Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 1 | <html> |
| 2 | <head> |
| 3 | <title>Clang Language Extensions</title> |
| 4 | <link type="text/css" rel="stylesheet" href="../menu.css" /> |
| 5 | <link type="text/css" rel="stylesheet" href="../content.css" /> |
| 6 | <style type="text/css"> |
| 7 | td { |
| 8 | vertical-align: top; |
| 9 | } |
| 10 | </style> |
| 11 | </head> |
| 12 | <body> |
| 13 | |
| 14 | <!--#include virtual="../menu.html.incl"--> |
| 15 | |
| 16 | <div id="content"> |
| 17 | |
| 18 | <h1>Clang Language Extensions</h1> |
| 19 | |
| 20 | <ul> |
| 21 | <li><a href="#intro">Introduction</a></li> |
Chris Lattner | 148772a | 2009-06-13 07:13:28 +0000 | [diff] [blame] | 22 | <li><a href="#feature_check">Feature Checking Macros</a></li> |
John Thompson | 92bd8c7 | 2009-11-02 22:28:12 +0000 | [diff] [blame] | 23 | <li><a href="#has_include">Include File Checking Macros</a></li> |
Chris Lattner | 81edc9f | 2009-04-13 02:45:46 +0000 | [diff] [blame] | 24 | <li><a href="#builtinmacros">Builtin Macros</a></li> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 25 | <li><a href="#vectors">Vectors and Extended Vectors</a></li> |
Ted Kremenek | 87774fd | 2009-12-03 02:04:01 +0000 | [diff] [blame^] | 26 | <li><a href="#checking_language_features">Checks for Standard Language Features</a></li> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 27 | <li><a href="#blocks">Blocks</a></li> |
Douglas Gregor | cb54d43 | 2009-02-13 00:57:04 +0000 | [diff] [blame] | 28 | <li><a href="#overloading-in-c">Function Overloading in C</a></li> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 29 | <li><a href="#builtins">Builtin Functions</a> |
| 30 | <ul> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 31 | <li><a href="#__builtin_shufflevector">__builtin_shufflevector</a></li> |
Chris Lattner | 21190d5 | 2009-09-21 03:09:59 +0000 | [diff] [blame] | 32 | <li><a href="#__builtin_unreachable">__builtin_unreachable</a></li> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 33 | </ul> |
| 34 | </li> |
Chris Lattner | 1177f91 | 2009-04-09 19:58:15 +0000 | [diff] [blame] | 35 | <li><a href="#targetspecific">Target-Specific Extensions</a> |
| 36 | <ul> |
| 37 | <li><a href="#x86-specific">X86/X86-64 Language Extensions</a></li> |
| 38 | </ul> |
| 39 | </li> |
Ted Kremenek | ed86931 | 2009-04-10 05:03:33 +0000 | [diff] [blame] | 40 | <li><a href="#analyzerspecific">Static Analysis-Specific Extensions</a> |
| 41 | <ul> |
| 42 | <li><a href="#analyzerattributes">Analyzer Attributes</a></li> |
| 43 | </ul> |
| 44 | </li> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 45 | </ul> |
| 46 | |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 47 | <!-- ======================================================================= --> |
| 48 | <h2 id="intro">Introduction</h2> |
| 49 | <!-- ======================================================================= --> |
| 50 | |
| 51 | <p>This document describes the language extensions provided by Clang. In |
Chris Lattner | 148772a | 2009-06-13 07:13:28 +0000 | [diff] [blame] | 52 | addition to the language extensions listed here, Clang aims to support a broad |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 53 | range of GCC extensions. Please see the <a |
| 54 | href="http://gcc.gnu.org/onlinedocs/gcc/C-Extensions.html">GCC manual</a> for |
| 55 | more information on these extensions.</p> |
| 56 | |
| 57 | <!-- ======================================================================= --> |
Chris Lattner | 148772a | 2009-06-13 07:13:28 +0000 | [diff] [blame] | 58 | <h2 id="feature_check">Feature Checking Macros</h2> |
| 59 | <!-- ======================================================================= --> |
| 60 | |
