blob: 41bca026c9eb0c4070651303fc4f8ae57183d853 [file] [log] [blame]
Reid Spencer5f016e22007-07-11 17:01:13 +00001//===----------------------------------------------------------------------===//
2// C Language Family Front-end
3//===----------------------------------------------------------------------===//
4 Chris Lattner
5
6I. Introduction:
7
8 clang: noun
9 1. A loud, resonant, metallic sound.
10 2. The strident call of a crane or goose.
11 3. C-language family front-end toolkit.
12
13 The world needs better compiler tools, tools which are built as libraries. This
14 design point allows reuse of the tools in new and novel ways. However, building
15 the tools as libraries isn't enough: they must have clean APIs, be as
16 decoupled from each other as possible, and be easy to modify/extend. This
17 requires clean layering, decent design, and avoiding tying the libraries to a
18 specific use. Oh yeah, did I mention that we want the resultant libraries to
19 be as fast as possible? :)
20
21 This front-end is built as a component of the LLVM toolkit that can be used
22 with the LLVM backend or independently of it. In this spirit, the API has been
23 carefully designed as the following components:
24
25 libsupport - Basic support library, reused from LLVM.
26 libsystem - System abstraction library, reused from LLVM.
27
28 libbasic - Diagnostics, SourceLocations, SourceBuffer abstraction,
29 file system caching for input source files. This depends on
30 libsupport and libsystem.
31 libast - Provides classes to represent the C AST, the C type system,
32 builtin functions, and various helpers for analyzing and
33 manipulating the AST (visitors, pretty printers, etc). This
34 library depends on libbasic.
35
36 liblex - C/C++/ObjC lexing and preprocessing, identifier hash table,
37 pragma handling, tokens, and macros. This depends on libbasic.
38 libparse - C (for now) parsing and local semantic analysis. This library
39 invokes coarse-grained 'Actions' provided by the client to do
40 stuff (e.g. libsema builds ASTs). This depends on liblex.
41 libsema - Provides a set of parser actions to build a standardized AST
42 for programs. AST's are 'streamed' out a top-level declaration
43 at a time, allowing clients to use decl-at-a-time processing,
44 build up entire translation units, or even build 'whole
45 program' ASTs depending on how they use the APIs. This depends
46 on libast and libparse.
47
48 libcodegen - Lower the AST to LLVM IR for optimization & codegen. Depends
49 on libast.
50 clang - An example driver, client of the libraries at various levels.
51 This depends on all these libraries, and on LLVM VMCore.
52
Chris Lattner3321f9f2007-07-11 18:58:19 +000053 This front-end has been intentionally built as a DAG of libraries, making it
54 easy to reuse individual parts or replace pieces if desired. For example, to
55 build a preprocessor, you take the Basic and Lexer libraries. If you want an
56 indexer, you take those plus the Parser library and provide some actions for
57 indexing. If you want a refactoring, static analysis, or source-to-source
58 compiler tool, it makes sense to take those plus the AST building and semantic
59 analyzer library. Finally, if you want to use this with the LLVM backend,
60 you'd take these components plus the AST to LLVM lowering code.
Reid Spencer5f016e22007-07-11 17:01:13 +000061
62 In the future I hope this toolkit will grow to include new and interesting
63 components, including a C++ front-end, ObjC support, and a whole lot of other
64 things.
65
66 Finally, it should be pointed out that the goal here is to build something that
67 is high-quality and industrial-strength: all the obnoxious features of the C
68 family must be correctly supported (trigraphs, preprocessor arcana, K&R-style
69 prototypes, GCC/MS extensions, etc). It cannot be used if it is not 'real'.
70
71
72II. Usage of clang driver:
73
74 * Basic Command-Line Options:
75 - Help: clang --help
76 - Standard GCC options accepted: -E, -I*, -i*, -pedantic, -std=c90, etc.
77 - To make diagnostics more gcc-like: -fno-caret-diagnostics -fno-show-column
78 - Enable metric printing: -stats
79
Chris Lattner3321f9f2007-07-11 18:58:19 +000080 * -fsyntax-only is currently the default mode.
Reid Spencer5f016e22007-07-11 17:01:13 +000081
Chris Lattner3321f9f2007-07-11 18:58:19 +000082 * -E mode works the same way as GCC.
Reid Spencer5f016e22007-07-11 17:01:13 +000083
Ted Kremeneka01a1ee2007-08-29 23:26:37 +000084 * -Eonly mode does all preprocessing, but does not print the output,
85 useful for timing the preprocessor.
Reid Spencer5f016e22007-07-11 17:01:13 +000086
Ted Kremeneka01a1ee2007-08-29 23:26:37 +000087 * -fsyntax-only is currently partially implemented, lacking some
88 semantic analysis (some errors and warnings are not produced).
Chris Lattner3321f9f2007-07-11 18:58:19 +000089
Ted Kremeneka01a1ee2007-08-29 23:26:37 +000090 * -parse-noop parses code without building an AST. This is useful
91 for timing the cost of the parser without including AST building
92 time.
