1 2008-12-29 Cameron Zwarich <cwzwarich@uwaterloo.ca>
3 Rubber-stamped by Sam Weinig.
5 Remove unused kjsyydebug #define.
9 2008-12-29 Cameron Zwarich <cwzwarich@uwaterloo.ca>
11 Reviewed by Oliver Hunt and Sam Weinig.
13 Bug 23029: REGRESSION (r39337): jsfunfuzz generates identical test files
14 <https://bugs.webkit.org/show_bug.cgi?id=23029>
15 <rdar://problem/6469185>
17 The unification of random number generation in r39337 resulted in random()
18 being initialized on Darwin, but rand() actually being used. Fix this by
19 making randomNumber() use random() instead of rand() on Darwin.
21 * wtf/RandomNumber.cpp:
24 2008-12-29 Sam Weinig <sam@webkit.org>
28 * runtime/Structure.cpp:
30 2008-12-29 Sam Weinig <sam@webkit.org>
32 Reviewed by Oliver Hunt.
34 Patch for https://bugs.webkit.org/show_bug.cgi?id=23026
35 Move the deleted offsets vector into the PropertyMap
37 Saves 3 words per Structure.
39 * runtime/PropertyMapHashTable.h:
40 * runtime/Structure.cpp:
41 (JSC::Structure::addPropertyTransition):
42 (JSC::Structure::changePrototypeTransition):
43 (JSC::Structure::getterSetterTransition):
44 (JSC::Structure::toDictionaryTransition):
45 (JSC::Structure::fromDictionaryTransition):
46 (JSC::Structure::copyPropertyTable):
47 (JSC::Structure::put):
48 (JSC::Structure::remove):
49 (JSC::Structure::rehashPropertyMapHashTable):
50 * runtime/Structure.h:
51 (JSC::Structure::propertyStorageSize):
53 2008-12-29 Cameron Zwarich <cwzwarich@uwaterloo.ca>
55 Reviewed by Oliver Hunt.
57 Change code using m_body.get() as a boolean to take advantage of the
58 implicit conversion of RefPtr to boolean.
60 * runtime/JSFunction.cpp:
61 (JSC::JSFunction::~JSFunction):
63 2008-12-28 Cameron Zwarich <cwzwarich@uwaterloo.ca>
65 Reviewed by Oliver Hunt.
67 Bug 22840: REGRESSION (r38349): Gmail doesn't load with profiling enabled
68 <https://bugs.webkit.org/show_bug.cgi?id=22840>
69 <rdar://problem/6468077>
71 * bytecompiler/BytecodeGenerator.cpp:
72 (JSC::BytecodeGenerator::emitNewArray): Add an assertion that the range
73 of registers passed to op_new_array is sequential.
74 (JSC::BytecodeGenerator::emitCall): Correct the relocation of registers
75 when emitting profiler hooks so that registers aren't leaked. Also, add
76 an assertion that the 'this' register is always ref'd (because it is),
77 remove the needless protection of the 'this' register when relocating,
78 and add an assertion that the range of registers passed to op_call for
79 function call arguments is sequential.
80 (JSC::BytecodeGenerator::emitConstruct): Correct the relocation of
81 registers when emitting profiler hooks so that registers aren't leaked.
82 Also, add an assertion that the range of registers passed to op_construct
83 for function call arguments is sequential.
85 2008-12-26 Mark Rowe <mrowe@apple.com>
87 Reviewed by Alexey Proskuryakov.
89 <rdar://problem/6467376> Race condition in WTF::currentThread can lead to a thread using two different identifiers during its lifetime
91 If a newly-created thread calls WTF::currentThread() before WTF::createThread calls establishIdentifierForPthreadHandle
92 then more than one identifier will be used for the same thread. We can avoid this by adding some extra synchronization
93 during thread creation that delays the execution of the thread function until the thread identifier has been set up, and
94 an assertion to catch this problem should it reappear in the future.
96 * wtf/Threading.cpp: Added.
97 (WTF::NewThreadContext::NewThreadContext):
98 (WTF::threadEntryPoint):
99 (WTF::createThread): Add cross-platform createThread function that delays the execution of the thread function until
100 after the thread identifier has been set up.
102 * wtf/ThreadingGtk.cpp:
103 (WTF::establishIdentifierForThread):
104 (WTF::createThreadInternal):
105 * wtf/ThreadingNone.cpp:
106 (WTF::createThreadInternal):
107 * wtf/ThreadingPthreads.cpp:
108 (WTF::establishIdentifierForPthreadHandle):
109 (WTF::createThreadInternal):
110 * wtf/ThreadingQt.cpp:
111 (WTF::identifierByQthreadHandle):
112 (WTF::establishIdentifierForThread):
113 (WTF::createThreadInternal):
114 * wtf/ThreadingWin.cpp:
115 (WTF::storeThreadHandleByIdentifier):
116 (WTF::createThreadInternal):
118 Add Threading.cpp to the build.
121 * JavaScriptCore.pri:
122 * JavaScriptCore.scons:
123 * JavaScriptCore.vcproj/WTF/WTF.vcproj:
124 * JavaScriptCore.xcodeproj/project.pbxproj:
125 * JavaScriptCoreSources.bkl:
127 2008-12-26 Sam Weinig <sam@webkit.org>
129 Reviewed by Alexey Proskuryakov.
131 Remove unused method.
133 * runtime/Structure.h: Remove mutableTypeInfo.
135 2008-12-22 Gavin Barraclough <barraclough@apple.com>
137 Reviewed by Oliver Hunt.
139 Fix rounding / bounds / signed comparison bug in ExecutableAllocator.
141 ExecutableAllocator::alloc assumed that m_freePtr would be aligned. This was
142 not always true, since the first allocation from an additional pool would not
143 be rounded up. Subsequent allocations would be unaligned, and too much memory
144 could be erroneously allocated from the pool, when the size requested was
145 available, but the size rounded up to word granularity was not available in the
146 pool. This may result in the value of m_freePtr being greater than m_end.
148 Under these circumstances, the unsigned check for space will always pass,
149 resulting in pointers to memory outside of the arena being returned, and
150 ultimately segfaulty goodness when attempting to memcpy the hot freshly jitted
151 code from the AssemblerBuffer.
153 https://bugs.webkit.org/show_bug.cgi?id=22974
154 ... and probably many, many more.
156 * jit/ExecutableAllocator.h:
157 (JSC::ExecutablePool::alloc):
158 (JSC::ExecutablePool::roundUpAllocationSize):
159 (JSC::ExecutablePool::ExecutablePool):
160 (JSC::ExecutablePool::poolAllocate):
162 2008-12-22 Sam Weinig <sam@webkit.org>
164 Reviewed by Gavin Barraclough.
166 Rename all uses of the term "repatch" to "patch".
168 * assembler/MacroAssembler.h:
169 (JSC::MacroAssembler::DataLabelPtr::patch):
170 (JSC::MacroAssembler::DataLabel32::patch):
171 (JSC::MacroAssembler::Jump::patch):
172 (JSC::MacroAssembler::PatchBuffer::PatchBuffer):
173 (JSC::MacroAssembler::PatchBuffer::setPtr):
174 (JSC::MacroAssembler::loadPtrWithAddressOffsetPatch):
175 (JSC::MacroAssembler::storePtrWithAddressOffsetPatch):
176 (JSC::MacroAssembler::storePtrWithPatch):
177 (JSC::MacroAssembler::jnePtrWithPatch):
178 * assembler/X86Assembler.h:
179 (JSC::X86Assembler::patchAddress):
180 (JSC::X86Assembler::patchImmediate):
181 (JSC::X86Assembler::patchPointer):
182 (JSC::X86Assembler::patchBranchOffset):
183 * interpreter/Interpreter.cpp:
184 (JSC::Interpreter::tryCTICachePutByID):
185 (JSC::Interpreter::tryCTICacheGetByID):
186 (JSC::Interpreter::cti_op_put_by_id):
187 (JSC::Interpreter::cti_op_get_by_id):
188 (JSC::Interpreter::cti_op_get_by_id_self_fail):
189 (JSC::Interpreter::cti_op_get_by_id_proto_list):
190 (JSC::Interpreter::cti_vm_dontLazyLinkCall):
192 (JSC::ctiPatchCallByReturnAddress):
193 (JSC::JIT::privateCompileMainPass):
194 (JSC::JIT::privateCompile):
195 (JSC::JIT::privateCompileCTIMachineTrampolines):
198 (JSC::JIT::unlinkCall):
199 (JSC::JIT::linkCall):
200 (JSC::JIT::compileOpCall):
201 * jit/JITPropertyAccess.cpp:
202 (JSC::JIT::compileGetByIdHotPath):
203 (JSC::JIT::compilePutByIdHotPath):
204 (JSC::JIT::compileGetByIdSlowCase):
205 (JSC::JIT::compilePutByIdSlowCase):
206 (JSC::JIT::privateCompilePutByIdTransition):
207 (JSC::JIT::patchGetByIdSelf):
208 (JSC::JIT::patchPutByIdReplace):
209 (JSC::JIT::privateCompilePatchGetArrayLength):
210 (JSC::JIT::privateCompileGetByIdSelf):
211 (JSC::JIT::privateCompileGetByIdProto):
212 (JSC::JIT::privateCompileGetByIdSelfList):
213 (JSC::JIT::privateCompileGetByIdProtoList):
214 (JSC::JIT::privateCompileGetByIdChainList):
215 (JSC::JIT::privateCompileGetByIdChain):
216 (JSC::JIT::privateCompilePutByIdReplace):
218 2008-12-22 Adam Roben <aroben@apple.com>
220 Build fix after r39428
223 (JSC::JIT::compileOpCallSlowCase): Added a missing MacroAssembler::
225 2008-12-22 Nikolas Zimmermann <nikolas.zimmermann@torchmobile.com>
227 Rubber-stamped by George Staikos.
229 Unify all TorchMobile copyright lines. Consolidate in a single line, as requested by Mark Rowe, some time ago.
231 * wtf/RandomNumber.cpp:
232 * wtf/RandomNumber.h:
233 * wtf/RandomNumberSeed.h:
