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 - Artificial, 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, ;; DWARF path discriminator value
323 i32 ;; Unique ID to identify blocks from a template function
326 This descriptor provides debug information about nested blocks within a
327 subprogram. The line number and column numbers are used to dinstinguish two
328 lexical blocks at same depth.
333 i32, ;; Tag = 11 (DW_TAG_lexical_block)
334 metadata,;; Source directory (including trailing slash) & file pair
335 metadata ;; Reference to the scope we're annotating with a file change
338 This descriptor provides a wrapper around a lexical scope to handle file
339 changes in the middle of a lexical block.
341 .. _format_basic_type:
343 Basic type descriptors
344 ^^^^^^^^^^^^^^^^^^^^^^
349 i32, ;; Tag = 36 (DW_TAG_base_type)
350 metadata, ;; Source directory (including trailing slash) & file pair (may be null)
351 metadata, ;; Reference to context
352 metadata, ;; Name (may be "" for anonymous types)
353 i32, ;; Line number where defined (may be 0)
355 i64, ;; Alignment in bits
356 i64, ;; Offset in bits
358 i32 ;; DWARF type encoding
361 These descriptors define primitive types used in the code. Example ``int``,
362 ``bool`` and ``float``. The context provides the scope of the type, which is
363 usually the top level. Since basic types are not usually user defined the
364 context and line number can be left as NULL and 0. The size, alignment and
365 offset are expressed in bits and can be 64 bit values. The alignment is used
366 to round the offset when embedded in a :ref:`composite type
367 <format_composite_type>` (example to keep float doubles on 64 bit boundaries).
368 The offset is the bit offset if embedded in a :ref:`composite type
369 <format_composite_type>`.
371 The type encoding provides the details of the type. The values are typically
372 one of the following:
380 DW_ATE_signed_char = 6
382 DW_ATE_unsigned_char = 8
384 .. _format_derived_type:
386 Derived type descriptors
387 ^^^^^^^^^^^^^^^^^^^^^^^^
392 i32, ;; Tag (see below)
393 metadata, ;; Source directory (including trailing slash) & file pair (may be null)
394 metadata, ;; Reference to context
395 metadata, ;; Name (may be "" for anonymous types)
396 i32, ;; Line number where defined (may be 0)
398 i64, ;; Alignment in bits
399 i64, ;; Offset in bits
400 i32, ;; Flags to encode attributes, e.g. private
401 metadata, ;; Reference to type derived from
402 metadata, ;; (optional) Name of the Objective C property associated with
403 ;; Objective-C an ivar, or the type of which this
404 ;; pointer-to-member is pointing to members of.
405 metadata, ;; (optional) Name of the Objective C property getter selector.
406 metadata, ;; (optional) Name of the Objective C property setter selector.
407 i32 ;; (optional) Objective C property attributes.
410 These descriptors are used to define types derived from other types. The value
411 of the tag varies depending on the meaning. The following are possible tag
416 DW_TAG_formal_parameter = 5
418 DW_TAG_pointer_type = 15
419 DW_TAG_reference_type = 16
421 DW_TAG_ptr_to_member_type = 31
422 DW_TAG_const_type = 38
423 DW_TAG_volatile_type = 53
424 DW_TAG_restrict_type = 55
426 ``DW_TAG_member`` is used to define a member of a :ref:`composite type
427 <format_composite_type>` or :ref:`subprogram <format_subprograms>`. The type
428 of the member is the :ref:`derived type <format_derived_type>`.
429 ``DW_TAG_formal_parameter`` is used to define a member which is a formal
430 argument of a subprogram.
432 ``DW_TAG_typedef`` is used to provide a name for the derived type.
434 ``DW_TAG_pointer_type``, ``DW_TAG_reference_type``, ``DW_TAG_const_type``,
435 ``DW_TAG_volatile_type`` and ``DW_TAG_restrict_type`` are used to qualify the
436 :ref:`derived type <format_derived_type>`.
438 :ref:`Derived type <format_derived_type>` location can be determined from the
439 context and line number. The size, alignment and offset are expressed in bits
440 and can be 64 bit values. The alignment is used to round the offset when
441 embedded in a :ref:`composite type <format_composite_type>` (example to keep
442 float doubles on 64 bit boundaries.) The offset is the bit offset if embedded
443 in a :ref:`composite type <format_composite_type>`.
445 Note that the ``void *`` type is expressed as a type derived from NULL.
447 .. _format_composite_type:
449 Composite type descriptors
450 ^^^^^^^^^^^^^^^^^^^^^^^^^^
455 i32, ;; Tag (see below)
456 metadata, ;; Source directory (including trailing slash) & file pair (may be null)
457 metadata, ;; Reference to context
458 metadata, ;; Name (may be "" for anonymous types)
459 i32, ;; Line number where defined (may be 0)
461 i64, ;; Alignment in bits
462 i64, ;; Offset in bits
464 metadata, ;; Reference to type derived from
465 metadata, ;; Reference to array of member descriptors
466 i32, ;; Runtime languages
467 metadata, ;; Base type containing the vtable pointer for this type
468 metadata, ;; Template parameters
469 metadata ;; A unique identifier for type uniquing purpose (may be null)
472 These descriptors are used to define types that are composed of 0 or more
473 elements. The value of the tag varies depending on the meaning. The following
474 are possible tag values:
478 DW_TAG_array_type = 1
479 DW_TAG_enumeration_type = 4
480 DW_TAG_structure_type = 19
481 DW_TAG_union_type = 23
482 DW_TAG_subroutine_type = 21
483 DW_TAG_inheritance = 28
485 The vector flag indicates that an array type is a native packed vector.
487 The members of array types (tag = ``DW_TAG_array_type``) are
488 :ref:`subrange descriptors <format_subrange>`, each
489 representing the range of subscripts at that level of indexing.
491 The members of enumeration types (tag = ``DW_TAG_enumeration_type``) are
492 :ref:`enumerator descriptors <format_enumerator>`, each representing the
493 definition of enumeration value for the set. All enumeration type descriptors
494 are collected inside the named metadata ``!llvm.dbg.cu``.
496 The members of structure (tag = ``DW_TAG_structure_type``) or union (tag =
497 ``DW_TAG_union_type``) types are any one of the :ref:`basic
498 <format_basic_type>`, :ref:`derived <format_derived_type>` or :ref:`composite
499 <format_composite_type>` type descriptors, each representing a field member of
500 the structure or union.
502 For C++ classes (tag = ``DW_TAG_structure_type``), member descriptors provide
503 information about base classes, static members and member functions. If a
504 member is a :ref:`derived type descriptor <format_derived_type>` and has a tag
505 of ``DW_TAG_inheritance``, then the type represents a base class. If the member
506 of is a :ref:`global variable descriptor <format_global_variables>` then it
507 represents a static member. And, if the member is a :ref:`subprogram
508 descriptor <format_subprograms>` then it represents a member function. For
509 static members and member functions, ``getName()`` returns the members link or
510 the C++ mangled name. ``getDisplayName()`` the simplied version of the name.
512 The first member of subroutine (tag = ``DW_TAG_subroutine_type``) type elements
513 is the return type for the subroutine. The remaining elements are the formal
514 arguments to the subroutine.
516 :ref:`Composite type <format_composite_type>` location can be determined from
517 the context and line number. The size, alignment and offset are expressed in
518 bits and can be 64 bit values. The alignment is used to round the offset when
519 embedded in a :ref:`composite type <format_composite_type>` (as an example, to
520 keep float doubles on 64 bit boundaries). The offset is the bit offset if
521 embedded in a :ref:`composite type <format_composite_type>`.
531 i32, ;; Tag = 33 (DW_TAG_subrange_type)
536 These descriptors are used to define ranges of array subscripts for an array
537 :ref:`composite type <format_composite_type>`. The low value defines the lower
538 bounds typically zero for C/C++. The high value is the upper bounds. Values
539 are 64 bit. ``High - Low + 1`` is the size of the array. If ``Low > High``
540 the array bounds are not included in generated debugging information.
