1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
6 <title>Source Level Debugging with LLVM</title>
7 <link rel="stylesheet" href="_static/llvm.css" type="text/css">
11 <h1>Source Level Debugging with LLVM</h1>
13 <table class="layout" style="width:100%">
17 <li><a href="#introduction">Introduction</a>
19 <li><a href="#phil">Philosophy behind LLVM debugging information</a></li>
20 <li><a href="#consumers">Debug information consumers</a></li>
21 <li><a href="#debugopt">Debugging optimized code</a></li>
23 <li><a href="#format">Debugging information format</a>
25 <li><a href="#debug_info_descriptors">Debug information descriptors</a>
27 <li><a href="#format_compile_units">Compile unit descriptors</a></li>
28 <li><a href="#format_files">File descriptors</a></li>
29 <li><a href="#format_global_variables">Global variable descriptors</a></li>
30 <li><a href="#format_subprograms">Subprogram descriptors</a></li>
31 <li><a href="#format_blocks">Block descriptors</a></li>
32 <li><a href="#format_basic_type">Basic type descriptors</a></li>
33 <li><a href="#format_derived_type">Derived type descriptors</a></li>
34 <li><a href="#format_composite_type">Composite type descriptors</a></li>
35 <li><a href="#format_subrange">Subrange descriptors</a></li>
36 <li><a href="#format_enumeration">Enumerator descriptors</a></li>
37 <li><a href="#format_variables">Local variables</a></li>
39 <li><a href="#format_common_intrinsics">Debugger intrinsic functions</a>
41 <li><a href="#format_common_declare">llvm.dbg.declare</a></li>
42 <li><a href="#format_common_value">llvm.dbg.value</a></li>
45 <li><a href="#format_common_lifetime">Object lifetimes and scoping</a></li>
46 <li><a href="#ccxx_frontend">C/C++ front-end specific debug information</a>
48 <li><a href="#ccxx_compile_units">C/C++ source file information</a></li>
49 <li><a href="#ccxx_global_variable">C/C++ global variable information</a></li>
50 <li><a href="#ccxx_subprogram">C/C++ function information</a></li>
51 <li><a href="#ccxx_basic_types">C/C++ basic types</a></li>
52 <li><a href="#ccxx_derived_types">C/C++ derived types</a></li>
53 <li><a href="#ccxx_composite_types">C/C++ struct/union types</a></li>
54 <li><a href="#ccxx_enumeration_types">C/C++ enumeration types</a></li>
56 <li><a href="#llvmdwarfextension">LLVM Dwarf Extensions</a>
58 <li><a href="#objcproperty">Debugging Information Extension
59 for Objective C Properties</a>
61 <li><a href="#objcpropertyintroduction">Introduction</a></li>
62 <li><a href="#objcpropertyproposal">Proposal</a></li>
63 <li><a href="#objcpropertynewattributes">New DWARF Attributes</a></li>
64 <li><a href="#objcpropertynewconstants">New DWARF Constants</a></li>
67 <li><a href="#acceltable">Name Accelerator Tables</a>
69 <li><a href="#acceltableintroduction">Introduction</a></li>
70 <li><a href="#acceltablehashes">Hash Tables</a></li>
71 <li><a href="#acceltabledetails">Details</a></li>
72 <li><a href="#acceltablecontents">Contents</a></li>
73 <li><a href="#acceltableextensions">Language Extensions and File Format Changes</a></li>
82 <div class="doc_author">
83 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
84 and <a href="mailto:jlaskey@mac.com">Jim Laskey</a></p>
88 <!-- *********************************************************************** -->
89 <h2><a name="introduction">Introduction</a></h2>
90 <!-- *********************************************************************** -->
94 <p>This document is the central repository for all information pertaining to
95 debug information in LLVM. It describes the <a href="#format">actual format
96 that the LLVM debug information</a> takes, which is useful for those
97 interested in creating front-ends or dealing directly with the information.
98 Further, this document provides specific examples of what debug information
99 for C/C++ looks like.</p>
101 <!-- ======================================================================= -->
103 <a name="phil">Philosophy behind LLVM debugging information</a>
108 <p>The idea of the LLVM debugging information is to capture how the important
109 pieces of the source-language's Abstract Syntax Tree map onto LLVM code.
110 Several design aspects have shaped the solution that appears here. The
111 important ones are:</p>
114 <li>Debugging information should have very little impact on the rest of the
115 compiler. No transformations, analyses, or code generators should need to
116 be modified because of debugging information.</li>
118 <li>LLVM optimizations should interact in <a href="#debugopt">well-defined and
119 easily described ways</a> with the debugging information.</li>
121 <li>Because LLVM is designed to support arbitrary programming languages,
122 LLVM-to-LLVM tools should not need to know anything about the semantics of
123 the source-level-language.</li>
125 <li>Source-level languages are often <b>widely</b> different from one another.
126 LLVM should not put any restrictions of the flavor of the source-language,
127 and the debugging information should work with any language.</li>
129 <li>With code generator support, it should be possible to use an LLVM compiler
130 to compile a program to native machine code and standard debugging
131 formats. This allows compatibility with traditional machine-code level
132 debuggers, like GDB or DBX.</li>
135 <p>The approach used by the LLVM implementation is to use a small set
136 of <a href="#format_common_intrinsics">intrinsic functions</a> to define a
137 mapping between LLVM program objects and the source-level objects. The
138 description of the source-level program is maintained in LLVM metadata
139 in an <a href="#ccxx_frontend">implementation-defined format</a>
140 (the C/C++ front-end currently uses working draft 7 of
141 the <a href="http://www.eagercon.com/dwarf/dwarf3std.htm">DWARF 3
144 <p>When a program is being debugged, a debugger interacts with the user and
145 turns the stored debug information into source-language specific information.
146 As such, a debugger must be aware of the source-language, and is thus tied to
147 a specific language or family of languages.</p>
151 <!-- ======================================================================= -->
153 <a name="consumers">Debug information consumers</a>
158 <p>The role of debug information is to provide meta information normally
159 stripped away during the compilation process. This meta information provides
160 an LLVM user a relationship between generated code and the original program
163 <p>Currently, debug information is consumed by DwarfDebug to produce dwarf
164 information used by the gdb debugger. Other targets could use the same
165 information to produce stabs or other debug forms.</p>
167 <p>It would also be reasonable to use debug information to feed profiling tools
168 for analysis of generated code, or, tools for reconstructing the original
169 source from generated code.</p>
171 <p>TODO - expound a bit more.</p>
175 <!-- ======================================================================= -->
177 <a name="debugopt">Debugging optimized code</a>
182 <p>An extremely high priority of LLVM debugging information is to make it
183 interact well with optimizations and analysis. In particular, the LLVM debug
184 information provides the following guarantees:</p>
187 <li>LLVM debug information <b>always provides information to accurately read
188 the source-level state of the program</b>, regardless of which LLVM
189 optimizations have been run, and without any modification to the
190 optimizations themselves. However, some optimizations may impact the
191 ability to modify the current state of the program with a debugger, such
192 as setting program variables, or calling functions that have been
195 <li>As desired, LLVM optimizations can be upgraded to be aware of the LLVM
196 debugging information, allowing them to update the debugging information
197 as they perform aggressive optimizations. This means that, with effort,
198 the LLVM optimizers could optimize debug code just as well as non-debug
201 <li>LLVM debug information does not prevent optimizations from
202 happening (for example inlining, basic block reordering/merging/cleanup,
203 tail duplication, etc).</li>
205 <li>LLVM debug information is automatically optimized along with the rest of
206 the program, using existing facilities. For example, duplicate
207 information is automatically merged by the linker, and unused information
208 is automatically removed.</li>
211 <p>Basically, the debug information allows you to compile a program with
212 "<tt>-O0 -g</tt>" and get full debug information, allowing you to arbitrarily
213 modify the program as it executes from a debugger. Compiling a program with
214 "<tt>-O3 -g</tt>" gives you full debug information that is always available
215 and accurate for reading (e.g., you get accurate stack traces despite tail
216 call elimination and inlining), but you might lose the ability to modify the
217 program and call functions where were optimized out of the program, or
218 inlined away completely.</p>
220 <p><a href="TestingGuide.html#quicktestsuite">LLVM test suite</a> provides a
221 framework to test optimizer's handling of debugging information. It can be
224 <div class="doc_code">
226 % cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
231 <p>This will test impact of debugging information on optimization passes. If
232 debugging information influences optimization passes then it will be reported
233 as a failure. See <a href="TestingGuide.html">TestingGuide</a> for more
234 information on LLVM test infrastructure and how to run various tests.</p>
240 <!-- *********************************************************************** -->
242 <a name="format">Debugging information format</a>
244 <!-- *********************************************************************** -->
248 <p>LLVM debugging information has been carefully designed to make it possible
249 for the optimizer to optimize the program and debugging information without
250 necessarily having to know anything about debugging information. In
251 particular, the use of metadata avoids duplicated debugging information from
252 the beginning, and the global dead code elimination pass automatically
253 deletes debugging information for a function if it decides to delete the
256 <p>To do this, most of the debugging information (descriptors for types,
257 variables, functions, source files, etc) is inserted by the language
258 front-end in the form of LLVM metadata. </p>
260 <p>Debug information is designed to be agnostic about the target debugger and
261 debugging information representation (e.g. DWARF/Stabs/etc). It uses a
262 generic pass to decode the information that represents variables, types,
263 functions, namespaces, etc: this allows for arbitrary source-language
264 semantics and type-systems to be used, as long as there is a module
265 written for the target debugger to interpret the information. </p>
267 <p>To provide basic functionality, the LLVM debugger does have to make some
268 assumptions about the source-level language being debugged, though it keeps
269 these to a minimum. The only common features that the LLVM debugger assumes
270 exist are <a href="#format_files">source files</a>,
271 and <a href="#format_global_variables">program objects</a>. These abstract
272 objects are used by a debugger to form stack traces, show information about
273 local variables, etc.</p>
275 <p>This section of the documentation first describes the representation aspects
276 common to any source-language. The <a href="#ccxx_frontend">next section</a>
277 describes the data layout conventions used by the C and C++ front-ends.</p>
279 <!-- ======================================================================= -->
281 <a name="debug_info_descriptors">Debug information descriptors</a>
286 <p>In consideration of the complexity and volume of debug information, LLVM
287 provides a specification for well formed debug descriptors. </p>
289 <p>Consumers of LLVM debug information expect the descriptors for program
290 objects to start in a canonical format, but the descriptors can include
291 additional information appended at the end that is source-language
292 specific. All LLVM debugging information is versioned, allowing backwards
293 compatibility in the case that the core structures need to change in some
294 way. Also, all debugging information objects start with a tag to indicate
295 what type of object it is. The source-language is allowed to define its own
296 objects, by using unreserved tag numbers. We recommend using with tags in
297 the range 0x1000 through 0x2000 (there is a defined enum DW_TAG_user_base =
300 <p>The fields of debug descriptors used internally by LLVM
301 are restricted to only the simple data types <tt>i32</tt>, <tt>i1</tt>,
302 <tt>float</tt>, <tt>double</tt>, <tt>mdstring</tt> and <tt>mdnode</tt>. </p>
304 <div class="doc_code">
313 <p><a name="LLVMDebugVersion">The first field of a descriptor is always an
314 <tt>i32</tt> containing a tag value identifying the content of the
315 descriptor. The remaining fields are specific to the descriptor. The values
316 of tags are loosely bound to the tag values of DWARF information entries.
