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="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>
81 <img src="img/venusflytrap.jpg" alt="A leafy and green bug eater" width="247"
86 <div class="doc_author">
87 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
88 and <a href="mailto:jlaskey@mac.com">Jim Laskey</a></p>
92 <!-- *********************************************************************** -->
93 <h2><a name="introduction">Introduction</a></h2>
94 <!-- *********************************************************************** -->
98 <p>This document is the central repository for all information pertaining to
99 debug information in LLVM. It describes the <a href="#format">actual format
100 that the LLVM debug information</a> takes, which is useful for those
101 interested in creating front-ends or dealing directly with the information.
102 Further, this document provides specific examples of what debug information
103 for C/C++ looks like.</p>
105 <!-- ======================================================================= -->
107 <a name="phil">Philosophy behind LLVM debugging information</a>
112 <p>The idea of the LLVM debugging information is to capture how the important
113 pieces of the source-language's Abstract Syntax Tree map onto LLVM code.
114 Several design aspects have shaped the solution that appears here. The
115 important ones are:</p>
118 <li>Debugging information should have very little impact on the rest of the
119 compiler. No transformations, analyses, or code generators should need to
120 be modified because of debugging information.</li>
122 <li>LLVM optimizations should interact in <a href="#debugopt">well-defined and
123 easily described ways</a> with the debugging information.</li>
125 <li>Because LLVM is designed to support arbitrary programming languages,
126 LLVM-to-LLVM tools should not need to know anything about the semantics of
127 the source-level-language.</li>
129 <li>Source-level languages are often <b>widely</b> different from one another.
130 LLVM should not put any restrictions of the flavor of the source-language,
131 and the debugging information should work with any language.</li>
133 <li>With code generator support, it should be possible to use an LLVM compiler
134 to compile a program to native machine code and standard debugging
135 formats. This allows compatibility with traditional machine-code level
136 debuggers, like GDB or DBX.</li>
139 <p>The approach used by the LLVM implementation is to use a small set
140 of <a href="#format_common_intrinsics">intrinsic functions</a> to define a
141 mapping between LLVM program objects and the source-level objects. The
142 description of the source-level program is maintained in LLVM metadata
143 in an <a href="#ccxx_frontend">implementation-defined format</a>
144 (the C/C++ front-end currently uses working draft 7 of
145 the <a href="http://www.eagercon.com/dwarf/dwarf3std.htm">DWARF 3
148 <p>When a program is being debugged, a debugger interacts with the user and
149 turns the stored debug information into source-language specific information.
150 As such, a debugger must be aware of the source-language, and is thus tied to
151 a specific language or family of languages.</p>
155 <!-- ======================================================================= -->
157 <a name="consumers">Debug information consumers</a>
162 <p>The role of debug information is to provide meta information normally
163 stripped away during the compilation process. This meta information provides
164 an LLVM user a relationship between generated code and the original program
167 <p>Currently, debug information is consumed by DwarfDebug to produce dwarf
168 information used by the gdb debugger. Other targets could use the same
169 information to produce stabs or other debug forms.</p>
171 <p>It would also be reasonable to use debug information to feed profiling tools
172 for analysis of generated code, or, tools for reconstructing the original
173 source from generated code.</p>
175 <p>TODO - expound a bit more.</p>
179 <!-- ======================================================================= -->
181 <a name="debugopt">Debugging optimized code</a>
186 <p>An extremely high priority of LLVM debugging information is to make it
187 interact well with optimizations and analysis. In particular, the LLVM debug
188 information provides the following guarantees:</p>
191 <li>LLVM debug information <b>always provides information to accurately read
192 the source-level state of the program</b>, regardless of which LLVM
193 optimizations have been run, and without any modification to the
194 optimizations themselves. However, some optimizations may impact the
195 ability to modify the current state of the program with a debugger, such
196 as setting program variables, or calling functions that have been
199 <li>As desired, LLVM optimizations can be upgraded to be aware of the LLVM
200 debugging information, allowing them to update the debugging information
201 as they perform aggressive optimizations. This means that, with effort,
202 the LLVM optimizers could optimize debug code just as well as non-debug
205 <li>LLVM debug information does not prevent optimizations from
206 happening (for example inlining, basic block reordering/merging/cleanup,
207 tail duplication, etc).</li>
209 <li>LLVM debug information is automatically optimized along with the rest of
210 the program, using existing facilities. For example, duplicate
211 information is automatically merged by the linker, and unused information
212 is automatically removed.</li>
215 <p>Basically, the debug information allows you to compile a program with
216 "<tt>-O0 -g</tt>" and get full debug information, allowing you to arbitrarily
217 modify the program as it executes from a debugger. Compiling a program with
218 "<tt>-O3 -g</tt>" gives you full debug information that is always available
219 and accurate for reading (e.g., you get accurate stack traces despite tail
220 call elimination and inlining), but you might lose the ability to modify the
221 program and call functions where were optimized out of the program, or
222 inlined away completely.</p>
224 <p><a href="TestingGuide.html#quicktestsuite">LLVM test suite</a> provides a
225 framework to test optimizer's handling of debugging information. It can be
228 <div class="doc_code">
230 % cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
235 <p>This will test impact of debugging information on optimization passes. If
236 debugging information influences optimization passes then it will be reported
237 as a failure. See <a href="TestingGuide.html">TestingGuide</a> for more
238 information on LLVM test infrastructure and how to run various tests.</p>
244 <!-- *********************************************************************** -->
246 <a name="format">Debugging information format</a>
248 <!-- *********************************************************************** -->
252 <p>LLVM debugging information has been carefully designed to make it possible
253 for the optimizer to optimize the program and debugging information without
254 necessarily having to know anything about debugging information. In
255 particular, the use of metadata avoids duplicated debugging information from
256 the beginning, and the global dead code elimination pass automatically
257 deletes debugging information for a function if it decides to delete the
260 <p>To do this, most of the debugging information (descriptors for types,
261 variables, functions, source files, etc) is inserted by the language
262 front-end in the form of LLVM metadata. </p>
264 <p>Debug information is designed to be agnostic about the target debugger and
265 debugging information representation (e.g. DWARF/Stabs/etc). It uses a
266 generic pass to decode the information that represents variables, types,
267 functions, namespaces, etc: this allows for arbitrary source-language
268 semantics and type-systems to be used, as long as there is a module
269 written for the target debugger to interpret the information. </p>
271 <p>To provide basic functionality, the LLVM debugger does have to make some
272 assumptions about the source-level language being debugged, though it keeps
273 these to a minimum. The only common features that the LLVM debugger assumes
274 exist are <a href="#format_files">source files</a>,
275 and <a href="#format_global_variables">program objects</a>. These abstract
276 objects are used by a debugger to form stack traces, show information about
277 local variables, etc.</p>
279 <p>This section of the documentation first describes the representation aspects
280 common to any source-language. The <a href="#ccxx_frontend">next section</a>
281 describes the data layout conventions used by the C and C++ front-ends.</p>
283 <!-- ======================================================================= -->
285 <a name="debug_info_descriptors">Debug information descriptors</a>
290 <p>In consideration of the complexity and volume of debug information, LLVM
291 provides a specification for well formed debug descriptors. </p>
293 <p>Consumers of LLVM debug information expect the descriptors for program
294 objects to start in a canonical format, but the descriptors can include
295 additional information appended at the end that is source-language
296 specific. All LLVM debugging information is versioned, allowing backwards
297 compatibility in the case that the core structures need to change in some
298 way. Also, all debugging information objects start with a tag to indicate
299 what type of object it is. The source-language is allowed to define its own
300 objects, by using unreserved tag numbers. We recommend using with tags in
301 the range 0x1000 through 0x2000 (there is a defined enum DW_TAG_user_base =
304 <p>The fields of debug descriptors used internally by LLVM
305 are restricted to only the simple data types <tt>i32</tt>, <tt>i1</tt>,
306 <tt>float</tt>, <tt>double</tt>, <tt>mdstring</tt> and <tt>mdnode</tt>. </p>
308 <div class="doc_code">
317 <p><a name="LLVMDebugVersion">The first field of a descriptor is always an
318 <tt>i32</tt> containing a tag value identifying the content of the
319 descriptor. The remaining fields are specific to the descriptor. The values
320 of tags are loosely bound to the tag values of DWARF information entries.
321 However, that does not restrict the use of the information supplied to DWARF
322 targets. To facilitate versioning of debug information, the tag is augmented
323 with the current debug version (LLVMDebugVersion = 8 << 16 or
324 0x80000 or 524288.)</a></p>
326 <p>The details of the various descriptors follow.</p>
328 <!-- ======================================================================= -->
330 <a name="format_compile_units">Compile unit descriptors</a>
335 <div class="doc_code">
338 i32, ;; Tag = 17 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
339 ;; (DW_TAG_compile_unit)
340 i32, ;; Unused field.
341 i32, ;; DWARF language identifier (ex. DW_LANG_C89)
342 metadata, ;; Source file name
343 metadata, ;; Source file directory (includes trailing slash)
344 metadata ;; Producer (ex. "4.0.1 LLVM (LLVM research group)")
345 i1, ;; True if this is a main compile unit.
346 i1, ;; True if this is optimized.
348 i32 ;; Runtime version
349 metadata ;; List of enums types
350 metadata ;; List of retained types
351 metadata ;; List of subprograms
352 metadata ;; List of global variables
357 <p>These descriptors contain a source language ID for the file (we use the DWARF
358 3.0 ID numbers, such as <tt>DW_LANG_C89</tt>, <tt>DW_LANG_C_plus_plus</tt>,
359 <tt>DW_LANG_Cobol74</tt>, etc), three strings describing the filename,
360 working directory of the compiler, and an identifier string for the compiler
361 that produced it.</p>
363 <p>Compile unit descriptors provide the root context for objects declared in a
364 specific compilation unit. File descriptors are defined using this context.
365 These descriptors are collected by a named metadata
366 <tt>!llvm.dbg.cu</tt>. Compile unit descriptor keeps track of subprograms,
367 global variables and type information.