235 2008-12-21 Nikolas Zimmermann <nikolas.zimmermann@torchmobile.com>
237 Rubber-stamped by George Staikos.
239 Fix copyright of the new RandomNumber* files.
241 * wtf/RandomNumber.cpp:
242 * wtf/RandomNumber.h:
243 * wtf/RandomNumberSeed.h:
245 2008-12-21 Gavin Barraclough <barraclough@apple.com>
247 Reviewed by Oliver Hunt & Cameron Zwarich.
249 Add support for call and property access repatching on x86-64.
251 No change in performance on current configurations (2x impovement on v8-tests with JIT enabled on x86-64).
253 * assembler/MacroAssembler.h:
254 (JSC::MacroAssembler::DataLabelPtr::repatch):
255 (JSC::MacroAssembler::DataLabelPtr::operator X86Assembler::JmpDst):
256 (JSC::MacroAssembler::DataLabel32::repatch):
257 (JSC::MacroAssembler::RepatchBuffer::addressOf):
258 (JSC::MacroAssembler::add32):
259 (JSC::MacroAssembler::sub32):
260 (JSC::MacroAssembler::loadPtrWithAddressOffsetRepatch):
261 (JSC::MacroAssembler::storePtrWithAddressOffsetRepatch):
262 (JSC::MacroAssembler::jePtr):
263 (JSC::MacroAssembler::jnePtr):
264 (JSC::MacroAssembler::jnePtrWithRepatch):
265 (JSC::MacroAssembler::differenceBetween):
266 * assembler/X86Assembler.h:
267 (JSC::X86Assembler::addl_im):
268 (JSC::X86Assembler::subl_im):
269 (JSC::X86Assembler::cmpl_rm):
270 (JSC::X86Assembler::movq_rm_disp32):
271 (JSC::X86Assembler::movq_mr_disp32):
272 (JSC::X86Assembler::repatchPointer):
273 (JSC::X86Assembler::X86InstructionFormatter::oneByteOp64_disp32):
275 (JSC::JIT::privateCompile):
276 (JSC::JIT::privateCompileCTIMachineTrampolines):
279 (JSC::JIT::unlinkCall):
280 (JSC::JIT::linkCall):
281 (JSC::JIT::compileOpCall):
282 (JSC::JIT::compileOpCallSlowCase):
283 * jit/JITInlineMethods.h:
284 (JSC::JIT::restoreArgumentReferenceForTrampoline):
285 * jit/JITPropertyAccess.cpp:
286 (JSC::JIT::compileGetByIdHotPath):
287 (JSC::JIT::compileGetByIdSlowCase):
288 (JSC::JIT::compilePutByIdHotPath):
289 (JSC::JIT::compilePutByIdSlowCase):
290 (JSC::resizePropertyStorage):
291 (JSC::JIT::privateCompilePutByIdTransition):
292 (JSC::JIT::privateCompileGetByIdProto):
293 (JSC::JIT::privateCompileGetByIdProtoList):
294 (JSC::JIT::privateCompileGetByIdChainList):
295 (JSC::JIT::privateCompileGetByIdChain):
298 2008-12-20 Gavin Barraclough <barraclough@apple.com>
300 Reviewed by Oliver Hunt.
302 Port optimized property access generation to the MacroAssembler.
304 * assembler/MacroAssembler.h:
305 (JSC::MacroAssembler::AbsoluteAddress::AbsoluteAddress):
306 (JSC::MacroAssembler::DataLabelPtr::repatch):
307 (JSC::MacroAssembler::DataLabel32::DataLabel32):
308 (JSC::MacroAssembler::DataLabel32::repatch):
309 (JSC::MacroAssembler::Label::operator X86Assembler::JmpDst):
310 (JSC::MacroAssembler::Jump::repatch):
311 (JSC::MacroAssembler::JumpList::empty):
312 (JSC::MacroAssembler::RepatchBuffer::link):
313 (JSC::MacroAssembler::add32):
314 (JSC::MacroAssembler::and32):
315 (JSC::MacroAssembler::sub32):
316 (JSC::MacroAssembler::loadPtrWithAddressRepatch):
317 (JSC::MacroAssembler::storePtrWithAddressRepatch):
318 (JSC::MacroAssembler::push):
319 (JSC::MacroAssembler::ja32):
320 (JSC::MacroAssembler::jePtr):
321 (JSC::MacroAssembler::jnePtr):
322 (JSC::MacroAssembler::jnePtrWithRepatch):
323 (JSC::MacroAssembler::align):
324 (JSC::MacroAssembler::differenceBetween):
325 * assembler/X86Assembler.h:
326 (JSC::X86Assembler::movl_rm_disp32):
327 (JSC::X86Assembler::movl_mr_disp32):
328 (JSC::X86Assembler::X86InstructionFormatter::oneByteOp_disp32):
329 (JSC::X86Assembler::X86InstructionFormatter::memoryModRM):
331 (JSC::ctiRepatchCallByReturnAddress):
332 (JSC::JIT::privateCompileMainPass):
333 (JSC::JIT::privateCompile):
334 (JSC::JIT::privateCompileCTIMachineTrampolines):
336 * jit/JITPropertyAccess.cpp:
337 (JSC::JIT::compileGetByIdHotPath):
338 (JSC::JIT::compileGetByIdSlowCase):
339 (JSC::JIT::compilePutByIdHotPath):
340 (JSC::JIT::compilePutByIdSlowCase):
341 (JSC::resizePropertyStorage):
342 (JSC::JIT::privateCompilePutByIdTransition):
343 (JSC::JIT::patchGetByIdSelf):
344 (JSC::JIT::patchPutByIdReplace):
345 (JSC::JIT::privateCompilePatchGetArrayLength):
346 (JSC::JIT::privateCompileGetByIdSelf):
347 (JSC::JIT::privateCompileGetByIdProto):
348 (JSC::JIT::privateCompileGetByIdSelfList):
349 (JSC::JIT::privateCompileGetByIdProtoList):
350 (JSC::JIT::privateCompileGetByIdChainList):
351 (JSC::JIT::privateCompileGetByIdChain):
352 (JSC::JIT::privateCompilePutByIdReplace):
354 (WTF::RefCountedBase::addressOfCount):
356 2008-12-19 Gustavo Noronha Silva <gns@gnome.org>
358 Reviewed by Holger Freyther.
360 https://bugs.webkit.org/show_bug.cgi?id=22686
362 Added file which was missing to the javascriptcore_sources
363 variable, so that it shows up in the tarball created by `make
368 2008-12-19 Holger Hans Peter Freyther <zecke@selfish.org>
370 Reviewed by Antti Koivisto.
372 Build fix when building JS API tests with a c89 c compiler
374 Do not use C++ style comments and convert them to C comments.
378 2008-12-18 Gavin Barraclough <barraclough@apple.com>
380 Reviewed by Sam Weinig.
382 Same as last revision, adding cases for pre & post inc & dec.
384 https://bugs.webkit.org/show_bug.cgi?id=22928
387 (JSC::JIT::privateCompileMainPass):
389 2008-12-18 Gavin Barraclough <barraclough@apple.com>
391 Reviewed by Sam Weinig.
393 Fixes for the JIT's handling of JSImmediate values on x86-64.
394 On 64-bit systems, the code in JSImmediate.h relies on the upper
395 bits of a JSImmediate being a sign extension of the low 32-bits.
396 This was not being enforced by the JIT, since a number of inline
397 operations were being performed on 32-bit values in registers, and
398 when a 32-bit result is written to a register on x86-64 the value
399 is zero-extended to 64-bits.
401 This fix honors previous behavoir. A better fix in the long run
402 (when the JIT is enabled by default) may be to change JSImmediate.h
403 so it no longer relies on the upper bits of the pointer,... though
404 if we're going to change JSImmediate.h for 64-bit, we probably may
405 as well change the format so that the full range of 32-bit ints can
406 be stored, rather than just 31-bits.
408 https://bugs.webkit.org/show_bug.cgi?id=22925
410 * assembler/MacroAssembler.h:
411 (JSC::MacroAssembler::addPtr):
412 (JSC::MacroAssembler::andPtr):
413 (JSC::MacroAssembler::orPtr):
414 (JSC::MacroAssembler::or32):
415 (JSC::MacroAssembler::xor32):
416 (JSC::MacroAssembler::xorPtr):
417 (JSC::MacroAssembler::signExtend32ToPtr):
418 * assembler/X86Assembler.h:
419 (JSC::X86Assembler::):
420 (JSC::X86Assembler::andq_rr):
421 (JSC::X86Assembler::andq_ir):
422 (JSC::X86Assembler::orq_rr):
423 (JSC::X86Assembler::xorq_ir):
424 (JSC::X86Assembler::movsxd_rr):
426 (JSC::JIT::privateCompileMainPass):
427 * jit/JITInlineMethods.h:
428 (JSC::JIT::emitFastArithReTagImmediate):
429 (JSC::JIT::emitFastArithPotentiallyReTagImmediate):
430 (JSC::JIT::emitFastArithImmToInt):