317 However, that does not restrict the use of the information supplied to DWARF
318 targets. To facilitate versioning of debug information, the tag is augmented
319 with the current debug version (LLVMDebugVersion = 8 << 16 or
320 0x80000 or 524288.)</a></p>
322 <p>The details of the various descriptors follow.</p>
324 <!-- ======================================================================= -->
326 <a name="format_compile_units">Compile unit descriptors</a>
331 <div class="doc_code">
334 i32, ;; Tag = 17 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
335 ;; (DW_TAG_compile_unit)
336 i32, ;; Unused field.
337 i32, ;; DWARF language identifier (ex. DW_LANG_C89)
338 metadata, ;; Source file name
339 metadata, ;; Source file directory (includes trailing slash)
340 metadata ;; Producer (ex. "4.0.1 LLVM (LLVM research group)")
341 i1, ;; True if this is a main compile unit.
342 i1, ;; True if this is optimized.
344 i32 ;; Runtime version
345 metadata ;; List of enums types
346 metadata ;; List of retained types
347 metadata ;; List of subprograms
348 metadata ;; List of global variables
353 <p>These descriptors contain a source language ID for the file (we use the DWARF
354 3.0 ID numbers, such as <tt>DW_LANG_C89</tt>, <tt>DW_LANG_C_plus_plus</tt>,
355 <tt>DW_LANG_Cobol74</tt>, etc), three strings describing the filename,
356 working directory of the compiler, and an identifier string for the compiler
357 that produced it.</p>
359 <p>Compile unit descriptors provide the root context for objects declared in a
360 specific compilation unit. File descriptors are defined using this context.
361 These descriptors are collected by a named metadata
362 <tt>!llvm.dbg.cu</tt>. Compile unit descriptor keeps track of subprograms,
363 global variables and type information.
367 <!-- ======================================================================= -->
369 <a name="format_files">File descriptors</a>
374 <div class="doc_code">
377 i32, ;; Tag = 41 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
378 ;; (DW_TAG_file_type)
379 metadata, ;; Source file name
380 metadata, ;; Source file directory (includes trailing slash)
386 <p>These descriptors contain information for a file. Global variables and top
387 level functions would be defined using this context.k File descriptors also
388 provide context for source line correspondence. </p>
390 <p>Each input file is encoded as a separate file descriptor in LLVM debugging
391 information output. </p>
395 <!-- ======================================================================= -->
397 <a name="format_global_variables">Global variable descriptors</a>
402 <div class="doc_code">
405 i32, ;; Tag = 52 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
407 i32, ;; Unused field.
408 metadata, ;; Reference to context descriptor
410 metadata, ;; Display name (fully qualified C++ name)
411 metadata, ;; MIPS linkage name (for C++)
412 metadata, ;; Reference to file where defined
413 i32, ;; Line number where defined
414 metadata, ;; Reference to type descriptor
415 i1, ;; True if the global is local to compile unit (static)
416 i1, ;; True if the global is defined in the compile unit (not extern)
417 {}* ;; Reference to the global variable
422 <p>These descriptors provide debug information about globals variables. The
423 provide details such as name, type and where the variable is defined. All
424 global variables are collected inside the named metadata
425 <tt>!llvm.dbg.cu</tt>.</p>
429 <!-- ======================================================================= -->
431 <a name="format_subprograms">Subprogram descriptors</a>
436 <div class="doc_code">
439 i32, ;; Tag = 46 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
440 ;; (DW_TAG_subprogram)
441 i32, ;; Unused field.
442 metadata, ;; Reference to context descriptor
444 metadata, ;; Display name (fully qualified C++ name)
445 metadata, ;; MIPS linkage name (for C++)
446 metadata, ;; Reference to file where defined
447 i32, ;; Line number where defined
448 metadata, ;; Reference to type descriptor
449 i1, ;; True if the global is local to compile unit (static)
450 i1, ;; True if the global is defined in the compile unit (not extern)
451 i32, ;; Line number where the scope of the subprogram begins
452 i32, ;; Virtuality, e.g. dwarf::DW_VIRTUALITY__virtual
453 i32, ;; Index into a virtual function
454 metadata, ;; indicates which base type contains the vtable pointer for the
456 i32, ;; Flags - Artifical, Private, Protected, Explicit, Prototyped.
458 Function *,;; Pointer to LLVM function
459 metadata, ;; Lists function template parameters
460 metadata ;; Function declaration descriptor
461 metadata ;; List of function variables
466 <p>These descriptors provide debug information about functions, methods and
467 subprograms. They provide details such as name, return types and the source
468 location where the subprogram is defined.
473 <!-- ======================================================================= -->
475 <a name="format_blocks">Block descriptors</a>
480 <div class="doc_code">
483 i32, ;; Tag = 11 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_lexical_block)
484 metadata,;; Reference to context descriptor
486 i32, ;; Column number
487 metadata,;; Reference to source file
488 i32 ;; Unique ID to identify blocks from a template function
493 <p>This descriptor provides debug information about nested blocks within a
494 subprogram. The line number and column numbers are used to dinstinguish
495 two lexical blocks at same depth. </p>
497 <div class="doc_code">
500 i32, ;; Tag = 11 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_lexical_block)
501 metadata ;; Reference to the scope we're annotating with a file change
502 metadata,;; Reference to the file the scope is enclosed in.
507 <p>This descriptor provides a wrapper around a lexical scope to handle file
508 changes in the middle of a lexical block.</p>
512 <!-- ======================================================================= -->
514 <a name="format_basic_type">Basic type descriptors</a>
519 <div class="doc_code">
522 i32, ;; Tag = 36 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
523 ;; (DW_TAG_base_type)
524 metadata, ;; Reference to context
525 metadata, ;; Name (may be "" for anonymous types)
526 metadata, ;; Reference to file where defined (may be NULL)
527 i32, ;; Line number where defined (may be 0)
529 i64, ;; Alignment in bits
530 i64, ;; Offset in bits
532 i32 ;; DWARF type encoding
537 <p>These descriptors define primitive types used in the code. Example int, bool
538 and float. The context provides the scope of the type, which is usually the
539 top level. Since basic types are not usually user defined the context
540 and line number can be left as NULL and 0. The size, alignment and offset
541 are expressed in bits and can be 64 bit values. The alignment is used to
542 round the offset when embedded in a
543 <a href="#format_composite_type">composite type</a> (example to keep float
544 doubles on 64 bit boundaries.) The offset is the bit offset if embedded in
545 a <a href="#format_composite_type">composite type</a>.</p>
547 <p>The type encoding provides the details of the type. The values are typically
548 one of the following:</p>
550 <div class="doc_code">
556 DW_ATE_signed_char = 6
558 DW_ATE_unsigned_char = 8
564 <!-- ======================================================================= -->
566 <a name="format_derived_type">Derived type descriptors</a>
571 <div class="doc_code">
574 i32, ;; Tag (see below)
575 metadata, ;; Reference to context
576 metadata, ;; Name (may be "" for anonymous types)
577 metadata, ;; Reference to file where defined (may be NULL)
578 i32, ;; Line number where defined (may be 0)
580 i64, ;; Alignment in bits
581 i64, ;; Offset in bits
582 i32, ;; Flags to encode attributes, e.g. private
583 metadata, ;; Reference to type derived from
584 metadata, ;; (optional) Name of the Objective C property associated with
585 ;; Objective-C an ivar
586 metadata, ;; (optional) Name of the Objective C property getter selector.
587 metadata, ;; (optional) Name of the Objective C property setter selector.
588 i32 ;; (optional) Objective C property attributes.
593 <p>These descriptors are used to define types derived from other types. The
594 value of the tag varies depending on the meaning. The following are possible
597 <div class="doc_code">
599 DW_TAG_formal_parameter = 5
601 DW_TAG_pointer_type = 15
602 DW_TAG_reference_type = 16
604 DW_TAG_const_type = 38
605 DW_TAG_volatile_type = 53
606 DW_TAG_restrict_type = 55
610 <p><tt>DW_TAG_member</tt> is used to define a member of
611 a <a href="#format_composite_type">composite type</a>
612 or <a href="#format_subprograms">subprogram</a>. The type of the member is
613 the <a href="#format_derived_type">derived
614 type</a>. <tt>DW_TAG_formal_parameter</tt> is used to define a member which
615 is a formal argument of a subprogram.</p>
617 <p><tt>DW_TAG_typedef</tt> is used to provide a name for the derived type.</p>
619 <p><tt>DW_TAG_pointer_type</tt>, <tt>DW_TAG_reference_type</tt>,
620 <tt>DW_TAG_const_type</tt>, <tt>DW_TAG_volatile_type</tt> and
621 <tt>DW_TAG_restrict_type</tt> are used to qualify
622 the <a href="#format_derived_type">derived type</a>. </p>
624 <p><a href="#format_derived_type">Derived type</a> location can be determined
625 from the context and line number. The size, alignment and offset are
626 expressed in bits and can be 64 bit values. The alignment is used to round
627 the offset when embedded in a <a href="#format_composite_type">composite
628 type</a> (example to keep float doubles on 64 bit boundaries.) The offset is
629 the bit offset if embedded in a <a href="#format_composite_type">composite
632 <p>Note that the <tt>void *</tt> type is expressed as a type derived from NULL.