542 .. _format_enumerator:
544 Enumerator descriptors
545 ^^^^^^^^^^^^^^^^^^^^^^
550 i32, ;; Tag = 40 (DW_TAG_enumerator)
555 These descriptors are used to define members of an enumeration :ref:`composite
556 type <format_composite_type>`, it associates the name to the value.
564 i32, ;; Tag (see below)
567 metadata, ;; Reference to file where defined
568 i32, ;; 24 bit - Line number where defined
569 ;; 8 bit - Argument number. 1 indicates 1st argument.
570 metadata, ;; Reference to the type descriptor
572 metadata ;; (optional) Reference to inline location
575 These descriptors are used to define variables local to a sub program. The
576 value of the tag depends on the usage of the variable:
580 DW_TAG_auto_variable = 256
581 DW_TAG_arg_variable = 257
583 An auto variable is any variable declared in the body of the function. An
584 argument variable is any variable that appears as a formal argument to the
587 The context is either the subprogram or block where the variable is defined.
588 Name the source variable name. Context and line indicate where the variable
589 was defined. Type descriptor defines the declared type of the variable.
591 .. _format_common_intrinsics:
593 Debugger intrinsic functions
594 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
596 LLVM uses several intrinsic functions (name prefixed with "``llvm.dbg``") to
597 provide debug information at various points in generated code.
604 void %llvm.dbg.declare(metadata, metadata)
606 This intrinsic provides information about a local element (e.g., variable).
607 The first argument is metadata holding the alloca for the variable. The second
608 argument is metadata containing a description of the variable.
615 void %llvm.dbg.value(metadata, i64, metadata)
617 This intrinsic provides information when a user source variable is set to a new
618 value. The first argument is the new value (wrapped as metadata). The second
619 argument is the offset in the user source variable where the new value is
620 written. The third argument is metadata containing a description of the user
623 Object lifetimes and scoping
624 ============================
626 In many languages, the local variables in functions can have their lifetimes or
627 scopes limited to a subset of a function. In the C family of languages, for
628 example, variables are only live (readable and writable) within the source
629 block that they are defined in. In functional languages, values are only
630 readable after they have been defined. Though this is a very obvious concept,
631 it is non-trivial to model in LLVM, because it has no notion of scoping in this
632 sense, and does not want to be tied to a language's scoping rules.
634 In order to handle this, the LLVM debug format uses the metadata attached to
635 llvm instructions to encode line number and scoping information. Consider the
636 following C fragment, for example:
650 Compiled to LLVM, this function would be represented like this:
654 define void @foo() #0 {
656 %X = alloca i32, align 4
657 %Y = alloca i32, align 4
658 %Z = alloca i32, align 4
659 call void @llvm.dbg.declare(metadata !{i32* %X}, metadata !10), !dbg !12
660 ; [debug line = 2:7] [debug variable = X]
661 store i32 21, i32* %X, align 4, !dbg !12
662 call void @llvm.dbg.declare(metadata !{i32* %Y}, metadata !13), !dbg !14
663 ; [debug line = 3:7] [debug variable = Y]
664 store i32 22, i32* %Y, align 4, !dbg !14
665 call void @llvm.dbg.declare(metadata !{i32* %Z}, metadata !15), !dbg !17
666 ; [debug line = 5:9] [debug variable = Z]
667 store i32 23, i32* %Z, align 4, !dbg !17
668 %0 = load i32* %X, align 4, !dbg !18
670 store i32 %0, i32* %Z, align 4, !dbg !18
671 %1 = load i32* %Y, align 4, !dbg !19
673 store i32 %1, i32* %X, align 4, !dbg !19
677 ; Function Attrs: nounwind readnone
678 declare void @llvm.dbg.declare(metadata, metadata) #1
680 attributes #0 = { nounwind ssp uwtable "less-precise-fpmad"="false"
681 "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"
682 "no-infs-fp-math"="false" "no-nans-fp-math"="false"
683 "stack-protector-buffer-size"="8" "unsafe-fp-math"="false"
684 "use-soft-float"="false" }
685 attributes #1 = { nounwind readnone }
688 !llvm.module.flags = !{!8}
691 !0 = metadata !{i32 786449, metadata !1, i32 12,
692 metadata !"clang version 3.4 (trunk 193128) (llvm/trunk 193139)",
693 i1 false, metadata !"", i32 0, metadata !2, metadata !2, metadata !3,
694 metadata !2, metadata !2, metadata !""} ; [ DW_TAG_compile_unit ] \
695 [/private/tmp/foo.c] \
697 !1 = metadata !{metadata !"t.c", metadata !"/private/tmp"}
698 !2 = metadata !{i32 0}
699 !3 = metadata !{metadata !4}
700 !4 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"foo",
701 metadata !"foo", metadata !"", i32 1, metadata !6,
702 i1 false, i1 true, i32 0, i32 0, null, i32 0, i1 false,
703 void ()* @foo, null, null, metadata !2, i32 1}
704 ; [ DW_TAG_subprogram ] [line 1] [def] [foo]
705 !5 = metadata !{i32 786473, metadata !1} ; [ DW_TAG_file_type ] \
707 !6 = metadata !{i32 786453, i32 0, null, metadata !"", i32 0, i64 0, i64 0,
708 i64 0, i32 0, null, metadata !7, i32 0, null, null, null}
709 ; [ DW_TAG_subroutine_type ] \
710 [line 0, size 0, align 0, offset 0] [from ]
711 !7 = metadata !{null}
712 !8 = metadata !{i32 2, metadata !"Dwarf Version", i32 2}
713 !9 = metadata !{metadata !"clang version 3.4 (trunk 193128) (llvm/trunk 193139)"}
714 !10 = metadata !{i32 786688, metadata !4, metadata !"X", metadata !5, i32 2,
715 metadata !11, i32 0, i32 0} ; [ DW_TAG_auto_variable ] [X] \
717 !11 = metadata !{i32 786468, null, null, metadata !"int", i32 0, i64 32,
718 i64 32, i64 0, i32 0, i32 5} ; [ DW_TAG_base_type ] [int] \
719 [line 0, size 32, align 32, offset 0, enc DW_ATE_signed]
720 !12 = metadata !{i32 2, i32 0, metadata !4, null}
721 !13 = metadata !{i32 786688, metadata !4, metadata !"Y", metadata !5, i32 3,
722 metadata !11, i32 0, i32 0} ; [ DW_TAG_auto_variable ] [Y] \
724 !14 = metadata !{i32 3, i32 0, metadata !4, null}
725 !15 = metadata !{i32 786688, metadata !16, metadata !"Z", metadata !5, i32 5,
726 metadata !11, i32 0, i32 0} ; [ DW_TAG_auto_variable ] [Z] \
728 !16 = metadata !{i32 786443, metadata !1, metadata !4, i32 4, i32 0, i32 0,
730 ; [ DW_TAG_lexical_block ] [/private/tmp/t.c]
731 !17 = metadata !{i32 5, i32 0, metadata !16, null}
732 !18 = metadata !{i32 6, i32 0, metadata !16, null}
733 !19 = metadata !{i32 8, i32 0, metadata !4, null} ; [ DW_TAG_imported_declaration ]
734 !20 = metadata !{i32 9, i32 0, metadata !4, null}
736 This example illustrates a few important details about LLVM debugging
737 information. In particular, it shows how the ``llvm.dbg.declare`` intrinsic and
738 location information, which are attached to an instruction, are applied
739 together to allow a debugger to analyze the relationship between statements,
740 variable definitions, and the code used to implement the function.
744 call void @llvm.dbg.declare(metadata !{i32* %X}, metadata !10), !dbg !12
745 ; [debug line = 2:7] [debug variable = X]
747 The first intrinsic ``%llvm.dbg.declare`` encodes debugging information for the
748 variable ``X``. The metadata ``!dbg !12`` attached to the intrinsic provides
749 scope information for the variable ``X``.