371 <!-- ======================================================================= -->
373 <a name="format_files">File descriptors</a>
378 <div class="doc_code">
381 i32, ;; Tag = 41 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
382 ;; (DW_TAG_file_type)
383 metadata, ;; Source file name
384 metadata, ;; Source file directory (includes trailing slash)
390 <p>These descriptors contain information for a file. Global variables and top
391 level functions would be defined using this context.k File descriptors also
392 provide context for source line correspondence. </p>
394 <p>Each input file is encoded as a separate file descriptor in LLVM debugging
395 information output. </p>
399 <!-- ======================================================================= -->
401 <a name="format_global_variables">Global variable descriptors</a>
406 <div class="doc_code">
409 i32, ;; Tag = 52 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
411 i32, ;; Unused field.
412 metadata, ;; Reference to context descriptor
414 metadata, ;; Display name (fully qualified C++ name)
415 metadata, ;; MIPS linkage name (for C++)
416 metadata, ;; Reference to file where defined
417 i32, ;; Line number where defined
418 metadata, ;; Reference to type descriptor
419 i1, ;; True if the global is local to compile unit (static)
420 i1, ;; True if the global is defined in the compile unit (not extern)
421 {}* ;; Reference to the global variable
426 <p>These descriptors provide debug information about globals variables. The
427 provide details such as name, type and where the variable is defined. All
428 global variables are collected inside the named metadata
429 <tt>!llvm.dbg.cu</tt>.</p>
433 <!-- ======================================================================= -->
435 <a name="format_subprograms">Subprogram descriptors</a>
440 <div class="doc_code">
443 i32, ;; Tag = 46 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
444 ;; (DW_TAG_subprogram)
445 i32, ;; Unused field.
446 metadata, ;; Reference to context descriptor
448 metadata, ;; Display name (fully qualified C++ name)
449 metadata, ;; MIPS linkage name (for C++)
450 metadata, ;; Reference to file where defined
451 i32, ;; Line number where defined
452 metadata, ;; Reference to type descriptor
453 i1, ;; True if the global is local to compile unit (static)
454 i1, ;; True if the global is defined in the compile unit (not extern)
455 i32, ;; Virtuality, e.g. dwarf::DW_VIRTUALITY__virtual
456 i32, ;; Index into a virtual function
457 metadata, ;; indicates which base type contains the vtable pointer for the
459 i32, ;; Flags - Artifical, Private, Protected, Explicit, Prototyped.
461 Function *,;; Pointer to LLVM function
462 metadata, ;; Lists function template parameters
463 metadata ;; Function declaration descriptor
464 metadata ;; List of function variables
469 <p>These descriptors provide debug information about functions, methods and
470 subprograms. They provide details such as name, return types and the source
471 location where the subprogram is defined.
476 <!-- ======================================================================= -->
478 <a name="format_blocks">Block descriptors</a>
483 <div class="doc_code">
486 i32, ;; Tag = 11 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_lexical_block)
487 metadata,;; Reference to context descriptor
489 i32, ;; Column number
490 metadata,;; Reference to source file
491 i32 ;; Unique ID to identify blocks from a template function
496 <p>This descriptor provides debug information about nested blocks within a
497 subprogram. The line number and column numbers are used to dinstinguish
498 two lexical blocks at same depth. </p>
500 <div class="doc_code">
503 i32, ;; Tag = 11 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_lexical_block)
504 metadata ;; Reference to the scope we're annotating with a file change
505 metadata,;; Reference to the file the scope is enclosed in.
510 <p>This descriptor provides a wrapper around a lexical scope to handle file
511 changes in the middle of a lexical block.</p>
515 <!-- ======================================================================= -->
517 <a name="format_basic_type">Basic type descriptors</a>
522 <div class="doc_code">
525 i32, ;; Tag = 36 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
526 ;; (DW_TAG_base_type)
527 metadata, ;; Reference to context
528 metadata, ;; Name (may be "" for anonymous types)
529 metadata, ;; Reference to file where defined (may be NULL)
530 i32, ;; Line number where defined (may be 0)
532 i64, ;; Alignment in bits
533 i64, ;; Offset in bits
535 i32 ;; DWARF type encoding
540 <p>These descriptors define primitive types used in the code. Example int, bool
541 and float. The context provides the scope of the type, which is usually the
542 top level. Since basic types are not usually user defined the context
543 and line number can be left as NULL and 0. The size, alignment and offset
544 are expressed in bits and can be 64 bit values. The alignment is used to
545 round the offset when embedded in a
546 <a href="#format_composite_type">composite type</a> (example to keep float
547 doubles on 64 bit boundaries.) The offset is the bit offset if embedded in
548 a <a href="#format_composite_type">composite type</a>.</p>
550 <p>The type encoding provides the details of the type. The values are typically
551 one of the following:</p>
553 <div class="doc_code">
559 DW_ATE_signed_char = 6
561 DW_ATE_unsigned_char = 8
567 <!-- ======================================================================= -->
569 <a name="format_derived_type">Derived type descriptors</a>
574 <div class="doc_code">
577 i32, ;; Tag (see below)
578 metadata, ;; Reference to context
579 metadata, ;; Name (may be "" for anonymous types)
580 metadata, ;; Reference to file where defined (may be NULL)
581 i32, ;; Line number where defined (may be 0)
583 i64, ;; Alignment in bits
584 i64, ;; Offset in bits
585 i32, ;; Flags to encode attributes, e.g. private
586 metadata, ;; Reference to type derived from
587 metadata, ;; (optional) Name of the Objective C property associated with
588 ;; Objective-C an ivar
589 metadata, ;; (optional) Name of the Objective C property getter selector.
590 metadata, ;; (optional) Name of the Objective C property setter selector.
591 i32 ;; (optional) Objective C property attributes.
596 <p>These descriptors are used to define types derived from other types. The
597 value of the tag varies depending on the meaning. The following are possible
600 <div class="doc_code">
602 DW_TAG_formal_parameter = 5
604 DW_TAG_pointer_type = 15
605 DW_TAG_reference_type = 16
607 DW_TAG_const_type = 38
608 DW_TAG_volatile_type = 53
609 DW_TAG_restrict_type = 55
613 <p><tt>DW_TAG_member</tt> is used to define a member of
614 a <a href="#format_composite_type">composite type</a>
615 or <a href="#format_subprograms">subprogram</a>. The type of the member is
616 the <a href="#format_derived_type">derived
617 type</a>. <tt>DW_TAG_formal_parameter</tt> is used to define a member which
618 is a formal argument of a subprogram.</p>
620 <p><tt>DW_TAG_typedef</tt> is used to provide a name for the derived type.</p>
622 <p><tt>DW_TAG_pointer_type</tt>, <tt>DW_TAG_reference_type</tt>,
623 <tt>DW_TAG_const_type</tt>, <tt>DW_TAG_volatile_type</tt> and
624 <tt>DW_TAG_restrict_type</tt> are used to qualify
625 the <a href="#format_derived_type">derived type</a>. </p>
627 <p><a href="#format_derived_type">Derived type</a> location can be determined
628 from the context and line number. The size, alignment and offset are
629 expressed in bits and can be 64 bit values. The alignment is used to round
630 the offset when embedded in a <a href="#format_composite_type">composite
631 type</a> (example to keep float doubles on 64 bit boundaries.) The offset is
632 the bit offset if embedded in a <a href="#format_composite_type">composite
635 <p>Note that the <tt>void *</tt> type is expressed as a type derived from NULL.
640 <!-- ======================================================================= -->
642 <a name="format_composite_type">Composite type descriptors</a>
647 <div class="doc_code">
650 i32, ;; Tag (see below)
651 metadata, ;; Reference to context
652 metadata, ;; Name (may be "" for anonymous types)
653 metadata, ;; Reference to file where defined (may be NULL)
654 i32, ;; Line number where defined (may be 0)
656 i64, ;; Alignment in bits
657 i64, ;; Offset in bits
659 metadata, ;; Reference to type derived from
660 metadata, ;; Reference to array of member descriptors
661 i32 ;; Runtime languages
666 <p>These descriptors are used to define types that are composed of 0 or more
667 elements. The value of the tag varies depending on the meaning. The following
668 are possible tag values:</p>
670 <div class="doc_code">
672 DW_TAG_array_type = 1
673 DW_TAG_enumeration_type = 4
674 DW_TAG_structure_type = 19
675 DW_TAG_union_type = 23
676 DW_TAG_vector_type = 259
677 DW_TAG_subroutine_type = 21
678 DW_TAG_inheritance = 28
682 <p>The vector flag indicates that an array type is a native packed vector.</p>
684 <p>The members of array types (tag = <tt>DW_TAG_array_type</tt>) or vector types
685 (tag = <tt>DW_TAG_vector_type</tt>) are <a href="#format_subrange">subrange
686 descriptors</a>, each representing the range of subscripts at that level of
689 <p>The members of enumeration types (tag = <tt>DW_TAG_enumeration_type</tt>) are
690 <a href="#format_enumeration">enumerator descriptors</a>, each representing
691 the definition of enumeration value for the set. All enumeration type
692 descriptors are collected inside the named metadata
693 <tt>!llvm.dbg.cu</tt>.</p>
695 <p>The members of structure (tag = <tt>DW_TAG_structure_type</tt>) or union (tag
696 = <tt>DW_TAG_union_type</tt>) types are any one of
697 the <a href="#format_basic_type">basic</a>,
698 <a href="#format_derived_type">derived</a>
699 or <a href="#format_composite_type">composite</a> type descriptors, each
700 representing a field member of the structure or union.</p>
702 <p>For C++ classes (tag = <tt>DW_TAG_structure_type</tt>), member descriptors
703 provide information about base classes, static members and member
704 functions. If a member is a <a href="#format_derived_type">derived type
705 descriptor</a> and has a tag of <tt>DW_TAG_inheritance</tt>, then the type
706 represents a base class. If the member of is
707 a <a href="#format_global_variables">global variable descriptor</a> then it
708 represents a static member. And, if the member is
709 a <a href="#format_subprograms">subprogram descriptor</a> then it represents
710 a member function. For static members and member
711 functions, <tt>getName()</tt> returns the members link or the C++ mangled
712 name. <tt>getDisplayName()</tt> the simplied version of the name.</p>
714 <p>The first member of subroutine (tag = <tt>DW_TAG_subroutine_type</tt>) type
715 elements is the return type for the subroutine. The remaining elements are
716 the formal arguments to the subroutine.</p>
718 <p><a href="#format_composite_type">Composite type</a> location can be
719 determined from the context and line number. The size, alignment and
720 offset are expressed in bits and can be 64 bit values. The alignment is used
721 to round the offset when embedded in
722 a <a href="#format_composite_type">composite type</a> (as an example, to keep
723 float doubles on 64 bit boundaries.) The offset is the bit offset if embedded
724 in a <a href="#format_composite_type">composite type</a>.</p>
728 <!-- ======================================================================= -->
730 <a name="format_subrange">Subrange descriptors</a>
735 <div class="doc_code">
738 i32, ;; Tag = 33 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_subrange_type)
745 <p>These descriptors are used to define ranges of array subscripts for an array
746 <a href="#format_composite_type">composite type</a>. The low value defines
747 the lower bounds typically zero for C/C++. The high value is the upper
748 bounds. Values are 64 bit. High - low + 1 is the size of the array. If low
749 > high the array bounds are not included in generated debugging information.
