y = sext i1 true to i32 ; yields i32:-1
5351 %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7>
5353 '``fptrunc .. to``' Instruction
5354 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5361 <result> = fptrunc <ty> <value> to <ty2> ; yields ty2
5366 The '``fptrunc``' instruction truncates ``value`` to type ``ty2``.
5371 The '``fptrunc``' instruction takes a :ref:`floating point <t_floating>`
5372 value to cast and a :ref:`floating point <t_floating>` type to cast it to.
5373 The size of ``value`` must be larger than the size of ``ty2``. This
5374 implies that ``fptrunc`` cannot be used to make a *no-op cast*.
5379 The '``fptrunc``' instruction truncates a ``value`` from a larger
5380 :ref:`floating point <t_floating>` type to a smaller :ref:`floating
5381 point <t_floating>` type. If the value cannot fit within the
5382 destination type, ``ty2``, then the results are undefined.
5387 .. code-block:: llvm
5389 %X = fptrunc double 123.0 to float ; yields float:123.0
5390 %Y = fptrunc double 1.0E+300 to float ; yields undefined
5392 '``fpext .. to``' Instruction
5393 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5400 <result> = fpext <ty> <value> to <ty2> ; yields ty2
5405 The '``fpext``' extends a floating point ``value`` to a larger floating
5411 The '``fpext``' instruction takes a :ref:`floating point <t_floating>`
5412 ``value`` to cast, and a :ref:`floating point <t_floating>` type to cast it
5413 to. The source type must be smaller than the destination type.
5418 The '``fpext``' instruction extends the ``value`` from a smaller
5419 :ref:`floating point <t_floating>` type to a larger :ref:`floating
5420 point <t_floating>` type. The ``fpext`` cannot be used to make a
5421 *no-op cast* because it always changes bits. Use ``bitcast`` to make a
5422 *no-op cast* for a floating point cast.
5427 .. code-block:: llvm
5429 %X = fpext float 3.125 to double ; yields double:3.125000e+00
5430 %Y = fpext double %X to fp128 ; yields fp128:0xL00000000000000004000900000000000
5432 '``fptoui .. to``' Instruction
5433 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5440 <result> = fptoui <ty> <value> to <ty2> ; yields ty2
5445 The '``fptoui``' converts a floating point ``value`` to its unsigned
5446 integer equivalent of type ``ty2``.
5451 The '``fptoui``' instruction takes a value to cast, which must be a
5452 scalar or vector :ref:`floating point <t_floating>` value, and a type to
5453 cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If
5454 ``ty`` is a vector floating point type, ``ty2`` must be a vector integer
5455 type with the same number of elements as ``ty``
5460 The '``fptoui``' instruction converts its :ref:`floating
5461 point <t_floating>` operand into the nearest (rounding towards zero)
5462 unsigned integer value. If the value cannot fit in ``ty2``, the results
5468 .. code-block:: llvm
5470 %X = fptoui double 123.0 to i32 ; yields i32:123
5471 %Y = fptoui float 1.0E+300 to i1 ; yields undefined:1
5472 %Z = fptoui float 1.04E+17 to i8 ; yields undefined:1
5474 '``fptosi .. to``' Instruction
5475 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5482 <result> = fptosi <ty> <value> to <ty2> ; yields ty2
5487 The '``fptosi``' instruction converts :ref:`floating point <t_floating>`
5488 ``value`` to type ``ty2``.
5493 The '``fptosi``' instruction takes a value to cast, which must be a
5494 scalar or vector :ref:`floating point <t_floating>` value, and a type to
5495 cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If
5496 ``ty`` is a vector floating point type, ``ty2`` must be a vector integer
5497 type with the same number of elements as ``ty``
5502 The '``fptosi``' instruction converts its :ref:`floating
5503 point <t_floating>` operand into the nearest (rounding towards zero)
5504 signed integer value. If the value cannot fit in ``ty2``, the results
5510 .. code-block:: llvm
5512 %X = fptosi double -123.0 to i32 ; yields i32:-123
5513 %Y = fptosi float 1.0E-247 to i1 ; yields undefined:1
5514 %Z = fptosi float 1.04E+17 to i8 ; yields undefined:1
5516 '``uitofp .. to``' Instruction
5517 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5524 <result> = uitofp <ty> <value> to <ty2> ; yields ty2
5529 The '``uitofp``' instruction regards ``value`` as an unsigned integer
5530 and converts that value to the ``ty2`` type.
5535 The '``uitofp``' instruction takes a value to cast, which must be a
5536 scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to
5537 ``ty2``, which must be an :ref:`floating point <t_floating>` type. If
5538 ``ty`` is a vector integer type, ``ty2`` must be a vector floating point
5539 type with the same number of elements as ``ty``
5544 The '``uitofp``' instruction interprets its operand as an unsigned
5545 integer quantity and converts it to the corresponding floating point
5546 value. If the value cannot fit in the floating point value, the results
5552 .. code-block:: llvm
5554 %X = uitofp i32 257 to float ; yields float:257.0
5555 %Y = uitofp i8 -1 to double ; yields double:255.0
5557 '``sitofp .. to``' Instruction
5558 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5565 <result> = sitofp <ty> <value> to <ty2> ; yields ty2
5570 The '``sitofp``' instruction regards ``value`` as a signed integer and
5571 converts that value to the ``ty2`` type.
5576 The '``sitofp``' instruction takes a value to cast, which must be a
5577 scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to
5578 ``ty2``, which must be an :ref:`floating point <t_floating>` type. If
5579 ``ty`` is a vector integer type, ``ty2`` must be a vector floating point
5580 type with the same number of elements as ``ty``
5585 The '``sitofp``' instruction interprets its operand as a signed integer
5586 quantity and converts it to the corresponding floating point value. If
5587 the value cannot fit in the floating point value, the results are
5593 .. code-block:: llvm
5595 %X = sitofp i32 257 to float ; yields float:257.0
5596 %Y = sitofp i8 -1 to double ; yields double:-1.0
5600 '``ptrtoint .. to``' Instruction
5601 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5608 <result> = ptrtoint <ty> <value> to <ty2> ; yields ty2
5613 The '``ptrtoint``' instruction converts the pointer or a vector of
5614 pointers ``value`` to the integer (or vector of integers) type ``ty2``.
5619 The '``ptrtoint``' instruction takes a ``value`` to cast, which must be
5620 a a value of type :ref:`pointer <t_pointer>` or a vector of pointers, and a
5621 type to cast it to ``ty2``, which must be an :ref:`integer <t_integer>` or
5622 a vector of integers type.
5627 The '``ptrtoint``' instruction converts ``value`` to integer type
5628 ``ty2`` by interpreting the pointer value as an integer and either
5629 truncating or zero extending that value to the size of the integer type.
5630 If ``value`` is smaller than ``ty2`` then a zero extension is done. If
5631 ``value`` is larger than ``ty2`` then a truncation is done. If they are
5632 the same size, then nothing is done (*no-op cast*) other than a type
5638 .. code-block:: llvm
5640 %X = ptrtoint i32* %P to i8 ; yields truncation on 32-bit architecture
5641 %Y = ptrtoint i32* %P to i64 ; yields zero extension on 32-bit architecture
5642 %Z = ptrtoint <4 x i32*> %P to <4 x i64>; yields vector zero extension for a vector of addresses on 32-bit architecture
5646 '``inttoptr .. to``' Instruction
5647 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5654 <result> = inttoptr <ty> <value> to <ty2> ; yields ty2
5659 The '``inttoptr``' instruction converts an integer ``value`` to a
5660 pointer type, ``ty2``.
5665 The '``inttoptr``' instruction takes an :ref:`integer <t_integer>` value to
5666 cast, and a type to cast it to, which must be a :ref:`pointer <t_pointer>`
5672 The '``inttoptr``' instruction converts ``value`` to type ``ty2`` by
5673 applying either a zero extension or a truncation depending on the size
5674 of the integer ``value``. If ``value`` is larger than the size of a
5675 pointer then a truncation is done. If ``value`` is smaller than the size
5676 of a pointer then a zero extension is done. If they are the same size,
5677 nothing is done (*no-op cast*).
5682 .. code-block:: llvm
5684 %X = inttoptr i32 255 to i32* ; yields zero extension on 64-bit architecture
5685 %Y = inttoptr i32 255 to i32* ; yields no-op on 32-bit architecture
5686 %Z = inttoptr i64 0 to i32* ; yields truncation on 32-bit architecture
5687 %Z = inttoptr <4 x i32> %G to <4 x i8*>; yields truncation of vector G to four pointers
5691 '``bitcast .. to``' Instruction
5692 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5699 <result> = bitcast <ty> <value> to <ty2> ; yields ty2
5704 The '``bitcast``' instruction converts ``value`` to type ``ty2`` without
5710 The '``bitcast``' instruction takes a value to cast, which must be a
5711 non-aggregate first class value, and a type to cast it to, which must
5712 also be a non-aggregate :ref:`first class <t_firstclass>` type. The
5713 bit sizes of ``value`` and the destination type, ``ty2``, must be
5714 identical. If the source type is a pointer, the destination type must
5715 also be a pointer of the same size. This instruction supports bitwise
5716 conversion of vectors to integers and to vectors of other types (as
5717 long as they have the same size).
5722 The '``bitcast``' instruction converts ``value`` to type ``ty2``. It
5723 is always a *no-op cast* because no bits change with this
5724 conversion. The conversion is done as if the ``value`` had been stored
5725 to memory and read back as type ``ty2``. Pointer (or vector of
5726 pointers) types may only be converted to other pointer (or vector of
5727 pointers) types with this instruction if the pointer sizes are
5728 equal. To convert pointers to other types, use the :ref:`inttoptr
5729 <i_inttoptr>` or :ref:`ptrtoint <i_ptrtoint>` instructions first.
5734 .. code-block:: llvm
5736 %X = bitcast i8 255 to i8 ; yields i8 :-1
5737 %Y = bitcast i32* %x to sint* ; yields sint*:%x
5738 %Z = bitcast <2 x int> %V to i64; ; yields i64: %V
5739 %Z = bitcast <2 x i32*> %V to <2 x i64*> ; yields <2 x i64*>
5746 The instructions in this category are the "miscellaneous" instructions,
5747 which defy better classification.
5751 '``icmp``' Instruction
5752 ^^^^^^^^^^^^^^^^^^^^^^
5759 <result> = icmp <cond> <ty> <op1>, <op2> ; yields {i1} or {<N x i1>}:result
5764 The '``icmp``' instruction returns a boolean value or a vector of
5765 boolean values based on comparison of its two integer, integer vector,
5766 pointer, or pointer vector operands.
5771 The '``icmp``' instruction takes three operands. The first operand is
5772 the condition code indicating the kind of comparison to perform. It is
5773 not a value, just a keyword. The possible condition code are:
5776 #. ``ne``: not equal
5777 #. ``ugt``: unsigned greater than
5778 #. ``uge``: unsigned greater or equal
5779 #. ``ult``: unsigned less than
5780 #. ``ule``: unsigned less or equal
5781 #. ``sgt``: signed greater than
5782 #. ``sge``: signed greater or equal
5783 #. ``slt``: signed less than
5784 #. ``sle``: signed less or equal
5786 The remaining two arguments must be :ref:`integer <t_integer>` or
5787 :ref:`pointer <t_pointer>` or integer :ref:`vector <t_vector>` typed. They
5788 must also be identical types.
