1 ==============================
2 LLVM Language Reference Manual
3 ==============================
12 This document is a reference manual for the LLVM assembly language. LLVM
13 is a Static Single Assignment (SSA) based representation that provides
14 type safety, low-level operations, flexibility, and the capability of
15 representing 'all' high-level languages cleanly. It is the common code
16 representation used throughout all phases of the LLVM compilation
22 The LLVM code representation is designed to be used in three different
23 forms: as an in-memory compiler IR, as an on-disk bitcode representation
24 (suitable for fast loading by a Just-In-Time compiler), and as a human
25 readable assembly language representation. This allows LLVM to provide a
26 powerful intermediate representation for efficient compiler
27 transformations and analysis, while providing a natural means to debug
28 and visualize the transformations. The three different forms of LLVM are
29 all equivalent. This document describes the human readable
30 representation and notation.
32 The LLVM representation aims to be light-weight and low-level while
33 being expressive, typed, and extensible at the same time. It aims to be
34 a "universal IR" of sorts, by being at a low enough level that
35 high-level ideas may be cleanly mapped to it (similar to how
36 microprocessors are "universal IR's", allowing many source languages to
37 be mapped to them). By providing type information, LLVM can be used as
38 the target of optimizations: for example, through pointer analysis, it
39 can be proven that a C automatic variable is never accessed outside of
40 the current function, allowing it to be promoted to a simple SSA value
41 instead of a memory location.
48 It is important to note that this document describes 'well formed' LLVM
49 assembly language. There is a difference between what the parser accepts
50 and what is considered 'well formed'. For example, the following
51 instruction is syntactically okay, but not well formed:
57 because the definition of ``%x`` does not dominate all of its uses. The
58 LLVM infrastructure provides a verification pass that may be used to
59 verify that an LLVM module is well formed. This pass is automatically
60 run by the parser after parsing input assembly and by the optimizer
61 before it outputs bitcode. The violations pointed out by the verifier
62 pass indicate bugs in transformation passes or input to the parser.
69 LLVM identifiers come in two basic types: global and local. Global
70 identifiers (functions, global variables) begin with the ``'@'``
71 character. Local identifiers (register names, types) begin with the
72 ``'%'`` character. Additionally, there are three different formats for
73 identifiers, for different purposes:
75 #. Named values are represented as a string of characters with their
76 prefix. For example, ``%foo``, ``@DivisionByZero``,
77 ``%a.really.long.identifier``. The actual regular expression used is
78 '``[%@][-a-zA-Z$._][-a-zA-Z$._0-9]*``'. Identifiers that require other
79 characters in their names can be surrounded with quotes. Special
80 characters may be escaped using ``"\xx"`` where ``xx`` is the ASCII
81 code for the character in hexadecimal. In this way, any character can
82 be used in a name value, even quotes themselves. The ``"\01"`` prefix
83 can be used on global variables to suppress mangling.
84 #. Unnamed values are represented as an unsigned numeric value with
85 their prefix. For example, ``%12``, ``@2``, ``%44``.
86 #. Constants, which are described in the section Constants_ below.
88 LLVM requires that values start with a prefix for two reasons: Compilers
89 don't need to worry about name clashes with reserved words, and the set
90 of reserved words may be expanded in the future without penalty.
91 Additionally, unnamed identifiers allow a compiler to quickly come up
92 with a temporary variable without having to avoid symbol table
95 Reserved words in LLVM are very similar to reserved words in other
96 languages. There are keywords for different opcodes ('``add``',
97 '``bitcast``', '``ret``', etc...), for primitive type names ('``void``',
98 '``i32``', etc...), and others. These reserved words cannot conflict
99 with variable names, because none of them start with a prefix character
100 (``'%'`` or ``'@'``).
102 Here is an example of LLVM code to multiply the integer variable
109 %result = mul i32 %X, 8
111 After strength reduction:
115 %result = shl i32 %X, 3
121 %0 = add i32 %X, %X ; yields i32:%0
122 %1 = add i32 %0, %0 ; yields i32:%1
123 %result = add i32 %1, %1
125 This last way of multiplying ``%X`` by 8 illustrates several important
126 lexical features of LLVM:
128 #. Comments are delimited with a '``;``' and go until the end of line.
129 #. Unnamed temporaries are created when the result of a computation is
130 not assigned to a named value.
131 #. Unnamed temporaries are numbered sequentially (using a per-function
132 incrementing counter, starting with 0). Note that basic blocks and unnamed
133 function parameters are included in this numbering. For example, if the
134 entry basic block is not given a label name and all function parameters are
135 named, then it will get number 0.
137 It also shows a convention that we follow in this document. When
138 demonstrating instructions, we will follow an instruction with a comment
139 that defines the type and name of value produced.
147 LLVM programs are composed of ``Module``'s, each of which is a
148 translation unit of the input programs. Each module consists of
149 functions, global variables, and symbol table entries. Modules may be
150 combined together with the LLVM linker, which merges function (and
151 global variable) definitions, resolves forward declarations, and merges
152 symbol table entries. Here is an example of the "hello world" module:
156 ; Declare the string constant as a global constant.
157 @.str = private unnamed_addr constant [13 x i8] c"hello world\0A\00"
159 ; External declaration of the puts function
160 declare i32 @puts(i8* nocapture) nounwind
162 ; Definition of main function
163 define i32 @main() { ; i32()*
164 ; Convert [13 x i8]* to i8 *...
165 %cast210 = getelementptr [13 x i8], [13 x i8]* @.str, i64 0, i64 0
167 ; Call puts function to write out the string to stdout.
168 call i32 @puts(i8* %cast210)
173 !0 = !{i32 42, null, !"string"}
176 This example is made up of a :ref:`global variable <globalvars>` named
177 "``.str``", an external declaration of the "``puts``" function, a
178 :ref:`function definition <functionstructure>` for "``main``" and
179 :ref:`named metadata <namedmetadatastructure>` "``foo``".
181 In general, a module is made up of a list of global values (where both
182 functions and global variables are global values). Global values are
183 represented by a pointer to a memory location (in this case, a pointer
184 to an array of char, and a pointer to a function), and have one of the
185 following :ref:`linkage types <linkage>`.
192 All Global Variables and Functions have one of the following types of
196 Global values with "``private``" linkage are only directly
197 accessible by objects in the current module. In particular, linking
198 code into a module with an private global value may cause the
199 private to be renamed as necessary to avoid collisions. Because the
200 symbol is private to the module, all references can be updated. This
201 doesn't show up in any symbol table in the object file.
203 Similar to private, but the value shows as a local symbol
204 (``STB_LOCAL`` in the case of ELF) in the object file. This
205 corresponds to the notion of the '``static``' keyword in C.
206 ``available_externally``
207 Globals with "``available_externally``" linkage are never emitted
208 into the object file corresponding to the LLVM module. They exist to
209 allow inlining and other optimizations to take place given knowledge
210 of the definition of the global, which is known to be somewhere
211 outside the module. Globals with ``available_externally`` linkage
212 are allowed to be discarded at will, and are otherwise the same as
213 ``linkonce_odr``. This linkage type is only allowed on definitions,
216 Globals with "``linkonce``" linkage are merged with other globals of
217 the same name when linkage occurs. This can be used to implement
218 some forms of inline functions, templates, or other code which must
219 be generated in each translation unit that uses it, but where the
220 body may be overridden with a more definitive definition later.
221 Unreferenced ``linkonce`` globals are allowed to be discarded. Note
222 that ``linkonce`` linkage does not actually allow the optimizer to
223 inline the body of this function into callers because it doesn't
224 know if this definition of the function is the definitive definition
225 within the program or whether it will be overridden by a stronger
226 definition. To enable inlining and other optimizations, use
227 "``linkonce_odr``" linkage.
229 "``weak``" linkage has the same merging semantics as ``linkonce``
230 linkage, except that unreferenced globals with ``weak`` linkage may
231 not be discarded. This is used for globals that are declared "weak"
234 "``common``" linkage is most similar to "``weak``" linkage, but they
235 are used for tentative definitions in C, such as "``int X;``" at
236 global scope. Symbols with "``common``" linkage are merged in the
237 same way as ``weak symbols``, and they may not be deleted if
238 unreferenced. ``common`` symbols may not have an explicit section,
239 must have a zero initializer, and may not be marked
240 ':ref:`constant <globalvars>`'. Functions and aliases may not have
243 .. _linkage_appending:
246 "``appending``" linkage may only be applied to global variables of
247 pointer to array type. When two global variables with appending
248 linkage are linked together, the two global arrays are appended
249 together. This is the LLVM, typesafe, equivalent of having the
250 system linker append together "sections" with identical names when
253 The semantics of this linkage follow the ELF object file model: the
254 symbol is weak until linked, if not linked, the symbol becomes null
255 instead of being an undefined reference.
256 ``linkonce_odr``, ``weak_odr``
257 Some languages allow differing globals to be merged, such as two
258 functions with different semantics. Other languages, such as
259 ``C++``, ensure that only equivalent globals are ever merged (the
260 "one definition rule" --- "ODR"). Such languages can use the
261 ``linkonce_odr`` and ``weak_odr`` linkage types to indicate that the
262 global will only be merged with equivalent globals. These linkage
263 types are otherwise the same as their non-``odr`` versions.
265 If none of the above identifiers are used, the global is externally
266 visible, meaning that it participates in linkage and can be used to
267 resolve external symbol references.
269 It is illegal for a function *declaration* to have any linkage type
270 other than ``external`` or ``extern_weak``.
277 LLVM :ref:`functions <functionstructure>`, :ref:`calls <i_call>` and
278 :ref:`invokes <i_invoke>` can all have an optional calling convention
279 specified for the call. The calling convention of any pair of dynamic
280 caller/callee must match, or the behavior of the program is undefined.
281 The following calling conventions are supported by LLVM, and more may be
284 "``ccc``" - The C calling convention
285 This calling convention (the default if no other calling convention
286 is specified) matches the target C calling conventions. This calling
287 convention supports varargs function calls and tolerates some
288 mismatch in the declared prototype and implemented declaration of
289 the function (as does normal C).
290 "``fastcc``" - The fast calling convention
291 This calling convention attempts to make calls as fast as possible
292 (e.g. by passing things in registers). This calling convention
293 allows the target to use whatever tricks it wants to produce fast
294 code for the target, without having to conform to an externally
295 specified ABI (Application Binary Interface). `Tail calls can only
296 be optimized when this, the GHC or the HiPE convention is
297 used. <CodeGenerator.html#id80>`_ This calling convention does not
298 support varargs and requires the prototype of all callees to exactly
299 match the prototype of the function definition.
300 "``coldcc``" - The cold calling convention
301 This calling convention attempts to make code in the caller as
302 efficient as possible under the assumption that the call is not
303 commonly executed. As such, these calls often preserve all registers
304 so that the call does not break any live ranges in the caller side.
305 This calling convention does not support varargs and requires the
306 prototype of all callees to exactly match the prototype of the
307 function definition. Furthermore the inliner doesn't consider such function
309 "``cc 10``" - GHC convention
310 This calling convention has been implemented specifically for use by
311 the `Glasgow Haskell Compiler (GHC) <http://www.haskell.org/ghc>`_.
312 It passes everything in registers, going to extremes to achieve this
313 by disabling callee save registers. This calling convention should
314 not be used lightly but only for specific situations such as an
315 alternative to the *register pinning* performance technique often
316 used when implementing functional programming languages. At the
317 moment only X86 supports this convention and it has the following
320 - On *X86-32* only supports up to 4 bit type parameters. No
321 floating point types are supported.
322 - On *X86-64* only supports up to 10 bit type parameters and 6
323 floating point parameters.
325 This calling convention supports `tail call
326 optimization <CodeGenerator.html#id80>`_ but requires both the
327 caller and callee are using it.
328 "``cc 11``" - The HiPE calling convention
329 This calling convention has been implemented specifically for use by
330 the `High-Performance Erlang
331 (HiPE) <http://www.it.uu.se/research/group/hipe/>`_ compiler, *the*
332 native code compiler of the `Ericsson's Open Source Erlang/OTP
333 system <http://www.erlang.org/download.shtml>`_. It uses more
334 registers for argument passing than the ordinary C calling
335 convention and defines no callee-saved registers. The calling
336 convention properly supports `tail call
337 optimization <CodeGenerator.html#id80>`_ but requires that both the
338 caller and the callee use it. It uses a *register pinning*
339 mechanism, similar to GHC's convention, for keeping frequently
340 accessed runtime components pinned to specific hardware registers.
341 At the moment only X86 supports this convention (both 32 and 64
343 "``webkit_jscc``" - WebKit's JavaScript calling convention
344 This calling convention has been implemented for `WebKit FTL JIT
345 <https://trac.webkit.org/wiki/FTLJIT>`_. It passes arguments on the
346 stack right to left (as cdecl does), and returns a value in the
347 platform's customary return register.
348 "``anyregcc``" - Dynamic calling convention for code patching
349 This is a special convention that supports patching an arbitrary code
350 sequence in place of a call site. This convention forces the call
351 arguments into registers but allows them to be dynamically
352 allocated. This can currently only be used with calls to
353 llvm.experimental.patchpoint because only this intrinsic records
354 the location of its arguments in a side table. See :doc:`StackMaps`.
355 "``preserve_mostcc``" - The `PreserveMost` calling convention
356 This calling convention attempts to make the code in the caller as
357 unintrusive as possible. This convention behaves identically to the `C`
358 calling convention on how arguments and return values are passed, but it
359 uses a different set of caller/callee-saved registers. This alleviates the
360 burden of saving and recovering a large register set before and after the
361 call in the caller. If the arguments are passed in callee-saved registers,
362 then they will be preserved by the callee across the call. This doesn't
363 apply for values returned in callee-saved registers.
365 - On X86-64 the callee preserves all general purpose registers, except for
366 R11. R11 can be used as a scratch register. Floating-point registers
367 (XMMs/YMMs) are not preserved and need to be saved by the caller.
369 The idea behind this convention is to support calls to runtime functions
370 that have a hot path and a cold path. The hot path is usually a small piece
371 of code that doesn't use many registers. The cold path might need to call out to
372 another function and therefore only needs to preserve the caller-saved
373 registers, which haven't already been saved by the caller. The
374 `PreserveMost` calling convention is very similar to the `cold` calling
375 convention in terms of caller/callee-saved registers, but they are used for
376 different types of function calls. `coldcc` is for function calls that are
377 rarely executed, whereas `preserve_mostcc` function calls are intended to be
378 on the hot path and definitely executed a lot. Furthermore `preserve_mostcc`
379 doesn't prevent the inliner from inlining the function call.
381 This calling convention will be used by a future version of the ObjectiveC
382 runtime and should therefore still be considered experimental at this time.
383 Although this convention was created to optimize certain runtime calls to
384 the ObjectiveC runtime, it is not limited to this runtime and might be used
385 by other runtimes in the future too. The current implementation only
386 supports X86-64, but the intention is to support more architectures in the
388 "``preserve_allcc``" - The `PreserveAll` calling convention
389 This calling convention attempts to make the code in the caller even less
390 intrusive than the `PreserveMost` calling convention. This calling
391 convention also behaves identical to the `C` calling convention on how
392 arguments and return values are passed, but it uses a different set of
393 caller/callee-saved registers. This removes the burden of saving and
394 recovering a large register set before and after the call in the caller. If
395 the arguments are passed in callee-saved registers, then they will be
396 preserved by the callee across the call. This doesn't apply for values
397 returned in callee-saved registers.
399 - On X86-64 the callee preserves all general purpose registers, except for
400 R11. R11 can be used as a scratch register. Furthermore it also preserves
401 all floating-point registers (XMMs/YMMs).
403 The idea behind this convention is to support calls to runtime functions
404 that don't need to call out to any other functions.
406 This calling convention, like the `PreserveMost` calling convention, will be
407 used by a future version of the ObjectiveC runtime and should be considered
408 experimental at this time.
409 "``cc <n>``" - Numbered convention
410 Any calling convention may be specified by number, allowing
411 target-specific calling conventions to be used. Target specific
412 calling conventions start at 64.
414 More calling conventions can be added/defined on an as-needed basis, to
415 support Pascal conventions or any other well-known target-independent
418 .. _visibilitystyles:
423 All Global Variables and Functions have one of the following visibility
426 "``default``" - Default style
427 On targets that use the ELF object file format, default visibility
428 means that the declaration is visible to other modules and, in
429 shared libraries, means that the declared entity may be overridden.
430 On Darwin, default visibility means that the declaration is visible
431 to other modules. Default visibility corresponds to "external
432 linkage" in the language.
433 "``hidden``" - Hidden style
434 Two declarations of an object with hidden visibility refer to the
435 same object if they are in the same shared object. Usually, hidden
436 visibility indicates that the symbol will not be placed into the
437 dynamic symbol table, so no other module (executable or shared
438 library) can reference it directly.
439 "``protected``" - Protected style
440 On ELF, protected visibility indicates that the symbol will be
441 placed in the dynamic symbol table, but that references within the
442 defining module will bind to the local symbol. That is, the symbol
443 cannot be overridden by another module.
445 A symbol with ``internal`` or ``private`` linkage must have ``default``
453 All Global Variables, Functions and Aliases can have one of the following
457 "``dllimport``" causes the compiler to reference a function or variable via
458 a global pointer to a pointer that is set up by the DLL exporting the
459 symbol. On Microsoft Windows targets, the pointer name is formed by
460 combining ``__imp_`` and the function or variable name.
462 "``dllexport``" causes the compiler to provide a global pointer to a pointer
463 in a DLL, so that it can be referenced with the ``dllimport`` attribute. On
464 Microsoft Windows targets, the pointer name is formed by combining
465 ``__imp_`` and the function or variable name. Since this storage class
466 exists for defining a dll interface, the compiler, assembler and linker know
467 it is externally referenced and must refrain from deleting the symbol.
471 Thread Local Storage Models
472 ---------------------------
474 A variable may be defined as ``thread_local``, which means that it will
475 not be shared by threads (each thread will have a separated copy of the
476 variable). Not all targets support thread-local variables. Optionally, a
477 TLS model may be specified:
480 For variables that are only used within the current shared library.
482 For variables in modules that will not be loaded dynamically.
484 For variables defined in the executable and only used within it.
486 If no explicit model is given, the "general dynamic" model is used.
488 The models correspond to the ELF TLS models; see `ELF Handling For
489 Thread-Local Storage <http://people.redhat.com/drepper/tls.pdf>`_ for
490 more information on under which circumstances the different models may
491 be used. The target may choose a different TLS model if the specified
492 model is not supported, or if a better choice of model can be made.
494 A model can also be specified in a alias, but then it only governs how
495 the alias is accessed. It will not have any effect in the aliasee.
502 LLVM IR allows you to specify both "identified" and "literal" :ref:`structure
503 types <t_struct>`. Literal types are uniqued structurally, but identified types
504 are never uniqued. An :ref:`opaque structural type <t_opaque>` can also be used
505 to forward declare a type that is not yet available.
507 An example of a identified structure specification is:
511 %mytype = type { %mytype*, i32 }
513 Prior to the LLVM 3.0 release, identified types were structurally uniqued. Only
514 literal types are uniqued in recent versions of LLVM.
521 Global variables define regions of memory allocated at compilation time
524 Global variable definitions must be initialized.
526 Global variables in other translation units can also be declared, in which
527 case they don't have an initializer.
529 Either global variable definitions or declarations may have an explicit section
530 to be placed in and may have an optional explicit alignment specified.
532 A variable may be defined as a global ``constant``, which indicates that
533 the contents of the variable will **never** be modified (enabling better
534 optimization, allowing the global data to be placed in the read-only
535 section of an executable, etc). Note that variables that need runtime
536 initialization cannot be marked ``constant`` as there is a store to the
539 LLVM explicitly allows *declarations* of global variables to be marked
540 constant, even if the final definition of the global is not. This
541 capability can be used to enable slightly better optimization of the
542 program, but requires the language definition to guarantee that
543 optimizations based on the 'constantness' are valid for the translation
544 units that do not include the definition.
546 As SSA values, global variables define pointer values that are in scope
547 (i.e. they dominate) all basic blocks in the program. Global variables
548 always define a pointer to their "content" type because they describe a
549 region of memory, and all memory objects in LLVM are accessed through
552 Global variables can be marked with ``unnamed_addr`` which indicates
553 that the address is not significant, only the content. Constants marked
554 like this can be merged with other constants if they have the same
555 initializer. Note that a constant with significant address *can* be
556 merged with a ``unnamed_addr`` constant, the result being a constant
557 whose address is significant.
559 A global variable may be declared to reside in a target-specific
560 numbered address space. For targets that support them, address spaces
561 may affect how optimizations are performed and/or what target
562 instructions are used to access the variable. The default address space
563 is zero. The address space qualifier must precede any other attributes.
565 LLVM allows an explicit section to be specified for globals. If the
566 target supports it, it will emit globals to the section specified.
567 Additionally, the global can placed in a comdat if the target has the necessary
570 By default, global initializers are optimized by assuming that global
571 variables defined within the module are not modified from their
572 initial values before the start of the global initializer. This is
573 true even for variables potentially accessible from outside the
574 module, including those with external linkage or appearing in
575 ``@llvm.used`` or dllexported variables. This assumption may be suppressed
576 by marking the variable with ``externally_initialized``.
578 An explicit alignment may be specified for a global, which must be a
579 power of 2. If not present, or if the alignment is set to zero, the
580 alignment of the global is set by the target to whatever it feels
581 convenient. If an explicit alignment is specified, the global is forced
582 to have exactly that alignment. Targets and optimizers are not allowed
583 to over-align the global if the global has an assigned section. In this
584 case, the extra alignment could be observable: for example, code could
585 assume that the globals are densely packed in their section and try to
586 iterate over them as an array, alignment padding would break this
587 iteration. The maximum alignment is ``1 << 29``.
589 Globals can also have a :ref:`DLL storage class <dllstorageclass>`.
591 Variables and aliases can have a
592 :ref:`Thread Local Storage Model <tls_model>`.
596 [@<GlobalVarName> =] [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal]
597 [unnamed_addr] [AddrSpace] [ExternallyInitialized]
598 <global | constant> <Type> [<InitializerConstant>]
599 [, section "name"] [, comdat [($name)]]
600 [, align <Alignment>]
602 For example, the following defines a global in a numbered address space
603 with an initializer, section, and alignment:
607 @G = addrspace(5) constant float 1.0, section "foo", align 4
609 The following example just declares a global variable
613 @G = external global i32
615 The following example defines a thread-local global with the
616 ``initialexec`` TLS model:
620 @G = thread_local(initialexec) global i32 0, align 4
622 .. _functionstructure:
627 LLVM function definitions consist of the "``define``" keyword, an
628 optional :ref:`linkage type <linkage>`, an optional :ref:`visibility
629 style <visibility>`, an optional :ref:`DLL storage class <dllstorageclass>`,
630 an optional :ref:`calling convention <callingconv>`,
631 an optional ``unnamed_addr`` attribute, a return type, an optional
632 :ref:`parameter attribute <paramattrs>` for the return type, a function
633 name, a (possibly empty) argument list (each with optional :ref:`parameter
634 attributes <paramattrs>`), optional :ref:`function attributes <fnattrs>`,
635 an optional section, an optional alignment,
636 an optional :ref:`comdat <langref_comdats>`,
637 an optional :ref:`garbage collector name <gc>`, an optional :ref:`prefix <prefixdata>`,
638 an optional :ref:`prologue <prologuedata>`,
639 an optional :ref:`personality <personalityfn>`,
640 an opening curly brace, a list of basic blocks, and a closing curly brace.
642 LLVM function declarations consist of the "``declare``" keyword, an
643 optional :ref:`linkage type <linkage>`, an optional :ref:`visibility
644 style <visibility>`, an optional :ref:`DLL storage class <dllstorageclass>`,
645 an optional :ref:`calling convention <callingconv>`,
646 an optional ``unnamed_addr`` attribute, a return type, an optional
647 :ref:`parameter attribute <paramattrs>` for the return type, a function
648 name, a possibly empty list of arguments, an optional alignment, an optional
649 :ref:`garbage collector name <gc>`, an optional :ref:`prefix <prefixdata>`,
650 and an optional :ref:`prologue <prologuedata>`.
652 A function definition contains a list of basic blocks, forming the CFG (Control
653 Flow Graph) for the function. Each basic block may optionally start with a label
654 (giving the basic block a symbol table entry), contains a list of instructions,
655 and ends with a :ref:`terminator <terminators>` instruction (such as a branch or
656 function return). If an explicit label is not provided, a block is assigned an
657 implicit numbered label, using the next value from the same counter as used for
658 unnamed temporaries (:ref:`see above<identifiers>`). For example, if a function
659 entry block does not have an explicit label, it will be assigned label "%0",
660 then the first unnamed temporary in that block will be "%1", etc.
662 The first basic block in a function is special in two ways: it is
663 immediately executed on entrance to the function, and it is not allowed
664 to have predecessor basic blocks (i.e. there can not be any branches to
665 the entry block of a function). Because the block can have no
666 predecessors, it also cannot have any :ref:`PHI nodes <i_phi>`.
668 LLVM allows an explicit section to be specified for functions. If the
669 target supports it, it will emit functions to the section specified.
670 Additionally, the function can be placed in a COMDAT.
672 An explicit alignment may be specified for a function. If not present,
673 or if the alignment is set to zero, the alignment of the function is set
674 by the target to whatever it feels convenient. If an explicit alignment
675 is specified, the function is forced to have at least that much
676 alignment. All alignments must be a power of 2.
678 If the ``unnamed_addr`` attribute is given, the address is known to not
679 be significant and two identical functions can be merged.
683 define [linkage] [visibility] [DLLStorageClass]
685 <ResultType> @<FunctionName> ([argument list])
686 [unnamed_addr] [fn Attrs] [section "name"] [comdat [($name)]]
687 [align N] [gc] [prefix Constant] [prologue Constant]
688 [personality Constant] { ... }
690 The argument list is a comma seperated sequence of arguments where each
691 argument is of the following form
695 <type> [parameter Attrs] [name]
703 Aliases, unlike function or variables, don't create any new data. They
704 are just a new symbol and metadata for an existing position.
706 Aliases have a name and an aliasee that is either a global value or a
709 Aliases may have an optional :ref:`linkage type <linkage>`, an optional
710 :ref:`visibility style <visibility>`, an optional :ref:`DLL storage class
711 <dllstorageclass>` and an optional :ref:`tls model <tls_model>`.
715 @<Name> = [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal] [unnamed_addr] alias <AliaseeTy> @<Aliasee>
717 The linkage must be one of ``private``, ``internal``, ``linkonce``, ``weak``,
718 ``linkonce_odr``, ``weak_odr``, ``external``. Note that some system linkers
719 might not correctly handle dropping a weak symbol that is aliased.
721 Aliases that are not ``unnamed_addr`` are guaranteed to have the same address as
722 the aliasee expression. ``unnamed_addr`` ones are only guaranteed to point
725 Since aliases are only a second name, some restrictions apply, of which
726 some can only be checked when producing an object file:
728 * The expression defining the aliasee must be computable at assembly
729 time. Since it is just a name, no relocations can be used.
731 * No alias in the expression can be weak as the possibility of the
732 intermediate alias being overridden cannot be represented in an
735 * No global value in the expression can be a declaration, since that
736 would require a relocation, which is not possible.
743 Comdat IR provides access to COFF and ELF object file COMDAT functionality.
745 Comdats have a name which represents the COMDAT key. All global objects that
746 specify this key will only end up in the final object file if the linker chooses
747 that key over some other key. Aliases are placed in the same COMDAT that their
748 aliasee computes to, if any.
750 Comdats have a selection kind to provide input on how the linker should
751 choose between keys in two different object files.
755 $<Name> = comdat SelectionKind
757 The selection kind must be one of the following:
760 The linker may choose any COMDAT key, the choice is arbitrary.
762 The linker may choose any COMDAT key but the sections must contain the
765 The linker will choose the section containing the largest COMDAT key.
767 The linker requires that only section with this COMDAT key exist.
769 The linker may choose any COMDAT key but the sections must contain the
772 Note that the Mach-O platform doesn't support COMDATs and ELF only supports
773 ``any`` as a selection kind.
775 Here is an example of a COMDAT group where a function will only be selected if
776 the COMDAT key's section is the largest:
780 $foo = comdat largest
781 @foo = global i32 2, comdat($foo)
783 define void @bar() comdat($foo) {
787 As a syntactic sugar the ``$name`` can be omitted if the name is the same as
793 @foo = global i32 2, comdat
796 In a COFF object file, this will create a COMDAT section with selection kind
797 ``IMAGE_COMDAT_SELECT_LARGEST`` containing the contents of the ``@foo`` symbol
798 and another COMDAT section with selection kind
799 ``IMAGE_COMDAT_SELECT_ASSOCIATIVE`` which is associated with the first COMDAT
800 section and contains the contents of the ``@bar`` symbol.
802 There are some restrictions on the properties of the global object.
803 It, or an alias to it, must have the same name as the COMDAT group when
805 The contents and size of this object may be used during link-time to determine
806 which COMDAT groups get selected depending on the selection kind.
807 Because the name of the object must match the name of the COMDAT group, the
808 linkage of the global object must not be local; local symbols can get renamed
809 if a collision occurs in the symbol table.
811 The combined use of COMDATS and section attributes may yield surprising results.
818 @g1 = global i32 42, section "sec", comdat($foo)
819 @g2 = global i32 42, section "sec", comdat($bar)
821 From the object file perspective, this requires the creation of two sections
822 with the same name. This is necessary because both globals belong to different
823 COMDAT groups and COMDATs, at the object file level, are represented by
826 Note that certain IR constructs like global variables and functions may
827 create COMDATs in the object file in addition to any which are specified using
828 COMDAT IR. This arises when the code generator is configured to emit globals
829 in individual sections (e.g. when `-data-sections` or `-function-sections`
830 is supplied to `llc`).
832 .. _namedmetadatastructure:
837 Named metadata is a collection of metadata. :ref:`Metadata
838 nodes <metadata>` (but not metadata strings) are the only valid
839 operands for a named metadata.
841 #. Named metadata are represented as a string of characters with the
842 metadata prefix. The rules for metadata names are the same as for
843 identifiers, but quoted names are not allowed. ``"\xx"`` type escapes
844 are still valid, which allows any character to be part of a name.
848 ; Some unnamed metadata nodes, which are referenced by the named metadata.
853 !name = !{!0, !1, !2}
860 The return type and each parameter of a function type may have a set of
861 *parameter attributes* associated with them. Parameter attributes are
862 used to communicate additional information about the result or
863 parameters of a function. Parameter attributes are considered to be part
864 of the function, not of the function type, so functions with different
865 parameter attributes can have the same function type.
867 Parameter attributes are simple keywords that follow the type specified.
868 If multiple parameter attributes are needed, they are space separated.
873 declare i32 @printf(i8* noalias nocapture, ...)
874 declare i32 @atoi(i8 zeroext)
875 declare signext i8 @returns_signed_char()
877 Note that any attributes for the function result (``nounwind``,
878 ``readonly``) come immediately after the argument list.
880 Currently, only the following parameter attributes are defined:
883 This indicates to the code generator that the parameter or return
884 value should be zero-extended to the extent required by the target's
885 ABI (which is usually 32-bits, but is 8-bits for a i1 on x86-64) by
886 the caller (for a parameter) or the callee (for a return value).
888 This indicates to the code generator that the parameter or return
889 value should be sign-extended to the extent required by the target's
890 ABI (which is usually 32-bits) by the caller (for a parameter) or
891 the callee (for a return value).
893 This indicates that this parameter or return value should be treated
894 in a special target-dependent fashion during while emitting code for
895 a function call or return (usually, by putting it in a register as
896 opposed to memory, though some targets use it to distinguish between
897 two different kinds of registers). Use of this attribute is
900 This indicates that the pointer parameter should really be passed by
901 value to the function. The attribute implies that a hidden copy of
902 the pointee is made between the caller and the callee, so the callee
903 is unable to modify the value in the caller. This attribute is only
904 valid on LLVM pointer arguments. It is generally used to pass
905 structs and arrays by value, but is also valid on pointers to
906 scalars. The copy is considered to belong to the caller not the
907 callee (for example, ``readonly`` functions should not write to
908 ``byval`` parameters). This is not a valid attribute for return
911 The byval attribute also supports specifying an alignment with the
912 align attribute. It indicates the alignment of the stack slot to
913 form and the known alignment of the pointer specified to the call
914 site. If the alignment is not specified, then the code generator
915 makes a target-specific assumption.
921 The ``inalloca`` argument attribute allows the caller to take the
922 address of outgoing stack arguments. An ``inalloca`` argument must
923 be a pointer to stack memory produced by an ``alloca`` instruction.
924 The alloca, or argument allocation, must also be tagged with the
925 inalloca keyword. Only the last argument may have the ``inalloca``
926 attribute, and that argument is guaranteed to be passed in memory.
928 An argument allocation may be used by a call at most once because
929 the call may deallocate it. The ``inalloca`` attribute cannot be
930 used in conjunction with other attributes that affect argument
931 storage, like ``inreg``, ``nest``, ``sret``, or ``byval``. The
932 ``inalloca`` attribute also disables LLVM's implicit lowering of
933 large aggregate return values, which means that frontend authors
934 must lower them with ``sret`` pointers.
936 When the call site is reached, the argument allocation must have
937 been the most recent stack allocation that is still live, or the
938 results are undefined. It is possible to allocate additional stack
939 space after an argument allocation and before its call site, but it
940 must be cleared off with :ref:`llvm.stackrestore
943 See :doc:`InAlloca` for more information on how to use this
947 This indicates that the pointer parameter specifies the address of a
948 structure that is the return value of the function in the source
949 program. This pointer must be guaranteed by the caller to be valid:
950 loads and stores to the structure may be assumed by the callee
951 not to trap and to be properly aligned. This may only be applied to
952 the first parameter. This is not a valid attribute for return
956 This indicates that the pointer value may be assumed by the optimizer to
957 have the specified alignment.
959 Note that this attribute has additional semantics when combined with the
965 This indicates that objects accessed via pointer values
966 :ref:`based <pointeraliasing>` on the argument or return value are not also
967 accessed, during the execution of the function, via pointer values not
968 *based* on the argument or return value. The attribute on a return value
969 also has additional semantics described below. The caller shares the
970 responsibility with the callee for ensuring that these requirements are met.
971 For further details, please see the discussion of the NoAlias response in
972 :ref:`alias analysis <Must, May, or No>`.
974 Note that this definition of ``noalias`` is intentionally similar
975 to the definition of ``restrict`` in C99 for function arguments.
977 For function return values, C99's ``restrict`` is not meaningful,
978 while LLVM's ``noalias`` is. Furthermore, the semantics of the ``noalias``
979 attribute on return values are stronger than the semantics of the attribute
980 when used on function arguments. On function return values, the ``noalias``
981 attribute indicates that the function acts like a system memory allocation
982 function, returning a pointer to allocated storage disjoint from the
983 storage for any other object accessible to the caller.
986 This indicates that the callee does not make any copies of the
987 pointer that outlive the callee itself. This is not a valid
988 attribute for return values.
993 This indicates that the pointer parameter can be excised using the
994 :ref:`trampoline intrinsics <int_trampoline>`. This is not a valid
995 attribute for return values and can only be applied to one parameter.
998 This indicates that the function always returns the argument as its return
999 value. This is an optimization hint to the code generator when generating
1000 the caller, allowing tail call optimization and omission of register saves
1001 and restores in some cases; it is not checked or enforced when generating
1002 the callee. The parameter and the function return type must be valid
1003 operands for the :ref:`bitcast instruction <i_bitcast>`. This is not a
1004 valid attribute for return values and can only be applied to one parameter.
1007 This indicates that the parameter or return pointer is not null. This
1008 attribute may only be applied to pointer typed parameters. This is not
1009 checked or enforced by LLVM, the caller must ensure that the pointer
1010 passed in is non-null, or the callee must ensure that the returned pointer
1013 ``dereferenceable(<n>)``
1014 This indicates that the parameter or return pointer is dereferenceable. This
1015 attribute may only be applied to pointer typed parameters. A pointer that
1016 is dereferenceable can be loaded from speculatively without a risk of
1017 trapping. The number of bytes known to be dereferenceable must be provided
1018 in parentheses. It is legal for the number of bytes to be less than the
1019 size of the pointee type. The ``nonnull`` attribute does not imply
1020 dereferenceability (consider a pointer to one element past the end of an
1021 array), however ``dereferenceable(<n>)`` does imply ``nonnull`` in
1022 ``addrspace(0)`` (which is the default address space).
1024 ``dereferenceable_or_null(<n>)``
1025 This indicates that the parameter or return value isn't both
1026 non-null and non-dereferenceable (up to ``<n>`` bytes) at the same
1027 time. All non-null pointers tagged with
1028 ``dereferenceable_or_null(<n>)`` are ``dereferenceable(<n>)``.
1029 For address space 0 ``dereferenceable_or_null(<n>)`` implies that
1030 a pointer is exactly one of ``dereferenceable(<n>)`` or ``null``,
1031 and in other address spaces ``dereferenceable_or_null(<n>)``
1032 implies that a pointer is at least one of ``dereferenceable(<n>)``
1033 or ``null`` (i.e. it may be both ``null`` and
1034 ``dereferenceable(<n>)``). This attribute may only be applied to
1035 pointer typed parameters.
1039 Garbage Collector Strategy Names
1040 --------------------------------
1042 Each function may specify a garbage collector strategy name, which is simply a
1045 .. code-block:: llvm
1047 define void @f() gc "name" { ... }
1049 The supported values of *name* includes those :ref:`built in to LLVM
1050 <builtin-gc-strategies>` and any provided by loaded plugins. Specifying a GC
1051 strategy will cause the compiler to alter its output in order to support the
1052 named garbage collection algorithm. Note that LLVM itself does not contain a
1053 garbage collector, this functionality is restricted to generating machine code
1054 which can interoperate with a collector provided externally.
1061 Prefix data is data associated with a function which the code
1062 generator will emit immediately before the function's entrypoint.
1063 The purpose of this feature is to allow frontends to associate
1064 language-specific runtime metadata with specific functions and make it
1065 available through the function pointer while still allowing the
1066 function pointer to be called.
1068 To access the data for a given function, a program may bitcast the
1069 function pointer to a pointer to the constant's type and dereference
1070 index -1. This implies that the IR symbol points just past the end of
1071 the prefix data. For instance, take the example of a function annotated
1072 with a single ``i32``,
1074 .. code-block:: llvm
1076 define void @f() prefix i32 123 { ... }
1078 The prefix data can be referenced as,
1080 .. code-block:: llvm
1082 %0 = bitcast void* () @f to i32*
1083 %a = getelementptr inbounds i32, i32* %0, i32 -1
1084 %b = load i32, i32* %a
1086 Prefix data is laid out as if it were an initializer for a global variable
1087 of the prefix data's type. The function will be placed such that the
1088 beginning of the prefix data is aligned. This means that if the size
1089 of the prefix data is not a multiple of the alignment size, the
1090 function's entrypoint will not be aligned. If alignment of the
1091 function's entrypoint is desired, padding must be added to the prefix
1094 A function may have prefix data but no body. This has similar semantics
1095 to the ``available_externally`` linkage in that the data may be used by the
1096 optimizers but will not be emitted in the object file.
1103 The ``prologue`` attribute allows arbitrary code (encoded as bytes) to
1104 be inserted prior to the function body. This can be used for enabling
1105 function hot-patching and instrumentation.
1107 To maintain the semantics of ordinary function calls, the prologue data must
1108 have a particular format. Specifically, it must begin with a sequence of
1109 bytes which decode to a sequence of machine instructions, valid for the
1110 module's target, which transfer control to the point immediately succeeding
1111 the prologue data, without performing any other visible action. This allows
1112 the inliner and other passes to reason about the semantics of the function
1113 definition without needing to reason about the prologue data. Obviously this
1114 makes the format of the prologue data highly target dependent.
1116 A trivial example of valid prologue data for the x86 architecture is ``i8 144``,
1117 which encodes the ``nop`` instruction:
1119 .. code-block:: llvm
1121 define void @f() prologue i8 144 { ... }
1123 Generally prologue data can be formed by encoding a relative branch instruction
1124 which skips the metadata, as in this example of valid prologue data for the
1125 x86_64 architecture, where the first two bytes encode ``jmp .+10``:
1127 .. code-block:: llvm
1129 %0 = type <{ i8, i8, i8* }>
1131 define void @f() prologue %0 <{ i8 235, i8 8, i8* @md}> { ... }
1133 A function may have prologue data but no body. This has similar semantics
1134 to the ``available_externally`` linkage in that the data may be used by the
1135 optimizers but will not be emitted in the object file.
1139 Personality Function
1140 --------------------
1142 The ``personality`` attribute permits functions to specify what function
1143 to use for exception handling.
1150 Attribute groups are groups of attributes that are referenced by objects within
1151 the IR. They are important for keeping ``.ll`` files readable, because a lot of
1152 functions will use the same set of attributes. In the degenerative case of a
1153 ``.ll`` file that corresponds to a single ``.c`` file, the single attribute
1154 group will capture the important command line flags used to build that file.
1156 An attribute group is a module-level object. To use an attribute group, an
1157 object references the attribute group's ID (e.g. ``#37``). An object may refer
1158 to more than one attribute group. In that situation, the attributes from the
1159 different groups are merged.
1161 Here is an example of attribute groups for a function that should always be
1162 inlined, has a stack alignment of 4, and which shouldn't use SSE instructions:
1164 .. code-block:: llvm
1166 ; Target-independent attributes:
1167 attributes #0 = { alwaysinline alignstack=4 }
1169 ; Target-dependent attributes:
1170 attributes #1 = { "no-sse" }
1172 ; Function @f has attributes: alwaysinline, alignstack=4, and "no-sse".
1173 define void @f() #0 #1 { ... }
1180 Function attributes are set to communicate additional information about
1181 a function. Function attributes are considered to be part of the
1182 function, not of the function type, so functions with different function
1183 attributes can have the same function type.
1185 Function attributes are simple keywords that follow the type specified.
1186 If multiple attributes are needed, they are space separated. For
1189 .. code-block:: llvm
1191 define void @f() noinline { ... }
1192 define void @f() alwaysinline { ... }
1193 define void @f() alwaysinline optsize { ... }
1194 define void @f() optsize { ... }
1197 This attribute indicates that, when emitting the prologue and
1198 epilogue, the backend should forcibly align the stack pointer.
1199 Specify the desired alignment, which must be a power of two, in
1202 This attribute indicates that the inliner should attempt to inline
1203 this function into callers whenever possible, ignoring any active
1204 inlining size threshold for this caller.
1206 This indicates that the callee function at a call site should be
1207 recognized as a built-in function, even though the function's declaration
1208 uses the ``nobuiltin`` attribute. This is only valid at call sites for
1209 direct calls to functions that are declared with the ``nobuiltin``
1212 This attribute indicates that this function is rarely called. When
1213 computing edge weights, basic blocks post-dominated by a cold
1214 function call are also considered to be cold; and, thus, given low
1217 This attribute indicates that the callee is dependent on a convergent
1218 thread execution pattern under certain parallel execution models.
1219 Transformations that are execution model agnostic may only move or
1220 tranform this call if the final location is control equivalent to its
1221 original position in the program, where control equivalence is defined as
1222 A dominates B and B post-dominates A, or vice versa.
1224 This attribute indicates that the source code contained a hint that
1225 inlining this function is desirable (such as the "inline" keyword in
1226 C/C++). It is just a hint; it imposes no requirements on the
1229 This attribute indicates that the function should be added to a
1230 jump-instruction table at code-generation time, and that all address-taken
1231 references to this function should be replaced with a reference to the
1232 appropriate jump-instruction-table function pointer. Note that this creates
1233 a new pointer for the original function, which means that code that depends
1234 on function-pointer identity can break. So, any function annotated with
1235 ``jumptable`` must also be ``unnamed_addr``.
1237 This attribute suggests that optimization passes and code generator
1238 passes make choices that keep the code size of this function as small
1239 as possible and perform optimizations that may sacrifice runtime
1240 performance in order to minimize the size of the generated code.
1242 This attribute disables prologue / epilogue emission for the
1243 function. This can have very system-specific consequences.
1245 This indicates that the callee function at a call site is not recognized as
1246 a built-in function. LLVM will retain the original call and not replace it
1247 with equivalent code based on the semantics of the built-in function, unless
1248 the call site uses the ``builtin`` attribute. This is valid at call sites
1249 and on function declarations and definitions.
1251 This attribute indicates that calls to the function cannot be
1252 duplicated. A call to a ``noduplicate`` function may be moved
1253 within its parent function, but may not be duplicated within
1254 its parent function.
