1 ================================
2 Source Level Debugging with LLVM
3 ================================
5 .. sectionauthor:: Chris Lattner <sabre@nondot.org> and Jim Laskey <jlaskey@mac.com>
13 This document is the central repository for all information pertaining to debug
14 information in LLVM. It describes the :ref:`actual format that the LLVM debug
15 information takes <format>`, which is useful for those interested in creating
16 front-ends or dealing directly with the information. Further, this document
17 provides specific examples of what debug information for C/C++ looks like.
19 Philosophy behind LLVM debugging information
20 --------------------------------------------
22 The idea of the LLVM debugging information is to capture how the important
23 pieces of the source-language's Abstract Syntax Tree map onto LLVM code.
24 Several design aspects have shaped the solution that appears here. The
27 * Debugging information should have very little impact on the rest of the
28 compiler. No transformations, analyses, or code generators should need to
29 be modified because of debugging information.
31 * LLVM optimizations should interact in :ref:`well-defined and easily described
32 ways <intro_debugopt>` with the debugging information.
34 * Because LLVM is designed to support arbitrary programming languages,
35 LLVM-to-LLVM tools should not need to know anything about the semantics of
36 the source-level-language.
38 * Source-level languages are often **widely** different from one another.
39 LLVM should not put any restrictions of the flavor of the source-language,
40 and the debugging information should work with any language.
42 * With code generator support, it should be possible to use an LLVM compiler
43 to compile a program to native machine code and standard debugging
44 formats. This allows compatibility with traditional machine-code level
45 debuggers, like GDB or DBX.
47 The approach used by the LLVM implementation is to use a small set of
48 :ref:`intrinsic functions <format_common_intrinsics>` to define a mapping
49 between LLVM program objects and the source-level objects. The description of
50 the source-level program is maintained in LLVM metadata in an
51 :ref:`implementation-defined format <ccxx_frontend>` (the C/C++ front-end
52 currently uses working draft 7 of the `DWARF 3 standard
53 <http://www.eagercon.com/dwarf/dwarf3std.htm>`_).
55 When a program is being debugged, a debugger interacts with the user and turns
56 the stored debug information into source-language specific information. As
57 such, a debugger must be aware of the source-language, and is thus tied to a
58 specific language or family of languages.
60 Debug information consumers
61 ---------------------------
63 The role of debug information is to provide meta information normally stripped
64 away during the compilation process. This meta information provides an LLVM
65 user a relationship between generated code and the original program source
68 Currently, debug information is consumed by DwarfDebug to produce dwarf
69 information used by the gdb debugger. Other targets could use the same
70 information to produce stabs or other debug forms.
72 It would also be reasonable to use debug information to feed profiling tools
73 for analysis of generated code, or, tools for reconstructing the original
74 source from generated code.
76 TODO - expound a bit more.
80 Debugging optimized code
81 ------------------------
83 An extremely high priority of LLVM debugging information is to make it interact
84 well with optimizations and analysis. In particular, the LLVM debug
85 information provides the following guarantees:
87 * LLVM debug information **always provides information to accurately read
88 the source-level state of the program**, regardless of which LLVM
89 optimizations have been run, and without any modification to the
90 optimizations themselves. However, some optimizations may impact the
91 ability to modify the current state of the program with a debugger, such
92 as setting program variables, or calling functions that have been
95 * As desired, LLVM optimizations can be upgraded to be aware of the LLVM
96 debugging information, allowing them to update the debugging information
97 as they perform aggressive optimizations. This means that, with effort,
98 the LLVM optimizers could optimize debug code just as well as non-debug
101 * LLVM debug information does not prevent optimizations from
102 happening (for example inlining, basic block reordering/merging/cleanup,
103 tail duplication, etc).
105 * LLVM debug information is automatically optimized along with the rest of
106 the program, using existing facilities. For example, duplicate
107 information is automatically merged by the linker, and unused information
108 is automatically removed.
110 Basically, the debug information allows you to compile a program with
111 "``-O0 -g``" and get full debug information, allowing you to arbitrarily modify
112 the program as it executes from a debugger. Compiling a program with
113 "``-O3 -g``" gives you full debug information that is always available and
114 accurate for reading (e.g., you get accurate stack traces despite tail call
115 elimination and inlining), but you might lose the ability to modify the program
116 and call functions where were optimized out of the program, or inlined away
119 :ref:`LLVM test suite <test-suite-quickstart>` provides a framework to test
120 optimizer's handling of debugging information. It can be run like this:
124 % cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
127 This will test impact of debugging information on optimization passes. If
128 debugging information influences optimization passes then it will be reported
129 as a failure. See :doc:`TestingGuide` for more information on LLVM test
130 infrastructure and how to run various tests.
134 Debugging information format
135 ============================
137 LLVM debugging information has been carefully designed to make it possible for
138 the optimizer to optimize the program and debugging information without
139 necessarily having to know anything about debugging information. In
140 particular, the use of metadata avoids duplicated debugging information from
141 the beginning, and the global dead code elimination pass automatically deletes
142 debugging information for a function if it decides to delete the function.
144 To do this, most of the debugging information (descriptors for types,
145 variables, functions, source files, etc) is inserted by the language front-end
146 in the form of LLVM metadata.
148 Debug information is designed to be agnostic about the target debugger and
149 debugging information representation (e.g. DWARF/Stabs/etc). It uses a generic
150 pass to decode the information that represents variables, types, functions,
151 namespaces, etc: this allows for arbitrary source-language semantics and
152 type-systems to be used, as long as there is a module written for the target
153 debugger to interpret the information.
155 To provide basic functionality, the LLVM debugger does have to make some
156 assumptions about the source-level language being debugged, though it keeps
157 these to a minimum. The only common features that the LLVM debugger assumes
158 exist are :ref:`source files <format_files>`, and :ref:`program objects
159 <format_global_variables>`. These abstract objects are used by a debugger to
160 form stack traces, show information about local variables, etc.
162 This section of the documentation first describes the representation aspects
163 common to any source-language. :ref:`ccxx_frontend` describes the data layout
164 conventions used by the C and C++ front-ends.
166 Debug information descriptors
167 -----------------------------
169 In consideration of the complexity and volume of debug information, LLVM
170 provides a specification for well formed debug descriptors.
172 Consumers of LLVM debug information expect the descriptors for program objects
173 to start in a canonical format, but the descriptors can include additional
174 information appended at the end that is source-language specific. All LLVM
175 debugging information is versioned, allowing backwards compatibility in the
176 case that the core structures need to change in some way. Also, all debugging
177 information objects start with a tag to indicate what type of object it is.
178 The source-language is allowed to define its own objects, by using unreserved
179 tag numbers. We recommend using with tags in the range 0x1000 through 0x2000
180 (there is a defined ``enum DW_TAG_user_base = 0x1000``.)
182 The fields of debug descriptors used internally by LLVM are restricted to only
183 the simple data types ``i32``, ``i1``, ``float``, ``double``, ``mdstring`` and
193 <a name="LLVMDebugVersion">The first field of a descriptor is always an
194 ``i32`` containing a tag value identifying the content of the descriptor.
195 The remaining fields are specific to the descriptor. The values of tags are
196 loosely bound to the tag values of DWARF information entries. However, that
197 does not restrict the use of the information supplied to DWARF targets. To
198 facilitate versioning of debug information, the tag is augmented with the
199 current debug version (``LLVMDebugVersion = 8 << 16`` or 0x80000 or
202 The details of the various descriptors follow.
204 Compile unit descriptors
205 ^^^^^^^^^^^^^^^^^^^^^^^^
210 i32, ;; Tag = 17 + LLVMDebugVersion (DW_TAG_compile_unit)
211 i32, ;; Unused field.
212 i32, ;; DWARF language identifier (ex. DW_LANG_C89)
213 metadata, ;; Source file name
214 metadata, ;; Source file directory (includes trailing slash)
215 metadata ;; Producer (ex. "4.0.1 LLVM (LLVM research group)")
216 i1, ;; True if this is a main compile unit.
217 i1, ;; True if this is optimized.
219 i32 ;; Runtime version
220 metadata ;; List of enums types
221 metadata ;; List of retained types
222 metadata ;; List of subprograms
223 metadata ;; List of global variables
226 These descriptors contain a source language ID for the file (we use the DWARF
227 3.0 ID numbers, such as ``DW_LANG_C89``, ``DW_LANG_C_plus_plus``,
228 ``DW_LANG_Cobol74``, etc), three strings describing the filename, working
229 directory of the compiler, and an identifier string for the compiler that
232 Compile unit descriptors provide the root context for objects declared in a
233 specific compilation unit. File descriptors are defined using this context.
234 These descriptors are collected by a named metadata ``!llvm.dbg.cu``. They
235 keep track of subprograms, global variables and type information.
245 i32, ;; Tag = 41 + LLVMDebugVersion (DW_TAG_file_type)
246 metadata, ;; Source file name
247 metadata, ;; Source file directory (includes trailing slash)
251 These descriptors contain information for a file. Global variables and top
252 level functions would be defined using this context. File descriptors also
253 provide context for source line correspondence.
