internals.html revision 1.1
1<?xml version="1.0" encoding="UTF-8" standalone="no"?> 2<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> 3<html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>Porting to New Hardware or Operating Systems</title><meta name="generator" content="DocBook XSL Stylesheets V1.75.2" /><meta name="keywords" content=" ISO C++ , internals " /><meta name="keywords" content=" ISO C++ , library " /><link rel="home" href="../spine.html" title="The GNU C++ Library Documentation" /><link rel="up" href="appendix_porting.html" title="Appendix��B.�� Porting and Maintenance" /><link rel="prev" href="appendix_porting.html" title="Appendix��B.�� Porting and Maintenance" /><link rel="next" href="test.html" title="Test" /></head><body><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">Porting to New Hardware or Operating Systems</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="appendix_porting.html">Prev</a>��</td><th width="60%" align="center">Appendix��B.�� 4 Porting and Maintenance 5 6</th><td width="20%" align="right">��<a accesskey="n" href="test.html">Next</a></td></tr></table><hr /></div><div class="sect1" title="Porting to New Hardware or Operating Systems"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="appendix.porting.internals"></a>Porting to New Hardware or Operating Systems</h2></div></div></div><p> 7</p><p>This document explains how to port libstdc++ (the GNU C++ library) to 8a new target. 9</p><p>In order to make the GNU C++ library (libstdc++) work with a new 10target, you must edit some configuration files and provide some new 11header files. Unless this is done, libstdc++ will use generic 12settings which may not be correct for your target; even if they are 13correct, they will likely be inefficient. 14 </p><p>Before you get started, make sure that you have a working C library on 15your target. The C library need not precisely comply with any 16particular standard, but should generally conform to the requirements 17imposed by the ANSI/ISO standard. 18 </p><p>In addition, you should try to verify that the C++ compiler generally 19works. It is difficult to test the C++ compiler without a working 20library, but you should at least try some minimal test cases. 21 </p><p>(Note that what we think of as a "target," the library refers to as 22a "host." The comment at the top of <code class="code">configure.ac</code> explains why.) 23 </p><div class="sect2" title="Operating System"><div class="titlepage"><div><div><h3 class="title"><a id="internals.os"></a>Operating System</h3></div></div></div><p>If you are porting to a new operating system (as opposed to a new chip 24using an existing operating system), you will need to create a new 25directory in the <code class="code">config/os</code> hierarchy. For example, the IRIX 26configuration files are all in <code class="code">config/os/irix</code>. There is no set 27way to organize the OS configuration directory. For example, 28<code class="code">config/os/solaris/solaris-2.6</code> and 29<code class="code">config/os/solaris/solaris-2.7</code> are used as configuration 30directories for these two versions of Solaris. On the other hand, both 31Solaris 2.7 and Solaris 2.8 use the <code class="code">config/os/solaris/solaris-2.7</code> 32directory. The important information is that there needs to be a 33directory under <code class="code">config/os</code> to store the files for your operating 34system. 35</p><p>You might have to change the <code class="code">configure.host</code> file to ensure that 36your new directory is activated. Look for the switch statement that sets 37<code class="code">os_include_dir</code>, and add a pattern to handle your operating system 38if the default will not suffice. The switch statement switches on only 39the OS portion of the standard target triplet; e.g., the <code class="code">solaris2.8</code> 40in <code class="code">sparc-sun-solaris2.8</code>. If the new directory is named after the 41OS portion of the triplet (the default), then nothing needs to be changed. 42 </p><p>The first file to create in this directory, should be called 43<code class="code">os_defines.h</code>. This file contains basic macro definitions 44that are required to allow the C++ library to work with your C library. 45 </p><p>Several libstdc++ source files unconditionally define the macro 46<code class="code">_POSIX_SOURCE</code>. On many systems, defining this macro causes 47large portions of the C library header files to be eliminated 48at preprocessing time. Therefore, you may have to <code class="code">#undef</code> this 49macro, or define other macros (like <code class="code">_LARGEFILE_SOURCE</code> or 50<code class="code">__EXTENSIONS__</code>). You won't know what macros to define or 51undefine at this point; you'll have to try compiling the library and 52seeing what goes wrong. If you see errors about calling functions 53that have not been declared, look in your C library headers to see if 54the functions are declared there, and then figure out what macros you 55need to define. You will need to add them to the 56<code class="code">CPLUSPLUS_CPP_SPEC</code> macro in the GCC configuration file for your 57target. It will not work to simply define these macros in 58<code class="code">os_defines.h</code>. 