| 61 | <p>Language extensions can be very useful, but only if you know you can depend |
| 62 | on them. In order to allow fine-grain features checks, we support two builtin |
| 63 | function-like macros. This allows you to directly test for a feature in your |
| 64 | code without having to resort to something like autoconf or fragile "compiler |
| 65 | version checks".</p> |
| 66 | |
| 67 | <!-- ======================================================================= --> |
| 68 | <h3 id="__has_builtin">__has_builtin</h3> |
| 69 | <!-- ======================================================================= --> |
| 70 | |
| 71 | <p>This function-like macro takes a single identifier argument that is the name |
| 72 | of a builtin function. It evaluates to 1 if the builtin is supported or 0 if |
| 73 | not. It can be used like this:</p> |
| 74 | |
| 75 | <blockquote> |
| 76 | <pre> |
| 77 | #ifndef __has_builtin // Optional of course. |
| 78 | #define __has_builtin(x) 0 // Compatibility with non-clang compilers. |
| 79 | #endif |
| 80 | |
| 81 | ... |
| 82 | #if __has_builtin(__builtin_trap) |
| 83 | __builtin_trap(); |
| 84 | #else |
| 85 | abort(); |
| 86 | #endif |
| 87 | ... |
| 88 | </pre> |
| 89 | </blockquote> |
| 90 | |
| 91 | |
| 92 | <!-- ======================================================================= --> |
| 93 | <h3 id="__has_feature">__has_feature</h3> |
| 94 | <!-- ======================================================================= --> |
| 95 | |
| 96 | <p>This function-like macro takes a single identifier argument that is the name |
| 97 | of a feature. It evaluates to 1 if the feature is supported or 0 if not. It |
| 98 | can be used like this:</p> |
| 99 | |
| 100 | <blockquote> |
| 101 | <pre> |
| 102 | #ifndef __has_feature // Optional of course. |
| 103 | #define __has_feature(x) 0 // Compatibility with non-clang compilers. |
| 104 | #endif |
| 105 | |
| 106 | ... |
| 107 | #if __has_feature(attribute_overloadable) || \ |
| 108 | __has_feature(blocks) |
| 109 | ... |
| 110 | #endif |
| 111 | ... |
| 112 | </pre> |
| 113 | </blockquote> |
| 114 | |
| 115 | <p>The feature tag is described along with the language feature below.</p> |
| 116 | |
John Thompson | 92bd8c7 | 2009-11-02 22:28:12 +0000 | [diff] [blame] | 117 | <!-- ======================================================================= --> |
| 118 | <h2 id="has_include">Include File Checking Macros</h2> |
| 119 | <!-- ======================================================================= --> |
| 120 | |
| 121 | <p>Not all developments systems have the same include files. |
| 122 | The <a href="#__has_include">__has_include</a> and |
| 123 | <a href="#__has_include_next">__has_include_next</a> macros allow you to |
| 124 | check for the existence of an include file before doing |
| 125 | a possibly failing #include directive.</p> |
| 126 | |
| 127 | <!-- ======================================================================= --> |
| 128 | <h3 id="__has_include">__has_include</h3> |
| 129 | <!-- ======================================================================= --> |
| 130 | |
| 131 | <p>This function-like macro takes a single file name string argument that |
| 132 | is the name of an include file. It evaluates to 1 if the file can |
| 133 | be found using the include paths, or 0 otherwise:</p> |
| 134 | |
| 135 | <blockquote> |
| 136 | <pre> |
| 137 | // Note the two possible file name string formats. |
| 138 | #if __has_include("myinclude.h") && __has_include(<stdint.h>) |
| 139 | # include "myinclude.h" |
| 140 | #endif |
| 141 | |
| 142 | // To avoid problem with non-clang compilers not having this macro. |
| 143 | #if defined(__has_include) && __has_include("myinclude.h") |
| 144 | # include "myinclude.h" |
| 145 | #endif |