1256 A function containing a ``noduplicate`` call may still
1257 be an inlining candidate, provided that the call is not
1258 duplicated by inlining. That implies that the function has
1259 internal linkage and only has one call site, so the original
1260 call is dead after inlining.
1262 This attributes disables implicit floating point instructions.
1264 This attribute indicates that the inliner should never inline this
1265 function in any situation. This attribute may not be used together
1266 with the ``alwaysinline`` attribute.
1268 This attribute suppresses lazy symbol binding for the function. This
1269 may make calls to the function faster, at the cost of extra program
1270 startup time if the function is not called during program startup.
1272 This attribute indicates that the code generator should not use a
1273 red zone, even if the target-specific ABI normally permits it.
1275 This function attribute indicates that the function never returns
1276 normally. This produces undefined behavior at runtime if the
1277 function ever does dynamically return.
1279 This function attribute indicates that the function never raises an
1280 exception. If the function does raise an exception, its runtime
1281 behavior is undefined. However, functions marked nounwind may still
1282 trap or generate asynchronous exceptions. Exception handling schemes
1283 that are recognized by LLVM to handle asynchronous exceptions, such
1284 as SEH, will still provide their implementation defined semantics.
1286 This function attribute indicates that the function is not optimized
1287 by any optimization or code generator passes with the
1288 exception of interprocedural optimization passes.
1289 This attribute cannot be used together with the ``alwaysinline``
1290 attribute; this attribute is also incompatible
1291 with the ``minsize`` attribute and the ``optsize`` attribute.
1293 This attribute requires the ``noinline`` attribute to be specified on
1294 the function as well, so the function is never inlined into any caller.
1295 Only functions with the ``alwaysinline`` attribute are valid
1296 candidates for inlining into the body of this function.
1298 This attribute suggests that optimization passes and code generator
1299 passes make choices that keep the code size of this function low,
1300 and otherwise do optimizations specifically to reduce code size as
1301 long as they do not significantly impact runtime performance.
1303 On a function, this attribute indicates that the function computes its
1304 result (or decides to unwind an exception) based strictly on its arguments,
1305 without dereferencing any pointer arguments or otherwise accessing
1306 any mutable state (e.g. memory, control registers, etc) visible to
1307 caller functions. It does not write through any pointer arguments
1308 (including ``byval`` arguments) and never changes any state visible
1309 to callers. This means that it cannot unwind exceptions by calling
1310 the ``C++`` exception throwing methods.
1312 On an argument, this attribute indicates that the function does not
1313 dereference that pointer argument, even though it may read or write the
1314 memory that the pointer points to if accessed through other pointers.
1316 On a function, this attribute indicates that the function does not write
1317 through any pointer arguments (including ``byval`` arguments) or otherwise
1318 modify any state (e.g. memory, control registers, etc) visible to
1319 caller functions. It may dereference pointer arguments and read
1320 state that may be set in the caller. A readonly function always
1321 returns the same value (or unwinds an exception identically) when
1322 called with the same set of arguments and global state. It cannot
1323 unwind an exception by calling the ``C++`` exception throwing
1326 On an argument, this attribute indicates that the function does not write
1327 through this pointer argument, even though it may write to the memory that
1328 the pointer points to.
1330 This attribute indicates that this function can return twice. The C
1331 ``setjmp`` is an example of such a function. The compiler disables
1332 some optimizations (like tail calls) in the caller of these
1335 This attribute indicates that
1336 `SafeStack <http://clang.llvm.org/docs/SafeStack.html>`_
1337 protection is enabled for this function.
1339 If a function that has a ``safestack`` attribute is inlined into a
1340 function that doesn't have a ``safestack`` attribute or which has an
1341 ``ssp``, ``sspstrong`` or ``sspreq`` attribute, then the resulting
1342 function will have a ``safestack`` attribute.
1343 ``sanitize_address``
1344 This attribute indicates that AddressSanitizer checks
1345 (dynamic address safety analysis) are enabled for this function.
1347 This attribute indicates that MemorySanitizer checks (dynamic detection
1348 of accesses to uninitialized memory) are enabled for this function.
1350 This attribute indicates that ThreadSanitizer checks
1351 (dynamic thread safety analysis) are enabled for this function.
1353 This attribute indicates that the function should emit a stack
1354 smashing protector. It is in the form of a "canary" --- a random value
1355 placed on the stack before the local variables that's checked upon
1356 return from the function to see if it has been overwritten. A
1357 heuristic is used to determine if a function needs stack protectors
1358 or not. The heuristic used will enable protectors for functions with:
1360 - Character arrays larger than ``ssp-buffer-size`` (default 8).
1361 - Aggregates containing character arrays larger than ``ssp-buffer-size``.
1362 - Calls to alloca() with variable sizes or constant sizes greater than
1363 ``ssp-buffer-size``.
1365 Variables that are identified as requiring a protector will be arranged
1366 on the stack such that they are adjacent to the stack protector guard.
1368 If a function that has an ``ssp`` attribute is inlined into a
1369 function that doesn't have an ``ssp`` attribute, then the resulting
1370 function will have an ``ssp`` attribute.
1372 This attribute indicates that the function should *always* emit a
1373 stack smashing protector. This overrides the ``ssp`` function
1376 Variables that are identified as requiring a protector will be arranged
1377 on the stack such that they are adjacent to the stack protector guard.
1378 The specific layout rules are:
1380 #. Large arrays and structures containing large arrays
1381 (``>= ssp-buffer-size``) are closest to the stack protector.
1382 #. Small arrays and structures containing small arrays
1383 (``< ssp-buffer-size``) are 2nd closest to the protector.
1384 #. Variables that have had their address taken are 3rd closest to the
1387 If a function that has an ``sspreq`` attribute is inlined into a
1388 function that doesn't have an ``sspreq`` attribute or which has an
1389 ``ssp`` or ``sspstrong`` attribute, then the resulting function will have
1390 an ``sspreq`` attribute.
1392 This attribute indicates that the function should emit a stack smashing
1393 protector. This attribute causes a strong heuristic to be used when
1394 determining if a function needs stack protectors. The strong heuristic
1395 will enable protectors for functions with:
1397 - Arrays of any size and type
1398 - Aggregates containing an array of any size and type.
1399 - Calls to alloca().
1400 - Local variables that have had their address taken.
1402 Variables that are identified as requiring a protector will be arranged
1403 on the stack such that they are adjacent to the stack protector guard.
1404 The specific layout rules are:
1406 #. Large arrays and structures containing large arrays
1407 (``>= ssp-buffer-size``) are closest to the stack protector.
1408 #. Small arrays and structures containing small arrays
1409 (``< ssp-buffer-size``) are 2nd closest to the protector.
1410 #. Variables that have had their address taken are 3rd closest to the
1413 This overrides the ``ssp`` function attribute.
1415 If a function that has an ``sspstrong`` attribute is inlined into a
1416 function that doesn't have an ``sspstrong`` attribute, then the
1417 resulting function will have an ``sspstrong`` attribute.
1419 This attribute indicates that the function will delegate to some other
1420 function with a tail call. The prototype of a thunk should not be used for
1421 optimization purposes. The caller is expected to cast the thunk prototype to
1422 match the thunk target prototype.
1424 This attribute indicates that the ABI being targeted requires that
1425 an unwind table entry be produce for this function even if we can
1426 show that no exceptions passes by it. This is normally the case for
1427 the ELF x86-64 abi, but it can be disabled for some compilation
1432 Module-Level Inline Assembly
1433 ----------------------------
1435 Modules may contain "module-level inline asm" blocks, which corresponds
1436 to the GCC "file scope inline asm" blocks. These blocks are internally
1437 concatenated by LLVM and treated as a single unit, but may be separated
1438 in the ``.ll`` file if desired. The syntax is very simple:
1440 .. code-block:: llvm
1442 module asm "inline asm code goes here"
1443 module asm "more can go here"
1445 The strings can contain any character by escaping non-printable
1446 characters. The escape sequence used is simply "\\xx" where "xx" is the
1447 two digit hex code for the number.
1449 Note that the assembly string *must* be parseable by LLVM's integrated assembler
1450 (unless it is disabled), even when emitting a ``.s`` file.
1452 .. _langref_datalayout:
1457 A module may specify a target specific data layout string that specifies
1458 how data is to be laid out in memory. The syntax for the data layout is
1461 .. code-block:: llvm
1463 target datalayout = "layout specification"
1465 The *layout specification* consists of a list of specifications
1466 separated by the minus sign character ('-'). Each specification starts
1467 with a letter and may include other information after the letter to
1468 define some aspect of the data layout. The specifications accepted are
1472 Specifies that the target lays out data in big-endian form. That is,
1473 the bits with the most significance have the lowest address
1476 Specifies that the target lays out data in little-endian form. That
1477 is, the bits with the least significance have the lowest address
1480 Specifies the natural alignment of the stack in bits. Alignment
1481 promotion of stack variables is limited to the natural stack
1482 alignment to avoid dynamic stack realignment. The stack alignment
1483 must be a multiple of 8-bits. If omitted, the natural stack
1484 alignment defaults to "unspecified", which does not prevent any
1485 alignment promotions.
1486 ``p[n]:<size>:<abi>:<pref>``
1487 This specifies the *size* of a pointer and its ``<abi>`` and
1488 ``<pref>``\erred alignments for address space ``n``. All sizes are in
1489 bits. The address space, ``n`` is optional, and if not specified,
1490 denotes the default address space 0. The value of ``n`` must be
1491 in the range [1,2^23).
1492 ``i<size>:<abi>:<pref>``
1493 This specifies the alignment for an integer type of a given bit
1494 ``<size>``. The value of ``<size>`` must be in the range [1,2^23).
1495 ``v<size>:<abi>:<pref>``
1496 This specifies the alignment for a vector type of a given bit
1498 ``f<size>:<abi>:<pref>``
1499 This specifies the alignment for a floating point type of a given bit
1500 ``<size>``. Only values of ``<size>`` that are supported by the target
1501 will work. 32 (float) and 64 (double) are supported on all targets; 80
1502 or 128 (different flavors of long double) are also supported on some
1505 This specifies the alignment for an object of aggregate type.
1507 If present, specifies that llvm names are mangled in the output. The
1510 * ``e``: ELF mangling: Private symbols get a ``.L`` prefix.
1511 * ``m``: Mips mangling: Private symbols get a ``$`` prefix.
1512 * ``o``: Mach-O mangling: Private symbols get ``L`` prefix. Other
1513 symbols get a ``_`` prefix.
1514 * ``w``: Windows COFF prefix: Similar to Mach-O, but stdcall and fastcall
1515 functions also get a suffix based on the frame size.
1516 ``n<size1>:<size2>:<size3>...``
1517 This specifies a set of native integer widths for the target CPU in
1518 bits. For example, it might contain ``n32`` for 32-bit PowerPC,
1519 ``n32:64`` for PowerPC 64, or ``n8:16:32:64`` for X86-64. Elements of
1520 this set are considered to support most general arithmetic operations
1523 On every specification that takes a ``<abi>:<pref>``, specifying the
1524 ``<pref>`` alignment is optional. If omitted, the preceding ``:``
1525 should be omitted too and ``<pref>`` will be equal to ``<abi>``.
1527 When constructing the data layout for a given target, LLVM starts with a
1528 default set of specifications which are then (possibly) overridden by
1529 the specifications in the ``datalayout`` keyword. The default
1530 specifications are given in this list:
1532 - ``E`` - big endian
1533 - ``p:64:64:64`` - 64-bit pointers with 64-bit alignment.
1534 - ``p[n]:64:64:64`` - Other address spaces are assumed to be the
1535 same as the default address space.
1536 - ``S0`` - natural stack alignment is unspecified
1537 - ``i1:8:8`` - i1 is 8-bit (byte) aligned
1538 - ``i8:8:8`` - i8 is 8-bit (byte) aligned
1539 - ``i16:16:16`` - i16 is 16-bit aligned
1540 - ``i32:32:32`` - i32 is 32-bit aligned
1541 - ``i64:32:64`` - i64 has ABI alignment of 32-bits but preferred
1542 alignment of 64-bits
1543 - ``f16:16:16`` - half is 16-bit aligned
1544 - ``f32:32:32`` - float is 32-bit aligned
1545 - ``f64:64:64`` - double is 64-bit aligned
1546 - ``f128:128:128`` - quad is 128-bit aligned
1547 - ``v64:64:64`` - 64-bit vector is 64-bit aligned
1548 - ``v128:128:128`` - 128-bit vector is 128-bit aligned
1549 - ``a:0:64`` - aggregates are 64-bit aligned
1551 When LLVM is determining the alignment for a given type, it uses the
1554 #. If the type sought is an exact match for one of the specifications,
1555 that specification is used.
1556 #. If no match is found, and the type sought is an integer type, then
1557 the smallest integer type that is larger than the bitwidth of the
1558 sought type is used. If none of the specifications are larger than
1559 the bitwidth then the largest integer type is used. For example,
1560 given the default specifications above, the i7 type will use the
1561 alignment of i8 (next largest) while both i65 and i256 will use the
1562 alignment of i64 (largest specified).
1563 #. If no match is found, and the type sought is a vector type, then the
1564 largest vector type that is smaller than the sought vector type will
1565 be used as a fall back. This happens because <128 x double> can be
1566 implemented in terms of 64 <2 x double>, for example.
1568 The function of the data layout string may not be what you expect.
1569 Notably, this is not a specification from the frontend of what alignment
1570 the code generator should use.
1572 Instead, if specified, the target data layout is required to match what
1573 the ultimate *code generator* expects. This string is used by the
1574 mid-level optimizers to improve code, and this only works if it matches
1575 what the ultimate code generator uses. There is no way to generate IR
1576 that does not embed this target-specific detail into the IR. If you
1577 don't specify the string, the default specifications will be used to
1578 generate a Data Layout and the optimization phases will operate
1579 accordingly and introduce target specificity into the IR with respect to
1580 these default specifications.
1587 A module may specify a target triple string that describes the target
1588 host. The syntax for the target triple is simply:
1590 .. code-block:: llvm
1592 target triple = "x86_64-apple-macosx10.7.0"
1594 The *target triple* string consists of a series of identifiers delimited
1595 by the minus sign character ('-'). The canonical forms are:
1599 ARCHITECTURE-VENDOR-OPERATING_SYSTEM
1600 ARCHITECTURE-VENDOR-OPERATING_SYSTEM-ENVIRONMENT
1602 This information is passed along to the backend so that it generates
1603 code for the proper architecture. It's possible to override this on the
1604 command line with the ``-mtriple`` command line option.
1606 .. _pointeraliasing:
1608 Pointer Aliasing Rules
1609 ----------------------
1611 Any memory access must be done through a pointer value associated with
1612 an address range of the memory access, otherwise the behavior is
1613 undefined. Pointer values are associated with address ranges according
1614 to the following rules:
1616 - A pointer value is associated with the addresses associated with any
1617 value it is *based* on.
1618 - An address of a global variable is associated with the address range
1619 of the variable's storage.
1620 - The result value of an allocation instruction is associated with the
1621 address range of the allocated storage.
1622 - A null pointer in the default address-space is associated with no
1624 - An integer constant other than zero or a pointer value returned from
1625 a function not defined within LLVM may be associated with address
1626 ranges allocated through mechanisms other than those provided by
1627 LLVM. Such ranges shall not overlap with any ranges of addresses
1628 allocated by mechanisms provided by LLVM.
1630 A pointer value is *based* on another pointer value according to the
1633 - A pointer value formed from a ``getelementptr`` operation is *based*
1634 on the first value operand of the ``getelementptr``.
1635 - The result value of a ``bitcast`` is *based* on the operand of the
1637 - A pointer value formed by an ``inttoptr`` is *based* on all pointer
1638 values that contribute (directly or indirectly) to the computation of
1639 the pointer's value.
1640 - The "*based* on" relationship is transitive.
1642 Note that this definition of *"based"* is intentionally similar to the
1643 definition of *"based"* in C99, though it is slightly weaker.
1645 LLVM IR does not associate types with memory. The result type of a
1646 ``load`` merely indicates the size and alignment of the memory from
1647 which to load, as well as the interpretation of the value. The first
1648 operand type of a ``store`` similarly only indicates the size and
1649 alignment of the store.
1651 Consequently, type-based alias analysis, aka TBAA, aka
1652 ``-fstrict-aliasing``, is not applicable to general unadorned LLVM IR.
1653 :ref:`Metadata <metadata>` may be used to encode additional information
1654 which specialized optimization passes may use to implement type-based
1659 Volatile Memory Accesses
1660 ------------------------
1662 Certain memory accesses, such as :ref:`load <i_load>`'s,
1663 :ref:`store <i_store>`'s, and :ref:`llvm.memcpy <int_memcpy>`'s may be
1664 marked ``volatile``. The optimizers must not change the number of
1665 volatile operations or change their order of execution relative to other
1666 volatile operations. The optimizers *may* change the order of volatile
1667 operations relative to non-volatile operations. This is not Java's
1668 "volatile" and has no cross-thread synchronization behavior.
1670 IR-level volatile loads and stores cannot safely be optimized into
1671 llvm.memcpy or llvm.memmove intrinsics even when those intrinsics are
1672 flagged volatile. Likewise, the backend should never split or merge
1673 target-legal volatile load/store instructions.
1675 .. admonition:: Rationale
1677 Platforms may rely on volatile loads and stores of natively supported
1678 data width to be executed as single instruction. For example, in C
1679 this holds for an l-value of volatile primitive type with native
1680 hardware support, but not necessarily for aggregate types. The
1681 frontend upholds these expectations, which are intentionally
1682 unspecified in the IR. The rules above ensure that IR transformation
1683 do not violate the frontend's contract with the language.
1687 Memory Model for Concurrent Operations
1688 --------------------------------------
1690 The LLVM IR does not define any way to start parallel threads of
1691 execution or to register signal handlers. Nonetheless, there are
1692 platform-specific ways to create them, and we define LLVM IR's behavior
1693 in their presence. This model is inspired by the C++0x memory model.
1695 For a more informal introduction to this model, see the :doc:`Atomics`.
1697 We define a *happens-before* partial order as the least partial order
1700 - Is a superset of single-thread program order, and
1701 - When a *synchronizes-with* ``b``, includes an edge from ``a`` to
1702 ``b``. *Synchronizes-with* pairs are introduced by platform-specific
1703 techniques, like pthread locks, thread creation, thread joining,
1704 etc., and by atomic instructions. (See also :ref:`Atomic Memory Ordering
1705 Constraints <ordering>`).
1707 Note that program order does not introduce *happens-before* edges
1708 between a thread and signals executing inside that thread.
1710 Every (defined) read operation (load instructions, memcpy, atomic
1711 loads/read-modify-writes, etc.) R reads a series of bytes written by
1712 (defined) write operations (store instructions, atomic
1713 stores/read-modify-writes, memcpy, etc.). For the purposes of this
1714 section, initialized globals are considered to have a write of the
1715 initializer which is atomic and happens before any other read or write
1716 of the memory in question. For each byte of a read R, R\ :sub:`byte`
1717 may see any write to the same byte, except:
1719 - If write\ :sub:`1` happens before write\ :sub:`2`, and
1720 write\ :sub:`2` happens before R\ :sub:`byte`, then
1721 R\ :sub:`byte` does not see write\ :sub:`1`.
1722 - If R\ :sub:`byte` happens before write\ :sub:`3`, then
1723 R\ :sub:`byte` does not see write\ :sub:`3`.
1725 Given that definition, R\ :sub:`byte` is defined as follows:
1727 - If R is volatile, the result is target-dependent. (Volatile is
1728 supposed to give guarantees which can support ``sig_atomic_t`` in
1729 C/C++, and may be used for accesses to addresses that do not behave
1730 like normal memory. It does not generally provide cross-thread
1732 - Otherwise, if there is no write to the same byte that happens before
1733 R\ :sub:`byte`, R\ :sub:`byte` returns ``undef`` for that byte.
1734 - Otherwise, if R\ :sub:`byte` may see exactly one write,
1735 R\ :sub:`byte` returns the value written by that write.
1736 - Otherwise, if R is atomic, and all the writes R\ :sub:`byte` may
1737 see are atomic, it chooses one of the values written. See the :ref:`Atomic
1738 Memory Ordering Constraints <ordering>` section for additional
1739 constraints on how the choice is made.
1740 - Otherwise R\ :sub:`byte` returns ``undef``.
1742 R returns the value composed of the series of bytes it read. This
1743 implies that some bytes within the value may be ``undef`` **without**
1744 the entire value being ``undef``. Note that this only defines the
1745 semantics of the operation; it doesn't mean that targets will emit more
1746 than one instruction to read the series of bytes.
1748 Note that in cases where none of the atomic intrinsics are used, this
1749 model places only one restriction on IR transformations on top of what
1750 is required for single-threaded execution: introducing a store to a byte
1751 which might not otherwise be stored is not allowed in general.
1752 (Specifically, in the case where another thread might write to and read
1753 from an address, introducing a store can change a load that may see
1754 exactly one write into a load that may see multiple writes.)
1758 Atomic Memory Ordering Constraints
1759 ----------------------------------
1761 Atomic instructions (:ref:`cmpxchg <i_cmpxchg>`,
1762 :ref:`atomicrmw <i_atomicrmw>`, :ref:`fence <i_fence>`,
1763 :ref:`atomic load <i_load>`, and :ref:`atomic store <i_store>`) take
1764 ordering parameters that determine which other atomic instructions on
1765 the same address they *synchronize with*. These semantics are borrowed
1766 from Java and C++0x, but are somewhat more colloquial. If these
1767 descriptions aren't precise enough, check those specs (see spec
1768 references in the :doc:`atomics guide <Atomics>`).
1769 :ref:`fence <i_fence>` instructions treat these orderings somewhat
1770 differently since they don't take an address. See that instruction's
1771 documentation for details.
1773 For a simpler introduction to the ordering constraints, see the
1777 The set of values that can be read is governed by the happens-before
1778 partial order. A value cannot be read unless some operation wrote
1779 it. This is intended to provide a guarantee strong enough to model
1780 Java's non-volatile shared variables. This ordering cannot be
1781 specified for read-modify-write operations; it is not strong enough
1782 to make them atomic in any interesting way.
1784 In addition to the guarantees of ``unordered``, there is a single
1785 total order for modifications by ``monotonic`` operations on each
1786 address. All modification orders must be compatible with the
1787 happens-before order. There is no guarantee that the modification
1788 orders can be combined to a global total order for the whole program
1789 (and this often will not be possible). The read in an atomic
1790 read-modify-write operation (:ref:`cmpxchg <i_cmpxchg>` and
1791 :ref:`atomicrmw <i_atomicrmw>`) reads the value in the modification
1792 order immediately before the value it writes. If one atomic read
1793 happens before another atomic read of the same address, the later
1794 read must see the same value or a later value in the address's
1795 modification order. This disallows reordering of ``monotonic`` (or
1796 stronger) operations on the same address. If an address is written
1797 ``monotonic``-ally by one thread, and other threads ``monotonic``-ally
1798 read that address repeatedly, the other threads must eventually see
1799 the write. This corresponds to the C++0x/C1x
1800 ``memory_order_relaxed``.
1802 In addition to the guarantees of ``monotonic``, a
1803 *synchronizes-with* edge may be formed with a ``release`` operation.
1804 This is intended to model C++'s ``memory_order_acquire``.
1806 In addition to the guarantees of ``monotonic``, if this operation
1807 writes a value which is subsequently read by an ``acquire``
1808 operation, it *synchronizes-with* that operation. (This isn't a
1809 complete description; see the C++0x definition of a release
1810 sequence.) This corresponds to the C++0x/C1x
1811 ``memory_order_release``.
1812 ``acq_rel`` (acquire+release)
1813 Acts as both an ``acquire`` and ``release`` operation on its
1814 address. This corresponds to the C++0x/C1x ``memory_order_acq_rel``.
1815 ``seq_cst`` (sequentially consistent)
1816 In addition to the guarantees of ``acq_rel`` (``acquire`` for an
1817 operation that only reads, ``release`` for an operation that only
1818 writes), there is a global total order on all
1819 sequentially-consistent operations on all addresses, which is
1820 consistent with the *happens-before* partial order and with the
1821 modification orders of all the affected addresses. Each
1822 sequentially-consistent read sees the last preceding write to the
1823 same address in this global order. This corresponds to the C++0x/C1x
1824 ``memory_order_seq_cst`` and Java volatile.
1828 If an atomic operation is marked ``singlethread``, it only *synchronizes
1829 with* or participates in modification and seq\_cst total orderings with
1830 other operations running in the same thread (for example, in signal
1838 LLVM IR floating-point binary ops (:ref:`fadd <i_fadd>`,
1839 :ref:`fsub <i_fsub>`, :ref:`fmul <i_fmul>`, :ref:`fdiv <i_fdiv>`,
1840 :ref:`frem <i_frem>`) have the following flags that can be set to enable
1841 otherwise unsafe floating point operations
1844 No NaNs - Allow optimizations to assume the arguments and result are not
1845 NaN. Such optimizations are required to retain defined behavior over
1846 NaNs, but the value of the result is undefined.
1849 No Infs - Allow optimizations to assume the arguments and result are not
1850 +/-Inf. Such optimizations are required to retain defined behavior over
1851 +/-Inf, but the value of the result is undefined.
1854 No Signed Zeros - Allow optimizations to treat the sign of a zero
1855 argument or result as insignificant.
1858 Allow Reciprocal - Allow optimizations to use the reciprocal of an
1859 argument rather than perform division.
1862 Fast - Allow algebraically equivalent transformations that may
1863 dramatically change results in floating point (e.g. reassociate). This
1864 flag implies all the others.
1868 Use-list Order Directives
1869 -------------------------
1871 Use-list directives encode the in-memory order of each use-list, allowing the
1872 order to be recreated. ``<order-indexes>`` is a comma-separated list of
1873 indexes that are assigned to the referenced value's uses. The referenced
1874 value's use-list is immediately sorted by these indexes.
1876 Use-list directives may appear at function scope or global scope. They are not
1877 instructions, and have no effect on the semantics of the IR. When they're at
1878 function scope, they must appear after the terminator of the final basic block.
1880 If basic blocks have their address taken via ``blockaddress()`` expressions,
1881 ``uselistorder_bb`` can be used to reorder their use-lists from outside their
1888 uselistorder <ty> <value>, { <order-indexes> }
1889 uselistorder_bb @function, %block { <order-indexes> }
1895 define void @foo(i32 %arg1, i32 %arg2) {
1897 ; ... instructions ...
1899 ; ... instructions ...
1901 ; At function scope.
1902 uselistorder i32 %arg1, { 1, 0, 2 }
1903 uselistorder label %bb, { 1, 0 }
1907 uselistorder i32* @global, { 1, 2, 0 }
1908 uselistorder i32 7, { 1, 0 }
1909 uselistorder i32 (i32) @bar, { 1, 0 }
1910 uselistorder_bb @foo, %bb, { 5, 1, 3, 2, 0, 4 }
1917 The LLVM type system is one of the most important features of the
1918 intermediate representation. Being typed enables a number of
1919 optimizations to be performed on the intermediate representation
1920 directly, without having to do extra analyses on the side before the
1921 transformation. A strong type system makes it easier to read the
1922 generated code and enables novel analyses and transformations that are
1923 not feasible to perform on normal three address code representations.
1933 The void type does not represent any value and has no size.
1951 The function type can be thought of as a function signature. It consists of a
1952 return type and a list of formal parameter types. The return type of a function
1953 type is a void type or first class type --- except for :ref:`label <t_label>`
1954 and :ref:`metadata <t_metadata>` types.
1960 <returntype> (<parameter list>)
1962 ...where '``<parameter list>``' is a comma-separated list of type
1963 specifiers. Optionally, the parameter list may include a type ``...``, which
1964 indicates that the function takes a variable number of arguments. Variable
1965 argument functions can access their arguments with the :ref:`variable argument
1966 handling intrinsic <int_varargs>` functions. '``<returntype>``' is any type
1967 except :ref:`label <t_label>` and :ref:`metadata <t_metadata>`.
1971 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1972 | ``i32 (i32)`` | function taking an ``i32``, returning an ``i32`` |
1973 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1974 | ``float (i16, i32 *) *`` | :ref:`Pointer <t_pointer>` to a function that takes an ``i16`` and a :ref:`pointer <t_pointer>` to ``i32``, returning ``float``. |
1975 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1976 | ``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. |
1977 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1978 | ``{i32, i32} (i32)`` | A function taking an ``i32``, returning a :ref:`structure <t_struct>` containing two ``i32`` values |
1979 +---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1986 The :ref:`first class <t_firstclass>` types are perhaps the most important.
1987 Values of these types are the only ones which can be produced by
1995 These are the types that are valid in registers from CodeGen's perspective.
2004 The integer type is a very simple type that simply specifies an
2005 arbitrary bit width for the integer type desired. Any bit width from 1
2006 bit to 2\ :sup:`23`\ -1 (about 8 million) can be specified.
2014 The number of bits the integer will occupy is specified by the ``N``
2020 +----------------+------------------------------------------------+
2021 | ``i1`` | a single-bit integer. |
2022 +----------------+------------------------------------------------+
2023 | ``i32`` | a 32-bit integer. |
2024 +----------------+------------------------------------------------+
2025 | ``i1942652`` | a really big integer of over 1 million bits. |
2026 +----------------+------------------------------------------------+
2030 Floating Point Types
2031 """"""""""""""""""""
2040 - 16-bit floating point value
2043 - 32-bit floating point value
2046 - 64-bit floating point value
2049 - 128-bit floating point value (112-bit mantissa)
2052 - 80-bit floating point value (X87)
2055 - 128-bit floating point value (two 64-bits)
2062 The x86_mmx type represents a value held in an MMX register on an x86
2063 machine. The operations allowed on it are quite limited: parameters and
2064 return values, load and store, and bitcast. User-specified MMX
2065 instructions are represented as intrinsic or asm calls with arguments
2066 and/or results of this type. There are no arrays, vectors or constants
2083 The pointer type is used to specify memory locations. Pointers are
2084 commonly used to reference objects in memory.
2086 Pointer types may have an optional address space attribute defining the
2087 numbered address space where the pointed-to object resides. The default
2088 address space is number zero. The semantics of non-zero address spaces
2089 are target-specific.
2091 Note that LLVM does not permit pointers to void (``void*``) nor does it
2092 permit pointers to labels (``label*``). Use ``i8*`` instead.
2102 +-------------------------+--------------------------------------------------------------------------------------------------------------+
2103 | ``[4 x i32]*`` | A :ref:`pointer <t_pointer>` to :ref:`array <t_array>` of four ``i32`` values. |
2104 +-------------------------+--------------------------------------------------------------------------------------------------------------+
2105 | ``i32 (i32*) *`` | A :ref:`pointer <t_pointer>` to a :ref:`function <t_function>` that takes an ``i32*``, returning an ``i32``. |
2106 +-------------------------+--------------------------------------------------------------------------------------------------------------+
2107 | ``i32 addrspace(5)*`` | A :ref:`pointer <t_pointer>` to an ``i32`` value that resides in address space #5. |
2108 +-------------------------+--------------------------------------------------------------------------------------------------------------+
2117 A vector type is a simple derived type that represents a vector of
2118 elements. Vector types are used when multiple primitive data are
2119 operated in parallel using a single instruction (SIMD). A vector type
2120 requires a size (number of elements) and an underlying primitive data
2121 type. Vector types are considered :ref:`first class <t_firstclass>`.
2127 < <# elements> x <elementtype> >
2129 The number of elements is a constant integer value larger than 0;
2130 elementtype may be any integer, floating point or pointer type. Vectors
2131 of size zero are not allowed.
2135 +-------------------+--------------------------------------------------+
2136 | ``<4 x i32>`` | Vector of 4 32-bit integer values. |
2137 +-------------------+--------------------------------------------------+
2138 | ``<8 x float>`` | Vector of 8 32-bit floating-point values. |
2139 +-------------------+--------------------------------------------------+
2140 | ``<2 x i64>`` | Vector of 2 64-bit integer values. |
2141 +-------------------+--------------------------------------------------+
2142 | ``<4 x i64*>`` | Vector of 4 pointers to 64-bit integer values. |
2143 +-------------------+--------------------------------------------------+
2152 The label type represents code labels.
2167 The metadata type represents embedded metadata. No derived types may be
2168 created from metadata except for :ref:`function <t_function>` arguments.
2181 Aggregate Types are a subset of derived types that can contain multiple
2182 member types. :ref:`Arrays <t_array>` and :ref:`structs <t_struct>` are
2183 aggregate types. :ref:`Vectors <t_vector>` are not considered to be
2193 The array type is a very simple derived type that arranges elements
2194 sequentially in memory. The array type requires a size (number of
2195 elements) and an underlying data type.
2201 [<# elements> x <elementtype>]
2203 The number of elements is a constant integer value; ``elementtype`` may
2204 be any type with a size.
2208 +------------------+--------------------------------------+
2209 | ``[40 x i32]`` | Array of 40 32-bit integer values. |
2210 +------------------+--------------------------------------+
2211 | ``[41 x i32]`` | Array of 41 32-bit integer values. |
2212 +------------------+--------------------------------------+
2213 | ``[4 x i8]`` | Array of 4 8-bit integer values. |
2214 +------------------+--------------------------------------+
2216 Here are some examples of multidimensional arrays:
2218 +-----------------------------+----------------------------------------------------------+
2219 | ``[3 x [4 x i32]]`` | 3x4 array of 32-bit integer values. |
2220 +-----------------------------+----------------------------------------------------------+
2221 | ``[12 x [10 x float]]`` | 12x10 array of single precision floating point values. |
2222 +-----------------------------+----------------------------------------------------------+
2223 | ``[2 x [3 x [4 x i16]]]`` | 2x3x4 array of 16-bit integer values. |
2224 +-----------------------------+----------------------------------------------------------+
2226 There is no restriction on indexing beyond the end of the array implied
2227 by a static type (though there are restrictions on indexing beyond the
2228 bounds of an allocated object in some cases). This means that
2229 single-dimension 'variable sized array' addressing can be implemented in
2230 LLVM with a zero length array type. An implementation of 'pascal style
2231 arrays' in LLVM could use the type "``{ i32, [0 x float]}``", for
2241 The structure type is used to represent a collection of data members
2242 together in memory. The elements of a structure may be any type that has
2245 Structures in memory are accessed using '``load``' and '``store``' by
2246 getting a pointer to a field with the '``getelementptr``' instruction.
2247 Structures in registers are accessed using the '``extractvalue``' and
2248 '``insertvalue``' instructions.
2250 Structures may optionally be "packed" structures, which indicate that
2251 the alignment of the struct is one byte, and that there is no padding
2252 between the elements. In non-packed structs, padding between field types
2253 is inserted as defined by the DataLayout string in the module, which is
2254 required to match what the underlying code generator expects.
2256 Structures can either be "literal" or "identified". A literal structure
2257 is defined inline with other types (e.g. ``{i32, i32}*``) whereas
2258 identified types are always defined at the top level with a name.
2259 Literal types are uniqued by their contents and can never be recursive
2260 or opaque since there is no way to write one. Identified types can be
2261 recursive, can be opaqued, and are never uniqued.
2267 %T1 = type { <type list> } ; Identified normal struct type
2268 %T2 = type <{ <type list> }> ; Identified packed struct type
2272 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2273 | ``{ i32, i32, i32 }`` | A triple of three ``i32`` values |
2274 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2275 | ``{ 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``. |
2276 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2277 | ``<{ i8, i32 }>`` | A packed struct known to be 5 bytes in size. |
2278 +------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2282 Opaque Structure Types
2283 """"""""""""""""""""""
2287 Opaque structure types are used to represent named structure types that
2288 do not have a body specified. This corresponds (for example) to the C
2289 notion of a forward declared structure.
2300 +--------------+-------------------+
2301 | ``opaque`` | An opaque type. |
2302 +--------------+-------------------+
2309 LLVM has several different basic types of constants. This section
2310 describes them all and their syntax.
2315 **Boolean constants**
2316 The two strings '``true``' and '``false``' are both valid constants
2318 **Integer constants**
2319 Standard integers (such as '4') are constants of the
2320 :ref:`integer <t_integer>` type. Negative numbers may be used with
2322 **Floating point constants**
2323 Floating point constants use standard decimal notation (e.g.
2324 123.421), exponential notation (e.g. 1.23421e+2), or a more precise
2325 hexadecimal notation (see below). The assembler requires the exact
2326 decimal value of a floating-point constant. For example, the
2327 assembler accepts 1.25 but rejects 1.3 because 1.3 is a repeating
2328 decimal in binary. Floating point constants must have a :ref:`floating
2329 point <t_floating>` type.
2330 **Null pointer constants**
2331 The identifier '``null``' is recognized as a null pointer constant
2332 and must be of :ref:`pointer type <t_pointer>`.
2334 The one non-intuitive notation for constants is the hexadecimal form of
2335 floating point constants. For example, the form
2336 '``double 0x432ff973cafa8000``' is equivalent to (but harder to read
2337 than) '``double 4.5e+15``'. The only time hexadecimal floating point
2338 constants are required (and the only time that they are generated by the
2339 disassembler) is when a floating point constant must be emitted but it
2340 cannot be represented as a decimal floating point number in a reasonable
2341 number of digits. For example, NaN's, infinities, and other special
2342 values are represented in their IEEE hexadecimal format so that assembly
2343 and disassembly do not cause any bits to change in the constants.
2345 When using the hexadecimal form, constants of types half, float, and
2346 double are represented using the 16-digit form shown above (which
2347 matches the IEEE754 representation for double); half and float values
2348 must, however, be exactly representable as IEEE 754 half and single
2349 precision, respectively. Hexadecimal format is always used for long
2350 double, and there are three forms of long double. The 80-bit format used
2351 by x86 is represented as ``0xK`` followed by 20 hexadecimal digits. The
2352 128-bit format used by PowerPC (two adjacent doubles) is represented by
2353 ``0xM`` followed by 32 hexadecimal digits. The IEEE 128-bit format is
2354 represented by ``0xL`` followed by 32 hexadecimal digits. Long doubles
2355 will only work if they match the long double format on your target.
2356 The IEEE 16-bit format (half precision) is represented by ``0xH``
2357 followed by 4 hexadecimal digits. All hexadecimal formats are big-endian
2358 (sign bit at the left).
2360 There are no constants of type x86_mmx.
2362 .. _complexconstants:
2367 Complex constants are a (potentially recursive) combination of simple
2368 constants and smaller complex constants.
2370 **Structure constants**
2371 Structure constants are represented with notation similar to
2372 structure type definitions (a comma separated list of elements,
2373 surrounded by braces (``{}``)). For example:
2374 "``{ i32 4, float 17.0, i32* @G }``", where "``@G``" is declared as
2375 "``@G = external global i32``". Structure constants must have
2376 :ref:`structure type <t_struct>`, and the number and types of elements
2377 must match those specified by the type.
2379 Array constants are represented with notation similar to array type
2380 definitions (a comma separated list of elements, surrounded by
2381 square brackets (``[]``)). For example:
2382 "``[ i32 42, i32 11, i32 74 ]``". Array constants must have
2383 :ref:`array type <t_array>`, and the number and types of elements must
2384 match those specified by the type. As a special case, character array
2385 constants may also be represented as a double-quoted string using the ``c``
2386 prefix. For example: "``c"Hello World\0A\00"``".
2387 **Vector constants**
2388 Vector constants are represented with notation similar to vector
2389 type definitions (a comma separated list of elements, surrounded by
2390 less-than/greater-than's (``<>``)). For example:
2391 "``< i32 42, i32 11, i32 74, i32 100 >``". Vector constants
2392 must have :ref:`vector type <t_vector>`, and the number and types of
2393 elements must match those specified by the type.
2394 **Zero initialization**
2395 The string '``zeroinitializer``' can be used to zero initialize a
2396 value to zero of *any* type, including scalar and
2397 :ref:`aggregate <t_aggregate>` types. This is often used to avoid
2398 having to print large zero initializers (e.g. for large arrays) and
2399 is always exactly equivalent to using explicit zero initializers.
2401 A metadata node is a constant tuple without types. For example:
2402 "``!{!0, !{!2, !0}, !"test"}``". Metadata can reference constant values,
2403 for example: "``!{!0, i32 0, i8* @global, i64 (i64)* @function, !"str"}``".
2404 Unlike other typed constants that are meant to be interpreted as part of
2405 the instruction stream, metadata is a place to attach additional
2406 information such as debug info.
2408 Global Variable and Function Addresses
2409 --------------------------------------
2411 The addresses of :ref:`global variables <globalvars>` and
2412 :ref:`functions <functionstructure>` are always implicitly valid
2413 (link-time) constants. These constants are explicitly referenced when
2414 the :ref:`identifier for the global <identifiers>` is used and always have
2415 :ref:`pointer <t_pointer>` type. For example, the following is a legal LLVM
2418 .. code-block:: llvm
2422 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2429 The string '``undef``' can be used anywhere a constant is expected, and
2430 indicates that the user of the value may receive an unspecified
2431 bit-pattern. Undefined values may be of any type (other than '``label``'
2432 or '``void``') and be used anywhere a constant is permitted.
2434 Undefined values are useful because they indicate to the compiler that
2435 the program is well defined no matter what value is used. This gives the
2436 compiler more freedom to optimize. Here are some examples of
2437 (potentially surprising) transformations that are valid (in pseudo IR):
2439 .. code-block:: llvm
2449 This is safe because all of the output bits are affected by the undef
2450 bits. Any output bit can have a zero or one depending on the input bits.
2452 .. code-block:: llvm
2463 These logical operations have bits that are not always affected by the
2464 input. For example, if ``%X`` has a zero bit, then the output of the
2465 '``and``' operation will always be a zero for that bit, no matter what
2466 the corresponding bit from the '``undef``' is. As such, it is unsafe to
2467 optimize or assume that the result of the '``and``' is '``undef``'.
2468 However, it is safe to assume that all bits of the '``undef``' could be
2469 0, and optimize the '``and``' to 0. Likewise, it is safe to assume that
2470 all the bits of the '``undef``' operand to the '``or``' could be set,
2471 allowing the '``or``' to be folded to -1.
2473 .. code-block:: llvm
2475 %A = select undef, %X, %Y
2476 %B = select undef, 42, %Y
2477 %C = select %X, %Y, undef
2487 This set of examples shows that undefined '``select``' (and conditional
2488 branch) conditions can go *either way*, but they have to come from one
2489 of the two operands. In the ``%A`` example, if ``%X`` and ``%Y`` were
2490 both known to have a clear low bit, then ``%A`` would have to have a
2491 cleared low bit. However, in the ``%C`` example, the optimizer is
2492 allowed to assume that the '``undef``' operand could be the same as
2493 ``%Y``, allowing the whole '``select``' to be eliminated.
2495 .. code-block:: llvm
2497 %A = xor undef, undef
2514 This example points out that two '``undef``' operands are not
2515 necessarily the same. This can be surprising to people (and also matches
2516 C semantics) where they assume that "``X^X``" is always zero, even if
2517 ``X`` is undefined. This isn't true for a number of reasons, but the
2518 short answer is that an '``undef``' "variable" can arbitrarily change
2519 its value over its "live range". This is true because the variable
2520 doesn't actually *have a live range*. Instead, the value is logically
2521 read from arbitrary registers that happen to be around when needed, so
2522 the value is not necessarily consistent over time. In fact, ``%A`` and
2523 ``%C`` need to have the same semantics or the core LLVM "replace all
2524 uses with" concept would not hold.
2526 .. code-block:: llvm
2534 These examples show the crucial difference between an *undefined value*
2535 and *undefined behavior*. An undefined value (like '``undef``') is
2536 allowed to have an arbitrary bit-pattern. This means that the ``%A``
2537 operation can be constant folded to '``undef``', because the '``undef``'
2538 could be an SNaN, and ``fdiv`` is not (currently) defined on SNaN's.
2539 However, in the second example, we can make a more aggressive
2540 assumption: because the ``undef`` is allowed to be an arbitrary value,
2541 we are allowed to assume that it could be zero. Since a divide by zero
2542 has *undefined behavior*, we are allowed to assume that the operation
2543 does not execute at all. This allows us to delete the divide and all
2544 code after it. Because the undefined operation "can't happen", the
2545 optimizer can assume that it occurs in dead code.
2547 .. code-block:: llvm
2549 a: store undef -> %X
2550 b: store %X -> undef
2555 These examples reiterate the ``fdiv`` example: a store *of* an undefined
2556 value can be assumed to not have any effect; we can assume that the
2557 value is overwritten with bits that happen to match what was already
2558 there. However, a store *to* an undefined location could clobber
2559 arbitrary memory, therefore, it has undefined behavior.
