1============================== 2LLVM Language Reference Manual 3============================== 4 5.. contents:: 6 :local: 7 :depth: 4 8 9Abstract 10======== 11 12This document is a reference manual for the LLVM assembly language. LLVM 13is a Static Single Assignment (SSA) based representation that provides 14type safety, low-level operations, flexibility, and the capability of 15representing 'all' high-level languages cleanly. It is the common code 16representation used throughout all phases of the LLVM compilation 17strategy. 18 19Introduction 20============ 21 22The LLVM code representation is designed to be used in three different 23forms: as an in-memory compiler IR, as an on-disk bitcode representation 24(suitable for fast loading by a Just-In-Time compiler), and as a human 25readable assembly language representation. This allows LLVM to provide a 26powerful intermediate representation for efficient compiler 27transformations and analysis, while providing a natural means to debug 28and visualize the transformations. The three different forms of LLVM are 29all equivalent. This document describes the human readable 30representation and notation. 31 32The LLVM representation aims to be light-weight and low-level while 33being expressive, typed, and extensible at the same time. It aims to be 34a "universal IR" of sorts, by being at a low enough level that 35high-level ideas may be cleanly mapped to it (similar to how 36microprocessors are "universal IR's", allowing many source languages to 37be mapped to them). By providing type information, LLVM can be used as 38the target of optimizations: for example, through pointer analysis, it 39can be proven that a C automatic variable is never accessed outside of 40the current function, allowing it to be promoted to a simple SSA value 41instead of a memory location. 42 43.. _wellformed: 44 45Well-Formedness 46--------------- 47 48It is important to note that this document describes 'well formed' LLVM 49assembly language. There is a difference between what the parser accepts 50and what is considered 'well formed'. For example, the following 51instruction is syntactically okay, but not well formed: 52 53.. code-block:: llvm 54 55 %x = add i32 1, %x 56 57because the definition of ``%x`` does not dominate all of its uses. The 58LLVM infrastructure provides a verification pass that may be used to 59verify that an LLVM module is well formed. This pass is automatically 60run by the parser after parsing input assembly and by the optimizer 61before it outputs bitcode. The violations pointed out by the verifier 62pass indicate bugs in transformation passes or input to the parser. 63 64.. _identifiers: 65 66Identifiers 67=========== 68 69LLVM identifiers come in two basic types: global and local. Global 70identifiers (functions, global variables) begin with the ``'@'`` 71character. Local identifiers (register names, types) begin with the 72``'%'`` character. Additionally, there are three different formats for 73identifiers, for different purposes: 74 75#. Named values are represented as a string of characters with their 76 prefix. For example, ``%foo``, ``@DivisionByZero``, 77 ``%a.really.long.identifier``. The actual regular expression used is 78 '``[%@][-a-zA-Z$._][-a-zA-Z$._0-9]*``'. Identifiers that require other 79 characters in their names can be surrounded with quotes. Special 80 characters may be escaped using ``"\xx"`` where ``xx`` is the ASCII 81 code for the character in hexadecimal. In this way, any character can 82 be used in a name value, even quotes themselves. The ``"\01"`` prefix 83 can be used on global variables to suppress mangling. 84#. Unnamed values are represented as an unsigned numeric value with 85 their prefix. For example, ``%12``, ``@2``, ``%44``. 86#. Constants, which are described in the section Constants_ below. 87 88LLVM requires that values start with a prefix for two reasons: Compilers 89don't need to worry about name clashes with reserved words, and the set 90of reserved words may be expanded in the future without penalty. 91Additionally, unnamed identifiers allow a compiler to quickly come up 92with a temporary variable without having to avoid symbol table 93conflicts. 94 95Reserved words in LLVM are very similar to reserved words in other 96languages. There are keywords for different opcodes ('``add``', 97'``bitcast``', '``ret``', etc...), for primitive type names ('``void``', 98'``i32``', etc...), and others. These reserved words cannot conflict 99with variable names, because none of them start with a prefix character 100(``'%'`` or ``'@'``). 101 102Here is an example of LLVM code to multiply the integer variable 103'``%X``' by 8: 104 105The easy way: 106 107.. code-block:: llvm 108 109 %result = mul i32 %X, 8 110 111After strength reduction: 112 113.. code-block:: llvm 114 115 %result = shl i32 %X, 3 116 117And the hard way: 118 119.. code-block:: llvm 120 121 %0 = add i32 %X, %X ; yields i32:%0 122 %1 = add i32 %0, %0 ; yields i32:%1 123 %result = add i32 %1, %1 124 125This last way of multiplying ``%X`` by 8 illustrates several important 126lexical features of LLVM: 127 128#. Comments are delimited with a '``;``' and go until the end of line. 129#. Unnamed temporaries are created when the result of a computation is 130 not assigned to a named value. 131#. Unnamed temporaries are numbered sequentially (using a per-function 132 incrementing counter, starting with 0). Note that basic blocks and unnamed 133 function parameters are included in this numbering. For example, if the 134 entry basic block is not given a label name and all function parameters are 135 named, then it will get number 0. 136 137It also shows a convention that we follow in this document. When 138demonstrating instructions, we will follow an instruction with a comment 139that defines the type and name of value produced. 140 141High Level Structure 142==================== 143 144Module Structure 145---------------- 146 147LLVM programs are composed of ``Module``'s, each of which is a 148translation unit of the input programs. Each module consists of 149functions, global variables, and symbol table entries. Modules may be 150combined together with the LLVM linker, which merges function (and 151global variable) definitions, resolves forward declarations, and merges 152symbol table entries. Here is an example of the "hello world" module: 153 154.. code-block:: llvm 155 156 ; Declare the string constant as a global constant. 157 @.str = private unnamed_addr constant [13 x i8] c"hello world\0A\00" 158 159 ; External declaration of the puts function 160 declare i32 @puts(i8* nocapture) nounwind 161 162 ; Definition of main function 163 define i32 @main() { ; i32()* 164 ; Convert [13 x i8]* to i8 *... 165 %cast210 = getelementptr [13 x i8], [13 x i8]* @.str, i64 0, i64 0 166 167 ; Call puts function to write out the string to stdout. 168 call i32 @puts(i8* %cast210) 169 ret i32 0 170 } 171 172 ; Named metadata 173 !0 = !{i32 42, null, !"string"} 174 !foo = !{!0} 175 176This example is made up of a :ref:`global variable <globalvars>` named 177"``.str``", an external declaration of the "``puts``" function, a 178:ref:`function definition <functionstructure>` for "``main``" and 179:ref:`named metadata <namedmetadatastructure>` "``foo``". 180 181In general, a module is made up of a list of global values (where both 182functions and global variables are global values). Global values are 183represented by a pointer to a memory location (in this case, a pointer 184to an array of char, and a pointer to a function), and have one of the 185following :ref:`linkage types <linkage>`. 186 187.. _linkage: 188 189Linkage Types 190------------- 191 192All Global Variables and Functions have one of the following types of 193linkage: 194 195``private`` 196 Global values with "``private``" linkage are only directly 197 accessible by objects in the current module. In particular, linking 198 code into a module with an private global value may cause the 199 private to be renamed as necessary to avoid collisions. Because the 200 symbol is private to the module, all references can be updated. This 201 doesn't show up in any symbol table in the object file. 202``internal`` 203 Similar to private, but the value shows as a local symbol 204 (``STB_LOCAL`` in the case of ELF) in the object file. This 205 corresponds to the notion of the '``static``' keyword in C. 206``available_externally`` 207 Globals with "``available_externally``" linkage are never emitted 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, 214 not declarations. 215``linkonce`` 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. 228``weak`` 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" 232 in C source code. 233``common`` 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 241 common linkage. 242 243.. _linkage_appending: 244 245``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 251 .o files are linked. 252``extern_weak`` 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. 264``external`` 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. 268 269It is illegal for a function *declaration* to have any linkage type 270other than ``external`` or ``extern_weak``. 271 272.. _callingconv: 273 274Calling Conventions 275------------------- 276 277LLVM :ref:`functions <functionstructure>`, :ref:`calls <i_call>` and 278:ref:`invokes <i_invoke>` can all have an optional calling convention 279specified for the call. The calling convention of any pair of dynamic 280caller/callee must match, or the behavior of the program is undefined. 281The following calling conventions are supported by LLVM, and more may be 282added in the future: 283 284"``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 308 calls for inlining. 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 318 limitations: 319 320 - 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. 324 325 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 64 342 bit). 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. 364 365 - 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. 368 369 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. 380 381 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 387 future. 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. 398 399 - 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). 402 403 The idea behind this convention is to support calls to runtime functions 404 that don't need to call out to any other functions. 405 406 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. 413 414More calling conventions can be added/defined on an as-needed basis, to 415support Pascal conventions or any other well-known target-independent 416convention. 417 418.. _visibilitystyles: 419 420Visibility Styles 421----------------- 422 423All Global Variables and Functions have one of the following visibility 424styles: 425 426"``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. 444 445A symbol with ``internal`` or ``private`` linkage must have ``default`` 446visibility. 447 448.. _dllstorageclass: 449 450DLL Storage Classes 451------------------- 452 453All Global Variables, Functions and Aliases can have one of the following 454DLL storage class: 455 456``dllimport`` 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. 461``dllexport`` 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. 468 469.. _tls_model: 470 471Thread Local Storage Models 472--------------------------- 473 474A variable may be defined as ``thread_local``, which means that it will 475not be shared by threads (each thread will have a separated copy of the 476variable). Not all targets support thread-local variables. Optionally, a 477TLS model may be specified: 478 479``localdynamic`` 480 For variables that are only used within the current shared library. 481``initialexec`` 482 For variables in modules that will not be loaded dynamically. 483``localexec`` 484 For variables defined in the executable and only used within it. 485 486If no explicit model is given, the "general dynamic" model is used. 487 488The models correspond to the ELF TLS models; see `ELF Handling For 489Thread-Local Storage <http://people.redhat.com/drepper/tls.pdf>`_ for 490more information on under which circumstances the different models may 491be used. The target may choose a different TLS model if the specified 492model is not supported, or if a better choice of model can be made. 493 494A model can also be specified in a alias, but then it only governs how 495the alias is accessed. It will not have any effect in the aliasee. 496 497.. _namedtypes: 498 499Structure Types 500--------------- 501 502LLVM IR allows you to specify both "identified" and "literal" :ref:`structure 503types <t_struct>`. Literal types are uniqued structurally, but identified types 504are never uniqued. An :ref:`opaque structural type <t_opaque>` can also be used 505to forward declare a type that is not yet available. 506 507An example of a identified structure specification is: 508 509.. code-block:: llvm 510 511 %mytype = type { %mytype*, i32 } 512 513Prior to the LLVM 3.0 release, identified types were structurally uniqued. Only 514literal types are uniqued in recent versions of LLVM. 515 516.. _globalvars: 517 518Global Variables 519---------------- 520 521Global variables define regions of memory allocated at compilation time 522instead of run-time. 523 524Global variable definitions must be initialized. 525 526Global variables in other translation units can also be declared, in which 527case they don't have an initializer. 528 529Either global variable definitions or declarations may have an explicit section 530to be placed in and may have an optional explicit alignment specified. 531 532A variable may be defined as a global ``constant``, which indicates that 533the contents of the variable will **never** be modified (enabling better 534optimization, allowing the global data to be placed in the read-only 535section of an executable, etc). Note that variables that need runtime 536initialization cannot be marked ``constant`` as there is a store to the 537variable. 538 539LLVM explicitly allows *declarations* of global variables to be marked 540constant, even if the final definition of the global is not. This 541capability can be used to enable slightly better optimization of the 542program, but requires the language definition to guarantee that 543optimizations based on the 'constantness' are valid for the translation 544units that do not include the definition. 545 546As SSA values, global variables define pointer values that are in scope 547(i.e. they dominate) all basic blocks in the program. Global variables 548always define a pointer to their "content" type because they describe a 549region of memory, and all memory objects in LLVM are accessed through 550pointers. 551 552Global variables can be marked with ``unnamed_addr`` which indicates 553that the address is not significant, only the content. Constants marked 554like this can be merged with other constants if they have the same 555initializer. Note that a constant with significant address *can* be 556merged with a ``unnamed_addr`` constant, the result being a constant 557whose address is significant. 558 559A global variable may be declared to reside in a target-specific 560numbered address space. For targets that support them, address spaces 561may affect how optimizations are performed and/or what target 562instructions are used to access the variable. The default address space 563is zero. The address space qualifier must precede any other attributes. 564 565LLVM allows an explicit section to be specified for globals. If the 566target supports it, it will emit globals to the section specified. 567Additionally, the global can placed in a comdat if the target has the necessary 568support. 569 570By default, global initializers are optimized by assuming that global 571variables defined within the module are not modified from their 572initial values before the start of the global initializer. This is 573true even for variables potentially accessible from outside the 574module, including those with external linkage or appearing in 575``@llvm.used`` or dllexported variables. This assumption may be suppressed 576by marking the variable with ``externally_initialized``. 577 578An explicit alignment may be specified for a global, which must be a 579power of 2. If not present, or if the alignment is set to zero, the 580alignment of the global is set by the target to whatever it feels 581convenient. If an explicit alignment is specified, the global is forced 582to have exactly that alignment. Targets and optimizers are not allowed 583to over-align the global if the global has an assigned section. In this 584case, the extra alignment could be observable: for example, code could 585assume that the globals are densely packed in their section and try to 586iterate over them as an array, alignment padding would break this 587iteration. The maximum alignment is ``1 << 29``. 588 589Globals can also have a :ref:`DLL storage class <dllstorageclass>`. 590 591Variables and aliases can have a 592:ref:`Thread Local Storage Model <tls_model>`. 593 594Syntax:: 595 596 [@<GlobalVarName> =] [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal] 597 [unnamed_addr] [AddrSpace] [ExternallyInitialized] 598 <global | constant> <Type> [<InitializerConstant>] 599 [, section "name"] [, comdat [($name)]] 600 [, align <Alignment>] 601 602For example, the following defines a global in a numbered address space 603with an initializer, section, and alignment: 604 605.. code-block:: llvm 606 607 @G = addrspace(5) constant float 1.0, section "foo", align 4 608 609The following example just declares a global variable 610 611.. code-block:: llvm 612 613 @G = external global i32 614 615The following example defines a thread-local global with the 616``initialexec`` TLS model: 617 618.. code-block:: llvm 619 620 @G = thread_local(initialexec) global i32 0, align 4 621 622.. _functionstructure: 623 624Functions 625--------- 626 627LLVM function definitions consist of the "``define``" keyword, an 628optional :ref:`linkage type <linkage>`, an optional :ref:`visibility 629style <visibility>`, an optional :ref:`DLL storage class <dllstorageclass>`, 630an optional :ref:`calling convention <callingconv>`, 631an optional ``unnamed_addr`` attribute, a return type, an optional 632:ref:`parameter attribute <paramattrs>` for the return type, a function 633name, a (possibly empty) argument list (each with optional :ref:`parameter 634attributes <paramattrs>`), optional :ref:`function attributes <fnattrs>`, 635an optional section, an optional alignment, 636an optional :ref:`comdat <langref_comdats>`, 637an optional :ref:`garbage collector name <gc>`, an optional :ref:`prefix <prefixdata>`, 638an optional :ref:`prologue <prologuedata>`, an opening 639curly brace, a list of basic blocks, and a closing curly brace. 640 641LLVM function declarations consist of the "``declare``" keyword, an 642optional :ref:`linkage type <linkage>`, an optional :ref:`visibility 643style <visibility>`, an optional :ref:`DLL storage class <dllstorageclass>`, 644an optional :ref:`calling convention <callingconv>`, 645an optional ``unnamed_addr`` attribute, a return type, an optional 646:ref:`parameter attribute <paramattrs>` for the return type, a function 647name, a possibly empty list of arguments, an optional alignment, an optional 648:ref:`garbage collector name <gc>`, an optional :ref:`prefix <prefixdata>`, 649and an optional :ref:`prologue <prologuedata>`. 650 651A function definition contains a list of basic blocks, forming the CFG (Control 652Flow Graph) for the function. Each basic block may optionally start with a label 653(giving the basic block a symbol table entry), contains a list of instructions, 654and ends with a :ref:`terminator <terminators>` instruction (such as a branch or 655function return). If an explicit label is not provided, a block is assigned an 656implicit numbered label, using the next value from the same counter as used for 657unnamed temporaries (:ref:`see above<identifiers>`). For example, if a function 658entry block does not have an explicit label, it will be assigned label "%0", 659then the first unnamed temporary in that block will be "%1", etc. 660 661The first basic block in a function is special in two ways: it is 662immediately executed on entrance to the function, and it is not allowed 663to have predecessor basic blocks (i.e. there can not be any branches to 664the entry block of a function). Because the block can have no 665predecessors, it also cannot have any :ref:`PHI nodes <i_phi>`. 666 667LLVM allows an explicit section to be specified for functions. If the 668target supports it, it will emit functions to the section specified. 669Additionally, the function can be placed in a COMDAT. 670 671An explicit alignment may be specified for a function. If not present, 672or if the alignment is set to zero, the alignment of the function is set 673by the target to whatever it feels convenient. If an explicit alignment 674is specified, the function is forced to have at least that much 675alignment. All alignments must be a power of 2. 676 677If the ``unnamed_addr`` attribute is given, the address is known to not 678be significant and two identical functions can be merged. 679 680Syntax:: 681 682 define [linkage] [visibility] [DLLStorageClass] 683 [cconv] [ret attrs] 684 <ResultType> @<FunctionName> ([argument list]) 685 [unnamed_addr] [fn Attrs] [section "name"] [comdat [($name)]] 686 [align N] [gc] [prefix Constant] [prologue Constant] { ... } 687 688The argument list is a comma seperated sequence of arguments where each 689argument is of the following form 690 691Syntax:: 692 693 <type> [parameter Attrs] [name] 694 695 696.. _langref_aliases: 697 698Aliases 699------- 700 701Aliases, unlike function or variables, don't create any new data. They 702are just a new symbol and metadata for an existing position. 703 704Aliases have a name and an aliasee that is either a global value or a 705constant expression. 706 707Aliases may have an optional :ref:`linkage type <linkage>`, an optional 708:ref:`visibility style <visibility>`, an optional :ref:`DLL storage class 709<dllstorageclass>` and an optional :ref:`tls model <tls_model>`. 710 711Syntax:: 712 713 @<Name> = [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal] [unnamed_addr] alias <AliaseeTy> @<Aliasee> 714 715The linkage must be one of ``private``, ``internal``, ``linkonce``, ``weak``, 716``linkonce_odr``, ``weak_odr``, ``external``. Note that some system linkers 717might not correctly handle dropping a weak symbol that is aliased. 718 719Aliases that are not ``unnamed_addr`` are guaranteed to have the same address as 720the aliasee expression. ``unnamed_addr`` ones are only guaranteed to point 721to the same content. 722 723Since aliases are only a second name, some restrictions apply, of which 724some can only be checked when producing an object file: 725 726* The expression defining the aliasee must be computable at assembly 727 time. Since it is just a name, no relocations can be used. 728 729* No alias in the expression can be weak as the possibility of the 730 intermediate alias being overridden cannot be represented in an 731 object file. 732 733* No global value in the expression can be a declaration, since that 734 would require a relocation, which is not possible. 735 736.. _langref_comdats: 737 738Comdats 739------- 740 741Comdat IR provides access to COFF and ELF object file COMDAT functionality. 742 743Comdats have a name which represents the COMDAT key. All global objects that 744specify this key will only end up in the final object file if the linker chooses 745that key over some other key. Aliases are placed in the same COMDAT that their 746aliasee computes to, if any. 747 748Comdats have a selection kind to provide input on how the linker should 749choose between keys in two different object files. 750 751Syntax:: 752 753 $<Name> = comdat SelectionKind 754 755The selection kind must be one of the following: 756 757``any`` 758 The linker may choose any COMDAT key, the choice is arbitrary. 759``exactmatch`` 760 The linker may choose any COMDAT key but the sections must contain the 761 same data. 762``largest`` 763 The linker will choose the section containing the largest COMDAT key. 764``noduplicates`` 765 The linker requires that only section with this COMDAT key exist. 766``samesize`` 767 The linker may choose any COMDAT key but the sections must contain the 768 same amount of data. 769 770Note that the Mach-O platform doesn't support COMDATs and ELF only supports 771``any`` as a selection kind. 772 773Here is an example of a COMDAT group where a function will only be selected if 774the COMDAT key's section is the largest: 775 776.. code-block:: llvm 777 778 $foo = comdat largest 779 @foo = global i32 2, comdat($foo) 780 781 define void @bar() comdat($foo) { 782 ret void 783 } 784 785As a syntactic sugar the ``$name`` can be omitted if the name is the same as 786the global name: 787 788.. code-block:: llvm 789 790 $foo = comdat any 791 @foo = global i32 2, comdat 792 793 794In a COFF object file, this will create a COMDAT section with selection kind 795``IMAGE_COMDAT_SELECT_LARGEST`` containing the contents of the ``@foo`` symbol 796and another COMDAT section with selection kind 797``IMAGE_COMDAT_SELECT_ASSOCIATIVE`` which is associated with the first COMDAT 798section and contains the contents of the ``@bar`` symbol. 799 800There are some restrictions on the properties of the global object. 801It, or an alias to it, must have the same name as the COMDAT group when 802targeting COFF. 803The contents and size of this object may be used during link-time to determine 804which COMDAT groups get selected depending on the selection kind. 805Because the name of the object must match the name of the COMDAT group, the 806linkage of the global object must not be local; local symbols can get renamed 807if a collision occurs in the symbol table. 808 809The combined use of COMDATS and section attributes may yield surprising results. 810For example: 811 812.. code-block:: llvm 813 814 $foo = comdat any 815 $bar = comdat any 816 @g1 = global i32 42, section "sec", comdat($foo) 817 @g2 = global i32 42, section "sec", comdat($bar) 818 819From the object file perspective, this requires the creation of two sections 820with the same name. This is necessary because both globals belong to different 821COMDAT groups and COMDATs, at the object file level, are represented by 822sections. 823 824Note that certain IR constructs like global variables and functions may create 825COMDATs in the object file in addition to any which are specified using COMDAT 826IR. This arises, for example, when a global variable has linkonce_odr linkage. 827 828.. _namedmetadatastructure: 829 830Named Metadata 831-------------- 832 833Named metadata is a collection of metadata. :ref:`Metadata 834nodes <metadata>` (but not metadata strings) are the only valid 835operands for a named metadata. 836 837Syntax:: 838 839 ; Some unnamed metadata nodes, which are referenced by the named metadata. 840 !0 = !{!"zero"} 841 !1 = !{!"one"} 842 !2 = !{!"two"} 843 ; A named metadata. 844 !name = !{!0, !1, !2} 845 846.. _paramattrs: 847 848Parameter Attributes 849-------------------- 850 851The return type and each parameter of a function type may have a set of 852*parameter attributes* associated with them. Parameter attributes are 853used to communicate additional information about the result or 854parameters of a function. Parameter attributes are considered to be part 855of the function, not of the function type, so functions with different 856parameter attributes can have the same function type. 857 858Parameter attributes are simple keywords that follow the type specified. 859If multiple parameter attributes are needed, they are space separated. 860For example: 861 862.. code-block:: llvm 863 864 declare i32 @printf(i8* noalias nocapture, ...) 865 declare i32 @atoi(i8 zeroext) 866 declare signext i8 @returns_signed_char() 867 868Note that any attributes for the function result (``nounwind``, 869``readonly``) come immediately after the argument list. 870 871Currently, only the following parameter attributes are defined: 872 873``zeroext`` 874 This indicates to the code generator that the parameter or return 875 value should be zero-extended to the extent required by the target's 876 ABI (which is usually 32-bits, but is 8-bits for a i1 on x86-64) by 877 the caller (for a parameter) or the callee (for a return value). 878``signext`` 879 This indicates to the code generator that the parameter or return 880 value should be sign-extended to the extent required by the target's 881 ABI (which is usually 32-bits) by the caller (for a parameter) or 882 the callee (for a return value). 883``inreg`` 884 This indicates that this parameter or return value should be treated 885 in a special target-dependent fashion during while emitting code for 886 a function call or return (usually, by putting it in a register as 887 opposed to memory, though some targets use it to distinguish between 888 two different kinds of registers). Use of this attribute is 889 target-specific. 890``byval`` 891 This indicates that the pointer parameter should really be passed by 892 value to the function. The attribute implies that a hidden copy of 893 the pointee is made between the caller and the callee, so the callee 894 is unable to modify the value in the caller. This attribute is only 895 valid on LLVM pointer arguments. It is generally used to pass 896 structs and arrays by value, but is also valid on pointers to 897 scalars. The copy is considered to belong to the caller not the 898 callee (for example, ``readonly`` functions should not write to 899 ``byval`` parameters). This is not a valid attribute for return 900 values. 901 902 The byval attribute also supports specifying an alignment with the 903 align attribute. It indicates the alignment of the stack slot to 904 form and the known alignment of the pointer specified to the call 905 site. If the alignment is not specified, then the code generator 906 makes a target-specific assumption. 907 908.. _attr_inalloca: 909 910``inalloca`` 911 912 The ``inalloca`` argument attribute allows the caller to take the 913 address of outgoing stack arguments. An ``inalloca`` argument must 914 be a pointer to stack memory produced by an ``alloca`` instruction. 915 The alloca, or argument allocation, must also be tagged with the 916 inalloca keyword. Only the last argument may have the ``inalloca`` 917 attribute, and that argument is guaranteed to be passed in memory. 918 919 An argument allocation may be used by a call at most once because 920 the call may deallocate it. The ``inalloca`` attribute cannot be 921 used in conjunction with other attributes that affect argument 922 storage, like ``inreg``, ``nest``, ``sret``, or ``byval``. The 923 ``inalloca`` attribute also disables LLVM's implicit lowering of 924 large aggregate return values, which means that frontend authors 925 must lower them with ``sret`` pointers. 926 927 When the call site is reached, the argument allocation must have 928 been the most recent stack allocation that is still live, or the 929 results are undefined. It is possible to allocate additional stack 930 space after an argument allocation and before its call site, but it 931 must be cleared off with :ref:`llvm.stackrestore 932 <int_stackrestore>`. 933 934 See :doc:`InAlloca` for more information on how to use this 935 attribute. 936 937``sret`` 938 This indicates that the pointer parameter specifies the address of a 939 structure that is the return value of the function in the source 940 program. This pointer must be guaranteed by the caller to be valid: 941 loads and stores to the structure may be assumed by the callee 942 not to trap and to be properly aligned. This may only be applied to 943 the first parameter. This is not a valid attribute for return 944 values. 945 946``align <n>`` 947 This indicates that the pointer value may be assumed by the optimizer to 948 have the specified alignment. 949 950 Note that this attribute has additional semantics when combined with the 951 ``byval`` attribute. 952 953.. _noalias: 954 955``noalias`` 956 This indicates that objects accessed via pointer values 957 :ref:`based <pointeraliasing>` on the argument or return value are not also 958 accessed, during the execution of the function, via pointer values not 959 *based* on the argument or return value. The attribute on a return value 960 also has additional semantics described below. The caller shares the 961 responsibility with the callee for ensuring that these requirements are met. 962 For further details, please see the discussion of the NoAlias response in 963 :ref:`alias analysis <Must, May, or No>`. 964 965 Note that this definition of ``noalias`` is intentionally similar 966 to the definition of ``restrict`` in C99 for function arguments. 967 968 For function return values, C99's ``restrict`` is not meaningful, 969 while LLVM's ``noalias`` is. Furthermore, the semantics of the ``noalias`` 970 attribute on return values are stronger than the semantics of the attribute 971 when used on function arguments. On function return values, the ``noalias`` 972 attribute indicates that the function acts like a system memory allocation 973 function, returning a pointer to allocated storage disjoint from the 974 storage for any other object accessible to the caller. 975 976``nocapture`` 977 This indicates that the callee does not make any copies of the 978 pointer that outlive the callee itself. This is not a valid 979 attribute for return values. 980 981.. _nest: 982 983``nest`` 984 This indicates that the pointer parameter can be excised using the 985 :ref:`trampoline intrinsics <int_trampoline>`. This is not a valid 986 attribute for return values and can only be applied to one parameter. 987 988``returned`` 989 This indicates that the function always returns the argument as its return 990 value. This is an optimization hint to the code generator when generating 991 the caller, allowing tail call optimization and omission of register saves 992 and restores in some cases; it is not checked or enforced when generating 993 the callee. The parameter and the function return type must be valid 994 operands for the :ref:`bitcast instruction <i_bitcast>`. This is not a 995 valid attribute for return values and can only be applied to one parameter. 996 997``nonnull`` 998 This indicates that the parameter or return pointer is not null. This 999 attribute may only be applied to pointer typed parameters. This is not 1000 checked or enforced by LLVM, the caller must ensure that the pointer 1001 passed in is non-null, or the callee must ensure that the returned pointer 1002 is non-null. 1003 1004``dereferenceable(<n>)`` 1005 This indicates that the parameter or return pointer is dereferenceable. This 1006 attribute may only be applied to pointer typed parameters. A pointer that 1007 is dereferenceable can be loaded from speculatively without a risk of 1008 trapping. The number of bytes known to be dereferenceable must be provided 1009 in parentheses. It is legal for the number of bytes to be less than the 1010 size of the pointee type. The ``nonnull`` attribute does not imply 1011 dereferenceability (consider a pointer to one element past the end of an 1012 array), however ``dereferenceable(<n>)`` does imply ``nonnull`` in 1013 ``addrspace(0)`` (which is the default address space). 1014 1015``dereferenceable_or_null(<n>)`` 1016 This indicates that the parameter or return value isn't both 1017 non-null and non-dereferenceable (up to ``<n>`` bytes) at the same 1018 time. All non-null pointers tagged with 1019 ``dereferenceable_or_null(<n>)`` are ``dereferenceable(<n>)``. 1020 For address space 0 ``dereferenceable_or_null(<n>)`` implies that 1021 a pointer is exactly one of ``dereferenceable(<n>)`` or ``null``, 1022 and in other address spaces ``dereferenceable_or_null(<n>)`` 1023 implies that a pointer is at least one of ``dereferenceable(<n>)`` 1024 or ``null`` (i.e. it may be both ``null`` and 1025 ``dereferenceable(<n>)``). This attribute may only be applied to 1026 pointer typed parameters. 1027 1028.. _gc: 1029 1030Garbage Collector Strategy Names 1031-------------------------------- 1032 1033Each function may specify a garbage collector strategy name, which is simply a 1034string: 1035 1036.. code-block:: llvm 1037 1038 define void @f() gc "name" { ... } 1039 1040The supported values of *name* includes those :ref:`built in to LLVM 1041<builtin-gc-strategies>` and any provided by loaded plugins. Specifying a GC 1042strategy will cause the compiler to alter its output in order to support the 1043named garbage collection algorithm. Note that LLVM itself does not contain a 1044garbage collector, this functionality is restricted to generating machine code 1045which can interoperate with a collector provided externally. 1046 1047.. _prefixdata: 1048 1049Prefix Data 1050----------- 1051 1052Prefix data is data associated with a function which the code 1053generator will emit immediately before the function's entrypoint. 1054The purpose of this feature is to allow frontends to associate 1055language-specific runtime metadata with specific functions and make it 1056available through the function pointer while still allowing the 1057function pointer to be called. 1058 1059To access the data for a given function, a program may bitcast the 1060function pointer to a pointer to the constant's type and dereference 1061index -1. This implies that the IR symbol points just past the end of 1062the prefix data. For instance, take the example of a function annotated 1063with a single ``i32``, 1064 1065.. code-block:: llvm 1066 1067 define void @f() prefix i32 123 { ... } 1068 1069The prefix data can be referenced as, 1070 1071.. code-block:: llvm 1072 1073 %0 = bitcast void* () @f to i32* 1074 %a = getelementptr inbounds i32, i32* %0, i32 -1 1075 %b = load i32, i32* %a 1076 1077Prefix data is laid out as if it were an initializer for a global variable 1078of the prefix data's type. The function will be placed such that the 1079beginning of the prefix data is aligned. This means that if the size 1080of the prefix data is not a multiple of the alignment size, the 1081function's entrypoint will not be aligned. If alignment of the 1082function's entrypoint is desired, padding must be added to the prefix 1083data. 1084 1085A function may have prefix data but no body. This has similar semantics 1086to the ``available_externally`` linkage in that the data may be used by the 1087optimizers but will not be emitted in the object file. 1088 1089.. _prologuedata: 1090 1091Prologue Data 1092------------- 1093 1094The ``prologue`` attribute allows arbitrary code (encoded as bytes) to 1095be inserted prior to the function body. This can be used for enabling 1096function hot-patching and instrumentation. 1097 1098To maintain the semantics of ordinary function calls, the prologue data must 1099have a particular format. Specifically, it must begin with a sequence of 1100bytes which decode to a sequence of machine instructions, valid for the 1101module's target, which transfer control to the point immediately succeeding 1102the prologue data, without performing any other visible action. This allows 1103the inliner and other passes to reason about the semantics of the function 1104definition without needing to reason about the prologue data. Obviously this 1105makes the format of the prologue data highly target dependent. 1106 1107A trivial example of valid prologue data for the x86 architecture is ``i8 144``, 1108which encodes the ``nop`` instruction: 1109 1110.. code-block:: llvm 1111 1112 define void @f() prologue i8 144 { ... } 1113 1114Generally prologue data can be formed by encoding a relative branch instruction 1115which skips the metadata, as in this example of valid prologue data for the 1116x86_64 architecture, where the first two bytes encode ``jmp .+10``: 1117 1118.. code-block:: llvm 1119 1120 %0 = type <{ i8, i8, i8* }> 1121 1122 define void @f() prologue %0 <{ i8 235, i8 8, i8* @md}> { ... } 1123 1124A function may have prologue data but no body. This has similar semantics 1125to the ``available_externally`` linkage in that the data may be used by the 1126optimizers but will not be emitted in the object file. 1127 1128.. _attrgrp: 1129 1130Attribute Groups 1131---------------- 1132 1133Attribute groups are groups of attributes that are referenced by objects within 1134the IR. They are important for keeping ``.ll`` files readable, because a lot of 1135functions will use the same set of attributes. In the degenerative case of a 1136``.ll`` file that corresponds to a single ``.c`` file, the single attribute 1137group will capture the important command line flags used to build that file. 1138 1139An attribute group is a module-level object. To use an attribute group, an 1140object references the attribute group's ID (e.g. ``#37``). An object may refer 1141to more than one attribute group. In that situation, the attributes from the 1142different groups are merged. 1143 1144Here is an example of attribute groups for a function that should always be 1145inlined, has a stack alignment of 4, and which shouldn't use SSE instructions: 1146 1147.. code-block:: llvm 1148 1149 ; Target-independent attributes: 1150 attributes #0 = { alwaysinline alignstack=4 } 1151 1152 ; Target-dependent attributes: 1153 attributes #1 = { "no-sse" } 1154 1155 ; Function @f has attributes: alwaysinline, alignstack=4, and "no-sse". 1156 define void @f() #0 #1 { ... } 1157 1158.. _fnattrs: 1159 1160Function Attributes 1161------------------- 1162 1163Function attributes are set to communicate additional information about 1164a function. Function attributes are considered to be part of the 1165function, not of the function type, so functions with different function 1166attributes can have the same function type. 1167 1168Function attributes are simple keywords that follow the type specified. 1169If multiple attributes are needed, they are space separated. For 1170example: 1171 1172.. code-block:: llvm 1173 1174 define void @f() noinline { ... } 1175 define void @f() alwaysinline { ... } 1176 define void @f() alwaysinline optsize { ... } 1177 define void @f() optsize { ... } 1178 1179``alignstack(<n>)`` 1180 This attribute indicates that, when emitting the prologue and 1181 epilogue, the backend should forcibly align the stack pointer. 1182 Specify the desired alignment, which must be a power of two, in 1183 parentheses. 1184``alwaysinline`` 1185 This attribute indicates that the inliner should attempt to inline 1186 this function into callers whenever possible, ignoring any active 1187 inlining size threshold for this caller. 1188``builtin`` 1189 This indicates that the callee function at a call site should be 1190 recognized as a built-in function, even though the function's declaration 1191 uses the ``nobuiltin`` attribute. This is only valid at call sites for 1192 direct calls to functions that are declared with the ``nobuiltin`` 1193 attribute. 1194``cold`` 1195 This attribute indicates that this function is rarely called. When 1196 computing edge weights, basic blocks post-dominated by a cold 1197 function call are also considered to be cold; and, thus, given low 1198 weight. 1199``inlinehint`` 1200 This attribute indicates that the source code contained a hint that 1201 inlining this function is desirable (such as the "inline" keyword in 1202 C/C++). It is just a hint; it imposes no requirements on the 1203 inliner. 1204``jumptable`` 1205 This attribute indicates that the function should be added to a 1206 jump-instruction table at code-generation time, and that all address-taken 1207 references to this function should be replaced with a reference to the 1208 appropriate jump-instruction-table function pointer. Note that this creates 1209 a new pointer for the original function, which means that code that depends 1210 on function-pointer identity can break. So, any function annotated with 1211 ``jumptable`` must also be ``unnamed_addr``. 1212``minsize`` 1213 This attribute suggests that optimization passes and code generator 1214 passes make choices that keep the code size of this function as small 1215 as possible and perform optimizations that may sacrifice runtime 1216 performance in order to minimize the size of the generated code. 1217``naked`` 1218 This attribute disables prologue / epilogue emission for the 1219 function. This can have very system-specific consequences. 1220``nobuiltin`` 1221 This indicates that the callee function at a call site is not recognized as 1222 a built-in function. LLVM will retain the original call and not replace it 1223 with equivalent code based on the semantics of the built-in function, unless 1224 the call site uses the ``builtin`` attribute. This is valid at call sites 1225 and on function declarations and definitions. 1226``noduplicate`` 1227 This attribute indicates that calls to the function cannot be 1228 duplicated. A call to a ``noduplicate`` function may be moved 1229 within its parent function, but may not be duplicated within 1230 its parent function. 1231 1232 A function containing a ``noduplicate`` call may still 1233 be an inlining candidate, provided that the call is not 1234 duplicated by inlining. That implies that the function has 1235 internal linkage and only has one call site, so the original 1236 call is dead after inlining. 1237``noimplicitfloat`` 1238 This attributes disables implicit floating point instructions. 1239``noinline`` 1240 This attribute indicates that the inliner should never inline this 1241 function in any situation. This attribute may not be used together 1242 with the ``alwaysinline`` attribute. 1243``nonlazybind`` 1244 This attribute suppresses lazy symbol binding for the function. This 1245 may make calls to the function faster, at the cost of extra program 1246 startup time if the function is not called during program startup. 1247``noredzone`` 1248 This attribute indicates that the code generator should not use a 1249 red zone, even if the target-specific ABI normally permits it. 1250``noreturn`` 1251 This function attribute indicates that the function never returns 1252 normally. This produces undefined behavior at runtime if the 1253 function ever does dynamically return. 