docs/SourceLevelDebugging.html

   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
   2                       "http://www.w3.org/TR/html4/strict.dtd">
   3 <html>
   4 <head>
   5   <title>Source Level Debugging with LLVM</title>
   6   <link rel="stylesheet" href="llvm.css" type="text/css">
   7 </head>
   8 <body>
   9
  10 <div class="doc_title">Source Level Debugging with LLVM</div>
  11
  12 <table class="layout" style="width:100%">
  13   <tr class="layout">
  14     <td class="left">
  15 <ul>
  16   <li><a href="#introduction">Introduction</a>
  17   <ol>
  18     <li><a href="#phil">Philosophy behind LLVM debugging information</a></li>
  19     <li><a href="#consumers">Debug information consumers</a></li>
  20     <li><a href="#debugopt">Debugging optimized code</a></li>
  21   </ol></li>
  22   <li><a href="#format">Debugging information format</a>
  23   <ol>
  24     <li><a href="#debug_info_descriptors">Debug information descriptors</a>
  25     <ul>
  26       <li><a href="#format_anchors">Anchor descriptors</a></li>
  27       <li><a href="#format_compile_units">Compile unit descriptors</a></li>
  28       <li><a href="#format_global_variables">Global variable descriptors</a></li>
  29       <li><a href="#format_subprograms">Subprogram descriptors</a></li>
  30       <li><a href="#format_blocks">Block descriptors</a></li>
  31       <li><a href="#format_basic_type">Basic type descriptors</a></li>
  32       <li><a href="#format_derived_type">Derived type descriptors</a></li>
  33       <li><a href="#format_composite_type">Composite type descriptors</a></li>
  34       <li><a href="#format_subrange">Subrange descriptors</a></li>
  35       <li><a href="#format_enumeration">Enumerator descriptors</a></li>
  36       <li><a href="#format_variables">Local variables</a></li>
  37     </ul></li>
  38     <li><a href="#format_common_intrinsics">Debugger intrinsic functions</a>
  39       <ul>
  40       <li><a href="#format_common_stoppoint">llvm.dbg.stoppoint</a></li>
  41       <li><a href="#format_common_func_start">llvm.dbg.func.start</a></li>
  42       <li><a href="#format_common_region_start">llvm.dbg.region.start</a></li>
  43       <li><a href="#format_common_region_end">llvm.dbg.region.end</a></li>
  44       <li><a href="#format_common_declare">llvm.dbg.declare</a></li>
  45     </ul></li>
  46     <li><a href="#format_common_stoppoints">Representing stopping points in the
  47                                            source program</a></li>
  48   </ol></li>
  49   <li><a href="#ccxx_frontend">C/C++ front-end specific debug information</a>
  50   <ol>
  51     <li><a href="#ccxx_compile_units">C/C++ source file information</a></li>
  52     <li><a href="#ccxx_global_variable">C/C++ global variable information</a></li>
  53     <li><a href="#ccxx_subprogram">C/C++ function information</a></li>
  54     <li><a href="#ccxx_basic_types">C/C++ basic types</a></li>
  55     <li><a href="#ccxx_derived_types">C/C++ derived types</a></li>
  56     <li><a href="#ccxx_composite_types">C/C++ struct/union types</a></li>
  57     <li><a href="#ccxx_enumeration_types">C/C++ enumeration types</a></li>
  58   </ol></li>
  59 </ul>
  60 </td>
  61 <td class="right">
  62 <img src="img/venusflytrap.jpg" alt="A leafy and green bug eater" width="247"
  63 height="369">
  64 </td>
  65 </tr></table>
  66
  67 <div class="doc_author">
  68   <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
  69             and <a href="mailto:jlaskey@mac.com">Jim Laskey</a></p>
  70 </div>
  71
  72
  73 <!-- *********************************************************************** -->
  74 <div class="doc_section"><a name="introduction">Introduction</a></div>
  75 <!-- *********************************************************************** -->
  76
  77 <div class="doc_text">
  78
  79 <p>This document is the central repository for all information pertaining to
  80 debug information in LLVM.  It describes the <a href="#format">actual format
  81 that the LLVM debug information</a> takes, which is useful for those interested
  82 in creating front-ends or dealing directly with the information.  Further, this
  83 document provides specifc examples of what debug information for C/C++.</p>
  84
  85 </div>
  86
  87 <!-- ======================================================================= -->
  88 <div class="doc_subsection">
  89   <a name="phil">Philosophy behind LLVM debugging information</a>
  90 </div>
  91
  92 <div class="doc_text">
  93
  94 <p>The idea of the LLVM debugging information is to capture how the important
  95 pieces of the source-language's Abstract Syntax Tree map onto LLVM code.
  96 Several design aspects have shaped the solution that appears here.  The
  97 important ones are:</p>
  98
  99 <ul>
 100 <li>Debugging information should have very little impact on the rest of the
 101 compiler.  No transformations, analyses, or code generators should need to be
 102 modified because of debugging information.</li>
 103
 104 <li>LLVM optimizations should interact in <a href="#debugopt">well-defined and
 105 easily described ways</a> with the debugging information.</li>
 106
 107 <li>Because LLVM is designed to support arbitrary programming languages,
 108 LLVM-to-LLVM tools should not need to know anything about the semantics of the
 109 source-level-language.</li>
 110
 111 <li>Source-level languages are often <b>widely</b> different from one another.
 112 LLVM should not put any restrictions of the flavor of the source-language, and
 113 the debugging information should work with any language.</li>
 114
 115 <li>With code generator support, it should be possible to use an LLVM compiler
 116 to compile a program to native machine code and standard debugging formats.
 117 This allows compatibility with traditional machine-code level debuggers, like
 118 GDB or DBX.</li>
 119
 120 </ul>
 121
 122 <p>The approach used by the LLVM implementation is to use a small set of <a
 123 href="#format_common_intrinsics">intrinsic functions</a> to define a mapping
 124 between LLVM program objects and the source-level objects.  The description of
 125 the source-level program is maintained in LLVM global variables in an <a
 126 href="#ccxx_frontend">implementation-defined format</a> (the C/C++ front-end
 127 currently uses working draft 7 of the <a
 128 href="http://www.eagercon.com/dwarf/dwarf3std.htm">Dwarf 3 standard</a>).</p>
 129
 130 <p>When a program is being debugged, a debugger interacts with the user and
 131 turns the stored debug information into source-language specific information.
 132 As such, a debugger must be aware of the source-language, and is thus tied to
 133 a specific language of family of languages.</p>
 134
 135 </div>
 136
 137 <!-- ======================================================================= -->
 138 <div class="doc_subsection">
 139   <a name="consumers">Debug information consumers</a>
 140 </div>
 141
 142 <div class="doc_text">
 143 <p>The role of debug information is to provide meta information normally
 144 stripped away during the compilation process.  This meta information provides an
 145 llvm user a relationship between generated code and the original program source
 146 code.</p>
 147
 148 <p>Currently, debug information is consumed by the DwarfWriter to produce dwarf
 149 information used by the gdb debugger.  Other targets could use the same
 150 information to produce stabs or other debug forms.</p>
 151
 152 <p>It would also be reasonable to use debug information to feed profiling tools
 153 for analysis of generated code, or, tools for reconstructing the original source
 154 from generated code.</p>
 155
 156 <p>TODO - expound a bit more.</p>
 157
 158 </div>
 159
 160 <!-- ======================================================================= -->
 161 <div class="doc_subsection">
 162   <a name="debugopt">Debugging optimized code</a>
 163 </div>
 164
 165 <div class="doc_text">
 166
 167 <p>An extremely high priority of LLVM debugging information is to make it
 168 interact well with optimizations and analysis.  In particular, the LLVM debug
 169 information provides the following guarantees:</p>
 170
 171 <ul>
 172
 173 <li>LLVM debug information <b>always provides information to accurately read the
 174 source-level state of the program</b>, regardless of which LLVM optimizations
 175 have been run, and without any modification to the optimizations themselves.