255 Each input file is encoded as a separate file descriptor in LLVM debugging
258 .. _format_global_variables:
260 Global variable descriptors
261 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
266 i32, ;; Tag = 52 + LLVMDebugVersion (DW_TAG_variable)
267 i32, ;; Unused field.
268 metadata, ;; Reference to context descriptor
270 metadata, ;; Display name (fully qualified C++ name)
271 metadata, ;; MIPS linkage name (for C++)
272 metadata, ;; Reference to file where defined
273 i32, ;; Line number where defined
274 metadata, ;; Reference to type descriptor
275 i1, ;; True if the global is local to compile unit (static)
276 i1, ;; True if the global is defined in the compile unit (not extern)
277 {}* ;; Reference to the global variable
280 These descriptors provide debug information about globals variables. They
281 provide details such as name, type and where the variable is defined. All
282 global variables are collected inside the named metadata ``!llvm.dbg.cu``.
284 .. _format_subprograms:
286 Subprogram descriptors
287 ^^^^^^^^^^^^^^^^^^^^^^
292 i32, ;; Tag = 46 + LLVMDebugVersion (DW_TAG_subprogram)
293 i32, ;; Unused field.
294 metadata, ;; Reference to context descriptor
296 metadata, ;; Display name (fully qualified C++ name)
297 metadata, ;; MIPS linkage name (for C++)
298 metadata, ;; Reference to file where defined
299 i32, ;; Line number where defined
300 metadata, ;; Reference to type descriptor
301 i1, ;; True if the global is local to compile unit (static)
302 i1, ;; True if the global is defined in the compile unit (not extern)
303 i32, ;; Line number where the scope of the subprogram begins
304 i32, ;; Virtuality, e.g. dwarf::DW_VIRTUALITY__virtual
305 i32, ;; Index into a virtual function
306 metadata, ;; indicates which base type contains the vtable pointer for the
308 i32, ;; Flags - Artifical, Private, Protected, Explicit, Prototyped.
310 Function * , ;; Pointer to LLVM function
311 metadata, ;; Lists function template parameters
312 metadata, ;; Function declaration descriptor
313 metadata ;; List of function variables
316 These descriptors provide debug information about functions, methods and
317 subprograms. They provide details such as name, return types and the source
318 location where the subprogram is defined.
326 i32, ;; Tag = 11 + LLVMDebugVersion (DW_TAG_lexical_block)
327 metadata,;; Reference to context descriptor
329 i32, ;; Column number
330 metadata,;; Reference to source file
331 i32 ;; Unique ID to identify blocks from a template function
334 This descriptor provides debug information about nested blocks within a
335 subprogram. The line number and column numbers are used to dinstinguish two
336 lexical blocks at same depth.
341 i32, ;; Tag = 11 + LLVMDebugVersion (DW_TAG_lexical_block)
342 metadata ;; Reference to the scope we're annotating with a file change
343 metadata,;; Reference to the file the scope is enclosed in.
346 This descriptor provides a wrapper around a lexical scope to handle file
347 changes in the middle of a lexical block.
349 .. _format_basic_type:
351 Basic type descriptors
352 ^^^^^^^^^^^^^^^^^^^^^^
357 i32, ;; Tag = 36 + LLVMDebugVersion (DW_TAG_base_type)
358 metadata, ;; Reference to context
359 metadata, ;; Name (may be "" for anonymous types)
360 metadata, ;; Reference to file where defined (may be NULL)
361 i32, ;; Line number where defined (may be 0)
363 i64, ;; Alignment in bits
364 i64, ;; Offset in bits
366 i32 ;; DWARF type encoding
369 These descriptors define primitive types used in the code. Example ``int``,
370 ``bool`` and ``float``. The context provides the scope of the type, which is
371 usually the top level. Since basic types are not usually user defined the
372 context and line number can be left as NULL and 0. The size, alignment and
373 offset are expressed in bits and can be 64 bit values. The alignment is used
374 to round the offset when embedded in a :ref:`composite type
375 <format_composite_type>` (example to keep float doubles on 64 bit boundaries).
376 The offset is the bit offset if embedded in a :ref:`composite type
377 <format_composite_type>`.
379 The type encoding provides the details of the type. The values are typically
380 one of the following:
388 DW_ATE_signed_char = 6
390 DW_ATE_unsigned_char = 8
392 .. _format_derived_type:
394 Derived type descriptors
395 ^^^^^^^^^^^^^^^^^^^^^^^^
400 i32, ;; Tag (see below)
401 metadata, ;; Reference to context
402 metadata, ;; Name (may be "" for anonymous types)
403 metadata, ;; Reference to file where defined (may be NULL)
404 i32, ;; Line number where defined (may be 0)
406 i64, ;; Alignment in bits
407 i64, ;; Offset in bits
408 i32, ;; Flags to encode attributes, e.g. private
409 metadata, ;; Reference to type derived from
410 metadata, ;; (optional) Name of the Objective C property associated with
411 ;; Objective-C an ivar, or the type of which this
412 ;; pointer-to-member is pointing to members of.
413 metadata, ;; (optional) Name of the Objective C property getter selector.
414 metadata, ;; (optional) Name of the Objective C property setter selector.
415 i32 ;; (optional) Objective C property attributes.
418 These descriptors are used to define types derived from other types. The value
419 of the tag varies depending on the meaning. The following are possible tag
424 DW_TAG_formal_parameter = 5
426 DW_TAG_pointer_type = 15
427 DW_TAG_reference_type = 16
429 DW_TAG_ptr_to_member_type = 31
430 DW_TAG_const_type = 38
431 DW_TAG_volatile_type = 53
432 DW_TAG_restrict_type = 55
434 ``DW_TAG_member`` is used to define a member of a :ref:`composite type
435 <format_composite_type>` or :ref:`subprogram <format_subprograms>`. The type
436 of the member is the :ref:`derived type <format_derived_type>`.
437 ``DW_TAG_formal_parameter`` is used to define a member which is a formal
438 argument of a subprogram.
440 ``DW_TAG_typedef`` is used to provide a name for the derived type.
442 ``DW_TAG_pointer_type``, ``DW_TAG_reference_type``, ``DW_TAG_const_type``,
443 ``DW_TAG_volatile_type`` and ``DW_TAG_restrict_type`` are used to qualify the
444 :ref:`derived type <format_derived_type>`.
446 :ref:`Derived type <format_derived_type>` location can be determined from the
447 context and line number. The size, alignment and offset are expressed in bits
448 and can be 64 bit values. The alignment is used to round the offset when
449 embedded in a :ref:`composite type <format_composite_type>` (example to keep
450 float doubles on 64 bit boundaries.) The offset is the bit offset if embedded
451 in a :ref:`composite type <format_composite_type>`.
453 Note that the ``void *`` type is expressed as a type derived from NULL.
455 .. _format_composite_type:
457 Composite type descriptors
458 ^^^^^^^^^^^^^^^^^^^^^^^^^^
463 i32, ;; Tag (see below)
464 metadata, ;; Reference to context
465 metadata, ;; Name (may be "" for anonymous types)
466 metadata, ;; Reference to file where defined (may be NULL)
467 i32, ;; Line number where defined (may be 0)
469 i64, ;; Alignment in bits
470 i64, ;; Offset in bits
472 metadata, ;; Reference to type derived from
473 metadata, ;; Reference to array of member descriptors
474 i32 ;; Runtime languages
477 These descriptors are used to define types that are composed of 0 or more
478 elements. The value of the tag varies depending on the meaning. The following
479 are possible tag values:
483 DW_TAG_array_type = 1
484 DW_TAG_enumeration_type = 4
485 DW_TAG_structure_type = 19
486 DW_TAG_union_type = 23
487 DW_TAG_subroutine_type = 21
488 DW_TAG_inheritance = 28
490 The vector flag indicates that an array type is a native packed vector.
492 The members of array types (tag = ``DW_TAG_array_type``) are
493 :ref:`subrange descriptors <format_subrange>`, each
494 representing the range of subscripts at that level of indexing.
496 The members of enumeration types (tag = ``DW_TAG_enumeration_type``) are
497 :ref:`enumerator descriptors <format_enumerator>`, each representing the
498 definition of enumeration value for the set. All enumeration type descriptors
499 are collected inside the named metadata ``!llvm.dbg.cu``.
501 The members of structure (tag = ``DW_TAG_structure_type``) or union (tag =
502 ``DW_TAG_union_type``) types are any one of the :ref:`basic
503 <format_basic_type>`, :ref:`derived <format_derived_type>` or :ref:`composite
504 <format_composite_type>` type descriptors, each representing a field member of
505 the structure or union.
507 For C++ classes (tag = ``DW_TAG_structure_type``), member descriptors provide
508 information about base classes, static members and member functions. If a
509 member is a :ref:`derived type descriptor <format_derived_type>` and has a tag
510 of ``DW_TAG_inheritance``, then the type represents a base class. If the member
511 of is a :ref:`global variable descriptor <format_global_variables>` then it
512 represents a static member. And, if the member is a :ref:`subprogram
513 descriptor <format_subprograms>` then it represents a member function. For
514 static members and member functions, ``getName()`` returns the members link or
515 the C++ mangled name. ``getDisplayName()`` the simplied version of the name.
517 The first member of subroutine (tag = ``DW_TAG_subroutine_type``) type elements
518 is the return type for the subroutine. The remaining elements are the formal
519 arguments to the subroutine.