637 <!-- ======================================================================= -->
639 <a name="format_composite_type">Composite type descriptors</a>
644 <div class="doc_code">
647 i32, ;; Tag (see below)
648 metadata, ;; Reference to context
649 metadata, ;; Name (may be "" for anonymous types)
650 metadata, ;; Reference to file where defined (may be NULL)
651 i32, ;; Line number where defined (may be 0)
653 i64, ;; Alignment in bits
654 i64, ;; Offset in bits
656 metadata, ;; Reference to type derived from
657 metadata, ;; Reference to array of member descriptors
658 i32 ;; Runtime languages
663 <p>These descriptors are used to define types that are composed of 0 or more
664 elements. The value of the tag varies depending on the meaning. The following
665 are possible tag values:</p>
667 <div class="doc_code">
669 DW_TAG_array_type = 1
670 DW_TAG_enumeration_type = 4
671 DW_TAG_structure_type = 19
672 DW_TAG_union_type = 23
673 DW_TAG_vector_type = 259
674 DW_TAG_subroutine_type = 21
675 DW_TAG_inheritance = 28
679 <p>The vector flag indicates that an array type is a native packed vector.</p>
681 <p>The members of array types (tag = <tt>DW_TAG_array_type</tt>) or vector types
682 (tag = <tt>DW_TAG_vector_type</tt>) are <a href="#format_subrange">subrange
683 descriptors</a>, each representing the range of subscripts at that level of
686 <p>The members of enumeration types (tag = <tt>DW_TAG_enumeration_type</tt>) are
687 <a href="#format_enumeration">enumerator descriptors</a>, each representing
688 the definition of enumeration value for the set. All enumeration type
689 descriptors are collected inside the named metadata
690 <tt>!llvm.dbg.cu</tt>.</p>
692 <p>The members of structure (tag = <tt>DW_TAG_structure_type</tt>) or union (tag
693 = <tt>DW_TAG_union_type</tt>) types are any one of
694 the <a href="#format_basic_type">basic</a>,
695 <a href="#format_derived_type">derived</a>
696 or <a href="#format_composite_type">composite</a> type descriptors, each
697 representing a field member of the structure or union.</p>
699 <p>For C++ classes (tag = <tt>DW_TAG_structure_type</tt>), member descriptors
700 provide information about base classes, static members and member
701 functions. If a member is a <a href="#format_derived_type">derived type
702 descriptor</a> and has a tag of <tt>DW_TAG_inheritance</tt>, then the type
703 represents a base class. If the member of is
704 a <a href="#format_global_variables">global variable descriptor</a> then it
705 represents a static member. And, if the member is
706 a <a href="#format_subprograms">subprogram descriptor</a> then it represents
707 a member function. For static members and member
708 functions, <tt>getName()</tt> returns the members link or the C++ mangled
709 name. <tt>getDisplayName()</tt> the simplied version of the name.</p>
711 <p>The first member of subroutine (tag = <tt>DW_TAG_subroutine_type</tt>) type
712 elements is the return type for the subroutine. The remaining elements are
713 the formal arguments to the subroutine.</p>
715 <p><a href="#format_composite_type">Composite type</a> location can be
716 determined from the context and line number. The size, alignment and
717 offset are expressed in bits and can be 64 bit values. The alignment is used
718 to round the offset when embedded in
719 a <a href="#format_composite_type">composite type</a> (as an example, to keep
720 float doubles on 64 bit boundaries.) The offset is the bit offset if embedded
721 in a <a href="#format_composite_type">composite type</a>.</p>
725 <!-- ======================================================================= -->
727 <a name="format_subrange">Subrange descriptors</a>
732 <div class="doc_code">
735 i32, ;; Tag = 33 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_subrange_type)
742 <p>These descriptors are used to define ranges of array subscripts for an array
743 <a href="#format_composite_type">composite type</a>. The low value defines
744 the lower bounds typically zero for C/C++. The high value is the upper
745 bounds. Values are 64 bit. High - low + 1 is the size of the array. If low
746 > high the array bounds are not included in generated debugging information.
751 <!-- ======================================================================= -->
753 <a name="format_enumeration">Enumerator descriptors</a>
758 <div class="doc_code">
761 i32, ;; Tag = 40 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
762 ;; (DW_TAG_enumerator)
769 <p>These descriptors are used to define members of an
770 enumeration <a href="#format_composite_type">composite type</a>, it
771 associates the name to the value.</p>
775 <!-- ======================================================================= -->
777 <a name="format_variables">Local variables</a>
782 <div class="doc_code">
785 i32, ;; Tag (see below)
788 metadata, ;; Reference to file where defined
789 i32, ;; 24 bit - Line number where defined
790 ;; 8 bit - Argument number. 1 indicates 1st argument.
791 metadata, ;; Type descriptor
793 metadata ;; (optional) Reference to inline location
798 <p>These descriptors are used to define variables local to a sub program. The
799 value of the tag depends on the usage of the variable:</p>
801 <div class="doc_code">
803 DW_TAG_auto_variable = 256
804 DW_TAG_arg_variable = 257
805 DW_TAG_return_variable = 258
809 <p>An auto variable is any variable declared in the body of the function. An
810 argument variable is any variable that appears as a formal argument to the
811 function. A return variable is used to track the result of a function and
812 has no source correspondent.</p>
814 <p>The context is either the subprogram or block where the variable is defined.
815 Name the source variable name. Context and line indicate where the
816 variable was defined. Type descriptor defines the declared type of the
823 <!-- ======================================================================= -->
825 <a name="format_common_intrinsics">Debugger intrinsic functions</a>
830 <p>LLVM uses several intrinsic functions (name prefixed with "llvm.dbg") to
831 provide debug information at various points in generated code.</p>
833 <!-- ======================================================================= -->
835 <a name="format_common_declare">llvm.dbg.declare</a>
840 void %<a href="#format_common_declare">llvm.dbg.declare</a>(metadata, metadata)
843 <p>This intrinsic provides information about a local element (e.g., variable). The
844 first argument is metadata holding the alloca for the variable. The
845 second argument is metadata containing a description of the variable.</p>
848 <!-- ======================================================================= -->
850 <a name="format_common_value">llvm.dbg.value</a>
855 void %<a href="#format_common_value">llvm.dbg.value</a>(metadata, i64, metadata)
858 <p>This intrinsic provides information when a user source variable is set to a
859 new value. The first argument is the new value (wrapped as metadata). The
860 second argument is the offset in the user source variable where the new value
861 is written. The third argument is metadata containing a description of the
862 user source variable.</p>
867 <!-- ======================================================================= -->
869 <a name="format_common_lifetime">Object lifetimes and scoping</a>
873 <p>In many languages, the local variables in functions can have their lifetimes
874 or scopes limited to a subset of a function. In the C family of languages,
875 for example, variables are only live (readable and writable) within the
876 source block that they are defined in. In functional languages, values are
877 only readable after they have been defined. Though this is a very obvious
878 concept, it is non-trivial to model in LLVM, because it has no notion of
879 scoping in this sense, and does not want to be tied to a language's scoping
882 <p>In order to handle this, the LLVM debug format uses the metadata attached to
883 llvm instructions to encode line number and scoping information. Consider
884 the following C fragment, for example:</p>
886 <div class="doc_code">
900 <p>Compiled to LLVM, this function would be represented like this:</p>
902 <div class="doc_code">
904 define void @foo() nounwind ssp {
906 %X = alloca i32, align 4 ; <i32*> [#uses=4]
907 %Y = alloca i32, align 4 ; <i32*> [#uses=4]
908 %Z = alloca i32, align 4 ; <i32*> [#uses=3]
909 %0 = bitcast i32* %X to {}* ; <{}*> [#uses=1]
910 call void @llvm.dbg.declare(metadata !{i32 * %X}, metadata !0), !dbg !7
911 store i32 21, i32* %X, !dbg !8
912 %1 = bitcast i32* %Y to {}* ; <{}*> [#uses=1]
913 call void @llvm.dbg.declare(metadata !{i32 * %Y}, metadata !9), !dbg !10
914 store i32 22, i32* %Y, !dbg !11
915 %2 = bitcast i32* %Z to {}* ; <{}*> [#uses=1]
916 call void @llvm.dbg.declare(metadata !{i32 * %Z}, metadata !12), !dbg !14
917 store i32 23, i32* %Z, !dbg !15
918 %tmp = load i32* %X, !dbg !16 ; <i32> [#uses=1]
919 %tmp1 = load i32* %Y, !dbg !16 ; <i32> [#uses=1]
920 %add = add nsw i32 %tmp, %tmp1, !dbg !16 ; <i32> [#uses=1]
921 store i32 %add, i32* %Z, !dbg !16
922 %tmp2 = load i32* %Y, !dbg !17 ; <i32> [#uses=1]
923 store i32 %tmp2, i32* %X, !dbg !17
927 declare void @llvm.dbg.declare(metadata, metadata) nounwind readnone
929 !0 = metadata !{i32 459008, metadata !1, metadata !"X",
930 metadata !3, i32 2, metadata !6}; [ DW_TAG_auto_variable ]
931 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
932 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo", metadata !"foo",
933 metadata !"foo", metadata !3, i32 1, metadata !4,
934 i1 false, i1 true}; [DW_TAG_subprogram ]
935 !3 = metadata !{i32 458769, i32 0, i32 12, metadata !"foo.c",