754 <!-- ======================================================================= -->
756 <a name="format_enumeration">Enumerator descriptors</a>
761 <div class="doc_code">
764 i32, ;; Tag = 40 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
765 ;; (DW_TAG_enumerator)
772 <p>These descriptors are used to define members of an
773 enumeration <a href="#format_composite_type">composite type</a>, it
774 associates the name to the value.</p>
778 <!-- ======================================================================= -->
780 <a name="format_variables">Local variables</a>
785 <div class="doc_code">
788 i32, ;; Tag (see below)
791 metadata, ;; Reference to file where defined
792 i32, ;; 24 bit - Line number where defined
793 ;; 8 bit - Argument number. 1 indicates 1st argument.
794 metadata, ;; Type descriptor
796 metadata ;; (optional) Reference to inline location
801 <p>These descriptors are used to define variables local to a sub program. The
802 value of the tag depends on the usage of the variable:</p>
804 <div class="doc_code">
806 DW_TAG_auto_variable = 256
807 DW_TAG_arg_variable = 257
808 DW_TAG_return_variable = 258
812 <p>An auto variable is any variable declared in the body of the function. An
813 argument variable is any variable that appears as a formal argument to the
814 function. A return variable is used to track the result of a function and
815 has no source correspondent.</p>
817 <p>The context is either the subprogram or block where the variable is defined.
818 Name the source variable name. Context and line indicate where the
819 variable was defined. Type descriptor defines the declared type of the
826 <!-- ======================================================================= -->
828 <a name="format_common_intrinsics">Debugger intrinsic functions</a>
833 <p>LLVM uses several intrinsic functions (name prefixed with "llvm.dbg") to
834 provide debug information at various points in generated code.</p>
836 <!-- ======================================================================= -->
838 <a name="format_common_declare">llvm.dbg.declare</a>
843 void %<a href="#format_common_declare">llvm.dbg.declare</a>(metadata, metadata)
846 <p>This intrinsic provides information about a local element (e.g., variable). The
847 first argument is metadata holding the alloca for the variable. The
848 second argument is metadata containing a description of the variable.</p>
851 <!-- ======================================================================= -->
853 <a name="format_common_value">llvm.dbg.value</a>
858 void %<a href="#format_common_value">llvm.dbg.value</a>(metadata, i64, metadata)
861 <p>This intrinsic provides information when a user source variable is set to a
862 new value. The first argument is the new value (wrapped as metadata). The
863 second argument is the offset in the user source variable where the new value
864 is written. The third argument is metadata containing a description of the
865 user source variable.</p>
870 <!-- ======================================================================= -->
872 <a name="format_common_lifetime">Object lifetimes and scoping</a>
876 <p>In many languages, the local variables in functions can have their lifetimes
877 or scopes limited to a subset of a function. In the C family of languages,
878 for example, variables are only live (readable and writable) within the
879 source block that they are defined in. In functional languages, values are
880 only readable after they have been defined. Though this is a very obvious
881 concept, it is non-trivial to model in LLVM, because it has no notion of
882 scoping in this sense, and does not want to be tied to a language's scoping
885 <p>In order to handle this, the LLVM debug format uses the metadata attached to
886 llvm instructions to encode line number and scoping information. Consider
887 the following C fragment, for example:</p>
889 <div class="doc_code">
903 <p>Compiled to LLVM, this function would be represented like this:</p>
905 <div class="doc_code">
907 define void @foo() nounwind ssp {
909 %X = alloca i32, align 4 ; <i32*> [#uses=4]
910 %Y = alloca i32, align 4 ; <i32*> [#uses=4]
911 %Z = alloca i32, align 4 ; <i32*> [#uses=3]
912 %0 = bitcast i32* %X to {}* ; <{}*> [#uses=1]
913 call void @llvm.dbg.declare(metadata !{i32 * %X}, metadata !0), !dbg !7
914 store i32 21, i32* %X, !dbg !8
915 %1 = bitcast i32* %Y to {}* ; <{}*> [#uses=1]
916 call void @llvm.dbg.declare(metadata !{i32 * %Y}, metadata !9), !dbg !10
917 store i32 22, i32* %Y, !dbg !11
918 %2 = bitcast i32* %Z to {}* ; <{}*> [#uses=1]
919 call void @llvm.dbg.declare(metadata !{i32 * %Z}, metadata !12), !dbg !14
920 store i32 23, i32* %Z, !dbg !15
921 %tmp = load i32* %X, !dbg !16 ; <i32> [#uses=1]
922 %tmp1 = load i32* %Y, !dbg !16 ; <i32> [#uses=1]
923 %add = add nsw i32 %tmp, %tmp1, !dbg !16 ; <i32> [#uses=1]
924 store i32 %add, i32* %Z, !dbg !16
925 %tmp2 = load i32* %Y, !dbg !17 ; <i32> [#uses=1]
926 store i32 %tmp2, i32* %X, !dbg !17
930 declare void @llvm.dbg.declare(metadata, metadata) nounwind readnone
932 !0 = metadata !{i32 459008, metadata !1, metadata !"X",
933 metadata !3, i32 2, metadata !6}; [ DW_TAG_auto_variable ]
934 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
935 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo", metadata !"foo",
936 metadata !"foo", metadata !3, i32 1, metadata !4,
937 i1 false, i1 true}; [DW_TAG_subprogram ]
938 !3 = metadata !{i32 458769, i32 0, i32 12, metadata !"foo.c",
939 metadata !"/private/tmp", metadata !"clang 1.1", i1 true,
940 i1 false, metadata !"", i32 0}; [DW_TAG_compile_unit ]
941 !4 = metadata !{i32 458773, metadata !3, metadata !"", null, i32 0, i64 0, i64 0,
942 i64 0, i32 0, null, metadata !5, i32 0}; [DW_TAG_subroutine_type ]
943 !5 = metadata !{null}
944 !6 = metadata !{i32 458788, metadata !3, metadata !"int", metadata !3, i32 0,
945 i64 32, i64 32, i64 0, i32 0, i32 5}; [DW_TAG_base_type ]
946 !7 = metadata !{i32 2, i32 7, metadata !1, null}
947 !8 = metadata !{i32 2, i32 3, metadata !1, null}
948 !9 = metadata !{i32 459008, metadata !1, metadata !"Y", metadata !3, i32 3,
949 metadata !6}; [ DW_TAG_auto_variable ]
950 !10 = metadata !{i32 3, i32 7, metadata !1, null}
951 !11 = metadata !{i32 3, i32 3, metadata !1, null}
952 !12 = metadata !{i32 459008, metadata !13, metadata !"Z", metadata !3, i32 5,
953 metadata !6}; [ DW_TAG_auto_variable ]
954 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
955 !14 = metadata !{i32 5, i32 9, metadata !13, null}
956 !15 = metadata !{i32 5, i32 5, metadata !13, null}
957 !16 = metadata !{i32 6, i32 5, metadata !13, null}
958 !17 = metadata !{i32 8, i32 3, metadata !1, null}
959 !18 = metadata !{i32 9, i32 1, metadata !2, null}
963 <p>This example illustrates a few important details about LLVM debugging
964 information. In particular, it shows how the <tt>llvm.dbg.declare</tt>
965 intrinsic and location information, which are attached to an instruction,
966 are applied together to allow a debugger to analyze the relationship between
967 statements, variable definitions, and the code used to implement the
970 <div class="doc_code">
972 call void @llvm.dbg.declare(metadata, metadata !0), !dbg !7
976 <p>The first intrinsic
977 <tt>%<a href="#format_common_declare">llvm.dbg.declare</a></tt>
978 encodes debugging information for the variable <tt>X</tt>. The metadata
979 <tt>!dbg !7</tt> attached to the intrinsic provides scope information for the
980 variable <tt>X</tt>.</p>
982 <div class="doc_code">
984 !7 = metadata !{i32 2, i32 7, metadata !1, null}
985 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
986 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo",
987 metadata !"foo", metadata !"foo", metadata !3, i32 1,
988 metadata !4, i1 false, i1 true}; [DW_TAG_subprogram ]
992 <p>Here <tt>!7</tt> is metadata providing location information. It has four
993 fields: line number, column number, scope, and original scope. The original
994 scope represents inline location if this instruction is inlined inside a
995 caller, and is null otherwise. In this example, scope is encoded by
996 <tt>!1</tt>. <tt>!1</tt> represents a lexical block inside the scope
997 <tt>!2</tt>, where <tt>!2</tt> is a
998 <a href="#format_subprograms">subprogram descriptor</a>. This way the
999 location information attached to the intrinsics indicates that the
1000 variable <tt>X</tt> is declared at line number 2 at a function level scope in
1001 function <tt>foo</tt>.</p>
1003 <p>Now lets take another example.</p>
1005 <div class="doc_code">
1007 call void @llvm.dbg.declare(metadata, metadata !12), !dbg !14
1011 <p>The second intrinsic
1012 <tt>%<a href="#format_common_declare">llvm.dbg.declare</a></tt>
1013 encodes debugging information for variable <tt>Z</tt>. The metadata
1014 <tt>!dbg !14</tt> attached to the intrinsic provides scope information for
1015 the variable <tt>Z</tt>.</p>
1017 <div class="doc_code">
1019 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
1020 !14 = metadata !{i32 5, i32 9, metadata !13, null}
1024 <p>Here <tt>!14</tt> indicates that <tt>Z</tt> is declared at line number 5 and
1025 column number 9 inside of lexical scope <tt>!13</tt>. The lexical scope
1026 itself resides inside of lexical scope <tt>!1</tt> described above.</p>
1028 <p>The scope information attached with each instruction provides a
1029 straightforward way to find instructions covered by a scope.</p>
1035 <!-- *********************************************************************** -->
1037 <a name="ccxx_frontend">C/C++ front-end specific debug information</a>
1039 <!-- *********************************************************************** -->
1043 <p>The C and C++ front-ends represent information about the program in a format
1044 that is effectively identical
1045 to <a href="http://www.eagercon.com/dwarf/dwarf3std.htm">DWARF 3.0</a> in
1046 terms of information content. This allows code generators to trivially
1047 support native debuggers by generating standard dwarf information, and
1048 contains enough information for non-dwarf targets to translate it as
1051 <p>This section describes the forms used to represent C and C++ programs. Other
1052 languages could pattern themselves after this (which itself is tuned to
1053 representing programs in the same way that DWARF 3 does), or they could
1054 choose to provide completely different forms if they don't fit into the DWARF
1055 model. As support for debugging information gets added to the various LLVM
1056 source-language front-ends, the information used should be documented
1059 <p>The following sections provide examples of various C/C++ constructs and the
1060 debug information that would best describe those constructs.</p>
1062 <!-- ======================================================================= -->
1064 <a name="ccxx_compile_units">C/C++ source file information</a>
1069 <p>Given the source files <tt>MySource.cpp</tt> and <tt>MyHeader.h</tt> located
1070 in the directory <tt>/Users/mine/sources</tt>, the following code:</p>
1072 <div class="doc_code">
1074 #include "MyHeader.h"
1076 int main(int argc, char *argv[]) {
1082 <p>a C/C++ front-end would generate the following descriptors:</p>
1084 <div class="doc_code">
1088 ;; Define the compile unit for the main source file "/Users/mine/sources/MySource.cpp".
