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></li>
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>
66 <li><a href="#acceltable">Name Accelerator Tables</a></li>
68 <li><a href="#acceltableintroduction">Introduction</a></li>
69 <li><a href="#acceltablehashes">Hash Tables</a></li>
70 <li><a href="#acceltabledetails">Details</a></li>
71 <li><a href="#acceltablecontents">Contents</a></li>
72 <li><a href="#acceltableextensions">Language Extensions and File Format Changes</a></li>
79 <img src="img/venusflytrap.jpg" alt="A leafy and green bug eater" width="247"
84 <div class="doc_author">
85 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
86 and <a href="mailto:jlaskey@mac.com">Jim Laskey</a></p>
90 <!-- *********************************************************************** -->
91 <h2><a name="introduction">Introduction</a></h2>
92 <!-- *********************************************************************** -->
96 <p>This document is the central repository for all information pertaining to
97 debug information in LLVM. It describes the <a href="#format">actual format
98 that the LLVM debug information</a> takes, which is useful for those
99 interested in creating front-ends or dealing directly with the information.
100 Further, this document provides specific examples of what debug information
101 for C/C++ looks like.</p>
103 <!-- ======================================================================= -->
105 <a name="phil">Philosophy behind LLVM debugging information</a>
110 <p>The idea of the LLVM debugging information is to capture how the important
111 pieces of the source-language's Abstract Syntax Tree map onto LLVM code.
112 Several design aspects have shaped the solution that appears here. The
113 important ones are:</p>
116 <li>Debugging information should have very little impact on the rest of the
117 compiler. No transformations, analyses, or code generators should need to
118 be modified because of debugging information.</li>
120 <li>LLVM optimizations should interact in <a href="#debugopt">well-defined and
121 easily described ways</a> with the debugging information.</li>
123 <li>Because LLVM is designed to support arbitrary programming languages,
124 LLVM-to-LLVM tools should not need to know anything about the semantics of
125 the source-level-language.</li>
127 <li>Source-level languages are often <b>widely</b> different from one another.
128 LLVM should not put any restrictions of the flavor of the source-language,
129 and the debugging information should work with any language.</li>
131 <li>With code generator support, it should be possible to use an LLVM compiler
132 to compile a program to native machine code and standard debugging
133 formats. This allows compatibility with traditional machine-code level
134 debuggers, like GDB or DBX.</li>
137 <p>The approach used by the LLVM implementation is to use a small set
138 of <a href="#format_common_intrinsics">intrinsic functions</a> to define a
139 mapping between LLVM program objects and the source-level objects. The
140 description of the source-level program is maintained in LLVM metadata
141 in an <a href="#ccxx_frontend">implementation-defined format</a>
142 (the C/C++ front-end currently uses working draft 7 of
143 the <a href="http://www.eagercon.com/dwarf/dwarf3std.htm">DWARF 3
146 <p>When a program is being debugged, a debugger interacts with the user and
147 turns the stored debug information into source-language specific information.
148 As such, a debugger must be aware of the source-language, and is thus tied to
149 a specific language or family of languages.</p>
153 <!-- ======================================================================= -->
155 <a name="consumers">Debug information consumers</a>
160 <p>The role of debug information is to provide meta information normally
161 stripped away during the compilation process. This meta information provides
162 an LLVM user a relationship between generated code and the original program
165 <p>Currently, debug information is consumed by DwarfDebug to produce dwarf
166 information used by the gdb debugger. Other targets could use the same
167 information to produce stabs or other debug forms.</p>
169 <p>It would also be reasonable to use debug information to feed profiling tools
170 for analysis of generated code, or, tools for reconstructing the original
171 source from generated code.</p>
173 <p>TODO - expound a bit more.</p>
177 <!-- ======================================================================= -->
179 <a name="debugopt">Debugging optimized code</a>
184 <p>An extremely high priority of LLVM debugging information is to make it
185 interact well with optimizations and analysis. In particular, the LLVM debug
186 information provides the following guarantees:</p>
189 <li>LLVM debug information <b>always provides information to accurately read
190 the source-level state of the program</b>, regardless of which LLVM
191 optimizations have been run, and without any modification to the
192 optimizations themselves. However, some optimizations may impact the
193 ability to modify the current state of the program with a debugger, such
194 as setting program variables, or calling functions that have been
197 <li>As desired, LLVM optimizations can be upgraded to be aware of the LLVM
198 debugging information, allowing them to update the debugging information
199 as they perform aggressive optimizations. This means that, with effort,
200 the LLVM optimizers could optimize debug code just as well as non-debug
203 <li>LLVM debug information does not prevent optimizations from
204 happening (for example inlining, basic block reordering/merging/cleanup,
205 tail duplication, etc).</li>
207 <li>LLVM debug information is automatically optimized along with the rest of
208 the program, using existing facilities. For example, duplicate
209 information is automatically merged by the linker, and unused information
210 is automatically removed.</li>
213 <p>Basically, the debug information allows you to compile a program with
214 "<tt>-O0 -g</tt>" and get full debug information, allowing you to arbitrarily
215 modify the program as it executes from a debugger. Compiling a program with
216 "<tt>-O3 -g</tt>" gives you full debug information that is always available
217 and accurate for reading (e.g., you get accurate stack traces despite tail
218 call elimination and inlining), but you might lose the ability to modify the
219 program and call functions where were optimized out of the program, or
220 inlined away completely.</p>
222 <p><a href="TestingGuide.html#quicktestsuite">LLVM test suite</a> provides a
223 framework to test optimizer's handling of debugging information. It can be
226 <div class="doc_code">
228 % cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
233 <p>This will test impact of debugging information on optimization passes. If
234 debugging information influences optimization passes then it will be reported
235 as a failure. See <a href="TestingGuide.html">TestingGuide</a> for more
236 information on LLVM test infrastructure and how to run various tests.</p>
242 <!-- *********************************************************************** -->
244 <a name="format">Debugging information format</a>
246 <!-- *********************************************************************** -->
250 <p>LLVM debugging information has been carefully designed to make it possible
251 for the optimizer to optimize the program and debugging information without
252 necessarily having to know anything about debugging information. In
253 particular, the use of metadata avoids duplicated debugging information from
254 the beginning, and the global dead code elimination pass automatically
255 deletes debugging information for a function if it decides to delete the
258 <p>To do this, most of the debugging information (descriptors for types,
259 variables, functions, source files, etc) is inserted by the language
260 front-end in the form of LLVM metadata. </p>
262 <p>Debug information is designed to be agnostic about the target debugger and
263 debugging information representation (e.g. DWARF/Stabs/etc). It uses a
264 generic pass to decode the information that represents variables, types,
265 functions, namespaces, etc: this allows for arbitrary source-language
266 semantics and type-systems to be used, as long as there is a module
267 written for the target debugger to interpret the information. </p>
269 <p>To provide basic functionality, the LLVM debugger does have to make some
270 assumptions about the source-level language being debugged, though it keeps
271 these to a minimum. The only common features that the LLVM debugger assumes
272 exist are <a href="#format_files">source files</a>,
273 and <a href="#format_global_variables">program objects</a>. These abstract
274 objects are used by a debugger to form stack traces, show information about
275 local variables, etc.</p>
277 <p>This section of the documentation first describes the representation aspects
278 common to any source-language. The <a href="#ccxx_frontend">next section</a>
279 describes the data layout conventions used by the C and C++ front-ends.</p>
281 <!-- ======================================================================= -->
283 <a name="debug_info_descriptors">Debug information descriptors</a>
288 <p>In consideration of the complexity and volume of debug information, LLVM
289 provides a specification for well formed debug descriptors. </p>
291 <p>Consumers of LLVM debug information expect the descriptors for program
292 objects to start in a canonical format, but the descriptors can include
293 additional information appended at the end that is source-language
294 specific. All LLVM debugging information is versioned, allowing backwards
295 compatibility in the case that the core structures need to change in some
296 way. Also, all debugging information objects start with a tag to indicate
297 what type of object it is. The source-language is allowed to define its own
298 objects, by using unreserved tag numbers. We recommend using with tags in
299 the range 0x1000 through 0x2000 (there is a defined enum DW_TAG_user_base =
302 <p>The fields of debug descriptors used internally by LLVM
303 are restricted to only the simple data types <tt>i32</tt>, <tt>i1</tt>,
304 <tt>float</tt>, <tt>double</tt>, <tt>mdstring</tt> and <tt>mdnode</tt>. </p>
306 <div class="doc_code">
315 <p><a name="LLVMDebugVersion">The first field of a descriptor is always an
316 <tt>i32</tt> containing a tag value identifying the content of the
317 descriptor. The remaining fields are specific to the descriptor. The values
318 of tags are loosely bound to the tag values of DWARF information entries.
319 However, that does not restrict the use of the information supplied to DWARF
320 targets. To facilitate versioning of debug information, the tag is augmented
321 with the current debug version (LLVMDebugVersion = 8 << 16 or
322 0x80000 or 524288.)</a></p>
324 <p>The details of the various descriptors follow.</p>
326 <!-- ======================================================================= -->
328 <a name="format_compile_units">Compile unit descriptors</a>
333 <div class="doc_code">
336 i32, ;; Tag = 17 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
337 ;; (DW_TAG_compile_unit)
338 i32, ;; Unused field.
339 i32, ;; DWARF language identifier (ex. DW_LANG_C89)
340 metadata, ;; Source file name
341 metadata, ;; Source file directory (includes trailing slash)
342 metadata ;; Producer (ex. "4.0.1 LLVM (LLVM research group)")
343 i1, ;; True if this is a main compile unit.
