1 ================================
2 Source Level Debugging with LLVM
3 ================================
11 This document is the central repository for all information pertaining to debug
12 information in LLVM. It describes the :ref:`actual format that the LLVM debug
13 information takes <format>`, which is useful for those interested in creating
14 front-ends or dealing directly with the information. Further, this document
15 provides specific examples of what debug information for C/C++ looks like.
17 Philosophy behind LLVM debugging information
18 --------------------------------------------
20 The idea of the LLVM debugging information is to capture how the important
21 pieces of the source-language's Abstract Syntax Tree map onto LLVM code.
22 Several design aspects have shaped the solution that appears here. The
25 * Debugging information should have very little impact on the rest of the
26 compiler. No transformations, analyses, or code generators should need to
27 be modified because of debugging information.
29 * LLVM optimizations should interact in :ref:`well-defined and easily described
30 ways <intro_debugopt>` with the debugging information.
32 * Because LLVM is designed to support arbitrary programming languages,
33 LLVM-to-LLVM tools should not need to know anything about the semantics of
34 the source-level-language.
36 * Source-level languages are often **widely** different from one another.
37 LLVM should not put any restrictions of the flavor of the source-language,
38 and the debugging information should work with any language.
40 * With code generator support, it should be possible to use an LLVM compiler
41 to compile a program to native machine code and standard debugging
42 formats. This allows compatibility with traditional machine-code level
43 debuggers, like GDB or DBX.
45 The approach used by the LLVM implementation is to use a small set of
46 :ref:`intrinsic functions <format_common_intrinsics>` to define a mapping
47 between LLVM program objects and the source-level objects. The description of
48 the source-level program is maintained in LLVM metadata in an
49 :ref:`implementation-defined format <ccxx_frontend>` (the C/C++ front-end
50 currently uses working draft 7 of the `DWARF 3 standard
51 <http://www.eagercon.com/dwarf/dwarf3std.htm>`_).
53 When a program is being debugged, a debugger interacts with the user and turns
54 the stored debug information into source-language specific information. As
55 such, a debugger must be aware of the source-language, and is thus tied to a
56 specific language or family of languages.
58 Debug information consumers
59 ---------------------------
61 The role of debug information is to provide meta information normally stripped
62 away during the compilation process. This meta information provides an LLVM
63 user a relationship between generated code and the original program source
66 Currently, debug information is consumed by DwarfDebug to produce dwarf
67 information used by the gdb debugger. Other targets could use the same
68 information to produce stabs or other debug forms.
70 It would also be reasonable to use debug information to feed profiling tools
71 for analysis of generated code, or, tools for reconstructing the original
72 source from generated code.
74 TODO - expound a bit more.
78 Debugging optimized code
79 ------------------------
81 An extremely high priority of LLVM debugging information is to make it interact
82 well with optimizations and analysis. In particular, the LLVM debug
83 information provides the following guarantees:
85 * LLVM debug information **always provides information to accurately read
86 the source-level state of the program**, regardless of which LLVM
87 optimizations have been run, and without any modification to the
88 optimizations themselves. However, some optimizations may impact the
89 ability to modify the current state of the program with a debugger, such
90 as setting program variables, or calling functions that have been
93 * As desired, LLVM optimizations can be upgraded to be aware of the LLVM
94 debugging information, allowing them to update the debugging information
95 as they perform aggressive optimizations. This means that, with effort,
96 the LLVM optimizers could optimize debug code just as well as non-debug
99 * LLVM debug information does not prevent optimizations from
100 happening (for example inlining, basic block reordering/merging/cleanup,
101 tail duplication, etc).
103 * LLVM debug information is automatically optimized along with the rest of
104 the program, using existing facilities. For example, duplicate
105 information is automatically merged by the linker, and unused information
106 is automatically removed.
108 Basically, the debug information allows you to compile a program with
109 "``-O0 -g``" and get full debug information, allowing you to arbitrarily modify
110 the program as it executes from a debugger. Compiling a program with
111 "``-O3 -g``" gives you full debug information that is always available and
112 accurate for reading (e.g., you get accurate stack traces despite tail call
113 elimination and inlining), but you might lose the ability to modify the program
114 and call functions where were optimized out of the program, or inlined away
117 :ref:`LLVM test suite <test-suite-quickstart>` provides a framework to test
118 optimizer's handling of debugging information. It can be run like this:
122 % cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
125 This will test impact of debugging information on optimization passes. If
126 debugging information influences optimization passes then it will be reported
127 as a failure. See :doc:`TestingGuide` for more information on LLVM test
128 infrastructure and how to run various tests.
132 Debugging information format
133 ============================
135 LLVM debugging information has been carefully designed to make it possible for
136 the optimizer to optimize the program and debugging information without
137 necessarily having to know anything about debugging information. In
138 particular, the use of metadata avoids duplicated debugging information from
139 the beginning, and the global dead code elimination pass automatically deletes
140 debugging information for a function if it decides to delete the function.
142 To do this, most of the debugging information (descriptors for types,
143 variables, functions, source files, etc) is inserted by the language front-end
144 in the form of LLVM metadata.
146 Debug information is designed to be agnostic about the target debugger and
147 debugging information representation (e.g. DWARF/Stabs/etc). It uses a generic
148 pass to decode the information that represents variables, types, functions,
149 namespaces, etc: this allows for arbitrary source-language semantics and
150 type-systems to be used, as long as there is a module written for the target
151 debugger to interpret the information.
153 To provide basic functionality, the LLVM debugger does have to make some
154 assumptions about the source-level language being debugged, though it keeps
155 these to a minimum. The only common features that the LLVM debugger assumes
156 exist are :ref:`source files <format_files>`, and :ref:`program objects
157 <format_global_variables>`. These abstract objects are used by a debugger to
158 form stack traces, show information about local variables, etc.
160 This section of the documentation first describes the representation aspects
161 common to any source-language. :ref:`ccxx_frontend` describes the data layout
162 conventions used by the C and C++ front-ends.
164 Debug information descriptors
165 -----------------------------
167 In consideration of the complexity and volume of debug information, LLVM
168 provides a specification for well formed debug descriptors.
170 Consumers of LLVM debug information expect the descriptors for program objects
171 to start in a canonical format, but the descriptors can include additional
172 information appended at the end that is source-language specific. All debugging
173 information objects start with a tag to indicate what type of object it is.
174 The source-language is allowed to define its own objects, by using unreserved
175 tag numbers. We recommend using with tags in the range 0x1000 through 0x2000
176 (there is a defined ``enum DW_TAG_user_base = 0x1000``.)
178 The fields of debug descriptors used internally by LLVM are restricted to only
179 the simple data types ``i32``, ``i1``, ``float``, ``double``, ``mdstring`` and
189 <a name="LLVMDebugVersion">The first field of a descriptor is always an
190 ``i32`` containing a tag value identifying the content of the descriptor.
191 The remaining fields are specific to the descriptor. The values of tags are
192 loosely bound to the tag values of DWARF information entries. However, that
193 does not restrict the use of the information supplied to DWARF targets.
195 The details of the various descriptors follow.
197 Compile unit descriptors
198 ^^^^^^^^^^^^^^^^^^^^^^^^
203 i32, ;; Tag = 17 (DW_TAG_compile_unit)
204 metadata, ;; Source directory (including trailing slash) & file pair
205 i32, ;; DWARF language identifier (ex. DW_LANG_C89)
206 metadata ;; Producer (ex. "4.0.1 LLVM (LLVM research group)")
207 i1, ;; True if this is optimized.
209 i32 ;; Runtime version
210 metadata ;; List of enums types
211 metadata ;; List of retained types
212 metadata ;; List of subprograms
213 metadata ;; List of global variables
214 metadata ;; List of imported entities
215 metadata ;; Split debug filename
218 These descriptors contain a source language ID for the file (we use the DWARF
219 3.0 ID numbers, such as ``DW_LANG_C89``, ``DW_LANG_C_plus_plus``,
220 ``DW_LANG_Cobol74``, etc), a reference to a metadata node containing a pair of
221 strings for the source file name and the working directory, as well as an
222 identifier string for the compiler that produced it.
224 Compile unit descriptors provide the root context for objects declared in a
225 specific compilation unit. File descriptors are defined using this context.
226 These descriptors are collected by a named metadata ``!llvm.dbg.cu``. They
227 keep track of subprograms, global variables, type information, and imported
228 entities (declarations and namespaces).
238 i32, ;; Tag = 41 (DW_TAG_file_type)
239 metadata, ;; Source directory (including trailing slash) & file pair
242 These descriptors contain information for a file. Global variables and top
243 level functions would be defined using this context. File descriptors also
244 provide context for source line correspondence.
246 Each input file is encoded as a separate file descriptor in LLVM debugging
249 .. _format_global_variables:
251 Global variable descriptors
252 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
257 i32, ;; Tag = 52 (DW_TAG_variable)
258 i32, ;; Unused field.
