Subversion Repositories Kolibri OS

Rev

Blame | Last modification | View Log | RSS feed

  1. <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
  2. <html lang="en">
  3. <head>
  4.   <meta http-equiv="content-type" content="text/html; charset=utf-8">
  5.   <title>llvmpipe</title>
  6.   <link rel="stylesheet" type="text/css" href="mesa.css">
  7. </head>
  8. <body>
  9.  
  10. <div class="header">
  11.   <h1>The Mesa 3D Graphics Library</h1>
  12. </div>
  13.  
  14. <iframe src="contents.html"></iframe>
  15. <div class="content">
  16.  
  17. <h1>Introduction</h1>
  18.  
  19. <p>
  20. The Gallium llvmpipe driver is a software rasterizer that uses LLVM to
  21. do runtime code generation.
  22. Shaders, point/line/triangle rasterization and vertex processing are
  23. implemented with LLVM IR which is translated to x86 or x86-64 machine
  24. code.
  25. Also, the driver is multithreaded to take advantage of multiple CPU cores
  26. (up to 8 at this time).
  27. It's the fastest software rasterizer for Mesa.
  28. </p>
  29.  
  30.  
  31. <h1>Requirements</h1>
  32.  
  33. <ul>
  34. <li>
  35.   <p>An x86 or amd64 processor; 64-bit mode recommended.</p>
  36.   <p>
  37.   Support for SSE2 is strongly encouraged.  Support for SSSE3 and SSE4.1 will
  38.   yield the most efficient code.  The fewer features the CPU has the more
  39.   likely is that you run into underperforming, buggy, or incomplete code.
  40.   </p>
  41.   <p>
  42.   See /proc/cpuinfo to know what your CPU supports.
  43.   </p>
  44. </li>
  45. <li>
  46.   <p>LLVM: version 3.4 recommended; 3.3 or later required.</p>
  47.   <p>
  48.   For Linux, on a recent Debian based distribution do:
  49.   </p>
  50. <pre>
  51.     aptitude install llvm-dev
  52. </pre>
  53.   <p>
  54.   For a RPM-based distribution do:
  55.   </p>
  56. <pre>
  57.     yum install llvm-devel
  58. </pre>
  59.  
  60.   <p>
  61.   For Windows you will need to build LLVM from source with MSVC or MINGW
  62.   (either natively or through cross compilers) and CMake, and set the LLVM
  63.   environment variable to the directory you installed it to.
  64.  
  65.   LLVM will be statically linked, so when building on MSVC it needs to be
  66.   built with a matching CRT as Mesa, and you'll need to pass
  67.    <code>-DLLVM_USE_CRT_xxx=yyy</code> as described below.
  68.    </p>
  69.  
  70.    <table border="1">
  71.      <tr>
  72.        <th rowspan="2">LLVM build-type</th>
  73.        <th colspan="2" align="center">Mesa build-type</th>
  74.      </tr>
  75.      <tr>
  76.        <th>debug,checked</th>
  77.        <th>release,profile</th>
  78.      </tr>
  79.      <tr>
  80.        <th>Debug</th>
  81.        <td><code>-DLLVM_USE_CRT_DEBUG=MTd</code></td>
  82.        <td><code>-DLLVM_USE_CRT_DEBUG=MT</code></td>
  83.      </tr>
  84.      <tr>
  85.        <th>Release</th>
  86.        <td><code>-DLLVM_USE_CRT_RELEASE=MTd</code></td>
  87.        <td><code>-DLLVM_USE_CRT_RELEASE=MT</code></td>
  88.      </tr>
  89.    </table>
  90.  
  91.    <p>
  92.    You can build only the x86 target by passing -DLLVM_TARGETS_TO_BUILD=X86
  93.    to cmake.
  94.    </p>
  95. </li>
  96.  
  97. <li>
  98.    <p>scons (optional)</p>
  99. </li>
  100. </ul>
  101.  
  102.  
  103. <h1>Building</h1>
  104.  
  105. To build everything on Linux invoke scons as:
  106.  
  107. <pre>
  108.   scons build=debug libgl-xlib
  109. </pre>
  110.  
  111. Alternatively, you can build it with GNU make, if you prefer, by invoking it as
  112.  
  113. <pre>
  114.   make linux-llvm
  115. </pre>
  116.  
  117. but the rest of these instructions assume that scons is used.
  118.  
  119. For Windows the procedure is similar except the target:
  120.  
  121. <pre>
  122.   scons platform=windows build=debug libgl-gdi
  123. </pre>
  124.  
  125.  
  126. <h1>Using</h1>
  127.  
  128. <h2>Linux</h2>
  129.  