| 146 | </pre> |
| 147 | </blockquote> |
| 148 | |
| 149 | <p>To test for this feature, use #if defined(__has_include).</p> |
| 150 | |
| 151 | <!-- ======================================================================= --> |
| 152 | <h3 id="__has_include_next">__has_include_next</h3> |
| 153 | <!-- ======================================================================= --> |
| 154 | |
| 155 | <p>This function-like macro takes a single file name string argument that |
| 156 | is the name of an include file. It is like __has_include except that it |
| 157 | looks for the second instance of the given file found in the include |
| 158 | paths. It evaluates to 1 if the second instance of the file can |
| 159 | be found using the include paths, or 0 otherwise:</p> |
| 160 | |
| 161 | <blockquote> |
| 162 | <pre> |
| 163 | // Note the two possible file name string formats. |
| 164 | #if __has_include_next("myinclude.h") && __has_include_next(<stdint.h>) |
| 165 | # include_next "myinclude.h" |
| 166 | #endif |
| 167 | |
| 168 | // To avoid problem with non-clang compilers not having this macro. |
| 169 | #if defined(__has_include_next) && __has_include_next("myinclude.h") |
| 170 | # include_next "myinclude.h" |
| 171 | #endif |
| 172 | </pre> |
| 173 | </blockquote> |
| 174 | |
| 175 | <p>Note that __has_include_next, like the GNU extension |
| 176 | #include_next directive, is intended for use in headers only, |
| 177 | and will issue a warning if used in the top-level compilation |
| 178 | file. A warning will also be issued if an absolute path |
| 179 | is used in the file argument.</p> |
Chris Lattner | 148772a | 2009-06-13 07:13:28 +0000 | [diff] [blame] | 180 | |
| 181 | <!-- ======================================================================= --> |
Chris Lattner | 81edc9f | 2009-04-13 02:45:46 +0000 | [diff] [blame] | 182 | <h2 id="builtinmacros">Builtin Macros</h2> |
| 183 | <!-- ======================================================================= --> |
| 184 | |
| 185 | <p>__BASE_FILE__, __INCLUDE_LEVEL__, __TIMESTAMP__, __COUNTER__</p> |
| 186 | |
| 187 | <!-- ======================================================================= --> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 188 | <h2 id="vectors">Vectors and Extended Vectors</h2> |
| 189 | <!-- ======================================================================= --> |
| 190 | |
| 191 | <p>Supports the GCC vector extensions, plus some stuff like V[1]. ext_vector |
| 192 | with V.xyzw syntax and other tidbits. See also <a |
| 193 | href="#__builtin_shufflevector">__builtin_shufflevector</a>.</p> |
| 194 | |
Chris Lattner | 148772a | 2009-06-13 07:13:28 +0000 | [diff] [blame] | 195 | <p>Query for this feature with __has_feature(attribute_ext_vector_type).</p> |
| 196 | |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 197 | <!-- ======================================================================= --> |
Ted Kremenek | 87774fd | 2009-12-03 02:04:01 +0000 | [diff] [blame^] | 198 | <h2 id="checking_language_features">Checks for Standard Language Features</h2> |
| 199 | <!-- ======================================================================= --> |
| 200 | |
| 201 | <p>The <tt>__has_feature</tt> macro can be used to query if certain standard language features are |
| 202 | enabled. Those features are listed here.</p> |
| 203 | |
| 204 | <h3>C++ exceptions</h3> |
| 205 | |
| 206 | Use <tt>__has_feature(cxx_exceptions)</tt> to determine if C++ exceptions have been enabled. For |
| 207 | example, compiling code with <tt>-fexceptions</tt> enables C++ exceptions. |
| 208 | |
| 209 | <h3>C++ RTTI</h3> |
| 210 | |
| 211 | Use <tt>__has_feature(cxx_rtt)</tt> to determine if C++ RTTI has been enabled. For example, |