344 i1, ;; True if this is optimized.
346 i32 ;; Runtime version
347 metadata ;; List of enums types
348 metadata ;; List of retained types
349 metadata ;; List of subprograms
350 metadata ;; List of global variables
355 <p>These descriptors contain a source language ID for the file (we use the DWARF
356 3.0 ID numbers, such as <tt>DW_LANG_C89</tt>, <tt>DW_LANG_C_plus_plus</tt>,
357 <tt>DW_LANG_Cobol74</tt>, etc), three strings describing the filename,
358 working directory of the compiler, and an identifier string for the compiler
359 that produced it.</p>
361 <p>Compile unit descriptors provide the root context for objects declared in a
362 specific compilation unit. File descriptors are defined using this context.
363 These descriptors are collected by a named metadata
364 <tt>!llvm.dbg.cu</tt>. Compile unit descriptor keeps track of subprograms,
365 global variables and type information.
369 <!-- ======================================================================= -->
371 <a name="format_files">File descriptors</a>
376 <div class="doc_code">
379 i32, ;; Tag = 41 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
380 ;; (DW_TAG_file_type)
381 metadata, ;; Source file name
382 metadata, ;; Source file directory (includes trailing slash)
388 <p>These descriptors contain information for a file. Global variables and top
389 level functions would be defined using this context.k File descriptors also
390 provide context for source line correspondence. </p>
392 <p>Each input file is encoded as a separate file descriptor in LLVM debugging
393 information output. </p>
397 <!-- ======================================================================= -->
399 <a name="format_global_variables">Global variable descriptors</a>
404 <div class="doc_code">
407 i32, ;; Tag = 52 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
409 i32, ;; Unused field.
410 metadata, ;; Reference to context descriptor
412 metadata, ;; Display name (fully qualified C++ name)
413 metadata, ;; MIPS linkage name (for C++)
414 metadata, ;; Reference to file where defined
415 i32, ;; Line number where defined
416 metadata, ;; Reference to type descriptor
417 i1, ;; True if the global is local to compile unit (static)
418 i1, ;; True if the global is defined in the compile unit (not extern)
419 {}* ;; Reference to the global variable
424 <p>These descriptors provide debug information about globals variables. The
425 provide details such as name, type and where the variable is defined. All
426 global variables are collected by named metadata <tt>!llvm.dbg.gv</tt>.</p>
430 <!-- ======================================================================= -->
432 <a name="format_subprograms">Subprogram descriptors</a>
437 <div class="doc_code">
440 i32, ;; Tag = 46 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
441 ;; (DW_TAG_subprogram)
442 i32, ;; Unused field.
443 metadata, ;; Reference to context descriptor
445 metadata, ;; Display name (fully qualified C++ name)
446 metadata, ;; MIPS linkage name (for C++)
447 metadata, ;; Reference to file where defined
448 i32, ;; Line number where defined
449 metadata, ;; Reference to type descriptor
450 i1, ;; True if the global is local to compile unit (static)
451 i1, ;; True if the global is defined in the compile unit (not extern)
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 by named metadata <tt>!llvm.dbg.enum</tt>.</p>
691 <p>The members of structure (tag = <tt>DW_TAG_structure_type</tt>) or union (tag
692 = <tt>DW_TAG_union_type</tt>) types are any one of
693 the <a href="#format_basic_type">basic</a>,
694 <a href="#format_derived_type">derived</a>
695 or <a href="#format_composite_type">composite</a> type descriptors, each
696 representing a field member of the structure or union.</p>
698 <p>For C++ classes (tag = <tt>DW_TAG_structure_type</tt>), member descriptors
699 provide information about base classes, static members and member
700 functions. If a member is a <a href="#format_derived_type">derived type
701 descriptor</a> and has a tag of <tt>DW_TAG_inheritance</tt>, then the type
702 represents a base class. If the member of is
703 a <a href="#format_global_variables">global variable descriptor</a> then it
704 represents a static member. And, if the member is
705 a <a href="#format_subprograms">subprogram descriptor</a> then it represents
706 a member function. For static members and member
707 functions, <tt>getName()</tt> returns the members link or the C++ mangled
708 name. <tt>getDisplayName()</tt> the simplied version of the name.</p>
710 <p>The first member of subroutine (tag = <tt>DW_TAG_subroutine_type</tt>) type
711 elements is the return type for the subroutine. The remaining elements are
712 the formal arguments to the subroutine.</p>
714 <p><a href="#format_composite_type">Composite type</a> location can be
715 determined from the context and line number. The size, alignment and
716 offset are expressed in bits and can be 64 bit values. The alignment is used
717 to round the offset when embedded in
718 a <a href="#format_composite_type">composite type</a> (as an example, to keep
719 float doubles on 64 bit boundaries.) The offset is the bit offset if embedded
720 in a <a href="#format_composite_type">composite type</a>.</p>
724 <!-- ======================================================================= -->
726 <a name="format_subrange">Subrange descriptors</a>
731 <div class="doc_code">
734 i32, ;; Tag = 33 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_subrange_type)
741 <p>These descriptors are used to define ranges of array subscripts for an array
742 <a href="#format_composite_type">composite type</a>. The low value defines
743 the lower bounds typically zero for C/C++. The high value is the upper
744 bounds. Values are 64 bit. High - low + 1 is the size of the array. If low
745 > high the array bounds are not included in generated debugging information.
750 <!-- ======================================================================= -->
752 <a name="format_enumeration">Enumerator descriptors</a>
757 <div class="doc_code">
760 i32, ;; Tag = 40 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
761 ;; (DW_TAG_enumerator)
768 <p>These descriptors are used to define members of an
769 enumeration <a href="#format_composite_type">composite type</a>, it
770 associates the name to the value.</p>
774 <!-- ======================================================================= -->
776 <a name="format_variables">Local variables</a>
781 <div class="doc_code">
784 i32, ;; Tag (see below)
787 metadata, ;; Reference to file where defined
788 i32, ;; 24 bit - Line number where defined
789 ;; 8 bit - Argument number. 1 indicates 1st argument.
790 metadata, ;; Type descriptor
792 metadata ;; (optional) Reference to inline location
797 <p>These descriptors are used to define variables local to a sub program. The
798 value of the tag depends on the usage of the variable:</p>
800 <div class="doc_code">
802 DW_TAG_auto_variable = 256
803 DW_TAG_arg_variable = 257
804 DW_TAG_return_variable = 258
808 <p>An auto variable is any variable declared in the body of the function. An
809 argument variable is any variable that appears as a formal argument to the
810 function. A return variable is used to track the result of a function and
811 has no source correspondent.</p>
813 <p>The context is either the subprogram or block where the variable is defined.