259 metadata, ;; Reference to context descriptor
261 metadata, ;; Display name (fully qualified C++ name)
262 metadata, ;; MIPS linkage name (for C++)
263 metadata, ;; Reference to file where defined
264 i32, ;; Line number where defined
265 metadata, ;; Reference to type descriptor
266 i1, ;; True if the global is local to compile unit (static)
267 i1, ;; True if the global is defined in the compile unit (not extern)
268 {}*, ;; Reference to the global variable
269 metadata, ;; The static member declaration, if any
272 These descriptors provide debug information about globals variables. They
273 provide details such as name, type and where the variable is defined. All
274 global variables are collected inside the named metadata ``!llvm.dbg.cu``.
276 .. _format_subprograms:
278 Subprogram descriptors
279 ^^^^^^^^^^^^^^^^^^^^^^
284 i32, ;; Tag = 46 (DW_TAG_subprogram)
285 metadata, ;; Source directory (including trailing slash) & file pair
286 metadata, ;; Reference to context descriptor
288 metadata, ;; Display name (fully qualified C++ name)
289 metadata, ;; MIPS linkage name (for C++)
290 i32, ;; Line number where defined
291 metadata, ;; Reference to type descriptor
292 i1, ;; True if the global is local to compile unit (static)
293 i1, ;; True if the global is defined in the compile unit (not extern)
294 i32, ;; Virtuality, e.g. dwarf::DW_VIRTUALITY__virtual
295 i32, ;; Index into a virtual function
296 metadata, ;; indicates which base type contains the vtable pointer for the
298 i32, ;; Flags - Artifical, Private, Protected, Explicit, Prototyped.
300 Function * , ;; Pointer to LLVM function
301 metadata, ;; Lists function template parameters
302 metadata, ;; Function declaration descriptor
303 metadata, ;; List of function variables
304 i32 ;; Line number where the scope of the subprogram begins
307 These descriptors provide debug information about functions, methods and
308 subprograms. They provide details such as name, return types and the source
309 location where the subprogram is defined.
317 i32, ;; Tag = 11 (DW_TAG_lexical_block)
318 metadata,;; Source directory (including trailing slash) & file pair
319 metadata,;; Reference to context descriptor
321 i32, ;; Column number
322 i32 ;; Unique ID to identify blocks from a template function
325 This descriptor provides debug information about nested blocks within a
326 subprogram. The line number and column numbers are used to dinstinguish two
327 lexical blocks at same depth.
332 i32, ;; Tag = 11 (DW_TAG_lexical_block)
333 metadata,;; Source directory (including trailing slash) & file pair
334 metadata ;; Reference to the scope we're annotating with a file change
337 This descriptor provides a wrapper around a lexical scope to handle file
338 changes in the middle of a lexical block.
340 .. _format_basic_type:
342 Basic type descriptors
343 ^^^^^^^^^^^^^^^^^^^^^^
348 i32, ;; Tag = 36 (DW_TAG_base_type)
349 metadata,;; Source directory (including trailing slash) & file pair (may be null)
350 metadata, ;; Reference to context
351 metadata, ;; Name (may be "" for anonymous types)
352 i32, ;; Line number where defined (may be 0)
354 i64, ;; Alignment in bits
355 i64, ;; Offset in bits
357 i32 ;; DWARF type encoding
360 These descriptors define primitive types used in the code. Example ``int``,
361 ``bool`` and ``float``. The context provides the scope of the type, which is
362 usually the top level. Since basic types are not usually user defined the
363 context and line number can be left as NULL and 0. The size, alignment and
364 offset are expressed in bits and can be 64 bit values. The alignment is used
365 to round the offset when embedded in a :ref:`composite type
366 <format_composite_type>` (example to keep float doubles on 64 bit boundaries).
367 The offset is the bit offset if embedded in a :ref:`composite type
368 <format_composite_type>`.
370 The type encoding provides the details of the type. The values are typically
371 one of the following:
379 DW_ATE_signed_char = 6
381 DW_ATE_unsigned_char = 8
383 .. _format_derived_type:
385 Derived type descriptors
386 ^^^^^^^^^^^^^^^^^^^^^^^^
391 i32, ;; Tag (see below)
392 metadata,;; Source directory (including trailing slash) & file pair (may be null)
393 metadata, ;; Reference to context
394 metadata, ;; Name (may be "" for anonymous types)
395 i32, ;; Line number where defined (may be 0)
397 i64, ;; Alignment in bits
398 i64, ;; Offset in bits
399 i32, ;; Flags to encode attributes, e.g. private
400 metadata, ;; Reference to type derived from
401 metadata, ;; (optional) Name of the Objective C property associated with
402 ;; Objective-C an ivar, or the type of which this
403 ;; pointer-to-member is pointing to members of.
404 metadata, ;; (optional) Name of the Objective C property getter selector.
405 metadata, ;; (optional) Name of the Objective C property setter selector.
406 i32 ;; (optional) Objective C property attributes.
409 These descriptors are used to define types derived from other types. The value
410 of the tag varies depending on the meaning. The following are possible tag
415 DW_TAG_formal_parameter = 5
417 DW_TAG_pointer_type = 15
418 DW_TAG_reference_type = 16
420 DW_TAG_ptr_to_member_type = 31
421 DW_TAG_const_type = 38
422 DW_TAG_volatile_type = 53
423 DW_TAG_restrict_type = 55
425 ``DW_TAG_member`` is used to define a member of a :ref:`composite type
426 <format_composite_type>` or :ref:`subprogram <format_subprograms>`. The type
427 of the member is the :ref:`derived type <format_derived_type>`.
428 ``DW_TAG_formal_parameter`` is used to define a member which is a formal
429 argument of a subprogram.
431 ``DW_TAG_typedef`` is used to provide a name for the derived type.
433 ``DW_TAG_pointer_type``, ``DW_TAG_reference_type``, ``DW_TAG_const_type``,
434 ``DW_TAG_volatile_type`` and ``DW_TAG_restrict_type`` are used to qualify the
435 :ref:`derived type <format_derived_type>`.
437 :ref:`Derived type <format_derived_type>` location can be determined from the
438 context and line number. The size, alignment and offset are expressed in bits
439 and can be 64 bit values. The alignment is used to round the offset when
440 embedded in a :ref:`composite type <format_composite_type>` (example to keep
441 float doubles on 64 bit boundaries.) The offset is the bit offset if embedded
442 in a :ref:`composite type <format_composite_type>`.
444 Note that the ``void *`` type is expressed as a type derived from NULL.
446 .. _format_composite_type:
448 Composite type descriptors
449 ^^^^^^^^^^^^^^^^^^^^^^^^^^
454 i32, ;; Tag (see below)
455 metadata,;; Source directory (including trailing slash) & file pair (may be null)
456 metadata, ;; Reference to context
457 metadata, ;; Name (may be "" for anonymous types)
458 i32, ;; Line number where defined (may be 0)
460 i64, ;; Alignment in bits
461 i64, ;; Offset in bits
463 metadata, ;; Reference to type derived from
464 metadata, ;; Reference to array of member descriptors
465 i32 ;; Runtime languages
466 metadata, ;; Base type containing the vtable pointer for this type
467 metadata ;; Template parameters
470 These descriptors are used to define types that are composed of 0 or more
471 elements. The value of the tag varies depending on the meaning. The following
472 are possible tag values:
476 DW_TAG_array_type = 1
477 DW_TAG_enumeration_type = 4
478 DW_TAG_structure_type = 19
479 DW_TAG_union_type = 23
480 DW_TAG_subroutine_type = 21
481 DW_TAG_inheritance = 28
483 The vector flag indicates that an array type is a native packed vector.
485 The members of array types (tag = ``DW_TAG_array_type``) are
486 :ref:`subrange descriptors <format_subrange>`, each
487 representing the range of subscripts at that level of indexing.
489 The members of enumeration types (tag = ``DW_TAG_enumeration_type``) are
490 :ref:`enumerator descriptors <format_enumerator>`, each representing the
491 definition of enumeration value for the set. All enumeration type descriptors
492 are collected inside the named metadata ``!llvm.dbg.cu``.
494 The members of structure (tag = ``DW_TAG_structure_type``) or union (tag =
495 ``DW_TAG_union_type``) types are any one of the :ref:`basic
496 <format_basic_type>`, :ref:`derived <format_derived_type>` or :ref:`composite
497 <format_composite_type>` type descriptors, each representing a field member of
498 the structure or union.
500 For C++ classes (tag = ``DW_TAG_structure_type``), member descriptors provide
501 information about base classes, static members and member functions. If a
502 member is a :ref:`derived type descriptor <format_derived_type>` and has a tag
503 of ``DW_TAG_inheritance``, then the type represents a base class. If the member
504 of is a :ref:`global variable descriptor <format_global_variables>` then it
505 represents a static member. And, if the member is a :ref:`subprogram
506 descriptor <format_subprograms>` then it represents a member function. For
507 static members and member functions, ``getName()`` returns the members link or
508 the C++ mangled name. ``getDisplayName()`` the simplied version of the name.
510 The first member of subroutine (tag = ``DW_TAG_subroutine_type``) type elements
511 is the return type for the subroutine. The remaining elements are the formal
512 arguments to the subroutine.
514 :ref:`Composite type <format_composite_type>` location can be determined from
515 the context and line number. The size, alignment and offset are expressed in
516 bits and can be 64 bit values. The alignment is used to round the offset when
517 embedded in a :ref:`composite type <format_composite_type>` (as an example, to
518 keep float doubles on 64 bit boundaries). The offset is the bit offset if
519 embedded in a :ref:`composite type <format_composite_type>`.