1093 i32 4, ;; Language Id
1094 metadata !"MySource.cpp",
1095 metadata !"/Users/mine/sources",
1096 metadata !"4.2.1 (Based on Apple Inc. build 5649) (LLVM build 00)",
1097 i1 true, ;; Main Compile Unit
1098 i1 false, ;; Optimized compile unit
1099 metadata !"", ;; Compiler flags
1100 i32 0} ;; Runtime version
1103 ;; Define the file for the file "/Users/mine/sources/MySource.cpp".
1107 metadata !"MySource.cpp",
1108 metadata !"/Users/mine/sources",
1109 metadata !2 ;; Compile unit
1113 ;; Define the file for the file "/Users/mine/sources/Myheader.h"
1117 metadata !"Myheader.h"
1118 metadata !"/Users/mine/sources",
1119 metadata !2 ;; Compile unit
1126 <p>llvm::Instruction provides easy access to metadata attached with an
1127 instruction. One can extract line number information encoded in LLVM IR
1128 using <tt>Instruction::getMetadata()</tt> and
1129 <tt>DILocation::getLineNumber()</tt>.
1131 if (MDNode *N = I->getMetadata("dbg")) { // Here I is an LLVM instruction
1132 DILocation Loc(N); // DILocation is in DebugInfo.h
1133 unsigned Line = Loc.getLineNumber();
1134 StringRef File = Loc.getFilename();
1135 StringRef Dir = Loc.getDirectory();
1140 <!-- ======================================================================= -->
1142 <a name="ccxx_global_variable">C/C++ global variable information</a>
1147 <p>Given an integer global variable declared as follows:</p>
1149 <div class="doc_code">
1155 <p>a C/C++ front-end would generate the following descriptors:</p>
1157 <div class="doc_code">
1160 ;; Define the global itself.
1162 %MyGlobal = global int 100
1165 ;; List of debug info of globals
1167 !llvm.dbg.cu = !{!0}
1169 ;; Define the compile unit.
1174 metadata !"foo.cpp", ;; File
1175 metadata !"/Volumes/Data/tmp", ;; Directory
1176 metadata !"clang version 3.1 ", ;; Producer
1177 i1 true, ;; Deprecated field
1178 i1 false, ;; "isOptimized"?
1179 metadata !"", ;; Flags
1180 i32 0, ;; Runtime Version
1181 metadata !1, ;; Enum Types
1182 metadata !1, ;; Retained Types
1183 metadata !1, ;; Subprograms
1184 metadata !3 ;; Global Variables
1185 } ; [ DW_TAG_compile_unit ]
1187 ;; The Array of Global Variables
1197 ;; Define the global variable itself.
1203 metadata !"MyGlobal", ;; Name
1204 metadata !"MyGlobal", ;; Display Name
1205 metadata !"", ;; Linkage Name
1206 metadata !6, ;; File
1208 metadata !7, ;; Type
1209 i32 0, ;; IsLocalToUnit
1210 i32 1, ;; IsDefinition
1211 i32* @MyGlobal ;; LLVM-IR Value
1212 } ; [ DW_TAG_variable ]
1219 metadata !"foo.cpp", ;; File
1220 metadata !"/Volumes/Data/tmp", ;; Directory
1222 } ; [ DW_TAG_file_type ]
1230 metadata !"int", ;; Name
1233 i64 32, ;; Size in Bits
1234 i64 32, ;; Align in Bits
1238 } ; [ DW_TAG_base_type ]
1245 <!-- ======================================================================= -->
1247 <a name="ccxx_subprogram">C/C++ function information</a>
1252 <p>Given a function declared as follows:</p>
1254 <div class="doc_code">
1256 int main(int argc, char *argv[]) {
1262 <p>a C/C++ front-end would generate the following descriptors:</p>
1264 <div class="doc_code">
1267 ;; Define the anchor for subprograms. Note that the second field of the
1268 ;; anchor is 46, which is the same as the tag for subprograms
1269 ;; (46 = DW_TAG_subprogram.)
1274 metadata !1, ;; Context
1275 metadata !"main", ;; Name
1276 metadata !"main", ;; Display name
1277 metadata !"main", ;; Linkage name
1278 metadata !1, ;; File
1279 i32 1, ;; Line number
1280 metadata !4, ;; Type
1281 i1 false, ;; Is local
1282 i1 true, ;; Is definition
1283 i32 0, ;; Virtuality attribute, e.g. pure virtual function
1284 i32 0, ;; Index into virtual table for C++ methods
1285 i32 0, ;; Type that holds virtual table.
1287 i1 false, ;; True if this function is optimized
1288 Function *, ;; Pointer to llvm::Function
1289 null ;; Function template parameters
1292 ;; Define the subprogram itself.
1294 define i32 @main(i32 %argc, i8** %argv) {
1302 <!-- ======================================================================= -->
1304 <a name="ccxx_basic_types">C/C++ basic types</a>
1309 <p>The following are the basic type descriptors for C/C++ core types:</p>
1311 <!-- ======================================================================= -->
1313 <a name="ccxx_basic_type_bool">bool</a>
1318 <div class="doc_code">
1322 metadata !1, ;; Context
1323 metadata !"bool", ;; Name
1324 metadata !1, ;; File
1325 i32 0, ;; Line number
1326 i64 8, ;; Size in Bits
1327 i64 8, ;; Align in Bits
1328 i64 0, ;; Offset in Bits
1337 <!-- ======================================================================= -->
1339 <a name="ccxx_basic_char">char</a>
1344 <div class="doc_code">
1348 metadata !1, ;; Context
1349 metadata !"char", ;; Name
1350 metadata !1, ;; File
1351 i32 0, ;; Line number
1352 i64 8, ;; Size in Bits
1353 i64 8, ;; Align in Bits
1354 i64 0, ;; Offset in Bits
1363 <!-- ======================================================================= -->
1365 <a name="ccxx_basic_unsigned_char">unsigned char</a>
1370 <div class="doc_code">
1374 metadata !1, ;; Context
1375 metadata !"unsigned char",
1376 metadata !1, ;; File
1377 i32 0, ;; Line number
1378 i64 8, ;; Size in Bits
1379 i64 8, ;; Align in Bits
1380 i64 0, ;; Offset in Bits
1389 <!-- ======================================================================= -->
1391 <a name="ccxx_basic_short">short</a>
1396 <div class="doc_code">
1400 metadata !1, ;; Context
1401 metadata !"short int",
1402 metadata !1, ;; File
1403 i32 0, ;; Line number
1404 i64 16, ;; Size in Bits
1405 i64 16, ;; Align in Bits
1406 i64 0, ;; Offset in Bits
1415 <!-- ======================================================================= -->
1417 <a name="ccxx_basic_unsigned_short">unsigned short</a>
1422 <div class="doc_code">
1426 metadata !1, ;; Context
1427 metadata !"short unsigned int",
1428 metadata !1, ;; File
1429 i32 0, ;; Line number
1430 i64 16, ;; Size in Bits
1431 i64 16, ;; Align in Bits
1432 i64 0, ;; Offset in Bits
1441 <!-- ======================================================================= -->
1443 <a name="ccxx_basic_int">int</a>
1448 <div class="doc_code">
1452 metadata !1, ;; Context
1453 metadata !"int", ;; Name
1454 metadata !1, ;; File
1455 i32 0, ;; Line number
1456 i64 32, ;; Size in Bits
1457 i64 32, ;; Align in Bits
1458 i64 0, ;; Offset in Bits
1466 <!-- ======================================================================= -->
1468 <a name="ccxx_basic_unsigned_int">unsigned int</a>
1473 <div class="doc_code">
1477 metadata !1, ;; Context
1478 metadata !"unsigned int",
1479 metadata !1, ;; File
1480 i32 0, ;; Line number
1481 i64 32, ;; Size in Bits
1482 i64 32, ;; Align in Bits
1483 i64 0, ;; Offset in Bits
1492 <!-- ======================================================================= -->
1494 <a name="ccxx_basic_long_long">long long</a>
1499 <div class="doc_code">
1503 metadata !1, ;; Context
1504 metadata !"long long int",
1505 metadata !1, ;; File
1506 i32 0, ;; Line number
1507 i64 64, ;; Size in Bits
1508 i64 64, ;; Align in Bits
1509 i64 0, ;; Offset in Bits
1518 <!-- ======================================================================= -->
1520 <a name="ccxx_basic_unsigned_long_long">unsigned long long</a>
1525 <div class="doc_code">
1529 metadata !1, ;; Context
1530 metadata !"long long unsigned int",
1531 metadata !1, ;; File
1532 i32 0, ;; Line number
1533 i64 64, ;; Size in Bits
1534 i64 64, ;; Align in Bits
1535 i64 0, ;; Offset in Bits
1544 <!-- ======================================================================= -->
1546 <a name="ccxx_basic_float">float</a>
1551 <div class="doc_code">
1555 metadata !1, ;; Context
1557 metadata !1, ;; File
1558 i32 0, ;; Line number
1559 i64 32, ;; Size in Bits
1560 i64 32, ;; Align in Bits
1561 i64 0, ;; Offset in Bits
1570 <!-- ======================================================================= -->
1572 <a name="ccxx_basic_double">double</a>
1577 <div class="doc_code">
1581 metadata !1, ;; Context
1582 metadata !"double",;; Name
1583 metadata !1, ;; File
1584 i32 0, ;; Line number
1585 i64 64, ;; Size in Bits
1586 i64 64, ;; Align in Bits
1587 i64 0, ;; Offset in Bits
1598 <!-- ======================================================================= -->
1600 <a name="ccxx_derived_types">C/C++ derived types</a>
1605 <p>Given the following as an example of C/C++ derived type:</p>
1607 <div class="doc_code">
1609 typedef const int *IntPtr;
1613 <p>a C/C++ front-end would generate the following descriptors:</p>
1615 <div class="doc_code">
1618 ;; Define the typedef "IntPtr".