1254``nounwind`` 1255 This function attribute indicates that the function never raises an 1256 exception. If the function does raise an exception, its runtime 1257 behavior is undefined. However, functions marked nounwind may still 1258 trap or generate asynchronous exceptions. Exception handling schemes 1259 that are recognized by LLVM to handle asynchronous exceptions, such 1260 as SEH, will still provide their implementation defined semantics. 1261``optnone`` 1262 This function attribute indicates that the function is not optimized 1263 by any optimization or code generator passes with the 1264 exception of interprocedural optimization passes. 1265 This attribute cannot be used together with the ``alwaysinline`` 1266 attribute; this attribute is also incompatible 1267 with the ``minsize`` attribute and the ``optsize`` attribute. 1268 1269 This attribute requires the ``noinline`` attribute to be specified on 1270 the function as well, so the function is never inlined into any caller. 1271 Only functions with the ``alwaysinline`` attribute are valid 1272 candidates for inlining into the body of this function. 1273``optsize`` 1274 This attribute suggests that optimization passes and code generator 1275 passes make choices that keep the code size of this function low, 1276 and otherwise do optimizations specifically to reduce code size as 1277 long as they do not significantly impact runtime performance. 1278``readnone`` 1279 On a function, this attribute indicates that the function computes its 1280 result (or decides to unwind an exception) based strictly on its arguments, 1281 without dereferencing any pointer arguments or otherwise accessing 1282 any mutable state (e.g. memory, control registers, etc) visible to 1283 caller functions. It does not write through any pointer arguments 1284 (including ``byval`` arguments) and never changes any state visible 1285 to callers. This means that it cannot unwind exceptions by calling 1286 the ``C++`` exception throwing methods. 1287 1288 On an argument, this attribute indicates that the function does not 1289 dereference that pointer argument, even though it may read or write the 1290 memory that the pointer points to if accessed through other pointers. 1291``readonly`` 1292 On a function, this attribute indicates that the function does not write 1293 through any pointer arguments (including ``byval`` arguments) or otherwise 1294 modify any state (e.g. memory, control registers, etc) visible to 1295 caller functions. It may dereference pointer arguments and read 1296 state that may be set in the caller. A readonly function always 1297 returns the same value (or unwinds an exception identically) when 1298 called with the same set of arguments and global state. It cannot 1299 unwind an exception by calling the ``C++`` exception throwing 1300 methods. 1301 1302 On an argument, this attribute indicates that the function does not write 1303 through this pointer argument, even though it may write to the memory that 1304 the pointer points to. 1305``returns_twice`` 1306 This attribute indicates that this function can return twice. The C 1307 ``setjmp`` is an example of such a function. The compiler disables 1308 some optimizations (like tail calls) in the caller of these 1309 functions. 1310``sanitize_address`` 1311 This attribute indicates that AddressSanitizer checks 1312 (dynamic address safety analysis) are enabled for this function. 1313``sanitize_memory`` 1314 This attribute indicates that MemorySanitizer checks (dynamic detection 1315 of accesses to uninitialized memory) are enabled for this function. 1316``sanitize_thread`` 1317 This attribute indicates that ThreadSanitizer checks 1318 (dynamic thread safety analysis) are enabled for this function. 1319``ssp`` 1320 This attribute indicates that the function should emit a stack 1321 smashing protector. It is in the form of a "canary" --- a random value 1322 placed on the stack before the local variables that's checked upon 1323 return from the function to see if it has been overwritten. A 1324 heuristic is used to determine if a function needs stack protectors 1325 or not. The heuristic used will enable protectors for functions with: 1326 1327 - Character arrays larger than ``ssp-buffer-size`` (default 8). 1328 - Aggregates containing character arrays larger than ``ssp-buffer-size``. 1329 - Calls to alloca() with variable sizes or constant sizes greater than 1330 ``ssp-buffer-size``. 1331 1332 Variables that are identified as requiring a protector will be arranged 1333 on the stack such that they are adjacent to the stack protector guard. 1334 1335 If a function that has an ``ssp`` attribute is inlined into a 1336 function that doesn't have an ``ssp`` attribute, then the resulting 1337 function will have an ``ssp`` attribute. 1338``sspreq`` 1339 This attribute indicates that the function should *always* emit a 1340 stack smashing protector. This overrides the ``ssp`` function 1341 attribute. 1342 1343 Variables that are identified as requiring a protector will be arranged 1344 on the stack such that they are adjacent to the stack protector guard. 1345 The specific layout rules are: 1346 1347 #. Large arrays and structures containing large arrays 1348 (``>= ssp-buffer-size``) are closest to the stack protector. 1349 #. Small arrays and structures containing small arrays 1350 (``< ssp-buffer-size``) are 2nd closest to the protector. 1351 #. Variables that have had their address taken are 3rd closest to the 1352 protector. 1353 1354 If a function that has an ``sspreq`` attribute is inlined into a 1355 function that doesn't have an ``sspreq`` attribute or which has an 1356 ``ssp`` or ``sspstrong`` attribute, then the resulting function will have 1357 an ``sspreq`` attribute. 1358``sspstrong`` 1359 This attribute indicates that the function should emit a stack smashing 1360 protector. This attribute causes a strong heuristic to be used when 1361 determining if a function needs stack protectors. The strong heuristic 1362 will enable protectors for functions with: 1363 1364 - Arrays of any size and type 1365 - Aggregates containing an array of any size and type. 1366 - Calls to alloca(). 1367 - Local variables that have had their address taken. 1368 1369 Variables that are identified as requiring a protector will be arranged 1370 on the stack such that they are adjacent to the stack protector guard. 1371 The specific layout rules are: 1372 1373 #. Large arrays and structures containing large arrays 1374 (``>= ssp-buffer-size``) are closest to the stack protector. 1375 #. Small arrays and structures containing small arrays 1376 (``< ssp-buffer-size``) are 2nd closest to the protector. 1377 #. Variables that have had their address taken are 3rd closest to the 1378 protector. 1379 1380 This overrides the ``ssp`` function attribute. 1381 1382 If a function that has an ``sspstrong`` attribute is inlined into a 1383 function that doesn't have an ``sspstrong`` attribute, then the 1384 resulting function will have an ``sspstrong`` attribute. 1385``"thunk"`` 1386 This attribute indicates that the function will delegate to some other 1387 function with a tail call. The prototype of a thunk should not be used for 1388 optimization purposes. The caller is expected to cast the thunk prototype to 1389 match the thunk target prototype. 1390``uwtable`` 1391 This attribute indicates that the ABI being targeted requires that 1392 an unwind table entry be produce for this function even if we can 1393 show that no exceptions passes by it. This is normally the case for 1394 the ELF x86-64 abi, but it can be disabled for some compilation 1395 units. 1396 1397.. _moduleasm: 1398 1399Module-Level Inline Assembly 1400---------------------------- 1401 1402Modules may contain "module-level inline asm" blocks, which corresponds 1403to the GCC "file scope inline asm" blocks. These blocks are internally 1404concatenated by LLVM and treated as a single unit, but may be separated 1405in the ``.ll`` file if desired. The syntax is very simple: 1406 1407.. code-block:: llvm 1408 1409 module asm "inline asm code goes here" 1410 module asm "more can go here" 1411 1412The strings can contain any character by escaping non-printable 1413characters. The escape sequence used is simply "\\xx" where "xx" is the 1414two digit hex code for the number. 1415 1416The inline asm code is simply printed to the machine code .s file when 1417assembly code is generated. 1418 1419.. _langref_datalayout: 1420 1421Data Layout 1422----------- 1423 1424A module may specify a target specific data layout string that specifies 1425how data is to be laid out in memory. The syntax for the data layout is 1426simply: 1427 1428.. code-block:: llvm 1429 1430 target datalayout = "layout specification" 1431 1432The *layout specification* consists of a list of specifications 1433separated by the minus sign character ('-'). Each specification starts 1434with a letter and may include other information after the letter to 1435define some aspect of the data layout. The specifications accepted are 1436as follows: 1437 1438``E`` 1439 Specifies that the target lays out data in big-endian form. That is, 1440 the bits with the most significance have the lowest address 1441 location. 1442``e`` 1443 Specifies that the target lays out data in little-endian form. That 1444 is, the bits with the least significance have the lowest address 1445 location. 1446``S<size>`` 1447 Specifies the natural alignment of the stack in bits. Alignment 1448 promotion of stack variables is limited to the natural stack 1449 alignment to avoid dynamic stack realignment. The stack alignment 1450 must be a multiple of 8-bits. If omitted, the natural stack 1451 alignment defaults to "unspecified", which does not prevent any 1452 alignment promotions. 1453``p[n]:<size>:<abi>:<pref>`` 1454 This specifies the *size* of a pointer and its ``<abi>`` and 1455 ``<pref>``\erred alignments for address space ``n``. All sizes are in 1456 bits. The address space, ``n`` is optional, and if not specified, 1457 denotes the default address space 0. The value of ``n`` must be 1458 in the range [1,2^23). 1459``i<size>:<abi>:<pref>`` 1460 This specifies the alignment for an integer type of a given bit 1461 ``<size>``. The value of ``<size>`` must be in the range [1,2^23). 1462``v<size>:<abi>:<pref>`` 1463 This specifies the alignment for a vector type of a given bit 1464 ``<size>``. 1465``f<size>:<abi>:<pref>`` 1466 This specifies the alignment for a floating point type of a given bit 1467 ``<size>``. Only values of ``<size>`` that are supported by the target 1468 will work. 32 (float) and 64 (double) are supported on all targets; 80 1469 or 128 (different flavors of long double) are also supported on some 1470 targets. 1471``a:<abi>:<pref>`` 1472 This specifies the alignment for an object of aggregate type. 1473``m:<mangling>`` 1474 If present, specifies that llvm names are mangled in the output. The 1475 options are 1476 1477 * ``e``: ELF mangling: Private symbols get a ``.L`` prefix. 1478 * ``m``: Mips mangling: Private symbols get a ``$`` prefix. 1479 * ``o``: Mach-O mangling: Private symbols get ``L`` prefix. Other 1480 symbols get a ``_`` prefix. 1481 * ``w``: Windows COFF prefix: Similar to Mach-O, but stdcall and fastcall 1482 functions also get a suffix based on the frame size. 1483``n<size1>:<size2>:<size3>...`` 1484 This specifies a set of native integer widths for the target CPU in 1485 bits. For example, it might contain ``n32`` for 32-bit PowerPC, 1486 ``n32:64`` for PowerPC 64, or ``n8:16:32:64`` for X86-64. Elements of 1487 this set are considered to support most general arithmetic operations 1488 efficiently. 1489 1490On every specification that takes a ``<abi>:<pref>``, specifying the 1491``<pref>`` alignment is optional. If omitted, the preceding ``:`` 1492should be omitted too and ``<pref>`` will be equal to ``<abi>``. 1493 1494When constructing the data layout for a given target, LLVM starts with a 1495default set of specifications which are then (possibly) overridden by 1496the specifications in the ``datalayout`` keyword. The default 1497specifications are given in this list: 1498 1499- ``E`` - big endian 1500- ``p:64:64:64`` - 64-bit pointers with 64-bit alignment. 1501- ``p[n]:64:64:64`` - Other address spaces are assumed to be the 1502 same as the default address space. 1503- ``S0`` - natural stack alignment is unspecified 1504- ``i1:8:8`` - i1 is 8-bit (byte) aligned 1505- ``i8:8:8`` - i8 is 8-bit (byte) aligned 1506- ``i16:16:16`` - i16 is 16-bit aligned 1507- ``i32:32:32`` - i32 is 32-bit aligned 1508- ``i64:32:64`` - i64 has ABI alignment of 32-bits but preferred 1509 alignment of 64-bits 1510- ``f16:16:16`` - half is 16-bit aligned 1511- ``f32:32:32`` - float is 32-bit aligned 1512- ``f64:64:64`` - double is 64-bit aligned 1513- ``f128:128:128`` - quad is 128-bit aligned 1514- ``v64:64:64`` - 64-bit vector is 64-bit aligned 1515- ``v128:128:128`` - 128-bit vector is 128-bit aligned 1516- ``a:0:64`` - aggregates are 64-bit aligned 1517 1518When LLVM is determining the alignment for a given type, it uses the 1519following rules: 1520 1521#. If the type sought is an exact match for one of the specifications, 1522 that specification is used. 1523#. If no match is found, and the type sought is an integer type, then 1524 the smallest integer type that is larger than the bitwidth of the 1525 sought type is used. If none of the specifications are larger than 1526 the bitwidth then the largest integer type is used. For example, 1527 given the default specifications above, the i7 type will use the 1528 alignment of i8 (next largest) while both i65 and i256 will use the 1529 alignment of i64 (largest specified). 1530#. If no match is found, and the type sought is a vector type, then the 1531 largest vector type that is smaller than the sought vector type will 1532 be used as a fall back. This happens because <128 x double> can be 1533 implemented in terms of 64 <2 x double>, for example. 1534 1535The function of the data layout string may not be what you expect. 1536Notably, this is not a specification from the frontend of what alignment 1537the code generator should use. 1538 1539Instead, if specified, the target data layout is required to match what 1540the ultimate *code generator* expects. This string is used by the 1541mid-level optimizers to improve code, and this only works if it matches 1542what the ultimate code generator uses. There is no way to generate IR 1543that does not embed this target-specific detail into the IR. If you 1544don't specify the string, the default specifications will be used to 1545generate a Data Layout and the optimization phases will operate 1546accordingly and introduce target specificity into the IR with respect to 1547these default specifications. 1548 1549.. _langref_triple: 1550 1551Target Triple 1552------------- 1553 1554A module may specify a target triple string that describes the target 1555host. The syntax for the target triple is simply: 1556 1557.. code-block:: llvm 1558 1559 target triple = "x86_64-apple-macosx10.7.0" 1560 1561The *target triple* string consists of a series of identifiers delimited 1562by the minus sign character ('-'). The canonical forms are: 1563 1564:: 1565 1566 ARCHITECTURE-VENDOR-OPERATING_SYSTEM 1567 ARCHITECTURE-VENDOR-OPERATING_SYSTEM-ENVIRONMENT 1568 1569This information is passed along to the backend so that it generates 1570code for the proper architecture. It's possible to override this on the 1571command line with the ``-mtriple`` command line option. 1572 1573.. _pointeraliasing: 1574 1575Pointer Aliasing Rules 1576---------------------- 1577 1578Any memory access must be done through a pointer value associated with 1579an address range of the memory access, otherwise the behavior is 1580undefined. Pointer values are associated with address ranges according 1581to the following rules: 1582 1583- A pointer value is associated with the addresses associated with any 1584 value it is *based* on. 1585- An address of a global variable is associated with the address range 1586 of the variable's storage. 1587- The result value of an allocation instruction is associated with the 1588 address range of the allocated storage. 1589- A null pointer in the default address-space is associated with no 1590 address. 1591- An integer constant other than zero or a pointer value returned from 1592 a function not defined within LLVM may be associated with address 1593 ranges allocated through mechanisms other than those provided by 1594 LLVM. Such ranges shall not overlap with any ranges of addresses 1595 allocated by mechanisms provided by LLVM. 1596 1597A pointer value is *based* on another pointer value according to the 1598following rules: 1599 1600- A pointer value formed from a ``getelementptr`` operation is *based* 1601 on the first value operand of the ``getelementptr``. 1602- The result value of a ``bitcast`` is *based* on the operand of the 1603 ``bitcast``. 1604- A pointer value formed by an ``inttoptr`` is *based* on all pointer 1605 values that contribute (directly or indirectly) to the computation of 1606 the pointer's value. 1607- The "*based* on" relationship is transitive. 1608 1609Note that this definition of *"based"* is intentionally similar to the 1610definition of *"based"* in C99, though it is slightly weaker. 1611 1612LLVM IR does not associate types with memory. The result type of a 1613``load`` merely indicates the size and alignment of the memory from 1614which to load, as well as the interpretation of the value. The first 1615operand type of a ``store`` similarly only indicates the size and 1616alignment of the store. 1617 1618Consequently, type-based alias analysis, aka TBAA, aka 1619``-fstrict-aliasing``, is not applicable to general unadorned LLVM IR. 1620:ref:`Metadata <metadata>` may be used to encode additional information 1621which specialized optimization passes may use to implement type-based 1622alias analysis. 1623 1624.. _volatile: 1625 1626Volatile Memory Accesses 1627------------------------ 1628 1629Certain memory accesses, such as :ref:`load <i_load>`'s, 1630:ref:`store <i_store>`'s, and :ref:`llvm.memcpy <int_memcpy>`'s may be 1631marked ``volatile``. The optimizers must not change the number of 1632volatile operations or change their order of execution relative to other 1633volatile operations. The optimizers *may* change the order of volatile 1634operations relative to non-volatile operations. This is not Java's 1635"volatile" and has no cross-thread synchronization behavior. 1636 1637IR-level volatile loads and stores cannot safely be optimized into 1638llvm.memcpy or llvm.memmove intrinsics even when those intrinsics are 1639flagged volatile. Likewise, the backend should never split or merge 1640target-legal volatile load/store instructions. 1641 1642.. admonition:: Rationale 1643 1644 Platforms may rely on volatile loads and stores of natively supported 1645 data width to be executed as single instruction. For example, in C 1646 this holds for an l-value of volatile primitive type with native 1647 hardware support, but not necessarily for aggregate types. The 1648 frontend upholds these expectations, which are intentionally 1649 unspecified in the IR. The rules above ensure that IR transformation 1650 do not violate the frontend's contract with the language. 1651 1652.. _memmodel: 1653 1654Memory Model for Concurrent Operations 1655-------------------------------------- 1656 1657The LLVM IR does not define any way to start parallel threads of 1658execution or to register signal handlers. Nonetheless, there are 1659platform-specific ways to create them, and we define LLVM IR's behavior 1660in their presence. This model is inspired by the C++0x memory model. 1661 1662For a more informal introduction to this model, see the :doc:`Atomics`. 1663 1664We define a *happens-before* partial order as the least partial order 1665that 1666 1667- Is a superset of single-thread program order, and 1668- When a *synchronizes-with* ``b``, includes an edge from ``a`` to 1669 ``b``. *Synchronizes-with* pairs are introduced by platform-specific 1670 techniques, like pthread locks, thread creation, thread joining, 1671 etc., and by atomic instructions. (See also :ref:`Atomic Memory Ordering 1672 Constraints <ordering>`). 1673 1674Note that program order does not introduce *happens-before* edges 1675between a thread and signals executing inside that thread. 1676 1677Every (defined) read operation (load instructions, memcpy, atomic 1678loads/read-modify-writes, etc.) R reads a series of bytes written by 1679(defined) write operations (store instructions, atomic 1680stores/read-modify-writes, memcpy, etc.). For the purposes of this 1681section, initialized globals are considered to have a write of the 1682initializer which is atomic and happens before any other read or write 1683of the memory in question. For each byte of a read R, R\ :sub:`byte` 1684may see any write to the same byte, except: 1685 1686- If write\ :sub:`1` happens before write\ :sub:`2`, and 1687 write\ :sub:`2` happens before R\ :sub:`byte`, then 1688 R\ :sub:`byte` does not see write\ :sub:`1`. 1689- If R\ :sub:`byte` happens before write\ :sub:`3`, then 1690 R\ :sub:`byte` does not see write\ :sub:`3`. 1691 1692Given that definition, R\ :sub:`byte` is defined as follows: 1693 1694- If R is volatile, the result is target-dependent. (Volatile is 1695 supposed to give guarantees which can support ``sig_atomic_t`` in 1696 C/C++, and may be used for accesses to addresses that do not behave 1697 like normal memory. It does not generally provide cross-thread 1698 synchronization.) 1699- Otherwise, if there is no write to the same byte that happens before 1700 R\ :sub:`byte`, R\ :sub:`byte` returns ``undef`` for that byte. 1701- Otherwise, if R\ :sub:`byte` may see exactly one write, 1702 R\ :sub:`byte` returns the value written by that write. 1703- Otherwise, if R is atomic, and all the writes R\ :sub:`byte` may 1704 see are atomic, it chooses one of the values written. See the :ref:`Atomic 1705 Memory Ordering Constraints <ordering>` section for additional 1706 constraints on how the choice is made. 1707- Otherwise R\ :sub:`byte` returns ``undef``. 1708 1709R returns the value composed of the series of bytes it read. This 1710implies that some bytes within the value may be ``undef`` **without** 1711the entire value being ``undef``. Note that this only defines the 1712semantics of the operation; it doesn't mean that targets will emit more 1713than one instruction to read the series of bytes. 1714 1715Note that in cases where none of the atomic intrinsics are used, this 1716model places only one restriction on IR transformations on top of what 1717is required for single-threaded execution: introducing a store to a byte 1718which might not otherwise be stored is not allowed in general. 1719(Specifically, in the case where another thread might write to and read 1720from an address, introducing a store can change a load that may see 1721exactly one write into a load that may see multiple writes.) 1722 1723.. _ordering: 1724 1725Atomic Memory Ordering Constraints 1726---------------------------------- 1727 1728Atomic instructions (:ref:`cmpxchg <i_cmpxchg>`, 1729:ref:`atomicrmw <i_atomicrmw>`, :ref:`fence <i_fence>`, 1730:ref:`atomic load <i_load>`, and :ref:`atomic store <i_store>`) take 1731ordering parameters that determine which other atomic instructions on 1732the same address they *synchronize with*. These semantics are borrowed 1733from Java and C++0x, but are somewhat more colloquial. If these 1734descriptions aren't precise enough, check those specs (see spec 1735references in the :doc:`atomics guide <Atomics>`). 1736:ref:`fence <i_fence>` instructions treat these orderings somewhat 1737differently since they don't take an address. See that instruction's 1738documentation for details. 1739 1740For a simpler introduction to the ordering constraints, see the 1741:doc:`Atomics`. 1742 1743``unordered`` 1744 The set of values that can be read is governed by the happens-before 1745 partial order. A value cannot be read unless some operation wrote 1746 it. This is intended to provide a guarantee strong enough to model 1747 Java's non-volatile shared variables. This ordering cannot be 1748 specified for read-modify-write operations; it is not strong enough 1749 to make them atomic in any interesting way. 1750``monotonic`` 1751 In addition to the guarantees of ``unordered``, there is a single 1752 total order for modifications by ``monotonic`` operations on each 1753 address. All modification orders must be compatible with the 1754 happens-before order. There is no guarantee that the modification 1755 orders can be combined to a global total order for the whole program 1756 (and this often will not be possible). The read in an atomic 1757 read-modify-write operation (:ref:`cmpxchg <i_cmpxchg>` and 1758 :ref:`atomicrmw <i_atomicrmw>`) reads the value in the modification 1759 order immediately before the value it writes. If one atomic read 1760 happens before another atomic read of the same address, the later 1761 read must see the same value or a later value in the address's 1762 modification order. This disallows reordering of ``monotonic`` (or 1763 stronger) operations on the same address. If an address is written 1764 ``monotonic``-ally by one thread, and other threads ``monotonic``-ally 1765 read that address repeatedly, the other threads must eventually see 1766 the write. This corresponds to the C++0x/C1x 1767 ``memory_order_relaxed``. 1768``acquire`` 1769 In addition to the guarantees of ``monotonic``, a 1770 *synchronizes-with* edge may be formed with a ``release`` operation. 1771 This is intended to model C++'s ``memory_order_acquire``. 1772``release`` 1773 In addition to the guarantees of ``monotonic``, if this operation 1774 writes a value which is subsequently read by an ``acquire`` 1775 operation, it *synchronizes-with* that operation. (This isn't a 1776 complete description; see the C++0x definition of a release 1777 sequence.) This corresponds to the C++0x/C1x 1778 ``memory_order_release``. 1779``acq_rel`` (acquire+release) 1780 Acts as both an ``acquire`` and ``release`` operation on its 1781 address. This corresponds to the C++0x/C1x ``memory_order_acq_rel``. 1782``seq_cst`` (sequentially consistent) 1783 In addition to the guarantees of ``acq_rel`` (``acquire`` for an 1784 operation that only reads, ``release`` for an operation that only 1785 writes), there is a global total order on all 1786 sequentially-consistent operations on all addresses, which is 1787 consistent with the *happens-before* partial order and with the 1788 modification orders of all the affected addresses. Each 1789 sequentially-consistent read sees the last preceding write to the 1790 same address in this global order. This corresponds to the C++0x/C1x 1791 ``memory_order_seq_cst`` and Java volatile. 1792 1793.. _singlethread: 1794 1795If an atomic operation is marked ``singlethread``, it only *synchronizes 1796with* or participates in modification and seq\_cst total orderings with 1797other operations running in the same thread (for example, in signal 1798handlers). 1799 1800.. _fastmath: 1801 1802Fast-Math Flags 1803--------------- 1804 1805LLVM IR floating-point binary ops (:ref:`fadd <i_fadd>`, 1806:ref:`fsub <i_fsub>`, :ref:`fmul <i_fmul>`, :ref:`fdiv <i_fdiv>`, 1807:ref:`frem <i_frem>`) have the following flags that can be set to enable 1808otherwise unsafe floating point operations 1809 1810``nnan`` 1811 No NaNs - Allow optimizations to assume the arguments and result are not 1812 NaN. Such optimizations are required to retain defined behavior over 1813 NaNs, but the value of the result is undefined. 1814 1815``ninf`` 1816 No Infs - Allow optimizations to assume the arguments and result are not 1817 +/-Inf. Such optimizations are required to retain defined behavior over 1818 +/-Inf, but the value of the result is undefined. 1819 1820``nsz`` 1821 No Signed Zeros - Allow optimizations to treat the sign of a zero 1822 argument or result as insignificant. 1823 1824``arcp`` 1825 Allow Reciprocal - Allow optimizations to use the reciprocal of an 1826 argument rather than perform division. 1827 1828``fast`` 1829 Fast - Allow algebraically equivalent transformations that may 1830 dramatically change results in floating point (e.g. reassociate). This 1831 flag implies all the others. 1832 1833.. _uselistorder: 1834 1835Use-list Order Directives 1836------------------------- 1837 1838Use-list directives encode the in-memory order of each use-list, allowing the 1839order to be recreated. ``<order-indexes>`` is a comma-separated list of 1840indexes that are assigned to the referenced value's uses. The referenced 1841value's use-list is immediately sorted by these indexes. 1842 1843Use-list directives may appear at function scope or global scope. They are not 1844instructions, and have no effect on the semantics of the IR. When they're at 1845function scope, they must appear after the terminator of the final basic block. 1846 1847If basic blocks have their address taken via ``blockaddress()`` expressions, 1848``uselistorder_bb`` can be used to reorder their use-lists from outside their 1849function's scope. 1850 1851:Syntax: 1852 1853:: 1854 1855 uselistorder <ty> <value>, { <order-indexes> } 1856 uselistorder_bb @function, %block { <order-indexes> } 1857 1858:Examples: 1859 1860:: 1861 1862 define void @foo(i32 %arg1, i32 %arg2) { 1863 entry: 1864 ; ... instructions ... 1865 bb: 1866 ; ... instructions ... 1867 1868 ; At function scope. 1869 uselistorder i32 %arg1, { 1, 0, 2 } 1870 uselistorder label %bb, { 1, 0 } 1871 } 1872 1873 ; At global scope. 1874 uselistorder i32* @global, { 1, 2, 0 } 1875 uselistorder i32 7, { 1, 0 } 1876 uselistorder i32 (i32) @bar, { 1, 0 } 1877 uselistorder_bb @foo, %bb, { 5, 1, 3, 2, 0, 4 } 1878 1879.. _typesystem: 1880 1881Type System 1882=========== 1883 1884The LLVM type system is one of the most important features of the 1885intermediate representation. Being typed enables a number of 1886optimizations to be performed on the intermediate representation 1887directly, without having to do extra analyses on the side before the 1888transformation. A strong type system makes it easier to read the 1889generated code and enables novel analyses and transformations that are 1890not feasible to perform on normal three address code representations. 1891 1892.. _t_void: 1893 1894Void Type 1895--------- 1896 1897:Overview: 1898 1899 1900The void type does not represent any value and has no size. 1901 1902:Syntax: 1903 1904 1905:: 1906 1907 void 1908 1909 1910.. _t_function: 1911 1912Function Type 1913------------- 1914 1915:Overview: 1916 1917 1918The function type can be thought of as a function signature. It consists of a 1919return type and a list of formal parameter types. The return type of a function 1920type is a void type or first class type --- except for :ref:`label <t_label>` 1921and :ref:`metadata <t_metadata>` types. 1922 1923:Syntax: 1924 1925:: 1926 1927 <returntype> (<parameter list>) 1928 1929...where '``<parameter list>``' is a comma-separated list of type 1930specifiers. Optionally, the parameter list may include a type ``...``, which 1931indicates that the function takes a variable number of arguments. Variable 1932argument functions can access their arguments with the :ref:`variable argument 1933handling intrinsic <int_varargs>` functions. '``<returntype>``' is any type 1934except :ref:`label <t_label>` and :ref:`metadata <t_metadata>`. 1935 1936:Examples: 1937 1938+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 1939| ``i32 (i32)`` | function taking an ``i32``, returning an ``i32`` | 1940+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 1941| ``float (i16, i32 *) *`` | :ref:`Pointer <t_pointer>` to a function that takes an ``i16`` and a :ref:`pointer <t_pointer>` to ``i32``, returning ``float``. | 1942+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 1943| ``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. | 1944+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 1945| ``{i32, i32} (i32)`` | A function taking an ``i32``, returning a :ref:`structure <t_struct>` containing two ``i32`` values | 1946+---------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 1947 1948.. _t_firstclass: 1949 1950First Class Types 1951----------------- 1952 1953The :ref:`first class <t_firstclass>` types are perhaps the most important. 1954Values of these types are the only ones which can be produced by 1955instructions. 1956 1957.. _t_single_value: 1958 1959Single Value Types 1960^^^^^^^^^^^^^^^^^^ 1961 1962These are the types that are valid in registers from CodeGen's perspective. 1963 1964.. _t_integer: 1965 1966Integer Type 1967"""""""""""" 1968 1969:Overview: 1970 1971The integer type is a very simple type that simply specifies an 1972arbitrary bit width for the integer type desired. Any bit width from 1 1973bit to 2\ :sup:`23`\ -1 (about 8 million) can be specified. 1974 1975:Syntax: 1976 1977:: 1978 1979 iN 1980 1981The number of bits the integer will occupy is specified by the ``N`` 1982value. 1983 1984Examples: 1985********* 1986 1987+----------------+------------------------------------------------+ 1988| ``i1`` | a single-bit integer. | 1989+----------------+------------------------------------------------+ 1990| ``i32`` | a 32-bit integer. | 1991+----------------+------------------------------------------------+ 1992| ``i1942652`` | a really big integer of over 1 million bits. | 1993+----------------+------------------------------------------------+ 1994 1995.. _t_floating: 1996 1997Floating Point Types 1998"""""""""""""""""""" 1999 2000.. list-table:: 2001 :header-rows: 1 2002 2003 * - Type 2004 - Description 2005 2006 * - ``half`` 2007 - 16-bit floating point value 2008 2009 * - ``float`` 2010 - 32-bit floating point value 2011 2012 * - ``double`` 2013 - 64-bit floating point value 2014 2015 * - ``fp128`` 2016 - 128-bit floating point value (112-bit mantissa) 2017 2018 * - ``x86_fp80`` 2019 - 80-bit floating point value (X87) 2020 2021 * - ``ppc_fp128`` 2022 - 128-bit floating point value (two 64-bits) 2023 2024X86_mmx Type 2025"""""""""""" 2026 2027:Overview: 2028 2029The x86_mmx type represents a value held in an MMX register on an x86 2030machine. The operations allowed on it are quite limited: parameters and 2031return values, load and store, and bitcast. User-specified MMX 2032instructions are represented as intrinsic or asm calls with arguments 2033and/or results of this type. There are no arrays, vectors or constants 2034of this type. 2035 2036:Syntax: 2037 2038:: 2039 2040 x86_mmx 2041 2042 2043.. _t_pointer: 2044 2045Pointer Type 2046"""""""""""" 2047 2048:Overview: 2049 2050The pointer type is used to specify memory locations. Pointers are 2051commonly used to reference objects in memory. 2052 2053Pointer types may have an optional address space attribute defining the 2054numbered address space where the pointed-to object resides. The default 2055address space is number zero. The semantics of non-zero address spaces 2056are target-specific. 2057 2058Note that LLVM does not permit pointers to void (``void*``) nor does it 2059permit pointers to labels (``label*``). Use ``i8*`` instead. 2060 2061:Syntax: 2062 2063:: 2064 2065 <type> * 2066 2067:Examples: 2068 2069+-------------------------+--------------------------------------------------------------------------------------------------------------+ 2070| ``[4 x i32]*`` | A :ref:`pointer <t_pointer>` to :ref:`array <t_array>` of four ``i32`` values. | 2071+-------------------------+--------------------------------------------------------------------------------------------------------------+ 2072| ``i32 (i32*) *`` | A :ref:`pointer <t_pointer>` to a :ref:`function <t_function>` that takes an ``i32*``, returning an ``i32``. | 2073+-------------------------+--------------------------------------------------------------------------------------------------------------+ 2074| ``i32 addrspace(5)*`` | A :ref:`pointer <t_pointer>` to an ``i32`` value that resides in address space #5. | 2075+-------------------------+--------------------------------------------------------------------------------------------------------------+ 2076 2077.. _t_vector: 2078 2079Vector Type 2080""""""""""" 2081 2082:Overview: 2083 2084A vector type is a simple derived type that represents a vector of 2085elements. Vector types are used when multiple primitive data are 2086operated in parallel using a single instruction (SIMD). A vector type 2087requires a size (number of elements) and an underlying primitive data 2088type. Vector types are considered :ref:`first class <t_firstclass>`. 2089 2090:Syntax: 2091 2092:: 2093 2094 < <# elements> x <elementtype> > 2095 2096The number of elements is a constant integer value larger than 0; 2097elementtype may be any integer, floating point or pointer type. Vectors 2098of size zero are not allowed. 2099 2100:Examples: 2101 2102+-------------------+--------------------------------------------------+ 2103| ``<4 x i32>`` | Vector of 4 32-bit integer values. | 2104+-------------------+--------------------------------------------------+ 2105| ``<8 x float>`` | Vector of 8 32-bit floating-point values. | 2106+-------------------+--------------------------------------------------+ 2107| ``<2 x i64>`` | Vector of 2 64-bit integer values. | 2108+-------------------+--------------------------------------------------+ 2109| ``<4 x i64*>`` | Vector of 4 pointers to 64-bit integer values. | 2110+-------------------+--------------------------------------------------+ 2111 2112.. _t_label: 2113 2114Label Type 2115^^^^^^^^^^ 2116 2117:Overview: 2118 2119The label type represents code labels. 2120 2121:Syntax: 2122 2123:: 2124 2125 label 2126 2127.. _t_metadata: 2128 2129Metadata Type 2130^^^^^^^^^^^^^ 2131 2132:Overview: 2133 2134The metadata type represents embedded metadata. No derived types may be 2135created from metadata except for :ref:`function <t_function>` arguments. 2136 2137:Syntax: 2138 2139:: 2140 2141 metadata 2142 2143.. _t_aggregate: 2144 2145Aggregate Types 2146^^^^^^^^^^^^^^^ 2147 2148Aggregate Types are a subset of derived types that can contain multiple 2149member types. :ref:`Arrays <t_array>` and :ref:`structs <t_struct>` are 2150aggregate types. :ref:`Vectors <t_vector>` are not considered to be 2151aggregate types. 2152 2153.. _t_array: 2154 2155Array Type 2156"""""""""" 2157 2158:Overview: 2159 2160The array type is a very simple derived type that arranges elements 2161sequentially in memory. The array type requires a size (number of 2162elements) and an underlying data type. 2163 2164:Syntax: 2165 2166:: 2167 2168 [<# elements> x <elementtype>] 2169 2170The number of elements is a constant integer value; ``elementtype`` may 2171be any type with a size. 2172 2173:Examples: 2174 2175+------------------+--------------------------------------+ 2176| ``[40 x i32]`` | Array of 40 32-bit integer values. | 2177+------------------+--------------------------------------+ 2178| ``[41 x i32]`` | Array of 41 32-bit integer values. | 2179+------------------+--------------------------------------+ 2180| ``[4 x i8]`` | Array of 4 8-bit integer values. | 2181+------------------+--------------------------------------+ 2182 2183Here are some examples of multidimensional arrays: 2184 2185+-----------------------------+----------------------------------------------------------+ 2186| ``[3 x [4 x i32]]`` | 3x4 array of 32-bit integer values. | 2187+-----------------------------+----------------------------------------------------------+ 2188| ``[12 x [10 x float]]`` | 12x10 array of single precision floating point values. | 2189+-----------------------------+----------------------------------------------------------+ 2190| ``[2 x [3 x [4 x i16]]]`` | 2x3x4 array of 16-bit integer values. | 2191+-----------------------------+----------------------------------------------------------+ 2192 2193There is no restriction on indexing beyond the end of the array implied 2194by a static type (though there are restrictions on indexing beyond the 2195bounds of an allocated object in some cases). This means that 2196single-dimension 'variable sized array' addressing can be implemented in 2197LLVM with a zero length array type. An implementation of 'pascal style 2198arrays' in LLVM could use the type "``{ i32, [0 x float]}``", for 2199example. 2200 2201.. _t_struct: 2202 2203Structure Type 2204"""""""""""""" 2205 2206:Overview: 2207 2208The structure type is used to represent a collection of data members 2209together in memory. The elements of a structure may be any type that has 2210a size. 2211 2212Structures in memory are accessed using '``load``' and '``store``' by 2213getting a pointer to a field with the '``getelementptr``' instruction. 2214Structures in registers are accessed using the '``extractvalue``' and 2215'``insertvalue``' instructions. 2216 2217Structures may optionally be "packed" structures, which indicate that 2218the alignment of the struct is one byte, and that there is no padding 2219between the elements. In non-packed structs, padding between field types 2220is inserted as defined by the DataLayout string in the module, which is 2221required to match what the underlying code generator expects. 2222 2223Structures can either be "literal" or "identified". A literal structure 2224is defined inline with other types (e.g. ``{i32, i32}*``) whereas 2225identified types are always defined at the top level with a name. 2226Literal types are uniqued by their contents and can never be recursive 2227or opaque since there is no way to write one. Identified types can be 2228recursive, can be opaqued, and are never uniqued. 2229 2230:Syntax: 2231 2232:: 2233 2234 %T1 = type { <type list> } ; Identified normal struct type 2235 %T2 = type <{ <type list> }> ; Identified packed struct type 2236 2237:Examples: 2238 2239+------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 2240| ``{ i32, i32, i32 }`` | A triple of three ``i32`` values | 2241+------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 2242| ``{ 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``. | 2243+------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 2244| ``<{ i8, i32 }>`` | A packed struct known to be 5 bytes in size. | 2245+------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ 2246 2247.. _t_opaque: 2248 2249Opaque Structure Types 2250"""""""""""""""""""""" 2251 2252:Overview: 2253 2254Opaque structure types are used to represent named structure types that 2255do not have a body specified. This corresponds (for example) to the C 2256notion of a forward declared structure. 2257 2258:Syntax: 2259 2260:: 2261 2262 %X = type opaque 2263 %52 = type opaque 2264 2265:Examples: 2266 2267+--------------+-------------------+ 2268| ``opaque`` | An opaque type. | 2269+--------------+-------------------+ 2270 2271.. _constants: 2272 2273Constants 2274========= 2275 2276LLVM has several different basic types of constants. This section 2277describes them all and their syntax. 2278 2279Simple Constants 2280---------------- 2281 2282**Boolean constants** 2283 The two strings '``true``' and '``false``' are both valid constants 2284 of the ``i1`` type. 2285**Integer constants** 2286 Standard integers (such as '4') are constants of the 2287 :ref:`integer <t_integer>` type. Negative numbers may be used with 2288 integer types. 2289**Floating point constants** 2290 Floating point constants use standard decimal notation (e.g. 2291 123.421), exponential notation (e.g. 1.23421e+2), or a more precise 2292 hexadecimal notation (see below). The assembler requires the exact 2293 decimal value of a floating-point constant. For example, the 2294 assembler accepts 1.25 but rejects 1.3 because 1.3 is a repeating 2295 decimal in binary. Floating point constants must have a :ref:`floating 2296 point <t_floating>` type. 2297**Null pointer constants** 2298 The identifier '``null``' is recognized as a null pointer constant 2299 and must be of :ref:`pointer type <t_pointer>`. 2300 2301The one non-intuitive notation for constants is the hexadecimal form of 2302floating point constants. For example, the form 2303'``double 0x432ff973cafa8000``' is equivalent to (but harder to read 2304than) '``double 4.5e+15``'. The only time hexadecimal floating point 2305constants are required (and the only time that they are generated by the 2306disassembler) is when a floating point constant must be emitted but it 2307cannot be represented as a decimal floating point number in a reasonable 2308number of digits. For example, NaN's, infinities, and other special 2309values are represented in their IEEE hexadecimal format so that assembly 2310and disassembly do not cause any bits to change in the constants. 2311 2312When using the hexadecimal form, constants of types half, float, and 2313double are represented using the 16-digit form shown above (which 2314matches the IEEE754 representation for double); half and float values 2315must, however, be exactly representable as IEEE 754 half and single 2316precision, respectively. Hexadecimal format is always used for long 2317double, and there are three forms of long double. The 80-bit format used 2318by x86 is represented as ``0xK`` followed by 20 hexadecimal digits. The 2319128-bit format used by PowerPC (two adjacent doubles) is represented by 2320``0xM`` followed by 32 hexadecimal digits. The IEEE 128-bit format is 2321represented by ``0xL`` followed by 32 hexadecimal digits. Long doubles 2322will only work if they match the long double format on your target. 2323The IEEE 16-bit format (half precision) is represented by ``0xH`` 2324followed by 4 hexadecimal digits. All hexadecimal formats are big-endian 2325(sign bit at the left). 2326 2327There are no constants of type x86_mmx. 2328 2329.. _complexconstants: 2330 2331Complex Constants 2332----------------- 2333 2334Complex constants are a (potentially recursive) combination of simple 2335constants and smaller complex constants. 2336 2337**Structure constants** 2338 Structure constants are represented with notation similar to 2339 structure type definitions (a comma separated list of elements, 2340 surrounded by braces (``{}``)). For example: 2341 "``{ i32 4, float 17.0, i32* @G }``", where "``@G``" is declared as 2342 "``@G = external global i32``". Structure constants must have 2343 :ref:`structure type <t_struct>`, and the number and types of elements 2344 must match those specified by the type. 2345**Array constants** 2346 Array constants are represented with notation similar to array type 2347 definitions (a comma separated list of elements, surrounded by 2348 square brackets (``[]``)). For example: 2349 "``[ i32 42, i32 11, i32 74 ]``". Array constants must have 2350 :ref:`array type <t_array>`, and the number and types of elements must 2351 match those specified by the type. As a special case, character array 2352 constants may also be represented as a double-quoted string using the ``c`` 2353 prefix. For example: "``c"Hello World\0A\00"``". 