936 metadata !"/private/tmp", metadata !"clang 1.1", i1 true,
937 i1 false, metadata !"", i32 0}; [DW_TAG_compile_unit ]
938 !4 = metadata !{i32 458773, metadata !3, metadata !"", null, i32 0, i64 0, i64 0,
939 i64 0, i32 0, null, metadata !5, i32 0}; [DW_TAG_subroutine_type ]
940 !5 = metadata !{null}
941 !6 = metadata !{i32 458788, metadata !3, metadata !"int", metadata !3, i32 0,
942 i64 32, i64 32, i64 0, i32 0, i32 5}; [DW_TAG_base_type ]
943 !7 = metadata !{i32 2, i32 7, metadata !1, null}
944 !8 = metadata !{i32 2, i32 3, metadata !1, null}
945 !9 = metadata !{i32 459008, metadata !1, metadata !"Y", metadata !3, i32 3,
946 metadata !6}; [ DW_TAG_auto_variable ]
947 !10 = metadata !{i32 3, i32 7, metadata !1, null}
948 !11 = metadata !{i32 3, i32 3, metadata !1, null}
949 !12 = metadata !{i32 459008, metadata !13, metadata !"Z", metadata !3, i32 5,
950 metadata !6}; [ DW_TAG_auto_variable ]
951 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
952 !14 = metadata !{i32 5, i32 9, metadata !13, null}
953 !15 = metadata !{i32 5, i32 5, metadata !13, null}
954 !16 = metadata !{i32 6, i32 5, metadata !13, null}
955 !17 = metadata !{i32 8, i32 3, metadata !1, null}
956 !18 = metadata !{i32 9, i32 1, metadata !2, null}
960 <p>This example illustrates a few important details about LLVM debugging
961 information. In particular, it shows how the <tt>llvm.dbg.declare</tt>
962 intrinsic and location information, which are attached to an instruction,
963 are applied together to allow a debugger to analyze the relationship between
964 statements, variable definitions, and the code used to implement the
967 <div class="doc_code">
969 call void @llvm.dbg.declare(metadata, metadata !0), !dbg !7
973 <p>The first intrinsic
974 <tt>%<a href="#format_common_declare">llvm.dbg.declare</a></tt>
975 encodes debugging information for the variable <tt>X</tt>. The metadata
976 <tt>!dbg !7</tt> attached to the intrinsic provides scope information for the
977 variable <tt>X</tt>.</p>
979 <div class="doc_code">
981 !7 = metadata !{i32 2, i32 7, metadata !1, null}
982 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
983 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo",
984 metadata !"foo", metadata !"foo", metadata !3, i32 1,
985 metadata !4, i1 false, i1 true}; [DW_TAG_subprogram ]
989 <p>Here <tt>!7</tt> is metadata providing location information. It has four
990 fields: line number, column number, scope, and original scope. The original
991 scope represents inline location if this instruction is inlined inside a
992 caller, and is null otherwise. In this example, scope is encoded by
993 <tt>!1</tt>. <tt>!1</tt> represents a lexical block inside the scope
994 <tt>!2</tt>, where <tt>!2</tt> is a
995 <a href="#format_subprograms">subprogram descriptor</a>. This way the
996 location information attached to the intrinsics indicates that the
997 variable <tt>X</tt> is declared at line number 2 at a function level scope in
998 function <tt>foo</tt>.</p>
1000 <p>Now lets take another example.</p>
1002 <div class="doc_code">
1004 call void @llvm.dbg.declare(metadata, metadata !12), !dbg !14
1008 <p>The second intrinsic
1009 <tt>%<a href="#format_common_declare">llvm.dbg.declare</a></tt>
1010 encodes debugging information for variable <tt>Z</tt>. The metadata
1011 <tt>!dbg !14</tt> attached to the intrinsic provides scope information for
1012 the variable <tt>Z</tt>.</p>
1014 <div class="doc_code">
1016 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
1017 !14 = metadata !{i32 5, i32 9, metadata !13, null}
1021 <p>Here <tt>!14</tt> indicates that <tt>Z</tt> is declared at line number 5 and
1022 column number 9 inside of lexical scope <tt>!13</tt>. The lexical scope
1023 itself resides inside of lexical scope <tt>!1</tt> described above.</p>
1025 <p>The scope information attached with each instruction provides a
1026 straightforward way to find instructions covered by a scope.</p>
1032 <!-- *********************************************************************** -->
1034 <a name="ccxx_frontend">C/C++ front-end specific debug information</a>
1036 <!-- *********************************************************************** -->
1040 <p>The C and C++ front-ends represent information about the program in a format
1041 that is effectively identical
1042 to <a href="http://www.eagercon.com/dwarf/dwarf3std.htm">DWARF 3.0</a> in
1043 terms of information content. This allows code generators to trivially
1044 support native debuggers by generating standard dwarf information, and
1045 contains enough information for non-dwarf targets to translate it as
1048 <p>This section describes the forms used to represent C and C++ programs. Other
1049 languages could pattern themselves after this (which itself is tuned to
1050 representing programs in the same way that DWARF 3 does), or they could
1051 choose to provide completely different forms if they don't fit into the DWARF
1052 model. As support for debugging information gets added to the various LLVM
1053 source-language front-ends, the information used should be documented
1056 <p>The following sections provide examples of various C/C++ constructs and the
1057 debug information that would best describe those constructs.</p>
1059 <!-- ======================================================================= -->
1061 <a name="ccxx_compile_units">C/C++ source file information</a>
1066 <p>Given the source files <tt>MySource.cpp</tt> and <tt>MyHeader.h</tt> located
1067 in the directory <tt>/Users/mine/sources</tt>, the following code:</p>
1069 <div class="doc_code">
1071 #include "MyHeader.h"
1073 int main(int argc, char *argv[]) {
1079 <p>a C/C++ front-end would generate the following descriptors:</p>
1081 <div class="doc_code">
1085 ;; Define the compile unit for the main source file "/Users/mine/sources/MySource.cpp".
1090 i32 4, ;; Language Id
1091 metadata !"MySource.cpp",
1092 metadata !"/Users/mine/sources",
1093 metadata !"4.2.1 (Based on Apple Inc. build 5649) (LLVM build 00)",
1094 i1 true, ;; Main Compile Unit
1095 i1 false, ;; Optimized compile unit
1096 metadata !"", ;; Compiler flags
1097 i32 0} ;; Runtime version
1100 ;; Define the file for the file "/Users/mine/sources/MySource.cpp".
1104 metadata !"MySource.cpp",
1105 metadata !"/Users/mine/sources",
1106 metadata !2 ;; Compile unit
1110 ;; Define the file for the file "/Users/mine/sources/Myheader.h"
1114 metadata !"Myheader.h"
1115 metadata !"/Users/mine/sources",
1116 metadata !2 ;; Compile unit
1123 <p>llvm::Instruction provides easy access to metadata attached with an
1124 instruction. One can extract line number information encoded in LLVM IR
1125 using <tt>Instruction::getMetadata()</tt> and
1126 <tt>DILocation::getLineNumber()</tt>.
1128 if (MDNode *N = I->getMetadata("dbg")) { // Here I is an LLVM instruction
1129 DILocation Loc(N); // DILocation is in DebugInfo.h
1130 unsigned Line = Loc.getLineNumber();
1131 StringRef File = Loc.getFilename();
1132 StringRef Dir = Loc.getDirectory();
1137 <!-- ======================================================================= -->
1139 <a name="ccxx_global_variable">C/C++ global variable information</a>
1144 <p>Given an integer global variable declared as follows:</p>
1146 <div class="doc_code">
1152 <p>a C/C++ front-end would generate the following descriptors:</p>
1154 <div class="doc_code">
1157 ;; Define the global itself.
1159 %MyGlobal = global int 100
1162 ;; List of debug info of globals
1164 !llvm.dbg.cu = !{!0}
1166 ;; Define the compile unit.
1171 metadata !"foo.cpp", ;; File
1172 metadata !"/Volumes/Data/tmp", ;; Directory
1173 metadata !"clang version 3.1 ", ;; Producer
1174 i1 true, ;; Deprecated field
1175 i1 false, ;; "isOptimized"?
1176 metadata !"", ;; Flags
1177 i32 0, ;; Runtime Version
1178 metadata !1, ;; Enum Types
1179 metadata !1, ;; Retained Types
1180 metadata !1, ;; Subprograms
1181 metadata !3 ;; Global Variables
1182 } ; [ DW_TAG_compile_unit ]
1184 ;; The Array of Global Variables
1194 ;; Define the global variable itself.
1200 metadata !"MyGlobal", ;; Name
1201 metadata !"MyGlobal", ;; Display Name
1202 metadata !"", ;; Linkage Name
1203 metadata !6, ;; File
1205 metadata !7, ;; Type
1206 i32 0, ;; IsLocalToUnit
1207 i32 1, ;; IsDefinition
1208 i32* @MyGlobal ;; LLVM-IR Value
1209 } ; [ DW_TAG_variable ]
1216 metadata !"foo.cpp", ;; File
1217 metadata !"/Volumes/Data/tmp", ;; Directory
1219 } ; [ DW_TAG_file_type ]
1227 metadata !"int", ;; Name
1230 i64 32, ;; Size in Bits
1231 i64 32, ;; Align in Bits
1235 } ; [ DW_TAG_base_type ]
1242 <!-- ======================================================================= -->
1244 <a name="ccxx_subprogram">C/C++ function information</a>
1249 <p>Given a function declared as follows:</p>
1251 <div class="doc_code">
1253 int main(int argc, char *argv[]) {
1259 <p>a C/C++ front-end would generate the following descriptors:</p>
1261 <div class="doc_code">
1264 ;; Define the anchor for subprograms. Note that the second field of the
1265 ;; anchor is 46, which is the same as the tag for subprograms
1266 ;; (46 = DW_TAG_subprogram.)
1271 metadata !1, ;; Context
1272 metadata !"main", ;; Name
1273 metadata !"main", ;; Display name
1274 metadata !"main", ;; Linkage name
1275 metadata !1, ;; File
1276 i32 1, ;; Line number
1277 metadata !4, ;; Type
1278 i1 false, ;; Is local
1279 i1 true, ;; Is definition
1280 i32 0, ;; Virtuality attribute, e.g. pure virtual function
1281 i32 0, ;; Index into virtual table for C++ methods
1282 i32 0, ;; Type that holds virtual table.
1284 i1 false, ;; True if this function is optimized
1285 Function *, ;; Pointer to llvm::Function
1286 null ;; Function template parameters
1289 ;; Define the subprogram itself.
