Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 1 | <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" |
| 2 | "http://www.w3.org/TR/html4/strict.dtd"> |
| 3 | <html> |
| 4 | <head> |
| 5 | <title>System Library</title> |
| 6 | <link rel="stylesheet" href="llvm.css" type="text/css"> |
| 7 | </head> |
| 8 | <body> |
| 9 | |
Misha Brukman | ebbbf25 | 2004-07-21 18:02:43 +0000 | [diff] [blame] | 10 | <div class="doc_title">System Library</div> |
Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 11 | |
Misha Brukman | ebbbf25 | 2004-07-21 18:02:43 +0000 | [diff] [blame] | 12 | <div class="doc_warning"> |
| 13 | <p>Warning: This document is a work in progress.</p> |
| 14 | </div> |
Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 15 | |
| 16 | <ul> |
| 17 | <li><a href="#abstract">Abstract</a></li> |
| 18 | <li><a href="#requirements">System Library Requirements</a> |
| 19 | <ol> |
| 20 | <li><a href="#headers">Hide System Header Files</a></li> |
| 21 | <li><a href="#nofunc">No Exposed Functions</a></li> |
| 22 | <li><a href="#nodata">No Exposed Data</a></li> |
| 23 | <li><a href="#xcptns">No Exceptions</a></li> |
| 24 | <li><a href="#errors">Standard Error Codes</a></li> |
| 25 | <li><a href="#overhead">Minimize Overhead</a></li> |
| 26 | </ol></li> |
| 27 | <li><a href="#design">System Library Design</a> |
| 28 | <ol> |
| 29 | <li><a href="#opaque">Use Opaque Classes</a></li> |
| 30 | <li><a href="#common">Common Implementations</a></li> |
| 31 | <li><a href="#multi_imps">Multiple Implementations</a></li> |
| 32 | <li><a href="#lowlevel">Use Low Level Interfaces</a></li> |
| 33 | <li><a href="#memalloc">No Memory Allocation</a></li> |
| 34 | <li><a href="#virtuals">No Virtual Methods</a></li> |
| 35 | </ol></li> |
| 36 | <li><a href="#detail">System Library Details</a> |
| 37 | <ol> |
| 38 | <li><a href="#bug">Tracking Bugzilla Bug: 351</a></li> |
| 39 | <li><a href="#refimpl">Reference Implementatation</a></li> |
| 40 | </ol></li> |
| 41 | </ul> |
| 42 | |
| 43 | <div class="doc_author"> |
Misha Brukman | 1721784 | 2004-07-21 18:04:27 +0000 | [diff] [blame] | 44 | <p>Written by <a href="rspencer@x10sys.com">Reid Spencer</a></p> |
Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 45 | </div> |
| 46 | |
| 47 | |
| 48 | <!-- *********************************************************************** --> |
| 49 | <div class="doc_section"><a name="abstract">Abstract</a></div> |
| 50 | <div class="doc_text"> |
| 51 | <p>This document describes the requirements, design, and implementation |
| 52 | details of LLVM's System Library. The library is composed of the header files |
| 53 | in <tt>llvm/include/llvm/System</tt> and the source files in |
| 54 | <tt>llvm/lib/System</tt>. The goal of this library is to completely shield |
| 55 | LLVM from the variations in operating system interfaces. By centralizing |
| 56 | LLVM's use of operating system interfaces, we make it possible for the LLVM |
| 57 | tool chain and runtime libraries to be more easily ported to new platforms. |
| 58 | The library also unclutters the rest of LLVM from #ifdef use and special |
| 59 | cases for specific operating systems.</p> |
| 60 | <p>The System Library was donated to LLVM by Reid Spencer who formulated the |
| 61 | original design as part of the eXtensible Programming System (XPS) which is |
| 62 | based, in part, on LLVM.</p> |
| 63 | </div> |
| 64 | |
| 65 | <!-- *********************************************************************** --> |
| 66 | <div class="doc_section"> |
| 67 | <a name="requirements">System Library Requirements</a> |
| 68 | </div> |
| 69 | <div class="doc_text"> |
| 70 | <p>The System library's requirements are aimed at shielding LLVM from the |
| 71 | variations in operating system interfaces. The following sections define the |
| 72 | requirements needed to fulfill this objective.</p> |
| 73 | </div> |
| 74 | |
| 75 | <!-- ======================================================================= --> |
| 76 | <div class="doc_subsection"><a name="headers">Hide System Header Files</a></div> |
| 77 | <div class="doc_text"> |
| 78 | <p>To be written.</p> |
| 79 | </div> |
| 80 | |
| 81 | <!-- ======================================================================= --> |
| 82 | <div class="doc_subsection"><a name="nofunc">No Exposed Functions</a></div> |
| 83 | <div class="doc_text"> |
| 84 | <p>To be written.</p> |
| 85 | </div> |
| 86 | |