2354**Vector constants** 2355 Vector constants are represented with notation similar to vector 2356 type definitions (a comma separated list of elements, surrounded by 2357 less-than/greater-than's (``<>``)). For example: 2358 "``< i32 42, i32 11, i32 74, i32 100 >``". Vector constants 2359 must have :ref:`vector type <t_vector>`, and the number and types of 2360 elements must match those specified by the type. 2361**Zero initialization** 2362 The string '``zeroinitializer``' can be used to zero initialize a 2363 value to zero of *any* type, including scalar and 2364 :ref:`aggregate <t_aggregate>` types. This is often used to avoid 2365 having to print large zero initializers (e.g. for large arrays) and 2366 is always exactly equivalent to using explicit zero initializers. 2367**Metadata node** 2368 A metadata node is a constant tuple without types. For example: 2369 "``!{!0, !{!2, !0}, !"test"}``". Metadata can reference constant values, 2370 for example: "``!{!0, i32 0, i8* @global, i64 (i64)* @function, !"str"}``". 2371 Unlike other typed constants that are meant to be interpreted as part of 2372 the instruction stream, metadata is a place to attach additional 2373 information such as debug info. 2374 2375Global Variable and Function Addresses 2376-------------------------------------- 2377 2378The addresses of :ref:`global variables <globalvars>` and 2379:ref:`functions <functionstructure>` are always implicitly valid 2380(link-time) constants. These constants are explicitly referenced when 2381the :ref:`identifier for the global <identifiers>` is used and always have 2382:ref:`pointer <t_pointer>` type. For example, the following is a legal LLVM 2383file: 2384 2385.. code-block:: llvm 2386 2387 @X = global i32 17 2388 @Y = global i32 42 2389 @Z = global [2 x i32*] [ i32* @X, i32* @Y ] 2390 2391.. _undefvalues: 2392 2393Undefined Values 2394---------------- 2395 2396The string '``undef``' can be used anywhere a constant is expected, and 2397indicates that the user of the value may receive an unspecified 2398bit-pattern. Undefined values may be of any type (other than '``label``' 2399or '``void``') and be used anywhere a constant is permitted. 2400 2401Undefined values are useful because they indicate to the compiler that 2402the program is well defined no matter what value is used. This gives the 2403compiler more freedom to optimize. Here are some examples of 2404(potentially surprising) transformations that are valid (in pseudo IR): 2405 2406.. code-block:: llvm 2407 2408 %A = add %X, undef 2409 %B = sub %X, undef 2410 %C = xor %X, undef 2411 Safe: 2412 %A = undef 2413 %B = undef 2414 %C = undef 2415 2416This is safe because all of the output bits are affected by the undef 2417bits. Any output bit can have a zero or one depending on the input bits. 2418 2419.. code-block:: llvm 2420 2421 %A = or %X, undef 2422 %B = and %X, undef 2423 Safe: 2424 %A = -1 2425 %B = 0 2426 Unsafe: 2427 %A = undef 2428 %B = undef 2429 2430These logical operations have bits that are not always affected by the 2431input. For example, if ``%X`` has a zero bit, then the output of the 2432'``and``' operation will always be a zero for that bit, no matter what 2433the corresponding bit from the '``undef``' is. As such, it is unsafe to 2434optimize or assume that the result of the '``and``' is '``undef``'. 2435However, it is safe to assume that all bits of the '``undef``' could be 24360, and optimize the '``and``' to 0. Likewise, it is safe to assume that 2437all the bits of the '``undef``' operand to the '``or``' could be set, 2438allowing the '``or``' to be folded to -1. 2439 2440.. code-block:: llvm 2441 2442 %A = select undef, %X, %Y 2443 %B = select undef, 42, %Y 2444 %C = select %X, %Y, undef 2445 Safe: 2446 %A = %X (or %Y) 2447 %B = 42 (or %Y) 2448 %C = %Y 2449 Unsafe: 2450 %A = undef 2451 %B = undef 2452 %C = undef 2453 2454This set of examples shows that undefined '``select``' (and conditional 2455branch) conditions can go *either way*, but they have to come from one 2456of the two operands. In the ``%A`` example, if ``%X`` and ``%Y`` were 2457both known to have a clear low bit, then ``%A`` would have to have a 2458cleared low bit. However, in the ``%C`` example, the optimizer is 2459allowed to assume that the '``undef``' operand could be the same as 2460``%Y``, allowing the whole '``select``' to be eliminated. 2461 2462.. code-block:: llvm 2463 2464 %A = xor undef, undef 2465 2466 %B = undef 2467 %C = xor %B, %B 2468 2469 %D = undef 2470 %E = icmp slt %D, 4 2471 %F = icmp gte %D, 4 2472 2473 Safe: 2474 %A = undef 2475 %B = undef 2476 %C = undef 2477 %D = undef 2478 %E = undef 2479 %F = undef 2480 2481This example points out that two '``undef``' operands are not 2482necessarily the same. This can be surprising to people (and also matches 2483C semantics) where they assume that "``X^X``" is always zero, even if 2484``X`` is undefined. This isn't true for a number of reasons, but the 2485short answer is that an '``undef``' "variable" can arbitrarily change 2486its value over its "live range". This is true because the variable 2487doesn't actually *have a live range*. Instead, the value is logically 2488read from arbitrary registers that happen to be around when needed, so 2489the value is not necessarily consistent over time. In fact, ``%A`` and 2490``%C`` need to have the same semantics or the core LLVM "replace all 2491uses with" concept would not hold. 2492 2493.. code-block:: llvm 2494 2495 %A = fdiv undef, %X 2496 %B = fdiv %X, undef 2497 Safe: 2498 %A = undef 2499 b: unreachable 2500 2501These examples show the crucial difference between an *undefined value* 2502and *undefined behavior*. An undefined value (like '``undef``') is 2503allowed to have an arbitrary bit-pattern. This means that the ``%A`` 2504operation can be constant folded to '``undef``', because the '``undef``' 2505could be an SNaN, and ``fdiv`` is not (currently) defined on SNaN's. 2506However, in the second example, we can make a more aggressive 2507assumption: because the ``undef`` is allowed to be an arbitrary value, 2508we are allowed to assume that it could be zero. Since a divide by zero 2509has *undefined behavior*, we are allowed to assume that the operation 2510does not execute at all. This allows us to delete the divide and all 2511code after it. Because the undefined operation "can't happen", the 2512optimizer can assume that it occurs in dead code. 2513 2514.. code-block:: llvm 2515 2516 a: store undef -> %X 2517 b: store %X -> undef 2518 Safe: 2519 a: <deleted> 2520 b: unreachable 2521 2522These examples reiterate the ``fdiv`` example: a store *of* an undefined 2523value can be assumed to not have any effect; we can assume that the 2524value is overwritten with bits that happen to match what was already 2525there. However, a store *to* an undefined location could clobber 2526arbitrary memory, therefore, it has undefined behavior. 2527 2528.. _poisonvalues: 2529 2530Poison Values 2531------------- 2532 2533Poison values are similar to :ref:`undef values <undefvalues>`, however 2534they also represent the fact that an instruction or constant expression 2535that cannot evoke side effects has nevertheless detected a condition 2536that results in undefined behavior. 2537 2538There is currently no way of representing a poison value in the IR; they 2539only exist when produced by operations such as :ref:`add <i_add>` with 2540the ``nsw`` flag. 2541 2542Poison value behavior is defined in terms of value *dependence*: 2543 2544- Values other than :ref:`phi <i_phi>` nodes depend on their operands. 2545- :ref:`Phi <i_phi>` nodes depend on the operand corresponding to 2546 their dynamic predecessor basic block. 2547- Function arguments depend on the corresponding actual argument values 2548 in the dynamic callers of their functions. 2549- :ref:`Call <i_call>` instructions depend on the :ref:`ret <i_ret>` 2550 instructions that dynamically transfer control back to them. 2551- :ref:`Invoke <i_invoke>` instructions depend on the 2552 :ref:`ret <i_ret>`, :ref:`resume <i_resume>`, or exception-throwing 2553 call instructions that dynamically transfer control back to them. 2554- Non-volatile loads and stores depend on the most recent stores to all 2555 of the referenced memory addresses, following the order in the IR 2556 (including loads and stores implied by intrinsics such as 2557 :ref:`@llvm.memcpy <int_memcpy>`.) 2558- An instruction with externally visible side effects depends on the 2559 most recent preceding instruction with externally visible side 2560 effects, following the order in the IR. (This includes :ref:`volatile 2561 operations <volatile>`.) 2562- An instruction *control-depends* on a :ref:`terminator 2563 instruction <terminators>` if the terminator instruction has 2564 multiple successors and the instruction is always executed when 2565 control transfers to one of the successors, and may not be executed 2566 when control is transferred to another. 2567- Additionally, an instruction also *control-depends* on a terminator 2568 instruction if the set of instructions it otherwise depends on would 2569 be different if the terminator had transferred control to a different 2570 successor. 2571- Dependence is transitive. 2572 2573Poison values have the same behavior as :ref:`undef values <undefvalues>`, 2574with the additional effect that any instruction that has a *dependence* 2575on a poison value has undefined behavior. 2576 2577Here are some examples: 2578 2579.. code-block:: llvm 2580 2581 entry: 2582 %poison = sub nuw i32 0, 1 ; Results in a poison value. 2583 %still_poison = and i32 %poison, 0 ; 0, but also poison. 2584 %poison_yet_again = getelementptr i32, i32* @h, i32 %still_poison 2585 store i32 0, i32* %poison_yet_again ; memory at @h[0] is poisoned 2586 2587 store i32 %poison, i32* @g ; Poison value stored to memory. 2588 %poison2 = load i32, i32* @g ; Poison value loaded back from memory. 2589 2590 store volatile i32 %poison, i32* @g ; External observation; undefined behavior. 2591 2592 %narrowaddr = bitcast i32* @g to i16* 2593 %wideaddr = bitcast i32* @g to i64* 2594 %poison3 = load i16, i16* %narrowaddr ; Returns a poison value. 2595 %poison4 = load i64, i64* %wideaddr ; Returns a poison value. 2596 2597 %cmp = icmp slt i32 %poison, 0 ; Returns a poison value. 2598 br i1 %cmp, label %true, label %end ; Branch to either destination. 2599 2600 true: 2601 store volatile i32 0, i32* @g ; This is control-dependent on %cmp, so 2602 ; it has undefined behavior. 2603 br label %end 2604 2605 end: 2606 %p = phi i32 [ 0, %entry ], [ 1, %true ] 2607 ; Both edges into this PHI are 2608 ; control-dependent on %cmp, so this 2609 ; always results in a poison value. 2610 2611 store volatile i32 0, i32* @g ; This would depend on the store in %true 2612 ; if %cmp is true, or the store in %entry 2613 ; otherwise, so this is undefined behavior. 2614 2615 br i1 %cmp, label %second_true, label %second_end 2616 ; The same branch again, but this time the 2617 ; true block doesn't have side effects. 2618 2619 second_true: 2620 ; No side effects! 2621 ret void 2622 2623 second_end: 2624 store volatile i32 0, i32* @g ; This time, the instruction always depends 2625 ; on the store in %end. Also, it is 2626 ; control-equivalent to %end, so this is 2627 ; well-defined (ignoring earlier undefined 2628 ; behavior in this example). 2629 2630.. _blockaddress: 2631 2632Addresses of Basic Blocks 2633------------------------- 2634 2635``blockaddress(@function, %block)`` 2636 2637The '``blockaddress``' constant computes the address of the specified 2638basic block in the specified function, and always has an ``i8*`` type. 2639Taking the address of the entry block is illegal. 2640 2641This value only has defined behavior when used as an operand to the 2642':ref:`indirectbr <i_indirectbr>`' instruction, or for comparisons 2643against null. Pointer equality tests between labels addresses results in 2644undefined behavior --- though, again, comparison against null is ok, and 2645no label is equal to the null pointer. This may be passed around as an 2646opaque pointer sized value as long as the bits are not inspected. This 2647allows ``ptrtoint`` and arithmetic to be performed on these values so 2648long as the original value is reconstituted before the ``indirectbr`` 2649instruction. 2650 2651Finally, some targets may provide defined semantics when using the value 2652as the operand to an inline assembly, but that is target specific. 2653 2654.. _constantexprs: 2655 2656Constant Expressions 2657-------------------- 2658 2659Constant expressions are used to allow expressions involving other 2660constants to be used as constants. Constant expressions may be of any 2661:ref:`first class <t_firstclass>` type and may involve any LLVM operation 2662that does not have side effects (e.g. load and call are not supported). 2663The following is the syntax for constant expressions: 2664 2665``trunc (CST to TYPE)`` 2666 Truncate a constant to another type. The bit size of CST must be 2667 larger than the bit size of TYPE. Both types must be integers. 2668``zext (CST to TYPE)`` 2669 Zero extend a constant to another type. The bit size of CST must be 2670 smaller than the bit size of TYPE. Both types must be integers. 2671``sext (CST to TYPE)`` 2672 Sign extend a constant to another type. The bit size of CST must be 2673 smaller than the bit size of TYPE. Both types must be integers. 2674``fptrunc (CST to TYPE)`` 2675 Truncate a floating point constant to another floating point type. 2676 The size of CST must be larger than the size of TYPE. Both types 2677 must be floating point. 2678``fpext (CST to TYPE)`` 2679 Floating point extend a constant to another type. The size of CST 2680 must be smaller or equal to the size of TYPE. Both types must be 2681 floating point. 2682``fptoui (CST to TYPE)`` 2683 Convert a floating point constant to the corresponding unsigned 2684 integer constant. TYPE must be a scalar or vector integer type. CST 2685 must be of scalar or vector floating point type. Both CST and TYPE 2686 must be scalars, or vectors of the same number of elements. If the 2687 value won't fit in the integer type, the results are undefined. 2688``fptosi (CST to TYPE)`` 2689 Convert a floating point constant to the corresponding signed 2690 integer constant. TYPE must be a scalar or vector integer type. CST 2691 must be of scalar or vector floating point type. Both CST and TYPE 2692 must be scalars, or vectors of the same number of elements. If the 2693 value won't fit in the integer type, the results are undefined. 2694``uitofp (CST to TYPE)`` 2695 Convert an unsigned integer constant to the corresponding floating 2696 point constant. TYPE must be a scalar or vector floating point type. 2697 CST must be of scalar or vector integer type. Both CST and TYPE must 2698 be scalars, or vectors of the same number of elements. If the value 2699 won't fit in the floating point type, the results are undefined. 2700``sitofp (CST to TYPE)`` 2701 Convert a signed integer constant to the corresponding floating 2702 point constant. TYPE must be a scalar or vector floating point type. 2703 CST must be of scalar or vector integer type. Both CST and TYPE must 2704 be scalars, or vectors of the same number of elements. If the value 2705 won't fit in the floating point type, the results are undefined. 2706``ptrtoint (CST to TYPE)`` 2707 Convert a pointer typed constant to the corresponding integer 2708 constant. ``TYPE`` must be an integer type. ``CST`` must be of 2709 pointer type. The ``CST`` value is zero extended, truncated, or 2710 unchanged to make it fit in ``TYPE``. 2711``inttoptr (CST to TYPE)`` 2712 Convert an integer constant to a pointer constant. TYPE must be a 2713 pointer type. CST must be of integer type. The CST value is zero 2714 extended, truncated, or unchanged to make it fit in a pointer size. 2715 This one is *really* dangerous! 2716``bitcast (CST to TYPE)`` 2717 Convert a constant, CST, to another TYPE. The constraints of the 2718 operands are the same as those for the :ref:`bitcast 2719 instruction <i_bitcast>`. 2720``addrspacecast (CST to TYPE)`` 2721 Convert a constant pointer or constant vector of pointer, CST, to another 2722 TYPE in a different address space. The constraints of the operands are the 2723 same as those for the :ref:`addrspacecast instruction <i_addrspacecast>`. 2724``getelementptr (TY, CSTPTR, IDX0, IDX1, ...)``, ``getelementptr inbounds (TY, CSTPTR, IDX0, IDX1, ...)`` 2725 Perform the :ref:`getelementptr operation <i_getelementptr>` on 2726 constants. As with the :ref:`getelementptr <i_getelementptr>` 2727 instruction, the index list may have zero or more indexes, which are 2728 required to make sense for the type of "pointer to TY". 2729``select (COND, VAL1, VAL2)`` 2730 Perform the :ref:`select operation <i_select>` on constants. 2731``icmp COND (VAL1, VAL2)`` 2732 Performs the :ref:`icmp operation <i_icmp>` on constants. 2733``fcmp COND (VAL1, VAL2)`` 2734 Performs the :ref:`fcmp operation <i_fcmp>` on constants. 2735``extractelement (VAL, IDX)`` 2736 Perform the :ref:`extractelement operation <i_extractelement>` on 2737 constants. 2738``insertelement (VAL, ELT, IDX)`` 2739 Perform the :ref:`insertelement operation <i_insertelement>` on 2740 constants. 2741``shufflevector (VEC1, VEC2, IDXMASK)`` 2742 Perform the :ref:`shufflevector operation <i_shufflevector>` on 2743 constants. 2744``extractvalue (VAL, IDX0, IDX1, ...)`` 2745 Perform the :ref:`extractvalue operation <i_extractvalue>` on 2746 constants. The index list is interpreted in a similar manner as 2747 indices in a ':ref:`getelementptr <i_getelementptr>`' operation. At 2748 least one index value must be specified. 2749``insertvalue (VAL, ELT, IDX0, IDX1, ...)`` 2750 Perform the :ref:`insertvalue operation <i_insertvalue>` on constants. 2751 The index list is interpreted in a similar manner as indices in a 2752 ':ref:`getelementptr <i_getelementptr>`' operation. At least one index 2753 value must be specified. 2754``OPCODE (LHS, RHS)`` 2755 Perform the specified operation of the LHS and RHS constants. OPCODE 2756 may be any of the :ref:`binary <binaryops>` or :ref:`bitwise 2757 binary <bitwiseops>` operations. The constraints on operands are 2758 the same as those for the corresponding instruction (e.g. no bitwise 2759 operations on floating point values are allowed). 2760 2761Other Values 2762============ 2763 2764.. _inlineasmexprs: 2765 2766Inline Assembler Expressions 2767---------------------------- 2768 2769LLVM supports inline assembler expressions (as opposed to :ref:`Module-Level 2770Inline Assembly <moduleasm>`) through the use of a special value. This 2771value represents the inline assembler as a string (containing the 2772instructions to emit), a list of operand constraints (stored as a 2773string), a flag that indicates whether or not the inline asm expression 2774has side effects, and a flag indicating whether the function containing 2775the asm needs to align its stack conservatively. An example inline 2776assembler expression is: 2777 2778.. code-block:: llvm 2779 2780 i32 (i32) asm "bswap $0", "=r,r" 2781 2782Inline assembler expressions may **only** be used as the callee operand 2783of a :ref:`call <i_call>` or an :ref:`invoke <i_invoke>` instruction. 2784Thus, typically we have: 2785 2786.. code-block:: llvm 2787 2788 %X = call i32 asm "bswap $0", "=r,r"(i32 %Y) 2789 2790Inline asms with side effects not visible in the constraint list must be 2791marked as having side effects. This is done through the use of the 2792'``sideeffect``' keyword, like so: 2793 2794.. code-block:: llvm 2795 2796 call void asm sideeffect "eieio", ""() 2797 2798In some cases inline asms will contain code that will not work unless 2799the stack is aligned in some way, such as calls or SSE instructions on 2800x86, yet will not contain code that does that alignment within the asm. 2801The compiler should make conservative assumptions about what the asm 2802might contain and should generate its usual stack alignment code in the 2803prologue if the '``alignstack``' keyword is present: 2804 2805.. code-block:: llvm 2806 2807 call void asm alignstack "eieio", ""() 2808 2809Inline asms also support using non-standard assembly dialects. The 2810assumed dialect is ATT. When the '``inteldialect``' keyword is present, 2811the inline asm is using the Intel dialect. Currently, ATT and Intel are 2812the only supported dialects. An example is: 2813 2814.. code-block:: llvm 2815 2816 call void asm inteldialect "eieio", ""() 2817 2818If multiple keywords appear the '``sideeffect``' keyword must come 2819first, the '``alignstack``' keyword second and the '``inteldialect``' 2820keyword last. 2821 2822Inline Asm Metadata 2823^^^^^^^^^^^^^^^^^^^ 2824 2825The call instructions that wrap inline asm nodes may have a 2826"``!srcloc``" MDNode attached to it that contains a list of constant 2827integers. If present, the code generator will use the integer as the 2828location cookie value when report errors through the ``LLVMContext`` 2829error reporting mechanisms. This allows a front-end to correlate backend 2830errors that occur with inline asm back to the source code that produced 2831it. For example: 2832 2833.. code-block:: llvm 2834 2835 call void asm sideeffect "something bad", ""(), !srcloc !42 2836 ... 2837 !42 = !{ i32 1234567 } 2838 2839It is up to the front-end to make sense of the magic numbers it places 2840in the IR. If the MDNode contains multiple constants, the code generator 2841will use the one that corresponds to the line of the asm that the error 2842occurs on. 2843 2844.. _metadata: 2845 2846Metadata 2847======== 2848 2849LLVM IR allows metadata to be attached to instructions in the program 2850that can convey extra information about the code to the optimizers and 2851code generator. One example application of metadata is source-level 2852debug information. There are two metadata primitives: strings and nodes. 2853 2854Metadata does not have a type, and is not a value. If referenced from a 2855``call`` instruction, it uses the ``metadata`` type. 2856 2857All metadata are identified in syntax by a exclamation point ('``!``'). 2858 2859.. _metadata-string: 2860 2861Metadata Nodes and Metadata Strings 2862----------------------------------- 2863 2864A metadata string is a string surrounded by double quotes. It can 2865contain any character by escaping non-printable characters with 2866"``\xx``" where "``xx``" is the two digit hex code. For example: 2867"``!"test\00"``". 2868 2869Metadata nodes are represented with notation similar to structure 2870constants (a comma separated list of elements, surrounded by braces and 2871preceded by an exclamation point). Metadata nodes can have any values as 2872their operand. For example: 2873 2874.. code-block:: llvm 2875 2876 !{ !"test\00", i32 10} 2877 2878Metadata nodes that aren't uniqued use the ``distinct`` keyword. For example: 2879 2880.. code-block:: llvm 2881 2882 !0 = distinct !{!"test\00", i32 10} 2883 2884``distinct`` nodes are useful when nodes shouldn't be merged based on their 2885content. They can also occur when transformations cause uniquing collisions 2886when metadata operands change. 2887 2888A :ref:`named metadata <namedmetadatastructure>` is a collection of 2889metadata nodes, which can be looked up in the module symbol table. For 2890example: 2891 2892.. code-block:: llvm 2893 2894 !foo = !{!4, !3} 2895 2896Metadata can be used as function arguments. Here ``llvm.dbg.value`` 2897function is using two metadata arguments: 2898 2899.. code-block:: llvm 2900 2901 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25) 2902 2903Metadata can be attached with an instruction. Here metadata ``!21`` is 2904attached to the ``add`` instruction using the ``!dbg`` identifier: 2905 2906.. code-block:: llvm 2907 2908 %indvar.next = add i64 %indvar, 1, !dbg !21 2909 2910More information about specific metadata nodes recognized by the 2911optimizers and code generator is found below. 2912 2913.. _specialized-metadata: 2914 2915Specialized Metadata Nodes 2916^^^^^^^^^^^^^^^^^^^^^^^^^^ 2917 2918Specialized metadata nodes are custom data structures in metadata (as opposed 2919to generic tuples). Their fields are labelled, and can be specified in any 2920order. 2921 2922These aren't inherently debug info centric, but currently all the specialized 2923metadata nodes are related to debug info. 2924 2925.. _MDCompileUnit: 2926 2927MDCompileUnit 2928""""""""""""" 2929 2930``MDCompileUnit`` nodes represent a compile unit. The ``enums:``, 2931``retainedTypes:``, ``subprograms:``, ``globals:`` and ``imports:`` fields are 2932tuples containing the debug info to be emitted along with the compile unit, 2933regardless of code optimizations (some nodes are only emitted if there are 2934references to them from instructions). 2935 2936.. code-block:: llvm 2937 2938 !0 = !MDCompileUnit(language: DW_LANG_C99, file: !1, producer: "clang", 2939 isOptimized: true, flags: "-O2", runtimeVersion: 2, 2940 splitDebugFilename: "abc.debug", emissionKind: 1, 2941 enums: !2, retainedTypes: !3, subprograms: !4, 2942 globals: !5, imports: !6) 2943 2944Compile unit descriptors provide the root scope for objects declared in a 2945specific compilation unit. File descriptors are defined using this scope. 2946These descriptors are collected by a named metadata ``!llvm.dbg.cu``. They 2947keep track of subprograms, global variables, type information, and imported 2948entities (declarations and namespaces). 2949 2950.. _MDFile: 2951 2952MDFile 2953"""""" 2954 2955``MDFile`` nodes represent files. The ``filename:`` can include slashes. 2956 2957.. code-block:: llvm 2958 2959 !0 = !MDFile(filename: "path/to/file", directory: "/path/to/dir") 2960 2961Files are sometimes used in ``scope:`` fields, and are the only valid target 2962for ``file:`` fields. 2963 2964.. _MDLocation: 2965 2966MDBasicType 2967""""""""""" 2968 2969``MDBasicType`` nodes represent primitive types, such as ``int``, ``bool`` and 2970``float``. ``tag:`` defaults to ``DW_TAG_base_type``. 2971 2972.. code-block:: llvm 2973 2974 !0 = !MDBasicType(name: "unsigned char", size: 8, align: 8, 2975 encoding: DW_ATE_unsigned_char) 2976 !1 = !MDBasicType(tag: DW_TAG_unspecified_type, name: "decltype(nullptr)") 2977 2978The ``encoding:`` describes the details of the type. Usually it's one of the 2979following: 2980 2981.. code-block:: llvm 2982 2983 DW_ATE_address = 1 2984 DW_ATE_boolean = 2 2985 DW_ATE_float = 4 2986 DW_ATE_signed = 5 2987 DW_ATE_signed_char = 6 2988 DW_ATE_unsigned = 7 2989 DW_ATE_unsigned_char = 8 2990 2991.. _MDSubroutineType: 2992 2993MDSubroutineType 2994"""""""""""""""" 2995 2996``MDSubroutineType`` nodes represent subroutine types. Their ``types:`` field 2997refers to a tuple; the first operand is the return type, while the rest are the 2998types of the formal arguments in order. If the first operand is ``null``, that 2999represents a function with no return value (such as ``void foo() {}`` in C++). 3000 3001.. code-block:: llvm 3002 3003 !0 = !BasicType(name: "int", size: 32, align: 32, DW_ATE_signed) 3004 !1 = !BasicType(name: "char", size: 8, align: 8, DW_ATE_signed_char) 3005 !2 = !MDSubroutineType(types: !{null, !0, !1}) ; void (int, char) 3006 3007.. _MDDerivedType: 3008 3009MDDerivedType 3010""""""""""""" 3011 3012``MDDerivedType`` nodes represent types derived from other types, such as 3013qualified types. 3014 3015.. code-block:: llvm 3016 3017 !0 = !MDBasicType(name: "unsigned char", size: 8, align: 8, 3018 encoding: DW_ATE_unsigned_char) 3019 !1 = !MDDerivedType(tag: DW_TAG_pointer_type, baseType: !0, size: 32, 3020 align: 32) 3021 3022The following ``tag:`` values are valid: 3023 3024.. code-block:: llvm 3025 3026 DW_TAG_formal_parameter = 5 3027 DW_TAG_member = 13 3028 DW_TAG_pointer_type = 15 3029 DW_TAG_reference_type = 16 3030 DW_TAG_typedef = 22 3031 DW_TAG_ptr_to_member_type = 31 3032 DW_TAG_const_type = 38 3033 DW_TAG_volatile_type = 53 3034 DW_TAG_restrict_type = 55 3035 3036``DW_TAG_member`` is used to define a member of a :ref:`composite type 3037<MDCompositeType>` or :ref:`subprogram <MDSubprogram>`. The type of the member 3038is the ``baseType:``. The ``offset:`` is the member's bit offset. 3039``DW_TAG_formal_parameter`` is used to define a member which is a formal 3040argument of a subprogram. 3041 3042``DW_TAG_typedef`` is used to provide a name for the ``baseType:``. 3043 3044``DW_TAG_pointer_type``, ``DW_TAG_reference_type``, ``DW_TAG_const_type``, 3045``DW_TAG_volatile_type`` and ``DW_TAG_restrict_type`` are used to qualify the 3046``baseType:``. 3047 3048Note that the ``void *`` type is expressed as a type derived from NULL. 3049 3050.. _MDCompositeType: 3051 3052MDCompositeType 3053""""""""""""""" 3054 3055``MDCompositeType`` nodes represent types composed of other types, like 3056structures and unions. ``elements:`` points to a tuple of the composed types. 3057 3058If the source language supports ODR, the ``identifier:`` field gives the unique 3059identifier used for type merging between modules. When specified, other types 3060can refer to composite types indirectly via a :ref:`metadata string 3061<metadata-string>` that matches their identifier. 3062 3063.. code-block:: llvm 3064 3065 !0 = !MDEnumerator(name: "SixKind", value: 7) 3066 !1 = !MDEnumerator(name: "SevenKind", value: 7) 3067 !2 = !MDEnumerator(name: "NegEightKind", value: -8) 3068 !3 = !MDCompositeType(tag: DW_TAG_enumeration_type, name: "Enum", file: !12, 3069 line: 2, size: 32, align: 32, identifier: "_M4Enum", 3070 elements: !{!0, !1, !2}) 3071 3072The following ``tag:`` values are valid: 3073 3074.. code-block:: llvm 3075 3076 DW_TAG_array_type = 1 3077 DW_TAG_class_type = 2 3078 DW_TAG_enumeration_type = 4 3079 DW_TAG_structure_type = 19 3080 DW_TAG_union_type = 23 3081 DW_TAG_subroutine_type = 21 3082 DW_TAG_inheritance = 28 3083 3084 3085For ``DW_TAG_array_type``, the ``elements:`` should be :ref:`subrange 3086descriptors <MDSubrange>`, each representing the range of subscripts at that 3087level of indexing. The ``DIFlagVector`` flag to ``flags:`` indicates that an 3088array type is a native packed vector. 3089 3090For ``DW_TAG_enumeration_type``, the ``elements:`` should be :ref:`enumerator 3091descriptors <MDEnumerator>`, each representing the definition of an enumeration 3092value for the set. All enumeration type descriptors are collected in the 3093``enums:`` field of the :ref:`compile unit <MDCompileUnit>`. 3094 3095For ``DW_TAG_structure_type``, ``DW_TAG_class_type``, and 3096``DW_TAG_union_type``, the ``elements:`` should be :ref:`derived types 3097<MDDerivedType>` with ``tag: DW_TAG_member`` or ``tag: DW_TAG_inheritance``. 3098 3099.. _MDSubrange: 3100 3101MDSubrange 3102"""""""""" 3103 3104``MDSubrange`` nodes are the elements for ``DW_TAG_array_type`` variants of 3105:ref:`MDCompositeType`. ``count: -1`` indicates an empty array. 3106 3107.. code-block:: llvm 3108 3109 !0 = !MDSubrange(count: 5, lowerBound: 0) ; array counting from 0 3110 !1 = !MDSubrange(count: 5, lowerBound: 1) ; array counting from 1 3111 !2 = !MDSubrange(count: -1) ; empty array. 3112 3113.. _MDEnumerator: 3114 3115MDEnumerator 3116"""""""""""" 3117 3118``MDEnumerator`` nodes are the elements for ``DW_TAG_enumeration_type`` 3119variants of :ref:`MDCompositeType`. 3120 3121.. code-block:: llvm 3122 3123 !0 = !MDEnumerator(name: "SixKind", value: 7) 3124 !1 = !MDEnumerator(name: "SevenKind", value: 7) 3125 !2 = !MDEnumerator(name: "NegEightKind", value: -8) 3126 3127MDTemplateTypeParameter 3128""""""""""""""""""""""" 3129 3130``MDTemplateTypeParameter`` nodes represent type parameters to generic source 3131language constructs. They are used (optionally) in :ref:`MDCompositeType` and 3132:ref:`MDSubprogram` ``templateParams:`` fields. 3133 3134.. code-block:: llvm 3135 3136 !0 = !MDTemplateTypeParameter(name: "Ty", type: !1) 3137 3138MDTemplateValueParameter 3139"""""""""""""""""""""""" 3140 3141``MDTemplateValueParameter`` nodes represent value parameters to generic source 3142language constructs. ``tag:`` defaults to ``DW_TAG_template_value_parameter``, 3143but if specified can also be set to ``DW_TAG_GNU_template_template_param`` or 3144``DW_TAG_GNU_template_param_pack``. They are used (optionally) in 3145:ref:`MDCompositeType` and :ref:`MDSubprogram` ``templateParams:`` fields. 3146 3147.. code-block:: llvm 3148 3149 !0 = !MDTemplateValueParameter(name: "Ty", type: !1, value: i32 7) 3150 3151MDNamespace 3152""""""""""" 3153 3154``MDNamespace`` nodes represent namespaces in the source language. 3155 3156.. code-block:: llvm 3157 3158 !0 = !MDNamespace(name: "myawesomeproject", scope: !1, file: !2, line: 7) 3159 3160MDGlobalVariable 3161"""""""""""""""" 3162 3163``MDGlobalVariable`` nodes represent global variables in the source language. 3164 3165.. code-block:: llvm 3166 3167 !0 = !MDGlobalVariable(name: "foo", linkageName: "foo", scope: !1, 3168 file: !2, line: 7, type: !3, isLocal: true, 3169 isDefinition: false, variable: i32* @foo, 3170 declaration: !4) 3171 3172All global variables should be referenced by the `globals:` field of a 3173:ref:`compile unit <MDCompileUnit>`. 3174 3175.. _MDSubprogram: 3176 3177MDSubprogram 3178"""""""""""" 3179 3180``MDSubprogram`` nodes represent functions from the source language. The 3181``variables:`` field points at :ref:`variables <MDLocalVariable>` that must be 3182retained, even if their IR counterparts are optimized out of the IR. The 3183``type:`` field must point at an :ref:`MDSubroutineType`. 3184 3185.. code-block:: llvm 3186 3187 !0 = !MDSubprogram(name: "foo", linkageName: "_Zfoov", scope: !1, 3188 file: !2, line: 7, type: !3, isLocal: true, 3189 isDefinition: false, scopeLine: 8, containingType: !4, 3190 virtuality: DW_VIRTUALITY_pure_virtual, virtualIndex: 10, 3191 flags: DIFlagPrototyped, isOptimized: true, 3192 function: void ()* @_Z3foov, 3193 templateParams: !5, declaration: !6, variables: !7) 3194 3195.. _MDLexicalBlock: 3196 3197MDLexicalBlock 3198"""""""""""""" 3199 3200``MDLexicalBlock`` nodes describe nested blocks within a :ref:`subprogram 3201<MDSubprogram>`. The line number and column numbers are used to dinstinguish 3202two lexical blocks at same depth. They are valid targets for ``scope:`` 3203fields. 3204 3205.. code-block:: llvm 3206 3207 !0 = distinct !MDLexicalBlock(scope: !1, file: !2, line: 7, column: 35) 3208 3209Usually lexical blocks are ``distinct`` to prevent node merging based on 3210operands. 3211 3212.. _MDLexicalBlockFile: 3213 3214MDLexicalBlockFile 3215"""""""""""""""""" 3216 3217``MDLexicalBlockFile`` nodes are used to discriminate between sections of a 3218:ref:`lexical block <MDLexicalBlock>`. The ``file:`` field can be changed to 3219indicate textual inclusion, or the ``discriminator:`` field can be used to 3220discriminate between control flow within a single block in the source language. 3221 3222.. code-block:: llvm 3223 3224 !0 = !MDLexicalBlock(scope: !3, file: !4, line: 7, column: 35) 3225 !1 = !MDLexicalBlockFile(scope: !0, file: !4, discriminator: 0) 3226 !2 = !MDLexicalBlockFile(scope: !0, file: !4, discriminator: 1) 3227 3228MDLocation 3229"""""""""" 3230 3231``MDLocation`` nodes represent source debug locations. The ``scope:`` field is 3232mandatory, and points at an :ref:`MDLexicalBlockFile`, an 3233:ref:`MDLexicalBlock`, or an :ref:`MDSubprogram`. 3234 3235.. code-block:: llvm 3236 3237 !0 = !MDLocation(line: 2900, column: 42, scope: !1, inlinedAt: !2) 3238 3239.. _MDLocalVariable: 3240 3241MDLocalVariable 3242""""""""""""""" 3243 3244``MDLocalVariable`` nodes represent local variables in the source language. 3245Instead of ``DW_TAG_variable``, they use LLVM-specific fake tags to 3246discriminate between local variables (``DW_TAG_auto_variable``) and subprogram 3247arguments (``DW_TAG_arg_variable``). In the latter case, the ``arg:`` field 3248specifies the argument position, and this variable will be included in the 3249``variables:`` field of its :ref:`MDSubprogram`. 3250 3251.. code-block:: llvm 3252 3253 !0 = !MDLocalVariable(tag: DW_TAG_arg_variable, name: "this", arg: 0, 3254 scope: !3, file: !2, line: 7, type: !3, 3255 flags: DIFlagArtificial) 3256 !1 = !MDLocalVariable(tag: DW_TAG_arg_variable, name: "x", arg: 1, 3257 scope: !4, file: !2, line: 7, type: !3) 3258 !1 = !MDLocalVariable(tag: DW_TAG_auto_variable, name: "y", 3259 scope: !5, file: !2, line: 7, type: !3) 3260 3261MDExpression 3262"""""""""""" 3263 3264``MDExpression`` nodes represent DWARF expression sequences. They are used in 3265:ref:`debug intrinsics<dbg_intrinsics>` (such as ``llvm.dbg.declare``) to 3266describe how the referenced LLVM variable relates to the source language 3267variable. 3268 3269The current supported vocabulary is limited: 3270 3271- ``DW_OP_deref`` dereferences the working expression. 3272- ``DW_OP_plus, 93`` adds ``93`` to the working expression. 3273- ``DW_OP_bit_piece, 16, 8`` specifies the offset and size (``16`` and ``8`` 3274 here, respectively) of the variable piece from the working expression. 3275 3276.. code-block:: llvm 3277 3278 !0 = !MDExpression(DW_OP_deref) 3279 !1 = !MDExpression(DW_OP_plus, 3) 3280 !2 = !MDExpression(DW_OP_bit_piece, 3, 7) 3281 !3 = !MDExpression(DW_OP_deref, DW_OP_plus, 3, DW_OP_bit_piece, 3, 7) 3282 3283MDObjCProperty 3284"""""""""""""" 3285 3286``MDObjCProperty`` nodes represent Objective-C property nodes. 3287 3288.. code-block:: llvm 3289 3290 !3 = !MDObjCProperty(name: "foo", file: !1, line: 7, setter: "setFoo", 3291 getter: "getFoo", attributes: 7, type: !2) 3292 3293MDImportedEntity 3294"""""""""""""""" 3295 3296``MDImportedEntity`` nodes represent entities (such as modules) imported into a 3297compile unit. 3298 3299.. code-block:: llvm 3300 3301 !2 = !MDImportedEntity(tag: DW_TAG_imported_module, name: "foo", scope: !0, 3302 entity: !1, line: 7) 3303 3304'``tbaa``' Metadata 3305^^^^^^^^^^^^^^^^^^^ 3306 3307In LLVM IR, memory does not have types, so LLVM's own type system is not 3308suitable for doing TBAA. Instead, metadata is added to the IR to 3309describe a type system of a higher level language. This can be used to 3310implement typical C/C++ TBAA, but it can also be used to implement 3311custom alias analysis behavior for other languages. 3312 3313The current metadata format is very simple. TBAA metadata nodes have up 3314to three fields, e.g.: 3315 3316.. code-block:: llvm 3317 3318 !0 = !{ !"an example type tree" } 3319 !1 = !{ !"int", !0 } 3320 !2 = !{ !"float", !0 } 3321 !3 = !{ !"const float", !2, i64 1 } 3322 3323The first field is an identity field. It can be any value, usually a 3324metadata string, which uniquely identifies the type. The most important 3325name in the tree is the name of the root node. Two trees with different 3326root node names are entirely disjoint, even if they have leaves with 3327common names. 3328 3329The second field identifies the type's parent node in the tree, or is 3330null or omitted for a root node. A type is considered to alias all of 3331its descendants and all of its ancestors in the tree. Also, a type is 3332considered to alias all types in other trees, so that bitcode produced 3333from multiple front-ends is handled conservatively. 3334 3335If the third field is present, it's an integer which if equal to 1 3336indicates that the type is "constant" (meaning 3337``pointsToConstantMemory`` should return true; see `other useful 3338AliasAnalysis methods <AliasAnalysis.html#OtherItfs>`_). 3339 3340'``tbaa.struct``' Metadata 3341^^^^^^^^^^^^^^^^^^^^^^^^^^ 3342 3343The :ref:`llvm.memcpy <int_memcpy>` is often used to implement 3344aggregate assignment operations in C and similar languages, however it 3345is defined to copy a contiguous region of memory, which is more than 3346strictly necessary for aggregate types which contain holes due to 3347padding. Also, it doesn't contain any TBAA information about the fields 3348of the aggregate. 3349 3350``!tbaa.struct`` metadata can describe which memory subregions in a 3351memcpy are padding and what the TBAA tags of the struct are. 3352 3353The current metadata format is very simple. ``!tbaa.struct`` metadata 3354nodes are a list of operands which are in conceptual groups of three. 3355For each group of three, the first operand gives the byte offset of a 3356field in bytes, the second gives its size in bytes, and the third gives 3357its tbaa tag. e.g.: 3358 3359.. code-block:: llvm 3360 3361 !4 = !{ i64 0, i64 4, !1, i64 8, i64 4, !2 } 3362 3363This describes a struct with two fields. The first is at offset 0 bytes 3364with size 4 bytes, and has tbaa tag !1. The second is at offset 8 bytes 3365and has size 4 bytes and has tbaa tag !2. 3366 3367Note that the fields need not be contiguous. In this example, there is a 33684 byte gap between the two fields. This gap represents padding which 3369does not carry useful data and need not be preserved. 3370 3371'``noalias``' and '``alias.scope``' Metadata 3372^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3373 3374``noalias`` and ``alias.scope`` metadata provide the ability to specify generic 3375noalias memory-access sets. This means that some collection of memory access 3376instructions (loads, stores, memory-accessing calls, etc.) that carry 3377``noalias`` metadata can specifically be specified not to alias with some other 3378collection of memory access instructions that carry ``alias.scope`` metadata. 3379Each type of metadata specifies a list of scopes where each scope has an id and 3380a domain. When evaluating an aliasing query, if for some domain, the set 3381of scopes with that domain in one instruction's ``alias.scope`` list is a 3382subset of (or equal to) the set of scopes for that domain in another 3383instruction's ``noalias`` list, then the two memory accesses are assumed not to 3384alias. 3385 3386The metadata identifying each domain is itself a list containing one or two 3387entries. The first entry is the name of the domain. Note that if the name is a 3388string then it can be combined accross functions and translation units. A 3389self-reference can be used to create globally unique domain names. A 3390descriptive string may optionally be provided as a second list entry. 3391 3392The metadata identifying each scope is also itself a list containing two or 3393three entries. The first entry is the name of the scope. Note that if the name 3394is a string then it can be combined accross functions and translation units. A 3395self-reference can be used to create globally unique scope names. A metadata 3396reference to the scope's domain is the second entry. A descriptive string may 3397optionally be provided as a third list entry. 3398 3399For example, 3400 3401.. code-block:: llvm 3402 3403 ; Two scope domains: 3404 !0 = !{!0} 3405 !1 = !{!1} 3406 3407 ; Some scopes in these domains: 3408 !2 = !{!2, !0} 3409 !3 = !{!3, !0} 3410 !4 = !{!4, !1} 3411 3412 ; Some scope lists: 3413 !5 = !{!4} ; A list containing only scope !4 3414 !6 = !{!4, !3, !2} 3415 !7 = !{!3} 3416 3417 ; These two instructions don't alias: 3418 %0 = load float, float* %c, align 4, !alias.scope !5 3419 store float %0, float* %arrayidx.i, align 4, !noalias !5 3420 3421 ; These two instructions also don't alias (for domain !1, the set of scopes 3422 ; in the !alias.scope equals that in the !noalias list): 3423 %2 = load float, float* %c, align 4, !alias.scope !5 3424 store float %2, float* %arrayidx.i2, align 4, !noalias !6 3425 3426 ; These two instructions don't alias (for domain !0, the set of scopes in 3427 ; the !noalias list is not a superset of, or equal to, the scopes in the 3428 ; !alias.scope list): 3429 %2 = load float, float* %c, align 4, !alias.scope !6 3430 store float %0, float* %arrayidx.i, align 4, !noalias !7 3431 3432'``fpmath``' Metadata 3433^^^^^^^^^^^^^^^^^^^^^ 3434 3435``fpmath`` metadata may be attached to any instruction of floating point 3436type. It can be used to express the maximum acceptable error in the 3437result of that instruction, in ULPs, thus potentially allowing the 3438compiler to use a more efficient but less accurate method of computing 3439it. ULP is defined as follows: 3440 3441 If ``x`` is a real number that lies between two finite consecutive 3442 floating-point numbers ``a`` and ``b``, without being equal to one 3443 of them, then ``ulp(x) = |b - a|``, otherwise ``ulp(x)`` is the 3444 distance between the two non-equal finite floating-point numbers 3445 nearest ``x``. Moreover, ``ulp(NaN)`` is ``NaN``. 3446 3447The metadata node shall consist of a single positive floating point 3448number representing the maximum relative error, for example: 3449 3450.. code-block:: llvm 3451 3452 !0 = !{ float 2.5 } ; maximum acceptable inaccuracy is 2.5 ULPs 3453 3454.. _range-metadata: 3455 3456'``range``' Metadata 3457^^^^^^^^^^^^^^^^^^^^ 3458 3459``range`` metadata may be attached only to ``load``, ``call`` and ``invoke`` of 3460integer types. It expresses the possible ranges the loaded value or the value 3461returned by the called function at this call site is in. The ranges are 3462represented with a flattened list of integers. The loaded value or the value 3463returned is known to be in the union of the ranges defined by each consecutive 3464pair. Each pair has the following properties: 3465 3466- The type must match the type loaded by the instruction. 3467- The pair ``a,b`` represents the range ``[a,b)``. 3468- Both ``a`` and ``b`` are constants. 3469- The range is allowed to wrap. 3470- The range should not represent the full or empty set. That is, 3471 ``a!=b``. 3472 3473In addition, the pairs must be in signed order of the lower bound and 3474they must be non-contiguous. 3475 3476Examples: 3477 3478.. code-block:: llvm 3479 3480 %a = load i8, i8* %x, align 1, !range !0 ; Can only be 0 or 1 3481 %b = load i8, i8* %y, align 1, !range !1 ; Can only be 255 (-1), 0 or 1 3482 %c = call i8 @foo(), !range !2 ; Can only be 0, 1, 3, 4 or 5 3483 %d = invoke i8 @bar() to label %cont 3484 unwind label %lpad, !range !3 ; Can only be -2, -1, 3, 4 or 5 3485 ... 3486 !0 = !{ i8 0, i8 2 } 3487 !1 = !{ i8 255, i8 2 } 3488 !2 = !{ i8 0, i8 2, i8 3, i8 6 } 3489 !3 = !{ i8 -2, i8 0, i8 3, i8 6 } 3490 3491'``llvm.loop``' 3492^^^^^^^^^^^^^^^ 3493 3494It is sometimes useful to attach information to loop constructs. Currently, 3495loop metadata is implemented as metadata attached to the branch instruction 3496in the loop latch block. This type of metadata refer to a metadata node that is 3497guaranteed to be separate for each loop. The loop identifier metadata is 3498specified with the name ``llvm.loop``. 