521 :ref:`Composite type <format_composite_type>` location can be determined from
522 the context and line number. The size, alignment and offset are expressed in
523 bits and can be 64 bit values. The alignment is used to round the offset when
524 embedded in a :ref:`composite type <format_composite_type>` (as an example, to
525 keep float doubles on 64 bit boundaries). The offset is the bit offset if
526 embedded in a :ref:`composite type <format_composite_type>`.
536 i32, ;; Tag = 33 + LLVMDebugVersion (DW_TAG_subrange_type)
541 These descriptors are used to define ranges of array subscripts for an array
542 :ref:`composite type <format_composite_type>`. The low value defines the lower
543 bounds typically zero for C/C++. The high value is the upper bounds. Values
544 are 64 bit. ``High - Low + 1`` is the size of the array. If ``Low > High``
545 the array bounds are not included in generated debugging information.
547 .. _format_enumerator:
549 Enumerator descriptors
550 ^^^^^^^^^^^^^^^^^^^^^^
555 i32, ;; Tag = 40 + LLVMDebugVersion (DW_TAG_enumerator)
560 These descriptors are used to define members of an enumeration :ref:`composite
561 type <format_composite_type>`, it associates the name to the value.
569 i32, ;; Tag (see below)
572 metadata, ;; Reference to file where defined
573 i32, ;; 24 bit - Line number where defined
574 ;; 8 bit - Argument number. 1 indicates 1st argument.
575 metadata, ;; Type descriptor
577 metadata ;; (optional) Reference to inline location
580 These descriptors are used to define variables local to a sub program. The
581 value of the tag depends on the usage of the variable:
585 DW_TAG_auto_variable = 256
586 DW_TAG_arg_variable = 257
588 An auto variable is any variable declared in the body of the function. An
589 argument variable is any variable that appears as a formal argument to the
592 The context is either the subprogram or block where the variable is defined.
593 Name the source variable name. Context and line indicate where the variable
594 was defined. Type descriptor defines the declared type of the variable.
596 .. _format_common_intrinsics:
598 Debugger intrinsic functions
599 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
601 LLVM uses several intrinsic functions (name prefixed with "``llvm.dbg``") to
602 provide debug information at various points in generated code.
609 void %llvm.dbg.declare(metadata, metadata)
611 This intrinsic provides information about a local element (e.g., variable).
612 The first argument is metadata holding the alloca for the variable. The second
613 argument is metadata containing a description of the variable.
620 void %llvm.dbg.value(metadata, i64, metadata)
622 This intrinsic provides information when a user source variable is set to a new
623 value. The first argument is the new value (wrapped as metadata). The second
624 argument is the offset in the user source variable where the new value is
625 written. The third argument is metadata containing a description of the user
628 Object lifetimes and scoping
629 ============================
631 In many languages, the local variables in functions can have their lifetimes or
632 scopes limited to a subset of a function. In the C family of languages, for
633 example, variables are only live (readable and writable) within the source
634 block that they are defined in. In functional languages, values are only
635 readable after they have been defined. Though this is a very obvious concept,
636 it is non-trivial to model in LLVM, because it has no notion of scoping in this
637 sense, and does not want to be tied to a language's scoping rules.
639 In order to handle this, the LLVM debug format uses the metadata attached to
640 llvm instructions to encode line number and scoping information. Consider the
641 following C fragment, for example:
655 Compiled to LLVM, this function would be represented like this:
659 define void @foo() nounwind ssp {
661 %X = alloca i32, align 4 ; <i32*> [#uses=4]
662 %Y = alloca i32, align 4 ; <i32*> [#uses=4]
663 %Z = alloca i32, align 4 ; <i32*> [#uses=3]
664 %0 = bitcast i32* %X to {}* ; <{}*> [#uses=1]
665 call void @llvm.dbg.declare(metadata !{i32 * %X}, metadata !0), !dbg !7
666 store i32 21, i32* %X, !dbg !8
667 %1 = bitcast i32* %Y to {}* ; <{}*> [#uses=1]
668 call void @llvm.dbg.declare(metadata !{i32 * %Y}, metadata !9), !dbg !10
669 store i32 22, i32* %Y, !dbg !11
670 %2 = bitcast i32* %Z to {}* ; <{}*> [#uses=1]
671 call void @llvm.dbg.declare(metadata !{i32 * %Z}, metadata !12), !dbg !14
672 store i32 23, i32* %Z, !dbg !15
673 %tmp = load i32* %X, !dbg !16 ; <i32> [#uses=1]
674 %tmp1 = load i32* %Y, !dbg !16 ; <i32> [#uses=1]
675 %add = add nsw i32 %tmp, %tmp1, !dbg !16 ; <i32> [#uses=1]
676 store i32 %add, i32* %Z, !dbg !16
677 %tmp2 = load i32* %Y, !dbg !17 ; <i32> [#uses=1]
678 store i32 %tmp2, i32* %X, !dbg !17
682 declare void @llvm.dbg.declare(metadata, metadata) nounwind readnone
684 !0 = metadata !{i32 459008, metadata !1, metadata !"X",
685 metadata !3, i32 2, metadata !6}; [ DW_TAG_auto_variable ]
686 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
687 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo", metadata !"foo",
688 metadata !"foo", metadata !3, i32 1, metadata !4,
689 i1 false, i1 true}; [DW_TAG_subprogram ]
690 !3 = metadata !{i32 458769, i32 0, i32 12, metadata !"foo.c",
691 metadata !"/private/tmp", metadata !"clang 1.1", i1 true,
692 i1 false, metadata !"", i32 0}; [DW_TAG_compile_unit ]
693 !4 = metadata !{i32 458773, metadata !3, metadata !"", null, i32 0, i64 0, i64 0,
694 i64 0, i32 0, null, metadata !5, i32 0}; [DW_TAG_subroutine_type ]
695 !5 = metadata !{null}
696 !6 = metadata !{i32 458788, metadata !3, metadata !"int", metadata !3, i32 0,
697 i64 32, i64 32, i64 0, i32 0, i32 5}; [DW_TAG_base_type ]
698 !7 = metadata !{i32 2, i32 7, metadata !1, null}
699 !8 = metadata !{i32 2, i32 3, metadata !1, null}
700 !9 = metadata !{i32 459008, metadata !1, metadata !"Y", metadata !3, i32 3,
701 metadata !6}; [ DW_TAG_auto_variable ]
702 !10 = metadata !{i32 3, i32 7, metadata !1, null}
703 !11 = metadata !{i32 3, i32 3, metadata !1, null}
704 !12 = metadata !{i32 459008, metadata !13, metadata !"Z", metadata !3, i32 5,
705 metadata !6}; [ DW_TAG_auto_variable ]
706 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
707 !14 = metadata !{i32 5, i32 9, metadata !13, null}
708 !15 = metadata !{i32 5, i32 5, metadata !13, null}
709 !16 = metadata !{i32 6, i32 5, metadata !13, null}
710 !17 = metadata !{i32 8, i32 3, metadata !1, null}
711 !18 = metadata !{i32 9, i32 1, metadata !2, null}
713 This example illustrates a few important details about LLVM debugging
714 information. In particular, it shows how the ``llvm.dbg.declare`` intrinsic and
715 location information, which are attached to an instruction, are applied
716 together to allow a debugger to analyze the relationship between statements,
717 variable definitions, and the code used to implement the function.
721 call void @llvm.dbg.declare(metadata, metadata !0), !dbg !7
723 The first intrinsic ``%llvm.dbg.declare`` encodes debugging information for the
724 variable ``X``. The metadata ``!dbg !7`` attached to the intrinsic provides
725 scope information for the variable ``X``.
729 !7 = metadata !{i32 2, i32 7, metadata !1, null}
730 !1 = metadata !{i32 458763, metadata !2}; [DW_TAG_lexical_block ]
731 !2 = metadata !{i32 458798, i32 0, metadata !3, metadata !"foo",
732 metadata !"foo", metadata !"foo", metadata !3, i32 1,
733 metadata !4, i1 false, i1 true}; [DW_TAG_subprogram ]
735 Here ``!7`` is metadata providing location information. It has four fields:
736 line number, column number, scope, and original scope. The original scope
737 represents inline location if this instruction is inlined inside a caller, and
738 is null otherwise. In this example, scope is encoded by ``!1``. ``!1``
739 represents a lexical block inside the scope ``!2``, where ``!2`` is a
740 :ref:`subprogram descriptor <format_subprograms>`. This way the location
741 information attached to the intrinsics indicates that the variable ``X`` is
742 declared at line number 2 at a function level scope in function ``foo``.
744 Now lets take another example.
748 call void @llvm.dbg.declare(metadata, metadata !12), !dbg !14
750 The second intrinsic ``%llvm.dbg.declare`` encodes debugging information for
751 variable ``Z``. The metadata ``!dbg !14`` attached to the intrinsic provides
752 scope information for the variable ``Z``.
756 !13 = metadata !{i32 458763, metadata !1}; [DW_TAG_lexical_block ]
757 !14 = metadata !{i32 5, i32 9, metadata !13, null}
759 Here ``!14`` indicates that ``Z`` is declared at line number 5 and
760 column number 9 inside of lexical scope ``!13``. The lexical scope itself
761 resides inside of lexical scope ``!1`` described above.