 176 However, some optimizations may impact the ability to modify the current state
 177 of the program with a debugger, such as setting program variables, or calling
 178 function that have been deleted.</li>
 179
 180 <li>LLVM optimizations gracefully interact with debugging information.  If they
 181 are not aware of debug information, they are automatically disabled as necessary
 182 in the cases that would invalidate the debug info.  This retains the LLVM
 183 features making it easy to write new transformations.</li>
 184
 185 <li>As desired, LLVM optimizations can be upgraded to be aware of the LLVM
 186 debugging information, allowing them to update the debugging information as they
 187 perform aggressive optimizations.  This means that, with effort, the LLVM
 188 optimizers could optimize debug code just as well as non-debug code.</li>
 189
 190 <li>LLVM debug information does not prevent many important optimizations from
 191 happening (for example inlining, basic block reordering/merging/cleanup, tail
 192 duplication, etc), further reducing the amount of the compiler that eventually
 193 is "aware" of debugging information.</li>
 194
 195 <li>LLVM debug information is automatically optimized along with the rest of the
 196 program, using existing facilities.  For example, duplicate information is
 197 automatically merged by the linker, and unused information is automatically
 198 removed.</li>
 199
 200 </ul>
 201
 202 <p>Basically, the debug information allows you to compile a program with
 203 "<tt>-O0 -g</tt>" and get full debug information, allowing you to arbitrarily
 204 modify the program as it executes from a debugger.  Compiling a program with
 205 "<tt>-O3 -g</tt>" gives you full debug information that is always available and
 206 accurate for reading (e.g., you get accurate stack traces despite tail call
 207 elimination and inlining), but you might lose the ability to modify the program
 208 and call functions where were optimized out of the program, or inlined away
 209 completely.</p>
 210
 211 </div>
 212
 213 <!-- *********************************************************************** -->
 214 <div class="doc_section">
 215   <a name="format">Debugging information format</a>
 216 </div>
 217 <!-- *********************************************************************** -->
 218
 219 <div class="doc_text">
 220
 221 <p>LLVM debugging information has been carefully designed to make it possible
 222 for the optimizer to optimize the program and debugging information without
 223 necessarily having to know anything about debugging information.  In particular,
 224 the global constant merging pass automatically eliminates duplicated debugging
 225 information (often caused by header files), the global dead code elimination
 226 pass automatically deletes debugging information for a function if it decides to
 227 delete the function, and the linker eliminates debug information when it merges
 228 <tt>linkonce</tt> functions.</p>
 229
 230 <p>To do this, most of the debugging information (descriptors for types,
 231 variables, functions, source files, etc) is inserted by the language front-end
 232 in the form of LLVM global variables.  These LLVM global variables are no
 233 different from any other global variables, except that they have a web of LLVM
 234 intrinsic functions that point to them.  If the last references to a particular
 235 piece of debugging information are deleted (for example, by the
 236 <tt>-globaldce</tt> pass), the extraneous debug information will automatically
 237 become dead and be removed by the optimizer.</p>
 238
 239 <p>Debug information is designed to be agnostic about the target debugger and
 240 debugging information representation (e.g. DWARF/Stabs/etc).  It uses a generic
 241 machine debug information pass to decode the information that represents
 242 variables, types, functions, namespaces, etc: this allows for arbitrary
 243 source-language semantics and type-systems to be used, as long as there is a
 244 module written for the target debugger to interpret the information. In
 245 addition, debug global variables are declared in the <tt>"llvm.metadata"</tt>
 246 section.  All values declared in this section are stripped away after target
 247 debug information is constructed and before the program object is emitted.</p>
 248
 249 <p>To provide basic functionality, the LLVM debugger does have to make some
 250 assumptions about the source-level language being debugged, though it keeps
 251 these to a minimum.  The only common features that the LLVM debugger assumes
 252 exist are <a href="#format_compile_units">source files</a>, and <a
 253 href="#format_global_variables">program objects</a>.  These abstract objects are
 254 used by a debugger to form stack traces, show information about local
 255 variables, etc.</p>
 256
 257 <p>This section of the documentation first describes the representation aspects
 258 common to any source-language.  The <a href="#ccxx_frontend">next section</a>
 259 describes the data layout conventions used by the C and C++ front-ends.</p>
 260
 261 </div>
 262
 263 <!-- ======================================================================= -->
 264 <div class="doc_subsection">
 265   <a name="debug_info_descriptors">Debug information descriptors</a>
 266 </div>
 267
 268 <div class="doc_text">
 269 <p>In consideration of the complexity and volume of debug information, LLVM
 270 provides a specification for well formed debug global variables.  The constant
 271 value of each of these globals is one of a limited set of structures, known as
 272 debug descriptors.</p>
 273
 274 <p>Consumers of LLVM debug information expect the descriptors for program
 275 objects to start in a canonical format, but the descriptors can include
 276 additional information appended at the end that is source-language specific. All
 277 LLVM debugging information is versioned, allowing backwards compatibility in the
 278 case that the core structures need to change in some way.  Also, all debugging
 279 information objects start with a tag to indicate what type of object it is.  The
 280 source-language is allowed to define its own objects, by using unreserved tag
 281 numbers.  We recommend using with tags in the range 0x1000 thru 0x2000 (there is
 282 a defined enum DW_TAG_user_base = 0x1000.)</p>
 283
 284 <p>The fields of debug descriptors used internally by LLVM (MachineModuleInfo)
 285 are restricted to only the simple data types <tt>int</tt>, <tt>uint</tt>,
 286 <tt>bool</tt>, <tt>float</tt>, <tt>double</tt>, <tt>sbyte*</tt> and <tt> { }*
 287 </tt>.  References to arbitrary values are handled using a <tt> { }* </tt> and a
 288 cast to <tt> { }* </tt> expression; typically references to other field
 289 descriptors, arrays of descriptors or global variables.</p>
 290
 291 <pre>
 292   %llvm.dbg.object.type = type {
 293     uint,   ;; A tag
 294     ...
 295   }
 296 </pre>
 297
 298 <p><a name="LLVMDebugVersion">The first field of a descriptor is always an
 299 <tt>uint</tt> containing a tag value identifying the content of the descriptor.
 300 The remaining fields are specific to the descriptor.  The values of tags are
 301 loosely bound to the tag values of Dwarf information entries.  However, that
 302 does not restrict the use of the information supplied to Dwarf targets.  To
 303 facilitate versioning of debug information, the tag is augmented with the
 304 current debug version (LLVMDebugVersion = 4 << 16 or 0x40000 or 262144.)</a></p>
 305
 306 <p>The details of the various descriptors follow.</p>
 307
 308 </div>
 309
 310 <!-- ======================================================================= -->
 311 <div class="doc_subsubsection">
 312   <a name="format_anchors">Anchor descriptors</a>
 313 </div>
 314
 315 <div class="doc_text">
 316
 317 <pre>
 318   %<a href="#format_anchors">llvm.dbg.anchor.type</a> = type {
 319     uint,   ;; Tag = 0 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
 320     uint    ;; Tag of descriptors grouped by the anchor
 321   }
 322 </pre>
 323
 324 <p>One important aspect of the LLVM debug representation is that it allows the
 325 LLVM debugger to efficiently index all of the global objects without having the
 326 scan the program.  To do this, all of the global objects use "anchor"
 327 descriptors with designated names.  All of the global objects of a particular
 328 type (e.g., compile units) contain a pointer to the anchor.  This pointer allows
 329 a debugger to use def-use chains to find all global objects of that type.</p>
 330
 331 <p>The following names are recognized as anchors by LLVM:</p>
 332
 333 <pre>
 334   %<a href="#format_compile_units">llvm.dbg.compile_units</a>       = linkonce constant %<a href="#format_anchors">llvm.dbg.anchor.type</a>  { uint 0, uint 17 } ;; DW_TAG_compile_unit
 335   %<a href="#format_global_variables">llvm.dbg.global_variables</a>    = linkonce constant %<a href="#format_anchors">llvm.dbg.anchor.type</a>  { uint 0, uint 52 } ;; DW_TAG_variable
 336   %<a href="#format_subprograms">llvm.dbg.subprograms</a>         = linkonce constant %<a href="#format_anchors">llvm.dbg.anchor.type</a>  { uint 0, uint 46 } ;; DW_TAG_subprogram
 337 </pre>
 338
 339 <p>Using anchors in this way (where the compile unit descriptor points to the
 340 anchors, as opposed to having a list of compile unit descriptors) allows for the
 341 standard dead global elimination and merging passes to automatically remove
 342 unused debugging information.  If the globals were kept track of through lists,
 343 there would always be an object pointing to the descriptors, thus would never be
 344 deleted.</p>
 345
 346 </div>
 347
 348 <!-- ======================================================================= -->
 349 <div class="doc_subsubsection">
 350   <a name="format_compile_units">Compile unit descriptors</a>
 351 </div>
 352
 353 <div class="doc_text">
 354
 355 <pre>
 356   %<a href="#format_compile_units">llvm.dbg.compile_unit.type</a> = type {
 357     uint,   ;; Tag = 17 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_compile_unit)
 358     {  }*,  ;; Compile unit anchor = cast = (%<a href="#format_anchors">llvm.dbg.anchor.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_units</a> to {  }*)
 359     uint,   ;; Dwarf language identifier (ex. DW_LANG_C89)
 360     sbyte*, ;; Source file name
 361     sbyte*, ;; Source file directory (includes trailing slash)
 362     sbyte*  ;; Producer (ex. "4.0.1 LLVM (LLVM research group)")
 363   }
 364 </pre>
 365
 366 <p>These descriptors contain a source language ID for the file (we use the Dwarf
 367 3.0 ID numbers, such as <tt>DW_LANG_C89</tt>, <tt>DW_LANG_C_plus_plus</tt>,
 368 <tt>DW_LANG_Cobol74</tt>, etc), three strings describing the filename, working
 369 directory of the compiler, and an identifier string for the compiler that
 370 produced it.</p>
 371
 372 <p> Compile unit descriptors provide the root context for objects declared in a
 373 specific source file.  Global variables and top level functions would be defined
 374 using this context.  Compile unit descriptors also provide context for source
 375 line correspondence.</p>
 376
 377 </div>
 378
 379 <!-- ======================================================================= -->
 380 <div class="doc_subsubsection">
 381   <a name="format_global_variables">Global variable descriptors</a>
 382 </div>
 383
 384 <div class="doc_text">
 385
 386 <pre>
 387   %<a href="#format_global_variables">llvm.dbg.global_variable.type</a> = type {
 388     uint,   ;; Tag = 52 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_variable)
 389     {  }*,  ;; Global variable anchor = cast (%<a href="#format_anchors">llvm.dbg.anchor.type</a>* %<a href="#format_global_variables">llvm.dbg.global_variables</a> to {  }*),
