changes.html 17 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419
  1. <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
  2. <html>
  3. <head>
  4. <title>LuaJIT Change History</title>
  5. <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
  6. <meta name="Author" content="Mike Pall">
  7. <meta name="Copyright" content="Copyright (C) 2005-2010, Mike Pall">
  8. <meta name="Language" content="en">
  9. <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
  10. <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
  11. <style type="text/css">
  12. div.major { max-width: 600px; padding: 1em; margin: 1em 0 1em 0; }
  13. </style>
  14. </head>
  15. <body>
  16. <div id="site">
  17. <a href="http://luajit.org"><span>Lua<span id="logo">JIT</span></span></a>
  18. </div>
  19. <div id="head">
  20. <h1>LuaJIT Change History</h1>
  21. </div>
  22. <div id="nav">
  23. <ul><li>
  24. <a href="luajit.html">LuaJIT</a>
  25. <ul><li>
  26. <a href="install.html">Installation</a>
  27. </li><li>
  28. <a href="running.html">Running</a>
  29. </li><li>
  30. <a href="api.html">API Extensions</a>
  31. </li></ul>
  32. </li><li>
  33. <a href="status.html">Status</a>
  34. <ul><li>
  35. <a class="current" href="changes.html">Changes</a>
  36. </li></ul>
  37. </li><li>
  38. <a href="faq.html">FAQ</a>
  39. </li><li>
  40. <a href="http://luajit.org/performance.html">Performance <span class="ext">&raquo;</span></a>
  41. </li><li>
  42. <a href="http://luajit.org/download.html">Download <span class="ext">&raquo;</span></a>
  43. </li></ul>
  44. </div>
  45. <div id="main">
  46. <p>
  47. This is a list of changes between the released versions of LuaJIT.<br>
  48. The current <span style="color: #c00000;">development version</span> is <strong>LuaJIT&nbsp;2.0.0-beta5</strong>.<br>
  49. The current <span style="color: #0000c0;">stable version</span> is <strong>LuaJIT&nbsp;1.1.6</strong>.
  50. </p>
  51. <p>
  52. Please check the
  53. <a href="http://luajit.org/changes.html"><span class="ext">&raquo;</span>&nbsp;Online Change History</a>
  54. to see whether newer versions are available.
  55. </p>
  56. <div class="major" style="background: #ffd0d0;">
  57. <h2 id="LuaJIT-2.0.0-beta5">LuaJIT 2.0.0-beta5 &mdash; 2010-08-24</h2>
  58. <ul>
  59. <li>Correctness and completeness:
  60. <ul>
  61. <li>Fix trace exit dispatch to function headers.</li>
  62. <li>Fix Windows and OSX builds with LUAJIT_DISABLE_JIT.</li>
  63. <li>Reorganize and fix placement of generated machine code on x64.</li>
  64. <li>Fix TNEW in x64 interpreter.</li>
  65. <li>Do not eliminate PHIs for values only referenced from side exits.</li>
  66. <li>OS-independent canonicalization of strings for non-finite numbers.</li>
  67. <li>Fix <tt>string.char()</tt> range check on x64.</li>
  68. <li>Fix <tt>tostring()</tt> resolving within <tt>print()</tt>.</li>
  69. <li>Fix error handling for <tt>next()</tt>.</li>
  70. <li>Fix passing of constant arguments to external calls on x64.</li>
  71. <li>Fix interpreter argument check for two-argument SSE math functions.</li>
  72. <li>Fix C frame chain corruption caused by <tt>lua_cpcall()</tt>.</li>
  73. <li>Fix return from <tt>pcall()</tt> within active hook.</li>
  74. </ul></li>