814 Name the source variable name. Context and line indicate where the
815 variable was defined. Type descriptor defines the declared type of the
822 <!-- ======================================================================= -->
824 <a name="format_common_intrinsics">Debugger intrinsic functions</a>
829 <p>LLVM uses several intrinsic functions (name prefixed with "llvm.dbg") to
830 provide debug information at various points in generated code.</p>
832 <!-- ======================================================================= -->
834 <a name="format_common_declare">llvm.dbg.declare</a>
839 void %<a href="#format_common_declare">llvm.dbg.declare</a>(metadata, metadata)
842 <p>This intrinsic provides information about a local element (e.g., variable). The
843 first argument is metadata holding the alloca for the variable. The
844 second argument is metadata containing a description of the variable.</p>
847 <!-- ======================================================================= -->
849 <a name="format_common_value">llvm.dbg.value</a>
854 void %<a href="#format_common_value">llvm.dbg.value</a>(metadata, i64, metadata)
857 <p>This intrinsic provides information when a user source variable is set to a
858 new value. The first argument is the new value (wrapped as metadata). The
859 second argument is the offset in the user source variable where the new value
860 is written. The third argument is metadata containing a description of the
861 user source variable.</p>
866 <!-- ======================================================================= -->
868 <a name="format_common_lifetime">Object lifetimes and scoping</a>
872 <p>In many languages, the local variables in functions can have their lifetimes
873 or scopes limited to a subset of a function. In the C family of languages,
874 for example, variables are only live (readable and writable) within the
875 source block that they are defined in. In functional languages, values are
876 only readable after they have been defined. Though this is a very obvious
877 concept, it is non-trivial to model in LLVM, because it has no notion of
878 scoping in this sense, and does not want to be tied to a language's scoping
881 <p>In order to handle this, the LLVM debug format uses the metadata attached to
882 llvm instructions to encode line number and scoping information. Consider
883 the following C fragment, for example:</p>
885 <div class="doc_code">
899 <p>Compiled to LLVM, this function would be represented like this:</p>
901 <div class="doc_code">
903 define void @foo() nounwind ssp {
905 %X = alloca i32, align 4 ; <i32*> [#uses=4]
906 %Y = alloca i32, align 4 ; <i32*> [#uses=4]
907 %Z = alloca i32, align 4 ; <i32*> [#uses=3]
908 %0 = bitcast i32* %X to {}* ; <{}*> [#uses=1]
909 call void @llvm.dbg.declare(metadata !{i32 * %X}, metadata !0), !dbg !7
910 store i32 21, i32* %X, !dbg !8
911 %1 = bitcast i32* %Y to {}* ; <{}*> [#uses=1]
912 call void @llvm.dbg.declare(metadata !{i32 * %Y}, metadata !9), !dbg !10
913 store i32 22, i32* %Y, !dbg !11
914 %2 = bitcast i32* %Z to {}* ; <{}*> [#uses=1]
915 call void @llvm.dbg.declare(metadata !{i32 * %Z}, metadata !12), !dbg !14
916 store i32 23, i32* %Z, !dbg !15
917 %tmp = load i32* %X, !dbg !16 ; <i32> [#uses=1]
918 %tmp1 = load i32* %Y, !dbg !16 ; <i32> [#uses=1]
919 %add = add nsw i32 %tmp, %tmp1, !dbg !16 ; <i32> [#uses=1]
920 store i32 %add, i32* %Z, !dbg !16
921 %tmp2 = load i32* %Y, !dbg !17 ; <i32> [#uses=1]
922 store i32 %tmp2, i32* %X, !dbg !17
926 declare void @llvm.dbg.declare(metadata, metadata) nounwind readnone
928 !0 = metadata !{i32 459008, metadata !1, metadata !"X",
929 metadata !3, i32 2, metadata !6}; [ DW_TAG_auto_variable ]
930 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
931 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo", metadata !"foo",
932 metadata !"foo", metadata !3, i32 1, metadata !4,
933 i1 false, i1 true}; [DW_TAG_subprogram ]
934 !3 = metadata !{i32 458769, i32 0, i32 12, metadata !"foo.c",
935 metadata !"/private/tmp", metadata !"clang 1.1", i1 true,
936 i1 false, metadata !"", i32 0}; [DW_TAG_compile_unit ]
937 !4 = metadata !{i32 458773, metadata !3, metadata !"", null, i32 0, i64 0, i64 0,
938 i64 0, i32 0, null, metadata !5, i32 0}; [DW_TAG_subroutine_type ]
939 !5 = metadata !{null}
940 !6 = metadata !{i32 458788, metadata !3, metadata !"int", metadata !3, i32 0,
941 i64 32, i64 32, i64 0, i32 0, i32 5}; [DW_TAG_base_type ]
942 !7 = metadata !{i32 2, i32 7, metadata !1, null}
943 !8 = metadata !{i32 2, i32 3, metadata !1, null}
944 !9 = metadata !{i32 459008, metadata !1, metadata !"Y", metadata !3, i32 3,
945 metadata !6}; [ DW_TAG_auto_variable ]
946 !10 = metadata !{i32 3, i32 7, metadata !1, null}
947 !11 = metadata !{i32 3, i32 3, metadata !1, null}
948 !12 = metadata !{i32 459008, metadata !13, metadata !"Z", metadata !3, i32 5,
949 metadata !6}; [ DW_TAG_auto_variable ]
950 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
951 !14 = metadata !{i32 5, i32 9, metadata !13, null}
952 !15 = metadata !{i32 5, i32 5, metadata !13, null}
953 !16 = metadata !{i32 6, i32 5, metadata !13, null}
954 !17 = metadata !{i32 8, i32 3, metadata !1, null}
955 !18 = metadata !{i32 9, i32 1, metadata !2, null}
959 <p>This example illustrates a few important details about LLVM debugging
960 information. In particular, it shows how the <tt>llvm.dbg.declare</tt>
961 intrinsic and location information, which are attached to an instruction,
962 are applied together to allow a debugger to analyze the relationship between
963 statements, variable definitions, and the code used to implement the
966 <div class="doc_code">
968 call void @llvm.dbg.declare(metadata, metadata !0), !dbg !7
972 <p>The first intrinsic
973 <tt>%<a href="#format_common_declare">llvm.dbg.declare</a></tt>
974 encodes debugging information for the variable <tt>X</tt>. The metadata
975 <tt>!dbg !7</tt> attached to the intrinsic provides scope information for the
976 variable <tt>X</tt>.</p>
978 <div class="doc_code">
980 !7 = metadata !{i32 2, i32 7, metadata !1, null}
981 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
982 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo",
983 metadata !"foo", metadata !"foo", metadata !3, i32 1,
984 metadata !4, i1 false, i1 true}; [DW_TAG_subprogram ]
988 <p>Here <tt>!7</tt> is metadata providing location information. It has four
989 fields: line number, column number, scope, and original scope. The original
990 scope represents inline location if this instruction is inlined inside a
991 caller, and is null otherwise. In this example, scope is encoded by
992 <tt>!1</tt>. <tt>!1</tt> represents a lexical block inside the scope
993 <tt>!2</tt>, where <tt>!2</tt> is a
994 <a href="#format_subprograms">subprogram descriptor</a>. This way the
995 location information attached to the intrinsics indicates that the
996 variable <tt>X</tt> is declared at line number 2 at a function level scope in
997 function <tt>foo</tt>.</p>
999 <p>Now lets take another example.</p>
1001 <div class="doc_code">
1003 call void @llvm.dbg.declare(metadata, metadata !12), !dbg !14
1007 <p>The second intrinsic
1008 <tt>%<a href="#format_common_declare">llvm.dbg.declare</a></tt>
1009 encodes debugging information for variable <tt>Z</tt>. The metadata
1010 <tt>!dbg !14</tt> attached to the intrinsic provides scope information for
1011 the variable <tt>Z</tt>.</p>
1013 <div class="doc_code">
1015 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
1016 !14 = metadata !{i32 5, i32 9, metadata !13, null}
1020 <p>Here <tt>!14</tt> indicates that <tt>Z</tt> is declared at line number 5 and
1021 column number 9 inside of lexical scope <tt>!13</tt>. The lexical scope
1022 itself resides inside of lexical scope <tt>!1</tt> described above.</p>
1024 <p>The scope information attached with each instruction provides a
1025 straightforward way to find instructions covered by a scope.</p>
1031 <!-- *********************************************************************** -->
1033 <a name="ccxx_frontend">C/C++ front-end specific debug information</a>
1035 <!-- *********************************************************************** -->
1039 <p>The C and C++ front-ends represent information about the program in a format
1040 that is effectively identical
1041 to <a href="http://www.eagercon.com/dwarf/dwarf3std.htm">DWARF 3.0</a> in
1042 terms of information content. This allows code generators to trivially
1043 support native debuggers by generating standard dwarf information, and
1044 contains enough information for non-dwarf targets to translate it as
1047 <p>This section describes the forms used to represent C and C++ programs. Other
1048 languages could pattern themselves after this (which itself is tuned to
1049 representing programs in the same way that DWARF 3 does), or they could
1050 choose to provide completely different forms if they don't fit into the DWARF
1051 model. As support for debugging information gets added to the various LLVM
1052 source-language front-ends, the information used should be documented
1055 <p>The following sections provide examples of various C/C++ constructs and the
1056 debug information that would best describe those constructs.</p>
1058 <!-- ======================================================================= -->
1060 <a name="ccxx_compile_units">C/C++ source file information</a>
1065 <p>Given the source files <tt>MySource.cpp</tt> and <tt>MyHeader.h</tt> located
1066 in the directory <tt>/Users/mine/sources</tt>, the following code:</p>
1068 <div class="doc_code">
1070 #include "MyHeader.h"
1072 int main(int argc, char *argv[]) {
1078 <p>a C/C++ front-end would generate the following descriptors:</p>
1080 <div class="doc_code">
1084 ;; Define the compile unit for the main source file "/Users/mine/sources/MySource.cpp".
1089 i32 4, ;; Language Id
1090 metadata !"MySource.cpp",
1091 metadata !"/Users/mine/sources",
1092 metadata !"4.2.1 (Based on Apple Inc. build 5649) (LLVM build 00)",
1093 i1 true, ;; Main Compile Unit
1094 i1 false, ;; Optimized compile unit
1095 metadata !"", ;; Compiler flags
1096 i32 0} ;; Runtime version
1099 ;; Define the file for the file "/Users/mine/sources/MySource.cpp".
1103 metadata !"MySource.cpp",
1104 metadata !"/Users/mine/sources",
1105 metadata !2 ;; Compile unit
1109 ;; Define the file for the file "/Users/mine/sources/Myheader.h"
1113 metadata !"Myheader.h"
1114 metadata !"/Users/mine/sources",
1115 metadata !2 ;; Compile unit
1122 <p>llvm::Instruction provides easy access to metadata attached with an
1123 instruction. One can extract line number information encoded in LLVM IR
1124 using <tt>Instruction::getMetadata()</tt> and
1125 <tt>DILocation::getLineNumber()</tt>.
1127 if (MDNode *N = I->getMetadata("dbg")) { // Here I is an LLVM instruction
1128 DILocation Loc(N); // DILocation is in DebugInfo.h
1129 unsigned Line = Loc.getLineNumber();
1130 StringRef File = Loc.getFilename();
1131 StringRef Dir = Loc.getDirectory();
1136 <!-- ======================================================================= -->
1138 <a name="ccxx_global_variable">C/C++ global variable information</a>
1143 <p>Given an integer global variable declared as follows:</p>
1145 <div class="doc_code">
1151 <p>a C/C++ front-end would generate the following descriptors:</p>
1153 <div class="doc_code">
1156 ;; Define the global itself.
1158 %MyGlobal = global int 100
1161 ;; List of debug info of globals
1163 !llvm.dbg.gv = !{!0}
1166 ;; Define the global variable descriptor. Note the reference to the global
1167 ;; variable anchor and the global variable itself.