 390     {  }*,  ;; Reference to context descriptor
 391     sbyte*, ;; Name
 392     sbyte*, ;; Display name (fully qualified C++ name)
 393     sbyte*, ;; MIPS linkage name (for C++)
 394     {  }*,  ;; Reference to compile unit where defined
 395     uint,   ;; Line number where defined
 396     {  }*,  ;; Reference to type descriptor
 397     bool,   ;; True if the global is local to compile unit (static)
 398     bool,   ;; True if the global is defined in the compile unit (not extern)
 399     {  }*   ;; Reference to the global variable
 400   }
 401 </pre>
 402
 403 <p>These descriptors provide debug information about globals variables.  The
 404 provide details such as name, type and where the variable is defined.</p>
 405
 406 </div>
 407
 408 <!-- ======================================================================= -->
 409 <div class="doc_subsubsection">
 410   <a name="format_subprograms">Subprogram descriptors</a>
 411 </div>
 412
 413 <div class="doc_text">
 414
 415 <pre>
 416   %<a href="#format_subprograms">llvm.dbg.subprogram.type</a> = type {
 417     uint,   ;; Tag = 46 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_subprogram)
 418     {  }*,  ;; Subprogram anchor = cast (%<a href="#format_anchors">llvm.dbg.anchor.type</a>* %<a href="#format_subprograms">llvm.dbg.subprograms</a> to {  }*),
 419     {  }*,  ;; Reference to context descriptor
 420     sbyte*, ;; Name
 421     sbyte*, ;; Display name (fully qualified C++ name)
 422     sbyte*, ;; MIPS linkage name (for C++)
 423     {  }*,  ;; Reference to compile unit where defined
 424     uint,   ;; Line number where defined
 425     {  }*,  ;; Reference to type descriptor
 426     bool,   ;; True if the global is local to compile unit (static)
 427     bool    ;; True if the global is defined in the compile unit (not extern)
 428   }
 429 </pre>
 430
 431 <p>These descriptors provide debug information about functions, methods and
 432 subprograms.  They provide details such as name, return types and the source
 433 location where the subprogram is defined.</p>
 434
 435 </div>
 436 <!-- ======================================================================= -->
 437 <div class="doc_subsubsection">
 438   <a name="format_blocks">Block descriptors</a>
 439 </div>
 440
 441 <div class="doc_text">
 442
 443 <pre>
 444   %<a href="#format_blocks">llvm.dbg.block</a> = type {
 445     uint,   ;; Tag = 13 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_lexical_block)
 446     {  }*   ;; Reference to context descriptor
 447   }
 448 </pre>
 449
 450 <p>These descriptors provide debug information about nested blocks within a
 451 subprogram.  The array of member descriptors is used to define local variables
 452 and deeper nested blocks.</p>
 453
 454 </div>
 455
 456 <!-- ======================================================================= -->
 457 <div class="doc_subsubsection">
 458   <a name="format_basic_type">Basic type descriptors</a>
 459 </div>
 460
 461 <div class="doc_text">
 462
 463 <pre>
 464   %<a href="#format_basic_type">llvm.dbg.basictype.type</a> = type {
 465     uint,   ;; Tag = 36 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_base_type)
 466     {  }*,  ;; Reference to context (typically a compile unit)
 467     sbyte*, ;; Name (may be "" for anonymous types)
 468     {  }*,  ;; Reference to compile unit where defined (may be NULL)
 469     uint,   ;; Line number where defined (may be 0)
 470     uint,   ;; Size in bits
 471     uint,   ;; Alignment in bits
 472     uint,   ;; Offset in bits
 473     uint    ;; Dwarf type encoding
 474   }
 475 </pre>
 476
 477 <p>These descriptors define primitive types used in the code. Example int, bool
 478 and float.  The context provides the scope of the type, which is usually the top
 479 level.  Since basic types are not usually user defined the compile unit and line
 480 number can be left as NULL and 0.  The size, alignment and offset are expressed
 481 in bits and can be 64 bit values.  The alignment is used to round the offset
 482 when embedded in a <a href="#format_composite_type">composite type</a>
 483 (example to keep float doubles on 64 bit boundaries.) The offset is the bit
 484 offset if embedded in a <a href="#format_composite_type">composite
 485 type</a>.</p>
 486
 487 <p>The type encoding provides the details of the type.  The values are typically
 488 one of the following;</p>
 489
 490 <pre>
 491   DW_ATE_address = 1
 492   DW_ATE_boolean = 2
 493   DW_ATE_float = 4
 494   DW_ATE_signed = 5
 495   DW_ATE_signed_char = 6
 496   DW_ATE_unsigned = 7
 497   DW_ATE_unsigned_char = 8
 498 </pre>
 499
 500 </div>
 501
 502 <!-- ======================================================================= -->
 503 <div class="doc_subsubsection">
 504   <a name="format_derived_type">Derived type descriptors</a>
 505 </div>
 506
 507 <div class="doc_text">
 508
 509 <pre>
 510   %<a href="#format_derived_type">llvm.dbg.derivedtype.type</a> = type {
 511     uint,   ;; Tag (see below)
 512     {  }*,  ;; Reference to context
 513     sbyte*, ;; Name (may be "" for anonymous types)
 514     {  }*,  ;; Reference to compile unit where defined (may be NULL)
 515     uint,   ;; Line number where defined (may be 0)
 516     uint,   ;; Size in bits
 517     uint,   ;; Alignment in bits
 518     uint,   ;; Offset in bits
 519     {  }*   ;; Reference to type derived from
 520   }
 521 </pre>
 522
 523 <p>These descriptors are used to define types derived from other types.  The
 524 value of the tag varies depending on the meaning.  The following are possible
 525 tag values;</p>
 526
 527 <pre>
 528   DW_TAG_formal_parameter = 5
 529   DW_TAG_member = 13
 530   DW_TAG_pointer_type = 15
 531   DW_TAG_reference_type = 16
 532   DW_TAG_typedef = 22
 533   DW_TAG_const_type = 38
 534   DW_TAG_volatile_type = 53
 535   DW_TAG_restrict_type = 55
 536 </pre>
 537
 538 <p> <tt>DW_TAG_member</tt> is used to define a member of a <a
 539 href="#format_composite_type">composite type</a> or <a
 540 href="#format_subprograms">subprogram</a>.  The type of the member is the <a
 541 href="#format_derived_type">derived type</a>. <tt>DW_TAG_formal_parameter</tt>
 542 is used to define a member which is a formal argument of a subprogram.</p>
 543
 544 <p><tt>DW_TAG_typedef</tt> is used to
 545 provide a name for the derived type.</p>
 546
 547 <p><tt>DW_TAG_pointer_type</tt>,
 548 <tt>DW_TAG_reference_type</tt>, <tt>DW_TAG_const_type</tt>,
 549 <tt>DW_TAG_volatile_type</tt> and <tt>DW_TAG_restrict_type</tt> are used to
 550 qualify the <a href="#format_derived_type">derived type</a>. </p>
 551
 552 <p><a href="#format_derived_type">Derived type</a> location can be determined
 553 from the compile unit and line number.  The size, alignment and offset are
 554 expressed in bits and can be 64 bit values.  The alignment is used to round the
 555 offset when embedded in a <a href="#format_composite_type">composite type</a>
 556 (example to keep float doubles on 64 bit boundaries.) The offset is the bit
 557 offset if embedded in a <a href="#format_composite_type">composite
 558 type</a>.</p>
 559
 560 <p>Note that the <tt>void *</tt> type is expressed as a
 561 <tt>llvm.dbg.derivedtype.type</tt> with tag of <tt>DW_TAG_pointer_type</tt> and
 562 NULL derived type.</p>
 563
 564 </div>
 565
 566 <!-- ======================================================================= -->
 567 <div class="doc_subsubsection">
 568   <a name="format_composite_type">Composite type descriptors</a>
 569 </div>
 570
 571 <div class="doc_text">
 572
 573 <pre>
 574   %<a href="#format_composite_type">llvm.dbg.compositetype.type</a> = type {
 575     uint,   ;; Tag (see below)
 576     {  }*,  ;; Reference to context
 577     sbyte*, ;; Name (may be "" for anonymous types)
 578     {  }*,  ;; Reference to compile unit where defined (may be NULL)
 579     uint,   ;; Line number where defined (may be 0)
 580     uint,   ;; Size in bits
 581     uint,   ;; Alignment in bits
 582     uint,   ;; Offset in bits
 583     {  }*   ;; Reference to array of member descriptors
 584   }
 585 </pre>
 586
 587 <p>These descriptors are used to define types that are composed of 0 or more
 588 elements.  The value of the tag varies depending on the meaning.  The following
 589 are possible tag values;</p>
 590
 591 <pre>
 592   DW_TAG_array_type = 1
 593   DW_TAG_enumeration_type = 4
 594   DW_TAG_structure_type = 19
 595   DW_TAG_union_type = 23
 596   DW_TAG_vector_type = 259
 597   DW_TAG_subroutine_type = 46
 598   DW_TAG_inheritance = 26
 599 </pre>
 600
 601 <p>The vector flag indicates that an array type is a native packed vector.</p>
 602
 603 <p>The members of array types (tag = <tt>DW_TAG_array_type</tt>) or vector types
 604 (tag = <tt>DW_TAG_vector_type</tt>) are <a href="#format_subrange">subrange
 605 descriptors</a>, each representing the range of subscripts at that level of
 606 indexing.</p>
 607
 608 <p>The members of enumeration types (tag = <tt>DW_TAG_enumeration_type</tt>) are
 609 <a href="#format_enumeration">enumerator descriptors</a>, each representing the
 610 definition of enumeration value
 611 for the set.</p>
 612
 613 <p>The members of structure (tag = <tt>DW_TAG_structure_type</tt>) or union (tag
 614 = <tt>DW_TAG_union_type</tt>) types are any one of the <a
 615 href="#format_basic_type">basic</a>, <a href="#format_derived_type">derived</a>
 616 or <a href="#format_composite_type">composite</a> type descriptors, each
 617 representing a field member of the structure or union.</p>
 618
 619 <p>For C++ classes (tag = <tt>DW_TAG_structure_type</tt>), member descriptors
 620 provide information about base classes, static members and member functions. If
 621 a member is a <a href="#format_derived_type">derived type descriptor</a> and has
 622 a tag of <tt>DW_TAG_inheritance</tt>, then the type represents a base class. If
 623 the member of is a <a href="#format_global_variables">global variable
 624 descriptor</a> then it represents a static member.  And, if the member is a <a
 625 href="#format_subprograms">subprogram descriptor</a> then it represents a member
 626 function.  For static members and member functions, <tt>getName()</tt> returns
 627 the members link or the C++ mangled name.  <tt>getDisplayName()</tt> the
 628 simplied version of the name.</p>
 629
 630 <p>The first member of subroutine (tag = <tt>DW_TAG_subroutine_type</tt>)
 631 type elements is the return type for the subroutine.  The remaining
 632 elements are the formal arguments to the subroutine.</p>
 633
 634 <p><a href="#format_composite_type">Composite type</a> location can be
 635 determined from the compile unit and line number.  The size, alignment and
 636 offset are expressed in bits and can be 64 bit values.  The alignment is used to
 637 round the offset when embedded in a <a href="#format_composite_type">composite
 638 type</a> (as an example, to keep float doubles on 64 bit boundaries.) The offset
 639 is the bit offset if embedded in a <a href="#format_composite_type">composite
 640 type</a>.</p>
 641
 642 </div>
 643
 644 <!-- ======================================================================= -->
 645 <div class="doc_subsubsection">
 646   <a name="format_subrange">Subrange descriptors</a>
 647 </div>
 648
 649 <div class="doc_text">
 650
 651 <pre>
 652   %<a href="#format_subrange">llvm.dbg.subrange.type</a> = type {
 653     uint,   ;; Tag = 33 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_subrange_type)
 654     uint,   ;; Low value
 655     uint    ;; High value
 656   }
 657 </pre>
 658
 659 <p>These descriptors are used to define ranges of array subscripts for an array
 660 <a href="#format_composite_type">composite type</a>.  The low value defines the
 661 lower bounds typically zero for C/C++.  The high value is the upper bounds.