5793 The '``icmp``' compares ``op1`` and ``op2`` according to the condition
5794 code given as ``cond``. The comparison performed always yields either an
5795 :ref:`i1 <t_integer>` or vector of ``i1`` result, as follows:
5797 #. ``eq``: yields ``true`` if the operands are equal, ``false``
5798 otherwise. No sign interpretation is necessary or performed.
5799 #. ``ne``: yields ``true`` if the operands are unequal, ``false``
5800 otherwise. No sign interpretation is necessary or performed.
5801 #. ``ugt``: interprets the operands as unsigned values and yields
5802 ``true`` if ``op1`` is greater than ``op2``.
5803 #. ``uge``: interprets the operands as unsigned values and yields
5804 ``true`` if ``op1`` is greater than or equal to ``op2``.
5805 #. ``ult``: interprets the operands as unsigned values and yields
5806 ``true`` if ``op1`` is less than ``op2``.
5807 #. ``ule``: interprets the operands as unsigned values and yields
5808 ``true`` if ``op1`` is less than or equal to ``op2``.
5809 #. ``sgt``: interprets the operands as signed values and yields ``true``
5810 if ``op1`` is greater than ``op2``.
5811 #. ``sge``: interprets the operands as signed values and yields ``true``
5812 if ``op1`` is greater than or equal to ``op2``.
5813 #. ``slt``: interprets the operands as signed values and yields ``true``
5814 if ``op1`` is less than ``op2``.
5815 #. ``sle``: interprets the operands as signed values and yields ``true``
5816 if ``op1`` is less than or equal to ``op2``.
5818 If the operands are :ref:`pointer <t_pointer>` typed, the pointer values
5819 are compared as if they were integers.
5821 If the operands are integer vectors, then they are compared element by
5822 element. The result is an ``i1`` vector with the same number of elements
5823 as the values being compared. Otherwise, the result is an ``i1``.
5828 .. code-block:: llvm
5830 <result> = icmp eq i32 4, 5 ; yields: result=false
5831 <result> = icmp ne float* %X, %X ; yields: result=false
5832 <result> = icmp ult i16 4, 5 ; yields: result=true
5833 <result> = icmp sgt i16 4, 5 ; yields: result=false
5834 <result> = icmp ule i16 -4, 5 ; yields: result=false
5835 <result> = icmp sge i16 4, 5 ; yields: result=false
5837 Note that the code generator does not yet support vector types with the
5838 ``icmp`` instruction.
5842 '``fcmp``' Instruction
5843 ^^^^^^^^^^^^^^^^^^^^^^
5850 <result> = fcmp <cond> <ty> <op1>, <op2> ; yields {i1} or {<N x i1>}:result
5855 The '``fcmp``' instruction returns a boolean value or vector of boolean
5856 values based on comparison of its operands.
5858 If the operands are floating point scalars, then the result type is a
5859 boolean (:ref:`i1 <t_integer>`).
5861 If the operands are floating point vectors, then the result type is a
5862 vector of boolean with the same number of elements as the operands being
5868 The '``fcmp``' instruction takes three operands. The first operand is
5869 the condition code indicating the kind of comparison to perform. It is
5870 not a value, just a keyword. The possible condition code are:
5872 #. ``false``: no comparison, always returns false
5873 #. ``oeq``: ordered and equal
5874 #. ``ogt``: ordered and greater than
5875 #. ``oge``: ordered and greater than or equal
5876 #. ``olt``: ordered and less than
5877 #. ``ole``: ordered and less than or equal
5878 #. ``one``: ordered and not equal
5879 #. ``ord``: ordered (no nans)
5880 #. ``ueq``: unordered or equal
5881 #. ``ugt``: unordered or greater than
5882 #. ``uge``: unordered or greater than or equal
5883 #. ``ult``: unordered or less than
5884 #. ``ule``: unordered or less than or equal
5885 #. ``une``: unordered or not equal
5886 #. ``uno``: unordered (either nans)
5887 #. ``true``: no comparison, always returns true
5889 *Ordered* means that neither operand is a QNAN while *unordered* means
5890 that either operand may be a QNAN.
5892 Each of ``val1`` and ``val2`` arguments must be either a :ref:`floating
5893 point <t_floating>` type or a :ref:`vector <t_vector>` of floating point
5894 type. They must have identical types.
5899 The '``fcmp``' instruction compares ``op1`` and ``op2`` according to the
5900 condition code given as ``cond``. If the operands are vectors, then the
5901 vectors are compared element by element. Each comparison performed
5902 always yields an :ref:`i1 <t_integer>` result, as follows:
5904 #. ``false``: always yields ``false``, regardless of operands.
5905 #. ``oeq``: yields ``true`` if both operands are not a QNAN and ``op1``
5906 is equal to ``op2``.
5907 #. ``ogt``: yields ``true`` if both operands are not a QNAN and ``op1``
5908 is greater than ``op2``.
5909 #. ``oge``: yields ``true`` if both operands are not a QNAN and ``op1``
5910 is greater than or equal to ``op2``.
5911 #. ``olt``: yields ``true`` if both operands are not a QNAN and ``op1``
5912 is less than ``op2``.
5913 #. ``ole``: yields ``true`` if both operands are not a QNAN and ``op1``
5914 is less than or equal to ``op2``.
5915 #. ``one``: yields ``true`` if both operands are not a QNAN and ``op1``
5916 is not equal to ``op2``.
5917 #. ``ord``: yields ``true`` if both operands are not a QNAN.
5918 #. ``ueq``: yields ``true`` if either operand is a QNAN or ``op1`` is
5920 #. ``ugt``: yields ``true`` if either operand is a QNAN or ``op1`` is
5921 greater than ``op2``.
5922 #. ``uge``: yields ``true`` if either operand is a QNAN or ``op1`` is
5923 greater than or equal to ``op2``.
5924 #. ``ult``: yields ``true`` if either operand is a QNAN or ``op1`` is
5926 #. ``ule``: yields ``true`` if either operand is a QNAN or ``op1`` is
5927 less than or equal to ``op2``.
5928 #. ``une``: yields ``true`` if either operand is a QNAN or ``op1`` is
5929 not equal to ``op2``.
5930 #. ``uno``: yields ``true`` if either operand is a QNAN.
5931 #. ``true``: always yields ``true``, regardless of operands.
5936 .. code-block:: llvm
5938 <result> = fcmp oeq float 4.0, 5.0 ; yields: result=false
5939 <result> = fcmp one float 4.0, 5.0 ; yields: result=true
5940 <result> = fcmp olt float 4.0, 5.0 ; yields: result=true
5941 <result> = fcmp ueq double 1.0, 2.0 ; yields: result=false
5943 Note that the code generator does not yet support vector types with the
5944 ``fcmp`` instruction.
5948 '``phi``' Instruction
5949 ^^^^^^^^^^^^^^^^^^^^^
5956 <result> = phi <ty> [ <val0>, <label0>], ...
5961 The '``phi``' instruction is used to implement the φ node in the SSA
5962 graph representing the function.
5967 The type of the incoming values is specified with the first type field.
5968 After this, the '``phi``' instruction takes a list of pairs as
5969 arguments, with one pair for each predecessor basic block of the current
5970 block. Only values of :ref:`first class <t_firstclass>` type may be used as
5971 the value arguments to the PHI node. Only labels may be used as the
5974 There must be no non-phi instructions between the start of a basic block
5975 and the PHI instructions: i.e. PHI instructions must be first in a basic
5978 For the purposes of the SSA form, the use of each incoming value is
5979 deemed to occur on the edge from the corresponding predecessor block to
5980 the current block (but after any definition of an '``invoke``'
5981 instruction's return value on the same edge).
5986 At runtime, the '``phi``' instruction logically takes on the value
5987 specified by the pair corresponding to the predecessor basic block that
5988 executed just prior to the current block.
5993 .. code-block:: llvm
5995 Loop: ; Infinite loop that counts from 0 on up...
5996 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5997 %nextindvar = add i32 %indvar, 1
6002 '``select``' Instruction
6003 ^^^^^^^^^^^^^^^^^^^^^^^^
6010 <result> = select selty <cond>, <ty> <val1>, <ty> <val2> ; yields ty
6012 selty is either i1 or {<N x i1>}
6017 The '``select``' instruction is used to choose one value based on a
6018 condition, without branching.
6023 The '``select``' instruction requires an 'i1' value or a vector of 'i1'
6024 values indicating the condition, and two values of the same :ref:`first
6025 class <t_firstclass>` type. If the val1/val2 are vectors and the
6026 condition is a scalar, then entire vectors are selected, not individual
6032 If the condition is an i1 and it evaluates to 1, the instruction returns
6033 the first value argument; otherwise, it returns the second value
6036 If the condition is a vector of i1, then the value arguments must be
6037 vectors of the same size, and the selection is done element by element.
6042 .. code-block:: llvm
6044 %X = select i1 true, i8 17, i8 42 ; yields i8:17
6048 '``call``' Instruction
6049 ^^^^^^^^^^^^^^^^^^^^^^
6056 <result> = [tail] call [cconv] [ret attrs] <ty> [<fnty>*] <fnptrval>(<function args>) [fn attrs]
6061 The '``call``' instruction represents a simple function call.
6066 This instruction requires several arguments:
6068 #. The optional "tail" marker indicates that the callee function does
6069 not access any allocas or varargs in the caller. Note that calls may
6070 be marked "tail" even if they do not occur before a
6071 :ref:`ret <i_ret>` instruction. If the "tail" marker is present, the
6072 function call is eligible for tail call optimization, but `might not
6073 in fact be optimized into a jump <CodeGenerator.html#tailcallopt>`_.
6074 The code generator may optimize calls marked "tail" with either 1)
6075 automatic `sibling call
6076 optimization <CodeGenerator.html#sibcallopt>`_ when the caller and
6077 callee have matching signatures, or 2) forced tail call optimization
6078 when the following extra requirements are met:
6080 - Caller and callee both have the calling convention ``fastcc``.
6081 - The call is in tail position (ret immediately follows call and ret
6082 uses value of call or is void).
6083 - Option ``-tailcallopt`` is enabled, or
6084 ``llvm::GuaranteedTailCallOpt`` is ``true``.
6085 - `Platform specific constraints are
6086 met. <CodeGenerator.html#tailcallopt>`_
6088 #. The optional "cconv" marker indicates which :ref:`calling
6089 convention <callingconv>` the call should use. If none is
6090 specified, the call defaults to using C calling conventions. The
6091 calling convention of the call must match the calling convention of
6092 the target function, or else the behavior is undefined.
6093 #. The optional :ref:`Parameter Attributes <paramattrs>` list for return
6094 values. Only '``zeroext``', '``signext``', and '``inreg``' attributes
6096 #. '``ty``': the type of the call instruction itself which is also the
6097 type of the return value. Functions that return no value are marked
6099 #. '``fnty``': shall be the signature of the pointer to function value
6100 being invoked. The argument types must match the types implied by
6101 this signature. This type can be omitted if the function is not
6102 varargs and if the function type does not return a pointer to a
6104 #. '``fnptrval``': An LLVM value containing a pointer to a function to
6105 be invoked. In most cases, this is a direct function invocation, but
6106 indirect ``call``'s are just as possible, calling an arbitrary pointer
6108 #. '``function args``': argument list whose types match the function
6109 signature argument types and parameter attributes. All arguments must
6110 be of :ref:`first class <t_firstclass>` type. If the function signature
6111 indicates the function accepts a variable number of arguments, the
6112 extra arguments can be specified.
6113 #. The optional :ref:`function attributes <fnattrs>` list. Only
6114 '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``'
6115 attributes are valid here.