1172 metadata !1, ;; Context
1173 metadata !"MyGlobal", ;; Name
1174 metadata !"MyGlobal", ;; Display Name
1175 metadata !"MyGlobal", ;; Linkage Name
1176 metadata !3, ;; Compile Unit
1177 i32 1, ;; Line Number
1178 metadata !4, ;; Type
1179 i1 false, ;; Is a local variable
1180 i1 true, ;; Is this a definition
1181 i32* @MyGlobal ;; The global variable
1185 ;; Define the basic type of 32 bit signed integer. Note that since int is an
1186 ;; intrinsic type the source file is NULL and line 0.
1190 metadata !1, ;; Context
1191 metadata !"int", ;; Name
1192 metadata !1, ;; File
1193 i32 0, ;; Line number
1194 i64 32, ;; Size in Bits
1195 i64 32, ;; Align in Bits
1196 i64 0, ;; Offset in Bits
1206 <!-- ======================================================================= -->
1208 <a name="ccxx_subprogram">C/C++ function information</a>
1213 <p>Given a function declared as follows:</p>
1215 <div class="doc_code">
1217 int main(int argc, char *argv[]) {
1223 <p>a C/C++ front-end would generate the following descriptors:</p>
1225 <div class="doc_code">
1228 ;; Define the anchor for subprograms. Note that the second field of the
1229 ;; anchor is 46, which is the same as the tag for subprograms
1230 ;; (46 = DW_TAG_subprogram.)
1235 metadata !1, ;; Context
1236 metadata !"main", ;; Name
1237 metadata !"main", ;; Display name
1238 metadata !"main", ;; Linkage name
1239 metadata !1, ;; File
1240 i32 1, ;; Line number
1241 metadata !4, ;; Type
1242 i1 false, ;; Is local
1243 i1 true, ;; Is definition
1244 i32 0, ;; Virtuality attribute, e.g. pure virtual function
1245 i32 0, ;; Index into virtual table for C++ methods
1246 i32 0, ;; Type that holds virtual table.
1248 i1 false, ;; True if this function is optimized
1249 Function *, ;; Pointer to llvm::Function
1250 null ;; Function template parameters
1253 ;; Define the subprogram itself.
1255 define i32 @main(i32 %argc, i8** %argv) {
1263 <!-- ======================================================================= -->
1265 <a name="ccxx_basic_types">C/C++ basic types</a>
1270 <p>The following are the basic type descriptors for C/C++ core types:</p>
1272 <!-- ======================================================================= -->
1274 <a name="ccxx_basic_type_bool">bool</a>
1279 <div class="doc_code">
1283 metadata !1, ;; Context
1284 metadata !"bool", ;; Name
1285 metadata !1, ;; File
1286 i32 0, ;; Line number
1287 i64 8, ;; Size in Bits
1288 i64 8, ;; Align in Bits
1289 i64 0, ;; Offset in Bits
1298 <!-- ======================================================================= -->
1300 <a name="ccxx_basic_char">char</a>
1305 <div class="doc_code">
1309 metadata !1, ;; Context
1310 metadata !"char", ;; Name
1311 metadata !1, ;; File
1312 i32 0, ;; Line number
1313 i64 8, ;; Size in Bits
1314 i64 8, ;; Align in Bits
1315 i64 0, ;; Offset in Bits
1324 <!-- ======================================================================= -->
1326 <a name="ccxx_basic_unsigned_char">unsigned char</a>
1331 <div class="doc_code">
1335 metadata !1, ;; Context
1336 metadata !"unsigned char",
1337 metadata !1, ;; File
1338 i32 0, ;; Line number
1339 i64 8, ;; Size in Bits
1340 i64 8, ;; Align in Bits
1341 i64 0, ;; Offset in Bits
1350 <!-- ======================================================================= -->
1352 <a name="ccxx_basic_short">short</a>
1357 <div class="doc_code">
1361 metadata !1, ;; Context
1362 metadata !"short int",
1363 metadata !1, ;; File
1364 i32 0, ;; Line number
1365 i64 16, ;; Size in Bits
1366 i64 16, ;; Align in Bits
1367 i64 0, ;; Offset in Bits
1376 <!-- ======================================================================= -->
1378 <a name="ccxx_basic_unsigned_short">unsigned short</a>
1383 <div class="doc_code">
1387 metadata !1, ;; Context
1388 metadata !"short unsigned int",
1389 metadata !1, ;; File
1390 i32 0, ;; Line number
1391 i64 16, ;; Size in Bits
1392 i64 16, ;; Align in Bits
1393 i64 0, ;; Offset in Bits
1402 <!-- ======================================================================= -->
1404 <a name="ccxx_basic_int">int</a>
1409 <div class="doc_code">
1413 metadata !1, ;; Context
1414 metadata !"int", ;; Name
1415 metadata !1, ;; File
1416 i32 0, ;; Line number
1417 i64 32, ;; Size in Bits
1418 i64 32, ;; Align in Bits
1419 i64 0, ;; Offset in Bits
1427 <!-- ======================================================================= -->
1429 <a name="ccxx_basic_unsigned_int">unsigned int</a>
1434 <div class="doc_code">
1438 metadata !1, ;; Context
1439 metadata !"unsigned int",
1440 metadata !1, ;; File
1441 i32 0, ;; Line number
1442 i64 32, ;; Size in Bits
1443 i64 32, ;; Align in Bits
1444 i64 0, ;; Offset in Bits
1453 <!-- ======================================================================= -->
1455 <a name="ccxx_basic_long_long">long long</a>
1460 <div class="doc_code">
1464 metadata !1, ;; Context
1465 metadata !"long long int",
1466 metadata !1, ;; File
1467 i32 0, ;; Line number
1468 i64 64, ;; Size in Bits
1469 i64 64, ;; Align in Bits
1470 i64 0, ;; Offset in Bits
1479 <!-- ======================================================================= -->
1481 <a name="ccxx_basic_unsigned_long_long">unsigned long long</a>
1486 <div class="doc_code">
1490 metadata !1, ;; Context
1491 metadata !"long long unsigned int",
1492 metadata !1, ;; File
1493 i32 0, ;; Line number
1494 i64 64, ;; Size in Bits
1495 i64 64, ;; Align in Bits
1496 i64 0, ;; Offset in Bits
1505 <!-- ======================================================================= -->
1507 <a name="ccxx_basic_float">float</a>
1512 <div class="doc_code">
1516 metadata !1, ;; Context
1518 metadata !1, ;; File
1519 i32 0, ;; Line number
1520 i64 32, ;; Size in Bits
1521 i64 32, ;; Align in Bits
1522 i64 0, ;; Offset in Bits
1531 <!-- ======================================================================= -->
1533 <a name="ccxx_basic_double">double</a>
1538 <div class="doc_code">
1542 metadata !1, ;; Context
1543 metadata !"double",;; Name
1544 metadata !1, ;; File
1545 i32 0, ;; Line number
1546 i64 64, ;; Size in Bits
1547 i64 64, ;; Align in Bits
1548 i64 0, ;; Offset in Bits
1559 <!-- ======================================================================= -->
1561 <a name="ccxx_derived_types">C/C++ derived types</a>
1566 <p>Given the following as an example of C/C++ derived type:</p>
1568 <div class="doc_code">
1570 typedef const int *IntPtr;
1574 <p>a C/C++ front-end would generate the following descriptors:</p>
1576 <div class="doc_code">
1579 ;; Define the typedef "IntPtr".
1583 metadata !1, ;; Context
1584 metadata !"IntPtr", ;; Name
1585 metadata !3, ;; File
1586 i32 0, ;; Line number
1587 i64 0, ;; Size in bits
1588 i64 0, ;; Align in bits
1589 i64 0, ;; Offset in bits
1591 metadata !4 ;; Derived From type
1595 ;; Define the pointer type.
1599 metadata !1, ;; Context
1600 metadata !"", ;; Name
1601 metadata !1, ;; File
1602 i32 0, ;; Line number
1603 i64 64, ;; Size in bits
1604 i64 64, ;; Align in bits
1605 i64 0, ;; Offset in bits
1607 metadata !5 ;; Derived From type
1610 ;; Define the const type.
1614 metadata !1, ;; Context
1615 metadata !"", ;; Name
1616 metadata !1, ;; File
1617 i32 0, ;; Line number
1618 i64 32, ;; Size in bits
1619 i64 32, ;; Align in bits
1620 i64 0, ;; Offset in bits
1622 metadata !6 ;; Derived From type
1625 ;; Define the int type.
1629 metadata !1, ;; Context
1630 metadata !"int", ;; Name
1631 metadata !1, ;; File
1632 i32 0, ;; Line number
1633 i64 32, ;; Size in bits
1634 i64 32, ;; Align in bits
1635 i64 0, ;; Offset in bits
1644 <!-- ======================================================================= -->
1646 <a name="ccxx_composite_types">C/C++ struct/union types</a>
1651 <p>Given the following as an example of C/C++ struct type:</p>
1653 <div class="doc_code">
1663 <p>a C/C++ front-end would generate the following descriptors:</p>
1665 <div class="doc_code">
1668 ;; Define basic type for unsigned int.
1672 metadata !1, ;; Context
1673 metadata !"unsigned int",
1674 metadata !1, ;; File
1675 i32 0, ;; Line number
1676 i64 32, ;; Size in Bits
1677 i64 32, ;; Align in Bits
1678 i64 0, ;; Offset in Bits
1683 ;; Define composite type for struct Color.