3499 3500The loop identifier metadata is implemented using a metadata that refers to 3501itself to avoid merging it with any other identifier metadata, e.g., 3502during module linkage or function inlining. That is, each loop should refer 3503to their own identification metadata even if they reside in separate functions. 3504The following example contains loop identifier metadata for two separate loop 3505constructs: 3506 3507.. code-block:: llvm 3508 3509 !0 = !{!0} 3510 !1 = !{!1} 3511 3512The loop identifier metadata can be used to specify additional 3513per-loop metadata. Any operands after the first operand can be treated 3514as user-defined metadata. For example the ``llvm.loop.unroll.count`` 3515suggests an unroll factor to the loop unroller: 3516 3517.. code-block:: llvm 3518 3519 br i1 %exitcond, label %._crit_edge, label %.lr.ph, !llvm.loop !0 3520 ... 3521 !0 = !{!0, !1} 3522 !1 = !{!"llvm.loop.unroll.count", i32 4} 3523 3524'``llvm.loop.vectorize``' and '``llvm.loop.interleave``' 3525^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3526 3527Metadata prefixed with ``llvm.loop.vectorize`` or ``llvm.loop.interleave`` are 3528used to control per-loop vectorization and interleaving parameters such as 3529vectorization width and interleave count. These metadata should be used in 3530conjunction with ``llvm.loop`` loop identification metadata. The 3531``llvm.loop.vectorize`` and ``llvm.loop.interleave`` metadata are only 3532optimization hints and the optimizer will only interleave and vectorize loops if 3533it believes it is safe to do so. The ``llvm.mem.parallel_loop_access`` metadata 3534which contains information about loop-carried memory dependencies can be helpful 3535in determining the safety of these transformations. 3536 3537'``llvm.loop.interleave.count``' Metadata 3538^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3539 3540This metadata suggests an interleave count to the loop interleaver. 3541The first operand is the string ``llvm.loop.interleave.count`` and the 3542second operand is an integer specifying the interleave count. For 3543example: 3544 3545.. code-block:: llvm 3546 3547 !0 = !{!"llvm.loop.interleave.count", i32 4} 3548 3549Note that setting ``llvm.loop.interleave.count`` to 1 disables interleaving 3550multiple iterations of the loop. If ``llvm.loop.interleave.count`` is set to 0 3551then the interleave count will be determined automatically. 3552 3553'``llvm.loop.vectorize.enable``' Metadata 3554^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3555 3556This metadata selectively enables or disables vectorization for the loop. The 3557first operand is the string ``llvm.loop.vectorize.enable`` and the second operand 3558is a bit. If the bit operand value is 1 vectorization is enabled. A value of 35590 disables vectorization: 3560 3561.. code-block:: llvm 3562 3563 !0 = !{!"llvm.loop.vectorize.enable", i1 0} 3564 !1 = !{!"llvm.loop.vectorize.enable", i1 1} 3565 3566'``llvm.loop.vectorize.width``' Metadata 3567^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3568 3569This metadata sets the target width of the vectorizer. The first 3570operand is the string ``llvm.loop.vectorize.width`` and the second 3571operand is an integer specifying the width. For example: 3572 3573.. code-block:: llvm 3574 3575 !0 = !{!"llvm.loop.vectorize.width", i32 4} 3576 3577Note that setting ``llvm.loop.vectorize.width`` to 1 disables 3578vectorization of the loop. If ``llvm.loop.vectorize.width`` is set to 35790 or if the loop does not have this metadata the width will be 3580determined automatically. 3581 3582'``llvm.loop.unroll``' 3583^^^^^^^^^^^^^^^^^^^^^^ 3584 3585Metadata prefixed with ``llvm.loop.unroll`` are loop unrolling 3586optimization hints such as the unroll factor. ``llvm.loop.unroll`` 3587metadata should be used in conjunction with ``llvm.loop`` loop 3588identification metadata. The ``llvm.loop.unroll`` metadata are only 3589optimization hints and the unrolling will only be performed if the 3590optimizer believes it is safe to do so. 3591 3592'``llvm.loop.unroll.count``' Metadata 3593^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3594 3595This metadata suggests an unroll factor to the loop unroller. The 3596first operand is the string ``llvm.loop.unroll.count`` and the second 3597operand is a positive integer specifying the unroll factor. For 3598example: 3599 3600.. code-block:: llvm 3601 3602 !0 = !{!"llvm.loop.unroll.count", i32 4} 3603 3604If the trip count of the loop is less than the unroll count the loop 3605will be partially unrolled. 3606 3607'``llvm.loop.unroll.disable``' Metadata 3608^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3609 3610This metadata either disables loop unrolling. The metadata has a single operand 3611which is the string ``llvm.loop.unroll.disable``. For example: 3612 3613.. code-block:: llvm 3614 3615 !0 = !{!"llvm.loop.unroll.disable"} 3616 3617'``llvm.loop.unroll.runtime.disable``' Metadata 3618^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3619 3620This metadata either disables runtime loop unrolling. The metadata has a single 3621operand which is the string ``llvm.loop.unroll.runtime.disable``. For example: 3622 3623.. code-block:: llvm 3624 3625 !0 = !{!"llvm.loop.unroll.runtime.disable"} 3626 3627'``llvm.loop.unroll.full``' Metadata 3628^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3629 3630This metadata either suggests that the loop should be unrolled fully. The 3631metadata has a single operand which is the string ``llvm.loop.unroll.disable``. 3632For example: 3633 3634.. code-block:: llvm 3635 3636 !0 = !{!"llvm.loop.unroll.full"} 3637 3638'``llvm.mem``' 3639^^^^^^^^^^^^^^^ 3640 3641Metadata types used to annotate memory accesses with information helpful 3642for optimizations are prefixed with ``llvm.mem``. 3643 3644'``llvm.mem.parallel_loop_access``' Metadata 3645^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 3646 3647The ``llvm.mem.parallel_loop_access`` metadata refers to a loop identifier, 3648or metadata containing a list of loop identifiers for nested loops. 3649The metadata is attached to memory accessing instructions and denotes that 3650no loop carried memory dependence exist between it and other instructions denoted 3651with the same loop identifier. 3652 3653Precisely, given two instructions ``m1`` and ``m2`` that both have the 3654``llvm.mem.parallel_loop_access`` metadata, with ``L1`` and ``L2`` being the 3655set of loops associated with that metadata, respectively, then there is no loop 3656carried dependence between ``m1`` and ``m2`` for loops in both ``L1`` and 3657``L2``. 3658 3659As a special case, if all memory accessing instructions in a loop have 3660``llvm.mem.parallel_loop_access`` metadata that refers to that loop, then the 3661loop has no loop carried memory dependences and is considered to be a parallel 3662loop. 3663 3664Note that if not all memory access instructions have such metadata referring to 3665the loop, then the loop is considered not being trivially parallel. Additional 3666memory dependence analysis is required to make that determination. As a fail 3667safe mechanism, this causes loops that were originally parallel to be considered 3668sequential (if optimization passes that are unaware of the parallel semantics 3669insert new memory instructions into the loop body). 3670 3671Example of a loop that is considered parallel due to its correct use of 3672both ``llvm.loop`` and ``llvm.mem.parallel_loop_access`` 3673metadata types that refer to the same loop identifier metadata. 3674 3675.. code-block:: llvm 3676 3677 for.body: 3678 ... 3679 %val0 = load i32, i32* %arrayidx, !llvm.mem.parallel_loop_access !0 3680 ... 3681 store i32 %val0, i32* %arrayidx1, !llvm.mem.parallel_loop_access !0 3682 ... 3683 br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0 3684 3685 for.end: 3686 ... 3687 !0 = !{!0} 3688 3689It is also possible to have nested parallel loops. In that case the 3690memory accesses refer to a list of loop identifier metadata nodes instead of 3691the loop identifier metadata node directly: 3692 3693.. code-block:: llvm 3694 3695 outer.for.body: 3696 ... 3697 %val1 = load i32, i32* %arrayidx3, !llvm.mem.parallel_loop_access !2 3698 ... 3699 br label %inner.for.body 3700 3701 inner.for.body: 3702 ... 3703 %val0 = load i32, i32* %arrayidx1, !llvm.mem.parallel_loop_access !0 3704 ... 3705 store i32 %val0, i32* %arrayidx2, !llvm.mem.parallel_loop_access !0 3706 ... 3707 br i1 %exitcond, label %inner.for.end, label %inner.for.body, !llvm.loop !1 3708 3709 inner.for.end: 3710 ... 3711 store i32 %val1, i32* %arrayidx4, !llvm.mem.parallel_loop_access !2 3712 ... 3713 br i1 %exitcond, label %outer.for.end, label %outer.for.body, !llvm.loop !2 3714 3715 outer.for.end: ; preds = %for.body 3716 ... 3717 !0 = !{!1, !2} ; a list of loop identifiers 3718 !1 = !{!1} ; an identifier for the inner loop 3719 !2 = !{!2} ; an identifier for the outer loop 3720 3721'``llvm.bitsets``' 3722^^^^^^^^^^^^^^^^^^ 3723 3724The ``llvm.bitsets`` global metadata is used to implement 3725:doc:`bitsets <BitSets>`. 3726 3727Module Flags Metadata 3728===================== 3729 3730Information about the module as a whole is difficult to convey to LLVM's 3731subsystems. The LLVM IR isn't sufficient to transmit this information. 3732The ``llvm.module.flags`` named metadata exists in order to facilitate 3733this. These flags are in the form of key / value pairs --- much like a 3734dictionary --- making it easy for any subsystem who cares about a flag to 3735look it up. 3736 3737The ``llvm.module.flags`` metadata contains a list of metadata triplets. 3738Each triplet has the following form: 3739 3740- The first element is a *behavior* flag, which specifies the behavior 3741 when two (or more) modules are merged together, and it encounters two 3742 (or more) metadata with the same ID. The supported behaviors are 3743 described below. 3744- The second element is a metadata string that is a unique ID for the 3745 metadata. Each module may only have one flag entry for each unique ID (not 3746 including entries with the **Require** behavior). 3747- The third element is the value of the flag. 3748 3749When two (or more) modules are merged together, the resulting 3750``llvm.module.flags`` metadata is the union of the modules' flags. That is, for 3751each unique metadata ID string, there will be exactly one entry in the merged 3752modules ``llvm.module.flags`` metadata table, and the value for that entry will 3753be determined by the merge behavior flag, as described below. The only exception 3754is that entries with the *Require* behavior are always preserved. 3755 3756The following behaviors are supported: 3757 3758.. list-table:: 3759 :header-rows: 1 3760 :widths: 10 90 3761 3762 * - Value 3763 - Behavior 3764 3765 * - 1 3766 - **Error** 3767 Emits an error if two values disagree, otherwise the resulting value 3768 is that of the operands. 3769 3770 * - 2 3771 - **Warning** 3772 Emits a warning if two values disagree. The result value will be the 3773 operand for the flag from the first module being linked. 3774 3775 * - 3 3776 - **Require** 3777 Adds a requirement that another module flag be present and have a 3778 specified value after linking is performed. The value must be a 3779 metadata pair, where the first element of the pair is the ID of the 3780 module flag to be restricted, and the second element of the pair is 3781 the value the module flag should be restricted to. This behavior can 3782 be used to restrict the allowable results (via triggering of an 3783 error) of linking IDs with the **Override** behavior. 3784 3785 * - 4 3786 - **Override** 3787 Uses the specified value, regardless of the behavior or value of the 3788 other module. If both modules specify **Override**, but the values 3789 differ, an error will be emitted. 3790 3791 * - 5 3792 - **Append** 3793 Appends the two values, which are required to be metadata nodes. 3794 3795 * - 6 3796 - **AppendUnique** 3797 Appends the two values, which are required to be metadata 3798 nodes. However, duplicate entries in the second list are dropped 3799 during the append operation. 3800 3801It is an error for a particular unique flag ID to have multiple behaviors, 3802except in the case of **Require** (which adds restrictions on another metadata 3803value) or **Override**. 3804 3805An example of module flags: 3806 3807.. code-block:: llvm 3808 3809 !0 = !{ i32 1, !"foo", i32 1 } 3810 !1 = !{ i32 4, !"bar", i32 37 } 3811 !2 = !{ i32 2, !"qux", i32 42 } 3812 !3 = !{ i32 3, !"qux", 3813 !{ 3814 !"foo", i32 1 3815 } 3816 } 3817 !llvm.module.flags = !{ !0, !1, !2, !3 } 3818 3819- Metadata ``!0`` has the ID ``!"foo"`` and the value '1'. The behavior 3820 if two or more ``!"foo"`` flags are seen is to emit an error if their 3821 values are not equal. 3822 3823- Metadata ``!1`` has the ID ``!"bar"`` and the value '37'. The 3824 behavior if two or more ``!"bar"`` flags are seen is to use the value 3825 '37'. 3826 3827- Metadata ``!2`` has the ID ``!"qux"`` and the value '42'. The 3828 behavior if two or more ``!"qux"`` flags are seen is to emit a 3829 warning if their values are not equal. 3830 3831- Metadata ``!3`` has the ID ``!"qux"`` and the value: 3832 3833 :: 3834 3835 !{ !"foo", i32 1 } 3836 3837 The behavior is to emit an error if the ``llvm.module.flags`` does not 3838 contain a flag with the ID ``!"foo"`` that has the value '1' after linking is 3839 performed. 3840 3841Objective-C Garbage Collection Module Flags Metadata 3842---------------------------------------------------- 3843 3844On the Mach-O platform, Objective-C stores metadata about garbage 3845collection in a special section called "image info". The metadata 3846consists of a version number and a bitmask specifying what types of 3847garbage collection are supported (if any) by the file. If two or more 3848modules are linked together their garbage collection metadata needs to 3849be merged rather than appended together. 3850 3851The Objective-C garbage collection module flags metadata consists of the 3852following key-value pairs: 3853 3854.. list-table:: 3855 :header-rows: 1 3856 :widths: 30 70 3857 3858 * - Key 3859 - Value 3860 3861 * - ``Objective-C Version`` 3862 - **[Required]** --- The Objective-C ABI version. Valid values are 1 and 2. 3863 3864 * - ``Objective-C Image Info Version`` 3865 - **[Required]** --- The version of the image info section. Currently 3866 always 0. 3867 3868 * - ``Objective-C Image Info Section`` 3869 - **[Required]** --- The section to place the metadata. Valid values are 3870 ``"__OBJC, __image_info, regular"`` for Objective-C ABI version 1, and 3871 ``"__DATA,__objc_imageinfo, regular, no_dead_strip"`` for 3872 Objective-C ABI version 2. 3873 3874 * - ``Objective-C Garbage Collection`` 3875 - **[Required]** --- Specifies whether garbage collection is supported or 3876 not. Valid values are 0, for no garbage collection, and 2, for garbage 3877 collection supported. 3878 3879 * - ``Objective-C GC Only`` 3880 - **[Optional]** --- Specifies that only garbage collection is supported. 3881 If present, its value must be 6. This flag requires that the 3882 ``Objective-C Garbage Collection`` flag have the value 2. 3883 3884Some important flag interactions: 3885 3886- If a module with ``Objective-C Garbage Collection`` set to 0 is 3887 merged with a module with ``Objective-C Garbage Collection`` set to 3888 2, then the resulting module has the 3889 ``Objective-C Garbage Collection`` flag set to 0. 3890- A module with ``Objective-C Garbage Collection`` set to 0 cannot be 3891 merged with a module with ``Objective-C GC Only`` set to 6. 3892 3893Automatic Linker Flags Module Flags Metadata 3894-------------------------------------------- 3895 3896Some targets support embedding flags to the linker inside individual object 3897files. Typically this is used in conjunction with language extensions which 3898allow source files to explicitly declare the libraries they depend on, and have 3899these automatically be transmitted to the linker via object files. 3900 3901These flags are encoded in the IR using metadata in the module flags section, 3902using the ``Linker Options`` key. The merge behavior for this flag is required 3903to be ``AppendUnique``, and the value for the key is expected to be a metadata 3904node which should be a list of other metadata nodes, each of which should be a 3905list of metadata strings defining linker options. 3906 3907For example, the following metadata section specifies two separate sets of 3908linker options, presumably to link against ``libz`` and the ``Cocoa`` 3909framework:: 3910 3911 !0 = !{ i32 6, !"Linker Options", 3912 !{ 3913 !{ !"-lz" }, 3914 !{ !"-framework", !"Cocoa" } } } 3915 !llvm.module.flags = !{ !0 } 3916 3917The metadata encoding as lists of lists of options, as opposed to a collapsed 3918list of options, is chosen so that the IR encoding can use multiple option 3919strings to specify e.g., a single library, while still having that specifier be 3920preserved as an atomic element that can be recognized by a target specific 3921assembly writer or object file emitter. 3922 3923Each individual option is required to be either a valid option for the target's 3924linker, or an option that is reserved by the target specific assembly writer or 3925object file emitter. No other aspect of these options is defined by the IR. 3926 3927C type width Module Flags Metadata 3928---------------------------------- 3929 3930The ARM backend emits a section into each generated object file describing the 3931options that it was compiled with (in a compiler-independent way) to prevent 3932linking incompatible objects, and to allow automatic library selection. Some 3933of these options are not visible at the IR level, namely wchar_t width and enum 3934width. 3935 3936To pass this information to the backend, these options are encoded in module 3937flags metadata, using the following key-value pairs: 3938 3939.. list-table:: 3940 :header-rows: 1 3941 :widths: 30 70 3942 3943 * - Key 3944 - Value 3945 3946 * - short_wchar 3947 - * 0 --- sizeof(wchar_t) == 4 3948 * 1 --- sizeof(wchar_t) == 2 3949 3950 * - short_enum 3951 - * 0 --- Enums are at least as large as an ``int``. 3952 * 1 --- Enums are stored in the smallest integer type which can 3953 represent all of its values. 3954 3955For example, the following metadata section specifies that the module was 3956compiled with a ``wchar_t`` width of 4 bytes, and the underlying type of an 3957enum is the smallest type which can represent all of its values:: 3958 3959 !llvm.module.flags = !{!0, !1} 3960 !0 = !{i32 1, !"short_wchar", i32 1} 3961 !1 = !{i32 1, !"short_enum", i32 0} 3962 3963.. _intrinsicglobalvariables: 3964 3965Intrinsic Global Variables 3966========================== 3967 3968LLVM has a number of "magic" global variables that contain data that 3969affect code generation or other IR semantics. These are documented here. 3970All globals of this sort should have a section specified as 3971"``llvm.metadata``". This section and all globals that start with 3972"``llvm.``" are reserved for use by LLVM. 3973 3974.. _gv_llvmused: 3975 3976The '``llvm.used``' Global Variable 3977----------------------------------- 3978 3979The ``@llvm.used`` global is an array which has 3980:ref:`appending linkage <linkage_appending>`. This array contains a list of 3981pointers to named global variables, functions and aliases which may optionally 3982have a pointer cast formed of bitcast or getelementptr. For example, a legal 3983use of it is: 3984 3985.. code-block:: llvm 3986 3987 @X = global i8 4 3988 @Y = global i32 123 3989 3990 @llvm.used = appending global [2 x i8*] [ 3991 i8* @X, 3992 i8* bitcast (i32* @Y to i8*) 3993 ], section "llvm.metadata" 3994 3995If a symbol appears in the ``@llvm.used`` list, then the compiler, assembler, 3996and linker are required to treat the symbol as if there is a reference to the 3997symbol that it cannot see (which is why they have to be named). For example, if 3998a variable has internal linkage and no references other than that from the 3999``@llvm.used`` list, it cannot be deleted. This is commonly used to represent 4000references from inline asms and other things the compiler cannot "see", and 4001corresponds to "``attribute((used))``" in GNU C. 4002 4003On some targets, the code generator must emit a directive to the 4004assembler or object file to prevent the assembler and linker from 4005molesting the symbol. 4006 4007.. _gv_llvmcompilerused: 4008 4009The '``llvm.compiler.used``' Global Variable 4010-------------------------------------------- 4011 4012The ``@llvm.compiler.used`` directive is the same as the ``@llvm.used`` 4013directive, except that it only prevents the compiler from touching the 4014symbol. On targets that support it, this allows an intelligent linker to 4015optimize references to the symbol without being impeded as it would be 4016by ``@llvm.used``. 4017 4018This is a rare construct that should only be used in rare circumstances, 4019and should not be exposed to source languages. 4020 4021.. _gv_llvmglobalctors: 4022 4023The '``llvm.global_ctors``' Global Variable 4024------------------------------------------- 4025 4026.. code-block:: llvm 4027 4028 %0 = type { i32, void ()*, i8* } 4029 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor, i8* @data }] 4030 4031The ``@llvm.global_ctors`` array contains a list of constructor 4032functions, priorities, and an optional associated global or function. 4033The functions referenced by this array will be called in ascending order 4034of priority (i.e. lowest first) when the module is loaded. The order of 4035functions with the same priority is not defined. 4036 4037If the third field is present, non-null, and points to a global variable 4038or function, the initializer function will only run if the associated 4039data from the current module is not discarded. 4040 4041.. _llvmglobaldtors: 4042 4043The '``llvm.global_dtors``' Global Variable 4044------------------------------------------- 4045 4046.. code-block:: llvm 4047 4048 %0 = type { i32, void ()*, i8* } 4049 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor, i8* @data }] 4050 4051The ``@llvm.global_dtors`` array contains a list of destructor 4052functions, priorities, and an optional associated global or function. 4053The functions referenced by this array will be called in descending 4054order of priority (i.e. highest first) when the module is unloaded. The 4055order of functions with the same priority is not defined. 4056 4057If the third field is present, non-null, and points to a global variable 4058or function, the destructor function will only run if the associated 4059data from the current module is not discarded. 4060 4061Instruction Reference 4062===================== 4063 4064The LLVM instruction set consists of several different classifications 4065of instructions: :ref:`terminator instructions <terminators>`, :ref:`binary 4066instructions <binaryops>`, :ref:`bitwise binary 4067instructions <bitwiseops>`, :ref:`memory instructions <memoryops>`, and 4068:ref:`other instructions <otherops>`. 4069 4070.. _terminators: 4071 4072Terminator Instructions 4073----------------------- 4074 4075As mentioned :ref:`previously <functionstructure>`, every basic block in a 4076program ends with a "Terminator" instruction, which indicates which 4077block should be executed after the current block is finished. These 4078terminator instructions typically yield a '``void``' value: they produce 4079control flow, not values (the one exception being the 4080':ref:`invoke <i_invoke>`' instruction). 4081 4082The terminator instructions are: ':ref:`ret <i_ret>`', 4083':ref:`br <i_br>`', ':ref:`switch <i_switch>`', 4084':ref:`indirectbr <i_indirectbr>`', ':ref:`invoke <i_invoke>`', 4085':ref:`resume <i_resume>`', and ':ref:`unreachable <i_unreachable>`'. 4086 4087.. _i_ret: 4088 4089'``ret``' Instruction 4090^^^^^^^^^^^^^^^^^^^^^ 4091 4092Syntax: 4093""""""" 4094 4095:: 4096 4097 ret <type> <value> ; Return a value from a non-void function 4098 ret void ; Return from void function 4099 4100Overview: 4101""""""""" 4102 4103The '``ret``' instruction is used to return control flow (and optionally 4104a value) from a function back to the caller. 4105 4106There are two forms of the '``ret``' instruction: one that returns a 4107value and then causes control flow, and one that just causes control 4108flow to occur. 4109 4110Arguments: 4111"""""""""" 4112 4113The '``ret``' instruction optionally accepts a single argument, the 4114return value. The type of the return value must be a ':ref:`first 4115class <t_firstclass>`' type. 4116 4117A function is not :ref:`well formed <wellformed>` if it it has a non-void 4118return type and contains a '``ret``' instruction with no return value or 4119a return value with a type that does not match its type, or if it has a 4120void return type and contains a '``ret``' instruction with a return 4121value. 4122 4123Semantics: 4124"""""""""" 4125 4126When the '``ret``' instruction is executed, control flow returns back to 4127the calling function's context. If the caller is a 4128":ref:`call <i_call>`" instruction, execution continues at the 4129instruction after the call. If the caller was an 4130":ref:`invoke <i_invoke>`" instruction, execution continues at the 4131beginning of the "normal" destination block. If the instruction returns 4132a value, that value shall set the call or invoke instruction's return 4133value. 4134 4135Example: 4136"""""""" 4137 4138.. code-block:: llvm 4139 4140 ret i32 5 ; Return an integer value of 5 4141 ret void ; Return from a void function 4142 ret { i32, i8 } { i32 4, i8 2 } ; Return a struct of values 4 and 2 4143 4144.. _i_br: 4145 4146'``br``' Instruction 4147^^^^^^^^^^^^^^^^^^^^ 4148 4149Syntax: 4150""""""" 4151 4152:: 4153 4154 br i1 <cond>, label <iftrue>, label <iffalse> 4155 br label <dest> ; Unconditional branch 4156 4157Overview: 4158""""""""" 4159 4160The '``br``' instruction is used to cause control flow to transfer to a 4161different basic block in the current function. There are two forms of 4162this instruction, corresponding to a conditional branch and an 4163unconditional branch. 4164 4165Arguments: 4166"""""""""" 4167 4168The conditional branch form of the '``br``' instruction takes a single 4169'``i1``' value and two '``label``' values. The unconditional form of the 4170'``br``' instruction takes a single '``label``' value as a target. 4171 4172Semantics: 4173"""""""""" 4174 4175Upon execution of a conditional '``br``' instruction, the '``i1``' 4176argument is evaluated. If the value is ``true``, control flows to the 4177'``iftrue``' ``label`` argument. If "cond" is ``false``, control flows 4178to the '``iffalse``' ``label`` argument. 4179 4180Example: 4181"""""""" 4182 4183.. code-block:: llvm 4184 4185 Test: 4186 %cond = icmp eq i32 %a, %b 4187 br i1 %cond, label %IfEqual, label %IfUnequal 4188 IfEqual: 4189 ret i32 1 4190 IfUnequal: 4191 ret i32 0 4192 4193.. _i_switch: 4194 4195'``switch``' Instruction 4196^^^^^^^^^^^^^^^^^^^^^^^^ 4197 4198Syntax: 4199""""""" 4200 4201:: 4202 4203 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ] 4204 4205Overview: 4206""""""""" 4207 4208The '``switch``' instruction is used to transfer control flow to one of 4209several different places. It is a generalization of the '``br``' 4210instruction, allowing a branch to occur to one of many possible 4211destinations. 4212 4213Arguments: 4214"""""""""" 4215 4216The '``switch``' instruction uses three parameters: an integer 4217comparison value '``value``', a default '``label``' destination, and an 4218array of pairs of comparison value constants and '``label``'s. The table 4219is not allowed to contain duplicate constant entries. 4220 4221Semantics: 4222"""""""""" 4223 4224The ``switch`` instruction specifies a table of values and destinations. 4225When the '``switch``' instruction is executed, this table is searched 4226for the given value. If the value is found, control flow is transferred 4227to the corresponding destination; otherwise, control flow is transferred 4228to the default destination. 4229 4230Implementation: 4231""""""""""""""" 4232 4233Depending on properties of the target machine and the particular 4234``switch`` instruction, this instruction may be code generated in 4235different ways. For example, it could be generated as a series of 4236chained conditional branches or with a lookup table. 4237 4238Example: 4239"""""""" 4240 4241.. code-block:: llvm 4242 4243 ; Emulate a conditional br instruction 4244 %Val = zext i1 %value to i32 4245 switch i32 %Val, label %truedest [ i32 0, label %falsedest ] 4246 4247 ; Emulate an unconditional br instruction 4248 switch i32 0, label %dest [ ] 4249 4250 ; Implement a jump table: 4251 switch i32 %val, label %otherwise [ i32 0, label %onzero 4252 i32 1, label %onone 4253 i32 2, label %ontwo ] 4254 4255.. _i_indirectbr: 4256 4257'``indirectbr``' Instruction 4258^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 4259 4260Syntax: 4261""""""" 4262 4263:: 4264 4265 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ] 4266 4267Overview: 4268""""""""" 4269 4270The '``indirectbr``' instruction implements an indirect branch to a 4271label within the current function, whose address is specified by 4272"``address``". Address must be derived from a 4273:ref:`blockaddress <blockaddress>` constant. 4274 4275Arguments: 4276"""""""""" 4277 4278The '``address``' argument is the address of the label to jump to. The 4279rest of the arguments indicate the full set of possible destinations 4280that the address may point to. Blocks are allowed to occur multiple 4281times in the destination list, though this isn't particularly useful. 4282 4283This destination list is required so that dataflow analysis has an 4284accurate understanding of the CFG. 4285 4286Semantics: 4287"""""""""" 4288 4289Control transfers to the block specified in the address argument. All 4290possible destination blocks must be listed in the label list, otherwise 4291this instruction has undefined behavior. This implies that jumps to 4292labels defined in other functions have undefined behavior as well. 4293 4294Implementation: 4295""""""""""""""" 4296 4297This is typically implemented with a jump through a register. 4298 4299Example: 4300"""""""" 4301 4302.. code-block:: llvm 4303 4304 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ] 4305 4306.. _i_invoke: 4307 4308'``invoke``' Instruction 4309^^^^^^^^^^^^^^^^^^^^^^^^ 4310 4311Syntax: 4312""""""" 4313 4314:: 4315 4316 <result> = invoke [cconv] [ret attrs] <ptr to function ty> <function ptr val>(<function args>) [fn attrs] 4317 to label <normal label> unwind label <exception label> 4318 4319Overview: 4320""""""""" 4321 4322The '``invoke``' instruction causes control to transfer to a specified 4323function, with the possibility of control flow transfer to either the 4324'``normal``' label or the '``exception``' label. If the callee function 4325returns with the "``ret``" instruction, control flow will return to the 4326"normal" label. If the callee (or any indirect callees) returns via the 4327":ref:`resume <i_resume>`" instruction or other exception handling 4328mechanism, control is interrupted and continued at the dynamically 4329nearest "exception" label. 4330 4331The '``exception``' label is a `landing 4332pad <ExceptionHandling.html#overview>`_ for the exception. As such, 4333'``exception``' label is required to have the 4334":ref:`landingpad <i_landingpad>`" instruction, which contains the 4335information about the behavior of the program after unwinding happens, 4336as its first non-PHI instruction. The restrictions on the 4337"``landingpad``" instruction's tightly couples it to the "``invoke``" 4338instruction, so that the important information contained within the 4339"``landingpad``" instruction can't be lost through normal code motion. 4340 4341Arguments: 4342"""""""""" 4343 4344This instruction requires several arguments: 4345 4346#. The optional "cconv" marker indicates which :ref:`calling 4347 convention <callingconv>` the call should use. If none is 4348 specified, the call defaults to using C calling conventions. 4349#. The optional :ref:`Parameter Attributes <paramattrs>` list for return 4350 values. Only '``zeroext``', '``signext``', and '``inreg``' attributes 4351 are valid here. 4352#. '``ptr to function ty``': shall be the signature of the pointer to 4353 function value being invoked. In most cases, this is a direct 4354 function invocation, but indirect ``invoke``'s are just as possible, 4355 branching off an arbitrary pointer to function value. 4356#. '``function ptr val``': An LLVM value containing a pointer to a 4357 function to be invoked. 4358#. '``function args``': argument list whose types match the function 4359 signature argument types and parameter attributes. All arguments must 4360 be of :ref:`first class <t_firstclass>` type. If the function signature 4361 indicates the function accepts a variable number of arguments, the 4362 extra arguments can be specified. 4363#. '``normal label``': the label reached when the called function 4364 executes a '``ret``' instruction. 4365#. '``exception label``': the label reached when a callee returns via 4366 the :ref:`resume <i_resume>` instruction or other exception handling 4367 mechanism. 4368#. The optional :ref:`function attributes <fnattrs>` list. Only 4369 '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``' 4370 attributes are valid here. 4371 4372Semantics: 4373"""""""""" 4374 4375This instruction is designed to operate as a standard '``call``' 4376instruction in most regards. The primary difference is that it 4377establishes an association with a label, which is used by the runtime 4378library to unwind the stack. 4379 4380This instruction is used in languages with destructors to ensure that 4381proper cleanup is performed in the case of either a ``longjmp`` or a 4382thrown exception. Additionally, this is important for implementation of 4383'``catch``' clauses in high-level languages that support them. 4384 4385For the purposes of the SSA form, the definition of the value returned 4386by the '``invoke``' instruction is deemed to occur on the edge from the 4387current block to the "normal" label. If the callee unwinds then no 4388return value is available. 4389 4390Example: 4391"""""""" 4392 4393.. code-block:: llvm 4394 4395 %retval = invoke i32 @Test(i32 15) to label %Continue 4396 unwind label %TestCleanup ; i32:retval set 4397 %retval = invoke coldcc i32 %Testfnptr(i32 15) to label %Continue 4398 unwind label %TestCleanup ; i32:retval set 4399 4400.. _i_resume: 4401 4402'``resume``' Instruction 4403^^^^^^^^^^^^^^^^^^^^^^^^ 4404 4405Syntax: 4406""""""" 4407 4408:: 4409 4410 resume <type> <value> 4411 4412Overview: 4413""""""""" 4414 4415The '``resume``' instruction is a terminator instruction that has no 4416successors. 4417 4418Arguments: 4419"""""""""" 4420 4421The '``resume``' instruction requires one argument, which must have the 4422same type as the result of any '``landingpad``' instruction in the same 4423function. 4424 4425Semantics: 4426"""""""""" 4427 4428The '``resume``' instruction resumes propagation of an existing 4429(in-flight) exception whose unwinding was interrupted with a 4430:ref:`landingpad <i_landingpad>` instruction. 4431 4432Example: 4433"""""""" 4434 4435.. code-block:: llvm 4436 4437 resume { i8*, i32 } %exn 4438 4439.. _i_unreachable: 4440 4441'``unreachable``' Instruction 4442^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 4443 4444Syntax: 4445""""""" 4446 4447:: 4448 4449 unreachable 4450 4451Overview: 4452""""""""" 4453 4454The '``unreachable``' instruction has no defined semantics. This 4455instruction is used to inform the optimizer that a particular portion of 4456the code is not reachable. This can be used to indicate that the code 4457after a no-return function cannot be reached, and other facts. 4458 4459Semantics: 4460"""""""""" 4461 4462The '``unreachable``' instruction has no defined semantics. 4463 4464.. _binaryops: 4465 4466Binary Operations 4467----------------- 4468 4469Binary operators are used to do most of the computation in a program. 4470They require two operands of the same type, execute an operation on 4471them, and produce a single value. The operands might represent multiple 4472data, as is the case with the :ref:`vector <t_vector>` data type. The 4473result value has the same type as its operands. 4474 4475There are several different binary operators: 4476 4477.. _i_add: 4478 4479'``add``' Instruction 4480^^^^^^^^^^^^^^^^^^^^^ 4481 4482Syntax: 4483""""""" 4484 4485:: 4486 4487 <result> = add <ty> <op1>, <op2> ; yields ty:result 4488 <result> = add nuw <ty> <op1>, <op2> ; yields ty:result 4489 <result> = add nsw <ty> <op1>, <op2> ; yields ty:result 4490 <result> = add nuw nsw <ty> <op1>, <op2> ; yields ty:result 4491 4492Overview: 4493""""""""" 4494 4495The '``add``' instruction returns the sum of its two operands. 4496 4497Arguments: 4498"""""""""" 4499 4500The two arguments to the '``add``' instruction must be 4501:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 4502arguments must have identical types. 4503 4504Semantics: 4505"""""""""" 4506 4507The value produced is the integer sum of the two operands. 4508 4509If the sum has unsigned overflow, the result returned is the 4510mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of 4511the result. 4512 4513Because LLVM integers use a two's complement representation, this 4514instruction is appropriate for both signed and unsigned integers. 4515 4516``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap", 4517respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the 4518result value of the ``add`` is a :ref:`poison value <poisonvalues>` if 4519unsigned and/or signed overflow, respectively, occurs. 4520 4521Example: 4522"""""""" 4523 4524.. code-block:: llvm 4525 4526 <result> = add i32 4, %var ; yields i32:result = 4 + %var 4527 4528.. _i_fadd: 4529 4530'``fadd``' Instruction 4531^^^^^^^^^^^^^^^^^^^^^^ 4532 4533Syntax: 4534""""""" 4535 4536:: 4537 4538 <result> = fadd [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result 4539 4540Overview: 4541""""""""" 4542 4543The '``fadd``' instruction returns the sum of its two operands. 4544 4545Arguments: 4546"""""""""" 4547 4548The two arguments to the '``fadd``' instruction must be :ref:`floating 4549point <t_floating>` or :ref:`vector <t_vector>` of floating point values. 4550Both arguments must have identical types. 4551 4552Semantics: 4553"""""""""" 4554 4555The value produced is the floating point sum of the two operands. This 4556instruction can also take any number of :ref:`fast-math flags <fastmath>`, 4557which are optimization hints to enable otherwise unsafe floating point 4558optimizations: 4559 4560Example: 4561"""""""" 4562 4563.. code-block:: llvm 4564 4565 <result> = fadd float 4.0, %var ; yields float:result = 4.0 + %var 4566 4567'``sub``' Instruction 4568^^^^^^^^^^^^^^^^^^^^^ 4569 4570Syntax: 4571""""""" 4572 4573:: 4574 4575 <result> = sub <ty> <op1>, <op2> ; yields ty:result 4576 <result> = sub nuw <ty> <op1>, <op2> ; yields ty:result 4577 <result> = sub nsw <ty> <op1>, <op2> ; yields ty:result 4578 <result> = sub nuw nsw <ty> <op1>, <op2> ; yields ty:result 4579 4580Overview: 4581""""""""" 4582 4583The '``sub``' instruction returns the difference of its two operands. 4584 4585Note that the '``sub``' instruction is used to represent the '``neg``' 4586instruction present in most other intermediate representations. 4587 4588Arguments: 4589"""""""""" 4590 4591The two arguments to the '``sub``' instruction must be 4592:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 4593arguments must have identical types. 4594 4595Semantics: 4596"""""""""" 4597 4598The value produced is the integer difference of the two operands. 4599 4600If the difference has unsigned overflow, the result returned is the 4601mathematical result modulo 2\ :sup:`n`\ , where n is the bit width of 4602the result. 4603 4604Because LLVM integers use a two's complement representation, this 4605instruction is appropriate for both signed and unsigned integers. 4606 4607``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap", 4608respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the 4609result value of the ``sub`` is a :ref:`poison value <poisonvalues>` if 4610unsigned and/or signed overflow, respectively, occurs. 4611 4612Example: 4613"""""""" 4614 4615.. code-block:: llvm 4616 4617 <result> = sub i32 4, %var ; yields i32:result = 4 - %var 4618 <result> = sub i32 0, %val ; yields i32:result = -%var 4619 4620.. _i_fsub: 4621 4622'``fsub``' Instruction 4623^^^^^^^^^^^^^^^^^^^^^^ 4624 4625Syntax: 4626""""""" 4627 4628:: 4629 4630 <result> = fsub [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result 4631 4632Overview: 4633""""""""" 4634 4635The '``fsub``' instruction returns the difference of its two operands. 4636 4637Note that the '``fsub``' instruction is used to represent the '``fneg``' 4638instruction present in most other intermediate representations. 4639 4640Arguments: 4641"""""""""" 4642 4643The two arguments to the '``fsub``' instruction must be :ref:`floating 4644point <t_floating>` or :ref:`vector <t_vector>` of floating point values. 4645Both arguments must have identical types. 4646 4647Semantics: 4648"""""""""" 4649 4650The value produced is the floating point difference of the two operands. 4651This instruction can also take any number of :ref:`fast-math 4652flags <fastmath>`, which are optimization hints to enable otherwise 4653unsafe floating point optimizations: 4654 4655Example: 4656"""""""" 4657 4658.. code-block:: llvm 4659 4660 <result> = fsub float 4.0, %var ; yields float:result = 4.0 - %var 4661 <result> = fsub float -0.0, %val ; yields float:result = -%var 4662 4663'``mul``' Instruction 4664^^^^^^^^^^^^^^^^^^^^^ 4665 4666Syntax: 4667""""""" 4668 4669:: 4670 4671 <result> = mul <ty> <op1>, <op2> ; yields ty:result 4672 <result> = mul nuw <ty> <op1>, <op2> ; yields ty:result 4673 <result> = mul nsw <ty> <op1>, <op2> ; yields ty:result 4674 <result> = mul nuw nsw <ty> <op1>, <op2> ; yields ty:result 4675 4676Overview: 4677""""""""" 4678 4679The '``mul``' instruction returns the product of its two operands. 4680 4681Arguments: 4682"""""""""" 4683 4684The two arguments to the '``mul``' instruction must be 4685:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 4686arguments must have identical types. 4687 4688Semantics: 4689"""""""""" 4690 4691The value produced is the integer product of the two operands. 4692 4693If the result of the multiplication has unsigned overflow, the result 4694returned is the mathematical result modulo 2\ :sup:`n`\ , where n is the 4695bit width of the result. 4696 4697Because LLVM integers use a two's complement representation, and the 4698result is the same width as the operands, this instruction returns the 4699correct result for both signed and unsigned integers. If a full product 4700(e.g. ``i32`` * ``i32`` -> ``i64``) is needed, the operands should be 4701sign-extended or zero-extended as appropriate to the width of the full 4702product. 4703 4704``nuw`` and ``nsw`` stand for "No Unsigned Wrap" and "No Signed Wrap", 4705respectively. If the ``nuw`` and/or ``nsw`` keywords are present, the 4706result value of the ``mul`` is a :ref:`poison value <poisonvalues>` if 4707unsigned and/or signed overflow, respectively, occurs. 4708 4709Example: 4710"""""""" 4711 4712.. code-block:: llvm 4713 4714 <result> = mul i32 4, %var ; yields i32:result = 4 * %var 4715 4716.. _i_fmul: 4717 4718'``fmul``' Instruction 4719^^^^^^^^^^^^^^^^^^^^^^ 4720 4721Syntax: 4722""""""" 4723 4724:: 4725 4726 <result> = fmul [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result 4727 4728Overview: 4729""""""""" 4730 4731The '``fmul``' instruction returns the product of its two operands. 4732 4733Arguments: 4734"""""""""" 4735 4736The two arguments to the '``fmul``' instruction must be :ref:`floating 4737point <t_floating>` or :ref:`vector <t_vector>` of floating point values. 4738Both arguments must have identical types. 4739 4740Semantics: 4741"""""""""" 4742 4743The value produced is the floating point product of the two operands. 4744This instruction can also take any number of :ref:`fast-math 4745flags <fastmath>`, which are optimization hints to enable otherwise 4746unsafe floating point optimizations: 4747 4748Example: 4749"""""""" 4750 4751.. code-block:: llvm 4752 4753 <result> = fmul float 4.0, %var ; yields float:result = 4.0 * %var 4754 4755'``udiv``' Instruction 4756^^^^^^^^^^^^^^^^^^^^^^ 4757 4758Syntax: 4759""""""" 4760 4761:: 4762 4763 <result> = udiv <ty> <op1>, <op2> ; yields ty:result 4764 <result> = udiv exact <ty> <op1>, <op2> ; yields ty:result 4765 4766Overview: 4767""""""""" 4768 4769The '``udiv``' instruction returns the quotient of its two operands. 4770 4771Arguments: 4772"""""""""" 4773 4774The two arguments to the '``udiv``' instruction must be 4775:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 4776arguments must have identical types. 4777 4778Semantics: 4779"""""""""" 4780 4781The value produced is the unsigned integer quotient of the two operands. 4782 4783Note that unsigned integer division and signed integer division are 4784distinct operations; for signed integer division, use '``sdiv``'. 4785 4786Division by zero leads to undefined behavior. 4787 4788If the ``exact`` keyword is present, the result value of the ``udiv`` is 4789a :ref:`poison value <poisonvalues>` if %op1 is not a multiple of %op2 (as 4790such, "((a udiv exact b) mul b) == a"). 4791 4792Example: 4793"""""""" 4794 4795.. code-block:: llvm 4796 4797 <result> = udiv i32 4, %var ; yields i32:result = 4 / %var 4798 4799'``sdiv``' Instruction 4800^^^^^^^^^^^^^^^^^^^^^^ 4801 4802Syntax: 4803""""""" 4804 4805:: 4806 4807 <result> = sdiv <ty> <op1>, <op2> ; yields ty:result 4808 <result> = sdiv exact <ty> <op1>, <op2> ; yields ty:result 4809 4810Overview: 4811""""""""" 4812 4813The '``sdiv``' instruction returns the quotient of its two operands. 4814 4815Arguments: 4816"""""""""" 4817 4818The two arguments to the '``sdiv``' instruction must be 4819:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 4820arguments must have identical types. 