6120 The '``call``' instruction is used to cause control flow to transfer to
6121 a specified function, with its incoming arguments bound to the specified
6122 values. Upon a '``ret``' instruction in the called function, control
6123 flow continues with the instruction after the function call, and the
6124 return value of the function is bound to the result argument.
6129 .. code-block:: llvm
6131 %retval = call i32 @test(i32 %argc)
6132 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) ; yields i32
6133 %X = tail call i32 @foo() ; yields i32
6134 %Y = tail call fastcc i32 @foo() ; yields i32
6135 call void %foo(i8 97 signext)
6137 %struct.A = type { i32, i8 }
6138 %r = call %struct.A @foo() ; yields { 32, i8 }
6139 %gr = extractvalue %struct.A %r, 0 ; yields i32
6140 %gr1 = extractvalue %struct.A %r, 1 ; yields i8
6141 %Z = call void @foo() noreturn ; indicates that %foo never returns normally
6142 %ZZ = call zeroext i32 @bar() ; Return value is %zero extended
6144 llvm treats calls to some functions with names and arguments that match
6145 the standard C99 library as being the C99 library functions, and may
6146 perform optimizations or generate code for them under that assumption.
6147 This is something we'd like to change in the future to provide better
6148 support for freestanding environments and non-C-based languages.
6152 '``va_arg``' Instruction
6153 ^^^^^^^^^^^^^^^^^^^^^^^^
6160 <resultval> = va_arg <va_list*> <arglist>, <argty>
6165 The '``va_arg``' instruction is used to access arguments passed through
6166 the "variable argument" area of a function call. It is used to implement
6167 the ``va_arg`` macro in C.
6172 This instruction takes a ``va_list*`` value and the type of the
6173 argument. It returns a value of the specified argument type and
6174 increments the ``va_list`` to point to the next argument. The actual
6175 type of ``va_list`` is target specific.
6180 The '``va_arg``' instruction loads an argument of the specified type
6181 from the specified ``va_list`` and causes the ``va_list`` to point to
6182 the next argument. For more information, see the variable argument
6183 handling :ref:`Intrinsic Functions <int_varargs>`.
6185 It is legal for this instruction to be called in a function which does
6186 not take a variable number of arguments, for example, the ``vfprintf``
6189 ``va_arg`` is an LLVM instruction instead of an :ref:`intrinsic
6190 function <intrinsics>` because it takes a type as an argument.
6195 See the :ref:`variable argument processing <int_varargs>` section.
6197 Note that the code generator does not yet fully support va\_arg on many
6198 targets. Also, it does not currently support va\_arg with aggregate
6199 types on any target.
6203 '``landingpad``' Instruction
6204 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6211 <resultval> = landingpad <resultty> personality <type> <pers_fn> <clause>+
6212 <resultval> = landingpad <resultty> personality <type> <pers_fn> cleanup <clause>*
6214 <clause> := catch <type> <value>
6215 <clause> := filter <array constant type> <array constant>
6220 The '``landingpad``' instruction is used by `LLVM's exception handling
6221 system <ExceptionHandling.html#overview>`_ to specify that a basic block
6222 is a landing pad --- one where the exception lands, and corresponds to the
6223 code found in the ``catch`` portion of a ``try``/``catch`` sequence. It
6224 defines values supplied by the personality function (``pers_fn``) upon
6225 re-entry to the function. The ``resultval`` has the type ``resultty``.
6230 This instruction takes a ``pers_fn`` value. This is the personality
6231 function associated with the unwinding mechanism. The optional
6232 ``cleanup`` flag indicates that the landing pad block is a cleanup.
6234 A ``clause`` begins with the clause type --- ``catch`` or ``filter`` --- and
6235 contains the global variable representing the "type" that may be caught
6236 or filtered respectively. Unlike the ``catch`` clause, the ``filter``
6237 clause takes an array constant as its argument. Use
6238 "``[0 x i8**] undef``" for a filter which cannot throw. The
6239 '``landingpad``' instruction must contain *at least* one ``clause`` or
6240 the ``cleanup`` flag.
6245 The '``landingpad``' instruction defines the values which are set by the
6246 personality function (``pers_fn``) upon re-entry to the function, and
6247 therefore the "result type" of the ``landingpad`` instruction. As with
6248 calling conventions, how the personality function results are
6249 represented in LLVM IR is target specific.
6251 The clauses are applied in order from top to bottom. If two
6252 ``landingpad`` instructions are merged together through inlining, the
6253 clauses from the calling function are appended to the list of clauses.
6254 When the call stack is being unwound due to an exception being thrown,
6255 the exception is compared against each ``clause`` in turn. If it doesn't
6256 match any of the clauses, and the ``cleanup`` flag is not set, then
6257 unwinding continues further up the call stack.
6259 The ``landingpad`` instruction has several restrictions:
6261 - A landing pad block is a basic block which is the unwind destination
6262 of an '``invoke``' instruction.
6263 - A landing pad block must have a '``landingpad``' instruction as its
6264 first non-PHI instruction.
6265 - There can be only one '``landingpad``' instruction within the landing
6267 - A basic block that is not a landing pad block may not include a
6268 '``landingpad``' instruction.
6269 - All '``landingpad``' instructions in a function must have the same
6270 personality function.
6275 .. code-block:: llvm
6277 ;; A landing pad which can catch an integer.
6278 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6280 ;; A landing pad that is a cleanup.
6281 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6283 ;; A landing pad which can catch an integer and can only throw a double.
6284 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6286 filter [1 x i8**] [@_ZTId]
6293 LLVM supports the notion of an "intrinsic function". These functions
6294 have well known names and semantics and are required to follow certain
6295 restrictions. Overall, these intrinsics represent an extension mechanism
6296 for the LLVM language that does not require changing all of the
6297 transformations in LLVM when adding to the language (or the bitcode
6298 reader/writer, the parser, etc...).
6300 Intrinsic function names must all start with an "``llvm.``" prefix. This
6301 prefix is reserved in LLVM for intrinsic names; thus, function names may
6302 not begin with this prefix. Intrinsic functions must always be external
6303 functions: you cannot define the body of intrinsic functions. Intrinsic
6304 functions may only be used in call or invoke instructions: it is illegal
6305 to take the address of an intrinsic function. Additionally, because
6306 intrinsic functions are part of the LLVM language, it is required if any
6307 are added that they be documented here.
6309 Some intrinsic functions can be overloaded, i.e., the intrinsic
6310 represents a family of functions that perform the same operation but on
6311 different data types. Because LLVM can represent over 8 million
6312 different integer types, overloading is used commonly to allow an
6313 intrinsic function to operate on any integer type. One or more of the
6314 argument types or the result type can be overloaded to accept any
6315 integer type. Argument types may also be defined as exactly matching a
6316 previous argument's type or the result type. This allows an intrinsic
6317 function which accepts multiple arguments, but needs all of them to be
6318 of the same type, to only be overloaded with respect to a single
6319 argument or the result.
6321 Overloaded intrinsics will have the names of its overloaded argument
6322 types encoded into its function name, each preceded by a period. Only
6323 those types which are overloaded result in a name suffix. Arguments
6324 whose type is matched against another type do not. For example, the
6325 ``llvm.ctpop`` function can take an integer of any width and returns an
6326 integer of exactly the same integer width. This leads to a family of
6327 functions such as ``i8 @llvm.ctpop.i8(i8 %val)`` and
6328 ``i29 @llvm.ctpop.i29(i29 %val)``. Only one type, the return type, is
6329 overloaded, and only one type suffix is required. Because the argument's
6330 type is matched against the return type, it does not require its own
6333 To learn how to add an intrinsic function, please see the `Extending
6334 LLVM Guide <ExtendingLLVM.html>`_.
6338 Variable Argument Handling Intrinsics
6339 -------------------------------------
6341 Variable argument support is defined in LLVM with the
6342 :ref:`va_arg <i_va_arg>` instruction and these three intrinsic
6343 functions. These functions are related to the similarly named macros
6344 defined in the ``<stdarg.h>`` header file.
6346 All of these functions operate on arguments that use a target-specific
6347 value type "``va_list``". The LLVM assembly language reference manual
6348 does not define what this type is, so all transformations should be
6349 prepared to handle these functions regardless of the type used.
6351 This example shows how the :ref:`va_arg <i_va_arg>` instruction and the
6352 variable argument handling intrinsic functions are used.
6354 .. code-block:: llvm
6356 define i32 @test(i32 %X, ...) {
6357 ; Initialize variable argument processing
6359 %ap2 = bitcast i8** %ap to i8*
6360 call void @llvm.va_start(i8* %ap2)
6362 ; Read a single integer argument
6363 %tmp = va_arg i8** %ap, i32
6365 ; Demonstrate usage of llvm.va_copy and llvm.va_end
6367 %aq2 = bitcast i8** %aq to i8*
6368 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
6369 call void @llvm.va_end(i8* %aq2)
6371 ; Stop processing of arguments.
6372 call void @llvm.va_end(i8* %ap2)
6376 declare void @llvm.va_start(i8*)
6377 declare void @llvm.va_copy(i8*, i8*)
6378 declare void @llvm.va_end(i8*)
6382 '``llvm.va_start``' Intrinsic
6383 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6390 declare void @llvm.va_start(i8* <arglist>)
6395 The '``llvm.va_start``' intrinsic initializes ``*<arglist>`` for
6396 subsequent use by ``va_arg``.
6401 The argument is a pointer to a ``va_list`` element to initialize.
6406 The '``llvm.va_start``' intrinsic works just like the ``va_start`` macro
6407 available in C. In a target-dependent way, it initializes the
6408 ``va_list`` element to which the argument points, so that the next call
6409 to ``va_arg`` will produce the first variable argument passed to the
6410 function. Unlike the C ``va_start`` macro, this intrinsic does not need
6411 to know the last argument of the function as the compiler can figure
6414 '``llvm.va_end``' Intrinsic
6415 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
6422 declare void @llvm.va_end(i8* <arglist>)
6427 The '``llvm.va_end``' intrinsic destroys ``*<arglist>``, which has been
6428 initialized previously with ``llvm.va_start`` or ``llvm.va_copy``.
6433 The argument is a pointer to a ``va_list`` to destroy.
6438 The '``llvm.va_end``' intrinsic works just like the ``va_end`` macro
6439 available in C. In a target-dependent way, it destroys the ``va_list``
6440 element to which the argument points. Calls to
6441 :ref:`llvm.va_start <int_va_start>` and
6442 :ref:`llvm.va_copy <int_va_copy>` must be matched exactly with calls to
6447 '``llvm.va_copy``' Intrinsic
6448 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6455 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
6460 The '``llvm.va_copy``' intrinsic copies the current argument position
6461 from the source argument list to the destination argument list.
6466 The first argument is a pointer to a ``va_list`` element to initialize.
6467 The second argument is a pointer to a ``va_list`` element to copy from.
6472 The '``llvm.va_copy``' intrinsic works just like the ``va_copy`` macro
6473 available in C. In a target-dependent way, it copies the source
6474 ``va_list`` element into the destination ``va_list`` element. This
6475 intrinsic is necessary because the `` llvm.va_start`` intrinsic may be
6476 arbitrarily complex and require, for example, memory allocation.
6478 Accurate Garbage Collection Intrinsics
6479 --------------------------------------
6481 LLVM support for `Accurate Garbage Collection <GarbageCollection.html>`_
6482 (GC) requires the implementation and generation of these intrinsics.
6483 These intrinsics allow identification of :ref:`GC roots on the
6484 stack <int_gcroot>`, as well as garbage collector implementations that
6485 require :ref:`read <int_gcread>` and :ref:`write <int_gcwrite>` barriers.