1291 define i32 @main(i32 %argc, i8** %argv) {
1299 <!-- ======================================================================= -->
1301 <a name="ccxx_basic_types">C/C++ basic types</a>
1306 <p>The following are the basic type descriptors for C/C++ core types:</p>
1308 <!-- ======================================================================= -->
1310 <a name="ccxx_basic_type_bool">bool</a>
1315 <div class="doc_code">
1319 metadata !1, ;; Context
1320 metadata !"bool", ;; Name
1321 metadata !1, ;; File
1322 i32 0, ;; Line number
1323 i64 8, ;; Size in Bits
1324 i64 8, ;; Align in Bits
1325 i64 0, ;; Offset in Bits
1334 <!-- ======================================================================= -->
1336 <a name="ccxx_basic_char">char</a>
1341 <div class="doc_code">
1345 metadata !1, ;; Context
1346 metadata !"char", ;; Name
1347 metadata !1, ;; File
1348 i32 0, ;; Line number
1349 i64 8, ;; Size in Bits
1350 i64 8, ;; Align in Bits
1351 i64 0, ;; Offset in Bits
1360 <!-- ======================================================================= -->
1362 <a name="ccxx_basic_unsigned_char">unsigned char</a>
1367 <div class="doc_code">
1371 metadata !1, ;; Context
1372 metadata !"unsigned char",
1373 metadata !1, ;; File
1374 i32 0, ;; Line number
1375 i64 8, ;; Size in Bits
1376 i64 8, ;; Align in Bits
1377 i64 0, ;; Offset in Bits
1386 <!-- ======================================================================= -->
1388 <a name="ccxx_basic_short">short</a>
1393 <div class="doc_code">
1397 metadata !1, ;; Context
1398 metadata !"short int",
1399 metadata !1, ;; File
1400 i32 0, ;; Line number
1401 i64 16, ;; Size in Bits
1402 i64 16, ;; Align in Bits
1403 i64 0, ;; Offset in Bits
1412 <!-- ======================================================================= -->
1414 <a name="ccxx_basic_unsigned_short">unsigned short</a>
1419 <div class="doc_code">
1423 metadata !1, ;; Context
1424 metadata !"short unsigned int",
1425 metadata !1, ;; File
1426 i32 0, ;; Line number
1427 i64 16, ;; Size in Bits
1428 i64 16, ;; Align in Bits
1429 i64 0, ;; Offset in Bits
1438 <!-- ======================================================================= -->
1440 <a name="ccxx_basic_int">int</a>
1445 <div class="doc_code">
1449 metadata !1, ;; Context
1450 metadata !"int", ;; Name
1451 metadata !1, ;; File
1452 i32 0, ;; Line number
1453 i64 32, ;; Size in Bits
1454 i64 32, ;; Align in Bits
1455 i64 0, ;; Offset in Bits
1463 <!-- ======================================================================= -->
1465 <a name="ccxx_basic_unsigned_int">unsigned int</a>
1470 <div class="doc_code">
1474 metadata !1, ;; Context
1475 metadata !"unsigned int",
1476 metadata !1, ;; File
1477 i32 0, ;; Line number
1478 i64 32, ;; Size in Bits
1479 i64 32, ;; Align in Bits
1480 i64 0, ;; Offset in Bits
1489 <!-- ======================================================================= -->
1491 <a name="ccxx_basic_long_long">long long</a>
1496 <div class="doc_code">
1500 metadata !1, ;; Context
1501 metadata !"long long int",
1502 metadata !1, ;; File
1503 i32 0, ;; Line number
1504 i64 64, ;; Size in Bits
1505 i64 64, ;; Align in Bits
1506 i64 0, ;; Offset in Bits
1515 <!-- ======================================================================= -->
1517 <a name="ccxx_basic_unsigned_long_long">unsigned long long</a>
1522 <div class="doc_code">
1526 metadata !1, ;; Context
1527 metadata !"long long unsigned int",
1528 metadata !1, ;; File
1529 i32 0, ;; Line number
1530 i64 64, ;; Size in Bits
1531 i64 64, ;; Align in Bits
1532 i64 0, ;; Offset in Bits
1541 <!-- ======================================================================= -->
1543 <a name="ccxx_basic_float">float</a>
1548 <div class="doc_code">
1552 metadata !1, ;; Context
1554 metadata !1, ;; File
1555 i32 0, ;; Line number
1556 i64 32, ;; Size in Bits
1557 i64 32, ;; Align in Bits
1558 i64 0, ;; Offset in Bits
1567 <!-- ======================================================================= -->
1569 <a name="ccxx_basic_double">double</a>
1574 <div class="doc_code">
1578 metadata !1, ;; Context
1579 metadata !"double",;; Name
1580 metadata !1, ;; File
1581 i32 0, ;; Line number
1582 i64 64, ;; Size in Bits
1583 i64 64, ;; Align in Bits
1584 i64 0, ;; Offset in Bits
1595 <!-- ======================================================================= -->
1597 <a name="ccxx_derived_types">C/C++ derived types</a>
1602 <p>Given the following as an example of C/C++ derived type:</p>
1604 <div class="doc_code">
1606 typedef const int *IntPtr;
1610 <p>a C/C++ front-end would generate the following descriptors:</p>
1612 <div class="doc_code">
1615 ;; Define the typedef "IntPtr".
1619 metadata !1, ;; Context
1620 metadata !"IntPtr", ;; Name
1621 metadata !3, ;; File
1622 i32 0, ;; Line number
1623 i64 0, ;; Size in bits
1624 i64 0, ;; Align in bits
1625 i64 0, ;; Offset in bits
1627 metadata !4 ;; Derived From type
1631 ;; Define the pointer type.
1635 metadata !1, ;; Context
1636 metadata !"", ;; Name
1637 metadata !1, ;; File
1638 i32 0, ;; Line number
1639 i64 64, ;; Size in bits
1640 i64 64, ;; Align in bits
1641 i64 0, ;; Offset in bits
1643 metadata !5 ;; Derived From type
1646 ;; Define the const type.
1650 metadata !1, ;; Context
1651 metadata !"", ;; Name
1652 metadata !1, ;; File
1653 i32 0, ;; Line number
1654 i64 32, ;; Size in bits
1655 i64 32, ;; Align in bits
1656 i64 0, ;; Offset in bits
1658 metadata !6 ;; Derived From type
1661 ;; Define the int type.
1665 metadata !1, ;; Context
1666 metadata !"int", ;; Name
1667 metadata !1, ;; File
1668 i32 0, ;; Line number
1669 i64 32, ;; Size in bits
1670 i64 32, ;; Align in bits
1671 i64 0, ;; Offset in bits
1680 <!-- ======================================================================= -->
1682 <a name="ccxx_composite_types">C/C++ struct/union types</a>
1687 <p>Given the following as an example of C/C++ struct type:</p>
1689 <div class="doc_code">
1699 <p>a C/C++ front-end would generate the following descriptors:</p>
1701 <div class="doc_code">
1704 ;; Define basic type for unsigned int.
1708 metadata !1, ;; Context
1709 metadata !"unsigned int",
1710 metadata !1, ;; File
1711 i32 0, ;; Line number
1712 i64 32, ;; Size in Bits
1713 i64 32, ;; Align in Bits
1714 i64 0, ;; Offset in Bits
1719 ;; Define composite type for struct Color.
1723 metadata !1, ;; Context
1724 metadata !"Color", ;; Name
1725 metadata !1, ;; Compile unit
1726 i32 1, ;; Line number
1727 i64 96, ;; Size in bits
1728 i64 32, ;; Align in bits
1729 i64 0, ;; Offset in bits
1731 null, ;; Derived From
1732 metadata !3, ;; Elements
1733 i32 0 ;; Runtime Language
1737 ;; Define the Red field.
1741 metadata !1, ;; Context
1742 metadata !"Red", ;; Name
1743 metadata !1, ;; File
1744 i32 2, ;; Line number
1745 i64 32, ;; Size in bits
1746 i64 32, ;; Align in bits
1747 i64 0, ;; Offset in bits
1749 metadata !5 ;; Derived From type
1753 ;; Define the Green field.
1757 metadata !1, ;; Context
1758 metadata !"Green", ;; Name
1759 metadata !1, ;; File
1760 i32 3, ;; Line number
1761 i64 32, ;; Size in bits
1762 i64 32, ;; Align in bits
1763 i64 32, ;; Offset in bits
1765 metadata !5 ;; Derived From type
1769 ;; Define the Blue field.
1773 metadata !1, ;; Context
1774 metadata !"Blue", ;; Name
1775 metadata !1, ;; File
1776 i32 4, ;; Line number
1777 i64 32, ;; Size in bits
1778 i64 32, ;; Align in bits
1779 i64 64, ;; Offset in bits
1781 metadata !5 ;; Derived From type
1785 ;; Define the array of fields used by the composite type Color.
1787 !3 = metadata !{metadata !4, metadata !6, metadata !7}
1793 <!-- ======================================================================= -->
1795 <a name="ccxx_enumeration_types">C/C++ enumeration types</a>
1800 <p>Given the following as an example of C/C++ enumeration type:</p>
1802 <div class="doc_code">
1812 <p>a C/C++ front-end would generate the following descriptors:</p>
1814 <div class="doc_code">
1817 ;; Define composite type for enum Trees
1821 metadata !1, ;; Context
1822 metadata !"Trees", ;; Name
1823 metadata !1, ;; File
1824 i32 1, ;; Line number
1825 i64 32, ;; Size in bits
1826 i64 32, ;; Align in bits
1827 i64 0, ;; Offset in bits
1829 null, ;; Derived From type
1830 metadata !3, ;; Elements
1831 i32 0 ;; Runtime language
1835 ;; Define the array of enumerators used by composite type Trees.
1837 !3 = metadata !{metadata !4, metadata !5, metadata !6}
1840 ;; Define Spruce enumerator.
1842 !4 = metadata !{i32 524328, metadata !"Spruce", i64 100}
1845 ;; Define Oak enumerator.
1847 !5 = metadata !{i32 524328, metadata !"Oak", i64 200}
1850 ;; Define Maple enumerator.
1852 !6 = metadata !{i32 524328, metadata !"Maple", i64 300}
1862 <!-- *********************************************************************** -->
1864 <a name="llvmdwarfextension">Debugging information format</a>
1866 <!-- *********************************************************************** -->
1868 <!-- ======================================================================= -->
1870 <a name="objcproperty">Debugging Information Extension for Objective C Properties</a>
1873 <!-- *********************************************************************** -->
1875 <a name="objcpropertyintroduction">Introduction</a>
1877 <!-- *********************************************************************** -->
1880 <p>Objective C provides a simpler way to declare and define accessor methods
1881 using declared properties. The language provides features to declare a
1882 property and to let compiler synthesize accessor methods.
1885 <p>The debugger lets developer inspect Objective C interfaces and their
1886 instance variables and class variables. However, the debugger does not know
1887 anything about the properties defined in Objective C interfaces. The debugger
1888 consumes information generated by compiler in DWARF format. The format does
1889 not support encoding of Objective C properties. This proposal describes DWARF
1890 extensions to encode Objective C properties, which the debugger can use to let
1891 developers inspect Objective C properties.
1897 <!-- *********************************************************************** -->
1899 <a name="objcpropertyproposal">Proposal</a>
1901 <!-- *********************************************************************** -->
1904 <p>Objective C properties exist separately from class members. A property
1905 can be defined only by "setter" and "getter" selectors, and
1906 be calculated anew on each access. Or a property can just be a direct access
1907 to some declared ivar. Finally it can have an ivar "automatically
1908 synthesized" for it by the compiler, in which case the property can be
1909 referred to in user code directly using the standard C dereference syntax as
1910 well as through the property "dot" syntax, but there is no entry in
1911 the @interface declaration corresponding to this ivar.
1914 To facilitate debugging, these properties we will add a new DWARF TAG into the
1915 DW_TAG_structure_type definition for the class to hold the description of a
1916 given property, and a set of DWARF attributes that provide said description.
1917 The property tag will also contain the name and declared type of the property.
1920 If there is a related ivar, there will also be a DWARF property attribute placed
1921 in the DW_TAG_member DIE for that ivar referring back to the property TAG for
1922 that property. And in the case where the compiler synthesizes the ivar directly,
1923 the compiler is expected to generate a DW_TAG_member for that ivar (with the
1924 DW_AT_artificial set to 1), whose name will be the name used to access this
1925 ivar directly in code, and with the property attribute pointing back to the
1926 property it is backing.
