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, ;; Line number where the scope of the subprogram begins
456 i32, ;; Virtuality, e.g. dwarf::DW_VIRTUALITY__virtual
457 i32, ;; Index into a virtual function
458 metadata, ;; indicates which base type contains the vtable pointer for the
460 i32, ;; Flags - Artifical, Private, Protected, Explicit, Prototyped.
462 Function *,;; Pointer to LLVM function
463 metadata, ;; Lists function template parameters
464 metadata ;; Function declaration descriptor
465 metadata ;; List of function variables
470 <p>These descriptors provide debug information about functions, methods and
471 subprograms. They provide details such as name, return types and the source
472 location where the subprogram is defined.
477 <!-- ======================================================================= -->
479 <a name="format_blocks">Block descriptors</a>
484 <div class="doc_code">
487 i32, ;; Tag = 11 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_lexical_block)
488 metadata,;; Reference to context descriptor
490 i32, ;; Column number
491 metadata,;; Reference to source file
492 i32 ;; Unique ID to identify blocks from a template function
497 <p>This descriptor provides debug information about nested blocks within a
498 subprogram. The line number and column numbers are used to dinstinguish
499 two lexical blocks at same depth. </p>
501 <div class="doc_code">
504 i32, ;; Tag = 11 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_lexical_block)
505 metadata ;; Reference to the scope we're annotating with a file change
506 metadata,;; Reference to the file the scope is enclosed in.
511 <p>This descriptor provides a wrapper around a lexical scope to handle file
512 changes in the middle of a lexical block.</p>
516 <!-- ======================================================================= -->
518 <a name="format_basic_type">Basic type descriptors</a>
523 <div class="doc_code">
526 i32, ;; Tag = 36 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
527 ;; (DW_TAG_base_type)
528 metadata, ;; Reference to context
529 metadata, ;; Name (may be "" for anonymous types)
530 metadata, ;; Reference to file where defined (may be NULL)
531 i32, ;; Line number where defined (may be 0)
533 i64, ;; Alignment in bits
534 i64, ;; Offset in bits
536 i32 ;; DWARF type encoding
541 <p>These descriptors define primitive types used in the code. Example int, bool
542 and float. The context provides the scope of the type, which is usually the
543 top level. Since basic types are not usually user defined the context
544 and line number can be left as NULL and 0. The size, alignment and offset
545 are expressed in bits and can be 64 bit values. The alignment is used to
546 round the offset when embedded in a
547 <a href="#format_composite_type">composite type</a> (example to keep float
548 doubles on 64 bit boundaries.) The offset is the bit offset if embedded in
549 a <a href="#format_composite_type">composite type</a>.</p>
551 <p>The type encoding provides the details of the type. The values are typically
552 one of the following:</p>
554 <div class="doc_code">
560 DW_ATE_signed_char = 6
562 DW_ATE_unsigned_char = 8
568 <!-- ======================================================================= -->
570 <a name="format_derived_type">Derived type descriptors</a>
575 <div class="doc_code">
578 i32, ;; Tag (see below)
579 metadata, ;; Reference to context
580 metadata, ;; Name (may be "" for anonymous types)
581 metadata, ;; Reference to file where defined (may be NULL)
582 i32, ;; Line number where defined (may be 0)
584 i64, ;; Alignment in bits
585 i64, ;; Offset in bits
586 i32, ;; Flags to encode attributes, e.g. private
587 metadata, ;; Reference to type derived from
588 metadata, ;; (optional) Name of the Objective C property associated with
589 ;; Objective-C an ivar
590 metadata, ;; (optional) Name of the Objective C property getter selector.
591 metadata, ;; (optional) Name of the Objective C property setter selector.
592 i32 ;; (optional) Objective C property attributes.
597 <p>These descriptors are used to define types derived from other types. The
598 value of the tag varies depending on the meaning. The following are possible
601 <div class="doc_code">
603 DW_TAG_formal_parameter = 5
605 DW_TAG_pointer_type = 15
606 DW_TAG_reference_type = 16
608 DW_TAG_const_type = 38
609 DW_TAG_volatile_type = 53
610 DW_TAG_restrict_type = 55
614 <p><tt>DW_TAG_member</tt> is used to define a member of
615 a <a href="#format_composite_type">composite type</a>
616 or <a href="#format_subprograms">subprogram</a>. The type of the member is
617 the <a href="#format_derived_type">derived
618 type</a>. <tt>DW_TAG_formal_parameter</tt> is used to define a member which
619 is a formal argument of a subprogram.</p>
621 <p><tt>DW_TAG_typedef</tt> is used to provide a name for the derived type.</p>
623 <p><tt>DW_TAG_pointer_type</tt>, <tt>DW_TAG_reference_type</tt>,
624 <tt>DW_TAG_const_type</tt>, <tt>DW_TAG_volatile_type</tt> and
625 <tt>DW_TAG_restrict_type</tt> are used to qualify
626 the <a href="#format_derived_type">derived type</a>. </p>
628 <p><a href="#format_derived_type">Derived type</a> location can be determined
629 from the context and line number. The size, alignment and offset are
630 expressed in bits and can be 64 bit values. The alignment is used to round
631 the offset when embedded in a <a href="#format_composite_type">composite
632 type</a> (example to keep float doubles on 64 bit boundaries.) The offset is
633 the bit offset if embedded in a <a href="#format_composite_type">composite
636 <p>Note that the <tt>void *</tt> type is expressed as a type derived from NULL.
641 <!-- ======================================================================= -->
643 <a name="format_composite_type">Composite type descriptors</a>
648 <div class="doc_code">
651 i32, ;; Tag (see below)
652 metadata, ;; Reference to context
653 metadata, ;; Name (may be "" for anonymous types)
654 metadata, ;; Reference to file where defined (may be NULL)
655 i32, ;; Line number where defined (may be 0)
657 i64, ;; Alignment in bits
658 i64, ;; Offset in bits
660 metadata, ;; Reference to type derived from
661 metadata, ;; Reference to array of member descriptors
662 i32 ;; Runtime languages
667 <p>These descriptors are used to define types that are composed of 0 or more
668 elements. The value of the tag varies depending on the meaning. The following
669 are possible tag values:</p>
671 <div class="doc_code">
673 DW_TAG_array_type = 1
674 DW_TAG_enumeration_type = 4
675 DW_TAG_structure_type = 19
676 DW_TAG_union_type = 23
677 DW_TAG_vector_type = 259
678 DW_TAG_subroutine_type = 21
679 DW_TAG_inheritance = 28
683 <p>The vector flag indicates that an array type is a native packed vector.</p>
685 <p>The members of array types (tag = <tt>DW_TAG_array_type</tt>) or vector types
686 (tag = <tt>DW_TAG_vector_type</tt>) are <a href="#format_subrange">subrange
687 descriptors</a>, each representing the range of subscripts at that level of
690 <p>The members of enumeration types (tag = <tt>DW_TAG_enumeration_type</tt>) are
691 <a href="#format_enumeration">enumerator descriptors</a>, each representing
692 the definition of enumeration value for the set. All enumeration type
693 descriptors are collected inside the named metadata
694 <tt>!llvm.dbg.cu</tt>.</p>
696 <p>The members of structure (tag = <tt>DW_TAG_structure_type</tt>) or union (tag
697 = <tt>DW_TAG_union_type</tt>) types are any one of
698 the <a href="#format_basic_type">basic</a>,
699 <a href="#format_derived_type">derived</a>
700 or <a href="#format_composite_type">composite</a> type descriptors, each
701 representing a field member of the structure or union.</p>
703 <p>For C++ classes (tag = <tt>DW_TAG_structure_type</tt>), member descriptors
704 provide information about base classes, static members and member
705 functions. If a member is a <a href="#format_derived_type">derived type
706 descriptor</a> and has a tag of <tt>DW_TAG_inheritance</tt>, then the type
707 represents a base class. If the member of is
708 a <a href="#format_global_variables">global variable descriptor</a> then it
709 represents a static member. And, if the member is
710 a <a href="#format_subprograms">subprogram descriptor</a> then it represents
711 a member function. For static members and member
712 functions, <tt>getName()</tt> returns the members link or the C++ mangled
713 name. <tt>getDisplayName()</tt> the simplied version of the name.</p>
715 <p>The first member of subroutine (tag = <tt>DW_TAG_subroutine_type</tt>) type
716 elements is the return type for the subroutine. The remaining elements are
717 the formal arguments to the subroutine.</p>
719 <p><a href="#format_composite_type">Composite type</a> location can be
720 determined from the context and line number. The size, alignment and
721 offset are expressed in bits and can be 64 bit values. The alignment is used
722 to round the offset when embedded in
723 a <a href="#format_composite_type">composite type</a> (as an example, to keep
724 float doubles on 64 bit boundaries.) The offset is the bit offset if embedded
725 in a <a href="#format_composite_type">composite type</a>.</p>
729 <!-- ======================================================================= -->
731 <a name="format_subrange">Subrange descriptors</a>
736 <div class="doc_code">
739 i32, ;; Tag = 33 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> (DW_TAG_subrange_type)
746 <p>These descriptors are used to define ranges of array subscripts for an array
747 <a href="#format_composite_type">composite type</a>. The low value defines
748 the lower bounds typically zero for C/C++. The high value is the upper
749 bounds. Values are 64 bit. High - low + 1 is the size of the array. If low
750 > high the array bounds are not included in generated debugging information.