| 212 | compiling code with <tt>-fno-rtti</tt> disables the use of RTTI. |
| 213 | |
| 214 | <!-- ======================================================================= --> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 215 | <h2 id="blocks">Blocks</h2> |
| 216 | <!-- ======================================================================= --> |
| 217 | |
Chris Lattner | a7dbdf5 | 2009-03-09 07:03:22 +0000 | [diff] [blame] | 218 | <p>The syntax and high level language feature description is in <a |
| 219 | href="BlockLanguageSpec.txt">BlockLanguageSpec.txt</a>. Implementation and ABI |
| 220 | details for the clang implementation are in <a |
| 221 | href="BlockImplementation.txt">BlockImplementation.txt</a>.</p> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 222 | |
Chris Lattner | 148772a | 2009-06-13 07:13:28 +0000 | [diff] [blame] | 223 | |
| 224 | <p>Query for this feature with __has_feature(blocks).</p> |
| 225 | |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 226 | <!-- ======================================================================= --> |
Douglas Gregor | cb54d43 | 2009-02-13 00:57:04 +0000 | [diff] [blame] | 227 | <h2 id="overloading-in-c">Function Overloading in C</h2> |
| 228 | <!-- ======================================================================= --> |
| 229 | |
Chris Lattner | f161d41 | 2009-02-13 21:51:45 +0000 | [diff] [blame] | 230 | <p>Clang provides support for C++ function overloading in C. Function |
| 231 | overloading in C is introduced using the <tt>overloadable</tt> attribute. For |
| 232 | example, one might provide several overloaded versions of a <tt>tgsin</tt> |
| 233 | function that invokes the appropriate standard function computing the sine of a |
| 234 | value with <tt>float</tt>, <tt>double</tt>, or <tt>long double</tt> |
| 235 | precision:</p> |
Douglas Gregor | cb54d43 | 2009-02-13 00:57:04 +0000 | [diff] [blame] | 236 | |
| 237 | <blockquote> |
| 238 | <pre> |
| 239 | #include <math.h> |
| 240 | float <b>__attribute__((overloadable))</b> tgsin(float x) { return sinf(x); } |
| 241 | double <b>__attribute__((overloadable))</b> tgsin(double x) { return sin(x); } |
| 242 | long double <b>__attribute__((overloadable))</b> tgsin(long double x) { return sinl(x); } |
| 243 | </pre> |
| 244 | </blockquote> |
| 245 | |
| 246 | <p>Given these declarations, one can call <tt>tgsin</tt> with a |
| 247 | <tt>float</tt> value to receive a <tt>float</tt> result, with a |
| 248 | <tt>double</tt> to receive a <tt>double</tt> result, etc. Function |
| 249 | overloading in C follows the rules of C++ function overloading to pick |
| 250 | the best overload given the call arguments, with a few C-specific |
| 251 | semantics:</p> |
| 252 | <ul> |
| 253 | <li>Conversion from <tt>float</tt> or <tt>double</tt> to <tt>long |
| 254 | double</tt> is ranked as a floating-point promotion (per C99) rather |
| 255 | than as a floating-point conversion (as in C++).</li> |
| 256 | |
| 257 | <li>A conversion from a pointer of type <tt>T*</tt> to a pointer of type |
| 258 | <tt>U*</tt> is considered a pointer conversion (with conversion |
| 259 | rank) if <tt>T</tt> and <tt>U</tt> are compatible types.</li> |
| 260 | |
| 261 | <li>A conversion from type <tt>T</tt> to a value of type <tt>U</tt> |
| 262 | is permitted if <tt>T</tt> and <tt>U</tt> are compatible types. This |
| 263 | conversion is given "conversion" rank.</li> |
| 264 | </ul> |
| 265 | |
| 266 | <p>The declaration of <tt>overloadable</tt> functions is restricted to |
| 267 | function declarations and definitions. Most importantly, if any |
| 268 | function with a given name is given the <tt>overloadable</tt> |
| 269 | attribute, then all function declarations and definitions with that |
