diff options
Diffstat (limited to 'docs')
| -rw-r--r-- | docs/PTHInternals.html | 220 |
1 files changed, 220 insertions, 0 deletions
diff --git a/docs/PTHInternals.html b/docs/PTHInternals.html new file mode 100644 index 0000000000..7714fb91b2 --- /dev/null +++ b/docs/PTHInternals.html @@ -0,0 +1,220 @@ +<html> + <head> + <title>Pretokenized Headers (PTH)</title> + <link type="text/css" rel="stylesheet" href="../menu.css" /> + <link type="text/css" rel="stylesheet" href="../content.css" /> + <style type="text/css"> + td { + vertical-align: top; + } + </style> +</head> +<body> + +<!--#include virtual="../menu.html.incl"--> + +<div id="content"> +<h1>Pretokenized Headers</h1> + +<p> <a href="http://en.wikipedia.org/wiki/Precompiled_header">Precompiled +headers</a> is a general approach employed by many compilers to reduce +compilation time. The underlying motivation of the approach is that within a +codebase frequently the same (and often large) header files are included by +multiple source files. Consequently, compile times can often be greatly improved +by caching some of the (redundant) work done by a compiler to process headers. +Precompiled header files, which represent one of possibly many ways to implement +this optimization, are literally files that represent an on-disk cache that +contains the vital information necessary to reduce some (or all) of the work +needed to process a corresponding header file. While details of precompiled +headers vary between compilers, precompiled headers have been shown to be a +highly effective at speeding up program compilation on systems with very large +system headers (e.g., Mac OS X).</p> + +<p>Clang supports an implementation of precompiled headers known as +<em>pre-tokenized headers</em> (PTH). Clang's pre-tokenized headers support most +of same interfaces as GCC's pre-compiled headers (as well as others) but are +completely different in their implementation. This pages first describes the +interface for using PTH and then briefly elaborates on their design and +implementation.</p> + + +<h2>Using Pretokenized Headers (High-level Interface)</h2> + +<p>The high-level interface to generate a PTH file is the same as GCC's:</p> + +<pre> + $ gcc -x c-header test.h -o test.h.gch + $ clang -x c-header test.h -o test.h.pth +</pre> + +<p>A PTH file can then be used as a prefix header when a <tt>-include</tt> +option is passed to <tt>clang</tt>:</p> + +<pre> + $ clang -include test.h test.c -o test +</pre> + +<p>The <tt>clang</tt> driver will first check if a PTH file for <tt>test.h</tt> +is available; if so, the contents of <tt>test.h</tt> (and the files it includes) +will be processed from the PTH file. Otherwise, <tt>clang</tt> falls back to +directly processing the content of <tt>test.h</tt>. This mirrors the behavior of +GCC.</p> + +<p><b>NOTE:</b> <tt>clang</tt> does <em>not</em> automatically used PTH files +for headers that are directly included within a source file. For example:</p> + +<pre> + $ clang -x c-header test.h -o test.h.pth + $ cat test.c + #include "test.h" + $ clang test.c -o test +</pre> + +<p>In this example, <tt>clang</tt> will not automatically use the PTH file for +<tt>test.h</tt> since <tt>test.h</tt> was included directly in the source file +and not specified on the command line using <tt>-include</tt>.</p> + +<h2>Using Pretokenized Headers (Low-level Interface)</h2> + +<p>The low-level Clang driver, <tt>clang-cc</tt>, supports three command line +options for generating and using PTH files.<p> + +<p>To generate PTH files using <tt>clang-cc</tt>, use the option <tt>-emit-pth</tt>: + +<pre> + $ clang-cc test.h -emit-pth -o test.h.pth +</pre> + +<p>This option is transparently used by <tt>clang</tt> when generating PTH +files. Similarly, PTH files can be used as prefix headers using the <tt>-include-pth</tt> option:</p> + +<pre> + $ clang-cc -include-pth test.h.pth test.c -o test.s +</pre> + +<p>Alternatively, Clang's PTH files can be used as a raw "token-cache" +(or "content" cache) of the source included by the original header +file. This means that the contents of the PTH file are searched as substitutes +for <em>any</em> source files that are used by <tt>clang-cc</tt> to process a +source file. This is done by specifying the <tt>-token-cache</tt> option:</p> + +<pre> + $ cat test.h + #include<stdio.h> + $ clang-cc -emit-pth test.h -o test.h.pth + $ cat test.c + #include "test.h" + $ clang-cc test.c -o test -token-cache test.h.pth +</pre> + +<p>In this example the contents of <tt>stdio.h</tt> (and the files it includes) +will be retrieved from <tt>test.h.pth</tt>, as the PTH file is being used in +this case as a raw cache of the contents of <tt>test.h</tt>. This is a low-level +interface used to both implement the high-level PTH interface as well as to +provide alternative means to use PTH-style caching.</p> + +<h2>PTH Design and Implementation</h2> + +<p>Unlike GCC's precompiled headers, which cache the full ASTs and preprocessor +state of a header file, Clang's pretokenized header files mainly cache the raw +lexer <em>tokens</em> that are needed to segment the stream of characters in a +source file into keywords, identifiers, and operators. Consequently, PTH serves +to mainly directly speed up the lexing and preprocessing of a source file, while +parsing and type-checking must be completely redone every time a PTH file is +used.