755 <!-- ======================================================================= -->
757 <a name="format_enumeration">Enumerator descriptors</a>
762 <div class="doc_code">
765 i32, ;; Tag = 40 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a>
766 ;; (DW_TAG_enumerator)
773 <p>These descriptors are used to define members of an
774 enumeration <a href="#format_composite_type">composite type</a>, it
775 associates the name to the value.</p>
779 <!-- ======================================================================= -->
781 <a name="format_variables">Local variables</a>
786 <div class="doc_code">
789 i32, ;; Tag (see below)
792 metadata, ;; Reference to file where defined
793 i32, ;; 24 bit - Line number where defined
794 ;; 8 bit - Argument number. 1 indicates 1st argument.
795 metadata, ;; Type descriptor
797 metadata ;; (optional) Reference to inline location
802 <p>These descriptors are used to define variables local to a sub program. The
803 value of the tag depends on the usage of the variable:</p>
805 <div class="doc_code">
807 DW_TAG_auto_variable = 256
808 DW_TAG_arg_variable = 257
809 DW_TAG_return_variable = 258
813 <p>An auto variable is any variable declared in the body of the function. An
814 argument variable is any variable that appears as a formal argument to the
815 function. A return variable is used to track the result of a function and
816 has no source correspondent.</p>
818 <p>The context is either the subprogram or block where the variable is defined.
819 Name the source variable name. Context and line indicate where the
820 variable was defined. Type descriptor defines the declared type of the
827 <!-- ======================================================================= -->
829 <a name="format_common_intrinsics">Debugger intrinsic functions</a>
834 <p>LLVM uses several intrinsic functions (name prefixed with "llvm.dbg") to
835 provide debug information at various points in generated code.</p>
837 <!-- ======================================================================= -->
839 <a name="format_common_declare">llvm.dbg.declare</a>
844 void %<a href="#format_common_declare">llvm.dbg.declare</a>(metadata, metadata)
847 <p>This intrinsic provides information about a local element (e.g., variable). The
848 first argument is metadata holding the alloca for the variable. The
849 second argument is metadata containing a description of the variable.</p>
852 <!-- ======================================================================= -->
854 <a name="format_common_value">llvm.dbg.value</a>
859 void %<a href="#format_common_value">llvm.dbg.value</a>(metadata, i64, metadata)
862 <p>This intrinsic provides information when a user source variable is set to a
863 new value. The first argument is the new value (wrapped as metadata). The
864 second argument is the offset in the user source variable where the new value
865 is written. The third argument is metadata containing a description of the
866 user source variable.</p>
871 <!-- ======================================================================= -->
873 <a name="format_common_lifetime">Object lifetimes and scoping</a>
877 <p>In many languages, the local variables in functions can have their lifetimes
878 or scopes limited to a subset of a function. In the C family of languages,
879 for example, variables are only live (readable and writable) within the
880 source block that they are defined in. In functional languages, values are
881 only readable after they have been defined. Though this is a very obvious
882 concept, it is non-trivial to model in LLVM, because it has no notion of
883 scoping in this sense, and does not want to be tied to a language's scoping
886 <p>In order to handle this, the LLVM debug format uses the metadata attached to
887 llvm instructions to encode line number and scoping information. Consider
888 the following C fragment, for example:</p>
890 <div class="doc_code">
904 <p>Compiled to LLVM, this function would be represented like this:</p>
906 <div class="doc_code">
908 define void @foo() nounwind ssp {
910 %X = alloca i32, align 4 ; <i32*> [#uses=4]
911 %Y = alloca i32, align 4 ; <i32*> [#uses=4]
912 %Z = alloca i32, align 4 ; <i32*> [#uses=3]
913 %0 = bitcast i32* %X to {}* ; <{}*> [#uses=1]
914 call void @llvm.dbg.declare(metadata !{i32 * %X}, metadata !0), !dbg !7
915 store i32 21, i32* %X, !dbg !8
916 %1 = bitcast i32* %Y to {}* ; <{}*> [#uses=1]
917 call void @llvm.dbg.declare(metadata !{i32 * %Y}, metadata !9), !dbg !10
918 store i32 22, i32* %Y, !dbg !11
919 %2 = bitcast i32* %Z to {}* ; <{}*> [#uses=1]
920 call void @llvm.dbg.declare(metadata !{i32 * %Z}, metadata !12), !dbg !14
921 store i32 23, i32* %Z, !dbg !15
922 %tmp = load i32* %X, !dbg !16 ; <i32> [#uses=1]
923 %tmp1 = load i32* %Y, !dbg !16 ; <i32> [#uses=1]
924 %add = add nsw i32 %tmp, %tmp1, !dbg !16 ; <i32> [#uses=1]
925 store i32 %add, i32* %Z, !dbg !16
926 %tmp2 = load i32* %Y, !dbg !17 ; <i32> [#uses=1]
927 store i32 %tmp2, i32* %X, !dbg !17
931 declare void @llvm.dbg.declare(metadata, metadata) nounwind readnone
933 !0 = metadata !{i32 459008, metadata !1, metadata !"X",
934 metadata !3, i32 2, metadata !6}; [ DW_TAG_auto_variable ]
935 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
936 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo", metadata !"foo",
937 metadata !"foo", metadata !3, i32 1, metadata !4,
938 i1 false, i1 true}; [DW_TAG_subprogram ]
939 !3 = metadata !{i32 458769, i32 0, i32 12, metadata !"foo.c",
940 metadata !"/private/tmp", metadata !"clang 1.1", i1 true,
941 i1 false, metadata !"", i32 0}; [DW_TAG_compile_unit ]
942 !4 = metadata !{i32 458773, metadata !3, metadata !"", null, i32 0, i64 0, i64 0,
943 i64 0, i32 0, null, metadata !5, i32 0}; [DW_TAG_subroutine_type ]
944 !5 = metadata !{null}
945 !6 = metadata !{i32 458788, metadata !3, metadata !"int", metadata !3, i32 0,
946 i64 32, i64 32, i64 0, i32 0, i32 5}; [DW_TAG_base_type ]
947 !7 = metadata !{i32 2, i32 7, metadata !1, null}
948 !8 = metadata !{i32 2, i32 3, metadata !1, null}
949 !9 = metadata !{i32 459008, metadata !1, metadata !"Y", metadata !3, i32 3,
950 metadata !6}; [ DW_TAG_auto_variable ]
951 !10 = metadata !{i32 3, i32 7, metadata !1, null}
952 !11 = metadata !{i32 3, i32 3, metadata !1, null}
953 !12 = metadata !{i32 459008, metadata !13, metadata !"Z", metadata !3, i32 5,
954 metadata !6}; [ DW_TAG_auto_variable ]
955 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
956 !14 = metadata !{i32 5, i32 9, metadata !13, null}
957 !15 = metadata !{i32 5, i32 5, metadata !13, null}
958 !16 = metadata !{i32 6, i32 5, metadata !13, null}
959 !17 = metadata !{i32 8, i32 3, metadata !1, null}
960 !18 = metadata !{i32 9, i32 1, metadata !2, null}
964 <p>This example illustrates a few important details about LLVM debugging
965 information. In particular, it shows how the <tt>llvm.dbg.declare</tt>
966 intrinsic and location information, which are attached to an instruction,
967 are applied together to allow a debugger to analyze the relationship between
968 statements, variable definitions, and the code used to implement the
971 <div class="doc_code">
973 call void @llvm.dbg.declare(metadata, metadata !0), !dbg !7
977 <p>The first intrinsic
978 <tt>%<a href="#format_common_declare">llvm.dbg.declare</a></tt>
979 encodes debugging information for the variable <tt>X</tt>. The metadata
980 <tt>!dbg !7</tt> attached to the intrinsic provides scope information for the
981 variable <tt>X</tt>.</p>
983 <div class="doc_code">
985 !7 = metadata !{i32 2, i32 7, metadata !1, null}