763 The scope information attached with each instruction provides a straightforward
764 way to find instructions covered by a scope.
768 C/C++ front-end specific debug information
769 ==========================================
771 The C and C++ front-ends represent information about the program in a format
772 that is effectively identical to `DWARF 3.0
773 <http://www.eagercon.com/dwarf/dwarf3std.htm>`_ in terms of information
774 content. This allows code generators to trivially support native debuggers by
775 generating standard dwarf information, and contains enough information for
776 non-dwarf targets to translate it as needed.
778 This section describes the forms used to represent C and C++ programs. Other
779 languages could pattern themselves after this (which itself is tuned to
780 representing programs in the same way that DWARF 3 does), or they could choose
781 to provide completely different forms if they don't fit into the DWARF model.
782 As support for debugging information gets added to the various LLVM
783 source-language front-ends, the information used should be documented here.
785 The following sections provide examples of various C/C++ constructs and the
786 debug information that would best describe those constructs.
788 C/C++ source file information
789 -----------------------------
791 Given the source files ``MySource.cpp`` and ``MyHeader.h`` located in the
792 directory ``/Users/mine/sources``, the following code:
796 #include "MyHeader.h"
798 int main(int argc, char *argv[]) {
802 a C/C++ front-end would generate the following descriptors:
808 ;; Define the compile unit for the main source file "/Users/mine/sources/MySource.cpp".
813 i32 4, ;; Language Id
814 metadata !"MySource.cpp",
815 metadata !"/Users/mine/sources",
816 metadata !"4.2.1 (Based on Apple Inc. build 5649) (LLVM build 00)",
817 i1 true, ;; Main Compile Unit
818 i1 false, ;; Optimized compile unit
819 metadata !"", ;; Compiler flags
820 i32 0} ;; Runtime version
823 ;; Define the file for the file "/Users/mine/sources/MySource.cpp".
827 metadata !"MySource.cpp",
828 metadata !"/Users/mine/sources",
829 metadata !2 ;; Compile unit
833 ;; Define the file for the file "/Users/mine/sources/Myheader.h"
837 metadata !"Myheader.h"
838 metadata !"/Users/mine/sources",
839 metadata !2 ;; Compile unit
844 ``llvm::Instruction`` provides easy access to metadata attached with an
845 instruction. One can extract line number information encoded in LLVM IR using
846 ``Instruction::getMetadata()`` and ``DILocation::getLineNumber()``.
850 if (MDNode *N = I->getMetadata("dbg")) { // Here I is an LLVM instruction
851 DILocation Loc(N); // DILocation is in DebugInfo.h
852 unsigned Line = Loc.getLineNumber();
853 StringRef File = Loc.getFilename();
854 StringRef Dir = Loc.getDirectory();
857 C/C++ global variable information
858 ---------------------------------
860 Given an integer global variable declared as follows:
866 a C/C++ front-end would generate the following descriptors:
871 ;; Define the global itself.
873 %MyGlobal = global int 100
876 ;; List of debug info of globals
880 ;; Define the compile unit.
885 metadata !"foo.cpp", ;; File
886 metadata !"/Volumes/Data/tmp", ;; Directory
887 metadata !"clang version 3.1 ", ;; Producer
888 i1 true, ;; Deprecated field
889 i1 false, ;; "isOptimized"?
890 metadata !"", ;; Flags
891 i32 0, ;; Runtime Version
892 metadata !1, ;; Enum Types
893 metadata !1, ;; Retained Types
894 metadata !1, ;; Subprograms
895 metadata !3 ;; Global Variables
896 } ; [ DW_TAG_compile_unit ]
898 ;; The Array of Global Variables
908 ;; Define the global variable itself.
914 metadata !"MyGlobal", ;; Name
915 metadata !"MyGlobal", ;; Display Name
916 metadata !"", ;; Linkage Name
920 i32 0, ;; IsLocalToUnit
921 i32 1, ;; IsDefinition
922 i32* @MyGlobal ;; LLVM-IR Value
923 } ; [ DW_TAG_variable ]
930 metadata !"foo.cpp", ;; File
931 metadata !"/Volumes/Data/tmp", ;; Directory
933 } ; [ DW_TAG_file_type ]
941 metadata !"int", ;; Name
944 i64 32, ;; Size in Bits
945 i64 32, ;; Align in Bits
949 } ; [ DW_TAG_base_type ]
951 C/C++ function information
952 --------------------------
954 Given a function declared as follows:
958 int main(int argc, char *argv[]) {
962 a C/C++ front-end would generate the following descriptors:
967 ;; Define the anchor for subprograms. Note that the second field of the
968 ;; anchor is 46, which is the same as the tag for subprograms
969 ;; (46 = DW_TAG_subprogram.)
974 metadata !1, ;; Context
975 metadata !"main", ;; Name
976 metadata !"main", ;; Display name
977 metadata !"main", ;; Linkage name
979 i32 1, ;; Line number
981 i1 false, ;; Is local
982 i1 true, ;; Is definition
983 i32 0, ;; Virtuality attribute, e.g. pure virtual function
984 i32 0, ;; Index into virtual table for C++ methods
985 i32 0, ;; Type that holds virtual table.
987 i1 false, ;; True if this function is optimized
988 Function *, ;; Pointer to llvm::Function
989 null ;; Function template parameters
992 ;; Define the subprogram itself.
994 define i32 @main(i32 %argc, i8** %argv) {
1001 The following are the basic type descriptors for C/C++ core types:
1006 .. code-block:: llvm
1010 metadata !1, ;; Context
1011 metadata !"bool", ;; Name
1012 metadata !1, ;; File
1013 i32 0, ;; Line number
1014 i64 8, ;; Size in Bits
1015 i64 8, ;; Align in Bits
1016 i64 0, ;; Offset in Bits
1024 .. code-block:: llvm
1028 metadata !1, ;; Context
1029 metadata !"char", ;; Name
1030 metadata !1, ;; File
1031 i32 0, ;; Line number
1032 i64 8, ;; Size in Bits
1033 i64 8, ;; Align in Bits
1034 i64 0, ;; Offset in Bits
1042 .. code-block:: llvm
1046 metadata !1, ;; Context
1047 metadata !"unsigned char",
1048 metadata !1, ;; File
1049 i32 0, ;; Line number
1050 i64 8, ;; Size in Bits
1051 i64 8, ;; Align in Bits
1052 i64 0, ;; Offset in Bits
1060 .. code-block:: llvm
1064 metadata !1, ;; Context
1065 metadata !"short int",
1066 metadata !1, ;; File
1067 i32 0, ;; Line number
1068 i64 16, ;; Size in Bits
1069 i64 16, ;; Align in Bits
1070 i64 0, ;; Offset in Bits
1078 .. code-block:: llvm
1082 metadata !1, ;; Context
1083 metadata !"short unsigned int",
1084 metadata !1, ;; File
1085 i32 0, ;; Line number
1086 i64 16, ;; Size in Bits
1087 i64 16, ;; Align in Bits
1088 i64 0, ;; Offset in Bits
1096 .. code-block:: llvm
1100 metadata !1, ;; Context
1101 metadata !"int", ;; Name
1102 metadata !1, ;; File
1103 i32 0, ;; Line number
1104 i64 32, ;; Size in Bits
1105 i64 32, ;; Align in Bits
1106 i64 0, ;; Offset in Bits
1114 .. code-block:: llvm
1118 metadata !1, ;; Context
1119 metadata !"unsigned int",
1120 metadata !1, ;; File
1121 i32 0, ;; Line number
1122 i64 32, ;; Size in Bits
1123 i64 32, ;; Align in Bits
1124 i64 0, ;; Offset in Bits
1132 .. code-block:: llvm
1136 metadata !1, ;; Context
1137 metadata !"long long int",
1138 metadata !1, ;; File
1139 i32 0, ;; Line number
1140 i64 64, ;; Size in Bits
1141 i64 64, ;; Align in Bits
1142 i64 0, ;; Offset in Bits
1150 .. code-block:: llvm
1154 metadata !1, ;; Context
1155 metadata !"long long unsigned int",
1156 metadata !1, ;; File
1157 i32 0, ;; Line number
1158 i64 64, ;; Size in Bits
1159 i64 64, ;; Align in Bits
1160 i64 0, ;; Offset in Bits
1168 .. code-block:: llvm
1172 metadata !1, ;; Context
1174 metadata !1, ;; File
1175 i32 0, ;; Line number
1176 i64 32, ;; Size in Bits
1177 i64 32, ;; Align in Bits
1178 i64 0, ;; Offset in Bits
1186 .. code-block:: llvm
1190 metadata !1, ;; Context
1191 metadata !"double",;; Name
1192 metadata !1, ;; File
1193 i32 0, ;; Line number
1194 i64 64, ;; Size in Bits
1195 i64 64, ;; Align in Bits
1196 i64 0, ;; Offset in Bits
1204 Given the following as an example of C/C++ derived type:
1208 typedef const int *IntPtr;
1210 a C/C++ front-end would generate the following descriptors:
1212 .. code-block:: llvm
1215 ;; Define the typedef "IntPtr".