753 !12 = metadata !{i32 2, i32 0, metadata !4, null}
754 !4 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"foo",
755 metadata !"foo", metadata !"", i32 1, metadata !6,
756 i1 false, i1 true, i32 0, i32 0, null, i32 0, i1 false,
757 void ()* @foo, null, null, metadata !2, i32 1}
758 ; [ DW_TAG_subprogram ] [line 1] [def] [foo]
760 Here ``!12`` is metadata providing location information. It has four fields:
761 line number, column number, scope, and original scope. The original scope
762 represents inline location if this instruction is inlined inside a caller, and
763 is null otherwise. In this example, scope is encoded by ``!4``, a
764 :ref:`subprogram descriptor <format_subprograms>`. This way the location
765 information attached to the intrinsics indicates that the variable ``X`` is
766 declared at line number 2 at a function level scope in function ``foo``.
768 Now lets take another example.
772 call void @llvm.dbg.declare(metadata !{i32* %Z}, metadata !15), !dbg !17
773 ; [debug line = 5:9] [debug variable = Z]
775 The third intrinsic ``%llvm.dbg.declare`` encodes debugging information for
776 variable ``Z``. The metadata ``!dbg !17`` attached to the intrinsic provides
777 scope information for the variable ``Z``.
781 !16 = metadata !{i32 786443, metadata !1, metadata !4, i32 4, i32 0, i32 0,
783 ; [ DW_TAG_lexical_block ] [/private/tmp/t.c]
784 !17 = metadata !{i32 5, i32 0, metadata !16, null}
786 Here ``!15`` indicates that ``Z`` is declared at line number 5 and
787 column number 0 inside of lexical scope ``!16``. The lexical scope itself
788 resides inside of subprogram ``!4`` described above.
790 The scope information attached with each instruction provides a straightforward
791 way to find instructions covered by a scope.
795 C/C++ front-end specific debug information
796 ==========================================
798 The C and C++ front-ends represent information about the program in a format
799 that is effectively identical to `DWARF 3.0
800 <http://www.eagercon.com/dwarf/dwarf3std.htm>`_ in terms of information
801 content. This allows code generators to trivially support native debuggers by
802 generating standard dwarf information, and contains enough information for
803 non-dwarf targets to translate it as needed.
805 This section describes the forms used to represent C and C++ programs. Other
806 languages could pattern themselves after this (which itself is tuned to
807 representing programs in the same way that DWARF 3 does), or they could choose
808 to provide completely different forms if they don't fit into the DWARF model.
809 As support for debugging information gets added to the various LLVM
810 source-language front-ends, the information used should be documented here.
812 The following sections provide examples of various C/C++ constructs and the
813 debug information that would best describe those constructs.
815 C/C++ source file information
816 -----------------------------
818 Given the source files ``MySource.cpp`` and ``MyHeader.h`` located in the
819 directory ``/Users/mine/sources``, the following code:
823 #include "MyHeader.h"
825 int main(int argc, char *argv[]) {
829 a C/C++ front-end would generate the following descriptors:
835 ;; Define the compile unit for the main source file "/Users/mine/sources/MySource.cpp".
839 metadata !1, ;; File/directory name
840 i32 4, ;; Language Id
841 metadata !"clang version 3.4 ",
842 i1 false, ;; Optimized compile unit
843 metadata !"", ;; Compiler flags
844 i32 0, ;; Runtime version
845 metadata !2, ;; Enumeration types
846 metadata !2, ;; Retained types
847 metadata !3, ;; Subprograms
848 metadata !2, ;; Global variables
849 metadata !2, ;; Imported entities (declarations and namespaces)
850 metadata !"" ;; Split debug filename
854 ;; Define the file for the file "/Users/mine/sources/MySource.cpp".
857 metadata !"MySource.cpp",
858 metadata !"/Users/mine/sources"
866 ;; Define the file for the file "/Users/mine/sources/Myheader.h"
873 metadata !"./MyHeader.h",
874 metadata !"/Users/mine/sources",
879 ``llvm::Instruction`` provides easy access to metadata attached with an
880 instruction. One can extract line number information encoded in LLVM IR using
881 ``Instruction::getMetadata()`` and ``DILocation::getLineNumber()``.
885 if (MDNode *N = I->getMetadata("dbg")) { // Here I is an LLVM instruction
886 DILocation Loc(N); // DILocation is in DebugInfo.h
887 unsigned Line = Loc.getLineNumber();
888 StringRef File = Loc.getFilename();
889 StringRef Dir = Loc.getDirectory();
892 C/C++ global variable information
893 ---------------------------------
895 Given an integer global variable declared as follows:
901 a C/C++ front-end would generate the following descriptors:
906 ;; Define the global itself.
908 %MyGlobal = global int 100
911 ;; List of debug info of globals
915 ;; Define the compile unit.
920 metadata !"foo.cpp", ;; File
921 metadata !"/Volumes/Data/tmp", ;; Directory
922 metadata !"clang version 3.1 ", ;; Producer
923 i1 true, ;; Deprecated field
924 i1 false, ;; "isOptimized"?
925 metadata !"", ;; Flags
926 i32 0, ;; Runtime Version
927 metadata !1, ;; Enum Types
928 metadata !1, ;; Retained Types
929 metadata !1, ;; Subprograms
930 metadata !3, ;; Global Variables
931 metadata !1, ;; Imported entities
932 "", ;; Split debug filename
933 } ; [ DW_TAG_compile_unit ]
935 ;; The Array of Global Variables
941 ;; Define the global variable itself.
947 metadata !"MyGlobal", ;; Name
948 metadata !"MyGlobal", ;; Display Name
949 metadata !"", ;; Linkage Name
953 i32 0, ;; IsLocalToUnit
954 i32 1, ;; IsDefinition
955 i32* @MyGlobal, ;; LLVM-IR Value
956 null ;; Static member declaration
957 } ; [ DW_TAG_variable ]
963 metadata !"foo.cpp", ;; File
964 metadata !"/Volumes/Data/tmp", ;; Directory
968 metadata !5 ;; Unused
969 } ; [ DW_TAG_file_type ]
978 metadata !"int", ;; Name
980 i64 32, ;; Size in Bits
981 i64 32, ;; Align in Bits
985 } ; [ DW_TAG_base_type ]
987 C/C++ function information
988 --------------------------
990 Given a function declared as follows:
994 int main(int argc, char *argv[]) {
998 a C/C++ front-end would generate the following descriptors:
1000 .. code-block:: llvm
1003 ;; Define the anchor for subprograms.
1007 metadata !1, ;; File
1008 metadata !1, ;; Context
1009 metadata !"main", ;; Name
1010 metadata !"main", ;; Display name
1011 metadata !"main", ;; Linkage name
1012 i32 1, ;; Line number
1013 metadata !4, ;; Type
1014 i1 false, ;; Is local
1015 i1 true, ;; Is definition
1016 i32 0, ;; Virtuality attribute, e.g. pure virtual function
1017 i32 0, ;; Index into virtual table for C++ methods
1018 i32 0, ;; Type that holds virtual table.
1020 i1 false, ;; True if this function is optimized
1021 Function *, ;; Pointer to llvm::Function
1022 null, ;; Function template parameters
1023 null, ;; List of function variables (emitted when optimizing)
1024 1 ;; Line number of the opening '{' of the function
1027 ;; Define the subprogram itself.