 662 Values are 64 bit.  High - low + 1 is the size of the array.  If
 663 low == high the array will be unbounded.</p>
 664
 665 </div>
 666
 667 <!-- ======================================================================= -->
 668 <div class="doc_subsubsection">
 669   <a name="format_enumeration">Enumerator descriptors</a>
 670 </div>
 671
 672 <div class="doc_text">
 673
 674 <pre>
 675   %<a href="#format_enumeration">llvm.dbg.enumerator.type</a> = type {
 676     uint,   ;; Tag = 40 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_enumerator)
 677     sbyte*, ;; Name
 678     uint    ;; Value
 679   }
 680 </pre>
 681
 682 <p>These descriptors are used to define members of an enumeration <a
 683 href="#format_composite_type">composite type</a>, it associates the name to the
 684 value.</p>
 685
 686 </div>
 687
 688 <!-- ======================================================================= -->
 689 <div class="doc_subsubsection">
 690   <a name="format_variables">Local variables</a>
 691 </div>
 692
 693 <div class="doc_text">
 694 <pre>
 695   %<a href="#format_variables">llvm.dbg.variable.type</a> = type {
 696     uint,    ;; Tag (see below)
 697     {  }*,   ;; Context
 698     sbyte*,  ;; Name
 699     {  }*,   ;; Reference to compile unit where defined
 700     uint,    ;; Line number where defined
 701     {  }*    ;; Type descriptor
 702   }
 703 </pre>
 704
 705 <p>These descriptors are used to define variables local to a sub program.  The
 706 value of the tag depends on the usage of the variable;</p>
 707
 708 <pre>
 709   DW_TAG_auto_variable = 256
 710   DW_TAG_arg_variable = 257
 711   DW_TAG_return_variable = 258
 712 </pre>
 713
 714 <p>An auto variable is any variable declared in the body of the function.  An
 715 argument variable is any variable that appears as a formal argument to the
 716 function.  A return variable is used to track the result of a function and has
 717 no source correspondent.</p>
 718
 719 <p>The context is either the subprogram or block where the variable is defined.
 720 Name the source variable name.  Compile unit and line indicate where the
 721 variable was defined. Type descriptor defines the declared type of the
 722 variable.</p>
 723
 724 </div>
 725
 726 <!-- ======================================================================= -->
 727 <div class="doc_subsection">
 728   <a name="format_common_intrinsics">Debugger intrinsic functions</a>
 729 </div>
 730
 731 <div class="doc_text">
 732
 733 <p>LLVM uses several intrinsic functions (name prefixed with "llvm.dbg") to
 734 provide debug information at various points in generated code.</p>
 735
 736 </div>
 737
 738 <!-- ======================================================================= -->
 739 <div class="doc_subsubsection">
 740   <a name="format_common_stoppoint">llvm.dbg.stoppoint</a>
 741 </div>
 742
 743 <div class="doc_text">
 744 <pre>
 745   void %<a href="#format_common_stoppoint">llvm.dbg.stoppoint</a>( uint, uint, { }* )
 746 </pre>
 747
 748 <p>This intrinsic is used to provide correspondence between the source file and
 749 the generated code.  The first argument is the line number (base 1), second
 750 argument is the column number (0 if unknown) and the third argument the source
 751 <tt>%<a href="#format_compile_units">llvm.dbg.compile_unit</a>*</tt> cast to a
 752 <tt>{ }*</tt>.  Code following a call to this intrinsic will have been defined
 753 in close proximity of the line, column and file.  This information holds until
 754 the next call to <tt>%<a
 755 href="#format_common_stoppoint">lvm.dbg.stoppoint</a></tt>.</p>
 756
 757 </div>
 758
 759 <!-- ======================================================================= -->
 760 <div class="doc_subsubsection">
 761   <a name="format_common_func_start">llvm.dbg.func.start</a>
 762 </div>
 763
 764 <div class="doc_text">
 765 <pre>
 766   void %<a href="#format_common_func_start">llvm.dbg.func.start</a>( { }* )
 767 </pre>
 768
 769 <p>This intrinsic is used to link the debug information in <tt>%<a
 770 href="#format_subprograms">llvm.dbg.subprogram</a></tt> to the function. It also
 771 defines the beginning of the function's declarative region (scope.)  The
 772 intrinsic should be called early in the function after the all the alloca
 773 instructions.  It should be paired off with a closing <tt>%<a
 774 href="#format_common_region_end">llvm.dbg.region.end</a></tt>.  The function's
 775 single argument is the <tt>%<a
 776 href="#format_subprograms">llvm.dbg.subprogram.type</a></tt>.</p>
 777
 778 </div>
 779
 780 <!-- ======================================================================= -->
 781 <div class="doc_subsubsection">
 782   <a name="format_common_region_start">llvm.dbg.region.start</a>
 783 </div>
 784
 785 <div class="doc_text">
 786 <pre>
 787   void %<a href="#format_common_region_start">llvm.dbg.region.start</a>( { }* )
 788 </pre>
 789
 790 <p>This intrinsic is used to define the beginning of a declarative scope (ex.
 791 block) for local language elements.  It should be paired off with a closing
 792 <tt>%<a href="#format_common_region_end">llvm.dbg.region.end</a></tt>.  The
 793 function's single argument is the <tt>%<a
 794 href="#format_blocks">llvm.dbg.block</a></tt> which is starting.</p>
 795
 796
 797 </div>
 798
 799 <!-- ======================================================================= -->
 800 <div class="doc_subsubsection">
 801   <a name="format_common_region_end">llvm.dbg.region.end</a>
 802 </div>
 803
 804 <div class="doc_text">
 805 <pre>
 806   void %<a href="#format_common_region_end">llvm.dbg.region.end</a>( { }* )
 807 </pre>
 808
 809 <p>This intrinsic is used to define the end of a declarative scope (ex. block)
 810 for local language elements.  It should be paired off with an opening <tt>%<a
 811 href="#format_common_region_start">llvm.dbg.region.start</a></tt> or <tt>%<a
 812 href="#format_common_func_start">llvm.dbg.func.start</a></tt>.  The function's
 813 single argument is either the <tt>%<a
 814 href="#format_blocks">llvm.dbg.block</a></tt> or the <tt>%<a
 815 href="#format_subprograms">llvm.dbg.subprogram.type</a></tt> which is
 816 ending.</p>
 817
 818 </div>
 819
 820 <!-- ======================================================================= -->
 821 <div class="doc_subsubsection">
 822   <a name="format_common_declare">llvm.dbg.declare</a>
 823 </div>
 824
 825 <div class="doc_text">
 826 <pre>
 827   void %<a href="#format_common_declare">llvm.dbg.declare</a>( { } *, { }* )
 828 </pre>
 829
 830 <p>This intrinsic provides information about a local element (ex. variable.) The
 831 first argument is the alloca for the variable, cast to a <tt>{ }*</tt>. The
 832 second argument is the <tt>%<a
 833 href="#format_variables">llvm.dbg.variable</a></tt> containing the description
 834 of the variable, also cast to a <tt>{ }*</tt>.</p>
 835
 836 </div>
 837
 838 <!-- ======================================================================= -->
 839 <div class="doc_subsection">
 840   <a name="format_common_stoppoints">
 841      Representing stopping points in the source program
 842   </a>
 843 </div>
 844
 845 <div class="doc_text">
 846
 847 <p>LLVM debugger "stop points" are a key part of the debugging representation
 848 that allows the LLVM to maintain simple semantics for <a
 849 href="#debugopt">debugging optimized code</a>.  The basic idea is that the
 850 front-end inserts calls to the <a
 851 href="#format_common_stoppoint">%<tt>llvm.dbg.stoppoint</tt></a> intrinsic
 852 function at every point in the program where a debugger should be able to
 853 inspect the program (these correspond to places a debugger stops when you
 854 "<tt>step</tt>" through it).  The front-end can choose to place these as
 855 fine-grained as it would like (for example, before every subexpression
 856 evaluated), but it is recommended to only put them after every source statement
 857 that includes executable code.</p>
 858
 859 <p>Using calls to this intrinsic function to demark legal points for the
 860 debugger to inspect the program automatically disables any optimizations that
 861 could potentially confuse debugging information.  To non-debug-information-aware
 862 transformations, these calls simply look like calls to an external function,
 863 which they must assume to do anything (including reading or writing to any part
 864 of reachable memory).  On the other hand, it does not impact many optimizations,
 865 such as code motion of non-trapping instructions, nor does it impact
 866 optimization of subexpressions, code duplication transformations, or basic-block
 867 reordering transformations.</p>
 868
 869 </div>
 870
 871
 872 <!-- ======================================================================= -->
 873 <div class="doc_subsection">
 874   <a name="format_common_lifetime">Object lifetimes and scoping</a>
 875 </div>
 876
 877 <div class="doc_text">
 878 <p>In many languages, the local variables in functions can have their lifetime
 879 or scope limited to a subset of a function.  In the C family of languages, for
 880 example, variables are only live (readable and writable) within the source block
 881 that they are defined in.  In functional languages, values are only readable
 882 after they have been defined.  Though this is a very obvious concept, it is also
 883 non-trivial to model in LLVM, because it has no notion of scoping in this sense,
 884 and does not want to be tied to a language's scoping rules.</p>
 885
 886 <p>In order to handle this, the LLVM debug format uses the notion of "regions"
 887 of a function, delineated by calls to intrinsic functions.  These intrinsic
 888 functions define new regions of the program and indicate when the region
 889 lifetime expires.  Consider the following C fragment, for example:</p>
 890
 891 <pre>
 892 1.  void foo() {
 893 2.    int X = ...;
 894 3.    int Y = ...;
 895 4.    {
 896 5.      int Z = ...;
 897 6.      ...