1219 metadata !1, ;; Context
1220 metadata !"IntPtr", ;; Name
1221 metadata !3, ;; File
1222 i32 0, ;; Line number
1223 i64 0, ;; Size in bits
1224 i64 0, ;; Align in bits
1225 i64 0, ;; Offset in bits
1227 metadata !4 ;; Derived From type
1230 ;; Define the pointer type.
1234 metadata !1, ;; Context
1235 metadata !"", ;; Name
1236 metadata !1, ;; File
1237 i32 0, ;; Line number
1238 i64 64, ;; Size in bits
1239 i64 64, ;; Align in bits
1240 i64 0, ;; Offset in bits
1242 metadata !5 ;; Derived From type
1245 ;; Define the const type.
1249 metadata !1, ;; Context
1250 metadata !"", ;; Name
1251 metadata !1, ;; File
1252 i32 0, ;; Line number
1253 i64 32, ;; Size in bits
1254 i64 32, ;; Align in bits
1255 i64 0, ;; Offset in bits
1257 metadata !6 ;; Derived From type
1260 ;; Define the int type.
1264 metadata !1, ;; Context
1265 metadata !"int", ;; Name
1266 metadata !1, ;; File
1267 i32 0, ;; Line number
1268 i64 32, ;; Size in bits
1269 i64 32, ;; Align in bits
1270 i64 0, ;; Offset in bits
1275 C/C++ struct/union types
1276 ------------------------
1278 Given the following as an example of C/C++ struct type:
1288 a C/C++ front-end would generate the following descriptors:
1290 .. code-block:: llvm
1293 ;; Define basic type for unsigned int.
1297 metadata !1, ;; Context
1298 metadata !"unsigned int",
1299 metadata !1, ;; File
1300 i32 0, ;; Line number
1301 i64 32, ;; Size in Bits
1302 i64 32, ;; Align in Bits
1303 i64 0, ;; Offset in Bits
1308 ;; Define composite type for struct Color.
1312 metadata !1, ;; Context
1313 metadata !"Color", ;; Name
1314 metadata !1, ;; Compile unit
1315 i32 1, ;; Line number
1316 i64 96, ;; Size in bits
1317 i64 32, ;; Align in bits
1318 i64 0, ;; Offset in bits
1320 null, ;; Derived From
1321 metadata !3, ;; Elements
1322 i32 0 ;; Runtime Language
1326 ;; Define the Red field.
1330 metadata !1, ;; Context
1331 metadata !"Red", ;; Name
1332 metadata !1, ;; File
1333 i32 2, ;; Line number
1334 i64 32, ;; Size in bits
1335 i64 32, ;; Align in bits
1336 i64 0, ;; Offset in bits
1338 metadata !5 ;; Derived From type
1342 ;; Define the Green field.
1346 metadata !1, ;; Context
1347 metadata !"Green", ;; Name
1348 metadata !1, ;; File
1349 i32 3, ;; Line number
1350 i64 32, ;; Size in bits
1351 i64 32, ;; Align in bits
1352 i64 32, ;; Offset in bits
1354 metadata !5 ;; Derived From type
1358 ;; Define the Blue field.
1362 metadata !1, ;; Context
1363 metadata !"Blue", ;; Name
1364 metadata !1, ;; File
1365 i32 4, ;; Line number
1366 i64 32, ;; Size in bits
1367 i64 32, ;; Align in bits
1368 i64 64, ;; Offset in bits
1370 metadata !5 ;; Derived From type
1374 ;; Define the array of fields used by the composite type Color.
1376 !3 = metadata !{metadata !4, metadata !6, metadata !7}
1378 C/C++ enumeration types
1379 -----------------------
1381 Given the following as an example of C/C++ enumeration type:
1391 a C/C++ front-end would generate the following descriptors:
1393 .. code-block:: llvm
1396 ;; Define composite type for enum Trees
1400 metadata !1, ;; Context
1401 metadata !"Trees", ;; Name
1402 metadata !1, ;; File
1403 i32 1, ;; Line number
1404 i64 32, ;; Size in bits
1405 i64 32, ;; Align in bits
1406 i64 0, ;; Offset in bits
1408 null, ;; Derived From type
1409 metadata !3, ;; Elements
1410 i32 0 ;; Runtime language
1414 ;; Define the array of enumerators used by composite type Trees.
1416 !3 = metadata !{metadata !4, metadata !5, metadata !6}
1419 ;; Define Spruce enumerator.
1421 !4 = metadata !{i32 524328, metadata !"Spruce", i64 100}
1424 ;; Define Oak enumerator.
1426 !5 = metadata !{i32 524328, metadata !"Oak", i64 200}
1429 ;; Define Maple enumerator.
1431 !6 = metadata !{i32 524328, metadata !"Maple", i64 300}
1433 Debugging information format
1434 ============================
1436 Debugging Information Extension for Objective C Properties
1437 ----------------------------------------------------------
1442 Objective C provides a simpler way to declare and define accessor methods using
1443 declared properties. The language provides features to declare a property and
1444 to let compiler synthesize accessor methods.
1446 The debugger lets developer inspect Objective C interfaces and their instance
1447 variables and class variables. However, the debugger does not know anything
1448 about the properties defined in Objective C interfaces. The debugger consumes
1449 information generated by compiler in DWARF format. The format does not support
1450 encoding of Objective C properties. This proposal describes DWARF extensions to
1451 encode Objective C properties, which the debugger can use to let developers
1452 inspect Objective C properties.
1457 Objective C properties exist separately from class members. A property can be
1458 defined only by "setter" and "getter" selectors, and be calculated anew on each
1459 access. Or a property can just be a direct access to some declared ivar.
1460 Finally it can have an ivar "automatically synthesized" for it by the compiler,
1461 in which case the property can be referred to in user code directly using the
1462 standard C dereference syntax as well as through the property "dot" syntax, but
1463 there is no entry in the ``@interface`` declaration corresponding to this ivar.
1465 To facilitate debugging, these properties we will add a new DWARF TAG into the
1466 ``DW_TAG_structure_type`` definition for the class to hold the description of a
1467 given property, and a set of DWARF attributes that provide said description.
1468 The property tag will also contain the name and declared type of the property.
1470 If there is a related ivar, there will also be a DWARF property attribute placed
1471 in the ``DW_TAG_member`` DIE for that ivar referring back to the property TAG
1472 for that property. And in the case where the compiler synthesizes the ivar
1473 directly, the compiler is expected to generate a ``DW_TAG_member`` for that
1474 ivar (with the ``DW_AT_artificial`` set to 1), whose name will be the name used
1475 to access this ivar directly in code, and with the property attribute pointing
1476 back to the property it is backing.
1478 The following examples will serve as illustration for our discussion:
1480 .. code-block:: objc
1492 @synthesize p2 = n2;
1495 This produces the following DWARF (this is a "pseudo dwarfdump" output):
1497 .. code-block:: none
1499 0x00000100: TAG_structure_type [7] *
1500 AT_APPLE_runtime_class( 0x10 )
1502 AT_decl_file( "Objc_Property.m" )
1505 0x00000110 TAG_APPLE_property
1507 AT_type ( {0x00000150} ( int ) )
1509 0x00000120: TAG_APPLE_property
1511 AT_type ( {0x00000150} ( int ) )
1513 0x00000130: TAG_member [8]
1515 AT_APPLE_property ( {0x00000110} "p1" )
1516 AT_type( {0x00000150} ( int ) )
1517 AT_artificial ( 0x1 )
1519 0x00000140: TAG_member [8]
1521 AT_APPLE_property ( {0x00000120} "p2" )
1522 AT_type( {0x00000150} ( int ) )
1524 0x00000150: AT_type( ( int ) )
1526 Note, the current convention is that the name of the ivar for an
1527 auto-synthesized property is the name of the property from which it derives
1528 with an underscore prepended, as is shown in the example. But we actually
1529 don't need to know this convention, since we are given the name of the ivar
1532 Also, it is common practice in ObjC to have different property declarations in
1533 the @interface and @implementation - e.g. to provide a read-only property in
1534 the interface,and a read-write interface in the implementation. In that case,
1535 the compiler should emit whichever property declaration will be in force in the
1536 current translation unit.
1538 Developers can decorate a property with attributes which are encoded using
1539 ``DW_AT_APPLE_property_attribute``.
1541 .. code-block:: objc
1543 @property (readonly, nonatomic) int pr;
1545 .. code-block:: none
1547 TAG_APPLE_property [8]
1549 AT_type ( {0x00000147} (int) )
1550 AT_APPLE_property_attribute (DW_APPLE_PROPERTY_readonly, DW_APPLE_PROPERTY_nonatomic)
1552 The setter and getter method names are attached to the property using
1553 ``DW_AT_APPLE_property_setter`` and ``DW_AT_APPLE_property_getter`` attributes.