| 270 | name (and in that scope) must have the <tt>overloadable</tt> |
Chris Lattner | f161d41 | 2009-02-13 21:51:45 +0000 | [diff] [blame] | 271 | attribute. This rule even applies to redeclarations of functions whose original |
| 272 | declaration had the <tt>overloadable</tt> attribute, e.g.,</p> |
Douglas Gregor | cb54d43 | 2009-02-13 00:57:04 +0000 | [diff] [blame] | 273 | |
| 274 | <blockquote> |
| 275 | <pre> |
| 276 | int f(int) __attribute__((overloadable)); |
| 277 | float f(float); <i>// error: declaration of "f" must have the "overloadable" attribute</i> |
| 278 | |
| 279 | int g(int) __attribute__((overloadable)); |
| 280 | int g(int) { } <i>// error: redeclaration of "g" must also have the "overloadable" attribute</i> |
| 281 | </pre> |
| 282 | </blockquote> |
| 283 | |
Douglas Gregor | 965acbb | 2009-02-18 07:07:28 +0000 | [diff] [blame] | 284 | <p>Functions marked <tt>overloadable</tt> must have |
| 285 | prototypes. Therefore, the following code is ill-formed:</p> |
| 286 | |
| 287 | <blockquote> |
| 288 | <pre> |
| 289 | int h() __attribute__((overloadable)); <i>// error: h does not have a prototype</i> |
| 290 | </pre> |
| 291 | </blockquote> |
| 292 | |
| 293 | <p>However, <tt>overloadable</tt> functions are allowed to use a |
| 294 | ellipsis even if there are no named parameters (as is permitted in C++). This feature is particularly useful when combined with the <tt>unavailable</tt> attribute:</p> |
| 295 | |
| 296 | <blockquote> |
| 297 | <pre> |
Chris Lattner | 0224680 | 2009-02-18 22:27:46 +0000 | [diff] [blame] | 298 | void honeypot(...) __attribute__((overloadable, unavailable)); <i>// calling me is an error</i> |
Douglas Gregor | 965acbb | 2009-02-18 07:07:28 +0000 | [diff] [blame] | 299 | </pre> |
| 300 | </blockquote> |
| 301 | |
Douglas Gregor | cb54d43 | 2009-02-13 00:57:04 +0000 | [diff] [blame] | 302 | <p>Functions declared with the <tt>overloadable</tt> attribute have |
| 303 | their names mangled according to the same rules as C++ function |
| 304 | names. For example, the three <tt>tgsin</tt> functions in our |
| 305 | motivating example get the mangled names <tt>_Z5tgsinf</tt>, |
| 306 | <tt>_Z5tgsind</tt>, and <tt>Z5tgsine</tt>, respectively. There are two |
| 307 | caveats to this use of name mangling:</p> |
| 308 | |
| 309 | <ul> |
| 310 | |
| 311 | <li>Future versions of Clang may change the name mangling of |
| 312 | functions overloaded in C, so you should not depend on an specific |
| 313 | mangling. To be completely safe, we strongly urge the use of |
| 314 | <tt>static inline</tt> with <tt>overloadable</tt> functions.</li> |
| 315 | |
| 316 | <li>The <tt>overloadable</tt> attribute has almost no meaning when |
| 317 | used in C++, because names will already be mangled and functions are |
| 318 | already overloadable. However, when an <tt>overloadable</tt> |
| 319 | function occurs within an <tt>extern "C"</tt> linkage specification, |
| 320 | it's name <i>will</i> be mangled in the same way as it would in |
| 321 | C.</li> |
| 322 | </ul> |
| 323 | |
Chris Lattner | 148772a | 2009-06-13 07:13:28 +0000 | [diff] [blame] | 324 | <p>Query for this feature with __has_feature(attribute_overloadable).</p> |
| 325 | |
| 326 | |
Douglas Gregor | cb54d43 | 2009-02-13 00:57:04 +0000 | [diff] [blame] | 327 | <!-- ======================================================================= --> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 328 | <h2 id="builtins">Builtin Functions</h2> |
| 329 | <!-- ======================================================================= --> |
| 330 | |
| 331 | <p>Clang supports a number of builtin library functions with the same syntax as |
| 332 | GCC, including things like <tt>__builtin_nan</tt>, |