529 i32, ;; Tag = 33 (DW_TAG_subrange_type)
534 These descriptors are used to define ranges of array subscripts for an array
535 :ref:`composite type <format_composite_type>`. The low value defines the lower
536 bounds typically zero for C/C++. The high value is the upper bounds. Values
537 are 64 bit. ``High - Low + 1`` is the size of the array. If ``Low > High``
538 the array bounds are not included in generated debugging information.
540 .. _format_enumerator:
542 Enumerator descriptors
543 ^^^^^^^^^^^^^^^^^^^^^^
548 i32, ;; Tag = 40 (DW_TAG_enumerator)
553 These descriptors are used to define members of an enumeration :ref:`composite
554 type <format_composite_type>`, it associates the name to the value.
562 i32, ;; Tag (see below)
565 metadata, ;; Reference to file where defined
566 i32, ;; 24 bit - Line number where defined
567 ;; 8 bit - Argument number. 1 indicates 1st argument.
568 metadata, ;; Type descriptor
570 metadata ;; (optional) Reference to inline location
573 These descriptors are used to define variables local to a sub program. The
574 value of the tag depends on the usage of the variable:
578 DW_TAG_auto_variable = 256
579 DW_TAG_arg_variable = 257
581 An auto variable is any variable declared in the body of the function. An
582 argument variable is any variable that appears as a formal argument to the
585 The context is either the subprogram or block where the variable is defined.
586 Name the source variable name. Context and line indicate where the variable
587 was defined. Type descriptor defines the declared type of the variable.
589 .. _format_common_intrinsics:
591 Debugger intrinsic functions
592 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
594 LLVM uses several intrinsic functions (name prefixed with "``llvm.dbg``") to
595 provide debug information at various points in generated code.
602 void %llvm.dbg.declare(metadata, metadata)
604 This intrinsic provides information about a local element (e.g., variable).
605 The first argument is metadata holding the alloca for the variable. The second
606 argument is metadata containing a description of the variable.
613 void %llvm.dbg.value(metadata, i64, metadata)
615 This intrinsic provides information when a user source variable is set to a new
616 value. The first argument is the new value (wrapped as metadata). The second
617 argument is the offset in the user source variable where the new value is
618 written. The third argument is metadata containing a description of the user
621 Object lifetimes and scoping
622 ============================
624 In many languages, the local variables in functions can have their lifetimes or
625 scopes limited to a subset of a function. In the C family of languages, for
626 example, variables are only live (readable and writable) within the source
627 block that they are defined in. In functional languages, values are only
628 readable after they have been defined. Though this is a very obvious concept,
629 it is non-trivial to model in LLVM, because it has no notion of scoping in this
630 sense, and does not want to be tied to a language's scoping rules.
632 In order to handle this, the LLVM debug format uses the metadata attached to
633 llvm instructions to encode line number and scoping information. Consider the
634 following C fragment, for example:
648 Compiled to LLVM, this function would be represented like this:
652 define void @_Z3foov() #0 {
654 %X = alloca i32, align 4 ; [#uses=3 type=i32*]
655 %Y = alloca i32, align 4 ; [#uses=2 type=i32*]
656 %Z = alloca i32, align 4 ; [#uses=2 type=i32*]
657 call void @llvm.dbg.declare(metadata !{i32* %X}, metadata !8), !dbg !10
658 ; [debug line = 2:7] [debug variable = X]
659 store i32 21, i32* %X, align 4, !dbg !11 ; [debug line = 2:13]
660 call void @llvm.dbg.declare(metadata !{i32* %Y}, metadata !12), !dbg !13
661 ; [debug line = 3:7] [debug variable = Y]
662 store i32 22, i32* %Y, align 4, !dbg !14 ; [debug line = 3:13]
663 call void @llvm.dbg.declare(metadata !{i32* %Z}, metadata !15), !dbg !17
664 ; [debug line = 5:9] [debug variable = Z]
665 store i32 23, i32* %Z, align 4, !dbg !18 ; [debug line = 5:15]
666 %0 = load i32* %X, align 4, !dbg !19 ; [#uses=1 type=i32] \
668 store i32 %0, i32* %Z, align 4, !dbg !19 ; [debug line = 6:5]
669 %1 = load i32* %Y, align 4, !dbg !20 ; [#uses=1 type=i32] \
671 store i32 %1, i32* %X, align 4, !dbg !20 ; [debug line = 8:3]
672 ret void, !dbg !21 ; [debug line = 9:1]
676 ; Function Attrs: nounwind readnone
677 declare void @llvm.dbg.declare(metadata, metadata) #1
679 attributes #0 = { optsize zeroext "less-precise-fpmad"="false"
680 "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"="true"
681 "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false"
682 "use-soft-float"="false" }
683 attributes #1 = { nounwind readnone }
687 !0 = metadata !{i32 786449, metadata !1, i32 12,
688 metadata !"clang version 3.4 ", i1 false, metadata !"", i32 0,
689 metadata !2, metadata !2, metadata !3, metadata !2,
690 metadata !2, metadata !""} ; [ DW_TAG_compile_unit ] \
691 [/private/tmp/foo.c] \
693 !1 = metadata !{metadata !"foo.c", metadata !"/private/tmp"}
694 !2 = metadata !{i32 0}
695 !3 = metadata !{metadata !4}
696 !4 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"foo",
697 metadata !"foo", metadata !"_Z3foov", i32 1, metadata !6,
698 i1 false, i1 true, i32 0, i32 0, null, i32 256, i1 false,
699 void ()* @_Z3foov, null, null, metadata !2, i32 1}
700 ; [ DW_TAG_subprogram ] [line 1] [def] [foo]
701 !5 = metadata !{i32 786473, metadata !1} ; [ DW_TAG_file_type ] \
703 !6 = metadata !{i32 786453, i32 0, i32 0, metadata !"", i32 0, i64 0, i64 0,
704 i64 0, i32 0, null, metadata !7, i32 0, i32 0}
705 ; [ DW_TAG_subroutine_type ] \
706 [line 0, size 0, align 0, offset 0] [from ]
707 !7 = metadata !{null}
708 !8 = metadata !{i32 786688, metadata !4, metadata !"X", metadata !5, i32 2, \
709 metadata !9, i32 0, i32 0} ; [ DW_TAG_auto_variable ] [X] \
711 !9 = metadata !{i32 786468, null, null, metadata !"int", i32 0, i64 32, \
712 i64 32, i64 0, i32 0, i32 5} ; [ DW_TAG_base_type ] [int] \
713 [line 0, size 32, align 32, offset 0, enc DW_ATE_signed]
714 !10 = metadata !{i32 2, i32 7, metadata !4, null}
715 !11 = metadata !{i32 2, i32 13, metadata !4, null}
716 !12 = metadata !{i32 786688, metadata !4, metadata !"Y", metadata !5, i32 3, \
717 metadata !9, i32 0, i32 0} ; [ DW_TAG_auto_variable ] [Y] \
719 !13 = metadata !{i32 3, i32 7, metadata !4, null}
720 !14 = metadata !{i32 3, i32 13, metadata !4, null}
721 !15 = metadata !{i32 786688, metadata !16, metadata !"Z", metadata !5, i32 5, \
722 metadata !9, i32 0, i32 0} ; [ DW_TAG_auto_variable ] [Z] \
724 !16 = metadata !{i32 786443, metadata !1, metadata !4, i32 4, i32 3, i32 0}
725 ; [ DW_TAG_lexical_block ] [/private/tmp/foo.c]
726 !17 = metadata !{i32 5, i32 9, metadata !16, null}
727 !18 = metadata !{i32 5, i32 15, metadata !16, null}
728 !19 = metadata !{i32 6, i32 5, metadata !16, null}
729 !20 = metadata !{i32 8, i32 3, metadata !4, null}
730 !21 = metadata !{i32 9, i32 1, metadata !4, null}
732 This example illustrates a few important details about LLVM debugging
733 information. In particular, it shows how the ``llvm.dbg.declare`` intrinsic and
734 location information, which are attached to an instruction, are applied
735 together to allow a debugger to analyze the relationship between statements,
736 variable definitions, and the code used to implement the function.
740 call void @llvm.dbg.declare(metadata !{i32* %X}, metadata !8), !dbg !10
741 ; [debug line = 2:7] [debug variable = X]
743 The first intrinsic ``%llvm.dbg.declare`` encodes debugging information for the
744 variable ``X``. The metadata ``!dbg !10`` attached to the intrinsic provides
745 scope information for the variable ``X``.
749 !10 = metadata !{i32 2, i32 7, metadata !4, null}
750 !4 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"foo",
751 metadata !"foo", metadata !"_Z3foov", i32 1, metadata !6,
752 i1 false, i1 true, i32 0, i32 0, null, i32 256, i1 false,
753 void ()* @_Z3foov, null, null, metadata !2, i32 1}
754 ; [ DW_TAG_subprogram ] [line 1] [def] [foo]
756 Here ``!10`` is metadata providing location information. It has four fields:
757 line number, column number, scope, and original scope. The original scope
758 represents inline location if this instruction is inlined inside a caller, and
759 is null otherwise. In this example, scope is encoded by ``!4``, a
760 :ref:`subprogram descriptor <format_subprograms>`. This way the location
761 information attached to the intrinsics indicates that the variable ``X`` is
762 declared at line number 2 at a function level scope in function ``foo``.