  75. <li>Structural and performance enhancements:
  76. <ul>
  77. <li>Replace on-trace GC frame syncing with interpreter exit.</li>
  78. <li>Improve hash lookup specialization by not removing dead keys during GC.</li>
  79. <li>Turn traces into true GC objects.</li>
  80. <li>Avoid starting a GC cycle immediately after library init.</li>
  81. <li>Add weak guards to improve dead-code elimination.</li>
  82. <li>Speed up string interning.</li>
  83. </ul></li>
  84. </ul>
  85. <h2 id="LuaJIT-2.0.0-beta4">LuaJIT 2.0.0-beta4 &mdash; 2010-03-28</h2>
  86. <ul>
  87. <li>Correctness and completeness:
  88. <ul>
  89. <li>Fix precondition for on-trace creation of table keys.</li>
  90. <li>Fix <tt>{f()}</tt> on x64 when table is resized.</li>
  91. <li>Fix folding of ordered comparisons with same references.</li>
  92. <li>Fix snapshot restores for multi-result bytecodes.</li>
  93. <li>Fix potential hang when recording bytecode with nested closures.</li>
  94. <li>Fix recording of <tt>getmetatable()</tt>, <tt>tonumber()</tt> and bad argument types.</li>
  95. <li>Fix SLOAD fusion across returns to lower frames.</li>
  96. </ul></li>
  97. <li>Structural and performance enhancements:
  98. <ul>
  99. <li>Add array bounds check elimination. <tt>-Oabc</tt> is enabled by default.</li>
  100. <li>More tuning for x64, e.g. smaller table objects.</li>
  101. </ul></li>
  102. </ul>
  103. <h2 id="LuaJIT-2.0.0-beta3">LuaJIT 2.0.0-beta3 &mdash; 2010-03-07</h2>
  104. <ul>
  105. <li>LuaJIT x64 port:</li>
  106. <ul>
  107. <li>Port integrated memory allocator to Linux/x64, Windows/x64 and OSX/x64.</li>
  108. <li>Port interpreter and JIT compiler to x64.</li>
  109. <li>Port DynASM to x64.</li>
  110. <li>Many 32/64 bit cleanups in the VM.</li>
  111. <li>Allow building the interpreter with either x87 or SSE2 arithmetics.</li>
  112. <li>Add external unwinding and C++ exception interop (default on x64).</li>
  113. </ul></li>
  114. <li>Correctness and completeness:
  115. <ul>
  116. <li>Fix constructor bytecode generation for certain conditional values.</li>
  117. <li>Fix some cases of ordered string comparisons.</li>
  118. <li>Fix <tt>lua_tocfunction()</tt>.</li>
  119. <li>Fix cutoff register in JMP bytecode for some conditional expressions.</li>
  120. <li>Fix PHI marking algorithm for references from variant slots.</li>
  121. <li>Fix <tt>package.cpath</tt> for non-default PREFIX.</li>
  122. <li>Fix DWARF2 frame unwind information for interpreter on OSX.</li>
  123. <li>Drive the GC forward on string allocations in the parser.</li>
  124. <li>Implement call/return hooks (zero-cost if disabled).</li>
  125. <li>Implement yield from C hooks.</li>
  126. <li>Disable JIT compiler on older non-SSE2 CPUs instead of aborting.</li>
  127. </ul></li>
  128. <li>Structural and performance enhancements:
  129. <ul>
  130. <li>Compile recursive code (tail-, up- and down-recursion).</li>
  131. <li>Improve heuristics for bytecode penalties and blacklisting.</li>
  132. <li>Split CALL/FUNC recording and clean up fast function call semantics.</li>
  133. <li>Major redesign of internal function call handling.</li>
  134. <li>Improve FOR loop const specialization and integerness checks.</li>
  135. <li>Switch to pre-initialized stacks. Avoid frame-clearing.</li>