2566 Poison values are similar to :ref:`undef values <undefvalues>`, however
2567 they also represent the fact that an instruction or constant expression
2568 that cannot evoke side effects has nevertheless detected a condition
2569 that results in undefined behavior.
2571 There is currently no way of representing a poison value in the IR; they
2572 only exist when produced by operations such as :ref:`add <i_add>` with
2575 Poison value behavior is defined in terms of value *dependence*:
2577 - Values other than :ref:`phi <i_phi>` nodes depend on their operands.
2578 - :ref:`Phi <i_phi>` nodes depend on the operand corresponding to
2579 their dynamic predecessor basic block.
2580 - Function arguments depend on the corresponding actual argument values
2581 in the dynamic callers of their functions.
2582 - :ref:`Call <i_call>` instructions depend on the :ref:`ret <i_ret>`
2583 instructions that dynamically transfer control back to them.
2584 - :ref:`Invoke <i_invoke>` instructions depend on the
2585 :ref:`ret <i_ret>`, :ref:`resume <i_resume>`, or exception-throwing
2586 call instructions that dynamically transfer control back to them.
2587 - Non-volatile loads and stores depend on the most recent stores to all
2588 of the referenced memory addresses, following the order in the IR
2589 (including loads and stores implied by intrinsics such as
2590 :ref:`@llvm.memcpy <int_memcpy>`.)
2591 - An instruction with externally visible side effects depends on the
2592 most recent preceding instruction with externally visible side
2593 effects, following the order in the IR. (This includes :ref:`volatile
2594 operations <volatile>`.)
2595 - An instruction *control-depends* on a :ref:`terminator
2596 instruction <terminators>` if the terminator instruction has
2597 multiple successors and the instruction is always executed when
2598 control transfers to one of the successors, and may not be executed
2599 when control is transferred to another.
2600 - Additionally, an instruction also *control-depends* on a terminator
2601 instruction if the set of instructions it otherwise depends on would
2602 be different if the terminator had transferred control to a different
2604 - Dependence is transitive.
2606 Poison values have the same behavior as :ref:`undef values <undefvalues>`,
2607 with the additional effect that any instruction that has a *dependence*
2608 on a poison value has undefined behavior.
2610 Here are some examples:
2612 .. code-block:: llvm
2615 %poison = sub nuw i32 0, 1 ; Results in a poison value.
2616 %still_poison = and i32 %poison, 0 ; 0, but also poison.
2617 %poison_yet_again = getelementptr i32, i32* @h, i32 %still_poison
2618 store i32 0, i32* %poison_yet_again ; memory at @h[0] is poisoned
2620 store i32 %poison, i32* @g ; Poison value stored to memory.
2621 %poison2 = load i32, i32* @g ; Poison value loaded back from memory.
2623 store volatile i32 %poison, i32* @g ; External observation; undefined behavior.
2625 %narrowaddr = bitcast i32* @g to i16*
2626 %wideaddr = bitcast i32* @g to i64*
2627 %poison3 = load i16, i16* %narrowaddr ; Returns a poison value.
2628 %poison4 = load i64, i64* %wideaddr ; Returns a poison value.
2630 %cmp = icmp slt i32 %poison, 0 ; Returns a poison value.
2631 br i1 %cmp, label %true, label %end ; Branch to either destination.
2634 store volatile i32 0, i32* @g ; This is control-dependent on %cmp, so
2635 ; it has undefined behavior.
2639 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2640 ; Both edges into this PHI are
2641 ; control-dependent on %cmp, so this
2642 ; always results in a poison value.
2644 store volatile i32 0, i32* @g ; This would depend on the store in %true
2645 ; if %cmp is true, or the store in %entry
2646 ; otherwise, so this is undefined behavior.
2648 br i1 %cmp, label %second_true, label %second_end
2649 ; The same branch again, but this time the
2650 ; true block doesn't have side effects.
2657 store volatile i32 0, i32* @g ; This time, the instruction always depends
2658 ; on the store in %end. Also, it is
2659 ; control-equivalent to %end, so this is
2660 ; well-defined (ignoring earlier undefined
2661 ; behavior in this example).
2665 Addresses of Basic Blocks
2666 -------------------------
2668 ``blockaddress(@function, %block)``
2670 The '``blockaddress``' constant computes the address of the specified
2671 basic block in the specified function, and always has an ``i8*`` type.
2672 Taking the address of the entry block is illegal.
2674 This value only has defined behavior when used as an operand to the
2675 ':ref:`indirectbr <i_indirectbr>`' instruction, or for comparisons
2676 against null. Pointer equality tests between labels addresses results in
2677 undefined behavior --- though, again, comparison against null is ok, and
2678 no label is equal to the null pointer. This may be passed around as an
2679 opaque pointer sized value as long as the bits are not inspected. This
2680 allows ``ptrtoint`` and arithmetic to be performed on these values so
2681 long as the original value is reconstituted before the ``indirectbr``
2684 Finally, some targets may provide defined semantics when using the value
2685 as the operand to an inline assembly, but that is target specific.
2689 Constant Expressions
2690 --------------------
2692 Constant expressions are used to allow expressions involving other
2693 constants to be used as constants. Constant expressions may be of any
2694 :ref:`first class <t_firstclass>` type and may involve any LLVM operation
2695 that does not have side effects (e.g. load and call are not supported).
2696 The following is the syntax for constant expressions:
2698 ``trunc (CST to TYPE)``
2699 Truncate a constant to another type. The bit size of CST must be
2700 larger than the bit size of TYPE. Both types must be integers.
2701 ``zext (CST to TYPE)``
2702 Zero extend a constant to another type. The bit size of CST must be
2703 smaller than the bit size of TYPE. Both types must be integers.
2704 ``sext (CST to TYPE)``
2705 Sign extend a constant to another type. The bit size of CST must be
2706 smaller than the bit size of TYPE. Both types must be integers.
2707 ``fptrunc (CST to TYPE)``
2708 Truncate a floating point constant to another floating point type.
2709 The size of CST must be larger than the size of TYPE. Both types
2710 must be floating point.
2711 ``fpext (CST to TYPE)``
2712 Floating point extend a constant to another type. The size of CST
2713 must be smaller or equal to the size of TYPE. Both types must be
2715 ``fptoui (CST to TYPE)``
2716 Convert a floating point constant to the corresponding unsigned
2717 integer constant. TYPE must be a scalar or vector integer type. CST
2718 must be of scalar or vector floating point type. Both CST and TYPE
2719 must be scalars, or vectors of the same number of elements. If the
2720 value won't fit in the integer type, the results are undefined.
2721 ``fptosi (CST to TYPE)``
2722 Convert a floating point constant to the corresponding signed
2723 integer constant. TYPE must be a scalar or vector integer type. CST
2724 must be of scalar or vector floating point type. Both CST and TYPE
2725 must be scalars, or vectors of the same number of elements. If the
2726 value won't fit in the integer type, the results are undefined.
2727 ``uitofp (CST to TYPE)``
2728 Convert an unsigned integer constant to the corresponding floating
2729 point constant. TYPE must be a scalar or vector floating point type.
2730 CST must be of scalar or vector integer type. Both CST and TYPE must
2731 be scalars, or vectors of the same number of elements. If the value
2732 won't fit in the floating point type, the results are undefined.
2733 ``sitofp (CST to TYPE)``
2734 Convert a signed integer constant to the corresponding floating
2735 point constant. TYPE must be a scalar or vector floating point type.
2736 CST must be of scalar or vector integer type. Both CST and TYPE must
2737 be scalars, or vectors of the same number of elements. If the value
2738 won't fit in the floating point type, the results are undefined.
2739 ``ptrtoint (CST to TYPE)``
2740 Convert a pointer typed constant to the corresponding integer
2741 constant. ``TYPE`` must be an integer type. ``CST`` must be of
2742 pointer type. The ``CST`` value is zero extended, truncated, or
2743 unchanged to make it fit in ``TYPE``.
2744 ``inttoptr (CST to TYPE)``
2745 Convert an integer constant to a pointer constant. TYPE must be a
2746 pointer type. CST must be of integer type. The CST value is zero
2747 extended, truncated, or unchanged to make it fit in a pointer size.
2748 This one is *really* dangerous!
2749 ``bitcast (CST to TYPE)``
2750 Convert a constant, CST, to another TYPE. The constraints of the
2751 operands are the same as those for the :ref:`bitcast
2752 instruction <i_bitcast>`.
2753 ``addrspacecast (CST to TYPE)``
2754 Convert a constant pointer or constant vector of pointer, CST, to another
2755 TYPE in a different address space. The constraints of the operands are the
2756 same as those for the :ref:`addrspacecast instruction <i_addrspacecast>`.
2757 ``getelementptr (TY, CSTPTR, IDX0, IDX1, ...)``, ``getelementptr inbounds (TY, CSTPTR, IDX0, IDX1, ...)``
2758 Perform the :ref:`getelementptr operation <i_getelementptr>` on
2759 constants. As with the :ref:`getelementptr <i_getelementptr>`
2760 instruction, the index list may have zero or more indexes, which are
2761 required to make sense for the type of "pointer to TY".
2762 ``select (COND, VAL1, VAL2)``
2763 Perform the :ref:`select operation <i_select>` on constants.
2764 ``icmp COND (VAL1, VAL2)``
2765 Performs the :ref:`icmp operation <i_icmp>` on constants.
2766 ``fcmp COND (VAL1, VAL2)``
2767 Performs the :ref:`fcmp operation <i_fcmp>` on constants.
2768 ``extractelement (VAL, IDX)``
2769 Perform the :ref:`extractelement operation <i_extractelement>` on
2771 ``insertelement (VAL, ELT, IDX)``
2772 Perform the :ref:`insertelement operation <i_insertelement>` on
2774 ``shufflevector (VEC1, VEC2, IDXMASK)``
2775 Perform the :ref:`shufflevector operation <i_shufflevector>` on
2777 ``extractvalue (VAL, IDX0, IDX1, ...)``
2778 Perform the :ref:`extractvalue operation <i_extractvalue>` on
2779 constants. The index list is interpreted in a similar manner as
2780 indices in a ':ref:`getelementptr <i_getelementptr>`' operation. At
2781 least one index value must be specified.
2782 ``insertvalue (VAL, ELT, IDX0, IDX1, ...)``
2783 Perform the :ref:`insertvalue operation <i_insertvalue>` on constants.
2784 The index list is interpreted in a similar manner as indices in a
2785 ':ref:`getelementptr <i_getelementptr>`' operation. At least one index
2786 value must be specified.
2787 ``OPCODE (LHS, RHS)``
2788 Perform the specified operation of the LHS and RHS constants. OPCODE
2789 may be any of the :ref:`binary <binaryops>` or :ref:`bitwise
2790 binary <bitwiseops>` operations. The constraints on operands are
2791 the same as those for the corresponding instruction (e.g. no bitwise
2792 operations on floating point values are allowed).
2799 Inline Assembler Expressions
2800 ----------------------------
2802 LLVM supports inline assembler expressions (as opposed to :ref:`Module-Level
2803 Inline Assembly <moduleasm>`) through the use of a special value. This value
2804 represents the inline assembler as a template string (containing the
2805 instructions to emit), a list of operand constraints (stored as a string), a
2806 flag that indicates whether or not the inline asm expression has side effects,
2807 and a flag indicating whether the function containing the asm needs to align its
2808 stack conservatively.
2810 The template string supports argument substitution of the operands using "``$``"
2811 followed by a number, to indicate substitution of the given register/memory
2812 location, as specified by the constraint string. "``${NUM:MODIFIER}``" may also
2813 be used, where ``MODIFIER`` is a target-specific annotation for how to print the
2814 operand (See :ref:`inline-asm-modifiers`).
2816 A literal "``$``" may be included by using "``$$``" in the template. To include
2817 other special characters into the output, the usual "``\XX``" escapes may be
2818 used, just as in other strings. Note that after template substitution, the
2819 resulting assembly string is parsed by LLVM's integrated assembler unless it is
2820 disabled -- even when emitting a ``.s`` file -- and thus must contain assembly
2821 syntax known to LLVM.
2823 LLVM's support for inline asm is modeled closely on the requirements of Clang's
2824 GCC-compatible inline-asm support. Thus, the feature-set and the constraint and
2825 modifier codes listed here are similar or identical to those in GCC's inline asm
2826 support. However, to be clear, the syntax of the template and constraint strings
2827 described here is *not* the same as the syntax accepted by GCC and Clang, and,
2828 while most constraint letters are passed through as-is by Clang, some get
2829 translated to other codes when converting from the C source to the LLVM
2832 An example inline assembler expression is:
2834 .. code-block:: llvm
2836 i32 (i32) asm "bswap $0", "=r,r"
2838 Inline assembler expressions may **only** be used as the callee operand
2839 of a :ref:`call <i_call>` or an :ref:`invoke <i_invoke>` instruction.
2840 Thus, typically we have:
2842 .. code-block:: llvm
2844 %X = call i32 asm "bswap $0", "=r,r"(i32 %Y)
2846 Inline asms with side effects not visible in the constraint list must be
2847 marked as having side effects. This is done through the use of the
2848 '``sideeffect``' keyword, like so:
2850 .. code-block:: llvm
2852 call void asm sideeffect "eieio", ""()
2854 In some cases inline asms will contain code that will not work unless
2855 the stack is aligned in some way, such as calls or SSE instructions on
2856 x86, yet will not contain code that does that alignment within the asm.
2857 The compiler should make conservative assumptions about what the asm
2858 might contain and should generate its usual stack alignment code in the
2859 prologue if the '``alignstack``' keyword is present:
2861 .. code-block:: llvm
2863 call void asm alignstack "eieio", ""()
2865 Inline asms also support using non-standard assembly dialects. The
2866 assumed dialect is ATT. When the '``inteldialect``' keyword is present,
2867 the inline asm is using the Intel dialect. Currently, ATT and Intel are
2868 the only supported dialects. An example is:
2870 .. code-block:: llvm
2872 call void asm inteldialect "eieio", ""()
2874 If multiple keywords appear the '``sideeffect``' keyword must come
2875 first, the '``alignstack``' keyword second and the '``inteldialect``'
2878 Inline Asm Constraint String
2879 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2881 The constraint list is a comma-separated string, each element containing one or
2882 more constraint codes.
2884 For each element in the constraint list an appropriate register or memory
2885 operand will be chosen, and it will be made available to assembly template
2886 string expansion as ``$0`` for the first constraint in the list, ``$1`` for the
2889 There are three different types of constraints, which are distinguished by a
2890 prefix symbol in front of the constraint code: Output, Input, and Clobber. The
2891 constraints must always be given in that order: outputs first, then inputs, then
2892 clobbers. They cannot be intermingled.
2894 There are also three different categories of constraint codes:
2896 - Register constraint. This is either a register class, or a fixed physical
2897 register. This kind of constraint will allocate a register, and if necessary,
2898 bitcast the argument or result to the appropriate type.
2899 - Memory constraint. This kind of constraint is for use with an instruction
2900 taking a memory operand. Different constraints allow for different addressing
2901 modes used by the target.
2902 - Immediate value constraint. This kind of constraint is for an integer or other
2903 immediate value which can be rendered directly into an instruction. The
2904 various target-specific constraints allow the selection of a value in the
2905 proper range for the instruction you wish to use it with.
2910 Output constraints are specified by an "``=``" prefix (e.g. "``=r``"). This
2911 indicates that the assembly will write to this operand, and the operand will
2912 then be made available as a return value of the ``asm`` expression. Output
2913 constraints do not consume an argument from the call instruction. (Except, see
2914 below about indirect outputs).
2916 Normally, it is expected that no output locations are written to by the assembly
2917 expression until *all* of the inputs have been read. As such, LLVM may assign
2918 the same register to an output and an input. If this is not safe (e.g. if the
2919 assembly contains two instructions, where the first writes to one output, and
2920 the second reads an input and writes to a second output), then the "``&``"
2921 modifier must be used (e.g. "``=&r``") to specify that the output is an
2922 "early-clobber" output. Marking an ouput as "early-clobber" ensures that LLVM
2923 will not use the same register for any inputs (other than an input tied to this
2929 Input constraints do not have a prefix -- just the constraint codes. Each input
2930 constraint will consume one argument from the call instruction. It is not
2931 permitted for the asm to write to any input register or memory location (unless
2932 that input is tied to an output). Note also that multiple inputs may all be
2933 assigned to the same register, if LLVM can determine that they necessarily all
2934 contain the same value.
2936 Instead of providing a Constraint Code, input constraints may also "tie"
2937 themselves to an output constraint, by providing an integer as the constraint
2938 string. Tied inputs still consume an argument from the call instruction, and
2939 take up a position in the asm template numbering as is usual -- they will simply
2940 be constrained to always use the same register as the output they've been tied
2941 to. For example, a constraint string of "``=r,0``" says to assign a register for
2942 output, and use that register as an input as well (it being the 0'th
2945 It is permitted to tie an input to an "early-clobber" output. In that case, no
2946 *other* input may share the same register as the input tied to the early-clobber
2947 (even when the other input has the same value).
2949 You may only tie an input to an output which has a register constraint, not a
2950 memory constraint. Only a single input may be tied to an output.
2952 There is also an "interesting" feature which deserves a bit of explanation: if a
2953 register class constraint allocates a register which is too small for the value
2954 type operand provided as input, the input value will be split into multiple
2955 registers, and all of them passed to the inline asm.
2957 However, this feature is often not as useful as you might think.
2959 Firstly, the registers are *not* guaranteed to be consecutive. So, on those
2960 architectures that have instructions which operate on multiple consecutive
2961 instructions, this is not an appropriate way to support them. (e.g. the 32-bit
2962 SparcV8 has a 64-bit load, which instruction takes a single 32-bit register. The
2963 hardware then loads into both the named register, and the next register. This
2964 feature of inline asm would not be useful to support that.)
2966 A few of the targets provide a template string modifier allowing explicit access
2967 to the second register of a two-register operand (e.g. MIPS ``L``, ``M``, and
2968 ``D``). On such an architecture, you can actually access the second allocated
2969 register (yet, still, not any subsequent ones). But, in that case, you're still
2970 probably better off simply splitting the value into two separate operands, for
2971 clarity. (e.g. see the description of the ``A`` constraint on X86, which,
2972 despite existing only for use with this feature, is not really a good idea to
2975 Indirect inputs and outputs
2976 """""""""""""""""""""""""""
2978 Indirect output or input constraints can be specified by the "``*``" modifier
2979 (which goes after the "``=``" in case of an output). This indicates that the asm
2980 will write to or read from the contents of an *address* provided as an input
2981 argument. (Note that in this way, indirect outputs act more like an *input* than
2982 an output: just like an input, they consume an argument of the call expression,
2983 rather than producing a return value. An indirect output constraint is an
2984 "output" only in that the asm is expected to write to the contents of the input
2985 memory location, instead of just read from it).
2987 This is most typically used for memory constraint, e.g. "``=*m``", to pass the
2988 address of a variable as a value.
2990 It is also possible to use an indirect *register* constraint, but only on output
2991 (e.g. "``=*r``"). This will cause LLVM to allocate a register for an output
2992 value normally, and then, separately emit a store to the address provided as
2993 input, after the provided inline asm. (It's not clear what value this
2994 functionality provides, compared to writing the store explicitly after the asm
2995 statement, and it can only produce worse code, since it bypasses many
2996 optimization passes. I would recommend not using it.)
3002 A clobber constraint is indicated by a "``~``" prefix. A clobber does not
3003 consume an input operand, nor generate an output. Clobbers cannot use any of the
3004 general constraint code letters -- they may use only explicit register
3005 constraints, e.g. "``~{eax}``". The one exception is that a clobber string of
3006 "``~{memory}``" indicates that the assembly writes to arbitrary undeclared
3007 memory locations -- not only the memory pointed to by a declared indirect
3013 After a potential prefix comes constraint code, or codes.
3015 A Constraint Code is either a single letter (e.g. "``r``"), a "``^``" character
3016 followed by two letters (e.g. "``^wc``"), or "``{``" register-name "``}``"
3019 The one and two letter constraint codes are typically chosen to be the same as
3020 GCC's constraint codes.
3022 A single constraint may include one or more than constraint code in it, leaving
3023 it up to LLVM to choose which one to use. This is included mainly for
3024 compatibility with the translation of GCC inline asm coming from clang.
3026 There are two ways to specify alternatives, and either or both may be used in an
3027 inline asm constraint list:
3029 1) Append the codes to each other, making a constraint code set. E.g. "``im``"
3030 or "``{eax}m``". This means "choose any of the options in the set". The
3031 choice of constraint is made independently for each constraint in the
3034 2) Use "``|``" between constraint code sets, creating alternatives. Every
3035 constraint in the constraint list must have the same number of alternative
3036 sets. With this syntax, the same alternative in *all* of the items in the
3037 constraint list will be chosen together.
3039 Putting those together, you might have a two operand constraint string like
3040 ``"rm|r,ri|rm"``. This indicates that if operand 0 is ``r`` or ``m``, then
3041 operand 1 may be one of ``r`` or ``i``. If operand 0 is ``r``, then operand 1
3042 may be one of ``r`` or ``m``. But, operand 0 and 1 cannot both be of type m.
3044 However, the use of either of the alternatives features is *NOT* recommended, as
3045 LLVM is not able to make an intelligent choice about which one to use. (At the
3046 point it currently needs to choose, not enough information is available to do so
3047 in a smart way.) Thus, it simply tries to make a choice that's most likely to
3048 compile, not one that will be optimal performance. (e.g., given "``rm``", it'll
3049 always choose to use memory, not registers). And, if given multiple registers,
3050 or multiple register classes, it will simply choose the first one. (In fact, it
3051 doesn't currently even ensure explicitly specified physical registers are
3052 unique, so specifying multiple physical registers as alternatives, like
3053 ``{r11}{r12},{r11}{r12}``, will assign r11 to both operands, not at all what was
3056 Supported Constraint Code List
3057 """"""""""""""""""""""""""""""
3059 The constraint codes are, in general, expected to behave the same way they do in
3060 GCC. LLVM's support is often implemented on an 'as-needed' basis, to support C
3061 inline asm code which was supported by GCC. A mismatch in behavior between LLVM
3062 and GCC likely indicates a bug in LLVM.
3064 Some constraint codes are typically supported by all targets:
3066 - ``r``: A register in the target's general purpose register class.
3067 - ``m``: A memory address operand. It is target-specific what addressing modes
3068 are supported, typical examples are register, or register + register offset,
3069 or register + immediate offset (of some target-specific size).
3070 - ``i``: An integer constant (of target-specific width). Allows either a simple
3071 immediate, or a relocatable value.
3072 - ``n``: An integer constant -- *not* including relocatable values.
3073 - ``s``: An integer constant, but allowing *only* relocatable values.
3074 - ``X``: Allows an operand of any kind, no constraint whatsoever. Typically
3075 useful to pass a label for an asm branch or call.
3077 .. FIXME: but that surely isn't actually okay to jump out of an asm
3078 block without telling llvm about the control transfer???)
3080 - ``{register-name}``: Requires exactly the named physical register.
3082 Other constraints are target-specific:
3086 - ``z``: An immediate integer 0. Outputs ``WZR`` or ``XZR``, as appropriate.
3087 - ``I``: An immediate integer valid for an ``ADD`` or ``SUB`` instruction,
3088 i.e. 0 to 4095 with optional shift by 12.
3089 - ``J``: An immediate integer that, when negated, is valid for an ``ADD`` or
3090 ``SUB`` instruction, i.e. -1 to -4095 with optional left shift by 12.
3091 - ``K``: An immediate integer that is valid for the 'bitmask immediate 32' of a
3092 logical instruction like ``AND``, ``EOR``, or ``ORR`` with a 32-bit register.
3093 - ``L``: An immediate integer that is valid for the 'bitmask immediate 64' of a
3094 logical instruction like ``AND``, ``EOR``, or ``ORR`` with a 64-bit register.
3095 - ``M``: An immediate integer for use with the ``MOV`` assembly alias on a
3096 32-bit register. This is a superset of ``K``: in addition to the bitmask
3097 immediate, also allows immediate integers which can be loaded with a single
3098 ``MOVZ`` or ``MOVL`` instruction.
3099 - ``N``: An immediate integer for use with the ``MOV`` assembly alias on a
3100 64-bit register. This is a superset of ``L``.
3101 - ``Q``: Memory address operand must be in a single register (no
3102 offsets). (However, LLVM currently does this for the ``m`` constraint as
3104 - ``r``: A 32 or 64-bit integer register (W* or X*).
3105 - ``w``: A 32, 64, or 128-bit floating-point/SIMD register.
3106 - ``x``: A lower 128-bit floating-point/SIMD register (``V0`` to ``V15``).
3110 - ``r``: A 32 or 64-bit integer register.
3111 - ``[0-9]v``: The 32-bit VGPR register, number 0-9.
3112 - ``[0-9]s``: The 32-bit SGPR register, number 0-9.
3117 - ``Q``, ``Um``, ``Un``, ``Uq``, ``Us``, ``Ut``, ``Uv``, ``Uy``: Memory address
3118 operand. Treated the same as operand ``m``, at the moment.
3120 ARM and ARM's Thumb2 mode:
3122 - ``j``: An immediate integer between 0 and 65535 (valid for ``MOVW``)
3123 - ``I``: An immediate integer valid for a data-processing instruction.
3124 - ``J``: An immediate integer between -4095 and 4095.
3125 - ``K``: An immediate integer whose bitwise inverse is valid for a
3126 data-processing instruction. (Can be used with template modifier "``B``" to
3127 print the inverted value).
3128 - ``L``: An immediate integer whose negation is valid for a data-processing
3129 instruction. (Can be used with template modifier "``n``" to print the negated
3131 - ``M``: A power of two or a integer between 0 and 32.
3132 - ``N``: Invalid immediate constraint.
3133 - ``O``: Invalid immediate constraint.
3134 - ``r``: A general-purpose 32-bit integer register (``r0-r15``).
3135 - ``l``: In Thumb2 mode, low 32-bit GPR registers (``r0-r7``). In ARM mode, same
3137 - ``h``: In Thumb2 mode, a high 32-bit GPR register (``r8-r15``). In ARM mode,
3139 - ``w``: A 32, 64, or 128-bit floating-point/SIMD register: ``s0-s31``,
3140 ``d0-d31``, or ``q0-q15``.
3141 - ``x``: A 32, 64, or 128-bit floating-point/SIMD register: ``s0-s15``,
3142 ``d0-d7``, or ``q0-q3``.
3143 - ``t``: A floating-point/SIMD register, only supports 32-bit values:
3148 - ``I``: An immediate integer between 0 and 255.
3149 - ``J``: An immediate integer between -255 and -1.
3150 - ``K``: An immediate integer between 0 and 255, with optional left-shift by
3152 - ``L``: An immediate integer between -7 and 7.
3153 - ``M``: An immediate integer which is a multiple of 4 between 0 and 1020.
3154 - ``N``: An immediate integer between 0 and 31.
3155 - ``O``: An immediate integer which is a multiple of 4 between -508 and 508.
3156 - ``r``: A low 32-bit GPR register (``r0-r7``).
3157 - ``l``: A low 32-bit GPR register (``r0-r7``).
3158 - ``h``: A high GPR register (``r0-r7``).
3159 - ``w``: A 32, 64, or 128-bit floating-point/SIMD register: ``s0-s31``,
3160 ``d0-d31``, or ``q0-q15``.
3161 - ``x``: A 32, 64, or 128-bit floating-point/SIMD register: ``s0-s15``,
3162 ``d0-d7``, or ``q0-q3``.
3163 - ``t``: A floating-point/SIMD register, only supports 32-bit values:
3169 - ``o``, ``v``: A memory address operand, treated the same as constraint ``m``,
3171 - ``r``: A 32 or 64-bit register.
3175 - ``r``: An 8 or 16-bit register.
3179 - ``I``: An immediate signed 16-bit integer.
3180 - ``J``: An immediate integer zero.
3181 - ``K``: An immediate unsigned 16-bit integer.
3182 - ``L``: An immediate 32-bit integer, where the lower 16 bits are 0.
3183 - ``N``: An immediate integer between -65535 and -1.
3184 - ``O``: An immediate signed 15-bit integer.
3185 - ``P``: An immediate integer between 1 and 65535.
3186 - ``m``: A memory address operand. In MIPS-SE mode, allows a base address
3187 register plus 16-bit immediate offset. In MIPS mode, just a base register.
3188 - ``R``: A memory address operand. In MIPS-SE mode, allows a base address
3189 register plus a 9-bit signed offset. In MIPS mode, the same as constraint
3191 - ``ZC``: A memory address operand, suitable for use in a ``pref``, ``ll``, or
3192 ``sc`` instruction on the given subtarget (details vary).
3193 - ``r``, ``d``, ``y``: A 32 or 64-bit GPR register.
3194 - ``f``: A 32 or 64-bit FPU register (``F0-F31``), or a 128-bit MSA register
3196 - ``c``: A 32-bit or 64-bit GPR register suitable for indirect jump (always
3198 - ``l``: The ``lo`` register, 32 or 64-bit.
3203 - ``b``: A 1-bit integer register.
3204 - ``c`` or ``h``: A 16-bit integer register.
3205 - ``r``: A 32-bit integer register.
3206 - ``l`` or ``N``: A 64-bit integer register.
3207 - ``f``: A 32-bit float register.
3208 - ``d``: A 64-bit float register.
3213 - ``I``: An immediate signed 16-bit integer.
3214 - ``J``: An immediate unsigned 16-bit integer, shifted left 16 bits.
3215 - ``K``: An immediate unsigned 16-bit integer.
3216 - ``L``: An immediate signed 16-bit integer, shifted left 16 bits.
3217 - ``M``: An immediate integer greater than 31.
3218 - ``N``: An immediate integer that is an exact power of 2.
3219 - ``O``: The immediate integer constant 0.
3220 - ``P``: An immediate integer constant whose negation is a signed 16-bit
3222 - ``es``, ``o``, ``Q``, ``Z``, ``Zy``: A memory address operand, currently
3223 treated the same as ``m``.
3224 - ``r``: A 32 or 64-bit integer register.
3225 - ``b``: A 32 or 64-bit integer register, excluding ``R0`` (that is:
3227 - ``f``: A 32 or 64-bit float register (``F0-F31``), or when QPX is enabled, a
3228 128 or 256-bit QPX register (``Q0-Q31``; aliases the ``F`` registers).
3229 - ``v``: For ``4 x f32`` or ``4 x f64`` types, when QPX is enabled, a
3230 128 or 256-bit QPX register (``Q0-Q31``), otherwise a 128-bit
3231 altivec vector register (``V0-V31``).
3233 .. FIXME: is this a bug that v accepts QPX registers? I think this
3234 is supposed to only use the altivec vector registers?
3236 - ``y``: Condition register (``CR0-CR7``).
3237 - ``wc``: An individual CR bit in a CR register.
3238 - ``wa``, ``wd``, ``wf``: Any 128-bit VSX vector register, from the full VSX
3239 register set (overlapping both the floating-point and vector register files).
3240 - ``ws``: A 32 or 64-bit floating point register, from the full VSX register
3245 - ``I``: An immediate 13-bit signed integer.
3246 - ``r``: A 32-bit integer register.
3250 - ``I``: An immediate unsigned 8-bit integer.
3251 - ``J``: An immediate unsigned 12-bit integer.
3252 - ``K``: An immediate signed 16-bit integer.
3253 - ``L``: An immediate signed 20-bit integer.
3254 - ``M``: An immediate integer 0x7fffffff.
3255 - ``Q``, ``R``, ``S``, ``T``: A memory address operand, treated the same as
3256 ``m``, at the moment.
3257 - ``r`` or ``d``: A 32, 64, or 128-bit integer register.
3258 - ``a``: A 32, 64, or 128-bit integer address register (excludes R0, which in an
3259 address context evaluates as zero).
3260 - ``h``: A 32-bit value in the high part of a 64bit data register
3262 - ``f``: A 32, 64, or 128-bit floating point register.
3266 - ``I``: An immediate integer between 0 and 31.
3267 - ``J``: An immediate integer between 0 and 64.
3268 - ``K``: An immediate signed 8-bit integer.
3269 - ``L``: An immediate integer, 0xff or 0xffff or (in 64-bit mode only)
3271 - ``M``: An immediate integer between 0 and 3.
3272 - ``N``: An immediate unsigned 8-bit integer.
3273 - ``O``: An immediate integer between 0 and 127.
3274 - ``e``: An immediate 32-bit signed integer.
3275 - ``Z``: An immediate 32-bit unsigned integer.
3276 - ``o``, ``v``: Treated the same as ``m``, at the moment.
3277 - ``q``: An 8, 16, 32, or 64-bit register which can be accessed as an 8-bit
3278 ``l`` integer register. On X86-32, this is the ``a``, ``b``, ``c``, and ``d``
3279 registers, and on X86-64, it is all of the integer registers.
3280 - ``Q``: An 8, 16, 32, or 64-bit register which can be accessed as an 8-bit
3281 ``h`` integer register. This is the ``a``, ``b``, ``c``, and ``d`` registers.
3282 - ``r`` or ``l``: An 8, 16, 32, or 64-bit integer register.
3283 - ``R``: An 8, 16, 32, or 64-bit "legacy" integer register -- one which has
3284 existed since i386, and can be accessed without the REX prefix.
3285 - ``f``: A 32, 64, or 80-bit '387 FPU stack pseudo-register.
3286 - ``y``: A 64-bit MMX register, if MMX is enabled.
3287 - ``x``: If SSE is enabled: a 32 or 64-bit scalar operand, or 128-bit vector
3288 operand in a SSE register. If AVX is also enabled, can also be a 256-bit
3289 vector operand in an AVX register. If AVX-512 is also enabled, can also be a
3290 512-bit vector operand in an AVX512 register, Otherwise, an error.
3291 - ``Y``: The same as ``x``, if *SSE2* is enabled, otherwise an error.
3292 - ``A``: Special case: allocates EAX first, then EDX, for a single operand (in
3293 32-bit mode, a 64-bit integer operand will get split into two registers). It
3294 is not recommended to use this constraint, as in 64-bit mode, the 64-bit
3295 operand will get allocated only to RAX -- if two 32-bit operands are needed,
3296 you're better off splitting it yourself, before passing it to the asm
3301 - ``r``: A 32-bit integer register.
3304 .. _inline-asm-modifiers:
3306 Asm template argument modifiers
3307 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
3309 In the asm template string, modifiers can be used on the operand reference, like
3312 The modifiers are, in general, expected to behave the same way they do in
3313 GCC. LLVM's support is often implemented on an 'as-needed' basis, to support C
3314 inline asm code which was supported by GCC. A mismatch in behavior between LLVM
3315 and GCC likely indicates a bug in LLVM.
3319 - ``c``: Print an immediate integer constant unadorned, without
3320 the target-specific immediate punctuation (e.g. no ``$`` prefix).
3321 - ``n``: Negate and print immediate integer constant unadorned, without the
3322 target-specific immediate punctuation (e.g. no ``$`` prefix).
3323 - ``l``: Print as an unadorned label, without the target-specific label
3324 punctuation (e.g. no ``$`` prefix).
3328 - ``w``: Print a GPR register with a ``w*`` name instead of ``x*`` name. E.g.,
3329 instead of ``x30``, print ``w30``.
3330 - ``x``: Print a GPR register with a ``x*`` name. (this is the default, anyhow).
3331 - ``b``, ``h``, ``s``, ``d``, ``q``: Print a floating-point/SIMD register with a
3332 ``b*``, ``h*``, ``s*``, ``d*``, or ``q*`` name, rather than the default of
3341 - ``a``: Print an operand as an address (with ``[`` and ``]`` surrounding a
3345 - ``y``: Print a VFP single-precision register as an indexed double (e.g. print
3346 as ``d4[1]`` instead of ``s9``)
3347 - ``B``: Bitwise invert and print an immediate integer constant without ``#``
3349 - ``L``: Print the low 16-bits of an immediate integer constant.
3350 - ``M``: Print as a register set suitable for ldm/stm. Also prints *all*
3351 register operands subsequent to the specified one (!), so use carefully.
3352 - ``Q``: Print the low-order register of a register-pair, or the low-order
3353 register of a two-register operand.
3354 - ``R``: Print the high-order register of a register-pair, or the high-order
3355 register of a two-register operand.
3356 - ``H``: Print the second register of a register-pair. (On a big-endian system,
3357 ``H`` is equivalent to ``Q``, and on little-endian system, ``H`` is equivalent
3360 .. FIXME: H doesn't currently support printing the second register
3361 of a two-register operand.
3363 - ``e``: Print the low doubleword register of a NEON quad register.
3364 - ``f``: Print the high doubleword register of a NEON quad register.
3365 - ``m``: Print the base register of a memory operand without the ``[`` and ``]``
3370 - ``L``: Print the second register of a two-register operand. Requires that it
3371 has been allocated consecutively to the first.
3373 .. FIXME: why is it restricted to consecutive ones? And there's
3374 nothing that ensures that happens, is there?
3376 - ``I``: Print the letter 'i' if the operand is an integer constant, otherwise
3377 nothing. Used to print 'addi' vs 'add' instructions.
3381 No additional modifiers.
3385 - ``X``: Print an immediate integer as hexadecimal
3386 - ``x``: Print the low 16 bits of an immediate integer as hexadecimal.
3387 - ``d``: Print an immediate integer as decimal.
3388 - ``m``: Subtract one and print an immediate integer as decimal.
3389 - ``z``: Print $0 if an immediate zero, otherwise print normally.
3390 - ``L``: Print the low-order register of a two-register operand, or prints the
3391 address of the low-order word of a double-word memory operand.
3393 .. FIXME: L seems to be missing memory operand support.
3395 - ``M``: Print the high-order register of a two-register operand, or prints the
3396 address of the high-order word of a double-word memory operand.
3398 .. FIXME: M seems to be missing memory operand support.
3400 - ``D``: Print the second register of a two-register operand, or prints the
3401 second word of a double-word memory operand. (On a big-endian system, ``D`` is
3402 equivalent to ``L``, and on little-endian system, ``D`` is equivalent to
3412 - ``L``: Print the second register of a two-register operand. Requires that it
3413 has been allocated consecutively to the first.
3415 .. FIXME: why is it restricted to consecutive ones? And there's
3416 nothing that ensures that happens, is there?
3418 - ``I``: Print the letter 'i' if the operand is an integer constant, otherwise
3419 nothing. Used to print 'addi' vs 'add' instructions.
3420 - ``y``: For a memory operand, prints formatter for a two-register X-form
3421 instruction. (Currently always prints ``r0,OPERAND``).
3422 - ``U``: Prints 'u' if the memory operand is an update form, and nothing
3423 otherwise. (NOTE: LLVM does not support update form, so this will currently
3424 always print nothing)
3425 - ``X``: Prints 'x' if the memory operand is an indexed form. (NOTE: LLVM does
3426 not support indexed form, so this will currently always print nothing)
3434 SystemZ implements only ``n``, and does *not* support any of the other
3435 target-independent modifiers.
3439 - ``c``: Print an unadorned integer or symbol name. (The latter is
3440 target-specific behavior for this typically target-independent modifier).
3441 - ``A``: Print a register name with a '``*``' before it.
3442 - ``b``: Print an 8-bit register name (e.g. ``al``); do nothing on a memory
3444 - ``h``: Print the upper 8-bit register name (e.g. ``ah``); do nothing on a
3446 - ``w``: Print the 16-bit register name (e.g. ``ax``); do nothing on a memory
3448 - ``k``: Print the 32-bit register name (e.g. ``eax``); do nothing on a memory
3450 - ``q``: Print the 64-bit register name (e.g. ``rax``), if 64-bit registers are
3451 available, otherwise the 32-bit register name; do nothing on a memory operand.
3452 - ``n``: Negate and print an unadorned integer, or, for operands other than an
3453 immediate integer (e.g. a relocatable symbol expression), print a '-' before
3454 the operand. (The behavior for relocatable symbol expressions is a
3455 target-specific behavior for this typically target-independent modifier)
3456 - ``H``: Print a memory reference with additional offset +8.
3457 - ``P``: Print a memory reference or operand for use as the argument of a call
3458 instruction. (E.g. omit ``(rip)``, even though it's PC-relative.)
3462 No additional modifiers.
3468 The call instructions that wrap inline asm nodes may have a
3469 "``!srcloc``" MDNode attached to it that contains a list of constant
3470 integers. If present, the code generator will use the integer as the
3471 location cookie value when report errors through the ``LLVMContext``
3472 error reporting mechanisms. This allows a front-end to correlate backend
3473 errors that occur with inline asm back to the source code that produced
3476 .. code-block:: llvm
3478 call void asm sideeffect "something bad", ""(), !srcloc !42
3480 !42 = !{ i32 1234567 }
3482 It is up to the front-end to make sense of the magic numbers it places
3483 in the IR. If the MDNode contains multiple constants, the code generator
3484 will use the one that corresponds to the line of the asm that the error
3492 LLVM IR allows metadata to be attached to instructions in the program
3493 that can convey extra information about the code to the optimizers and
3494 code generator. One example application of metadata is source-level
3495 debug information. There are two metadata primitives: strings and nodes.
3497 Metadata does not have a type, and is not a value. If referenced from a
3498 ``call`` instruction, it uses the ``metadata`` type.
3500 All metadata are identified in syntax by a exclamation point ('``!``').
3502 .. _metadata-string:
3504 Metadata Nodes and Metadata Strings
3505 -----------------------------------
3507 A metadata string is a string surrounded by double quotes. It can
3508 contain any character by escaping non-printable characters with
3509 "``\xx``" where "``xx``" is the two digit hex code. For example:
3512 Metadata nodes are represented with notation similar to structure
3513 constants (a comma separated list of elements, surrounded by braces and
3514 preceded by an exclamation point). Metadata nodes can have any values as
3515 their operand. For example:
3517 .. code-block:: llvm
3519 !{ !"test\00", i32 10}
3521 Metadata nodes that aren't uniqued use the ``distinct`` keyword. For example:
3523 .. code-block:: llvm
3525 !0 = distinct !{!"test\00", i32 10}
3527 ``distinct`` nodes are useful when nodes shouldn't be merged based on their
3528 content. They can also occur when transformations cause uniquing collisions
3529 when metadata operands change.
3531 A :ref:`named metadata <namedmetadatastructure>` is a collection of
3532 metadata nodes, which can be looked up in the module symbol table. For
3535 .. code-block:: llvm
3539 Metadata can be used as function arguments. Here ``llvm.dbg.value``
3540 function is using two metadata arguments:
3542 .. code-block:: llvm
3544 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
3546 Metadata can be attached with an instruction. Here metadata ``!21`` is
3547 attached to the ``add`` instruction using the ``!dbg`` identifier:
3549 .. code-block:: llvm
3551 %indvar.next = add i64 %indvar, 1, !dbg !21
3553 More information about specific metadata nodes recognized by the
3554 optimizers and code generator is found below.
3556 .. _specialized-metadata:
3558 Specialized Metadata Nodes
3559 ^^^^^^^^^^^^^^^^^^^^^^^^^^
3561 Specialized metadata nodes are custom data structures in metadata (as opposed
3562 to generic tuples). Their fields are labelled, and can be specified in any
3565 These aren't inherently debug info centric, but currently all the specialized
3566 metadata nodes are related to debug info.
3573 ``DICompileUnit`` nodes represent a compile unit. The ``enums:``,
3574 ``retainedTypes:``, ``subprograms:``, ``globals:`` and ``imports:`` fields are
3575 tuples containing the debug info to be emitted along with the compile unit,
3576 regardless of code optimizations (some nodes are only emitted if there are
3577 references to them from instructions).
3579 .. code-block:: llvm
3581 !0 = !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang",
3582 isOptimized: true, flags: "-O2", runtimeVersion: 2,
3583 splitDebugFilename: "abc.debug", emissionKind: 1,
3584 enums: !2, retainedTypes: !3, subprograms: !4,
3585 globals: !5, imports: !6)
3587 Compile unit descriptors provide the root scope for objects declared in a
3588 specific compilation unit. File descriptors are defined using this scope.
3589 These descriptors are collected by a named metadata ``!llvm.dbg.cu``. They
3590 keep track of subprograms, global variables, type information, and imported
3591 entities (declarations and namespaces).
3598 ``DIFile`` nodes represent files. The ``filename:`` can include slashes.
3600 .. code-block:: llvm
3602 !0 = !DIFile(filename: "path/to/file", directory: "/path/to/dir")
3604 Files are sometimes used in ``scope:`` fields, and are the only valid target
3605 for ``file:`` fields.
3612 ``DIBasicType`` nodes represent primitive types, such as ``int``, ``bool`` and
3613 ``float``. ``tag:`` defaults to ``DW_TAG_base_type``.