 898 7.    }
 899 8.    ...
 900 9.  }
 901 </pre>
 902
 903 <p>Compiled to LLVM, this function would be represented like this:</p>
 904
 905 <pre>
 906 void %foo() {
 907 entry:
 908     %X = alloca int
 909     %Y = alloca int
 910     %Z = alloca int
 911
 912     ...
 913
 914     call void %<a href="#format_common_func_start">llvm.dbg.func.start</a>( %<a href="#format_subprograms">llvm.dbg.subprogram.type</a>* %llvm.dbg.subprogram )
 915
 916     call void %<a href="#format_common_stoppoint">llvm.dbg.stoppoint</a>( uint 2, uint 2, %<a href="#format_compile_units">llvm.dbg.compile_unit</a>* %llvm.dbg.compile_unit )
 917
 918     call void %<a href="#format_common_declare">llvm.dbg.declare</a>({}* %X, ...)
 919     call void %<a href="#format_common_declare">llvm.dbg.declare</a>({}* %Y, ...)
 920
 921     <i>;; Evaluate expression on line 2, assigning to X.</i>
 922
 923     call void %<a href="#format_common_stoppoint">llvm.dbg.stoppoint</a>( uint 3, uint 2, %<a href="#format_compile_units">llvm.dbg.compile_unit</a>* %llvm.dbg.compile_unit )
 924
 925     <i>;; Evaluate expression on line 3, assigning to Y.</i>
 926
 927     call void %<a href="#format_common_stoppoint">llvm.region.start</a>()
 928     call void %<a href="#format_common_stoppoint">llvm.dbg.stoppoint</a>( uint 5, uint 4, %<a href="#format_compile_units">llvm.dbg.compile_unit</a>* %llvm.dbg.compile_unit )
 929     call void %<a href="#format_common_declare">llvm.dbg.declare</a>({}* %X, ...)
 930
 931     <i>;; Evaluate expression on line 5, assigning to Z.</i>
 932
 933     call void %<a href="#format_common_stoppoint">llvm.dbg.stoppoint</a>( uint 7, uint 2, %<a href="#format_compile_units">llvm.dbg.compile_unit</a>* %llvm.dbg.compile_unit )
 934     call void %<a href="#format_common_region_end">llvm.region.end</a>()
 935
 936     call void %<a href="#format_common_stoppoint">llvm.dbg.stoppoint</a>( uint 9, uint 2, %<a href="#format_compile_units">llvm.dbg.compile_unit</a>* %llvm.dbg.compile_unit )
 937
 938     call void %<a href="#format_common_region_end">llvm.region.end</a>()
 939
 940     ret void
 941 }
 942 </pre>
 943
 944 <p>This example illustrates a few important details about the LLVM debugging
 945 information.  In particular, it shows how the various intrinsics are applied
 946 together to allow a debugger to analyze the relationship between statements,
 947 variable definitions, and the code used to implement the function.</p>
 948
 949 <p>The first intrinsic <tt>%<a
 950 href="#format_common_func_start">llvm.dbg.func.start</a></tt> provides
 951 a link with the <a href="#format_subprograms">subprogram descriptor</a>
 952 containing the details of this function.  This call also defines the beginning
 953 of the function region, bounded by the <tt>%<a
 954 href="#format_common_region_end">llvm.region.end</a></tt> at the end of
 955 the function.  This region is used to bracket the lifetime of variables declared
 956 within.  For a function, this outer region defines a new stack frame whose
 957 lifetime ends when the region is ended.</p>
 958
 959 <p>It is possible to define inner regions for short term variables by using the
 960 %<a href="#format_common_stoppoint"><tt>llvm.region.start</tt></a> and <a
 961 href="#format_common_region_end"><tt>%llvm.region.end</tt></a> to bound a
 962 region.  The inner region in this example would be for the block containing the
 963 declaration of Z.</p>
 964
 965 <p>Using regions to represent the boundaries of source-level functions allow
 966 LLVM interprocedural optimizations to arbitrarily modify LLVM functions without
 967 having to worry about breaking mapping information between the LLVM code and the
 968 and source-level program.  In particular, the inliner requires no modification
 969 to support inlining with debugging information: there is no explicit correlation
 970 drawn between LLVM functions and their source-level counterparts (note however,
 971 that if the inliner inlines all instances of a non-strong-linkage function into
 972 its caller that it will not be possible for the user to manually invoke the
 973 inlined function from a debugger).</p>
 974
 975 <p>Once the function has been defined, the <a
 976 href="#format_common_stoppoint"><tt>stopping point</tt></a> corresponding to
 977 line #2 (column #2) of the function is encountered.  At this point in the
 978 function, <b>no</b> local variables are live.  As lines 2 and 3 of the example
 979 are executed, their variable definitions are introduced into the program using
 980 %<a href="#format_common_declare"><tt>llvm.dbg.declare</tt></a>, without the
 981 need to specify a new region.  These variables do not require new regions to be
 982 introduced because they go out of scope at the same point in the program: line
 983 9.</p>
 984
 985 <p>In contrast, the <tt>Z</tt> variable goes out of scope at a different time,
 986 on line 7.  For this reason, it is defined within the inner region, which kills
 987 the availability of <tt>Z</tt> before the code for line 8 is executed.  In this
 988 way, regions can support arbitrary source-language scoping rules, as long as
 989 they can only be nested (ie, one scope cannot partially overlap with a part of
 990 another scope).</p>
 991
 992 <p>It is worth noting that this scoping mechanism is used to control scoping of
 993 all declarations, not just variable declarations.  For example, the scope of a
 994 C++ using declaration is controlled with this and could change how name lookup is
 995 performed.</p>
 996
 997 </div>
 998
 999
1000
1001 <!-- *********************************************************************** -->
1002 <div class="doc_section">
1003   <a name="ccxx_frontend">C/C++ front-end specific debug information</a>
1004 </div>
1005 <!-- *********************************************************************** -->
1006
1007 <div class="doc_text">
1008
1009 <p>The C and C++ front-ends represent information about the program in a format
1010 that is effectively identical to <a
1011 href="http://www.eagercon.com/dwarf/dwarf3std.htm">Dwarf 3.0</a> in terms of
1012 information content.  This allows code generators to trivially support native
1013 debuggers by generating standard dwarf information, and contains enough
1014 information for non-dwarf targets to translate it as needed.</p>
1015
1016 <p>This section describes the forms used to represent C and C++ programs. Other
1017 languages could pattern themselves after this (which itself is tuned to
1018 representing programs in the same way that Dwarf 3 does), or they could choose
1019 to provide completely different forms if they don't fit into the Dwarf model.
1020 As support for debugging information gets added to the various LLVM
1021 source-language front-ends, the information used should be documented here.</p>
1022
1023 <p>The following sections provide examples of various C/C++ constructs and the
1024 debug information that would best describe those constructs.</p>
1025
1026 </div>
1027
1028 <!-- ======================================================================= -->
1029 <div class="doc_subsection">
1030   <a name="ccxx_compile_units">C/C++ source file information</a>
1031 </div>
1032
1033 <div class="doc_text">
1034
1035 <p>Given the source files "MySource.cpp" and "MyHeader.h" located in the
1036 directory "/Users/mine/sources", the following code;</p>
1037
1038 <pre>
1039 #include "MyHeader.h"
1040
1041 int main(int argc, char *argv[]) {
1042   return 0;
1043 }
1044 </pre>
1045
1046 <p>a C/C++ front-end would generate the following descriptors;</p>
1047
1048 <pre>
1049 ...
1050 ;;
1051 ;; Define types used.  In this case we need one for compile unit anchors and one
1052 ;; for compile units.
1053 ;;
1054 %<a href="#format_anchors">llvm.dbg.anchor.type</a> = type { uint, uint }
1055 %<a href="#format_compile_units">llvm.dbg.compile_unit.type</a> = type { uint, {  }*, uint, uint, sbyte*, sbyte*, sbyte* }
1056 ...
1057 ;;
1058 ;; Define the anchor for compile units.  Note that the second field of the
1059 ;; anchor is 17, which is the same as the tag for compile units
1060 ;; (17 = DW_TAG_compile_unit.)
1061 ;;
1062 %<a href="#format_compile_units">llvm.dbg.compile_units</a> = linkonce constant %<a href="#format_anchors">llvm.dbg.anchor.type</a> { uint 0, uint 17 }, section "llvm.metadata"
1063
1064 ;;
1065 ;; Define the compile unit for the source file "/Users/mine/sources/MySource.cpp".