  136. <li>Colocation of prototypes and related data: bytecode, constants, debug info.</li>
  137. <li>Cleanup parser and streamline bytecode generation.</li>
  138. <li>Add support for weak IR references to register allocator.</li>
  139. <li>Switch to compressed, extensible snapshots.</li>
  140. <li>Compile returns to frames below the start frame.</li>
  141. <li>Improve alias analysis of upvalues using a disambiguation hash value.</li>
  142. <li>Compile floor/ceil/trunc to SSE2 helper calls or SSE4.1 instructions.</li>
  143. <li>Add generic C call handling to IR and backend.</li>
  144. <li>Improve KNUM fuse vs. load heuristics.</li>
  145. <li>Compile various <tt>io.*()</tt> functions.</li>
  146. <li>Compile <tt>math.sinh()</tt>, <tt>math.cosh()</tt>, <tt>math.tanh()</tt>
  147. and <tt>math.random()</tt>.</li>
  148. </ul></li>
  149. </ul>
  150. <h2 id="LuaJIT-2.0.0-beta2">LuaJIT 2.0.0-beta2 &mdash; 2009-11-09</h2>
  151. <ul>
  152. <li>Reorganize build system. Build static+shared library on POSIX.</li>
  153. <li>Allow C++ exception conversion on all platforms
  154. using a wrapper function.</li>
  155. <li>Automatically catch C++ exceptions and rethrow Lua error
  156. (DWARF2 only).</li>
  157. <li>Check for the correct x87 FPU precision at strategic points.</li>
  158. <li>Always use wrappers for libm functions.</li>
  159. <li>Resurrect metamethod name strings before copying them.</li>
  160. <li>Mark current trace, even if compiler is idle.</li>
  161. <li>Ensure FILE metatable is created only once.</li>
  162. <li>Fix type comparisons when different integer types are involved.</li>
  163. <li>Fix <tt>getmetatable()</tt> recording.</li>
  164. <li>Fix TDUP with dead keys in template table.</li>
  165. <li><tt>jit.flush(tr)</tt> returns status.
  166. Prevent manual flush of a trace that's still linked.</li>
  167. <li>Improve register allocation heuristics for invariant references.</li>
  168. <li>Compile the push/pop variants of <tt>table.insert()</tt> and
  169. <tt>table.remove()</tt>.</li>
  170. <li>Compatibility with MSVC <tt>link&nbsp/debug</tt>.</li>
  171. <li>Fix <tt>lua_iscfunction()</tt>.</li>
  172. <li>Fix <tt>math.random()</tt> when compiled with <tt>-fpic</tt> (OSX).</li>
  173. <li>Fix <tt>table.maxn()</tt>.</li>
  174. <li>Bump <tt>MACOSX_DEPLOYMENT_TARGET</tt> to <tt>10.4</tt></li>
  175. <li><tt>luaL_check*()</tt> and <tt>luaL_opt*()</tt> now support
  176. negative arguments, too.<br>
  177. This matches the behavior of Lua 5.1, but not the specification.</li>
  178. </ul>
  179. <h2 id="LuaJIT-2.0.0-beta1">LuaJIT 2.0.0-beta1 &mdash; 2009-10-31</h2>
  180. <ul>
  181. <li>This is the first public release of LuaJIT 2.0.</li>
  182. <li>The whole VM has been rewritten from the ground up, so there's
  183. no point in listing differences over earlier versions.</li>
  184. </ul>
  185. </div>
  186. <div class="major" style="background: #d0d0ff;">
  187. <h2 id="LuaJIT-1.1.6">LuaJIT 1.1.6 &mdash; 2010-03-28</h2>
  188. <ul>
  189. <li>Added fixes for the
  190. <a href="http://www.lua.org/bugs.html#5.1.4"><span class="ext">&raquo;</span>&nbsp;currently known bugs in Lua 5.1.4</a>.</li>
  191. <li>Removed wrong GC check in <tt>jit_createstate()</tt>.