</p> + +<h3>Basic Design Tradeoffs</h3> + +<p>In the long term there are plans to provide an alternate PCH implementation +for Clang that also caches the work for parsing and type checking the contents +of header files. The current implementation of PCH in Clang as pretokenized +header files was motivated by the following factors:<p> + +<ul> +<li><p><em>Language independence</em>: PTH files are (roughly) language +independent. They work with any language that Clang's lexer can handle, +including C, Objective-C, and (in the early stages) C++. This means development +on language features at the parsing level or above (which is basically almost +all interesting pieces) does not require PTH to be modified.</p></li> + +<li><em>Simple design</em>: Relatively speaking, PTH has a simple design and +implementation, making it easy to test. Further, because the machinery for PTH +resides at the lower-levels of the Clang library stack it is fairly +straightforward to profile and optimize.</li> +</ul> + +<p>Further, compared to GCC's PCH implementation (which is the dominate +precompiled header file implementation that Clang can be directly compared +against) the PTH design in Clang yields several attractive features:</p> + +<ul> + +<li><p><em>Architecture independence</em>: In contrast to GCC's PCH files (and +those of several other compilers), Clang's PTH files are architecture +independent, requiring only a single PTH file when building an program for +multiple architectures.</p> + +<p>For example, on Mac OS X one may wish to +compile a "universal binary" that runs on PowerPC, 32-bit Intel +(i386), and 64-bit Intel architectures. In contrast, GCC requires a PCH file for +each architecture, as the definitions of types in the AST are +architecture-specific. Since a Clang PTH file essentially represents a lexical +cache of header files, a single PTH file can be safely used when compiling for +multiple architectures. This can also reduce compile times because only a single +PTH file needs to be generated during a build instead of several.</p></li> + +<li><p><em>Reduced memory pressure</em>: Similar to GCC, +Clang reads PTH files via the use of memory mapping (i.e., <tt>mmap</tt>). +Clang, however, memory maps PTH files as read-only, meaning that multiple +invocations of <tt>clang-cc</tt> can share the same pages in memory from a +memory-mapped PTH file. In comparison, GCC also memory maps its PCH files but +also modifies those pages in memory, incurring the copy-on-write costs. The +read-only nature of PTH can greatly reduce memory pressure for builds involving +multiple cores, thus improving overall scalability.</p></li> + +</ul> + +<p>Despite these strengths, PTH's simple design suffers some algorithmic +handicaps compared to other PCH strategies such as those used by GCC. While PTH +can greatly speed up the processing time of a header file, the amount of work +required to process a header file is still roughly linear in the size of the +header file. In contrast, the amount of work done by GCC to process a +precompiled header is (theoretically) constant (the ASTs for the header are +literally memory mapped into the compiler). This means that only the pieces of +the header file that are referenced by the source file including the header are +the only ones the compiler needs to process during actual compilation. While +GCC's particular implementation of PCH mitigates some of these algorithmic +strengths via the use of copy-on-write pages, the approach itself can +fundamentally dominate at an algorithmic level, especially when one considers +header files of arbitrary size.</p> + +<p>Consequently, as alluded earlier, there are plans to potentially implement an +alternative PCH implementation for Clang based on the lazy deserialization of +ASTs. This approach would theoretically have the same constant-time algorithmic +advantages just mentioned but would also retain some of the strengths of PTH +such as reduced memory pressure (ideal for multi-core builds).</p> + +<h3>Internal PTH Optimizations</h3> + +<p>While the main optimization employed by PTH is to reduce lexing time of +header files by caching pre-lexed tokens, PTH also employs several other +optimizations to speed up the processing of header files:</p> + +<ul> + +<li><p><em><tt>stat</tt> caching</em>: PTH files cache information obtained via +calls to <tt>stat</tt> that <tt>clang-cc</tt> uses to resolve which files are +included by <tt>#include</tt> directives. This greatly reduces the overhead +involved in context-switching to the kernel to resolve included files.</p></li> + +<li><p><em>Fasting skipping of <tt>#ifdef</tt>...<tt>#endif</tt> chains</em>: +PTH files record the basic structure of nested preprocessor blocks. When the +condition of the preprocessor block is false, all of its tokens are immediately +skipped instead of requiring them to be handled by Clang's +preprocessor.</p></li> + +</ul> + +</div> +</body> +</html> |