1622 metadata !1, ;; Context
1623 metadata !"IntPtr", ;; Name
1624 metadata !3, ;; File
1625 i32 0, ;; Line number
1626 i64 0, ;; Size in bits
1627 i64 0, ;; Align in bits
1628 i64 0, ;; Offset in bits
1630 metadata !4 ;; Derived From type
1634 ;; Define the pointer type.
1638 metadata !1, ;; Context
1639 metadata !"", ;; Name
1640 metadata !1, ;; File
1641 i32 0, ;; Line number
1642 i64 64, ;; Size in bits
1643 i64 64, ;; Align in bits
1644 i64 0, ;; Offset in bits
1646 metadata !5 ;; Derived From type
1649 ;; Define the const type.
1653 metadata !1, ;; Context
1654 metadata !"", ;; Name
1655 metadata !1, ;; File
1656 i32 0, ;; Line number
1657 i64 32, ;; Size in bits
1658 i64 32, ;; Align in bits
1659 i64 0, ;; Offset in bits
1661 metadata !6 ;; Derived From type
1664 ;; Define the int type.
1668 metadata !1, ;; Context
1669 metadata !"int", ;; Name
1670 metadata !1, ;; File
1671 i32 0, ;; Line number
1672 i64 32, ;; Size in bits
1673 i64 32, ;; Align in bits
1674 i64 0, ;; Offset in bits
1683 <!-- ======================================================================= -->
1685 <a name="ccxx_composite_types">C/C++ struct/union types</a>
1690 <p>Given the following as an example of C/C++ struct type:</p>
1692 <div class="doc_code">
1702 <p>a C/C++ front-end would generate the following descriptors:</p>
1704 <div class="doc_code">
1707 ;; Define basic type for unsigned int.
1711 metadata !1, ;; Context
1712 metadata !"unsigned int",
1713 metadata !1, ;; File
1714 i32 0, ;; Line number
1715 i64 32, ;; Size in Bits
1716 i64 32, ;; Align in Bits
1717 i64 0, ;; Offset in Bits
1722 ;; Define composite type for struct Color.
1726 metadata !1, ;; Context
1727 metadata !"Color", ;; Name
1728 metadata !1, ;; Compile unit
1729 i32 1, ;; Line number
1730 i64 96, ;; Size in bits
1731 i64 32, ;; Align in bits
1732 i64 0, ;; Offset in bits
1734 null, ;; Derived From
1735 metadata !3, ;; Elements
1736 i32 0 ;; Runtime Language
1740 ;; Define the Red field.
1744 metadata !1, ;; Context
1745 metadata !"Red", ;; Name
1746 metadata !1, ;; File
1747 i32 2, ;; Line number
1748 i64 32, ;; Size in bits
1749 i64 32, ;; Align in bits
1750 i64 0, ;; Offset in bits
1752 metadata !5 ;; Derived From type
1756 ;; Define the Green field.
1760 metadata !1, ;; Context
1761 metadata !"Green", ;; Name
1762 metadata !1, ;; File
1763 i32 3, ;; Line number
1764 i64 32, ;; Size in bits
1765 i64 32, ;; Align in bits
1766 i64 32, ;; Offset in bits
1768 metadata !5 ;; Derived From type
1772 ;; Define the Blue field.
1776 metadata !1, ;; Context
1777 metadata !"Blue", ;; Name
1778 metadata !1, ;; File
1779 i32 4, ;; Line number
1780 i64 32, ;; Size in bits
1781 i64 32, ;; Align in bits
1782 i64 64, ;; Offset in bits
1784 metadata !5 ;; Derived From type
1788 ;; Define the array of fields used by the composite type Color.
1790 !3 = metadata !{metadata !4, metadata !6, metadata !7}
1796 <!-- ======================================================================= -->
1798 <a name="ccxx_enumeration_types">C/C++ enumeration types</a>
1803 <p>Given the following as an example of C/C++ enumeration type:</p>
1805 <div class="doc_code">
1815 <p>a C/C++ front-end would generate the following descriptors:</p>
1817 <div class="doc_code">
1820 ;; Define composite type for enum Trees
1824 metadata !1, ;; Context
1825 metadata !"Trees", ;; Name
1826 metadata !1, ;; File
1827 i32 1, ;; Line number
1828 i64 32, ;; Size in bits
1829 i64 32, ;; Align in bits
1830 i64 0, ;; Offset in bits
1832 null, ;; Derived From type
1833 metadata !3, ;; Elements
1834 i32 0 ;; Runtime language
1838 ;; Define the array of enumerators used by composite type Trees.
1840 !3 = metadata !{metadata !4, metadata !5, metadata !6}
1843 ;; Define Spruce enumerator.
1845 !4 = metadata !{i32 524328, metadata !"Spruce", i64 100}
1848 ;; Define Oak enumerator.
1850 !5 = metadata !{i32 524328, metadata !"Oak", i64 200}
1853 ;; Define Maple enumerator.
1855 !6 = metadata !{i32 524328, metadata !"Maple", i64 300}
1865 <!-- *********************************************************************** -->
1867 <a name="llvmdwarfextension">Debugging information format</a>
1869 <!-- *********************************************************************** -->
1871 <!-- ======================================================================= -->
1873 <a name="objcproperty">Debugging Information Extension for Objective C Properties</a>
1876 <!-- *********************************************************************** -->
1878 <a name="objcpropertyintroduction">Introduction</a>
1880 <!-- *********************************************************************** -->
1883 <p>Objective C provides a simpler way to declare and define accessor methods
1884 using declared properties. The language provides features to declare a
1885 property and to let compiler synthesize accessor methods.
1888 <p>The debugger lets developer inspect Objective C interfaces and their
1889 instance variables and class variables. However, the debugger does not know
1890 anything about the properties defined in Objective C interfaces. The debugger
1891 consumes information generated by compiler in DWARF format. The format does
1892 not support encoding of Objective C properties. This proposal describes DWARF
1893 extensions to encode Objective C properties, which the debugger can use to let
1894 developers inspect Objective C properties.
1900 <!-- *********************************************************************** -->
1902 <a name="objcpropertyproposal">Proposal</a>
1904 <!-- *********************************************************************** -->
1907 <p>Objective C properties exist separately from class members. A property
1908 can be defined only by "setter" and "getter" selectors, and
1909 be calculated anew on each access. Or a property can just be a direct access
1910 to some declared ivar. Finally it can have an ivar "automatically
1911 synthesized" for it by the compiler, in which case the property can be
1912 referred to in user code directly using the standard C dereference syntax as
1913 well as through the property "dot" syntax, but there is no entry in
1914 the @interface declaration corresponding to this ivar.
1917 To facilitate debugging, these properties we will add a new DWARF TAG into the
1918 DW_TAG_structure_type definition for the class to hold the description of a
1919 given property, and a set of DWARF attributes that provide said description.
1920 The property tag will also contain the name and declared type of the property.
1923 If there is a related ivar, there will also be a DWARF property attribute placed
1924 in the DW_TAG_member DIE for that ivar referring back to the property TAG for
1925 that property. And in the case where the compiler synthesizes the ivar directly,
1926 the compiler is expected to generate a DW_TAG_member for that ivar (with the
1927 DW_AT_artificial set to 1), whose name will be the name used to access this
1928 ivar directly in code, and with the property attribute pointing back to the
1929 property it is backing.
1932 The following examples will serve as illustration for our discussion:
1935 <div class="doc_code">
1947 @synthesize p2 = n2;
1953 This produces the following DWARF (this is a "pseudo dwarfdump" output):
1955 <div class="doc_code">
1957 0x00000100: TAG_structure_type [7] *
1958 AT_APPLE_runtime_class( 0x10 )
1960 AT_decl_file( "Objc_Property.m" )
1963 0x00000110 TAG_APPLE_property
1965 AT_type ( {0x00000150} ( int ) )
1967 0x00000120: TAG_APPLE_property
1969 AT_type ( {0x00000150} ( int ) )
1971 0x00000130: TAG_member [8]
1973 AT_APPLE_property ( {0x00000110} "p1" )
1974 AT_type( {0x00000150} ( int ) )
1975 AT_artificial ( 0x1 )
1977 0x00000140: TAG_member [8]
1979 AT_APPLE_property ( {0x00000120} "p2" )
1980 AT_type( {0x00000150} ( int ) )
1982 0x00000150: AT_type( ( int ) )
1986 <p> Note, the current convention is that the name of the ivar for an
1987 auto-synthesized property is the name of the property from which it derives with
1988 an underscore prepended, as is shown in the example.
1989 But we actually don't need to know this convention, since we are given the name
1990 of the ivar directly.