3615 .. code-block:: llvm
3617 !0 = !DIBasicType(name: "unsigned char", size: 8, align: 8,
3618 encoding: DW_ATE_unsigned_char)
3619 !1 = !DIBasicType(tag: DW_TAG_unspecified_type, name: "decltype(nullptr)")
3621 The ``encoding:`` describes the details of the type. Usually it's one of the
3624 .. code-block:: llvm
3630 DW_ATE_signed_char = 6
3632 DW_ATE_unsigned_char = 8
3634 .. _DISubroutineType:
3639 ``DISubroutineType`` nodes represent subroutine types. Their ``types:`` field
3640 refers to a tuple; the first operand is the return type, while the rest are the
3641 types of the formal arguments in order. If the first operand is ``null``, that
3642 represents a function with no return value (such as ``void foo() {}`` in C++).
3644 .. code-block:: llvm
3646 !0 = !BasicType(name: "int", size: 32, align: 32, DW_ATE_signed)
3647 !1 = !BasicType(name: "char", size: 8, align: 8, DW_ATE_signed_char)
3648 !2 = !DISubroutineType(types: !{null, !0, !1}) ; void (int, char)
3655 ``DIDerivedType`` nodes represent types derived from other types, such as
3658 .. code-block:: llvm
3660 !0 = !DIBasicType(name: "unsigned char", size: 8, align: 8,
3661 encoding: DW_ATE_unsigned_char)
3662 !1 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !0, size: 32,
3665 The following ``tag:`` values are valid:
3667 .. code-block:: llvm
3669 DW_TAG_formal_parameter = 5
3671 DW_TAG_pointer_type = 15
3672 DW_TAG_reference_type = 16
3674 DW_TAG_ptr_to_member_type = 31
3675 DW_TAG_const_type = 38
3676 DW_TAG_volatile_type = 53
3677 DW_TAG_restrict_type = 55
3679 ``DW_TAG_member`` is used to define a member of a :ref:`composite type
3680 <DICompositeType>` or :ref:`subprogram <DISubprogram>`. The type of the member
3681 is the ``baseType:``. The ``offset:`` is the member's bit offset.
3682 ``DW_TAG_formal_parameter`` is used to define a member which is a formal
3683 argument of a subprogram.
3685 ``DW_TAG_typedef`` is used to provide a name for the ``baseType:``.
3687 ``DW_TAG_pointer_type``, ``DW_TAG_reference_type``, ``DW_TAG_const_type``,
3688 ``DW_TAG_volatile_type`` and ``DW_TAG_restrict_type`` are used to qualify the
3691 Note that the ``void *`` type is expressed as a type derived from NULL.
3693 .. _DICompositeType:
3698 ``DICompositeType`` nodes represent types composed of other types, like
3699 structures and unions. ``elements:`` points to a tuple of the composed types.
3701 If the source language supports ODR, the ``identifier:`` field gives the unique
3702 identifier used for type merging between modules. When specified, other types
3703 can refer to composite types indirectly via a :ref:`metadata string
3704 <metadata-string>` that matches their identifier.
3706 .. code-block:: llvm
3708 !0 = !DIEnumerator(name: "SixKind", value: 7)
3709 !1 = !DIEnumerator(name: "SevenKind", value: 7)
3710 !2 = !DIEnumerator(name: "NegEightKind", value: -8)
3711 !3 = !DICompositeType(tag: DW_TAG_enumeration_type, name: "Enum", file: !12,
3712 line: 2, size: 32, align: 32, identifier: "_M4Enum",
3713 elements: !{!0, !1, !2})
3715 The following ``tag:`` values are valid:
3717 .. code-block:: llvm
3719 DW_TAG_array_type = 1
3720 DW_TAG_class_type = 2
3721 DW_TAG_enumeration_type = 4
3722 DW_TAG_structure_type = 19
3723 DW_TAG_union_type = 23
3724 DW_TAG_subroutine_type = 21
3725 DW_TAG_inheritance = 28
3728 For ``DW_TAG_array_type``, the ``elements:`` should be :ref:`subrange
3729 descriptors <DISubrange>`, each representing the range of subscripts at that
3730 level of indexing. The ``DIFlagVector`` flag to ``flags:`` indicates that an
3731 array type is a native packed vector.
3733 For ``DW_TAG_enumeration_type``, the ``elements:`` should be :ref:`enumerator
3734 descriptors <DIEnumerator>`, each representing the definition of an enumeration
3735 value for the set. All enumeration type descriptors are collected in the
3736 ``enums:`` field of the :ref:`compile unit <DICompileUnit>`.
3738 For ``DW_TAG_structure_type``, ``DW_TAG_class_type``, and
3739 ``DW_TAG_union_type``, the ``elements:`` should be :ref:`derived types
3740 <DIDerivedType>` with ``tag: DW_TAG_member`` or ``tag: DW_TAG_inheritance``.
3747 ``DISubrange`` nodes are the elements for ``DW_TAG_array_type`` variants of
3748 :ref:`DICompositeType`. ``count: -1`` indicates an empty array.
3750 .. code-block:: llvm
3752 !0 = !DISubrange(count: 5, lowerBound: 0) ; array counting from 0
3753 !1 = !DISubrange(count: 5, lowerBound: 1) ; array counting from 1
3754 !2 = !DISubrange(count: -1) ; empty array.
3761 ``DIEnumerator`` nodes are the elements for ``DW_TAG_enumeration_type``
3762 variants of :ref:`DICompositeType`.
3764 .. code-block:: llvm
3766 !0 = !DIEnumerator(name: "SixKind", value: 7)
3767 !1 = !DIEnumerator(name: "SevenKind", value: 7)
3768 !2 = !DIEnumerator(name: "NegEightKind", value: -8)
3770 DITemplateTypeParameter
3771 """""""""""""""""""""""
3773 ``DITemplateTypeParameter`` nodes represent type parameters to generic source
3774 language constructs. They are used (optionally) in :ref:`DICompositeType` and
3775 :ref:`DISubprogram` ``templateParams:`` fields.
3777 .. code-block:: llvm
3779 !0 = !DITemplateTypeParameter(name: "Ty", type: !1)
3781 DITemplateValueParameter
3782 """"""""""""""""""""""""
3784 ``DITemplateValueParameter`` nodes represent value parameters to generic source
3785 language constructs. ``tag:`` defaults to ``DW_TAG_template_value_parameter``,
3786 but if specified can also be set to ``DW_TAG_GNU_template_template_param`` or
3787 ``DW_TAG_GNU_template_param_pack``. They are used (optionally) in
3788 :ref:`DICompositeType` and :ref:`DISubprogram` ``templateParams:`` fields.
3790 .. code-block:: llvm
3792 !0 = !DITemplateValueParameter(name: "Ty", type: !1, value: i32 7)
3797 ``DINamespace`` nodes represent namespaces in the source language.
3799 .. code-block:: llvm
3801 !0 = !DINamespace(name: "myawesomeproject", scope: !1, file: !2, line: 7)
3806 ``DIGlobalVariable`` nodes represent global variables in the source language.
3808 .. code-block:: llvm
3810 !0 = !DIGlobalVariable(name: "foo", linkageName: "foo", scope: !1,
3811 file: !2, line: 7, type: !3, isLocal: true,
3812 isDefinition: false, variable: i32* @foo,
3815 All global variables should be referenced by the `globals:` field of a
3816 :ref:`compile unit <DICompileUnit>`.
3823 ``DISubprogram`` nodes represent functions from the source language. The
3824 ``variables:`` field points at :ref:`variables <DILocalVariable>` that must be
3825 retained, even if their IR counterparts are optimized out of the IR. The
3826 ``type:`` field must point at an :ref:`DISubroutineType`.
3828 .. code-block:: llvm
3830 !0 = !DISubprogram(name: "foo", linkageName: "_Zfoov", scope: !1,
3831 file: !2, line: 7, type: !3, isLocal: true,
3832 isDefinition: false, scopeLine: 8, containingType: !4,
3833 virtuality: DW_VIRTUALITY_pure_virtual, virtualIndex: 10,
3834 flags: DIFlagPrototyped, isOptimized: true,
3835 function: void ()* @_Z3foov,
3836 templateParams: !5, declaration: !6, variables: !7)
3843 ``DILexicalBlock`` nodes describe nested blocks within a :ref:`subprogram
3844 <DISubprogram>`. The line number and column numbers are used to dinstinguish
3845 two lexical blocks at same depth. They are valid targets for ``scope:``
3848 .. code-block:: llvm
3850 !0 = distinct !DILexicalBlock(scope: !1, file: !2, line: 7, column: 35)
3852 Usually lexical blocks are ``distinct`` to prevent node merging based on
3855 .. _DILexicalBlockFile:
3860 ``DILexicalBlockFile`` nodes are used to discriminate between sections of a
3861 :ref:`lexical block <DILexicalBlock>`. The ``file:`` field can be changed to
3862 indicate textual inclusion, or the ``discriminator:`` field can be used to
3863 discriminate between control flow within a single block in the source language.
3865 .. code-block:: llvm
3867 !0 = !DILexicalBlock(scope: !3, file: !4, line: 7, column: 35)
3868 !1 = !DILexicalBlockFile(scope: !0, file: !4, discriminator: 0)
3869 !2 = !DILexicalBlockFile(scope: !0, file: !4, discriminator: 1)
3876 ``DILocation`` nodes represent source debug locations. The ``scope:`` field is
3877 mandatory, and points at an :ref:`DILexicalBlockFile`, an
3878 :ref:`DILexicalBlock`, or an :ref:`DISubprogram`.
3880 .. code-block:: llvm
3882 !0 = !DILocation(line: 2900, column: 42, scope: !1, inlinedAt: !2)
3884 .. _DILocalVariable:
3889 ``DILocalVariable`` nodes represent local variables in the source language.
3890 Instead of ``DW_TAG_variable``, they use LLVM-specific fake tags to
3891 discriminate between local variables (``DW_TAG_auto_variable``) and subprogram
3892 arguments (``DW_TAG_arg_variable``). In the latter case, the ``arg:`` field
3893 specifies the argument position, and this variable will be included in the
3894 ``variables:`` field of its :ref:`DISubprogram`.
3896 .. code-block:: llvm
3898 !0 = !DILocalVariable(tag: DW_TAG_arg_variable, name: "this", arg: 0,
3899 scope: !3, file: !2, line: 7, type: !3,
3900 flags: DIFlagArtificial)
3901 !1 = !DILocalVariable(tag: DW_TAG_arg_variable, name: "x", arg: 1,
3902 scope: !4, file: !2, line: 7, type: !3)
3903 !1 = !DILocalVariable(tag: DW_TAG_auto_variable, name: "y",
3904 scope: !5, file: !2, line: 7, type: !3)
3909 ``DIExpression`` nodes represent DWARF expression sequences. They are used in
3910 :ref:`debug intrinsics<dbg_intrinsics>` (such as ``llvm.dbg.declare``) to
3911 describe how the referenced LLVM variable relates to the source language
3914 The current supported vocabulary is limited:
3916 - ``DW_OP_deref`` dereferences the working expression.
3917 - ``DW_OP_plus, 93`` adds ``93`` to the working expression.
3918 - ``DW_OP_bit_piece, 16, 8`` specifies the offset and size (``16`` and ``8``
3919 here, respectively) of the variable piece from the working expression.
3921 .. code-block:: llvm
3923 !0 = !DIExpression(DW_OP_deref)
3924 !1 = !DIExpression(DW_OP_plus, 3)
3925 !2 = !DIExpression(DW_OP_bit_piece, 3, 7)
3926 !3 = !DIExpression(DW_OP_deref, DW_OP_plus, 3, DW_OP_bit_piece, 3, 7)
3931 ``DIObjCProperty`` nodes represent Objective-C property nodes.
3933 .. code-block:: llvm
3935 !3 = !DIObjCProperty(name: "foo", file: !1, line: 7, setter: "setFoo",
3936 getter: "getFoo", attributes: 7, type: !2)
3941 ``DIImportedEntity`` nodes represent entities (such as modules) imported into a
3944 .. code-block:: llvm
3946 !2 = !DIImportedEntity(tag: DW_TAG_imported_module, name: "foo", scope: !0,
3947 entity: !1, line: 7)
3952 In LLVM IR, memory does not have types, so LLVM's own type system is not
3953 suitable for doing TBAA. Instead, metadata is added to the IR to
3954 describe a type system of a higher level language. This can be used to
3955 implement typical C/C++ TBAA, but it can also be used to implement
3956 custom alias analysis behavior for other languages.
3958 The current metadata format is very simple. TBAA metadata nodes have up
3959 to three fields, e.g.:
3961 .. code-block:: llvm
3963 !0 = !{ !"an example type tree" }
3964 !1 = !{ !"int", !0 }
3965 !2 = !{ !"float", !0 }
3966 !3 = !{ !"const float", !2, i64 1 }
3968 The first field is an identity field. It can be any value, usually a
3969 metadata string, which uniquely identifies the type. The most important
3970 name in the tree is the name of the root node. Two trees with different
3971 root node names are entirely disjoint, even if they have leaves with
3974 The second field identifies the type's parent node in the tree, or is
3975 null or omitted for a root node. A type is considered to alias all of
3976 its descendants and all of its ancestors in the tree. Also, a type is
3977 considered to alias all types in other trees, so that bitcode produced
3978 from multiple front-ends is handled conservatively.
3980 If the third field is present, it's an integer which if equal to 1
3981 indicates that the type is "constant" (meaning
3982 ``pointsToConstantMemory`` should return true; see `other useful
3983 AliasAnalysis methods <AliasAnalysis.html#OtherItfs>`_).
3985 '``tbaa.struct``' Metadata
3986 ^^^^^^^^^^^^^^^^^^^^^^^^^^
3988 The :ref:`llvm.memcpy <int_memcpy>` is often used to implement
3989 aggregate assignment operations in C and similar languages, however it
3990 is defined to copy a contiguous region of memory, which is more than
3991 strictly necessary for aggregate types which contain holes due to
3992 padding. Also, it doesn't contain any TBAA information about the fields
3995 ``!tbaa.struct`` metadata can describe which memory subregions in a
3996 memcpy are padding and what the TBAA tags of the struct are.
3998 The current metadata format is very simple. ``!tbaa.struct`` metadata
3999 nodes are a list of operands which are in conceptual groups of three.
4000 For each group of three, the first operand gives the byte offset of a
4001 field in bytes, the second gives its size in bytes, and the third gives
4004 .. code-block:: llvm
4006 !4 = !{ i64 0, i64 4, !1, i64 8, i64 4, !2 }
4008 This describes a struct with two fields. The first is at offset 0 bytes
4009 with size 4 bytes, and has tbaa tag !1. The second is at offset 8 bytes
4010 and has size 4 bytes and has tbaa tag !2.
4012 Note that the fields need not be contiguous. In this example, there is a
4013 4 byte gap between the two fields. This gap represents padding which
4014 does not carry useful data and need not be preserved.
4016 '``noalias``' and '``alias.scope``' Metadata
4017 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4019 ``noalias`` and ``alias.scope`` metadata provide the ability to specify generic
4020 noalias memory-access sets. This means that some collection of memory access
4021 instructions (loads, stores, memory-accessing calls, etc.) that carry
4022 ``noalias`` metadata can specifically be specified not to alias with some other
4023 collection of memory access instructions that carry ``alias.scope`` metadata.
4024 Each type of metadata specifies a list of scopes where each scope has an id and
4025 a domain. When evaluating an aliasing query, if for some domain, the set
4026 of scopes with that domain in one instruction's ``alias.scope`` list is a
4027 subset of (or equal to) the set of scopes for that domain in another
4028 instruction's ``noalias`` list, then the two memory accesses are assumed not to
4031 The metadata identifying each domain is itself a list containing one or two
4032 entries. The first entry is the name of the domain. Note that if the name is a
4033 string then it can be combined accross functions and translation units. A
4034 self-reference can be used to create globally unique domain names. A
4035 descriptive string may optionally be provided as a second list entry.
4037 The metadata identifying each scope is also itself a list containing two or
4038 three entries. The first entry is the name of the scope. Note that if the name
4039 is a string then it can be combined accross functions and translation units. A
4040 self-reference can be used to create globally unique scope names. A metadata
4041 reference to the scope's domain is the second entry. A descriptive string may
4042 optionally be provided as a third list entry.
4046 .. code-block:: llvm
4048 ; Two scope domains:
4052 ; Some scopes in these domains:
4058 !5 = !{!4} ; A list containing only scope !4
4062 ; These two instructions don't alias:
4063 %0 = load float, float* %c, align 4, !alias.scope !5
4064 store float %0, float* %arrayidx.i, align 4, !noalias !5
4066 ; These two instructions also don't alias (for domain !1, the set of scopes
4067 ; in the !alias.scope equals that in the !noalias list):
4068 %2 = load float, float* %c, align 4, !alias.scope !5
4069 store float %2, float* %arrayidx.i2, align 4, !noalias !6
4071 ; These two instructions may alias (for domain !0, the set of scopes in
4072 ; the !noalias list is not a superset of, or equal to, the scopes in the
4073 ; !alias.scope list):
4074 %2 = load float, float* %c, align 4, !alias.scope !6
4075 store float %0, float* %arrayidx.i, align 4, !noalias !7
4077 '``fpmath``' Metadata
4078 ^^^^^^^^^^^^^^^^^^^^^
4080 ``fpmath`` metadata may be attached to any instruction of floating point
4081 type. It can be used to express the maximum acceptable error in the
4082 result of that instruction, in ULPs, thus potentially allowing the
4083 compiler to use a more efficient but less accurate method of computing
4084 it. ULP is defined as follows:
4086 If ``x`` is a real number that lies between two finite consecutive
4087 floating-point numbers ``a`` and ``b``, without being equal to one
4088 of them, then ``ulp(x) = |b - a|``, otherwise ``ulp(x)`` is the
4089 distance between the two non-equal finite floating-point numbers
4090 nearest ``x``. Moreover, ``ulp(NaN)`` is ``NaN``.
4092 The metadata node shall consist of a single positive floating point
4093 number representing the maximum relative error, for example:
4095 .. code-block:: llvm
4097 !0 = !{ float 2.5 } ; maximum acceptable inaccuracy is 2.5 ULPs
4101 '``range``' Metadata
4102 ^^^^^^^^^^^^^^^^^^^^
4104 ``range`` metadata may be attached only to ``load``, ``call`` and ``invoke`` of
4105 integer types. It expresses the possible ranges the loaded value or the value
4106 returned by the called function at this call site is in. The ranges are
4107 represented with a flattened list of integers. The loaded value or the value
4108 returned is known to be in the union of the ranges defined by each consecutive
4109 pair. Each pair has the following properties:
4111 - The type must match the type loaded by the instruction.
4112 - The pair ``a,b`` represents the range ``[a,b)``.
4113 - Both ``a`` and ``b`` are constants.
4114 - The range is allowed to wrap.
4115 - The range should not represent the full or empty set. That is,
4118 In addition, the pairs must be in signed order of the lower bound and
4119 they must be non-contiguous.
4123 .. code-block:: llvm
4125 %a = load i8, i8* %x, align 1, !range !0 ; Can only be 0 or 1
4126 %b = load i8, i8* %y, align 1, !range !1 ; Can only be 255 (-1), 0 or 1
4127 %c = call i8 @foo(), !range !2 ; Can only be 0, 1, 3, 4 or 5
4128 %d = invoke i8 @bar() to label %cont
4129 unwind label %lpad, !range !3 ; Can only be -2, -1, 3, 4 or 5
4131 !0 = !{ i8 0, i8 2 }
4132 !1 = !{ i8 255, i8 2 }
4133 !2 = !{ i8 0, i8 2, i8 3, i8 6 }
4134 !3 = !{ i8 -2, i8 0, i8 3, i8 6 }
4139 It is sometimes useful to attach information to loop constructs. Currently,
4140 loop metadata is implemented as metadata attached to the branch instruction
4141 in the loop latch block. This type of metadata refer to a metadata node that is
4142 guaranteed to be separate for each loop. The loop identifier metadata is
4143 specified with the name ``llvm.loop``.
4145 The loop identifier metadata is implemented using a metadata that refers to
4146 itself to avoid merging it with any other identifier metadata, e.g.,
4147 during module linkage or function inlining. That is, each loop should refer
4148 to their own identification metadata even if they reside in separate functions.
4149 The following example contains loop identifier metadata for two separate loop
4152 .. code-block:: llvm
4157 The loop identifier metadata can be used to specify additional
4158 per-loop metadata. Any operands after the first operand can be treated
4159 as user-defined metadata. For example the ``llvm.loop.unroll.count``
4160 suggests an unroll factor to the loop unroller:
4162 .. code-block:: llvm
4164 br i1 %exitcond, label %._crit_edge, label %.lr.ph, !llvm.loop !0
4167 !1 = !{!"llvm.loop.unroll.count", i32 4}
4169 '``llvm.loop.vectorize``' and '``llvm.loop.interleave``'
4170 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4172 Metadata prefixed with ``llvm.loop.vectorize`` or ``llvm.loop.interleave`` are
4173 used to control per-loop vectorization and interleaving parameters such as
4174 vectorization width and interleave count. These metadata should be used in
4175 conjunction with ``llvm.loop`` loop identification metadata. The
4176 ``llvm.loop.vectorize`` and ``llvm.loop.interleave`` metadata are only
4177 optimization hints and the optimizer will only interleave and vectorize loops if
4178 it believes it is safe to do so. The ``llvm.mem.parallel_loop_access`` metadata
4179 which contains information about loop-carried memory dependencies can be helpful
4180 in determining the safety of these transformations.
4182 '``llvm.loop.interleave.count``' Metadata
4183 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4185 This metadata suggests an interleave count to the loop interleaver.
4186 The first operand is the string ``llvm.loop.interleave.count`` and the
4187 second operand is an integer specifying the interleave count. For
4190 .. code-block:: llvm
4192 !0 = !{!"llvm.loop.interleave.count", i32 4}
4194 Note that setting ``llvm.loop.interleave.count`` to 1 disables interleaving
4195 multiple iterations of the loop. If ``llvm.loop.interleave.count`` is set to 0
4196 then the interleave count will be determined automatically.
4198 '``llvm.loop.vectorize.enable``' Metadata
4199 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4201 This metadata selectively enables or disables vectorization for the loop. The
4202 first operand is the string ``llvm.loop.vectorize.enable`` and the second operand
4203 is a bit. If the bit operand value is 1 vectorization is enabled. A value of
4204 0 disables vectorization:
4206 .. code-block:: llvm
4208 !0 = !{!"llvm.loop.vectorize.enable", i1 0}
4209 !1 = !{!"llvm.loop.vectorize.enable", i1 1}
4211 '``llvm.loop.vectorize.width``' Metadata
4212 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4214 This metadata sets the target width of the vectorizer. The first
4215 operand is the string ``llvm.loop.vectorize.width`` and the second
4216 operand is an integer specifying the width. For example:
4218 .. code-block:: llvm
4220 !0 = !{!"llvm.loop.vectorize.width", i32 4}
4222 Note that setting ``llvm.loop.vectorize.width`` to 1 disables
4223 vectorization of the loop. If ``llvm.loop.vectorize.width`` is set to
4224 0 or if the loop does not have this metadata the width will be
4225 determined automatically.
4227 '``llvm.loop.unroll``'
4228 ^^^^^^^^^^^^^^^^^^^^^^
4230 Metadata prefixed with ``llvm.loop.unroll`` are loop unrolling
4231 optimization hints such as the unroll factor. ``llvm.loop.unroll``
4232 metadata should be used in conjunction with ``llvm.loop`` loop
4233 identification metadata. The ``llvm.loop.unroll`` metadata are only
4234 optimization hints and the unrolling will only be performed if the
4235 optimizer believes it is safe to do so.
4237 '``llvm.loop.unroll.count``' Metadata
4238 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4240 This metadata suggests an unroll factor to the loop unroller. The
4241 first operand is the string ``llvm.loop.unroll.count`` and the second
4242 operand is a positive integer specifying the unroll factor. For
4245 .. code-block:: llvm
4247 !0 = !{!"llvm.loop.unroll.count", i32 4}
4249 If the trip count of the loop is less than the unroll count the loop
4250 will be partially unrolled.
4252 '``llvm.loop.unroll.disable``' Metadata
4253 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4255 This metadata disables loop unrolling. The metadata has a single operand
4256 which is the string ``llvm.loop.unroll.disable``. For example:
4258 .. code-block:: llvm
4260 !0 = !{!"llvm.loop.unroll.disable"}
4262 '``llvm.loop.unroll.runtime.disable``' Metadata
4263 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4265 This metadata disables runtime loop unrolling. The metadata has a single
4266 operand which is the string ``llvm.loop.unroll.runtime.disable``. For example:
4268 .. code-block:: llvm
4270 !0 = !{!"llvm.loop.unroll.runtime.disable"}
4272 '``llvm.loop.unroll.full``' Metadata
4273 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4275 This metadata suggests that the loop should be unrolled fully. The
4276 metadata has a single operand which is the string ``llvm.loop.unroll.full``.
4279 .. code-block:: llvm
4281 !0 = !{!"llvm.loop.unroll.full"}
4286 Metadata types used to annotate memory accesses with information helpful
4287 for optimizations are prefixed with ``llvm.mem``.
4289 '``llvm.mem.parallel_loop_access``' Metadata
4290 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4292 The ``llvm.mem.parallel_loop_access`` metadata refers to a loop identifier,
4293 or metadata containing a list of loop identifiers for nested loops.
4294 The metadata is attached to memory accessing instructions and denotes that
4295 no loop carried memory dependence exist between it and other instructions denoted
4296 with the same loop identifier.
4298 Precisely, given two instructions ``m1`` and ``m2`` that both have the
4299 ``llvm.mem.parallel_loop_access`` metadata, with ``L1`` and ``L2`` being the
4300 set of loops associated with that metadata, respectively, then there is no loop
4301 carried dependence between ``m1`` and ``m2`` for loops in both ``L1`` and
4304 As a special case, if all memory accessing instructions in a loop have
4305 ``llvm.mem.parallel_loop_access`` metadata that refers to that loop, then the
4306 loop has no loop carried memory dependences and is considered to be a parallel
4309 Note that if not all memory access instructions have such metadata referring to
4310 the loop, then the loop is considered not being trivially parallel. Additional
4311 memory dependence analysis is required to make that determination. As a fail
4312 safe mechanism, this causes loops that were originally parallel to be considered
4313 sequential (if optimization passes that are unaware of the parallel semantics
4314 insert new memory instructions into the loop body).
4316 Example of a loop that is considered parallel due to its correct use of
4317 both ``llvm.loop`` and ``llvm.mem.parallel_loop_access``
4318 metadata types that refer to the same loop identifier metadata.
4320 .. code-block:: llvm
4324 %val0 = load i32, i32* %arrayidx, !llvm.mem.parallel_loop_access !0
4326 store i32 %val0, i32* %arrayidx1, !llvm.mem.parallel_loop_access !0
4328 br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0
4334 It is also possible to have nested parallel loops. In that case the
4335 memory accesses refer to a list of loop identifier metadata nodes instead of
4336 the loop identifier metadata node directly:
4338 .. code-block:: llvm
4342 %val1 = load i32, i32* %arrayidx3, !llvm.mem.parallel_loop_access !2
4344 br label %inner.for.body
4348 %val0 = load i32, i32* %arrayidx1, !llvm.mem.parallel_loop_access !0
4350 store i32 %val0, i32* %arrayidx2, !llvm.mem.parallel_loop_access !0
4352 br i1 %exitcond, label %inner.for.end, label %inner.for.body, !llvm.loop !1
4356 store i32 %val1, i32* %arrayidx4, !llvm.mem.parallel_loop_access !2
4358 br i1 %exitcond, label %outer.for.end, label %outer.for.body, !llvm.loop !2
4360 outer.for.end: ; preds = %for.body
4362 !0 = !{!1, !2} ; a list of loop identifiers
4363 !1 = !{!1} ; an identifier for the inner loop
4364 !2 = !{!2} ; an identifier for the outer loop
4369 The ``llvm.bitsets`` global metadata is used to implement
4370 :doc:`bitsets <BitSets>`.
4372 Module Flags Metadata
4373 =====================
4375 Information about the module as a whole is difficult to convey to LLVM's
4376 subsystems. The LLVM IR isn't sufficient to transmit this information.
4377 The ``llvm.module.flags`` named metadata exists in order to facilitate
4378 this. These flags are in the form of key / value pairs --- much like a
4379 dictionary --- making it easy for any subsystem who cares about a flag to
4382 The ``llvm.module.flags`` metadata contains a list of metadata triplets.
4383 Each triplet has the following form:
4385 - The first element is a *behavior* flag, which specifies the behavior
4386 when two (or more) modules are merged together, and it encounters two
4387 (or more) metadata with the same ID. The supported behaviors are
4389 - The second element is a metadata string that is a unique ID for the
4390 metadata. Each module may only have one flag entry for each unique ID (not
4391 including entries with the **Require** behavior).
4392 - The third element is the value of the flag.
4394 When two (or more) modules are merged together, the resulting
4395 ``llvm.module.flags`` metadata is the union of the modules' flags. That is, for
4396 each unique metadata ID string, there will be exactly one entry in the merged
4397 modules ``llvm.module.flags`` metadata table, and the value for that entry will
4398 be determined by the merge behavior flag, as described below. The only exception
4399 is that entries with the *Require* behavior are always preserved.
4401 The following behaviors are supported:
4412 Emits an error if two values disagree, otherwise the resulting value
4413 is that of the operands.
4417 Emits a warning if two values disagree. The result value will be the
4418 operand for the flag from the first module being linked.
4422 Adds a requirement that another module flag be present and have a
4423 specified value after linking is performed. The value must be a
4424 metadata pair, where the first element of the pair is the ID of the
4425 module flag to be restricted, and the second element of the pair is
4426 the value the module flag should be restricted to. This behavior can
4427 be used to restrict the allowable results (via triggering of an
4428 error) of linking IDs with the **Override** behavior.
4432 Uses the specified value, regardless of the behavior or value of the
4433 other module. If both modules specify **Override**, but the values
4434 differ, an error will be emitted.
4438 Appends the two values, which are required to be metadata nodes.
4442 Appends the two values, which are required to be metadata
4443 nodes. However, duplicate entries in the second list are dropped
4444 during the append operation.
4446 It is an error for a particular unique flag ID to have multiple behaviors,
4447 except in the case of **Require** (which adds restrictions on another metadata
4448 value) or **Override**.
4450 An example of module flags:
4452 .. code-block:: llvm
4454 !0 = !{ i32 1, !"foo", i32 1 }
4455 !1 = !{ i32 4, !"bar", i32 37 }
4456 !2 = !{ i32 2, !"qux", i32 42 }
4457 !3 = !{ i32 3, !"qux",
4462 !llvm.module.flags = !{ !0, !1, !2, !3 }
4464 - Metadata ``!0`` has the ID ``!"foo"`` and the value '1'. The behavior
4465 if two or more ``!"foo"`` flags are seen is to emit an error if their
4466 values are not equal.
4468 - Metadata ``!1`` has the ID ``!"bar"`` and the value '37'. The
4469 behavior if two or more ``!"bar"`` flags are seen is to use the value
4472 - Metadata ``!2`` has the ID ``!"qux"`` and the value '42'. The
4473 behavior if two or more ``!"qux"`` flags are seen is to emit a
4474 warning if their values are not equal.
4476 - Metadata ``!3`` has the ID ``!"qux"`` and the value:
4482 The behavior is to emit an error if the ``llvm.module.flags`` does not
4483 contain a flag with the ID ``!"foo"`` that has the value '1' after linking is
4486 Objective-C Garbage Collection Module Flags Metadata
4487 ----------------------------------------------------
4489 On the Mach-O platform, Objective-C stores metadata about garbage
4490 collection in a special section called "image info". The metadata
4491 consists of a version number and a bitmask specifying what types of
4492 garbage collection are supported (if any) by the file. If two or more
4493 modules are linked together their garbage collection metadata needs to
4494 be merged rather than appended together.
4496 The Objective-C garbage collection module flags metadata consists of the
4497 following key-value pairs:
4506 * - ``Objective-C Version``
4507 - **[Required]** --- The Objective-C ABI version. Valid values are 1 and 2.
4509 * - ``Objective-C Image Info Version``
4510 - **[Required]** --- The version of the image info section. Currently
4513 * - ``Objective-C Image Info Section``
4514 - **[Required]** --- The section to place the metadata. Valid values are
4515 ``"__OBJC, __image_info, regular"`` for Objective-C ABI version 1, and
4516 ``"__DATA,__objc_imageinfo, regular, no_dead_strip"`` for
4517 Objective-C ABI version 2.
4519 * - ``Objective-C Garbage Collection``
4520 - **[Required]** --- Specifies whether garbage collection is supported or
4521 not. Valid values are 0, for no garbage collection, and 2, for garbage
4522 collection supported.
4524 * - ``Objective-C GC Only``
4525 - **[Optional]** --- Specifies that only garbage collection is supported.
4526 If present, its value must be 6. This flag requires that the
4527 ``Objective-C Garbage Collection`` flag have the value 2.
4529 Some important flag interactions:
4531 - If a module with ``Objective-C Garbage Collection`` set to 0 is
4532 merged with a module with ``Objective-C Garbage Collection`` set to
4533 2, then the resulting module has the
4534 ``Objective-C Garbage Collection`` flag set to 0.
4535 - A module with ``Objective-C Garbage Collection`` set to 0 cannot be
4536 merged with a module with ``Objective-C GC Only`` set to 6.
4538 Automatic Linker Flags Module Flags Metadata
4539 --------------------------------------------
4541 Some targets support embedding flags to the linker inside individual object
4542 files. Typically this is used in conjunction with language extensions which
4543 allow source files to explicitly declare the libraries they depend on, and have
4544 these automatically be transmitted to the linker via object files.
4546 These flags are encoded in the IR using metadata in the module flags section,
4547 using the ``Linker Options`` key. The merge behavior for this flag is required
4548 to be ``AppendUnique``, and the value for the key is expected to be a metadata
4549 node which should be a list of other metadata nodes, each of which should be a
4550 list of metadata strings defining linker options.
4552 For example, the following metadata section specifies two separate sets of
4553 linker options, presumably to link against ``libz`` and the ``Cocoa``
4556 !0 = !{ i32 6, !"Linker Options",
4559 !{ !"-framework", !"Cocoa" } } }
4560 !llvm.module.flags = !{ !0 }
4562 The metadata encoding as lists of lists of options, as opposed to a collapsed
4563 list of options, is chosen so that the IR encoding can use multiple option
4564 strings to specify e.g., a single library, while still having that specifier be
4565 preserved as an atomic element that can be recognized by a target specific
4566 assembly writer or object file emitter.
4568 Each individual option is required to be either a valid option for the target's
4569 linker, or an option that is reserved by the target specific assembly writer or
4570 object file emitter. No other aspect of these options is defined by the IR.
4572 C type width Module Flags Metadata
4573 ----------------------------------
4575 The ARM backend emits a section into each generated object file describing the
4576 options that it was compiled with (in a compiler-independent way) to prevent
4577 linking incompatible objects, and to allow automatic library selection. Some
4578 of these options are not visible at the IR level, namely wchar_t width and enum
4581 To pass this information to the backend, these options are encoded in module
4582 flags metadata, using the following key-value pairs:
4592 - * 0 --- sizeof(wchar_t) == 4
4593 * 1 --- sizeof(wchar_t) == 2
4596 - * 0 --- Enums are at least as large as an ``int``.
4597 * 1 --- Enums are stored in the smallest integer type which can
4598 represent all of its values.
4600 For example, the following metadata section specifies that the module was
4601 compiled with a ``wchar_t`` width of 4 bytes, and the underlying type of an
4602 enum is the smallest type which can represent all of its values::
4604 !llvm.module.flags = !{!0, !1}
4605 !0 = !{i32 1, !"short_wchar", i32 1}
4606 !1 = !{i32 1, !"short_enum", i32 0}
4608 .. _intrinsicglobalvariables:
4610 Intrinsic Global Variables
4611 ==========================
4613 LLVM has a number of "magic" global variables that contain data that
4614 affect code generation or other IR semantics. These are documented here.
4615 All globals of this sort should have a section specified as
4616 "``llvm.metadata``". This section and all globals that start with
4617 "``llvm.``" are reserved for use by LLVM.
4621 The '``llvm.used``' Global Variable
4622 -----------------------------------
4624 The ``@llvm.used`` global is an array which has
4625 :ref:`appending linkage <linkage_appending>`. This array contains a list of
4626 pointers to named global variables, functions and aliases which may optionally
4627 have a pointer cast formed of bitcast or getelementptr. For example, a legal
4630 .. code-block:: llvm
4635 @llvm.used = appending global [2 x i8*] [
4637 i8* bitcast (i32* @Y to i8*)
4638 ], section "llvm.metadata"
4640 If a symbol appears in the ``@llvm.used`` list, then the compiler, assembler,
4641 and linker are required to treat the symbol as if there is a reference to the
4642 symbol that it cannot see (which is why they have to be named). For example, if
4643 a variable has internal linkage and no references other than that from the
4644 ``@llvm.used`` list, it cannot be deleted. This is commonly used to represent
4645 references from inline asms and other things the compiler cannot "see", and
4646 corresponds to "``attribute((used))``" in GNU C.
4648 On some targets, the code generator must emit a directive to the
4649 assembler or object file to prevent the assembler and linker from
4650 molesting the symbol.
4652 .. _gv_llvmcompilerused:
4654 The '``llvm.compiler.used``' Global Variable
4655 --------------------------------------------
4657 The ``@llvm.compiler.used`` directive is the same as the ``@llvm.used``
4658 directive, except that it only prevents the compiler from touching the
4659 symbol. On targets that support it, this allows an intelligent linker to
4660 optimize references to the symbol without being impeded as it would be
4663 This is a rare construct that should only be used in rare circumstances,
4664 and should not be exposed to source languages.
4666 .. _gv_llvmglobalctors:
4668 The '``llvm.global_ctors``' Global Variable
4669 -------------------------------------------
4671 .. code-block:: llvm
4673 %0 = type { i32, void ()*, i8* }
4674 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor, i8* @data }]
4676 The ``@llvm.global_ctors`` array contains a list of constructor
4677 functions, priorities, and an optional associated global or function.
4678 The functions referenced by this array will be called in ascending order
4679 of priority (i.e. lowest first) when the module is loaded. The order of
4680 functions with the same priority is not defined.
4682 If the third field is present, non-null, and points to a global variable
4683 or function, the initializer function will only run if the associated
4684 data from the current module is not discarded.
4686 .. _llvmglobaldtors:
4688 The '``llvm.global_dtors``' Global Variable
4689 -------------------------------------------
4691 .. code-block:: llvm
4693 %0 = type { i32, void ()*, i8* }
4694 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor, i8* @data }]
4696 The ``@llvm.global_dtors`` array contains a list of destructor
4697 functions, priorities, and an optional associated global or function.
4698 The functions referenced by this array will be called in descending
4699 order of priority (i.e. highest first) when the module is unloaded. The
4700 order of functions with the same priority is not defined.
4702 If the third field is present, non-null, and points to a global variable
4703 or function, the destructor function will only run if the associated
4704 data from the current module is not discarded.
4706 Instruction Reference
4707 =====================
4709 The LLVM instruction set consists of several different classifications
4710 of instructions: :ref:`terminator instructions <terminators>`, :ref:`binary
4711 instructions <binaryops>`, :ref:`bitwise binary
4712 instructions <bitwiseops>`, :ref:`memory instructions <memoryops>`, and
4713 :ref:`other instructions <otherops>`.
4717 Terminator Instructions
4718 -----------------------
4720 As mentioned :ref:`previously <functionstructure>`, every basic block in a
4721 program ends with a "Terminator" instruction, which indicates which
4722 block should be executed after the current block is finished. These
4723 terminator instructions typically yield a '``void``' value: they produce
4724 control flow, not values (the one exception being the
4725 ':ref:`invoke <i_invoke>`' instruction).
4727 The terminator instructions are: ':ref:`ret <i_ret>`',
4728 ':ref:`br <i_br>`', ':ref:`switch <i_switch>`',
4729 ':ref:`indirectbr <i_indirectbr>`', ':ref:`invoke <i_invoke>`',
4730 ':ref:`resume <i_resume>`', ':ref:`catchblock <i_catchblock>`',
4731 ':ref:`catchendblock <i_catchendblock>`',
4732 ':ref:`catchret <i_catchret>`',
4733 ':ref:`cleanupret <i_cleanupret>`',
4734 ':ref:`terminateblock <i_terminateblock>`',
4735 and ':ref:`unreachable <i_unreachable>`'.
4739 '``ret``' Instruction
4740 ^^^^^^^^^^^^^^^^^^^^^
4747 ret <type> <value> ; Return a value from a non-void function
4748 ret void ; Return from void function
4753 The '``ret``' instruction is used to return control flow (and optionally
4754 a value) from a function back to the caller.
4756 There are two forms of the '``ret``' instruction: one that returns a
4757 value and then causes control flow, and one that just causes control
4763 The '``ret``' instruction optionally accepts a single argument, the
4764 return value. The type of the return value must be a ':ref:`first
4765 class <t_firstclass>`' type.
4767 A function is not :ref:`well formed <wellformed>` if it it has a non-void
4768 return type and contains a '``ret``' instruction with no return value or
4769 a return value with a type that does not match its type, or if it has a
4770 void return type and contains a '``ret``' instruction with a return
4776 When the '``ret``' instruction is executed, control flow returns back to
4777 the calling function's context. If the caller is a
4778 ":ref:`call <i_call>`" instruction, execution continues at the
4779 instruction after the call. If the caller was an
4780 ":ref:`invoke <i_invoke>`" instruction, execution continues at the
4781 beginning of the "normal" destination block. If the instruction returns
4782 a value, that value shall set the call or invoke instruction's return
4788 .. code-block:: llvm
4790 ret i32 5 ; Return an integer value of 5
4791 ret void ; Return from a void function
4792 ret { i32, i8 } { i32 4, i8 2 } ; Return a struct of values 4 and 2
4796 '``br``' Instruction
4797 ^^^^^^^^^^^^^^^^^^^^
4804 br i1 <cond>, label <iftrue>, label <iffalse>
4805 br label <dest> ; Unconditional branch
4810 The '``br``' instruction is used to cause control flow to transfer to a
4811 different basic block in the current function. There are two forms of
4812 this instruction, corresponding to a conditional branch and an
4813 unconditional branch.
4818 The conditional branch form of the '``br``' instruction takes a single
4819 '``i1``' value and two '``label``' values. The unconditional form of the
4820 '``br``' instruction takes a single '``label``' value as a target.
4825 Upon execution of a conditional '``br``' instruction, the '``i1``'
4826 argument is evaluated. If the value is ``true``, control flows to the
4827 '``iftrue``' ``label`` argument. If "cond" is ``false``, control flows
4828 to the '``iffalse``' ``label`` argument.
4833 .. code-block:: llvm
4836 %cond = icmp eq i32 %a, %b
4837 br i1 %cond, label %IfEqual, label %IfUnequal
4845 '``switch``' Instruction
4846 ^^^^^^^^^^^^^^^^^^^^^^^^
4853 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
4858 The '``switch``' instruction is used to transfer control flow to one of
4859 several different places. It is a generalization of the '``br``'
4860 instruction, allowing a branch to occur to one of many possible
4866 The '``switch``' instruction uses three parameters: an integer
4867 comparison value '``value``', a default '``label``' destination, and an
4868 array of pairs of comparison value constants and '``label``'s. The table
4869 is not allowed to contain duplicate constant entries.
4874 The ``switch`` instruction specifies a table of values and destinations.
4875 When the '``switch``' instruction is executed, this table is searched
4876 for the given value. If the value is found, control flow is transferred
4877 to the corresponding destination; otherwise, control flow is transferred
4878 to the default destination.
4883 Depending on properties of the target machine and the particular
4884 ``switch`` instruction, this instruction may be code generated in
4885 different ways. For example, it could be generated as a series of
4886 chained conditional branches or with a lookup table.
4891 .. code-block:: llvm
4893 ; Emulate a conditional br instruction
4894 %Val = zext i1 %value to i32
4895 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
4897 ; Emulate an unconditional br instruction
4898 switch i32 0, label %dest [ ]
4900 ; Implement a jump table:
4901 switch i32 %val, label %otherwise [ i32 0, label %onzero
4903 i32 2, label %ontwo ]
4907 '``indirectbr``' Instruction
4908 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4915 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
4920 The '``indirectbr``' instruction implements an indirect branch to a
4921 label within the current function, whose address is specified by
4922 "``address``". Address must be derived from a
4923 :ref:`blockaddress <blockaddress>` constant.