432 2008-12-18 Gavin Barraclough <barraclough@apple.com>
434 Reviewed by Sam Weinig.
436 Just a tidy up - rename & refactor some the #defines configuring the JIT.
438 * interpreter/Interpreter.cpp:
439 (JSC::Interpreter::cti_op_convert_this):
440 (JSC::Interpreter::cti_op_end):
441 (JSC::Interpreter::cti_op_add):
442 (JSC::Interpreter::cti_op_pre_inc):
443 (JSC::Interpreter::cti_timeout_check):
444 (JSC::Interpreter::cti_register_file_check):
445 (JSC::Interpreter::cti_op_loop_if_less):
446 (JSC::Interpreter::cti_op_loop_if_lesseq):
447 (JSC::Interpreter::cti_op_new_object):
448 (JSC::Interpreter::cti_op_put_by_id_generic):
449 (JSC::Interpreter::cti_op_get_by_id_generic):
450 (JSC::Interpreter::cti_op_put_by_id):
451 (JSC::Interpreter::cti_op_put_by_id_second):
452 (JSC::Interpreter::cti_op_put_by_id_fail):
453 (JSC::Interpreter::cti_op_get_by_id):
454 (JSC::Interpreter::cti_op_get_by_id_second):
455 (JSC::Interpreter::cti_op_get_by_id_self_fail):
456 (JSC::Interpreter::cti_op_get_by_id_proto_list):
457 (JSC::Interpreter::cti_op_get_by_id_proto_list_full):
458 (JSC::Interpreter::cti_op_get_by_id_proto_fail):
459 (JSC::Interpreter::cti_op_get_by_id_array_fail):
460 (JSC::Interpreter::cti_op_get_by_id_string_fail):
461 (JSC::Interpreter::cti_op_instanceof):
462 (JSC::Interpreter::cti_op_del_by_id):
463 (JSC::Interpreter::cti_op_mul):
464 (JSC::Interpreter::cti_op_new_func):
465 (JSC::Interpreter::cti_op_call_JSFunction):
466 (JSC::Interpreter::cti_op_call_arityCheck):
467 (JSC::Interpreter::cti_vm_dontLazyLinkCall):
468 (JSC::Interpreter::cti_vm_lazyLinkCall):
469 (JSC::Interpreter::cti_op_push_activation):
470 (JSC::Interpreter::cti_op_call_NotJSFunction):
471 (JSC::Interpreter::cti_op_create_arguments):
472 (JSC::Interpreter::cti_op_create_arguments_no_params):
473 (JSC::Interpreter::cti_op_tear_off_activation):
474 (JSC::Interpreter::cti_op_tear_off_arguments):
475 (JSC::Interpreter::cti_op_profile_will_call):
476 (JSC::Interpreter::cti_op_profile_did_call):
477 (JSC::Interpreter::cti_op_ret_scopeChain):
478 (JSC::Interpreter::cti_op_new_array):
479 (JSC::Interpreter::cti_op_resolve):
480 (JSC::Interpreter::cti_op_construct_JSConstruct):
481 (JSC::Interpreter::cti_op_construct_NotJSConstruct):
482 (JSC::Interpreter::cti_op_get_by_val):
483 (JSC::Interpreter::cti_op_resolve_func):
484 (JSC::Interpreter::cti_op_sub):
485 (JSC::Interpreter::cti_op_put_by_val):
486 (JSC::Interpreter::cti_op_put_by_val_array):
487 (JSC::Interpreter::cti_op_lesseq):
488 (JSC::Interpreter::cti_op_loop_if_true):
489 (JSC::Interpreter::cti_op_negate):
490 (JSC::Interpreter::cti_op_resolve_base):
491 (JSC::Interpreter::cti_op_resolve_skip):
492 (JSC::Interpreter::cti_op_resolve_global):
493 (JSC::Interpreter::cti_op_div):
494 (JSC::Interpreter::cti_op_pre_dec):
495 (JSC::Interpreter::cti_op_jless):
496 (JSC::Interpreter::cti_op_not):
497 (JSC::Interpreter::cti_op_jtrue):
498 (JSC::Interpreter::cti_op_post_inc):
499 (JSC::Interpreter::cti_op_eq):
500 (JSC::Interpreter::cti_op_lshift):
501 (JSC::Interpreter::cti_op_bitand):
502 (JSC::Interpreter::cti_op_rshift):
503 (JSC::Interpreter::cti_op_bitnot):
504 (JSC::Interpreter::cti_op_resolve_with_base):
505 (JSC::Interpreter::cti_op_new_func_exp):
506 (JSC::Interpreter::cti_op_mod):
507 (JSC::Interpreter::cti_op_less):
508 (JSC::Interpreter::cti_op_neq):
509 (JSC::Interpreter::cti_op_post_dec):
510 (JSC::Interpreter::cti_op_urshift):
511 (JSC::Interpreter::cti_op_bitxor):
512 (JSC::Interpreter::cti_op_new_regexp):
513 (JSC::Interpreter::cti_op_bitor):
514 (JSC::Interpreter::cti_op_call_eval):
515 (JSC::Interpreter::cti_op_throw):
516 (JSC::Interpreter::cti_op_get_pnames):
517 (JSC::Interpreter::cti_op_next_pname):
518 (JSC::Interpreter::cti_op_push_scope):
519 (JSC::Interpreter::cti_op_pop_scope):
520 (JSC::Interpreter::cti_op_typeof):
521 (JSC::Interpreter::cti_op_is_undefined):
522 (JSC::Interpreter::cti_op_is_boolean):
523 (JSC::Interpreter::cti_op_is_number):
524 (JSC::Interpreter::cti_op_is_string):
525 (JSC::Interpreter::cti_op_is_object):
526 (JSC::Interpreter::cti_op_is_function):
527 (JSC::Interpreter::cti_op_stricteq):
528 (JSC::Interpreter::cti_op_nstricteq):
529 (JSC::Interpreter::cti_op_to_jsnumber):
530 (JSC::Interpreter::cti_op_in):
531 (JSC::Interpreter::cti_op_push_new_scope):
532 (JSC::Interpreter::cti_op_jmp_scopes):
533 (JSC::Interpreter::cti_op_put_by_index):
534 (JSC::Interpreter::cti_op_switch_imm):
535 (JSC::Interpreter::cti_op_switch_char):
536 (JSC::Interpreter::cti_op_switch_string):
537 (JSC::Interpreter::cti_op_del_by_val):
538 (JSC::Interpreter::cti_op_put_getter):
539 (JSC::Interpreter::cti_op_put_setter):
540 (JSC::Interpreter::cti_op_new_error):
541 (JSC::Interpreter::cti_op_debug):
542 (JSC::Interpreter::cti_vm_throw):
543 * interpreter/Interpreter.h:
546 (JSC::JIT::privateCompileMainPass):
547 (JSC::JIT::privateCompile):
549 * jit/JITInlineMethods.h:
550 (JSC::JIT::restoreArgumentReference):
551 (JSC::JIT::restoreArgumentReferenceForTrampoline):
554 2008-12-18 Cameron Zwarich <zwarich@apple.com>
556 Reviewed by Geoff Garen.
558 Bug 21855: REGRESSION (r37323): Gmail complains about popup blocking when opening a link
559 <https://bugs.webkit.org/show_bug.cgi?id=21855>
560 <rdar://problem/6278244>
562 Move DynamicGlobalObjectScope to JSGlobalObject.h so that it can be used
565 * interpreter/Interpreter.cpp:
566 * runtime/JSGlobalObject.h:
567 (JSC::DynamicGlobalObjectScope::DynamicGlobalObjectScope):
568 (JSC::DynamicGlobalObjectScope::~DynamicGlobalObjectScope):
570 2008-12-17 Geoffrey Garen <ggaren@apple.com>
572 Reviewed by Gavin Barraclough.
574 Fixed https://bugs.webkit.org/show_bug.cgi?id=22393
575 Segfault when caching property accesses to primitive cells.
577 Changed some asObject casts to asCell casts in cases where a primitive
578 value may be a cell and not an object.
580 Re-enabled property caching for primitives in cases where it had been
581 disabled because of this bug.
583 Updated a comment to better explain something Darin thought needed
584 explaining in an old patch review.
586 * interpreter/Interpreter.cpp:
587 (JSC::countPrototypeChainEntriesAndCheckForProxies):
588 (JSC::Interpreter::tryCacheGetByID):
589 (JSC::Interpreter::tryCTICacheGetByID):
590 (JSC::Interpreter::cti_op_get_by_id_self_fail):
591 (JSC::Interpreter::cti_op_get_by_id_proto_list):
593 2008-12-17 Gavin Barraclough <barraclough@apple.com>
595 Reviewed by Cameron Zwarich.
597 Fixes for Sunspider failures with the JIT enabled on x86-64.
599 * assembler/MacroAssembler.h:
600 Switch the order of the RegisterID & Address form of je32, to keep it consistent with jne32.
603 * jit/JITInlineMethods.h:
604 Port the m_ctiVirtualCall tramopline generation to use the MacroAssembler interface.
606 Fix bug in the non-optimizing code path, vptr check should have been to the memory address pointer
607 to by the register, not to the register itself.
608 * wrec/WRECGenerator.cpp:
609 See assembler/MacroAssembler.h, above.
611 2008-12-17 Gavin Barraclough <barraclough@apple.com>
613 Reviewed by Sam Weinig.
615 print("Hello, 64-bit jitted world!");
616 Get hello-world working through the JIT, on x86-64.
618 * assembler/X86Assembler.h:
619 Fix encoding of opcode + RegisterID format instructions for 64-bit.
620 * interpreter/Interpreter.cpp:
621 * interpreter/Interpreter.h:
622 Make VoidPtrPair actually be a pair of void*s.
623 (Possibly should make this change for 32-bit Mac platforms, too - but won't change 32-bit behaviour in this patch).
626 Provide names for the timeoutCheckRegister & callFrameRegister on x86-64,
627 force x86-64 ctiTrampoline arguments onto the stack,
628 implement the asm trampolines for x86-64,
629 implement the restoreArgumentReference methods for x86-64 calling conventions.
631 * jit/JITInlineMethods.h:
633 Add switch settings to ENABLE(JIT), on PLATFORM(X86_64) (currently still disabled).
635 2008-12-17 Sam Weinig <sam@webkit.org>
637 Reviewed by Gavin Barraclough.
639 Add more CodeBlock statistics.
641 * bytecode/CodeBlock.cpp:
642 (JSC::CodeBlock::dumpStatistics):
644 2008-12-17 Sam Weinig <sam@webkit.org>
646 Reviewed by Darin Adler.
648 Fix for https://bugs.webkit.org/show_bug.cgi?id=22897
649 <rdar://problem/6428342>
650 Look into feasibility of discarding bytecode after native codegen
652 Clear the bytecode Instruction vector at the end JIT generation.
654 Saves 4.8 MB on Membuster head.
656 * bytecode/CodeBlock.cpp:
657 (JSC::CodeBlock::dump): Add logging for the case that someone tries
658 to dump the instructions of a CodeBlock that has had its bytecode
660 (JSC::CodeBlock::CodeBlock): Initialize the instructionCount
661 (JSC::CodeBlock::handlerForBytecodeOffset): Use instructionCount instead
662 of the size of the instruction vector in the assertion.