764 Now lets take another example.
768 call void @llvm.dbg.declare(metadata !{i32* %Z}, metadata !15), !dbg !17
769 ; [debug line = 5:9] [debug variable = Z]
771 The third intrinsic ``%llvm.dbg.declare`` encodes debugging information for
772 variable ``Z``. The metadata ``!dbg !17`` attached to the intrinsic provides
773 scope information for the variable ``Z``.
777 !16 = metadata !{i32 786443, metadata !1, metadata !4, i32 4, i32 3, i32 0}
778 ; [ DW_TAG_lexical_block ] [/private/tmp/foo.c]
779 !17 = metadata !{i32 5, i32 9, metadata !16, null}
781 Here ``!15`` indicates that ``Z`` is declared at line number 5 and
782 column number 9 inside of lexical scope ``!16``. The lexical scope itself
783 resides inside of subprogram ``!4`` described above.
785 The scope information attached with each instruction provides a straightforward
786 way to find instructions covered by a scope.
790 C/C++ front-end specific debug information
791 ==========================================
793 The C and C++ front-ends represent information about the program in a format
794 that is effectively identical to `DWARF 3.0
795 <http://www.eagercon.com/dwarf/dwarf3std.htm>`_ in terms of information
796 content. This allows code generators to trivially support native debuggers by
797 generating standard dwarf information, and contains enough information for
798 non-dwarf targets to translate it as needed.
800 This section describes the forms used to represent C and C++ programs. Other
801 languages could pattern themselves after this (which itself is tuned to
802 representing programs in the same way that DWARF 3 does), or they could choose
803 to provide completely different forms if they don't fit into the DWARF model.
804 As support for debugging information gets added to the various LLVM
805 source-language front-ends, the information used should be documented here.
807 The following sections provide examples of various C/C++ constructs and the
808 debug information that would best describe those constructs.
810 C/C++ source file information
811 -----------------------------
813 Given the source files ``MySource.cpp`` and ``MyHeader.h`` located in the
814 directory ``/Users/mine/sources``, the following code:
818 #include "MyHeader.h"
820 int main(int argc, char *argv[]) {
824 a C/C++ front-end would generate the following descriptors:
830 ;; Define the compile unit for the main source file "/Users/mine/sources/MySource.cpp".
834 metadata !1, ;; File/directory name
835 i32 4, ;; Language Id
836 metadata !"clang version 3.4 ",
837 i1 false, ;; Optimized compile unit
838 metadata !"", ;; Compiler flags
839 i32 0, ;; Runtime version
840 metadata !2, ;; Enumeration types
841 metadata !2, ;; Retained types
842 metadata !3, ;; Subprograms
843 metadata !2, ;; Global variables
844 metadata !2, ;; Imported entities (declarations and namespaces)
845 metadata !"" ;; Split debug filename
849 ;; Define the file for the file "/Users/mine/sources/MySource.cpp".
852 metadata !"MySource.cpp",
853 metadata !"/Users/mine/sources"
861 ;; Define the file for the file "/Users/mine/sources/Myheader.h"
868 metadata !"./MyHeader.h",
869 metadata !"/Users/mine/sources",
874 ``llvm::Instruction`` provides easy access to metadata attached with an
875 instruction. One can extract line number information encoded in LLVM IR using
876 ``Instruction::getMetadata()`` and ``DILocation::getLineNumber()``.
880 if (MDNode *N = I->getMetadata("dbg")) { // Here I is an LLVM instruction
881 DILocation Loc(N); // DILocation is in DebugInfo.h
882 unsigned Line = Loc.getLineNumber();
883 StringRef File = Loc.getFilename();
884 StringRef Dir = Loc.getDirectory();
887 C/C++ global variable information
888 ---------------------------------
890 Given an integer global variable declared as follows:
896 a C/C++ front-end would generate the following descriptors:
901 ;; Define the global itself.
903 %MyGlobal = global int 100
906 ;; List of debug info of globals
910 ;; Define the compile unit.
915 metadata !"foo.cpp", ;; File
916 metadata !"/Volumes/Data/tmp", ;; Directory
917 metadata !"clang version 3.1 ", ;; Producer
918 i1 true, ;; Deprecated field
919 i1 false, ;; "isOptimized"?
920 metadata !"", ;; Flags
921 i32 0, ;; Runtime Version
922 metadata !1, ;; Enum Types
923 metadata !1, ;; Retained Types
924 metadata !1, ;; Subprograms
925 metadata !3, ;; Global Variables
926 metadata !1, ;; Imported entities
927 "", ;; Split debug filename
928 } ; [ DW_TAG_compile_unit ]
930 ;; The Array of Global Variables
936 ;; Define the global variable itself.
942 metadata !"MyGlobal", ;; Name
943 metadata !"MyGlobal", ;; Display Name
944 metadata !"", ;; Linkage Name
948 i32 0, ;; IsLocalToUnit
949 i32 1, ;; IsDefinition
950 i32* @MyGlobal, ;; LLVM-IR Value
951 null ;; Static member declaration
952 } ; [ DW_TAG_variable ]
958 metadata !"foo.cpp", ;; File
959 metadata !"/Volumes/Data/tmp", ;; Directory
963 metadata !5 ;; Unused
964 } ; [ DW_TAG_file_type ]
973 metadata !"int", ;; Name
975 i64 32, ;; Size in Bits
976 i64 32, ;; Align in Bits
980 } ; [ DW_TAG_base_type ]
982 C/C++ function information
983 --------------------------
985 Given a function declared as follows:
989 int main(int argc, char *argv[]) {
993 a C/C++ front-end would generate the following descriptors:
998 ;; Define the anchor for subprograms.
1002 metadata !1, ;; File
1003 metadata !1, ;; Context
1004 metadata !"main", ;; Name
1005 metadata !"main", ;; Display name
1006 metadata !"main", ;; Linkage name
1007 i32 1, ;; Line number
1008 metadata !4, ;; Type
1009 i1 false, ;; Is local
1010 i1 true, ;; Is definition
1011 i32 0, ;; Virtuality attribute, e.g. pure virtual function
1012 i32 0, ;; Index into virtual table for C++ methods
1013 i32 0, ;; Type that holds virtual table.
1015 i1 false, ;; True if this function is optimized
1016 Function *, ;; Pointer to llvm::Function
1017 null, ;; Function template parameters
1018 null, ;; List of function variables (emitted when optimizing)
1019 1 ;; Line number of the opening '{' of the function
1022 ;; Define the subprogram itself.
1024 define i32 @main(i32 %argc, i8** %argv) {
1031 The following are the basic type descriptors for C/C++ core types:
1036 .. code-block:: llvm
1042 metadata !"bool", ;; Name
1043 i32 0, ;; Line number
1044 i64 8, ;; Size in Bits
1045 i64 8, ;; Align in Bits
1046 i64 0, ;; Offset in Bits
1054 .. code-block:: llvm
1060 metadata !"char", ;; Name
1061 i32 0, ;; Line number
1062 i64 8, ;; Size in Bits
1063 i64 8, ;; Align in Bits
1064 i64 0, ;; Offset in Bits
1072 .. code-block:: llvm
1078 metadata !"unsigned char",
1079 i32 0, ;; Line number
1080 i64 8, ;; Size in Bits
1081 i64 8, ;; Align in Bits
1082 i64 0, ;; Offset in Bits
1090 .. code-block:: llvm
1096 metadata !"short int",
1097 i32 0, ;; Line number
1098 i64 16, ;; Size in Bits
1099 i64 16, ;; Align in Bits
1100 i64 0, ;; Offset in Bits
1108 .. code-block:: llvm
1114 metadata !"short unsigned int",
1115 i32 0, ;; Line number
1116 i64 16, ;; Size in Bits
1117 i64 16, ;; Align in Bits
1118 i64 0, ;; Offset in Bits
1126 .. code-block:: llvm
1132 metadata !"int", ;; Name
1133 i32 0, ;; Line number
1134 i64 32, ;; Size in Bits
1135 i64 32, ;; Align in Bits
1136 i64 0, ;; Offset in Bits
1144 .. code-block:: llvm
1150 metadata !"unsigned int",
1151 i32 0, ;; Line number
1152 i64 32, ;; Size in Bits
1153 i64 32, ;; Align in Bits
1154 i64 0, ;; Offset in Bits
1162 .. code-block:: llvm
1168 metadata !"long long int",
1169 i32 0, ;; Line number
1170 i64 64, ;; Size in Bits
1171 i64 64, ;; Align in Bits
1172 i64 0, ;; Offset in Bits
1180 .. code-block:: llvm
1186 metadata !"long long unsigned int",
1187 i32 0, ;; Line number
1188 i64 64, ;; Size in Bits
1189 i64 64, ;; Align in Bits
1190 i64 0, ;; Offset in Bits
1198 .. code-block:: llvm
1205 i32 0, ;; Line number
1206 i64 32, ;; Size in Bits
1207 i64 32, ;; Align in Bits
1208 i64 0, ;; Offset in Bits
1216 .. code-block:: llvm
1222 metadata !"double",;; Name
1223 i32 0, ;; Line number
1224 i64 64, ;; Size in Bits
1225 i64 64, ;; Align in Bits
1226 i64 0, ;; Offset in Bits
1234 Given the following as an example of C/C++ derived type:
1238 typedef const int *IntPtr;
1240 a C/C++ front-end would generate the following descriptors:
1242 .. code-block:: llvm
1245 ;; Define the typedef "IntPtr".
