• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1=======================================
2LLVM's Optional Rich Disassembly Output
3=======================================
4
5.. contents::
6   :local:
7
8Introduction
9============
10
11LLVM's default disassembly output is raw text. To allow consumers more ability
12to introspect the instructions' textual representation or to reformat for a more
13user friendly display there is an optional rich disassembly output.
14
15This optional output is sufficient to reference into individual portions of the
16instruction text. This is intended for clients like disassemblers, list file
17generators, and pretty-printers, which need more than the raw instructions and
18the ability to print them.
19
20To provide this functionality the assembly text is marked up with annotations.
21The markup is simple enough in syntax to be robust even in the case of version
22mismatches between consumers and producers. That is, the syntax generally does
23not carry semantics beyond "this text has an annotation," so consumers can
24simply ignore annotations they do not understand or do not care about.
25
26After calling ``LLVMCreateDisasm()`` to create a disassembler context the
27optional output is enable with this call:
28
29.. code-block:: c
30
31    LLVMSetDisasmOptions(DC, LLVMDisassembler_Option_UseMarkup);
32
33Then subsequent calls to ``LLVMDisasmInstruction()`` will return output strings
34with the marked up annotations.
35
36Instruction Annotations
37=======================
38
39.. _contextual markups:
40
41Contextual markups
42------------------
43
44Annoated assembly display will supply contextual markup to help clients more
45efficiently implement things like pretty printers. Most markup will be target
46independent, so clients can effectively provide good display without any target
47specific knowledge.
48
49Annotated assembly goes through the normal instruction printer, but optionally
50includes contextual tags on portions of the instruction string. An annotation
51is any '<' '>' delimited section of text(1).
52
53.. code-block:: bat
54
55    annotation: '<' tag-name tag-modifier-list ':' annotated-text '>'
56    tag-name: identifier
57    tag-modifier-list: comma delimited identifier list
58
59The tag-name is an identifier which gives the type of the annotation. For the
60first pass, this will be very simple, with memory references, registers, and
61immediates having the tag names "mem", "reg", and "imm", respectively.
62
63The tag-modifier-list is typically additional target-specific context, such as
64register class.
65
66Clients should accept and ignore any tag-names or tag-modifiers they do not
67understand, allowing the annotations to grow in richness without breaking older
68clients.
69
70For example, a possible annotation of an ARM load of a stack-relative location
71might be annotated as:
72
73.. code-block:: text
74
75   ldr <reg gpr:r0>, <mem regoffset:[<reg gpr:sp>, <imm:#4>]>
76
77
781: For assembly dialects in which '<' and/or '>' are legal tokens, a literal token is escaped by following immediately with a repeat of the character.  For example, a literal '<' character is output as '<<' in an annotated assembly string.
79
80C API Details
81-------------
82
83The intended consumers of this information use the C API, therefore the new C
84API function for the disassembler will be added to provide an option to produce
85disassembled instructions with annotations, ``LLVMSetDisasmOptions()`` and the
86``LLVMDisassembler_Option_UseMarkup`` option (see above).
87