59 </p><p>At this time, there are a few libstdc++-specific macros which may be 60defined: 61 </p><p><code class="code">_GLIBCXX_USE_C99_CHECK</code> may be defined to 1 to check C99 62function declarations (which are not covered by specialization below) 63found in system headers against versions found in the library headers 64derived from the standard. 65 </p><p><code class="code">_GLIBCXX_USE_C99_DYNAMIC</code> may be defined to an expression that 66yields 0 if and only if the system headers are exposing proper support 67for C99 functions (which are not covered by specialization below). If 68defined, it must be 0 while bootstrapping the compiler/rebuilding the 69library. 70 </p><p><code class="code">_GLIBCXX_USE_C99_LONG_LONG_CHECK</code> may be defined to 1 to check 71the set of C99 long long function declarations found in system headers 72against versions found in the library headers derived from the 73standard. 74 75 </p><p><code class="code">_GLIBCXX_USE_C99_LONG_LONG_DYNAMIC</code> may be defined to an 76expression that yields 0 if and only if the system headers are 77exposing proper support for the set of C99 long long functions. If 78defined, it must be 0 while bootstrapping the compiler/rebuilding the 79library. 80 </p><p><code class="code">_GLIBCXX_USE_C99_FP_MACROS_DYNAMIC</code> may be defined to an 81expression that yields 0 if and only if the system headers 82are exposing proper support for the related set of macros. If defined, 83it must be 0 while bootstrapping the compiler/rebuilding the library. 84 </p><p><code class="code">_GLIBCXX_USE_C99_FLOAT_TRANSCENDENTALS_CHECK</code> may be defined 85to 1 to check the related set of function declarations found in system 86headers against versions found in the library headers derived from 87the standard. 88 </p><p><code class="code">_GLIBCXX_USE_C99_FLOAT_TRANSCENDENTALS_DYNAMIC</code> may be defined 89to an expression that yields 0 if and only if the system headers 90are exposing proper support for the related set of functions. If defined, 91it must be 0 while bootstrapping the compiler/rebuilding the library. 92 </p><p>Finally, you should bracket the entire file in an include-guard, like 93this: 94 </p><pre class="programlisting"> 95 96#ifndef _GLIBCXX_OS_DEFINES 97#define _GLIBCXX_OS_DEFINES 98... 99#endif 100</pre><p>We recommend copying an existing <code class="code">os_defines.h</code> to use as a 101starting point. 102 </p></div><div class="sect2" title="CPU"><div class="titlepage"><div><div><h3 class="title"><a id="internals.cpu"></a>CPU</h3></div></div></div><p>If you are porting to a new chip (as opposed to a new operating system 103running on an existing chip), you will need to create a new directory in the 104<code class="code">config/cpu</code> hierarchy. Much like the <a class="link" href="internals.html#internals.os" title="Operating System">Operating system</a> setup, 105there are no strict rules on how to organize the CPU configuration 106directory, but careful naming choices will allow the configury to find your 107setup files without explicit help. 108</p><p>We recommend that for a target triplet <code class="code"><CPU>-<vendor>-<OS></code>, you 109name your configuration directory <code class="code">config/cpu/<CPU></code>. If you do this, 110the configury will find the directory by itself. Otherwise you will need to 111edit the <code class="code">configure.host</code> file and, in the switch statement that sets 112<code class="code">cpu_include_dir</code>, add a pattern to handle your chip. 113 </p><p>Note that some chip families share a single configuration directory, for 114example, <code class="code">alpha</code>, <code class="code">alphaev5</code>, and <code class="code">alphaev6</code> all use the 115<code class="code">config/cpu/alpha</code> directory, and there is an entry in the 116<code class="code">configure.host</code> switch statement to handle this. 117 </p><p>The <code class="code">cpu_include_dir</code> sets default locations for the files controlling 118<a class="link" href="internals.html#internals.thread_safety" title="Thread Safety">Thread safety</a> and <a class="link" href="internals.html#internals.numeric_limits" title="Numeric Limits">Numeric limits</a>, if the defaults are not 119appropriate for your chip. 120 </p></div><div class="sect2" title="Character Types"><div class="titlepage"><div><div><h3 class="title"><a id="internals.char_types"></a>Character Types</h3></div></div></div><p>The library requires that you provide three header files to implement 121character classification, analogous to that provided by the C libraries 122<code class="code"><ctype.h></code> header. You can model these on the files provided in 123<code class="code">config/os/generic</code>. However, these files will almost 124certainly need some modification. 