ext_ffi.html 9.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322
  1. <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
  2. <html>
  3. <head>
  4. <title>FFI Library</title>
  5. <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
  6. <meta name="Copyright" content="Copyright (C) 2005-2023">
  7. <meta name="Language" content="en">
  8. <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
  9. <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
  10. </head>
  11. <body>
  12. <div id="site">
  13. <a href="https://luajit.org"><span>Lua<span id="logo">JIT</span></span></a>
  14. </div>
  15. <div id="head">
  16. <h1>FFI Library</h1>
  17. </div>
  18. <div id="nav">
  19. <ul><li>
  20. <a href="luajit.html">LuaJIT</a>
  21. <ul><li>
  22. <a href="https://luajit.org/download.html">Download <span class="ext">&raquo;</span></a>
  23. </li><li>
  24. <a href="install.html">Installation</a>
  25. </li><li>
  26. <a href="running.html">Running</a>
  27. </li></ul>
  28. </li><li>
  29. <a href="extensions.html">Extensions</a>
  30. <ul><li>
  31. <a class="current" href="ext_ffi.html">FFI Library</a>
  32. <ul><li>
  33. <a href="ext_ffi_tutorial.html">FFI Tutorial</a>
  34. </li><li>
  35. <a href="ext_ffi_api.html">ffi.* API</a>
  36. </li><li>
  37. <a href="ext_ffi_semantics.html">FFI Semantics</a>
  38. </li></ul>
  39. </li><li>
  40. <a href="ext_jit.html">jit.* Library</a>
  41. </li><li>
  42. <a href="ext_c_api.html">Lua/C API</a>
  43. </li></ul>
  44. </li><li>
  45. <a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
  46. </li><li>
  47. <a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
  48. </li><li>
  49. <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
  50. </li></ul>
  51. </div>
  52. <div id="main">
  53. <p>
  54. The FFI library allows <b>calling external C&nbsp;functions</b> and
  55. <b>using C&nbsp;data structures</b> from pure Lua code.
  56. </p>
  57. <p>
  58. The FFI library largely obviates the need to write tedious manual
  59. Lua/C bindings in C. No need to learn a separate binding language
  60. &mdash; <b>it parses plain C&nbsp;declarations!</b> These can be
  61. cut-n-pasted from C&nbsp;header files or reference manuals. It's up to
  62. the task of binding large libraries without the need for dealing with
  63. fragile binding generators.
  64. </p>
  65. <p>
  66. The FFI library is tightly integrated into LuaJIT (it's not available
  67. as a separate module). The code generated by the JIT-compiler for
  68. accesses to C&nbsp;data structures from Lua code is on par with the
  69. code a C&nbsp;compiler would generate. Calls to C&nbsp;functions can
  70. be inlined in JIT-compiled code, unlike calls to functions bound via
  71. the classic Lua/C API.
  72. </p>
  73. <p>
  74. This page gives a short introduction to the usage of the FFI library.
  75. <em>Please use the FFI sub-topics in the navigation bar to learn more.</em>
  76. </p>
  77. <h2 id="call">Motivating Example: Calling External C Functions</h2>
  78. <p>
  79. It's really easy to call an external C&nbsp;library function:
  80. </p>
  81. <pre class="code mark">
  82. <span class="codemark">&#9312;
  83. &#9313;
  84. &#9314;</span>local ffi = require("ffi")
  85. ffi.cdef[[
  86. <span style="color:#00a000;">int printf(const char *fmt, ...);</span>
  87. ]]
  88. ffi.C.printf("Hello %s!", "world")
  89. </pre>
  90. <p>
  91. So, let's pick that apart:
  92. </p>
  93. <p>
  94. <span class="mark">&#9312;</span> Load the FFI library.
  95. </p>
  96. <p>
  97. <span class="mark">&#9313;</span> Add a C&nbsp;declaration
  98. for the function. The part inside the double-brackets (in green) is
  99. just standard C&nbsp;syntax.