986 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
987 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo",
988 metadata !"foo", metadata !"foo", metadata !3, i32 1,
989 metadata !4, i1 false, i1 true}; [DW_TAG_subprogram ]
993 <p>Here <tt>!7</tt> is metadata providing location information. It has four
994 fields: line number, column number, scope, and original scope. The original
995 scope represents inline location if this instruction is inlined inside a
996 caller, and is null otherwise. In this example, scope is encoded by
997 <tt>!1</tt>. <tt>!1</tt> represents a lexical block inside the scope
998 <tt>!2</tt>, where <tt>!2</tt> is a
999 <a href="#format_subprograms">subprogram descriptor</a>. This way the
1000 location information attached to the intrinsics indicates that the
1001 variable <tt>X</tt> is declared at line number 2 at a function level scope in
1002 function <tt>foo</tt>.</p>
1004 <p>Now lets take another example.</p>
1006 <div class="doc_code">
1008 call void @llvm.dbg.declare(metadata, metadata !12), !dbg !14
1012 <p>The second intrinsic
1013 <tt>%<a href="#format_common_declare">llvm.dbg.declare</a></tt>
1014 encodes debugging information for variable <tt>Z</tt>. The metadata
1015 <tt>!dbg !14</tt> attached to the intrinsic provides scope information for
1016 the variable <tt>Z</tt>.</p>
1018 <div class="doc_code">
1020 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
1021 !14 = metadata !{i32 5, i32 9, metadata !13, null}
1025 <p>Here <tt>!14</tt> indicates that <tt>Z</tt> is declared at line number 5 and
1026 column number 9 inside of lexical scope <tt>!13</tt>. The lexical scope
1027 itself resides inside of lexical scope <tt>!1</tt> described above.</p>
1029 <p>The scope information attached with each instruction provides a
1030 straightforward way to find instructions covered by a scope.</p>
1036 <!-- *********************************************************************** -->
1038 <a name="ccxx_frontend">C/C++ front-end specific debug information</a>
1040 <!-- *********************************************************************** -->
1044 <p>The C and C++ front-ends represent information about the program in a format
1045 that is effectively identical
1046 to <a href="http://www.eagercon.com/dwarf/dwarf3std.htm">DWARF 3.0</a> in
1047 terms of information content. This allows code generators to trivially
1048 support native debuggers by generating standard dwarf information, and
1049 contains enough information for non-dwarf targets to translate it as
1052 <p>This section describes the forms used to represent C and C++ programs. Other
1053 languages could pattern themselves after this (which itself is tuned to
1054 representing programs in the same way that DWARF 3 does), or they could
1055 choose to provide completely different forms if they don't fit into the DWARF
1056 model. As support for debugging information gets added to the various LLVM
1057 source-language front-ends, the information used should be documented
1060 <p>The following sections provide examples of various C/C++ constructs and the
1061 debug information that would best describe those constructs.</p>
1063 <!-- ======================================================================= -->
1065 <a name="ccxx_compile_units">C/C++ source file information</a>
1070 <p>Given the source files <tt>MySource.cpp</tt> and <tt>MyHeader.h</tt> located
1071 in the directory <tt>/Users/mine/sources</tt>, the following code:</p>
1073 <div class="doc_code">
1075 #include "MyHeader.h"
1077 int main(int argc, char *argv[]) {
1083 <p>a C/C++ front-end would generate the following descriptors:</p>
1085 <div class="doc_code">
1089 ;; Define the compile unit for the main source file "/Users/mine/sources/MySource.cpp".
1094 i32 4, ;; Language Id
1095 metadata !"MySource.cpp",
1096 metadata !"/Users/mine/sources",
1097 metadata !"4.2.1 (Based on Apple Inc. build 5649) (LLVM build 00)",
1098 i1 true, ;; Main Compile Unit
1099 i1 false, ;; Optimized compile unit
1100 metadata !"", ;; Compiler flags
1101 i32 0} ;; Runtime version
1104 ;; Define the file for the file "/Users/mine/sources/MySource.cpp".
1108 metadata !"MySource.cpp",
1109 metadata !"/Users/mine/sources",
1110 metadata !2 ;; Compile unit
1114 ;; Define the file for the file "/Users/mine/sources/Myheader.h"
1118 metadata !"Myheader.h"
1119 metadata !"/Users/mine/sources",
1120 metadata !2 ;; Compile unit
1127 <p>llvm::Instruction provides easy access to metadata attached with an
1128 instruction. One can extract line number information encoded in LLVM IR
1129 using <tt>Instruction::getMetadata()</tt> and
1130 <tt>DILocation::getLineNumber()</tt>.
1132 if (MDNode *N = I->getMetadata("dbg")) { // Here I is an LLVM instruction
1133 DILocation Loc(N); // DILocation is in DebugInfo.h
1134 unsigned Line = Loc.getLineNumber();
1135 StringRef File = Loc.getFilename();
1136 StringRef Dir = Loc.getDirectory();
1141 <!-- ======================================================================= -->
1143 <a name="ccxx_global_variable">C/C++ global variable information</a>
1148 <p>Given an integer global variable declared as follows:</p>
1150 <div class="doc_code">
1156 <p>a C/C++ front-end would generate the following descriptors:</p>
1158 <div class="doc_code">
1161 ;; Define the global itself.
1163 %MyGlobal = global int 100
1166 ;; List of debug info of globals
1168 !llvm.dbg.cu = !{!0}
1170 ;; Define the compile unit.
1175 metadata !"foo.cpp", ;; File
1176 metadata !"/Volumes/Data/tmp", ;; Directory
1177 metadata !"clang version 3.1 ", ;; Producer
1178 i1 true, ;; Deprecated field
1179 i1 false, ;; "isOptimized"?
1180 metadata !"", ;; Flags
1181 i32 0, ;; Runtime Version
1182 metadata !1, ;; Enum Types
1183 metadata !1, ;; Retained Types
1184 metadata !1, ;; Subprograms
1185 metadata !3 ;; Global Variables
1186 } ; [ DW_TAG_compile_unit ]
1188 ;; The Array of Global Variables
1198 ;; Define the global variable itself.
1204 metadata !"MyGlobal", ;; Name
1205 metadata !"MyGlobal", ;; Display Name
1206 metadata !"", ;; Linkage Name
1207 metadata !6, ;; File
1209 metadata !7, ;; Type
1210 i32 0, ;; IsLocalToUnit
1211 i32 1, ;; IsDefinition
1212 i32* @MyGlobal ;; LLVM-IR Value
1213 } ; [ DW_TAG_variable ]
1220 metadata !"foo.cpp", ;; File
1221 metadata !"/Volumes/Data/tmp", ;; Directory
1223 } ; [ DW_TAG_file_type ]
1231 metadata !"int", ;; Name
1234 i64 32, ;; Size in Bits
1235 i64 32, ;; Align in Bits
1239 } ; [ DW_TAG_base_type ]
1246 <!-- ======================================================================= -->
1248 <a name="ccxx_subprogram">C/C++ function information</a>
1253 <p>Given a function declared as follows:</p>
1255 <div class="doc_code">
1257 int main(int argc, char *argv[]) {
1263 <p>a C/C++ front-end would generate the following descriptors:</p>
1265 <div class="doc_code">
1268 ;; Define the anchor for subprograms. Note that the second field of the
1269 ;; anchor is 46, which is the same as the tag for subprograms
1270 ;; (46 = DW_TAG_subprogram.)
1275 metadata !1, ;; Context
1276 metadata !"main", ;; Name
1277 metadata !"main", ;; Display name
1278 metadata !"main", ;; Linkage name
1279 metadata !1, ;; File
1280 i32 1, ;; Line number
1281 metadata !4, ;; Type
1282 i1 false, ;; Is local
1283 i1 true, ;; Is definition
1284 i32 0, ;; Virtuality attribute, e.g. pure virtual function
1285 i32 0, ;; Index into virtual table for C++ methods
1286 i32 0, ;; Type that holds virtual table.
1288 i1 false, ;; True if this function is optimized
1289 Function *, ;; Pointer to llvm::Function
1290 null ;; Function template parameters
1293 ;; Define the subprogram itself.