6486 Front-ends for type-safe garbage collected languages should generate
6487 these intrinsics to make use of the LLVM garbage collectors. For more
6488 details, see `Accurate Garbage Collection with
6489 LLVM <GarbageCollection.html>`_.
6491 The garbage collection intrinsics only operate on objects in the generic
6492 address space (address space zero).
6496 '``llvm.gcroot``' Intrinsic
6497 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
6504 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
6509 The '``llvm.gcroot``' intrinsic declares the existence of a GC root to
6510 the code generator, and allows some metadata to be associated with it.
6515 The first argument specifies the address of a stack object that contains
6516 the root pointer. The second pointer (which must be either a constant or
6517 a global value address) contains the meta-data to be associated with the
6523 At runtime, a call to this intrinsic stores a null pointer into the
6524 "ptrloc" location. At compile-time, the code generator generates
6525 information to allow the runtime to find the pointer at GC safe points.
6526 The '``llvm.gcroot``' intrinsic may only be used in a function which
6527 :ref:`specifies a GC algorithm <gc>`.
6531 '``llvm.gcread``' Intrinsic
6532 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
6539 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
6544 The '``llvm.gcread``' intrinsic identifies reads of references from heap
6545 locations, allowing garbage collector implementations that require read
6551 The second argument is the address to read from, which should be an
6552 address allocated from the garbage collector. The first object is a
6553 pointer to the start of the referenced object, if needed by the language
6554 runtime (otherwise null).
6559 The '``llvm.gcread``' intrinsic has the same semantics as a load
6560 instruction, but may be replaced with substantially more complex code by
6561 the garbage collector runtime, as needed. The '``llvm.gcread``'
6562 intrinsic may only be used in a function which :ref:`specifies a GC
6567 '``llvm.gcwrite``' Intrinsic
6568 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6575 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
6580 The '``llvm.gcwrite``' intrinsic identifies writes of references to heap
6581 locations, allowing garbage collector implementations that require write
6582 barriers (such as generational or reference counting collectors).
6587 The first argument is the reference to store, the second is the start of
6588 the object to store it to, and the third is the address of the field of
6589 Obj to store to. If the runtime does not require a pointer to the
6590 object, Obj may be null.
6595 The '``llvm.gcwrite``' intrinsic has the same semantics as a store
6596 instruction, but may be replaced with substantially more complex code by
6597 the garbage collector runtime, as needed. The '``llvm.gcwrite``'
6598 intrinsic may only be used in a function which :ref:`specifies a GC
6601 Code Generator Intrinsics
6602 -------------------------
6604 These intrinsics are provided by LLVM to expose special features that
6605 may only be implemented with code generator support.
6607 '``llvm.returnaddress``' Intrinsic
6608 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6615 declare i8 *@llvm.returnaddress(i32 <level>)
6620 The '``llvm.returnaddress``' intrinsic attempts to compute a
6621 target-specific value indicating the return address of the current
6622 function or one of its callers.
6627 The argument to this intrinsic indicates which function to return the
6628 address for. Zero indicates the calling function, one indicates its
6629 caller, etc. The argument is **required** to be a constant integer
6635 The '``llvm.returnaddress``' intrinsic either returns a pointer
6636 indicating the return address of the specified call frame, or zero if it
6637 cannot be identified. The value returned by this intrinsic is likely to
6638 be incorrect or 0 for arguments other than zero, so it should only be
6639 used for debugging purposes.
6641 Note that calling this intrinsic does not prevent function inlining or
6642 other aggressive transformations, so the value returned may not be that
6643 of the obvious source-language caller.
6645 '``llvm.frameaddress``' Intrinsic
6646 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6653 declare i8* @llvm.frameaddress(i32 <level>)
6658 The '``llvm.frameaddress``' intrinsic attempts to return the
6659 target-specific frame pointer value for the specified stack frame.
6664 The argument to this intrinsic indicates which function to return the
6665 frame pointer for. Zero indicates the calling function, one indicates
6666 its caller, etc. The argument is **required** to be a constant integer
6672 The '``llvm.frameaddress``' intrinsic either returns a pointer
6673 indicating the frame address of the specified call frame, or zero if it
6674 cannot be identified. The value returned by this intrinsic is likely to
6675 be incorrect or 0 for arguments other than zero, so it should only be
6676 used for debugging purposes.
6678 Note that calling this intrinsic does not prevent function inlining or
6679 other aggressive transformations, so the value returned may not be that
6680 of the obvious source-language caller.
6684 '``llvm.stacksave``' Intrinsic
6685 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6692 declare i8* @llvm.stacksave()
6697 The '``llvm.stacksave``' intrinsic is used to remember the current state
6698 of the function stack, for use with
6699 :ref:`llvm.stackrestore <int_stackrestore>`. This is useful for
6700 implementing language features like scoped automatic variable sized
6706 This intrinsic returns a opaque pointer value that can be passed to
6707 :ref:`llvm.stackrestore <int_stackrestore>`. When an
6708 ``llvm.stackrestore`` intrinsic is executed with a value saved from
6709 ``llvm.stacksave``, it effectively restores the state of the stack to
6710 the state it was in when the ``llvm.stacksave`` intrinsic executed. In
6711 practice, this pops any :ref:`alloca <i_alloca>` blocks from the stack that
6712 were allocated after the ``llvm.stacksave`` was executed.
6714 .. _int_stackrestore:
6716 '``llvm.stackrestore``' Intrinsic
6717 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6724 declare void @llvm.stackrestore(i8* %ptr)
6729 The '``llvm.stackrestore``' intrinsic is used to restore the state of
6730 the function stack to the state it was in when the corresponding
6731 :ref:`llvm.stacksave <int_stacksave>` intrinsic executed. This is
6732 useful for implementing language features like scoped automatic variable
6733 sized arrays in C99.
6738 See the description for :ref:`llvm.stacksave <int_stacksave>`.
6740 '``llvm.prefetch``' Intrinsic
6741 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6748 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
6753 The '``llvm.prefetch``' intrinsic is a hint to the code generator to
6754 insert a prefetch instruction if supported; otherwise, it is a noop.
6755 Prefetches have no effect on the behavior of the program but can change
6756 its performance characteristics.
6761 ``address`` is the address to be prefetched, ``rw`` is the specifier
6762 determining if the fetch should be for a read (0) or write (1), and
6763 ``locality`` is a temporal locality specifier ranging from (0) - no
6764 locality, to (3) - extremely local keep in cache. The ``cache type``
6765 specifies whether the prefetch is performed on the data (1) or
6766 instruction (0) cache. The ``rw``, ``locality`` and ``cache type``
6767 arguments must be constant integers.
6772 This intrinsic does not modify the behavior of the program. In
6773 particular, prefetches cannot trap and do not produce a value. On
6774 targets that support this intrinsic, the prefetch can provide hints to
6775 the processor cache for better performance.
6777 '``llvm.pcmarker``' Intrinsic
6778 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6785 declare void @llvm.pcmarker(i32 <id>)
6790 The '``llvm.pcmarker``' intrinsic is a method to export a Program
6791 Counter (PC) in a region of code to simulators and other tools. The
6792 method is target specific, but it is expected that the marker will use
6793 exported symbols to transmit the PC of the marker. The marker makes no
6794 guarantees that it will remain with any specific instruction after
6795 optimizations. It is possible that the presence of a marker will inhibit
6796 optimizations. The intended use is to be inserted after optimizations to
6797 allow correlations of simulation runs.
6802 ``id`` is a numerical id identifying the marker.
6807 This intrinsic does not modify the behavior of the program. Backends
6808 that do not support this intrinsic may ignore it.
6810 '``llvm.readcyclecounter``' Intrinsic
6811 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6818 declare i64 @llvm.readcyclecounter()
6823 The '``llvm.readcyclecounter``' intrinsic provides access to the cycle
6824 counter register (or similar low latency, high accuracy clocks) on those
6825 targets that support it. On X86, it should map to RDTSC. On Alpha, it
6826 should map to RPCC. As the backing counters overflow quickly (on the
6827 order of 9 seconds on alpha), this should only be used for small
6833 When directly supported, reading the cycle counter should not modify any
6834 memory. Implementations are allowed to either return a application
6835 specific value or a system wide value. On backends without support, this
6836 is lowered to a constant 0.
6838 Note that runtime support may be conditional on the privilege-level code is
6839 running at and the host platform.
6841 Standard C Library Intrinsics
6842 -----------------------------
6844 LLVM provides intrinsics for a few important standard C library
6845 functions. These intrinsics allow source-language front-ends to pass
6846 information about the alignment of the pointer arguments to the code
6847 generator, providing opportunity for more efficient code generation.
6851 '``llvm.memcpy``' Intrinsic
6852 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
6857 This is an overloaded intrinsic. You can use ``llvm.memcpy`` on any
6858 integer bit width and for different address spaces. Not all targets
6859 support all bit widths however.
6863 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6864 i32 <len>, i32 <align>, i1 <isvolatile>)
6865 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6866 i64 <len>, i32 <align>, i1 <isvolatile>)
6871 The '``llvm.memcpy.*``' intrinsics copy a block of memory from the
6872 source location to the destination location.
6874 Note that, unlike the standard libc function, the ``llvm.memcpy.*``
6875 intrinsics do not return a value, takes extra alignment/isvolatile
6876 arguments and the pointers can be in specified address spaces.
6881 The first argument is a pointer to the destination, the second is a
6882 pointer to the source. The third argument is an integer argument
6883 specifying the number of bytes to copy, the fourth argument is the
6884 alignment of the source and destination locations, and the fifth is a
6885 boolean indicating a volatile access.
6887 If the call to this intrinsic has an alignment value that is not 0 or 1,
6888 then the caller guarantees that both the source and destination pointers
6889 are aligned to that boundary.
6891 If the ``isvolatile`` parameter is ``true``, the ``llvm.memcpy`` call is
6892 a :ref:`volatile operation <volatile>`. The detailed access behavior is not
6893 very cleanly specified and it is unwise to depend on it.
6898 The '``llvm.memcpy.*``' intrinsics copy a block of memory from the
6899 source location to the destination location, which are not allowed to
6900 overlap. It copies "len" bytes of memory over. If the argument is known
6901 to be aligned to some boundary, this can be specified as the fourth
6902 argument, otherwise it should be set to 0 or 1 (both meaning no alignment).
6904 '``llvm.memmove``' Intrinsic
6905 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6910 This is an overloaded intrinsic. You can use llvm.memmove on any integer
6911 bit width and for different address space. Not all targets support all
6916 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6917 i32 <len>, i32 <align>, i1 <isvolatile>)
6918 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6919 i64 <len>, i32 <align>, i1 <isvolatile>)
6924 The '``llvm.memmove.*``' intrinsics move a block of memory from the
6925 source location to the destination location. It is similar to the
6926 '``llvm.memcpy``' intrinsic but allows the two memory locations to
6929 Note that, unlike the standard libc function, the ``llvm.memmove.*``
6930 intrinsics do not return a value, takes extra alignment/isvolatile
6931 arguments and the pointers can be in specified address spaces.
6936 The first argument is a pointer to the destination, the second is a
6937 pointer to the source. The third argument is an integer argument
6938 specifying the number of bytes to copy, the fourth argument is the
6939 alignment of the source and destination locations, and the fifth is a
6940 boolean indicating a volatile access.
6942 If the call to this intrinsic has an alignment value that is not 0 or 1,
6943 then the caller guarantees that the source and destination pointers are
6944 aligned to that boundary.