1029 define i32 @main(i32 %argc, i8** %argv) {
1036 The following are the basic type descriptors for C/C++ core types:
1041 .. code-block:: llvm
1047 metadata !"bool", ;; Name
1048 i32 0, ;; Line number
1049 i64 8, ;; Size in Bits
1050 i64 8, ;; Align in Bits
1051 i64 0, ;; Offset in Bits
1059 .. code-block:: llvm
1065 metadata !"char", ;; Name
1066 i32 0, ;; Line number
1067 i64 8, ;; Size in Bits
1068 i64 8, ;; Align in Bits
1069 i64 0, ;; Offset in Bits
1077 .. code-block:: llvm
1083 metadata !"unsigned char",
1084 i32 0, ;; Line number
1085 i64 8, ;; Size in Bits
1086 i64 8, ;; Align in Bits
1087 i64 0, ;; Offset in Bits
1095 .. code-block:: llvm
1101 metadata !"short int",
1102 i32 0, ;; Line number
1103 i64 16, ;; Size in Bits
1104 i64 16, ;; Align in Bits
1105 i64 0, ;; Offset in Bits
1113 .. code-block:: llvm
1119 metadata !"short unsigned int",
1120 i32 0, ;; Line number
1121 i64 16, ;; Size in Bits
1122 i64 16, ;; Align in Bits
1123 i64 0, ;; Offset in Bits
1131 .. code-block:: llvm
1137 metadata !"int", ;; Name
1138 i32 0, ;; Line number
1139 i64 32, ;; Size in Bits
1140 i64 32, ;; Align in Bits
1141 i64 0, ;; Offset in Bits
1149 .. code-block:: llvm
1155 metadata !"unsigned int",
1156 i32 0, ;; Line number
1157 i64 32, ;; Size in Bits
1158 i64 32, ;; Align in Bits
1159 i64 0, ;; Offset in Bits
1167 .. code-block:: llvm
1173 metadata !"long long int",
1174 i32 0, ;; Line number
1175 i64 64, ;; Size in Bits
1176 i64 64, ;; Align in Bits
1177 i64 0, ;; Offset in Bits
1185 .. code-block:: llvm
1191 metadata !"long long unsigned int",
1192 i32 0, ;; Line number
1193 i64 64, ;; Size in Bits
1194 i64 64, ;; Align in Bits
1195 i64 0, ;; Offset in Bits
1203 .. code-block:: llvm
1210 i32 0, ;; Line number
1211 i64 32, ;; Size in Bits
1212 i64 32, ;; Align in Bits
1213 i64 0, ;; Offset in Bits
1221 .. code-block:: llvm
1227 metadata !"double",;; Name
1228 i32 0, ;; Line number
1229 i64 64, ;; Size in Bits
1230 i64 64, ;; Align in Bits
1231 i64 0, ;; Offset in Bits
1239 Given the following as an example of C/C++ derived type:
1243 typedef const int *IntPtr;
1245 a C/C++ front-end would generate the following descriptors:
1247 .. code-block:: llvm
1250 ;; Define the typedef "IntPtr".
1254 metadata !3, ;; File
1255 metadata !1, ;; Context
1256 metadata !"IntPtr", ;; Name
1257 i32 0, ;; Line number
1258 i64 0, ;; Size in bits
1259 i64 0, ;; Align in bits
1260 i64 0, ;; Offset in bits
1262 metadata !4 ;; Derived From type
1265 ;; Define the pointer type.
1271 metadata !"", ;; Name
1272 i32 0, ;; Line number
1273 i64 64, ;; Size in bits
1274 i64 64, ;; Align in bits
1275 i64 0, ;; Offset in bits
1277 metadata !5 ;; Derived From type
1280 ;; Define the const type.
1286 metadata !"", ;; Name
1287 i32 0, ;; Line number
1288 i64 0, ;; Size in bits
1289 i64 0, ;; Align in bits
1290 i64 0, ;; Offset in bits
1292 metadata !6 ;; Derived From type
1295 ;; Define the int type.
1301 metadata !"int", ;; Name
1302 i32 0, ;; Line number
1303 i64 32, ;; Size in bits
1304 i64 32, ;; Align in bits
1305 i64 0, ;; Offset in bits
1310 C/C++ struct/union types
1311 ------------------------
1313 Given the following as an example of C/C++ struct type:
1323 a C/C++ front-end would generate the following descriptors:
1325 .. code-block:: llvm
1328 ;; Define basic type for unsigned int.
1334 metadata !"unsigned int",
1335 i32 0, ;; Line number
1336 i64 32, ;; Size in Bits
1337 i64 32, ;; Align in Bits
1338 i64 0, ;; Offset in Bits
1343 ;; Define composite type for struct Color.
1347 metadata !1, ;; Compile unit
1349 metadata !"Color", ;; Name
1350 i32 1, ;; Line number
1351 i64 96, ;; Size in bits
1352 i64 32, ;; Align in bits
1353 i64 0, ;; Offset in bits
1355 null, ;; Derived From
1356 metadata !3, ;; Elements
1357 i32 0, ;; Runtime Language
1358 null, ;; Base type containing the vtable pointer for this type
1359 null ;; Template parameters
1363 ;; Define the Red field.
1367 metadata !1, ;; File
1368 metadata !1, ;; Context
1369 metadata !"Red", ;; Name
1370 i32 2, ;; Line number
1371 i64 32, ;; Size in bits
1372 i64 32, ;; Align in bits
1373 i64 0, ;; Offset in bits
1375 metadata !5 ;; Derived From type
1379 ;; Define the Green field.
1383 metadata !1, ;; File
1384 metadata !1, ;; Context
1385 metadata !"Green", ;; Name
1386 i32 3, ;; Line number
1387 i64 32, ;; Size in bits
1388 i64 32, ;; Align in bits
1389 i64 32, ;; Offset in bits
1391 metadata !5 ;; Derived From type
1395 ;; Define the Blue field.
1399 metadata !1, ;; File
1400 metadata !1, ;; Context
1401 metadata !"Blue", ;; Name
1402 i32 4, ;; Line number
1403 i64 32, ;; Size in bits
1404 i64 32, ;; Align in bits
1405 i64 64, ;; Offset in bits
1407 metadata !5 ;; Derived From type
1411 ;; Define the array of fields used by the composite type Color.
1413 !3 = metadata !{metadata !4, metadata !6, metadata !7}
1415 C/C++ enumeration types
1416 -----------------------
1418 Given the following as an example of C/C++ enumeration type:
1428 a C/C++ front-end would generate the following descriptors:
1430 .. code-block:: llvm
1433 ;; Define composite type for enum Trees
1437 metadata !1, ;; File
1438 metadata !1, ;; Context
1439 metadata !"Trees", ;; Name
1440 i32 1, ;; Line number
1441 i64 32, ;; Size in bits
1442 i64 32, ;; Align in bits
1443 i64 0, ;; Offset in bits
1445 null, ;; Derived From type
1446 metadata !3, ;; Elements
1447 i32 0 ;; Runtime language
1451 ;; Define the array of enumerators used by composite type Trees.
1453 !3 = metadata !{metadata !4, metadata !5, metadata !6}
1456 ;; Define Spruce enumerator.
1458 !4 = metadata !{i32 786472, metadata !"Spruce", i64 100}
1461 ;; Define Oak enumerator.
1463 !5 = metadata !{i32 786472, metadata !"Oak", i64 200}
1466 ;; Define Maple enumerator.
1468 !6 = metadata !{i32 786472, metadata !"Maple", i64 300}
1470 Debugging information format
1471 ============================
1473 Debugging Information Extension for Objective C Properties
1474 ----------------------------------------------------------
1479 Objective C provides a simpler way to declare and define accessor methods using
1480 declared properties. The language provides features to declare a property and
1481 to let compiler synthesize accessor methods.
1483 The debugger lets developer inspect Objective C interfaces and their instance
1484 variables and class variables. However, the debugger does not know anything
1485 about the properties defined in Objective C interfaces. The debugger consumes
1486 information generated by compiler in DWARF format. The format does not support
1487 encoding of Objective C properties. This proposal describes DWARF extensions to
1488 encode Objective C properties, which the debugger can use to let developers
1489 inspect Objective C properties.