1295 define i32 @main(i32 %argc, i8** %argv) {
1303 <!-- ======================================================================= -->
1305 <a name="ccxx_basic_types">C/C++ basic types</a>
1310 <p>The following are the basic type descriptors for C/C++ core types:</p>
1312 <!-- ======================================================================= -->
1314 <a name="ccxx_basic_type_bool">bool</a>
1319 <div class="doc_code">
1323 metadata !1, ;; Context
1324 metadata !"bool", ;; Name
1325 metadata !1, ;; File
1326 i32 0, ;; Line number
1327 i64 8, ;; Size in Bits
1328 i64 8, ;; Align in Bits
1329 i64 0, ;; Offset in Bits
1338 <!-- ======================================================================= -->
1340 <a name="ccxx_basic_char">char</a>
1345 <div class="doc_code">
1349 metadata !1, ;; Context
1350 metadata !"char", ;; Name
1351 metadata !1, ;; File
1352 i32 0, ;; Line number
1353 i64 8, ;; Size in Bits
1354 i64 8, ;; Align in Bits
1355 i64 0, ;; Offset in Bits
1364 <!-- ======================================================================= -->
1366 <a name="ccxx_basic_unsigned_char">unsigned char</a>
1371 <div class="doc_code">
1375 metadata !1, ;; Context
1376 metadata !"unsigned char",
1377 metadata !1, ;; File
1378 i32 0, ;; Line number
1379 i64 8, ;; Size in Bits
1380 i64 8, ;; Align in Bits
1381 i64 0, ;; Offset in Bits
1390 <!-- ======================================================================= -->
1392 <a name="ccxx_basic_short">short</a>
1397 <div class="doc_code">
1401 metadata !1, ;; Context
1402 metadata !"short int",
1403 metadata !1, ;; File
1404 i32 0, ;; Line number
1405 i64 16, ;; Size in Bits
1406 i64 16, ;; Align in Bits
1407 i64 0, ;; Offset in Bits
1416 <!-- ======================================================================= -->
1418 <a name="ccxx_basic_unsigned_short">unsigned short</a>
1423 <div class="doc_code">
1427 metadata !1, ;; Context
1428 metadata !"short unsigned int",
1429 metadata !1, ;; File
1430 i32 0, ;; Line number
1431 i64 16, ;; Size in Bits
1432 i64 16, ;; Align in Bits
1433 i64 0, ;; Offset in Bits
1442 <!-- ======================================================================= -->
1444 <a name="ccxx_basic_int">int</a>
1449 <div class="doc_code">
1453 metadata !1, ;; Context
1454 metadata !"int", ;; Name
1455 metadata !1, ;; File
1456 i32 0, ;; Line number
1457 i64 32, ;; Size in Bits
1458 i64 32, ;; Align in Bits
1459 i64 0, ;; Offset in Bits
1467 <!-- ======================================================================= -->
1469 <a name="ccxx_basic_unsigned_int">unsigned int</a>
1474 <div class="doc_code">
1478 metadata !1, ;; Context
1479 metadata !"unsigned int",
1480 metadata !1, ;; File
1481 i32 0, ;; Line number
1482 i64 32, ;; Size in Bits
1483 i64 32, ;; Align in Bits
1484 i64 0, ;; Offset in Bits
1493 <!-- ======================================================================= -->
1495 <a name="ccxx_basic_long_long">long long</a>
1500 <div class="doc_code">
1504 metadata !1, ;; Context
1505 metadata !"long long int",
1506 metadata !1, ;; File
1507 i32 0, ;; Line number
1508 i64 64, ;; Size in Bits
1509 i64 64, ;; Align in Bits
1510 i64 0, ;; Offset in Bits
1519 <!-- ======================================================================= -->
1521 <a name="ccxx_basic_unsigned_long_long">unsigned long long</a>
1526 <div class="doc_code">
1530 metadata !1, ;; Context
1531 metadata !"long long unsigned int",
1532 metadata !1, ;; File
1533 i32 0, ;; Line number
1534 i64 64, ;; Size in Bits
1535 i64 64, ;; Align in Bits
1536 i64 0, ;; Offset in Bits
1545 <!-- ======================================================================= -->
1547 <a name="ccxx_basic_float">float</a>
1552 <div class="doc_code">
1556 metadata !1, ;; Context
1558 metadata !1, ;; File
1559 i32 0, ;; Line number
1560 i64 32, ;; Size in Bits
1561 i64 32, ;; Align in Bits
1562 i64 0, ;; Offset in Bits
1571 <!-- ======================================================================= -->
1573 <a name="ccxx_basic_double">double</a>
1578 <div class="doc_code">
1582 metadata !1, ;; Context
1583 metadata !"double",;; Name
1584 metadata !1, ;; File
1585 i32 0, ;; Line number
1586 i64 64, ;; Size in Bits
1587 i64 64, ;; Align in Bits
1588 i64 0, ;; Offset in Bits
1599 <!-- ======================================================================= -->
1601 <a name="ccxx_derived_types">C/C++ derived types</a>
1606 <p>Given the following as an example of C/C++ derived type:</p>
1608 <div class="doc_code">
1610 typedef const int *IntPtr;
1614 <p>a C/C++ front-end would generate the following descriptors:</p>
1616 <div class="doc_code">
1619 ;; Define the typedef "IntPtr".
1623 metadata !1, ;; Context
1624 metadata !"IntPtr", ;; Name
1625 metadata !3, ;; File
1626 i32 0, ;; Line number
1627 i64 0, ;; Size in bits
1628 i64 0, ;; Align in bits
1629 i64 0, ;; Offset in bits
1631 metadata !4 ;; Derived From type
1635 ;; Define the pointer type.
1639 metadata !1, ;; Context
1640 metadata !"", ;; Name
1641 metadata !1, ;; File
1642 i32 0, ;; Line number
1643 i64 64, ;; Size in bits
1644 i64 64, ;; Align in bits
1645 i64 0, ;; Offset in bits
1647 metadata !5 ;; Derived From type
1650 ;; Define the const type.
1654 metadata !1, ;; Context
1655 metadata !"", ;; Name
1656 metadata !1, ;; File
1657 i32 0, ;; Line number
1658 i64 32, ;; Size in bits
1659 i64 32, ;; Align in bits
1660 i64 0, ;; Offset in bits
1662 metadata !6 ;; Derived From type
1665 ;; Define the int type.
1669 metadata !1, ;; Context
1670 metadata !"int", ;; Name
1671 metadata !1, ;; File
1672 i32 0, ;; Line number
1673 i64 32, ;; Size in bits
1674 i64 32, ;; Align in bits
1675 i64 0, ;; Offset in bits
1684 <!-- ======================================================================= -->
1686 <a name="ccxx_composite_types">C/C++ struct/union types</a>
1691 <p>Given the following as an example of C/C++ struct type:</p>
1693 <div class="doc_code">
1703 <p>a C/C++ front-end would generate the following descriptors:</p>
1705 <div class="doc_code">
1708 ;; Define basic type for unsigned int.
1712 metadata !1, ;; Context
1713 metadata !"unsigned int",
1714 metadata !1, ;; File
1715 i32 0, ;; Line number
1716 i64 32, ;; Size in Bits
1717 i64 32, ;; Align in Bits
1718 i64 0, ;; Offset in Bits
1723 ;; Define composite type for struct Color.
1727 metadata !1, ;; Context
1728 metadata !"Color", ;; Name
1729 metadata !1, ;; Compile unit
1730 i32 1, ;; Line number
1731 i64 96, ;; Size in bits
1732 i64 32, ;; Align in bits
1733 i64 0, ;; Offset in bits
1735 null, ;; Derived From
1736 metadata !3, ;; Elements
1737 i32 0 ;; Runtime Language
1741 ;; Define the Red field.
1745 metadata !1, ;; Context
1746 metadata !"Red", ;; Name
1747 metadata !1, ;; File
1748 i32 2, ;; Line number
1749 i64 32, ;; Size in bits
1750 i64 32, ;; Align in bits
1751 i64 0, ;; Offset in bits
1753 metadata !5 ;; Derived From type
1757 ;; Define the Green field.
1761 metadata !1, ;; Context
1762 metadata !"Green", ;; Name
1763 metadata !1, ;; File
1764 i32 3, ;; Line number
1765 i64 32, ;; Size in bits
1766 i64 32, ;; Align in bits
1767 i64 32, ;; Offset in bits
1769 metadata !5 ;; Derived From type
1773 ;; Define the Blue field.
1777 metadata !1, ;; Context
1778 metadata !"Blue", ;; Name
1779 metadata !1, ;; File
1780 i32 4, ;; Line number
1781 i64 32, ;; Size in bits
1782 i64 32, ;; Align in bits
1783 i64 64, ;; Offset in bits
1785 metadata !5 ;; Derived From type
1789 ;; Define the array of fields used by the composite type Color.
1791 !3 = metadata !{metadata !4, metadata !6, metadata !7}
1797 <!-- ======================================================================= -->
1799 <a name="ccxx_enumeration_types">C/C++ enumeration types</a>
1804 <p>Given the following as an example of C/C++ enumeration type:</p>
1806 <div class="doc_code">
1816 <p>a C/C++ front-end would generate the following descriptors:</p>
1818 <div class="doc_code">
1821 ;; Define composite type for enum Trees
1825 metadata !1, ;; Context
1826 metadata !"Trees", ;; Name
1827 metadata !1, ;; File
1828 i32 1, ;; Line number
1829 i64 32, ;; Size in bits
1830 i64 32, ;; Align in bits
1831 i64 0, ;; Offset in bits
1833 null, ;; Derived From type
1834 metadata !3, ;; Elements
1835 i32 0 ;; Runtime language
1839 ;; Define the array of enumerators used by composite type Trees.