4928 The '``address``' argument is the address of the label to jump to. The
4929 rest of the arguments indicate the full set of possible destinations
4930 that the address may point to. Blocks are allowed to occur multiple
4931 times in the destination list, though this isn't particularly useful.
4933 This destination list is required so that dataflow analysis has an
4934 accurate understanding of the CFG.
4939 Control transfers to the block specified in the address argument. All
4940 possible destination blocks must be listed in the label list, otherwise
4941 this instruction has undefined behavior. This implies that jumps to
4942 labels defined in other functions have undefined behavior as well.
4947 This is typically implemented with a jump through a register.
4952 .. code-block:: llvm
4954 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
4958 '``invoke``' Instruction
4959 ^^^^^^^^^^^^^^^^^^^^^^^^
4966 <result> = invoke [cconv] [ret attrs] <ptr to function ty> <function ptr val>(<function args>) [fn attrs]
4967 to label <normal label> unwind label <exception label>
4972 The '``invoke``' instruction causes control to transfer to a specified
4973 function, with the possibility of control flow transfer to either the
4974 '``normal``' label or the '``exception``' label. If the callee function
4975 returns with the "``ret``" instruction, control flow will return to the
4976 "normal" label. If the callee (or any indirect callees) returns via the
4977 ":ref:`resume <i_resume>`" instruction or other exception handling
4978 mechanism, control is interrupted and continued at the dynamically
4979 nearest "exception" label.
4981 The '``exception``' label is a `landing
4982 pad <ExceptionHandling.html#overview>`_ for the exception. As such,
4983 '``exception``' label is required to have the
4984 ":ref:`landingpad <i_landingpad>`" instruction, which contains the
4985 information about the behavior of the program after unwinding happens,
4986 as its first non-PHI instruction. The restrictions on the
4987 "``landingpad``" instruction's tightly couples it to the "``invoke``"
4988 instruction, so that the important information contained within the
4989 "``landingpad``" instruction can't be lost through normal code motion.
4994 This instruction requires several arguments:
4996 #. The optional "cconv" marker indicates which :ref:`calling
4997 convention <callingconv>` the call should use. If none is
4998 specified, the call defaults to using C calling conventions.
4999 #. The optional :ref:`Parameter Attributes <paramattrs>` list for return
5000 values. Only '``zeroext``', '``signext``', and '``inreg``' attributes
5002 #. '``ptr to function ty``': shall be the signature of the pointer to
5003 function value being invoked. In most cases, this is a direct
5004 function invocation, but indirect ``invoke``'s are just as possible,
5005 branching off an arbitrary pointer to function value.
5006 #. '``function ptr val``': An LLVM value containing a pointer to a
5007 function to be invoked.
5008 #. '``function args``': argument list whose types match the function
5009 signature argument types and parameter attributes. All arguments must
5010 be of :ref:`first class <t_firstclass>` type. If the function signature
5011 indicates the function accepts a variable number of arguments, the
5012 extra arguments can be specified.
5013 #. '``normal label``': the label reached when the called function
5014 executes a '``ret``' instruction.
5015 #. '``exception label``': the label reached when a callee returns via
5016 the :ref:`resume <i_resume>` instruction or other exception handling
5018 #. The optional :ref:`function attributes <fnattrs>` list. Only
5019 '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``'
5020 attributes are valid here.
5025 This instruction is designed to operate as a standard '``call``'
5026 instruction in most regards. The primary difference is that it
5027 establishes an association with a label, which is used by the runtime
5028 library to unwind the stack.
5030 This instruction is used in languages with destructors to ensure that
5031 proper cleanup is performed in the case of either a ``longjmp`` or a
5032 thrown exception. Additionally, this is important for implementation of
5033 '``catch``' clauses in high-level languages that support them.
5035 For the purposes of the SSA form, the definition of the value returned
5036 by the '``invoke``' instruction is deemed to occur on the edge from the
5037 current block to the "normal" label. If the callee unwinds then no
5038 return value is available.
5043 .. code-block:: llvm
5045 %retval = invoke i32 @Test(i32 15) to label %Continue
5046 unwind label %TestCleanup ; i32:retval set
5047 %retval = invoke coldcc i32 %Testfnptr(i32 15) to label %Continue
5048 unwind label %TestCleanup ; i32:retval set
5052 '``resume``' Instruction
5053 ^^^^^^^^^^^^^^^^^^^^^^^^
5060 resume <type> <value>
5065 The '``resume``' instruction is a terminator instruction that has no
5071 The '``resume``' instruction requires one argument, which must have the
5072 same type as the result of any '``landingpad``' instruction in the same
5078 The '``resume``' instruction resumes propagation of an existing
5079 (in-flight) exception whose unwinding was interrupted with a
5080 :ref:`landingpad <i_landingpad>` instruction.
5085 .. code-block:: llvm
5087 resume { i8*, i32 } %exn
5091 '``catchblock``' Instruction
5092 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5099 <resultval> = catchblock <resultty> [<args>*]
5100 to label <normal label> unwind label <exception label>
5105 The '``catchblock``' instruction is used by `LLVM's exception handling
5106 system <ExceptionHandling.html#overview>`_ to specify that a basic block
5107 is a catch block --- one where a personality routine attempts to transfer
5108 control to catch an exception.
5109 The ``args`` correspond to whatever information the personality
5110 routine requires to know if this is an appropriate place to catch the
5111 exception. Control is tranfered to the ``exception`` label if the
5112 ``catchblock`` is not an appropriate handler for the in-flight exception.
5113 The ``normal`` label should contain the code found in the ``catch``
5114 portion of a ``try``/``catch`` sequence. It defines values supplied by
5115 the :ref:`personality function <personalityfn>` upon re-entry to the
5116 function. The ``resultval`` has the type ``resultty``.
5121 The instruction takes a list of arbitrary values which are interpreted
5122 by the :ref:`personality function <personalityfn>`.
5124 The ``catchblock`` must be provided a ``normal`` label to transfer control
5125 to if the ``catchblock`` matches the exception and an ``exception``
5126 label to transfer control to if it doesn't.
5131 The '``catchblock``' instruction defines the values which are set by the
5132 :ref:`personality function <personalityfn>` upon re-entry to the function, and
5133 therefore the "result type" of the ``catchblock`` instruction. As with
5134 calling conventions, how the personality function results are
5135 represented in LLVM IR is target specific.
5137 When the call stack is being unwound due to an exception being thrown,
5138 the exception is compared against the ``args``. If it doesn't match,
5139 then control is transfered to the ``exception`` basic block.
5141 The ``catchblock`` instruction has several restrictions:
5143 - A catch block is a basic block which is the unwind destination of
5144 an exceptional instruction.
5145 - A catch block must have a '``catchblock``' instruction as its
5146 first non-PHI instruction.
5147 - A catch block's ``exception`` edge must refer to a catch block or a
5149 - There can be only one '``catchblock``' instruction within the
5151 - A basic block that is not a catch block may not include a
5152 '``catchblock``' instruction.
5157 .. code-block:: llvm
5159 ;; A catch block which can catch an integer.
5160 %res = catchblock { i8*, i32 } [i8** @_ZTIi]
5161 to label %int.handler unwind label %terminate
5163 .. _i_catchendblock:
5165 '``catchendblock``' Instruction
5166 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5173 catchendblock unwind label <nextaction>
5174 catchendblock unwind to caller
5179 The '``catchendblock``' instruction is used by `LLVM's exception handling
5180 system <ExceptionHandling.html#overview>`_ to communicate to the
5181 :ref:`personality function <personalityfn>` which invokes are associated
5182 with a chain of :ref:`catchblock <i_catchblock>` instructions.
5184 The ``nextaction`` label indicates where control should transfer to if
5185 none of the ``catchblock`` instructions are suitable for catching the
5186 in-flight exception.
5188 If a ``nextaction`` label is not present, the instruction unwinds out of
5189 its parent function. The
5190 :ref:`personality function <personalityfn>` will continue processing
5191 exception handling actions in the caller.
5196 The instruction optionally takes a label, ``nextaction``, indicating
5197 where control should transfer to if none of the preceding
5198 ``catchblock`` instructions are suitable for the in-flight exception.
5203 When the call stack is being unwound due to an exception being thrown
5204 and none of the constituent ``catchblock`` instructions match, then
5205 control is transfered to ``nextaction`` if it is present. If it is not
5206 present, control is transfered to the caller.
5208 The ``catchendblock`` instruction has several restrictions:
5210 - A catch-end block is a basic block which is the unwind destination of
5211 an exceptional instruction.
5212 - A catch-end block must have a '``catchendblock``' instruction as its
5213 first non-PHI instruction.
5214 - There can be only one '``catchendblock``' instruction within the
5216 - A basic block that is not a catch-end block may not include a
5217 '``catchendblock``' instruction.
5218 - Exactly one catch block may unwind to a ``catchendblock``.
5223 .. code-block:: llvm
5225 catchendblock unwind label %terminate
5226 catchendblock unwind to caller
5230 '``catchret``' Instruction
5231 ^^^^^^^^^^^^^^^^^^^^^^^^^^
5238 catchret label <normal>
5243 The '``catchret``' instruction is a terminator instruction that has a
5250 The '``catchret``' instruction requires one argument which specifies
5251 where control will transfer to next.
5256 The '``catchret``' instruction ends the existing (in-flight) exception
5257 whose unwinding was interrupted with a
5258 :ref:`catchblock <i_catchblock>` instruction.
5259 The :ref:`personality function <personalityfn>` gets a chance to execute
5260 arbitrary code to, for example, run a C++ destructor.
5261 Control then transfers to ``normal``.
5266 .. code-block:: llvm
5268 catchret label %continue
5272 '``cleanupret``' Instruction
5273 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5280 cleanupret <type> <value> unwind label <continue>
5281 cleanupret <type> <value> unwind to caller
5286 The '``cleanupret``' instruction is a terminator instruction that has
5287 an optional successor.
5293 The '``cleanupret``' instruction requires one argument, which must have the
5294 same type as the result of any '``cleanupblock``' instruction in the same
5295 function. It also has an optional successor, ``continue``.
5300 The '``cleanupret``' instruction indicates to the
5301 :ref:`personality function <personalityfn>` that one
5302 :ref:`cleanupblock <i_cleanupblock>` it transferred control to has ended.
5303 It transfers control to ``continue`` or unwinds out of the function.
5308 .. code-block:: llvm
5310 cleanupret void unwind to caller
5311 cleanupret { i8*, i32 } %exn unwind label %continue
5313 .. _i_terminateblock:
5315 '``terminateblock``' Instruction
5316 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5323 terminateblock [<args>*] unwind label <exception label>
5324 terminateblock [<args>*] unwind to caller
5329 The '``terminateblock``' instruction is used by `LLVM's exception handling
5330 system <ExceptionHandling.html#overview>`_ to specify that a basic block
5331 is a terminate block --- one where a personality routine may decide to
5332 terminate the program.
5333 The ``args`` correspond to whatever information the personality
5334 routine requires to know if this is an appropriate place to terminate the
5335 program. Control is transferred to the ``exception`` label if the
5336 personality routine decides not to terminate the program for the
5337 in-flight exception.
5342 The instruction takes a list of arbitrary values which are interpreted
5343 by the :ref:`personality function <personalityfn>`.
5345 The ``terminateblock`` may be given an ``exception`` label to
5346 transfer control to if the in-flight exception matches the ``args``.
5351 When the call stack is being unwound due to an exception being thrown,
5352 the exception is compared against the ``args``. If it matches,
5353 then control is transfered to the ``exception`` basic block. Otherwise,
5354 the program is terminated via personality-specific means. Typically,
5355 the first argument to ``terminateblock`` specifies what function the
5356 personality should defer to in order to terminate the program.
5358 The ``terminateblock`` instruction has several restrictions:
5360 - A terminate block is a basic block which is the unwind destination of
5361 an exceptional instruction.
5362 - A terminate block must have a '``terminateblock``' instruction as its
5363 first non-PHI instruction.
5364 - There can be only one '``terminateblock``' instruction within the
5366 - A basic block that is not a terminate block may not include a
5367 '``terminateblock``' instruction.
5372 .. code-block:: llvm
5374 ;; A terminate block which only permits integers.
5375 terminateblock [i8** @_ZTIi] unwind label %continue
5379 '``unreachable``' Instruction
5380 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
5392 The '``unreachable``' instruction has no defined semantics. This
5393 instruction is used to inform the optimizer that a particular portion of
5394 the code is not reachable. This can be used to indicate that the code
5395 after a no-return function cannot be reached, and other facts.
5400 The '``unreachable``' instruction has no defined semantics.
5407 Binary operators are used to do most of the computation in a program.
5408 They require two operands of the same type, execute an operation on
5409 them, and produce a single value. The operands might represent multiple
5410 data, as is the case with the :ref:`vector <t_vector>` data type. The
5411 result value has the same type as its operands.
5413 There are several different binary operators:
5417 '``add``' Instruction
5418 ^^^^^^^^^^^^^^^^^^^^^
5425 <result> = add <ty> <op1>, <op2> ; yields ty:result
5426 <result> = add nuw <ty> <op1>, <op2> ; yields ty:result
5427 <result> = add nsw <ty> <op1>, <op2> ; yields ty:result
5428 <result> = add nuw nsw <ty> <op1>, <op2> ; yields ty:result
5433 The '``add``' instruction returns the sum of its two operands.
5438 The two arguments to the '``add``' instruction must be
5439 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5440 arguments must have identical types.
5445 The value produced is the integer sum of the two operands.
5447 If the sum has unsigned overflow, the result returned is the
5448 mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of
5451 Because LLVM integers use a two's complement representation, this
5452 instruction is appropriate for both signed and unsigned integers.
5454 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
5455 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
5456 result value of the ``add`` is a :ref:`poison value <poisonvalues>` if
5457 unsigned and/or signed overflow, respectively, occurs.
5462 .. code-block:: llvm
5464 <result> = add i32 4, %var ; yields i32:result = 4 + %var
5468 '``fadd``' Instruction
5469 ^^^^^^^^^^^^^^^^^^^^^^
5476 <result> = fadd [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
5481 The '``fadd``' instruction returns the sum of its two operands.
5486 The two arguments to the '``fadd``' instruction must be :ref:`floating
5487 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
5488 Both arguments must have identical types.
5493 The value produced is the floating point sum of the two operands. This
5494 instruction can also take any number of :ref:`fast-math flags <fastmath>`,
5495 which are optimization hints to enable otherwise unsafe floating point
5501 .. code-block:: llvm
5503 <result> = fadd float 4.0, %var ; yields float:result = 4.0 + %var
5505 '``sub``' Instruction
5506 ^^^^^^^^^^^^^^^^^^^^^
5513 <result> = sub <ty> <op1>, <op2> ; yields ty:result
5514 <result> = sub nuw <ty> <op1>, <op2> ; yields ty:result
5515 <result> = sub nsw <ty> <op1>, <op2> ; yields ty:result
5516 <result> = sub nuw nsw <ty> <op1>, <op2> ; yields ty:result
5521 The '``sub``' instruction returns the difference of its two operands.
5523 Note that the '``sub``' instruction is used to represent the '``neg``'
5524 instruction present in most other intermediate representations.
5529 The two arguments to the '``sub``' instruction must be
5530 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5531 arguments must have identical types.
5536 The value produced is the integer difference of the two operands.
5538 If the difference has unsigned overflow, the result returned is the
5539 mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of
5542 Because LLVM integers use a two's complement representation, this
5543 instruction is appropriate for both signed and unsigned integers.
5545 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
5546 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
5547 result value of the ``sub`` is a :ref:`poison value <poisonvalues>` if
5548 unsigned and/or signed overflow, respectively, occurs.
5553 .. code-block:: llvm
5555 <result> = sub i32 4, %var ; yields i32:result = 4 - %var
5556 <result> = sub i32 0, %val ; yields i32:result = -%var
5560 '``fsub``' Instruction
5561 ^^^^^^^^^^^^^^^^^^^^^^
5568 <result> = fsub [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
5573 The '``fsub``' instruction returns the difference of its two operands.
5575 Note that the '``fsub``' instruction is used to represent the '``fneg``'
5576 instruction present in most other intermediate representations.
5581 The two arguments to the '``fsub``' instruction must be :ref:`floating
5582 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
5583 Both arguments must have identical types.
5588 The value produced is the floating point difference of the two operands.
5589 This instruction can also take any number of :ref:`fast-math
5590 flags <fastmath>`, which are optimization hints to enable otherwise
5591 unsafe floating point optimizations:
5596 .. code-block:: llvm
5598 <result> = fsub float 4.0, %var ; yields float:result = 4.0 - %var
5599 <result> = fsub float -0.0, %val ; yields float:result = -%var
5601 '``mul``' Instruction
5602 ^^^^^^^^^^^^^^^^^^^^^
5609 <result> = mul <ty> <op1>, <op2> ; yields ty:result
5610 <result> = mul nuw <ty> <op1>, <op2> ; yields ty:result
5611 <result> = mul nsw <ty> <op1>, <op2> ; yields ty:result
5612 <result> = mul nuw nsw <ty> <op1>, <op2> ; yields ty:result
5617 The '``mul``' instruction returns the product of its two operands.
5622 The two arguments to the '``mul``' instruction must be
5623 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5624 arguments must have identical types.
5629 The value produced is the integer product of the two operands.
5631 If the result of the multiplication has unsigned overflow, the result
5632 returned is the mathematical result modulo 2\ :sup:`n`\ , where n is the
5633 bit width of the result.
5635 Because LLVM integers use a two's complement representation, and the
5636 result is the same width as the operands, this instruction returns the
5637 correct result for both signed and unsigned integers. If a full product
5638 (e.g. ``i32`` * ``i32`` -> ``i64``) is needed, the operands should be
5639 sign-extended or zero-extended as appropriate to the width of the full
5642 ``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap",
5643 respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the
5644 result value of the ``mul`` is a :ref:`poison value <poisonvalues>` if
5645 unsigned and/or signed overflow, respectively, occurs.
5650 .. code-block:: llvm
5652 <result> = mul i32 4, %var ; yields i32:result = 4 * %var
5656 '``fmul``' Instruction
5657 ^^^^^^^^^^^^^^^^^^^^^^
5664 <result> = fmul [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
5669 The '``fmul``' instruction returns the product of its two operands.
5674 The two arguments to the '``fmul``' instruction must be :ref:`floating
5675 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
5676 Both arguments must have identical types.
5681 The value produced is the floating point product of the two operands.
5682 This instruction can also take any number of :ref:`fast-math
5683 flags <fastmath>`, which are optimization hints to enable otherwise
5684 unsafe floating point optimizations:
5689 .. code-block:: llvm
5691 <result> = fmul float 4.0, %var ; yields float:result = 4.0 * %var
5693 '``udiv``' Instruction
5694 ^^^^^^^^^^^^^^^^^^^^^^
5701 <result> = udiv <ty> <op1>, <op2> ; yields ty:result
5702 <result> = udiv exact <ty> <op1>, <op2> ; yields ty:result
5707 The '``udiv``' instruction returns the quotient of its two operands.
5712 The two arguments to the '``udiv``' instruction must be
5713 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5714 arguments must have identical types.
5719 The value produced is the unsigned integer quotient of the two operands.
5721 Note that unsigned integer division and signed integer division are
5722 distinct operations; for signed integer division, use '``sdiv``'.
5724 Division by zero leads to undefined behavior.
5726 If the ``exact`` keyword is present, the result value of the ``udiv`` is
5727 a :ref:`poison value <poisonvalues>` if %op1 is not a multiple of %op2 (as
5728 such, "((a udiv exact b) mul b) == a").
5733 .. code-block:: llvm
5735 <result> = udiv i32 4, %var ; yields i32:result = 4 / %var
5737 '``sdiv``' Instruction
5738 ^^^^^^^^^^^^^^^^^^^^^^
5745 <result> = sdiv <ty> <op1>, <op2> ; yields ty:result
5746 <result> = sdiv exact <ty> <op1>, <op2> ; yields ty:result
5751 The '``sdiv``' instruction returns the quotient of its two operands.
5756 The two arguments to the '``sdiv``' instruction must be
5757 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5758 arguments must have identical types.
5763 The value produced is the signed integer quotient of the two operands
5764 rounded towards zero.
5766 Note that signed integer division and unsigned integer division are
5767 distinct operations; for unsigned integer division, use '``udiv``'.
5769 Division by zero leads to undefined behavior. Overflow also leads to
5770 undefined behavior; this is a rare case, but can occur, for example, by
5771 doing a 32-bit division of -2147483648 by -1.
5773 If the ``exact`` keyword is present, the result value of the ``sdiv`` is
5774 a :ref:`poison value <poisonvalues>` if the result would be rounded.
5779 .. code-block:: llvm
5781 <result> = sdiv i32 4, %var ; yields i32:result = 4 / %var
5785 '``fdiv``' Instruction
5786 ^^^^^^^^^^^^^^^^^^^^^^
5793 <result> = fdiv [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
5798 The '``fdiv``' instruction returns the quotient of its two operands.
5803 The two arguments to the '``fdiv``' instruction must be :ref:`floating
5804 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
5805 Both arguments must have identical types.
5810 The value produced is the floating point quotient of the two operands.
5811 This instruction can also take any number of :ref:`fast-math
5812 flags <fastmath>`, which are optimization hints to enable otherwise
5813 unsafe floating point optimizations:
5818 .. code-block:: llvm
5820 <result> = fdiv float 4.0, %var ; yields float:result = 4.0 / %var
5822 '``urem``' Instruction
5823 ^^^^^^^^^^^^^^^^^^^^^^
5830 <result> = urem <ty> <op1>, <op2> ; yields ty:result
5835 The '``urem``' instruction returns the remainder from the unsigned
5836 division of its two arguments.
5841 The two arguments to the '``urem``' instruction must be
5842 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5843 arguments must have identical types.
5848 This instruction returns the unsigned integer *remainder* of a division.
5849 This instruction always performs an unsigned division to get the
5852 Note that unsigned integer remainder and signed integer remainder are
5853 distinct operations; for signed integer remainder, use '``srem``'.
5855 Taking the remainder of a division by zero leads to undefined behavior.
5860 .. code-block:: llvm
5862 <result> = urem i32 4, %var ; yields i32:result = 4 % %var
5864 '``srem``' Instruction
5865 ^^^^^^^^^^^^^^^^^^^^^^
5872 <result> = srem <ty> <op1>, <op2> ; yields ty:result
5877 The '``srem``' instruction returns the remainder from the signed
5878 division of its two operands. This instruction can also take
5879 :ref:`vector <t_vector>` versions of the values in which case the elements
5885 The two arguments to the '``srem``' instruction must be
5886 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
5887 arguments must have identical types.
5892 This instruction returns the *remainder* of a division (where the result
5893 is either zero or has the same sign as the dividend, ``op1``), not the
5894 *modulo* operator (where the result is either zero or has the same sign
5895 as the divisor, ``op2``) of a value. For more information about the
5896 difference, see `The Math
5897 Forum <http://mathforum.org/dr.math/problems/anne.4.28.99.html>`_. For a
5898 table of how this is implemented in various languages, please see
5900 operation <http://en.wikipedia.org/wiki/Modulo_operation>`_.
5902 Note that signed integer remainder and unsigned integer remainder are
5903 distinct operations; for unsigned integer remainder, use '``urem``'.
5905 Taking the remainder of a division by zero leads to undefined behavior.
5906 Overflow also leads to undefined behavior; this is a rare case, but can
5907 occur, for example, by taking the remainder of a 32-bit division of
5908 -2147483648 by -1. (The remainder doesn't actually overflow, but this
5909 rule lets srem be implemented using instructions that return both the
5910 result of the division and the remainder.)
5915 .. code-block:: llvm
5917 <result> = srem i32 4, %var ; yields i32:result = 4 % %var
5921 '``frem``' Instruction
5922 ^^^^^^^^^^^^^^^^^^^^^^
5929 <result> = frem [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
5934 The '``frem``' instruction returns the remainder from the division of
5940 The two arguments to the '``frem``' instruction must be :ref:`floating
5941 point <t_floating>` or :ref:`vector <t_vector>` of floating point values.
5942 Both arguments must have identical types.
5947 This instruction returns the *remainder* of a division. The remainder
5948 has the same sign as the dividend. This instruction can also take any
5949 number of :ref:`fast-math flags <fastmath>`, which are optimization hints
5950 to enable otherwise unsafe floating point optimizations:
5955 .. code-block:: llvm
5957 <result> = frem float 4.0, %var ; yields float:result = 4.0 % %var
5961 Bitwise Binary Operations
5962 -------------------------
5964 Bitwise binary operators are used to do various forms of bit-twiddling
5965 in a program. They are generally very efficient instructions and can
5966 commonly be strength reduced from other instructions. They require two
5967 operands of the same type, execute an operation on them, and produce a
5968 single value. The resulting value is the same type as its operands.
5970 '``shl``' Instruction
5971 ^^^^^^^^^^^^^^^^^^^^^
5978 <result> = shl <ty> <op1>, <op2> ; yields ty:result
5979 <result> = shl nuw <ty> <op1>, <op2> ; yields ty:result
5980 <result> = shl nsw <ty> <op1>, <op2> ; yields ty:result
5981 <result> = shl nuw nsw <ty> <op1>, <op2> ; yields ty:result
5986 The '``shl``' instruction returns the first operand shifted to the left
5987 a specified number of bits.
5992 Both arguments to the '``shl``' instruction must be the same
5993 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
5994 '``op2``' is treated as an unsigned value.
5999 The value produced is ``op1`` \* 2\ :sup:`op2` mod 2\ :sup:`n`,
6000 where ``n`` is the width of the result. If ``op2`` is (statically or
6001 dynamically) equal to or larger than the number of bits in
6002 ``op1``, the result is undefined. If the arguments are vectors, each
6003 vector element of ``op1`` is shifted by the corresponding shift amount
6006 If the ``nuw`` keyword is present, then the shift produces a :ref:`poison
6007 value <poisonvalues>` if it shifts out any non-zero bits. If the
6008 ``nsw`` keyword is present, then the shift produces a :ref:`poison
6009 value <poisonvalues>` if it shifts out any bits that disagree with the
6010 resultant sign bit. As such, NUW/NSW have the same semantics as they
6011 would if the shift were expressed as a mul instruction with the same
6012 nsw/nuw bits in (mul %op1, (shl 1, %op2)).
6017 .. code-block:: llvm
6019 <result> = shl i32 4, %var ; yields i32: 4 << %var
6020 <result> = shl i32 4, 2 ; yields i32: 16
6021 <result> = shl i32 1, 10 ; yields i32: 1024
6022 <result> = shl i32 1, 32 ; undefined
6023 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 2, i32 4>
6025 '``lshr``' Instruction
6026 ^^^^^^^^^^^^^^^^^^^^^^
6033 <result> = lshr <ty> <op1>, <op2> ; yields ty:result
6034 <result> = lshr exact <ty> <op1>, <op2> ; yields ty:result
6039 The '``lshr``' instruction (logical shift right) returns the first
6040 operand shifted to the right a specified number of bits with zero fill.
6045 Both arguments to the '``lshr``' instruction must be the same
6046 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
6047 '``op2``' is treated as an unsigned value.
6052 This instruction always performs a logical shift right operation. The
6053 most significant bits of the result will be filled with zero bits after
6054 the shift. If ``op2`` is (statically or dynamically) equal to or larger
6055 than the number of bits in ``op1``, the result is undefined. If the
6056 arguments are vectors, each vector element of ``op1`` is shifted by the
6057 corresponding shift amount in ``op2``.
6059 If the ``exact`` keyword is present, the result value of the ``lshr`` is
6060 a :ref:`poison value <poisonvalues>` if any of the bits shifted out are
6066 .. code-block:: llvm
6068 <result> = lshr i32 4, 1 ; yields i32:result = 2
6069 <result> = lshr i32 4, 2 ; yields i32:result = 1
6070 <result> = lshr i8 4, 3 ; yields i8:result = 0
6071 <result> = lshr i8 -2, 1 ; yields i8:result = 0x7F
6072 <result> = lshr i32 1, 32 ; undefined
6073 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1>
6075 '``ashr``' Instruction
6076 ^^^^^^^^^^^^^^^^^^^^^^
6083 <result> = ashr <ty> <op1>, <op2> ; yields ty:result
6084 <result> = ashr exact <ty> <op1>, <op2> ; yields ty:result
6089 The '``ashr``' instruction (arithmetic shift right) returns the first
6090 operand shifted to the right a specified number of bits with sign
6096 Both arguments to the '``ashr``' instruction must be the same
6097 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type.
6098 '``op2``' is treated as an unsigned value.
6103 This instruction always performs an arithmetic shift right operation,
6104 The most significant bits of the result will be filled with the sign bit
6105 of ``op1``. If ``op2`` is (statically or dynamically) equal to or larger
6106 than the number of bits in ``op1``, the result is undefined. If the
6107 arguments are vectors, each vector element of ``op1`` is shifted by the
6108 corresponding shift amount in ``op2``.
6110 If the ``exact`` keyword is present, the result value of the ``ashr`` is
6111 a :ref:`poison value <poisonvalues>` if any of the bits shifted out are
6117 .. code-block:: llvm
6119 <result> = ashr i32 4, 1 ; yields i32:result = 2
6120 <result> = ashr i32 4, 2 ; yields i32:result = 1
6121 <result> = ashr i8 4, 3 ; yields i8:result = 0
6122 <result> = ashr i8 -2, 1 ; yields i8:result = -1
6123 <result> = ashr i32 1, 32 ; undefined
6124 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> ; yields: result=<2 x i32> < i32 -1, i32 0>
6126 '``and``' Instruction
6127 ^^^^^^^^^^^^^^^^^^^^^
6134 <result> = and <ty> <op1>, <op2> ; yields ty:result
6139 The '``and``' instruction returns the bitwise logical and of its two
6145 The two arguments to the '``and``' instruction must be
6146 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6147 arguments must have identical types.
6152 The truth table used for the '``and``' instruction is:
6169 .. code-block:: llvm
6171 <result> = and i32 4, %var ; yields i32:result = 4 & %var
6172 <result> = and i32 15, 40 ; yields i32:result = 8
6173 <result> = and i32 4, 8 ; yields i32:result = 0
6175 '``or``' Instruction
6176 ^^^^^^^^^^^^^^^^^^^^
6183 <result> = or <ty> <op1>, <op2> ; yields ty:result
6188 The '``or``' instruction returns the bitwise logical inclusive or of its
6194 The two arguments to the '``or``' instruction must be
6195 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6196 arguments must have identical types.
6201 The truth table used for the '``or``' instruction is:
6220 <result> = or i32 4, %var ; yields i32:result = 4 | %var
6221 <result> = or i32 15, 40 ; yields i32:result = 47
6222 <result> = or i32 4, 8 ; yields i32:result = 12
6224 '``xor``' Instruction
6225 ^^^^^^^^^^^^^^^^^^^^^
6232 <result> = xor <ty> <op1>, <op2> ; yields ty:result
6237 The '``xor``' instruction returns the bitwise logical exclusive or of
6238 its two operands. The ``xor`` is used to implement the "one's
6239 complement" operation, which is the "~" operator in C.
6244 The two arguments to the '``xor``' instruction must be
6245 :ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both
6246 arguments must have identical types.
6251 The truth table used for the '``xor``' instruction is:
6268 .. code-block:: llvm
6270 <result> = xor i32 4, %var ; yields i32:result = 4 ^ %var
6271 <result> = xor i32 15, 40 ; yields i32:result = 39
6272 <result> = xor i32 4, 8 ; yields i32:result = 12
6273 <result> = xor i32 %V, -1 ; yields i32:result = ~%V
6278 LLVM supports several instructions to represent vector operations in a
6279 target-independent manner. These instructions cover the element-access
6280 and vector-specific operations needed to process vectors effectively.
6281 While LLVM does directly support these vector operations, many
6282 sophisticated algorithms will want to use target-specific intrinsics to
6283 take full advantage of a specific target.
6285 .. _i_extractelement:
6287 '``extractelement``' Instruction
6288 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6295 <result> = extractelement <n x <ty>> <val>, <ty2> <idx> ; yields <ty>
6300 The '``extractelement``' instruction extracts a single scalar element
6301 from a vector at a specified index.
6306 The first operand of an '``extractelement``' instruction is a value of
6307 :ref:`vector <t_vector>` type. The second operand is an index indicating
6308 the position from which to extract the element. The index may be a
6309 variable of any integer type.
6314 The result is a scalar of the same type as the element type of ``val``.
6315 Its value is the value at position ``idx`` of ``val``. If ``idx``
6316 exceeds the length of ``val``, the results are undefined.
6321 .. code-block:: llvm
6323 <result> = extractelement <4 x i32> %vec, i32 0 ; yields i32
6325 .. _i_insertelement:
6327 '``insertelement``' Instruction
6328 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6335 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, <ty2> <idx> ; yields <n x <ty>>
6340 The '``insertelement``' instruction inserts a scalar element into a
6341 vector at a specified index.
6346 The first operand of an '``insertelement``' instruction is a value of
6347 :ref:`vector <t_vector>` type. The second operand is a scalar value whose
6348 type must equal the element type of the first operand. The third operand
6349 is an index indicating the position at which to insert the value. The
6350 index may be a variable of any integer type.
6355 The result is a vector of the same type as ``val``. Its element values
6356 are those of ``val`` except at position ``idx``, where it gets the value
6357 ``elt``. If ``idx`` exceeds the length of ``val``, the results are
6363 .. code-block:: llvm
6365 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 ; yields <4 x i32>
6367 .. _i_shufflevector:
6369 '``shufflevector``' Instruction
6370 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6377 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> ; yields <m x <ty>>
6382 The '``shufflevector``' instruction constructs a permutation of elements
6383 from two input vectors, returning a vector with the same element type as
6384 the input and length that is the same as the shuffle mask.
6389 The first two operands of a '``shufflevector``' instruction are vectors
6390 with the same type. The third argument is a shuffle mask whose element
6391 type is always 'i32'. The result of the instruction is a vector whose
6392 length is the same as the shuffle mask and whose element type is the
6393 same as the element type of the first two operands.
6395 The shuffle mask operand is required to be a constant vector with either
6396 constant integer or undef values.
6401 The elements of the two input vectors are numbered from left to right
6402 across both of the vectors. The shuffle mask operand specifies, for each
6403 element of the result vector, which element of the two input vectors the
6404 result element gets. The element selector may be undef (meaning "don't
6405 care") and the second operand may be undef if performing a shuffle from
6411 .. code-block:: llvm
6413 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
6414 <4 x i32> <i32 0, i32 4, i32 1, i32 5> ; yields <4 x i32>
6415 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
6416 <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32> - Identity shuffle.
6417 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
6418 <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32>
6419 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
6420 <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 > ; yields <8 x i32>
6422 Aggregate Operations
6423 --------------------
6425 LLVM supports several instructions for working with
6426 :ref:`aggregate <t_aggregate>` values.
6430 '``extractvalue``' Instruction
6431 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6438 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
6443 The '``extractvalue``' instruction extracts the value of a member field
6444 from an :ref:`aggregate <t_aggregate>` value.
6449 The first operand of an '``extractvalue``' instruction is a value of
6450 :ref:`struct <t_struct>` or :ref:`array <t_array>` type. The operands are
6451 constant indices to specify which value to extract in a similar manner
6452 as indices in a '``getelementptr``' instruction.
6454 The major differences to ``getelementptr`` indexing are:
6456 - Since the value being indexed is not a pointer, the first index is
6457 omitted and assumed to be zero.
6458 - At least one index must be specified.
6459 - Not only struct indices but also array indices must be in bounds.
6464 The result is the value at the position in the aggregate specified by
6470 .. code-block:: llvm
6472 <result> = extractvalue {i32, float} %agg, 0 ; yields i32
6476 '``insertvalue``' Instruction
6477 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6484 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* ; yields <aggregate type>
6489 The '``insertvalue``' instruction inserts a value into a member field in
6490 an :ref:`aggregate <t_aggregate>` value.
6495 The first operand of an '``insertvalue``' instruction is a value of
6496 :ref:`struct <t_struct>` or :ref:`array <t_array>` type. The second operand is
6497 a first-class value to insert. The following operands are constant
6498 indices indicating the position at which to insert the value in a
6499 similar manner as indices in a '``extractvalue``' instruction. The value
6500 to insert must have the same type as the value identified by the
6506 The result is an aggregate of the same type as ``val``. Its value is
6507 that of ``val`` except that the value at the position specified by the
6508 indices is that of ``elt``.
6513 .. code-block:: llvm
6515 %agg1 = insertvalue {i32, float} undef, i32 1, 0 ; yields {i32 1, float undef}
6516 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 ; yields {i32 1, float %val}
6517 %agg3 = insertvalue {i32, {float}} undef, float %val, 1, 0 ; yields {i32 undef, {float %val}}
6521 Memory Access and Addressing Operations
6522 ---------------------------------------
6524 A key design point of an SSA-based representation is how it represents
6525 memory. In LLVM, no memory locations are in SSA form, which makes things
6526 very simple. This section describes how to read, write, and allocate
6531 '``alloca``' Instruction
6532 ^^^^^^^^^^^^^^^^^^^^^^^^
6539 <result> = alloca [inalloca] <type> [, <ty> <NumElements>] [, align <alignment>] ; yields type*:result
6544 The '``alloca``' instruction allocates memory on the stack frame of the
6545 currently executing function, to be automatically released when this
6546 function returns to its caller. The object is always allocated in the
6547 generic address space (address space zero).
6552 The '``alloca``' instruction allocates ``sizeof(<type>)*NumElements``
6553 bytes of memory on the runtime stack, returning a pointer of the
6554 appropriate type to the program. If "NumElements" is specified, it is
6555 the number of elements allocated, otherwise "NumElements" is defaulted
6556 to be one. If a constant alignment is specified, the value result of the
6557 allocation is guaranteed to be aligned to at least that boundary. The
6558 alignment may not be greater than ``1 << 29``. If not specified, or if
6559 zero, the target can choose to align the allocation on any convenient
6560 boundary compatible with the type.
6562 '``type``' may be any sized type.
6567 Memory is allocated; a pointer is returned. The operation is undefined
6568 if there is insufficient stack space for the allocation. '``alloca``'d
6569 memory is automatically released when the function returns. The
6570 '``alloca``' instruction is commonly used to represent automatic
6571 variables that must have an address available. When the function returns
6572 (either with the ``ret`` or ``resume`` instructions), the memory is
6573 reclaimed. Allocating zero bytes is legal, but the result is undefined.
6574 The order in which memory is allocated (ie., which way the stack grows)
6580 .. code-block:: llvm
6582 %ptr = alloca i32 ; yields i32*:ptr
6583 %ptr = alloca i32, i32 4 ; yields i32*:ptr
6584 %ptr = alloca i32, i32 4, align 1024 ; yields i32*:ptr
6585 %ptr = alloca i32, align 1024 ; yields i32*:ptr
6589 '``load``' Instruction
6590 ^^^^^^^^^^^^^^^^^^^^^^
6597 <result> = load [volatile] <ty>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>][, !invariant.load !<index>][, !nonnull !<index>][, !dereferenceable !<index>][, !dereferenceable_or_null !<index>]
6598 <result> = load atomic [volatile] <ty>* <pointer> [singlethread] <ordering>, align <alignment>
6599 !<index> = !{ i32 1 }
6604 The '``load``' instruction is used to read from memory.
6609 The argument to the ``load`` instruction specifies the memory address
6610 from which to load. The type specified must be a :ref:`first
6611 class <t_firstclass>` type. If the ``load`` is marked as ``volatile``,
6612 then the optimizer is not allowed to modify the number or order of
6613 execution of this ``load`` with other :ref:`volatile
6614 operations <volatile>`.
6616 If the ``load`` is marked as ``atomic``, it takes an extra
6617 :ref:`ordering <ordering>` and optional ``singlethread`` argument. The
6618 ``release`` and ``acq_rel`` orderings are not valid on ``load``
6619 instructions. Atomic loads produce :ref:`defined <memmodel>` results
6620 when they may see multiple atomic stores. The type of the pointee must
6621 be an integer type whose bit width is a power of two greater than or
6622 equal to eight and less than or equal to a target-specific size limit.
6623 ``align`` must be explicitly specified on atomic loads, and the load has
6624 undefined behavior if the alignment is not set to a value which is at
6625 least the size in bytes of the pointee. ``!nontemporal`` does not have
6626 any defined semantics for atomic loads.
6628 The optional constant ``align`` argument specifies the alignment of the
6629 operation (that is, the alignment of the memory address). A value of 0
6630 or an omitted ``align`` argument means that the operation has the ABI
6631 alignment for the target. It is the responsibility of the code emitter
6632 to ensure that the alignment information is correct. Overestimating the
6633 alignment results in undefined behavior. Underestimating the alignment
6634 may produce less efficient code. An alignment of 1 is always safe. The
6635 maximum possible alignment is ``1 << 29``.
6637 The optional ``!nontemporal`` metadata must reference a single
6638 metadata name ``<index>`` corresponding to a metadata node with one
6639 ``i32`` entry of value 1. The existence of the ``!nontemporal``
6640 metadata on the instruction tells the optimizer and code generator
6641 that this load is not expected to be reused in the cache. The code
6642 generator may select special instructions to save cache bandwidth, such
6643 as the ``MOVNT`` instruction on x86.
6645 The optional ``!invariant.load`` metadata must reference a single
6646 metadata name ``<index>`` corresponding to a metadata node with no
6647 entries. The existence of the ``!invariant.load`` metadata on the
6648 instruction tells the optimizer and code generator that the address
6649 operand to this load points to memory which can be assumed unchanged.
6650 Being invariant does not imply that a location is dereferenceable,
6651 but it does imply that once the location is known dereferenceable
6652 its value is henceforth unchanging.
6654 The optional ``!nonnull`` metadata must reference a single
6655 metadata name ``<index>`` corresponding to a metadata node with no
6656 entries. The existence of the ``!nonnull`` metadata on the
6657 instruction tells the optimizer that the value loaded is known to
6658 never be null. This is analogous to the ''nonnull'' attribute
6659 on parameters and return values. This metadata can only be applied
6660 to loads of a pointer type.
6662 The optional ``!dereferenceable`` metadata must reference a single
6663 metadata name ``<index>`` corresponding to a metadata node with one ``i64``
6664 entry. The existence of the ``!dereferenceable`` metadata on the instruction
6665 tells the optimizer that the value loaded is known to be dereferenceable.
6666 The number of bytes known to be dereferenceable is specified by the integer
6667 value in the metadata node. This is analogous to the ''dereferenceable''
6668 attribute on parameters and return values. This metadata can only be applied
6669 to loads of a pointer type.
6671 The optional ``!dereferenceable_or_null`` metadata must reference a single
6672 metadata name ``<index>`` corresponding to a metadata node with one ``i64``
6673 entry. The existence of the ``!dereferenceable_or_null`` metadata on the
6674 instruction tells the optimizer that the value loaded is known to be either
6675 dereferenceable or null.
6676 The number of bytes known to be dereferenceable is specified by the integer
6677 value in the metadata node. This is analogous to the ''dereferenceable_or_null''
6678 attribute on parameters and return values. This metadata can only be applied
6679 to loads of a pointer type.
6684 The location of memory pointed to is loaded. If the value being loaded
6685 is of scalar type then the number of bytes read does not exceed the
6686 minimum number of bytes needed to hold all bits of the type. For
6687 example, loading an ``i24`` reads at most three bytes. When loading a
6688 value of a type like ``i20`` with a size that is not an integral number
6689 of bytes, the result is undefined if the value was not originally
6690 written using a store of the same type.
6695 .. code-block:: llvm
6697 %ptr = alloca i32 ; yields i32*:ptr
6698 store i32 3, i32* %ptr ; yields void
6699 %val = load i32, i32* %ptr ; yields i32:val = i32 3
6703 '``store``' Instruction
6704 ^^^^^^^^^^^^^^^^^^^^^^^
6711 store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] ; yields void
6712 store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> ; yields void
6717 The '``store``' instruction is used to write to memory.
6722 There are two arguments to the ``store`` instruction: a value to store
6723 and an address at which to store it. The type of the ``<pointer>``
6724 operand must be a pointer to the :ref:`first class <t_firstclass>` type of
6725 the ``<value>`` operand. If the ``store`` is marked as ``volatile``,
6726 then the optimizer is not allowed to modify the number or order of
6727 execution of this ``store`` with other :ref:`volatile
6728 operations <volatile>`.
6730 If the ``store`` is marked as ``atomic``, it takes an extra
6731 :ref:`ordering <ordering>` and optional ``singlethread`` argument. The
6732 ``acquire`` and ``acq_rel`` orderings aren't valid on ``store``
6733 instructions. Atomic loads produce :ref:`defined <memmodel>` results
6734 when they may see multiple atomic stores. The type of the pointee must
6735 be an integer type whose bit width is a power of two greater than or
6736 equal to eight and less than or equal to a target-specific size limit.