1494 Objective C properties exist separately from class members. A property can be
1495 defined only by "setter" and "getter" selectors, and be calculated anew on each
1496 access. Or a property can just be a direct access to some declared ivar.
1497 Finally it can have an ivar "automatically synthesized" for it by the compiler,
1498 in which case the property can be referred to in user code directly using the
1499 standard C dereference syntax as well as through the property "dot" syntax, but
1500 there is no entry in the ``@interface`` declaration corresponding to this ivar.
1502 To facilitate debugging, these properties we will add a new DWARF TAG into the
1503 ``DW_TAG_structure_type`` definition for the class to hold the description of a
1504 given property, and a set of DWARF attributes that provide said description.
1505 The property tag will also contain the name and declared type of the property.
1507 If there is a related ivar, there will also be a DWARF property attribute placed
1508 in the ``DW_TAG_member`` DIE for that ivar referring back to the property TAG
1509 for that property. And in the case where the compiler synthesizes the ivar
1510 directly, the compiler is expected to generate a ``DW_TAG_member`` for that
1511 ivar (with the ``DW_AT_artificial`` set to 1), whose name will be the name used
1512 to access this ivar directly in code, and with the property attribute pointing
1513 back to the property it is backing.
1515 The following examples will serve as illustration for our discussion:
1517 .. code-block:: objc
1529 @synthesize p2 = n2;
1532 This produces the following DWARF (this is a "pseudo dwarfdump" output):
1534 .. code-block:: none
1536 0x00000100: TAG_structure_type [7] *
1537 AT_APPLE_runtime_class( 0x10 )
1539 AT_decl_file( "Objc_Property.m" )
1542 0x00000110 TAG_APPLE_property
1544 AT_type ( {0x00000150} ( int ) )
1546 0x00000120: TAG_APPLE_property
1548 AT_type ( {0x00000150} ( int ) )
1550 0x00000130: TAG_member [8]
1552 AT_APPLE_property ( {0x00000110} "p1" )
1553 AT_type( {0x00000150} ( int ) )
1554 AT_artificial ( 0x1 )
1556 0x00000140: TAG_member [8]
1558 AT_APPLE_property ( {0x00000120} "p2" )
1559 AT_type( {0x00000150} ( int ) )
1561 0x00000150: AT_type( ( int ) )
1563 Note, the current convention is that the name of the ivar for an
1564 auto-synthesized property is the name of the property from which it derives
1565 with an underscore prepended, as is shown in the example. But we actually
1566 don't need to know this convention, since we are given the name of the ivar
1569 Also, it is common practice in ObjC to have different property declarations in
1570 the @interface and @implementation - e.g. to provide a read-only property in
1571 the interface,and a read-write interface in the implementation. In that case,
1572 the compiler should emit whichever property declaration will be in force in the
1573 current translation unit.
1575 Developers can decorate a property with attributes which are encoded using
1576 ``DW_AT_APPLE_property_attribute``.
1578 .. code-block:: objc
1580 @property (readonly, nonatomic) int pr;
1582 .. code-block:: none
1584 TAG_APPLE_property [8]
1586 AT_type ( {0x00000147} (int) )
1587 AT_APPLE_property_attribute (DW_APPLE_PROPERTY_readonly, DW_APPLE_PROPERTY_nonatomic)
1589 The setter and getter method names are attached to the property using
1590 ``DW_AT_APPLE_property_setter`` and ``DW_AT_APPLE_property_getter`` attributes.
1592 .. code-block:: objc
1595 @property (setter=myOwnP3Setter:) int p3;
1596 -(void)myOwnP3Setter:(int)a;
1601 -(void)myOwnP3Setter:(int)a{ }
1604 The DWARF for this would be:
1606 .. code-block:: none
1608 0x000003bd: TAG_structure_type [7] *
1609 AT_APPLE_runtime_class( 0x10 )
1611 AT_decl_file( "Objc_Property.m" )
1614 0x000003cd TAG_APPLE_property
1616 AT_APPLE_property_setter ( "myOwnP3Setter:" )
1617 AT_type( {0x00000147} ( int ) )
1619 0x000003f3: TAG_member [8]
1621 AT_type ( {0x00000147} ( int ) )
1622 AT_APPLE_property ( {0x000003cd} )
1623 AT_artificial ( 0x1 )
1628 +-----------------------+--------+
1630 +=======================+========+
1631 | DW_TAG_APPLE_property | 0x4200 |
1632 +-----------------------+--------+
1634 New DWARF Attributes
1635 ^^^^^^^^^^^^^^^^^^^^
1637 +--------------------------------+--------+-----------+
1638 | Attribute | Value | Classes |
1639 +================================+========+===========+
1640 | DW_AT_APPLE_property | 0x3fed | Reference |
1641 +--------------------------------+--------+-----------+
1642 | DW_AT_APPLE_property_getter | 0x3fe9 | String |
1643 +--------------------------------+--------+-----------+
1644 | DW_AT_APPLE_property_setter | 0x3fea | String |
1645 +--------------------------------+--------+-----------+
1646 | DW_AT_APPLE_property_attribute | 0x3feb | Constant |
1647 +--------------------------------+--------+-----------+
1652 +--------------------------------+-------+
1654 +================================+=======+
1655 | DW_AT_APPLE_PROPERTY_readonly | 0x1 |
1656 +--------------------------------+-------+
1657 | DW_AT_APPLE_PROPERTY_readwrite | 0x2 |
1658 +--------------------------------+-------+
1659 | DW_AT_APPLE_PROPERTY_assign | 0x4 |
1660 +--------------------------------+-------+
1661 | DW_AT_APPLE_PROPERTY_retain | 0x8 |
1662 +--------------------------------+-------+
1663 | DW_AT_APPLE_PROPERTY_copy | 0x10 |
1664 +--------------------------------+-------+
1665 | DW_AT_APPLE_PROPERTY_nonatomic | 0x20 |
1666 +--------------------------------+-------+
1668 Name Accelerator Tables
1669 -----------------------
1674 The "``.debug_pubnames``" and "``.debug_pubtypes``" formats are not what a
1675 debugger needs. The "``pub``" in the section name indicates that the entries
1676 in the table are publicly visible names only. This means no static or hidden
1677 functions show up in the "``.debug_pubnames``". No static variables or private
1678 class variables are in the "``.debug_pubtypes``". Many compilers add different
1679 things to these tables, so we can't rely upon the contents between gcc, icc, or
1682 The typical query given by users tends not to match up with the contents of
1683 these tables. For example, the DWARF spec states that "In the case of the name
1684 of a function member or static data member of a C++ structure, class or union,
1685 the name presented in the "``.debug_pubnames``" section is not the simple name
1686 given by the ``DW_AT_name attribute`` of the referenced debugging information
1687 entry, but rather the fully qualified name of the data or function member."
1688 So the only names in these tables for complex C++ entries is a fully
1689 qualified name. Debugger users tend not to enter their search strings as
1690 "``a::b::c(int,const Foo&) const``", but rather as "``c``", "``b::c``" , or
1691 "``a::b::c``". So the name entered in the name table must be demangled in
1692 order to chop it up appropriately and additional names must be manually entered
1693 into the table to make it effective as a name lookup table for debuggers to
1696 All debuggers currently ignore the "``.debug_pubnames``" table as a result of
1697 its inconsistent and useless public-only name content making it a waste of
1698 space in the object file. These tables, when they are written to disk, are not
1699 sorted in any way, leaving every debugger to do its own parsing and sorting.
1700 These tables also include an inlined copy of the string values in the table
1701 itself making the tables much larger than they need to be on disk, especially
1702 for large C++ programs.
1704 Can't we just fix the sections by adding all of the names we need to this
1705 table? No, because that is not what the tables are defined to contain and we
1706 won't know the difference between the old bad tables and the new good tables.