1929 The following examples will serve as illustration for our discussion:
1932 <div class="doc_code">
1944 @synthesize p2 = n2;
1950 This produces the following DWARF (this is a "pseudo dwarfdump" output):
1952 <div class="doc_code">
1954 0x00000100: TAG_structure_type [7] *
1955 AT_APPLE_runtime_class( 0x10 )
1957 AT_decl_file( "Objc_Property.m" )
1960 0x00000110 TAG_APPLE_property
1962 AT_type ( {0x00000150} ( int ) )
1964 0x00000120: TAG_APPLE_property
1966 AT_type ( {0x00000150} ( int ) )
1968 0x00000130: TAG_member [8]
1970 AT_APPLE_property ( {0x00000110} "p1" )
1971 AT_type( {0x00000150} ( int ) )
1972 AT_artificial ( 0x1 )
1974 0x00000140: TAG_member [8]
1976 AT_APPLE_property ( {0x00000120} "p2" )
1977 AT_type( {0x00000150} ( int ) )
1979 0x00000150: AT_type( ( int ) )
1983 <p> Note, the current convention is that the name of the ivar for an
1984 auto-synthesized property is the name of the property from which it derives with
1985 an underscore prepended, as is shown in the example.
1986 But we actually don't need to know this convention, since we are given the name
1987 of the ivar directly.
1991 Also, it is common practice in ObjC to have different property declarations in
1992 the @interface and @implementation - e.g. to provide a read-only property in
1993 the interface,and a read-write interface in the implementation. In that case,
1994 the compiler should emit whichever property declaration will be in force in the
1995 current translation unit.
1998 <p> Developers can decorate a property with attributes which are encoded using
1999 DW_AT_APPLE_property_attribute.
2002 <div class="doc_code">
2004 @property (readonly, nonatomic) int pr;
2008 Which produces a property tag:
2010 <div class="doc_code">
2012 TAG_APPLE_property [8]
2014 AT_type ( {0x00000147} (int) )
2015 AT_APPLE_property_attribute (DW_APPLE_PROPERTY_readonly, DW_APPLE_PROPERTY_nonatomic)
2019 <p> The setter and getter method names are attached to the property using
2020 DW_AT_APPLE_property_setter and DW_AT_APPLE_property_getter attributes.
2022 <div class="doc_code">
2025 @property (setter=myOwnP3Setter:) int p3;
2026 -(void)myOwnP3Setter:(int)a;
2031 -(void)myOwnP3Setter:(int)a{ }
2037 The DWARF for this would be:
2039 <div class="doc_code">
2041 0x000003bd: TAG_structure_type [7] *
2042 AT_APPLE_runtime_class( 0x10 )
2044 AT_decl_file( "Objc_Property.m" )
2047 0x000003cd TAG_APPLE_property
2049 AT_APPLE_property_setter ( "myOwnP3Setter:" )
2050 AT_type( {0x00000147} ( int ) )
2052 0x000003f3: TAG_member [8]
2054 AT_type ( {0x00000147} ( int ) )
2055 AT_APPLE_property ( {0x000003cd} )
2056 AT_artificial ( 0x1 )
2062 <!-- *********************************************************************** -->
2064 <a name="objcpropertynewtags">New DWARF Tags</a>
2066 <!-- *********************************************************************** -->
2069 <table border="1" cellspacing="0">
2077 <td>DW_TAG_APPLE_property</td>
2084 <!-- *********************************************************************** -->
2086 <a name="objcpropertynewattributes">New DWARF Attributes</a>
2088 <!-- *********************************************************************** -->
2091 <table border="1" cellspacing="0">
2101 <td>DW_AT_APPLE_property</td>
2106 <td>DW_AT_APPLE_property_getter</td>
2111 <td>DW_AT_APPLE_property_setter</td>
2116 <td>DW_AT_APPLE_property_attribute</td>
2124 <!-- *********************************************************************** -->
2126 <a name="objcpropertynewconstants">New DWARF Constants</a>
2128 <!-- *********************************************************************** -->
2131 <table border="1" cellspacing="0">
2139 <td>DW_AT_APPLE_PROPERTY_readonly</td>
2143 <td>DW_AT_APPLE_PROPERTY_readwrite</td>
2147 <td>DW_AT_APPLE_PROPERTY_assign</td>
2151 <td>DW_AT_APPLE_PROPERTY_retain</td>
2155 <td>DW_AT_APPLE_PROPERTY_copy</td>
2159 <td>DW_AT_APPLE_PROPERTY_nonatomic</td>
2167 <!-- ======================================================================= -->
2169 <a name="acceltable">Name Accelerator Tables</a>
2171 <!-- ======================================================================= -->
2173 <!-- ======================================================================= -->
2175 <a name="acceltableintroduction">Introduction</a>
2177 <!-- ======================================================================= -->
2179 <p>The .debug_pubnames and .debug_pubtypes formats are not what a debugger
2180 needs. The "pub" in the section name indicates that the entries in the
2181 table are publicly visible names only. This means no static or hidden
2182 functions show up in the .debug_pubnames. No static variables or private class
2183 variables are in the .debug_pubtypes. Many compilers add different things to
2184 these tables, so we can't rely upon the contents between gcc, icc, or clang.</p>
2186 <p>The typical query given by users tends not to match up with the contents of
2187 these tables. For example, the DWARF spec states that "In the case of the
2188 name of a function member or static data member of a C++ structure, class or
2189 union, the name presented in the .debug_pubnames section is not the simple
2190 name given by the DW_AT_name attribute of the referenced debugging information
2191 entry, but rather the fully qualified name of the data or function member."
2192 So the only names in these tables for complex C++ entries is a fully
2193 qualified name. Debugger users tend not to enter their search strings as
2194 "a::b::c(int,const Foo&) const", but rather as "c", "b::c" , or "a::b::c". So
2195 the name entered in the name table must be demangled in order to chop it up
2196 appropriately and additional names must be manually entered into the table
2197 to make it effective as a name lookup table for debuggers to use.</p>
2199 <p>All debuggers currently ignore the .debug_pubnames table as a result of
2200 its inconsistent and useless public-only name content making it a waste of
2201 space in the object file. These tables, when they are written to disk, are
2202 not sorted in any way, leaving every debugger to do its own parsing
2203 and sorting. These tables also include an inlined copy of the string values
2204 in the table itself making the tables much larger than they need to be on
2205 disk, especially for large C++ programs.</p>
2207 <p>Can't we just fix the sections by adding all of the names we need to this
2208 table? No, because that is not what the tables are defined to contain and we
2209 won't know the difference between the old bad tables and the new good tables.
2210 At best we could make our own renamed sections that contain all of the data
2213 <p>These tables are also insufficient for what a debugger like LLDB needs.
2214 LLDB uses clang for its expression parsing where LLDB acts as a PCH. LLDB is
2215 then often asked to look for type "foo" or namespace "bar", or list items in
2216 namespace "baz". Namespaces are not included in the pubnames or pubtypes
2217 tables. Since clang asks a lot of questions when it is parsing an expression,
2218 we need to be very fast when looking up names, as it happens a lot. Having new
2219 accelerator tables that are optimized for very quick lookups will benefit
2220 this type of debugging experience greatly.</p>
2222 <p>We would like to generate name lookup tables that can be mapped into
2223 memory from disk, and used as is, with little or no up-front parsing. We would
2224 also be able to control the exact content of these different tables so they
2225 contain exactly what we need. The Name Accelerator Tables were designed
2226 to fix these issues. In order to solve these issues we need to:</p>
2229 <li>Have a format that can be mapped into memory from disk and used as is</li>
2230 <li>Lookups should be very fast</li>
2231 <li>Extensible table format so these tables can be made by many producers</li>
2232 <li>Contain all of the names needed for typical lookups out of the box</li>
2233 <li>Strict rules for the contents of tables</li>
2236 <p>Table size is important and the accelerator table format should allow the
2237 reuse of strings from common string tables so the strings for the names are
2238 not duplicated. We also want to make sure the table is ready to be used as-is
2239 by simply mapping the table into memory with minimal header parsing.</p>
2241 <p>The name lookups need to be fast and optimized for the kinds of lookups
2242 that debuggers tend to do. Optimally we would like to touch as few parts of
2243 the mapped table as possible when doing a name lookup and be able to quickly
2244 find the name entry we are looking for, or discover there are no matches. In
2245 the case of debuggers we optimized for lookups that fail most of the time.</p>
2247 <p>Each table that is defined should have strict rules on exactly what is in
2248 the accelerator tables and documented so clients can rely on the content.</p>
2252 <!-- ======================================================================= -->
2254 <a name="acceltablehashes">Hash Tables</a>
2256 <!-- ======================================================================= -->
2259 <h5>Standard Hash Tables</h5>
2261 <p>Typical hash tables have a header, buckets, and each bucket points to the
2265 <div class="doc_code">
2277 <p>The BUCKETS are an array of offsets to DATA for each hash:</p>
2279 <div class="doc_code">
2282 | 0x00001000 | BUCKETS[0]
2283 | 0x00002000 | BUCKETS[1]
2284 | 0x00002200 | BUCKETS[2]
2285 | 0x000034f0 | BUCKETS[3]
2287 | 0xXXXXXXXX | BUCKETS[n_buckets]
2292 <p>So for bucket[3] in the example above, we have an offset into the table
2293 0x000034f0 which points to a chain of entries for the bucket. Each bucket
2294 must contain a next pointer, full 32 bit hash value, the string itself,
2295 and the data for the current string value.</p>
2297 <div class="doc_code">
2300 0x000034f0: | 0x00003500 | next pointer
2301 | 0x12345678 | 32 bit hash
2302 | "erase" | string value
2303 | data[n] | HashData for this bucket
2305 0x00003500: | 0x00003550 | next pointer
2306 | 0x29273623 | 32 bit hash
2307 | "dump" | string value
2308 | data[n] | HashData for this bucket
2310 0x00003550: | 0x00000000 | next pointer
2311 | 0x82638293 | 32 bit hash
2312 | "main" | string value
2313 | data[n] | HashData for this bucket
2318 <p>The problem with this layout for debuggers is that we need to optimize for
2319 the negative lookup case where the symbol we're searching for is not present.