125</p><p>The first file to write is <code class="code">ctype_base.h</code>. This file provides 126some very basic information about character classification. The libstdc++ 127library assumes that your C library implements <code class="code"><ctype.h></code> by using 128a table (indexed by character code) containing integers, where each of 129these integers is a bit-mask indicating whether the character is 130upper-case, lower-case, alphabetic, etc. The <code class="code">ctype_base.h</code> 131file gives the type of the integer, and the values of the various bit 132masks. You will have to peer at your own <code class="code"><ctype.h></code> to figure out 133how to define the values required by this file. 134 </p><p>The <code class="code">ctype_base.h</code> header file does not need include guards. 135It should contain a single <code class="code">struct</code> definition called 136<code class="code">ctype_base</code>. This <code class="code">struct</code> should contain two type 137declarations, and one enumeration declaration, like this example, taken 138from the IRIX configuration: 139 </p><pre class="programlisting"> 140 struct ctype_base 141 { 142 typedef unsigned int mask; 143 typedef int* __to_type; 144 145 enum 146 { 147 space = _ISspace, 148 print = _ISprint, 149 cntrl = _IScntrl, 150 upper = _ISupper, 151 lower = _ISlower, 152 alpha = _ISalpha, 153 digit = _ISdigit, 154 punct = _ISpunct, 155 xdigit = _ISxdigit, 156 alnum = _ISalnum, 157 graph = _ISgraph 158 }; 159 }; 160</pre><p>The <code class="code">mask</code> type is the type of the elements in the table. If your 161C library uses a table to map lower-case numbers to upper-case numbers, 162and vice versa, you should define <code class="code">__to_type</code> to be the type of the 163elements in that table. If you don't mind taking a minor performance 164penalty, or if your library doesn't implement <code class="code">toupper</code> and 165<code class="code">tolower</code> in this way, you can pick any pointer-to-integer type, 166but you must still define the type. 167</p><p>The enumeration should give definitions for all the values in the above 168example, using the values from your native <code class="code"><ctype.h></code>. They can 169be given symbolically (as above), or numerically, if you prefer. You do 170not have to include <code class="code"><ctype.h></code> in this header; it will always be 171included before <code class="code">ctype_base.h</code> is included. 172 </p><p>The next file to write is <code class="code">ctype_noninline.h</code>, which also does 173not require include guards. This file defines a few member functions 174that will be included in <code class="code">include/bits/locale_facets.h</code>. The first 175function that must be written is the <code class="code">ctype<char>::ctype</code> 176constructor. Here is the IRIX example: 177 </p><pre class="programlisting"> 178ctype<char>::ctype(const mask* __table = 0, bool __del = false, 179 size_t __refs = 0) 180 : _Ctype_nois<char>(__refs), _M_del(__table != 0 && __del), 181 _M_toupper(NULL), 182 _M_tolower(NULL), 183 _M_ctable(NULL), 184 _M_table(!__table 185 ? (const mask*) (__libc_attr._ctype_tbl->_class + 1) 186 : __table) 187 { } 188</pre><p>There are two parts of this that you might choose to alter. The first, 189and most important, is the line involving <code class="code">__libc_attr</code>. That is 190IRIX system-dependent code that gets the base of the table mapping 191character codes to attributes. You need to substitute code that obtains 192the address of this table on your system. If you want to use your 193operating system's tables to map upper-case letters to lower-case, and 194vice versa, you should initialize <code class="code">_M_toupper</code> and 195<code class="code">_M_tolower</code> with those tables, in similar fashion. 196</p><p>Now, you have to write two functions to convert from upper-case to 197lower-case, and vice versa. Here are the IRIX versions: 198 </p><pre class="programlisting"> 199 char 200 ctype<char>::do_toupper(char __c) const 201 { return _toupper(__c); } 202 203 char 204 ctype<char>::do_tolower(char __c) const 205 { return _tolower(__c); } 206</pre><p>Your C library provides equivalents to IRIX's <code class="code">_toupper</code> and 207<code class="code">_tolower</code>. If you initialized <code class="code">_M_toupper</code> and 208<code class="code">_M_tolower</code> above, then you could use those tables instead. 