663 (JSC::CodeBlock::lineNumberForBytecodeOffset): Ditto.
664 (JSC::CodeBlock::expressionRangeForBytecodeOffset): Ditto.
665 (JSC::CodeBlock::getByIdExceptionInfoForBytecodeOffset): Ditto.
666 (JSC::CodeBlock::functionRegisterForBytecodeOffset): Ditto.
667 * bytecode/CodeBlock.h:
668 (JSC::CodeBlock::setInstructionCount): Store the instruction vector size
669 in debug builds for assertions.
670 * bytecompiler/BytecodeGenerator.cpp:
671 (JSC::BytecodeGenerator::generate):
673 (JSC::JIT::privateCompile): Clear the bytecode vector unless we
674 have compiled with Opcode sampling where we will continue to require it
676 2008-12-17 Cary Clark <caryclark@google.com>
678 Reviewed by Darin Adler.
679 Landed by Adam Barth.
681 Add ENABLE_TEXT_CARET to permit the ANDROID platform
682 to invalidate and draw the caret in a separate thread.
685 Default ENABLE_TEXT_CARET to 1.
687 2008-12-17 Alexey Proskuryakov <ap@webkit.org>
689 Reviewed by Darin Adler.
691 Don't use unique context group in JSGlobalContextCreate() on Tiger or Leopard, take two.
693 * API/JSContextRef.cpp: The previous patch that claimed to do this was making Tiger and
694 Leopard always use unique context group instead.
696 2008-12-16 Sam Weinig <sam@webkit.org>
698 Reviewed by Geoffrey Garen.
700 Fix for https://bugs.webkit.org/show_bug.cgi?id=22838
701 Remove dependency on the bytecode Instruction buffer in Interpreter::throwException
702 Part of <rdar://problem/6428342>
704 * bytecode/CodeBlock.cpp:
705 (JSC::CodeBlock::functionRegisterForBytecodeOffset): Added. Function to get
706 a function Register index in a callFrame for a bytecode offset.
707 (JSC::CodeBlock::shrinkToFit): Shrink m_getByIdExceptionInfo and m_functionRegisterInfos.
708 * bytecode/CodeBlock.h:
709 (JSC::FunctionRegisterInfo::FunctionRegisterInfo): Added.
710 (JSC::CodeBlock::addFunctionRegisterInfo):
711 * bytecompiler/BytecodeGenerator.cpp:
712 (JSC::BytecodeGenerator::emitCall):
713 * interpreter/Interpreter.cpp:
714 (JSC::Interpreter::throwException): Use functionRegisterForBytecodeOffset in JIT
717 2008-12-16 Sam Weinig <sam@webkit.org>
719 Reviewed by Gavin Barraclough.
721 Fix for https://bugs.webkit.org/show_bug.cgi?id=22837
722 Remove dependency on the bytecode Instruction buffer in Interpreter::cti_op_call_NotJSFunction
723 Part of <rdar://problem/6428342>
725 * interpreter/CallFrame.h: Added comment regarding returnPC storing a void*.
726 * interpreter/Interpreter.cpp:
727 (JSC::bytecodeOffsetForPC): We no longer have any cases of the PC
728 being in the instruction stream for JIT, so we can remove the check.
729 (JSC::Interpreter::cti_op_call_NotJSFunction): Use the CTI_RETURN_ADDRESS
730 as the call frame returnPC as it is only necessary for looking up when
731 throwing an exception.
732 * interpreter/RegisterFile.h:
733 (JSC::RegisterFile::): Added comment regarding returnPC storing a void*.
734 * jit/JIT.h: Remove ARG_instr4.
736 (JSC::JIT::compileOpCallSetupArgs): Don't pass the instruction pointer.
738 2008-12-16 Darin Adler <darin@apple.com>
740 Reviewed and landed by Cameron Zwarich.
742 Preparatory work for fixing
744 Bug 22887: Make UString::Rep use RefCounted rather than implementing its own ref counting
745 <https://bugs.webkit.org/show_bug.cgi?id=22887>
747 Change the various string translators used by Identifier:add() so that
748 they never zero the ref count of a newly created UString::Rep.
750 * runtime/Identifier.cpp:
751 (JSC::CStringTranslator::translate):
752 (JSC::Identifier::add):
753 (JSC::UCharBufferTranslator::translate):
755 2008-12-16 Gavin Barraclough <barraclough@apple.com>
759 * assembler/AssemblerBuffer.h:
761 2008-12-16 Gavin Barraclough <barraclough@apple.com>
763 Reviewed by Cameron Zwarich.
765 Make the JIT compile on x86-64.
766 This largely involves populting the missing calls in MacroAssembler.h.
767 In addition some reinterpret_casts need removing from the JIT, and the
768 repatching property access code will need to be fully compiled out for
769 now. The changes in interpret.cpp are to reorder the functions so that
770 the _generic forms come before all other property access methods, and
771 then to place all property access methods other than the generic forms
772 under control of the ENABLE_JIT_OPTIMIZE_PROPERTY_ACCESS macro.
774 No performance impact.
776 * assembler/AssemblerBuffer.h:
777 (JSC::AssemblerBuffer::putInt64Unchecked):
778 * assembler/MacroAssembler.h:
779 (JSC::MacroAssembler::loadPtr):
780 (JSC::MacroAssembler::load32):
781 (JSC::MacroAssembler::storePtr):
782 (JSC::MacroAssembler::storePtrWithRepatch):
783 (JSC::MacroAssembler::store32):
784 (JSC::MacroAssembler::poke):
785 (JSC::MacroAssembler::move):
786 (JSC::MacroAssembler::testImm64):
787 (JSC::MacroAssembler::jePtr):
788 (JSC::MacroAssembler::jnePtr):
789 (JSC::MacroAssembler::jnzPtr):
790 (JSC::MacroAssembler::jzPtr):
791 * assembler/X86Assembler.h:
792 (JSC::X86Assembler::):
793 (JSC::X86Assembler::cmpq_rr):
794 (JSC::X86Assembler::cmpq_rm):
795 (JSC::X86Assembler::cmpq_im):
796 (JSC::X86Assembler::testq_i32m):
797 (JSC::X86Assembler::movl_mEAX):
798 (JSC::X86Assembler::movl_i32r):
799 (JSC::X86Assembler::movl_EAXm):
800 (JSC::X86Assembler::movq_rm):
801 (JSC::X86Assembler::movq_mEAX):
802 (JSC::X86Assembler::movq_mr):
803 (JSC::X86Assembler::movq_i64r):
804 (JSC::X86Assembler::movl_mr):
805 (JSC::X86Assembler::X86InstructionFormatter::oneByteOp64):
806 (JSC::X86Assembler::X86InstructionFormatter::immediate64):
807 * interpreter/Interpreter.cpp:
808 (JSC::Interpreter::cti_op_put_by_id_generic):
809 (JSC::Interpreter::cti_op_get_by_id_generic):
810 (JSC::Interpreter::cti_op_put_by_id):
811 (JSC::Interpreter::cti_op_put_by_id_second):
813 (JSC::JIT::privateCompileMainPass):
814 (JSC::JIT::privateCompile):
815 (JSC::JIT::privateCompileCTIMachineTrampolines):
817 (JSC::JIT::compileOpCallSetupArgs):
818 (JSC::JIT::compileOpCall):
819 * jit/JITPropertyAccess.cpp:
820 (JSC::JIT::compileGetByIdHotPath):
821 (JSC::JIT::compilePutByIdHotPath):
822 * runtime/JSImmediate.h:
823 (JSC::JSImmediate::makeInt):
825 2008-12-16 Cameron Zwarich <zwarich@apple.com>
827 Reviewed by Darin Adler.
829 Bug 22869: REGRESSION (r38407): http://news.cnet.com/8301-13579_3-9953533-37.html crashes
830 <https://bugs.webkit.org/show_bug.cgi?id=22869>
831 <rdar://problem/6402499>
833 Before r38407, Structure::m_nameInPrevious was ref'd due to it being
834 stored in a PropertyMap. However, PropertyMaps are created lazily after
835 r38407, so Structure::m_nameInPrevious is not necessarily ref'd while
836 it is being used. Making it a RefPtr instead of a raw pointer fixes
839 Unfortunately, the crash in the bug is rather intermittent, and it is
840 impossible to add an assertion in UString::Ref::ref() to catch this bug
841 because some users of UString::Rep deliberately zero out the reference
842 count. Therefore, there is no layout test accompanying this bug fix.
844 * runtime/Structure.cpp:
845 (JSC::Structure::~Structure): Use get().
846 (JSC::Structure::materializePropertyMap): Use get().
847 (JSC::Structure::addPropertyTransitionToExistingStructure): Use get().
848 (JSC::Structure::addPropertyTransition): Use get().
849 * runtime/Structure.h: Make Structure::m_nameInPrevious a RefPtr instead
852 2008-12-16 Nikolas Zimmermann <nikolas.zimmermann@torchmobile.com>
854 Not reviewed. Attempt to fix win build. No 'using namespace WTF' in this file, needs manual WTF:: prefix.
855 Not sure why the build works as is here.
857 * runtime/MathObject.cpp:
858 (JSC::mathProtoFuncRandom):
860 2008-12-16 Nikolas Zimmermann <nikolas.zimmermann@torchmobile.com>
862 Reviewed by Darin Adler.
864 Fixes: https://bugs.webkit.org/show_bug.cgi?id=22876
866 Unify random number generation in JavaScriptCore & WebCore, by introducing
867 wtf/RandomNumber.h and moving wtf_random/wtf_random_init out of MathExtras.h.
869 wtf_random_init() has been renamed to initializeRandomNumberGenerator() and
870 lives in it's own private header: wtf/RandomNumberSeed.h, only intended to
871 be used from within JavaScriptCore.
873 wtf_random() has been renamed to randomNumber() and lives in a public header
874 wtf/RandomNumber.h, usable from within JavaScriptCore & WebCore. It encapsulates
875 the code taking care of initializing the random number generator (only when
876 building without ENABLE(JSC_MULTIPLE_THREADS), otherwhise initializeThreading()
877 already took care of that).