| 87 | <!-- ======================================================================= --> |
| 88 | <div class="doc_subsection"><a name="nodata">No Exposed Data</a></div> |
| 89 | <div class="doc_text"> |
| 90 | <p>To be written.</p> |
| 91 | </div> |
| 92 | |
| 93 | <!-- ======================================================================= --> |
| 94 | <div class="doc_subsection"><a name="xcptns">No Exceptions</a></div> |
| 95 | <div class="doc_text"> |
| 96 | <p>To be written.</p> |
| 97 | </div> |
| 98 | |
| 99 | <!-- ======================================================================= --> |
| 100 | <div class="doc_subsection"><a name="errors">Standard Error Codes</a></div> |
| 101 | <div class="doc_text"> |
| 102 | <p>To be written.</p> |
| 103 | </div> |
| 104 | |
| 105 | <!-- ======================================================================= --> |
| 106 | <div class="doc_subsection"><a name="overhead">Minimize Overhead</a></div> |
| 107 | <div class="doc_text"> |
| 108 | <p>To be written.</p> |
| 109 | </div> |
| 110 | |
| 111 | <!-- *********************************************************************** --> |
| 112 | <div class="doc_section"><a name="design">System Library Design</a></div> |
| 113 | <div class="doc_text"> |
| 114 | <p>In order to fulfill the requirements of the system library, strict design |
| 115 | objectives must be maintained in the library as it evolves. The goal here |
| 116 | is to provide interfaces to operating system concepts (files, memory maps, |
| 117 | sockets, signals, locking, etc) efficiently and in such a way that the |
| 118 | remainder of LLVM is completely operating system agnostic.</p> |
| 119 | </div> |
| 120 | |
| 121 | <!-- ======================================================================= --> |
| 122 | <div class="doc_subsection"><a name="opaque">Use Opaque Classes</a></div> |
| 123 | <div class="doc_text"> |
| 124 | <p>no public data</p> |
| 125 | <p>onlyprimitive typed private/protected data</p> |
| 126 | <p>data size is "right" for platform, not max of all platforms</p> |
| 127 | <p>each class corresponds to O/S concept</p> |
| 128 | </div> |
| 129 | |
| 130 | <!-- ======================================================================= --> |
| 131 | <div class="doc_subsection"><a name="common">Common Implementations</a></div> |
| 132 | <div class="doc_text"> |
| 133 | <p>To be written.</p> |
| 134 | </div> |
| 135 | |
| 136 | <!-- ======================================================================= --> |
| 137 | <div class="doc_subsection"> |
| 138 | <a name="multi_imps">Multiple Implementations</a> |
| 139 | </div> |
| 140 | <div class="doc_text"> |
| 141 | <p>To be written.</p> |
| 142 | </div> |
| 143 | |
| 144 | <!-- ======================================================================= --> |
| 145 | <div class="doc_subsection"> |
| 146 | <a name="low_level">Use Low Level Interfaces</a> |
| 147 | </div> |
| 148 | <div class="doc_text"> |
| 149 | <p>To be written.</p> |
| 150 | </div> |
| 151 | |
| 152 | <!-- ======================================================================= --> |
| 153 | <div class="doc_subsection"><a name="memalloc">No Memory Allocation</a></div> |
| 154 | <div class="doc_text"> |
| 155 | <p>To be written.</p> |
| 156 | </div> |
| 157 | |
| 158 | <!-- ======================================================================= --> |
| 159 | <div class="doc_subsection"><a name="virtuals">No Virtual Methods</a></div> |
| 160 | <div class="doc_text"> |
| 161 | <p>To be written.</p> |
| 162 | </div> |
| 163 | |
| 164 | <!-- *********************************************************************** --> |
| 165 | <div class="doc_section"><a name="detail">System Library Details</a></div> |
| 166 | <div class="doc_text"> |
| 167 | <p>To be written.</p> |
| 168 | </div> |
| 169 | |
| 170 | <!-- ======================================================================= --> |
Chris Lattner | 05540a7 | 2004-07-17 18:50:19 +0000 | [diff] [blame] | 171 | <div class="doc_subsection"><a name="bug">Bug 351</a></div> |
Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 172 | <div class="doc_text"> |
Chris Lattner | 05540a7 | 2004-07-17 18:50:19 +0000 | [diff] [blame] | 173 | <p>See <a href="http://llvm.cs.uiuc.edu/PR351">bug 351</a> |
Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 174 | for further details on the progress of this work</p> |
| 175 | </div> |
| 176 | |