1555 .. code-block:: objc
1558 @property (setter=myOwnP3Setter:) int p3;
1559 -(void)myOwnP3Setter:(int)a;
1564 -(void)myOwnP3Setter:(int)a{ }
1567 The DWARF for this would be:
1569 .. code-block:: none
1571 0x000003bd: TAG_structure_type [7] *
1572 AT_APPLE_runtime_class( 0x10 )
1574 AT_decl_file( "Objc_Property.m" )
1577 0x000003cd TAG_APPLE_property
1579 AT_APPLE_property_setter ( "myOwnP3Setter:" )
1580 AT_type( {0x00000147} ( int ) )
1582 0x000003f3: TAG_member [8]
1584 AT_type ( {0x00000147} ( int ) )
1585 AT_APPLE_property ( {0x000003cd} )
1586 AT_artificial ( 0x1 )
1591 +-----------------------+--------+
1593 +=======================+========+
1594 | DW_TAG_APPLE_property | 0x4200 |
1595 +-----------------------+--------+
1597 New DWARF Attributes
1598 ^^^^^^^^^^^^^^^^^^^^
1600 +--------------------------------+--------+-----------+
1601 | Attribute | Value | Classes |
1602 +================================+========+===========+
1603 | DW_AT_APPLE_property | 0x3fed | Reference |
1604 +--------------------------------+--------+-----------+
1605 | DW_AT_APPLE_property_getter | 0x3fe9 | String |
1606 +--------------------------------+--------+-----------+
1607 | DW_AT_APPLE_property_setter | 0x3fea | String |
1608 +--------------------------------+--------+-----------+
1609 | DW_AT_APPLE_property_attribute | 0x3feb | Constant |
1610 +--------------------------------+--------+-----------+
1615 +--------------------------------+-------+
1617 +================================+=======+
1618 | DW_AT_APPLE_PROPERTY_readonly | 0x1 |
1619 +--------------------------------+-------+
1620 | DW_AT_APPLE_PROPERTY_readwrite | 0x2 |
1621 +--------------------------------+-------+
1622 | DW_AT_APPLE_PROPERTY_assign | 0x4 |
1623 +--------------------------------+-------+
1624 | DW_AT_APPLE_PROPERTY_retain | 0x8 |
1625 +--------------------------------+-------+
1626 | DW_AT_APPLE_PROPERTY_copy | 0x10 |
1627 +--------------------------------+-------+
1628 | DW_AT_APPLE_PROPERTY_nonatomic | 0x20 |
1629 +--------------------------------+-------+
1631 Name Accelerator Tables
1632 -----------------------
1637 The "``.debug_pubnames``" and "``.debug_pubtypes``" formats are not what a
1638 debugger needs. The "``pub``" in the section name indicates that the entries
1639 in the table are publicly visible names only. This means no static or hidden
1640 functions show up in the "``.debug_pubnames``". No static variables or private
1641 class variables are in the "``.debug_pubtypes``". Many compilers add different
1642 things to these tables, so we can't rely upon the contents between gcc, icc, or
1645 The typical query given by users tends not to match up with the contents of
1646 these tables. For example, the DWARF spec states that "In the case of the name
1647 of a function member or static data member of a C++ structure, class or union,
1648 the name presented in the "``.debug_pubnames``" section is not the simple name
1649 given by the ``DW_AT_name attribute`` of the referenced debugging information
1650 entry, but rather the fully qualified name of the data or function member."
1651 So the only names in these tables for complex C++ entries is a fully
1652 qualified name. Debugger users tend not to enter their search strings as
1653 "``a::b::c(int,const Foo&) const``", but rather as "``c``", "``b::c``" , or
1654 "``a::b::c``". So the name entered in the name table must be demangled in
1655 order to chop it up appropriately and additional names must be manually entered
1656 into the table to make it effective as a name lookup table for debuggers to
1659 All debuggers currently ignore the "``.debug_pubnames``" table as a result of
1660 its inconsistent and useless public-only name content making it a waste of
1661 space in the object file. These tables, when they are written to disk, are not
1662 sorted in any way, leaving every debugger to do its own parsing and sorting.
1663 These tables also include an inlined copy of the string values in the table
1664 itself making the tables much larger than they need to be on disk, especially
1665 for large C++ programs.
1667 Can't we just fix the sections by adding all of the names we need to this
1668 table? No, because that is not what the tables are defined to contain and we
1669 won't know the difference between the old bad tables and the new good tables.
1670 At best we could make our own renamed sections that contain all of the data we
1673 These tables are also insufficient for what a debugger like LLDB needs. LLDB
1674 uses clang for its expression parsing where LLDB acts as a PCH. LLDB is then
1675 often asked to look for type "``foo``" or namespace "``bar``", or list items in
1676 namespace "``baz``". Namespaces are not included in the pubnames or pubtypes
1677 tables. Since clang asks a lot of questions when it is parsing an expression,
1678 we need to be very fast when looking up names, as it happens a lot. Having new
1679 accelerator tables that are optimized for very quick lookups will benefit this
1680 type of debugging experience greatly.
1682 We would like to generate name lookup tables that can be mapped into memory
1683 from disk, and used as is, with little or no up-front parsing. We would also
1684 be able to control the exact content of these different tables so they contain
1685 exactly what we need. The Name Accelerator Tables were designed to fix these
1686 issues. In order to solve these issues we need to:
1688 * Have a format that can be mapped into memory from disk and used as is
1689 * Lookups should be very fast
1690 * Extensible table format so these tables can be made by many producers
1691 * Contain all of the names needed for typical lookups out of the box
1692 * Strict rules for the contents of tables
1694 Table size is important and the accelerator table format should allow the reuse
1695 of strings from common string tables so the strings for the names are not
1696 duplicated. We also want to make sure the table is ready to be used as-is by
1697 simply mapping the table into memory with minimal header parsing.
1699 The name lookups need to be fast and optimized for the kinds of lookups that
1700 debuggers tend to do. Optimally we would like to touch as few parts of the
1701 mapped table as possible when doing a name lookup and be able to quickly find
1702 the name entry we are looking for, or discover there are no matches. In the
1703 case of debuggers we optimized for lookups that fail most of the time.
1705 Each table that is defined should have strict rules on exactly what is in the
1706 accelerator tables and documented so clients can rely on the content.
1711 Standard Hash Tables
1712 """"""""""""""""""""
1714 Typical hash tables have a header, buckets, and each bucket points to the
1717 .. code-block:: none
1727 The BUCKETS are an array of offsets to DATA for each hash:
1729 .. code-block:: none
1732 | 0x00001000 | BUCKETS[0]
1733 | 0x00002000 | BUCKETS[1]
1734 | 0x00002200 | BUCKETS[2]
1735 | 0x000034f0 | BUCKETS[3]
1737 | 0xXXXXXXXX | BUCKETS[n_buckets]
1740 So for ``bucket[3]`` in the example above, we have an offset into the table
1741 0x000034f0 which points to a chain of entries for the bucket. Each bucket must
1742 contain a next pointer, full 32 bit hash value, the string itself, and the data
1743 for the current string value.
1745 .. code-block:: none
1748 0x000034f0: | 0x00003500 | next pointer
1749 | 0x12345678 | 32 bit hash
1750 | "erase" | string value
1751 | data[n] | HashData for this bucket
1753 0x00003500: | 0x00003550 | next pointer
1754 | 0x29273623 | 32 bit hash
1755 | "dump" | string value
1756 | data[n] | HashData for this bucket
1758 0x00003550: | 0x00000000 | next pointer
1759 | 0x82638293 | 32 bit hash
1760 | "main" | string value
1761 | data[n] | HashData for this bucket
1764 The problem with this layout for debuggers is that we need to optimize for the
1765 negative lookup case where the symbol we're searching for is not present. So
1766 if we were to lookup "``printf``" in the table above, we would make a 32 hash
1767 for "``printf``", it might match ``bucket[3]``. We would need to go to the
1768 offset 0x000034f0 and start looking to see if our 32 bit hash matches. To do
1769 so, we need to read the next pointer, then read the hash, compare it, and skip
1770 to the next bucket. Each time we are skipping many bytes in memory and
1771 touching new cache pages just to do the compare on the full 32 bit hash. All
1772 of these accesses then tell us that we didn't have a match.
1777 To solve the issues mentioned above we have structured the hash tables a bit
1778 differently: a header, buckets, an array of all unique 32 bit hash values,
1779 followed by an array of hash value data offsets, one for each hash value, then
1780 the data for all hash values:
1782 .. code-block:: none
1796 The ``BUCKETS`` in the name tables are an index into the ``HASHES`` array. By
1797 making all of the full 32 bit hash values contiguous in memory, we allow
1798 ourselves to efficiently check for a match while touching as little memory as
1799 possible. Most often checking the 32 bit hash values is as far as the lookup
1800 goes. If it does match, it usually is a match with no collisions. So for a
1801 table with "``n_buckets``" buckets, and "``n_hashes``" unique 32 bit hash
1802 values, we can clarify the contents of the ``BUCKETS``, ``HASHES`` and
1805 .. code-block:: none
1807 .-------------------------.