1707 At best we could make our own renamed sections that contain all of the data we
1710 These tables are also insufficient for what a debugger like LLDB needs. LLDB
1711 uses clang for its expression parsing where LLDB acts as a PCH. LLDB is then
1712 often asked to look for type "``foo``" or namespace "``bar``", or list items in
1713 namespace "``baz``". Namespaces are not included in the pubnames or pubtypes
1714 tables. Since clang asks a lot of questions when it is parsing an expression,
1715 we need to be very fast when looking up names, as it happens a lot. Having new
1716 accelerator tables that are optimized for very quick lookups will benefit this
1717 type of debugging experience greatly.
1719 We would like to generate name lookup tables that can be mapped into memory
1720 from disk, and used as is, with little or no up-front parsing. We would also
1721 be able to control the exact content of these different tables so they contain
1722 exactly what we need. The Name Accelerator Tables were designed to fix these
1723 issues. In order to solve these issues we need to:
1725 * Have a format that can be mapped into memory from disk and used as is
1726 * Lookups should be very fast
1727 * Extensible table format so these tables can be made by many producers
1728 * Contain all of the names needed for typical lookups out of the box
1729 * Strict rules for the contents of tables
1731 Table size is important and the accelerator table format should allow the reuse
1732 of strings from common string tables so the strings for the names are not
1733 duplicated. We also want to make sure the table is ready to be used as-is by
1734 simply mapping the table into memory with minimal header parsing.
1736 The name lookups need to be fast and optimized for the kinds of lookups that
1737 debuggers tend to do. Optimally we would like to touch as few parts of the
1738 mapped table as possible when doing a name lookup and be able to quickly find
1739 the name entry we are looking for, or discover there are no matches. In the
1740 case of debuggers we optimized for lookups that fail most of the time.
1742 Each table that is defined should have strict rules on exactly what is in the
1743 accelerator tables and documented so clients can rely on the content.
1748 Standard Hash Tables
1749 """"""""""""""""""""
1751 Typical hash tables have a header, buckets, and each bucket points to the
1754 .. code-block:: none
1764 The BUCKETS are an array of offsets to DATA for each hash:
1766 .. code-block:: none
1769 | 0x00001000 | BUCKETS[0]
1770 | 0x00002000 | BUCKETS[1]
1771 | 0x00002200 | BUCKETS[2]
1772 | 0x000034f0 | BUCKETS[3]
1774 | 0xXXXXXXXX | BUCKETS[n_buckets]
1777 So for ``bucket[3]`` in the example above, we have an offset into the table
1778 0x000034f0 which points to a chain of entries for the bucket. Each bucket must
1779 contain a next pointer, full 32 bit hash value, the string itself, and the data
1780 for the current string value.
1782 .. code-block:: none
1785 0x000034f0: | 0x00003500 | next pointer
1786 | 0x12345678 | 32 bit hash
1787 | "erase" | string value
1788 | data[n] | HashData for this bucket
1790 0x00003500: | 0x00003550 | next pointer
1791 | 0x29273623 | 32 bit hash
1792 | "dump" | string value
1793 | data[n] | HashData for this bucket
1795 0x00003550: | 0x00000000 | next pointer
1796 | 0x82638293 | 32 bit hash
1797 | "main" | string value
1798 | data[n] | HashData for this bucket
1801 The problem with this layout for debuggers is that we need to optimize for the
1802 negative lookup case where the symbol we're searching for is not present. So
1803 if we were to lookup "``printf``" in the table above, we would make a 32 hash
1804 for "``printf``", it might match ``bucket[3]``. We would need to go to the
1805 offset 0x000034f0 and start looking to see if our 32 bit hash matches. To do
1806 so, we need to read the next pointer, then read the hash, compare it, and skip
1807 to the next bucket. Each time we are skipping many bytes in memory and
1808 touching new cache pages just to do the compare on the full 32 bit hash. All
1809 of these accesses then tell us that we didn't have a match.
1814 To solve the issues mentioned above we have structured the hash tables a bit
1815 differently: a header, buckets, an array of all unique 32 bit hash values,
1816 followed by an array of hash value data offsets, one for each hash value, then
1817 the data for all hash values:
1819 .. code-block:: none
1833 The ``BUCKETS`` in the name tables are an index into the ``HASHES`` array. By
1834 making all of the full 32 bit hash values contiguous in memory, we allow
1835 ourselves to efficiently check for a match while touching as little memory as
1836 possible. Most often checking the 32 bit hash values is as far as the lookup
1837 goes. If it does match, it usually is a match with no collisions. So for a
1838 table with "``n_buckets``" buckets, and "``n_hashes``" unique 32 bit hash
1839 values, we can clarify the contents of the ``BUCKETS``, ``HASHES`` and
1842 .. code-block:: none
1844 .-------------------------.
1845 | HEADER.magic | uint32_t
1846 | HEADER.version | uint16_t
1847 | HEADER.hash_function | uint16_t
1848 | HEADER.bucket_count | uint32_t
1849 | HEADER.hashes_count | uint32_t
1850 | HEADER.header_data_len | uint32_t
1851 | HEADER_DATA | HeaderData
1852 |-------------------------|
1853 | BUCKETS | uint32_t[n_buckets] // 32 bit hash indexes
1854 |-------------------------|
1855 | HASHES | uint32_t[n_hashes] // 32 bit hash values
1856 |-------------------------|
1857 | OFFSETS | uint32_t[n_hashes] // 32 bit offsets to hash value data
1858 |-------------------------|
1860 `-------------------------'
1862 So taking the exact same data from the standard hash example above we end up
1865 .. code-block:: none
1875 | ... | BUCKETS[n_buckets]
1877 | 0x........ | HASHES[0]
1878 | 0x........ | HASHES[1]
1879 | 0x........ | HASHES[2]
1880 | 0x........ | HASHES[3]
1881 | 0x........ | HASHES[4]
1882 | 0x........ | HASHES[5]
1883 | 0x12345678 | HASHES[6] hash for BUCKETS[3]
1884 | 0x29273623 | HASHES[7] hash for BUCKETS[3]
1885 | 0x82638293 | HASHES[8] hash for BUCKETS[3]
1886 | 0x........ | HASHES[9]
1887 | 0x........ | HASHES[10]
1888 | 0x........ | HASHES[11]
1889 | 0x........ | HASHES[12]
1890 | 0x........ | HASHES[13]
1891 | 0x........ | HASHES[n_hashes]
1893 | 0x........ | OFFSETS[0]
1894 | 0x........ | OFFSETS[1]
1895 | 0x........ | OFFSETS[2]
1896 | 0x........ | OFFSETS[3]
1897 | 0x........ | OFFSETS[4]
1898 | 0x........ | OFFSETS[5]
1899 | 0x000034f0 | OFFSETS[6] offset for BUCKETS[3]
1900 | 0x00003500 | OFFSETS[7] offset for BUCKETS[3]
1901 | 0x00003550 | OFFSETS[8] offset for BUCKETS[3]
1902 | 0x........ | OFFSETS[9]
1903 | 0x........ | OFFSETS[10]
1904 | 0x........ | OFFSETS[11]
1905 | 0x........ | OFFSETS[12]
1906 | 0x........ | OFFSETS[13]
1907 | 0x........ | OFFSETS[n_hashes]
1915 0x000034f0: | 0x00001203 | .debug_str ("erase")
1916 | 0x00000004 | A 32 bit array count - number of HashData with name "erase"
1917 | 0x........ | HashData[0]
1918 | 0x........ | HashData[1]
1919 | 0x........ | HashData[2]
1920 | 0x........ | HashData[3]
1921 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1923 0x00003500: | 0x00001203 | String offset into .debug_str ("collision")
1924 | 0x00000002 | A 32 bit array count - number of HashData with name "collision"
1925 | 0x........ | HashData[0]
1926 | 0x........ | HashData[1]
1927 | 0x00001203 | String offset into .debug_str ("dump")
1928 | 0x00000003 | A 32 bit array count - number of HashData with name "dump"
1929 | 0x........ | HashData[0]
1930 | 0x........ | HashData[1]
1931 | 0x........ | HashData[2]
1932 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1934 0x00003550: | 0x00001203 | String offset into .debug_str ("main")
1935 | 0x00000009 | A 32 bit array count - number of HashData with name "main"
1936 | 0x........ | HashData[0]
1937 | 0x........ | HashData[1]
1938 | 0x........ | HashData[2]
1939 | 0x........ | HashData[3]
1940 | 0x........ | HashData[4]
1941 | 0x........ | HashData[5]
1942 | 0x........ | HashData[6]
1943 | 0x........ | HashData[7]
1944 | 0x........ | HashData[8]
1945 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1948 So we still have all of the same data, we just organize it more efficiently for
1949 debugger lookup. If we repeat the same "``printf``" lookup from above, we
1950 would hash "``printf``" and find it matches ``BUCKETS[3]`` by taking the 32 bit
1951 hash value and modulo it by ``n_buckets``. ``BUCKETS[3]`` contains "6" which
1952 is the index into the ``HASHES`` table. We would then compare any consecutive
1953 32 bit hashes values in the ``HASHES`` array as long as the hashes would be in
1954 ``BUCKETS[3]``. We do this by verifying that each subsequent hash value modulo
1955 ``n_buckets`` is still 3. In the case of a failed lookup we would access the
1956 memory for ``BUCKETS[3]``, and then compare a few consecutive 32 bit hashes
1957 before we know that we have no match. We don't end up marching through
1958 multiple words of memory and we really keep the number of processor data cache
1959 lines being accessed as small as possible.