6946 If the ``isvolatile`` parameter is ``true``, the ``llvm.memmove`` call
6947 is a :ref:`volatile operation <volatile>`. The detailed access behavior is
6948 not very cleanly specified and it is unwise to depend on it.
6953 The '``llvm.memmove.*``' intrinsics copy a block of memory from the
6954 source location to the destination location, which may overlap. It
6955 copies "len" bytes of memory over. If the argument is known to be
6956 aligned to some boundary, this can be specified as the fourth argument,
6957 otherwise it should be set to 0 or 1 (both meaning no alignment).
6959 '``llvm.memset.*``' Intrinsics
6960 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6965 This is an overloaded intrinsic. You can use llvm.memset on any integer
6966 bit width and for different address spaces. However, not all targets
6967 support all bit widths.
6971 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
6972 i32 <len>, i32 <align>, i1 <isvolatile>)
6973 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
6974 i64 <len>, i32 <align>, i1 <isvolatile>)
6979 The '``llvm.memset.*``' intrinsics fill a block of memory with a
6980 particular byte value.
6982 Note that, unlike the standard libc function, the ``llvm.memset``
6983 intrinsic does not return a value and takes extra alignment/volatile
6984 arguments. Also, the destination can be in an arbitrary address space.
6989 The first argument is a pointer to the destination to fill, the second
6990 is the byte value with which to fill it, the third argument is an
6991 integer argument specifying the number of bytes to fill, and the fourth
6992 argument is the known alignment of the destination location.
6994 If the call to this intrinsic has an alignment value that is not 0 or 1,
6995 then the caller guarantees that the destination pointer is aligned to
6998 If the ``isvolatile`` parameter is ``true``, the ``llvm.memset`` call is
6999 a :ref:`volatile operation <volatile>`. The detailed access behavior is not
7000 very cleanly specified and it is unwise to depend on it.
7005 The '``llvm.memset.*``' intrinsics fill "len" bytes of memory starting
7006 at the destination location. If the argument is known to be aligned to
7007 some boundary, this can be specified as the fourth argument, otherwise
7008 it should be set to 0 or 1 (both meaning no alignment).
7010 '``llvm.sqrt.*``' Intrinsic
7011 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7016 This is an overloaded intrinsic. You can use ``llvm.sqrt`` on any
7017 floating point or vector of floating point type. Not all targets support
7022 declare float @llvm.sqrt.f32(float %Val)
7023 declare double @llvm.sqrt.f64(double %Val)
7024 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
7025 declare fp128 @llvm.sqrt.f128(fp128 %Val)
7026 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
7031 The '``llvm.sqrt``' intrinsics return the sqrt of the specified operand,
7032 returning the same value as the libm '``sqrt``' functions would. Unlike
7033 ``sqrt`` in libm, however, ``llvm.sqrt`` has undefined behavior for
7034 negative numbers other than -0.0 (which allows for better optimization,
7035 because there is no need to worry about errno being set).
7036 ``llvm.sqrt(-0.0)`` is defined to return -0.0 like IEEE sqrt.
7041 The argument and return value are floating point numbers of the same
7047 This function returns the sqrt of the specified operand if it is a
7048 nonnegative floating point number.
7050 '``llvm.powi.*``' Intrinsic
7051 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7056 This is an overloaded intrinsic. You can use ``llvm.powi`` on any
7057 floating point or vector of floating point type. Not all targets support
7062 declare float @llvm.powi.f32(float %Val, i32 %power)
7063 declare double @llvm.powi.f64(double %Val, i32 %power)
7064 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
7065 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
7066 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
7071 The '``llvm.powi.*``' intrinsics return the first operand raised to the
7072 specified (positive or negative) power. The order of evaluation of
7073 multiplications is not defined. When a vector of floating point type is
7074 used, the second argument remains a scalar integer value.
7079 The second argument is an integer power, and the first is a value to
7080 raise to that power.
7085 This function returns the first value raised to the second power with an
7086 unspecified sequence of rounding operations.
7088 '``llvm.sin.*``' Intrinsic
7089 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7094 This is an overloaded intrinsic. You can use ``llvm.sin`` on any
7095 floating point or vector of floating point type. Not all targets support
7100 declare float @llvm.sin.f32(float %Val)
7101 declare double @llvm.sin.f64(double %Val)
7102 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
7103 declare fp128 @llvm.sin.f128(fp128 %Val)
7104 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
7109 The '``llvm.sin.*``' intrinsics return the sine of the operand.
7114 The argument and return value are floating point numbers of the same
7120 This function returns the sine of the specified operand, returning the
7121 same values as the libm ``sin`` functions would, and handles error
7122 conditions in the same way.
7124 '``llvm.cos.*``' Intrinsic
7125 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7130 This is an overloaded intrinsic. You can use ``llvm.cos`` on any
7131 floating point or vector of floating point type. Not all targets support
7136 declare float @llvm.cos.f32(float %Val)
7137 declare double @llvm.cos.f64(double %Val)
7138 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
7139 declare fp128 @llvm.cos.f128(fp128 %Val)
7140 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
7145 The '``llvm.cos.*``' intrinsics return the cosine of the operand.
7150 The argument and return value are floating point numbers of the same
7156 This function returns the cosine of the specified operand, returning the
7157 same values as the libm ``cos`` functions would, and handles error
7158 conditions in the same way.
7160 '``llvm.pow.*``' Intrinsic
7161 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7166 This is an overloaded intrinsic. You can use ``llvm.pow`` on any
7167 floating point or vector of floating point type. Not all targets support
7172 declare float @llvm.pow.f32(float %Val, float %Power)
7173 declare double @llvm.pow.f64(double %Val, double %Power)
7174 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
7175 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
7176 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
7181 The '``llvm.pow.*``' intrinsics return the first operand raised to the
7182 specified (positive or negative) power.
7187 The second argument is a floating point power, and the first is a value
7188 to raise to that power.
7193 This function returns the first value raised to the second power,
7194 returning the same values as the libm ``pow`` functions would, and
7195 handles error conditions in the same way.
7197 '``llvm.exp.*``' Intrinsic
7198 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7203 This is an overloaded intrinsic. You can use ``llvm.exp`` on any
7204 floating point or vector of floating point type. Not all targets support
7209 declare float @llvm.exp.f32(float %Val)
7210 declare double @llvm.exp.f64(double %Val)
7211 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
7212 declare fp128 @llvm.exp.f128(fp128 %Val)
7213 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
7218 The '``llvm.exp.*``' intrinsics perform the exp function.
7223 The argument and return value are floating point numbers of the same
7229 This function returns the same values as the libm ``exp`` functions
7230 would, and handles error conditions in the same way.
7232 '``llvm.exp2.*``' Intrinsic
7233 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7238 This is an overloaded intrinsic. You can use ``llvm.exp2`` on any
7239 floating point or vector of floating point type. Not all targets support
7244 declare float @llvm.exp2.f32(float %Val)
7245 declare double @llvm.exp2.f64(double %Val)
7246 declare x86_fp80 @llvm.exp2.f80(x86_fp80 %Val)
7247 declare fp128 @llvm.exp2.f128(fp128 %Val)
7248 declare ppc_fp128 @llvm.exp2.ppcf128(ppc_fp128 %Val)
7253 The '``llvm.exp2.*``' intrinsics perform the exp2 function.
7258 The argument and return value are floating point numbers of the same
7264 This function returns the same values as the libm ``exp2`` functions
7265 would, and handles error conditions in the same way.
7267 '``llvm.log.*``' Intrinsic
7268 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7273 This is an overloaded intrinsic. You can use ``llvm.log`` on any
7274 floating point or vector of floating point type. Not all targets support
7279 declare float @llvm.log.f32(float %Val)
7280 declare double @llvm.log.f64(double %Val)
7281 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
7282 declare fp128 @llvm.log.f128(fp128 %Val)
7283 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
7288 The '``llvm.log.*``' intrinsics perform the log function.
7293 The argument and return value are floating point numbers of the same
7299 This function returns the same values as the libm ``log`` functions
7300 would, and handles error conditions in the same way.
7302 '``llvm.log10.*``' Intrinsic
7303 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7308 This is an overloaded intrinsic. You can use ``llvm.log10`` on any
7309 floating point or vector of floating point type. Not all targets support
7314 declare float @llvm.log10.f32(float %Val)
7315 declare double @llvm.log10.f64(double %Val)
7316 declare x86_fp80 @llvm.log10.f80(x86_fp80 %Val)
7317 declare fp128 @llvm.log10.f128(fp128 %Val)
7318 declare ppc_fp128 @llvm.log10.ppcf128(ppc_fp128 %Val)
7323 The '``llvm.log10.*``' intrinsics perform the log10 function.
7328 The argument and return value are floating point numbers of the same
7334 This function returns the same values as the libm ``log10`` functions
7335 would, and handles error conditions in the same way.
7337 '``llvm.log2.*``' Intrinsic
7338 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7343 This is an overloaded intrinsic. You can use ``llvm.log2`` on any
7344 floating point or vector of floating point type. Not all targets support
7349 declare float @llvm.log2.f32(float %Val)
7350 declare double @llvm.log2.f64(double %Val)
7351 declare x86_fp80 @llvm.log2.f80(x86_fp80 %Val)
7352 declare fp128 @llvm.log2.f128(fp128 %Val)
7353 declare ppc_fp128 @llvm.log2.ppcf128(ppc_fp128 %Val)
7358 The '``llvm.log2.*``' intrinsics perform the log2 function.
7363 The argument and return value are floating point numbers of the same
7369 This function returns the same values as the libm ``log2`` functions
7370 would, and handles error conditions in the same way.
7372 '``llvm.fma.*``' Intrinsic
7373 ^^^^^^^^^^^^^^^^^^^^^^^^^^
7378 This is an overloaded intrinsic. You can use ``llvm.fma`` on any
7379 floating point or vector of floating point type. Not all targets support
7384 declare float @llvm.fma.f32(float %a, float %b, float %c)
7385 declare double @llvm.fma.f64(double %a, double %b, double %c)
7386 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
7387 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
7388 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
7393 The '``llvm.fma.*``' intrinsics perform the fused multiply-add
7399 The argument and return value are floating point numbers of the same
7405 This function returns the same values as the libm ``fma`` functions
7408 '``llvm.fabs.*``' Intrinsic
7409 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7414 This is an overloaded intrinsic. You can use ``llvm.fabs`` on any
7415 floating point or vector of floating point type. Not all targets support
7420 declare float @llvm.fabs.f32(float %Val)
7421 declare double @llvm.fabs.f64(double %Val)
7422 declare x86_fp80 @llvm.fabs.f80(x86_fp80 %Val)
7423 declare fp128 @llvm.fabs.f128(fp128 %Val)
7424 declare ppc_fp128 @llvm.fabs.ppcf128(ppc_fp128 %Val)
7429 The '``llvm.fabs.*``' intrinsics return the absolute value of the
7435 The argument and return value are floating point numbers of the same
7441 This function returns the same values as the libm ``fabs`` functions
7442 would, and handles error conditions in the same way.
7444 '``llvm.copysign.*``' Intrinsic
7445 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7450 This is an overloaded intrinsic. You can use ``llvm.copysign`` on any
7451 floating point or vector of floating point type. Not all targets support
7456 declare float @llvm.copysign.f32(float %Mag, float %Sgn)
7457 declare double @llvm.copysign.f64(double %Mag, double %Sgn)
7458 declare x86_fp80 @llvm.copysign.f80(x86_fp80 %Mag, x86_fp80 %Sgn)
7459 declare fp128 @llvm.copysign.f128(fp128 %Mag, fp128 %Sgn)
7460 declare ppc_fp128 @llvm.copysign.ppcf128(ppc_fp128 %Mag, ppc_fp128 %Sgn)
7465 The '``llvm.copysign.*``' intrinsics return a value with the magnitude of the
7466 first operand and the sign of the second operand.