Reid Spencer5f016e22007-07-11 17:01:13 +000093
Ted Kremeneka01a1ee2007-08-29 23:26:37 +000094 * -parse-ast builds ASTs, but doesn't print them. This is most
95 useful for timing AST building vs -parse-noop.
Reid Spencer5f016e22007-07-11 17:01:13 +000096
Chris Lattner3321f9f2007-07-11 18:58:19 +000097 * -parse-ast-print pretty prints most expression and statements nodes.
Reid Spencer5f016e22007-07-11 17:01:13 +000098
Ted Kremeneka01a1ee2007-08-29 23:26:37 +000099 * -parse-ast-check checks that diagnostic messages that are expected
100 are reported and that those which are reported are expected.
101
102 * -dump-cfg builds ASTs and then CFGs. CFGs are then pretty-printed.
103
104 * -view-cfg builds ASTs and then CFGs. CFGs are then visualized by
105 invoking Graphviz.
106
107 For more information on getting Graphviz to work with clang/LLVM,
108 see: http://llvm.org/docs/ProgrammersManual.html#ViewGraph
Reid Spencer5f016e22007-07-11 17:01:13 +0000109
Chris Lattner3321f9f2007-07-11 18:58:19 +0000110
Reid Spencer5f016e22007-07-11 17:01:13 +0000111III. Current advantages over GCC:
112
113 * Column numbers are fully tracked (no 256 col limit, no GCC-style pruning).
114 * All diagnostics have column numbers, includes 'caret diagnostics', and they
115 highlight regions of interesting code (e.g. the LHS and RHS of a binop).
116 * Full diagnostic customization by client (can format diagnostics however they
117 like, e.g. in an IDE or refactoring tool) through DiagnosticClient interface.
118 * Built as a framework, can be reused by multiple tools.
119 * All languages supported linked into same library (no cc1,cc1obj, ...).
120 * mmap's code in read-only, does not dirty the pages like GCC (mem footprint).
121 * LLVM License, can be linked into non-GPL projects.
122 * Full diagnostic control, per diagnostic. Diagnostics are identified by ID.
123 * Significantly faster than GCC at semantic analysis, parsing, preprocessing
124 and lexing.
125 * Defers exposing platform-specific stuff to as late as possible, tracks use of
126 platform-specific features (e.g. #ifdef PPC) to allow 'portable bytecodes'.
127 * The lexer doesn't rely on the "lexer hack": it has no notion of scope and
128 does not categorize identifiers as types or variables -- this is up to the
129 parser to decide.
130
131Potential Future Features:
132
133 * Fine grained diag control within the source (#pragma enable/disable warning).
134 * Better token tracking within macros? (Token came from this line, which is
135 a macro argument instantiated here, recursively instantiated here).
136 * Fast #import with a module system.
137 * Dependency tracking: change to header file doesn't recompile every function
138 that texually depends on it: recompile only those functions that need it.
Chris Lattner3321f9f2007-07-11 18:58:19 +0000139 This is aka 'incremental parsing'.
Reid Spencer5f016e22007-07-11 17:01:13 +0000140
141
142IV. Missing Functionality / Improvements
143
144clang driver:
Chris Lattner3321f9f2007-07-11 18:58:19 +0000145 * Include search paths are hard-coded into the driver. Doh.
Reid Spencer5f016e22007-07-11 17:01:13 +0000146
147File Manager:
Chris Lattner3321f9f2007-07-11 18:58:19 +0000148 * Reduce syscalls for reduced compile time, see NOTES.txt.
Reid Spencer5f016e22007-07-11 17:01:13 +0000149
150Lexer:
151 * Source character mapping. GCC supports ASCII and UTF-8.
152 See GCC options: -ftarget-charset and -ftarget-wide-charset.
153 * Universal character support. Experimental in GCC, enabled with
154 -fextended-identifiers.
155 * -fpreprocessed mode.
156
157Preprocessor:
158 * Know about apple header maps.
159 * #assert/#unassert
160 * #line / #file directives (currently accepted and ignored).
161 * MSExtension: "L#param" stringizes to a wide string literal.
162 * Charize extension: "#define F(o) #@o F(a)" -> 'a'.
163 * Consider merging the parser's expression parser into the preprocessor to
164 eliminate duplicate code.
165 * Add support for -M*
166
167Traditional Preprocessor:
Chris Lattner3321f9f2007-07-11 18:58:19 +0000168 * Currently, we have none. :)
Reid Spencer5f016e22007-07-11 17:01:13 +0000169
170Parser:
171 * C90/K&R modes are only partially implemented.
Chris Lattner3321f9f2007-07-11 18:58:19 +0000172 * __extension__ is currently just skipped and ignored.
Reid Spencer5f016e22007-07-11 17:01:13 +0000173 * "initializers", GCC inline asm.
174
175Semantic Analysis:
176 * Perhaps 75% done.
177
Chris Lattner3321f9f2007-07-11 18:58:19 +0000178LLVM Code Gen:
179 * Still very early.
Reid Spencer5f016e22007-07-11 17:01:13 +0000180