| 333 | <tt>__builtin_constant_p</tt>, <tt>__builtin_choose_expr</tt>, |
| 334 | <tt>__builtin_types_compatible_p</tt>, <tt>__sync_fetch_and_add</tt>, etc. In |
| 335 | addition to the GCC builtins, Clang supports a number of builtins that GCC does |
| 336 | not, which are listed here.</p> |
| 337 | |
| 338 | <p>Please note that Clang does not and will not support all of the GCC builtins |
| 339 | for vector operations. Instead of using builtins, you should use the functions |
| 340 | defined in target-specific header files like <tt><xmmintrin.h></tt>, which |
| 341 | define portable wrappers for these. Many of the Clang versions of these |
| 342 | functions are implemented directly in terms of <a href="#vectors">extended |
| 343 | vector support</a> instead of builtins, in order to reduce the number of |
| 344 | builtins that we need to implement.</p> |
| 345 | |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 346 | <!-- ======================================================================= --> |
Chris Lattner | 6f72da5 | 2009-02-13 20:00:20 +0000 | [diff] [blame] | 347 | <h3 id="__builtin_shufflevector">__builtin_shufflevector</h3> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 348 | <!-- ======================================================================= --> |
| 349 | |
Chris Lattner | aad826b | 2009-09-16 18:56:12 +0000 | [diff] [blame] | 350 | <p><tt>__builtin_shufflevector</tt> is used to express generic vector |
Chris Lattner | 6f72da5 | 2009-02-13 20:00:20 +0000 | [diff] [blame] | 351 | permutation/shuffle/swizzle operations. This builtin is also very important for |
| 352 | the implementation of various target-specific header files like |
| 353 | <tt><xmmintrin.h></tt>. |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 354 | </p> |
| 355 | |
| 356 | <p><b>Syntax:</b></p> |
| 357 | |
| 358 | <pre> |
Chris Lattner | 6f72da5 | 2009-02-13 20:00:20 +0000 | [diff] [blame] | 359 | __builtin_shufflevector(vec1, vec2, index1, index2, ...) |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 360 | </pre> |
| 361 | |
| 362 | <p><b>Examples:</b></p> |
| 363 | |
| 364 | <pre> |
Chris Lattner | 6f72da5 | 2009-02-13 20:00:20 +0000 | [diff] [blame] | 365 | // Identity operation - return 4-element vector V1. |
| 366 | __builtin_shufflevector(V1, V1, 0, 1, 2, 3) |
| 367 | |
| 368 | // "Splat" element 0 of V1 into a 4-element result. |
| 369 | __builtin_shufflevector(V1, V1, 0, 0, 0, 0) |
| 370 | |
| 371 | // Reverse 4-element vector V1. |
| 372 | __builtin_shufflevector(V1, V1, 3, 2, 1, 0) |
| 373 | |
| 374 | // Concatenate every other element of 4-element vectors V1 and V2. |
| 375 | __builtin_shufflevector(V1, V2, 0, 2, 4, 6) |
| 376 | |
| 377 | // Concatenate every other element of 8-element vectors V1 and V2. |
| 378 | __builtin_shufflevector(V1, V2, 0, 2, 4, 6, 8, 10, 12, 14) |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 379 | </pre> |
| 380 | |
| 381 | <p><b>Description:</b></p> |
| 382 | |
Chris Lattner | 6f72da5 | 2009-02-13 20:00:20 +0000 | [diff] [blame] | 383 | <p>The first two arguments to __builtin_shufflevector are vectors that have the |
| 384 | same element type. The remaining arguments are a list of integers that specify |
| 385 | the elements indices of the first two vectors that should be extracted and |
| 386 | returned in a new vector. These element indices are numbered sequentially |
| 387 | starting with the first vector, continuing into the second vector. Thus, if |
| 388 | vec1 is a 4-element vector, index 5 would refer to the second element of vec2. |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 389 | </p> |
| 390 | |
Chris Lattner | 6f72da5 | 2009-02-13 20:00:20 +0000 | [diff] [blame] | 391 | <p>The result of __builtin_shufflevector is a vector |