  100. </p>
  101. <p>
  102. <span class="mark">&#9314;</span> Call the named
  103. C&nbsp;function &mdash; Yes, it's that simple!
  104. </p>
  105. <p style="font-size: 8pt;">
  106. Actually, what goes on behind the scenes is far from simple: <span
  107. style="color:#4040c0;">&#9314;</span> makes use of the standard
  108. C&nbsp;library namespace <tt>ffi.C</tt>. Indexing this namespace with
  109. a symbol name (<tt>"printf"</tt>) automatically binds it to the
  110. standard C&nbsp;library. The result is a special kind of object which,
  111. when called, runs the <tt>printf</tt> function. The arguments passed
  112. to this function are automatically converted from Lua objects to the
  113. corresponding C&nbsp;types.
  114. </p>
  115. <p>
  116. Ok, so maybe the use of <tt>printf()</tt> wasn't such a spectacular
  117. example. You could have done that with <tt>io.write()</tt> and
  118. <tt>string.format()</tt>, too. But you get the idea ...
  119. </p>
  120. <p>
  121. So here's something to pop up a message box on Windows:
  122. </p>
  123. <pre class="code">
  124. local ffi = require("ffi")
  125. ffi.cdef[[
  126. <span style="color:#00a000;">int MessageBoxA(void *w, const char *txt, const char *cap, int type);</span>
  127. ]]
  128. ffi.C.MessageBoxA(nil, "Hello world!", "Test", 0)
  129. </pre>
  130. <p>
  131. Bing! Again, that was far too easy, no?
  132. </p>
  133. <p style="font-size: 8pt;">
  134. Compare this with the effort required to bind that function using the
  135. classic Lua/C API: create an extra C&nbsp;file, add a C&nbsp;function
  136. that retrieves and checks the argument types passed from Lua and calls
  137. the actual C&nbsp;function, add a list of module functions and their
  138. names, add a <tt>luaopen_*</tt> function and register all module
  139. functions, compile and link it into a shared library (DLL), move it to
  140. the proper path, add Lua code that loads the module aaaand ... finally
  141. call the binding function. Phew!
  142. </p>
  143. <h2 id="cdata">Motivating Example: Using C Data Structures</h2>
  144. <p>
  145. The FFI library allows you to create and access C&nbsp;data
  146. structures. Of course, the main use for this is for interfacing with
  147. C&nbsp;functions. But they can be used stand-alone, too.
  148. </p>
  149. <p>
  150. Lua is built upon high-level data types. They are flexible, extensible
  151. and dynamic. That's why we all love Lua so much. Alas, this can be
  152. inefficient for certain tasks, where you'd really want a low-level
  153. data type. E.g. a large array of a fixed structure needs to be
  154. implemented with a big table holding lots of tiny tables. This imposes
  155. both a substantial memory overhead as well as a performance overhead.
  156. </p>
  157. <p>
  158. Here's a sketch of a library that operates on color images, plus a
  159. simple benchmark. First, the plain Lua version:
  160. </p>
  161. <pre class="code">
  162. local floor = math.floor
  163. local function image_ramp_green(n)
  164. local img = {}
  165. local f = 255/(n-1)
  166. for i=1,n do
  167. img[i] = { red = 0, green = floor((i-1)*f), blue = 0, alpha = 255 }
  168. end
  169. return img
  170. end
  171. local function image_to_gray(img, n)
  172. for i=1,n do
  173. local y = floor(0.3*img[i].red + 0.59*img[i].green + 0.11*img[i].blue)
  174. img[i].red = y; img[i].green = y; img[i].blue = y
  175. end
  176. end
  177. local N = 400*400
  178. local img = image_ramp_green(N)
  179. for i=1,1000 do
  180. image_to_gray(img, N)
  181. end
  182. </pre>
  183. <p>
  184. This creates a table with 160.000 pixels, each of which is a table
  185. holding four number values in the range of 0-255. First, an image with
  186. a green ramp is created (1D for simplicity), then the image is
  187. converted to grayscale 1000 times. Yes, that's silly, but I was in
  188. need of a simple example ...