1249 metadata !3, ;; File
1250 metadata !1, ;; Context
1251 metadata !"IntPtr", ;; Name
1252 i32 0, ;; Line number
1253 i64 0, ;; Size in bits
1254 i64 0, ;; Align in bits
1255 i64 0, ;; Offset in bits
1257 metadata !4 ;; Derived From type
1260 ;; Define the pointer type.
1266 metadata !"", ;; Name
1267 i32 0, ;; Line number
1268 i64 64, ;; Size in bits
1269 i64 64, ;; Align in bits
1270 i64 0, ;; Offset in bits
1272 metadata !5 ;; Derived From type
1275 ;; Define the const type.
1281 metadata !"", ;; Name
1282 i32 0, ;; Line number
1283 i64 0, ;; Size in bits
1284 i64 0, ;; Align in bits
1285 i64 0, ;; Offset in bits
1287 metadata !6 ;; Derived From type
1290 ;; Define the int type.
1296 metadata !"int", ;; Name
1297 i32 0, ;; Line number
1298 i64 32, ;; Size in bits
1299 i64 32, ;; Align in bits
1300 i64 0, ;; Offset in bits
1305 C/C++ struct/union types
1306 ------------------------
1308 Given the following as an example of C/C++ struct type:
1318 a C/C++ front-end would generate the following descriptors:
1320 .. code-block:: llvm
1323 ;; Define basic type for unsigned int.
1329 metadata !"unsigned int",
1330 i32 0, ;; Line number
1331 i64 32, ;; Size in Bits
1332 i64 32, ;; Align in Bits
1333 i64 0, ;; Offset in Bits
1338 ;; Define composite type for struct Color.
1342 metadata !1, ;; Compile unit
1344 metadata !"Color", ;; Name
1345 i32 1, ;; Line number
1346 i64 96, ;; Size in bits
1347 i64 32, ;; Align in bits
1348 i64 0, ;; Offset in bits
1350 null, ;; Derived From
1351 metadata !3, ;; Elements
1352 i32 0, ;; Runtime Language
1353 null, ;; Base type containing the vtable pointer for this type
1354 null ;; Template parameters
1358 ;; Define the Red field.
1362 metadata !1, ;; File
1363 metadata !1, ;; Context
1364 metadata !"Red", ;; Name
1365 i32 2, ;; Line number
1366 i64 32, ;; Size in bits
1367 i64 32, ;; Align in bits
1368 i64 0, ;; Offset in bits
1370 metadata !5 ;; Derived From type
1374 ;; Define the Green field.
1378 metadata !1, ;; File
1379 metadata !1, ;; Context
1380 metadata !"Green", ;; Name
1381 i32 3, ;; Line number
1382 i64 32, ;; Size in bits
1383 i64 32, ;; Align in bits
1384 i64 32, ;; Offset in bits
1386 metadata !5 ;; Derived From type
1390 ;; Define the Blue field.
1394 metadata !1, ;; File
1395 metadata !1, ;; Context
1396 metadata !"Blue", ;; Name
1397 i32 4, ;; Line number
1398 i64 32, ;; Size in bits
1399 i64 32, ;; Align in bits
1400 i64 64, ;; Offset in bits
1402 metadata !5 ;; Derived From type
1406 ;; Define the array of fields used by the composite type Color.
1408 !3 = metadata !{metadata !4, metadata !6, metadata !7}
1410 C/C++ enumeration types
1411 -----------------------
1413 Given the following as an example of C/C++ enumeration type:
1423 a C/C++ front-end would generate the following descriptors:
1425 .. code-block:: llvm
1428 ;; Define composite type for enum Trees
1432 metadata !1, ;; File
1433 metadata !1, ;; Context
1434 metadata !"Trees", ;; Name
1435 i32 1, ;; Line number
1436 i64 32, ;; Size in bits
1437 i64 32, ;; Align in bits
1438 i64 0, ;; Offset in bits
1440 null, ;; Derived From type
1441 metadata !3, ;; Elements
1442 i32 0 ;; Runtime language
1446 ;; Define the array of enumerators used by composite type Trees.
1448 !3 = metadata !{metadata !4, metadata !5, metadata !6}
1451 ;; Define Spruce enumerator.
1453 !4 = metadata !{i32 786472, metadata !"Spruce", i64 100}
1456 ;; Define Oak enumerator.
1458 !5 = metadata !{i32 786472, metadata !"Oak", i64 200}
1461 ;; Define Maple enumerator.
1463 !6 = metadata !{i32 786472, metadata !"Maple", i64 300}
1465 Debugging information format
1466 ============================
1468 Debugging Information Extension for Objective C Properties
1469 ----------------------------------------------------------
1474 Objective C provides a simpler way to declare and define accessor methods using
1475 declared properties. The language provides features to declare a property and
1476 to let compiler synthesize accessor methods.
1478 The debugger lets developer inspect Objective C interfaces and their instance
1479 variables and class variables. However, the debugger does not know anything
1480 about the properties defined in Objective C interfaces. The debugger consumes
1481 information generated by compiler in DWARF format. The format does not support
1482 encoding of Objective C properties. This proposal describes DWARF extensions to
1483 encode Objective C properties, which the debugger can use to let developers
1484 inspect Objective C properties.
1489 Objective C properties exist separately from class members. A property can be
1490 defined only by "setter" and "getter" selectors, and be calculated anew on each
1491 access. Or a property can just be a direct access to some declared ivar.
1492 Finally it can have an ivar "automatically synthesized" for it by the compiler,
1493 in which case the property can be referred to in user code directly using the
1494 standard C dereference syntax as well as through the property "dot" syntax, but
1495 there is no entry in the ``@interface`` declaration corresponding to this ivar.
1497 To facilitate debugging, these properties we will add a new DWARF TAG into the
1498 ``DW_TAG_structure_type`` definition for the class to hold the description of a
1499 given property, and a set of DWARF attributes that provide said description.
1500 The property tag will also contain the name and declared type of the property.
1502 If there is a related ivar, there will also be a DWARF property attribute placed
1503 in the ``DW_TAG_member`` DIE for that ivar referring back to the property TAG
1504 for that property. And in the case where the compiler synthesizes the ivar
1505 directly, the compiler is expected to generate a ``DW_TAG_member`` for that
1506 ivar (with the ``DW_AT_artificial`` set to 1), whose name will be the name used
1507 to access this ivar directly in code, and with the property attribute pointing
1508 back to the property it is backing.
1510 The following examples will serve as illustration for our discussion:
1512 .. code-block:: objc
1524 @synthesize p2 = n2;
1527 This produces the following DWARF (this is a "pseudo dwarfdump" output):
1529 .. code-block:: none
1531 0x00000100: TAG_structure_type [7] *
1532 AT_APPLE_runtime_class( 0x10 )
1534 AT_decl_file( "Objc_Property.m" )
1537 0x00000110 TAG_APPLE_property
1539 AT_type ( {0x00000150} ( int ) )
1541 0x00000120: TAG_APPLE_property
1543 AT_type ( {0x00000150} ( int ) )
1545 0x00000130: TAG_member [8]
1547 AT_APPLE_property ( {0x00000110} "p1" )
1548 AT_type( {0x00000150} ( int ) )
1549 AT_artificial ( 0x1 )
1551 0x00000140: TAG_member [8]
1553 AT_APPLE_property ( {0x00000120} "p2" )
1554 AT_type( {0x00000150} ( int ) )
1556 0x00000150: AT_type( ( int ) )
1558 Note, the current convention is that the name of the ivar for an
1559 auto-synthesized property is the name of the property from which it derives
1560 with an underscore prepended, as is shown in the example. But we actually
1561 don't need to know this convention, since we are given the name of the ivar
1564 Also, it is common practice in ObjC to have different property declarations in
1565 the @interface and @implementation - e.g. to provide a read-only property in
1566 the interface,and a read-write interface in the implementation. In that case,
1567 the compiler should emit whichever property declaration will be in force in the
1568 current translation unit.
1570 Developers can decorate a property with attributes which are encoded using
1571 ``DW_AT_APPLE_property_attribute``.
1573 .. code-block:: objc
1575 @property (readonly, nonatomic) int pr;
1577 .. code-block:: none
1579 TAG_APPLE_property [8]
1581 AT_type ( {0x00000147} (int) )
1582 AT_APPLE_property_attribute (DW_APPLE_PROPERTY_readonly, DW_APPLE_PROPERTY_nonatomic)
1584 The setter and getter method names are attached to the property using
1585 ``DW_AT_APPLE_property_setter`` and ``DW_AT_APPLE_property_getter`` attributes.