879 Functional change on darwin: Use random() instead of rand(), as it got a larger
880 period (more randomness). HTMLFormElement already contains this implementation
881 and I just moved it in randomNumber(), as special case for PLATFORM(DARWIN).
883 * GNUmakefile.am: Add RandomNumber.(cpp/h) / RandomNumberSeed.h.
884 * JavaScriptCore.exp: Ditto.
885 * JavaScriptCore.pri: Ditto.
886 * JavaScriptCore.scons: Ditto.
887 * JavaScriptCore.vcproj/WTF/WTF.vcproj: Ditto.
888 * JavaScriptCore.xcodeproj/project.pbxproj: Ditto.
889 * JavaScriptCoreSources.bkl: Ditto.
890 * runtime/MathObject.cpp: Use new WTF::randomNumber() functionality.
891 (JSC::mathProtoFuncRandom):
892 * wtf/MathExtras.h: Move wtf_random / wtf_random_init to new files.
893 * wtf/RandomNumber.cpp: Added.
895 * wtf/RandomNumber.h: Added.
896 * wtf/RandomNumberSeed.h: Added. Internal usage within JSC only.
897 (WTF::initializeRandomNumberGenerator):
898 * wtf/ThreadingGtk.cpp: Rename wtf_random_init() to initializeRandomNumberGenerator().
899 (WTF::initializeThreading):
900 * wtf/ThreadingPthreads.cpp: Ditto.
901 (WTF::initializeThreading):
902 * wtf/ThreadingQt.cpp: Ditto.
903 (WTF::initializeThreading):
904 * wtf/ThreadingWin.cpp: Ditto.
905 (WTF::initializeThreading):
907 2008-12-16 Yael Aharon <yael.aharon@nokia.com>
909 Reviewed by Tor Arne Vestbø.
913 * JavaScriptCore.pri:
915 2008-12-15 Mark Rowe <mrowe@apple.com>
917 Reviewed by Cameron Zwarich.
919 Fix the build with GCC 4.0.
921 * Configurations/JavaScriptCore.xcconfig: GCC 4.0 appears to have a bug when compiling with -funwind-tables on,
922 so don't use it with that compiler version.
924 2008-12-15 Mark Rowe <mrowe@apple.com>
926 Rubber-stamped by Cameron Zwarich.
928 <rdar://problem/6289933> Change WebKit-related projects to build with GCC 4.2 on Leopard.
930 * Configurations/Base.xcconfig:
931 * Configurations/DebugRelease.xcconfig:
933 2008-12-15 Alexey Proskuryakov <ap@webkit.org>
935 Reviewed by Darin Adler.
937 Don't use unique context group in JSGlobalContextCreate() on Tiger or Leopard.
939 * API/JSContextRef.cpp: (JSGlobalContextCreate):
941 2008-12-15 Alexey Proskuryakov <ap@webkit.org>
943 Reviewed by Darin Adler.
945 <rdar://problem/6445089> Mach ports leak from worker threads
947 * interpreter/Interpreter.cpp: (JSC::getCPUTime):
948 Deallocate the thread self port.
950 2008-12-15 Gavin Barraclough <barraclough@apple.com>
952 Reviewed by Mark Rowe.
954 Construct stack frames in JIT code, so that backtracing can still work.
955 <rdar://problem/6447870> JIT should play nice with attempts to take stack traces
959 (JSC::JIT::privateCompileMainPass):
961 2008-12-15 Mark Rowe <mrowe@apple.com>
963 Reviewed by Gavin Barraclough.
965 <rdar://problem/6402262> JavaScriptCore needs exception handling tables in order to get stack traces without frame pointers
967 * Configurations/JavaScriptCore.xcconfig:
969 2008-12-15 Gavin Barraclough <barraclough@apple.com>
971 Rubber stamped by Mark Rowe.
973 Revert r39226 / Bug 22818: Unify JIT callback argument access OS X / Windows
974 This causes Acid3 failures – reverting for now & will revisit later.
975 https://bugs.webkit.org/show_bug.cgi?id=22873
977 * interpreter/Interpreter.h:
979 (JSC::JIT::privateCompileCTIMachineTrampolines):
981 * jit/JITInlineMethods.h:
982 (JSC::JIT::restoreArgumentReference):
983 (JSC::JIT::restoreArgumentReferenceForTrampoline):
984 (JSC::JIT::emitCTICall_internal):
985 * jit/JITPropertyAccess.cpp:
986 (JSC::JIT::privateCompilePutByIdTransition):
989 2008-12-15 Darin Adler <darin@apple.com>
991 Reviewed by Sam Weinig.
993 - fix <rdar://problem/6427048> crash due to infinite recursion after setting window.__proto__ = window
995 Replaced toGlobalObject with the more generally useful unwrappedObject and used it to
996 fix the cycle detection code in put(__proto__).
998 * JavaScriptCore.exp: Updated.
1000 * runtime/JSGlobalObject.cpp: Removed toGlobalObject. We now use unwrappedObject instead.
1001 * runtime/JSGlobalObject.h:
1002 (JSC::JSGlobalObject::isGlobalObject): Ditto.
1004 * runtime/JSGlobalObjectFunctions.cpp:
1005 (JSC::globalFuncEval): Use unwrappedObject and isGlobalObject here rather than toGlobalObject.
1007 * runtime/JSObject.cpp:
1008 (JSC::JSObject::put): Rewrote prototype cycle checking loop. Use unwrappedObject in the loop now.
1009 (JSC::JSObject::unwrappedObject): Replaced toGlobalObject with this new function.
1010 * runtime/JSObject.h: More of the same.
1012 2008-12-15 Steve Falkenburg <sfalken@apple.com>
1016 Visual Studio requires visibility of forward declarations to match class declaration.
1018 * assembler/X86Assembler.h:
1020 2008-12-15 Gustavo Noronha Silva <kov@kov.eti.br>
1022 Reviewed by Mark Rowe.
1024 https://bugs.webkit.org/show_bug.cgi?id=22686
1030 2008-12-15 Gavin Barraclough <barraclough@apple.com>
1032 Reviewed by Geoff Garen.
1034 Add support to X86Assembler emitting instructions that access all 16 registers on x86-64.
1035 Add a new formating class, that is reponsible for both emitting the opcode bytes and the
1036 ModRm bytes of an instruction in a single call; this can insert the REX byte as necessary
1037 before the opcode, but has access to the register numbers to build the REX.
1039 * assembler/AssemblerBuffer.h:
1040 (JSC::AssemblerBuffer::isAligned):
1041 (JSC::AssemblerBuffer::data):
1042 * assembler/MacroAssembler.h:
1043 (JSC::MacroAssembler::addPtr):
1044 (JSC::MacroAssembler::add32):
1045 (JSC::MacroAssembler::and32):
1046 (JSC::MacroAssembler::or32):
1047 (JSC::MacroAssembler::sub32):
1048 (JSC::MacroAssembler::xor32):
1049 (JSC::MacroAssembler::loadPtr):
1050 (JSC::MacroAssembler::load32):
1051 (JSC::MacroAssembler::load16):
1052 (JSC::MacroAssembler::storePtr):
1053 (JSC::MacroAssembler::storePtrWithRepatch):
1054 (JSC::MacroAssembler::store32):
1055 (JSC::MacroAssembler::pop):
1056 (JSC::MacroAssembler::push):
1057 (JSC::MacroAssembler::compareImm32ForBranch):
1058 (JSC::MacroAssembler::compareImm32ForBranchEquality):
1059 (JSC::MacroAssembler::testImm32):
1060 (JSC::MacroAssembler::jae32):
1061 (JSC::MacroAssembler::jb32):
1062 (JSC::MacroAssembler::je16):
1063 (JSC::MacroAssembler::jg32):
1064 (JSC::MacroAssembler::jnePtr):
1065 (JSC::MacroAssembler::jne32):
1066 (JSC::MacroAssembler::jump):
1067 * assembler/X86Assembler.h:
1069 (JSC::X86Assembler::):
1070 (JSC::X86Assembler::size):
1071 (JSC::X86Assembler::push_r):
1072 (JSC::X86Assembler::pop_r):
1073 (JSC::X86Assembler::push_i32):
1074 (JSC::X86Assembler::push_m):
1075 (JSC::X86Assembler::pop_m):
1076 (JSC::X86Assembler::addl_rr):
1077 (JSC::X86Assembler::addl_mr):
1078 (JSC::X86Assembler::addl_ir):
1079 (JSC::X86Assembler::addq_ir):
1080 (JSC::X86Assembler::addl_im):
1081 (JSC::X86Assembler::andl_rr):
1082 (JSC::X86Assembler::andl_ir):
1083 (JSC::X86Assembler::orl_rr):
1084 (JSC::X86Assembler::orl_mr):
1085 (JSC::X86Assembler::orl_ir):
1086 (JSC::X86Assembler::subl_rr):
1087 (JSC::X86Assembler::subl_mr):
1088 (JSC::X86Assembler::subl_ir):
1089 (JSC::X86Assembler::subl_im):
1090 (JSC::X86Assembler::xorl_rr):
1091 (JSC::X86Assembler::xorl_ir):
1092 (JSC::X86Assembler::sarl_i8r):
1093 (JSC::X86Assembler::sarl_CLr):
1094 (JSC::X86Assembler::shll_i8r):
1095 (JSC::X86Assembler::shll_CLr):
1096 (JSC::X86Assembler::imull_rr):
1097 (JSC::X86Assembler::imull_i32r):
1098 (JSC::X86Assembler::idivl_r):
1099 (JSC::X86Assembler::cmpl_rr):
1100 (JSC::X86Assembler::cmpl_rm):
1101 (JSC::X86Assembler::cmpl_mr):
1102 (JSC::X86Assembler::cmpl_ir):
1103 (JSC::X86Assembler::cmpl_ir_force32):
1104 (JSC::X86Assembler::cmpl_im):
1105 (JSC::X86Assembler::cmpl_im_force32):
1106 (JSC::X86Assembler::cmpw_rm):
1107 (JSC::X86Assembler::testl_rr):
1108 (JSC::X86Assembler::testl_i32r):
1109 (JSC::X86Assembler::testl_i32m):
1110 (JSC::X86Assembler::testq_rr):
1111 (JSC::X86Assembler::testq_i32r):