1841 !3 = metadata !{metadata !4, metadata !5, metadata !6}
1844 ;; Define Spruce enumerator.
1846 !4 = metadata !{i32 524328, metadata !"Spruce", i64 100}
1849 ;; Define Oak enumerator.
1851 !5 = metadata !{i32 524328, metadata !"Oak", i64 200}
1854 ;; Define Maple enumerator.
1856 !6 = metadata !{i32 524328, metadata !"Maple", i64 300}
1866 <!-- *********************************************************************** -->
1868 <a name="llvmdwarfextension">Debugging information format</a>
1870 <!-- *********************************************************************** -->
1872 <!-- ======================================================================= -->
1874 <a name="objcproperty">Debugging Information Extension for Objective C Properties</a>
1877 <!-- *********************************************************************** -->
1879 <a name="objcpropertyintroduction">Introduction</a>
1881 <!-- *********************************************************************** -->
1884 <p>Objective C provides a simpler way to declare and define accessor methods
1885 using declared properties. The language provides features to declare a
1886 property and to let compiler synthesize accessor methods.
1889 <p>The debugger lets developer inspect Objective C interfaces and their
1890 instance variables and class variables. However, the debugger does not know
1891 anything about the properties defined in Objective C interfaces. The debugger
1892 consumes information generated by compiler in DWARF format. The format does
1893 not support encoding of Objective C properties. This proposal describes DWARF
1894 extensions to encode Objective C properties, which the debugger can use to let
1895 developers inspect Objective C properties.
1901 <!-- *********************************************************************** -->
1903 <a name="objcpropertyproposal">Proposal</a>
1905 <!-- *********************************************************************** -->
1908 <p>Objective C properties exist separately from class members. A property
1909 can be defined only by "setter" and "getter" selectors, and
1910 be calculated anew on each access. Or a property can just be a direct access
1911 to some declared ivar. Finally it can have an ivar "automatically
1912 synthesized" for it by the compiler, in which case the property can be
1913 referred to in user code directly using the standard C dereference syntax as
1914 well as through the property "dot" syntax, but there is no entry in
1915 the @interface declaration corresponding to this ivar.
1918 To facilitate debugging, these properties we will add a new DWARF TAG into the
1919 DW_TAG_structure_type definition for the class to hold the description of a
1920 given property, and a set of DWARF attributes that provide said description.
1921 The property tag will also contain the name and declared type of the property.
1924 If there is a related ivar, there will also be a DWARF property attribute placed
1925 in the DW_TAG_member DIE for that ivar referring back to the property TAG for
1926 that property. And in the case where the compiler synthesizes the ivar directly,
1927 the compiler is expected to generate a DW_TAG_member for that ivar (with the
1928 DW_AT_artificial set to 1), whose name will be the name used to access this
1929 ivar directly in code, and with the property attribute pointing back to the
1930 property it is backing.
1933 The following examples will serve as illustration for our discussion:
1936 <div class="doc_code">
1948 @synthesize p2 = n2;
1954 This produces the following DWARF (this is a "pseudo dwarfdump" output):
1956 <div class="doc_code">
1958 0x00000100: TAG_structure_type [7] *
1959 AT_APPLE_runtime_class( 0x10 )
1961 AT_decl_file( "Objc_Property.m" )
1964 0x00000110 TAG_APPLE_property
1966 AT_type ( {0x00000150} ( int ) )
1968 0x00000120: TAG_APPLE_property
1970 AT_type ( {0x00000150} ( int ) )
1972 0x00000130: TAG_member [8]
1974 AT_APPLE_property ( {0x00000110} "p1" )
1975 AT_type( {0x00000150} ( int ) )
1976 AT_artificial ( 0x1 )
1978 0x00000140: TAG_member [8]
1980 AT_APPLE_property ( {0x00000120} "p2" )
1981 AT_type( {0x00000150} ( int ) )
1983 0x00000150: AT_type( ( int ) )
1987 <p> Note, the current convention is that the name of the ivar for an
1988 auto-synthesized property is the name of the property from which it derives with
1989 an underscore prepended, as is shown in the example.
1990 But we actually don't need to know this convention, since we are given the name
1991 of the ivar directly.
1995 Also, it is common practice in ObjC to have different property declarations in
1996 the @interface and @implementation - e.g. to provide a read-only property in
1997 the interface,and a read-write interface in the implementation. In that case,
1998 the compiler should emit whichever property declaration will be in force in the
1999 current translation unit.
2002 <p> Developers can decorate a property with attributes which are encoded using
2003 DW_AT_APPLE_property_attribute.
2006 <div class="doc_code">
2008 @property (readonly, nonatomic) int pr;
2012 Which produces a property tag:
2014 <div class="doc_code">
2016 TAG_APPLE_property [8]
2018 AT_type ( {0x00000147} (int) )
2019 AT_APPLE_property_attribute (DW_APPLE_PROPERTY_readonly, DW_APPLE_PROPERTY_nonatomic)
2023 <p> The setter and getter method names are attached to the property using
2024 DW_AT_APPLE_property_setter and DW_AT_APPLE_property_getter attributes.
2026 <div class="doc_code">
2029 @property (setter=myOwnP3Setter:) int p3;
2030 -(void)myOwnP3Setter:(int)a;
2035 -(void)myOwnP3Setter:(int)a{ }
2041 The DWARF for this would be:
2043 <div class="doc_code">
2045 0x000003bd: TAG_structure_type [7] *
2046 AT_APPLE_runtime_class( 0x10 )
2048 AT_decl_file( "Objc_Property.m" )
2051 0x000003cd TAG_APPLE_property
2053 AT_APPLE_property_setter ( "myOwnP3Setter:" )
2054 AT_type( {0x00000147} ( int ) )
2056 0x000003f3: TAG_member [8]
2058 AT_type ( {0x00000147} ( int ) )
2059 AT_APPLE_property ( {0x000003cd} )
2060 AT_artificial ( 0x1 )
2066 <!-- *********************************************************************** -->
2068 <a name="objcpropertynewtags">New DWARF Tags</a>
2070 <!-- *********************************************************************** -->
2073 <table border="1" cellspacing="0">
2081 <td>DW_TAG_APPLE_property</td>
2088 <!-- *********************************************************************** -->
2090 <a name="objcpropertynewattributes">New DWARF Attributes</a>
2092 <!-- *********************************************************************** -->
2095 <table border="1" cellspacing="0">
2105 <td>DW_AT_APPLE_property</td>
2110 <td>DW_AT_APPLE_property_getter</td>
2115 <td>DW_AT_APPLE_property_setter</td>
2120 <td>DW_AT_APPLE_property_attribute</td>
2128 <!-- *********************************************************************** -->
2130 <a name="objcpropertynewconstants">New DWARF Constants</a>
2132 <!-- *********************************************************************** -->
2135 <table border="1" cellspacing="0">
2143 <td>DW_AT_APPLE_PROPERTY_readonly</td>
2147 <td>DW_AT_APPLE_PROPERTY_readwrite</td>
2151 <td>DW_AT_APPLE_PROPERTY_assign</td>
2155 <td>DW_AT_APPLE_PROPERTY_retain</td>
2159 <td>DW_AT_APPLE_PROPERTY_copy</td>
2163 <td>DW_AT_APPLE_PROPERTY_nonatomic</td>
2171 <!-- ======================================================================= -->
2173 <a name="acceltable">Name Accelerator Tables</a>
2175 <!-- ======================================================================= -->
2177 <!-- ======================================================================= -->
2179 <a name="acceltableintroduction">Introduction</a>
2181 <!-- ======================================================================= -->
2183 <p>The .debug_pubnames and .debug_pubtypes formats are not what a debugger
2184 needs. The "pub" in the section name indicates that the entries in the
2185 table are publicly visible names only. This means no static or hidden
2186 functions show up in the .debug_pubnames. No static variables or private class
2187 variables are in the .debug_pubtypes. Many compilers add different things to
2188 these tables, so we can't rely upon the contents between gcc, icc, or clang.</p>
2190 <p>The typical query given by users tends not to match up with the contents of
2191 these tables. For example, the DWARF spec states that "In the case of the
2192 name of a function member or static data member of a C++ structure, class or
2193 union, the name presented in the .debug_pubnames section is not the simple
2194 name given by the DW_AT_name attribute of the referenced debugging information
2195 entry, but rather the fully qualified name of the data or function member."