4821 4822Semantics: 4823"""""""""" 4824 4825The value produced is the signed integer quotient of the two operands 4826rounded towards zero. 4827 4828Note that signed integer division and unsigned integer division are 4829distinct operations; for unsigned integer division, use '``udiv``'. 4830 4831Division by zero leads to undefined behavior. Overflow also leads to 4832undefined behavior; this is a rare case, but can occur, for example, by 4833doing a 32-bit division of -2147483648 by -1. 4834 4835If the ``exact`` keyword is present, the result value of the ``sdiv`` is 4836a :ref:`poison value <poisonvalues>` if the result would be rounded. 4837 4838Example: 4839"""""""" 4840 4841.. code-block:: llvm 4842 4843 <result> = sdiv i32 4, %var ; yields i32:result = 4 / %var 4844 4845.. _i_fdiv: 4846 4847'``fdiv``' Instruction 4848^^^^^^^^^^^^^^^^^^^^^^ 4849 4850Syntax: 4851""""""" 4852 4853:: 4854 4855 <result> = fdiv [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result 4856 4857Overview: 4858""""""""" 4859 4860The '``fdiv``' instruction returns the quotient of its two operands. 4861 4862Arguments: 4863"""""""""" 4864 4865The two arguments to the '``fdiv``' instruction must be :ref:`floating 4866point <t_floating>` or :ref:`vector <t_vector>` of floating point values. 4867Both arguments must have identical types. 4868 4869Semantics: 4870"""""""""" 4871 4872The value produced is the floating point quotient of the two operands. 4873This instruction can also take any number of :ref:`fast-math 4874flags <fastmath>`, which are optimization hints to enable otherwise 4875unsafe floating point optimizations: 4876 4877Example: 4878"""""""" 4879 4880.. code-block:: llvm 4881 4882 <result> = fdiv float 4.0, %var ; yields float:result = 4.0 / %var 4883 4884'``urem``' Instruction 4885^^^^^^^^^^^^^^^^^^^^^^ 4886 4887Syntax: 4888""""""" 4889 4890:: 4891 4892 <result> = urem <ty> <op1>, <op2> ; yields ty:result 4893 4894Overview: 4895""""""""" 4896 4897The '``urem``' instruction returns the remainder from the unsigned 4898division of its two arguments. 4899 4900Arguments: 4901"""""""""" 4902 4903The two arguments to the '``urem``' instruction must be 4904:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 4905arguments must have identical types. 4906 4907Semantics: 4908"""""""""" 4909 4910This instruction returns the unsigned integer *remainder* of a division. 4911This instruction always performs an unsigned division to get the 4912remainder. 4913 4914Note that unsigned integer remainder and signed integer remainder are 4915distinct operations; for signed integer remainder, use '``srem``'. 4916 4917Taking the remainder of a division by zero leads to undefined behavior. 4918 4919Example: 4920"""""""" 4921 4922.. code-block:: llvm 4923 4924 <result> = urem i32 4, %var ; yields i32:result = 4 % %var 4925 4926'``srem``' Instruction 4927^^^^^^^^^^^^^^^^^^^^^^ 4928 4929Syntax: 4930""""""" 4931 4932:: 4933 4934 <result> = srem <ty> <op1>, <op2> ; yields ty:result 4935 4936Overview: 4937""""""""" 4938 4939The '``srem``' instruction returns the remainder from the signed 4940division of its two operands. This instruction can also take 4941:ref:`vector <t_vector>` versions of the values in which case the elements 4942must be integers. 4943 4944Arguments: 4945"""""""""" 4946 4947The two arguments to the '``srem``' instruction must be 4948:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 4949arguments must have identical types. 4950 4951Semantics: 4952"""""""""" 4953 4954This instruction returns the *remainder* of a division (where the result 4955is either zero or has the same sign as the dividend, ``op1``), not the 4956*modulo* operator (where the result is either zero or has the same sign 4957as the divisor, ``op2``) of a value. For more information about the 4958difference, see `The Math 4959Forum <http://mathforum.org/dr.math/problems/anne.4.28.99.html>`_. For a 4960table of how this is implemented in various languages, please see 4961`Wikipedia: modulo 4962operation <http://en.wikipedia.org/wiki/Modulo_operation>`_. 4963 4964Note that signed integer remainder and unsigned integer remainder are 4965distinct operations; for unsigned integer remainder, use '``urem``'. 4966 4967Taking the remainder of a division by zero leads to undefined behavior. 4968Overflow also leads to undefined behavior; this is a rare case, but can 4969occur, for example, by taking the remainder of a 32-bit division of 4970-2147483648 by -1. (The remainder doesn't actually overflow, but this 4971rule lets srem be implemented using instructions that return both the 4972result of the division and the remainder.) 4973 4974Example: 4975"""""""" 4976 4977.. code-block:: llvm 4978 4979 <result> = srem i32 4, %var ; yields i32:result = 4 % %var 4980 4981.. _i_frem: 4982 4983'``frem``' Instruction 4984^^^^^^^^^^^^^^^^^^^^^^ 4985 4986Syntax: 4987""""""" 4988 4989:: 4990 4991 <result> = frem [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result 4992 4993Overview: 4994""""""""" 4995 4996The '``frem``' instruction returns the remainder from the division of 4997its two operands. 4998 4999Arguments: 5000"""""""""" 5001 5002The two arguments to the '``frem``' instruction must be :ref:`floating 5003point <t_floating>` or :ref:`vector <t_vector>` of floating point values. 5004Both arguments must have identical types. 5005 5006Semantics: 5007"""""""""" 5008 5009This instruction returns the *remainder* of a division. The remainder 5010has the same sign as the dividend. This instruction can also take any 5011number of :ref:`fast-math flags <fastmath>`, which are optimization hints 5012to enable otherwise unsafe floating point optimizations: 5013 5014Example: 5015"""""""" 5016 5017.. code-block:: llvm 5018 5019 <result> = frem float 4.0, %var ; yields float:result = 4.0 % %var 5020 5021.. _bitwiseops: 5022 5023Bitwise Binary Operations 5024------------------------- 5025 5026Bitwise binary operators are used to do various forms of bit-twiddling 5027in a program. They are generally very efficient instructions and can 5028commonly be strength reduced from other instructions. They require two 5029operands of the same type, execute an operation on them, and produce a 5030single value. The resulting value is the same type as its operands. 5031 5032'``shl``' Instruction 5033^^^^^^^^^^^^^^^^^^^^^ 5034 5035Syntax: 5036""""""" 5037 5038:: 5039 5040 <result> = shl <ty> <op1>, <op2> ; yields ty:result 5041 <result> = shl nuw <ty> <op1>, <op2> ; yields ty:result 5042 <result> = shl nsw <ty> <op1>, <op2> ; yields ty:result 5043 <result> = shl nuw nsw <ty> <op1>, <op2> ; yields ty:result 5044 5045Overview: 5046""""""""" 5047 5048The '``shl``' instruction returns the first operand shifted to the left 5049a specified number of bits. 5050 5051Arguments: 5052"""""""""" 5053 5054Both arguments to the '``shl``' instruction must be the same 5055:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type. 5056'``op2``' is treated as an unsigned value. 5057 5058Semantics: 5059"""""""""" 5060 5061The value produced is ``op1`` \* 2\ :sup:`op2` mod 2\ :sup:`n`, 5062where ``n`` is the width of the result. If ``op2`` is (statically or 5063dynamically) negative or equal to or larger than the number of bits in 5064``op1``, the result is undefined. If the arguments are vectors, each 5065vector element of ``op1`` is shifted by the corresponding shift amount 5066in ``op2``. 5067 5068If the ``nuw`` keyword is present, then the shift produces a :ref:`poison 5069value <poisonvalues>` if it shifts out any non-zero bits. If the 5070``nsw`` keyword is present, then the shift produces a :ref:`poison 5071value <poisonvalues>` if it shifts out any bits that disagree with the 5072resultant sign bit. As such, NUW/NSW have the same semantics as they 5073would if the shift were expressed as a mul instruction with the same 5074nsw/nuw bits in (mul %op1, (shl 1, %op2)). 5075 5076Example: 5077"""""""" 5078 5079.. code-block:: llvm 5080 5081 <result> = shl i32 4, %var ; yields i32: 4 << %var 5082 <result> = shl i32 4, 2 ; yields i32: 16 5083 <result> = shl i32 1, 10 ; yields i32: 1024 5084 <result> = shl i32 1, 32 ; undefined 5085 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 2, i32 4> 5086 5087'``lshr``' Instruction 5088^^^^^^^^^^^^^^^^^^^^^^ 5089 5090Syntax: 5091""""""" 5092 5093:: 5094 5095 <result> = lshr <ty> <op1>, <op2> ; yields ty:result 5096 <result> = lshr exact <ty> <op1>, <op2> ; yields ty:result 5097 5098Overview: 5099""""""""" 5100 5101The '``lshr``' instruction (logical shift right) returns the first 5102operand shifted to the right a specified number of bits with zero fill. 5103 5104Arguments: 5105"""""""""" 5106 5107Both arguments to the '``lshr``' instruction must be the same 5108:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type. 5109'``op2``' is treated as an unsigned value. 5110 5111Semantics: 5112"""""""""" 5113 5114This instruction always performs a logical shift right operation. The 5115most significant bits of the result will be filled with zero bits after 5116the shift. If ``op2`` is (statically or dynamically) equal to or larger 5117than the number of bits in ``op1``, the result is undefined. If the 5118arguments are vectors, each vector element of ``op1`` is shifted by the 5119corresponding shift amount in ``op2``. 5120 5121If the ``exact`` keyword is present, the result value of the ``lshr`` is 5122a :ref:`poison value <poisonvalues>` if any of the bits shifted out are 5123non-zero. 5124 5125Example: 5126"""""""" 5127 5128.. code-block:: llvm 5129 5130 <result> = lshr i32 4, 1 ; yields i32:result = 2 5131 <result> = lshr i32 4, 2 ; yields i32:result = 1 5132 <result> = lshr i8 4, 3 ; yields i8:result = 0 5133 <result> = lshr i8 -2, 1 ; yields i8:result = 0x7F 5134 <result> = lshr i32 1, 32 ; undefined 5135 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1> 5136 5137'``ashr``' Instruction 5138^^^^^^^^^^^^^^^^^^^^^^ 5139 5140Syntax: 5141""""""" 5142 5143:: 5144 5145 <result> = ashr <ty> <op1>, <op2> ; yields ty:result 5146 <result> = ashr exact <ty> <op1>, <op2> ; yields ty:result 5147 5148Overview: 5149""""""""" 5150 5151The '``ashr``' instruction (arithmetic shift right) returns the first 5152operand shifted to the right a specified number of bits with sign 5153extension. 5154 5155Arguments: 5156"""""""""" 5157 5158Both arguments to the '``ashr``' instruction must be the same 5159:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer type. 5160'``op2``' is treated as an unsigned value. 5161 5162Semantics: 5163"""""""""" 5164 5165This instruction always performs an arithmetic shift right operation, 5166The most significant bits of the result will be filled with the sign bit 5167of ``op1``. If ``op2`` is (statically or dynamically) equal to or larger 5168than the number of bits in ``op1``, the result is undefined. If the 5169arguments are vectors, each vector element of ``op1`` is shifted by the 5170corresponding shift amount in ``op2``. 5171 5172If the ``exact`` keyword is present, the result value of the ``ashr`` is 5173a :ref:`poison value <poisonvalues>` if any of the bits shifted out are 5174non-zero. 5175 5176Example: 5177"""""""" 5178 5179.. code-block:: llvm 5180 5181 <result> = ashr i32 4, 1 ; yields i32:result = 2 5182 <result> = ashr i32 4, 2 ; yields i32:result = 1 5183 <result> = ashr i8 4, 3 ; yields i8:result = 0 5184 <result> = ashr i8 -2, 1 ; yields i8:result = -1 5185 <result> = ashr i32 1, 32 ; undefined 5186 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> ; yields: result=<2 x i32> < i32 -1, i32 0> 5187 5188'``and``' Instruction 5189^^^^^^^^^^^^^^^^^^^^^ 5190 5191Syntax: 5192""""""" 5193 5194:: 5195 5196 <result> = and <ty> <op1>, <op2> ; yields ty:result 5197 5198Overview: 5199""""""""" 5200 5201The '``and``' instruction returns the bitwise logical and of its two 5202operands. 5203 5204Arguments: 5205"""""""""" 5206 5207The two arguments to the '``and``' instruction must be 5208:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 5209arguments must have identical types. 5210 5211Semantics: 5212"""""""""" 5213 5214The truth table used for the '``and``' instruction is: 5215 5216+-----+-----+-----+ 5217| In0 | In1 | Out | 5218+-----+-----+-----+ 5219| 0 | 0 | 0 | 5220+-----+-----+-----+ 5221| 0 | 1 | 0 | 5222+-----+-----+-----+ 5223| 1 | 0 | 0 | 5224+-----+-----+-----+ 5225| 1 | 1 | 1 | 5226+-----+-----+-----+ 5227 5228Example: 5229"""""""" 5230 5231.. code-block:: llvm 5232 5233 <result> = and i32 4, %var ; yields i32:result = 4 & %var 5234 <result> = and i32 15, 40 ; yields i32:result = 8 5235 <result> = and i32 4, 8 ; yields i32:result = 0 5236 5237'``or``' Instruction 5238^^^^^^^^^^^^^^^^^^^^ 5239 5240Syntax: 5241""""""" 5242 5243:: 5244 5245 <result> = or <ty> <op1>, <op2> ; yields ty:result 5246 5247Overview: 5248""""""""" 5249 5250The '``or``' instruction returns the bitwise logical inclusive or of its 5251two operands. 5252 5253Arguments: 5254"""""""""" 5255 5256The two arguments to the '``or``' instruction must be 5257:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 5258arguments must have identical types. 5259 5260Semantics: 5261"""""""""" 5262 5263The truth table used for the '``or``' instruction is: 5264 5265+-----+-----+-----+ 5266| In0 | In1 | Out | 5267+-----+-----+-----+ 5268| 0 | 0 | 0 | 5269+-----+-----+-----+ 5270| 0 | 1 | 1 | 5271+-----+-----+-----+ 5272| 1 | 0 | 1 | 5273+-----+-----+-----+ 5274| 1 | 1 | 1 | 5275+-----+-----+-----+ 5276 5277Example: 5278"""""""" 5279 5280:: 5281 5282 <result> = or i32 4, %var ; yields i32:result = 4 | %var 5283 <result> = or i32 15, 40 ; yields i32:result = 47 5284 <result> = or i32 4, 8 ; yields i32:result = 12 5285 5286'``xor``' Instruction 5287^^^^^^^^^^^^^^^^^^^^^ 5288 5289Syntax: 5290""""""" 5291 5292:: 5293 5294 <result> = xor <ty> <op1>, <op2> ; yields ty:result 5295 5296Overview: 5297""""""""" 5298 5299The '``xor``' instruction returns the bitwise logical exclusive or of 5300its two operands. The ``xor`` is used to implement the "one's 5301complement" operation, which is the "~" operator in C. 5302 5303Arguments: 5304"""""""""" 5305 5306The two arguments to the '``xor``' instruction must be 5307:ref:`integer <t_integer>` or :ref:`vector <t_vector>` of integer values. Both 5308arguments must have identical types. 5309 5310Semantics: 5311"""""""""" 5312 5313The truth table used for the '``xor``' instruction is: 5314 5315+-----+-----+-----+ 5316| In0 | In1 | Out | 5317+-----+-----+-----+ 5318| 0 | 0 | 0 | 5319+-----+-----+-----+ 5320| 0 | 1 | 1 | 5321+-----+-----+-----+ 5322| 1 | 0 | 1 | 5323+-----+-----+-----+ 5324| 1 | 1 | 0 | 5325+-----+-----+-----+ 5326 5327Example: 5328"""""""" 5329 5330.. code-block:: llvm 5331 5332 <result> = xor i32 4, %var ; yields i32:result = 4 ^ %var 5333 <result> = xor i32 15, 40 ; yields i32:result = 39 5334 <result> = xor i32 4, 8 ; yields i32:result = 12 5335 <result> = xor i32 %V, -1 ; yields i32:result = ~%V 5336 5337Vector Operations 5338----------------- 5339 5340LLVM supports several instructions to represent vector operations in a 5341target-independent manner. These instructions cover the element-access 5342and vector-specific operations needed to process vectors effectively. 5343While LLVM does directly support these vector operations, many 5344sophisticated algorithms will want to use target-specific intrinsics to 5345take full advantage of a specific target. 5346 5347.. _i_extractelement: 5348 5349'``extractelement``' Instruction 5350^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 5351 5352Syntax: 5353""""""" 5354 5355:: 5356 5357 <result> = extractelement <n x <ty>> <val>, <ty2> <idx> ; yields <ty> 5358 5359Overview: 5360""""""""" 5361 5362The '``extractelement``' instruction extracts a single scalar element 5363from a vector at a specified index. 5364 5365Arguments: 5366"""""""""" 5367 5368The first operand of an '``extractelement``' instruction is a value of 5369:ref:`vector <t_vector>` type. The second operand is an index indicating 5370the position from which to extract the element. The index may be a 5371variable of any integer type. 5372 5373Semantics: 5374"""""""""" 5375 5376The result is a scalar of the same type as the element type of ``val``. 5377Its value is the value at position ``idx`` of ``val``. If ``idx`` 5378exceeds the length of ``val``, the results are undefined. 5379 5380Example: 5381"""""""" 5382 5383.. code-block:: llvm 5384 5385 <result> = extractelement <4 x i32> %vec, i32 0 ; yields i32 5386 5387.. _i_insertelement: 5388 5389'``insertelement``' Instruction 5390^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 5391 5392Syntax: 5393""""""" 5394 5395:: 5396 5397 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, <ty2> <idx> ; yields <n x <ty>> 5398 5399Overview: 5400""""""""" 5401 5402The '``insertelement``' instruction inserts a scalar element into a 5403vector at a specified index. 5404 5405Arguments: 5406"""""""""" 5407 5408The first operand of an '``insertelement``' instruction is a value of 5409:ref:`vector <t_vector>` type. The second operand is a scalar value whose 5410type must equal the element type of the first operand. The third operand 5411is an index indicating the position at which to insert the value. The 5412index may be a variable of any integer type. 5413 5414Semantics: 5415"""""""""" 5416 5417The result is a vector of the same type as ``val``. Its element values 5418are those of ``val`` except at position ``idx``, where it gets the value 5419``elt``. If ``idx`` exceeds the length of ``val``, the results are 5420undefined. 5421 5422Example: 5423"""""""" 5424 5425.. code-block:: llvm 5426 5427 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 ; yields <4 x i32> 5428 5429.. _i_shufflevector: 5430 5431'``shufflevector``' Instruction 5432^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 5433 5434Syntax: 5435""""""" 5436 5437:: 5438 5439 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> ; yields <m x <ty>> 5440 5441Overview: 5442""""""""" 5443 5444The '``shufflevector``' instruction constructs a permutation of elements 5445from two input vectors, returning a vector with the same element type as 5446the input and length that is the same as the shuffle mask. 5447 5448Arguments: 5449"""""""""" 5450 5451The first two operands of a '``shufflevector``' instruction are vectors 5452with the same type. The third argument is a shuffle mask whose element 5453type is always 'i32'. The result of the instruction is a vector whose 5454length is the same as the shuffle mask and whose element type is the 5455same as the element type of the first two operands. 5456 5457The shuffle mask operand is required to be a constant vector with either 5458constant integer or undef values. 5459 5460Semantics: 5461"""""""""" 5462 5463The elements of the two input vectors are numbered from left to right 5464across both of the vectors. The shuffle mask operand specifies, for each 5465element of the result vector, which element of the two input vectors the 5466result element gets. The element selector may be undef (meaning "don't 5467care") and the second operand may be undef if performing a shuffle from 5468only one vector. 5469 5470Example: 5471"""""""" 5472 5473.. code-block:: llvm 5474 5475 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2, 5476 <4 x i32> <i32 0, i32 4, i32 1, i32 5> ; yields <4 x i32> 5477 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef, 5478 <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32> - Identity shuffle. 5479 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef, 5480 <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32> 5481 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2, 5482 <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 > ; yields <8 x i32> 5483 5484Aggregate Operations 5485-------------------- 5486 5487LLVM supports several instructions for working with 5488:ref:`aggregate <t_aggregate>` values. 5489 5490.. _i_extractvalue: 5491 5492'``extractvalue``' Instruction 5493^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 5494 5495Syntax: 5496""""""" 5497 5498:: 5499 5500 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}* 5501 5502Overview: 5503""""""""" 5504 5505The '``extractvalue``' instruction extracts the value of a member field 5506from an :ref:`aggregate <t_aggregate>` value. 5507 5508Arguments: 5509"""""""""" 5510 5511The first operand of an '``extractvalue``' instruction is a value of 5512:ref:`struct <t_struct>` or :ref:`array <t_array>` type. The operands are 5513constant indices to specify which value to extract in a similar manner 5514as indices in a '``getelementptr``' instruction. 5515 5516The major differences to ``getelementptr`` indexing are: 5517 5518- Since the value being indexed is not a pointer, the first index is 5519 omitted and assumed to be zero. 5520- At least one index must be specified. 5521- Not only struct indices but also array indices must be in bounds. 5522 5523Semantics: 5524"""""""""" 5525 5526The result is the value at the position in the aggregate specified by 5527the index operands. 5528 5529Example: 5530"""""""" 5531 5532.. code-block:: llvm 5533 5534 <result> = extractvalue {i32, float} %agg, 0 ; yields i32 5535 5536.. _i_insertvalue: 5537 5538'``insertvalue``' Instruction 5539^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 5540 5541Syntax: 5542""""""" 5543 5544:: 5545 5546 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* ; yields <aggregate type> 5547 5548Overview: 5549""""""""" 5550 5551The '``insertvalue``' instruction inserts a value into a member field in 5552an :ref:`aggregate <t_aggregate>` value. 5553 5554Arguments: 5555"""""""""" 5556 5557The first operand of an '``insertvalue``' instruction is a value of 5558:ref:`struct <t_struct>` or :ref:`array <t_array>` type. The second operand is 5559a first-class value to insert. The following operands are constant 5560indices indicating the position at which to insert the value in a 5561similar manner as indices in a '``extractvalue``' instruction. The value 5562to insert must have the same type as the value identified by the 5563indices. 5564 5565Semantics: 5566"""""""""" 5567 5568The result is an aggregate of the same type as ``val``. Its value is 5569that of ``val`` except that the value at the position specified by the 5570indices is that of ``elt``. 5571 5572Example: 5573"""""""" 5574 5575.. code-block:: llvm 5576 5577 %agg1 = insertvalue {i32, float} undef, i32 1, 0 ; yields {i32 1, float undef} 5578 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 ; yields {i32 1, float %val} 5579 %agg3 = insertvalue {i32, {float}} undef, float %val, 1, 0 ; yields {i32 undef, {float %val}} 5580 5581.. _memoryops: 5582 5583Memory Access and Addressing Operations 5584--------------------------------------- 5585 5586A key design point of an SSA-based representation is how it represents 5587memory. In LLVM, no memory locations are in SSA form, which makes things 5588very simple. This section describes how to read, write, and allocate 5589memory in LLVM. 5590 5591.. _i_alloca: 5592 5593'``alloca``' Instruction 5594^^^^^^^^^^^^^^^^^^^^^^^^ 5595 5596Syntax: 5597""""""" 5598 5599:: 5600 5601 <result> = alloca [inalloca] <type> [, <ty> <NumElements>] [, align <alignment>] ; yields type*:result 5602 5603Overview: 5604""""""""" 5605 5606The '``alloca``' instruction allocates memory on the stack frame of the 5607currently executing function, to be automatically released when this 5608function returns to its caller. The object is always allocated in the 5609generic address space (address space zero). 5610 5611Arguments: 5612"""""""""" 5613 5614The '``alloca``' instruction allocates ``sizeof(<type>)*NumElements`` 5615bytes of memory on the runtime stack, returning a pointer of the 5616appropriate type to the program. If "NumElements" is specified, it is 5617the number of elements allocated, otherwise "NumElements" is defaulted 5618to be one. If a constant alignment is specified, the value result of the 5619allocation is guaranteed to be aligned to at least that boundary. The 5620alignment may not be greater than ``1 << 29``. If not specified, or if 5621zero, the target can choose to align the allocation on any convenient 5622boundary compatible with the type. 5623 5624'``type``' may be any sized type. 5625 5626Semantics: 5627"""""""""" 5628 5629Memory is allocated; a pointer is returned. The operation is undefined 5630if there is insufficient stack space for the allocation. '``alloca``'d 5631memory is automatically released when the function returns. The 5632'``alloca``' instruction is commonly used to represent automatic 5633variables that must have an address available. When the function returns 5634(either with the ``ret`` or ``resume`` instructions), the memory is 5635reclaimed. Allocating zero bytes is legal, but the result is undefined. 5636The order in which memory is allocated (ie., which way the stack grows) 5637is not specified. 5638 5639Example: 5640"""""""" 5641 5642.. code-block:: llvm 5643 5644 %ptr = alloca i32 ; yields i32*:ptr 5645 %ptr = alloca i32, i32 4 ; yields i32*:ptr 5646 %ptr = alloca i32, i32 4, align 1024 ; yields i32*:ptr 5647 %ptr = alloca i32, align 1024 ; yields i32*:ptr 5648 5649.. _i_load: 5650 5651'``load``' Instruction 5652^^^^^^^^^^^^^^^^^^^^^^ 5653 5654Syntax: 5655""""""" 5656 5657:: 5658 5659 <result> = load [volatile] <ty>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>][, !invariant.load !<index>][, !nonnull !<index>] 5660 <result> = load atomic [volatile] <ty>* <pointer> [singlethread] <ordering>, align <alignment> 5661 !<index> = !{ i32 1 } 5662 5663Overview: 5664""""""""" 5665 5666The '``load``' instruction is used to read from memory. 5667 5668Arguments: 5669"""""""""" 5670 5671The argument to the ``load`` instruction specifies the memory address 5672from which to load. The type specified must be a :ref:`first 5673class <t_firstclass>` type. If the ``load`` is marked as ``volatile``, 5674then the optimizer is not allowed to modify the number or order of 5675execution of this ``load`` with other :ref:`volatile 5676operations <volatile>`. 5677 5678If the ``load`` is marked as ``atomic``, it takes an extra 5679:ref:`ordering <ordering>` and optional ``singlethread`` argument. The 5680``release`` and ``acq_rel`` orderings are not valid on ``load`` 5681instructions. Atomic loads produce :ref:`defined <memmodel>` results 5682when they may see multiple atomic stores. The type of the pointee must 5683be an integer type whose bit width is a power of two greater than or 5684equal to eight and less than or equal to a target-specific size limit. 5685``align`` must be explicitly specified on atomic loads, and the load has 5686undefined behavior if the alignment is not set to a value which is at 5687least the size in bytes of the pointee. ``!nontemporal`` does not have 5688any defined semantics for atomic loads. 5689 5690The optional constant ``align`` argument specifies the alignment of the 5691operation (that is, the alignment of the memory address). A value of 0 5692or an omitted ``align`` argument means that the operation has the ABI 5693alignment for the target. It is the responsibility of the code emitter 5694to ensure that the alignment information is correct. Overestimating the 5695alignment results in undefined behavior. Underestimating the alignment 5696may produce less efficient code. An alignment of 1 is always safe. The 5697maximum possible alignment is ``1 << 29``. 5698 5699The optional ``!nontemporal`` metadata must reference a single 5700metadata name ``<index>`` corresponding to a metadata node with one 5701``i32`` entry of value 1. The existence of the ``!nontemporal`` 5702metadata on the instruction tells the optimizer and code generator 5703that this load is not expected to be reused in the cache. The code 5704generator may select special instructions to save cache bandwidth, such 5705as the ``MOVNT`` instruction on x86. 5706 5707The optional ``!invariant.load`` metadata must reference a single 5708metadata name ``<index>`` corresponding to a metadata node with no 5709entries. The existence of the ``!invariant.load`` metadata on the 5710instruction tells the optimizer and code generator that the address 5711operand to this load points to memory which can be assumed unchanged. 5712Being invariant does not imply that a location is dereferenceable, 5713but it does imply that once the location is known dereferenceable 5714its value is henceforth unchanging. 5715 5716The optional ``!nonnull`` metadata must reference a single 5717metadata name ``<index>`` corresponding to a metadata node with no 5718entries. The existence of the ``!nonnull`` metadata on the 5719instruction tells the optimizer that the value loaded is known to 5720never be null. This is analogous to the ''nonnull'' attribute 5721on parameters and return values. This metadata can only be applied 5722to loads of a pointer type. 5723 5724Semantics: 5725"""""""""" 5726 5727The location of memory pointed to is loaded. If the value being loaded 5728is of scalar type then the number of bytes read does not exceed the 5729minimum number of bytes needed to hold all bits of the type. For 5730example, loading an ``i24`` reads at most three bytes. When loading a 5731value of a type like ``i20`` with a size that is not an integral number 5732of bytes, the result is undefined if the value was not originally 5733written using a store of the same type. 5734 5735Examples: 5736""""""""" 5737 5738.. code-block:: llvm 5739 5740 %ptr = alloca i32 ; yields i32*:ptr 5741 store i32 3, i32* %ptr ; yields void 5742 %val = load i32, i32* %ptr ; yields i32:val = i32 3 5743 5744.. _i_store: 5745 5746'``store``' Instruction 5747^^^^^^^^^^^^^^^^^^^^^^^ 5748 5749Syntax: 5750""""""" 5751 5752:: 5753 5754 store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] ; yields void 5755 store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> ; yields void 5756 5757Overview: 5758""""""""" 5759 5760The '``store``' instruction is used to write to memory. 5761 5762Arguments: 5763"""""""""" 5764 5765There are two arguments to the ``store`` instruction: a value to store 5766and an address at which to store it. The type of the ``<pointer>`` 5767operand must be a pointer to the :ref:`first class <t_firstclass>` type of 5768the ``<value>`` operand. If the ``store`` is marked as ``volatile``, 5769then the optimizer is not allowed to modify the number or order of 5770execution of this ``store`` with other :ref:`volatile 5771operations <volatile>`. 5772 5773If the ``store`` is marked as ``atomic``, it takes an extra 5774:ref:`ordering <ordering>` and optional ``singlethread`` argument. The 5775``acquire`` and ``acq_rel`` orderings aren't valid on ``store`` 5776instructions. Atomic loads produce :ref:`defined <memmodel>` results 5777when they may see multiple atomic stores. The type of the pointee must 5778be an integer type whose bit width is a power of two greater than or 5779equal to eight and less than or equal to a target-specific size limit. 5780``align`` must be explicitly specified on atomic stores, and the store 5781has undefined behavior if the alignment is not set to a value which is 5782at least the size in bytes of the pointee. ``!nontemporal`` does not 5783have any defined semantics for atomic stores. 5784 5785The optional constant ``align`` argument specifies the alignment of the 5786operation (that is, the alignment of the memory address). A value of 0 5787or an omitted ``align`` argument means that the operation has the ABI 5788alignment for the target. It is the responsibility of the code emitter 5789to ensure that the alignment information is correct. Overestimating the 5790alignment results in undefined behavior. Underestimating the 5791alignment may produce less efficient code. An alignment of 1 is always 5792safe. The maximum possible alignment is ``1 << 29``. 5793 5794The optional ``!nontemporal`` metadata must reference a single metadata 5795name ``<index>`` corresponding to a metadata node with one ``i32`` entry of 5796value 1. The existence of the ``!nontemporal`` metadata on the instruction 5797tells the optimizer and code generator that this load is not expected to 5798be reused in the cache. The code generator may select special 5799instructions to save cache bandwidth, such as the MOVNT instruction on 5800x86. 5801 5802Semantics: 5803"""""""""" 5804 5805The contents of memory are updated to contain ``<value>`` at the 5806location specified by the ``<pointer>`` operand. If ``<value>`` is 5807of scalar type then the number of bytes written does not exceed the 5808minimum number of bytes needed to hold all bits of the type. For 5809example, storing an ``i24`` writes at most three bytes. When writing a 5810value of a type like ``i20`` with a size that is not an integral number 5811of bytes, it is unspecified what happens to the extra bits that do not 5812belong to the type, but they will typically be overwritten. 5813 5814Example: 5815"""""""" 5816 5817.. code-block:: llvm 5818 5819 %ptr = alloca i32 ; yields i32*:ptr 5820 store i32 3, i32* %ptr ; yields void 5821 %val = load i32* %ptr ; yields i32:val = i32 3 5822 5823.. _i_fence: 5824 5825'``fence``' Instruction 5826^^^^^^^^^^^^^^^^^^^^^^^ 5827 5828Syntax: 5829""""""" 5830 5831:: 5832 5833 fence [singlethread] <ordering> ; yields void 5834 5835Overview: 5836""""""""" 5837 5838The '``fence``' instruction is used to introduce happens-before edges 5839between operations. 5840 5841Arguments: 5842"""""""""" 5843 5844'``fence``' instructions take an :ref:`ordering <ordering>` argument which 5845defines what *synchronizes-with* edges they add. They can only be given 5846``acquire``, ``release``, ``acq_rel``, and ``seq_cst`` orderings. 5847 5848Semantics: 5849"""""""""" 5850 5851A fence A which has (at least) ``release`` ordering semantics 5852*synchronizes with* a fence B with (at least) ``acquire`` ordering 5853semantics if and only if there exist atomic operations X and Y, both 5854operating on some atomic object M, such that A is sequenced before X, X 5855modifies M (either directly or through some side effect of a sequence 5856headed by X), Y is sequenced before B, and Y observes M. This provides a 5857*happens-before* dependency between A and B. Rather than an explicit 5858``fence``, one (but not both) of the atomic operations X or Y might 5859provide a ``release`` or ``acquire`` (resp.) ordering constraint and 5860still *synchronize-with* the explicit ``fence`` and establish the 5861*happens-before* edge. 5862 5863A ``fence`` which has ``seq_cst`` ordering, in addition to having both 5864``acquire`` and ``release`` semantics specified above, participates in 5865the global program order of other ``seq_cst`` operations and/or fences. 5866 5867The optional ":ref:`singlethread <singlethread>`" argument specifies 5868that the fence only synchronizes with other fences in the same thread. 5869(This is useful for interacting with signal handlers.) 5870 5871Example: 5872"""""""" 5873 5874.. code-block:: llvm 5875 5876 fence acquire ; yields void 5877 fence singlethread seq_cst ; yields void 5878 5879.. _i_cmpxchg: 5880 5881'``cmpxchg``' Instruction 5882^^^^^^^^^^^^^^^^^^^^^^^^^ 5883 5884Syntax: 5885""""""" 5886 5887:: 5888 5889 cmpxchg [weak] [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <success ordering> <failure ordering> ; yields { ty, i1 } 5890 5891Overview: 5892""""""""" 5893 5894The '``cmpxchg``' instruction is used to atomically modify memory. It 5895loads a value in memory and compares it to a given value. If they are 5896equal, it tries to store a new value into the memory. 5897 5898Arguments: 5899"""""""""" 5900 5901There are three arguments to the '``cmpxchg``' instruction: an address 5902to operate on, a value to compare to the value currently be at that 5903address, and a new value to place at that address if the compared values 5904are equal. The type of '<cmp>' must be an integer type whose bit width 5905is a power of two greater than or equal to eight and less than or equal 5906to a target-specific size limit. '<cmp>' and '<new>' must have the same 5907type, and the type of '<pointer>' must be a pointer to that type. If the 5908``cmpxchg`` is marked as ``volatile``, then the optimizer is not allowed 5909to modify the number or order of execution of this ``cmpxchg`` with 5910other :ref:`volatile operations <volatile>`. 5911 5912The success and failure :ref:`ordering <ordering>` arguments specify how this 5913``cmpxchg`` synchronizes with other atomic operations. Both ordering parameters 5914must be at least ``monotonic``, the ordering constraint on failure must be no 5915stronger than that on success, and the failure ordering cannot be either 5916``release`` or ``acq_rel``. 5917 5918The optional "``singlethread``" argument declares that the ``cmpxchg`` 5919is only atomic with respect to code (usually signal handlers) running in 5920the same thread as the ``cmpxchg``. Otherwise the cmpxchg is atomic with 5921respect to all other code in the system. 5922 5923The pointer passed into cmpxchg must have alignment greater than or 5924equal to the size in memory of the operand. 5925 5926Semantics: 5927"""""""""" 5928 5929The contents of memory at the location specified by the '``<pointer>``' operand 5930is read and compared to '``<cmp>``'; if the read value is the equal, the 5931'``<new>``' is written. The original value at the location is returned, together 5932with a flag indicating success (true) or failure (false). 5933 5934If the cmpxchg operation is marked as ``weak`` then a spurious failure is 5935permitted: the operation may not write ``<new>`` even if the comparison 5936matched. 5937 5938If the cmpxchg operation is strong (the default), the i1 value is 1 if and only 5939if the value loaded equals ``cmp``. 5940 5941A successful ``cmpxchg`` is a read-modify-write instruction for the purpose of 5942identifying release sequences. A failed ``cmpxchg`` is equivalent to an atomic 5943load with an ordering parameter determined the second ordering parameter. 5944 5945Example: 5946"""""""" 5947 5948.. code-block:: llvm 5949 5950 entry: 5951 %orig = atomic load i32, i32* %ptr unordered ; yields i32 5952 br label %loop 5953 5954 loop: 5955 %cmp = phi i32 [ %orig, %entry ], [%old, %loop] 5956 %squared = mul i32 %cmp, %cmp 5957 %val_success = cmpxchg i32* %ptr, i32 %cmp, i32 %squared acq_rel monotonic ; yields { i32, i1 } 5958 %value_loaded = extractvalue { i32, i1 } %val_success, 0 5959 %success = extractvalue { i32, i1 } %val_success, 1 5960 br i1 %success, label %done, label %loop 5961 5962 done: 5963 ... 5964 5965.. _i_atomicrmw: 5966 5967'``atomicrmw``' Instruction 5968^^^^^^^^^^^^^^^^^^^^^^^^^^^ 5969 5970Syntax: 5971""""""" 5972 5973:: 5974 5975 atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering> ; yields ty 5976 5977Overview: 5978""""""""" 5979 5980The '``atomicrmw``' instruction is used to atomically modify memory. 5981 5982Arguments: 5983"""""""""" 5984 5985There are three arguments to the '``atomicrmw``' instruction: an 5986operation to apply, an address whose value to modify, an argument to the 5987operation. The operation must be one of the following keywords: 5988 5989- xchg 5990- add 5991- sub 5992- and 5993- nand 5994- or 5995- xor 5996- max 5997- min 5998- umax 5999- umin 6000 6001The type of '<value>' must be an integer type whose bit width is a power 6002of two greater than or equal to eight and less than or equal to a 6003target-specific size limit. The type of the '``<pointer>``' operand must 6004be a pointer to that type. If the ``atomicrmw`` is marked as 6005``volatile``, then the optimizer is not allowed to modify the number or 6006order of execution of this ``atomicrmw`` with other :ref:`volatile 6007operations <volatile>`. 6008 6009Semantics: 6010"""""""""" 6011 6012The contents of memory at the location specified by the '``<pointer>``' 6013operand are atomically read, modified, and written back. The original 6014value at the location is returned. The modification is specified by the 6015operation argument: 6016 6017- xchg: ``*ptr = val`` 6018- add: ``*ptr = *ptr + val`` 6019- sub: ``*ptr = *ptr - val`` 6020- and: ``*ptr = *ptr & val`` 6021- nand: ``*ptr = ~(*ptr & val)`` 6022- or: ``*ptr = *ptr | val`` 6023- xor: ``*ptr = *ptr ^ val`` 6024- max: ``*ptr = *ptr > val ? *ptr : val`` (using a signed comparison) 6025- min: ``*ptr = *ptr < val ? *ptr : val`` (using a signed comparison) 6026- umax: ``*ptr = *ptr > val ? *ptr : val`` (using an unsigned 6027 comparison) 6028- umin: ``*ptr = *ptr < val ? *ptr : val`` (using an unsigned 6029 comparison) 6030 6031Example: 6032"""""""" 6033 6034.. code-block:: llvm 6035 6036 %old = atomicrmw add i32* %ptr, i32 1 acquire ; yields i32 6037 6038.. _i_getelementptr: 6039 6040'``getelementptr``' Instruction 6041^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6042 6043Syntax: 6044""""""" 6045 6046:: 6047 6048 <result> = getelementptr <ty>, <ty>* <ptrval>{, <ty> <idx>}* 6049 <result> = getelementptr inbounds <ty>, <ty>* <ptrval>{, <ty> <idx>}* 6050 <result> = getelementptr <ty>, <ptr vector> <ptrval>, <vector index type> <idx> 6051 6052Overview: 6053""""""""" 6054 6055The '``getelementptr``' instruction is used to get the address of a 6056subelement of an :ref:`aggregate <t_aggregate>` data structure. It performs 6057address calculation only and does not access memory. 6058 6059Arguments: 6060"""""""""" 6061 6062The first argument is always a type used as the basis for the calculations. 6063The second argument is always a pointer or a vector of pointers, and is the 6064base address to start from. The remaining arguments are indices 6065that indicate which of the elements of the aggregate object are indexed. 6066The interpretation of each index is dependent on the type being indexed 6067into. The first index always indexes the pointer value given as the 6068first argument, the second index indexes a value of the type pointed to 6069(not necessarily the value directly pointed to, since the first index 6070can be non-zero), etc. The first type indexed into must be a pointer 6071value, subsequent types can be arrays, vectors, and structs. Note that 6072subsequent types being indexed into can never be pointers, since that 6073would require loading the pointer before continuing calculation. 6074 6075The type of each index argument depends on the type it is indexing into. 6076When indexing into a (optionally packed) structure, only ``i32`` integer 6077**constants** are allowed (when using a vector of indices they must all 6078be the **same** ``i32`` integer constant). When indexing into an array, 6079pointer or vector, integers of any width are allowed, and they are not 6080required to be constant. These integers are treated as signed values 6081where relevant. 6082 6083For example, let's consider a C code fragment and how it gets compiled 6084to LLVM: 6085 6086.. code-block:: c 6087 6088 struct RT { 6089 char A; 6090 int B[10][20]; 6091 char C; 6092 }; 6093 struct ST { 6094 int X; 6095 double Y; 6096 struct RT Z; 6097 }; 6098 6099 int *foo(struct ST *s) { 6100 return &s[1].Z.B[5][13]; 6101 } 6102 6103The LLVM code generated by Clang is: 6104 6105.. code-block:: llvm 6106 6107 %struct.RT = type { i8, [10 x [20 x i32]], i8 } 6108 %struct.ST = type { i32, double, %struct.RT } 6109 6110 define i32* @foo(%struct.ST* %s) nounwind uwtable readnone optsize ssp { 6111 entry: 6112 %arrayidx = getelementptr inbounds %struct.ST, %struct.ST* %s, i64 1, i32 2, i32 1, i64 5, i64 13 6113 ret i32* %arrayidx 6114 } 6115 6116Semantics: 6117"""""""""" 6118 6119In the example above, the first index is indexing into the 6120'``%struct.ST*``' type, which is a pointer, yielding a '``%struct.ST``' 6121= '``{ i32, double, %struct.RT }``' type, a structure. The second index 6122indexes into the third element of the structure, yielding a 6123'``%struct.RT``' = '``{ i8 , [10 x [20 x i32]], i8 }``' type, another 6124structure. The third index indexes into the second element of the 6125structure, yielding a '``[10 x [20 x i32]]``' type, an array. The two 6126dimensions of the array are subscripted into, yielding an '``i32``' 6127type. The '``getelementptr``' instruction returns a pointer to this 6128element, thus computing a value of '``i32*``' type. 6129 6130Note that it is perfectly legal to index partially through a structure, 6131returning a pointer to an inner element. Because of this, the LLVM code 6132for the given testcase is equivalent to: 6133 6134.. code-block:: llvm 6135 6136 define i32* @foo(%struct.ST* %s) { 6137 %t1 = getelementptr %struct.ST, %struct.ST* %s, i32 1 ; yields %struct.ST*:%t1 6138 %t2 = getelementptr %struct.ST, %struct.ST* %t1, i32 0, i32 2 ; yields %struct.RT*:%t2 6139 %t3 = getelementptr %struct.RT, %struct.RT* %t2, i32 0, i32 1 ; yields [10 x [20 x i32]]*:%t3 6140 %t4 = getelementptr [10 x [20 x i32]], [10 x [20 x i32]]* %t3, i32 0, i32 5 ; yields [20 x i32]*:%t4 6141 %t5 = getelementptr [20 x i32], [20 x i32]* %t4, i32 0, i32 13 ; yields i32*:%t5 6142 ret i32* %t5 6143 } 6144 6145If the ``inbounds`` keyword is present, the result value of the 6146``getelementptr`` is a :ref:`poison value <poisonvalues>` if the base 6147pointer is not an *in bounds* address of an allocated object, or if any 6148of the addresses that would be formed by successive addition of the 6149offsets implied by the indices to the base address with infinitely 6150precise signed arithmetic are not an *in bounds* address of that 6151allocated object. The *in bounds* addresses for an allocated object are 6152all the addresses that point into the object, plus the address one byte 6153past the end. In cases where the base is a vector of pointers the 6154``inbounds`` keyword applies to each of the computations element-wise. 6155 6156If the ``inbounds`` keyword is not present, the offsets are added to the 6157base address with silently-wrapping two's complement arithmetic. If the 6158offsets have a different width from the pointer, they are sign-extended 6159or truncated to the width of the pointer. The result value of the 6160``getelementptr`` may be outside the object pointed to by the base 6161pointer. The result value may not necessarily be used to access memory 6162though, even if it happens to point into allocated storage. See the 6163:ref:`Pointer Aliasing Rules <pointeraliasing>` section for more 6164information. 6165 6166The getelementptr instruction is often confusing. For some more insight 6167into how it works, see :doc:`the getelementptr FAQ <GetElementPtr>`. 