1112 (JSC::X86Assembler::testb_i8r):
1113 (JSC::X86Assembler::sete_r):
1114 (JSC::X86Assembler::setz_r):
1115 (JSC::X86Assembler::setne_r):
1116 (JSC::X86Assembler::setnz_r):
1117 (JSC::X86Assembler::cdq):
1118 (JSC::X86Assembler::xchgl_rr):
1119 (JSC::X86Assembler::movl_rr):
1120 (JSC::X86Assembler::movl_rm):
1121 (JSC::X86Assembler::movl_mr):
1122 (JSC::X86Assembler::movl_i32r):
1123 (JSC::X86Assembler::movl_i32m):
1124 (JSC::X86Assembler::movq_rr):
1125 (JSC::X86Assembler::movq_rm):
1126 (JSC::X86Assembler::movq_mr):
1127 (JSC::X86Assembler::movzwl_mr):
1128 (JSC::X86Assembler::movzbl_rr):
1129 (JSC::X86Assembler::leal_mr):
1130 (JSC::X86Assembler::call):
1131 (JSC::X86Assembler::jmp):
1132 (JSC::X86Assembler::jmp_r):
1133 (JSC::X86Assembler::jmp_m):
1134 (JSC::X86Assembler::jne):
1135 (JSC::X86Assembler::jnz):
1136 (JSC::X86Assembler::je):
1137 (JSC::X86Assembler::jl):
1138 (JSC::X86Assembler::jb):
1139 (JSC::X86Assembler::jle):
1140 (JSC::X86Assembler::jbe):
1141 (JSC::X86Assembler::jge):
1142 (JSC::X86Assembler::jg):
1143 (JSC::X86Assembler::ja):
1144 (JSC::X86Assembler::jae):
1145 (JSC::X86Assembler::jo):
1146 (JSC::X86Assembler::jp):
1147 (JSC::X86Assembler::js):
1148 (JSC::X86Assembler::addsd_rr):
1149 (JSC::X86Assembler::addsd_mr):
1150 (JSC::X86Assembler::cvtsi2sd_rr):
1151 (JSC::X86Assembler::cvttsd2si_rr):
1152 (JSC::X86Assembler::movd_rr):
1153 (JSC::X86Assembler::movsd_rm):
1154 (JSC::X86Assembler::movsd_mr):
1155 (JSC::X86Assembler::mulsd_rr):
1156 (JSC::X86Assembler::mulsd_mr):
1157 (JSC::X86Assembler::pextrw_irr):
1158 (JSC::X86Assembler::subsd_rr):
1159 (JSC::X86Assembler::subsd_mr):
1160 (JSC::X86Assembler::ucomis_rr):
1161 (JSC::X86Assembler::int3):
1162 (JSC::X86Assembler::ret):
1163 (JSC::X86Assembler::predictNotTaken):
1164 (JSC::X86Assembler::label):
1165 (JSC::X86Assembler::align):
1166 (JSC::X86Assembler::link):
1167 (JSC::X86Assembler::executableCopy):
1168 (JSC::X86Assembler::X86InstructionFormater::prefix):
1169 (JSC::X86Assembler::X86InstructionFormater::oneByteOp):
1170 (JSC::X86Assembler::X86InstructionFormater::twoByteOp):
1171 (JSC::X86Assembler::X86InstructionFormater::oneByteOp64):
1172 (JSC::X86Assembler::X86InstructionFormater::oneByteOp8):
1173 (JSC::X86Assembler::X86InstructionFormater::twoByteOp8):
1174 (JSC::X86Assembler::X86InstructionFormater::instructionImmediate8):
1175 (JSC::X86Assembler::X86InstructionFormater::instructionImmediate32):
1176 (JSC::X86Assembler::X86InstructionFormater::instructionRel32):
1177 (JSC::X86Assembler::X86InstructionFormater::size):
1178 (JSC::X86Assembler::X86InstructionFormater::isAligned):
1179 (JSC::X86Assembler::X86InstructionFormater::data):
1180 (JSC::X86Assembler::X86InstructionFormater::executableCopy):
1181 (JSC::X86Assembler::X86InstructionFormater::registerModRM):
1182 (JSC::X86Assembler::X86InstructionFormater::memoryModRM):
1184 (JSC::JIT::privateCompileMainPass):
1185 (JSC::JIT::privateCompile):
1186 (JSC::JIT::privateCompileCTIMachineTrampolines):
1187 * jit/JITArithmetic.cpp:
1188 (JSC::JIT::putDoubleResultToJSNumberCellOrJSImmediate):
1189 (JSC::JIT::compileBinaryArithOp):
1191 (JSC::JIT::compileOpCall):
1192 (JSC::JIT::compileOpCallSlowCase):
1193 * jit/JITPropertyAccess.cpp:
1194 (JSC::JIT::compileGetByIdHotPath):
1195 (JSC::JIT::compilePutByIdHotPath):
1196 (JSC::JIT::privateCompilePutByIdTransition):
1197 (JSC::JIT::privateCompilePatchGetArrayLength):
1198 (JSC::JIT::privateCompileGetByIdProto):
1199 (JSC::JIT::privateCompileGetByIdProtoList):
1200 (JSC::JIT::privateCompileGetByIdChainList):
1201 (JSC::JIT::privateCompileGetByIdChain):
1203 2008-12-15 Darin Adler <darin@apple.com>
1205 * interpreter/RegisterFile.h: Tweak include formatting.
1207 2008-12-15 Holger Hans Peter Freyther <zecke@selfish.org>
1211 * interpreter/RegisterFile.h: Include stdio.h for fprintf
1213 2008-12-15 Alexey Proskuryakov <ap@webkit.org>
1215 Reviewed by Oliver Hunt.
1217 <rdar://problem/6444455> Worker Thread crash running multiple workers for a moderate amount of time
1219 * interpreter/RegisterFile.h: (JSC::RegisterFile::RegisterFile):
1220 Improve error handling: if mmap fails, crash immediately, and print out the reason.
1222 2008-12-13 Gavin Barraclough <barraclough@apple.com>
1224 Reviewed by Cameron Zwarich.
1226 Re-enable WREC on 64-bit.
1227 Implements one of the MacroAssembler::jnzPtr methods, previously only implemented for 32-bit x86.
1229 https://bugs.webkit.org/show_bug.cgi?id=22849
1231 * assembler/MacroAssembler.h:
1232 (JSC::MacroAssembler::testImm64):
1233 (JSC::MacroAssembler::jnzPtr):
1234 * assembler/X86Assembler.h:
1235 (JSC::X86Assembler::testq_i32r):
1236 (JSC::X86Assembler::testq_rr):
1239 2008-12-13 Gavin Barraclough <barraclough@apple.com>
1243 * assembler/MacroAssembler.h:
1245 2008-12-13 Gavin Barraclough <barraclough@apple.com>
1247 Build fix only, no review.
1249 * bytecode/CodeBlock.h:
1251 2008-12-13 Gavin Barraclough <barraclough@apple.com>
1253 Reviewed by Cameron Zwarich.
1255 Port the remainder of the JIT, bar calling convention related code, and code
1256 implementing optimizations which can be disabled, to use the MacroAssembler.
1258 * assembler/MacroAssembler.h:
1259 (JSC::MacroAssembler::DataLabelPtr::DataLabelPtr):
1260 (JSC::MacroAssembler::RepatchBuffer::RepatchBuffer):
1261 (JSC::MacroAssembler::RepatchBuffer::link):
1262 (JSC::MacroAssembler::RepatchBuffer::addressOf):
1263 (JSC::MacroAssembler::RepatchBuffer::setPtr):
1264 (JSC::MacroAssembler::addPtr):
1265 (JSC::MacroAssembler::lshift32):
1266 (JSC::MacroAssembler::mod32):
1267 (JSC::MacroAssembler::rshift32):
1268 (JSC::MacroAssembler::storePtrWithRepatch):
1269 (JSC::MacroAssembler::jnzPtr):
1270 (JSC::MacroAssembler::jzPtr):
1271 (JSC::MacroAssembler::jump):
1272 (JSC::MacroAssembler::label):
1273 * assembler/X86Assembler.h:
1274 (JSC::X86Assembler::):
1275 (JSC::X86Assembler::xchgl_rr):
1276 (JSC::X86Assembler::jmp_m):
1277 (JSC::X86Assembler::repatchAddress):
1278 (JSC::X86Assembler::getRelocatedAddress):
1279 * bytecode/CodeBlock.cpp:
1280 (JSC::CodeBlock::CodeBlock):
1281 * bytecode/CodeBlock.h:
1282 (JSC::JITCodeRef::JITCodeRef):
1283 (JSC::CodeBlock::setJITCode):
1284 (JSC::CodeBlock::jitCode):
1285 (JSC::CodeBlock::executablePool):
1287 (JSC::JIT::privateCompileMainPass):
1288 (JSC::JIT::privateCompileLinkPass):
1289 (JSC::JIT::privateCompile):
1290 (JSC::JIT::privateCompileCTIMachineTrampolines):
1292 (JSC::CallRecord::CallRecord):
1293 (JSC::JumpTable::JumpTable):
1294 (JSC::JIT::emitCTICall):
1295 (JSC::JIT::JSRInfo::JSRInfo):
1296 * jit/JITArithmetic.cpp:
1298 * jit/JITInlineMethods.h:
1299 (JSC::JIT::emitNakedCall):
1300 (JSC::JIT::emitCTICall_internal):
1301 (JSC::JIT::checkStructure):
1302 (JSC::JIT::emitFastArithDeTagImmediateJumpIfZero):
1303 (JSC::JIT::addSlowCase):
1304 (JSC::JIT::addJump):
1305 (JSC::JIT::emitJumpSlowToHot):
1306 * jit/JITPropertyAccess.cpp:
1307 (JSC::JIT::privateCompileGetByIdChainList):
1308 (JSC::JIT::privateCompileGetByIdChain):
1310 2008-12-12 Cameron Zwarich <zwarich@apple.com>
1312 Reviewed by Sam Weinig.
1314 Fix the failures of the following layout tests, which regressed in
1317 fast/dom/StyleSheet/ownerNode-lifetime-2.html
1318 fast/xsl/transform-xhr-doc.xhtml
1320 The binary search in CodeBlock::getByIdExceptionInfoForBytecodeOffset()
1321 doesn't guarantee that it actually finds a match, so add an explicit check
1324 * bytecode/CodeBlock.cpp:
1325 (JSC::CodeBlock::getByIdExceptionInfoForBytecodeOffset):
1327 2008-12-12 Gavin Barraclough <barraclough@apple.com>
1329 Reviewed by Cameron Zwarich.
1331 Replace emitPutCallArg methods with emitPutJITStubArg methods. Primarily to make the argument numbering
1332 more sensible (1-based incrementing by 1, rather than 0-based incrementing by 4). The CTI name also seems
1333 to be being deprecated from the code generally.