7471 The arguments and return value are floating point numbers of the same
7477 This function returns the same values as the libm ``copysign``
7478 functions would, and handles error conditions in the same way.
7480 '``llvm.floor.*``' Intrinsic
7481 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7486 This is an overloaded intrinsic. You can use ``llvm.floor`` on any
7487 floating point or vector of floating point type. Not all targets support
7492 declare float @llvm.floor.f32(float %Val)
7493 declare double @llvm.floor.f64(double %Val)
7494 declare x86_fp80 @llvm.floor.f80(x86_fp80 %Val)
7495 declare fp128 @llvm.floor.f128(fp128 %Val)
7496 declare ppc_fp128 @llvm.floor.ppcf128(ppc_fp128 %Val)
7501 The '``llvm.floor.*``' intrinsics return the floor of the operand.
7506 The argument and return value are floating point numbers of the same
7512 This function returns the same values as the libm ``floor`` functions
7513 would, and handles error conditions in the same way.
7515 '``llvm.ceil.*``' Intrinsic
7516 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7521 This is an overloaded intrinsic. You can use ``llvm.ceil`` on any
7522 floating point or vector of floating point type. Not all targets support
7527 declare float @llvm.ceil.f32(float %Val)
7528 declare double @llvm.ceil.f64(double %Val)
7529 declare x86_fp80 @llvm.ceil.f80(x86_fp80 %Val)
7530 declare fp128 @llvm.ceil.f128(fp128 %Val)
7531 declare ppc_fp128 @llvm.ceil.ppcf128(ppc_fp128 %Val)
7536 The '``llvm.ceil.*``' intrinsics return the ceiling of the operand.
7541 The argument and return value are floating point numbers of the same
7547 This function returns the same values as the libm ``ceil`` functions
7548 would, and handles error conditions in the same way.
7550 '``llvm.trunc.*``' Intrinsic
7551 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7556 This is an overloaded intrinsic. You can use ``llvm.trunc`` on any
7557 floating point or vector of floating point type. Not all targets support
7562 declare float @llvm.trunc.f32(float %Val)
7563 declare double @llvm.trunc.f64(double %Val)
7564 declare x86_fp80 @llvm.trunc.f80(x86_fp80 %Val)
7565 declare fp128 @llvm.trunc.f128(fp128 %Val)
7566 declare ppc_fp128 @llvm.trunc.ppcf128(ppc_fp128 %Val)
7571 The '``llvm.trunc.*``' intrinsics returns the operand rounded to the
7572 nearest integer not larger in magnitude than the operand.
7577 The argument and return value are floating point numbers of the same
7583 This function returns the same values as the libm ``trunc`` functions
7584 would, and handles error conditions in the same way.
7586 '``llvm.rint.*``' Intrinsic
7587 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7592 This is an overloaded intrinsic. You can use ``llvm.rint`` on any
7593 floating point or vector of floating point type. Not all targets support
7598 declare float @llvm.rint.f32(float %Val)
7599 declare double @llvm.rint.f64(double %Val)
7600 declare x86_fp80 @llvm.rint.f80(x86_fp80 %Val)
7601 declare fp128 @llvm.rint.f128(fp128 %Val)
7602 declare ppc_fp128 @llvm.rint.ppcf128(ppc_fp128 %Val)
7607 The '``llvm.rint.*``' intrinsics returns the operand rounded to the
7608 nearest integer. It may raise an inexact floating-point exception if the
7609 operand isn't an integer.
7614 The argument and return value are floating point numbers of the same
7620 This function returns the same values as the libm ``rint`` functions
7621 would, and handles error conditions in the same way.
7623 '``llvm.nearbyint.*``' Intrinsic
7624 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7629 This is an overloaded intrinsic. You can use ``llvm.nearbyint`` on any
7630 floating point or vector of floating point type. Not all targets support
7635 declare float @llvm.nearbyint.f32(float %Val)
7636 declare double @llvm.nearbyint.f64(double %Val)
7637 declare x86_fp80 @llvm.nearbyint.f80(x86_fp80 %Val)
7638 declare fp128 @llvm.nearbyint.f128(fp128 %Val)
7639 declare ppc_fp128 @llvm.nearbyint.ppcf128(ppc_fp128 %Val)
7644 The '``llvm.nearbyint.*``' intrinsics returns the operand rounded to the
7650 The argument and return value are floating point numbers of the same
7656 This function returns the same values as the libm ``nearbyint``
7657 functions would, and handles error conditions in the same way.
7659 '``llvm.round.*``' Intrinsic
7660 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7665 This is an overloaded intrinsic. You can use ``llvm.round`` on any
7666 floating point or vector of floating point type. Not all targets support
7671 declare float @llvm.round.f32(float %Val)
7672 declare double @llvm.round.f64(double %Val)
7673 declare x86_fp80 @llvm.round.f80(x86_fp80 %Val)
7674 declare fp128 @llvm.round.f128(fp128 %Val)
7675 declare ppc_fp128 @llvm.round.ppcf128(ppc_fp128 %Val)
7680 The '``llvm.round.*``' intrinsics returns the operand rounded to the
7686 The argument and return value are floating point numbers of the same
7692 This function returns the same values as the libm ``round``
7693 functions would, and handles error conditions in the same way.
7695 Bit Manipulation Intrinsics
7696 ---------------------------
7698 LLVM provides intrinsics for a few important bit manipulation
7699 operations. These allow efficient code generation for some algorithms.
7701 '``llvm.bswap.*``' Intrinsics
7702 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7707 This is an overloaded intrinsic function. You can use bswap on any
7708 integer type that is an even number of bytes (i.e. BitWidth % 16 == 0).
7712 declare i16 @llvm.bswap.i16(i16 <id>)
7713 declare i32 @llvm.bswap.i32(i32 <id>)
7714 declare i64 @llvm.bswap.i64(i64 <id>)
7719 The '``llvm.bswap``' family of intrinsics is used to byte swap integer
7720 values with an even number of bytes (positive multiple of 16 bits).
7721 These are useful for performing operations on data that is not in the
7722 target's native byte order.
7727 The ``llvm.bswap.i16`` intrinsic returns an i16 value that has the high
7728 and low byte of the input i16 swapped. Similarly, the ``llvm.bswap.i32``
7729 intrinsic returns an i32 value that has the four bytes of the input i32
7730 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the
7731 returned i32 will have its bytes in 3, 2, 1, 0 order. The
7732 ``llvm.bswap.i48``, ``llvm.bswap.i64`` and other intrinsics extend this
7733 concept to additional even-byte lengths (6 bytes, 8 bytes and more,
7736 '``llvm.ctpop.*``' Intrinsic
7737 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7742 This is an overloaded intrinsic. You can use llvm.ctpop on any integer
7743 bit width, or on any vector with integer elements. Not all targets
7744 support all bit widths or vector types, however.
7748 declare i8 @llvm.ctpop.i8(i8 <src>)
7749 declare i16 @llvm.ctpop.i16(i16 <src>)
7750 declare i32 @llvm.ctpop.i32(i32 <src>)
7751 declare i64 @llvm.ctpop.i64(i64 <src>)
7752 declare i256 @llvm.ctpop.i256(i256 <src>)
7753 declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
7758 The '``llvm.ctpop``' family of intrinsics counts the number of bits set
7764 The only argument is the value to be counted. The argument may be of any
7765 integer type, or a vector with integer elements. The return type must
7766 match the argument type.
7771 The '``llvm.ctpop``' intrinsic counts the 1's in a variable, or within
7772 each element of a vector.
7774 '``llvm.ctlz.*``' Intrinsic
7775 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7780 This is an overloaded intrinsic. You can use ``llvm.ctlz`` on any
7781 integer bit width, or any vector whose elements are integers. Not all
7782 targets support all bit widths or vector types, however.
7786 declare i8 @llvm.ctlz.i8 (i8 <src>, i1 <is_zero_undef>)
7787 declare i16 @llvm.ctlz.i16 (i16 <src>, i1 <is_zero_undef>)
7788 declare i32 @llvm.ctlz.i32 (i32 <src>, i1 <is_zero_undef>)
7789 declare i64 @llvm.ctlz.i64 (i64 <src>, i1 <is_zero_undef>)
7790 declare i256 @llvm.ctlz.i256(i256 <src>, i1 <is_zero_undef>)
7791 declase <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
7796 The '``llvm.ctlz``' family of intrinsic functions counts the number of
7797 leading zeros in a variable.
7802 The first argument is the value to be counted. This argument may be of
7803 any integer type, or a vectory with integer element type. The return
7804 type must match the first argument type.
7806 The second argument must be a constant and is a flag to indicate whether
7807 the intrinsic should ensure that a zero as the first argument produces a
7808 defined result. Historically some architectures did not provide a
7809 defined result for zero values as efficiently, and many algorithms are
7810 now predicated on avoiding zero-value inputs.
7815 The '``llvm.ctlz``' intrinsic counts the leading (most significant)
7816 zeros in a variable, or within each element of the vector. If
7817 ``src == 0`` then the result is the size in bits of the type of ``src``
7818 if ``is_zero_undef == 0`` and ``undef`` otherwise. For example,
7819 ``llvm.ctlz(i32 2) = 30``.
7821 '``llvm.cttz.*``' Intrinsic
7822 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
7827 This is an overloaded intrinsic. You can use ``llvm.cttz`` on any
7828 integer bit width, or any vector of integer elements. Not all targets
7829 support all bit widths or vector types, however.
7833 declare i8 @llvm.cttz.i8 (i8 <src>, i1 <is_zero_undef>)
7834 declare i16 @llvm.cttz.i16 (i16 <src>, i1 <is_zero_undef>)
7835 declare i32 @llvm.cttz.i32 (i32 <src>, i1 <is_zero_undef>)
7836 declare i64 @llvm.cttz.i64 (i64 <src>, i1 <is_zero_undef>)
7837 declare i256 @llvm.cttz.i256(i256 <src>, i1 <is_zero_undef>)
7838 declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
7843 The '``llvm.cttz``' family of intrinsic functions counts the number of
7849 The first argument is the value to be counted. This argument may be of
7850 any integer type, or a vectory with integer element type. The return
7851 type must match the first argument type.
7853 The second argument must be a constant and is a flag to indicate whether
7854 the intrinsic should ensure that a zero as the first argument produces a
7855 defined result. Historically some architectures did not provide a
7856 defined result for zero values as efficiently, and many algorithms are
7857 now predicated on avoiding zero-value inputs.
7862 The '``llvm.cttz``' intrinsic counts the trailing (least significant)
7863 zeros in a variable, or within each element of a vector. If ``src == 0``
7864 then the result is the size in bits of the type of ``src`` if
7865 ``is_zero_undef == 0`` and ``undef`` otherwise. For example,
7866 ``llvm.cttz(2) = 1``.
7868 Arithmetic with Overflow Intrinsics
7869 -----------------------------------
7871 LLVM provides intrinsics for some arithmetic with overflow operations.
7873 '``llvm.sadd.with.overflow.*``' Intrinsics
7874 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7879 This is an overloaded intrinsic. You can use ``llvm.sadd.with.overflow``
7880 on any integer bit width.