| 392 | with the same element type as vec1/vec2 but that has an element count equal to |
| 393 | the number of indices specified. |
| 394 | </p> |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 395 | |
Chris Lattner | 21190d5 | 2009-09-21 03:09:59 +0000 | [diff] [blame] | 396 | <p>Query for this feature with __has_builtin(__builtin_shufflevector).</p> |
| 397 | |
| 398 | <!-- ======================================================================= --> |
| 399 | <h3 id="__builtin_unreachable">__builtin_unreachable</h3> |
| 400 | <!-- ======================================================================= --> |
| 401 | |
| 402 | <p><tt>__builtin_unreachable</tt> is used to indicate that a specific point in |
| 403 | the program cannot be reached, even if the compiler might otherwise think it |
| 404 | can. This is useful to improve optimization and eliminates certain warnings. |
| 405 | For example, without the <tt>__builtin_unreachable</tt> in the example below, |
| 406 | the compiler assumes that the inline asm can fall through and prints a "function |
| 407 | declared 'noreturn' should not return" warning. |
| 408 | </p> |
| 409 | |
| 410 | <p><b>Syntax:</b></p> |
| 411 | |
| 412 | <pre> |
| 413 | __builtin_unreachable() |
| 414 | </pre> |
| 415 | |
| 416 | <p><b>Example of Use:</b></p> |
| 417 | |
| 418 | <pre> |
| 419 | void myabort(void) __attribute__((noreturn)); |
| 420 | void myabort(void) { |
| 421 | asm("int3"); |
| 422 | __builtin_unreachable(); |
| 423 | } |
| 424 | </pre> |
| 425 | |
| 426 | <p><b>Description:</b></p> |
| 427 | |
| 428 | <p>The __builtin_unreachable() builtin has completely undefined behavior. Since |
| 429 | it has undefined behavior, it is a statement that it is never reached and the |
| 430 | optimizer can take advantage of this to produce better code. This builtin takes |
| 431 | no arguments and produces a void result. |
| 432 | </p> |
| 433 | |
| 434 | <p>Query for this feature with __has_builtin(__builtin_unreachable).</p> |
| 435 | |
| 436 | |
Chris Lattner | 1177f91 | 2009-04-09 19:58:15 +0000 | [diff] [blame] | 437 | <!-- ======================================================================= --> |
| 438 | <h2 id="targetspecific">Target-Specific Extensions</h2> |
| 439 | <!-- ======================================================================= --> |
| 440 | |
| 441 | <p>Clang supports some language features conditionally on some targets.</p> |
| 442 | |
| 443 | <!-- ======================================================================= --> |
| 444 | <h3 id="x86-specific">X86/X86-64 Language Extensions</h3> |
| 445 | <!-- ======================================================================= --> |
| 446 | |
| 447 | <p>The X86 backend has these language extensions:</p> |
| 448 | |
| 449 | <!-- ======================================================================= --> |
| 450 | <h4 id="x86-gs-segment">Memory references off the GS segment</h4> |
| 451 | <!-- ======================================================================= --> |
| 452 | |
| 453 | <p>Annotating a pointer with address space #256 causes it to be code generated |
Chris Lattner | a021e7c | 2009-05-05 18:54:47 +0000 | [diff] [blame] | 454 | relative to the X86 GS segment register, and address space #257 causes it to be |
| 455 | relative to the X86 FS segment. Note that this is a very very low-level |
| 456 | feature that should only be used if you know what you're doing (for example in |
| 457 | an OS kernel).</p> |
Chris Lattner | 1177f91 | 2009-04-09 19:58:15 +0000 | [diff] [blame] | 458 | |
| 459 | <p>Here is an example:</p> |
| 460 | |
| 461 | <pre> |
| 462 | #define GS_RELATIVE __attribute__((address_space(256))) |
| 463 | int foo(int GS_RELATIVE *P) { |
| 464 | return *P; |
| 465 | } |