  192. Thanks to Tim Mensch.</li>
  193. <li>Fixed bad assertions while compiling <tt>table.insert()</tt> and
  194. <tt>table.remove()</tt>.</li>
  195. </ul>
  196. <h2 id="LuaJIT-1.1.5">LuaJIT 1.1.5 &mdash; 2008-10-25</h2>
  197. <ul>
  198. <li>Merged with Lua 5.1.4. Fixes all
  199. <a href="http://www.lua.org/bugs.html#5.1.3"><span class="ext">&raquo;</span>&nbsp;known bugs in Lua 5.1.3</a>.</li>
  200. </ul>
  201. <h2 id="LuaJIT-1.1.4">LuaJIT 1.1.4 &mdash; 2008-02-05</h2>
  202. <ul>
  203. <li>Merged with Lua 5.1.3. Fixes all
  204. <a href="http://www.lua.org/bugs.html#5.1.2"><span class="ext">&raquo;</span>&nbsp;known bugs in Lua 5.1.2</a>.</li>
  205. <li>Fixed possible (but unlikely) stack corruption while compiling
  206. <tt>k^x</tt> expressions.</li>
  207. <li>Fixed DynASM template for cmpss instruction.</li>
  208. </ul>
  209. <h2 id="LuaJIT-1.1.3">LuaJIT 1.1.3 &mdash; 2007-05-24</h2>
  210. <ul>
  211. <li>Merged with Lua 5.1.2. Fixes all
  212. <a href="http://www.lua.org/bugs.html#5.1.1"><span class="ext">&raquo;</span>&nbsp;known bugs in Lua 5.1.1</a>.</li>
  213. <li>Merged pending Lua 5.1.x fixes: "return -nil" bug, spurious count hook call.</li>
  214. <li>Remove a (sometimes) wrong assertion in <tt>luaJIT_findpc()</tt>.</li>
  215. <li>DynASM now allows labels for displacements and <tt>.aword</tt>.</li>
  216. <li>Fix some compiler warnings for DynASM glue (internal API change).</li>
  217. <li>Correct naming for SSSE3 (temporarily known as SSE4) in DynASM and x86 disassembler.</li>
  218. <li>The loadable debug modules now handle redirection to stdout
  219. (e.g. <tt>-j&nbsp;trace=-</tt>).</li>
  220. </ul>
  221. <h2 id="LuaJIT-1.1.2">LuaJIT 1.1.2 &mdash; 2006-06-24</h2>
  222. <ul>
  223. <li>Fix MSVC inline assembly: use only local variables with
  224. <tt>lua_number2int()</tt>.</li>
  225. <li>Fix "attempt to call a thread value" bug on Mac OS X:
  226. make values of consts used as lightuserdata keys unique
  227. to avoid joining by the compiler/linker.</li>
  228. </ul>
  229. <h2 id="LuaJIT-1.1.1">LuaJIT 1.1.1 &mdash; 2006-06-20</h2>
  230. <ul>
  231. <li>Merged with Lua 5.1.1. Fixes all
  232. <a href="http://www.lua.org/bugs.html#5.1"><span class="ext">&raquo;</span>&nbsp;known bugs in Lua 5.1</a>.</li>
  233. <li>Enforce (dynamic) linker error for EXE/DLL version mismatches.</li>
  234. <li>Minor changes to DynASM: faster preprocessing, smaller encoding
  235. for some immediates.</li>
  236. </ul>
  237. <p>
  238. This release is in sync with Coco 1.1.1 (see the
  239. <a href="http://coco.luajit.org/changes.html"><span class="ext">&raquo;</span>&nbsp;Coco Change History</a>).
  240. </p>
  241. <h2 id="LuaJIT-1.1.0">LuaJIT 1.1.0 &mdash; 2006-03-13</h2>
  242. <ul>
  243. <li>Merged with Lua 5.1 (final).</li>
  244. <li>New JIT call frame setup:
  245. <ul>
  246. <li>The C stack is kept 16 byte aligned (faster).
  247. Mandatory for Mac OS X on Intel, too.</li>
  248. <li>Faster calling conventions for internal C helper functions.</li>
  249. <li>Better instruction scheduling for function prologue, OP_CALL and
  250. OP_RETURN.</li>
  251. </ul></li>
  252. <li>Miscellaneous optimizations:
  253. <ul>
  254. <li>Faster loads of FP constants. Remove narrow-to-wide store-to-load
  255. forwarding stalls.</li>
  256. <li>Use (scalar) SSE2 ops (if the CPU supports it) to speed up slot moves
  257. and FP to integer conversions.</li>
  258. <li>Optimized the two-argument form of <tt>OP_CONCAT</tt> (<tt>a..b</tt>).</li>
  259. <li>Inlined <tt>OP_MOD</tt> (<tt>a%b</tt>).