209</p><p>Finally, you have to provide two utility functions that convert strings 210of characters. The versions provided here will always work - but you 211could use specialized routines for greater performance if you have 212machinery to do that on your system: 213 </p><pre class="programlisting"> 214 const char* 215 ctype<char>::do_toupper(char* __low, const char* __high) const 216 { 217 while (__low < __high) 218 { 219 *__low = do_toupper(*__low); 220 ++__low; 221 } 222 return __high; 223 } 224 225 const char* 226 ctype<char>::do_tolower(char* __low, const char* __high) const 227 { 228 while (__low < __high) 229 { 230 *__low = do_tolower(*__low); 231 ++__low; 232 } 233 return __high; 234 } 235</pre><p>You must also provide the <code class="code">ctype_inline.h</code> file, which 236contains a few more functions. On most systems, you can just copy 237<code class="code">config/os/generic/ctype_inline.h</code> and use it on your system. 238 </p><p>In detail, the functions provided test characters for particular 239properties; they are analogous to the functions like <code class="code">isalpha</code> and 240<code class="code">islower</code> provided by the C library. 241 </p><p>The first function is implemented like this on IRIX: 242 </p><pre class="programlisting"> 243 bool 244 ctype<char>:: 245 is(mask __m, char __c) const throw() 246 { return (_M_table)[(unsigned char)(__c)] & __m; } 247</pre><p>The <code class="code">_M_table</code> is the table passed in above, in the constructor. 248This is the table that contains the bitmasks for each character. The 249implementation here should work on all systems. 250</p><p>The next function is: 251 </p><pre class="programlisting"> 252 const char* 253 ctype<char>:: 254 is(const char* __low, const char* __high, mask* __vec) const throw() 255 { 256 while (__low < __high) 257 *__vec++ = (_M_table)[(unsigned char)(*__low++)]; 258 return __high; 259 } 260</pre><p>This function is similar; it copies the masks for all the characters 261from <code class="code">__low</code> up until <code class="code">__high</code> into the vector given by 262<code class="code">__vec</code>. 263</p><p>The last two functions again are entirely generic: 264 </p><pre class="programlisting"> 265 const char* 266 ctype<char>:: 267 scan_is(mask __m, const char* __low, const char* __high) const throw() 268 { 269 while (__low < __high && !this->is(__m, *__low)) 270 ++__low; 271 return __low; 272 } 273 274 const char* 275 ctype<char>:: 276 scan_not(mask __m, const char* __low, const char* __high) const throw() 277 { 278 while (__low < __high && this->is(__m, *__low)) 279 ++__low; 280 return __low; 281 } 282</pre></div><div class="sect2" title="Thread Safety"><div class="titlepage"><div><div><h3 class="title"><a id="internals.thread_safety"></a>Thread Safety</h3></div></div></div><p>The C++ library string functionality requires a couple of atomic 283operations to provide thread-safety. If you don't take any special 284action, the library will use stub versions of these functions that are 285not thread-safe. They will work fine, unless your applications are 286multi-threaded. 287</p><p>If you want to provide custom, safe, versions of these functions, there 288are two distinct approaches. One is to provide a version for your CPU, 289using assembly language constructs. The other is to use the 290thread-safety primitives in your operating system. In either case, you 291make a file called <code class="code">atomicity.h</code>, and the variable 292<code class="code">ATOMICITYH</code> must point to this file. 293 </p><p>If you are using the assembly-language approach, put this code in 294<code class="code">config/cpu/<chip>/atomicity.h</code>, where chip is the name of 295your processor (see <a class="link" href="internals.html#internals.cpu" title="CPU">CPU</a>). No additional changes are necessary to 296locate the file in this case; <code class="code">ATOMICITYH</code> will be set by default. 297 </p><p>If you are using the operating system thread-safety primitives approach, 298you can also put this code in the same CPU directory, in which case no more 299work is needed to locate the file. For examples of this approach, 300see the <code class="code">atomicity.h</code> file for IRIX or IA64. 301 </p><p>Alternatively, if the primitives are more closely related to the OS 302than they are to the CPU, you can put the <code class="code">atomicity.h</code> file in 303the <a class="link" href="internals.html#internals.os" title="Operating System">Operating system</a> directory instead. In this case, you must 304edit <code class="code">configure.host</code>, and in the switch statement that handles 305operating systems, override the <code class="code">ATOMICITYH</code> variable to point to 306the appropriate <code class="code">os_include_dir</code>. For examples of this approach, 307see the <code class="code">atomicity.h</code> file for AIX. 308 </p><p>With those bits out of the way, you have to actually write 309<code class="code">atomicity.h</code> itself. This file should be wrapped in an 310include guard named <code class="code">_GLIBCXX_ATOMICITY_H</code>. It should define one 311type, and two functions. 