1994 Also, it is common practice in ObjC to have different property declarations in
1995 the @interface and @implementation - e.g. to provide a read-only property in
1996 the interface,and a read-write interface in the implementation. In that case,
1997 the compiler should emit whichever property declaration will be in force in the
1998 current translation unit.
2001 <p> Developers can decorate a property with attributes which are encoded using
2002 DW_AT_APPLE_property_attribute.
2005 <div class="doc_code">
2007 @property (readonly, nonatomic) int pr;
2011 Which produces a property tag:
2013 <div class="doc_code">
2015 TAG_APPLE_property [8]
2017 AT_type ( {0x00000147} (int) )
2018 AT_APPLE_property_attribute (DW_APPLE_PROPERTY_readonly, DW_APPLE_PROPERTY_nonatomic)
2022 <p> The setter and getter method names are attached to the property using
2023 DW_AT_APPLE_property_setter and DW_AT_APPLE_property_getter attributes.
2025 <div class="doc_code">
2028 @property (setter=myOwnP3Setter:) int p3;
2029 -(void)myOwnP3Setter:(int)a;
2034 -(void)myOwnP3Setter:(int)a{ }
2040 The DWARF for this would be:
2042 <div class="doc_code">
2044 0x000003bd: TAG_structure_type [7] *
2045 AT_APPLE_runtime_class( 0x10 )
2047 AT_decl_file( "Objc_Property.m" )
2050 0x000003cd TAG_APPLE_property
2052 AT_APPLE_property_setter ( "myOwnP3Setter:" )
2053 AT_type( {0x00000147} ( int ) )
2055 0x000003f3: TAG_member [8]
2057 AT_type ( {0x00000147} ( int ) )
2058 AT_APPLE_property ( {0x000003cd} )
2059 AT_artificial ( 0x1 )
2065 <!-- *********************************************************************** -->
2067 <a name="objcpropertynewtags">New DWARF Tags</a>
2069 <!-- *********************************************************************** -->
2072 <table border="1" cellspacing="0">
2080 <td>DW_TAG_APPLE_property</td>
2087 <!-- *********************************************************************** -->
2089 <a name="objcpropertynewattributes">New DWARF Attributes</a>
2091 <!-- *********************************************************************** -->
2094 <table border="1" cellspacing="0">
2104 <td>DW_AT_APPLE_property</td>
2109 <td>DW_AT_APPLE_property_getter</td>
2114 <td>DW_AT_APPLE_property_setter</td>
2119 <td>DW_AT_APPLE_property_attribute</td>
2127 <!-- *********************************************************************** -->
2129 <a name="objcpropertynewconstants">New DWARF Constants</a>
2131 <!-- *********************************************************************** -->
2134 <table border="1" cellspacing="0">
2142 <td>DW_AT_APPLE_PROPERTY_readonly</td>
2146 <td>DW_AT_APPLE_PROPERTY_readwrite</td>
2150 <td>DW_AT_APPLE_PROPERTY_assign</td>
2154 <td>DW_AT_APPLE_PROPERTY_retain</td>
2158 <td>DW_AT_APPLE_PROPERTY_copy</td>
2162 <td>DW_AT_APPLE_PROPERTY_nonatomic</td>
2170 <!-- ======================================================================= -->
2172 <a name="acceltable">Name Accelerator Tables</a>
2174 <!-- ======================================================================= -->
2176 <!-- ======================================================================= -->
2178 <a name="acceltableintroduction">Introduction</a>
2180 <!-- ======================================================================= -->
2182 <p>The .debug_pubnames and .debug_pubtypes formats are not what a debugger
2183 needs. The "pub" in the section name indicates that the entries in the
2184 table are publicly visible names only. This means no static or hidden
2185 functions show up in the .debug_pubnames. No static variables or private class
2186 variables are in the .debug_pubtypes. Many compilers add different things to
2187 these tables, so we can't rely upon the contents between gcc, icc, or clang.</p>
2189 <p>The typical query given by users tends not to match up with the contents of
2190 these tables. For example, the DWARF spec states that "In the case of the
2191 name of a function member or static data member of a C++ structure, class or
2192 union, the name presented in the .debug_pubnames section is not the simple
2193 name given by the DW_AT_name attribute of the referenced debugging information
2194 entry, but rather the fully qualified name of the data or function member."
2195 So the only names in these tables for complex C++ entries is a fully
2196 qualified name. Debugger users tend not to enter their search strings as
2197 "a::b::c(int,const Foo&) const", but rather as "c", "b::c" , or "a::b::c". So
2198 the name entered in the name table must be demangled in order to chop it up
2199 appropriately and additional names must be manually entered into the table
2200 to make it effective as a name lookup table for debuggers to use.</p>
2202 <p>All debuggers currently ignore the .debug_pubnames table as a result of
2203 its inconsistent and useless public-only name content making it a waste of
2204 space in the object file. These tables, when they are written to disk, are
2205 not sorted in any way, leaving every debugger to do its own parsing
2206 and sorting. These tables also include an inlined copy of the string values
2207 in the table itself making the tables much larger than they need to be on
2208 disk, especially for large C++ programs.</p>
2210 <p>Can't we just fix the sections by adding all of the names we need to this
2211 table? No, because that is not what the tables are defined to contain and we
2212 won't know the difference between the old bad tables and the new good tables.
2213 At best we could make our own renamed sections that contain all of the data
2216 <p>These tables are also insufficient for what a debugger like LLDB needs.
2217 LLDB uses clang for its expression parsing where LLDB acts as a PCH. LLDB is
2218 then often asked to look for type "foo" or namespace "bar", or list items in
2219 namespace "baz". Namespaces are not included in the pubnames or pubtypes
2220 tables. Since clang asks a lot of questions when it is parsing an expression,
2221 we need to be very fast when looking up names, as it happens a lot. Having new
2222 accelerator tables that are optimized for very quick lookups will benefit
2223 this type of debugging experience greatly.</p>
2225 <p>We would like to generate name lookup tables that can be mapped into
2226 memory from disk, and used as is, with little or no up-front parsing. We would
2227 also be able to control the exact content of these different tables so they
2228 contain exactly what we need. The Name Accelerator Tables were designed
2229 to fix these issues. In order to solve these issues we need to:</p>
2232 <li>Have a format that can be mapped into memory from disk and used as is</li>
2233 <li>Lookups should be very fast</li>
2234 <li>Extensible table format so these tables can be made by many producers</li>
2235 <li>Contain all of the names needed for typical lookups out of the box</li>
2236 <li>Strict rules for the contents of tables</li>
2239 <p>Table size is important and the accelerator table format should allow the
2240 reuse of strings from common string tables so the strings for the names are
2241 not duplicated. We also want to make sure the table is ready to be used as-is
2242 by simply mapping the table into memory with minimal header parsing.</p>
2244 <p>The name lookups need to be fast and optimized for the kinds of lookups
2245 that debuggers tend to do. Optimally we would like to touch as few parts of
2246 the mapped table as possible when doing a name lookup and be able to quickly
2247 find the name entry we are looking for, or discover there are no matches. In
2248 the case of debuggers we optimized for lookups that fail most of the time.</p>
2250 <p>Each table that is defined should have strict rules on exactly what is in
2251 the accelerator tables and documented so clients can rely on the content.</p>
2255 <!-- ======================================================================= -->
2257 <a name="acceltablehashes">Hash Tables</a>
2259 <!-- ======================================================================= -->
2262 <h5>Standard Hash Tables</h5>
2264 <p>Typical hash tables have a header, buckets, and each bucket points to the
2268 <div class="doc_code">
2280 <p>The BUCKETS are an array of offsets to DATA for each hash:</p>
2282 <div class="doc_code">
2285 | 0x00001000 | BUCKETS[0]
2286 | 0x00002000 | BUCKETS[1]
2287 | 0x00002200 | BUCKETS[2]
2288 | 0x000034f0 | BUCKETS[3]
2290 | 0xXXXXXXXX | BUCKETS[n_buckets]
2295 <p>So for bucket[3] in the example above, we have an offset into the table
2296 0x000034f0 which points to a chain of entries for the bucket. Each bucket
2297 must contain a next pointer, full 32 bit hash value, the string itself,
2298 and the data for the current string value.</p>
2300 <div class="doc_code">
2303 0x000034f0: | 0x00003500 | next pointer
2304 | 0x12345678 | 32 bit hash
2305 | "erase" | string value
2306 | data[n] | HashData for this bucket
2308 0x00003500: | 0x00003550 | next pointer
2309 | 0x29273623 | 32 bit hash
2310 | "dump" | string value
2311 | data[n] | HashData for this bucket
2313 0x00003550: | 0x00000000 | next pointer
2314 | 0x82638293 | 32 bit hash
2315 | "main" | string value
2316 | data[n] | HashData for this bucket
2321 <p>The problem with this layout for debuggers is that we need to optimize for
2322 the negative lookup case where the symbol we're searching for is not present.
2323 So if we were to lookup "printf" in the table above, we would make a 32 hash
2324 for "printf", it might match bucket[3]. We would need to go to the offset
2325 0x000034f0 and start looking to see if our 32 bit hash matches. To do so, we
2326 need to read the next pointer, then read the hash, compare it, and skip to
2327 the next bucket. Each time we are skipping many bytes in memory and touching
2328 new cache pages just to do the compare on the full 32 bit hash. All of these
2329 accesses then tell us that we didn't have a match.</p>
2331 <h5>Name Hash Tables</h5>
2333 <p>To solve the issues mentioned above we have structured the hash tables
2334 a bit differently: a header, buckets, an array of all unique 32 bit hash
2335 values, followed by an array of hash value data offsets, one for each hash
2336 value, then the data for all hash values:</p>
2338 <div class="doc_code">
2354 <p>The BUCKETS in the name tables are an index into the HASHES array. By
2355 making all of the full 32 bit hash values contiguous in memory, we allow
2356 ourselves to efficiently check for a match while touching as little
2357 memory as possible. Most often checking the 32 bit hash values is as far as
2358 the lookup goes. If it does match, it usually is a match with no collisions.
2359 So for a table with "n_buckets" buckets, and "n_hashes" unique 32 bit hash
2360 values, we can clarify the contents of the BUCKETS, HASHES and OFFSETS as:</p>
2362 <div class="doc_code">
2364 .-------------------------.