1687 metadata !1, ;; Context
1688 metadata !"Color", ;; Name
1689 metadata !1, ;; Compile unit
1690 i32 1, ;; Line number
1691 i64 96, ;; Size in bits
1692 i64 32, ;; Align in bits
1693 i64 0, ;; Offset in bits
1695 null, ;; Derived From
1696 metadata !3, ;; Elements
1697 i32 0 ;; Runtime Language
1701 ;; Define the Red field.
1705 metadata !1, ;; Context
1706 metadata !"Red", ;; Name
1707 metadata !1, ;; File
1708 i32 2, ;; Line number
1709 i64 32, ;; Size in bits
1710 i64 32, ;; Align in bits
1711 i64 0, ;; Offset in bits
1713 metadata !5 ;; Derived From type
1717 ;; Define the Green field.
1721 metadata !1, ;; Context
1722 metadata !"Green", ;; Name
1723 metadata !1, ;; File
1724 i32 3, ;; Line number
1725 i64 32, ;; Size in bits
1726 i64 32, ;; Align in bits
1727 i64 32, ;; Offset in bits
1729 metadata !5 ;; Derived From type
1733 ;; Define the Blue field.
1737 metadata !1, ;; Context
1738 metadata !"Blue", ;; Name
1739 metadata !1, ;; File
1740 i32 4, ;; Line number
1741 i64 32, ;; Size in bits
1742 i64 32, ;; Align in bits
1743 i64 64, ;; Offset in bits
1745 metadata !5 ;; Derived From type
1749 ;; Define the array of fields used by the composite type Color.
1751 !3 = metadata !{metadata !4, metadata !6, metadata !7}
1757 <!-- ======================================================================= -->
1759 <a name="ccxx_enumeration_types">C/C++ enumeration types</a>
1764 <p>Given the following as an example of C/C++ enumeration type:</p>
1766 <div class="doc_code">
1776 <p>a C/C++ front-end would generate the following descriptors:</p>
1778 <div class="doc_code">
1781 ;; Define composite type for enum Trees
1785 metadata !1, ;; Context
1786 metadata !"Trees", ;; Name
1787 metadata !1, ;; File
1788 i32 1, ;; Line number
1789 i64 32, ;; Size in bits
1790 i64 32, ;; Align in bits
1791 i64 0, ;; Offset in bits
1793 null, ;; Derived From type
1794 metadata !3, ;; Elements
1795 i32 0 ;; Runtime language
1799 ;; Define the array of enumerators used by composite type Trees.
1801 !3 = metadata !{metadata !4, metadata !5, metadata !6}
1804 ;; Define Spruce enumerator.
1806 !4 = metadata !{i32 524328, metadata !"Spruce", i64 100}
1809 ;; Define Oak enumerator.
1811 !5 = metadata !{i32 524328, metadata !"Oak", i64 200}
1814 ;; Define Maple enumerator.
1816 !6 = metadata !{i32 524328, metadata !"Maple", i64 300}
1826 <!-- *********************************************************************** -->
1828 <a name="llvmdwarfextension">Debugging information format</a>
1830 <!-- *********************************************************************** -->
1832 <!-- ======================================================================= -->
1834 <a name="objcproperty">Debugging Information Extension for Objective C
1838 <!-- *********************************************************************** -->
1840 <a name="objcpropertyintroduction">Introduction</a>
1842 <!-- *********************************************************************** -->
1845 <p>Objective C provides a simpler way to declare and define accessor methods
1846 using declared properties. The language provides features to declare a
1847 property and to let compiler synthesize accessor methods.
1850 <p>The debugger lets developer inspect Objective C interfaces and their
1851 instance variables and class variables. However, the debugger does not know
1852 anything about the properties defined in Objective C interfaces. The debugger
1853 consumes information generated by compiler in DWARF format. The format does
1854 not support encoding of Objective C properties. This proposal describes DWARF
1855 extensions to encode Objective C properties, which the debugger can use to let
1856 developers inspect Objective C properties.
1862 <!-- *********************************************************************** -->
1864 <a name="objcpropertyproposal">Proposal</a>
1866 <!-- *********************************************************************** -->
1869 <p>Objective C properties exist separately from class members. A property
1870 can be defined only by "setter" and "getter" selectors, and
1871 be calculated anew on each access. Or a property can just be a direct access
1872 to some declared ivar. Finally it can have an ivar "automatically
1873 synthesized" for it by the compiler, in which case the property can be
1874 referred to in user code directly using the standard C dereference syntax as
1875 well as through the property "dot" syntax, but there is no entry in
1876 the @interface declaration corresponding to this ivar.
1879 To facilitate debugging, these properties we will add a new DWARF TAG into the
1880 DW_TAG_structure_type definition for the class to hold the description of a
1881 given property, and a set of DWARF attributes that provide said description.
1882 The property tag will also contain the name and declared type of the property.
1885 If there is a related ivar, there will also be a DWARF property attribute placed
1886 in the DW_TAG_member DIE for that ivar referring back to the property TAG for
1887 that property. And in the case where the compiler synthesizes the ivar directly,
1888 the compiler is expected to generate a DW_TAG_member for that ivar (with the
1889 DW_AT_artificial set to 1), whose name will be the name used to access this
1890 ivar directly in code, and with the property attribute pointing back to the
1891 property it is backing.
1894 The following examples will serve as illustration for our discussion:
1897 <div class="doc_code">
1909 @synthesize p2 = n2;
1915 This produces the following DWARF (this is a "pseudo dwarfdump" output):
1917 <div class="doc_code">
1919 0x00000100: TAG_structure_type [7] *
1920 AT_APPLE_runtime_class( 0x10 )
1922 AT_decl_file( "Objc_Property.m" )
1925 0x00000110 TAG_APPLE_property
1927 AT_type ( {0x00000150} ( int ) )
1929 0x00000120: TAG_APPLE_property
1931 AT_type ( {0x00000150} ( int ) )
1933 0x00000130: TAG_member [8]
1935 AT_APPLE_property ( {0x00000110} "p1" )
1936 AT_type( {0x00000150} ( int ) )
1937 AT_artificial ( 0x1 )
1939 0x00000140: TAG_member [8]
1941 AT_APPLE_property ( {0x00000120} "p2" )
1942 AT_type( {0x00000150} ( int ) )
1944 0x00000150: AT_type( ( int ) )
1948 <p> Note, the current convention is that the name of the ivar for an
1949 auto-synthesized property is the name of the property from which it derives with
1950 an underscore prepended, as is shown in the example.
1951 But we actually don't need to know this convention, since we are given the name
1952 of the ivar directly.
1956 Also, it is common practice in ObjC to have different property declarations in
1957 the @interface and @implementation - e.g. to provide a read-only property in
1958 the interface,and a read-write interface in the implementation. In that case,
1959 the compiler should emit whichever property declaration will be in force in the
1960 current translation unit.
1963 <p> Developers can decorate a property with attributes which are encoded using
1964 DW_AT_APPLE_property_attribute.
1967 <div class="doc_code">
1969 @property (readonly, nonatomic) int pr;
1973 Which produces a property tag:
1975 <div class="doc_code">
1977 TAG_APPLE_property [8]
1979 AT_type ( {0x00000147} (int) )
1980 AT_APPLE_property_attribute (DW_APPLE_PROPERTY_readonly, DW_APPLE_PROPERTY_nonatomic)
1984 <p> The setter and getter method names are attached to the property using
1985 DW_AT_APPLE_property_setter and DW_AT_APPLE_property_getter attributes.
1987 <div class="doc_code">
1990 @property (setter=myOwnP3Setter:) int p3;
1991 -(void)myOwnP3Setter:(int)a;
1996 -(void)myOwnP3Setter:(int)a{ }
2002 The DWARF for this would be:
2004 <div class="doc_code">
2006 0x000003bd: TAG_structure_type [7] *
2007 AT_APPLE_runtime_class( 0x10 )
2009 AT_decl_file( "Objc_Property.m" )
2012 0x000003cd TAG_APPLE_property
2014 AT_APPLE_property_setter ( "myOwnP3Setter:" )
2015 AT_type( {0x00000147} ( int ) )
2017 0x000003f3: TAG_member [8]
2019 AT_type ( {0x00000147} ( int ) )
2020 AT_APPLE_property ( {0x000003cd} )
2021 AT_artificial ( 0x1 )
2027 <!-- *********************************************************************** -->
2029 <a name="objcpropertynewtags">New DWARF Tags</a>
2031 <!-- *********************************************************************** -->
2034 <table border="1" cellspacing="0">
2036 <th width=200 >TAG</th>
2037 <th width=200 >Value</th>
2040 <td width=200 >DW_TAG_APPLE_property</td>
2041 <td width=200 >0x4200</td>
2047 <!-- *********************************************************************** -->
2049 <a name="objcpropertynewattributes">New DWARF Attributes</a>
2051 <!-- *********************************************************************** -->
2054 <table border="1" cellspacing="0">
2056 <th width=200 >Attribute</th>
2057 <th width=200 >Value</th>
2058 <th width=200 >Classes</th>
2061 <td width=200 >DW_AT_APPLE_property</td>
2062 <td width=200 >0x3fed</td>
2063 <td width=200 >Reference</td>
2066 <td width=200 >DW_AT_APPLE_property_getter</td>
2067 <td width=200 >0x3fe9</td>
2068 <td width=200 >String</td>
2071 <td width=200 >DW_AT_APPLE_property_setter</td>
2072 <td width=200 >0x3fea</td>
2073 <td width=200 >String</td>
2076 <td width=200 >DW_AT_APPLE_property_attribute</td>
2077 <td width=200 >0x3feb</td>
2078 <td width=200 >Constant</td>
2084 <!-- *********************************************************************** -->
2086 <a name="objcpropertynewconstants">New DWARF Constants</a>
2088 <!-- *********************************************************************** -->
2091 <table border="1" cellspacing="0">
2093 <th width=200 >Name</th>
2094 <th width=200 >Value</th>
2097 <td width=200 >DW_AT_APPLE_PROPERTY_readonly</td>
2098 <td width=200 >0x1</td>
2101 <td width=200 >DW_AT_APPLE_PROPERTY_readwrite</td>
2102 <td width=200 >0x2</td>
2105 <td width=200 >DW_AT_APPLE_PROPERTY_assign</td>
2106 <td width=200 >0x4</td>
2109 <td width=200 >DW_AT_APPLE_PROPERTY_retain</td>
2110 <td width=200 >0x8</td>
2113 <td width=200 >DW_AT_APPLE_PROPERTY_copy</td>
2114 <td width=200 >0x10</td>
2117 <td width=200 >DW_AT_APPLE_PROPERTY_nonatomic</td>
2118 <td width=200 >0x20</td>
2127 <!-- ======================================================================= -->
2129 <a name="acceltable">Name Accelerator Tables</a>
2131 <!-- ======================================================================= -->
2132 <!-- ======================================================================= -->
2134 <a name="acceltableintroduction">Introduction</a>
2136 <!-- ======================================================================= -->
2138 <p>The .debug_pubnames and .debug_pubtypes formats are not what a debugger
2139 needs. The "pub" in the section name indicates that the entries in the
2140 table are publicly visible names only. This means no static or hidden
2141 functions show up in the .debug_pubnames. No static variables or private class
2142 variables are in the .debug_pubtypes. Many compilers add different things to
2143 these tables, so we can't rely upon the contents between gcc, icc, or clang.