7884 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
7885 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7886 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
7891 The '``llvm.sadd.with.overflow``' family of intrinsic functions perform
7892 a signed addition of the two arguments, and indicate whether an overflow
7893 occurred during the signed summation.
7898 The arguments (%a and %b) and the first element of the result structure
7899 may be of integer types of any bit width, but they must have the same
7900 bit width. The second element of the result structure must be of type
7901 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
7907 The '``llvm.sadd.with.overflow``' family of intrinsic functions perform
7908 a signed addition of the two variables. They return a structure --- the
7909 first element of which is the signed summation, and the second element
7910 of which is a bit specifying if the signed summation resulted in an
7916 .. code-block:: llvm
7918 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7919 %sum = extractvalue {i32, i1} %res, 0
7920 %obit = extractvalue {i32, i1} %res, 1
7921 br i1 %obit, label %overflow, label %normal
7923 '``llvm.uadd.with.overflow.*``' Intrinsics
7924 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7929 This is an overloaded intrinsic. You can use ``llvm.uadd.with.overflow``
7930 on any integer bit width.
7934 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
7935 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7936 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
7941 The '``llvm.uadd.with.overflow``' family of intrinsic functions perform
7942 an unsigned addition of the two arguments, and indicate whether a carry
7943 occurred during the unsigned summation.
7948 The arguments (%a and %b) and the first element of the result structure
7949 may be of integer types of any bit width, but they must have the same
7950 bit width. The second element of the result structure must be of type
7951 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
7957 The '``llvm.uadd.with.overflow``' family of intrinsic functions perform
7958 an unsigned addition of the two arguments. They return a structure --- the
7959 first element of which is the sum, and the second element of which is a
7960 bit specifying if the unsigned summation resulted in a carry.
7965 .. code-block:: llvm
7967 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7968 %sum = extractvalue {i32, i1} %res, 0
7969 %obit = extractvalue {i32, i1} %res, 1
7970 br i1 %obit, label %carry, label %normal
7972 '``llvm.ssub.with.overflow.*``' Intrinsics
7973 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7978 This is an overloaded intrinsic. You can use ``llvm.ssub.with.overflow``
7979 on any integer bit width.
7983 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
7984 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7985 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
7990 The '``llvm.ssub.with.overflow``' family of intrinsic functions perform
7991 a signed subtraction of the two arguments, and indicate whether an
7992 overflow occurred during the signed subtraction.
7997 The arguments (%a and %b) and the first element of the result structure
7998 may be of integer types of any bit width, but they must have the same
7999 bit width. The second element of the result structure must be of type
8000 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
8006 The '``llvm.ssub.with.overflow``' family of intrinsic functions perform
8007 a signed subtraction of the two arguments. They return a structure --- the
8008 first element of which is the subtraction, and the second element of
8009 which is a bit specifying if the signed subtraction resulted in an
8015 .. code-block:: llvm
8017 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
8018 %sum = extractvalue {i32, i1} %res, 0
8019 %obit = extractvalue {i32, i1} %res, 1
8020 br i1 %obit, label %overflow, label %normal
8022 '``llvm.usub.with.overflow.*``' Intrinsics
8023 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8028 This is an overloaded intrinsic. You can use ``llvm.usub.with.overflow``
8029 on any integer bit width.
8033 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
8034 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
8035 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
8040 The '``llvm.usub.with.overflow``' family of intrinsic functions perform
8041 an unsigned subtraction of the two arguments, and indicate whether an
8042 overflow occurred during the unsigned subtraction.
8047 The arguments (%a and %b) and the first element of the result structure
8048 may be of integer types of any bit width, but they must have the same
8049 bit width. The second element of the result structure must be of type
8050 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
8056 The '``llvm.usub.with.overflow``' family of intrinsic functions perform
8057 an unsigned subtraction of the two arguments. They return a structure ---
8058 the first element of which is the subtraction, and the second element of
8059 which is a bit specifying if the unsigned subtraction resulted in an
8065 .. code-block:: llvm
8067 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
8068 %sum = extractvalue {i32, i1} %res, 0
8069 %obit = extractvalue {i32, i1} %res, 1
8070 br i1 %obit, label %overflow, label %normal
8072 '``llvm.smul.with.overflow.*``' Intrinsics
8073 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8078 This is an overloaded intrinsic. You can use ``llvm.smul.with.overflow``
8079 on any integer bit width.
8083 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
8084 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
8085 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
8090 The '``llvm.smul.with.overflow``' family of intrinsic functions perform
8091 a signed multiplication of the two arguments, and indicate whether an
8092 overflow occurred during the signed multiplication.
8097 The arguments (%a and %b) and the first element of the result structure
8098 may be of integer types of any bit width, but they must have the same
8099 bit width. The second element of the result structure must be of type
8100 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
8106 The '``llvm.smul.with.overflow``' family of intrinsic functions perform
8107 a signed multiplication of the two arguments. They return a structure ---
8108 the first element of which is the multiplication, and the second element
8109 of which is a bit specifying if the signed multiplication resulted in an
8115 .. code-block:: llvm
8117 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
8118 %sum = extractvalue {i32, i1} %res, 0
8119 %obit = extractvalue {i32, i1} %res, 1
8120 br i1 %obit, label %overflow, label %normal
8122 '``llvm.umul.with.overflow.*``' Intrinsics
8123 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8128 This is an overloaded intrinsic. You can use ``llvm.umul.with.overflow``
8129 on any integer bit width.
8133 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
8134 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
8135 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
8140 The '``llvm.umul.with.overflow``' family of intrinsic functions perform
8141 a unsigned multiplication of the two arguments, and indicate whether an
8142 overflow occurred during the unsigned multiplication.
8147 The arguments (%a and %b) and the first element of the result structure
8148 may be of integer types of any bit width, but they must have the same
8149 bit width. The second element of the result structure must be of type
8150 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
8156 The '``llvm.umul.with.overflow``' family of intrinsic functions perform
8157 an unsigned multiplication of the two arguments. They return a structure ---
8158 the first element of which is the multiplication, and the second
8159 element of which is a bit specifying if the unsigned multiplication
8160 resulted in an overflow.
8165 .. code-block:: llvm
8167 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
8168 %sum = extractvalue {i32, i1} %res, 0
8169 %obit = extractvalue {i32, i1} %res, 1
8170 br i1 %obit, label %overflow, label %normal
8172 Specialised Arithmetic Intrinsics
8173 ---------------------------------
8175 '``llvm.fmuladd.*``' Intrinsic
8176 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8183 declare float @llvm.fmuladd.f32(float %a, float %b, float %c)
8184 declare double @llvm.fmuladd.f64(double %a, double %b, double %c)
8189 The '``llvm.fmuladd.*``' intrinsic functions represent multiply-add
8190 expressions that can be fused if the code generator determines that (a) the
8191 target instruction set has support for a fused operation, and (b) that the
8192 fused operation is more efficient than the equivalent, separate pair of mul
8193 and add instructions.
8198 The '``llvm.fmuladd.*``' intrinsics each take three arguments: two
8199 multiplicands, a and b, and an addend c.
8208 %0 = call float @llvm.fmuladd.f32(%a, %b, %c)
8210 is equivalent to the expression a \* b + c, except that rounding will
8211 not be performed between the multiplication and addition steps if the
8212 code generator fuses the operations. Fusion is not guaranteed, even if
8213 the target platform supports it. If a fused multiply-add is required the
8214 corresponding llvm.fma.\* intrinsic function should be used instead.
8219 .. code-block:: llvm
8221 %r2 = call float @llvm.fmuladd.f32(float %a, float %b, float %c) ; yields {float}:r2 = (a * b) + c
8223 Half Precision Floating Point Intrinsics
8224 ----------------------------------------
8226 For most target platforms, half precision floating point is a
8227 storage-only format. This means that it is a dense encoding (in memory)
8228 but does not support computation in the format.
8230 This means that code must first load the half-precision floating point
8231 value as an i16, then convert it to float with
8232 :ref:`llvm.convert.from.fp16 <int_convert_from_fp16>`. Computation can
8233 then be performed on the float value (including extending to double
8234 etc). To store the value back to memory, it is first converted to float
8235 if needed, then converted to i16 with
8236 :ref:`llvm.convert.to.fp16 <int_convert_to_fp16>`, then storing as an
8239 .. _int_convert_to_fp16:
8241 '``llvm.convert.to.fp16``' Intrinsic
8242 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8249 declare i16 @llvm.convert.to.fp16(f32 %a)
8254 The '``llvm.convert.to.fp16``' intrinsic function performs a conversion
8255 from single precision floating point format to half precision floating
8261 The intrinsic function contains single argument - the value to be
8267 The '``llvm.convert.to.fp16``' intrinsic function performs a conversion
8268 from single precision floating point format to half precision floating
8269 point format. The return value is an ``i16`` which contains the
8275 .. code-block:: llvm
8277 %res = call i16 @llvm.convert.to.fp16(f32 %a)
8278 store i16 %res, i16* @x, align 2
8280 .. _int_convert_from_fp16:
8282 '``llvm.convert.from.fp16``' Intrinsic
8283 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8290 declare f32 @llvm.convert.from.fp16(i16 %a)
8295 The '``llvm.convert.from.fp16``' intrinsic function performs a
8296 conversion from half precision floating point format to single precision
8297 floating point format.
8302 The intrinsic function contains single argument - the value to be
8308 The '``llvm.convert.from.fp16``' intrinsic function performs a
8309 conversion from half single precision floating point format to single
8310 precision floating point format. The input half-float value is
8311 represented by an ``i16`` value.
8316 .. code-block:: llvm
8318 %a = load i16* @x, align 2
8319 %res = call f32 @llvm.convert.from.fp16(i16 %a)
8324 The LLVM debugger intrinsics (which all start with ``llvm.dbg.``
8325 prefix), are described in the `LLVM Source Level
8326 Debugging <SourceLevelDebugging.html#format_common_intrinsics>`_
8329 Exception Handling Intrinsics
8330 -----------------------------
8332 The LLVM exception handling intrinsics (which all start with
8333 ``llvm.eh.`` prefix), are described in the `LLVM Exception
8334 Handling <ExceptionHandling.html#format_common_intrinsics>`_ document.
8338 Trampoline Intrinsics
8339 ---------------------
8341 These intrinsics make it possible to excise one parameter, marked with
8342 the :ref:`nest <nest>` attribute, from a function. The result is a
8343 callable function pointer lacking the nest parameter - the caller does
8344 not need to provide a value for it. Instead, the value to use is stored
8345 in advance in a "trampoline", a block of memory usually allocated on the
8346 stack, which also contains code to splice the nest value into the
8347 argument list. This is used to implement the GCC nested function address
8350 For example, if the function is ``i32 f(i8* nest %c, i32 %x, i32 %y)``
8351 then the resulting function pointer has signature ``i32 (i32, i32)*``.
8352 It can be created as follows:
8354 .. code-block:: llvm
8356 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
8357 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
8358 call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval)
8359 %p = call i8* @llvm.adjust.trampoline(i8* %tramp1)
8360 %fp = bitcast i8* %p to i32 (i32, i32)*
8362 The call ``%val = call i32 %fp(i32 %x, i32 %y)`` is then equivalent to
8363 ``%val = call i32 %f(i8* %nval, i32 %x, i32 %y)``.
8367 '``llvm.init.trampoline``' Intrinsic
8368 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8375 declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
8380 This fills the memory pointed to by ``tramp`` with executable code,
8381 turning it into a trampoline.