1808 | HEADER.magic | uint32_t
1809 | HEADER.version | uint16_t
1810 | HEADER.hash_function | uint16_t
1811 | HEADER.bucket_count | uint32_t
1812 | HEADER.hashes_count | uint32_t
1813 | HEADER.header_data_len | uint32_t
1814 | HEADER_DATA | HeaderData
1815 |-------------------------|
1816 | BUCKETS | uint32_t[bucket_count] // 32 bit hash indexes
1817 |-------------------------|
1818 | HASHES | uint32_t[hashes_count] // 32 bit hash values
1819 |-------------------------|
1820 | OFFSETS | uint32_t[hashes_count] // 32 bit offsets to hash value data
1821 |-------------------------|
1823 `-------------------------'
1825 So taking the exact same data from the standard hash example above we end up
1828 .. code-block:: none
1838 | ... | BUCKETS[n_buckets]
1840 | 0x........ | HASHES[0]
1841 | 0x........ | HASHES[1]
1842 | 0x........ | HASHES[2]
1843 | 0x........ | HASHES[3]
1844 | 0x........ | HASHES[4]
1845 | 0x........ | HASHES[5]
1846 | 0x12345678 | HASHES[6] hash for BUCKETS[3]
1847 | 0x29273623 | HASHES[7] hash for BUCKETS[3]
1848 | 0x82638293 | HASHES[8] hash for BUCKETS[3]
1849 | 0x........ | HASHES[9]
1850 | 0x........ | HASHES[10]
1851 | 0x........ | HASHES[11]
1852 | 0x........ | HASHES[12]
1853 | 0x........ | HASHES[13]
1854 | 0x........ | HASHES[n_hashes]
1856 | 0x........ | OFFSETS[0]
1857 | 0x........ | OFFSETS[1]
1858 | 0x........ | OFFSETS[2]
1859 | 0x........ | OFFSETS[3]
1860 | 0x........ | OFFSETS[4]
1861 | 0x........ | OFFSETS[5]
1862 | 0x000034f0 | OFFSETS[6] offset for BUCKETS[3]
1863 | 0x00003500 | OFFSETS[7] offset for BUCKETS[3]
1864 | 0x00003550 | OFFSETS[8] offset for BUCKETS[3]
1865 | 0x........ | OFFSETS[9]
1866 | 0x........ | OFFSETS[10]
1867 | 0x........ | OFFSETS[11]
1868 | 0x........ | OFFSETS[12]
1869 | 0x........ | OFFSETS[13]
1870 | 0x........ | OFFSETS[n_hashes]
1878 0x000034f0: | 0x00001203 | .debug_str ("erase")
1879 | 0x00000004 | A 32 bit array count - number of HashData with name "erase"
1880 | 0x........ | HashData[0]
1881 | 0x........ | HashData[1]
1882 | 0x........ | HashData[2]
1883 | 0x........ | HashData[3]
1884 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1886 0x00003500: | 0x00001203 | String offset into .debug_str ("collision")
1887 | 0x00000002 | A 32 bit array count - number of HashData with name "collision"
1888 | 0x........ | HashData[0]
1889 | 0x........ | HashData[1]
1890 | 0x00001203 | String offset into .debug_str ("dump")
1891 | 0x00000003 | A 32 bit array count - number of HashData with name "dump"
1892 | 0x........ | HashData[0]
1893 | 0x........ | HashData[1]
1894 | 0x........ | HashData[2]
1895 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1897 0x00003550: | 0x00001203 | String offset into .debug_str ("main")
1898 | 0x00000009 | A 32 bit array count - number of HashData with name "main"
1899 | 0x........ | HashData[0]
1900 | 0x........ | HashData[1]
1901 | 0x........ | HashData[2]
1902 | 0x........ | HashData[3]
1903 | 0x........ | HashData[4]
1904 | 0x........ | HashData[5]
1905 | 0x........ | HashData[6]
1906 | 0x........ | HashData[7]
1907 | 0x........ | HashData[8]
1908 | 0x00000000 | String offset into .debug_str (terminate data for hash)
1911 So we still have all of the same data, we just organize it more efficiently for
1912 debugger lookup. If we repeat the same "``printf``" lookup from above, we
1913 would hash "``printf``" and find it matches ``BUCKETS[3]`` by taking the 32 bit
1914 hash value and modulo it by ``n_buckets``. ``BUCKETS[3]`` contains "6" which
1915 is the index into the ``HASHES`` table. We would then compare any consecutive
1916 32 bit hashes values in the ``HASHES`` array as long as the hashes would be in
1917 ``BUCKETS[3]``. We do this by verifying that each subsequent hash value modulo
1918 ``n_buckets`` is still 3. In the case of a failed lookup we would access the
1919 memory for ``BUCKETS[3]``, and then compare a few consecutive 32 bit hashes
1920 before we know that we have no match. We don't end up marching through
1921 multiple words of memory and we really keep the number of processor data cache
1922 lines being accessed as small as possible.
1924 The string hash that is used for these lookup tables is the Daniel J.
1925 Bernstein hash which is also used in the ELF ``GNU_HASH`` sections. It is a
1926 very good hash for all kinds of names in programs with very few hash
1929 Empty buckets are designated by using an invalid hash index of ``UINT32_MAX``.
1934 These name hash tables are designed to be generic where specializations of the
1935 table get to define additional data that goes into the header ("``HeaderData``"),
1936 how the string value is stored ("``KeyType``") and the content of the data for each
1942 The header has a fixed part, and the specialized part. The exact format of the
1949 uint32_t magic; // 'HASH' magic value to allow endian detection
1950 uint16_t version; // Version number
1951 uint16_t hash_function; // The hash function enumeration that was used
1952 uint32_t bucket_count; // The number of buckets in this hash table
1953 uint32_t hashes_count; // The total number of unique hash values and hash data offsets in this table
1954 uint32_t header_data_len; // The bytes to skip to get to the hash indexes (buckets) for correct alignment
1955 // Specifically the length of the following HeaderData field - this does not
1956 // include the size of the preceding fields
1957 HeaderData header_data; // Implementation specific header data
1960 The header starts with a 32 bit "``magic``" value which must be ``'HASH'``
1961 encoded as an ASCII integer. This allows the detection of the start of the
1962 hash table and also allows the table's byte order to be determined so the table
1963 can be correctly extracted. The "``magic``" value is followed by a 16 bit
1964 ``version`` number which allows the table to be revised and modified in the
1965 future. The current version number is 1. ``hash_function`` is a ``uint16_t``
1966 enumeration that specifies which hash function was used to produce this table.
1967 The current values for the hash function enumerations include:
1971 enum HashFunctionType
1973 eHashFunctionDJB = 0u, // Daniel J Bernstein hash function
1976 ``bucket_count`` is a 32 bit unsigned integer that represents how many buckets
1977 are in the ``BUCKETS`` array. ``hashes_count`` is the number of unique 32 bit
1978 hash values that are in the ``HASHES`` array, and is the same number of offsets
1979 are contained in the ``OFFSETS`` array. ``header_data_len`` specifies the size
1980 in bytes of the ``HeaderData`` that is filled in by specialized versions of
1986 The header is followed by the buckets, hashes, offsets, and hash value data.
1992 uint32_t buckets[Header.bucket_count]; // An array of hash indexes into the "hashes[]" array below
1993 uint32_t hashes [Header.hashes_count]; // Every unique 32 bit hash for the entire table is in this table
1994 uint32_t offsets[Header.hashes_count]; // An offset that corresponds to each item in the "hashes[]" array above
1997 ``buckets`` is an array of 32 bit indexes into the ``hashes`` array. The
1998 ``hashes`` array contains all of the 32 bit hash values for all names in the
1999 hash table. Each hash in the ``hashes`` table has an offset in the ``offsets``
2000 array that points to the data for the hash value.
2002 This table setup makes it very easy to repurpose these tables to contain
2003 different data, while keeping the lookup mechanism the same for all tables.
2004 This layout also makes it possible to save the table to disk and map it in
2005 later and do very efficient name lookups with little or no parsing.
2007 DWARF lookup tables can be implemented in a variety of ways and can store a lot
2008 of information for each name. We want to make the DWARF tables extensible and
2009 able to store the data efficiently so we have used some of the DWARF features
2010 that enable efficient data storage to define exactly what kind of data we store
2013 The ``HeaderData`` contains a definition of the contents of each HashData chunk.
2014 We might want to store an offset to all of the debug information entries (DIEs)
2015 for each name. To keep things extensible, we create a list of items, or
2016 Atoms, that are contained in the data for each name. First comes the type of
2017 the data in each atom:
2024 eAtomTypeDIEOffset = 1u, // DIE offset, check form for encoding
2025 eAtomTypeCUOffset = 2u, // DIE offset of the compiler unit header that contains the item in question
2026 eAtomTypeTag = 3u, // DW_TAG_xxx value, should be encoded as DW_FORM_data1 (if no tags exceed 255) or DW_FORM_data2
2027 eAtomTypeNameFlags = 4u, // Flags from enum NameFlags
2028 eAtomTypeTypeFlags = 5u, // Flags from enum TypeFlags
2031 The enumeration values and their meanings are:
2033 .. code-block:: none
2035 eAtomTypeNULL - a termination atom that specifies the end of the atom list
2036 eAtomTypeDIEOffset - an offset into the .debug_info section for the DWARF DIE for this name
2037 eAtomTypeCUOffset - an offset into the .debug_info section for the CU that contains the DIE
2038 eAtomTypeDIETag - The DW_TAG_XXX enumeration value so you don't have to parse the DWARF to see what it is
2039 eAtomTypeNameFlags - Flags for functions and global variables (isFunction, isInlined, isExternal...)