2145 <p>The typical query given by users tends not to match up with the contents of
2146 these tables. For example, the DWARF spec states that "In the case of the
2147 name of a function member or static data member of a C++ structure, class or
2148 union, the name presented in the .debug_pubnames section is not the simple
2149 name given by the DW_AT_name attribute of the referenced debugging information
2150 entry, but rather the fully qualified name of the data or function member."
2151 So the only names in these tables for complex C++ entries is a fully
2152 qualified name. Debugger users tend not to enter their search strings as
2153 "a::b::c(int,const Foo&) const", but rather as "c", "b::c" , or "a::b::c". So
2154 the name entered in the name table must be demangled in order to chop it up
2155 appropriately and additional names must be manually entered into the table
2156 to make it effective as a name lookup table for debuggers to use.
2158 <p>All debuggers currently ignore the .debug_pubnames table as a result of
2159 its inconsistent and useless public-only name content making it a waste of
2160 space in the object file. These tables, when they are written to disk, are
2161 not sorted in any way, leaving every debugger to do its own parsing
2162 and sorting. These tables also include an inlined copy of the string values
2163 in the table itself making the tables much larger than they need to be on
2164 disk, especially for large C++ programs.
2166 <p>Can't we just fix the sections by adding all of the names we need to this
2167 table? No, because that is not what the tables are defined to contain and we
2168 won't know the difference between the old bad tables and the new good tables.
2169 At best we could make our own renamed sections that contain all of the data
2172 <p>These tables are also insufficient for what a debugger like LLDB needs.
2173 LLDB uses clang for its expression parsing where LLDB acts as a PCH. LLDB is
2174 then often asked to look for type "foo" or namespace "bar", or list items in
2175 namespace "baz". Namespaces are not included in the pubnames or pubtypes
2176 tables. Since clang asks a lot of questions when it is parsing an expression,
2177 we need to be very fast when looking up names, as it happens a lot. Having new
2178 accelerator tables that are optimized for very quick lookups will benefit
2179 this type of debugging experience greatly.
2181 <p>We would like to generate name lookup tables that can be mapped into
2182 memory from disk, and used as is, with little or no up-front parsing. We would
2183 also be able to control the exact content of these different tables so they
2184 contain exactly what we need. The Name Accelerator Tables were designed
2185 to fix these issues. In order to solve these issues we need to:
2187 <li>Have a format that can be mapped into memory from disk and used as is</li>
2188 <li>Lookups should be very fast</li>
2189 <li>Extensible table format so these tables can be made by many producers</li>
2190 <li>Contain all of the names needed for typical lookups out of the box</li>
2191 <li>Strict rules for the contents of tables</li>
2193 <p>Table size is important and the accelerator table format should allow the
2194 reuse of strings from common string tables so the strings for the names are
2195 not duplicated. We also want to make sure the table is ready to be used as-is
2196 by simply mapping the table into memory with minimal header parsing.
2198 <p>The name lookups need to be fast and optimized for the kinds of lookups
2199 that debuggers tend to do. Optimally we would like to touch as few parts of
2200 the mapped table as possible when doing a name lookup and be able to quickly
2201 find the name entry we are looking for, or discover there are no matches. In
2202 the case of debuggers we optimized for lookups that fail most of the time.
2204 <p>Each table that is defined should have strict rules on exactly what is in
2205 the accelerator tables and documented so clients can rely on the content.
2207 <!-- ======================================================================= -->
2209 <a name="acceltablehashes">Hash Tables</a>
2211 <!-- ======================================================================= -->
2213 <h5>Standard Hash Tables</h5>
2214 <p>Typical hash tables have a header, buckets, and each bucket points to the
2216 <div class="doc_code">
2227 <p>The BUCKETS are an array of offsets to DATA for each hash:
2228 <div class="doc_code">
2231 | 0x00001000 | BUCKETS[0]
2232 | 0x00002000 | BUCKETS[1]
2233 | 0x00002200 | BUCKETS[2]
2234 | 0x000034f0 | BUCKETS[3]
2236 | 0xXXXXXXXX | BUCKETS[n_buckets]
2240 <p>So for bucket[3] in the example above, we have an offset into the table
2241 0x000034f0 which points to a chain of entries for the bucket. Each bucket
2242 must contain a next pointer, full 32 bit hash value, the string itself,
2243 and the data for the current string value.
2244 <div class="doc_code">
2247 0x000034f0: | 0x00003500 | next pointer
2248 | 0x12345678 | 32 bit hash
2249 | "erase" | string value
2250 | data[n] | HashData for this bucket
2252 0x00003500: | 0x00003550 | next pointer
2253 | 0x29273623 | 32 bit hash
2254 | "dump" | string value
2255 | data[n] | HashData for this bucket
2257 0x00003550: | 0x00000000 | next pointer
2258 | 0x82638293 | 32 bit hash
2259 | "main" | string value
2260 | data[n] | HashData for this bucket
2264 <p>The problem with this layout for debuggers is that we need to optimize for
2265 the negative lookup case where the symbol we're searching for is not present.
2266 So if we were to lookup "printf" in the table above, we would make a 32 hash
2267 for "printf", it might match bucket[3]. We would need to go to the offset
2268 0x000034f0 and start looking to see if our 32 bit hash matches. To do so, we
2269 need to read the next pointer, then read the hash, compare it, and skip to
2270 the next bucket. Each time we are skipping many bytes in memory and touching
2271 new cache pages just to do the compare on the full 32 bit hash. All of these
2272 accesses then tell us that we didn't have a match.
2274 <h5>Name Hash Tables</h5>
2276 <p>To solve the issues mentioned above we have structured the hash tables
2277 a bit differently: a header, buckets, an array of all unique 32 bit hash
2278 values, followed by an array of hash value data offsets, one for each hash
2279 value, then the data for all hash values:
2280 <div class="doc_code">
2295 <p>The BUCKETS in the name tables are an index into the HASHES array. By
2296 making all of the full 32 bit hash values contiguous in memory, we allow
2297 ourselves to efficiently check for a match while touching as little
2298 memory as possible. Most often checking the 32 bit hash values is as far as
2299 the lookup goes. If it does match, it usually is a match with no collisions.
2300 So for a table with "n_buckets" buckets, and "n_hashes" unique 32 bit hash
2301 values, we can clarify the contents of the BUCKETS, HASHES and OFFSETS as:
2302 <div class="doc_code">
2304 .-------------------------.