1066 ;;
1067 %<a href="#format_compile_units">llvm.dbg.compile_unit1</a> = internal constant %<a href="#format_compile_units">llvm.dbg.compile_unit.type</a> {
1068     uint add(uint 17, uint 262144),
1069     {  }* cast (%<a href="#format_anchors">llvm.dbg.anchor.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_units</a> to {  }*),
1070     uint 1,
1071     uint 1,
1072     sbyte* getelementptr ([13 x sbyte]* %str1, int 0, int 0),
1073     sbyte* getelementptr ([21 x sbyte]* %str2, int 0, int 0),
1074     sbyte* getelementptr ([33 x sbyte]* %str3, int 0, int 0) }, section "llvm.metadata"
1075
1076 ;;
1077 ;; Define the compile unit for the header file "/Users/mine/sources/MyHeader.h".
1078 ;;
1079 %<a href="#format_compile_units">llvm.dbg.compile_unit2</a> = internal constant %<a href="#format_compile_units">llvm.dbg.compile_unit.type</a> {
1080     uint add(uint 17, uint 262144),
1081     {  }* cast (%<a href="#format_anchors">llvm.dbg.anchor.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_units</a> to {  }*),
1082     uint 1,
1083     uint 1,
1084     sbyte* getelementptr ([11 x sbyte]* %str4, int 0, int 0),
1085     sbyte* getelementptr ([21 x sbyte]* %str2, int 0, int 0),
1086     sbyte* getelementptr ([33 x sbyte]* %str3, int 0, int 0) }, section "llvm.metadata"
1087
1088 ;;
1089 ;; Define each of the strings used in the compile units.
1090 ;;
1091 %str1 = internal constant [13 x sbyte] c"MySource.cpp\00", section "llvm.metadata";
1092 %str2 = internal constant [21 x sbyte] c"/Users/mine/sources/\00", section "llvm.metadata";
1093 %str3 = internal constant [33 x sbyte] c"4.0.1 LLVM (LLVM research group)\00", section "llvm.metadata";
1094 %str4 = internal constant [11 x sbyte] c"MyHeader.h\00", section "llvm.metadata";
1095 ...
1096 </pre>
1097
1098 </div>
1099
1100 <!-- ======================================================================= -->
1101 <div class="doc_subsection">
1102   <a name="ccxx_global_variable">C/C++ global variable information</a>
1103 </div>
1104
1105 <div class="doc_text">
1106
1107 <p>Given an integer global variable declared as follows;</p>
1108
1109 <pre>
1110 int MyGlobal = 100;
1111 </pre>
1112
1113 <p>a C/C++ front-end would generate the following descriptors;</p>
1114
1115 <pre>
1116 ;;
1117 ;; Define types used. One for global variable anchors, one for the global
1118 ;; variable descriptor, one for the global's basic type and one for the global's
1119 ;; compile unit.
1120 ;;
1121 %<a href="#format_anchors">llvm.dbg.anchor.type</a> = type { uint, uint }
1122 %<a href="#format_global_variables">llvm.dbg.global_variable.type</a> = type { uint, {  }*, {  }*, sbyte*, {  }*, uint, {  }*, bool, bool, {  }*, uint }
1123 %<a href="#format_basic_type">llvm.dbg.basictype.type</a> = type { uint, {  }*, sbyte*, {  }*, int, uint, uint, uint, uint }
1124 %<a href="#format_compile_units">llvm.dbg.compile_unit.type</a> = ...
1125 ...
1126 ;;
1127 ;; Define the global itself.
1128 ;;
1129 %MyGlobal = global int 100
1130 ...
1131 ;;
1132 ;; Define the anchor for global variables.  Note that the second field of the
1133 ;; anchor is 52, which is the same as the tag for global variables
1134 ;; (52 = DW_TAG_variable.)
1135 ;;
1136 %<a href="#format_global_variables">llvm.dbg.global_variables</a> = linkonce constant %<a href="#format_anchors">llvm.dbg.anchor.type</a> { uint 0, uint 52 }, section "llvm.metadata"
1137
1138 ;;
1139 ;; Define the global variable descriptor.  Note the reference to the global
1140 ;; variable anchor and the global variable itself.
1141 ;;
1142 %<a href="#format_global_variables">llvm.dbg.global_variable</a> = internal constant %<a href="#format_global_variables">llvm.dbg.global_variable.type</a> {
1143     uint add(uint 52, uint 262144),
1144     {  }* cast (%<a href="#format_anchors">llvm.dbg.anchor.type</a>* %<a href="#format_global_variables">llvm.dbg.global_variables</a> to {  }*),
1145     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1146     sbyte* getelementptr ([9 x sbyte]* %str1, int 0, int 0),
1147     sbyte* getelementptr ([1 x sbyte]* %str2, int 0, int 0),
1148     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1149     uint 1,
1150     {  }* cast (%<a href="#format_basic_type">llvm.dbg.basictype.type</a>* %<a href="#format_basic_type">llvm.dbg.basictype</a> to {  }*),
1151     bool false,
1152     bool true,
1153     {  }* cast (int* %MyGlobal to {  }*) }, section "llvm.metadata"
1154
1155 ;;
1156 ;; Define the basic type of 32 bit signed integer.  Note that since int is an
1157 ;; intrinsic type the source file is NULL and line 0.
1158 ;;
1159 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1160     uint add(uint 36, uint 262144),
1161     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1162     sbyte* getelementptr ([4 x sbyte]* %str3, int 0, int 0),
1163     {  }* null,
1164     int 0,
1165     uint 32,
1166     uint 32,
1167     uint 0,
1168     uint 5 }, section "llvm.metadata"
1169
1170 ;;
1171 ;; Define the names of the global variable and basic type.
1172 ;;
1173 %str1 = internal constant [9 x sbyte] c"MyGlobal\00", section "llvm.metadata"
1174 %str2 = internal constant [1 x sbyte] c"\00", section "llvm.metadata"
1175 %str3 = internal constant [4 x sbyte] c"int\00", section "llvm.metadata"
1176 </pre>
1177
1178 </div>
1179
1180 <!-- ======================================================================= -->
1181 <div class="doc_subsection">
1182   <a name="ccxx_subprogram">C/C++ function information</a>
1183 </div>
1184
1185 <div class="doc_text">
1186
1187 <p>Given a function declared as follows;</p>
1188
1189 <pre>
1190 int main(int argc, char *argv[]) {
1191   return 0;
1192 }
1193 </pre>
1194
1195 <p>a C/C++ front-end would generate the following descriptors;</p>
1196
1197 <pre>
1198 ;;
1199 ;; Define types used. One for subprogram anchors, one for the subprogram
1200 ;; descriptor, one for the global's basic type and one for the subprogram's
1201 ;; compile unit.
1202 ;;
1203 %<a href="#format_subprograms">llvm.dbg.subprogram.type</a> = type { uint, {  }*, {  }*, sbyte*, {  }*, bool, bool }
1204 %<a href="#format_anchors">llvm.dbg.anchor.type</a> = type { uint, uint }
1205 %<a href="#format_compile_units">llvm.dbg.compile_unit.type</a> = ...
1206
1207 ;;
1208 ;; Define the anchor for subprograms.  Note that the second field of the
1209 ;; anchor is 46, which is the same as the tag for subprograms
1210 ;; (46 = DW_TAG_subprogram.)
1211 ;;
1212 %<a href="#format_subprograms">llvm.dbg.subprograms</a> = linkonce constant %<a href="#format_anchors">llvm.dbg.anchor.type</a> { uint 0, uint 46 }, section "llvm.metadata"
1213
1214 ;;
1215 ;; Define the descriptor for the subprogram.  TODO - more details.
1216 ;;
1217 %<a href="#format_subprograms">llvm.dbg.subprogram</a> = internal constant %<a href="#format_subprograms">llvm.dbg.subprogram.type</a> {
1218     uint add(uint 46, uint 262144),
1219     {  }* cast (%<a href="#format_anchors">llvm.dbg.anchor.type</a>* %<a href="#format_subprograms">llvm.dbg.subprograms</a> to {  }*),
1220     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1221     sbyte* getelementptr ([5 x sbyte]* %str1, int 0, int 0),
1222     sbyte* getelementptr ([1 x sbyte]* %str2, int 0, int 0),
1223     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1224     uint 1,
1225     {  }* null,
1226     bool false,
1227     bool true }, section "llvm.metadata"
1228
1229 ;;
1230 ;; Define the name of the subprogram.
1231 ;;
1232 %str1 = internal constant [5 x sbyte] c"main\00", section "llvm.metadata"
1233 %str2 = internal constant [1 x sbyte] c"\00", section "llvm.metadata"
1234
1235 ;;
1236 ;; Define the subprogram itself.
1237 ;;
1238 int %main(int %argc, sbyte** %argv) {
1239 ...