2320 So if we were to lookup "printf" in the table above, we would make a 32 hash
2321 for "printf", it might match bucket[3]. We would need to go to the offset
2322 0x000034f0 and start looking to see if our 32 bit hash matches. To do so, we
2323 need to read the next pointer, then read the hash, compare it, and skip to
2324 the next bucket. Each time we are skipping many bytes in memory and touching
2325 new cache pages just to do the compare on the full 32 bit hash. All of these
2326 accesses then tell us that we didn't have a match.</p>
2328 <h5>Name Hash Tables</h5>
2330 <p>To solve the issues mentioned above we have structured the hash tables
2331 a bit differently: a header, buckets, an array of all unique 32 bit hash
2332 values, followed by an array of hash value data offsets, one for each hash
2333 value, then the data for all hash values:</p>
2335 <div class="doc_code">
2351 <p>The BUCKETS in the name tables are an index into the HASHES array. By
2352 making all of the full 32 bit hash values contiguous in memory, we allow
2353 ourselves to efficiently check for a match while touching as little
2354 memory as possible. Most often checking the 32 bit hash values is as far as
2355 the lookup goes. If it does match, it usually is a match with no collisions.
2356 So for a table with "n_buckets" buckets, and "n_hashes" unique 32 bit hash
2357 values, we can clarify the contents of the BUCKETS, HASHES and OFFSETS as:</p>
2359 <div class="doc_code">
2361 .-------------------------.
2362 | HEADER.magic | uint32_t
2363 | HEADER.version | uint16_t
2364 | HEADER.hash_function | uint16_t
2365 | HEADER.bucket_count | uint32_t
2366 | HEADER.hashes_count | uint32_t
2367 | HEADER.header_data_len | uint32_t
2368 | HEADER_DATA | HeaderData
2369 |-------------------------|
2370 | BUCKETS | uint32_t[bucket_count] // 32 bit hash indexes
2371 |-------------------------|
2372 | HASHES | uint32_t[hashes_count] // 32 bit hash values
2373 |-------------------------|
2374 | OFFSETS | uint32_t[hashes_count] // 32 bit offsets to hash value data
2375 |-------------------------|
2377 `-------------------------'
2381 <p>So taking the exact same data from the standard hash example above we end up
2384 <div class="doc_code">
2394 | ... | BUCKETS[n_buckets]
2396 | 0x........ | HASHES[0]
2397 | 0x........ | HASHES[1]
2398 | 0x........ | HASHES[2]
2399 | 0x........ | HASHES[3]
2400 | 0x........ | HASHES[4]
2401 | 0x........ | HASHES[5]
2402 | 0x12345678 | HASHES[6] hash for BUCKETS[3]
2403 | 0x29273623 | HASHES[7] hash for BUCKETS[3]
2404 | 0x82638293 | HASHES[8] hash for BUCKETS[3]
2405 | 0x........ | HASHES[9]
2406 | 0x........ | HASHES[10]
2407 | 0x........ | HASHES[11]
2408 | 0x........ | HASHES[12]
2409 | 0x........ | HASHES[13]
2410 | 0x........ | HASHES[n_hashes]
2412 | 0x........ | OFFSETS[0]
2413 | 0x........ | OFFSETS[1]
2414 | 0x........ | OFFSETS[2]
2415 | 0x........ | OFFSETS[3]
2416 | 0x........ | OFFSETS[4]
2417 | 0x........ | OFFSETS[5]
2418 | 0x000034f0 | OFFSETS[6] offset for BUCKETS[3]
2419 | 0x00003500 | OFFSETS[7] offset for BUCKETS[3]
2420 | 0x00003550 | OFFSETS[8] offset for BUCKETS[3]
2421 | 0x........ | OFFSETS[9]
2422 | 0x........ | OFFSETS[10]
2423 | 0x........ | OFFSETS[11]
2424 | 0x........ | OFFSETS[12]
2425 | 0x........ | OFFSETS[13]
2426 | 0x........ | OFFSETS[n_hashes]
2434 0x000034f0: | 0x00001203 | .debug_str ("erase")
2435 | 0x00000004 | A 32 bit array count - number of HashData with name "erase"
2436 | 0x........ | HashData[0]
2437 | 0x........ | HashData[1]
2438 | 0x........ | HashData[2]
2439 | 0x........ | HashData[3]
2440 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2442 0x00003500: | 0x00001203 | String offset into .debug_str ("collision")
2443 | 0x00000002 | A 32 bit array count - number of HashData with name "collision"
2444 | 0x........ | HashData[0]
2445 | 0x........ | HashData[1]
2446 | 0x00001203 | String offset into .debug_str ("dump")
2447 | 0x00000003 | A 32 bit array count - number of HashData with name "dump"
2448 | 0x........ | HashData[0]
2449 | 0x........ | HashData[1]
2450 | 0x........ | HashData[2]
2451 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2453 0x00003550: | 0x00001203 | String offset into .debug_str ("main")
2454 | 0x00000009 | A 32 bit array count - number of HashData with name "main"
2455 | 0x........ | HashData[0]
2456 | 0x........ | HashData[1]
2457 | 0x........ | HashData[2]
2458 | 0x........ | HashData[3]
2459 | 0x........ | HashData[4]
2460 | 0x........ | HashData[5]
2461 | 0x........ | HashData[6]
2462 | 0x........ | HashData[7]
2463 | 0x........ | HashData[8]
2464 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2469 <p>So we still have all of the same data, we just organize it more efficiently
2470 for debugger lookup. If we repeat the same "printf" lookup from above, we
2471 would hash "printf" and find it matches BUCKETS[3] by taking the 32 bit hash
2472 value and modulo it by n_buckets. BUCKETS[3] contains "6" which is the index
2473 into the HASHES table. We would then compare any consecutive 32 bit hashes
2474 values in the HASHES array as long as the hashes would be in BUCKETS[3]. We
2475 do this by verifying that each subsequent hash value modulo n_buckets is still
2476 3. In the case of a failed lookup we would access the memory for BUCKETS[3], and
2477 then compare a few consecutive 32 bit hashes before we know that we have no match.
2478 We don't end up marching through multiple words of memory and we really keep the
2479 number of processor data cache lines being accessed as small as possible.</p>
2481 <p>The string hash that is used for these lookup tables is the Daniel J.
2482 Bernstein hash which is also used in the ELF GNU_HASH sections. It is a very
2483 good hash for all kinds of names in programs with very few hash collisions.</p>
2485 <p>Empty buckets are designated by using an invalid hash index of UINT32_MAX.</p>
2488 <!-- ======================================================================= -->
2490 <a name="acceltabledetails">Details</a>
2492 <!-- ======================================================================= -->
2494 <p>These name hash tables are designed to be generic where specializations of
2495 the table get to define additional data that goes into the header
2496 ("HeaderData"), how the string value is stored ("KeyType") and the content
2497 of the data for each hash value.</p>
2499 <h5>Header Layout</h5>
2500 <p>The header has a fixed part, and the specialized part. The exact format of
2502 <div class="doc_code">
2506 uint32_t magic; // 'HASH' magic value to allow endian detection
2507 uint16_t version; // Version number
2508 uint16_t hash_function; // The hash function enumeration that was used
2509 uint32_t bucket_count; // The number of buckets in this hash table
2510 uint32_t hashes_count; // The total number of unique hash values and hash data offsets in this table
2511 uint32_t header_data_len; // The bytes to skip to get to the hash indexes (buckets) for correct alignment
2512 // Specifically the length of the following HeaderData field - this does not
2513 // include the size of the preceding fields
2514 HeaderData header_data; // Implementation specific header data
2518 <p>The header starts with a 32 bit "magic" value which must be 'HASH' encoded as
2519 an ASCII integer. This allows the detection of the start of the hash table and
2520 also allows the table's byte order to be determined so the table can be
2521 correctly extracted. The "magic" value is followed by a 16 bit version number
2522 which allows the table to be revised and modified in the future. The current
2523 version number is 1. "hash_function" is a uint16_t enumeration that specifies
2524 which hash function was used to produce this table. The current values for the
2525 hash function enumerations include:</p>
2526 <div class="doc_code">
2528 enum HashFunctionType
2530 eHashFunctionDJB = 0u, // Daniel J Bernstein hash function
2534 <p>"bucket_count" is a 32 bit unsigned integer that represents how many buckets
2535 are in the BUCKETS array. "hashes_count" is the number of unique 32 bit hash
2536 values that are in the HASHES array, and is the same number of offsets are
2537 contained in the OFFSETS array. "header_data_len" specifies the size in
2538 bytes of the HeaderData that is filled in by specialized versions of this
2541 <h5>Fixed Lookup</h5>
2542 <p>The header is followed by the buckets, hashes, offsets, and hash value
2544 <div class="doc_code">
2548 uint32_t buckets[Header.bucket_count]; // An array of hash indexes into the "hashes[]" array below
2549 uint32_t hashes [Header.hashes_count]; // Every unique 32 bit hash for the entire table is in this table
2550 uint32_t offsets[Header.hashes_count]; // An offset that corresponds to each item in the "hashes[]" array above
2554 <p>"buckets" is an array of 32 bit indexes into the "hashes" array. The
2555 "hashes" array contains all of the 32 bit hash values for all names in the
2556 hash table. Each hash in the "hashes" table has an offset in the "offsets"
2557 array that points to the data for the hash value.</p>
2559 <p>This table setup makes it very easy to repurpose these tables to contain
2560 different data, while keeping the lookup mechanism the same for all tables.
2561 This layout also makes it possible to save the table to disk and map it in
2562 later and do very efficient name lookups with little or no parsing.</p>
2564 <p>DWARF lookup tables can be implemented in a variety of ways and can store
2565 a lot of information for each name. We want to make the DWARF tables
2566 extensible and able to store the data efficiently so we have used some of the
2567 DWARF features that enable efficient data storage to define exactly what kind
2568 of data we store for each name.</p>
2570 <p>The "HeaderData" contains a definition of the contents of each HashData
2571 chunk. We might want to store an offset to all of the debug information
2572 entries (DIEs) for each name. To keep things extensible, we create a list of
2573 items, or Atoms, that are contained in the data for each name. First comes the
2574 type of the data in each atom:</p>
2575 <div class="doc_code">
2580 eAtomTypeDIEOffset = 1u, // DIE offset, check form for encoding
2581 eAtomTypeCUOffset = 2u, // DIE offset of the compiler unit header that contains the item in question
2582 eAtomTypeTag = 3u, // DW_TAG_xxx value, should be encoded as DW_FORM_data1 (if no tags exceed 255) or DW_FORM_data2
2583 eAtomTypeNameFlags = 4u, // Flags from enum NameFlags
2584 eAtomTypeTypeFlags = 5u, // Flags from enum TypeFlags
2588 <p>The enumeration values and their meanings are:</p>
2589 <div class="doc_code">
2591 eAtomTypeNULL - a termination atom that specifies the end of the atom list
2592 eAtomTypeDIEOffset - an offset into the .debug_info section for the DWARF DIE for this name
2593 eAtomTypeCUOffset - an offset into the .debug_info section for the CU that contains the DIE
2594 eAtomTypeDIETag - The DW_TAG_XXX enumeration value so you don't have to parse the DWARF to see what it is
2595 eAtomTypeNameFlags - Flags for functions and global variables (isFunction, isInlined, isExternal...)