2365 | HEADER.magic | uint32_t
2366 | HEADER.version | uint16_t
2367 | HEADER.hash_function | uint16_t
2368 | HEADER.bucket_count | uint32_t
2369 | HEADER.hashes_count | uint32_t
2370 | HEADER.header_data_len | uint32_t
2371 | HEADER_DATA | HeaderData
2372 |-------------------------|
2373 | BUCKETS | uint32_t[n_buckets] // 32 bit hash indexes
2374 |-------------------------|
2375 | HASHES | uint32_t[n_buckets] // 32 bit hash values
2376 |-------------------------|
2377 | OFFSETS | uint32_t[n_buckets] // 32 bit offsets to hash value data
2378 |-------------------------|
2380 `-------------------------'
2384 <p>So taking the exact same data from the standard hash example above we end up
2387 <div class="doc_code">
2397 | ... | BUCKETS[n_buckets]
2399 | 0x........ | HASHES[0]
2400 | 0x........ | HASHES[1]
2401 | 0x........ | HASHES[2]
2402 | 0x........ | HASHES[3]
2403 | 0x........ | HASHES[4]
2404 | 0x........ | HASHES[5]
2405 | 0x12345678 | HASHES[6] hash for BUCKETS[3]
2406 | 0x29273623 | HASHES[7] hash for BUCKETS[3]
2407 | 0x82638293 | HASHES[8] hash for BUCKETS[3]
2408 | 0x........ | HASHES[9]
2409 | 0x........ | HASHES[10]
2410 | 0x........ | HASHES[11]
2411 | 0x........ | HASHES[12]
2412 | 0x........ | HASHES[13]
2413 | 0x........ | HASHES[n_hashes]
2415 | 0x........ | OFFSETS[0]
2416 | 0x........ | OFFSETS[1]
2417 | 0x........ | OFFSETS[2]
2418 | 0x........ | OFFSETS[3]
2419 | 0x........ | OFFSETS[4]
2420 | 0x........ | OFFSETS[5]
2421 | 0x000034f0 | OFFSETS[6] offset for BUCKETS[3]
2422 | 0x00003500 | OFFSETS[7] offset for BUCKETS[3]
2423 | 0x00003550 | OFFSETS[8] offset for BUCKETS[3]
2424 | 0x........ | OFFSETS[9]
2425 | 0x........ | OFFSETS[10]
2426 | 0x........ | OFFSETS[11]
2427 | 0x........ | OFFSETS[12]
2428 | 0x........ | OFFSETS[13]
2429 | 0x........ | OFFSETS[n_hashes]
2437 0x000034f0: | 0x00001203 | .debug_str ("erase")
2438 | 0x00000004 | A 32 bit array count - number of HashData with name "erase"
2439 | 0x........ | HashData[0]
2440 | 0x........ | HashData[1]
2441 | 0x........ | HashData[2]
2442 | 0x........ | HashData[3]
2443 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2445 0x00003500: | 0x00001203 | String offset into .debug_str ("collision")
2446 | 0x00000002 | A 32 bit array count - number of HashData with name "collision"
2447 | 0x........ | HashData[0]
2448 | 0x........ | HashData[1]
2449 | 0x00001203 | String offset into .debug_str ("dump")
2450 | 0x00000003 | A 32 bit array count - number of HashData with name "dump"
2451 | 0x........ | HashData[0]
2452 | 0x........ | HashData[1]
2453 | 0x........ | HashData[2]
2454 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2456 0x00003550: | 0x00001203 | String offset into .debug_str ("main")
2457 | 0x00000009 | A 32 bit array count - number of HashData with name "main"
2458 | 0x........ | HashData[0]
2459 | 0x........ | HashData[1]
2460 | 0x........ | HashData[2]
2461 | 0x........ | HashData[3]
2462 | 0x........ | HashData[4]
2463 | 0x........ | HashData[5]
2464 | 0x........ | HashData[6]
2465 | 0x........ | HashData[7]
2466 | 0x........ | HashData[8]
2467 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2472 <p>So we still have all of the same data, we just organize it more efficiently
2473 for debugger lookup. If we repeat the same "printf" lookup from above, we
2474 would hash "printf" and find it matches BUCKETS[3] by taking the 32 bit hash
2475 value and modulo it by n_buckets. BUCKETS[3] contains "6" which is the index
2476 into the HASHES table. We would then compare any consecutive 32 bit hashes
2477 values in the HASHES array as long as the hashes would be in BUCKETS[3]. We
2478 do this by verifying that each subsequent hash value modulo n_buckets is still
2479 3. In the case of a failed lookup we would access the memory for BUCKETS[3], and
2480 then compare a few consecutive 32 bit hashes before we know that we have no match.
2481 We don't end up marching through multiple words of memory and we really keep the
2482 number of processor data cache lines being accessed as small as possible.</p>
2484 <p>The string hash that is used for these lookup tables is the Daniel J.
2485 Bernstein hash which is also used in the ELF GNU_HASH sections. It is a very
2486 good hash for all kinds of names in programs with very few hash collisions.</p>
2488 <p>Empty buckets are designated by using an invalid hash index of UINT32_MAX.</p>
2491 <!-- ======================================================================= -->
2493 <a name="acceltabledetails">Details</a>
2495 <!-- ======================================================================= -->
2497 <p>These name hash tables are designed to be generic where specializations of
2498 the table get to define additional data that goes into the header
2499 ("HeaderData"), how the string value is stored ("KeyType") and the content
2500 of the data for each hash value.</p>
2502 <h5>Header Layout</h5>
2503 <p>The header has a fixed part, and the specialized part. The exact format of
2505 <div class="doc_code">
2509 uint32_t magic; // 'HASH' magic value to allow endian detection
2510 uint16_t version; // Version number
2511 uint16_t hash_function; // The hash function enumeration that was used
2512 uint32_t bucket_count; // The number of buckets in this hash table
2513 uint32_t hashes_count; // The total number of unique hash values and hash data offsets in this table
2514 uint32_t header_data_len; // The bytes to skip to get to the hash indexes (buckets) for correct alignment
2515 // Specifically the length of the following HeaderData field - this does not
2516 // include the size of the preceding fields
2517 HeaderData header_data; // Implementation specific header data
2521 <p>The header starts with a 32 bit "magic" value which must be 'HASH' encoded as
2522 an ASCII integer. This allows the detection of the start of the hash table and
2523 also allows the table's byte order to be determined so the table can be
2524 correctly extracted. The "magic" value is followed by a 16 bit version number
2525 which allows the table to be revised and modified in the future. The current
2526 version number is 1. "hash_function" is a uint16_t enumeration that specifies
2527 which hash function was used to produce this table. The current values for the
2528 hash function enumerations include:</p>
2529 <div class="doc_code">
2531 enum HashFunctionType
2533 eHashFunctionDJB = 0u, // Daniel J Bernstein hash function
2537 <p>"bucket_count" is a 32 bit unsigned integer that represents how many buckets
2538 are in the BUCKETS array. "hashes_count" is the number of unique 32 bit hash
2539 values that are in the HASHES array, and is the same number of offsets are
2540 contained in the OFFSETS array. "header_data_len" specifies the size in
2541 bytes of the HeaderData that is filled in by specialized versions of this
2544 <h5>Fixed Lookup</h5>
2545 <p>The header is followed by the buckets, hashes, offsets, and hash value
2547 <div class="doc_code">
2551 uint32_t buckets[Header.bucket_count]; // An array of hash indexes into the "hashes[]" array below
2552 uint32_t hashes [Header.hashes_count]; // Every unique 32 bit hash for the entire table is in this table
2553 uint32_t offsets[Header.hashes_count]; // An offset that corresponds to each item in the "hashes[]" array above
2557 <p>"buckets" is an array of 32 bit indexes into the "hashes" array. The
2558 "hashes" array contains all of the 32 bit hash values for all names in the
2559 hash table. Each hash in the "hashes" table has an offset in the "offsets"
2560 array that points to the data for the hash value.</p>
2562 <p>This table setup makes it very easy to repurpose these tables to contain
2563 different data, while keeping the lookup mechanism the same for all tables.
2564 This layout also makes it possible to save the table to disk and map it in
2565 later and do very efficient name lookups with little or no parsing.</p>
2567 <p>DWARF lookup tables can be implemented in a variety of ways and can store
2568 a lot of information for each name. We want to make the DWARF tables
2569 extensible and able to store the data efficiently so we have used some of the
2570 DWARF features that enable efficient data storage to define exactly what kind
2571 of data we store for each name.</p>
2573 <p>The "HeaderData" contains a definition of the contents of each HashData
2574 chunk. We might want to store an offset to all of the debug information
2575 entries (DIEs) for each name. To keep things extensible, we create a list of
2576 items, or Atoms, that are contained in the data for each name. First comes the
2577 type of the data in each atom:</p>
2578 <div class="doc_code">
2583 eAtomTypeDIEOffset = 1u, // DIE offset, check form for encoding
2584 eAtomTypeCUOffset = 2u, // DIE offset of the compiler unit header that contains the item in question
2585 eAtomTypeTag = 3u, // DW_TAG_xxx value, should be encoded as DW_FORM_data1 (if no tags exceed 255) or DW_FORM_data2
2586 eAtomTypeNameFlags = 4u, // Flags from enum NameFlags
2587 eAtomTypeTypeFlags = 5u, // Flags from enum TypeFlags
2591 <p>The enumeration values and their meanings are:</p>
2592 <div class="doc_code">
2594 eAtomTypeNULL - a termination atom that specifies the end of the atom list
2595 eAtomTypeDIEOffset - an offset into the .debug_info section for the DWARF DIE for this name
2596 eAtomTypeCUOffset - an offset into the .debug_info section for the CU that contains the DIE
2597 eAtomTypeDIETag - The DW_TAG_XXX enumeration value so you don't have to parse the DWARF to see what it is
2598 eAtomTypeNameFlags - Flags for functions and global variables (isFunction, isInlined, isExternal...)
2599 eAtomTypeTypeFlags - Flags for types (isCXXClass, isObjCClass, ...)