2305 | HEADER.magic | uint32_t
2306 | HEADER.version | uint16_t
2307 | HEADER.hash_function | uint16_t
2308 | HEADER.bucket_count | uint32_t
2309 | HEADER.hashes_count | uint32_t
2310 | HEADER.header_data_len | uint32_t
2311 | HEADER_DATA | HeaderData
2312 |-------------------------|
2313 | BUCKETS | uint32_t[n_buckets] // 32 bit hash indexes
2314 |-------------------------|
2315 | HASHES | uint32_t[n_buckets] // 32 bit hash values
2316 |-------------------------|
2317 | OFFSETS | uint32_t[n_buckets] // 32 bit offsets to hash value data
2318 |-------------------------|
2320 `-------------------------'
2323 <p>So taking the exact same data from the standard hash example above we end up
2325 <div class="doc_code">
2335 | ... | BUCKETS[n_buckets]
2337 | 0x........ | HASHES[0]
2338 | 0x........ | HASHES[1]
2339 | 0x........ | HASHES[2]
2340 | 0x........ | HASHES[3]
2341 | 0x........ | HASHES[4]
2342 | 0x........ | HASHES[5]
2343 | 0x12345678 | HASHES[6] hash for BUCKETS[3]
2344 | 0x29273623 | HASHES[7] hash for BUCKETS[3]
2345 | 0x82638293 | HASHES[8] hash for BUCKETS[3]
2346 | 0x........ | HASHES[9]
2347 | 0x........ | HASHES[10]
2348 | 0x........ | HASHES[11]
2349 | 0x........ | HASHES[12]
2350 | 0x........ | HASHES[13]
2351 | 0x........ | HASHES[n_hashes]
2353 | 0x........ | OFFSETS[0]
2354 | 0x........ | OFFSETS[1]
2355 | 0x........ | OFFSETS[2]
2356 | 0x........ | OFFSETS[3]
2357 | 0x........ | OFFSETS[4]
2358 | 0x........ | OFFSETS[5]
2359 | 0x000034f0 | OFFSETS[6] offset for BUCKETS[3]
2360 | 0x00003500 | OFFSETS[7] offset for BUCKETS[3]
2361 | 0x00003550 | OFFSETS[8] offset for BUCKETS[3]
2362 | 0x........ | OFFSETS[9]
2363 | 0x........ | OFFSETS[10]
2364 | 0x........ | OFFSETS[11]
2365 | 0x........ | OFFSETS[12]
2366 | 0x........ | OFFSETS[13]
2367 | 0x........ | OFFSETS[n_hashes]
2375 0x000034f0: | 0x00001203 | .debug_str ("erase")
2376 | 0x00000004 | A 32 bit array count - number of HashData with name "erase"
2377 | 0x........ | HashData[0]
2378 | 0x........ | HashData[1]
2379 | 0x........ | HashData[2]
2380 | 0x........ | HashData[3]
2381 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2383 0x00003500: | 0x00001203 | String offset into .debug_str ("collision")
2384 | 0x00000002 | A 32 bit array count - number of HashData with name "collision"
2385 | 0x........ | HashData[0]
2386 | 0x........ | HashData[1]
2387 | 0x00001203 | String offset into .debug_str ("dump")
2388 | 0x00000003 | A 32 bit array count - number of HashData with name "dump"
2389 | 0x........ | HashData[0]
2390 | 0x........ | HashData[1]
2391 | 0x........ | HashData[2]
2392 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2394 0x00003550: | 0x00001203 | String offset into .debug_str ("main")
2395 | 0x00000009 | A 32 bit array count - number of HashData with name "main"
2396 | 0x........ | HashData[0]
2397 | 0x........ | HashData[1]
2398 | 0x........ | HashData[2]
2399 | 0x........ | HashData[3]
2400 | 0x........ | HashData[4]
2401 | 0x........ | HashData[5]
2402 | 0x........ | HashData[6]
2403 | 0x........ | HashData[7]
2404 | 0x........ | HashData[8]
2405 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2409 <p>So we still have all of the same data, we just organize it more efficiently
2410 for debugger lookup. If we repeat the same "printf" lookup from above, we
2411 would hash "printf" and find it matches BUCKETS[3] by taking the 32 bit hash
2412 value and modulo it by n_buckets. BUCKETS[3] contains "6" which is the index
2413 into the HASHES table. We would then compare any consecutive 32 bit hashes
2414 values in the HASHES array as long as the hashes would be in BUCKETS[3]. We
2415 do this by verifying that each subsequent hash value modulo n_buckets is still
2416 3. In the case of a failed lookup we would access the memory for BUCKETS[3], and
2417 then compare a few consecutive 32 bit hashes before we know that we have no match.
2418 We don't end up marching through multiple words of memory and we really keep the
2419 number of processor data cache lines being accessed as small as possible.
2421 <p>The string hash that is used for these lookup tables is the Daniel J.
2422 Bernstein hash which is also used in the ELF GNU_HASH sections. It is a very
2423 good hash for all kinds of names in programs with very few hash collisions.
2425 <p>Empty buckets are designated by using an invalid hash index of UINT32_MAX.
2427 <!-- ======================================================================= -->
2429 <a name="acceltabledetails">Details</a>
2431 <!-- ======================================================================= -->
2433 <p>These name hash tables are designed to be generic where specializations of
2434 the table get to define additional data that goes into the header
2435 ("HeaderData"), how the string value is stored ("KeyType") and the content
2436 of the data for each hash value.
2438 <h5>Header Layout</h5>
2439 <p>The header has a fixed part, and the specialized part. The exact format of
2441 <div class="doc_code">
2445 uint32_t magic; // 'HASH' magic value to allow endian detection
2446 uint16_t version; // Version number
2447 uint16_t hash_function; // The hash function enumeration that was used
2448 uint32_t bucket_count; // The number of buckets in this hash table
2449 uint32_t hashes_count; // The total number of unique hash values and hash data offsets in this table
2450 uint32_t header_data_len; // The bytes to skip to get to the hash indexes (buckets) for correct alignment
2451 // Specifically the length of the following HeaderData field - this does not
2452 // include the size of the preceding fields
2453 HeaderData header_data; // Implementation specific header data
2457 <p>The header starts with a 32 bit "magic" value which must be 'HASH' encoded as
2458 an ASCII integer. This allows the detection of the start of the hash table and
2459 also allows the table's byte order to be determined so the table can be
2460 correctly extracted. The "magic" value is followed by a 16 bit version number
2461 which allows the table to be revised and modified in the future. The current
2462 version number is 1. "hash_function" is a uint16_t enumeration that specifies
2463 which hash function was used to produce this table. The current values for the
2464 hash function enumerations include:
2465 <div class="doc_code">
2467 enum HashFunctionType
2469 eHashFunctionDJB = 0u, // Daniel J Bernstein hash function
2473 <p>"bucket_count" is a 32 bit unsigned integer that represents how many buckets
2474 are in the BUCKETS array. "hashes_count" is the number of unique 32 bit hash
2475 values that are in the HASHES array, and is the same number of offsets are
2476 contained in the OFFSETS array. "header_data_len" specifies the size in
2477 bytes of the HeaderData that is filled in by specialized versions of this
2480 <h5>Fixed Lookup</h5>
2481 <p>The header is followed by the buckets, hashes, offsets, and hash value
2483 <div class="doc_code">
2487 uint32_t buckets[Header.bucket_count]; // An array of hash indexes into the "hashes[]" array below
2488 uint32_t hashes [Header.hashes_count]; // Every unique 32 bit hash for the entire table is in this table
2489 uint32_t offsets[Header.hashes_count]; // An offset that corresponds to each item in the "hashes[]" array above
2493 <p>"buckets" is an array of 32 bit indexes into the "hashes" array. The
2494 "hashes" array contains all of the 32 bit hash values for all names in the
2495 hash table. Each hash in the "hashes" table has an offset in the "offsets"
2496 array that points to the data for the hash value.
2498 <p>This table setup makes it very easy to repurpose these tables to contain
2499 different data, while keeping the lookup mechanism the same for all tables.
2500 This layout also makes it possible to save the table to disk and map it in
2501 later and do very efficient name lookups with little or no parsing.
2503 <p>DWARF lookup tables can be implemented in a variety of ways and can store
2504 a lot of information for each name. We want to make the DWARF tables
2505 extensible and able to store the data efficiently so we have used some of the
2506 DWARF features that enable efficient data storage to define exactly what kind
2507 of data we store for each name.
2509 <p>The "HeaderData" contains a definition of the contents of each HashData
2510 chunk. We might want to store an offset to all of the debug information
2511 entries (DIEs) for each name. To keep things extensible, we create a list of
2512 items, or Atoms, that are contained in the data for each name. First comes the
2513 type of the data in each atom:
2514 <div class="doc_code">
2519 eAtomTypeDIEOffset = 1u, // DIE offset, check form for encoding
2520 eAtomTypeCUOffset = 2u, // DIE offset of the compiler unit header that contains the item in question
2521 eAtomTypeTag = 3u, // DW_TAG_xxx value, should be encoded as DW_FORM_data1 (if no tags exceed 255) or DW_FORM_data2
2522 eAtomTypeNameFlags = 4u, // Flags from enum NameFlags
2523 eAtomTypeTypeFlags = 5u, // Flags from enum TypeFlags
2527 <p>The enumeration values and their meanings are:
2528 <div class="doc_code">
2530 eAtomTypeNULL - a termination atom that specifies the end of the atom list
2531 eAtomTypeDIEOffset - an offset into the .debug_info section for the DWARF DIE for this name
2532 eAtomTypeCUOffset - an offset into the .debug_info section for the CU that contains the DIE
2533 eAtomTypeDIETag - The DW_TAG_XXX enumeration value so you don't have to parse the DWARF to see what it is
2534 eAtomTypeNameFlags - Flags for functions and global variables (isFunction, isInlined, isExternal...)
2535 eAtomTypeTypeFlags - Flags for types (isCXXClass, isObjCClass, ...)
2538 <p>Then we allow each atom type to define the atom type and how the data for
2539 each atom type data is encoded:
2540 <div class="doc_code">
2544 uint16_t type; // AtomType enum value
2545 uint16_t form; // DWARF DW_FORM_XXX defines
2549 <p>The "form" type above is from the DWARF specification and defines the
2550 exact encoding of the data for the Atom type. See the DWARF specification for
2551 the DW_FORM_ definitions.
2552 <div class="doc_code">
2556 uint32_t die_offset_base;
2557 uint32_t atom_count;
2558 Atoms atoms[atom_count0];
2562 <p>"HeaderData" defines the base DIE offset that should be added to any atoms
2563 that are encoded using the DW_FORM_ref1, DW_FORM_ref2, DW_FORM_ref4,
2564 DW_FORM_ref8 or DW_FORM_ref_udata. It also defines what is contained in
2565 each "HashData" object -- Atom.form tells us how large each field will be in
2566 the HashData and the Atom.type tells us how this data should be interpreted.