  130. <p>On Linux, building will create a drop-in alternative for libGL.so into</p>
  131.  
  132. <pre>
  133.   build/foo/gallium/targets/libgl-xlib/libGL.so
  134. </pre>
  135. or
  136. <pre>
  137.   lib/gallium/libGL.so
  138. </pre>
  139.  
  140. <p>To use it set the LD_LIBRARY_PATH environment variable accordingly.</p>
  141.  
  142. <p>For performance evaluation pass build=release to scons, and use the corresponding
  143. lib directory without the "-debug" suffix.</p>
  144.  
  145.  
  146. <h2>Windows</h2>
  147.  
  148. <p>
  149. On Windows, building will create
  150. <code>build/windows-x86-debug/gallium/targets/libgl-gdi/opengl32.dll</code>
  151. which is a drop-in alternative for system's <code>opengl32.dll</code>.  To use
  152. it put it in the same directory as your application.  It can also be used by
  153. replacing the native ICD driver, but it's quite an advanced usage, so if you
  154. need to ask, don't even try it.
  155. </p>
  156.  
  157. <p>
  158. There is however an easy way to replace the OpenGL software renderer that comes
  159. with Microsoft Windows 7 (or later) with llvmpipe (that is, on systems without
  160. any OpenGL drivers):
  161. </p>
  162.  
  163. <ul>
  164.  <li><p>copy build/windows-x86-debug/gallium/targets/libgl-gdi/opengl32.dll to C:\Windows\SysWOW64\mesadrv.dll</p></li>
  165.  <li><p>load this registry settings:</p>
  166.  <pre>REGEDIT4
  167.  
  168. ; http://technet.microsoft.com/en-us/library/cc749368.aspx
  169. ; http://www.msfn.org/board/topic/143241-portable-windows-7-build-from-winpe-30/page-5#entry942596
  170. [HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\Microsoft\Windows NT\CurrentVersion\OpenGLDrivers\MSOGL]
  171. "DLL"="mesadrv.dll"
  172. "DriverVersion"=dword:00000001
  173. "Flags"=dword:00000001
  174. "Version"=dword:00000002
  175. </pre>
  176.  </li>
  177.  <li>Ditto for 64 bits drivers if you need them.</li>
  178. </ul>
  179.  
  180.  
  181. <h1>Profiling</h1>
  182.  
  183. <p>
  184. To profile llvmpipe you should build as
  185. </p>
  186. <pre>
  187.  scons build=profile &lt;same-as-before&gt;
  188. </pre>
  189.  
  190. <p>
  191. This will ensure that frame pointers are used both in C and JIT functions, and
  192. that no tail call optimizations are done by gcc.
  193. </p>
  194.  
  195. <h2>Linux perf integration</h2>
  196.  
  197. <p>
  198. On Linux, it is possible to have symbol resolution of JIT code with <a href="http://perf.wiki.kernel.org/">Linux perf</a>:
  199. </p>
  200.  
  201. <pre>
  202.         perf record -g /my/application
  203.         perf report
  204. </pre>
  205.  
  206. <p>
  207. When run inside Linux perf, llvmpipe will create a /tmp/perf-XXXXX.map file with
  208. symbol address table.  It also dumps assembly code to /tmp/perf-XXXXX.map.asm,
  209. which can be used by the bin/perf-annotate-jit script to produce disassembly of
  210. the generated code annotated with the samples.
  211. </p>
  212.  
  213. <p>You can obtain a call graph via
  214. <a href="http://code.google.com/p/jrfonseca/wiki/Gprof2Dot#linux_perf">Gprof2Dot</a>.</p>
  215.  
  216.  
  217. <h1>Unit testing</h1>
  218.  
  219. <p>
  220. Building will also create several unit tests in
  221. build/linux-???-debug/gallium/drivers/llvmpipe:
  222. </p>
  223.  
  224. <ul>
  225. <li> lp_test_blend: blending
  226. <li> lp_test_conv: SIMD vector conversion
  227. <li> lp_test_format: pixel unpacking/packing
  228. </ul>
  229.  
  230. <p>
  231. Some of this tests can output results and benchmarks to a tab-separated-file
  232. for posterior analysis, e.g.:
  233. </p>
  234. <pre>
  235.  build/linux-x86_64-debug/gallium/drivers/llvmpipe/lp_test_blend -o blend.tsv
  236. </pre>
  237.  
  238.  
  239. <h1>Development Notes</h1>
  240.  
  241. <ul>
  242. <li>
  243.  When looking to this code by the first time start in lp_state_fs.c, and
  244.  then skim through the lp_bld_* functions called in there, and the comments
  245.  at the top of the lp_bld_*.c functions.