6737 ``align`` must be explicitly specified on atomic stores, and the store
6738 has undefined behavior if the alignment is not set to a value which is
6739 at least the size in bytes of the pointee. ``!nontemporal`` does not
6740 have any defined semantics for atomic stores.
6742 The optional constant ``align`` argument specifies the alignment of the
6743 operation (that is, the alignment of the memory address). A value of 0
6744 or an omitted ``align`` argument means that the operation has the ABI
6745 alignment for the target. It is the responsibility of the code emitter
6746 to ensure that the alignment information is correct. Overestimating the
6747 alignment results in undefined behavior. Underestimating the
6748 alignment may produce less efficient code. An alignment of 1 is always
6749 safe. The maximum possible alignment is ``1 << 29``.
6751 The optional ``!nontemporal`` metadata must reference a single metadata
6752 name ``<index>`` corresponding to a metadata node with one ``i32`` entry of
6753 value 1. The existence of the ``!nontemporal`` metadata on the instruction
6754 tells the optimizer and code generator that this load is not expected to
6755 be reused in the cache. The code generator may select special
6756 instructions to save cache bandwidth, such as the MOVNT instruction on
6762 The contents of memory are updated to contain ``<value>`` at the
6763 location specified by the ``<pointer>`` operand. If ``<value>`` is
6764 of scalar type then the number of bytes written does not exceed the
6765 minimum number of bytes needed to hold all bits of the type. For
6766 example, storing an ``i24`` writes at most three bytes. When writing a
6767 value of a type like ``i20`` with a size that is not an integral number
6768 of bytes, it is unspecified what happens to the extra bits that do not
6769 belong to the type, but they will typically be overwritten.
6774 .. code-block:: llvm
6776 %ptr = alloca i32 ; yields i32*:ptr
6777 store i32 3, i32* %ptr ; yields void
6778 %val = load i32* %ptr ; yields i32:val = i32 3
6782 '``fence``' Instruction
6783 ^^^^^^^^^^^^^^^^^^^^^^^
6790 fence [singlethread] <ordering> ; yields void
6795 The '``fence``' instruction is used to introduce happens-before edges
6801 '``fence``' instructions take an :ref:`ordering <ordering>` argument which
6802 defines what *synchronizes-with* edges they add. They can only be given
6803 ``acquire``, ``release``, ``acq_rel``, and ``seq_cst`` orderings.
6808 A fence A which has (at least) ``release`` ordering semantics
6809 *synchronizes with* a fence B with (at least) ``acquire`` ordering
6810 semantics if and only if there exist atomic operations X and Y, both
6811 operating on some atomic object M, such that A is sequenced before X, X
6812 modifies M (either directly or through some side effect of a sequence
6813 headed by X), Y is sequenced before B, and Y observes M. This provides a
6814 *happens-before* dependency between A and B. Rather than an explicit
6815 ``fence``, one (but not both) of the atomic operations X or Y might
6816 provide a ``release`` or ``acquire`` (resp.) ordering constraint and
6817 still *synchronize-with* the explicit ``fence`` and establish the
6818 *happens-before* edge.
6820 A ``fence`` which has ``seq_cst`` ordering, in addition to having both
6821 ``acquire`` and ``release`` semantics specified above, participates in
6822 the global program order of other ``seq_cst`` operations and/or fences.
6824 The optional ":ref:`singlethread <singlethread>`" argument specifies
6825 that the fence only synchronizes with other fences in the same thread.
6826 (This is useful for interacting with signal handlers.)
6831 .. code-block:: llvm
6833 fence acquire ; yields void
6834 fence singlethread seq_cst ; yields void
6838 '``cmpxchg``' Instruction
6839 ^^^^^^^^^^^^^^^^^^^^^^^^^
6846 cmpxchg [weak] [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <success ordering> <failure ordering> ; yields { ty, i1 }
6851 The '``cmpxchg``' instruction is used to atomically modify memory. It
6852 loads a value in memory and compares it to a given value. If they are
6853 equal, it tries to store a new value into the memory.
6858 There are three arguments to the '``cmpxchg``' instruction: an address
6859 to operate on, a value to compare to the value currently be at that
6860 address, and a new value to place at that address if the compared values
6861 are equal. The type of '<cmp>' must be an integer type whose bit width
6862 is a power of two greater than or equal to eight and less than or equal
6863 to a target-specific size limit. '<cmp>' and '<new>' must have the same
6864 type, and the type of '<pointer>' must be a pointer to that type. If the
6865 ``cmpxchg`` is marked as ``volatile``, then the optimizer is not allowed
6866 to modify the number or order of execution of this ``cmpxchg`` with
6867 other :ref:`volatile operations <volatile>`.
6869 The success and failure :ref:`ordering <ordering>` arguments specify how this
6870 ``cmpxchg`` synchronizes with other atomic operations. Both ordering parameters
6871 must be at least ``monotonic``, the ordering constraint on failure must be no
6872 stronger than that on success, and the failure ordering cannot be either
6873 ``release`` or ``acq_rel``.
6875 The optional "``singlethread``" argument declares that the ``cmpxchg``
6876 is only atomic with respect to code (usually signal handlers) running in
6877 the same thread as the ``cmpxchg``. Otherwise the cmpxchg is atomic with
6878 respect to all other code in the system.
6880 The pointer passed into cmpxchg must have alignment greater than or
6881 equal to the size in memory of the operand.
6886 The contents of memory at the location specified by the '``<pointer>``' operand
6887 is read and compared to '``<cmp>``'; if the read value is the equal, the
6888 '``<new>``' is written. The original value at the location is returned, together
6889 with a flag indicating success (true) or failure (false).
6891 If the cmpxchg operation is marked as ``weak`` then a spurious failure is
6892 permitted: the operation may not write ``<new>`` even if the comparison
6895 If the cmpxchg operation is strong (the default), the i1 value is 1 if and only
6896 if the value loaded equals ``cmp``.
6898 A successful ``cmpxchg`` is a read-modify-write instruction for the purpose of
6899 identifying release sequences. A failed ``cmpxchg`` is equivalent to an atomic
6900 load with an ordering parameter determined the second ordering parameter.
6905 .. code-block:: llvm
6908 %orig = atomic load i32, i32* %ptr unordered ; yields i32
6912 %cmp = phi i32 [ %orig, %entry ], [%old, %loop]
6913 %squared = mul i32 %cmp, %cmp
6914 %val_success = cmpxchg i32* %ptr, i32 %cmp, i32 %squared acq_rel monotonic ; yields { i32, i1 }
6915 %value_loaded = extractvalue { i32, i1 } %val_success, 0
6916 %success = extractvalue { i32, i1 } %val_success, 1
6917 br i1 %success, label %done, label %loop
6924 '``atomicrmw``' Instruction
6925 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
6932 atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering> ; yields ty
6937 The '``atomicrmw``' instruction is used to atomically modify memory.
6942 There are three arguments to the '``atomicrmw``' instruction: an
6943 operation to apply, an address whose value to modify, an argument to the
6944 operation. The operation must be one of the following keywords:
6958 The type of '<value>' must be an integer type whose bit width is a power
6959 of two greater than or equal to eight and less than or equal to a
6960 target-specific size limit. The type of the '``<pointer>``' operand must
6961 be a pointer to that type. If the ``atomicrmw`` is marked as
6962 ``volatile``, then the optimizer is not allowed to modify the number or
6963 order of execution of this ``atomicrmw`` with other :ref:`volatile
6964 operations <volatile>`.
6969 The contents of memory at the location specified by the '``<pointer>``'
6970 operand are atomically read, modified, and written back. The original
6971 value at the location is returned. The modification is specified by the
6974 - xchg: ``*ptr = val``
6975 - add: ``*ptr = *ptr + val``
6976 - sub: ``*ptr = *ptr - val``
6977 - and: ``*ptr = *ptr & val``
6978 - nand: ``*ptr = ~(*ptr & val)``
6979 - or: ``*ptr = *ptr | val``
6980 - xor: ``*ptr = *ptr ^ val``
6981 - max: ``*ptr = *ptr > val ? *ptr : val`` (using a signed comparison)
6982 - min: ``*ptr = *ptr < val ? *ptr : val`` (using a signed comparison)
6983 - umax: ``*ptr = *ptr > val ? *ptr : val`` (using an unsigned
6985 - umin: ``*ptr = *ptr < val ? *ptr : val`` (using an unsigned
6991 .. code-block:: llvm
6993 %old = atomicrmw add i32* %ptr, i32 1 acquire ; yields i32
6995 .. _i_getelementptr:
6997 '``getelementptr``' Instruction
6998 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7005 <result> = getelementptr <ty>, <ty>* <ptrval>{, <ty> <idx>}*
7006 <result> = getelementptr inbounds <ty>, <ty>* <ptrval>{, <ty> <idx>}*
7007 <result> = getelementptr <ty>, <ptr vector> <ptrval>, <vector index type> <idx>
7012 The '``getelementptr``' instruction is used to get the address of a
7013 subelement of an :ref:`aggregate <t_aggregate>` data structure. It performs
7014 address calculation only and does not access memory. The instruction can also
7015 be used to calculate a vector of such addresses.
7020 The first argument is always a type used as the basis for the calculations.
7021 The second argument is always a pointer or a vector of pointers, and is the
7022 base address to start from. The remaining arguments are indices
7023 that indicate which of the elements of the aggregate object are indexed.
7024 The interpretation of each index is dependent on the type being indexed
7025 into. The first index always indexes the pointer value given as the
7026 first argument, the second index indexes a value of the type pointed to
7027 (not necessarily the value directly pointed to, since the first index
7028 can be non-zero), etc. The first type indexed into must be a pointer
7029 value, subsequent types can be arrays, vectors, and structs. Note that
7030 subsequent types being indexed into can never be pointers, since that
7031 would require loading the pointer before continuing calculation.
7033 The type of each index argument depends on the type it is indexing into.
7034 When indexing into a (optionally packed) structure, only ``i32`` integer
7035 **constants** are allowed (when using a vector of indices they must all
7036 be the **same** ``i32`` integer constant). When indexing into an array,
7037 pointer or vector, integers of any width are allowed, and they are not
7038 required to be constant. These integers are treated as signed values
7041 For example, let's consider a C code fragment and how it gets compiled
7057 int *foo(struct ST *s) {
7058 return &s[1].Z.B[5][13];
7061 The LLVM code generated by Clang is:
7063 .. code-block:: llvm
7065 %struct.RT = type { i8, [10 x [20 x i32]], i8 }
7066 %struct.ST = type { i32, double, %struct.RT }
7068 define i32* @foo(%struct.ST* %s) nounwind uwtable readnone optsize ssp {
7070 %arrayidx = getelementptr inbounds %struct.ST, %struct.ST* %s, i64 1, i32 2, i32 1, i64 5, i64 13
7077 In the example above, the first index is indexing into the
7078 '``%struct.ST*``' type, which is a pointer, yielding a '``%struct.ST``'
7079 = '``{ i32, double, %struct.RT }``' type, a structure. The second index
7080 indexes into the third element of the structure, yielding a
7081 '``%struct.RT``' = '``{ i8 , [10 x [20 x i32]], i8 }``' type, another
7082 structure. The third index indexes into the second element of the
7083 structure, yielding a '``[10 x [20 x i32]]``' type, an array. The two
7084 dimensions of the array are subscripted into, yielding an '``i32``'
7085 type. The '``getelementptr``' instruction returns a pointer to this
7086 element, thus computing a value of '``i32*``' type.
7088 Note that it is perfectly legal to index partially through a structure,
7089 returning a pointer to an inner element. Because of this, the LLVM code
7090 for the given testcase is equivalent to:
7092 .. code-block:: llvm
7094 define i32* @foo(%struct.ST* %s) {
7095 %t1 = getelementptr %struct.ST, %struct.ST* %s, i32 1 ; yields %struct.ST*:%t1
7096 %t2 = getelementptr %struct.ST, %struct.ST* %t1, i32 0, i32 2 ; yields %struct.RT*:%t2
7097 %t3 = getelementptr %struct.RT, %struct.RT* %t2, i32 0, i32 1 ; yields [10 x [20 x i32]]*:%t3
7098 %t4 = getelementptr [10 x [20 x i32]], [10 x [20 x i32]]* %t3, i32 0, i32 5 ; yields [20 x i32]*:%t4
7099 %t5 = getelementptr [20 x i32], [20 x i32]* %t4, i32 0, i32 13 ; yields i32*:%t5
7103 If the ``inbounds`` keyword is present, the result value of the
7104 ``getelementptr`` is a :ref:`poison value <poisonvalues>` if the base
7105 pointer is not an *in bounds* address of an allocated object, or if any
7106 of the addresses that would be formed by successive addition of the
7107 offsets implied by the indices to the base address with infinitely
7108 precise signed arithmetic are not an *in bounds* address of that
7109 allocated object. The *in bounds* addresses for an allocated object are
7110 all the addresses that point into the object, plus the address one byte
7111 past the end. In cases where the base is a vector of pointers the
7112 ``inbounds`` keyword applies to each of the computations element-wise.
7114 If the ``inbounds`` keyword is not present, the offsets are added to the
7115 base address with silently-wrapping two's complement arithmetic. If the
7116 offsets have a different width from the pointer, they are sign-extended
7117 or truncated to the width of the pointer. The result value of the
7118 ``getelementptr`` may be outside the object pointed to by the base
7119 pointer. The result value may not necessarily be used to access memory
7120 though, even if it happens to point into allocated storage. See the
7121 :ref:`Pointer Aliasing Rules <pointeraliasing>` section for more
7124 The getelementptr instruction is often confusing. For some more insight
7125 into how it works, see :doc:`the getelementptr FAQ <GetElementPtr>`.
7130 .. code-block:: llvm
7132 ; yields [12 x i8]*:aptr
7133 %aptr = getelementptr {i32, [12 x i8]}, {i32, [12 x i8]}* %saptr, i64 0, i32 1
7135 %vptr = getelementptr {i32, <2 x i8>}, {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
7137 %eptr = getelementptr [12 x i8], [12 x i8]* %aptr, i64 0, i32 1
7139 %iptr = getelementptr [10 x i32], [10 x i32]* @arr, i16 0, i16 0
7144 The ``getelementptr`` returns a vector of pointers, instead of a single address,
7145 when one or more of its arguments is a vector. In such cases, all vector
7146 arguments should have the same number of elements, and every scalar argument
7147 will be effectively broadcast into a vector during address calculation.
7149 .. code-block:: llvm
7151 ; All arguments are vectors:
7152 ; A[i] = ptrs[i] + offsets[i]*sizeof(i8)
7153 %A = getelementptr i8, <4 x i8*> %ptrs, <4 x i64> %offsets
7155 ; Add the same scalar offset to each pointer of a vector:
7156 ; A[i] = ptrs[i] + offset*sizeof(i8)
7157 %A = getelementptr i8, <4 x i8*> %ptrs, i64 %offset
7159 ; Add distinct offsets to the same pointer:
7160 ; A[i] = ptr + offsets[i]*sizeof(i8)
7161 %A = getelementptr i8, i8* %ptr, <4 x i64> %offsets
7163 ; In all cases described above the type of the result is <4 x i8*>
7165 The two following instructions are equivalent:
7167 .. code-block:: llvm
7169 getelementptr %struct.ST, <4 x %struct.ST*> %s, <4 x i64> %ind1,
7170 <4 x i32> <i32 2, i32 2, i32 2, i32 2>,
7171 <4 x i32> <i32 1, i32 1, i32 1, i32 1>,
7173 <4 x i64> <i64 13, i64 13, i64 13, i64 13>
7175 getelementptr %struct.ST, <4 x %struct.ST*> %s, <4 x i64> %ind1,
7176 i32 2, i32 1, <4 x i32> %ind4, i64 13
7178 Let's look at the C code, where the vector version of ``getelementptr``
7183 // Let's assume that we vectorize the following loop:
7184 double *A, B; int *C;
7185 for (int i = 0; i < size; ++i) {
7189 .. code-block:: llvm
7191 ; get pointers for 8 elements from array B
7192 %ptrs = getelementptr double, double* %B, <8 x i32> %C
7193 ; load 8 elements from array B into A
7194 %A = call <8 x double> @llvm.masked.gather.v8f64(<8 x double*> %ptrs,
7195 i32 8, <8 x i1> %mask, <8 x double> %passthru)
7197 Conversion Operations
7198 ---------------------
7200 The instructions in this category are the conversion instructions
7201 (casting) which all take a single operand and a type. They perform
7202 various bit conversions on the operand.
7204 '``trunc .. to``' Instruction
7205 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7212 <result> = trunc <ty> <value> to <ty2> ; yields ty2
7217 The '``trunc``' instruction truncates its operand to the type ``ty2``.
7222 The '``trunc``' instruction takes a value to trunc, and a type to trunc
7223 it to. Both types must be of :ref:`integer <t_integer>` types, or vectors
7224 of the same number of integers. The bit size of the ``value`` must be
7225 larger than the bit size of the destination type, ``ty2``. Equal sized
7226 types are not allowed.
7231 The '``trunc``' instruction truncates the high order bits in ``value``
7232 and converts the remaining bits to ``ty2``. Since the source size must
7233 be larger than the destination size, ``trunc`` cannot be a *no-op cast*.
7234 It will always truncate bits.
7239 .. code-block:: llvm
7241 %X = trunc i32 257 to i8 ; yields i8:1
7242 %Y = trunc i32 123 to i1 ; yields i1:true
7243 %Z = trunc i32 122 to i1 ; yields i1:false
7244 %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> ; yields <i8 8, i8 7>
7246 '``zext .. to``' Instruction
7247 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7254 <result> = zext <ty> <value> to <ty2> ; yields ty2
7259 The '``zext``' instruction zero extends its operand to type ``ty2``.
7264 The '``zext``' instruction takes a value to cast, and a type to cast it
7265 to. Both types must be of :ref:`integer <t_integer>` types, or vectors of
7266 the same number of integers. The bit size of the ``value`` must be
7267 smaller than the bit size of the destination type, ``ty2``.
7272 The ``zext`` fills the high order bits of the ``value`` with zero bits
7273 until it reaches the size of the destination type, ``ty2``.
7275 When zero extending from i1, the result will always be either 0 or 1.
7280 .. code-block:: llvm
7282 %X = zext i32 257 to i64 ; yields i64:257
7283 %Y = zext i1 true to i32 ; yields i32:1
7284 %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7>
7286 '``sext .. to``' Instruction
7287 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7294 <result> = sext <ty> <value> to <ty2> ; yields ty2
7299 The '``sext``' sign extends ``value`` to the type ``ty2``.
7304 The '``sext``' instruction takes a value to cast, and a type to cast it
7305 to. Both types must be of :ref:`integer <t_integer>` types, or vectors of
7306 the same number of integers. The bit size of the ``value`` must be
7307 smaller than the bit size of the destination type, ``ty2``.
7312 The '``sext``' instruction performs a sign extension by copying the sign
7313 bit (highest order bit) of the ``value`` until it reaches the bit size
7314 of the type ``ty2``.
7316 When sign extending from i1, the extension always results in -1 or 0.
7321 .. code-block:: llvm
7323 %X = sext i8 -1 to i16 ; yields i16 :65535
7324 %Y = sext i1 true to i32 ; yields i32:-1
7325 %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7>
7327 '``fptrunc .. to``' Instruction
7328 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7335 <result> = fptrunc <ty> <value> to <ty2> ; yields ty2
7340 The '``fptrunc``' instruction truncates ``value`` to type ``ty2``.
7345 The '``fptrunc``' instruction takes a :ref:`floating point <t_floating>`
7346 value to cast and a :ref:`floating point <t_floating>` type to cast it to.
7347 The size of ``value`` must be larger than the size of ``ty2``. This
7348 implies that ``fptrunc`` cannot be used to make a *no-op cast*.
7353 The '``fptrunc``' instruction truncates a ``value`` from a larger
7354 :ref:`floating point <t_floating>` type to a smaller :ref:`floating
7355 point <t_floating>` type. If the value cannot fit within the
7356 destination type, ``ty2``, then the results are undefined.
7361 .. code-block:: llvm
7363 %X = fptrunc double 123.0 to float ; yields float:123.0
7364 %Y = fptrunc double 1.0E+300 to float ; yields undefined
7366 '``fpext .. to``' Instruction
7367 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7374 <result> = fpext <ty> <value> to <ty2> ; yields ty2
7379 The '``fpext``' extends a floating point ``value`` to a larger floating
7385 The '``fpext``' instruction takes a :ref:`floating point <t_floating>`
7386 ``value`` to cast, and a :ref:`floating point <t_floating>` type to cast it
7387 to. The source type must be smaller than the destination type.
7392 The '``fpext``' instruction extends the ``value`` from a smaller
7393 :ref:`floating point <t_floating>` type to a larger :ref:`floating
7394 point <t_floating>` type. The ``fpext`` cannot be used to make a
7395 *no-op cast* because it always changes bits. Use ``bitcast`` to make a
7396 *no-op cast* for a floating point cast.
7401 .. code-block:: llvm
7403 %X = fpext float 3.125 to double ; yields double:3.125000e+00
7404 %Y = fpext double %X to fp128 ; yields fp128:0xL00000000000000004000900000000000
7406 '``fptoui .. to``' Instruction
7407 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7414 <result> = fptoui <ty> <value> to <ty2> ; yields ty2
7419 The '``fptoui``' converts a floating point ``value`` to its unsigned
7420 integer equivalent of type ``ty2``.
7425 The '``fptoui``' instruction takes a value to cast, which must be a
7426 scalar or vector :ref:`floating point <t_floating>` value, and a type to
7427 cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If
7428 ``ty`` is a vector floating point type, ``ty2`` must be a vector integer
7429 type with the same number of elements as ``ty``
7434 The '``fptoui``' instruction converts its :ref:`floating
7435 point <t_floating>` operand into the nearest (rounding towards zero)
7436 unsigned integer value. If the value cannot fit in ``ty2``, the results
7442 .. code-block:: llvm
7444 %X = fptoui double 123.0 to i32 ; yields i32:123
7445 %Y = fptoui float 1.0E+300 to i1 ; yields undefined:1
7446 %Z = fptoui float 1.04E+17 to i8 ; yields undefined:1
7448 '``fptosi .. to``' Instruction
7449 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7456 <result> = fptosi <ty> <value> to <ty2> ; yields ty2
7461 The '``fptosi``' instruction converts :ref:`floating point <t_floating>`
7462 ``value`` to type ``ty2``.
7467 The '``fptosi``' instruction takes a value to cast, which must be a
7468 scalar or vector :ref:`floating point <t_floating>` value, and a type to
7469 cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If
7470 ``ty`` is a vector floating point type, ``ty2`` must be a vector integer
7471 type with the same number of elements as ``ty``
7476 The '``fptosi``' instruction converts its :ref:`floating
7477 point <t_floating>` operand into the nearest (rounding towards zero)
7478 signed integer value. If the value cannot fit in ``ty2``, the results
7484 .. code-block:: llvm
7486 %X = fptosi double -123.0 to i32 ; yields i32:-123
7487 %Y = fptosi float 1.0E-247 to i1 ; yields undefined:1
7488 %Z = fptosi float 1.04E+17 to i8 ; yields undefined:1
7490 '``uitofp .. to``' Instruction
7491 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7498 <result> = uitofp <ty> <value> to <ty2> ; yields ty2
7503 The '``uitofp``' instruction regards ``value`` as an unsigned integer
7504 and converts that value to the ``ty2`` type.
7509 The '``uitofp``' instruction takes a value to cast, which must be a
7510 scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to
7511 ``ty2``, which must be an :ref:`floating point <t_floating>` type. If
7512 ``ty`` is a vector integer type, ``ty2`` must be a vector floating point
7513 type with the same number of elements as ``ty``
7518 The '``uitofp``' instruction interprets its operand as an unsigned
7519 integer quantity and converts it to the corresponding floating point
7520 value. If the value cannot fit in the floating point value, the results
7526 .. code-block:: llvm
7528 %X = uitofp i32 257 to float ; yields float:257.0
7529 %Y = uitofp i8 -1 to double ; yields double:255.0
7531 '``sitofp .. to``' Instruction
7532 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7539 <result> = sitofp <ty> <value> to <ty2> ; yields ty2
7544 The '``sitofp``' instruction regards ``value`` as a signed integer and
7545 converts that value to the ``ty2`` type.
7550 The '``sitofp``' instruction takes a value to cast, which must be a
7551 scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to
7552 ``ty2``, which must be an :ref:`floating point <t_floating>` type. If
7553 ``ty`` is a vector integer type, ``ty2`` must be a vector floating point
7554 type with the same number of elements as ``ty``
7559 The '``sitofp``' instruction interprets its operand as a signed integer
7560 quantity and converts it to the corresponding floating point value. If
7561 the value cannot fit in the floating point value, the results are
7567 .. code-block:: llvm
7569 %X = sitofp i32 257 to float ; yields float:257.0
7570 %Y = sitofp i8 -1 to double ; yields double:-1.0
7574 '``ptrtoint .. to``' Instruction
7575 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7582 <result> = ptrtoint <ty> <value> to <ty2> ; yields ty2
7587 The '``ptrtoint``' instruction converts the pointer or a vector of
7588 pointers ``value`` to the integer (or vector of integers) type ``ty2``.
7593 The '``ptrtoint``' instruction takes a ``value`` to cast, which must be
7594 a value of type :ref:`pointer <t_pointer>` or a vector of pointers, and a
7595 type to cast it to ``ty2``, which must be an :ref:`integer <t_integer>` or
7596 a vector of integers type.
7601 The '``ptrtoint``' instruction converts ``value`` to integer type
7602 ``ty2`` by interpreting the pointer value as an integer and either
7603 truncating or zero extending that value to the size of the integer type.
7604 If ``value`` is smaller than ``ty2`` then a zero extension is done. If
7605 ``value`` is larger than ``ty2`` then a truncation is done. If they are
7606 the same size, then nothing is done (*no-op cast*) other than a type
7612 .. code-block:: llvm
7614 %X = ptrtoint i32* %P to i8 ; yields truncation on 32-bit architecture
7615 %Y = ptrtoint i32* %P to i64 ; yields zero extension on 32-bit architecture
7616 %Z = ptrtoint <4 x i32*> %P to <4 x i64>; yields vector zero extension for a vector of addresses on 32-bit architecture
7620 '``inttoptr .. to``' Instruction
7621 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7628 <result> = inttoptr <ty> <value> to <ty2> ; yields ty2
7633 The '``inttoptr``' instruction converts an integer ``value`` to a
7634 pointer type, ``ty2``.
7639 The '``inttoptr``' instruction takes an :ref:`integer <t_integer>` value to
7640 cast, and a type to cast it to, which must be a :ref:`pointer <t_pointer>`
7646 The '``inttoptr``' instruction converts ``value`` to type ``ty2`` by
7647 applying either a zero extension or a truncation depending on the size
7648 of the integer ``value``. If ``value`` is larger than the size of a
7649 pointer then a truncation is done. If ``value`` is smaller than the size
7650 of a pointer then a zero extension is done. If they are the same size,
7651 nothing is done (*no-op cast*).
7656 .. code-block:: llvm
7658 %X = inttoptr i32 255 to i32* ; yields zero extension on 64-bit architecture
7659 %Y = inttoptr i32 255 to i32* ; yields no-op on 32-bit architecture
7660 %Z = inttoptr i64 0 to i32* ; yields truncation on 32-bit architecture
7661 %Z = inttoptr <4 x i32> %G to <4 x i8*>; yields truncation of vector G to four pointers
7665 '``bitcast .. to``' Instruction
7666 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7673 <result> = bitcast <ty> <value> to <ty2> ; yields ty2
7678 The '``bitcast``' instruction converts ``value`` to type ``ty2`` without
7684 The '``bitcast``' instruction takes a value to cast, which must be a
7685 non-aggregate first class value, and a type to cast it to, which must
7686 also be a non-aggregate :ref:`first class <t_firstclass>` type. The
7687 bit sizes of ``value`` and the destination type, ``ty2``, must be
7688 identical. If the source type is a pointer, the destination type must
7689 also be a pointer of the same size. This instruction supports bitwise
7690 conversion of vectors to integers and to vectors of other types (as
7691 long as they have the same size).
7696 The '``bitcast``' instruction converts ``value`` to type ``ty2``. It
7697 is always a *no-op cast* because no bits change with this
7698 conversion. The conversion is done as if the ``value`` had been stored
7699 to memory and read back as type ``ty2``. Pointer (or vector of
7700 pointers) types may only be converted to other pointer (or vector of
7701 pointers) types with the same address space through this instruction.
7702 To convert pointers to other types, use the :ref:`inttoptr <i_inttoptr>`
7703 or :ref:`ptrtoint <i_ptrtoint>` instructions first.
7708 .. code-block:: llvm
7710 %X = bitcast i8 255 to i8 ; yields i8 :-1
7711 %Y = bitcast i32* %x to sint* ; yields sint*:%x
7712 %Z = bitcast <2 x int> %V to i64; ; yields i64: %V
7713 %Z = bitcast <2 x i32*> %V to <2 x i64*> ; yields <2 x i64*>
7715 .. _i_addrspacecast:
7717 '``addrspacecast .. to``' Instruction
7718 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7725 <result> = addrspacecast <pty> <ptrval> to <pty2> ; yields pty2
7730 The '``addrspacecast``' instruction converts ``ptrval`` from ``pty`` in
7731 address space ``n`` to type ``pty2`` in address space ``m``.
7736 The '``addrspacecast``' instruction takes a pointer or vector of pointer value
7737 to cast and a pointer type to cast it to, which must have a different
7743 The '``addrspacecast``' instruction converts the pointer value
7744 ``ptrval`` to type ``pty2``. It can be a *no-op cast* or a complex
7745 value modification, depending on the target and the address space
7746 pair. Pointer conversions within the same address space must be
7747 performed with the ``bitcast`` instruction. Note that if the address space
7748 conversion is legal then both result and operand refer to the same memory
7754 .. code-block:: llvm
7756 %X = addrspacecast i32* %x to i32 addrspace(1)* ; yields i32 addrspace(1)*:%x
7757 %Y = addrspacecast i32 addrspace(1)* %y to i64 addrspace(2)* ; yields i64 addrspace(2)*:%y
7758 %Z = addrspacecast <4 x i32*> %z to <4 x float addrspace(3)*> ; yields <4 x float addrspace(3)*>:%z
7765 The instructions in this category are the "miscellaneous" instructions,
7766 which defy better classification.
7770 '``icmp``' Instruction
7771 ^^^^^^^^^^^^^^^^^^^^^^
7778 <result> = icmp <cond> <ty> <op1>, <op2> ; yields i1 or <N x i1>:result
7783 The '``icmp``' instruction returns a boolean value or a vector of
7784 boolean values based on comparison of its two integer, integer vector,
7785 pointer, or pointer vector operands.
7790 The '``icmp``' instruction takes three operands. The first operand is
7791 the condition code indicating the kind of comparison to perform. It is
7792 not a value, just a keyword. The possible condition code are:
7795 #. ``ne``: not equal
7796 #. ``ugt``: unsigned greater than
7797 #. ``uge``: unsigned greater or equal
7798 #. ``ult``: unsigned less than
7799 #. ``ule``: unsigned less or equal
7800 #. ``sgt``: signed greater than
7801 #. ``sge``: signed greater or equal
7802 #. ``slt``: signed less than
7803 #. ``sle``: signed less or equal
7805 The remaining two arguments must be :ref:`integer <t_integer>` or
7806 :ref:`pointer <t_pointer>` or integer :ref:`vector <t_vector>` typed. They
7807 must also be identical types.
7812 The '``icmp``' compares ``op1`` and ``op2`` according to the condition
7813 code given as ``cond``. The comparison performed always yields either an
7814 :ref:`i1 <t_integer>` or vector of ``i1`` result, as follows:
7816 #. ``eq``: yields ``true`` if the operands are equal, ``false``
7817 otherwise. No sign interpretation is necessary or performed.
7818 #. ``ne``: yields ``true`` if the operands are unequal, ``false``
7819 otherwise. No sign interpretation is necessary or performed.
7820 #. ``ugt``: interprets the operands as unsigned values and yields
7821 ``true`` if ``op1`` is greater than ``op2``.
7822 #. ``uge``: interprets the operands as unsigned values and yields
7823 ``true`` if ``op1`` is greater than or equal to ``op2``.
7824 #. ``ult``: interprets the operands as unsigned values and yields
7825 ``true`` if ``op1`` is less than ``op2``.
7826 #. ``ule``: interprets the operands as unsigned values and yields
7827 ``true`` if ``op1`` is less than or equal to ``op2``.
7828 #. ``sgt``: interprets the operands as signed values and yields ``true``
7829 if ``op1`` is greater than ``op2``.
7830 #. ``sge``: interprets the operands as signed values and yields ``true``
7831 if ``op1`` is greater than or equal to ``op2``.
7832 #. ``slt``: interprets the operands as signed values and yields ``true``
7833 if ``op1`` is less than ``op2``.
7834 #. ``sle``: interprets the operands as signed values and yields ``true``
7835 if ``op1`` is less than or equal to ``op2``.
7837 If the operands are :ref:`pointer <t_pointer>` typed, the pointer values
7838 are compared as if they were integers.
7840 If the operands are integer vectors, then they are compared element by
7841 element. The result is an ``i1`` vector with the same number of elements
7842 as the values being compared. Otherwise, the result is an ``i1``.
7847 .. code-block:: llvm
7849 <result> = icmp eq i32 4, 5 ; yields: result=false
7850 <result> = icmp ne float* %X, %X ; yields: result=false
7851 <result> = icmp ult i16 4, 5 ; yields: result=true
7852 <result> = icmp sgt i16 4, 5 ; yields: result=false
7853 <result> = icmp ule i16 -4, 5 ; yields: result=false
7854 <result> = icmp sge i16 4, 5 ; yields: result=false
7856 Note that the code generator does not yet support vector types with the
7857 ``icmp`` instruction.
7861 '``fcmp``' Instruction
7862 ^^^^^^^^^^^^^^^^^^^^^^
7869 <result> = fcmp <cond> <ty> <op1>, <op2> ; yields i1 or <N x i1>:result
7874 The '``fcmp``' instruction returns a boolean value or vector of boolean
7875 values based on comparison of its operands.
7877 If the operands are floating point scalars, then the result type is a
7878 boolean (:ref:`i1 <t_integer>`).
7880 If the operands are floating point vectors, then the result type is a
7881 vector of boolean with the same number of elements as the operands being
7887 The '``fcmp``' instruction takes three operands. The first operand is
7888 the condition code indicating the kind of comparison to perform. It is
7889 not a value, just a keyword. The possible condition code are:
7891 #. ``false``: no comparison, always returns false
7892 #. ``oeq``: ordered and equal
7893 #. ``ogt``: ordered and greater than
7894 #. ``oge``: ordered and greater than or equal
7895 #. ``olt``: ordered and less than
7896 #. ``ole``: ordered and less than or equal
7897 #. ``one``: ordered and not equal
7898 #. ``ord``: ordered (no nans)
7899 #. ``ueq``: unordered or equal
7900 #. ``ugt``: unordered or greater than
7901 #. ``uge``: unordered or greater than or equal
7902 #. ``ult``: unordered or less than
7903 #. ``ule``: unordered or less than or equal
7904 #. ``une``: unordered or not equal
7905 #. ``uno``: unordered (either nans)
7906 #. ``true``: no comparison, always returns true
7908 *Ordered* means that neither operand is a QNAN while *unordered* means
7909 that either operand may be a QNAN.
7911 Each of ``val1`` and ``val2`` arguments must be either a :ref:`floating
7912 point <t_floating>` type or a :ref:`vector <t_vector>` of floating point
7913 type. They must have identical types.
7918 The '``fcmp``' instruction compares ``op1`` and ``op2`` according to the
7919 condition code given as ``cond``. If the operands are vectors, then the
7920 vectors are compared element by element. Each comparison performed
7921 always yields an :ref:`i1 <t_integer>` result, as follows:
7923 #. ``false``: always yields ``false``, regardless of operands.
7924 #. ``oeq``: yields ``true`` if both operands are not a QNAN and ``op1``
7925 is equal to ``op2``.
7926 #. ``ogt``: yields ``true`` if both operands are not a QNAN and ``op1``
7927 is greater than ``op2``.
7928 #. ``oge``: yields ``true`` if both operands are not a QNAN and ``op1``
7929 is greater than or equal to ``op2``.
7930 #. ``olt``: yields ``true`` if both operands are not a QNAN and ``op1``
7931 is less than ``op2``.
7932 #. ``ole``: yields ``true`` if both operands are not a QNAN and ``op1``
7933 is less than or equal to ``op2``.
7934 #. ``one``: yields ``true`` if both operands are not a QNAN and ``op1``
7935 is not equal to ``op2``.
7936 #. ``ord``: yields ``true`` if both operands are not a QNAN.
7937 #. ``ueq``: yields ``true`` if either operand is a QNAN or ``op1`` is
7939 #. ``ugt``: yields ``true`` if either operand is a QNAN or ``op1`` is
7940 greater than ``op2``.
7941 #. ``uge``: yields ``true`` if either operand is a QNAN or ``op1`` is
7942 greater than or equal to ``op2``.
7943 #. ``ult``: yields ``true`` if either operand is a QNAN or ``op1`` is
7945 #. ``ule``: yields ``true`` if either operand is a QNAN or ``op1`` is
7946 less than or equal to ``op2``.
7947 #. ``une``: yields ``true`` if either operand is a QNAN or ``op1`` is
7948 not equal to ``op2``.
7949 #. ``uno``: yields ``true`` if either operand is a QNAN.
7950 #. ``true``: always yields ``true``, regardless of operands.
7955 .. code-block:: llvm
7957 <result> = fcmp oeq float 4.0, 5.0 ; yields: result=false
7958 <result> = fcmp one float 4.0, 5.0 ; yields: result=true
7959 <result> = fcmp olt float 4.0, 5.0 ; yields: result=true
7960 <result> = fcmp ueq double 1.0, 2.0 ; yields: result=false
7962 Note that the code generator does not yet support vector types with the
7963 ``fcmp`` instruction.
7967 '``phi``' Instruction
7968 ^^^^^^^^^^^^^^^^^^^^^
7975 <result> = phi <ty> [ <val0>, <label0>], ...
7980 The '``phi``' instruction is used to implement the φ node in the SSA
7981 graph representing the function.
7986 The type of the incoming values is specified with the first type field.
7987 After this, the '``phi``' instruction takes a list of pairs as
7988 arguments, with one pair for each predecessor basic block of the current
7989 block. Only values of :ref:`first class <t_firstclass>` type may be used as
7990 the value arguments to the PHI node. Only labels may be used as the
7993 There must be no non-phi instructions between the start of a basic block
7994 and the PHI instructions: i.e. PHI instructions must be first in a basic
7997 For the purposes of the SSA form, the use of each incoming value is
7998 deemed to occur on the edge from the corresponding predecessor block to
7999 the current block (but after any definition of an '``invoke``'
8000 instruction's return value on the same edge).
8005 At runtime, the '``phi``' instruction logically takes on the value
8006 specified by the pair corresponding to the predecessor basic block that
8007 executed just prior to the current block.
8012 .. code-block:: llvm
8014 Loop: ; Infinite loop that counts from 0 on up...
8015 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
8016 %nextindvar = add i32 %indvar, 1
8021 '``select``' Instruction
8022 ^^^^^^^^^^^^^^^^^^^^^^^^
8029 <result> = select selty <cond>, <ty> <val1>, <ty> <val2> ; yields ty
8031 selty is either i1 or {<N x i1>}
8036 The '``select``' instruction is used to choose one value based on a
8037 condition, without IR-level branching.
8042 The '``select``' instruction requires an 'i1' value or a vector of 'i1'
8043 values indicating the condition, and two values of the same :ref:`first
8044 class <t_firstclass>` type.
8049 If the condition is an i1 and it evaluates to 1, the instruction returns
8050 the first value argument; otherwise, it returns the second value
8053 If the condition is a vector of i1, then the value arguments must be
8054 vectors of the same size, and the selection is done element by element.
8056 If the condition is an i1 and the value arguments are vectors of the
8057 same size, then an entire vector is selected.
8062 .. code-block:: llvm
8064 %X = select i1 true, i8 17, i8 42 ; yields i8:17
8068 '``call``' Instruction
8069 ^^^^^^^^^^^^^^^^^^^^^^
8076 <result> = [tail | musttail] call [cconv] [ret attrs] <ty> [<fnty>*] <fnptrval>(<function args>) [fn attrs]
8081 The '``call``' instruction represents a simple function call.
8086 This instruction requires several arguments:
8088 #. The optional ``tail`` and ``musttail`` markers indicate that the optimizers
8089 should perform tail call optimization. The ``tail`` marker is a hint that
8090 `can be ignored <CodeGenerator.html#sibcallopt>`_. The ``musttail`` marker
8091 means that the call must be tail call optimized in order for the program to
8092 be correct. The ``musttail`` marker provides these guarantees:
8094 #. The call will not cause unbounded stack growth if it is part of a
8095 recursive cycle in the call graph.
8096 #. Arguments with the :ref:`inalloca <attr_inalloca>` attribute are
8099 Both markers imply that the callee does not access allocas or varargs from
8100 the caller. Calls marked ``musttail`` must obey the following additional
8103 - The call must immediately precede a :ref:`ret <i_ret>` instruction,
8104 or a pointer bitcast followed by a ret instruction.
8105 - The ret instruction must return the (possibly bitcasted) value
8106 produced by the call or void.
8107 - The caller and callee prototypes must match. Pointer types of
8108 parameters or return types may differ in pointee type, but not
8110 - The calling conventions of the caller and callee must match.
8111 - All ABI-impacting function attributes, such as sret, byval, inreg,
8112 returned, and inalloca, must match.
8113 - The callee must be varargs iff the caller is varargs. Bitcasting a
8114 non-varargs function to the appropriate varargs type is legal so
8115 long as the non-varargs prefixes obey the other rules.
8117 Tail call optimization for calls marked ``tail`` is guaranteed to occur if
8118 the following conditions are met:
8120 - Caller and callee both have the calling convention ``fastcc``.
8121 - The call is in tail position (ret immediately follows call and ret
8122 uses value of call or is void).
8123 - Option ``-tailcallopt`` is enabled, or
8124 ``llvm::GuaranteedTailCallOpt`` is ``true``.
8125 - `Platform-specific constraints are
8126 met. <CodeGenerator.html#tailcallopt>`_
8128 #. The optional "cconv" marker indicates which :ref:`calling
8129 convention <callingconv>` the call should use. If none is
8130 specified, the call defaults to using C calling conventions. The
8131 calling convention of the call must match the calling convention of
8132 the target function, or else the behavior is undefined.
8133 #. The optional :ref:`Parameter Attributes <paramattrs>` list for return
8134 values. Only '``zeroext``', '``signext``', and '``inreg``' attributes
8136 #. '``ty``': the type of the call instruction itself which is also the
8137 type of the return value. Functions that return no value are marked
8139 #. '``fnty``': shall be the signature of the pointer to function value
8140 being invoked. The argument types must match the types implied by
8141 this signature. This type can be omitted if the function is not
8142 varargs and if the function type does not return a pointer to a
8144 #. '``fnptrval``': An LLVM value containing a pointer to a function to
8145 be invoked. In most cases, this is a direct function invocation, but
8146 indirect ``call``'s are just as possible, calling an arbitrary pointer
8148 #. '``function args``': argument list whose types match the function
8149 signature argument types and parameter attributes. All arguments must
8150 be of :ref:`first class <t_firstclass>` type. If the function signature
8151 indicates the function accepts a variable number of arguments, the
8152 extra arguments can be specified.
8153 #. The optional :ref:`function attributes <fnattrs>` list. Only
8154 '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``'
8155 attributes are valid here.
8160 The '``call``' instruction is used to cause control flow to transfer to
8161 a specified function, with its incoming arguments bound to the specified
8162 values. Upon a '``ret``' instruction in the called function, control
8163 flow continues with the instruction after the function call, and the
8164 return value of the function is bound to the result argument.
8169 .. code-block:: llvm
8171 %retval = call i32 @test(i32 %argc)
8172 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) ; yields i32
8173 %X = tail call i32 @foo() ; yields i32
8174 %Y = tail call fastcc i32 @foo() ; yields i32
8175 call void %foo(i8 97 signext)
8177 %struct.A = type { i32, i8 }
8178 %r = call %struct.A @foo() ; yields { i32, i8 }
8179 %gr = extractvalue %struct.A %r, 0 ; yields i32
8180 %gr1 = extractvalue %struct.A %r, 1 ; yields i8
8181 %Z = call void @foo() noreturn ; indicates that %foo never returns normally
8182 %ZZ = call zeroext i32 @bar() ; Return value is %zero extended
8184 llvm treats calls to some functions with names and arguments that match
8185 the standard C99 library as being the C99 library functions, and may
8186 perform optimizations or generate code for them under that assumption.