1587 .. code-block:: objc
1590 @property (setter=myOwnP3Setter:) int p3;
1591 -(void)myOwnP3Setter:(int)a;
1596 -(void)myOwnP3Setter:(int)a{ }
1599 The DWARF for this would be:
1601 .. code-block:: none
1603 0x000003bd: TAG_structure_type [7] *
1604 AT_APPLE_runtime_class( 0x10 )
1606 AT_decl_file( "Objc_Property.m" )
1609 0x000003cd TAG_APPLE_property
1611 AT_APPLE_property_setter ( "myOwnP3Setter:" )
1612 AT_type( {0x00000147} ( int ) )
1614 0x000003f3: TAG_member [8]
1616 AT_type ( {0x00000147} ( int ) )
1617 AT_APPLE_property ( {0x000003cd} )
1618 AT_artificial ( 0x1 )
1623 +-----------------------+--------+
1625 +=======================+========+
1626 | DW_TAG_APPLE_property | 0x4200 |
1627 +-----------------------+--------+
1629 New DWARF Attributes
1630 ^^^^^^^^^^^^^^^^^^^^
1632 +--------------------------------+--------+-----------+
1633 | Attribute | Value | Classes |
1634 +================================+========+===========+
1635 | DW_AT_APPLE_property | 0x3fed | Reference |
1636 +--------------------------------+--------+-----------+
1637 | DW_AT_APPLE_property_getter | 0x3fe9 | String |
1638 +--------------------------------+--------+-----------+
1639 | DW_AT_APPLE_property_setter | 0x3fea | String |
1640 +--------------------------------+--------+-----------+
1641 | DW_AT_APPLE_property_attribute | 0x3feb | Constant |
1642 +--------------------------------+--------+-----------+
1647 +--------------------------------+-------+
1649 +================================+=======+
1650 | DW_AT_APPLE_PROPERTY_readonly | 0x1 |
1651 +--------------------------------+-------+
1652 | DW_AT_APPLE_PROPERTY_readwrite | 0x2 |
1653 +--------------------------------+-------+
1654 | DW_AT_APPLE_PROPERTY_assign | 0x4 |
1655 +--------------------------------+-------+
1656 | DW_AT_APPLE_PROPERTY_retain | 0x8 |
1657 +--------------------------------+-------+
1658 | DW_AT_APPLE_PROPERTY_copy | 0x10 |
1659 +--------------------------------+-------+
1660 | DW_AT_APPLE_PROPERTY_nonatomic | 0x20 |
1661 +--------------------------------+-------+
1663 Name Accelerator Tables
1664 -----------------------
1669 The "``.debug_pubnames``" and "``.debug_pubtypes``" formats are not what a
1670 debugger needs. The "``pub``" in the section name indicates that the entries
1671 in the table are publicly visible names only. This means no static or hidden
1672 functions show up in the "``.debug_pubnames``". No static variables or private
1673 class variables are in the "``.debug_pubtypes``". Many compilers add different
1674 things to these tables, so we can't rely upon the contents between gcc, icc, or
1677 The typical query given by users tends not to match up with the contents of
1678 these tables. For example, the DWARF spec states that "In the case of the name
1679 of a function member or static data member of a C++ structure, class or union,
1680 the name presented in the "``.debug_pubnames``" section is not the simple name
1681 given by the ``DW_AT_name attribute`` of the referenced debugging information
1682 entry, but rather the fully qualified name of the data or function member."
1683 So the only names in these tables for complex C++ entries is a fully
1684 qualified name. Debugger users tend not to enter their search strings as
1685 "``a::b::c(int,const Foo&) const``", but rather as "``c``", "``b::c``" , or
1686 "``a::b::c``". So the name entered in the name table must be demangled in
1687 order to chop it up appropriately and additional names must be manually entered
1688 into the table to make it effective as a name lookup table for debuggers to
1691 All debuggers currently ignore the "``.debug_pubnames``" table as a result of
1692 its inconsistent and useless public-only name content making it a waste of
1693 space in the object file. These tables, when they are written to disk, are not
1694 sorted in any way, leaving every debugger to do its own parsing and sorting.
1695 These tables also include an inlined copy of the string values in the table
1696 itself making the tables much larger than they need to be on disk, especially
1697 for large C++ programs.
1699 Can't we just fix the sections by adding all of the names we need to this
1700 table? No, because that is not what the tables are defined to contain and we
1701 won't know the difference between the old bad tables and the new good tables.
1702 At best we could make our own renamed sections that contain all of the data we
1705 These tables are also insufficient for what a debugger like LLDB needs. LLDB
1706 uses clang for its expression parsing where LLDB acts as a PCH. LLDB is then
1707 often asked to look for type "``foo``" or namespace "``bar``", or list items in
1708 namespace "``baz``". Namespaces are not included in the pubnames or pubtypes
1709 tables. Since clang asks a lot of questions when it is parsing an expression,
1710 we need to be very fast when looking up names, as it happens a lot. Having new
1711 accelerator tables that are optimized for very quick lookups will benefit this
1712 type of debugging experience greatly.
1714 We would like to generate name lookup tables that can be mapped into memory
1715 from disk, and used as is, with little or no up-front parsing. We would also
1716 be able to control the exact content of these different tables so they contain
1717 exactly what we need. The Name Accelerator Tables were designed to fix these
1718 issues. In order to solve these issues we need to:
1720 * Have a format that can be mapped into memory from disk and used as is
1721 * Lookups should be very fast
1722 * Extensible table format so these tables can be made by many producers
1723 * Contain all of the names needed for typical lookups out of the box
1724 * Strict rules for the contents of tables
1726 Table size is important and the accelerator table format should allow the reuse
1727 of strings from common string tables so the strings for the names are not
1728 duplicated. We also want to make sure the table is ready to be used as-is by
1729 simply mapping the table into memory with minimal header parsing.
1731 The name lookups need to be fast and optimized for the kinds of lookups that
1732 debuggers tend to do. Optimally we would like to touch as few parts of the
1733 mapped table as possible when doing a name lookup and be able to quickly find
1734 the name entry we are looking for, or discover there are no matches. In the
1735 case of debuggers we optimized for lookups that fail most of the time.
1737 Each table that is defined should have strict rules on exactly what is in the
1738 accelerator tables and documented so clients can rely on the content.
1743 Standard Hash Tables
1744 """"""""""""""""""""
1746 Typical hash tables have a header, buckets, and each bucket points to the
1749 .. code-block:: none
1759 The BUCKETS are an array of offsets to DATA for each hash:
1761 .. code-block:: none
1764 | 0x00001000 | BUCKETS[0]
1765 | 0x00002000 | BUCKETS[1]
1766 | 0x00002200 | BUCKETS[2]
1767 | 0x000034f0 | BUCKETS[3]
1769 | 0xXXXXXXXX | BUCKETS[n_buckets]
1772 So for ``bucket[3]`` in the example above, we have an offset into the table
1773 0x000034f0 which points to a chain of entries for the bucket. Each bucket must
1774 contain a next pointer, full 32 bit hash value, the string itself, and the data
1775 for the current string value.
1777 .. code-block:: none
1780 0x000034f0: | 0x00003500 | next pointer
1781 | 0x12345678 | 32 bit hash
1782 | "erase" | string value
1783 | data[n] | HashData for this bucket
1785 0x00003500: | 0x00003550 | next pointer
1786 | 0x29273623 | 32 bit hash
1787 | "dump" | string value
1788 | data[n] | HashData for this bucket
1790 0x00003550: | 0x00000000 | next pointer
1791 | 0x82638293 | 32 bit hash
1792 | "main" | string value
1793 | data[n] | HashData for this bucket
1796 The problem with this layout for debuggers is that we need to optimize for the
1797 negative lookup case where the symbol we're searching for is not present. So
1798 if we were to lookup "``printf``" in the table above, we would make a 32 hash
1799 for "``printf``", it might match ``bucket[3]``. We would need to go to the
1800 offset 0x000034f0 and start looking to see if our 32 bit hash matches. To do
1801 so, we need to read the next pointer, then read the hash, compare it, and skip
1802 to the next bucket. Each time we are skipping many bytes in memory and
1803 touching new cache pages just to do the compare on the full 32 bit hash. All
1804 of these accesses then tell us that we didn't have a match.
1809 To solve the issues mentioned above we have structured the hash tables a bit
1810 differently: a header, buckets, an array of all unique 32 bit hash values,
1811 followed by an array of hash value data offsets, one for each hash value, then
1812 the data for all hash values:
1814 .. code-block:: none
1828 The ``BUCKETS`` in the name tables are an index into the ``HASHES`` array. By
1829 making all of the full 32 bit hash values contiguous in memory, we allow
1830 ourselves to efficiently check for a match while touching as little memory as
1831 possible. Most often checking the 32 bit hash values is as far as the lookup
1832 goes. If it does match, it usually is a match with no collisions. So for a
1833 table with "``n_buckets``" buckets, and "``n_hashes``" unique 32 bit hash
1834 values, we can clarify the contents of the ``BUCKETS``, ``HASHES`` and
1837 .. code-block:: none
1839 .-------------------------.