2602 <p>Then we allow each atom type to define the atom type and how the data for
2603 each atom type data is encoded:</p>
2604 <div class="doc_code">
2608 uint16_t type; // AtomType enum value
2609 uint16_t form; // DWARF DW_FORM_XXX defines
2613 <p>The "form" type above is from the DWARF specification and defines the
2614 exact encoding of the data for the Atom type. See the DWARF specification for
2615 the DW_FORM_ definitions.</p>
2616 <div class="doc_code">
2620 uint32_t die_offset_base;
2621 uint32_t atom_count;
2622 Atoms atoms[atom_count0];
2626 <p>"HeaderData" defines the base DIE offset that should be added to any atoms
2627 that are encoded using the DW_FORM_ref1, DW_FORM_ref2, DW_FORM_ref4,
2628 DW_FORM_ref8 or DW_FORM_ref_udata. It also defines what is contained in
2629 each "HashData" object -- Atom.form tells us how large each field will be in
2630 the HashData and the Atom.type tells us how this data should be interpreted.</p>
2632 <p>For the current implementations of the ".apple_names" (all functions + globals),
2633 the ".apple_types" (names of all types that are defined), and the
2634 ".apple_namespaces" (all namespaces), we currently set the Atom array to be:</p>
2635 <div class="doc_code">
2637 HeaderData.atom_count = 1;
2638 HeaderData.atoms[0].type = eAtomTypeDIEOffset;
2639 HeaderData.atoms[0].form = DW_FORM_data4;
2642 <p>This defines the contents to be the DIE offset (eAtomTypeDIEOffset) that is
2643 encoded as a 32 bit value (DW_FORM_data4). This allows a single name to have
2644 multiple matching DIEs in a single file, which could come up with an inlined
2645 function for instance. Future tables could include more information about the
2646 DIE such as flags indicating if the DIE is a function, method, block,
2649 <p>The KeyType for the DWARF table is a 32 bit string table offset into the
2650 ".debug_str" table. The ".debug_str" is the string table for the DWARF which
2651 may already contain copies of all of the strings. This helps make sure, with
2652 help from the compiler, that we reuse the strings between all of the DWARF
2653 sections and keeps the hash table size down. Another benefit to having the
2654 compiler generate all strings as DW_FORM_strp in the debug info, is that
2655 DWARF parsing can be made much faster.</p>
2657 <p>After a lookup is made, we get an offset into the hash data. The hash data
2658 needs to be able to deal with 32 bit hash collisions, so the chunk of data
2659 at the offset in the hash data consists of a triple:</p>
2660 <div class="doc_code">
2663 uint32_t hash_data_count
2664 HashData[hash_data_count]
2667 <p>If "str_offset" is zero, then the bucket contents are done. 99.9% of the
2668 hash data chunks contain a single item (no 32 bit hash collision):</p>
2669 <div class="doc_code">
2672 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2673 | 0x00000004 | uint32_t HashData count
2674 | 0x........ | uint32_t HashData[0] DIE offset
2675 | 0x........ | uint32_t HashData[1] DIE offset
2676 | 0x........ | uint32_t HashData[2] DIE offset
2677 | 0x........ | uint32_t HashData[3] DIE offset
2678 | 0x00000000 | uint32_t KeyType (end of hash chain)
2682 <p>If there are collisions, you will have multiple valid string offsets:</p>
2683 <div class="doc_code">
2686 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2687 | 0x00000004 | uint32_t HashData count
2688 | 0x........ | uint32_t HashData[0] DIE offset
2689 | 0x........ | uint32_t HashData[1] DIE offset
2690 | 0x........ | uint32_t HashData[2] DIE offset
2691 | 0x........ | uint32_t HashData[3] DIE offset
2692 | 0x00002023 | uint32_t KeyType (.debug_str[0x0002023] => "print")
2693 | 0x00000002 | uint32_t HashData count
2694 | 0x........ | uint32_t HashData[0] DIE offset
2695 | 0x........ | uint32_t HashData[1] DIE offset
2696 | 0x00000000 | uint32_t KeyType (end of hash chain)
2700 <p>Current testing with real world C++ binaries has shown that there is around 1
2701 32 bit hash collision per 100,000 name entries.</p>
2703 <!-- ======================================================================= -->
2705 <a name="acceltablecontents">Contents</a>
2707 <!-- ======================================================================= -->
2709 <p>As we said, we want to strictly define exactly what is included in the
2710 different tables. For DWARF, we have 3 tables: ".apple_names", ".apple_types",
2711 and ".apple_namespaces".</p>
2713 <p>".apple_names" sections should contain an entry for each DWARF DIE whose
2714 DW_TAG is a DW_TAG_label, DW_TAG_inlined_subroutine, or DW_TAG_subprogram that
2715 has address attributes: DW_AT_low_pc, DW_AT_high_pc, DW_AT_ranges or
2716 DW_AT_entry_pc. It also contains DW_TAG_variable DIEs that have a DW_OP_addr
2717 in the location (global and static variables). All global and static variables
2718 should be included, including those scoped withing functions and classes. For
2719 example using the following code:</p>
2720 <div class="doc_code">
2730 <p>Both of the static "var" variables would be included in the table. All
2731 functions should emit both their full names and their basenames. For C or C++,
2732 the full name is the mangled name (if available) which is usually in the
2733 DW_AT_MIPS_linkage_name attribute, and the DW_AT_name contains the function
2734 basename. If global or static variables have a mangled name in a
2735 DW_AT_MIPS_linkage_name attribute, this should be emitted along with the
2736 simple name found in the DW_AT_name attribute.</p>
2738 <p>".apple_types" sections should contain an entry for each DWARF DIE whose
2741 <li>DW_TAG_array_type</li>
2742 <li>DW_TAG_class_type</li>
2743 <li>DW_TAG_enumeration_type</li>
2744 <li>DW_TAG_pointer_type</li>
2745 <li>DW_TAG_reference_type</li>
2746 <li>DW_TAG_string_type</li>
2747 <li>DW_TAG_structure_type</li>
2748 <li>DW_TAG_subroutine_type</li>
2749 <li>DW_TAG_typedef</li>
2750 <li>DW_TAG_union_type</li>
2751 <li>DW_TAG_ptr_to_member_type</li>
2752 <li>DW_TAG_set_type</li>
2753 <li>DW_TAG_subrange_type</li>
2754 <li>DW_TAG_base_type</li>
2755 <li>DW_TAG_const_type</li>
2756 <li>DW_TAG_constant</li>
2757 <li>DW_TAG_file_type</li>
2758 <li>DW_TAG_namelist</li>
2759 <li>DW_TAG_packed_type</li>
2760 <li>DW_TAG_volatile_type</li>
2761 <li>DW_TAG_restrict_type</li>
2762 <li>DW_TAG_interface_type</li>
2763 <li>DW_TAG_unspecified_type</li>
2764 <li>DW_TAG_shared_type</li>
2766 <p>Only entries with a DW_AT_name attribute are included, and the entry must
2767 not be a forward declaration (DW_AT_declaration attribute with a non-zero value).
2768 For example, using the following code:</p>
2769 <div class="doc_code">
2778 <p>We get a few type DIEs:</p>
2779 <div class="doc_code">
2781 0x00000067: TAG_base_type [5]
2782 AT_encoding( DW_ATE_signed )
2784 AT_byte_size( 0x04 )
2786 0x0000006e: TAG_pointer_type [6]
2787 AT_type( {0x00000067} ( int ) )
2788 AT_byte_size( 0x08 )
2791 <p>The DW_TAG_pointer_type is not included because it does not have a DW_AT_name.</p>
2793 <p>".apple_namespaces" section should contain all DW_TAG_namespace DIEs. If
2794 we run into a namespace that has no name this is an anonymous namespace,
2795 and the name should be output as "(anonymous namespace)" (without the quotes).
2796 Why? This matches the output of the abi::cxa_demangle() that is in the standard
2797 C++ library that demangles mangled names.</p>
2800 <!-- ======================================================================= -->
2802 <a name="acceltableextensions">Language Extensions and File Format Changes</a>
2804 <!-- ======================================================================= -->
2806 <h5>Objective-C Extensions</h5>
2807 <p>".apple_objc" section should contain all DW_TAG_subprogram DIEs for an
2808 Objective-C class. The name used in the hash table is the name of the
2809 Objective-C class itself. If the Objective-C class has a category, then an
2810 entry is made for both the class name without the category, and for the class
2811 name with the category. So if we have a DIE at offset 0x1234 with a name
2812 of method "-[NSString(my_additions) stringWithSpecialString:]", we would add
2813 an entry for "NSString" that points to DIE 0x1234, and an entry for
2814 "NSString(my_additions)" that points to 0x1234. This allows us to quickly
2815 track down all Objective-C methods for an Objective-C class when doing
2816 expressions. It is needed because of the dynamic nature of Objective-C where
2817 anyone can add methods to a class. The DWARF for Objective-C methods is also
2818 emitted differently from C++ classes where the methods are not usually
2819 contained in the class definition, they are scattered about across one or more
2820 compile units. Categories can also be defined in different shared libraries.
2821 So we need to be able to quickly find all of the methods and class functions
2822 given the Objective-C class name, or quickly find all methods and class
2823 functions for a class + category name. This table does not contain any selector
2824 names, it just maps Objective-C class names (or class names + category) to all
2825 of the methods and class functions. The selectors are added as function
2826 basenames in the .debug_names section.</p>
2828 <p>In the ".apple_names" section for Objective-C functions, the full name is the
2829 entire function name with the brackets ("-[NSString stringWithCString:]") and the
2830 basename is the selector only ("stringWithCString:").</p>
2832 <h5>Mach-O Changes</h5>
2833 <p>The sections names for the apple hash tables are for non mach-o files. For
2834 mach-o files, the sections should be contained in the "__DWARF" segment with
2835 names as follows:</p>
2837 <li>".apple_names" -> "__apple_names"</li>
2838 <li>".apple_types" -> "__apple_types"</li>
2839 <li>".apple_namespaces" -> "__apple_namespac" (16 character limit)</li>
2840 <li> ".apple_objc" -> "__apple_objc"</li>
2846 <!-- *********************************************************************** -->
2850 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2851 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2852 <a href="http://validator.w3.org/check/referer"><img
2853 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2855 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2856 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
2857 Last modified: $Date$