8386 The ``llvm.init.trampoline`` intrinsic takes three arguments, all
8387 pointers. The ``tramp`` argument must point to a sufficiently large and
8388 sufficiently aligned block of memory; this memory is written to by the
8389 intrinsic. Note that the size and the alignment are target-specific -
8390 LLVM currently provides no portable way of determining them, so a
8391 front-end that generates this intrinsic needs to have some
8392 target-specific knowledge. The ``func`` argument must hold a function
8393 bitcast to an ``i8*``.
8398 The block of memory pointed to by ``tramp`` is filled with target
8399 dependent code, turning it into a function. Then ``tramp`` needs to be
8400 passed to :ref:`llvm.adjust.trampoline <int_at>` to get a pointer which can
8401 be :ref:`bitcast (to a new function) and called <int_trampoline>`. The new
8402 function's signature is the same as that of ``func`` with any arguments
8403 marked with the ``nest`` attribute removed. At most one such ``nest``
8404 argument is allowed, and it must be of pointer type. Calling the new
8405 function is equivalent to calling ``func`` with the same argument list,
8406 but with ``nval`` used for the missing ``nest`` argument. If, after
8407 calling ``llvm.init.trampoline``, the memory pointed to by ``tramp`` is
8408 modified, then the effect of any later call to the returned function
8409 pointer is undefined.
8413 '``llvm.adjust.trampoline``' Intrinsic
8414 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8421 declare i8* @llvm.adjust.trampoline(i8* <tramp>)
8426 This performs any required machine-specific adjustment to the address of
8427 a trampoline (passed as ``tramp``).
8432 ``tramp`` must point to a block of memory which already has trampoline
8433 code filled in by a previous call to
8434 :ref:`llvm.init.trampoline <int_it>`.
8439 On some architectures the address of the code to be executed needs to be
8440 different to the address where the trampoline is actually stored. This
8441 intrinsic returns the executable address corresponding to ``tramp``
8442 after performing the required machine specific adjustments. The pointer
8443 returned can then be :ref:`bitcast and executed <int_trampoline>`.
8448 This class of intrinsics exists to information about the lifetime of
8449 memory objects and ranges where variables are immutable.
8451 '``llvm.lifetime.start``' Intrinsic
8452 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8459 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
8464 The '``llvm.lifetime.start``' intrinsic specifies the start of a memory
8470 The first argument is a constant integer representing the size of the
8471 object, or -1 if it is variable sized. The second argument is a pointer
8477 This intrinsic indicates that before this point in the code, the value
8478 of the memory pointed to by ``ptr`` is dead. This means that it is known
8479 to never be used and has an undefined value. A load from the pointer
8480 that precedes this intrinsic can be replaced with ``'undef'``.
8482 '``llvm.lifetime.end``' Intrinsic
8483 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8490 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
8495 The '``llvm.lifetime.end``' intrinsic specifies the end of a memory
8501 The first argument is a constant integer representing the size of the
8502 object, or -1 if it is variable sized. The second argument is a pointer
8508 This intrinsic indicates that after this point in the code, the value of
8509 the memory pointed to by ``ptr`` is dead. This means that it is known to
8510 never be used and has an undefined value. Any stores into the memory
8511 object following this intrinsic may be removed as dead.
8513 '``llvm.invariant.start``' Intrinsic
8514 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8521 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
8526 The '``llvm.invariant.start``' intrinsic specifies that the contents of
8527 a memory object will not change.
8532 The first argument is a constant integer representing the size of the
8533 object, or -1 if it is variable sized. The second argument is a pointer
8539 This intrinsic indicates that until an ``llvm.invariant.end`` that uses
8540 the return value, the referenced memory location is constant and
8543 '``llvm.invariant.end``' Intrinsic
8544 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8551 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
8556 The '``llvm.invariant.end``' intrinsic specifies that the contents of a
8557 memory object are mutable.
8562 The first argument is the matching ``llvm.invariant.start`` intrinsic.
8563 The second argument is a constant integer representing the size of the
8564 object, or -1 if it is variable sized and the third argument is a
8565 pointer to the object.
8570 This intrinsic indicates that the memory is mutable again.
8575 This class of intrinsics is designed to be generic and has no specific
8578 '``llvm.var.annotation``' Intrinsic
8579 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8586 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
8591 The '``llvm.var.annotation``' intrinsic.
8596 The first argument is a pointer to a value, the second is a pointer to a
8597 global string, the third is a pointer to a global string which is the
8598 source file name, and the last argument is the line number.
8603 This intrinsic allows annotation of local variables with arbitrary
8604 strings. This can be useful for special purpose optimizations that want
8605 to look for these annotations. These have no other defined use; they are
8606 ignored by code generation and optimization.
8608 '``llvm.ptr.annotation.*``' Intrinsic
8609 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8614 This is an overloaded intrinsic. You can use '``llvm.ptr.annotation``' on a
8615 pointer to an integer of any width. *NOTE* you must specify an address space for
8616 the pointer. The identifier for the default address space is the integer
8621 declare i8* @llvm.ptr.annotation.p<address space>i8(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
8622 declare i16* @llvm.ptr.annotation.p<address space>i16(i16* <val>, i8* <str>, i8* <str>, i32 <int>)
8623 declare i32* @llvm.ptr.annotation.p<address space>i32(i32* <val>, i8* <str>, i8* <str>, i32 <int>)
8624 declare i64* @llvm.ptr.annotation.p<address space>i64(i64* <val>, i8* <str>, i8* <str>, i32 <int>)
8625 declare i256* @llvm.ptr.annotation.p<address space>i256(i256* <val>, i8* <str>, i8* <str>, i32 <int>)
8630 The '``llvm.ptr.annotation``' intrinsic.
8635 The first argument is a pointer to an integer value of arbitrary bitwidth
8636 (result of some expression), the second is a pointer to a global string, the
8637 third is a pointer to a global string which is the source file name, and the
8638 last argument is the line number. It returns the value of the first argument.
8643 This intrinsic allows annotation of a pointer to an integer with arbitrary
8644 strings. This can be useful for special purpose optimizations that want to look
8645 for these annotations. These have no other defined use; they are ignored by code
8646 generation and optimization.
8648 '``llvm.annotation.*``' Intrinsic
8649 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8654 This is an overloaded intrinsic. You can use '``llvm.annotation``' on
8655 any integer bit width.
8659 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
8660 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
8661 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
8662 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
8663 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
8668 The '``llvm.annotation``' intrinsic.
8673 The first argument is an integer value (result of some expression), the
8674 second is a pointer to a global string, the third is a pointer to a
8675 global string which is the source file name, and the last argument is
8676 the line number. It returns the value of the first argument.
8681 This intrinsic allows annotations to be put on arbitrary expressions
8682 with arbitrary strings. This can be useful for special purpose
8683 optimizations that want to look for these annotations. These have no
8684 other defined use; they are ignored by code generation and optimization.
8686 '``llvm.trap``' Intrinsic
8687 ^^^^^^^^^^^^^^^^^^^^^^^^^
8694 declare void @llvm.trap() noreturn nounwind
8699 The '``llvm.trap``' intrinsic.
8709 This intrinsic is lowered to the target dependent trap instruction. If
8710 the target does not have a trap instruction, this intrinsic will be
8711 lowered to a call of the ``abort()`` function.
8713 '``llvm.debugtrap``' Intrinsic
8714 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8721 declare void @llvm.debugtrap() nounwind
8726 The '``llvm.debugtrap``' intrinsic.
8736 This intrinsic is lowered to code which is intended to cause an
8737 execution trap with the intention of requesting the attention of a
8740 '``llvm.stackprotector``' Intrinsic
8741 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8748 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
8753 The ``llvm.stackprotector`` intrinsic takes the ``guard`` and stores it
8754 onto the stack at ``slot``. The stack slot is adjusted to ensure that it
8755 is placed on the stack before local variables.
8760 The ``llvm.stackprotector`` intrinsic requires two pointer arguments.
8761 The first argument is the value loaded from the stack guard
8762 ``@__stack_chk_guard``. The second variable is an ``alloca`` that has
8763 enough space to hold the value of the guard.
8768 This intrinsic causes the prologue/epilogue inserter to force the position of
8769 the ``AllocaInst`` stack slot to be before local variables on the stack. This is
8770 to ensure that if a local variable on the stack is overwritten, it will destroy
8771 the value of the guard. When the function exits, the guard on the stack is
8772 checked against the original guard by ``llvm.stackprotectorcheck``. If they are
8773 different, then ``llvm.stackprotectorcheck`` causes the program to abort by
8774 calling the ``__stack_chk_fail()`` function.
8776 '``llvm.stackprotectorcheck``' Intrinsic
8777 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8784 declare void @llvm.stackprotectorcheck(i8** <guard>)
8789 The ``llvm.stackprotectorcheck`` intrinsic compares ``guard`` against an already
8790 created stack protector and if they are not equal calls the
8791 ``__stack_chk_fail()`` function.
8796 The ``llvm.stackprotectorcheck`` intrinsic requires one pointer argument, the
8797 the variable ``@__stack_chk_guard``.
8802 This intrinsic is provided to perform the stack protector check by comparing
8803 ``guard`` with the stack slot created by ``llvm.stackprotector`` and if the
8804 values do not match call the ``__stack_chk_fail()`` function.
8806 The reason to provide this as an IR level intrinsic instead of implementing it
8807 via other IR operations is that in order to perform this operation at the IR
8808 level without an intrinsic, one would need to create additional basic blocks to
8809 handle the success/failure cases. This makes it difficult to stop the stack
8810 protector check from disrupting sibling tail calls in Codegen. With this
8811 intrinsic, we are able to generate the stack protector basic blocks late in
8812 codegen after the tail call decision has occured.
8814 '``llvm.objectsize``' Intrinsic
8815 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8822 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <min>)
8823 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <min>)
8828 The ``llvm.objectsize`` intrinsic is designed to provide information to
8829 the optimizers to determine at compile time whether a) an operation
8830 (like memcpy) will overflow a buffer that corresponds to an object, or
8831 b) that a runtime check for overflow isn't necessary. An object in this
8832 context means an allocation of a specific class, structure, array, or
8838 The ``llvm.objectsize`` intrinsic takes two arguments. The first
8839 argument is a pointer to or into the ``object``. The second argument is
8840 a boolean and determines whether ``llvm.objectsize`` returns 0 (if true)
8841 or -1 (if false) when the object size is unknown. The second argument
8842 only accepts constants.
8847 The ``llvm.objectsize`` intrinsic is lowered to a constant representing
8848 the size of the object concerned. If the size cannot be determined at
8849 compile time, ``llvm.objectsize`` returns ``i32/i64 -1 or 0`` (depending
8850 on the ``min`` argument).
8852 '``llvm.expect``' Intrinsic
8853 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8860 declare i32 @llvm.expect.i32(i32 <val>, i32 <expected_val>)
8861 declare i64 @llvm.expect.i64(i64 <val>, i64 <expected_val>)
8866 The ``llvm.expect`` intrinsic provides information about expected (the
8867 most probable) value of ``val``, which can be used by optimizers.
8872 The ``llvm.expect`` intrinsic takes two arguments. The first argument is
8873 a value. The second argument is an expected value, this needs to be a
8874 constant value, variables are not allowed.
8879 This intrinsic is lowered to the ``val``.
8881 '``llvm.donothing``' Intrinsic
8882 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8889 declare void @llvm.donothing() nounwind readnone
8894 The ``llvm.donothing`` intrinsic doesn't perform any operation. It's the
8895 only intrinsic that can be called with an invoke instruction.
8905 This intrinsic does nothing, and it's removed by optimizers and ignored