| 466 | </pre> |
| 467 | |
| 468 | <p>Which compiles to (on X86-32):</p> |
| 469 | |
| 470 | <pre> |
| 471 | _foo: |
| 472 | movl 4(%esp), %eax |
| 473 | movl %gs:(%eax), %eax |
| 474 | ret |
| 475 | </pre> |
| 476 | |
Ted Kremenek | ed86931 | 2009-04-10 05:03:33 +0000 | [diff] [blame] | 477 | <!-- ======================================================================= --> |
| 478 | <h2 id="analyzerspecific">Static Analysis-Specific Extensions</h2> |
| 479 | <!-- ======================================================================= --> |
| 480 | |
| 481 | <p>Clang supports additional attributes that are useful for documenting program |
| 482 | invariants and rules for static analysis tools. The extensions documented here |
| 483 | are used by the <a |
| 484 | href="http://clang.llvm.org/StaticAnalysis.html">path-sensitive static analyzer |
| 485 | engine</a> that is part of Clang's Analysis library.</p> |
| 486 | |
| 487 | <!-- ======================================================================= --> |
| 488 | <h3 id="analyzerattributes">Analyzer Attributes</h3> |
| 489 | <!-- ======================================================================= --> |
| 490 | |
| 491 | <h4 id="attr_analyzer_noreturn"><tt>analyzer_noreturn</tt></h4> |
| 492 | |
| 493 | <p>Clang's static analysis engine understands the standard <tt>noreturn</tt> |
Ted Kremenek | 4df2114 | 2009-04-10 05:04:22 +0000 | [diff] [blame] | 494 | attribute. This attribute, which is typically affixed to a function prototype, |
| 495 | indicates that a call to a given function never returns. Function prototypes for |
| 496 | common functions like <tt>exit</tt> are typically annotated with this attribute, |
| 497 | as well as a variety of common assertion handlers. Users can educate the static |
| 498 | analyzer about their own custom assertion handles (thus cutting down on false |
| 499 | positives due to false paths) by marking their own "panic" functions |
| 500 | with this attribute.</p> |
Ted Kremenek | ed86931 | 2009-04-10 05:03:33 +0000 | [diff] [blame] | 501 | |
| 502 | <p>While useful, <tt>noreturn</tt> is not applicable in all cases. Sometimes |
Nick Lewycky | 625b586 | 2009-06-14 04:08:08 +0000 | [diff] [blame] | 503 | there are special functions that for all intents and purposes should be |
| 504 | considered panic functions (i.e., they are only called when an internal program |
| 505 | error occurs) but may actually return so that the program can fail gracefully. |
| 506 | The <tt>analyzer_noreturn</tt> attribute allows one to annotate such functions |
| 507 | as being interpreted as "no return" functions by the analyzer (thus |
Chris Lattner | 2893589 | 2009-04-10 05:54:56 +0000 | [diff] [blame] | 508 | pruning bogus paths) but will not affect compilation (as in the case of |
Ted Kremenek | ed86931 | 2009-04-10 05:03:33 +0000 | [diff] [blame] | 509 | <tt>noreturn</tt>).</p> |
| 510 | |
| 511 | <p><b>Usage</b>: The <tt>analyzer_noreturn</tt> attribute can be placed in the |
Chris Lattner | 2893589 | 2009-04-10 05:54:56 +0000 | [diff] [blame] | 512 | same places where the <tt>noreturn</tt> attribute can be placed. It is commonly |
Ted Kremenek | ed86931 | 2009-04-10 05:03:33 +0000 | [diff] [blame] | 513 | placed at the end of function prototypes:</p> |
| 514 | |
| 515 | <pre> |
| 516 | void foo() <b>__attribute__((analyzer_noreturn))</b>; |
Chris Lattner | 148772a | 2009-06-13 07:13:28 +0000 | [diff] [blame] | 517 | </pre> |
| 518 | |
| 519 | <p>Query for this feature with __has_feature(attribute_analyzer_noreturn).</p> |
| 520 | |
Ted Kremenek | ed86931 | 2009-04-10 05:03:33 +0000 | [diff] [blame] | 521 | |
Chris Lattner | 5ce933f | 2009-02-09 08:46:11 +0000 | [diff] [blame] | 522 | </div> |
| 523 | </body> |
| 524 | </html> |