6168 6169Example: 6170"""""""" 6171 6172.. code-block:: llvm 6173 6174 ; yields [12 x i8]*:aptr 6175 %aptr = getelementptr {i32, [12 x i8]}, {i32, [12 x i8]}* %saptr, i64 0, i32 1 6176 ; yields i8*:vptr 6177 %vptr = getelementptr {i32, <2 x i8>}, {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1 6178 ; yields i8*:eptr 6179 %eptr = getelementptr [12 x i8], [12 x i8]* %aptr, i64 0, i32 1 6180 ; yields i32*:iptr 6181 %iptr = getelementptr [10 x i32], [10 x i32]* @arr, i16 0, i16 0 6182 6183In cases where the pointer argument is a vector of pointers, each index 6184must be a vector with the same number of elements. For example: 6185 6186.. code-block:: llvm 6187 6188 %A = getelementptr i8, <4 x i8*> %ptrs, <4 x i64> %offsets, 6189 6190Conversion Operations 6191--------------------- 6192 6193The instructions in this category are the conversion instructions 6194(casting) which all take a single operand and a type. They perform 6195various bit conversions on the operand. 6196 6197'``trunc .. to``' Instruction 6198^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6199 6200Syntax: 6201""""""" 6202 6203:: 6204 6205 <result> = trunc <ty> <value> to <ty2> ; yields ty2 6206 6207Overview: 6208""""""""" 6209 6210The '``trunc``' instruction truncates its operand to the type ``ty2``. 6211 6212Arguments: 6213"""""""""" 6214 6215The '``trunc``' instruction takes a value to trunc, and a type to trunc 6216it to. Both types must be of :ref:`integer <t_integer>` types, or vectors 6217of the same number of integers. The bit size of the ``value`` must be 6218larger than the bit size of the destination type, ``ty2``. Equal sized 6219types are not allowed. 6220 6221Semantics: 6222"""""""""" 6223 6224The '``trunc``' instruction truncates the high order bits in ``value`` 6225and converts the remaining bits to ``ty2``. Since the source size must 6226be larger than the destination size, ``trunc`` cannot be a *no-op cast*. 6227It will always truncate bits. 6228 6229Example: 6230"""""""" 6231 6232.. code-block:: llvm 6233 6234 %X = trunc i32 257 to i8 ; yields i8:1 6235 %Y = trunc i32 123 to i1 ; yields i1:true 6236 %Z = trunc i32 122 to i1 ; yields i1:false 6237 %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> ; yields <i8 8, i8 7> 6238 6239'``zext .. to``' Instruction 6240^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6241 6242Syntax: 6243""""""" 6244 6245:: 6246 6247 <result> = zext <ty> <value> to <ty2> ; yields ty2 6248 6249Overview: 6250""""""""" 6251 6252The '``zext``' instruction zero extends its operand to type ``ty2``. 6253 6254Arguments: 6255"""""""""" 6256 6257The '``zext``' instruction takes a value to cast, and a type to cast it 6258to. Both types must be of :ref:`integer <t_integer>` types, or vectors of 6259the same number of integers. The bit size of the ``value`` must be 6260smaller than the bit size of the destination type, ``ty2``. 6261 6262Semantics: 6263"""""""""" 6264 6265The ``zext`` fills the high order bits of the ``value`` with zero bits 6266until it reaches the size of the destination type, ``ty2``. 6267 6268When zero extending from i1, the result will always be either 0 or 1. 6269 6270Example: 6271"""""""" 6272 6273.. code-block:: llvm 6274 6275 %X = zext i32 257 to i64 ; yields i64:257 6276 %Y = zext i1 true to i32 ; yields i32:1 6277 %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7> 6278 6279'``sext .. to``' Instruction 6280^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6281 6282Syntax: 6283""""""" 6284 6285:: 6286 6287 <result> = sext <ty> <value> to <ty2> ; yields ty2 6288 6289Overview: 6290""""""""" 6291 6292The '``sext``' sign extends ``value`` to the type ``ty2``. 6293 6294Arguments: 6295"""""""""" 6296 6297The '``sext``' instruction takes a value to cast, and a type to cast it 6298to. Both types must be of :ref:`integer <t_integer>` types, or vectors of 6299the same number of integers. The bit size of the ``value`` must be 6300smaller than the bit size of the destination type, ``ty2``. 6301 6302Semantics: 6303"""""""""" 6304 6305The '``sext``' instruction performs a sign extension by copying the sign 6306bit (highest order bit) of the ``value`` until it reaches the bit size 6307of the type ``ty2``. 6308 6309When sign extending from i1, the extension always results in -1 or 0. 6310 6311Example: 6312"""""""" 6313 6314.. code-block:: llvm 6315 6316 %X = sext i8 -1 to i16 ; yields i16 :65535 6317 %Y = sext i1 true to i32 ; yields i32:-1 6318 %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> ; yields <i32 8, i32 7> 6319 6320'``fptrunc .. to``' Instruction 6321^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6322 6323Syntax: 6324""""""" 6325 6326:: 6327 6328 <result> = fptrunc <ty> <value> to <ty2> ; yields ty2 6329 6330Overview: 6331""""""""" 6332 6333The '``fptrunc``' instruction truncates ``value`` to type ``ty2``. 6334 6335Arguments: 6336"""""""""" 6337 6338The '``fptrunc``' instruction takes a :ref:`floating point <t_floating>` 6339value to cast and a :ref:`floating point <t_floating>` type to cast it to. 6340The size of ``value`` must be larger than the size of ``ty2``. This 6341implies that ``fptrunc`` cannot be used to make a *no-op cast*. 6342 6343Semantics: 6344"""""""""" 6345 6346The '``fptrunc``' instruction truncates a ``value`` from a larger 6347:ref:`floating point <t_floating>` type to a smaller :ref:`floating 6348point <t_floating>` type. If the value cannot fit within the 6349destination type, ``ty2``, then the results are undefined. 6350 6351Example: 6352"""""""" 6353 6354.. code-block:: llvm 6355 6356 %X = fptrunc double 123.0 to float ; yields float:123.0 6357 %Y = fptrunc double 1.0E+300 to float ; yields undefined 6358 6359'``fpext .. to``' Instruction 6360^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6361 6362Syntax: 6363""""""" 6364 6365:: 6366 6367 <result> = fpext <ty> <value> to <ty2> ; yields ty2 6368 6369Overview: 6370""""""""" 6371 6372The '``fpext``' extends a floating point ``value`` to a larger floating 6373point value. 6374 6375Arguments: 6376"""""""""" 6377 6378The '``fpext``' instruction takes a :ref:`floating point <t_floating>` 6379``value`` to cast, and a :ref:`floating point <t_floating>` type to cast it 6380to. The source type must be smaller than the destination type. 6381 6382Semantics: 6383"""""""""" 6384 6385The '``fpext``' instruction extends the ``value`` from a smaller 6386:ref:`floating point <t_floating>` type to a larger :ref:`floating 6387point <t_floating>` type. The ``fpext`` cannot be used to make a 6388*no-op cast* because it always changes bits. Use ``bitcast`` to make a 6389*no-op cast* for a floating point cast. 6390 6391Example: 6392"""""""" 6393 6394.. code-block:: llvm 6395 6396 %X = fpext float 3.125 to double ; yields double:3.125000e+00 6397 %Y = fpext double %X to fp128 ; yields fp128:0xL00000000000000004000900000000000 6398 6399'``fptoui .. to``' Instruction 6400^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6401 6402Syntax: 6403""""""" 6404 6405:: 6406 6407 <result> = fptoui <ty> <value> to <ty2> ; yields ty2 6408 6409Overview: 6410""""""""" 6411 6412The '``fptoui``' converts a floating point ``value`` to its unsigned 6413integer equivalent of type ``ty2``. 6414 6415Arguments: 6416"""""""""" 6417 6418The '``fptoui``' instruction takes a value to cast, which must be a 6419scalar or vector :ref:`floating point <t_floating>` value, and a type to 6420cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If 6421``ty`` is a vector floating point type, ``ty2`` must be a vector integer 6422type with the same number of elements as ``ty`` 6423 6424Semantics: 6425"""""""""" 6426 6427The '``fptoui``' instruction converts its :ref:`floating 6428point <t_floating>` operand into the nearest (rounding towards zero) 6429unsigned integer value. If the value cannot fit in ``ty2``, the results 6430are undefined. 6431 6432Example: 6433"""""""" 6434 6435.. code-block:: llvm 6436 6437 %X = fptoui double 123.0 to i32 ; yields i32:123 6438 %Y = fptoui float 1.0E+300 to i1 ; yields undefined:1 6439 %Z = fptoui float 1.04E+17 to i8 ; yields undefined:1 6440 6441'``fptosi .. to``' Instruction 6442^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6443 6444Syntax: 6445""""""" 6446 6447:: 6448 6449 <result> = fptosi <ty> <value> to <ty2> ; yields ty2 6450 6451Overview: 6452""""""""" 6453 6454The '``fptosi``' instruction converts :ref:`floating point <t_floating>` 6455``value`` to type ``ty2``. 6456 6457Arguments: 6458"""""""""" 6459 6460The '``fptosi``' instruction takes a value to cast, which must be a 6461scalar or vector :ref:`floating point <t_floating>` value, and a type to 6462cast it to ``ty2``, which must be an :ref:`integer <t_integer>` type. If 6463``ty`` is a vector floating point type, ``ty2`` must be a vector integer 6464type with the same number of elements as ``ty`` 6465 6466Semantics: 6467"""""""""" 6468 6469The '``fptosi``' instruction converts its :ref:`floating 6470point <t_floating>` operand into the nearest (rounding towards zero) 6471signed integer value. If the value cannot fit in ``ty2``, the results 6472are undefined. 6473 6474Example: 6475"""""""" 6476 6477.. code-block:: llvm 6478 6479 %X = fptosi double -123.0 to i32 ; yields i32:-123 6480 %Y = fptosi float 1.0E-247 to i1 ; yields undefined:1 6481 %Z = fptosi float 1.04E+17 to i8 ; yields undefined:1 6482 6483'``uitofp .. to``' Instruction 6484^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6485 6486Syntax: 6487""""""" 6488 6489:: 6490 6491 <result> = uitofp <ty> <value> to <ty2> ; yields ty2 6492 6493Overview: 6494""""""""" 6495 6496The '``uitofp``' instruction regards ``value`` as an unsigned integer 6497and converts that value to the ``ty2`` type. 6498 6499Arguments: 6500"""""""""" 6501 6502The '``uitofp``' instruction takes a value to cast, which must be a 6503scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to 6504``ty2``, which must be an :ref:`floating point <t_floating>` type. If 6505``ty`` is a vector integer type, ``ty2`` must be a vector floating point 6506type with the same number of elements as ``ty`` 6507 6508Semantics: 6509"""""""""" 6510 6511The '``uitofp``' instruction interprets its operand as an unsigned 6512integer quantity and converts it to the corresponding floating point 6513value. If the value cannot fit in the floating point value, the results 6514are undefined. 6515 6516Example: 6517"""""""" 6518 6519.. code-block:: llvm 6520 6521 %X = uitofp i32 257 to float ; yields float:257.0 6522 %Y = uitofp i8 -1 to double ; yields double:255.0 6523 6524'``sitofp .. to``' Instruction 6525^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6526 6527Syntax: 6528""""""" 6529 6530:: 6531 6532 <result> = sitofp <ty> <value> to <ty2> ; yields ty2 6533 6534Overview: 6535""""""""" 6536 6537The '``sitofp``' instruction regards ``value`` as a signed integer and 6538converts that value to the ``ty2`` type. 6539 6540Arguments: 6541"""""""""" 6542 6543The '``sitofp``' instruction takes a value to cast, which must be a 6544scalar or vector :ref:`integer <t_integer>` value, and a type to cast it to 6545``ty2``, which must be an :ref:`floating point <t_floating>` type. If 6546``ty`` is a vector integer type, ``ty2`` must be a vector floating point 6547type with the same number of elements as ``ty`` 6548 6549Semantics: 6550"""""""""" 6551 6552The '``sitofp``' instruction interprets its operand as a signed integer 6553quantity and converts it to the corresponding floating point value. If 6554the value cannot fit in the floating point value, the results are 6555undefined. 6556 6557Example: 6558"""""""" 6559 6560.. code-block:: llvm 6561 6562 %X = sitofp i32 257 to float ; yields float:257.0 6563 %Y = sitofp i8 -1 to double ; yields double:-1.0 6564 6565.. _i_ptrtoint: 6566 6567'``ptrtoint .. to``' Instruction 6568^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6569 6570Syntax: 6571""""""" 6572 6573:: 6574 6575 <result> = ptrtoint <ty> <value> to <ty2> ; yields ty2 6576 6577Overview: 6578""""""""" 6579 6580The '``ptrtoint``' instruction converts the pointer or a vector of 6581pointers ``value`` to the integer (or vector of integers) type ``ty2``. 6582 6583Arguments: 6584"""""""""" 6585 6586The '``ptrtoint``' instruction takes a ``value`` to cast, which must be 6587a value of type :ref:`pointer <t_pointer>` or a vector of pointers, and a 6588type to cast it to ``ty2``, which must be an :ref:`integer <t_integer>` or 6589a vector of integers type. 6590 6591Semantics: 6592"""""""""" 6593 6594The '``ptrtoint``' instruction converts ``value`` to integer type 6595``ty2`` by interpreting the pointer value as an integer and either 6596truncating or zero extending that value to the size of the integer type. 6597If ``value`` is smaller than ``ty2`` then a zero extension is done. If 6598``value`` is larger than ``ty2`` then a truncation is done. If they are 6599the same size, then nothing is done (*no-op cast*) other than a type 6600change. 6601 6602Example: 6603"""""""" 6604 6605.. code-block:: llvm 6606 6607 %X = ptrtoint i32* %P to i8 ; yields truncation on 32-bit architecture 6608 %Y = ptrtoint i32* %P to i64 ; yields zero extension on 32-bit architecture 6609 %Z = ptrtoint <4 x i32*> %P to <4 x i64>; yields vector zero extension for a vector of addresses on 32-bit architecture 6610 6611.. _i_inttoptr: 6612 6613'``inttoptr .. to``' Instruction 6614^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6615 6616Syntax: 6617""""""" 6618 6619:: 6620 6621 <result> = inttoptr <ty> <value> to <ty2> ; yields ty2 6622 6623Overview: 6624""""""""" 6625 6626The '``inttoptr``' instruction converts an integer ``value`` to a 6627pointer type, ``ty2``. 6628 6629Arguments: 6630"""""""""" 6631 6632The '``inttoptr``' instruction takes an :ref:`integer <t_integer>` value to 6633cast, and a type to cast it to, which must be a :ref:`pointer <t_pointer>` 6634type. 6635 6636Semantics: 6637"""""""""" 6638 6639The '``inttoptr``' instruction converts ``value`` to type ``ty2`` by 6640applying either a zero extension or a truncation depending on the size 6641of the integer ``value``. If ``value`` is larger than the size of a 6642pointer then a truncation is done. If ``value`` is smaller than the size 6643of a pointer then a zero extension is done. If they are the same size, 6644nothing is done (*no-op cast*). 6645 6646Example: 6647"""""""" 6648 6649.. code-block:: llvm 6650 6651 %X = inttoptr i32 255 to i32* ; yields zero extension on 64-bit architecture 6652 %Y = inttoptr i32 255 to i32* ; yields no-op on 32-bit architecture 6653 %Z = inttoptr i64 0 to i32* ; yields truncation on 32-bit architecture 6654 %Z = inttoptr <4 x i32> %G to <4 x i8*>; yields truncation of vector G to four pointers 6655 6656.. _i_bitcast: 6657 6658'``bitcast .. to``' Instruction 6659^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6660 6661Syntax: 6662""""""" 6663 6664:: 6665 6666 <result> = bitcast <ty> <value> to <ty2> ; yields ty2 6667 6668Overview: 6669""""""""" 6670 6671The '``bitcast``' instruction converts ``value`` to type ``ty2`` without 6672changing any bits. 6673 6674Arguments: 6675"""""""""" 6676 6677The '``bitcast``' instruction takes a value to cast, which must be a 6678non-aggregate first class value, and a type to cast it to, which must 6679also be a non-aggregate :ref:`first class <t_firstclass>` type. The 6680bit sizes of ``value`` and the destination type, ``ty2``, must be 6681identical. If the source type is a pointer, the destination type must 6682also be a pointer of the same size. This instruction supports bitwise 6683conversion of vectors to integers and to vectors of other types (as 6684long as they have the same size). 6685 6686Semantics: 6687"""""""""" 6688 6689The '``bitcast``' instruction converts ``value`` to type ``ty2``. It 6690is always a *no-op cast* because no bits change with this 6691conversion. The conversion is done as if the ``value`` had been stored 6692to memory and read back as type ``ty2``. Pointer (or vector of 6693pointers) types may only be converted to other pointer (or vector of 6694pointers) types with the same address space through this instruction. 6695To convert pointers to other types, use the :ref:`inttoptr <i_inttoptr>` 6696or :ref:`ptrtoint <i_ptrtoint>` instructions first. 6697 6698Example: 6699"""""""" 6700 6701.. code-block:: llvm 6702 6703 %X = bitcast i8 255 to i8 ; yields i8 :-1 6704 %Y = bitcast i32* %x to sint* ; yields sint*:%x 6705 %Z = bitcast <2 x int> %V to i64; ; yields i64: %V 6706 %Z = bitcast <2 x i32*> %V to <2 x i64*> ; yields <2 x i64*> 6707 6708.. _i_addrspacecast: 6709 6710'``addrspacecast .. to``' Instruction 6711^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6712 6713Syntax: 6714""""""" 6715 6716:: 6717 6718 <result> = addrspacecast <pty> <ptrval> to <pty2> ; yields pty2 6719 6720Overview: 6721""""""""" 6722 6723The '``addrspacecast``' instruction converts ``ptrval`` from ``pty`` in 6724address space ``n`` to type ``pty2`` in address space ``m``. 6725 6726Arguments: 6727"""""""""" 6728 6729The '``addrspacecast``' instruction takes a pointer or vector of pointer value 6730to cast and a pointer type to cast it to, which must have a different 6731address space. 6732 6733Semantics: 6734"""""""""" 6735 6736The '``addrspacecast``' instruction converts the pointer value 6737``ptrval`` to type ``pty2``. It can be a *no-op cast* or a complex 6738value modification, depending on the target and the address space 6739pair. Pointer conversions within the same address space must be 6740performed with the ``bitcast`` instruction. Note that if the address space 6741conversion is legal then both result and operand refer to the same memory 6742location. 6743 6744Example: 6745"""""""" 6746 6747.. code-block:: llvm 6748 6749 %X = addrspacecast i32* %x to i32 addrspace(1)* ; yields i32 addrspace(1)*:%x 6750 %Y = addrspacecast i32 addrspace(1)* %y to i64 addrspace(2)* ; yields i64 addrspace(2)*:%y 6751 %Z = addrspacecast <4 x i32*> %z to <4 x float addrspace(3)*> ; yields <4 x float addrspace(3)*>:%z 6752 6753.. _otherops: 6754 6755Other Operations 6756---------------- 6757 6758The instructions in this category are the "miscellaneous" instructions, 6759which defy better classification. 6760 6761.. _i_icmp: 6762 6763'``icmp``' Instruction 6764^^^^^^^^^^^^^^^^^^^^^^ 6765 6766Syntax: 6767""""""" 6768 6769:: 6770 6771 <result> = icmp <cond> <ty> <op1>, <op2> ; yields i1 or <N x i1>:result 6772 6773Overview: 6774""""""""" 6775 6776The '``icmp``' instruction returns a boolean value or a vector of 6777boolean values based on comparison of its two integer, integer vector, 6778pointer, or pointer vector operands. 6779 6780Arguments: 6781"""""""""" 6782 6783The '``icmp``' instruction takes three operands. The first operand is 6784the condition code indicating the kind of comparison to perform. It is 6785not a value, just a keyword. The possible condition code are: 6786 6787#. ``eq``: equal 6788#. ``ne``: not equal 6789#. ``ugt``: unsigned greater than 6790#. ``uge``: unsigned greater or equal 6791#. ``ult``: unsigned less than 6792#. ``ule``: unsigned less or equal 6793#. ``sgt``: signed greater than 6794#. ``sge``: signed greater or equal 6795#. ``slt``: signed less than 6796#. ``sle``: signed less or equal 6797 6798The remaining two arguments must be :ref:`integer <t_integer>` or 6799:ref:`pointer <t_pointer>` or integer :ref:`vector <t_vector>` typed. They 6800must also be identical types. 6801 6802Semantics: 6803"""""""""" 6804 6805The '``icmp``' compares ``op1`` and ``op2`` according to the condition 6806code given as ``cond``. The comparison performed always yields either an 6807:ref:`i1 <t_integer>` or vector of ``i1`` result, as follows: 6808 6809#. ``eq``: yields ``true`` if the operands are equal, ``false`` 6810 otherwise. No sign interpretation is necessary or performed. 6811#. ``ne``: yields ``true`` if the operands are unequal, ``false`` 6812 otherwise. No sign interpretation is necessary or performed. 6813#. ``ugt``: interprets the operands as unsigned values and yields 6814 ``true`` if ``op1`` is greater than ``op2``. 6815#. ``uge``: interprets the operands as unsigned values and yields 6816 ``true`` if ``op1`` is greater than or equal to ``op2``. 6817#. ``ult``: interprets the operands as unsigned values and yields 6818 ``true`` if ``op1`` is less than ``op2``. 6819#. ``ule``: interprets the operands as unsigned values and yields 6820 ``true`` if ``op1`` is less than or equal to ``op2``. 6821#. ``sgt``: interprets the operands as signed values and yields ``true`` 6822 if ``op1`` is greater than ``op2``. 6823#. ``sge``: interprets the operands as signed values and yields ``true`` 6824 if ``op1`` is greater than or equal to ``op2``. 6825#. ``slt``: interprets the operands as signed values and yields ``true`` 6826 if ``op1`` is less than ``op2``. 6827#. ``sle``: interprets the operands as signed values and yields ``true`` 6828 if ``op1`` is less than or equal to ``op2``. 6829 6830If the operands are :ref:`pointer <t_pointer>` typed, the pointer values 6831are compared as if they were integers. 6832 6833If the operands are integer vectors, then they are compared element by 6834element. The result is an ``i1`` vector with the same number of elements 6835as the values being compared. Otherwise, the result is an ``i1``. 6836 6837Example: 6838"""""""" 6839 6840.. code-block:: llvm 6841 6842 <result> = icmp eq i32 4, 5 ; yields: result=false 6843 <result> = icmp ne float* %X, %X ; yields: result=false 6844 <result> = icmp ult i16 4, 5 ; yields: result=true 6845 <result> = icmp sgt i16 4, 5 ; yields: result=false 6846 <result> = icmp ule i16 -4, 5 ; yields: result=false 6847 <result> = icmp sge i16 4, 5 ; yields: result=false 6848 6849Note that the code generator does not yet support vector types with the 6850``icmp`` instruction. 6851 6852.. _i_fcmp: 6853 6854'``fcmp``' Instruction 6855^^^^^^^^^^^^^^^^^^^^^^ 6856 6857Syntax: 6858""""""" 6859 6860:: 6861 6862 <result> = fcmp <cond> <ty> <op1>, <op2> ; yields i1 or <N x i1>:result 6863 6864Overview: 6865""""""""" 6866 6867The '``fcmp``' instruction returns a boolean value or vector of boolean 6868values based on comparison of its operands. 6869 6870If the operands are floating point scalars, then the result type is a 6871boolean (:ref:`i1 <t_integer>`). 6872 6873If the operands are floating point vectors, then the result type is a 6874vector of boolean with the same number of elements as the operands being 6875compared. 6876 6877Arguments: 6878"""""""""" 6879 6880The '``fcmp``' instruction takes three operands. The first operand is 6881the condition code indicating the kind of comparison to perform. It is 6882not a value, just a keyword. The possible condition code are: 6883 6884#. ``false``: no comparison, always returns false 6885#. ``oeq``: ordered and equal 6886#. ``ogt``: ordered and greater than 6887#. ``oge``: ordered and greater than or equal 6888#. ``olt``: ordered and less than 6889#. ``ole``: ordered and less than or equal 6890#. ``one``: ordered and not equal 6891#. ``ord``: ordered (no nans) 6892#. ``ueq``: unordered or equal 6893#. ``ugt``: unordered or greater than 6894#. ``uge``: unordered or greater than or equal 6895#. ``ult``: unordered or less than 6896#. ``ule``: unordered or less than or equal 6897#. ``une``: unordered or not equal 6898#. ``uno``: unordered (either nans) 6899#. ``true``: no comparison, always returns true 6900 6901*Ordered* means that neither operand is a QNAN while *unordered* means 6902that either operand may be a QNAN. 6903 6904Each of ``val1`` and ``val2`` arguments must be either a :ref:`floating 6905point <t_floating>` type or a :ref:`vector <t_vector>` of floating point 6906type. They must have identical types. 6907 6908Semantics: 6909"""""""""" 6910 6911The '``fcmp``' instruction compares ``op1`` and ``op2`` according to the 6912condition code given as ``cond``. If the operands are vectors, then the 6913vectors are compared element by element. Each comparison performed 6914always yields an :ref:`i1 <t_integer>` result, as follows: 6915 6916#. ``false``: always yields ``false``, regardless of operands. 6917#. ``oeq``: yields ``true`` if both operands are not a QNAN and ``op1`` 6918 is equal to ``op2``. 6919#. ``ogt``: yields ``true`` if both operands are not a QNAN and ``op1`` 6920 is greater than ``op2``. 6921#. ``oge``: yields ``true`` if both operands are not a QNAN and ``op1`` 6922 is greater than or equal to ``op2``. 6923#. ``olt``: yields ``true`` if both operands are not a QNAN and ``op1`` 6924 is less than ``op2``. 6925#. ``ole``: yields ``true`` if both operands are not a QNAN and ``op1`` 6926 is less than or equal to ``op2``. 6927#. ``one``: yields ``true`` if both operands are not a QNAN and ``op1`` 6928 is not equal to ``op2``. 6929#. ``ord``: yields ``true`` if both operands are not a QNAN. 6930#. ``ueq``: yields ``true`` if either operand is a QNAN or ``op1`` is 6931 equal to ``op2``. 6932#. ``ugt``: yields ``true`` if either operand is a QNAN or ``op1`` is 6933 greater than ``op2``. 6934#. ``uge``: yields ``true`` if either operand is a QNAN or ``op1`` is 6935 greater than or equal to ``op2``. 6936#. ``ult``: yields ``true`` if either operand is a QNAN or ``op1`` is 6937 less than ``op2``. 6938#. ``ule``: yields ``true`` if either operand is a QNAN or ``op1`` is 6939 less than or equal to ``op2``. 6940#. ``une``: yields ``true`` if either operand is a QNAN or ``op1`` is 6941 not equal to ``op2``. 6942#. ``uno``: yields ``true`` if either operand is a QNAN. 6943#. ``true``: always yields ``true``, regardless of operands. 6944 6945Example: 6946"""""""" 6947 6948.. code-block:: llvm 6949 6950 <result> = fcmp oeq float 4.0, 5.0 ; yields: result=false 6951 <result> = fcmp one float 4.0, 5.0 ; yields: result=true 6952 <result> = fcmp olt float 4.0, 5.0 ; yields: result=true 6953 <result> = fcmp ueq double 1.0, 2.0 ; yields: result=false 6954 6955Note that the code generator does not yet support vector types with the 6956``fcmp`` instruction. 6957 6958.. _i_phi: 6959 6960'``phi``' Instruction 6961^^^^^^^^^^^^^^^^^^^^^ 6962 6963Syntax: 6964""""""" 6965 6966:: 6967 6968 <result> = phi <ty> [ <val0>, <label0>], ... 6969 6970Overview: 6971""""""""" 6972 6973The '``phi``' instruction is used to implement the φ node in the SSA 6974graph representing the function. 6975 6976Arguments: 6977"""""""""" 6978 6979The type of the incoming values is specified with the first type field. 6980After this, the '``phi``' instruction takes a list of pairs as 6981arguments, with one pair for each predecessor basic block of the current 6982block. Only values of :ref:`first class <t_firstclass>` type may be used as 6983the value arguments to the PHI node. Only labels may be used as the 6984label arguments. 6985 6986There must be no non-phi instructions between the start of a basic block 6987and the PHI instructions: i.e. PHI instructions must be first in a basic 6988block. 6989 6990For the purposes of the SSA form, the use of each incoming value is 6991deemed to occur on the edge from the corresponding predecessor block to 6992the current block (but after any definition of an '``invoke``' 6993instruction's return value on the same edge). 6994 6995Semantics: 6996"""""""""" 6997 6998At runtime, the '``phi``' instruction logically takes on the value 6999specified by the pair corresponding to the predecessor basic block that 7000executed just prior to the current block. 7001 7002Example: 7003"""""""" 7004 7005.. code-block:: llvm 7006 7007 Loop: ; Infinite loop that counts from 0 on up... 7008 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ] 7009 %nextindvar = add i32 %indvar, 1 7010 br label %Loop 7011 7012.. _i_select: 7013 7014'``select``' Instruction 7015^^^^^^^^^^^^^^^^^^^^^^^^ 7016 7017Syntax: 7018""""""" 7019 7020:: 7021 7022 <result> = select selty <cond>, <ty> <val1>, <ty> <val2> ; yields ty 7023 7024 selty is either i1 or {<N x i1>} 7025 7026Overview: 7027""""""""" 7028 7029The '``select``' instruction is used to choose one value based on a 7030condition, without IR-level branching. 7031 7032Arguments: 7033"""""""""" 7034 7035The '``select``' instruction requires an 'i1' value or a vector of 'i1' 7036values indicating the condition, and two values of the same :ref:`first 7037class <t_firstclass>` type. 7038 7039Semantics: 7040"""""""""" 7041 7042If the condition is an i1 and it evaluates to 1, the instruction returns 7043the first value argument; otherwise, it returns the second value 7044argument. 7045 7046If the condition is a vector of i1, then the value arguments must be 7047vectors of the same size, and the selection is done element by element. 7048 7049If the condition is an i1 and the value arguments are vectors of the 7050same size, then an entire vector is selected. 7051 7052Example: 7053"""""""" 7054 7055.. code-block:: llvm 7056 7057 %X = select i1 true, i8 17, i8 42 ; yields i8:17 7058 7059.. _i_call: 7060 7061'``call``' Instruction 7062^^^^^^^^^^^^^^^^^^^^^^ 7063 7064Syntax: 7065""""""" 7066 7067:: 7068 7069 <result> = [tail | musttail] call [cconv] [ret attrs] <ty> [<fnty>*] <fnptrval>(<function args>) [fn attrs] 7070 7071Overview: 7072""""""""" 7073 7074The '``call``' instruction represents a simple function call. 7075 7076Arguments: 7077"""""""""" 7078 7079This instruction requires several arguments: 7080 7081#. The optional ``tail`` and ``musttail`` markers indicate that the optimizers 7082 should perform tail call optimization. The ``tail`` marker is a hint that 7083 `can be ignored <CodeGenerator.html#sibcallopt>`_. The ``musttail`` marker 7084 means that the call must be tail call optimized in order for the program to 7085 be correct. The ``musttail`` marker provides these guarantees: 7086 7087 #. The call will not cause unbounded stack growth if it is part of a 7088 recursive cycle in the call graph. 7089 #. Arguments with the :ref:`inalloca <attr_inalloca>` attribute are 7090 forwarded in place. 7091 7092 Both markers imply that the callee does not access allocas or varargs from 7093 the caller. Calls marked ``musttail`` must obey the following additional 7094 rules: 7095 7096 - The call must immediately precede a :ref:`ret <i_ret>` instruction, 7097 or a pointer bitcast followed by a ret instruction. 7098 - The ret instruction must return the (possibly bitcasted) value 7099 produced by the call or void. 7100 - The caller and callee prototypes must match. Pointer types of 7101 parameters or return types may differ in pointee type, but not 7102 in address space. 7103 - The calling conventions of the caller and callee must match. 7104 - All ABI-impacting function attributes, such as sret, byval, inreg, 7105 returned, and inalloca, must match. 7106 - The callee must be varargs iff the caller is varargs. Bitcasting a 7107 non-varargs function to the appropriate varargs type is legal so 7108 long as the non-varargs prefixes obey the other rules. 7109 7110 Tail call optimization for calls marked ``tail`` is guaranteed to occur if 7111 the following conditions are met: 7112 7113 - Caller and callee both have the calling convention ``fastcc``. 7114 - The call is in tail position (ret immediately follows call and ret 7115 uses value of call or is void). 7116 - Option ``-tailcallopt`` is enabled, or 7117 ``llvm::GuaranteedTailCallOpt`` is ``true``. 7118 - `Platform-specific constraints are 7119 met. <CodeGenerator.html#tailcallopt>`_ 7120 7121#. The optional "cconv" marker indicates which :ref:`calling 7122 convention <callingconv>` the call should use. If none is 7123 specified, the call defaults to using C calling conventions. The 7124 calling convention of the call must match the calling convention of 7125 the target function, or else the behavior is undefined. 7126#. The optional :ref:`Parameter Attributes <paramattrs>` list for return 7127 values. Only '``zeroext``', '``signext``', and '``inreg``' attributes 7128 are valid here. 7129#. '``ty``': the type of the call instruction itself which is also the 7130 type of the return value. Functions that return no value are marked 7131 ``void``. 7132#. '``fnty``': shall be the signature of the pointer to function value 7133 being invoked. The argument types must match the types implied by 7134 this signature. This type can be omitted if the function is not 7135 varargs and if the function type does not return a pointer to a 7136 function. 7137#. '``fnptrval``': An LLVM value containing a pointer to a function to 7138 be invoked. In most cases, this is a direct function invocation, but 7139 indirect ``call``'s are just as possible, calling an arbitrary pointer 7140 to function value. 7141#. '``function args``': argument list whose types match the function 7142 signature argument types and parameter attributes. All arguments must 7143 be of :ref:`first class <t_firstclass>` type. If the function signature 7144 indicates the function accepts a variable number of arguments, the 7145 extra arguments can be specified. 7146#. The optional :ref:`function attributes <fnattrs>` list. Only 7147 '``noreturn``', '``nounwind``', '``readonly``' and '``readnone``' 7148 attributes are valid here. 7149 7150Semantics: 7151"""""""""" 7152 7153The '``call``' instruction is used to cause control flow to transfer to 7154a specified function, with its incoming arguments bound to the specified 7155values. Upon a '``ret``' instruction in the called function, control 7156flow continues with the instruction after the function call, and the 7157return value of the function is bound to the result argument. 7158 7159Example: 7160"""""""" 7161 7162.. code-block:: llvm 7163 7164 %retval = call i32 @test(i32 %argc) 7165 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) ; yields i32 7166 %X = tail call i32 @foo() ; yields i32 7167 %Y = tail call fastcc i32 @foo() ; yields i32 7168 call void %foo(i8 97 signext) 7169 7170 %struct.A = type { i32, i8 } 7171 %r = call %struct.A @foo() ; yields { i32, i8 } 7172 %gr = extractvalue %struct.A %r, 0 ; yields i32 7173 %gr1 = extractvalue %struct.A %r, 1 ; yields i8 7174 %Z = call void @foo() noreturn ; indicates that %foo never returns normally 7175 %ZZ = call zeroext i32 @bar() ; Return value is %zero extended 7176 7177llvm treats calls to some functions with names and arguments that match 7178the standard C99 library as being the C99 library functions, and may 7179perform optimizations or generate code for them under that assumption. 7180This is something we'd like to change in the future to provide better 7181support for freestanding environments and non-C-based languages. 7182 7183.. _i_va_arg: 7184 7185'``va_arg``' Instruction 7186^^^^^^^^^^^^^^^^^^^^^^^^ 7187 7188Syntax: 7189""""""" 7190 7191:: 7192 7193 <resultval> = va_arg <va_list*> <arglist>, <argty> 7194 7195Overview: 7196""""""""" 7197 7198The '``va_arg``' instruction is used to access arguments passed through 7199the "variable argument" area of a function call. It is used to implement 7200the ``va_arg`` macro in C. 7201 7202Arguments: 7203"""""""""" 7204 7205This instruction takes a ``va_list*`` value and the type of the 7206argument. It returns a value of the specified argument type and 7207increments the ``va_list`` to point to the next argument. The actual 7208type of ``va_list`` is target specific. 7209 7210Semantics: 7211"""""""""" 7212 7213The '``va_arg``' instruction loads an argument of the specified type 7214from the specified ``va_list`` and causes the ``va_list`` to point to 7215the next argument. For more information, see the variable argument 7216handling :ref:`Intrinsic Functions <int_varargs>`. 7217 7218It is legal for this instruction to be called in a function which does 7219not take a variable number of arguments, for example, the ``vfprintf`` 7220function. 7221 7222``va_arg`` is an LLVM instruction instead of an :ref:`intrinsic 7223function <intrinsics>` because it takes a type as an argument. 7224 7225Example: 7226"""""""" 7227 7228See the :ref:`variable argument processing <int_varargs>` section. 7229 7230Note that the code generator does not yet fully support va\_arg on many 7231targets. Also, it does not currently support va\_arg with aggregate 7232types on any target. 7233 7234.. _i_landingpad: 7235 7236'``landingpad``' Instruction 7237^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7238 7239Syntax: 7240""""""" 7241 7242:: 7243 7244 <resultval> = landingpad <resultty> personality <type> <pers_fn> <clause>+ 7245 <resultval> = landingpad <resultty> personality <type> <pers_fn> cleanup <clause>* 7246 7247 <clause> := catch <type> <value> 7248 <clause> := filter <array constant type> <array constant> 7249 7250Overview: 7251""""""""" 7252 7253The '``landingpad``' instruction is used by `LLVM's exception handling 7254system <ExceptionHandling.html#overview>`_ to specify that a basic block 7255is a landing pad --- one where the exception lands, and corresponds to the 7256code found in the ``catch`` portion of a ``try``/``catch`` sequence. It 7257defines values supplied by the personality function (``pers_fn``) upon 7258re-entry to the function. The ``resultval`` has the type ``resultty``. 7259 7260Arguments: 7261"""""""""" 7262 7263This instruction takes a ``pers_fn`` value. This is the personality 7264function associated with the unwinding mechanism. The optional 7265``cleanup`` flag indicates that the landing pad block is a cleanup. 7266 7267A ``clause`` begins with the clause type --- ``catch`` or ``filter`` --- and 7268contains the global variable representing the "type" that may be caught 7269or filtered respectively. Unlike the ``catch`` clause, the ``filter`` 7270clause takes an array constant as its argument. Use 7271"``[0 x i8**] undef``" for a filter which cannot throw. The 7272'``landingpad``' instruction must contain *at least* one ``clause`` or 7273the ``cleanup`` flag. 7274 7275Semantics: 7276"""""""""" 7277 7278The '``landingpad``' instruction defines the values which are set by the 7279personality function (``pers_fn``) upon re-entry to the function, and 7280therefore the "result type" of the ``landingpad`` instruction. As with 7281calling conventions, how the personality function results are 7282represented in LLVM IR is target specific. 7283 7284The clauses are applied in order from top to bottom. If two 7285``landingpad`` instructions are merged together through inlining, the 7286clauses from the calling function are appended to the list of clauses. 7287When the call stack is being unwound due to an exception being thrown, 7288the exception is compared against each ``clause`` in turn. If it doesn't 7289match any of the clauses, and the ``cleanup`` flag is not set, then 7290unwinding continues further up the call stack. 7291 7292The ``landingpad`` instruction has several restrictions: 7293 7294- A landing pad block is a basic block which is the unwind destination 7295 of an '``invoke``' instruction. 7296- A landing pad block must have a '``landingpad``' instruction as its 7297 first non-PHI instruction. 7298- There can be only one '``landingpad``' instruction within the landing 7299 pad block. 7300- A basic block that is not a landing pad block may not include a 7301 '``landingpad``' instruction. 7302- All '``landingpad``' instructions in a function must have the same 7303 personality function. 7304 7305Example: 7306"""""""" 7307 7308.. code-block:: llvm 7309 7310 ;; A landing pad which can catch an integer. 7311 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 7312 catch i8** @_ZTIi 7313 ;; A landing pad that is a cleanup. 7314 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 7315 cleanup 7316 ;; A landing pad which can catch an integer and can only throw a double. 7317 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0 7318 catch i8** @_ZTIi 7319 filter [1 x i8**] [@_ZTId] 7320 7321.. _intrinsics: 7322 7323Intrinsic Functions 7324=================== 7325 7326LLVM supports the notion of an "intrinsic function". These functions 7327have well known names and semantics and are required to follow certain 7328restrictions. Overall, these intrinsics represent an extension mechanism 7329for the LLVM language that does not require changing all of the 7330transformations in LLVM when adding to the language (or the bitcode 7331reader/writer, the parser, etc...). 7332 7333Intrinsic function names must all start with an "``llvm.``" prefix. This 7334prefix is reserved in LLVM for intrinsic names; thus, function names may 7335not begin with this prefix. Intrinsic functions must always be external 7336functions: you cannot define the body of intrinsic functions. Intrinsic 7337functions may only be used in call or invoke instructions: it is illegal 7338to take the address of an intrinsic function. Additionally, because 7339intrinsic functions are part of the LLVM language, it is required if any 7340are added that they be documented here. 7341 7342Some intrinsic functions can be overloaded, i.e., the intrinsic 7343represents a family of functions that perform the same operation but on 7344different data types. Because LLVM can represent over 8 million 7345different integer types, overloading is used commonly to allow an 7346intrinsic function to operate on any integer type. One or more of the 7347argument types or the result type can be overloaded to accept any 7348integer type. Argument types may also be defined as exactly matching a 7349previous argument's type or the result type. This allows an intrinsic 7350function which accepts multiple arguments, but needs all of them to be 7351of the same type, to only be overloaded with respect to a single 7352argument or the result. 7353 7354Overloaded intrinsics will have the names of its overloaded argument 7355types encoded into its function name, each preceded by a period. Only 7356those types which are overloaded result in a name suffix. Arguments 7357whose type is matched against another type do not. For example, the 7358``llvm.ctpop`` function can take an integer of any width and returns an 7359integer of exactly the same integer width. This leads to a family of 7360functions such as ``i8 @llvm.ctpop.i8(i8 %val)`` and 7361``i29 @llvm.ctpop.i29(i29 %val)``. Only one type, the return type, is 7362overloaded, and only one type suffix is required. Because the argument's 7363type is matched against the return type, it does not require its own 7364name suffix. 7365 7366To learn how to add an intrinsic function, please see the `Extending 7367LLVM Guide <ExtendingLLVM.html>`_. 7368 7369.. _int_varargs: 7370 7371Variable Argument Handling Intrinsics 7372------------------------------------- 7373 7374Variable argument support is defined in LLVM with the 7375:ref:`va_arg <i_va_arg>` instruction and these three intrinsic 7376functions. These functions are related to the similarly named macros 7377defined in the ``<stdarg.h>`` header file. 7378 7379All of these functions operate on arguments that use a target-specific 7380value type "``va_list``". The LLVM assembly language reference manual 7381does not define what this type is, so all transformations should be 7382prepared to handle these functions regardless of the type used. 7383 7384This example shows how the :ref:`va_arg <i_va_arg>` instruction and the 7385variable argument handling intrinsic functions are used. 7386 7387.. code-block:: llvm 7388 7389 ; This struct is different for every platform. For most platforms, 7390 ; it is merely an i8*. 7391 %struct.va_list = type { i8* } 7392 7393 ; For Unix x86_64 platforms, va_list is the following struct: 7394 ; %struct.va_list = type { i32, i32, i8*, i8* } 7395 7396 define i32 @test(i32 %X, ...) { 7397 ; Initialize variable argument processing 7398 %ap = alloca %struct.va_list 7399 %ap2 = bitcast %struct.va_list* %ap to i8* 7400 call void @llvm.va_start(i8* %ap2) 7401 7402 ; Read a single integer argument 7403 %tmp = va_arg i8* %ap2, i32 7404 7405 ; Demonstrate usage of llvm.va_copy and llvm.va_end 7406 %aq = alloca i8* 7407 %aq2 = bitcast i8** %aq to i8* 7408 call void @llvm.va_copy(i8* %aq2, i8* %ap2) 7409 call void @llvm.va_end(i8* %aq2) 7410 7411 ; Stop processing of arguments. 7412 call void @llvm.va_end(i8* %ap2) 7413 ret i32 %tmp 7414 } 7415 7416 declare void @llvm.va_start(i8*) 7417 declare void @llvm.va_copy(i8*, i8*) 7418 declare void @llvm.va_end(i8*) 7419 7420.. _int_va_start: 7421 7422'``llvm.va_start``' Intrinsic 7423^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7424 7425Syntax: 7426""""""" 7427 7428:: 7429 7430 declare void @llvm.va_start(i8* <arglist>) 7431 7432Overview: 7433""""""""" 7434 7435The '``llvm.va_start``' intrinsic initializes ``*<arglist>`` for 7436subsequent use by ``va_arg``. 7437 7438Arguments: 7439"""""""""" 7440 7441The argument is a pointer to a ``va_list`` element to initialize. 7442 7443Semantics: 7444"""""""""" 7445 7446The '``llvm.va_start``' intrinsic works just like the ``va_start`` macro 7447available in C. In a target-dependent way, it initializes the 7448``va_list`` element to which the argument points, so that the next call 7449to ``va_arg`` will produce the first variable argument passed to the 7450function. Unlike the C ``va_start`` macro, this intrinsic does not need 7451to know the last argument of the function as the compiler can figure 7452that out. 7453 7454'``llvm.va_end``' Intrinsic 7455^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7456 7457Syntax: 7458""""""" 7459 7460:: 7461 7462 declare void @llvm.va_end(i8* <arglist>) 7463 7464Overview: 7465""""""""" 7466 7467The '``llvm.va_end``' intrinsic destroys ``*<arglist>``, which has been 7468initialized previously with ``llvm.va_start`` or ``llvm.va_copy``. 7469 7470Arguments: 7471"""""""""" 7472 7473The argument is a pointer to a ``va_list`` to destroy. 7474 7475Semantics: 7476"""""""""" 7477 7478The '``llvm.va_end``' intrinsic works just like the ``va_end`` macro 7479available in C. In a target-dependent way, it destroys the ``va_list`` 7480element to which the argument points. Calls to 7481:ref:`llvm.va_start <int_va_start>` and 7482:ref:`llvm.va_copy <int_va_copy>` must be matched exactly with calls to 7483``llvm.va_end``. 7484 7485.. _int_va_copy: 7486 7487'``llvm.va_copy``' Intrinsic 7488^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7489 7490Syntax: 7491""""""" 7492 7493:: 7494 7495 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>) 7496 7497Overview: 7498""""""""" 7499 7500The '``llvm.va_copy``' intrinsic copies the current argument position 7501from the source argument list to the destination argument list. 7502 7503Arguments: 7504"""""""""" 7505 7506The first argument is a pointer to a ``va_list`` element to initialize. 7507The second argument is a pointer to a ``va_list`` element to copy from. 7508 7509Semantics: 7510"""""""""" 7511 7512The '``llvm.va_copy``' intrinsic works just like the ``va_copy`` macro 7513available in C. In a target-dependent way, it copies the source 7514``va_list`` element into the destination ``va_list`` element. This 7515intrinsic is necessary because the `` llvm.va_start`` intrinsic may be 7516arbitrarily complex and require, for example, memory allocation. 