1336 (JSC::JIT::privateCompileMainPass):
1337 (JSC::JIT::privateCompileSlowCases):
1338 (JSC::JIT::privateCompileCTIMachineTrampolines):
1340 * jit/JITArithmetic.cpp:
1341 (JSC::JIT::compileBinaryArithOp):
1342 (JSC::JIT::compileBinaryArithOpSlowCase):
1344 (JSC::JIT::compileOpCallSetupArgs):
1345 (JSC::JIT::compileOpCallEvalSetupArgs):
1346 (JSC::JIT::compileOpConstructSetupArgs):
1347 (JSC::JIT::compileOpCall):
1348 * jit/JITInlineMethods.h:
1349 (JSC::JIT::emitPutJITStubArg):
1350 (JSC::JIT::emitPutJITStubArgConstant):
1351 (JSC::JIT::emitGetJITStubArg):
1352 (JSC::JIT::emitPutJITStubArgFromVirtualRegister):
1353 * jit/JITPropertyAccess.cpp:
1354 (JSC::JIT::compileGetByIdHotPath):
1355 (JSC::JIT::compilePutByIdHotPath):
1356 (JSC::JIT::compileGetByIdSlowCase):
1357 (JSC::JIT::compilePutByIdSlowCase):
1359 2008-12-12 Gavin Barraclough <barraclough@apple.com>
1364 (JSC::JIT::privateCompileMainPass):
1365 (JSC::JIT::privateCompileSlowCases):
1366 (JSC::JIT::privateCompile):
1368 2008-12-12 Gavin Barraclough <barraclough@apple.com>
1370 Reviewed by Geoff Garen.
1372 Remove loop counter 'i' from the JIT generation passes, replace with a member m_bytecodeIndex.
1374 No impact on performance.
1377 (JSC::JIT::compileOpStrictEq):
1378 (JSC::JIT::emitSlowScriptCheck):
1379 (JSC::JIT::privateCompileMainPass):
1380 (JSC::JIT::privateCompileSlowCases):
1381 (JSC::JIT::privateCompile):
1383 (JSC::CallRecord::CallRecord):
1384 (JSC::JmpTable::JmpTable):
1385 (JSC::JIT::emitCTICall):
1386 * jit/JITArithmetic.cpp:
1387 (JSC::JIT::compileBinaryArithOp):
1388 (JSC::JIT::compileBinaryArithOpSlowCase):
1390 (JSC::JIT::compileOpCall):
1391 (JSC::JIT::compileOpCallSlowCase):
1392 * jit/JITInlineMethods.h:
1393 (JSC::JIT::emitGetVirtualRegister):
1394 (JSC::JIT::emitGetVirtualRegisters):
1395 (JSC::JIT::emitNakedCall):
1396 (JSC::JIT::emitCTICall_internal):
1397 (JSC::JIT::emitJumpSlowCaseIfJSCell):
1398 (JSC::JIT::emitJumpSlowCaseIfNotJSCell):
1399 (JSC::JIT::emitJumpSlowCaseIfNotImmNum):
1400 (JSC::JIT::emitJumpSlowCaseIfNotImmNums):
1401 (JSC::JIT::emitFastArithIntToImmOrSlowCase):
1402 (JSC::JIT::addSlowCase):
1403 (JSC::JIT::addJump):
1404 (JSC::JIT::emitJumpSlowToHot):
1405 * jit/JITPropertyAccess.cpp:
1406 (JSC::JIT::compileGetByIdHotPath):
1407 (JSC::JIT::compileGetByIdSlowCase):
1408 (JSC::JIT::compilePutByIdHotPath):
1409 (JSC::JIT::compilePutByIdSlowCase):
1411 2008-12-12 Sam Weinig <sam@webkit.org>
1413 Reviewed by Cameron Zwarich.
1415 <rdar://problem/6428342> Look into feasibility of discarding bytecode after native codegen
1417 Move more JIT functionality to using offsets into the Instruction buffer
1418 instead of raw pointers. Two to go!
1420 * interpreter/Interpreter.cpp:
1421 (JSC::bytecodeOffsetForPC): Rename from vPCForPC.
1422 (JSC::Interpreter::resolve): Pass offset to exception helper.
1423 (JSC::Interpreter::resolveSkip): Ditto.
1424 (JSC::Interpreter::resolveGlobal): Ditto.
1425 (JSC::Interpreter::resolveBaseAndProperty): Ditto.
1426 (JSC::Interpreter::resolveBaseAndFunc): Ditto.
1427 (JSC::isNotObject): Ditto.
1428 (JSC::Interpreter::unwindCallFrame): Call bytecodeOffsetForPC.
1429 (JSC::Interpreter::throwException): Use offsets instead of vPCs.
1430 (JSC::Interpreter::privateExecute): Pass offset to exception helper.
1431 (JSC::Interpreter::retrieveLastCaller): Ditto.
1432 (JSC::Interpreter::cti_op_instanceof): Ditto.
1433 (JSC::Interpreter::cti_op_call_NotJSFunction): Ditto.
1434 (JSC::Interpreter::cti_op_resolve): Pass offset to exception helper.
1435 (JSC::Interpreter::cti_op_construct_NotJSConstruct): Ditto.
1436 (JSC::Interpreter::cti_op_resolve_func): Ditto.
1437 (JSC::Interpreter::cti_op_resolve_skip): Ditto.
1438 (JSC::Interpreter::cti_op_resolve_global): Ditto.
1439 (JSC::Interpreter::cti_op_resolve_with_base): Ditto.
1440 (JSC::Interpreter::cti_op_throw): Ditto.
1441 (JSC::Interpreter::cti_op_in): Ditto.
1442 (JSC::Interpreter::cti_vm_throw): Ditto.
1443 * interpreter/Interpreter.h:
1446 (JSC::JIT::privateCompileMainPass): Don't pass unnecessary vPC to stub.
1447 * jit/JIT.h: Remove ARG_instr1 - ARG_instr3 and ARG_instr5 - ARG_instr6.
1449 (JSC::JIT::compileOpCallEvalSetupArgs): Don't pass unnecessary vPC to stub..
1450 (JSC::JIT::compileOpConstructSetupArgs): Ditto.
1452 * runtime/ExceptionHelpers.cpp:
1453 (JSC::createUndefinedVariableError): Take an offset instead of vPC.
1454 (JSC::createInvalidParamError): Ditto.
1455 (JSC::createNotAConstructorError): Ditto.
1456 (JSC::createNotAFunctionError): Ditto.
1457 (JSC::createNotAnObjectError): Ditto.
1458 * runtime/ExceptionHelpers.h:
1460 2008-12-12 Cameron Zwarich <zwarich@apple.com>
1462 Reviewed by Oliver Hunt.
1464 Bug 22835: Crash during bytecode generation when comparing to null
1465 <https://bugs.webkit.org/show_bug.cgi?id=22835>
1466 <rdar://problem/6286749>
1468 Change the special cases in bytecode generation for comparison to null
1469 to use tempDestination().
1472 (JSC::BinaryOpNode::emitBytecode):
1473 (JSC::EqualNode::emitBytecode):
1475 2008-12-12 Gavin Barraclough <barraclough@apple.com>
1477 Reviewed by Geoff Garen.
1479 Move slow-cases of JIT code generation over to the MacroAssembler interface.
1481 * assembler/MacroAssembler.h:
1482 (JSC::MacroAssembler::Label::Label):
1483 (JSC::MacroAssembler::jae32):
1484 (JSC::MacroAssembler::jg32):
1485 (JSC::MacroAssembler::jzPtr):
1487 (JSC::JIT::privateCompileSlowCases):
1488 (JSC::JIT::privateCompile):
1489 (JSC::JIT::emitGetVariableObjectRegister):
1490 (JSC::JIT::emitPutVariableObjectRegister):
1492 (JSC::SlowCaseEntry::SlowCaseEntry):
1493 (JSC::JIT::getSlowCase):
1494 (JSC::JIT::linkSlowCase):
1495 * jit/JITArithmetic.cpp:
1496 (JSC::JIT::compileBinaryArithOpSlowCase):
1498 (JSC::JIT::compileOpCallInitializeCallFrame):
1499 (JSC::JIT::compileOpCall):
1500 (JSC::JIT::compileOpCallSlowCase):
1501 * jit/JITInlineMethods.h:
1502 (JSC::JIT::emitJumpSlowCaseIfNotJSCell):
1503 (JSC::JIT::linkSlowCaseIfNotJSCell):
1504 * jit/JITPropertyAccess.cpp:
1505 (JSC::JIT::compileGetByIdHotPath):
1506 (JSC::JIT::compilePutByIdHotPath):
1507 (JSC::JIT::compileGetByIdSlowCase):
1508 (JSC::JIT::compilePutByIdSlowCase):
1510 2008-12-12 Cameron Zwarich <zwarich@apple.com>
1512 Reviewed by Sam Weinig.
1514 Bug 22828: Do not inspect bytecode instruction stream for op_get_by_id exception information
1515 <https://bugs.webkit.org/show_bug.cgi?id=22828>
1517 In order to remove the bytecode instruction stream after generating
1518 native code, all inspection of bytecode instructions at runtime must
1519 be removed. One particular instance of this is the special handling of
1520 exceptions thrown by the op_get_by_id emitted directly before an
1521 op_construct or an op_instanceof. This patch moves that information to
1522 an auxiliary data structure in CodeBlock.