2196 So the only names in these tables for complex C++ entries is a fully
2197 qualified name. Debugger users tend not to enter their search strings as
2198 "a::b::c(int,const Foo&) const", but rather as "c", "b::c" , or "a::b::c". So
2199 the name entered in the name table must be demangled in order to chop it up
2200 appropriately and additional names must be manually entered into the table
2201 to make it effective as a name lookup table for debuggers to use.</p>
2203 <p>All debuggers currently ignore the .debug_pubnames table as a result of
2204 its inconsistent and useless public-only name content making it a waste of
2205 space in the object file. These tables, when they are written to disk, are
2206 not sorted in any way, leaving every debugger to do its own parsing
2207 and sorting. These tables also include an inlined copy of the string values
2208 in the table itself making the tables much larger than they need to be on
2209 disk, especially for large C++ programs.</p>
2211 <p>Can't we just fix the sections by adding all of the names we need to this
2212 table? No, because that is not what the tables are defined to contain and we
2213 won't know the difference between the old bad tables and the new good tables.
2214 At best we could make our own renamed sections that contain all of the data
2217 <p>These tables are also insufficient for what a debugger like LLDB needs.
2218 LLDB uses clang for its expression parsing where LLDB acts as a PCH. LLDB is
2219 then often asked to look for type "foo" or namespace "bar", or list items in
2220 namespace "baz". Namespaces are not included in the pubnames or pubtypes
2221 tables. Since clang asks a lot of questions when it is parsing an expression,
2222 we need to be very fast when looking up names, as it happens a lot. Having new
2223 accelerator tables that are optimized for very quick lookups will benefit
2224 this type of debugging experience greatly.</p>
2226 <p>We would like to generate name lookup tables that can be mapped into
2227 memory from disk, and used as is, with little or no up-front parsing. We would
2228 also be able to control the exact content of these different tables so they
2229 contain exactly what we need. The Name Accelerator Tables were designed
2230 to fix these issues. In order to solve these issues we need to:</p>
2233 <li>Have a format that can be mapped into memory from disk and used as is</li>
2234 <li>Lookups should be very fast</li>
2235 <li>Extensible table format so these tables can be made by many producers</li>
2236 <li>Contain all of the names needed for typical lookups out of the box</li>
2237 <li>Strict rules for the contents of tables</li>
2240 <p>Table size is important and the accelerator table format should allow the
2241 reuse of strings from common string tables so the strings for the names are
2242 not duplicated. We also want to make sure the table is ready to be used as-is
2243 by simply mapping the table into memory with minimal header parsing.</p>
2245 <p>The name lookups need to be fast and optimized for the kinds of lookups
2246 that debuggers tend to do. Optimally we would like to touch as few parts of
2247 the mapped table as possible when doing a name lookup and be able to quickly
2248 find the name entry we are looking for, or discover there are no matches. In
2249 the case of debuggers we optimized for lookups that fail most of the time.</p>
2251 <p>Each table that is defined should have strict rules on exactly what is in
2252 the accelerator tables and documented so clients can rely on the content.</p>
2256 <!-- ======================================================================= -->
2258 <a name="acceltablehashes">Hash Tables</a>
2260 <!-- ======================================================================= -->
2263 <h5>Standard Hash Tables</h5>
2265 <p>Typical hash tables have a header, buckets, and each bucket points to the
2269 <div class="doc_code">
2281 <p>The BUCKETS are an array of offsets to DATA for each hash:</p>
2283 <div class="doc_code">
2286 | 0x00001000 | BUCKETS[0]
2287 | 0x00002000 | BUCKETS[1]
2288 | 0x00002200 | BUCKETS[2]
2289 | 0x000034f0 | BUCKETS[3]
2291 | 0xXXXXXXXX | BUCKETS[n_buckets]
2296 <p>So for bucket[3] in the example above, we have an offset into the table
2297 0x000034f0 which points to a chain of entries for the bucket. Each bucket
2298 must contain a next pointer, full 32 bit hash value, the string itself,
2299 and the data for the current string value.</p>
2301 <div class="doc_code">
2304 0x000034f0: | 0x00003500 | next pointer
2305 | 0x12345678 | 32 bit hash
2306 | "erase" | string value
2307 | data[n] | HashData for this bucket
2309 0x00003500: | 0x00003550 | next pointer
2310 | 0x29273623 | 32 bit hash
2311 | "dump" | string value
2312 | data[n] | HashData for this bucket
2314 0x00003550: | 0x00000000 | next pointer
2315 | 0x82638293 | 32 bit hash
2316 | "main" | string value
2317 | data[n] | HashData for this bucket
2322 <p>The problem with this layout for debuggers is that we need to optimize for
2323 the negative lookup case where the symbol we're searching for is not present.
2324 So if we were to lookup "printf" in the table above, we would make a 32 hash
2325 for "printf", it might match bucket[3]. We would need to go to the offset
2326 0x000034f0 and start looking to see if our 32 bit hash matches. To do so, we
2327 need to read the next pointer, then read the hash, compare it, and skip to
2328 the next bucket. Each time we are skipping many bytes in memory and touching
2329 new cache pages just to do the compare on the full 32 bit hash. All of these
2330 accesses then tell us that we didn't have a match.</p>
2332 <h5>Name Hash Tables</h5>
2334 <p>To solve the issues mentioned above we have structured the hash tables
2335 a bit differently: a header, buckets, an array of all unique 32 bit hash
2336 values, followed by an array of hash value data offsets, one for each hash
2337 value, then the data for all hash values:</p>
2339 <div class="doc_code">
2355 <p>The BUCKETS in the name tables are an index into the HASHES array. By
2356 making all of the full 32 bit hash values contiguous in memory, we allow
2357 ourselves to efficiently check for a match while touching as little
2358 memory as possible. Most often checking the 32 bit hash values is as far as
2359 the lookup goes. If it does match, it usually is a match with no collisions.
2360 So for a table with "n_buckets" buckets, and "n_hashes" unique 32 bit hash
2361 values, we can clarify the contents of the BUCKETS, HASHES and OFFSETS as:</p>
2363 <div class="doc_code">
2365 .-------------------------.
2366 | HEADER.magic | uint32_t
2367 | HEADER.version | uint16_t
2368 | HEADER.hash_function | uint16_t
2369 | HEADER.bucket_count | uint32_t
2370 | HEADER.hashes_count | uint32_t
2371 | HEADER.header_data_len | uint32_t
2372 | HEADER_DATA | HeaderData
2373 |-------------------------|
2374 | BUCKETS | uint32_t[n_buckets] // 32 bit hash indexes
2375 |-------------------------|
2376 | HASHES | uint32_t[n_buckets] // 32 bit hash values
2377 |-------------------------|
2378 | OFFSETS | uint32_t[n_buckets] // 32 bit offsets to hash value data
2379 |-------------------------|
2381 `-------------------------'
2385 <p>So taking the exact same data from the standard hash example above we end up
2388 <div class="doc_code">
2398 | ... | BUCKETS[n_buckets]
2400 | 0x........ | HASHES[0]
2401 | 0x........ | HASHES[1]
2402 | 0x........ | HASHES[2]
2403 | 0x........ | HASHES[3]
2404 | 0x........ | HASHES[4]
2405 | 0x........ | HASHES[5]
2406 | 0x12345678 | HASHES[6] hash for BUCKETS[3]
2407 | 0x29273623 | HASHES[7] hash for BUCKETS[3]
2408 | 0x82638293 | HASHES[8] hash for BUCKETS[3]
2409 | 0x........ | HASHES[9]
2410 | 0x........ | HASHES[10]
2411 | 0x........ | HASHES[11]
2412 | 0x........ | HASHES[12]
2413 | 0x........ | HASHES[13]
2414 | 0x........ | HASHES[n_hashes]
2416 | 0x........ | OFFSETS[0]
2417 | 0x........ | OFFSETS[1]
2418 | 0x........ | OFFSETS[2]
2419 | 0x........ | OFFSETS[3]
2420 | 0x........ | OFFSETS[4]
2421 | 0x........ | OFFSETS[5]
2422 | 0x000034f0 | OFFSETS[6] offset for BUCKETS[3]
2423 | 0x00003500 | OFFSETS[7] offset for BUCKETS[3]
2424 | 0x00003550 | OFFSETS[8] offset for BUCKETS[3]
2425 | 0x........ | OFFSETS[9]
2426 | 0x........ | OFFSETS[10]
2427 | 0x........ | OFFSETS[11]
2428 | 0x........ | OFFSETS[12]
2429 | 0x........ | OFFSETS[13]
2430 | 0x........ | OFFSETS[n_hashes]
2438 0x000034f0: | 0x00001203 | .debug_str ("erase")
2439 | 0x00000004 | A 32 bit array count - number of HashData with name "erase"
2440 | 0x........ | HashData[0]
2441 | 0x........ | HashData[1]
2442 | 0x........ | HashData[2]
2443 | 0x........ | HashData[3]
2444 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2446 0x00003500: | 0x00001203 | String offset into .debug_str ("collision")
2447 | 0x00000002 | A 32 bit array count - number of HashData with name "collision"
2448 | 0x........ | HashData[0]
2449 | 0x........ | HashData[1]
2450 | 0x00001203 | String offset into .debug_str ("dump")
2451 | 0x00000003 | A 32 bit array count - number of HashData with name "dump"
2452 | 0x........ | HashData[0]
2453 | 0x........ | HashData[1]
2454 | 0x........ | HashData[2]
2455 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2457 0x00003550: | 0x00001203 | String offset into .debug_str ("main")
2458 | 0x00000009 | A 32 bit array count - number of HashData with name "main"
2459 | 0x........ | HashData[0]
2460 | 0x........ | HashData[1]
2461 | 0x........ | HashData[2]
2462 | 0x........ | HashData[3]
2463 | 0x........ | HashData[4]
2464 | 0x........ | HashData[5]
2465 | 0x........ | HashData[6]
2466 | 0x........ | HashData[7]
2467 | 0x........ | HashData[8]
2468 | 0x00000000 | String offset into .debug_str (terminate data for hash)
2473 <p>So we still have all of the same data, we just organize it more efficiently
2474 for debugger lookup. If we repeat the same "printf" lookup from above, we
2475 would hash "printf" and find it matches BUCKETS[3] by taking the 32 bit hash
2476 value and modulo it by n_buckets. BUCKETS[3] contains "6" which is the index
2477 into the HASHES table. We would then compare any consecutive 32 bit hashes
2478 values in the HASHES array as long as the hashes would be in BUCKETS[3]. We
2479 do this by verifying that each subsequent hash value modulo n_buckets is still
2480 3. In the case of a failed lookup we would access the memory for BUCKETS[3], and
2481 then compare a few consecutive 32 bit hashes before we know that we have no match.