2596 eAtomTypeTypeFlags - Flags for types (isCXXClass, isObjCClass, ...)
2599 <p>Then we allow each atom type to define the atom type and how the data for
2600 each atom type data is encoded:</p>
2601 <div class="doc_code">
2605 uint16_t type; // AtomType enum value
2606 uint16_t form; // DWARF DW_FORM_XXX defines
2610 <p>The "form" type above is from the DWARF specification and defines the
2611 exact encoding of the data for the Atom type. See the DWARF specification for
2612 the DW_FORM_ definitions.</p>
2613 <div class="doc_code">
2617 uint32_t die_offset_base;
2618 uint32_t atom_count;
2619 Atoms atoms[atom_count0];
2623 <p>"HeaderData" defines the base DIE offset that should be added to any atoms
2624 that are encoded using the DW_FORM_ref1, DW_FORM_ref2, DW_FORM_ref4,
2625 DW_FORM_ref8 or DW_FORM_ref_udata. It also defines what is contained in
2626 each "HashData" object -- Atom.form tells us how large each field will be in
2627 the HashData and the Atom.type tells us how this data should be interpreted.</p>
2629 <p>For the current implementations of the ".apple_names" (all functions + globals),
2630 the ".apple_types" (names of all types that are defined), and the
2631 ".apple_namespaces" (all namespaces), we currently set the Atom array to be:</p>
2632 <div class="doc_code">
2634 HeaderData.atom_count = 1;
2635 HeaderData.atoms[0].type = eAtomTypeDIEOffset;
2636 HeaderData.atoms[0].form = DW_FORM_data4;
2639 <p>This defines the contents to be the DIE offset (eAtomTypeDIEOffset) that is
2640 encoded as a 32 bit value (DW_FORM_data4). This allows a single name to have
2641 multiple matching DIEs in a single file, which could come up with an inlined
2642 function for instance. Future tables could include more information about the
2643 DIE such as flags indicating if the DIE is a function, method, block,
2646 <p>The KeyType for the DWARF table is a 32 bit string table offset into the
2647 ".debug_str" table. The ".debug_str" is the string table for the DWARF which
2648 may already contain copies of all of the strings. This helps make sure, with
2649 help from the compiler, that we reuse the strings between all of the DWARF
2650 sections and keeps the hash table size down. Another benefit to having the
2651 compiler generate all strings as DW_FORM_strp in the debug info, is that
2652 DWARF parsing can be made much faster.</p>
2654 <p>After a lookup is made, we get an offset into the hash data. The hash data
2655 needs to be able to deal with 32 bit hash collisions, so the chunk of data
2656 at the offset in the hash data consists of a triple:</p>
2657 <div class="doc_code">
2660 uint32_t hash_data_count
2661 HashData[hash_data_count]
2664 <p>If "str_offset" is zero, then the bucket contents are done. 99.9% of the
2665 hash data chunks contain a single item (no 32 bit hash collision):</p>
2666 <div class="doc_code">
2669 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2670 | 0x00000004 | uint32_t HashData count
2671 | 0x........ | uint32_t HashData[0] DIE offset
2672 | 0x........ | uint32_t HashData[1] DIE offset
2673 | 0x........ | uint32_t HashData[2] DIE offset
2674 | 0x........ | uint32_t HashData[3] DIE offset
2675 | 0x00000000 | uint32_t KeyType (end of hash chain)
2679 <p>If there are collisions, you will have multiple valid string offsets:</p>
2680 <div class="doc_code">
2683 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2684 | 0x00000004 | uint32_t HashData count
2685 | 0x........ | uint32_t HashData[0] DIE offset
2686 | 0x........ | uint32_t HashData[1] DIE offset
2687 | 0x........ | uint32_t HashData[2] DIE offset
2688 | 0x........ | uint32_t HashData[3] DIE offset
2689 | 0x00002023 | uint32_t KeyType (.debug_str[0x0002023] => "print")
2690 | 0x00000002 | uint32_t HashData count
2691 | 0x........ | uint32_t HashData[0] DIE offset
2692 | 0x........ | uint32_t HashData[1] DIE offset
2693 | 0x00000000 | uint32_t KeyType (end of hash chain)
2697 <p>Current testing with real world C++ binaries has shown that there is around 1
2698 32 bit hash collision per 100,000 name entries.</p>
2700 <!-- ======================================================================= -->
2702 <a name="acceltablecontents">Contents</a>
2704 <!-- ======================================================================= -->
2706 <p>As we said, we want to strictly define exactly what is included in the
2707 different tables. For DWARF, we have 3 tables: ".apple_names", ".apple_types",
2708 and ".apple_namespaces".</p>
2710 <p>".apple_names" sections should contain an entry for each DWARF DIE whose
2711 DW_TAG is a DW_TAG_label, DW_TAG_inlined_subroutine, or DW_TAG_subprogram that
2712 has address attributes: DW_AT_low_pc, DW_AT_high_pc, DW_AT_ranges or
2713 DW_AT_entry_pc. It also contains DW_TAG_variable DIEs that have a DW_OP_addr
2714 in the location (global and static variables). All global and static variables
2715 should be included, including those scoped within functions and classes. For
2716 example using the following code:</p>
2717 <div class="doc_code">
2727 <p>Both of the static "var" variables would be included in the table. All
2728 functions should emit both their full names and their basenames. For C or C++,
2729 the full name is the mangled name (if available) which is usually in the
2730 DW_AT_MIPS_linkage_name attribute, and the DW_AT_name contains the function
2731 basename. If global or static variables have a mangled name in a
2732 DW_AT_MIPS_linkage_name attribute, this should be emitted along with the
2733 simple name found in the DW_AT_name attribute.</p>
2735 <p>".apple_types" sections should contain an entry for each DWARF DIE whose
2738 <li>DW_TAG_array_type</li>
2739 <li>DW_TAG_class_type</li>
2740 <li>DW_TAG_enumeration_type</li>
2741 <li>DW_TAG_pointer_type</li>
2742 <li>DW_TAG_reference_type</li>
2743 <li>DW_TAG_string_type</li>
2744 <li>DW_TAG_structure_type</li>
2745 <li>DW_TAG_subroutine_type</li>
2746 <li>DW_TAG_typedef</li>
2747 <li>DW_TAG_union_type</li>
2748 <li>DW_TAG_ptr_to_member_type</li>
2749 <li>DW_TAG_set_type</li>
2750 <li>DW_TAG_subrange_type</li>
2751 <li>DW_TAG_base_type</li>
2752 <li>DW_TAG_const_type</li>
2753 <li>DW_TAG_constant</li>
2754 <li>DW_TAG_file_type</li>
2755 <li>DW_TAG_namelist</li>
2756 <li>DW_TAG_packed_type</li>
2757 <li>DW_TAG_volatile_type</li>
2758 <li>DW_TAG_restrict_type</li>
2759 <li>DW_TAG_interface_type</li>
2760 <li>DW_TAG_unspecified_type</li>
2761 <li>DW_TAG_shared_type</li>
2763 <p>Only entries with a DW_AT_name attribute are included, and the entry must
2764 not be a forward declaration (DW_AT_declaration attribute with a non-zero value).
2765 For example, using the following code:</p>
2766 <div class="doc_code">
2775 <p>We get a few type DIEs:</p>
2776 <div class="doc_code">
2778 0x00000067: TAG_base_type [5]
2779 AT_encoding( DW_ATE_signed )
2781 AT_byte_size( 0x04 )
2783 0x0000006e: TAG_pointer_type [6]
2784 AT_type( {0x00000067} ( int ) )
2785 AT_byte_size( 0x08 )
2788 <p>The DW_TAG_pointer_type is not included because it does not have a DW_AT_name.</p>
2790 <p>".apple_namespaces" section should contain all DW_TAG_namespace DIEs. If
2791 we run into a namespace that has no name this is an anonymous namespace,
2792 and the name should be output as "(anonymous namespace)" (without the quotes).
2793 Why? This matches the output of the abi::cxa_demangle() that is in the standard
2794 C++ library that demangles mangled names.</p>
2797 <!-- ======================================================================= -->
2799 <a name="acceltableextensions">Language Extensions and File Format Changes</a>
2801 <!-- ======================================================================= -->
2803 <h5>Objective-C Extensions</h5>
2804 <p>".apple_objc" section should contain all DW_TAG_subprogram DIEs for an
2805 Objective-C class. The name used in the hash table is the name of the
2806 Objective-C class itself. If the Objective-C class has a category, then an
2807 entry is made for both the class name without the category, and for the class
2808 name with the category. So if we have a DIE at offset 0x1234 with a name
2809 of method "-[NSString(my_additions) stringWithSpecialString:]", we would add
2810 an entry for "NSString" that points to DIE 0x1234, and an entry for
2811 "NSString(my_additions)" that points to 0x1234. This allows us to quickly
2812 track down all Objective-C methods for an Objective-C class when doing
2813 expressions. It is needed because of the dynamic nature of Objective-C where
2814 anyone can add methods to a class. The DWARF for Objective-C methods is also
2815 emitted differently from C++ classes where the methods are not usually
2816 contained in the class definition, they are scattered about across one or more
2817 compile units. Categories can also be defined in different shared libraries.
2818 So we need to be able to quickly find all of the methods and class functions
2819 given the Objective-C class name, or quickly find all methods and class
2820 functions for a class + category name. This table does not contain any selector
2821 names, it just maps Objective-C class names (or class names + category) to all
2822 of the methods and class functions. The selectors are added as function
2823 basenames in the .debug_names section.</p>
2825 <p>In the ".apple_names" section for Objective-C functions, the full name is the
2826 entire function name with the brackets ("-[NSString stringWithCString:]") and the
2827 basename is the selector only ("stringWithCString:").</p>
2829 <h5>Mach-O Changes</h5>
2830 <p>The sections names for the apple hash tables are for non mach-o files. For
2831 mach-o files, the sections should be contained in the "__DWARF" segment with
2832 names as follows:</p>
2834 <li>".apple_names" -> "__apple_names"</li>
2835 <li>".apple_types" -> "__apple_types"</li>
2836 <li>".apple_namespaces" -> "__apple_namespac" (16 character limit)</li>
2837 <li> ".apple_objc" -> "__apple_objc"</li>
2843 <!-- *********************************************************************** -->
2847 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2848 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2849 <a href="http://validator.w3.org/check/referer"><img
2850 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2852 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2853 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
2854 Last modified: $Date$