  260. With better accuracy than the C variant, too.</li>
  261. <li>Inlined <tt>OP_POW</tt> (<tt>a^b</tt>). Unroll <tt>x^k</tt> or
  262. use <tt>k^x = 2^(log2(k)*x)</tt> or call <tt>pow()</tt>.</li>
  263. </ul></li>
  264. <li>Changes in the optimizer:
  265. <ul>
  266. <li>Improved hinting for table keys derived from table values
  267. (<tt>t1[t2[x]]</tt>).</li>
  268. <li>Lookup hinting now works with arbitrary object types and
  269. supports index chains, too.</li>
  270. <li>Generate type hints for arithmetic and comparison operators,
  271. OP_LEN, OP_CONCAT and OP_FORPREP.</li>
  272. <li>Remove several hint definitions in favour of a generic COMBINE hint.</li>
  273. <li>Complete rewrite of <tt>jit.opt_inline</tt> module
  274. (ex <tt>jit.opt_lib</tt>).</li>
  275. </ul></li>
  276. <li>Use adaptive deoptimization:
  277. <ul>
  278. <li>If runtime verification of a contract fails, the affected
  279. instruction is recompiled and patched on-the-fly.
  280. Regular programs will trigger deoptimization only occasionally.</li>
  281. <li>This avoids generating code for uncommon fallback cases
  282. most of the time. Generated code is up to 30% smaller compared to
  283. LuaJIT&nbsp;1.0.3.</li>
  284. <li>Deoptimization is used for many opcodes and contracts:
  285. <ul>
  286. <li>OP_CALL, OP_TAILCALL: type mismatch for callable.</li>
  287. <li>Inlined calls: closure mismatch, parameter number and type mismatches.</li>
  288. <li>OP_GETTABLE, OP_SETTABLE: table or key type and range mismatches.</li>
  289. <li>All arithmetic and comparison operators, OP_LEN, OP_CONCAT,
  290. OP_FORPREP: operand type and range mismatches.</li>
  291. </ul></li>
  292. <li>Complete redesign of the debug and traceback info
  293. (bytecode &harr; mcode) to support deoptimization.
  294. Much more flexible and needs only 50% of the space.</li>
  295. <li>The modules <tt>jit.trace</tt>, <tt>jit.dumphints</tt> and
  296. <tt>jit.dump</tt> handle deoptimization.</li>
  297. </ul></li>
  298. <li>Inlined many popular library functions
  299. (for commonly used arguments only):
  300. <ul>
  301. <li>Most <tt>math.*</tt> functions (the 18 most used ones)
  302. [2x-10x faster].</li>
  303. <li><tt>string.len</tt>, <tt>string.sub</tt> and <tt>string.char</tt>
  304. [2x-10x faster].</li>
  305. <li><tt>table.insert</tt>, <tt>table.remove</tt> and <tt>table.getn</tt>
  306. [3x-5x faster].</li>
  307. <li><tt>coroutine.yield</tt> and <tt>coroutine.resume</tt>
  308. [3x-5x faster].</li>
  309. <li><tt>pairs</tt>, <tt>ipairs</tt> and the corresponding iterators
  310. [8x-15x faster].</li>
  311. </ul></li>
  312. <li>Changes in the core and loadable modules and the stand-alone executable:
  313. <ul>
  314. <li>Added <tt>jit.version</tt>, <tt>jit.version_num</tt>
  315. and <tt>jit.arch</tt>.</li>
  316. <li>Reorganized some internal API functions (<tt>jit.util.*mcode*</tt>).</li>
  317. <li>The <tt>-j dump</tt> output now shows JSUB names, too.</li>
  318. <li>New x86 disassembler module written in pure Lua. No dependency
  319. on ndisasm anymore. Flexible API, very compact (500 lines)
  320. and complete (x87, MMX, SSE, SSE2, SSE3, SSSE3, privileged instructions).</li>
  321. <li><tt>luajit -v</tt> prints the LuaJIT version and copyright
  322. on a separate line.</li>
  323. </ul></li>
  324. <li>Added SSE, SSE2, SSE3 and SSSE3 support to DynASM.</li>
  325. <li>Miscellaneous doc changes. Added a section about
  326. <a href="install.html#embedding">embedding LuaJIT</a>.</li>
  327. </ul>
  328. <p>
  329. This release is in sync with Coco 1.1.0 (see the
  330. <a href="http://coco.luajit.org/changes.html"><span class="ext">&raquo;</span>&nbsp;Coco Change History</a>).