1240 }
1241 </pre>
1242
1243 </div>
1244
1245 <!-- ======================================================================= -->
1246 <div class="doc_subsection">
1247   <a name="ccxx_basic_types">C/C++ basic types</a>
1248 </div>
1249
1250 <div class="doc_text">
1251
1252 <p>The following are the basic type descriptors for C/C++ core types;</p>
1253
1254 </div>
1255
1256 <!-- ======================================================================= -->
1257 <div class="doc_subsubsection">
1258   <a name="ccxx_basic_type_bool">bool</a>
1259 </div>
1260
1261 <div class="doc_text">
1262
1263 <pre>
1264 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1265     uint add(uint 36, uint 262144),
1266     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1267     sbyte* getelementptr ([5 x sbyte]* %str1, int 0, int 0),
1268     {  }* null,
1269     int 0,
1270     uint 32,
1271     uint 32,
1272     uint 0,
1273     uint 2 }, section "llvm.metadata"
1274 %str1 = internal constant [5 x sbyte] c"bool\00", section "llvm.metadata"
1275 </pre>
1276
1277 </div>
1278
1279 <!-- ======================================================================= -->
1280 <div class="doc_subsubsection">
1281   <a name="ccxx_basic_char">char</a>
1282 </div>
1283
1284 <div class="doc_text">
1285
1286 <pre>
1287 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1288     uint add(uint 36, uint 262144),
1289     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1290     sbyte* getelementptr ([5 x sbyte]* %str1, int 0, int 0),
1291     {  }* null,
1292     int 0,
1293     uint 8,
1294     uint 8,
1295     uint 0,
1296     uint 6 }, section "llvm.metadata"
1297 %str1 = internal constant [5 x sbyte] c"char\00", section "llvm.metadata"
1298 </pre>
1299
1300 </div>
1301
1302 <!-- ======================================================================= -->
1303 <div class="doc_subsubsection">
1304   <a name="ccxx_basic_unsigned_char">unsigned char</a>
1305 </div>
1306
1307 <div class="doc_text">
1308
1309 <pre>
1310 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1311     uint add(uint 36, uint 262144),
1312     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1313     sbyte* getelementptr ([14 x sbyte]* %str1, int 0, int 0),
1314     {  }* null,
1315     int 0,
1316     uint 8,
1317     uint 8,
1318     uint 0,
1319     uint 8 }, section "llvm.metadata"
1320 %str1 = internal constant [14 x sbyte] c"unsigned char\00", section "llvm.metadata"
1321 </pre>
1322
1323 </div>
1324
1325 <!-- ======================================================================= -->
1326 <div class="doc_subsubsection">
1327   <a name="ccxx_basic_short">short</a>
1328 </div>
1329
1330 <div class="doc_text">
1331
1332 <pre>
1333 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1334     uint add(uint 36, uint 262144),
1335     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1336     sbyte* getelementptr ([10 x sbyte]* %str1, int 0, int 0),
1337     {  }* null,
1338     int 0,
1339     uint 16,
1340     uint 16,
1341     uint 0,
1342     uint 5 }, section "llvm.metadata"
1343 %str1 = internal constant [10 x sbyte] c"short int\00", section "llvm.metadata"
1344 </pre>
1345
1346 </div>
1347
1348 <!-- ======================================================================= -->
1349 <div class="doc_subsubsection">
1350   <a name="ccxx_basic_unsigned_short">unsigned short</a>
1351 </div>
1352
1353 <div class="doc_text">
1354
1355 <pre>
1356 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1357     uint add(uint 36, uint 262144),
1358     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1359     sbyte* getelementptr ([19 x sbyte]* %str1, int 0, int 0),
1360     {  }* null,
1361     int 0,
1362     uint 16,
1363     uint 16,
1364     uint 0,
1365     uint 7 }, section "llvm.metadata"
1366 %str1 = internal constant [19 x sbyte] c"short unsigned int\00", section "llvm.metadata"
1367 </pre>
1368
1369 </div>
1370
1371 <!-- ======================================================================= -->
1372 <div class="doc_subsubsection">
1373   <a name="ccxx_basic_int">int</a>
1374 </div>
1375
1376 <div class="doc_text">
1377
1378 <pre>
1379 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1380     uint add(uint 36, uint 262144),
1381     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1382     sbyte* getelementptr ([4 x sbyte]* %str1, int 0, int 0),
1383     {  }* null,
1384     int 0,
1385     uint 32,
1386     uint 32,
1387     uint 0,
1388     uint 5 }, section "llvm.metadata"
1389 %str1 = internal constant [4 x sbyte] c"int\00", section "llvm.metadata"
1390 </pre>
1391
1392 </div>
1393
1394 <!-- ======================================================================= -->
1395 <div class="doc_subsubsection">
1396   <a name="ccxx_basic_unsigned_int">unsigned int</a>
1397 </div>
1398
1399 <div class="doc_text">
1400
1401 <pre>
1402 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1403     uint add(uint 36, uint 262144),
1404     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1405     sbyte* getelementptr ([13 x sbyte]* %str1, int 0, int 0),
1406     {  }* null,
1407     int 0,
1408     uint 32,
1409     uint 32,
1410     uint 0,
1411     uint 7 }, section "llvm.metadata"
1412 %str1 = internal constant [13 x sbyte] c"unsigned int\00", section "llvm.metadata"
1413 </pre>
1414
1415 </div>
1416
1417 <!-- ======================================================================= -->
1418 <div class="doc_subsubsection">
1419   <a name="ccxx_basic_long_long">long long</a>
1420 </div>
1421
1422 <div class="doc_text">
1423
1424 <pre>
1425 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1426     uint add(uint 36, uint 262144),
1427     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1428     sbyte* getelementptr ([14 x sbyte]* %str1, int 0, int 0),
1429     {  }* null,
1430     int 0,
1431     uint 64,
1432     uint 64,
1433     uint 0,
1434     uint 5 }, section "llvm.metadata"
1435 %str1 = internal constant [14 x sbyte] c"long long int\00", section "llvm.metadata"
1436 </pre>
1437
1438 </div>
1439
1440 <!-- ======================================================================= -->
1441 <div class="doc_subsubsection">
1442   <a name="ccxx_basic_unsigned_long_long">unsigned long long</a>
1443 </div>
1444
1445 <div class="doc_text">
1446
1447 <pre>
1448 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1449     uint add(uint 36, uint 262144),
1450     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1451     sbyte* getelementptr ([23 x sbyte]* %str1, int 0, int 0),
1452     {  }* null,
1453     int 0,
1454     uint 64,
1455     uint 64,
1456     uint 0,
1457     uint 7 }, section "llvm.metadata"
1458 %str1 = internal constant [23 x sbyte] c"long long unsigned int\00", section "llvm.metadata"
1459 </pre>
1460
1461 </div>
1462
1463 <!-- ======================================================================= -->
1464 <div class="doc_subsubsection">
1465   <a name="ccxx_basic_float">float</a>
1466 </div>
1467
1468 <div class="doc_text">
1469
1470 <pre>
1471 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1472     uint add(uint 36, uint 262144),
1473     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1474     sbyte* getelementptr ([6 x sbyte]* %str1, int 0, int 0),
1475     {  }* null,
1476     int 0,
1477     uint 32,
1478     uint 32,
1479     uint 0,
1480     uint 4 }, section "llvm.metadata"
1481 %str1 = internal constant [6 x sbyte] c"float\00", section "llvm.metadata"
1482 </pre>
1483
1484 </div>
1485
1486 <!-- ======================================================================= -->
1487 <div class="doc_subsubsection">
1488   <a name="ccxx_basic_double">double</a>
1489 </div>
1490
1491 <div class="doc_text">
1492
1493 <pre>
1494 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1495     uint add(uint 36, uint 262144),
1496     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1497     sbyte* getelementptr ([7 x sbyte]* %str1, int 0, int 0),
1498     {  }* null,
1499     int 0,
1500     uint 64,
1501     uint 64,
1502     uint 0,
1503     uint 4 }, section "llvm.metadata"
1504 %str1 = internal constant [7 x sbyte] c"double\00", section "llvm.metadata"
1505 </pre>
1506
1507 </div>
1508
1509 <!-- ======================================================================= -->
1510 <div class="doc_subsection">
1511   <a name="ccxx_derived_types">C/C++ derived types</a>
1512 </div>
1513
1514 <div class="doc_text">
1515
1516 <p>Given the following as an example of C/C++ derived type;</p>
1517
1518 <pre>
1519 typedef const int *IntPtr;
1520 </pre>
1521
1522 <p>a C/C++ front-end would generate the following descriptors;</p>
1523
1524 <pre>
1525 ;;
1526 ;; Define the typedef "IntPtr".
1527 ;;
1528 %<a href="#format_derived_type">llvm.dbg.derivedtype1</a> = internal constant %<a href="#format_derived_type">llvm.dbg.derivedtype.type</a> {
1529     uint add(uint 22, uint 262144),
1530     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1531     sbyte* getelementptr ([7 x sbyte]* %str1, int 0, int 0),
1532     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1533     int 1,
1534     uint 0,
1535     uint 0,
1536     uint 0,
1537     {  }* cast (%<a href="#format_derived_type">llvm.dbg.derivedtype.type</a>* %<a href="#format_derived_type">llvm.dbg.derivedtype2</a> to {  }*) }, section "llvm.metadata"
1538 %str1 = internal constant [7 x sbyte] c"IntPtr\00", section "llvm.metadata"
1539
1540 ;;
1541 ;; Define the pointer type.
1542 ;;
1543 %<a href="#format_derived_type">llvm.dbg.derivedtype2</a> = internal constant %<a href="#format_derived_type">llvm.dbg.derivedtype.type</a> {
1544     uint add(uint 15, uint 262144),
1545     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1546     sbyte* null,
1547     {  }* null,
1548     int 0,
1549     uint 32,
1550     uint 32,
1551     uint 0,
1552     {  }* cast (%<a href="#format_derived_type">llvm.dbg.derivedtype.type</a>* %<a href="#format_derived_type">llvm.dbg.derivedtype3</a> to {  }*) }, section "llvm.metadata"
1553
1554 ;;
1555 ;; Define the const type.
1556 ;;
1557 %<a href="#format_derived_type">llvm.dbg.derivedtype3</a> = internal constant %<a href="#format_derived_type">llvm.dbg.derivedtype.type</a> {
1558     uint add(uint 38, uint 262144),
1559     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1560     sbyte* null,
1561     {  }* null,
1562     int 0,
1563     uint 0,
1564     uint 0,
1565     uint 0,
1566     {  }* cast (%<a href="#format_basic_type">llvm.dbg.basictype.type</a>* %<a href="#format_basic_type">llvm.dbg.basictype1</a> to {  }*) }, section "llvm.metadata"
1567
1568 ;;
1569 ;; Define the int type.