| 177 | <!-- ======================================================================= --> |
Reid Spencer | 9518920 | 2004-08-26 18:52:52 +0000 | [diff] [blame^] | 178 | <div class="doc_subsection"><a name="bug">Rationale For #include Hierarchy</a> |
| 179 | </div> |
| 180 | <div class="doc_text"> |
| 181 | <p>In order to provide different implementations of the lib/System interface |
| 182 | for different platforms, it is necessary for the library to "sense" which |
| 183 | operating system is being compiled for and conditionally compile only the |
| 184 | applicabe parts of the library. While several operating system wrapper |
| 185 | libraries (e.g. APR, ACE) choose to use #ifdef preprocessor statements in |
| 186 | combination with autoconf variable (HAVE_* family), lib/System chooses an |
| 187 | alternate strategy. <p> |
| 188 | <p>To put it succinctly, the lib/System strategy has traded "#ifdef hell" for |
| 189 | "#include hell". That is, a given implementation file defines one or more |
| 190 | functions for a particular operating system variant. The functions defined in |
| 191 | that file have no #ifdef's to disambiguate the platform since the file is only |
| 192 | compiled on one kind of platform. While this leads to the same function being |
| 193 | imlemented differently in different files, it is our contention that this |
| 194 | leads to better maintenance and easier portability.</p> |
| 195 | <p>For example, consider a function having different implementations on a |
| 196 | variety of platforms. Many wrapper libraries choose to deal with the different |
| 197 | implementations by using #ifdef, like this:</p> |
| 198 | <pre><tt> |
| 199 | void SomeFunction(void) { |
| 200 | #if defined __LINUX |
| 201 | // .. Linux implementation |
| 202 | #elif defined __WIN32 |
| 203 | // .. Win32 implementation |
| 204 | #elif defined __SunOS |
| 205 | // .. SunOS implementation |
| 206 | #else |
| 207 | #warning "Don't know how to implement SomeFunction on this platform" |
| 208 | #endif |
| 209 | } |
| 210 | </tt></pre> |
| 211 | <p>The problem with this is that its very messy to read, especially as the |
| 212 | number of operating systems and their variants grow. The above example is |
| 213 | actually tame compared to what can happen when the implementation depends on |
| 214 | specific flavors and versions of the operating system. In that case you end up |
| 215 | with multiple levels of nested #if statements. This is what we mean by "#ifdef |
| 216 | hell".</p> |
| 217 | <p>To avoid the situation above, we've choosen to locate all functions for a |
| 218 | given implementation file for a specific operating system into one place. This |
| 219 | has the following advantages:<p> |
| 220 | <ul> |
| 221 | <li>No "#ifdef hell"</li> |
| 222 | <li>When porting, the strategy is quite straight forward: copy the |
| 223 | implementation file from a similar operating system to a new directory and |
| 224 | re-implement them.<li> |
| 225 | <li>Correctness is helped during porting because the new operating system's |
| 226 | implementation is wholly contained in a separate directory. There's no |
| 227 | chance to make an error in the #if statements and affect some other |
| 228 | operating system's implementation.</li> |
| 229 | </ul> |
| 230 | <p>So, given that we have decided to use #include instead of #if to provide |
| 231 | platform specific implementations, there are actually three ways we can go |
| 232 | about doing this. None of them are perfect, but we believe we've chosen the |
| 233 | lesser of the three evils. Given that there is a variable named $OS which |
| 234 | names the platform for which we must build, here's a summary of the three |
| 235 | approaches we could use to determine the correct directory:</p> |
| 236 | <ol> |
| 237 | <li>Provide the compiler with a -I$(OS) on the command line. This could be |
| 238 | provided in only the lib/System makefile.</li> |
| 239 | <li>Use autoconf to transform #include statements in the implementation |
| 240 | files by using substitutions of @OS@. For example, if we had a file, |
| 241 | File.cpp.in, that contained "#include <@OS@/File.cpp>" this would get |
| 242 | transformed to "#include <actual/File.cpp>" where "actual" is the |
| 243 | actual name of the operating system</li> |
| 244 | <li>Create a link from $OBJ_DIR/platform to $SRC_DIR/$OS. This allows us to |
| 245 | use a generic directory name to get the correct platform, as in #include |