  331. </p>
  332. </div>
  333. <div class="major" style="background: #ffffd0;">
  334. <h2 id="LuaJIT-1.0.3">LuaJIT 1.0.3 &mdash; 2005-09-08</h2>
  335. <ul>
  336. <li>Even more docs.</li>
  337. <li>Unified closure checks in <tt>jit.*</tt>.</li>
  338. <li>Fixed some range checks in <tt>jit.util.*</tt>.</li>
  339. <li>Fixed __newindex call originating from <tt>jit_settable_str()</tt>.</li>
  340. <li>Merged with Lua 5.1 alpha (including early bug fixes).</li>
  341. </ul>
  342. <p>
  343. This is the first public release of LuaJIT.
  344. </p>
  345. <h2 id="LuaJIT-1.0.2">LuaJIT 1.0.2 &mdash; 2005-09-02</h2>
  346. <ul>
  347. <li>Add support for flushing the Valgrind translation cache <br>
  348. (<tt>MYCFLAGS= -DUSE_VALGRIND</tt>).</li>
  349. <li>Add support for freeing executable mcode memory to the <tt>mmap()</tt>-based
  350. variant for POSIX systems.</li>
  351. <li>Reorganized the C&nbsp;function signature handling in
  352. <tt>jit.opt_lib</tt>.</li>
  353. <li>Changed to index-based hints for inlining C&nbsp;functions.
  354. Still no support in the backend for inlining.</li>
  355. <li>Hardcode <tt>HEAP_CREATE_ENABLE_EXECUTE</tt> value if undefined.</li>
  356. <li>Misc. changes to the <tt>jit.*</tt> modules.</li>
  357. <li>Misc. changes to the Makefiles.</li>
  358. <li>Lots of new docs.</li>
  359. <li>Complete doc reorg.</li>
  360. </ul>
  361. <p>
  362. Not released because Lua 5.1 alpha came out today.
  363. </p>
  364. <h2 id="LuaJIT-1.0.1">LuaJIT 1.0.1 &mdash; 2005-08-31</h2>
  365. <ul>
  366. <li>Missing GC step in <tt>OP_CONCAT</tt>.</li>
  367. <li>Fix result handling for C &ndash;> JIT calls.</li>
  368. <li>Detect CPU feature bits.</li>
  369. <li>Encode conditional moves (<tt>fucomip</tt>) only when supported.</li>
  370. <li>Add fallback instructions for FP compares.</li>
  371. <li>Add support for <tt>LUA_COMPAT_VARARG</tt>. Still disabled by default.</li>
  372. <li>MSVC needs a specific place for the <tt>CALLBACK</tt> attribute
  373. (David Burgess).</li>
  374. <li>Misc. doc updates.</li>
  375. </ul>
  376. <p>
  377. Interim non-public release.
  378. Special thanks to Adam D. Moss for reporting most of the bugs.
  379. </p>
  380. <h2 id="LuaJIT-1.0.0">LuaJIT 1.0.0 &mdash; 2005-08-29</h2>
  381. <p>
  382. This is the initial non-public release of LuaJIT.
  383. </p>
  384. </div>
  385. <br class="flush">
  386. </div>
  387. <div id="foot">
  388. <hr class="hide">
  389. Copyright &copy; 2005-2010 Mike Pall
  390. <span class="noprint">
  391. &middot;
  392. <a href="contact.html">Contact</a>
  393. </span>
  394. </div>
  395. </body>
  396. </html>