2040 eAtomTypeTypeFlags - Flags for types (isCXXClass, isObjCClass, ...)
2042 Then we allow each atom type to define the atom type and how the data for each
2043 atom type data is encoded:
2049 uint16_t type; // AtomType enum value
2050 uint16_t form; // DWARF DW_FORM_XXX defines
2053 The ``form`` type above is from the DWARF specification and defines the exact
2054 encoding of the data for the Atom type. See the DWARF specification for the
2055 ``DW_FORM_`` definitions.
2061 uint32_t die_offset_base;
2062 uint32_t atom_count;
2063 Atoms atoms[atom_count0];
2066 ``HeaderData`` defines the base DIE offset that should be added to any atoms
2067 that are encoded using the ``DW_FORM_ref1``, ``DW_FORM_ref2``,
2068 ``DW_FORM_ref4``, ``DW_FORM_ref8`` or ``DW_FORM_ref_udata``. It also defines
2069 what is contained in each ``HashData`` object -- ``Atom.form`` tells us how large
2070 each field will be in the ``HashData`` and the ``Atom.type`` tells us how this data
2071 should be interpreted.
2073 For the current implementations of the "``.apple_names``" (all functions +
2074 globals), the "``.apple_types``" (names of all types that are defined), and
2075 the "``.apple_namespaces``" (all namespaces), we currently set the ``Atom``
2080 HeaderData.atom_count = 1;
2081 HeaderData.atoms[0].type = eAtomTypeDIEOffset;
2082 HeaderData.atoms[0].form = DW_FORM_data4;
2084 This defines the contents to be the DIE offset (eAtomTypeDIEOffset) that is
2085 encoded as a 32 bit value (DW_FORM_data4). This allows a single name to have
2086 multiple matching DIEs in a single file, which could come up with an inlined
2087 function for instance. Future tables could include more information about the
2088 DIE such as flags indicating if the DIE is a function, method, block,
2091 The KeyType for the DWARF table is a 32 bit string table offset into the
2092 ".debug_str" table. The ".debug_str" is the string table for the DWARF which
2093 may already contain copies of all of the strings. This helps make sure, with
2094 help from the compiler, that we reuse the strings between all of the DWARF
2095 sections and keeps the hash table size down. Another benefit to having the
2096 compiler generate all strings as DW_FORM_strp in the debug info, is that
2097 DWARF parsing can be made much faster.
2099 After a lookup is made, we get an offset into the hash data. The hash data
2100 needs to be able to deal with 32 bit hash collisions, so the chunk of data
2101 at the offset in the hash data consists of a triple:
2106 uint32_t hash_data_count
2107 HashData[hash_data_count]
2109 If "str_offset" is zero, then the bucket contents are done. 99.9% of the
2110 hash data chunks contain a single item (no 32 bit hash collision):
2112 .. code-block:: none
2115 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2116 | 0x00000004 | uint32_t HashData count
2117 | 0x........ | uint32_t HashData[0] DIE offset
2118 | 0x........ | uint32_t HashData[1] DIE offset
2119 | 0x........ | uint32_t HashData[2] DIE offset
2120 | 0x........ | uint32_t HashData[3] DIE offset
2121 | 0x00000000 | uint32_t KeyType (end of hash chain)
2124 If there are collisions, you will have multiple valid string offsets:
2126 .. code-block:: none
2129 | 0x00001023 | uint32_t KeyType (.debug_str[0x0001023] => "main")
2130 | 0x00000004 | uint32_t HashData count
2131 | 0x........ | uint32_t HashData[0] DIE offset
2132 | 0x........ | uint32_t HashData[1] DIE offset
2133 | 0x........ | uint32_t HashData[2] DIE offset
2134 | 0x........ | uint32_t HashData[3] DIE offset
2135 | 0x00002023 | uint32_t KeyType (.debug_str[0x0002023] => "print")
2136 | 0x00000002 | uint32_t HashData count
2137 | 0x........ | uint32_t HashData[0] DIE offset
2138 | 0x........ | uint32_t HashData[1] DIE offset
2139 | 0x00000000 | uint32_t KeyType (end of hash chain)
2142 Current testing with real world C++ binaries has shown that there is around 1
2143 32 bit hash collision per 100,000 name entries.
2148 As we said, we want to strictly define exactly what is included in the
2149 different tables. For DWARF, we have 3 tables: "``.apple_names``",
2150 "``.apple_types``", and "``.apple_namespaces``".
2152 "``.apple_names``" sections should contain an entry for each DWARF DIE whose
2153 ``DW_TAG`` is a ``DW_TAG_label``, ``DW_TAG_inlined_subroutine``, or
2154 ``DW_TAG_subprogram`` that has address attributes: ``DW_AT_low_pc``,
2155 ``DW_AT_high_pc``, ``DW_AT_ranges`` or ``DW_AT_entry_pc``. It also contains
2156 ``DW_TAG_variable`` DIEs that have a ``DW_OP_addr`` in the location (global and
2157 static variables). All global and static variables should be included,
2158 including those scoped within functions and classes. For example using the
2170 Both of the static ``var`` variables would be included in the table. All
2171 functions should emit both their full names and their basenames. For C or C++,
2172 the full name is the mangled name (if available) which is usually in the
2173 ``DW_AT_MIPS_linkage_name`` attribute, and the ``DW_AT_name`` contains the
2174 function basename. If global or static variables have a mangled name in a
2175 ``DW_AT_MIPS_linkage_name`` attribute, this should be emitted along with the
2176 simple name found in the ``DW_AT_name`` attribute.
2178 "``.apple_types``" sections should contain an entry for each DWARF DIE whose
2183 * DW_TAG_enumeration_type
2184 * DW_TAG_pointer_type
2185 * DW_TAG_reference_type
2186 * DW_TAG_string_type
2187 * DW_TAG_structure_type
2188 * DW_TAG_subroutine_type
2191 * DW_TAG_ptr_to_member_type
2193 * DW_TAG_subrange_type
2199 * DW_TAG_packed_type
2200 * DW_TAG_volatile_type
2201 * DW_TAG_restrict_type
2202 * DW_TAG_interface_type
2203 * DW_TAG_unspecified_type
2204 * DW_TAG_shared_type
2206 Only entries with a ``DW_AT_name`` attribute are included, and the entry must
2207 not be a forward declaration (``DW_AT_declaration`` attribute with a non-zero
2208 value). For example, using the following code:
2218 We get a few type DIEs:
2220 .. code-block:: none
2222 0x00000067: TAG_base_type [5]
2223 AT_encoding( DW_ATE_signed )
2225 AT_byte_size( 0x04 )
2227 0x0000006e: TAG_pointer_type [6]
2228 AT_type( {0x00000067} ( int ) )
2229 AT_byte_size( 0x08 )
2231 The DW_TAG_pointer_type is not included because it does not have a ``DW_AT_name``.
2233 "``.apple_namespaces``" section should contain all ``DW_TAG_namespace`` DIEs.
2234 If we run into a namespace that has no name this is an anonymous namespace, and
2235 the name should be output as "``(anonymous namespace)``" (without the quotes).
2236 Why? This matches the output of the ``abi::cxa_demangle()`` that is in the
2237 standard C++ library that demangles mangled names.
2240 Language Extensions and File Format Changes
2241 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2243 Objective-C Extensions
2244 """"""""""""""""""""""
2246 "``.apple_objc``" section should contain all ``DW_TAG_subprogram`` DIEs for an
2247 Objective-C class. The name used in the hash table is the name of the
2248 Objective-C class itself. If the Objective-C class has a category, then an
2249 entry is made for both the class name without the category, and for the class
2250 name with the category. So if we have a DIE at offset 0x1234 with a name of
2251 method "``-[NSString(my_additions) stringWithSpecialString:]``", we would add
2252 an entry for "``NSString``" that points to DIE 0x1234, and an entry for
2253 "``NSString(my_additions)``" that points to 0x1234. This allows us to quickly
2254 track down all Objective-C methods for an Objective-C class when doing
2255 expressions. It is needed because of the dynamic nature of Objective-C where
2256 anyone can add methods to a class. The DWARF for Objective-C methods is also
2257 emitted differently from C++ classes where the methods are not usually
2258 contained in the class definition, they are scattered about across one or more
2259 compile units. Categories can also be defined in different shared libraries.
2260 So we need to be able to quickly find all of the methods and class functions
2261 given the Objective-C class name, or quickly find all methods and class
2262 functions for a class + category name. This table does not contain any
2263 selector names, it just maps Objective-C class names (or class names +
2264 category) to all of the methods and class functions. The selectors are added
2265 as function basenames in the "``.debug_names``" section.
2267 In the "``.apple_names``" section for Objective-C functions, the full name is
2268 the entire function name with the brackets ("``-[NSString
2269 stringWithCString:]``") and the basename is the selector only
2270 ("``stringWithCString:``").
2275 The sections names for the apple hash tables are for non mach-o files. For
2276 mach-o files, the sections should be contained in the ``__DWARF`` segment with
2279 * "``.apple_names``" -> "``__apple_names``"
2280 * "``.apple_types``" -> "``__apple_types``"
2281 * "``.apple_namespaces``" -> "``__apple_namespac``" (16 character limit)
2282 * "``.apple_objc``" -> "``__apple_objc``"