1570 ;;
1571 %<a href="#format_basic_type">llvm.dbg.basictype1</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1572     uint add(uint 36, uint 262144),
1573     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1574     sbyte* getelementptr ([4 x sbyte]* %str2, int 0, int 0),
1575     {  }* null,
1576     int 0,
1577     uint 32,
1578     uint 32,
1579     uint 0,
1580     uint 5 }, section "llvm.metadata"
1581 %str2 = internal constant [4 x sbyte] c"int\00", section "llvm.metadata"
1582 </pre>
1583
1584 </div>
1585
1586 <!-- ======================================================================= -->
1587 <div class="doc_subsection">
1588   <a name="ccxx_composite_types">C/C++ struct/union types</a>
1589 </div>
1590
1591 <div class="doc_text">
1592
1593 <p>Given the following as an example of C/C++ struct type;</p>
1594
1595 <pre>
1596 struct Color {
1597   unsigned Red;
1598   unsigned Green;
1599   unsigned Blue;
1600 };
1601 </pre>
1602
1603 <p>a C/C++ front-end would generate the following descriptors;</p>
1604
1605 <pre>
1606 ;;
1607 ;; Define basic type for unsigned int.
1608 ;;
1609 %<a href="#format_basic_type">llvm.dbg.basictype</a> = internal constant %<a href="#format_basic_type">llvm.dbg.basictype.type</a> {
1610     uint add(uint 36, uint 262144),
1611     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1612     sbyte* getelementptr ([13 x sbyte]* %str1, int 0, int 0),
1613     {  }* null,
1614     int 0,
1615     uint 32,
1616     uint 32,
1617     uint 0,
1618     uint 7 }, section "llvm.metadata"
1619 %str1 = internal constant [13 x sbyte] c"unsigned int\00", section "llvm.metadata"
1620
1621 ;;
1622 ;; Define composite type for struct Color.
1623 ;;
1624 %<a href="#format_composite_type">llvm.dbg.compositetype</a> = internal constant %<a href="#format_composite_type">llvm.dbg.compositetype.type</a> {
1625     uint add(uint 19, uint 262144),
1626     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1627     sbyte* getelementptr ([6 x sbyte]* %str2, int 0, int 0),
1628     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1629     int 1,
1630     uint 96,
1631     uint 32,
1632     uint 0,
1633     {  }* null,
1634     {  }* cast ([3 x {  }*]* %llvm.dbg.array to {  }*) }, section "llvm.metadata"
1635 %str2 = internal constant [6 x sbyte] c"Color\00", section "llvm.metadata"
1636
1637 ;;
1638 ;; Define the Red field.
1639 ;;
1640 %<a href="#format_derived_type">llvm.dbg.derivedtype1</a> = internal constant %<a href="#format_derived_type">llvm.dbg.derivedtype.type</a> {
1641     uint add(uint 13, uint 262144),
1642     {  }* null,
1643     sbyte* getelementptr ([4 x sbyte]* %str3, int 0, int 0),
1644     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1645     int 2,
1646     uint 32,
1647     uint 32,
1648     uint 0,
1649     {  }* cast (%<a href="#format_basic_type">llvm.dbg.basictype.type</a>* %<a href="#format_basic_type">llvm.dbg.basictype</a> to {  }*) }, section "llvm.metadata"
1650 %str3 = internal constant [4 x sbyte] c"Red\00", section "llvm.metadata"
1651
1652 ;;
1653 ;; Define the Green field.
1654 ;;
1655 %<a href="#format_derived_type">llvm.dbg.derivedtype2</a> = internal constant %<a href="#format_derived_type">llvm.dbg.derivedtype.type</a> {
1656     uint add(uint 13, uint 262144),
1657     {  }* null,
1658     sbyte* getelementptr ([6 x sbyte]* %str4, int 0, int 0),
1659     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1660     int 3,
1661     uint 32,
1662     uint 32,
1663     uint 32,
1664     {  }* cast (%<a href="#format_basic_type">llvm.dbg.basictype.type</a>* %<a href="#format_basic_type">llvm.dbg.basictype</a> to {  }*) }, section "llvm.metadata"
1665 %str4 = internal constant [6 x sbyte] c"Green\00", section "llvm.metadata"
1666
1667 ;;
1668 ;; Define the Blue field.
1669 ;;
1670 %<a href="#format_derived_type">llvm.dbg.derivedtype3</a> = internal constant %<a href="#format_derived_type">llvm.dbg.derivedtype.type</a> {
1671     uint add(uint 13, uint 262144),
1672     {  }* null,
1673     sbyte* getelementptr ([5 x sbyte]* %str5, int 0, int 0),
1674     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1675     int 4,
1676     uint 32,
1677     uint 32,
1678     uint 64,
1679     {  }* cast (%<a href="#format_basic_type">llvm.dbg.basictype.type</a>* %<a href="#format_basic_type">llvm.dbg.basictype</a> to {  }*) }, section "llvm.metadata"
1680 %str5 = internal constant [5 x sbyte] c"Blue\00", section "llvm.metadata"
1681
1682 ;;
1683 ;; Define the array of fields used by the composite type Color.
1684 ;;
1685 %llvm.dbg.array = internal constant [3 x {  }*] [
1686       {  }* cast (%<a href="#format_derived_type">llvm.dbg.derivedtype.type</a>* %<a href="#format_derived_type">llvm.dbg.derivedtype1</a> to {  }*),
1687       {  }* cast (%<a href="#format_derived_type">llvm.dbg.derivedtype.type</a>* %<a href="#format_derived_type">llvm.dbg.derivedtype2</a> to {  }*),
1688       {  }* cast (%<a href="#format_derived_type">llvm.dbg.derivedtype.type</a>* %<a href="#format_derived_type">llvm.dbg.derivedtype3</a> to {  }*) ], section "llvm.metadata"
1689 </pre>
1690
1691 </div>
1692
1693 <!-- ======================================================================= -->
1694 <div class="doc_subsection">
1695   <a name="ccxx_enumeration_types">C/C++ enumeration types</a>
1696 </div>
1697
1698 <div class="doc_text">
1699
1700 <p>Given the following as an example of C/C++ enumeration type;</p>
1701
1702 <pre>
1703 enum Trees {
1704   Spruce = 100,
1705   Oak = 200,
1706   Maple = 300
1707 };
1708 </pre>
1709
1710 <p>a C/C++ front-end would generate the following descriptors;</p>
1711
1712 <pre>
1713 ;;
1714 ;; Define composite type for enum Trees
1715 ;;
1716 %<a href="#format_composite_type">llvm.dbg.compositetype</a> = internal constant %<a href="#format_composite_type">llvm.dbg.compositetype.type</a> {
1717     uint add(uint 4, uint 262144),
1718     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1719     sbyte* getelementptr ([6 x sbyte]* %str1, int 0, int 0),
1720     {  }* cast (%<a href="#format_compile_units">llvm.dbg.compile_unit.type</a>* %<a href="#format_compile_units">llvm.dbg.compile_unit</a> to {  }*),
1721     int 1,
1722     uint 32,
1723     uint 32,
1724     uint 0,
1725     {  }* null,
1726     {  }* cast ([3 x {  }*]* %llvm.dbg.array to {  }*) }, section "llvm.metadata"
1727 %str1 = internal constant [6 x sbyte] c"Trees\00", section "llvm.metadata"
1728
1729 ;;
1730 ;; Define Spruce enumerator.
1731 ;;
1732 %<a href="#format_enumeration">llvm.dbg.enumerator1</a> = internal constant %<a href="#format_enumeration">llvm.dbg.enumerator.type</a> {
1733     uint add(uint 40, uint 262144),
1734     sbyte* getelementptr ([7 x sbyte]* %str2, int 0, int 0),
1735     int 100 }, section "llvm.metadata"
1736 %str2 = internal constant [7 x sbyte] c"Spruce\00", section "llvm.metadata"
1737
1738 ;;
1739 ;; Define Oak enumerator.
1740 ;;
1741 %<a href="#format_enumeration">llvm.dbg.enumerator2</a> = internal constant %<a href="#format_enumeration">llvm.dbg.enumerator.type</a> {
1742     uint add(uint 40, uint 262144),
1743     sbyte* getelementptr ([4 x sbyte]* %str3, int 0, int 0),
1744     int 200 }, section "llvm.metadata"
1745 %str3 = internal constant [4 x sbyte] c"Oak\00", section "llvm.metadata"
1746
1747 ;;
1748 ;; Define Maple enumerator.
1749 ;;
1750 %<a href="#format_enumeration">llvm.dbg.enumerator3</a> = internal constant %<a href="#format_enumeration">llvm.dbg.enumerator.type</a> {
1751     uint add(uint 40, uint 262144),
1752     sbyte* getelementptr ([6 x sbyte]* %str4, int 0, int 0),
1753     int 300 }, section "llvm.metadata"
1754 %str4 = internal constant [6 x sbyte] c"Maple\00", section "llvm.metadata"
1755
1756 ;;
1757 ;; Define the array of enumerators used by composite type Trees.
1758 ;;
1759 %llvm.dbg.array = internal constant [3 x {  }*] [
1760   {  }* cast (%<a href="#format_enumeration">llvm.dbg.enumerator.type</a>* %<a href="#format_enumeration">llvm.dbg.enumerator1</a> to {  }*),
1761   {  }* cast (%<a href="#format_enumeration">llvm.dbg.enumerator.type</a>* %<a href="#format_enumeration">llvm.dbg.enumerator2</a> to {  }*),
1762   {  }* cast (%<a href="#format_enumeration">llvm.dbg.enumerator.type</a>* %<a href="#format_enumeration">llvm.dbg.enumerator3</a> to {  }*) ], section "llvm.metadata"
1763 </pre>
1764
1765 </div>
1766
1767 <!-- *********************************************************************** -->
1768
1769 <hr>
1770 <address>
1771   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1772   src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
1773   <a href="http://validator.w3.org/check/referer"><img
1774   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
1775
1776   <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1777   <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
1778   Last modified: $Date$
1779 </address>
1780
1781 </body>
1782 </html>