1840 | HEADER.magic | uint32_t
1841 | HEADER.version | uint16_t
1842 | HEADER.hash_function | uint16_t
1843 | HEADER.bucket_count | uint32_t
1844 | HEADER.hashes_count | uint32_t
1845 | HEADER.header_data_len | uint32_t
1846 | HEADER_DATA | HeaderData
1847 |-------------------------|
1848 | BUCKETS | uint32_t[n_buckets] // 32 bit hash indexes
1849 |-------------------------|
1850 | HASHES | uint32_t[n_hashes] // 32 bit hash values
1851 |-------------------------|
1852 | OFFSETS | uint32_t[n_hashes] // 32 bit offsets to hash value data
1853 |-------------------------|
1855 `-------------------------'
1857 So taking the exact same data from the standard hash example above we end up
1860 .. code-block:: none
1870 | ... | BUCKETS[n_buckets]
1872 | 0x........ | HASHES[0]
1873 | 0x........ | HASHES[1]
1874 | 0x........ | HASHES[2]
1875 | 0x........ | HASHES[3]
1876 | 0x........ | HASHES[4]
1877 | 0x........ | HASHES[5]
1878 | 0x12345678 | HASHES[6] hash for BUCKETS[3]
1879 | 0x29273623 | HASHES[7] hash for BUCKETS[3]
1880 | 0x82638293 | HASHES[8] hash for BUCKETS[3]
1881 | 0x........ | HASHES[9]
1882 | 0x........ | HASHES[10]
1883 | 0x........ | HASHES[11]
1884 | 0x........ | HASHES[12]
1885 | 0x........ | HASHES[13]
1886 | 0x........ | HASHES[n_hashes]
1888 | 0x........ | OFFSETS[0]
1889 | 0x........ | OFFSETS[1]
1890 | 0x........ | OFFSETS[2]
1891 | 0x........ | OFFSETS[3]
1892 | 0x........ | OFFSETS[4]
1893 | 0x........ | OFFSETS[5]
1894 | 0x000034f0 | OFFSETS[6] offset for BUCKETS[3]
1895 | 0x00003500 | OFFSETS[7] offset for BUCKETS[3]
1896 | 0x00003550 | OFFSETS[8] offset for BUCKETS[3]
1897 | 0x........ | OFFSETS[9]
1898 | 0x........ | OFFSETS[10]
1899 | 0x........ | OFFSETS[11]
1900 | 0x........ | OFFSETS[12]
1901 | 0x........ | OFFSETS[13]
1902 | 0x........ | OFFSETS[n_hashes]
1910 0x000034f0: | 0x00001203 | .debug_str ("erase")
1911 | 0x00000004 | A 32 bit array count - number of HashData with name "erase"
1912 | 0x........ | HashData[0]
1913 | 0x........ | HashData[1]
1914 | 0x........ | HashData[2]
1915 | 0x........ | HashData[3]
1916 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1918 0x00003500: | 0x00001203 | String offset into .debug_str ("collision")
1919 | 0x00000002 | A 32 bit array count - number of HashData with name "collision"
1920 | 0x........ | HashData[0]
1921 | 0x........ | HashData[1]
1922 | 0x00001203 | String offset into .debug_str ("dump")
1923 | 0x00000003 | A 32 bit array count - number of HashData with name "dump"
1924 | 0x........ | HashData[0]
1925 | 0x........ | HashData[1]
1926 | 0x........ | HashData[2]
1927 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1929 0x00003550: | 0x00001203 | String offset into .debug_str ("main")
1930 | 0x00000009 | A 32 bit array count - number of HashData with name "main"
1931 | 0x........ | HashData[0]
1932 | 0x........ | HashData[1]
1933 | 0x........ | HashData[2]
1934 | 0x........ | HashData[3]
1935 | 0x........ | HashData[4]
1936 | 0x........ | HashData[5]
1937 | 0x........ | HashData[6]
1938 | 0x........ | HashData[7]
1939 | 0x........ | HashData[8]
1940 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1943 So we still have all of the same data, we just organize it more efficiently for
1944 debugger lookup. If we repeat the same "``printf``" lookup from above, we
1945 would hash "``printf``" and find it matches ``BUCKETS[3]`` by taking the 32 bit
1946 hash value and modulo it by ``n_buckets``. ``BUCKETS[3]`` contains "6" which
1947 is the index into the ``HASHES`` table. We would then compare any consecutive
1948 32 bit hashes values in the ``HASHES`` array as long as the hashes would be in
1949 ``BUCKETS[3]``. We do this by verifying that each subsequent hash value modulo
1950 ``n_buckets`` is still 3. In the case of a failed lookup we would access the
1951 memory for ``BUCKETS[3]``, and then compare a few consecutive 32 bit hashes
1952 before we know that we have no match. We don't end up marching through
1953 multiple words of memory and we really keep the number of processor data cache
1954 lines being accessed as small as possible.
1956 The string hash that is used for these lookup tables is the Daniel J.
1957 Bernstein hash which is also used in the ELF ``GNU_HASH`` sections. It is a
1958 very good hash for all kinds of names in programs with very few hash
1961 Empty buckets are designated by using an invalid hash index of ``UINT32_MAX``.
1966 These name hash tables are designed to be generic where specializations of the
1967 table get to define additional data that goes into the header ("``HeaderData``"),
1968 how the string value is stored ("``KeyType``") and the content of the data for each
1974 The header has a fixed part, and the specialized part. The exact format of the
1981 uint32_t magic; // 'HASH' magic value to allow endian detection
1982 uint16_t version; // Version number
1983 uint16_t hash_function; // The hash function enumeration that was used
1984 uint32_t bucket_count; // The number of buckets in this hash table
1985 uint32_t hashes_count; // The total number of unique hash values and hash data offsets in this table
1986 uint32_t header_data_len; // The bytes to skip to get to the hash indexes (buckets) for correct alignment
1987 // Specifically the length of the following HeaderData field - this does not
1988 // include the size of the preceding fields
1989 HeaderData header_data; // Implementation specific header data
1992 The header starts with a 32 bit "``magic``" value which must be ``'HASH'``
1993 encoded as an ASCII integer. This allows the detection of the start of the
1994 hash table and also allows the table's byte order to be determined so the table
1995 can be correctly extracted. The "``magic``" value is followed by a 16 bit
1996 ``version`` number which allows the table to be revised and modified in the
1997 future. The current version number is 1. ``hash_function`` is a ``uint16_t``
1998 enumeration that specifies which hash function was used to produce this table.
1999 The current values for the hash function enumerations include:
2003 enum HashFunctionType
2005 eHashFunctionDJB = 0u, // Daniel J Bernstein hash function
2008 ``bucket_count`` is a 32 bit unsigned integer that represents how many buckets
2009 are in the ``BUCKETS`` array. ``hashes_count`` is the number of unique 32 bit
2010 hash values that are in the ``HASHES`` array, and is the same number of offsets
2011 are contained in the ``OFFSETS`` array. ``header_data_len`` specifies the size
2012 in bytes of the ``HeaderData`` that is filled in by specialized versions of
2018 The header is followed by the buckets, hashes, offsets, and hash value data.
2024 uint32_t buckets[Header.bucket_count]; // An array of hash indexes into the "hashes[]" array below
2025 uint32_t hashes [Header.hashes_count]; // Every unique 32 bit hash for the entire table is in this table
2026 uint32_t offsets[Header.hashes_count]; // An offset that corresponds to each item in the "hashes[]" array above
2029 ``buckets`` is an array of 32 bit indexes into the ``hashes`` array. The
2030 ``hashes`` array contains all of the 32 bit hash values for all names in the
2031 hash table. Each hash in the ``hashes`` table has an offset in the ``offsets``
2032 array that points to the data for the hash value.
2034 This table setup makes it very easy to repurpose these tables to contain
2035 different data, while keeping the lookup mechanism the same for all tables.
2036 This layout also makes it possible to save the table to disk and map it in
2037 later and do very efficient name lookups with little or no parsing.
2039 DWARF lookup tables can be implemented in a variety of ways and can store a lot
2040 of information for each name. We want to make the DWARF tables extensible and
2041 able to store the data efficiently so we have used some of the DWARF features
2042 that enable efficient data storage to define exactly what kind of data we store
2045 The ``HeaderData`` contains a definition of the contents of each HashData chunk.
2046 We might want to store an offset to all of the debug information entries (DIEs)
2047 for each name. To keep things extensible, we create a list of items, or
2048 Atoms, that are contained in the data for each name. First comes the type of
2049 the data in each atom:
2056 eAtomTypeDIEOffset = 1u, // DIE offset, check form for encoding
2057 eAtomTypeCUOffset = 2u, // DIE offset of the compiler unit header that contains the item in question
2058 eAtomTypeTag = 3u, // DW_TAG_xxx value, should be encoded as DW_FORM_data1 (if no tags exceed 255) or DW_FORM_data2
2059 eAtomTypeNameFlags = 4u, // Flags from enum NameFlags
2060 eAtomTypeTypeFlags = 5u, // Flags from enum TypeFlags
2063 The enumeration values and their meanings are:
2065 .. code-block:: none
2067 eAtomTypeNULL - a termination atom that specifies the end of the atom list
2068 eAtomTypeDIEOffset - an offset into the .debug_info section for the DWARF DIE for this name
2069 eAtomTypeCUOffset - an offset into the .debug_info section for the CU that contains the DIE
2070 eAtomTypeDIETag - The DW_TAG_XXX enumeration value so you don't have to parse the DWARF to see what it is
2071 eAtomTypeNameFlags - Flags for functions and global variables (isFunction, isInlined, isExternal...)