312 </p><p>The type is <code class="code">_Atomic_word</code>. Here is the version used on IRIX: 313 </p><pre class="programlisting"> 314typedef long _Atomic_word; 315</pre><p>This type must be a signed integral type supporting atomic operations. 316If you're using the OS approach, use the same type used by your system's 317primitives. Otherwise, use the type for which your CPU provides atomic 318primitives. 319</p><p>Then, you must provide two functions. The bodies of these functions 320must be equivalent to those provided here, but using atomic operations: 321 </p><pre class="programlisting"> 322 static inline _Atomic_word 323 __attribute__ ((__unused__)) 324 __exchange_and_add (_Atomic_word* __mem, int __val) 325 { 326 _Atomic_word __result = *__mem; 327 *__mem += __val; 328 return __result; 329 } 330 331 static inline void 332 __attribute__ ((__unused__)) 333 __atomic_add (_Atomic_word* __mem, int __val) 334 { 335 *__mem += __val; 336 } 337</pre></div><div class="sect2" title="Numeric Limits"><div class="titlepage"><div><div><h3 class="title"><a id="internals.numeric_limits"></a>Numeric Limits</h3></div></div></div><p>The C++ library requires information about the fundamental data types, 338such as the minimum and maximum representable values of each type. 339You can define each of these values individually, but it is usually 340easiest just to indicate how many bits are used in each of the data 341types and let the library do the rest. For information about the 342macros to define, see the top of <code class="code">include/bits/std_limits.h</code>. 343</p><p>If you need to define any macros, you can do so in <code class="code">os_defines.h</code>. 344However, if all operating systems for your CPU are likely to use the 345same values, you can provide a CPU-specific file instead so that you 346do not have to provide the same definitions for each operating system. 347To take that approach, create a new file called <code class="code">cpu_limits.h</code> in 348your CPU configuration directory (see <a class="link" href="internals.html#internals.cpu" title="CPU">CPU</a>). 349 </p></div><div class="sect2" title="Libtool"><div class="titlepage"><div><div><h3 class="title"><a id="internals.libtool"></a>Libtool</h3></div></div></div><p>The C++ library is compiled, archived and linked with libtool. 350Explaining the full workings of libtool is beyond the scope of this 351document, but there are a few, particular bits that are necessary for 352porting. 353</p><p>Some parts of the libstdc++ library are compiled with the libtool 354<code class="code">--tags CXX</code> option (the C++ definitions for libtool). Therefore, 355<code class="code">ltcf-cxx.sh</code> in the top-level directory needs to have the correct 356logic to compile and archive objects equivalent to the C version of libtool, 357<code class="code">ltcf-c.sh</code>. Some libtool targets have definitions for C but not 358for C++, or C++ definitions which have not been kept up to date. 359 </p><p>The C++ run-time library contains initialization code that needs to be 360run as the library is loaded. Often, that requires linking in special 361object files when the C++ library is built as a shared library, or 362taking other system-specific actions. 363 </p><p>The libstdc++ library is linked with the C version of libtool, even 364though it is a C++ library. Therefore, the C version of libtool needs to 365ensure that the run-time library initializers are run. The usual way to 366do this is to build the library using <code class="code">gcc -shared</code>. 367 </p><p>If you need to change how the library is linked, look at 368<code class="code">ltcf-c.sh</code> in the top-level directory. Find the switch statement 369that sets <code class="code">archive_cmds</code>. Here, adjust the setting for your 370operating system. 371 </p></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="appendix_porting.html">Prev</a>��</td><td width="20%" align="center"><a accesskey="u" href="appendix_porting.html">Up</a></td><td width="40%" align="right">��<a accesskey="n" href="test.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">Appendix��B.�� 372 Porting and Maintenance 373 374��</td><td width="20%" align="center"><a accesskey="h" href="../spine.html">Home</a></td><td width="40%" align="right" valign="top">��Test</td></tr></table></div></body></html> 375