1961 The string hash that is used for these lookup tables is the Daniel J.
1962 Bernstein hash which is also used in the ELF ``GNU_HASH`` sections. It is a
1963 very good hash for all kinds of names in programs with very few hash
1966 Empty buckets are designated by using an invalid hash index of ``UINT32_MAX``.
1971 These name hash tables are designed to be generic where specializations of the
1972 table get to define additional data that goes into the header ("``HeaderData``"),
1973 how the string value is stored ("``KeyType``") and the content of the data for each
1979 The header has a fixed part, and the specialized part. The exact format of the
1986 uint32_t magic; // 'HASH' magic value to allow endian detection
1987 uint16_t version; // Version number
1988 uint16_t hash_function; // The hash function enumeration that was used
1989 uint32_t bucket_count; // The number of buckets in this hash table
1990 uint32_t hashes_count; // The total number of unique hash values and hash data offsets in this table
1991 uint32_t header_data_len; // The bytes to skip to get to the hash indexes (buckets) for correct alignment
1992 // Specifically the length of the following HeaderData field - this does not
1993 // include the size of the preceding fields
1994 HeaderData header_data; // Implementation specific header data
1997 The header starts with a 32 bit "``magic``" value which must be ``'HASH'``
1998 encoded as an ASCII integer. This allows the detection of the start of the
1999 hash table and also allows the table's byte order to be determined so the table
2000 can be correctly extracted. The "``magic``" value is followed by a 16 bit
2001 ``version`` number which allows the table to be revised and modified in the
2002 future. The current version number is 1. ``hash_function`` is a ``uint16_t``
2003 enumeration that specifies which hash function was used to produce this table.
2004 The current values for the hash function enumerations include:
2008 enum HashFunctionType
2010 eHashFunctionDJB = 0u, // Daniel J Bernstein hash function
2013 ``bucket_count`` is a 32 bit unsigned integer that represents how many buckets
2014 are in the ``BUCKETS`` array. ``hashes_count`` is the number of unique 32 bit
2015 hash values that are in the ``HASHES`` array, and is the same number of offsets
2016 are contained in the ``OFFSETS`` array. ``header_data_len`` specifies the size
2017 in bytes of the ``HeaderData`` that is filled in by specialized versions of
2023 The header is followed by the buckets, hashes, offsets, and hash value data.
2029 uint32_t buckets[Header.bucket_count]; // An array of hash indexes into the "hashes[]" array below
2030 uint32_t hashes [Header.hashes_count]; // Every unique 32 bit hash for the entire table is in this table
2031 uint32_t offsets[Header.hashes_count]; // An offset that corresponds to each item in the "hashes[]" array above
2034 ``buckets`` is an array of 32 bit indexes into the ``hashes`` array. The
2035 ``hashes`` array contains all of the 32 bit hash values for all names in the
2036 hash table. Each hash in the ``hashes`` table has an offset in the ``offsets``
2037 array that points to the data for the hash value.
2039 This table setup makes it very easy to repurpose these tables to contain
2040 different data, while keeping the lookup mechanism the same for all tables.
2041 This layout also makes it possible to save the table to disk and map it in
2042 later and do very efficient name lookups with little or no parsing.
2044 DWARF lookup tables can be implemented in a variety of ways and can store a lot
2045 of information for each name. We want to make the DWARF tables extensible and
2046 able to store the data efficiently so we have used some of the DWARF features
2047 that enable efficient data storage to define exactly what kind of data we store
2050 The ``HeaderData`` contains a definition of the contents of each HashData chunk.
2051 We might want to store an offset to all of the debug information entries (DIEs)
2052 for each name. To keep things extensible, we create a list of items, or
2053 Atoms, that are contained in the data for each name. First comes the type of
2054 the data in each atom:
2061 eAtomTypeDIEOffset = 1u, // DIE offset, check form for encoding
2062 eAtomTypeCUOffset = 2u, // DIE offset of the compiler unit header that contains the item in question
2063 eAtomTypeTag = 3u, // DW_TAG_xxx value, should be encoded as DW_FORM_data1 (if no tags exceed 255) or DW_FORM_data2
2064 eAtomTypeNameFlags = 4u, // Flags from enum NameFlags
2065 eAtomTypeTypeFlags = 5u, // Flags from enum TypeFlags
2068 The enumeration values and their meanings are:
2070 .. code-block:: none
2072 eAtomTypeNULL - a termination atom that specifies the end of the atom list
2073 eAtomTypeDIEOffset - an offset into the .debug_info section for the DWARF DIE for this name
2074 eAtomTypeCUOffset - an offset into the .debug_info section for the CU that contains the DIE
2075 eAtomTypeDIETag - The DW_TAG_XXX enumeration value so you don't have to parse the DWARF to see what it is
2076 eAtomTypeNameFlags - Flags for functions and global variables (isFunction, isInlined, isExternal...)
2077 eAtomTypeTypeFlags - Flags for types (isCXXClass, isObjCClass, ...)
2079 Then we allow each atom type to define the atom type and how the data for each
2080 atom type data is encoded:
2086 uint16_t type; // AtomType enum value
2087 uint16_t form; // DWARF DW_FORM_XXX defines
2090 The ``form`` type above is from the DWARF specification and defines the exact
2091 encoding of the data for the Atom type. See the DWARF specification for the
2092 ``DW_FORM_`` definitions.
2098 uint32_t die_offset_base;
2099 uint32_t atom_count;
2100 Atoms atoms[atom_count0];
2103 ``HeaderData`` defines the base DIE offset that should be added to any atoms
2104 that are encoded using the ``DW_FORM_ref1``, ``DW_FORM_ref2``,
2105 ``DW_FORM_ref4``, ``DW_FORM_ref8`` or ``DW_FORM_ref_udata``. It also defines
2106 what is contained in each ``HashData`` object -- ``Atom.form`` tells us how large
2107 each field will be in the ``HashData`` and the ``Atom.type`` tells us how this data
2108 should be interpreted.