  246. </li>
  247. <li>
  248.  The driver-independent parts of the LLVM / Gallium code are found in
  249.  src/gallium/auxiliary/gallivm/.  The filenames and function prefixes
  250.  need to be renamed from "lp_bld_" to something else though.
  251. </li>
  252. <li>
  253.  We use LLVM-C bindings for now. They are not documented, but follow the C++
  254.  interfaces very closely, and appear to be complete enough for code
  255.  generation. See
  256.  <a href="http://npcontemplation.blogspot.com/2008/06/secret-of-llvm-c-bindings.html">
  257.  this stand-alone example</a>.  See the llvm-c/Core.h file for reference.
  258. </li>
  259. </ul>
  260.  
  261. <h1 id="recommended_reading">Recommended Reading</h1>
  262.  
  263. <ul>
  264.  <li>
  265.    <p>Rasterization</p>
  266.    <ul>
  267.      <li><a href="http://www.cs.unc.edu/~olano/papers/2dh-tri/">Triangle Scan Conversion using 2D Homogeneous Coordinates</a></li>
  268.      <li><a href="http://www.drdobbs.com/parallel/rasterization-on-larrabee/217200602">Rasterization on Larrabee</a> (<a href="http://devmaster.net/posts/2887/rasterization-on-larrabee">DevMaster copy</a>)</li>
  269.      <li><a href="http://devmaster.net/posts/6133/rasterization-using-half-space-functions">Rasterization using half-space functions</a></li>
  270.      <li><a href="http://devmaster.net/posts/6145/advanced-rasterization">Advanced Rasterization</a></li>
  271.      <li><a href="http://fgiesen.wordpress.com/2013/02/17/optimizing-sw-occlusion-culling-index/">Optimizing Software Occlusion Culling</a></li>
  272.    </ul>
  273.  </li>
  274.  <li>
  275.    <p>Texture sampling</p>
  276.    <ul>
  277.      <li><a href="http://chrishecker.com/Miscellaneous_Technical_Articles#Perspective_Texture_Mapping">Perspective Texture Mapping</a></li>
  278.      <li><a href="http://www.flipcode.com/archives/Texturing_As_In_Unreal.shtml">Texturing As In Unreal</a></li>
  279.      <li><a href="http://www.gamasutra.com/view/feature/3301/runtime_mipmap_filtering.php">Run-Time MIP-Map Filtering</a></li>
  280.      <li><a href="http://alt.3dcenter.org/artikel/2003/10-26_a_english.php">Will "brilinear" filtering persist?</a></li>
  281.      <li><a href="http://ixbtlabs.com/articles2/gffx/nv40-rx800-3.html">Trilinear filtering</a></li>
  282.      <li><a href="http://devmaster.net/posts/12785/texture-swizzling">Texture Swizzling</a></li>
  283.    </ul>
  284.  </li>
  285.  <li>
  286.    <p>SIMD</p>
  287.    <ul>
  288.      <li><a href="http://www.cdl.uni-saarland.de/projects/wfv/#header4">Whole-Function Vectorization</a></li>
  289.    </ul>
  290.  </li>
  291.  <li>
  292.    <p>Optimization</p>
  293.    <ul>
  294.      <li><a href="http://www.drdobbs.com/optimizing-pixomatic-for-modern-x86-proc/184405807">Optimizing Pixomatic For Modern x86 Processors</a></li>
  295.      <li><a href="http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html">Intel 64 and IA-32 Architectures Optimization Reference Manual</a></li>
  296.      <li><a href="http://www.agner.org/optimize/">Software optimization resources</a></li>
  297.      <li><a href="http://software.intel.com/en-us/articles/intel-intrinsics-guide">Intel Intrinsics Guide</a><li>
  298.    </ul>
  299.  </li>
  300.  <li>
  301.    <p>LLVM</p>
  302.    <ul>
  303.      <li><a href="http://llvm.org/docs/LangRef.html">LLVM Language Reference Manual</a></li>
  304.      <li><a href="http://npcontemplation.blogspot.co.uk/2008/06/secret-of-llvm-c-bindings.html">The secret of LLVM C bindings</a></li>
  305.    </ul>
  306.  </li>
  307.  <li>
  308.    <p>General</p>
  309.    <ul>
  310.      <li><a href="http://fgiesen.wordpress.com/2011/07/09/a-trip-through-the-graphics-pipeline-2011-index/">A trip through the Graphics Pipeline</a></li>
  311.      <li><a href="http://msdn.microsoft.com/en-us/library/gg615082.aspx#architecture">WARP Architecture and Performance</a></li>
  312.    </ul>
  313.  </li>
  314. </ul>
  315.  
  316. </div>
  317. </body>
  318. </html>
  319.