| 246 | <platform/File.cpp></li> |
| 247 | </ol> |
| 248 | <p>Let's look at the pitfalls of each approach.</p> |
| 249 | <p>In approach #1, we end up with some confusion as to what gets included. |
| 250 | Suppose we have lib/System/File.cpp that includes just File.cpp to get the |
| 251 | platform specific part of the implementation. In this case, the include |
| 252 | directive with the <> syntax will include the right file but the include |
| 253 | directive with the "" syntax will recursively include the same file, |
| 254 | lib/System/File.cpp. In the case of #include <File.cpp>, the -I options |
| 255 | to the compiler are searched first so it works. But in the #include "File.cpp" |
| 256 | case, the current directory is searched first. Furthermore, in both cases, |
| 257 | neither include directive documents which File.cpp is getting included.</p> |
| 258 | <p>In approach #2, we have the problem of needing to reconfigure repeatedly. |
| 259 | Developer's generally hate that and we don't want lib/System to be a thorn in |
| 260 | everyone's side because it will constantly need updating as operating systems |
| 261 | change and as new operating systems are added. The problem occurs when a new |
| 262 | implementation file is added to the library. First of all, you have to add a |
| 263 | file with the .in suffix, then you have to add that file name to the list of |
| 264 | configurable files in the autoconf/configure.ac file, then you have to run |
| 265 | AutoRegen.sh to rebuild the configure script, then you have to run the |
| 266 | configure script. This is deemed to be a pretty large hassle.</p> |
| 267 | <p>In approach #3, we have the problem that not all platforms support links. |
| 268 | Fortunately the autoconf macro used to create the link can compensate for |
| 269 | this. If a link can't be made, the configure script will copy the correct |
| 270 | directory from $BUILD_SRC_DIR to $BUILD_OBJ_DIR under the new name. The only |
| 271 | problem with this is that if a copy is made, the copy doesn't get updated if |
| 272 | the programmer adds or modifies files in the $BUILD_SRC_DIR. A reconfigure or |
| 273 | manual copying is needed to get things to compile.<p> |
| 274 | <p>The approach we have taken in lib/System is #3. Here's why:<p> |
| 275 | <ul> |
| 276 | <li>Approach #1 is rejected because it doesn't document what's actually |
| 277 | getting included and the potential for mistakes with alternate include |
| 278 | directive forms is high.</li> |
| 279 | <li>Approach #2 are both viable and only really impact development when new |
| 280 | files are added to the library.</li> |
| 281 | <li>However, approach #2 impacts every new file on every platform all the |
| 282 | time. With approach #3, only those platforms not supporting links will be |
| 283 | affected. The number of platforms not supporting links is very small and |
| 284 | they are generally archaic.</li> |
| 285 | <li>Given the above, approach #3 seems to have the least impact.</li> |
| 286 | </ul> |
| 287 | </div> |
| 288 | |
| 289 | <!-- ======================================================================= --> |
Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 290 | <div class="doc_subsection"> |
| 291 | <a name="refimpl">Reference Implementation</a> |
| 292 | </div> |
| 293 | <div class="doc_text"> |
| 294 | <p>The <tt>linux</tt> implementation of the system library will always be the |
| 295 | reference implementation. This means that (a) the concepts defined by the |
| 296 | linux must be identically replicated in the other implementations and (b) the |
| 297 | linux implementation must always be complete (provide implementations for all |
| 298 | concepts).</p> |
| 299 | </div> |
| 300 | |
| 301 | <!-- *********************************************************************** --> |
| 302 | |
| 303 | <hr> |
| 304 | <address> |
| 305 | <a href="http://jigsaw.w3.org/css-validator/check/referer"><img |
| 306 | src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a> |
| 307 | <a href="http://validator.w3.org/check/referer"><img |
| 308 | src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a> |
| 309 | |
| 310 | <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br> |
| 311 | <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br> |
| 312 | Last modified: $Date$ |
| 313 | </address> |
| 314 | </body> |
| 315 | </html> |