  189. </p>
  190. <p>
  191. And here's the FFI version. The modified parts have been marked in
  192. bold:
  193. </p>
  194. <pre class="code mark">
  195. <span class="codemark">&#9312;
  196. &#9313;
  197. &#9314;
  198. &#9315;
  199. &#9314;
  200. &#9316;</span><b>local ffi = require("ffi")
  201. ffi.cdef[[
  202. </b><span style="color:#00a000;">typedef struct { uint8_t red, green, blue, alpha; } rgba_pixel;</span><b>
  203. ]]</b>
  204. local function image_ramp_green(n)
  205. <b>local img = ffi.new("rgba_pixel[?]", n)</b>
  206. local f = 255/(n-1)
  207. for i=<b>0,n-1</b> do
  208. <b>img[i].green = i*f</b>
  209. <b>img[i].alpha = 255</b>
  210. end
  211. return img
  212. end
  213. local function image_to_grey(img, n)
  214. for i=<b>0,n-1</b> do
  215. local y = <b>0.3*img[i].red + 0.59*img[i].green + 0.11*img[i].blue</b>
  216. img[i].red = y; img[i].green = y; img[i].blue = y
  217. end
  218. end
  219. local N = 400*400
  220. local img = image_ramp_green(N)
  221. for i=1,1000 do
  222. image_to_grey(img, N)
  223. end
  224. </pre>
  225. <p>
  226. Ok, so that wasn't too difficult:
  227. </p>
  228. <p>
  229. <span class="mark">&#9312;</span> First, load the FFI
  230. library and declare the low-level data type. Here we choose a
  231. <tt>struct</tt> which holds four byte fields, one for each component
  232. of a 4x8&nbsp;bit RGBA pixel.
  233. </p>
  234. <p>
  235. <span class="mark">&#9313;</span> Creating the data
  236. structure with <tt>ffi.new()</tt> is straightforward &mdash; the
  237. <tt>'?'</tt> is a placeholder for the number of elements of a
  238. variable-length array.
  239. </p>
  240. <p>
  241. <span class="mark">&#9314;</span> C&nbsp;arrays are
  242. zero-based, so the indexes have to run from <tt>0</tt> to
  243. <tt>n-1</tt>. One might want to allocate one more element instead to
  244. simplify converting legacy code.
  245. </p>
  246. <p>
  247. <span class="mark">&#9315;</span> Since <tt>ffi.new()</tt>
  248. zero-fills the array by default, we only need to set the green and the
  249. alpha fields.
  250. </p>
  251. <p>
  252. <span class="mark">&#9316;</span> The calls to
  253. <tt>math.floor()</tt> can be omitted here, because floating-point
  254. numbers are already truncated towards zero when converting them to an
  255. integer. This happens implicitly when the number is stored in the
  256. fields of each pixel.
  257. </p>
  258. <p>
  259. Now let's have a look at the impact of the changes: first, memory
  260. consumption for the image is down from 22&nbsp;Megabytes to
  261. 640&nbsp;Kilobytes (400*400*4 bytes). That's a factor of 35x less! So,
  262. yes, tables do have a noticeable overhead. BTW: The original program
  263. would consume 40&nbsp;Megabytes in plain Lua (on x64).
  264. </p>
  265. <p>
  266. Next, performance: the pure Lua version runs in 9.57 seconds (52.9
  267. seconds with the Lua interpreter) and the FFI version runs in 0.48
  268. seconds on my machine (YMMV). That's a factor of 20x faster (110x
  269. faster than the Lua interpreter).
  270. </p>
  271. <p style="font-size: 8pt;">
  272. The avid reader may notice that converting the pure Lua version over
  273. to use array indexes for the colors (<tt>[1]</tt> instead of
  274. <tt>.red</tt>, <tt>[2]</tt> instead of <tt>.green</tt> etc.) ought to
  275. be more compact and faster. This is certainly true (by a factor of
  276. ~1.7x). Switching to a struct-of-arrays would help, too.
  277. </p>
  278. <p style="font-size: 8pt;">
  279. However, the resulting code would be less idiomatic and rather
  280. error-prone. And it still doesn't get even close to the performance of
  281. the FFI version of the code. Also, high-level data structures cannot
  282. be easily passed to other C&nbsp;functions, especially I/O functions,
  283. without undue conversion penalties.
  284. </p>
  285. <br class="flush">
  286. </div>
  287. <div id="foot">
  288. <hr class="hide">
  289. Copyright &copy; 2005-2023
  290. <span class="noprint">
  291. &middot;
  292. <a href="contact.html">Contact</a>
  293. </span>
  294. </div>
  295. </body>
  296. </html>