2072 eAtomTypeTypeFlags - Flags for types (isCXXClass, isObjCClass, ...)
2074 Then we allow each atom type to define the atom type and how the data for each
2075 atom type data is encoded:
2081 uint16_t type; // AtomType enum value
2082 uint16_t form; // DWARF DW_FORM_XXX defines
2085 The ``form`` type above is from the DWARF specification and defines the exact
2086 encoding of the data for the Atom type. See the DWARF specification for the
2087 ``DW_FORM_`` definitions.
2093 uint32_t die_offset_base;
2094 uint32_t atom_count;
2095 Atoms atoms[atom_count0];
2098 ``HeaderData`` defines the base DIE offset that should be added to any atoms
2099 that are encoded using the ``DW_FORM_ref1``, ``DW_FORM_ref2``,
2100 ``DW_FORM_ref4``, ``DW_FORM_ref8`` or ``DW_FORM_ref_udata``. It also defines
2101 what is contained in each ``HashData`` object -- ``Atom.form`` tells us how large
2102 each field will be in the ``HashData`` and the ``Atom.type`` tells us how this data
2103 should be interpreted.
2105 For the current implementations of the "``.apple_names``" (all functions +
2106 globals), the "``.apple_types``" (names of all types that are defined), and
2107 the "``.apple_namespaces``" (all namespaces), we currently set the ``Atom``
2112 HeaderData.atom_count = 1;
2113 HeaderData.atoms[0].type = eAtomTypeDIEOffset;
2114 HeaderData.atoms[0].form = DW_FORM_data4;
2116 This defines the contents to be the DIE offset (eAtomTypeDIEOffset) that is
2117 encoded as a 32 bit value (DW_FORM_data4). This allows a single name to have
2118 multiple matching DIEs in a single file, which could come up with an inlined
2119 function for instance. Future tables could include more information about the
2120 DIE such as flags indicating if the DIE is a function, method, block,
2123 The KeyType for the DWARF table is a 32 bit string table offset into the
2124 ".debug_str" table. The ".debug_str" is the string table for the DWARF which
2125 may already contain copies of all of the strings. This helps make sure, with
2126 help from the compiler, that we reuse the strings between all of the DWARF
2127 sections and keeps the hash table size down. Another benefit to having the
2128 compiler generate all strings as DW_FORM_strp in the debug info, is that
2129 DWARF parsing can be made much faster.
2131 After a lookup is made, we get an offset into the hash data. The hash data
2132 needs to be able to deal with 32 bit hash collisions, so the chunk of data
2133 at the offset in the hash data consists of a triple:
2138 uint32_t hash_data_count
2139 HashData[hash_data_count]
2141 If "str_offset" is zero, then the bucket contents are done. 99.9% of the
2142 hash data chunks contain a single item (no 32 bit hash collision):
2144 .. code-block:: none
2147 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2148 | 0x00000004 | uint32_t HashData count
2149 | 0x........ | uint32_t HashData[0] DIE offset
2150 | 0x........ | uint32_t HashData[1] DIE offset
2151 | 0x........ | uint32_t HashData[2] DIE offset
2152 | 0x........ | uint32_t HashData[3] DIE offset
2153 | 0x00000000 | uint32_t KeyType (end of hash chain)
2156 If there are collisions, you will have multiple valid string offsets:
2158 .. code-block:: none
2161 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2162 | 0x00000004 | uint32_t HashData count
2163 | 0x........ | uint32_t HashData[0] DIE offset
2164 | 0x........ | uint32_t HashData[1] DIE offset
2165 | 0x........ | uint32_t HashData[2] DIE offset
2166 | 0x........ | uint32_t HashData[3] DIE offset
2167 | 0x00002023 | uint32_t KeyType (.debug_str[0x0002023] => "print")
2168 | 0x00000002 | uint32_t HashData count
2169 | 0x........ | uint32_t HashData[0] DIE offset
2170 | 0x........ | uint32_t HashData[1] DIE offset
2171 | 0x00000000 | uint32_t KeyType (end of hash chain)
2174 Current testing with real world C++ binaries has shown that there is around 1
2175 32 bit hash collision per 100,000 name entries.
2180 As we said, we want to strictly define exactly what is included in the
2181 different tables. For DWARF, we have 3 tables: "``.apple_names``",
2182 "``.apple_types``", and "``.apple_namespaces``".
2184 "``.apple_names``" sections should contain an entry for each DWARF DIE whose
2185 ``DW_TAG`` is a ``DW_TAG_label``, ``DW_TAG_inlined_subroutine``, or
2186 ``DW_TAG_subprogram`` that has address attributes: ``DW_AT_low_pc``,
2187 ``DW_AT_high_pc``, ``DW_AT_ranges`` or ``DW_AT_entry_pc``. It also contains
2188 ``DW_TAG_variable`` DIEs that have a ``DW_OP_addr`` in the location (global and
2189 static variables). All global and static variables should be included,
2190 including those scoped within functions and classes. For example using the
2202 Both of the static ``var`` variables would be included in the table. All
2203 functions should emit both their full names and their basenames. For C or C++,
2204 the full name is the mangled name (if available) which is usually in the
2205 ``DW_AT_MIPS_linkage_name`` attribute, and the ``DW_AT_name`` contains the
2206 function basename. If global or static variables have a mangled name in a
2207 ``DW_AT_MIPS_linkage_name`` attribute, this should be emitted along with the
2208 simple name found in the ``DW_AT_name`` attribute.
2210 "``.apple_types``" sections should contain an entry for each DWARF DIE whose
2215 * DW_TAG_enumeration_type
2216 * DW_TAG_pointer_type
2217 * DW_TAG_reference_type
2218 * DW_TAG_string_type
2219 * DW_TAG_structure_type
2220 * DW_TAG_subroutine_type
2223 * DW_TAG_ptr_to_member_type
2225 * DW_TAG_subrange_type
2231 * DW_TAG_packed_type
2232 * DW_TAG_volatile_type
2233 * DW_TAG_restrict_type
2234 * DW_TAG_interface_type
2235 * DW_TAG_unspecified_type
2236 * DW_TAG_shared_type
2238 Only entries with a ``DW_AT_name`` attribute are included, and the entry must
2239 not be a forward declaration (``DW_AT_declaration`` attribute with a non-zero
2240 value). For example, using the following code:
2250 We get a few type DIEs:
2252 .. code-block:: none
2254 0x00000067: TAG_base_type [5]
2255 AT_encoding( DW_ATE_signed )
2257 AT_byte_size( 0x04 )
2259 0x0000006e: TAG_pointer_type [6]
2260 AT_type( {0x00000067} ( int ) )
2261 AT_byte_size( 0x08 )
2263 The DW_TAG_pointer_type is not included because it does not have a ``DW_AT_name``.
2265 "``.apple_namespaces``" section should contain all ``DW_TAG_namespace`` DIEs.
2266 If we run into a namespace that has no name this is an anonymous namespace, and
2267 the name should be output as "``(anonymous namespace)``" (without the quotes).
2268 Why? This matches the output of the ``abi::cxa_demangle()`` that is in the
2269 standard C++ library that demangles mangled names.
2272 Language Extensions and File Format Changes
2273 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2275 Objective-C Extensions
2276 """"""""""""""""""""""
2278 "``.apple_objc``" section should contain all ``DW_TAG_subprogram`` DIEs for an
2279 Objective-C class. The name used in the hash table is the name of the
2280 Objective-C class itself. If the Objective-C class has a category, then an
2281 entry is made for both the class name without the category, and for the class
2282 name with the category. So if we have a DIE at offset 0x1234 with a name of
2283 method "``-[NSString(my_additions) stringWithSpecialString:]``", we would add
2284 an entry for "``NSString``" that points to DIE 0x1234, and an entry for
2285 "``NSString(my_additions)``" that points to 0x1234. This allows us to quickly
2286 track down all Objective-C methods for an Objective-C class when doing
2287 expressions. It is needed because of the dynamic nature of Objective-C where
2288 anyone can add methods to a class. The DWARF for Objective-C methods is also
2289 emitted differently from C++ classes where the methods are not usually
2290 contained in the class definition, they are scattered about across one or more
2291 compile units. Categories can also be defined in different shared libraries.
2292 So we need to be able to quickly find all of the methods and class functions
2293 given the Objective-C class name, or quickly find all methods and class
2294 functions for a class + category name. This table does not contain any
2295 selector names, it just maps Objective-C class names (or class names +
2296 category) to all of the methods and class functions. The selectors are added
2297 as function basenames in the "``.debug_names``" section.
2299 In the "``.apple_names``" section for Objective-C functions, the full name is
2300 the entire function name with the brackets ("``-[NSString
2301 stringWithCString:]``") and the basename is the selector only
2302 ("``stringWithCString:``").
2307 The sections names for the apple hash tables are for non mach-o files. For
2308 mach-o files, the sections should be contained in the ``__DWARF`` segment with
2311 * "``.apple_names``" -> "``__apple_names``"
2312 * "``.apple_types``" -> "``__apple_types``"
2313 * "``.apple_namespaces``" -> "``__apple_namespac``" (16 character limit)
2314 * "``.apple_objc``" -> "``__apple_objc``"