8187 This is something we'd like to change in the future to provide better
8188 support for freestanding environments and non-C-based languages.
8192 '``va_arg``' Instruction
8193 ^^^^^^^^^^^^^^^^^^^^^^^^
8200 <resultval> = va_arg <va_list*> <arglist>, <argty>
8205 The '``va_arg``' instruction is used to access arguments passed through
8206 the "variable argument" area of a function call. It is used to implement
8207 the ``va_arg`` macro in C.
8212 This instruction takes a ``va_list*`` value and the type of the
8213 argument. It returns a value of the specified argument type and
8214 increments the ``va_list`` to point to the next argument. The actual
8215 type of ``va_list`` is target specific.
8220 The '``va_arg``' instruction loads an argument of the specified type
8221 from the specified ``va_list`` and causes the ``va_list`` to point to
8222 the next argument. For more information, see the variable argument
8223 handling :ref:`Intrinsic Functions <int_varargs>`.
8225 It is legal for this instruction to be called in a function which does
8226 not take a variable number of arguments, for example, the ``vfprintf``
8229 ``va_arg`` is an LLVM instruction instead of an :ref:`intrinsic
8230 function <intrinsics>` because it takes a type as an argument.
8235 See the :ref:`variable argument processing <int_varargs>` section.
8237 Note that the code generator does not yet fully support va\_arg on many
8238 targets. Also, it does not currently support va\_arg with aggregate
8239 types on any target.
8243 '``landingpad``' Instruction
8244 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8251 <resultval> = landingpad <resultty> <clause>+
8252 <resultval> = landingpad <resultty> cleanup <clause>*
8254 <clause> := catch <type> <value>
8255 <clause> := filter <array constant type> <array constant>
8260 The '``landingpad``' instruction is used by `LLVM's exception handling
8261 system <ExceptionHandling.html#overview>`_ to specify that a basic block
8262 is a landing pad --- one where the exception lands, and corresponds to the
8263 code found in the ``catch`` portion of a ``try``/``catch`` sequence. It
8264 defines values supplied by the :ref:`personality function <personalityfn>` upon
8265 re-entry to the function. The ``resultval`` has the type ``resultty``.
8271 ``cleanup`` flag indicates that the landing pad block is a cleanup.
8273 A ``clause`` begins with the clause type --- ``catch`` or ``filter`` --- and
8274 contains the global variable representing the "type" that may be caught
8275 or filtered respectively. Unlike the ``catch`` clause, the ``filter``
8276 clause takes an array constant as its argument. Use
8277 "``[0 x i8**] undef``" for a filter which cannot throw. The
8278 '``landingpad``' instruction must contain *at least* one ``clause`` or
8279 the ``cleanup`` flag.
8284 The '``landingpad``' instruction defines the values which are set by the
8285 :ref:`personality function <personalityfn>` upon re-entry to the function, and
8286 therefore the "result type" of the ``landingpad`` instruction. As with
8287 calling conventions, how the personality function results are
8288 represented in LLVM IR is target specific.
8290 The clauses are applied in order from top to bottom. If two
8291 ``landingpad`` instructions are merged together through inlining, the
8292 clauses from the calling function are appended to the list of clauses.
8293 When the call stack is being unwound due to an exception being thrown,
8294 the exception is compared against each ``clause`` in turn. If it doesn't
8295 match any of the clauses, and the ``cleanup`` flag is not set, then
8296 unwinding continues further up the call stack.
8298 The ``landingpad`` instruction has several restrictions:
8300 - A landing pad block is a basic block which is the unwind destination
8301 of an '``invoke``' instruction.
8302 - A landing pad block must have a '``landingpad``' instruction as its
8303 first non-PHI instruction.
8304 - There can be only one '``landingpad``' instruction within the landing
8306 - A basic block that is not a landing pad block may not include a
8307 '``landingpad``' instruction.
8312 .. code-block:: llvm
8314 ;; A landing pad which can catch an integer.
8315 %res = landingpad { i8*, i32 }
8317 ;; A landing pad that is a cleanup.
8318 %res = landingpad { i8*, i32 }
8320 ;; A landing pad which can catch an integer and can only throw a double.
8321 %res = landingpad { i8*, i32 }
8323 filter [1 x i8**] [@_ZTId]
8327 '``cleanupblock``' Instruction
8328 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8335 <resultval> = cleanupblock <resultty> [<args>*]
8340 The '``cleanupblock``' instruction is used by `LLVM's exception handling
8341 system <ExceptionHandling.html#overview>`_ to specify that a basic block
8342 is a cleanup block --- one where a personality routine attempts to
8343 transfer control to run cleanup actions.
8344 The ``args`` correspond to whatever additional
8345 information the :ref:`personality function <personalityfn>` requires to
8346 execute the cleanup.
8347 The ``resultval`` has the type ``resultty``.
8352 The instruction takes a list of arbitrary values which are interpreted
8353 by the :ref:`personality function <personalityfn>`.
8358 The '``cleanupblock``' instruction defines the values which are set by the
8359 :ref:`personality function <personalityfn>` upon re-entry to the function, and
8360 therefore the "result type" of the ``cleanupblock`` instruction. As with
8361 calling conventions, how the personality function results are
8362 represented in LLVM IR is target specific.
8364 When the call stack is being unwound due to an exception being thrown,
8365 the :ref:`personality function <personalityfn>` transfers control to the
8366 ``cleanupblock`` with the aid of the personality-specific arguments.
8368 The ``cleanupblock`` instruction has several restrictions:
8370 - A cleanup block is a basic block which is the unwind destination of
8371 an exceptional instruction.
8372 - A cleanup block must have a '``cleanupblock``' instruction as its
8373 first non-PHI instruction.
8374 - There can be only one '``cleanupblock``' instruction within the
8376 - A basic block that is not a cleanup block may not include a
8377 '``cleanupblock``' instruction.
8382 .. code-block:: llvm
8384 %res = cleanupblock { i8*, i32 } [label %nextaction]
8391 LLVM supports the notion of an "intrinsic function". These functions
8392 have well known names and semantics and are required to follow certain
8393 restrictions. Overall, these intrinsics represent an extension mechanism
8394 for the LLVM language that does not require changing all of the
8395 transformations in LLVM when adding to the language (or the bitcode
8396 reader/writer, the parser, etc...).
8398 Intrinsic function names must all start with an "``llvm.``" prefix. This
8399 prefix is reserved in LLVM for intrinsic names; thus, function names may
8400 not begin with this prefix. Intrinsic functions must always be external
8401 functions: you cannot define the body of intrinsic functions. Intrinsic
8402 functions may only be used in call or invoke instructions: it is illegal
8403 to take the address of an intrinsic function. Additionally, because
8404 intrinsic functions are part of the LLVM language, it is required if any
8405 are added that they be documented here.
8407 Some intrinsic functions can be overloaded, i.e., the intrinsic
8408 represents a family of functions that perform the same operation but on
8409 different data types. Because LLVM can represent over 8 million
8410 different integer types, overloading is used commonly to allow an
8411 intrinsic function to operate on any integer type. One or more of the
8412 argument types or the result type can be overloaded to accept any
8413 integer type. Argument types may also be defined as exactly matching a
8414 previous argument's type or the result type. This allows an intrinsic
8415 function which accepts multiple arguments, but needs all of them to be
8416 of the same type, to only be overloaded with respect to a single
8417 argument or the result.
8419 Overloaded intrinsics will have the names of its overloaded argument
8420 types encoded into its function name, each preceded by a period. Only
8421 those types which are overloaded result in a name suffix. Arguments
8422 whose type is matched against another type do not. For example, the
8423 ``llvm.ctpop`` function can take an integer of any width and returns an
8424 integer of exactly the same integer width. This leads to a family of
8425 functions such as ``i8 @llvm.ctpop.i8(i8 %val)`` and
8426 ``i29 @llvm.ctpop.i29(i29 %val)``. Only one type, the return type, is
8427 overloaded, and only one type suffix is required. Because the argument's
8428 type is matched against the return type, it does not require its own
8431 To learn how to add an intrinsic function, please see the `Extending
8432 LLVM Guide <ExtendingLLVM.html>`_.
8436 Variable Argument Handling Intrinsics
8437 -------------------------------------
8439 Variable argument support is defined in LLVM with the
8440 :ref:`va_arg <i_va_arg>` instruction and these three intrinsic
8441 functions. These functions are related to the similarly named macros
8442 defined in the ``<stdarg.h>`` header file.
8444 All of these functions operate on arguments that use a target-specific
8445 value type "``va_list``". The LLVM assembly language reference manual
8446 does not define what this type is, so all transformations should be
8447 prepared to handle these functions regardless of the type used.
8449 This example shows how the :ref:`va_arg <i_va_arg>` instruction and the
8450 variable argument handling intrinsic functions are used.
8452 .. code-block:: llvm
8454 ; This struct is different for every platform. For most platforms,
8455 ; it is merely an i8*.
8456 %struct.va_list = type { i8* }
8458 ; For Unix x86_64 platforms, va_list is the following struct:
8459 ; %struct.va_list = type { i32, i32, i8*, i8* }
8461 define i32 @test(i32 %X, ...) {
8462 ; Initialize variable argument processing
8463 %ap = alloca %struct.va_list
8464 %ap2 = bitcast %struct.va_list* %ap to i8*
8465 call void @llvm.va_start(i8* %ap2)
8467 ; Read a single integer argument
8468 %tmp = va_arg i8* %ap2, i32
8470 ; Demonstrate usage of llvm.va_copy and llvm.va_end
8472 %aq2 = bitcast i8** %aq to i8*
8473 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
8474 call void @llvm.va_end(i8* %aq2)
8476 ; Stop processing of arguments.
8477 call void @llvm.va_end(i8* %ap2)
8481 declare void @llvm.va_start(i8*)
8482 declare void @llvm.va_copy(i8*, i8*)
8483 declare void @llvm.va_end(i8*)
8487 '``llvm.va_start``' Intrinsic
8488 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8495 declare void @llvm.va_start(i8* <arglist>)
8500 The '``llvm.va_start``' intrinsic initializes ``*<arglist>`` for
8501 subsequent use by ``va_arg``.
8506 The argument is a pointer to a ``va_list`` element to initialize.
8511 The '``llvm.va_start``' intrinsic works just like the ``va_start`` macro
8512 available in C. In a target-dependent way, it initializes the
8513 ``va_list`` element to which the argument points, so that the next call
8514 to ``va_arg`` will produce the first variable argument passed to the
8515 function. Unlike the C ``va_start`` macro, this intrinsic does not need
8516 to know the last argument of the function as the compiler can figure
8519 '``llvm.va_end``' Intrinsic
8520 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8527 declare void @llvm.va_end(i8* <arglist>)
8532 The '``llvm.va_end``' intrinsic destroys ``*<arglist>``, which has been
8533 initialized previously with ``llvm.va_start`` or ``llvm.va_copy``.
8538 The argument is a pointer to a ``va_list`` to destroy.
8543 The '``llvm.va_end``' intrinsic works just like the ``va_end`` macro
8544 available in C. In a target-dependent way, it destroys the ``va_list``
8545 element to which the argument points. Calls to
8546 :ref:`llvm.va_start <int_va_start>` and
8547 :ref:`llvm.va_copy <int_va_copy>` must be matched exactly with calls to
8552 '``llvm.va_copy``' Intrinsic
8553 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8560 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
8565 The '``llvm.va_copy``' intrinsic copies the current argument position
8566 from the source argument list to the destination argument list.
8571 The first argument is a pointer to a ``va_list`` element to initialize.
8572 The second argument is a pointer to a ``va_list`` element to copy from.
8577 The '``llvm.va_copy``' intrinsic works just like the ``va_copy`` macro
8578 available in C. In a target-dependent way, it copies the source
8579 ``va_list`` element into the destination ``va_list`` element. This
8580 intrinsic is necessary because the `` llvm.va_start`` intrinsic may be
8581 arbitrarily complex and require, for example, memory allocation.
8583 Accurate Garbage Collection Intrinsics
8584 --------------------------------------
8586 LLVM's support for `Accurate Garbage Collection <GarbageCollection.html>`_
8587 (GC) requires the frontend to generate code containing appropriate intrinsic
8588 calls and select an appropriate GC strategy which knows how to lower these
8589 intrinsics in a manner which is appropriate for the target collector.
8591 These intrinsics allow identification of :ref:`GC roots on the
8592 stack <int_gcroot>`, as well as garbage collector implementations that
8593 require :ref:`read <int_gcread>` and :ref:`write <int_gcwrite>` barriers.
8594 Frontends for type-safe garbage collected languages should generate
8595 these intrinsics to make use of the LLVM garbage collectors. For more
8596 details, see `Garbage Collection with LLVM <GarbageCollection.html>`_.
8598 Experimental Statepoint Intrinsics
8599 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8601 LLVM provides an second experimental set of intrinsics for describing garbage
8602 collection safepoints in compiled code. These intrinsics are an alternative
8603 to the ``llvm.gcroot`` intrinsics, but are compatible with the ones for
8604 :ref:`read <int_gcread>` and :ref:`write <int_gcwrite>` barriers. The
8605 differences in approach are covered in the `Garbage Collection with LLVM
8606 <GarbageCollection.html>`_ documentation. The intrinsics themselves are
8607 described in :doc:`Statepoints`.
8611 '``llvm.gcroot``' Intrinsic
8612 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8619 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
8624 The '``llvm.gcroot``' intrinsic declares the existence of a GC root to
8625 the code generator, and allows some metadata to be associated with it.
8630 The first argument specifies the address of a stack object that contains
8631 the root pointer. The second pointer (which must be either a constant or
8632 a global value address) contains the meta-data to be associated with the
8638 At runtime, a call to this intrinsic stores a null pointer into the
8639 "ptrloc" location. At compile-time, the code generator generates
8640 information to allow the runtime to find the pointer at GC safe points.
8641 The '``llvm.gcroot``' intrinsic may only be used in a function which
8642 :ref:`specifies a GC algorithm <gc>`.
8646 '``llvm.gcread``' Intrinsic
8647 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
8654 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
8659 The '``llvm.gcread``' intrinsic identifies reads of references from heap
8660 locations, allowing garbage collector implementations that require read
8666 The second argument is the address to read from, which should be an
8667 address allocated from the garbage collector. The first object is a
8668 pointer to the start of the referenced object, if needed by the language
8669 runtime (otherwise null).
8674 The '``llvm.gcread``' intrinsic has the same semantics as a load
8675 instruction, but may be replaced with substantially more complex code by
8676 the garbage collector runtime, as needed. The '``llvm.gcread``'
8677 intrinsic may only be used in a function which :ref:`specifies a GC
8682 '``llvm.gcwrite``' Intrinsic
8683 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8690 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
8695 The '``llvm.gcwrite``' intrinsic identifies writes of references to heap
8696 locations, allowing garbage collector implementations that require write
8697 barriers (such as generational or reference counting collectors).
8702 The first argument is the reference to store, the second is the start of
8703 the object to store it to, and the third is the address of the field of
8704 Obj to store to. If the runtime does not require a pointer to the
8705 object, Obj may be null.
8710 The '``llvm.gcwrite``' intrinsic has the same semantics as a store
8711 instruction, but may be replaced with substantially more complex code by
8712 the garbage collector runtime, as needed. The '``llvm.gcwrite``'
8713 intrinsic may only be used in a function which :ref:`specifies a GC
8716 Code Generator Intrinsics
8717 -------------------------
8719 These intrinsics are provided by LLVM to expose special features that
8720 may only be implemented with code generator support.
8722 '``llvm.returnaddress``' Intrinsic
8723 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8730 declare i8 *@llvm.returnaddress(i32 <level>)
8735 The '``llvm.returnaddress``' intrinsic attempts to compute a
8736 target-specific value indicating the return address of the current
8737 function or one of its callers.
8742 The argument to this intrinsic indicates which function to return the
8743 address for. Zero indicates the calling function, one indicates its
8744 caller, etc. The argument is **required** to be a constant integer
8750 The '``llvm.returnaddress``' intrinsic either returns a pointer
8751 indicating the return address of the specified call frame, or zero if it
8752 cannot be identified. The value returned by this intrinsic is likely to
8753 be incorrect or 0 for arguments other than zero, so it should only be
8754 used for debugging purposes.
8756 Note that calling this intrinsic does not prevent function inlining or
8757 other aggressive transformations, so the value returned may not be that
8758 of the obvious source-language caller.
8760 '``llvm.frameaddress``' Intrinsic
8761 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8768 declare i8* @llvm.frameaddress(i32 <level>)
8773 The '``llvm.frameaddress``' intrinsic attempts to return the
8774 target-specific frame pointer value for the specified stack frame.
8779 The argument to this intrinsic indicates which function to return the
8780 frame pointer for. Zero indicates the calling function, one indicates
8781 its caller, etc. The argument is **required** to be a constant integer
8787 The '``llvm.frameaddress``' intrinsic either returns a pointer
8788 indicating the frame address of the specified call frame, or zero if it
8789 cannot be identified. The value returned by this intrinsic is likely to
8790 be incorrect or 0 for arguments other than zero, so it should only be
8791 used for debugging purposes.
8793 Note that calling this intrinsic does not prevent function inlining or
8794 other aggressive transformations, so the value returned may not be that
8795 of the obvious source-language caller.
8797 '``llvm.localescape``' and '``llvm.localrecover``' Intrinsics
8798 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8805 declare void @llvm.localescape(...)
8806 declare i8* @llvm.localrecover(i8* %func, i8* %fp, i32 %idx)
8811 The '``llvm.localescape``' intrinsic escapes offsets of a collection of static
8812 allocas, and the '``llvm.localrecover``' intrinsic applies those offsets to a
8813 live frame pointer to recover the address of the allocation. The offset is
8814 computed during frame layout of the caller of ``llvm.localescape``.
8819 All arguments to '``llvm.localescape``' must be pointers to static allocas or
8820 casts of static allocas. Each function can only call '``llvm.localescape``'
8821 once, and it can only do so from the entry block.
8823 The ``func`` argument to '``llvm.localrecover``' must be a constant
8824 bitcasted pointer to a function defined in the current module. The code
8825 generator cannot determine the frame allocation offset of functions defined in
8828 The ``fp`` argument to '``llvm.localrecover``' must be a frame pointer of a
8829 call frame that is currently live. The return value of '``llvm.localaddress``'
8830 is one way to produce such a value, but various runtimes also expose a suitable
8831 pointer in platform-specific ways.
8833 The ``idx`` argument to '``llvm.localrecover``' indicates which alloca passed to
8834 '``llvm.localescape``' to recover. It is zero-indexed.
8839 These intrinsics allow a group of functions to share access to a set of local
8840 stack allocations of a one parent function. The parent function may call the
8841 '``llvm.localescape``' intrinsic once from the function entry block, and the
8842 child functions can use '``llvm.localrecover``' to access the escaped allocas.
8843 The '``llvm.localescape``' intrinsic blocks inlining, as inlining changes where
8844 the escaped allocas are allocated, which would break attempts to use
8845 '``llvm.localrecover``'.
8847 .. _int_read_register:
8848 .. _int_write_register:
8850 '``llvm.read_register``' and '``llvm.write_register``' Intrinsics
8851 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8858 declare i32 @llvm.read_register.i32(metadata)
8859 declare i64 @llvm.read_register.i64(metadata)
8860 declare void @llvm.write_register.i32(metadata, i32 @value)
8861 declare void @llvm.write_register.i64(metadata, i64 @value)
8867 The '``llvm.read_register``' and '``llvm.write_register``' intrinsics
8868 provides access to the named register. The register must be valid on
8869 the architecture being compiled to. The type needs to be compatible
8870 with the register being read.
8875 The '``llvm.read_register``' intrinsic returns the current value of the
8876 register, where possible. The '``llvm.write_register``' intrinsic sets
8877 the current value of the register, where possible.
8879 This is useful to implement named register global variables that need
8880 to always be mapped to a specific register, as is common practice on
8881 bare-metal programs including OS kernels.
8883 The compiler doesn't check for register availability or use of the used
8884 register in surrounding code, including inline assembly. Because of that,
8885 allocatable registers are not supported.
8887 Warning: So far it only works with the stack pointer on selected
8888 architectures (ARM, AArch64, PowerPC and x86_64). Significant amount of
8889 work is needed to support other registers and even more so, allocatable
8894 '``llvm.stacksave``' Intrinsic
8895 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8902 declare i8* @llvm.stacksave()
8907 The '``llvm.stacksave``' intrinsic is used to remember the current state
8908 of the function stack, for use with
8909 :ref:`llvm.stackrestore <int_stackrestore>`. This is useful for
8910 implementing language features like scoped automatic variable sized
8916 This intrinsic returns a opaque pointer value that can be passed to
8917 :ref:`llvm.stackrestore <int_stackrestore>`. When an
8918 ``llvm.stackrestore`` intrinsic is executed with a value saved from
8919 ``llvm.stacksave``, it effectively restores the state of the stack to
8920 the state it was in when the ``llvm.stacksave`` intrinsic executed. In
8921 practice, this pops any :ref:`alloca <i_alloca>` blocks from the stack that
8922 were allocated after the ``llvm.stacksave`` was executed.
8924 .. _int_stackrestore:
8926 '``llvm.stackrestore``' Intrinsic
8927 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8934 declare void @llvm.stackrestore(i8* %ptr)
8939 The '``llvm.stackrestore``' intrinsic is used to restore the state of
8940 the function stack to the state it was in when the corresponding
8941 :ref:`llvm.stacksave <int_stacksave>` intrinsic executed. This is
8942 useful for implementing language features like scoped automatic variable
8943 sized arrays in C99.
8948 See the description for :ref:`llvm.stacksave <int_stacksave>`.
8950 '``llvm.prefetch``' Intrinsic
8951 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8958 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
8963 The '``llvm.prefetch``' intrinsic is a hint to the code generator to
8964 insert a prefetch instruction if supported; otherwise, it is a noop.
8965 Prefetches have no effect on the behavior of the program but can change
8966 its performance characteristics.
8971 ``address`` is the address to be prefetched, ``rw`` is the specifier
8972 determining if the fetch should be for a read (0) or write (1), and
8973 ``locality`` is a temporal locality specifier ranging from (0) - no
8974 locality, to (3) - extremely local keep in cache. The ``cache type``
8975 specifies whether the prefetch is performed on the data (1) or
8976 instruction (0) cache. The ``rw``, ``locality`` and ``cache type``
8977 arguments must be constant integers.
8982 This intrinsic does not modify the behavior of the program. In
8983 particular, prefetches cannot trap and do not produce a value. On
8984 targets that support this intrinsic, the prefetch can provide hints to
8985 the processor cache for better performance.
8987 '``llvm.pcmarker``' Intrinsic
8988 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8995 declare void @llvm.pcmarker(i32 <id>)
9000 The '``llvm.pcmarker``' intrinsic is a method to export a Program
9001 Counter (PC) in a region of code to simulators and other tools. The
9002 method is target specific, but it is expected that the marker will use
9003 exported symbols to transmit the PC of the marker. The marker makes no
9004 guarantees that it will remain with any specific instruction after
9005 optimizations. It is possible that the presence of a marker will inhibit
9006 optimizations. The intended use is to be inserted after optimizations to
9007 allow correlations of simulation runs.
9012 ``id`` is a numerical id identifying the marker.
9017 This intrinsic does not modify the behavior of the program. Backends
9018 that do not support this intrinsic may ignore it.
9020 '``llvm.readcyclecounter``' Intrinsic
9021 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9028 declare i64 @llvm.readcyclecounter()
9033 The '``llvm.readcyclecounter``' intrinsic provides access to the cycle
9034 counter register (or similar low latency, high accuracy clocks) on those
9035 targets that support it. On X86, it should map to RDTSC. On Alpha, it
9036 should map to RPCC. As the backing counters overflow quickly (on the
9037 order of 9 seconds on alpha), this should only be used for small
9043 When directly supported, reading the cycle counter should not modify any
9044 memory. Implementations are allowed to either return a application
9045 specific value or a system wide value. On backends without support, this
9046 is lowered to a constant 0.
9048 Note that runtime support may be conditional on the privilege-level code is
9049 running at and the host platform.
9051 '``llvm.clear_cache``' Intrinsic
9052 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9059 declare void @llvm.clear_cache(i8*, i8*)
9064 The '``llvm.clear_cache``' intrinsic ensures visibility of modifications
9065 in the specified range to the execution unit of the processor. On
9066 targets with non-unified instruction and data cache, the implementation
9067 flushes the instruction cache.
9072 On platforms with coherent instruction and data caches (e.g. x86), this
9073 intrinsic is a nop. On platforms with non-coherent instruction and data
9074 cache (e.g. ARM, MIPS), the intrinsic is lowered either to appropriate
9075 instructions or a system call, if cache flushing requires special
9078 The default behavior is to emit a call to ``__clear_cache`` from the run
9081 This instrinsic does *not* empty the instruction pipeline. Modifications
9082 of the current function are outside the scope of the intrinsic.
9084 '``llvm.instrprof_increment``' Intrinsic
9085 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9092 declare void @llvm.instrprof_increment(i8* <name>, i64 <hash>,
9093 i32 <num-counters>, i32 <index>)
9098 The '``llvm.instrprof_increment``' intrinsic can be emitted by a
9099 frontend for use with instrumentation based profiling. These will be
9100 lowered by the ``-instrprof`` pass to generate execution counts of a
9106 The first argument is a pointer to a global variable containing the
9107 name of the entity being instrumented. This should generally be the
9108 (mangled) function name for a set of counters.
9110 The second argument is a hash value that can be used by the consumer
9111 of the profile data to detect changes to the instrumented source, and
9112 the third is the number of counters associated with ``name``. It is an
9113 error if ``hash`` or ``num-counters`` differ between two instances of
9114 ``instrprof_increment`` that refer to the same name.
9116 The last argument refers to which of the counters for ``name`` should
9117 be incremented. It should be a value between 0 and ``num-counters``.
9122 This intrinsic represents an increment of a profiling counter. It will
9123 cause the ``-instrprof`` pass to generate the appropriate data
9124 structures and the code to increment the appropriate value, in a
9125 format that can be written out by a compiler runtime and consumed via
9126 the ``llvm-profdata`` tool.
9128 Standard C Library Intrinsics
9129 -----------------------------
9131 LLVM provides intrinsics for a few important standard C library
9132 functions. These intrinsics allow source-language front-ends to pass
9133 information about the alignment of the pointer arguments to the code
9134 generator, providing opportunity for more efficient code generation.
9138 '``llvm.memcpy``' Intrinsic
9139 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9144 This is an overloaded intrinsic. You can use ``llvm.memcpy`` on any
9145 integer bit width and for different address spaces. Not all targets
9146 support all bit widths however.
9150 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
9151 i32 <len>, i32 <align>, i1 <isvolatile>)
9152 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
9153 i64 <len>, i32 <align>, i1 <isvolatile>)
9158 The '``llvm.memcpy.*``' intrinsics copy a block of memory from the
9159 source location to the destination location.
9161 Note that, unlike the standard libc function, the ``llvm.memcpy.*``
9162 intrinsics do not return a value, takes extra alignment/isvolatile
9163 arguments and the pointers can be in specified address spaces.
9168 The first argument is a pointer to the destination, the second is a
9169 pointer to the source. The third argument is an integer argument
9170 specifying the number of bytes to copy, the fourth argument is the
9171 alignment of the source and destination locations, and the fifth is a
9172 boolean indicating a volatile access.
9174 If the call to this intrinsic has an alignment value that is not 0 or 1,
9175 then the caller guarantees that both the source and destination pointers
9176 are aligned to that boundary.
9178 If the ``isvolatile`` parameter is ``true``, the ``llvm.memcpy`` call is
9179 a :ref:`volatile operation <volatile>`. The detailed access behavior is not
9180 very cleanly specified and it is unwise to depend on it.
9185 The '``llvm.memcpy.*``' intrinsics copy a block of memory from the
9186 source location to the destination location, which are not allowed to
9187 overlap. It copies "len" bytes of memory over. If the argument is known
9188 to be aligned to some boundary, this can be specified as the fourth
9189 argument, otherwise it should be set to 0 or 1 (both meaning no alignment).
9191 '``llvm.memmove``' Intrinsic
9192 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9197 This is an overloaded intrinsic. You can use llvm.memmove on any integer
9198 bit width and for different address space. Not all targets support all
9203 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
9204 i32 <len>, i32 <align>, i1 <isvolatile>)
9205 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
9206 i64 <len>, i32 <align>, i1 <isvolatile>)
9211 The '``llvm.memmove.*``' intrinsics move a block of memory from the
9212 source location to the destination location. It is similar to the
9213 '``llvm.memcpy``' intrinsic but allows the two memory locations to
9216 Note that, unlike the standard libc function, the ``llvm.memmove.*``
9217 intrinsics do not return a value, takes extra alignment/isvolatile
9218 arguments and the pointers can be in specified address spaces.
9223 The first argument is a pointer to the destination, the second is a
9224 pointer to the source. The third argument is an integer argument
9225 specifying the number of bytes to copy, the fourth argument is the
9226 alignment of the source and destination locations, and the fifth is a
9227 boolean indicating a volatile access.
9229 If the call to this intrinsic has an alignment value that is not 0 or 1,
9230 then the caller guarantees that the source and destination pointers are
9231 aligned to that boundary.
9233 If the ``isvolatile`` parameter is ``true``, the ``llvm.memmove`` call
9234 is a :ref:`volatile operation <volatile>`. The detailed access behavior is
9235 not very cleanly specified and it is unwise to depend on it.
9240 The '``llvm.memmove.*``' intrinsics copy a block of memory from the
9241 source location to the destination location, which may overlap. It
9242 copies "len" bytes of memory over. If the argument is known to be
9243 aligned to some boundary, this can be specified as the fourth argument,
9244 otherwise it should be set to 0 or 1 (both meaning no alignment).
9246 '``llvm.memset.*``' Intrinsics
9247 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9252 This is an overloaded intrinsic. You can use llvm.memset on any integer
9253 bit width and for different address spaces. However, not all targets
9254 support all bit widths.
9258 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
9259 i32 <len>, i32 <align>, i1 <isvolatile>)
9260 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
9261 i64 <len>, i32 <align>, i1 <isvolatile>)
9266 The '``llvm.memset.*``' intrinsics fill a block of memory with a
9267 particular byte value.
9269 Note that, unlike the standard libc function, the ``llvm.memset``
9270 intrinsic does not return a value and takes extra alignment/volatile
9271 arguments. Also, the destination can be in an arbitrary address space.
9276 The first argument is a pointer to the destination to fill, the second
9277 is the byte value with which to fill it, the third argument is an
9278 integer argument specifying the number of bytes to fill, and the fourth
9279 argument is the known alignment of the destination location.
9281 If the call to this intrinsic has an alignment value that is not 0 or 1,
9282 then the caller guarantees that the destination pointer is aligned to
9285 If the ``isvolatile`` parameter is ``true``, the ``llvm.memset`` call is
9286 a :ref:`volatile operation <volatile>`. The detailed access behavior is not
9287 very cleanly specified and it is unwise to depend on it.
9292 The '``llvm.memset.*``' intrinsics fill "len" bytes of memory starting
9293 at the destination location. If the argument is known to be aligned to
9294 some boundary, this can be specified as the fourth argument, otherwise
9295 it should be set to 0 or 1 (both meaning no alignment).
9297 '``llvm.sqrt.*``' Intrinsic
9298 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9303 This is an overloaded intrinsic. You can use ``llvm.sqrt`` on any
9304 floating point or vector of floating point type. Not all targets support
9309 declare float @llvm.sqrt.f32(float %Val)
9310 declare double @llvm.sqrt.f64(double %Val)
9311 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
9312 declare fp128 @llvm.sqrt.f128(fp128 %Val)
9313 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
9318 The '``llvm.sqrt``' intrinsics return the sqrt of the specified operand,
9319 returning the same value as the libm '``sqrt``' functions would. Unlike
9320 ``sqrt`` in libm, however, ``llvm.sqrt`` has undefined behavior for
9321 negative numbers other than -0.0 (which allows for better optimization,
9322 because there is no need to worry about errno being set).
9323 ``llvm.sqrt(-0.0)`` is defined to return -0.0 like IEEE sqrt.
9328 The argument and return value are floating point numbers of the same
9334 This function returns the sqrt of the specified operand if it is a
9335 nonnegative floating point number.
9337 '``llvm.powi.*``' Intrinsic
9338 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9343 This is an overloaded intrinsic. You can use ``llvm.powi`` on any
9344 floating point or vector of floating point type. Not all targets support
9349 declare float @llvm.powi.f32(float %Val, i32 %power)
9350 declare double @llvm.powi.f64(double %Val, i32 %power)
9351 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
9352 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
9353 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
9358 The '``llvm.powi.*``' intrinsics return the first operand raised to the
9359 specified (positive or negative) power. The order of evaluation of
9360 multiplications is not defined. When a vector of floating point type is
9361 used, the second argument remains a scalar integer value.
9366 The second argument is an integer power, and the first is a value to
9367 raise to that power.
9372 This function returns the first value raised to the second power with an
9373 unspecified sequence of rounding operations.
9375 '``llvm.sin.*``' Intrinsic
9376 ^^^^^^^^^^^^^^^^^^^^^^^^^^
9381 This is an overloaded intrinsic. You can use ``llvm.sin`` on any
9382 floating point or vector of floating point type. Not all targets support
9387 declare float @llvm.sin.f32(float %Val)
9388 declare double @llvm.sin.f64(double %Val)
9389 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
9390 declare fp128 @llvm.sin.f128(fp128 %Val)
9391 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
9396 The '``llvm.sin.*``' intrinsics return the sine of the operand.
9401 The argument and return value are floating point numbers of the same
9407 This function returns the sine of the specified operand, returning the
9408 same values as the libm ``sin`` functions would, and handles error
9409 conditions in the same way.
9411 '``llvm.cos.*``' Intrinsic
9412 ^^^^^^^^^^^^^^^^^^^^^^^^^^
9417 This is an overloaded intrinsic. You can use ``llvm.cos`` on any
9418 floating point or vector of floating point type. Not all targets support
9423 declare float @llvm.cos.f32(float %Val)
9424 declare double @llvm.cos.f64(double %Val)
9425 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
9426 declare fp128 @llvm.cos.f128(fp128 %Val)
9427 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
9432 The '``llvm.cos.*``' intrinsics return the cosine of the operand.
9437 The argument and return value are floating point numbers of the same
9443 This function returns the cosine of the specified operand, returning the
9444 same values as the libm ``cos`` functions would, and handles error
9445 conditions in the same way.
9447 '``llvm.pow.*``' Intrinsic
9448 ^^^^^^^^^^^^^^^^^^^^^^^^^^
9453 This is an overloaded intrinsic. You can use ``llvm.pow`` on any
9454 floating point or vector of floating point type. Not all targets support
9459 declare float @llvm.pow.f32(float %Val, float %Power)
9460 declare double @llvm.pow.f64(double %Val, double %Power)
9461 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
9462 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
9463 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
9468 The '``llvm.pow.*``' intrinsics return the first operand raised to the
9469 specified (positive or negative) power.
9474 The second argument is a floating point power, and the first is a value
9475 to raise to that power.
9480 This function returns the first value raised to the second power,
9481 returning the same values as the libm ``pow`` functions would, and
9482 handles error conditions in the same way.
9484 '``llvm.exp.*``' Intrinsic
9485 ^^^^^^^^^^^^^^^^^^^^^^^^^^
9490 This is an overloaded intrinsic. You can use ``llvm.exp`` on any
9491 floating point or vector of floating point type. Not all targets support
9496 declare float @llvm.exp.f32(float %Val)
9497 declare double @llvm.exp.f64(double %Val)
9498 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
9499 declare fp128 @llvm.exp.f128(fp128 %Val)
9500 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
9505 The '``llvm.exp.*``' intrinsics perform the exp function.
9510 The argument and return value are floating point numbers of the same
9516 This function returns the same values as the libm ``exp`` functions
9517 would, and handles error conditions in the same way.
9519 '``llvm.exp2.*``' Intrinsic
9520 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9525 This is an overloaded intrinsic. You can use ``llvm.exp2`` on any
9526 floating point or vector of floating point type. Not all targets support
9531 declare float @llvm.exp2.f32(float %Val)
9532 declare double @llvm.exp2.f64(double %Val)
9533 declare x86_fp80 @llvm.exp2.f80(x86_fp80 %Val)
9534 declare fp128 @llvm.exp2.f128(fp128 %Val)
9535 declare ppc_fp128 @llvm.exp2.ppcf128(ppc_fp128 %Val)
9540 The '``llvm.exp2.*``' intrinsics perform the exp2 function.
9545 The argument and return value are floating point numbers of the same
9551 This function returns the same values as the libm ``exp2`` functions
9552 would, and handles error conditions in the same way.
9554 '``llvm.log.*``' Intrinsic
9555 ^^^^^^^^^^^^^^^^^^^^^^^^^^
9560 This is an overloaded intrinsic. You can use ``llvm.log`` on any
9561 floating point or vector of floating point type. Not all targets support
9566 declare float @llvm.log.f32(float %Val)
9567 declare double @llvm.log.f64(double %Val)
9568 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
9569 declare fp128 @llvm.log.f128(fp128 %Val)
9570 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
9575 The '``llvm.log.*``' intrinsics perform the log function.
9580 The argument and return value are floating point numbers of the same
9586 This function returns the same values as the libm ``log`` functions
9587 would, and handles error conditions in the same way.
9589 '``llvm.log10.*``' Intrinsic
9590 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9595 This is an overloaded intrinsic. You can use ``llvm.log10`` on any
9596 floating point or vector of floating point type. Not all targets support
9601 declare float @llvm.log10.f32(float %Val)
9602 declare double @llvm.log10.f64(double %Val)
9603 declare x86_fp80 @llvm.log10.f80(x86_fp80 %Val)
9604 declare fp128 @llvm.log10.f128(fp128 %Val)
9605 declare ppc_fp128 @llvm.log10.ppcf128(ppc_fp128 %Val)
9610 The '``llvm.log10.*``' intrinsics perform the log10 function.
9615 The argument and return value are floating point numbers of the same
9621 This function returns the same values as the libm ``log10`` functions
9622 would, and handles error conditions in the same way.
9624 '``llvm.log2.*``' Intrinsic
9625 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9630 This is an overloaded intrinsic. You can use ``llvm.log2`` on any
9631 floating point or vector of floating point type. Not all targets support
9636 declare float @llvm.log2.f32(float %Val)
9637 declare double @llvm.log2.f64(double %Val)
9638 declare x86_fp80 @llvm.log2.f80(x86_fp80 %Val)
9639 declare fp128 @llvm.log2.f128(fp128 %Val)
9640 declare ppc_fp128 @llvm.log2.ppcf128(ppc_fp128 %Val)
9645 The '``llvm.log2.*``' intrinsics perform the log2 function.
9650 The argument and return value are floating point numbers of the same
9656 This function returns the same values as the libm ``log2`` functions
9657 would, and handles error conditions in the same way.
9659 '``llvm.fma.*``' Intrinsic
9660 ^^^^^^^^^^^^^^^^^^^^^^^^^^
9665 This is an overloaded intrinsic. You can use ``llvm.fma`` on any
9666 floating point or vector of floating point type. Not all targets support
9671 declare float @llvm.fma.f32(float %a, float %b, float %c)
9672 declare double @llvm.fma.f64(double %a, double %b, double %c)
9673 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
9674 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
9675 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
9680 The '``llvm.fma.*``' intrinsics perform the fused multiply-add
9686 The argument and return value are floating point numbers of the same
9692 This function returns the same values as the libm ``fma`` functions
9693 would, and does not set errno.
9695 '``llvm.fabs.*``' Intrinsic
9696 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9701 This is an overloaded intrinsic. You can use ``llvm.fabs`` on any
9702 floating point or vector of floating point type. Not all targets support
9707 declare float @llvm.fabs.f32(float %Val)
9708 declare double @llvm.fabs.f64(double %Val)
9709 declare x86_fp80 @llvm.fabs.f80(x86_fp80 %Val)
9710 declare fp128 @llvm.fabs.f128(fp128 %Val)
9711 declare ppc_fp128 @llvm.fabs.ppcf128(ppc_fp128 %Val)
9716 The '``llvm.fabs.*``' intrinsics return the absolute value of the
9722 The argument and return value are floating point numbers of the same
9728 This function returns the same values as the libm ``fabs`` functions
9729 would, and handles error conditions in the same way.
9731 '``llvm.minnum.*``' Intrinsic
9732 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9737 This is an overloaded intrinsic. You can use ``llvm.minnum`` on any
9738 floating point or vector of floating point type. Not all targets support
9743 declare float @llvm.minnum.f32(float %Val0, float %Val1)
9744 declare double @llvm.minnum.f64(double %Val0, double %Val1)
9745 declare x86_fp80 @llvm.minnum.f80(x86_fp80 %Val0, x86_fp80 %Val1)
9746 declare fp128 @llvm.minnum.f128(fp128 %Val0, fp128 %Val1)
9747 declare ppc_fp128 @llvm.minnum.ppcf128(ppc_fp128 %Val0, ppc_fp128 %Val1)
9752 The '``llvm.minnum.*``' intrinsics return the minimum of the two
9759 The arguments and return value are floating point numbers of the same
9765 Follows the IEEE-754 semantics for minNum, which also match for libm's
9768 If either operand is a NaN, returns the other non-NaN operand. Returns
9769 NaN only if both operands are NaN. If the operands compare equal,
9770 returns a value that compares equal to both operands. This means that
9771 fmin(+/-0.0, +/-0.0) could return either -0.0 or 0.0.
9773 '``llvm.maxnum.*``' Intrinsic
9774 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9779 This is an overloaded intrinsic. You can use ``llvm.maxnum`` on any
9780 floating point or vector of floating point type. Not all targets support
9785 declare float @llvm.maxnum.f32(float %Val0, float %Val1l)
9786 declare double @llvm.maxnum.f64(double %Val0, double %Val1)
9787 declare x86_fp80 @llvm.maxnum.f80(x86_fp80 %Val0, x86_fp80 %Val1)
9788 declare fp128 @llvm.maxnum.f128(fp128 %Val0, fp128 %Val1)
9789 declare ppc_fp128 @llvm.maxnum.ppcf128(ppc_fp128 %Val0, ppc_fp128 %Val1)
9794 The '``llvm.maxnum.*``' intrinsics return the maximum of the two
9801 The arguments and return value are floating point numbers of the same
9806 Follows the IEEE-754 semantics for maxNum, which also match for libm's
9809 If either operand is a NaN, returns the other non-NaN operand. Returns
9810 NaN only if both operands are NaN. If the operands compare equal,
9811 returns a value that compares equal to both operands. This means that
9812 fmax(+/-0.0, +/-0.0) could return either -0.0 or 0.0.
9814 '``llvm.copysign.*``' Intrinsic
9815 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9820 This is an overloaded intrinsic. You can use ``llvm.copysign`` on any
9821 floating point or vector of floating point type. Not all targets support
9826 declare float @llvm.copysign.f32(float %Mag, float %Sgn)
9827 declare double @llvm.copysign.f64(double %Mag, double %Sgn)
9828 declare x86_fp80 @llvm.copysign.f80(x86_fp80 %Mag, x86_fp80 %Sgn)
9829 declare fp128 @llvm.copysign.f128(fp128 %Mag, fp128 %Sgn)
9830 declare ppc_fp128 @llvm.copysign.ppcf128(ppc_fp128 %Mag, ppc_fp128 %Sgn)
9835 The '``llvm.copysign.*``' intrinsics return a value with the magnitude of the
9836 first operand and the sign of the second operand.
9841 The arguments and return value are floating point numbers of the same
9847 This function returns the same values as the libm ``copysign``
9848 functions would, and handles error conditions in the same way.
9850 '``llvm.floor.*``' Intrinsic
9851 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9856 This is an overloaded intrinsic. You can use ``llvm.floor`` on any
9857 floating point or vector of floating point type. Not all targets support
9862 declare float @llvm.floor.f32(float %Val)
9863 declare double @llvm.floor.f64(double %Val)
9864 declare x86_fp80 @llvm.floor.f80(x86_fp80 %Val)
9865 declare fp128 @llvm.floor.f128(fp128 %Val)
9866 declare ppc_fp128 @llvm.floor.ppcf128(ppc_fp128 %Val)
9871 The '``llvm.floor.*``' intrinsics return the floor of the operand.
9876 The argument and return value are floating point numbers of the same
9882 This function returns the same values as the libm ``floor`` functions
9883 would, and handles error conditions in the same way.