2110 For the current implementations of the "``.apple_names``" (all functions +
2111 globals), the "``.apple_types``" (names of all types that are defined), and
2112 the "``.apple_namespaces``" (all namespaces), we currently set the ``Atom``
2117 HeaderData.atom_count = 1;
2118 HeaderData.atoms[0].type = eAtomTypeDIEOffset;
2119 HeaderData.atoms[0].form = DW_FORM_data4;
2121 This defines the contents to be the DIE offset (eAtomTypeDIEOffset) that is
2122 encoded as a 32 bit value (DW_FORM_data4). This allows a single name to have
2123 multiple matching DIEs in a single file, which could come up with an inlined
2124 function for instance. Future tables could include more information about the
2125 DIE such as flags indicating if the DIE is a function, method, block,
2128 The KeyType for the DWARF table is a 32 bit string table offset into the
2129 ".debug_str" table. The ".debug_str" is the string table for the DWARF which
2130 may already contain copies of all of the strings. This helps make sure, with
2131 help from the compiler, that we reuse the strings between all of the DWARF
2132 sections and keeps the hash table size down. Another benefit to having the
2133 compiler generate all strings as DW_FORM_strp in the debug info, is that
2134 DWARF parsing can be made much faster.
2136 After a lookup is made, we get an offset into the hash data. The hash data
2137 needs to be able to deal with 32 bit hash collisions, so the chunk of data
2138 at the offset in the hash data consists of a triple:
2143 uint32_t hash_data_count
2144 HashData[hash_data_count]
2146 If "str_offset" is zero, then the bucket contents are done. 99.9% of the
2147 hash data chunks contain a single item (no 32 bit hash collision):
2149 .. code-block:: none
2152 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2153 | 0x00000004 | uint32_t HashData count
2154 | 0x........ | uint32_t HashData[0] DIE offset
2155 | 0x........ | uint32_t HashData[1] DIE offset
2156 | 0x........ | uint32_t HashData[2] DIE offset
2157 | 0x........ | uint32_t HashData[3] DIE offset
2158 | 0x00000000 | uint32_t KeyType (end of hash chain)
2161 If there are collisions, you will have multiple valid string offsets:
2163 .. code-block:: none
2166 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2167 | 0x00000004 | uint32_t HashData count
2168 | 0x........ | uint32_t HashData[0] DIE offset
2169 | 0x........ | uint32_t HashData[1] DIE offset
2170 | 0x........ | uint32_t HashData[2] DIE offset
2171 | 0x........ | uint32_t HashData[3] DIE offset
2172 | 0x00002023 | uint32_t KeyType (.debug_str[0x0002023] => "print")
2173 | 0x00000002 | uint32_t HashData count
2174 | 0x........ | uint32_t HashData[0] DIE offset
2175 | 0x........ | uint32_t HashData[1] DIE offset
2176 | 0x00000000 | uint32_t KeyType (end of hash chain)
2179 Current testing with real world C++ binaries has shown that there is around 1
2180 32 bit hash collision per 100,000 name entries.
2185 As we said, we want to strictly define exactly what is included in the
2186 different tables. For DWARF, we have 3 tables: "``.apple_names``",
2187 "``.apple_types``", and "``.apple_namespaces``".
2189 "``.apple_names``" sections should contain an entry for each DWARF DIE whose
2190 ``DW_TAG`` is a ``DW_TAG_label``, ``DW_TAG_inlined_subroutine``, or
2191 ``DW_TAG_subprogram`` that has address attributes: ``DW_AT_low_pc``,
2192 ``DW_AT_high_pc``, ``DW_AT_ranges`` or ``DW_AT_entry_pc``. It also contains
2193 ``DW_TAG_variable`` DIEs that have a ``DW_OP_addr`` in the location (global and
2194 static variables). All global and static variables should be included,
2195 including those scoped within functions and classes. For example using the
2207 Both of the static ``var`` variables would be included in the table. All
2208 functions should emit both their full names and their basenames. For C or C++,
2209 the full name is the mangled name (if available) which is usually in the
2210 ``DW_AT_MIPS_linkage_name`` attribute, and the ``DW_AT_name`` contains the
2211 function basename. If global or static variables have a mangled name in a
2212 ``DW_AT_MIPS_linkage_name`` attribute, this should be emitted along with the
2213 simple name found in the ``DW_AT_name`` attribute.
2215 "``.apple_types``" sections should contain an entry for each DWARF DIE whose
2220 * DW_TAG_enumeration_type
2221 * DW_TAG_pointer_type
2222 * DW_TAG_reference_type
2223 * DW_TAG_string_type
2224 * DW_TAG_structure_type
2225 * DW_TAG_subroutine_type
2228 * DW_TAG_ptr_to_member_type
2230 * DW_TAG_subrange_type
2236 * DW_TAG_packed_type
2237 * DW_TAG_volatile_type
2238 * DW_TAG_restrict_type
2239 * DW_TAG_interface_type
2240 * DW_TAG_unspecified_type
2241 * DW_TAG_shared_type
2243 Only entries with a ``DW_AT_name`` attribute are included, and the entry must
2244 not be a forward declaration (``DW_AT_declaration`` attribute with a non-zero
2245 value). For example, using the following code:
2255 We get a few type DIEs:
2257 .. code-block:: none
2259 0x00000067: TAG_base_type [5]
2260 AT_encoding( DW_ATE_signed )
2262 AT_byte_size( 0x04 )
2264 0x0000006e: TAG_pointer_type [6]
2265 AT_type( {0x00000067} ( int ) )
2266 AT_byte_size( 0x08 )
2268 The DW_TAG_pointer_type is not included because it does not have a ``DW_AT_name``.
2270 "``.apple_namespaces``" section should contain all ``DW_TAG_namespace`` DIEs.
2271 If we run into a namespace that has no name this is an anonymous namespace, and
2272 the name should be output as "``(anonymous namespace)``" (without the quotes).
2273 Why? This matches the output of the ``abi::cxa_demangle()`` that is in the
2274 standard C++ library that demangles mangled names.
2277 Language Extensions and File Format Changes
2278 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2280 Objective-C Extensions
2281 """"""""""""""""""""""
2283 "``.apple_objc``" section should contain all ``DW_TAG_subprogram`` DIEs for an
2284 Objective-C class. The name used in the hash table is the name of the
2285 Objective-C class itself. If the Objective-C class has a category, then an
2286 entry is made for both the class name without the category, and for the class
2287 name with the category. So if we have a DIE at offset 0x1234 with a name of
2288 method "``-[NSString(my_additions) stringWithSpecialString:]``", we would add
2289 an entry for "``NSString``" that points to DIE 0x1234, and an entry for
2290 "``NSString(my_additions)``" that points to 0x1234. This allows us to quickly
2291 track down all Objective-C methods for an Objective-C class when doing
2292 expressions. It is needed because of the dynamic nature of Objective-C where
2293 anyone can add methods to a class. The DWARF for Objective-C methods is also
2294 emitted differently from C++ classes where the methods are not usually
2295 contained in the class definition, they are scattered about across one or more
2296 compile units. Categories can also be defined in different shared libraries.
2297 So we need to be able to quickly find all of the methods and class functions
2298 given the Objective-C class name, or quickly find all methods and class
2299 functions for a class + category name. This table does not contain any
2300 selector names, it just maps Objective-C class names (or class names +
2301 category) to all of the methods and class functions. The selectors are added
2302 as function basenames in the "``.debug_names``" section.
2304 In the "``.apple_names``" section for Objective-C functions, the full name is
2305 the entire function name with the brackets ("``-[NSString
2306 stringWithCString:]``") and the basename is the selector only
2307 ("``stringWithCString:``").
2312 The sections names for the apple hash tables are for non-mach-o files. For
2313 mach-o files, the sections should be contained in the ``__DWARF`` segment with
2316 * "``.apple_names``" -> "``__apple_names``"
2317 * "``.apple_types``" -> "``__apple_types``"
2318 * "``.apple_namespaces``" -> "``__apple_namespac``" (16 character limit)
2319 * "``.apple_objc``" -> "``__apple_objc``"