2482 We don't end up marching through multiple words of memory and we really keep the
2483 number of processor data cache lines being accessed as small as possible.</p>
2485 <p>The string hash that is used for these lookup tables is the Daniel J.
2486 Bernstein hash which is also used in the ELF GNU_HASH sections. It is a very
2487 good hash for all kinds of names in programs with very few hash collisions.</p>
2489 <p>Empty buckets are designated by using an invalid hash index of UINT32_MAX.</p>
2492 <!-- ======================================================================= -->
2494 <a name="acceltabledetails">Details</a>
2496 <!-- ======================================================================= -->
2498 <p>These name hash tables are designed to be generic where specializations of
2499 the table get to define additional data that goes into the header
2500 ("HeaderData"), how the string value is stored ("KeyType") and the content
2501 of the data for each hash value.</p>
2503 <h5>Header Layout</h5>
2504 <p>The header has a fixed part, and the specialized part. The exact format of
2506 <div class="doc_code">
2510 uint32_t magic; // 'HASH' magic value to allow endian detection
2511 uint16_t version; // Version number
2512 uint16_t hash_function; // The hash function enumeration that was used
2513 uint32_t bucket_count; // The number of buckets in this hash table
2514 uint32_t hashes_count; // The total number of unique hash values and hash data offsets in this table
2515 uint32_t header_data_len; // The bytes to skip to get to the hash indexes (buckets) for correct alignment
2516 // Specifically the length of the following HeaderData field - this does not
2517 // include the size of the preceding fields
2518 HeaderData header_data; // Implementation specific header data
2522 <p>The header starts with a 32 bit "magic" value which must be 'HASH' encoded as
2523 an ASCII integer. This allows the detection of the start of the hash table and
2524 also allows the table's byte order to be determined so the table can be
2525 correctly extracted. The "magic" value is followed by a 16 bit version number
2526 which allows the table to be revised and modified in the future. The current
2527 version number is 1. "hash_function" is a uint16_t enumeration that specifies
2528 which hash function was used to produce this table. The current values for the
2529 hash function enumerations include:</p>
2530 <div class="doc_code">
2532 enum HashFunctionType
2534 eHashFunctionDJB = 0u, // Daniel J Bernstein hash function
2538 <p>"bucket_count" is a 32 bit unsigned integer that represents how many buckets
2539 are in the BUCKETS array. "hashes_count" is the number of unique 32 bit hash
2540 values that are in the HASHES array, and is the same number of offsets are
2541 contained in the OFFSETS array. "header_data_len" specifies the size in
2542 bytes of the HeaderData that is filled in by specialized versions of this
2545 <h5>Fixed Lookup</h5>
2546 <p>The header is followed by the buckets, hashes, offsets, and hash value
2548 <div class="doc_code">
2552 uint32_t buckets[Header.bucket_count]; // An array of hash indexes into the "hashes[]" array below
2553 uint32_t hashes [Header.hashes_count]; // Every unique 32 bit hash for the entire table is in this table
2554 uint32_t offsets[Header.hashes_count]; // An offset that corresponds to each item in the "hashes[]" array above
2558 <p>"buckets" is an array of 32 bit indexes into the "hashes" array. The
2559 "hashes" array contains all of the 32 bit hash values for all names in the
2560 hash table. Each hash in the "hashes" table has an offset in the "offsets"
2561 array that points to the data for the hash value.</p>
2563 <p>This table setup makes it very easy to repurpose these tables to contain
2564 different data, while keeping the lookup mechanism the same for all tables.
2565 This layout also makes it possible to save the table to disk and map it in
2566 later and do very efficient name lookups with little or no parsing.</p>
2568 <p>DWARF lookup tables can be implemented in a variety of ways and can store
2569 a lot of information for each name. We want to make the DWARF tables
2570 extensible and able to store the data efficiently so we have used some of the
2571 DWARF features that enable efficient data storage to define exactly what kind
2572 of data we store for each name.</p>
2574 <p>The "HeaderData" contains a definition of the contents of each HashData
2575 chunk. We might want to store an offset to all of the debug information
2576 entries (DIEs) for each name. To keep things extensible, we create a list of
2577 items, or Atoms, that are contained in the data for each name. First comes the
2578 type of the data in each atom:</p>
2579 <div class="doc_code">
2584 eAtomTypeDIEOffset = 1u, // DIE offset, check form for encoding
2585 eAtomTypeCUOffset = 2u, // DIE offset of the compiler unit header that contains the item in question
2586 eAtomTypeTag = 3u, // DW_TAG_xxx value, should be encoded as DW_FORM_data1 (if no tags exceed 255) or DW_FORM_data2
2587 eAtomTypeNameFlags = 4u, // Flags from enum NameFlags
2588 eAtomTypeTypeFlags = 5u, // Flags from enum TypeFlags
2592 <p>The enumeration values and their meanings are:</p>
2593 <div class="doc_code">
2595 eAtomTypeNULL - a termination atom that specifies the end of the atom list
2596 eAtomTypeDIEOffset - an offset into the .debug_info section for the DWARF DIE for this name
2597 eAtomTypeCUOffset - an offset into the .debug_info section for the CU that contains the DIE
2598 eAtomTypeDIETag - The DW_TAG_XXX enumeration value so you don't have to parse the DWARF to see what it is
2599 eAtomTypeNameFlags - Flags for functions and global variables (isFunction, isInlined, isExternal...)
2600 eAtomTypeTypeFlags - Flags for types (isCXXClass, isObjCClass, ...)