7517 7518Accurate Garbage Collection Intrinsics 7519-------------------------------------- 7520 7521LLVM's support for `Accurate Garbage Collection <GarbageCollection.html>`_ 7522(GC) requires the frontend to generate code containing appropriate intrinsic 7523calls and select an appropriate GC strategy which knows how to lower these 7524intrinsics in a manner which is appropriate for the target collector. 7525 7526These intrinsics allow identification of :ref:`GC roots on the 7527stack <int_gcroot>`, as well as garbage collector implementations that 7528require :ref:`read <int_gcread>` and :ref:`write <int_gcwrite>` barriers. 7529Frontends for type-safe garbage collected languages should generate 7530these intrinsics to make use of the LLVM garbage collectors. For more 7531details, see `Garbage Collection with LLVM <GarbageCollection.html>`_. 7532 7533Experimental Statepoint Intrinsics 7534^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7535 7536LLVM provides an second experimental set of intrinsics for describing garbage 7537collection safepoints in compiled code. These intrinsics are an alternative 7538to the ``llvm.gcroot`` intrinsics, but are compatible with the ones for 7539:ref:`read <int_gcread>` and :ref:`write <int_gcwrite>` barriers. The 7540differences in approach are covered in the `Garbage Collection with LLVM 7541<GarbageCollection.html>`_ documentation. The intrinsics themselves are 7542described in :doc:`Statepoints`. 7543 7544.. _int_gcroot: 7545 7546'``llvm.gcroot``' Intrinsic 7547^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7548 7549Syntax: 7550""""""" 7551 7552:: 7553 7554 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata) 7555 7556Overview: 7557""""""""" 7558 7559The '``llvm.gcroot``' intrinsic declares the existence of a GC root to 7560the code generator, and allows some metadata to be associated with it. 7561 7562Arguments: 7563"""""""""" 7564 7565The first argument specifies the address of a stack object that contains 7566the root pointer. The second pointer (which must be either a constant or 7567a global value address) contains the meta-data to be associated with the 7568root. 7569 7570Semantics: 7571"""""""""" 7572 7573At runtime, a call to this intrinsic stores a null pointer into the 7574"ptrloc" location. At compile-time, the code generator generates 7575information to allow the runtime to find the pointer at GC safe points. 7576The '``llvm.gcroot``' intrinsic may only be used in a function which 7577:ref:`specifies a GC algorithm <gc>`. 7578 7579.. _int_gcread: 7580 7581'``llvm.gcread``' Intrinsic 7582^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7583 7584Syntax: 7585""""""" 7586 7587:: 7588 7589 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr) 7590 7591Overview: 7592""""""""" 7593 7594The '``llvm.gcread``' intrinsic identifies reads of references from heap 7595locations, allowing garbage collector implementations that require read 7596barriers. 7597 7598Arguments: 7599"""""""""" 7600 7601The second argument is the address to read from, which should be an 7602address allocated from the garbage collector. The first object is a 7603pointer to the start of the referenced object, if needed by the language 7604runtime (otherwise null). 7605 7606Semantics: 7607"""""""""" 7608 7609The '``llvm.gcread``' intrinsic has the same semantics as a load 7610instruction, but may be replaced with substantially more complex code by 7611the garbage collector runtime, as needed. The '``llvm.gcread``' 7612intrinsic may only be used in a function which :ref:`specifies a GC 7613algorithm <gc>`. 7614 7615.. _int_gcwrite: 7616 7617'``llvm.gcwrite``' Intrinsic 7618^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7619 7620Syntax: 7621""""""" 7622 7623:: 7624 7625 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2) 7626 7627Overview: 7628""""""""" 7629 7630The '``llvm.gcwrite``' intrinsic identifies writes of references to heap 7631locations, allowing garbage collector implementations that require write 7632barriers (such as generational or reference counting collectors). 7633 7634Arguments: 7635"""""""""" 7636 7637The first argument is the reference to store, the second is the start of 7638the object to store it to, and the third is the address of the field of 7639Obj to store to. If the runtime does not require a pointer to the 7640object, Obj may be null. 7641 7642Semantics: 7643"""""""""" 7644 7645The '``llvm.gcwrite``' intrinsic has the same semantics as a store 7646instruction, but may be replaced with substantially more complex code by 7647the garbage collector runtime, as needed. The '``llvm.gcwrite``' 7648intrinsic may only be used in a function which :ref:`specifies a GC 7649algorithm <gc>`. 7650 7651Code Generator Intrinsics 7652------------------------- 7653 7654These intrinsics are provided by LLVM to expose special features that 7655may only be implemented with code generator support. 7656 7657'``llvm.returnaddress``' Intrinsic 7658^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7659 7660Syntax: 7661""""""" 7662 7663:: 7664 7665 declare i8 *@llvm.returnaddress(i32 <level>) 7666 7667Overview: 7668""""""""" 7669 7670The '``llvm.returnaddress``' intrinsic attempts to compute a 7671target-specific value indicating the return address of the current 7672function or one of its callers. 7673 7674Arguments: 7675"""""""""" 7676 7677The argument to this intrinsic indicates which function to return the 7678address for. Zero indicates the calling function, one indicates its 7679caller, etc. The argument is **required** to be a constant integer 7680value. 7681 7682Semantics: 7683"""""""""" 7684 7685The '``llvm.returnaddress``' intrinsic either returns a pointer 7686indicating the return address of the specified call frame, or zero if it 7687cannot be identified. The value returned by this intrinsic is likely to 7688be incorrect or 0 for arguments other than zero, so it should only be 7689used for debugging purposes. 7690 7691Note that calling this intrinsic does not prevent function inlining or 7692other aggressive transformations, so the value returned may not be that 7693of the obvious source-language caller. 7694 7695'``llvm.frameaddress``' Intrinsic 7696^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7697 7698Syntax: 7699""""""" 7700 7701:: 7702 7703 declare i8* @llvm.frameaddress(i32 <level>) 7704 7705Overview: 7706""""""""" 7707 7708The '``llvm.frameaddress``' intrinsic attempts to return the 7709target-specific frame pointer value for the specified stack frame. 7710 7711Arguments: 7712"""""""""" 7713 7714The argument to this intrinsic indicates which function to return the 7715frame pointer for. Zero indicates the calling function, one indicates 7716its caller, etc. The argument is **required** to be a constant integer 7717value. 7718 7719Semantics: 7720"""""""""" 7721 7722The '``llvm.frameaddress``' intrinsic either returns a pointer 7723indicating the frame address of the specified call frame, or zero if it 7724cannot be identified. The value returned by this intrinsic is likely to 7725be incorrect or 0 for arguments other than zero, so it should only be 7726used for debugging purposes. 7727 7728Note that calling this intrinsic does not prevent function inlining or 7729other aggressive transformations, so the value returned may not be that 7730of the obvious source-language caller. 7731 7732'``llvm.frameescape``' and '``llvm.framerecover``' Intrinsics 7733^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7734 7735Syntax: 7736""""""" 7737 7738:: 7739 7740 declare void @llvm.frameescape(...) 7741 declare i8* @llvm.framerecover(i8* %func, i8* %fp, i32 %idx) 7742 7743Overview: 7744""""""""" 7745 7746The '``llvm.frameescape``' intrinsic escapes offsets of a collection of static 7747allocas, and the '``llvm.framerecover``' intrinsic applies those offsets to a 7748live frame pointer to recover the address of the allocation. The offset is 7749computed during frame layout of the caller of ``llvm.frameescape``. 7750 7751Arguments: 7752"""""""""" 7753 7754All arguments to '``llvm.frameescape``' must be pointers to static allocas or 7755casts of static allocas. Each function can only call '``llvm.frameescape``' 7756once, and it can only do so from the entry block. 7757 7758The ``func`` argument to '``llvm.framerecover``' must be a constant 7759bitcasted pointer to a function defined in the current module. The code 7760generator cannot determine the frame allocation offset of functions defined in 7761other modules. 7762 7763The ``fp`` argument to '``llvm.framerecover``' must be a frame 7764pointer of a call frame that is currently live. The return value of 7765'``llvm.frameaddress``' is one way to produce such a value, but most platforms 7766also expose the frame pointer through stack unwinding mechanisms. 7767 7768The ``idx`` argument to '``llvm.framerecover``' indicates which alloca passed to 7769'``llvm.frameescape``' to recover. It is zero-indexed. 7770 7771Semantics: 7772"""""""""" 7773 7774These intrinsics allow a group of functions to access one stack memory 7775allocation in an ancestor stack frame. The memory returned from 7776'``llvm.frameallocate``' may be allocated prior to stack realignment, so the 7777memory is only aligned to the ABI-required stack alignment. Each function may 7778only call '``llvm.frameallocate``' one or zero times from the function entry 7779block. The frame allocation intrinsic inhibits inlining, as any frame 7780allocations in the inlined function frame are likely to be at a different 7781offset from the one used by '``llvm.framerecover``' called with the 7782uninlined function. 7783 7784.. _int_read_register: 7785.. _int_write_register: 7786 7787'``llvm.read_register``' and '``llvm.write_register``' Intrinsics 7788^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7789 7790Syntax: 7791""""""" 7792 7793:: 7794 7795 declare i32 @llvm.read_register.i32(metadata) 7796 declare i64 @llvm.read_register.i64(metadata) 7797 declare void @llvm.write_register.i32(metadata, i32 @value) 7798 declare void @llvm.write_register.i64(metadata, i64 @value) 7799 !0 = !{!"sp\00"} 7800 7801Overview: 7802""""""""" 7803 7804The '``llvm.read_register``' and '``llvm.write_register``' intrinsics 7805provides access to the named register. The register must be valid on 7806the architecture being compiled to. The type needs to be compatible 7807with the register being read. 7808 7809Semantics: 7810"""""""""" 7811 7812The '``llvm.read_register``' intrinsic returns the current value of the 7813register, where possible. The '``llvm.write_register``' intrinsic sets 7814the current value of the register, where possible. 7815 7816This is useful to implement named register global variables that need 7817to always be mapped to a specific register, as is common practice on 7818bare-metal programs including OS kernels. 7819 7820The compiler doesn't check for register availability or use of the used 7821register in surrounding code, including inline assembly. Because of that, 7822allocatable registers are not supported. 7823 7824Warning: So far it only works with the stack pointer on selected 7825architectures (ARM, AArch64, PowerPC and x86_64). Significant amount of 7826work is needed to support other registers and even more so, allocatable 7827registers. 7828 7829.. _int_stacksave: 7830 7831'``llvm.stacksave``' Intrinsic 7832^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7833 7834Syntax: 7835""""""" 7836 7837:: 7838 7839 declare i8* @llvm.stacksave() 7840 7841Overview: 7842""""""""" 7843 7844The '``llvm.stacksave``' intrinsic is used to remember the current state 7845of the function stack, for use with 7846:ref:`llvm.stackrestore <int_stackrestore>`. This is useful for 7847implementing language features like scoped automatic variable sized 7848arrays in C99. 7849 7850Semantics: 7851"""""""""" 7852 7853This intrinsic returns a opaque pointer value that can be passed to 7854:ref:`llvm.stackrestore <int_stackrestore>`. When an 7855``llvm.stackrestore`` intrinsic is executed with a value saved from 7856``llvm.stacksave``, it effectively restores the state of the stack to 7857the state it was in when the ``llvm.stacksave`` intrinsic executed. In 7858practice, this pops any :ref:`alloca <i_alloca>` blocks from the stack that 7859were allocated after the ``llvm.stacksave`` was executed. 7860 7861.. _int_stackrestore: 7862 7863'``llvm.stackrestore``' Intrinsic 7864^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7865 7866Syntax: 7867""""""" 7868 7869:: 7870 7871 declare void @llvm.stackrestore(i8* %ptr) 7872 7873Overview: 7874""""""""" 7875 7876The '``llvm.stackrestore``' intrinsic is used to restore the state of 7877the function stack to the state it was in when the corresponding 7878:ref:`llvm.stacksave <int_stacksave>` intrinsic executed. This is 7879useful for implementing language features like scoped automatic variable 7880sized arrays in C99. 7881 7882Semantics: 7883"""""""""" 7884 7885See the description for :ref:`llvm.stacksave <int_stacksave>`. 7886 7887'``llvm.prefetch``' Intrinsic 7888^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7889 7890Syntax: 7891""""""" 7892 7893:: 7894 7895 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>) 7896 7897Overview: 7898""""""""" 7899 7900The '``llvm.prefetch``' intrinsic is a hint to the code generator to 7901insert a prefetch instruction if supported; otherwise, it is a noop. 7902Prefetches have no effect on the behavior of the program but can change 7903its performance characteristics. 7904 7905Arguments: 7906"""""""""" 7907 7908``address`` is the address to be prefetched, ``rw`` is the specifier 7909determining if the fetch should be for a read (0) or write (1), and 7910``locality`` is a temporal locality specifier ranging from (0) - no 7911locality, to (3) - extremely local keep in cache. The ``cache type`` 7912specifies whether the prefetch is performed on the data (1) or 7913instruction (0) cache. The ``rw``, ``locality`` and ``cache type`` 7914arguments must be constant integers. 7915 7916Semantics: 7917"""""""""" 7918 7919This intrinsic does not modify the behavior of the program. In 7920particular, prefetches cannot trap and do not produce a value. On 7921targets that support this intrinsic, the prefetch can provide hints to 7922the processor cache for better performance. 7923 7924'``llvm.pcmarker``' Intrinsic 7925^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7926 7927Syntax: 7928""""""" 7929 7930:: 7931 7932 declare void @llvm.pcmarker(i32 <id>) 7933 7934Overview: 7935""""""""" 7936 7937The '``llvm.pcmarker``' intrinsic is a method to export a Program 7938Counter (PC) in a region of code to simulators and other tools. The 7939method is target specific, but it is expected that the marker will use 7940exported symbols to transmit the PC of the marker. The marker makes no 7941guarantees that it will remain with any specific instruction after 7942optimizations. It is possible that the presence of a marker will inhibit 7943optimizations. The intended use is to be inserted after optimizations to 7944allow correlations of simulation runs. 7945 7946Arguments: 7947"""""""""" 7948 7949``id`` is a numerical id identifying the marker. 7950 7951Semantics: 7952"""""""""" 7953 7954This intrinsic does not modify the behavior of the program. Backends 7955that do not support this intrinsic may ignore it. 7956 7957'``llvm.readcyclecounter``' Intrinsic 7958^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7959 7960Syntax: 7961""""""" 7962 7963:: 7964 7965 declare i64 @llvm.readcyclecounter() 7966 7967Overview: 7968""""""""" 7969 7970The '``llvm.readcyclecounter``' intrinsic provides access to the cycle 7971counter register (or similar low latency, high accuracy clocks) on those 7972targets that support it. On X86, it should map to RDTSC. On Alpha, it 7973should map to RPCC. As the backing counters overflow quickly (on the 7974order of 9 seconds on alpha), this should only be used for small 7975timings. 7976 7977Semantics: 7978"""""""""" 7979 7980When directly supported, reading the cycle counter should not modify any 7981memory. Implementations are allowed to either return a application 7982specific value or a system wide value. On backends without support, this 7983is lowered to a constant 0. 7984 7985Note that runtime support may be conditional on the privilege-level code is 7986running at and the host platform. 7987 7988'``llvm.clear_cache``' Intrinsic 7989^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 7990 7991Syntax: 7992""""""" 7993 7994:: 7995 7996 declare void @llvm.clear_cache(i8*, i8*) 7997 7998Overview: 7999""""""""" 8000 8001The '``llvm.clear_cache``' intrinsic ensures visibility of modifications 8002in the specified range to the execution unit of the processor. On 8003targets with non-unified instruction and data cache, the implementation 8004flushes the instruction cache. 8005 8006Semantics: 8007"""""""""" 8008 8009On platforms with coherent instruction and data caches (e.g. x86), this 8010intrinsic is a nop. On platforms with non-coherent instruction and data 8011cache (e.g. ARM, MIPS), the intrinsic is lowered either to appropriate 8012instructions or a system call, if cache flushing requires special 8013privileges. 8014 8015The default behavior is to emit a call to ``__clear_cache`` from the run 8016time library. 8017 8018This instrinsic does *not* empty the instruction pipeline. Modifications 8019of the current function are outside the scope of the intrinsic. 8020 8021'``llvm.instrprof_increment``' Intrinsic 8022^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8023 8024Syntax: 8025""""""" 8026 8027:: 8028 8029 declare void @llvm.instrprof_increment(i8* <name>, i64 <hash>, 8030 i32 <num-counters>, i32 <index>) 8031 8032Overview: 8033""""""""" 8034 8035The '``llvm.instrprof_increment``' intrinsic can be emitted by a 8036frontend for use with instrumentation based profiling. These will be 8037lowered by the ``-instrprof`` pass to generate execution counts of a 8038program at runtime. 8039 8040Arguments: 8041"""""""""" 8042 8043The first argument is a pointer to a global variable containing the 8044name of the entity being instrumented. This should generally be the 8045(mangled) function name for a set of counters. 8046 8047The second argument is a hash value that can be used by the consumer 8048of the profile data to detect changes to the instrumented source, and 8049the third is the number of counters associated with ``name``. It is an 8050error if ``hash`` or ``num-counters`` differ between two instances of 8051``instrprof_increment`` that refer to the same name. 8052 8053The last argument refers to which of the counters for ``name`` should 8054be incremented. It should be a value between 0 and ``num-counters``. 8055 8056Semantics: 8057"""""""""" 8058 8059This intrinsic represents an increment of a profiling counter. It will 8060cause the ``-instrprof`` pass to generate the appropriate data 8061structures and the code to increment the appropriate value, in a 8062format that can be written out by a compiler runtime and consumed via 8063the ``llvm-profdata`` tool. 8064 8065Standard C Library Intrinsics 8066----------------------------- 8067 8068LLVM provides intrinsics for a few important standard C library 8069functions. These intrinsics allow source-language front-ends to pass 8070information about the alignment of the pointer arguments to the code 8071generator, providing opportunity for more efficient code generation. 8072 8073.. _int_memcpy: 8074 8075'``llvm.memcpy``' Intrinsic 8076^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8077 8078Syntax: 8079""""""" 8080 8081This is an overloaded intrinsic. You can use ``llvm.memcpy`` on any 8082integer bit width and for different address spaces. Not all targets 8083support all bit widths however. 8084 8085:: 8086 8087 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>, 8088 i32 <len>, i32 <align>, i1 <isvolatile>) 8089 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>, 8090 i64 <len>, i32 <align>, i1 <isvolatile>) 8091 8092Overview: 8093""""""""" 8094 8095The '``llvm.memcpy.*``' intrinsics copy a block of memory from the 8096source location to the destination location. 8097 8098Note that, unlike the standard libc function, the ``llvm.memcpy.*`` 8099intrinsics do not return a value, takes extra alignment/isvolatile 8100arguments and the pointers can be in specified address spaces. 8101 8102Arguments: 8103"""""""""" 8104 8105The first argument is a pointer to the destination, the second is a 8106pointer to the source. The third argument is an integer argument 8107specifying the number of bytes to copy, the fourth argument is the 8108alignment of the source and destination locations, and the fifth is a 8109boolean indicating a volatile access. 8110 8111If the call to this intrinsic has an alignment value that is not 0 or 1, 8112then the caller guarantees that both the source and destination pointers 8113are aligned to that boundary. 8114 8115If the ``isvolatile`` parameter is ``true``, the ``llvm.memcpy`` call is 8116a :ref:`volatile operation <volatile>`. The detailed access behavior is not 8117very cleanly specified and it is unwise to depend on it. 8118 8119Semantics: 8120"""""""""" 8121 8122The '``llvm.memcpy.*``' intrinsics copy a block of memory from the 8123source location to the destination location, which are not allowed to 8124overlap. It copies "len" bytes of memory over. If the argument is known 8125to be aligned to some boundary, this can be specified as the fourth 8126argument, otherwise it should be set to 0 or 1 (both meaning no alignment). 8127 8128'``llvm.memmove``' Intrinsic 8129^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8130 8131Syntax: 8132""""""" 8133 8134This is an overloaded intrinsic. You can use llvm.memmove on any integer 8135bit width and for different address space. Not all targets support all 8136bit widths however. 8137 8138:: 8139 8140 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>, 8141 i32 <len>, i32 <align>, i1 <isvolatile>) 8142 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>, 8143 i64 <len>, i32 <align>, i1 <isvolatile>) 8144 8145Overview: 8146""""""""" 8147 8148The '``llvm.memmove.*``' intrinsics move a block of memory from the 8149source location to the destination location. It is similar to the 8150'``llvm.memcpy``' intrinsic but allows the two memory locations to 8151overlap. 8152 8153Note that, unlike the standard libc function, the ``llvm.memmove.*`` 8154intrinsics do not return a value, takes extra alignment/isvolatile 8155arguments and the pointers can be in specified address spaces. 8156 8157Arguments: 8158"""""""""" 8159 8160The first argument is a pointer to the destination, the second is a 8161pointer to the source. The third argument is an integer argument 8162specifying the number of bytes to copy, the fourth argument is the 8163alignment of the source and destination locations, and the fifth is a 8164boolean indicating a volatile access. 8165 8166If the call to this intrinsic has an alignment value that is not 0 or 1, 8167then the caller guarantees that the source and destination pointers are 8168aligned to that boundary. 8169 8170If the ``isvolatile`` parameter is ``true``, the ``llvm.memmove`` call 8171is a :ref:`volatile operation <volatile>`. The detailed access behavior is 8172not very cleanly specified and it is unwise to depend on it. 8173 8174Semantics: 8175"""""""""" 8176 8177The '``llvm.memmove.*``' intrinsics copy a block of memory from the 8178source location to the destination location, which may overlap. It 8179copies "len" bytes of memory over. If the argument is known to be 8180aligned to some boundary, this can be specified as the fourth argument, 8181otherwise it should be set to 0 or 1 (both meaning no alignment). 8182 8183'``llvm.memset.*``' Intrinsics 8184^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8185 8186Syntax: 8187""""""" 8188 8189This is an overloaded intrinsic. You can use llvm.memset on any integer 8190bit width and for different address spaces. However, not all targets 8191support all bit widths. 8192 8193:: 8194 8195 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>, 8196 i32 <len>, i32 <align>, i1 <isvolatile>) 8197 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>, 8198 i64 <len>, i32 <align>, i1 <isvolatile>) 8199 8200Overview: 8201""""""""" 8202 8203The '``llvm.memset.*``' intrinsics fill a block of memory with a 8204particular byte value. 8205 8206Note that, unlike the standard libc function, the ``llvm.memset`` 8207intrinsic does not return a value and takes extra alignment/volatile 8208arguments. Also, the destination can be in an arbitrary address space. 8209 8210Arguments: 8211"""""""""" 8212 8213The first argument is a pointer to the destination to fill, the second 8214is the byte value with which to fill it, the third argument is an 8215integer argument specifying the number of bytes to fill, and the fourth 8216argument is the known alignment of the destination location. 8217 8218If the call to this intrinsic has an alignment value that is not 0 or 1, 8219then the caller guarantees that the destination pointer is aligned to 8220that boundary. 8221 8222If the ``isvolatile`` parameter is ``true``, the ``llvm.memset`` call is 8223a :ref:`volatile operation <volatile>`. The detailed access behavior is not 8224very cleanly specified and it is unwise to depend on it. 8225 8226Semantics: 8227"""""""""" 8228 8229The '``llvm.memset.*``' intrinsics fill "len" bytes of memory starting 8230at the destination location. If the argument is known to be aligned to 8231some boundary, this can be specified as the fourth argument, otherwise 8232it should be set to 0 or 1 (both meaning no alignment). 8233 8234'``llvm.sqrt.*``' Intrinsic 8235^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8236 8237Syntax: 8238""""""" 8239 8240This is an overloaded intrinsic. You can use ``llvm.sqrt`` on any 8241floating point or vector of floating point type. Not all targets support 8242all types however. 8243 8244:: 8245 8246 declare float @llvm.sqrt.f32(float %Val) 8247 declare double @llvm.sqrt.f64(double %Val) 8248 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val) 8249 declare fp128 @llvm.sqrt.f128(fp128 %Val) 8250 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val) 8251 8252Overview: 8253""""""""" 8254 8255The '``llvm.sqrt``' intrinsics return the sqrt of the specified operand, 8256returning the same value as the libm '``sqrt``' functions would. Unlike 8257``sqrt`` in libm, however, ``llvm.sqrt`` has undefined behavior for 8258negative numbers other than -0.0 (which allows for better optimization, 8259because there is no need to worry about errno being set). 8260``llvm.sqrt(-0.0)`` is defined to return -0.0 like IEEE sqrt. 8261 8262Arguments: 8263"""""""""" 8264 8265The argument and return value are floating point numbers of the same 8266type. 8267 8268Semantics: 8269"""""""""" 8270 8271This function returns the sqrt of the specified operand if it is a 8272nonnegative floating point number. 8273 8274'``llvm.powi.*``' Intrinsic 8275^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8276 8277Syntax: 8278""""""" 8279 8280This is an overloaded intrinsic. You can use ``llvm.powi`` on any 8281floating point or vector of floating point type. Not all targets support 8282all types however. 8283 8284:: 8285 8286 declare float @llvm.powi.f32(float %Val, i32 %power) 8287 declare double @llvm.powi.f64(double %Val, i32 %power) 8288 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power) 8289 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power) 8290 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power) 8291 8292Overview: 8293""""""""" 8294 8295The '``llvm.powi.*``' intrinsics return the first operand raised to the 8296specified (positive or negative) power. The order of evaluation of 8297multiplications is not defined. When a vector of floating point type is 8298used, the second argument remains a scalar integer value. 8299 8300Arguments: 8301"""""""""" 8302 8303The second argument is an integer power, and the first is a value to 8304raise to that power. 8305 8306Semantics: 8307"""""""""" 8308 8309This function returns the first value raised to the second power with an 8310unspecified sequence of rounding operations. 8311 8312'``llvm.sin.*``' Intrinsic 8313^^^^^^^^^^^^^^^^^^^^^^^^^^ 8314 8315Syntax: 8316""""""" 8317 8318This is an overloaded intrinsic. You can use ``llvm.sin`` on any 8319floating point or vector of floating point type. Not all targets support 8320all types however. 8321 8322:: 8323 8324 declare float @llvm.sin.f32(float %Val) 8325 declare double @llvm.sin.f64(double %Val) 8326 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val) 8327 declare fp128 @llvm.sin.f128(fp128 %Val) 8328 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val) 8329 8330Overview: 8331""""""""" 8332 8333The '``llvm.sin.*``' intrinsics return the sine of the operand. 8334 8335Arguments: 8336"""""""""" 8337 8338The argument and return value are floating point numbers of the same 8339type. 8340 8341Semantics: 8342"""""""""" 8343 8344This function returns the sine of the specified operand, returning the 8345same values as the libm ``sin`` functions would, and handles error 8346conditions in the same way. 8347 8348'``llvm.cos.*``' Intrinsic 8349^^^^^^^^^^^^^^^^^^^^^^^^^^ 8350 8351Syntax: 8352""""""" 8353 8354This is an overloaded intrinsic. You can use ``llvm.cos`` on any 8355floating point or vector of floating point type. Not all targets support 8356all types however. 8357 8358:: 8359 8360 declare float @llvm.cos.f32(float %Val) 8361 declare double @llvm.cos.f64(double %Val) 8362 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val) 8363 declare fp128 @llvm.cos.f128(fp128 %Val) 8364 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val) 8365 8366Overview: 8367""""""""" 8368 8369The '``llvm.cos.*``' intrinsics return the cosine of the operand. 8370 8371Arguments: 8372"""""""""" 8373 8374The argument and return value are floating point numbers of the same 8375type. 8376 8377Semantics: 8378"""""""""" 8379 8380This function returns the cosine of the specified operand, returning the 8381same values as the libm ``cos`` functions would, and handles error 8382conditions in the same way. 8383 8384'``llvm.pow.*``' Intrinsic 8385^^^^^^^^^^^^^^^^^^^^^^^^^^ 8386 8387Syntax: 8388""""""" 8389 8390This is an overloaded intrinsic. You can use ``llvm.pow`` on any 8391floating point or vector of floating point type. Not all targets support 8392all types however. 8393 8394:: 8395 8396 declare float @llvm.pow.f32(float %Val, float %Power) 8397 declare double @llvm.pow.f64(double %Val, double %Power) 8398 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power) 8399 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power) 8400 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power) 8401 8402Overview: 8403""""""""" 8404 8405The '``llvm.pow.*``' intrinsics return the first operand raised to the 8406specified (positive or negative) power. 8407 8408Arguments: 8409"""""""""" 8410 8411The second argument is a floating point power, and the first is a value 8412to raise to that power. 8413 8414Semantics: 8415"""""""""" 8416 8417This function returns the first value raised to the second power, 8418returning the same values as the libm ``pow`` functions would, and 8419handles error conditions in the same way. 8420 8421'``llvm.exp.*``' Intrinsic 8422^^^^^^^^^^^^^^^^^^^^^^^^^^ 8423 8424Syntax: 8425""""""" 8426 8427This is an overloaded intrinsic. You can use ``llvm.exp`` on any 8428floating point or vector of floating point type. Not all targets support 8429all types however. 8430 8431:: 8432 8433 declare float @llvm.exp.f32(float %Val) 8434 declare double @llvm.exp.f64(double %Val) 8435 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val) 8436 declare fp128 @llvm.exp.f128(fp128 %Val) 8437 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val) 8438 8439Overview: 8440""""""""" 8441 8442The '``llvm.exp.*``' intrinsics perform the exp function. 8443 8444Arguments: 8445"""""""""" 8446 8447The argument and return value are floating point numbers of the same 8448type. 8449 8450Semantics: 8451"""""""""" 8452 8453This function returns the same values as the libm ``exp`` functions 8454would, and handles error conditions in the same way. 8455 8456'``llvm.exp2.*``' Intrinsic 8457^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8458 8459Syntax: 8460""""""" 8461 8462This is an overloaded intrinsic. You can use ``llvm.exp2`` on any 8463floating point or vector of floating point type. Not all targets support 8464all types however. 8465 8466:: 8467 8468 declare float @llvm.exp2.f32(float %Val) 8469 declare double @llvm.exp2.f64(double %Val) 8470 declare x86_fp80 @llvm.exp2.f80(x86_fp80 %Val) 8471 declare fp128 @llvm.exp2.f128(fp128 %Val) 8472 declare ppc_fp128 @llvm.exp2.ppcf128(ppc_fp128 %Val) 8473 8474Overview: 8475""""""""" 8476 8477The '``llvm.exp2.*``' intrinsics perform the exp2 function. 8478 8479Arguments: 8480"""""""""" 8481 8482The argument and return value are floating point numbers of the same 8483type. 8484 8485Semantics: 8486"""""""""" 8487 8488This function returns the same values as the libm ``exp2`` functions 8489would, and handles error conditions in the same way. 8490 8491'``llvm.log.*``' Intrinsic 8492^^^^^^^^^^^^^^^^^^^^^^^^^^ 8493 8494Syntax: 8495""""""" 8496 8497This is an overloaded intrinsic. You can use ``llvm.log`` on any 8498floating point or vector of floating point type. Not all targets support 8499all types however. 8500 8501:: 8502 8503 declare float @llvm.log.f32(float %Val) 8504 declare double @llvm.log.f64(double %Val) 8505 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val) 8506 declare fp128 @llvm.log.f128(fp128 %Val) 8507 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val) 8508 8509Overview: 8510""""""""" 8511 8512The '``llvm.log.*``' intrinsics perform the log function. 8513 8514Arguments: 8515"""""""""" 8516 8517The argument and return value are floating point numbers of the same 8518type. 8519 8520Semantics: 8521"""""""""" 8522 8523This function returns the same values as the libm ``log`` functions 8524would, and handles error conditions in the same way. 8525 8526'``llvm.log10.*``' Intrinsic 8527^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8528 8529Syntax: 8530""""""" 8531 8532This is an overloaded intrinsic. You can use ``llvm.log10`` on any 8533floating point or vector of floating point type. Not all targets support 8534all types however. 8535 8536:: 8537 8538 declare float @llvm.log10.f32(float %Val) 8539 declare double @llvm.log10.f64(double %Val) 8540 declare x86_fp80 @llvm.log10.f80(x86_fp80 %Val) 8541 declare fp128 @llvm.log10.f128(fp128 %Val) 8542 declare ppc_fp128 @llvm.log10.ppcf128(ppc_fp128 %Val) 8543 8544Overview: 8545""""""""" 8546 8547The '``llvm.log10.*``' intrinsics perform the log10 function. 8548 8549Arguments: 8550"""""""""" 8551 8552The argument and return value are floating point numbers of the same 8553type. 8554 8555Semantics: 8556"""""""""" 8557 8558This function returns the same values as the libm ``log10`` functions 8559would, and handles error conditions in the same way. 8560 8561'``llvm.log2.*``' Intrinsic 8562^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8563 8564Syntax: 8565""""""" 8566 8567This is an overloaded intrinsic. You can use ``llvm.log2`` on any 8568floating point or vector of floating point type. Not all targets support 8569all types however. 8570 8571:: 8572 8573 declare float @llvm.log2.f32(float %Val) 8574 declare double @llvm.log2.f64(double %Val) 8575 declare x86_fp80 @llvm.log2.f80(x86_fp80 %Val) 8576 declare fp128 @llvm.log2.f128(fp128 %Val) 8577 declare ppc_fp128 @llvm.log2.ppcf128(ppc_fp128 %Val) 8578 8579Overview: 8580""""""""" 8581 8582The '``llvm.log2.*``' intrinsics perform the log2 function. 8583 8584Arguments: 8585"""""""""" 8586 8587The argument and return value are floating point numbers of the same 8588type. 8589 8590Semantics: 8591"""""""""" 8592 8593This function returns the same values as the libm ``log2`` functions 8594would, and handles error conditions in the same way. 8595 8596'``llvm.fma.*``' Intrinsic 8597^^^^^^^^^^^^^^^^^^^^^^^^^^ 8598 8599Syntax: 8600""""""" 8601 8602This is an overloaded intrinsic. You can use ``llvm.fma`` on any 8603floating point or vector of floating point type. Not all targets support 8604all types however. 8605 8606:: 8607 8608 declare float @llvm.fma.f32(float %a, float %b, float %c) 8609 declare double @llvm.fma.f64(double %a, double %b, double %c) 8610 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c) 8611 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c) 8612 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c) 8613 8614Overview: 8615""""""""" 8616 8617The '``llvm.fma.*``' intrinsics perform the fused multiply-add 8618operation. 8619 8620Arguments: 8621"""""""""" 8622 8623The argument and return value are floating point numbers of the same 8624type. 8625 8626Semantics: 8627"""""""""" 8628 8629This function returns the same values as the libm ``fma`` functions 8630would, and does not set errno. 8631 8632'``llvm.fabs.*``' Intrinsic 8633^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8634 8635Syntax: 8636""""""" 8637 8638This is an overloaded intrinsic. You can use ``llvm.fabs`` on any 8639floating point or vector of floating point type. Not all targets support 8640all types however. 8641 8642:: 8643 8644 declare float @llvm.fabs.f32(float %Val) 8645 declare double @llvm.fabs.f64(double %Val) 8646 declare x86_fp80 @llvm.fabs.f80(x86_fp80 %Val) 8647 declare fp128 @llvm.fabs.f128(fp128 %Val) 8648 declare ppc_fp128 @llvm.fabs.ppcf128(ppc_fp128 %Val) 8649 8650Overview: 8651""""""""" 8652 8653The '``llvm.fabs.*``' intrinsics return the absolute value of the 8654operand. 8655 8656Arguments: 8657"""""""""" 8658 8659The argument and return value are floating point numbers of the same 8660type. 8661 8662Semantics: 8663"""""""""" 8664 8665This function returns the same values as the libm ``fabs`` functions 8666would, and handles error conditions in the same way. 8667 8668'``llvm.minnum.*``' Intrinsic 8669^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8670 8671Syntax: 8672""""""" 8673 8674This is an overloaded intrinsic. You can use ``llvm.minnum`` on any 8675floating point or vector of floating point type. Not all targets support 8676all types however. 8677 8678:: 8679 8680 declare float @llvm.minnum.f32(float %Val0, float %Val1) 8681 declare double @llvm.minnum.f64(double %Val0, double %Val1) 8682 declare x86_fp80 @llvm.minnum.f80(x86_fp80 %Val0, x86_fp80 %Val1) 8683 declare fp128 @llvm.minnum.f128(fp128 %Val0, fp128 %Val1) 8684 declare ppc_fp128 @llvm.minnum.ppcf128(ppc_fp128 %Val0, ppc_fp128 %Val1) 8685 8686Overview: 8687""""""""" 8688 8689The '``llvm.minnum.*``' intrinsics return the minimum of the two 8690arguments. 8691 8692 8693Arguments: 8694"""""""""" 8695 8696The arguments and return value are floating point numbers of the same 8697type. 8698 8699Semantics: 8700"""""""""" 8701 8702Follows the IEEE-754 semantics for minNum, which also match for libm's 8703fmin. 8704 8705If either operand is a NaN, returns the other non-NaN operand. Returns 8706NaN only if both operands are NaN. If the operands compare equal, 8707returns a value that compares equal to both operands. This means that 8708fmin(+/-0.0, +/-0.0) could return either -0.0 or 0.0. 8709 8710'``llvm.maxnum.*``' Intrinsic 8711^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8712 8713Syntax: 8714""""""" 8715 8716This is an overloaded intrinsic. You can use ``llvm.maxnum`` on any 8717floating point or vector of floating point type. Not all targets support 8718all types however. 8719 8720:: 8721 8722 declare float @llvm.maxnum.f32(float %Val0, float %Val1l) 8723 declare double @llvm.maxnum.f64(double %Val0, double %Val1) 8724 declare x86_fp80 @llvm.maxnum.f80(x86_fp80 %Val0, x86_fp80 %Val1) 8725 declare fp128 @llvm.maxnum.f128(fp128 %Val0, fp128 %Val1) 8726 declare ppc_fp128 @llvm.maxnum.ppcf128(ppc_fp128 %Val0, ppc_fp128 %Val1) 8727 8728Overview: 8729""""""""" 8730 8731The '``llvm.maxnum.*``' intrinsics return the maximum of the two 8732arguments. 8733 8734 8735Arguments: 8736"""""""""" 8737 8738The arguments and return value are floating point numbers of the same 8739type. 8740 8741Semantics: 8742"""""""""" 8743Follows the IEEE-754 semantics for maxNum, which also match for libm's 8744fmax. 8745 8746If either operand is a NaN, returns the other non-NaN operand. Returns 8747NaN only if both operands are NaN. If the operands compare equal, 8748returns a value that compares equal to both operands. This means that 8749fmax(+/-0.0, +/-0.0) could return either -0.0 or 0.0. 8750 8751'``llvm.copysign.*``' Intrinsic 8752^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8753 8754Syntax: 8755""""""" 8756 8757This is an overloaded intrinsic. You can use ``llvm.copysign`` on any 8758floating point or vector of floating point type. Not all targets support 8759all types however. 8760 8761:: 8762 8763 declare float @llvm.copysign.f32(float %Mag, float %Sgn) 8764 declare double @llvm.copysign.f64(double %Mag, double %Sgn) 8765 declare x86_fp80 @llvm.copysign.f80(x86_fp80 %Mag, x86_fp80 %Sgn) 8766 declare fp128 @llvm.copysign.f128(fp128 %Mag, fp128 %Sgn) 8767 declare ppc_fp128 @llvm.copysign.ppcf128(ppc_fp128 %Mag, ppc_fp128 %Sgn) 8768 8769Overview: 8770""""""""" 8771 8772The '``llvm.copysign.*``' intrinsics return a value with the magnitude of the 8773first operand and the sign of the second operand. 8774 8775Arguments: 8776"""""""""" 8777 8778The arguments and return value are floating point numbers of the same 8779type. 8780 8781Semantics: 8782"""""""""" 8783 8784This function returns the same values as the libm ``copysign`` 8785functions would, and handles error conditions in the same way. 8786 8787'``llvm.floor.*``' Intrinsic 8788^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8789 8790Syntax: 8791""""""" 8792 8793This is an overloaded intrinsic. You can use ``llvm.floor`` on any 8794floating point or vector of floating point type. Not all targets support 8795all types however. 8796 8797:: 8798 8799 declare float @llvm.floor.f32(float %Val) 8800 declare double @llvm.floor.f64(double %Val) 8801 declare x86_fp80 @llvm.floor.f80(x86_fp80 %Val) 8802 declare fp128 @llvm.floor.f128(fp128 %Val) 8803 declare ppc_fp128 @llvm.floor.ppcf128(ppc_fp128 %Val) 8804 8805Overview: 8806""""""""" 8807 8808The '``llvm.floor.*``' intrinsics return the floor of the operand. 8809 8810Arguments: 8811"""""""""" 8812 8813The argument and return value are floating point numbers of the same 8814type. 8815 8816Semantics: 8817"""""""""" 8818 8819This function returns the same values as the libm ``floor`` functions 8820would, and handles error conditions in the same way. 8821 8822'``llvm.ceil.*``' Intrinsic 8823^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8824 8825Syntax: 8826""""""" 8827 8828This is an overloaded intrinsic. You can use ``llvm.ceil`` on any 8829floating point or vector of floating point type. Not all targets support 8830all types however. 8831 8832:: 8833 8834 declare float @llvm.ceil.f32(float %Val) 8835 declare double @llvm.ceil.f64(double %Val) 8836 declare x86_fp80 @llvm.ceil.f80(x86_fp80 %Val) 8837 declare fp128 @llvm.ceil.f128(fp128 %Val) 8838 declare ppc_fp128 @llvm.ceil.ppcf128(ppc_fp128 %Val) 8839 8840Overview: 8841""""""""" 8842 8843The '``llvm.ceil.*``' intrinsics return the ceiling of the operand. 8844 8845Arguments: 8846"""""""""" 8847 8848The argument and return value are floating point numbers of the same 8849type. 8850 8851Semantics: 8852"""""""""" 8853 8854This function returns the same values as the libm ``ceil`` functions 8855would, and handles error conditions in the same way. 8856 8857'``llvm.trunc.*``' Intrinsic 8858^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8859 8860Syntax: 8861""""""" 8862 8863This is an overloaded intrinsic. You can use ``llvm.trunc`` on any 8864floating point or vector of floating point type. Not all targets support 8865all types however. 8866 8867:: 8868 8869 declare float @llvm.trunc.f32(float %Val) 8870 declare double @llvm.trunc.f64(double %Val) 8871 declare x86_fp80 @llvm.trunc.f80(x86_fp80 %Val) 8872 declare fp128 @llvm.trunc.f128(fp128 %Val) 8873 declare ppc_fp128 @llvm.trunc.ppcf128(ppc_fp128 %Val) 8874 8875Overview: 8876""""""""" 8877 8878The '``llvm.trunc.*``' intrinsics returns the operand rounded to the 8879nearest integer not larger in magnitude than the operand. 8880 8881Arguments: 8882"""""""""" 8883 8884The argument and return value are floating point numbers of the same 8885type. 8886 8887Semantics: 8888"""""""""" 8889 8890This function returns the same values as the libm ``trunc`` functions 8891would, and handles error conditions in the same way. 8892 8893'``llvm.rint.*``' Intrinsic 8894^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8895 8896Syntax: 8897""""""" 8898 8899This is an overloaded intrinsic. You can use ``llvm.rint`` on any 8900floating point or vector of floating point type. Not all targets support 8901all types however. 8902 8903:: 8904 8905 declare float @llvm.rint.f32(float %Val) 8906 declare double @llvm.rint.f64(double %Val) 8907 declare x86_fp80 @llvm.rint.f80(x86_fp80 %Val) 8908 declare fp128 @llvm.rint.f128(fp128 %Val) 8909 declare ppc_fp128 @llvm.rint.ppcf128(ppc_fp128 %Val) 8910 8911Overview: 8912""""""""" 8913 8914The '``llvm.rint.*``' intrinsics returns the operand rounded to the 8915nearest integer. It may raise an inexact floating-point exception if the 8916operand isn't an integer. 8917 8918Arguments: 8919"""""""""" 8920 8921The argument and return value are floating point numbers of the same 8922type. 8923 8924Semantics: 8925"""""""""" 8926 8927This function returns the same values as the libm ``rint`` functions 8928would, and handles error conditions in the same way. 8929 8930'``llvm.nearbyint.*``' Intrinsic 8931^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8932 8933Syntax: 8934""""""" 8935 8936This is an overloaded intrinsic. You can use ``llvm.nearbyint`` on any 8937floating point or vector of floating point type. Not all targets support 8938all types however. 8939 8940:: 8941 8942 declare float @llvm.nearbyint.f32(float %Val) 8943 declare double @llvm.nearbyint.f64(double %Val) 8944 declare x86_fp80 @llvm.nearbyint.f80(x86_fp80 %Val) 8945 declare fp128 @llvm.nearbyint.f128(fp128 %Val) 8946 declare ppc_fp128 @llvm.nearbyint.ppcf128(ppc_fp128 %Val) 8947 8948Overview: 8949""""""""" 8950 8951The '``llvm.nearbyint.*``' intrinsics returns the operand rounded to the 8952nearest integer. 