2568 <p>For the current implementations of the ".apple_names" (all functions + globals),
2569 the ".apple_types" (names of all types that are defined), and the
2570 ".apple_namespaces" (all namespaces), we currently set the Atom array to be:
2571 <div class="doc_code">
2573 HeaderData.atom_count = 1;
2574 HeaderData.atoms[0].type = eAtomTypeDIEOffset;
2575 HeaderData.atoms[0].form = DW_FORM_data4;
2578 <p>This defines the contents to be the DIE offset (eAtomTypeDIEOffset) that is
2579 encoded as a 32 bit value (DW_FORM_data4). This allows a single name to have
2580 multiple matching DIEs in a single file, which could come up with an inlined
2581 function for instance. Future tables could include more information about the
2582 DIE such as flags indicating if the DIE is a function, method, block,
2585 <p>The KeyType for the DWARF table is a 32 bit string table offset into the
2586 ".debug_str" table. The ".debug_str" is the string table for the DWARF which
2587 may already contain copies of all of the strings. This helps make sure, with
2588 help from the compiler, that we reuse the strings between all of the DWARF
2589 sections and keeps the hash table size down. Another benefit to having the
2590 compiler generate all strings as DW_FORM_strp in the debug info, is that
2591 DWARF parsing can be made much faster.
2593 <p>After a lookup is made, we get an offset into the hash data. The hash data
2594 needs to be able to deal with 32 bit hash collisions, so the chunk of data
2595 at the offset in the hash data consists of a triple:
2596 <div class="doc_code">
2599 uint32_t hash_data_count
2600 HashData[hash_data_count]
2603 <p>If "str_offset" is zero, then the bucket contents are done. 99.9% of the
2604 hash data chunks contain a single item (no 32 bit hash collision):
2605 <div class="doc_code">
2608 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2609 | 0x00000004 | uint32_t HashData count
2610 | 0x........ | uint32_t HashData[0] DIE offset
2611 | 0x........ | uint32_t HashData[1] DIE offset
2612 | 0x........ | uint32_t HashData[2] DIE offset
2613 | 0x........ | uint32_t HashData[3] DIE offset
2614 | 0x00000000 | uint32_t KeyType (end of hash chain)
2618 <p>If there are collisions, you will have multiple valid string offsets:
2619 <div class="doc_code">
2622 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2623 | 0x00000004 | uint32_t HashData count
2624 | 0x........ | uint32_t HashData[0] DIE offset
2625 | 0x........ | uint32_t HashData[1] DIE offset
2626 | 0x........ | uint32_t HashData[2] DIE offset
2627 | 0x........ | uint32_t HashData[3] DIE offset
2628 | 0x00002023 | uint32_t KeyType (.debug_str[0x0002023] => "print")
2629 | 0x00000002 | uint32_t HashData count
2630 | 0x........ | uint32_t HashData[0] DIE offset
2631 | 0x........ | uint32_t HashData[1] DIE offset
2632 | 0x00000000 | uint32_t KeyType (end of hash chain)
2636 <p>Current testing with real world C++ binaries has shown that there is around 1
2637 32 bit hash collision per 100,000 name entries.
2639 <!-- ======================================================================= -->
2641 <a name="acceltablecontents">Contents</a>
2643 <!-- ======================================================================= -->
2645 <p>As we said, we want to strictly define exactly what is included in the
2646 different tables. For DWARF, we have 3 tables: ".apple_names", ".apple_types",
2647 and ".apple_namespaces".
2649 <p>".apple_names" sections should contain an entry for each DWARF DIE whose
2650 DW_TAG is a DW_TAG_label, DW_TAG_inlined_subroutine, or DW_TAG_subprogram that
2651 has address attributes: DW_AT_low_pc, DW_AT_high_pc, DW_AT_ranges or
2652 DW_AT_entry_pc. It also contains DW_TAG_variable DIEs that have a DW_OP_addr
2653 in the location (global and static variables). All global and static variables
2654 should be included, including those scoped withing functions and classes. For
2655 example using the following code:
2656 <div class="doc_code">
2666 <p>Both of the static "var" variables would be included in the table. All
2667 functions should emit both their full names and their basenames. For C or C++,
2668 the full name is the mangled name (if available) which is usually in the
2669 DW_AT_MIPS_linkage_name attribute, and the DW_AT_name contains the function
2670 basename. If global or static variables have a mangled name in a
2671 DW_AT_MIPS_linkage_name attribute, this should be emitted along with the
2672 simple name found in the DW_AT_name attribute.
2674 <p>".apple_types" sections should contain an entry for each DWARF DIE whose
2677 <li>DW_TAG_array_type</li>
2678 <li>DW_TAG_class_type</li>
2679 <li>DW_TAG_enumeration_type</li>
2680 <li>DW_TAG_pointer_type</li>
2681 <li>DW_TAG_reference_type</li>
2682 <li>DW_TAG_string_type</li>
2683 <li>DW_TAG_structure_type</li>
2684 <li>DW_TAG_subroutine_type</li>
2685 <li>DW_TAG_typedef</li>
2686 <li>DW_TAG_union_type</li>
2687 <li>DW_TAG_ptr_to_member_type</li>
2688 <li>DW_TAG_set_type</li>
2689 <li>DW_TAG_subrange_type</li>
2690 <li>DW_TAG_base_type</li>
2691 <li>DW_TAG_const_type</li>
2692 <li>DW_TAG_constant</li>
2693 <li>DW_TAG_file_type</li>
2694 <li>DW_TAG_namelist</li>
2695 <li>DW_TAG_packed_type</li>
2696 <li>DW_TAG_volatile_type</li>
2697 <li>DW_TAG_restrict_type</li>
2698 <li>DW_TAG_interface_type</li>
2699 <li>DW_TAG_unspecified_type</li>
2700 <li>DW_TAG_shared_type</li>
2702 <p>Only entries with a DW_AT_name attribute are included, and the entry must
2703 not be a forward declaration (DW_AT_declaration attribute with a non-zero value).
2704 For example, using the following code:
2705 <div class="doc_code">
2714 <p>We get a few type DIEs:
2715 <div class="doc_code">
2717 0x00000067: TAG_base_type [5]
2718 AT_encoding( DW_ATE_signed )
2720 AT_byte_size( 0x04 )
2722 0x0000006e: TAG_pointer_type [6]
2723 AT_type( {0x00000067} ( int ) )
2724 AT_byte_size( 0x08 )
2727 <p>The DW_TAG_pointer_type is not included because it does not have a DW_AT_name.
2729 <p>".apple_namespaces" section should contain all DW_TAG_namespace DIEs. If
2730 we run into a namespace that has no name this is an anonymous namespace,
2731 and the name should be output as "(anonymous namespace)" (without the quotes).
2732 Why? This matches the output of the abi::cxa_demangle() that is in the standard
2733 C++ library that demangles mangled names.
2736 <!-- ======================================================================= -->
2738 <a name="acceltableextensions">Language Extensions and File Format Changes</a>
2740 <!-- ======================================================================= -->
2742 <h5>Objective-C Extensions</h5>
2743 <p>".apple_objc" section should contain all DW_TAG_subprogram DIEs for an
2744 Objective-C class. The name used in the hash table is the name of the
2745 Objective-C class itself. If the Objective-C class has a category, then an
2746 entry is made for both the class name without the category, and for the class
2747 name with the category. So if we have a DIE at offset 0x1234 with a name
2748 of method "-[NSString(my_additions) stringWithSpecialString:]", we would add
2749 an entry for "NSString" that points to DIE 0x1234, and an entry for
2750 "NSString(my_additions)" that points to 0x1234. This allows us to quickly
2751 track down all Objective-C methods for an Objective-C class when doing
2752 expressions. It is needed because of the dynamic nature of Objective-C where
2753 anyone can add methods to a class. The DWARF for Objective-C methods is also
2754 emitted differently from C++ classes where the methods are not usually
2755 contained in the class definition, they are scattered about across one or more
2756 compile units. Categories can also be defined in different shared libraries.
2757 So we need to be able to quickly find all of the methods and class functions
2758 given the Objective-C class name, or quickly find all methods and class
2759 functions for a class + category name. This table does not contain any selector
2760 names, it just maps Objective-C class names (or class names + category) to all
2761 of the methods and class functions. The selectors are added as function
2762 basenames in the .debug_names section.
2764 <p>In the ".apple_names" section for Objective-C functions, the full name is the
2765 entire function name with the brackets ("-[NSString stringWithCString:]") and the
2766 basename is the selector only ("stringWithCString:").
2768 <h5>Mach-O Changes</h5>
2769 <p>The sections names for the apple hash tables are for non mach-o files. For
2770 mach-o files, the sections should be contained in the "__DWARF" segment with
2773 <li>".apple_names" -> "__apple_names"</li>
2774 <li>".apple_types" -> "__apple_types"</li>
2775 <li>".apple_namespaces" -> "__apple_namespac" (16 character limit)</li>
2776 <li> ".apple_objc" -> "__apple_objc"</li>
2781 <!-- *********************************************************************** -->
2785 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2786 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2787 <a href="http://validator.w3.org/check/referer"><img
2788 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2790 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2791 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
2792 Last modified: $Date$