9885 '``llvm.ceil.*``' Intrinsic
9886 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9891 This is an overloaded intrinsic. You can use ``llvm.ceil`` on any
9892 floating point or vector of floating point type. Not all targets support
9897 declare float @llvm.ceil.f32(float %Val)
9898 declare double @llvm.ceil.f64(double %Val)
9899 declare x86_fp80 @llvm.ceil.f80(x86_fp80 %Val)
9900 declare fp128 @llvm.ceil.f128(fp128 %Val)
9901 declare ppc_fp128 @llvm.ceil.ppcf128(ppc_fp128 %Val)
9906 The '``llvm.ceil.*``' intrinsics return the ceiling of the operand.
9911 The argument and return value are floating point numbers of the same
9917 This function returns the same values as the libm ``ceil`` functions
9918 would, and handles error conditions in the same way.
9920 '``llvm.trunc.*``' Intrinsic
9921 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9926 This is an overloaded intrinsic. You can use ``llvm.trunc`` on any
9927 floating point or vector of floating point type. Not all targets support
9932 declare float @llvm.trunc.f32(float %Val)
9933 declare double @llvm.trunc.f64(double %Val)
9934 declare x86_fp80 @llvm.trunc.f80(x86_fp80 %Val)
9935 declare fp128 @llvm.trunc.f128(fp128 %Val)
9936 declare ppc_fp128 @llvm.trunc.ppcf128(ppc_fp128 %Val)
9941 The '``llvm.trunc.*``' intrinsics returns the operand rounded to the
9942 nearest integer not larger in magnitude than the operand.
9947 The argument and return value are floating point numbers of the same
9953 This function returns the same values as the libm ``trunc`` functions
9954 would, and handles error conditions in the same way.
9956 '``llvm.rint.*``' Intrinsic
9957 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
9962 This is an overloaded intrinsic. You can use ``llvm.rint`` on any
9963 floating point or vector of floating point type. Not all targets support
9968 declare float @llvm.rint.f32(float %Val)
9969 declare double @llvm.rint.f64(double %Val)
9970 declare x86_fp80 @llvm.rint.f80(x86_fp80 %Val)
9971 declare fp128 @llvm.rint.f128(fp128 %Val)
9972 declare ppc_fp128 @llvm.rint.ppcf128(ppc_fp128 %Val)
9977 The '``llvm.rint.*``' intrinsics returns the operand rounded to the
9978 nearest integer. It may raise an inexact floating-point exception if the
9979 operand isn't an integer.
9984 The argument and return value are floating point numbers of the same
9990 This function returns the same values as the libm ``rint`` functions
9991 would, and handles error conditions in the same way.
9993 '``llvm.nearbyint.*``' Intrinsic
9994 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
9999 This is an overloaded intrinsic. You can use ``llvm.nearbyint`` on any
10000 floating point or vector of floating point type. Not all targets support
10005 declare float @llvm.nearbyint.f32(float %Val)
10006 declare double @llvm.nearbyint.f64(double %Val)
10007 declare x86_fp80 @llvm.nearbyint.f80(x86_fp80 %Val)
10008 declare fp128 @llvm.nearbyint.f128(fp128 %Val)
10009 declare ppc_fp128 @llvm.nearbyint.ppcf128(ppc_fp128 %Val)
10014 The '``llvm.nearbyint.*``' intrinsics returns the operand rounded to the
10020 The argument and return value are floating point numbers of the same
10026 This function returns the same values as the libm ``nearbyint``
10027 functions would, and handles error conditions in the same way.
10029 '``llvm.round.*``' Intrinsic
10030 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10035 This is an overloaded intrinsic. You can use ``llvm.round`` on any
10036 floating point or vector of floating point type. Not all targets support
10041 declare float @llvm.round.f32(float %Val)
10042 declare double @llvm.round.f64(double %Val)
10043 declare x86_fp80 @llvm.round.f80(x86_fp80 %Val)
10044 declare fp128 @llvm.round.f128(fp128 %Val)
10045 declare ppc_fp128 @llvm.round.ppcf128(ppc_fp128 %Val)
10050 The '``llvm.round.*``' intrinsics returns the operand rounded to the
10056 The argument and return value are floating point numbers of the same
10062 This function returns the same values as the libm ``round``
10063 functions would, and handles error conditions in the same way.
10065 Bit Manipulation Intrinsics
10066 ---------------------------
10068 LLVM provides intrinsics for a few important bit manipulation
10069 operations. These allow efficient code generation for some algorithms.
10071 '``llvm.bswap.*``' Intrinsics
10072 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10077 This is an overloaded intrinsic function. You can use bswap on any
10078 integer type that is an even number of bytes (i.e. BitWidth % 16 == 0).
10082 declare i16 @llvm.bswap.i16(i16 <id>)
10083 declare i32 @llvm.bswap.i32(i32 <id>)
10084 declare i64 @llvm.bswap.i64(i64 <id>)
10089 The '``llvm.bswap``' family of intrinsics is used to byte swap integer
10090 values with an even number of bytes (positive multiple of 16 bits).
10091 These are useful for performing operations on data that is not in the
10092 target's native byte order.
10097 The ``llvm.bswap.i16`` intrinsic returns an i16 value that has the high
10098 and low byte of the input i16 swapped. Similarly, the ``llvm.bswap.i32``
10099 intrinsic returns an i32 value that has the four bytes of the input i32
10100 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the
10101 returned i32 will have its bytes in 3, 2, 1, 0 order. The
10102 ``llvm.bswap.i48``, ``llvm.bswap.i64`` and other intrinsics extend this
10103 concept to additional even-byte lengths (6 bytes, 8 bytes and more,
10106 '``llvm.ctpop.*``' Intrinsic
10107 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10112 This is an overloaded intrinsic. You can use llvm.ctpop on any integer
10113 bit width, or on any vector with integer elements. Not all targets
10114 support all bit widths or vector types, however.
10118 declare i8 @llvm.ctpop.i8(i8 <src>)
10119 declare i16 @llvm.ctpop.i16(i16 <src>)
10120 declare i32 @llvm.ctpop.i32(i32 <src>)
10121 declare i64 @llvm.ctpop.i64(i64 <src>)
10122 declare i256 @llvm.ctpop.i256(i256 <src>)
10123 declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
10128 The '``llvm.ctpop``' family of intrinsics counts the number of bits set
10134 The only argument is the value to be counted. The argument may be of any
10135 integer type, or a vector with integer elements. The return type must
10136 match the argument type.
10141 The '``llvm.ctpop``' intrinsic counts the 1's in a variable, or within
10142 each element of a vector.
10144 '``llvm.ctlz.*``' Intrinsic
10145 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
10150 This is an overloaded intrinsic. You can use ``llvm.ctlz`` on any
10151 integer bit width, or any vector whose elements are integers. Not all
10152 targets support all bit widths or vector types, however.
10156 declare i8 @llvm.ctlz.i8 (i8 <src>, i1 <is_zero_undef>)
10157 declare i16 @llvm.ctlz.i16 (i16 <src>, i1 <is_zero_undef>)
10158 declare i32 @llvm.ctlz.i32 (i32 <src>, i1 <is_zero_undef>)
10159 declare i64 @llvm.ctlz.i64 (i64 <src>, i1 <is_zero_undef>)
10160 declare i256 @llvm.ctlz.i256(i256 <src>, i1 <is_zero_undef>)
10161 declase <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
10166 The '``llvm.ctlz``' family of intrinsic functions counts the number of
10167 leading zeros in a variable.
10172 The first argument is the value to be counted. This argument may be of
10173 any integer type, or a vector with integer element type. The return
10174 type must match the first argument type.
10176 The second argument must be a constant and is a flag to indicate whether
10177 the intrinsic should ensure that a zero as the first argument produces a
10178 defined result. Historically some architectures did not provide a
10179 defined result for zero values as efficiently, and many algorithms are
10180 now predicated on avoiding zero-value inputs.
10185 The '``llvm.ctlz``' intrinsic counts the leading (most significant)
10186 zeros in a variable, or within each element of the vector. If
10187 ``src == 0`` then the result is the size in bits of the type of ``src``
10188 if ``is_zero_undef == 0`` and ``undef`` otherwise. For example,
10189 ``llvm.ctlz(i32 2) = 30``.
10191 '``llvm.cttz.*``' Intrinsic
10192 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
10197 This is an overloaded intrinsic. You can use ``llvm.cttz`` on any
10198 integer bit width, or any vector of integer elements. Not all targets
10199 support all bit widths or vector types, however.
10203 declare i8 @llvm.cttz.i8 (i8 <src>, i1 <is_zero_undef>)
10204 declare i16 @llvm.cttz.i16 (i16 <src>, i1 <is_zero_undef>)
10205 declare i32 @llvm.cttz.i32 (i32 <src>, i1 <is_zero_undef>)
10206 declare i64 @llvm.cttz.i64 (i64 <src>, i1 <is_zero_undef>)
10207 declare i256 @llvm.cttz.i256(i256 <src>, i1 <is_zero_undef>)
10208 declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
10213 The '``llvm.cttz``' family of intrinsic functions counts the number of
10219 The first argument is the value to be counted. This argument may be of
10220 any integer type, or a vector with integer element type. The return
10221 type must match the first argument type.
10223 The second argument must be a constant and is a flag to indicate whether
10224 the intrinsic should ensure that a zero as the first argument produces a
10225 defined result. Historically some architectures did not provide a
10226 defined result for zero values as efficiently, and many algorithms are
10227 now predicated on avoiding zero-value inputs.
10232 The '``llvm.cttz``' intrinsic counts the trailing (least significant)
10233 zeros in a variable, or within each element of a vector. If ``src == 0``
10234 then the result is the size in bits of the type of ``src`` if
10235 ``is_zero_undef == 0`` and ``undef`` otherwise. For example,
10236 ``llvm.cttz(2) = 1``.
10240 Arithmetic with Overflow Intrinsics
10241 -----------------------------------
10243 LLVM provides intrinsics for some arithmetic with overflow operations.
10245 '``llvm.sadd.with.overflow.*``' Intrinsics
10246 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10251 This is an overloaded intrinsic. You can use ``llvm.sadd.with.overflow``
10252 on any integer bit width.
10256 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
10257 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
10258 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
10263 The '``llvm.sadd.with.overflow``' family of intrinsic functions perform
10264 a signed addition of the two arguments, and indicate whether an overflow
10265 occurred during the signed summation.
10270 The arguments (%a and %b) and the first element of the result structure
10271 may be of integer types of any bit width, but they must have the same
10272 bit width. The second element of the result structure must be of type
10273 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
10279 The '``llvm.sadd.with.overflow``' family of intrinsic functions perform
10280 a signed addition of the two variables. They return a structure --- the
10281 first element of which is the signed summation, and the second element
10282 of which is a bit specifying if the signed summation resulted in an
10288 .. code-block:: llvm
10290 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
10291 %sum = extractvalue {i32, i1} %res, 0
10292 %obit = extractvalue {i32, i1} %res, 1
10293 br i1 %obit, label %overflow, label %normal
10295 '``llvm.uadd.with.overflow.*``' Intrinsics
10296 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10301 This is an overloaded intrinsic. You can use ``llvm.uadd.with.overflow``
10302 on any integer bit width.
10306 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
10307 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
10308 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
10313 The '``llvm.uadd.with.overflow``' family of intrinsic functions perform
10314 an unsigned addition of the two arguments, and indicate whether a carry
10315 occurred during the unsigned summation.
10320 The arguments (%a and %b) and the first element of the result structure
10321 may be of integer types of any bit width, but they must have the same
10322 bit width. The second element of the result structure must be of type
10323 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
10329 The '``llvm.uadd.with.overflow``' family of intrinsic functions perform
10330 an unsigned addition of the two arguments. They return a structure --- the
10331 first element of which is the sum, and the second element of which is a
10332 bit specifying if the unsigned summation resulted in a carry.
10337 .. code-block:: llvm
10339 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
10340 %sum = extractvalue {i32, i1} %res, 0
10341 %obit = extractvalue {i32, i1} %res, 1
10342 br i1 %obit, label %carry, label %normal
10344 '``llvm.ssub.with.overflow.*``' Intrinsics
10345 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10350 This is an overloaded intrinsic. You can use ``llvm.ssub.with.overflow``
10351 on any integer bit width.
10355 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
10356 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
10357 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
10362 The '``llvm.ssub.with.overflow``' family of intrinsic functions perform
10363 a signed subtraction of the two arguments, and indicate whether an
10364 overflow occurred during the signed subtraction.
10369 The arguments (%a and %b) and the first element of the result structure
10370 may be of integer types of any bit width, but they must have the same
10371 bit width. The second element of the result structure must be of type
10372 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
10378 The '``llvm.ssub.with.overflow``' family of intrinsic functions perform
10379 a signed subtraction of the two arguments. They return a structure --- the
10380 first element of which is the subtraction, and the second element of
10381 which is a bit specifying if the signed subtraction resulted in an
10387 .. code-block:: llvm
10389 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
10390 %sum = extractvalue {i32, i1} %res, 0
10391 %obit = extractvalue {i32, i1} %res, 1
10392 br i1 %obit, label %overflow, label %normal
10394 '``llvm.usub.with.overflow.*``' Intrinsics
10395 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10400 This is an overloaded intrinsic. You can use ``llvm.usub.with.overflow``
10401 on any integer bit width.
10405 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
10406 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
10407 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
10412 The '``llvm.usub.with.overflow``' family of intrinsic functions perform
10413 an unsigned subtraction of the two arguments, and indicate whether an
10414 overflow occurred during the unsigned subtraction.
10419 The arguments (%a and %b) and the first element of the result structure
10420 may be of integer types of any bit width, but they must have the same
10421 bit width. The second element of the result structure must be of type
10422 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
10428 The '``llvm.usub.with.overflow``' family of intrinsic functions perform
10429 an unsigned subtraction of the two arguments. They return a structure ---
10430 the first element of which is the subtraction, and the second element of
10431 which is a bit specifying if the unsigned subtraction resulted in an
10437 .. code-block:: llvm
10439 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
10440 %sum = extractvalue {i32, i1} %res, 0
10441 %obit = extractvalue {i32, i1} %res, 1
10442 br i1 %obit, label %overflow, label %normal
10444 '``llvm.smul.with.overflow.*``' Intrinsics
10445 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10450 This is an overloaded intrinsic. You can use ``llvm.smul.with.overflow``
10451 on any integer bit width.
10455 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
10456 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
10457 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
10462 The '``llvm.smul.with.overflow``' family of intrinsic functions perform
10463 a signed multiplication of the two arguments, and indicate whether an
10464 overflow occurred during the signed multiplication.
10469 The arguments (%a and %b) and the first element of the result structure
10470 may be of integer types of any bit width, but they must have the same
10471 bit width. The second element of the result structure must be of type
10472 ``i1``. ``%a`` and ``%b`` are the two values that will undergo signed
10478 The '``llvm.smul.with.overflow``' family of intrinsic functions perform
10479 a signed multiplication of the two arguments. They return a structure ---
10480 the first element of which is the multiplication, and the second element
10481 of which is a bit specifying if the signed multiplication resulted in an
10487 .. code-block:: llvm
10489 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
10490 %sum = extractvalue {i32, i1} %res, 0
10491 %obit = extractvalue {i32, i1} %res, 1
10492 br i1 %obit, label %overflow, label %normal
10494 '``llvm.umul.with.overflow.*``' Intrinsics
10495 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10500 This is an overloaded intrinsic. You can use ``llvm.umul.with.overflow``
10501 on any integer bit width.
10505 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
10506 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
10507 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
10512 The '``llvm.umul.with.overflow``' family of intrinsic functions perform
10513 a unsigned multiplication of the two arguments, and indicate whether an
10514 overflow occurred during the unsigned multiplication.
10519 The arguments (%a and %b) and the first element of the result structure
10520 may be of integer types of any bit width, but they must have the same
10521 bit width. The second element of the result structure must be of type
10522 ``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned
10528 The '``llvm.umul.with.overflow``' family of intrinsic functions perform
10529 an unsigned multiplication of the two arguments. They return a structure ---
10530 the first element of which is the multiplication, and the second
10531 element of which is a bit specifying if the unsigned multiplication
10532 resulted in an overflow.
10537 .. code-block:: llvm
10539 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
10540 %sum = extractvalue {i32, i1} %res, 0
10541 %obit = extractvalue {i32, i1} %res, 1
10542 br i1 %obit, label %overflow, label %normal
10544 Specialised Arithmetic Intrinsics
10545 ---------------------------------
10547 '``llvm.fmuladd.*``' Intrinsic
10548 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10555 declare float @llvm.fmuladd.f32(float %a, float %b, float %c)
10556 declare double @llvm.fmuladd.f64(double %a, double %b, double %c)
10561 The '``llvm.fmuladd.*``' intrinsic functions represent multiply-add
10562 expressions that can be fused if the code generator determines that (a) the
10563 target instruction set has support for a fused operation, and (b) that the
10564 fused operation is more efficient than the equivalent, separate pair of mul
10565 and add instructions.
10570 The '``llvm.fmuladd.*``' intrinsics each take three arguments: two
10571 multiplicands, a and b, and an addend c.
10580 %0 = call float @llvm.fmuladd.f32(%a, %b, %c)
10582 is equivalent to the expression a \* b + c, except that rounding will
10583 not be performed between the multiplication and addition steps if the
10584 code generator fuses the operations. Fusion is not guaranteed, even if
10585 the target platform supports it. If a fused multiply-add is required the
10586 corresponding llvm.fma.\* intrinsic function should be used
10587 instead. This never sets errno, just as '``llvm.fma.*``'.
10592 .. code-block:: llvm
10594 %r2 = call float @llvm.fmuladd.f32(float %a, float %b, float %c) ; yields float:r2 = (a * b) + c
10596 Half Precision Floating Point Intrinsics
10597 ----------------------------------------
10599 For most target platforms, half precision floating point is a
10600 storage-only format. This means that it is a dense encoding (in memory)
10601 but does not support computation in the format.
10603 This means that code must first load the half-precision floating point
10604 value as an i16, then convert it to float with
10605 :ref:`llvm.convert.from.fp16 <int_convert_from_fp16>`. Computation can
10606 then be performed on the float value (including extending to double
10607 etc). To store the value back to memory, it is first converted to float
10608 if needed, then converted to i16 with
10609 :ref:`llvm.convert.to.fp16 <int_convert_to_fp16>`, then storing as an
10612 .. _int_convert_to_fp16:
10614 '``llvm.convert.to.fp16``' Intrinsic
10615 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10622 declare i16 @llvm.convert.to.fp16.f32(float %a)
10623 declare i16 @llvm.convert.to.fp16.f64(double %a)
10628 The '``llvm.convert.to.fp16``' intrinsic function performs a conversion from a
10629 conventional floating point type to half precision floating point format.
10634 The intrinsic function contains single argument - the value to be
10640 The '``llvm.convert.to.fp16``' intrinsic function performs a conversion from a
10641 conventional floating point format to half precision floating point format. The
10642 return value is an ``i16`` which contains the converted number.
10647 .. code-block:: llvm
10649 %res = call i16 @llvm.convert.to.fp16.f32(float %a)
10650 store i16 %res, i16* @x, align 2
10652 .. _int_convert_from_fp16:
10654 '``llvm.convert.from.fp16``' Intrinsic
10655 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10662 declare float @llvm.convert.from.fp16.f32(i16 %a)
10663 declare double @llvm.convert.from.fp16.f64(i16 %a)
10668 The '``llvm.convert.from.fp16``' intrinsic function performs a
10669 conversion from half precision floating point format to single precision
10670 floating point format.
10675 The intrinsic function contains single argument - the value to be
10681 The '``llvm.convert.from.fp16``' intrinsic function performs a
10682 conversion from half single precision floating point format to single
10683 precision floating point format. The input half-float value is
10684 represented by an ``i16`` value.
10689 .. code-block:: llvm
10691 %a = load i16, i16* @x, align 2
10692 %res = call float @llvm.convert.from.fp16(i16 %a)
10694 .. _dbg_intrinsics:
10696 Debugger Intrinsics
10697 -------------------
10699 The LLVM debugger intrinsics (which all start with ``llvm.dbg.``
10700 prefix), are described in the `LLVM Source Level
10701 Debugging <SourceLevelDebugging.html#format_common_intrinsics>`_
10704 Exception Handling Intrinsics
10705 -----------------------------
10707 The LLVM exception handling intrinsics (which all start with
10708 ``llvm.eh.`` prefix), are described in the `LLVM Exception
10709 Handling <ExceptionHandling.html#format_common_intrinsics>`_ document.
10711 .. _int_trampoline:
10713 Trampoline Intrinsics
10714 ---------------------
10716 These intrinsics make it possible to excise one parameter, marked with
10717 the :ref:`nest <nest>` attribute, from a function. The result is a
10718 callable function pointer lacking the nest parameter - the caller does
10719 not need to provide a value for it. Instead, the value to use is stored
10720 in advance in a "trampoline", a block of memory usually allocated on the
10721 stack, which also contains code to splice the nest value into the
10722 argument list. This is used to implement the GCC nested function address
10725 For example, if the function is ``i32 f(i8* nest %c, i32 %x, i32 %y)``
10726 then the resulting function pointer has signature ``i32 (i32, i32)*``.
10727 It can be created as follows:
10729 .. code-block:: llvm
10731 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
10732 %tramp1 = getelementptr [10 x i8], [10 x i8]* %tramp, i32 0, i32 0
10733 call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval)
10734 %p = call i8* @llvm.adjust.trampoline(i8* %tramp1)
10735 %fp = bitcast i8* %p to i32 (i32, i32)*
10737 The call ``%val = call i32 %fp(i32 %x, i32 %y)`` is then equivalent to
10738 ``%val = call i32 %f(i8* %nval, i32 %x, i32 %y)``.
10742 '``llvm.init.trampoline``' Intrinsic
10743 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10750 declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
10755 This fills the memory pointed to by ``tramp`` with executable code,
10756 turning it into a trampoline.
10761 The ``llvm.init.trampoline`` intrinsic takes three arguments, all
10762 pointers. The ``tramp`` argument must point to a sufficiently large and
10763 sufficiently aligned block of memory; this memory is written to by the
10764 intrinsic. Note that the size and the alignment are target-specific -
10765 LLVM currently provides no portable way of determining them, so a
10766 front-end that generates this intrinsic needs to have some
10767 target-specific knowledge. The ``func`` argument must hold a function
10768 bitcast to an ``i8*``.
10773 The block of memory pointed to by ``tramp`` is filled with target
10774 dependent code, turning it into a function. Then ``tramp`` needs to be
10775 passed to :ref:`llvm.adjust.trampoline <int_at>` to get a pointer which can
10776 be :ref:`bitcast (to a new function) and called <int_trampoline>`. The new
10777 function's signature is the same as that of ``func`` with any arguments
10778 marked with the ``nest`` attribute removed. At most one such ``nest``
10779 argument is allowed, and it must be of pointer type. Calling the new
10780 function is equivalent to calling ``func`` with the same argument list,
10781 but with ``nval`` used for the missing ``nest`` argument. If, after
10782 calling ``llvm.init.trampoline``, the memory pointed to by ``tramp`` is
10783 modified, then the effect of any later call to the returned function
10784 pointer is undefined.
10788 '``llvm.adjust.trampoline``' Intrinsic
10789 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10796 declare i8* @llvm.adjust.trampoline(i8* <tramp>)
10801 This performs any required machine-specific adjustment to the address of
10802 a trampoline (passed as ``tramp``).
10807 ``tramp`` must point to a block of memory which already has trampoline
10808 code filled in by a previous call to
10809 :ref:`llvm.init.trampoline <int_it>`.
10814 On some architectures the address of the code to be executed needs to be
10815 different than the address where the trampoline is actually stored. This
10816 intrinsic returns the executable address corresponding to ``tramp``
10817 after performing the required machine specific adjustments. The pointer
10818 returned can then be :ref:`bitcast and executed <int_trampoline>`.
10820 .. _int_mload_mstore:
10822 Masked Vector Load and Store Intrinsics
10823 ---------------------------------------
10825 LLVM provides intrinsics for predicated vector load and store operations. The predicate is specified by a mask operand, which holds one bit per vector element, switching the associated vector lane on or off. The memory addresses corresponding to the "off" lanes are not accessed. When all bits of the mask are on, the intrinsic is identical to a regular vector load or store. When all bits are off, no memory is accessed.
10829 '``llvm.masked.load.*``' Intrinsics
10830 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10834 This is an overloaded intrinsic. The loaded data is a vector of any integer or floating point data type.
10838 declare <16 x float> @llvm.masked.load.v16f32 (<16 x float>* <ptr>, i32 <alignment>, <16 x i1> <mask>, <16 x float> <passthru>)
10839 declare <2 x double> @llvm.masked.load.v2f64 (<2 x double>* <ptr>, i32 <alignment>, <2 x i1> <mask>, <2 x double> <passthru>)
10844 Reads a vector from memory according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes. The masked-off lanes in the result vector are taken from the corresponding lanes of the '``passthru``' operand.
10850 The first operand is the base pointer for the load. The second operand is the alignment of the source location. It must be a constant integer value. The third operand, mask, is a vector of boolean values with the same number of elements as the return type. The fourth is a pass-through value that is used to fill the masked-off lanes of the result. The return type, underlying type of the base pointer and the type of the '``passthru``' operand are the same vector types.
10856 The '``llvm.masked.load``' intrinsic is designed for conditional reading of selected vector elements in a single IR operation. It is useful for targets that support vector masked loads and allows vectorizing predicated basic blocks on these targets. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar load operations.
10857 The result of this operation is equivalent to a regular vector load instruction followed by a 'select' between the loaded and the passthru values, predicated on the same mask. However, using this intrinsic prevents exceptions on memory access to masked-off lanes.
10862 %res = call <16 x float> @llvm.masked.load.v16f32 (<16 x float>* %ptr, i32 4, <16 x i1>%mask, <16 x float> %passthru)
10864 ;; The result of the two following instructions is identical aside from potential memory access exception
10865 %loadlal = load <16 x float>, <16 x float>* %ptr, align 4
10866 %res = select <16 x i1> %mask, <16 x float> %loadlal, <16 x float> %passthru
10870 '``llvm.masked.store.*``' Intrinsics
10871 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10875 This is an overloaded intrinsic. The data stored in memory is a vector of any integer or floating point data type.
10879 declare void @llvm.masked.store.v8i32 (<8 x i32> <value>, <8 x i32> * <ptr>, i32 <alignment>, <8 x i1> <mask>)
10880 declare void @llvm.masked.store.v16f32(<16 x i32> <value>, <16 x i32>* <ptr>, i32 <alignment>, <16 x i1> <mask>)
10885 Writes a vector to memory according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes.
10890 The first operand is the vector value to be written to memory. The second operand is the base pointer for the store, it has the same underlying type as the value operand. The third operand is the alignment of the destination location. The fourth operand, mask, is a vector of boolean values. The types of the mask and the value operand must have the same number of vector elements.
10896 The '``llvm.masked.store``' intrinsics is designed for conditional writing of selected vector elements in a single IR operation. It is useful for targets that support vector masked store and allows vectorizing predicated basic blocks on these targets. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar store operations.
10897 The result of this operation is equivalent to a load-modify-store sequence. However, using this intrinsic prevents exceptions and data races on memory access to masked-off lanes.
10901 call void @llvm.masked.store.v16f32(<16 x float> %value, <16 x float>* %ptr, i32 4, <16 x i1> %mask)
10903 ;; The result of the following instructions is identical aside from potential data races and memory access exceptions
10904 %oldval = load <16 x float>, <16 x float>* %ptr, align 4
10905 %res = select <16 x i1> %mask, <16 x float> %value, <16 x float> %oldval
10906 store <16 x float> %res, <16 x float>* %ptr, align 4
10909 Masked Vector Gather and Scatter Intrinsics
10910 -------------------------------------------
10912 LLVM provides intrinsics for vector gather and scatter operations. They are similar to :ref:`Masked Vector Load and Store <int_mload_mstore>`, except they are designed for arbitrary memory accesses, rather than sequential memory accesses. Gather and scatter also employ a mask operand, which holds one bit per vector element, switching the associated vector lane on or off. The memory addresses corresponding to the "off" lanes are not accessed. When all bits are off, no memory is accessed.
10916 '``llvm.masked.gather.*``' Intrinsics
10917 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10921 This is an overloaded intrinsic. The loaded data are multiple scalar values of any integer or floating point data type gathered together into one vector.
10925 declare <16 x float> @llvm.masked.gather.v16f32 (<16 x float*> <ptrs>, i32 <alignment>, <16 x i1> <mask>, <16 x float> <passthru>)
10926 declare <2 x double> @llvm.masked.gather.v2f64 (<2 x double*> <ptrs>, i32 <alignment>, <2 x i1> <mask>, <2 x double> <passthru>)
10931 Reads scalar values from arbitrary memory locations and gathers them into one vector. The memory locations are provided in the vector of pointers '``ptrs``'. The memory is accessed according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes. The masked-off lanes in the result vector are taken from the corresponding lanes of the '``passthru``' operand.
10937 The first operand is a vector of pointers which holds all memory addresses to read. The second operand is an alignment of the source addresses. It must be a constant integer value. The third operand, mask, is a vector of boolean values with the same number of elements as the return type. The fourth is a pass-through value that is used to fill the masked-off lanes of the result. The return type, underlying type of the vector of pointers and the type of the '``passthru``' operand are the same vector types.
10943 The '``llvm.masked.gather``' intrinsic is designed for conditional reading of multiple scalar values from arbitrary memory locations in a single IR operation. It is useful for targets that support vector masked gathers and allows vectorizing basic blocks with data and control divergence. Other targets may support this intrinsic differently, for example by lowering it into a sequence of scalar load operations.
10944 The semantics of this operation are equivalent to a sequence of conditional scalar loads with subsequent gathering all loaded values into a single vector. The mask restricts memory access to certain lanes and facilitates vectorization of predicated basic blocks.
10949 %res = call <4 x double> @llvm.masked.gather.v4f64 (<4 x double*> %ptrs, i32 8, <4 x i1>%mask, <4 x double> <true, true, true, true>)
10951 ;; The gather with all-true mask is equivalent to the following instruction sequence
10952 %ptr0 = extractelement <4 x double*> %ptrs, i32 0
10953 %ptr1 = extractelement <4 x double*> %ptrs, i32 1
10954 %ptr2 = extractelement <4 x double*> %ptrs, i32 2
10955 %ptr3 = extractelement <4 x double*> %ptrs, i32 3
10957 %val0 = load double, double* %ptr0, align 8
10958 %val1 = load double, double* %ptr1, align 8
10959 %val2 = load double, double* %ptr2, align 8
10960 %val3 = load double, double* %ptr3, align 8
10962 %vec0 = insertelement <4 x double>undef, %val0, 0
10963 %vec01 = insertelement <4 x double>%vec0, %val1, 1
10964 %vec012 = insertelement <4 x double>%vec01, %val2, 2
10965 %vec0123 = insertelement <4 x double>%vec012, %val3, 3
10969 '``llvm.masked.scatter.*``' Intrinsics
10970 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10974 This is an overloaded intrinsic. The data stored in memory is a vector of any integer or floating point data type. Each vector element is stored in an arbitrary memory addresses. Scatter with overlapping addresses is guaranteed to be ordered from least-significant to most-significant element.
10978 declare void @llvm.masked.scatter.v8i32 (<8 x i32> <value>, <8 x i32*> <ptrs>, i32 <alignment>, <8 x i1> <mask>)
10979 declare void @llvm.masked.scatter.v16f32(<16 x i32> <value>, <16 x i32*> <ptrs>, i32 <alignment>, <16 x i1> <mask>)
10984 Writes each element from the value vector to the corresponding memory address. The memory addresses are represented as a vector of pointers. Writing is done according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes.
10989 The first operand is a vector value to be written to memory. The second operand is a vector of pointers, pointing to where the value elements should be stored. It has the same underlying type as the value operand. The third operand is an alignment of the destination addresses. The fourth operand, mask, is a vector of boolean values. The types of the mask and the value operand must have the same number of vector elements.
10995 The '``llvm.masked.scatter``' intrinsics is designed for writing selected vector elements to arbitrary memory addresses in a single IR operation. The operation may be conditional, when not all bits in the mask are switched on. It is useful for targets that support vector masked scatter and allows vectorizing basic blocks with data and control divergency. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar store operations.
10999 ;; This instruction unconditionaly stores data vector in multiple addresses
11000 call @llvm.masked.scatter.v8i32 (<8 x i32> %value, <8 x i32*> %ptrs, i32 4, <8 x i1> <true, true, .. true>)
11002 ;; It is equivalent to a list of scalar stores
11003 %val0 = extractelement <8 x i32> %value, i32 0
11004 %val1 = extractelement <8 x i32> %value, i32 1
11006 %val7 = extractelement <8 x i32> %value, i32 7
11007 %ptr0 = extractelement <8 x i32*> %ptrs, i32 0
11008 %ptr1 = extractelement <8 x i32*> %ptrs, i32 1
11010 %ptr7 = extractelement <8 x i32*> %ptrs, i32 7
11011 ;; Note: the order of the following stores is important when they overlap:
11012 store i32 %val0, i32* %ptr0, align 4
11013 store i32 %val1, i32* %ptr1, align 4
11015 store i32 %val7, i32* %ptr7, align 4
11021 This class of intrinsics provides information about the lifetime of
11022 memory objects and ranges where variables are immutable.
11026 '``llvm.lifetime.start``' Intrinsic
11027 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11034 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
11039 The '``llvm.lifetime.start``' intrinsic specifies the start of a memory
11045 The first argument is a constant integer representing the size of the
11046 object, or -1 if it is variable sized. The second argument is a pointer
11052 This intrinsic indicates that before this point in the code, the value
11053 of the memory pointed to by ``ptr`` is dead. This means that it is known
11054 to never be used and has an undefined value. A load from the pointer
11055 that precedes this intrinsic can be replaced with ``'undef'``.
11059 '``llvm.lifetime.end``' Intrinsic
11060 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11067 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
11072 The '``llvm.lifetime.end``' intrinsic specifies the end of a memory
11078 The first argument is a constant integer representing the size of the
11079 object, or -1 if it is variable sized. The second argument is a pointer
11085 This intrinsic indicates that after this point in the code, the value of
11086 the memory pointed to by ``ptr`` is dead. This means that it is known to
11087 never be used and has an undefined value. Any stores into the memory
11088 object following this intrinsic may be removed as dead.
11090 '``llvm.invariant.start``' Intrinsic
11091 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11098 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
11103 The '``llvm.invariant.start``' intrinsic specifies that the contents of
11104 a memory object will not change.
11109 The first argument is a constant integer representing the size of the
11110 object, or -1 if it is variable sized. The second argument is a pointer
11116 This intrinsic indicates that until an ``llvm.invariant.end`` that uses
11117 the return value, the referenced memory location is constant and
11120 '``llvm.invariant.end``' Intrinsic
11121 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11128 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
11133 The '``llvm.invariant.end``' intrinsic specifies that the contents of a
11134 memory object are mutable.
11139 The first argument is the matching ``llvm.invariant.start`` intrinsic.
11140 The second argument is a constant integer representing the size of the
11141 object, or -1 if it is variable sized and the third argument is a
11142 pointer to the object.
11147 This intrinsic indicates that the memory is mutable again.
11152 This class of intrinsics is designed to be generic and has no specific
11155 '``llvm.var.annotation``' Intrinsic
11156 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11163 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
11168 The '``llvm.var.annotation``' intrinsic.
11173 The first argument is a pointer to a value, the second is a pointer to a
11174 global string, the third is a pointer to a global string which is the
11175 source file name, and the last argument is the line number.
11180 This intrinsic allows annotation of local variables with arbitrary
11181 strings. This can be useful for special purpose optimizations that want
11182 to look for these annotations. These have no other defined use; they are
11183 ignored by code generation and optimization.
11185 '``llvm.ptr.annotation.*``' Intrinsic
11186 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11191 This is an overloaded intrinsic. You can use '``llvm.ptr.annotation``' on a
11192 pointer to an integer of any width. *NOTE* you must specify an address space for
11193 the pointer. The identifier for the default address space is the integer
11198 declare i8* @llvm.ptr.annotation.p<address space>i8(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
11199 declare i16* @llvm.ptr.annotation.p<address space>i16(i16* <val>, i8* <str>, i8* <str>, i32 <int>)
11200 declare i32* @llvm.ptr.annotation.p<address space>i32(i32* <val>, i8* <str>, i8* <str>, i32 <int>)
11201 declare i64* @llvm.ptr.annotation.p<address space>i64(i64* <val>, i8* <str>, i8* <str>, i32 <int>)
11202 declare i256* @llvm.ptr.annotation.p<address space>i256(i256* <val>, i8* <str>, i8* <str>, i32 <int>)
11207 The '``llvm.ptr.annotation``' intrinsic.
11212 The first argument is a pointer to an integer value of arbitrary bitwidth
11213 (result of some expression), the second is a pointer to a global string, the
11214 third is a pointer to a global string which is the source file name, and the
11215 last argument is the line number. It returns the value of the first argument.
11220 This intrinsic allows annotation of a pointer to an integer with arbitrary
11221 strings. This can be useful for special purpose optimizations that want to look
11222 for these annotations. These have no other defined use; they are ignored by code
11223 generation and optimization.
11225 '``llvm.annotation.*``' Intrinsic
11226 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11231 This is an overloaded intrinsic. You can use '``llvm.annotation``' on
11232 any integer bit width.
11236 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
11237 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
11238 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
11239 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
11240 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
11245 The '``llvm.annotation``' intrinsic.
11250 The first argument is an integer value (result of some expression), the
11251 second is a pointer to a global string, the third is a pointer to a
11252 global string which is the source file name, and the last argument is
11253 the line number. It returns the value of the first argument.
11258 This intrinsic allows annotations to be put on arbitrary expressions
11259 with arbitrary strings. This can be useful for special purpose
11260 optimizations that want to look for these annotations. These have no
11261 other defined use; they are ignored by code generation and optimization.
11263 '``llvm.trap``' Intrinsic
11264 ^^^^^^^^^^^^^^^^^^^^^^^^^
11271 declare void @llvm.trap() noreturn nounwind
11276 The '``llvm.trap``' intrinsic.
11286 This intrinsic is lowered to the target dependent trap instruction. If
11287 the target does not have a trap instruction, this intrinsic will be
11288 lowered to a call of the ``abort()`` function.
11290 '``llvm.debugtrap``' Intrinsic
11291 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11298 declare void @llvm.debugtrap() nounwind
11303 The '``llvm.debugtrap``' intrinsic.
11313 This intrinsic is lowered to code which is intended to cause an
11314 execution trap with the intention of requesting the attention of a
11317 '``llvm.stackprotector``' Intrinsic
11318 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11325 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
11330 The ``llvm.stackprotector`` intrinsic takes the ``guard`` and stores it
11331 onto the stack at ``slot``. The stack slot is adjusted to ensure that it
11332 is placed on the stack before local variables.
11337 The ``llvm.stackprotector`` intrinsic requires two pointer arguments.
11338 The first argument is the value loaded from the stack guard
11339 ``@__stack_chk_guard``. The second variable is an ``alloca`` that has
11340 enough space to hold the value of the guard.
11345 This intrinsic causes the prologue/epilogue inserter to force the position of
11346 the ``AllocaInst`` stack slot to be before local variables on the stack. This is
11347 to ensure that if a local variable on the stack is overwritten, it will destroy
11348 the value of the guard. When the function exits, the guard on the stack is
11349 checked against the original guard by ``llvm.stackprotectorcheck``. If they are
11350 different, then ``llvm.stackprotectorcheck`` causes the program to abort by
11351 calling the ``__stack_chk_fail()`` function.
11353 '``llvm.stackprotectorcheck``' Intrinsic
11354 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11361 declare void @llvm.stackprotectorcheck(i8** <guard>)
11366 The ``llvm.stackprotectorcheck`` intrinsic compares ``guard`` against an already
11367 created stack protector and if they are not equal calls the
11368 ``__stack_chk_fail()`` function.
11373 The ``llvm.stackprotectorcheck`` intrinsic requires one pointer argument, the
11374 the variable ``@__stack_chk_guard``.
11379 This intrinsic is provided to perform the stack protector check by comparing
11380 ``guard`` with the stack slot created by ``llvm.stackprotector`` and if the
11381 values do not match call the ``__stack_chk_fail()`` function.
11383 The reason to provide this as an IR level intrinsic instead of implementing it
11384 via other IR operations is that in order to perform this operation at the IR
11385 level without an intrinsic, one would need to create additional basic blocks to
11386 handle the success/failure cases. This makes it difficult to stop the stack
11387 protector check from disrupting sibling tail calls in Codegen. With this
11388 intrinsic, we are able to generate the stack protector basic blocks late in
11389 codegen after the tail call decision has occurred.
11391 '``llvm.objectsize``' Intrinsic
11392 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11399 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <min>)
11400 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <min>)
11405 The ``llvm.objectsize`` intrinsic is designed to provide information to
11406 the optimizers to determine at compile time whether a) an operation
11407 (like memcpy) will overflow a buffer that corresponds to an object, or
11408 b) that a runtime check for overflow isn't necessary. An object in this
11409 context means an allocation of a specific class, structure, array, or
11415 The ``llvm.objectsize`` intrinsic takes two arguments. The first
11416 argument is a pointer to or into the ``object``. The second argument is
11417 a boolean and determines whether ``llvm.objectsize`` returns 0 (if true)
11418 or -1 (if false) when the object size is unknown. The second argument
11419 only accepts constants.
11424 The ``llvm.objectsize`` intrinsic is lowered to a constant representing
11425 the size of the object concerned. If the size cannot be determined at
11426 compile time, ``llvm.objectsize`` returns ``i32/i64 -1 or 0`` (depending
11427 on the ``min`` argument).
11429 '``llvm.expect``' Intrinsic
11430 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
11435 This is an overloaded intrinsic. You can use ``llvm.expect`` on any
11440 declare i1 @llvm.expect.i1(i1 <val>, i1 <expected_val>)
11441 declare i32 @llvm.expect.i32(i32 <val>, i32 <expected_val>)
11442 declare i64 @llvm.expect.i64(i64 <val>, i64 <expected_val>)
11447 The ``llvm.expect`` intrinsic provides information about expected (the
11448 most probable) value of ``val``, which can be used by optimizers.
11453 The ``llvm.expect`` intrinsic takes two arguments. The first argument is
11454 a value. The second argument is an expected value, this needs to be a
11455 constant value, variables are not allowed.
11460 This intrinsic is lowered to the ``val``.
11464 '``llvm.assume``' Intrinsic
11465 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11472 declare void @llvm.assume(i1 %cond)
11477 The ``llvm.assume`` allows the optimizer to assume that the provided
11478 condition is true. This information can then be used in simplifying other parts
11484 The condition which the optimizer may assume is always true.
11489 The intrinsic allows the optimizer to assume that the provided condition is
11490 always true whenever the control flow reaches the intrinsic call. No code is
11491 generated for this intrinsic, and instructions that contribute only to the
11492 provided condition are not used for code generation. If the condition is
11493 violated during execution, the behavior is undefined.
11495 Note that the optimizer might limit the transformations performed on values
11496 used by the ``llvm.assume`` intrinsic in order to preserve the instructions
11497 only used to form the intrinsic's input argument. This might prove undesirable
11498 if the extra information provided by the ``llvm.assume`` intrinsic does not cause
11499 sufficient overall improvement in code quality. For this reason,
11500 ``llvm.assume`` should not be used to document basic mathematical invariants
11501 that the optimizer can otherwise deduce or facts that are of little use to the
11506 '``llvm.bitset.test``' Intrinsic
11507 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11514 declare i1 @llvm.bitset.test(i8* %ptr, metadata %bitset) nounwind readnone
11520 The first argument is a pointer to be tested. The second argument is a
11521 metadata string containing the name of a :doc:`bitset <BitSets>`.
11526 The ``llvm.bitset.test`` intrinsic tests whether the given pointer is a
11527 member of the given bitset.
11529 '``llvm.donothing``' Intrinsic
11530 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11537 declare void @llvm.donothing() nounwind readnone
11542 The ``llvm.donothing`` intrinsic doesn't perform any operation. It's one of only
11543 two intrinsics (besides ``llvm.experimental.patchpoint``) that can be called
11544 with an invoke instruction.
11554 This intrinsic does nothing, and it's removed by optimizers and ignored
11557 Stack Map Intrinsics
11558 --------------------
11560 LLVM provides experimental intrinsics to support runtime patching
11561 mechanisms commonly desired in dynamic language JITs. These intrinsics
11562 are described in :doc:`StackMaps`.