8953 8954Arguments: 8955"""""""""" 8956 8957The argument and return value are floating point numbers of the same 8958type. 8959 8960Semantics: 8961"""""""""" 8962 8963This function returns the same values as the libm ``nearbyint`` 8964functions would, and handles error conditions in the same way. 8965 8966'``llvm.round.*``' Intrinsic 8967^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8968 8969Syntax: 8970""""""" 8971 8972This is an overloaded intrinsic. You can use ``llvm.round`` on any 8973floating point or vector of floating point type. Not all targets support 8974all types however. 8975 8976:: 8977 8978 declare float @llvm.round.f32(float %Val) 8979 declare double @llvm.round.f64(double %Val) 8980 declare x86_fp80 @llvm.round.f80(x86_fp80 %Val) 8981 declare fp128 @llvm.round.f128(fp128 %Val) 8982 declare ppc_fp128 @llvm.round.ppcf128(ppc_fp128 %Val) 8983 8984Overview: 8985""""""""" 8986 8987The '``llvm.round.*``' intrinsics returns the operand rounded to the 8988nearest integer. 8989 8990Arguments: 8991"""""""""" 8992 8993The argument and return value are floating point numbers of the same 8994type. 8995 8996Semantics: 8997"""""""""" 8998 8999This function returns the same values as the libm ``round`` 9000functions would, and handles error conditions in the same way. 9001 9002Bit Manipulation Intrinsics 9003--------------------------- 9004 9005LLVM provides intrinsics for a few important bit manipulation 9006operations. These allow efficient code generation for some algorithms. 9007 9008'``llvm.bswap.*``' Intrinsics 9009^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9010 9011Syntax: 9012""""""" 9013 9014This is an overloaded intrinsic function. You can use bswap on any 9015integer type that is an even number of bytes (i.e. BitWidth % 16 == 0). 9016 9017:: 9018 9019 declare i16 @llvm.bswap.i16(i16 <id>) 9020 declare i32 @llvm.bswap.i32(i32 <id>) 9021 declare i64 @llvm.bswap.i64(i64 <id>) 9022 9023Overview: 9024""""""""" 9025 9026The '``llvm.bswap``' family of intrinsics is used to byte swap integer 9027values with an even number of bytes (positive multiple of 16 bits). 9028These are useful for performing operations on data that is not in the 9029target's native byte order. 9030 9031Semantics: 9032"""""""""" 9033 9034The ``llvm.bswap.i16`` intrinsic returns an i16 value that has the high 9035and low byte of the input i16 swapped. Similarly, the ``llvm.bswap.i32`` 9036intrinsic returns an i32 value that has the four bytes of the input i32 9037swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the 9038returned i32 will have its bytes in 3, 2, 1, 0 order. The 9039``llvm.bswap.i48``, ``llvm.bswap.i64`` and other intrinsics extend this 9040concept to additional even-byte lengths (6 bytes, 8 bytes and more, 9041respectively). 9042 9043'``llvm.ctpop.*``' Intrinsic 9044^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9045 9046Syntax: 9047""""""" 9048 9049This is an overloaded intrinsic. You can use llvm.ctpop on any integer 9050bit width, or on any vector with integer elements. Not all targets 9051support all bit widths or vector types, however. 9052 9053:: 9054 9055 declare i8 @llvm.ctpop.i8(i8 <src>) 9056 declare i16 @llvm.ctpop.i16(i16 <src>) 9057 declare i32 @llvm.ctpop.i32(i32 <src>) 9058 declare i64 @llvm.ctpop.i64(i64 <src>) 9059 declare i256 @llvm.ctpop.i256(i256 <src>) 9060 declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>) 9061 9062Overview: 9063""""""""" 9064 9065The '``llvm.ctpop``' family of intrinsics counts the number of bits set 9066in a value. 9067 9068Arguments: 9069"""""""""" 9070 9071The only argument is the value to be counted. The argument may be of any 9072integer type, or a vector with integer elements. The return type must 9073match the argument type. 9074 9075Semantics: 9076"""""""""" 9077 9078The '``llvm.ctpop``' intrinsic counts the 1's in a variable, or within 9079each element of a vector. 9080 9081'``llvm.ctlz.*``' Intrinsic 9082^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9083 9084Syntax: 9085""""""" 9086 9087This is an overloaded intrinsic. You can use ``llvm.ctlz`` on any 9088integer bit width, or any vector whose elements are integers. Not all 9089targets support all bit widths or vector types, however. 9090 9091:: 9092 9093 declare i8 @llvm.ctlz.i8 (i8 <src>, i1 <is_zero_undef>) 9094 declare i16 @llvm.ctlz.i16 (i16 <src>, i1 <is_zero_undef>) 9095 declare i32 @llvm.ctlz.i32 (i32 <src>, i1 <is_zero_undef>) 9096 declare i64 @llvm.ctlz.i64 (i64 <src>, i1 <is_zero_undef>) 9097 declare i256 @llvm.ctlz.i256(i256 <src>, i1 <is_zero_undef>) 9098 declase <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>) 9099 9100Overview: 9101""""""""" 9102 9103The '``llvm.ctlz``' family of intrinsic functions counts the number of 9104leading zeros in a variable. 9105 9106Arguments: 9107"""""""""" 9108 9109The first argument is the value to be counted. This argument may be of 9110any integer type, or a vector with integer element type. The return 9111type must match the first argument type. 9112 9113The second argument must be a constant and is a flag to indicate whether 9114the intrinsic should ensure that a zero as the first argument produces a 9115defined result. Historically some architectures did not provide a 9116defined result for zero values as efficiently, and many algorithms are 9117now predicated on avoiding zero-value inputs. 9118 9119Semantics: 9120"""""""""" 9121 9122The '``llvm.ctlz``' intrinsic counts the leading (most significant) 9123zeros in a variable, or within each element of the vector. If 9124``src == 0`` then the result is the size in bits of the type of ``src`` 9125if ``is_zero_undef == 0`` and ``undef`` otherwise. For example, 9126``llvm.ctlz(i32 2) = 30``. 9127 9128'``llvm.cttz.*``' Intrinsic 9129^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9130 9131Syntax: 9132""""""" 9133 9134This is an overloaded intrinsic. You can use ``llvm.cttz`` on any 9135integer bit width, or any vector of integer elements. Not all targets 9136support all bit widths or vector types, however. 9137 9138:: 9139 9140 declare i8 @llvm.cttz.i8 (i8 <src>, i1 <is_zero_undef>) 9141 declare i16 @llvm.cttz.i16 (i16 <src>, i1 <is_zero_undef>) 9142 declare i32 @llvm.cttz.i32 (i32 <src>, i1 <is_zero_undef>) 9143 declare i64 @llvm.cttz.i64 (i64 <src>, i1 <is_zero_undef>) 9144 declare i256 @llvm.cttz.i256(i256 <src>, i1 <is_zero_undef>) 9145 declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>) 9146 9147Overview: 9148""""""""" 9149 9150The '``llvm.cttz``' family of intrinsic functions counts the number of 9151trailing zeros. 9152 9153Arguments: 9154"""""""""" 9155 9156The first argument is the value to be counted. This argument may be of 9157any integer type, or a vector with integer element type. The return 9158type must match the first argument type. 9159 9160The second argument must be a constant and is a flag to indicate whether 9161the intrinsic should ensure that a zero as the first argument produces a 9162defined result. Historically some architectures did not provide a 9163defined result for zero values as efficiently, and many algorithms are 9164now predicated on avoiding zero-value inputs. 9165 9166Semantics: 9167"""""""""" 9168 9169The '``llvm.cttz``' intrinsic counts the trailing (least significant) 9170zeros in a variable, or within each element of a vector. If ``src == 0`` 9171then the result is the size in bits of the type of ``src`` if 9172``is_zero_undef == 0`` and ``undef`` otherwise. For example, 9173``llvm.cttz(2) = 1``. 9174 9175.. _int_overflow: 9176 9177Arithmetic with Overflow Intrinsics 9178----------------------------------- 9179 9180LLVM provides intrinsics for some arithmetic with overflow operations. 9181 9182'``llvm.sadd.with.overflow.*``' Intrinsics 9183^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9184 9185Syntax: 9186""""""" 9187 9188This is an overloaded intrinsic. You can use ``llvm.sadd.with.overflow`` 9189on any integer bit width. 9190 9191:: 9192 9193 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b) 9194 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b) 9195 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b) 9196 9197Overview: 9198""""""""" 9199 9200The '``llvm.sadd.with.overflow``' family of intrinsic functions perform 9201a signed addition of the two arguments, and indicate whether an overflow 9202occurred during the signed summation. 9203 9204Arguments: 9205"""""""""" 9206 9207The arguments (%a and %b) and the first element of the result structure 9208may be of integer types of any bit width, but they must have the same 9209bit width. The second element of the result structure must be of type 9210``i1``. ``%a`` and ``%b`` are the two values that will undergo signed 9211addition. 9212 9213Semantics: 9214"""""""""" 9215 9216The '``llvm.sadd.with.overflow``' family of intrinsic functions perform 9217a signed addition of the two variables. They return a structure --- the 9218first element of which is the signed summation, and the second element 9219of which is a bit specifying if the signed summation resulted in an 9220overflow. 9221 9222Examples: 9223""""""""" 9224 9225.. code-block:: llvm 9226 9227 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b) 9228 %sum = extractvalue {i32, i1} %res, 0 9229 %obit = extractvalue {i32, i1} %res, 1 9230 br i1 %obit, label %overflow, label %normal 9231 9232'``llvm.uadd.with.overflow.*``' Intrinsics 9233^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9234 9235Syntax: 9236""""""" 9237 9238This is an overloaded intrinsic. You can use ``llvm.uadd.with.overflow`` 9239on any integer bit width. 9240 9241:: 9242 9243 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b) 9244 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b) 9245 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) 9246 9247Overview: 9248""""""""" 9249 9250The '``llvm.uadd.with.overflow``' family of intrinsic functions perform 9251an unsigned addition of the two arguments, and indicate whether a carry 9252occurred during the unsigned summation. 9253 9254Arguments: 9255"""""""""" 9256 9257The arguments (%a and %b) and the first element of the result structure 9258may be of integer types of any bit width, but they must have the same 9259bit width. The second element of the result structure must be of type 9260``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned 9261addition. 9262 9263Semantics: 9264"""""""""" 9265 9266The '``llvm.uadd.with.overflow``' family of intrinsic functions perform 9267an unsigned addition of the two arguments. They return a structure --- the 9268first element of which is the sum, and the second element of which is a 9269bit specifying if the unsigned summation resulted in a carry. 9270 9271Examples: 9272""""""""" 9273 9274.. code-block:: llvm 9275 9276 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b) 9277 %sum = extractvalue {i32, i1} %res, 0 9278 %obit = extractvalue {i32, i1} %res, 1 9279 br i1 %obit, label %carry, label %normal 9280 9281'``llvm.ssub.with.overflow.*``' Intrinsics 9282^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9283 9284Syntax: 9285""""""" 9286 9287This is an overloaded intrinsic. You can use ``llvm.ssub.with.overflow`` 9288on any integer bit width. 9289 9290:: 9291 9292 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b) 9293 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b) 9294 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) 9295 9296Overview: 9297""""""""" 9298 9299The '``llvm.ssub.with.overflow``' family of intrinsic functions perform 9300a signed subtraction of the two arguments, and indicate whether an 9301overflow occurred during the signed subtraction. 9302 9303Arguments: 9304"""""""""" 9305 9306The arguments (%a and %b) and the first element of the result structure 9307may be of integer types of any bit width, but they must have the same 9308bit width. The second element of the result structure must be of type 9309``i1``. ``%a`` and ``%b`` are the two values that will undergo signed 9310subtraction. 9311 9312Semantics: 9313"""""""""" 9314 9315The '``llvm.ssub.with.overflow``' family of intrinsic functions perform 9316a signed subtraction of the two arguments. They return a structure --- the 9317first element of which is the subtraction, and the second element of 9318which is a bit specifying if the signed subtraction resulted in an 9319overflow. 9320 9321Examples: 9322""""""""" 9323 9324.. code-block:: llvm 9325 9326 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b) 9327 %sum = extractvalue {i32, i1} %res, 0 9328 %obit = extractvalue {i32, i1} %res, 1 9329 br i1 %obit, label %overflow, label %normal 9330 9331'``llvm.usub.with.overflow.*``' Intrinsics 9332^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9333 9334Syntax: 9335""""""" 9336 9337This is an overloaded intrinsic. You can use ``llvm.usub.with.overflow`` 9338on any integer bit width. 9339 9340:: 9341 9342 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b) 9343 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b) 9344 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b) 9345 9346Overview: 9347""""""""" 9348 9349The '``llvm.usub.with.overflow``' family of intrinsic functions perform 9350an unsigned subtraction of the two arguments, and indicate whether an 9351overflow occurred during the unsigned subtraction. 9352 9353Arguments: 9354"""""""""" 9355 9356The arguments (%a and %b) and the first element of the result structure 9357may be of integer types of any bit width, but they must have the same 9358bit width. The second element of the result structure must be of type 9359``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned 9360subtraction. 9361 9362Semantics: 9363"""""""""" 9364 9365The '``llvm.usub.with.overflow``' family of intrinsic functions perform 9366an unsigned subtraction of the two arguments. They return a structure --- 9367the first element of which is the subtraction, and the second element of 9368which is a bit specifying if the unsigned subtraction resulted in an 9369overflow. 9370 9371Examples: 9372""""""""" 9373 9374.. code-block:: llvm 9375 9376 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b) 9377 %sum = extractvalue {i32, i1} %res, 0 9378 %obit = extractvalue {i32, i1} %res, 1 9379 br i1 %obit, label %overflow, label %normal 9380 9381'``llvm.smul.with.overflow.*``' Intrinsics 9382^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9383 9384Syntax: 9385""""""" 9386 9387This is an overloaded intrinsic. You can use ``llvm.smul.with.overflow`` 9388on any integer bit width. 9389 9390:: 9391 9392 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b) 9393 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b) 9394 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b) 9395 9396Overview: 9397""""""""" 9398 9399The '``llvm.smul.with.overflow``' family of intrinsic functions perform 9400a signed multiplication of the two arguments, and indicate whether an 9401overflow occurred during the signed multiplication. 9402 9403Arguments: 9404"""""""""" 9405 9406The arguments (%a and %b) and the first element of the result structure 9407may be of integer types of any bit width, but they must have the same 9408bit width. The second element of the result structure must be of type 9409``i1``. ``%a`` and ``%b`` are the two values that will undergo signed 9410multiplication. 9411 9412Semantics: 9413"""""""""" 9414 9415The '``llvm.smul.with.overflow``' family of intrinsic functions perform 9416a signed multiplication of the two arguments. They return a structure --- 9417the first element of which is the multiplication, and the second element 9418of which is a bit specifying if the signed multiplication resulted in an 9419overflow. 9420 9421Examples: 9422""""""""" 9423 9424.. code-block:: llvm 9425 9426 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b) 9427 %sum = extractvalue {i32, i1} %res, 0 9428 %obit = extractvalue {i32, i1} %res, 1 9429 br i1 %obit, label %overflow, label %normal 9430 9431'``llvm.umul.with.overflow.*``' Intrinsics 9432^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9433 9434Syntax: 9435""""""" 9436 9437This is an overloaded intrinsic. You can use ``llvm.umul.with.overflow`` 9438on any integer bit width. 9439 9440:: 9441 9442 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b) 9443 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b) 9444 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b) 9445 9446Overview: 9447""""""""" 9448 9449The '``llvm.umul.with.overflow``' family of intrinsic functions perform 9450a unsigned multiplication of the two arguments, and indicate whether an 9451overflow occurred during the unsigned multiplication. 9452 9453Arguments: 9454"""""""""" 9455 9456The arguments (%a and %b) and the first element of the result structure 9457may be of integer types of any bit width, but they must have the same 9458bit width. The second element of the result structure must be of type 9459``i1``. ``%a`` and ``%b`` are the two values that will undergo unsigned 9460multiplication. 9461 9462Semantics: 9463"""""""""" 9464 9465The '``llvm.umul.with.overflow``' family of intrinsic functions perform 9466an unsigned multiplication of the two arguments. They return a structure --- 9467the first element of which is the multiplication, and the second 9468element of which is a bit specifying if the unsigned multiplication 9469resulted in an overflow. 9470 9471Examples: 9472""""""""" 9473 9474.. code-block:: llvm 9475 9476 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b) 9477 %sum = extractvalue {i32, i1} %res, 0 9478 %obit = extractvalue {i32, i1} %res, 1 9479 br i1 %obit, label %overflow, label %normal 9480 9481Specialised Arithmetic Intrinsics 9482--------------------------------- 9483 9484'``llvm.fmuladd.*``' Intrinsic 9485^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9486 9487Syntax: 9488""""""" 9489 9490:: 9491 9492 declare float @llvm.fmuladd.f32(float %a, float %b, float %c) 9493 declare double @llvm.fmuladd.f64(double %a, double %b, double %c) 9494 9495Overview: 9496""""""""" 9497 9498The '``llvm.fmuladd.*``' intrinsic functions represent multiply-add 9499expressions that can be fused if the code generator determines that (a) the 9500target instruction set has support for a fused operation, and (b) that the 9501fused operation is more efficient than the equivalent, separate pair of mul 9502and add instructions. 9503 9504Arguments: 9505"""""""""" 9506 9507The '``llvm.fmuladd.*``' intrinsics each take three arguments: two 9508multiplicands, a and b, and an addend c. 9509 9510Semantics: 9511"""""""""" 9512 9513The expression: 9514 9515:: 9516 9517 %0 = call float @llvm.fmuladd.f32(%a, %b, %c) 9518 9519is equivalent to the expression a \* b + c, except that rounding will 9520not be performed between the multiplication and addition steps if the 9521code generator fuses the operations. Fusion is not guaranteed, even if 9522the target platform supports it. If a fused multiply-add is required the 9523corresponding llvm.fma.\* intrinsic function should be used 9524instead. This never sets errno, just as '``llvm.fma.*``'. 9525 9526Examples: 9527""""""""" 9528 9529.. code-block:: llvm 9530 9531 %r2 = call float @llvm.fmuladd.f32(float %a, float %b, float %c) ; yields float:r2 = (a * b) + c 9532 9533Half Precision Floating Point Intrinsics 9534---------------------------------------- 9535 9536For most target platforms, half precision floating point is a 9537storage-only format. This means that it is a dense encoding (in memory) 9538but does not support computation in the format. 9539 9540This means that code must first load the half-precision floating point 9541value as an i16, then convert it to float with 9542:ref:`llvm.convert.from.fp16 <int_convert_from_fp16>`. Computation can 9543then be performed on the float value (including extending to double 9544etc). To store the value back to memory, it is first converted to float 9545if needed, then converted to i16 with 9546:ref:`llvm.convert.to.fp16 <int_convert_to_fp16>`, then storing as an 9547i16 value. 9548 9549.. _int_convert_to_fp16: 9550 9551'``llvm.convert.to.fp16``' Intrinsic 9552^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9553 9554Syntax: 9555""""""" 9556 9557:: 9558 9559 declare i16 @llvm.convert.to.fp16.f32(float %a) 9560 declare i16 @llvm.convert.to.fp16.f64(double %a) 9561 9562Overview: 9563""""""""" 9564 9565The '``llvm.convert.to.fp16``' intrinsic function performs a conversion from a 9566conventional floating point type to half precision floating point format. 9567 9568Arguments: 9569"""""""""" 9570 9571The intrinsic function contains single argument - the value to be 9572converted. 9573 9574Semantics: 9575"""""""""" 9576 9577The '``llvm.convert.to.fp16``' intrinsic function performs a conversion from a 9578conventional floating point format to half precision floating point format. The 9579return value is an ``i16`` which contains the converted number. 9580 9581Examples: 9582""""""""" 9583 9584.. code-block:: llvm 9585 9586 %res = call i16 @llvm.convert.to.fp16.f32(float %a) 9587 store i16 %res, i16* @x, align 2 9588 9589.. _int_convert_from_fp16: 9590 9591'``llvm.convert.from.fp16``' Intrinsic 9592^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9593 9594Syntax: 9595""""""" 9596 9597:: 9598 9599 declare float @llvm.convert.from.fp16.f32(i16 %a) 9600 declare double @llvm.convert.from.fp16.f64(i16 %a) 9601 9602Overview: 9603""""""""" 9604 9605The '``llvm.convert.from.fp16``' intrinsic function performs a 9606conversion from half precision floating point format to single precision 9607floating point format. 9608 9609Arguments: 9610"""""""""" 9611 9612The intrinsic function contains single argument - the value to be 9613converted. 9614 9615Semantics: 9616"""""""""" 9617 9618The '``llvm.convert.from.fp16``' intrinsic function performs a 9619conversion from half single precision floating point format to single 9620precision floating point format. The input half-float value is 9621represented by an ``i16`` value. 9622 9623Examples: 9624""""""""" 9625 9626.. code-block:: llvm 9627 9628 %a = load i16, i16* @x, align 2 9629 %res = call float @llvm.convert.from.fp16(i16 %a) 9630 9631.. _dbg_intrinsics: 9632 9633Debugger Intrinsics 9634------------------- 9635 9636The LLVM debugger intrinsics (which all start with ``llvm.dbg.`` 9637prefix), are described in the `LLVM Source Level 9638Debugging <SourceLevelDebugging.html#format_common_intrinsics>`_ 9639document. 9640 9641Exception Handling Intrinsics 9642----------------------------- 9643 9644The LLVM exception handling intrinsics (which all start with 9645``llvm.eh.`` prefix), are described in the `LLVM Exception 9646Handling <ExceptionHandling.html#format_common_intrinsics>`_ document. 9647 9648.. _int_trampoline: 9649 9650Trampoline Intrinsics 9651--------------------- 9652 9653These intrinsics make it possible to excise one parameter, marked with 9654the :ref:`nest <nest>` attribute, from a function. The result is a 9655callable function pointer lacking the nest parameter - the caller does 9656not need to provide a value for it. Instead, the value to use is stored 9657in advance in a "trampoline", a block of memory usually allocated on the 9658stack, which also contains code to splice the nest value into the 9659argument list. This is used to implement the GCC nested function address 9660extension. 9661 9662For example, if the function is ``i32 f(i8* nest %c, i32 %x, i32 %y)`` 9663then the resulting function pointer has signature ``i32 (i32, i32)*``. 9664It can be created as follows: 9665 9666.. code-block:: llvm 9667 9668 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86 9669 %tramp1 = getelementptr [10 x i8], [10 x i8]* %tramp, i32 0, i32 0 9670 call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval) 9671 %p = call i8* @llvm.adjust.trampoline(i8* %tramp1) 9672 %fp = bitcast i8* %p to i32 (i32, i32)* 9673 9674The call ``%val = call i32 %fp(i32 %x, i32 %y)`` is then equivalent to 9675``%val = call i32 %f(i8* %nval, i32 %x, i32 %y)``. 9676 9677.. _int_it: 9678 9679'``llvm.init.trampoline``' Intrinsic 9680^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9681 9682Syntax: 9683""""""" 9684 9685:: 9686 9687 declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>) 9688 9689Overview: 9690""""""""" 9691 9692This fills the memory pointed to by ``tramp`` with executable code, 9693turning it into a trampoline. 9694 9695Arguments: 9696"""""""""" 9697 9698The ``llvm.init.trampoline`` intrinsic takes three arguments, all 9699pointers. The ``tramp`` argument must point to a sufficiently large and 9700sufficiently aligned block of memory; this memory is written to by the 9701intrinsic. Note that the size and the alignment are target-specific - 9702LLVM currently provides no portable way of determining them, so a 9703front-end that generates this intrinsic needs to have some 9704target-specific knowledge. The ``func`` argument must hold a function 9705bitcast to an ``i8*``. 9706 9707Semantics: 9708"""""""""" 9709 9710The block of memory pointed to by ``tramp`` is filled with target 9711dependent code, turning it into a function. Then ``tramp`` needs to be 9712passed to :ref:`llvm.adjust.trampoline <int_at>` to get a pointer which can 9713be :ref:`bitcast (to a new function) and called <int_trampoline>`. The new 9714function's signature is the same as that of ``func`` with any arguments 9715marked with the ``nest`` attribute removed. At most one such ``nest`` 9716argument is allowed, and it must be of pointer type. Calling the new 9717function is equivalent to calling ``func`` with the same argument list, 9718but with ``nval`` used for the missing ``nest`` argument. If, after 9719calling ``llvm.init.trampoline``, the memory pointed to by ``tramp`` is 9720modified, then the effect of any later call to the returned function 9721pointer is undefined. 9722 9723.. _int_at: 9724 9725'``llvm.adjust.trampoline``' Intrinsic 9726^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9727 9728Syntax: 9729""""""" 9730 9731:: 9732 9733 declare i8* @llvm.adjust.trampoline(i8* <tramp>) 9734 9735Overview: 9736""""""""" 9737 9738This performs any required machine-specific adjustment to the address of 9739a trampoline (passed as ``tramp``). 9740 9741Arguments: 9742"""""""""" 9743 9744``tramp`` must point to a block of memory which already has trampoline 9745code filled in by a previous call to 9746:ref:`llvm.init.trampoline <int_it>`. 9747 9748Semantics: 9749"""""""""" 9750 9751On some architectures the address of the code to be executed needs to be 9752different than the address where the trampoline is actually stored. This 9753intrinsic returns the executable address corresponding to ``tramp`` 9754after performing the required machine specific adjustments. The pointer 9755returned can then be :ref:`bitcast and executed <int_trampoline>`. 9756 9757Masked Vector Load and Store Intrinsics 9758--------------------------------------- 9759 9760LLVM 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. 9761 9762.. _int_mload: 9763 9764'``llvm.masked.load.*``' Intrinsics 9765^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9766 9767Syntax: 9768""""""" 9769This is an overloaded intrinsic. The loaded data is a vector of any integer or floating point data type. 9770 9771:: 9772 9773 declare <16 x float> @llvm.masked.load.v16f32 (<16 x float>* <ptr>, i32 <alignment>, <16 x i1> <mask>, <16 x float> <passthru>) 9774 declare <2 x double> @llvm.masked.load.v2f64 (<2 x double>* <ptr>, i32 <alignment>, <2 x i1> <mask>, <2 x double> <passthru>) 9775 9776Overview: 9777""""""""" 9778 9779Reads 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 in the passthru operand. 9780 9781 9782Arguments: 9783"""""""""" 9784 9785The 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 'i1' 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 passthru operand are the same vector types. 9786 9787 9788Semantics: 9789"""""""""" 9790 9791The '``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. 9792The 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. 9793 9794 9795:: 9796 9797 %res = call <16 x float> @llvm.masked.load.v16f32 (<16 x float>* %ptr, i32 4, <16 x i1>%mask, <16 x float> %passthru) 9798 9799 ;; The result of the two following instructions is identical aside from potential memory access exception 9800 %loadlal = load <16 x float>, <16 x float>* %ptr, align 4 9801 %res = select <16 x i1> %mask, <16 x float> %loadlal, <16 x float> %passthru 9802 9803.. _int_mstore: 9804 9805'``llvm.masked.store.*``' Intrinsics 9806^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9807 9808Syntax: 9809""""""" 9810This is an overloaded intrinsic. The data stored in memory is a vector of any integer or floating point data type. 9811 9812:: 9813 9814 declare void @llvm.masked.store.v8i32 (<8 x i32> <value>, <8 x i32> * <ptr>, i32 <alignment>, <8 x i1> <mask>) 9815 declare void @llvm.masked.store.v16f32(<16 x i32> <value>, <16 x i32>* <ptr>, i32 <alignment>, <16 x i1> <mask>) 9816 9817Overview: 9818""""""""" 9819 9820Writes 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. 9821 9822Arguments: 9823"""""""""" 9824 9825The 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. 9826 9827 9828Semantics: 9829"""""""""" 9830 9831The '``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. 9832The 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. 9833 9834:: 9835 9836 call void @llvm.masked.store.v16f32(<16 x float> %value, <16 x float>* %ptr, i32 4, <16 x i1> %mask) 9837 9838 ;; The result of the following instructions is identical aside from potential data races and memory access exceptions 9839 %oldval = load <16 x float>, <16 x float>* %ptr, align 4 9840 %res = select <16 x i1> %mask, <16 x float> %value, <16 x float> %oldval 9841 store <16 x float> %res, <16 x float>* %ptr, align 4 9842 9843 9844Memory Use Markers 9845------------------ 9846 9847This class of intrinsics provides information about the lifetime of 9848memory objects and ranges where variables are immutable. 9849 9850.. _int_lifestart: 9851 9852'``llvm.lifetime.start``' Intrinsic 9853^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9854 9855Syntax: 9856""""""" 9857 9858:: 9859 9860 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>) 9861 9862Overview: 9863""""""""" 9864 9865The '``llvm.lifetime.start``' intrinsic specifies the start of a memory 9866object's lifetime. 9867 9868Arguments: 9869"""""""""" 9870 9871The first argument is a constant integer representing the size of the 9872object, or -1 if it is variable sized. The second argument is a pointer 9873to the object. 9874 9875Semantics: 9876"""""""""" 9877 9878This intrinsic indicates that before this point in the code, the value 9879of the memory pointed to by ``ptr`` is dead. This means that it is known 9880to never be used and has an undefined value. A load from the pointer 9881that precedes this intrinsic can be replaced with ``'undef'``. 9882 9883.. _int_lifeend: 9884 9885'``llvm.lifetime.end``' Intrinsic 9886^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9887 9888Syntax: 9889""""""" 9890 9891:: 9892 9893 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>) 9894 9895Overview: 9896""""""""" 9897 9898The '``llvm.lifetime.end``' intrinsic specifies the end of a memory 9899object's lifetime. 9900 9901Arguments: 9902"""""""""" 9903 9904The first argument is a constant integer representing the size of the 9905object, or -1 if it is variable sized. The second argument is a pointer 9906to the object. 9907 9908Semantics: 9909"""""""""" 9910 9911This intrinsic indicates that after this point in the code, the value of 9912the memory pointed to by ``ptr`` is dead. This means that it is known to 9913never be used and has an undefined value. Any stores into the memory 9914object following this intrinsic may be removed as dead. 9915 9916'``llvm.invariant.start``' Intrinsic 9917^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9918 9919Syntax: 9920""""""" 9921 9922:: 9923 9924 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>) 9925 9926Overview: 9927""""""""" 9928 9929The '``llvm.invariant.start``' intrinsic specifies that the contents of 9930a memory object will not change. 9931 9932Arguments: 9933"""""""""" 9934 9935The first argument is a constant integer representing the size of the 9936object, or -1 if it is variable sized. The second argument is a pointer 9937to the object. 9938 9939Semantics: 9940"""""""""" 9941 9942This intrinsic indicates that until an ``llvm.invariant.end`` that uses 9943the return value, the referenced memory location is constant and 9944unchanging. 9945 9946'``llvm.invariant.end``' Intrinsic 9947^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9948 9949Syntax: 9950""""""" 9951 9952:: 9953 9954 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>) 9955 9956Overview: 9957""""""""" 9958 9959The '``llvm.invariant.end``' intrinsic specifies that the contents of a 9960memory object are mutable. 9961 9962Arguments: 9963"""""""""" 9964 9965The first argument is the matching ``llvm.invariant.start`` intrinsic. 9966The second argument is a constant integer representing the size of the 9967object, or -1 if it is variable sized and the third argument is a 9968pointer to the object. 9969 9970Semantics: 9971"""""""""" 9972 9973This intrinsic indicates that the memory is mutable again. 9974 9975General Intrinsics 9976------------------ 9977 9978This class of intrinsics is designed to be generic and has no specific 9979purpose. 9980 9981'``llvm.var.annotation``' Intrinsic 9982^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9983 9984Syntax: 9985""""""" 9986 9987:: 9988 9989 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>) 9990 9991Overview: 9992""""""""" 9993 9994The '``llvm.var.annotation``' intrinsic. 9995 9996Arguments: 9997"""""""""" 9998 9999The first argument is a pointer to a value, the second is a pointer to a 10000global string, the third is a pointer to a global string which is the 10001source file name, and the last argument is the line number. 10002 10003Semantics: 10004"""""""""" 10005 10006This intrinsic allows annotation of local variables with arbitrary 10007strings. This can be useful for special purpose optimizations that want 10008to look for these annotations. These have no other defined use; they are 10009ignored by code generation and optimization. 10010 10011'``llvm.ptr.annotation.*``' Intrinsic 10012^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10013 10014Syntax: 10015""""""" 10016 10017This is an overloaded intrinsic. You can use '``llvm.ptr.annotation``' on a 10018pointer to an integer of any width. *NOTE* you must specify an address space for 10019the pointer. The identifier for the default address space is the integer 10020'``0``'. 10021 10022:: 10023 10024 declare i8* @llvm.ptr.annotation.p<address space>i8(i8* <val>, i8* <str>, i8* <str>, i32 <int>) 10025 declare i16* @llvm.ptr.annotation.p<address space>i16(i16* <val>, i8* <str>, i8* <str>, i32 <int>) 10026 declare i32* @llvm.ptr.annotation.p<address space>i32(i32* <val>, i8* <str>, i8* <str>, i32 <int>) 10027 declare i64* @llvm.ptr.annotation.p<address space>i64(i64* <val>, i8* <str>, i8* <str>, i32 <int>) 10028 declare i256* @llvm.ptr.annotation.p<address space>i256(i256* <val>, i8* <str>, i8* <str>, i32 <int>) 10029 10030Overview: 10031""""""""" 10032 10033The '``llvm.ptr.annotation``' intrinsic. 10034 10035Arguments: 10036"""""""""" 10037 10038The first argument is a pointer to an integer value of arbitrary bitwidth 10039(result of some expression), the second is a pointer to a global string, the 10040third is a pointer to a global string which is the source file name, and the 10041last argument is the line number. It returns the value of the first argument. 10042 10043Semantics: 10044"""""""""" 10045 10046This intrinsic allows annotation of a pointer to an integer with arbitrary 10047strings. This can be useful for special purpose optimizations that want to look 10048for these annotations. These have no other defined use; they are ignored by code 10049generation and optimization. 10050 10051'``llvm.annotation.*``' Intrinsic 10052^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10053 10054Syntax: 10055""""""" 10056 10057This is an overloaded intrinsic. You can use '``llvm.annotation``' on 10058any integer bit width. 10059 10060:: 10061 10062 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>) 10063 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>) 10064 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>) 10065 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>) 10066 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>) 10067 10068Overview: 10069""""""""" 10070 10071The '``llvm.annotation``' intrinsic. 10072 10073Arguments: 10074"""""""""" 10075 10076The first argument is an integer value (result of some expression), the 10077second is a pointer to a global string, the third is a pointer to a 10078global string which is the source file name, and the last argument is 10079the line number. It returns the value of the first argument. 10080 10081Semantics: 10082"""""""""" 10083 10084This intrinsic allows annotations to be put on arbitrary expressions 10085with arbitrary strings. This can be useful for special purpose 10086optimizations that want to look for these annotations. These have no 10087other defined use; they are ignored by code generation and optimization. 10088 10089'``llvm.trap``' Intrinsic 10090^^^^^^^^^^^^^^^^^^^^^^^^^ 10091 10092Syntax: 10093""""""" 10094 10095:: 10096 10097 declare void @llvm.trap() noreturn nounwind 10098 10099Overview: 10100""""""""" 10101 10102The '``llvm.trap``' intrinsic. 10103 10104Arguments: 10105"""""""""" 10106 10107None. 10108 10109Semantics: 10110"""""""""" 10111 10112This intrinsic is lowered to the target dependent trap instruction. If 10113the target does not have a trap instruction, this intrinsic will be 10114lowered to a call of the ``abort()`` function. 10115 10116'``llvm.debugtrap``' Intrinsic 10117^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10118 10119Syntax: 10120""""""" 10121 10122:: 10123 10124 declare void @llvm.debugtrap() nounwind 10125 10126Overview: 10127""""""""" 10128 10129The '``llvm.debugtrap``' intrinsic. 10130 10131Arguments: 10132"""""""""" 10133 10134None. 10135 10136Semantics: 10137"""""""""" 10138 10139This intrinsic is lowered to code which is intended to cause an 10140execution trap with the intention of requesting the attention of a 10141debugger. 10142 10143'``llvm.stackprotector``' Intrinsic 10144^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10145 10146Syntax: 10147""""""" 10148 10149:: 10150 10151 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>) 10152 10153Overview: 10154""""""""" 10155 10156The ``llvm.stackprotector`` intrinsic takes the ``guard`` and stores it 10157onto the stack at ``slot``. The stack slot is adjusted to ensure that it 10158is placed on the stack before local variables. 10159 10160Arguments: 10161"""""""""" 10162 10163The ``llvm.stackprotector`` intrinsic requires two pointer arguments. 10164The first argument is the value loaded from the stack guard 10165``@__stack_chk_guard``. The second variable is an ``alloca`` that has 10166enough space to hold the value of the guard. 10167 10168Semantics: 10169"""""""""" 10170 10171This intrinsic causes the prologue/epilogue inserter to force the position of 10172the ``AllocaInst`` stack slot to be before local variables on the stack. This is 10173to ensure that if a local variable on the stack is overwritten, it will destroy 10174the value of the guard. When the function exits, the guard on the stack is 10175checked against the original guard by ``llvm.stackprotectorcheck``. If they are 10176different, then ``llvm.stackprotectorcheck`` causes the program to abort by 10177calling the ``__stack_chk_fail()`` function. 10178 10179'``llvm.stackprotectorcheck``' Intrinsic 10180^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10181 10182Syntax: 10183""""""" 10184 10185:: 10186 10187 declare void @llvm.stackprotectorcheck(i8** <guard>) 10188 10189Overview: 10190""""""""" 10191 10192The ``llvm.stackprotectorcheck`` intrinsic compares ``guard`` against an already 10193created stack protector and if they are not equal calls the 10194``__stack_chk_fail()`` function. 10195 10196Arguments: 10197"""""""""" 10198 10199The ``llvm.stackprotectorcheck`` intrinsic requires one pointer argument, the 10200the variable ``@__stack_chk_guard``. 10201 10202Semantics: 10203"""""""""" 10204 10205This intrinsic is provided to perform the stack protector check by comparing 10206``guard`` with the stack slot created by ``llvm.stackprotector`` and if the 10207values do not match call the ``__stack_chk_fail()`` function. 10208 10209The reason to provide this as an IR level intrinsic instead of implementing it 10210via other IR operations is that in order to perform this operation at the IR 10211level without an intrinsic, one would need to create additional basic blocks to 10212handle the success/failure cases. This makes it difficult to stop the stack 10213protector check from disrupting sibling tail calls in Codegen. With this 10214intrinsic, we are able to generate the stack protector basic blocks late in 10215codegen after the tail call decision has occurred. 10216 10217'``llvm.objectsize``' Intrinsic 10218^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10219 10220Syntax: 10221""""""" 10222 10223:: 10224 10225 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <min>) 10226 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <min>) 10227 10228Overview: 10229""""""""" 10230 10231The ``llvm.objectsize`` intrinsic is designed to provide information to 10232the optimizers to determine at compile time whether a) an operation 10233(like memcpy) will overflow a buffer that corresponds to an object, or 10234b) that a runtime check for overflow isn't necessary. An object in this 10235context means an allocation of a specific class, structure, array, or 10236other object. 10237 10238Arguments: 10239"""""""""" 10240 10241The ``llvm.objectsize`` intrinsic takes two arguments. The first 10242argument is a pointer to or into the ``object``. The second argument is 10243a boolean and determines whether ``llvm.objectsize`` returns 0 (if true) 10244or -1 (if false) when the object size is unknown. The second argument 10245only accepts constants. 10246 10247Semantics: 10248"""""""""" 10249 10250The ``llvm.objectsize`` intrinsic is lowered to a constant representing 10251the size of the object concerned. If the size cannot be determined at 10252compile time, ``llvm.objectsize`` returns ``i32/i64 -1 or 0`` (depending 10253on the ``min`` argument). 10254 10255'``llvm.expect``' Intrinsic 10256^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10257 10258Syntax: 10259""""""" 10260 10261This is an overloaded intrinsic. You can use ``llvm.expect`` on any 10262integer bit width. 10263 10264:: 10265 10266 declare i1 @llvm.expect.i1(i1 <val>, i1 <expected_val>) 10267 declare i32 @llvm.expect.i32(i32 <val>, i32 <expected_val>) 10268 declare i64 @llvm.expect.i64(i64 <val>, i64 <expected_val>) 10269 10270Overview: 10271""""""""" 10272 10273The ``llvm.expect`` intrinsic provides information about expected (the 10274most probable) value of ``val``, which can be used by optimizers. 10275 10276Arguments: 10277"""""""""" 10278 10279The ``llvm.expect`` intrinsic takes two arguments. The first argument is 10280a value. The second argument is an expected value, this needs to be a 10281constant value, variables are not allowed. 10282 10283Semantics: 10284"""""""""" 10285 10286This intrinsic is lowered to the ``val``. 10287 10288'``llvm.assume``' Intrinsic 10289^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10290 10291Syntax: 10292""""""" 10293 10294:: 10295 10296 declare void @llvm.assume(i1 %cond) 10297 10298Overview: 10299""""""""" 10300 10301The ``llvm.assume`` allows the optimizer to assume that the provided 10302condition is true. This information can then be used in simplifying other parts 10303of the code. 10304 10305Arguments: 10306"""""""""" 10307 10308The condition which the optimizer may assume is always true. 10309 10310Semantics: 10311"""""""""" 10312 10313The intrinsic allows the optimizer to assume that the provided condition is 10314always true whenever the control flow reaches the intrinsic call. No code is 10315generated for this intrinsic, and instructions that contribute only to the 10316provided condition are not used for code generation. If the condition is 10317violated during execution, the behavior is undefined. 10318 10319Note that the optimizer might limit the transformations performed on values 10320used by the ``llvm.assume`` intrinsic in order to preserve the instructions 10321only used to form the intrinsic's input argument. This might prove undesirable 10322if the extra information provided by the ``llvm.assume`` intrinsic does not cause 10323sufficient overall improvement in code quality. For this reason, 10324``llvm.assume`` should not be used to document basic mathematical invariants 10325that the optimizer can otherwise deduce or facts that are of little use to the 10326optimizer. 10327 10328.. _bitset.test: 10329 10330'``llvm.bitset.test``' Intrinsic 10331^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10332 10333Syntax: 10334""""""" 10335 10336:: 10337 10338 declare i1 @llvm.bitset.test(i8* %ptr, metadata %bitset) nounwind readnone 10339 10340 10341Arguments: 10342"""""""""" 10343 10344The first argument is a pointer to be tested. The second argument is a 10345metadata string containing the name of a :doc:`bitset <BitSets>`. 10346 10347Overview: 10348""""""""" 10349 10350The ``llvm.bitset.test`` intrinsic tests whether the given pointer is a 10351member of the given bitset. 10352 10353'``llvm.donothing``' Intrinsic 10354^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 10355 10356Syntax: 10357""""""" 10358 10359:: 10360 10361 declare void @llvm.donothing() nounwind readnone 10362 10363Overview: 10364""""""""" 10365 10366The ``llvm.donothing`` intrinsic doesn't perform any operation. It's one of only 10367two intrinsics (besides ``llvm.experimental.patchpoint``) that can be called 10368with an invoke instruction. 10369 10370Arguments: 10371"""""""""" 10372 10373None. 10374 10375Semantics: 10376"""""""""" 10377 10378This intrinsic does nothing, and it's removed by optimizers and ignored 10379by codegen. 10380 10381Stack Map Intrinsics 10382-------------------- 10383 10384LLVM provides experimental intrinsics to support runtime patching 10385mechanisms commonly desired in dynamic language JITs. These intrinsics 10386are described in :doc:`StackMaps`. 10387