2603 <p>Then we allow each atom type to define the atom type and how the data for
2604 each atom type data is encoded:</p>
2605 <div class="doc_code">
2609 uint16_t type; // AtomType enum value
2610 uint16_t form; // DWARF DW_FORM_XXX defines
2614 <p>The "form" type above is from the DWARF specification and defines the
2615 exact encoding of the data for the Atom type. See the DWARF specification for
2616 the DW_FORM_ definitions.</p>
2617 <div class="doc_code">
2621 uint32_t die_offset_base;
2622 uint32_t atom_count;
2623 Atoms atoms[atom_count0];
2627 <p>"HeaderData" defines the base DIE offset that should be added to any atoms
2628 that are encoded using the DW_FORM_ref1, DW_FORM_ref2, DW_FORM_ref4,
2629 DW_FORM_ref8 or DW_FORM_ref_udata. It also defines what is contained in
2630 each "HashData" object -- Atom.form tells us how large each field will be in
2631 the HashData and the Atom.type tells us how this data should be interpreted.</p>
2633 <p>For the current implementations of the ".apple_names" (all functions + globals),
2634 the ".apple_types" (names of all types that are defined), and the
2635 ".apple_namespaces" (all namespaces), we currently set the Atom array to be:</p>
2636 <div class="doc_code">
2638 HeaderData.atom_count = 1;
2639 HeaderData.atoms[0].type = eAtomTypeDIEOffset;
2640 HeaderData.atoms[0].form = DW_FORM_data4;
2643 <p>This defines the contents to be the DIE offset (eAtomTypeDIEOffset) that is
2644 encoded as a 32 bit value (DW_FORM_data4). This allows a single name to have
2645 multiple matching DIEs in a single file, which could come up with an inlined
2646 function for instance. Future tables could include more information about the
2647 DIE such as flags indicating if the DIE is a function, method, block,
2650 <p>The KeyType for the DWARF table is a 32 bit string table offset into the
2651 ".debug_str" table. The ".debug_str" is the string table for the DWARF which
2652 may already contain copies of all of the strings. This helps make sure, with
2653 help from the compiler, that we reuse the strings between all of the DWARF
2654 sections and keeps the hash table size down. Another benefit to having the
2655 compiler generate all strings as DW_FORM_strp in the debug info, is that
2656 DWARF parsing can be made much faster.</p>
2658 <p>After a lookup is made, we get an offset into the hash data. The hash data
2659 needs to be able to deal with 32 bit hash collisions, so the chunk of data
2660 at the offset in the hash data consists of a triple:</p>
2661 <div class="doc_code">
2664 uint32_t hash_data_count
2665 HashData[hash_data_count]
2668 <p>If "str_offset" is zero, then the bucket contents are done. 99.9% of the
2669 hash data chunks contain a single item (no 32 bit hash collision):</p>
2670 <div class="doc_code">
2673 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2674 | 0x00000004 | uint32_t HashData count
2675 | 0x........ | uint32_t HashData[0] DIE offset
2676 | 0x........ | uint32_t HashData[1] DIE offset
2677 | 0x........ | uint32_t HashData[2] DIE offset
2678 | 0x........ | uint32_t HashData[3] DIE offset
2679 | 0x00000000 | uint32_t KeyType (end of hash chain)
2683 <p>If there are collisions, you will have multiple valid string offsets:</p>
2684 <div class="doc_code">
2687 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2688 | 0x00000004 | uint32_t HashData count
2689 | 0x........ | uint32_t HashData[0] DIE offset
2690 | 0x........ | uint32_t HashData[1] DIE offset
2691 | 0x........ | uint32_t HashData[2] DIE offset
2692 | 0x........ | uint32_t HashData[3] DIE offset
2693 | 0x00002023 | uint32_t KeyType (.debug_str[0x0002023] => "print")
2694 | 0x00000002 | uint32_t HashData count
2695 | 0x........ | uint32_t HashData[0] DIE offset
2696 | 0x........ | uint32_t HashData[1] DIE offset
2697 | 0x00000000 | uint32_t KeyType (end of hash chain)
2701 <p>Current testing with real world C++ binaries has shown that there is around 1
2702 32 bit hash collision per 100,000 name entries.</p>
2704 <!-- ======================================================================= -->
2706 <a name="acceltablecontents">Contents</a>
2708 <!-- ======================================================================= -->
2710 <p>As we said, we want to strictly define exactly what is included in the
2711 different tables. For DWARF, we have 3 tables: ".apple_names", ".apple_types",
2712 and ".apple_namespaces".</p>
2714 <p>".apple_names" sections should contain an entry for each DWARF DIE whose
2715 DW_TAG is a DW_TAG_label, DW_TAG_inlined_subroutine, or DW_TAG_subprogram that
2716 has address attributes: DW_AT_low_pc, DW_AT_high_pc, DW_AT_ranges or
2717 DW_AT_entry_pc. It also contains DW_TAG_variable DIEs that have a DW_OP_addr
2718 in the location (global and static variables). All global and static variables
2719 should be included, including those scoped withing functions and classes. For
2720 example using the following code:</p>
2721 <div class="doc_code">
2731 <p>Both of the static "var" variables would be included in the table. All
2732 functions should emit both their full names and their basenames. For C or C++,
2733 the full name is the mangled name (if available) which is usually in the
2734 DW_AT_MIPS_linkage_name attribute, and the DW_AT_name contains the function
2735 basename. If global or static variables have a mangled name in a
2736 DW_AT_MIPS_linkage_name attribute, this should be emitted along with the
2737 simple name found in the DW_AT_name attribute.</p>
2739 <p>".apple_types" sections should contain an entry for each DWARF DIE whose
2742 <li>DW_TAG_array_type</li>
2743 <li>DW_TAG_class_type</li>
2744 <li>DW_TAG_enumeration_type</li>
2745 <li>DW_TAG_pointer_type</li>
2746 <li>DW_TAG_reference_type</li>
2747 <li>DW_TAG_string_type</li>
2748 <li>DW_TAG_structure_type</li>
2749 <li>DW_TAG_subroutine_type</li>
2750 <li>DW_TAG_typedef</li>
2751 <li>DW_TAG_union_type</li>
2752 <li>DW_TAG_ptr_to_member_type</li>
2753 <li>DW_TAG_set_type</li>
2754 <li>DW_TAG_subrange_type</li>
2755 <li>DW_TAG_base_type</li>
2756 <li>DW_TAG_const_type</li>
2757 <li>DW_TAG_constant</li>
2758 <li>DW_TAG_file_type</li>
2759 <li>DW_TAG_namelist</li>
2760 <li>DW_TAG_packed_type</li>
2761 <li>DW_TAG_volatile_type</li>
2762 <li>DW_TAG_restrict_type</li>
2763 <li>DW_TAG_interface_type</li>
2764 <li>DW_TAG_unspecified_type</li>
2765 <li>DW_TAG_shared_type</li>
2767 <p>Only entries with a DW_AT_name attribute are included, and the entry must
2768 not be a forward declaration (DW_AT_declaration attribute with a non-zero value).
2769 For example, using the following code:</p>
2770 <div class="doc_code">
2779 <p>We get a few type DIEs:</p>
2780 <div class="doc_code">
2782 0x00000067: TAG_base_type [5]
2783 AT_encoding( DW_ATE_signed )
2785 AT_byte_size( 0x04 )
2787 0x0000006e: TAG_pointer_type [6]
2788 AT_type( {0x00000067} ( int ) )
2789 AT_byte_size( 0x08 )
2792 <p>The DW_TAG_pointer_type is not included because it does not have a DW_AT_name.</p>
2794 <p>".apple_namespaces" section should contain all DW_TAG_namespace DIEs. If
2795 we run into a namespace that has no name this is an anonymous namespace,
2796 and the name should be output as "(anonymous namespace)" (without the quotes).
2797 Why? This matches the output of the abi::cxa_demangle() that is in the standard
2798 C++ library that demangles mangled names.</p>
2801 <!-- ======================================================================= -->
2803 <a name="acceltableextensions">Language Extensions and File Format Changes</a>
2805 <!-- ======================================================================= -->
2807 <h5>Objective-C Extensions</h5>
2808 <p>".apple_objc" section should contain all DW_TAG_subprogram DIEs for an
2809 Objective-C class. The name used in the hash table is the name of the
2810 Objective-C class itself. If the Objective-C class has a category, then an
2811 entry is made for both the class name without the category, and for the class
2812 name with the category. So if we have a DIE at offset 0x1234 with a name
2813 of method "-[NSString(my_additions) stringWithSpecialString:]", we would add
2814 an entry for "NSString" that points to DIE 0x1234, and an entry for
2815 "NSString(my_additions)" that points to 0x1234. This allows us to quickly
2816 track down all Objective-C methods for an Objective-C class when doing
2817 expressions. It is needed because of the dynamic nature of Objective-C where
2818 anyone can add methods to a class. The DWARF for Objective-C methods is also
2819 emitted differently from C++ classes where the methods are not usually
2820 contained in the class definition, they are scattered about across one or more
2821 compile units. Categories can also be defined in different shared libraries.
2822 So we need to be able to quickly find all of the methods and class functions
2823 given the Objective-C class name, or quickly find all methods and class
2824 functions for a class + category name. This table does not contain any selector
2825 names, it just maps Objective-C class names (or class names + category) to all
2826 of the methods and class functions. The selectors are added as function
2827 basenames in the .debug_names section.</p>
2829 <p>In the ".apple_names" section for Objective-C functions, the full name is the
2830 entire function name with the brackets ("-[NSString stringWithCString:]") and the
2831 basename is the selector only ("stringWithCString:").</p>
2833 <h5>Mach-O Changes</h5>
2834 <p>The sections names for the apple hash tables are for non mach-o files. For
2835 mach-o files, the sections should be contained in the "__DWARF" segment with
2836 names as follows:</p>
2838 <li>".apple_names" -> "__apple_names"</li>
2839 <li>".apple_types" -> "__apple_types"</li>
2840 <li>".apple_namespaces" -> "__apple_namespac" (16 character limit)</li>
2841 <li> ".apple_objc" -> "__apple_objc"</li>
2847 <!-- *********************************************************************** -->
2851 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2852 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2853 <a href="http://validator.w3.org/check/referer"><img
2854 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2856 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2857 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
2858 Last modified: $Date$