1524 * bytecode/CodeBlock.cpp:
1525 (JSC::CodeBlock::getByIdExceptionInfoForBytecodeOffset):
1526 * bytecode/CodeBlock.h:
1527 (JSC::CodeBlock::addGetByIdExceptionInfo):
1528 * bytecompiler/BytecodeGenerator.cpp:
1529 (JSC::BytecodeGenerator::emitConstruct):
1530 * bytecompiler/BytecodeGenerator.h:
1531 (JSC::BytecodeGenerator::emitGetByIdExceptionInfo):
1533 (JSC::InstanceOfNode::emitBytecode):
1534 * runtime/ExceptionHelpers.cpp:
1535 (JSC::createNotAnObjectError):
1537 2008-12-12 Sam Weinig <sam@webkit.org>
1539 Reviewed by Geoffrey Garen.
1541 Change exception information accessors to take offsets into the bytecode
1542 instruction buffer instead of pointers so that they can work even even
1543 if the bytecode buffer is purged.
1545 * bytecode/CodeBlock.cpp:
1546 (JSC::instructionOffsetForNth):
1547 (JSC::CodeBlock::handlerForBytecodeOffset):
1548 (JSC::CodeBlock::lineNumberForBytecodeOffset):
1549 (JSC::CodeBlock::expressionRangeForBytecodeOffset):
1550 * bytecode/CodeBlock.h:
1551 * bytecode/SamplingTool.cpp:
1552 (JSC::SamplingTool::dump):
1553 * interpreter/Interpreter.cpp:
1554 (JSC::Interpreter::throwException):
1555 (JSC::Interpreter::privateExecute):
1556 (JSC::Interpreter::retrieveLastCaller):
1558 (JSC::JIT::privateCompileMainPass):
1559 * runtime/ExceptionHelpers.cpp:
1560 (JSC::createUndefinedVariableError):
1561 (JSC::createInvalidParamError):
1562 (JSC::createNotAConstructorError):
1563 (JSC::createNotAFunctionError):
1564 (JSC::createNotAnObjectError):
1566 2008-12-12 Geoffrey Garen <ggaren@apple.com>
1568 Reviewed by Cameron Zwarich.
1570 Tiny bit of refactoring in quantifier generation.
1572 * wrec/WRECGenerator.cpp:
1573 (JSC::WREC::Generator::generateNonGreedyQuantifier):
1574 (JSC::WREC::Generator::generateGreedyQuantifier):
1576 2008-12-11 Sam Weinig <sam@webkit.org>
1578 Reviewed by Geoffrey Garen.
1580 Remove dependancy on having the Instruction buffer in order to
1581 deref Structures used for property access and global resolves.
1582 Instead, we put references to the necessary Structures in auxiliary
1583 data structures on the CodeBlock. This is not an ideal solution,
1584 as we still pay for having the Structures in two places and we
1585 would like to eventually just hold on to offsets into the machine
1588 - Also removes CodeBlock bloat in non-JIT by #ifdefing the JIT
1589 only data structures.
1592 * JavaScriptCore.pri:
1593 * JavaScriptCore.scons:
1594 * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
1595 * JavaScriptCore.xcodeproj/project.pbxproj:
1596 * JavaScriptCoreSources.bkl:
1597 * bytecode/CodeBlock.cpp:
1598 (JSC::isGlobalResolve):
1599 (JSC::isPropertyAccess):
1600 (JSC::instructionOffsetForNth):
1601 (JSC::printGlobalResolveInfo):
1602 (JSC::printStructureStubInfo):
1603 (JSC::CodeBlock::printStructures):
1604 (JSC::CodeBlock::dump):
1605 (JSC::CodeBlock::~CodeBlock):
1606 (JSC::CodeBlock::shrinkToFit):
1607 * bytecode/CodeBlock.h:
1608 (JSC::GlobalResolveInfo::GlobalResolveInfo):
1610 (JSC::CodeBlock::instructions):
1611 (JSC::CodeBlock::getStubInfo):
1612 (JSC::CodeBlock::getBytecodeIndex):
1613 (JSC::CodeBlock::addPropertyAccessInstruction):
1614 (JSC::CodeBlock::addGlobalResolveInstruction):
1615 (JSC::CodeBlock::numberOfStructureStubInfos):
1616 (JSC::CodeBlock::addStructureStubInfo):
1617 (JSC::CodeBlock::structureStubInfo):
1618 (JSC::CodeBlock::addGlobalResolveInfo):
1619 (JSC::CodeBlock::globalResolveInfo):
1620 (JSC::CodeBlock::numberOfCallLinkInfos):
1621 (JSC::CodeBlock::addCallLinkInfo):
1622 (JSC::CodeBlock::callLinkInfo):
1623 * bytecode/Instruction.h:
1624 (JSC::PolymorphicAccessStructureList::PolymorphicStubInfo::set):
1625 (JSC::PolymorphicAccessStructureList::PolymorphicAccessStructureList):
1626 * bytecode/Opcode.h:
1628 * bytecode/StructureStubInfo.cpp: Copied from bytecode/CodeBlock.cpp.
1629 (JSC::StructureStubInfo::deref):
1630 * bytecode/StructureStubInfo.h: Copied from bytecode/CodeBlock.h.
1631 (JSC::StructureStubInfo::StructureStubInfo):
1632 (JSC::StructureStubInfo::initGetByIdSelf):
1633 (JSC::StructureStubInfo::initGetByIdProto):
1634 (JSC::StructureStubInfo::initGetByIdChain):
1635 (JSC::StructureStubInfo::initGetByIdSelfList):
1636 (JSC::StructureStubInfo::initGetByIdProtoList):
1637 (JSC::StructureStubInfo::initPutByIdTransition):
1638 (JSC::StructureStubInfo::initPutByIdReplace):
1639 (JSC::StructureStubInfo::):
1640 * bytecompiler/BytecodeGenerator.cpp:
1641 (JSC::BytecodeGenerator::emitResolve):
1642 (JSC::BytecodeGenerator::emitGetById):
1643 (JSC::BytecodeGenerator::emitPutById):
1644 (JSC::BytecodeGenerator::emitCall):
1645 (JSC::BytecodeGenerator::emitConstruct):
1646 (JSC::BytecodeGenerator::emitCatch):
1647 * interpreter/Interpreter.cpp:
1648 (JSC::Interpreter::tryCTICachePutByID):
1649 (JSC::Interpreter::tryCTICacheGetByID):
1650 (JSC::Interpreter::cti_op_get_by_id_self_fail):
1651 (JSC::getPolymorphicAccessStructureListSlot):
1652 (JSC::Interpreter::cti_op_get_by_id_proto_list):
1653 (JSC::Interpreter::cti_op_resolve_global):
1656 (JSC::JIT::privateCompileMainPass):
1657 (JSC::JIT::privateCompileSlowCases):
1658 (JSC::JIT::privateCompile):
1659 * jit/JITPropertyAccess.cpp:
1660 (JSC::JIT::compileGetByIdHotPath):
1661 (JSC::JIT::compilePutByIdHotPath):
1662 (JSC::JIT::compileGetByIdSlowCase):
1663 (JSC::JIT::compilePutByIdSlowCase):
1664 (JSC::JIT::privateCompileGetByIdSelfList):
1665 (JSC::JIT::privateCompileGetByIdProtoList):
1666 (JSC::JIT::privateCompileGetByIdChainList):
1668 2008-12-11 Gavin Barraclough <barraclough@apple.com>
1670 Reviewed by Oliver Hunt.
1672 Remove CTI_ARGUMENTS mode, use va_start implementation on Windows,
1673 unifying JIT callback (cti_*) argument access on OS X & Windows
1675 No performance impact.
1677 * interpreter/Interpreter.h:
1679 (JSC::JIT::privateCompileCTIMachineTrampolines):
1681 * jit/JITInlineMethods.h:
1682 (JSC::JIT::emitCTICall):
1683 * jit/JITPropertyAccess.cpp:
1684 (JSC::JIT::privateCompilePutByIdTransition):
1687 2008-12-11 Holger Freyther <zecke@selfish.org>
1689 Reviewed by Simon Hausmann.
1691 https://bugs.webkit.org/show_bug.cgi?id=20953
1693 For Qt it is not pratical to have a FontCache and GlyphPageTreeNode
1694 implementation. This is one of the reasons why the Qt port is currently not
1695 using WebCore/platform/graphics/Font.cpp. By allowing to not use
1696 the simple/fast-path the Qt port will be able to use it.
1698 Introduce USE(FONT_FAST_PATH) and define it for every port but the
1701 * wtf/Platform.h: Enable USE(FONT_FAST_PATH)
1703 2008-12-11 Gabor Loki <loki@inf.u-szeged.hu>
1705 Reviewed by Darin Adler and landed by Holger Freyther.
1707 <https://bugs.webkit.org/show_bug.cgi?id=22648>
1708 Fix threading on Qt-port and Gtk-port for Sampling tool.
1710 * wtf/ThreadingGtk.cpp:
1711 (WTF::waitForThreadCompletion):
1712 * wtf/ThreadingQt.cpp:
1713 (WTF::waitForThreadCompletion):
1715 2008-12-10 Cameron Zwarich <zwarich@apple.com>
1717 Reviewed by Oliver Hunt.
1719 Bug 22734: Debugger crashes when stepping into a function call in a return statement
1720 <https://bugs.webkit.org/show_bug.cgi?id=22734>
1721 <rdar://problem/6426796>
1723 * bytecompiler/BytecodeGenerator.cpp:
1724 (JSC::BytecodeGenerator::BytecodeGenerator): The DebuggerCallFrame uses
1725 the 'this' value stored in a callFrame, so op_convert_this should be
1726 emitted at the beginning of a function body when generating bytecode
1728 * debugger/DebuggerCallFrame.cpp:
1729 (JSC::DebuggerCallFrame::thisObject): The assertion inherent in the call
1730 to asObject() here is valid, because any 'this' value should have been
1731 converted to a JSObject*.
1733 2008-12-10 Gavin Barraclough <barraclough@apple.com>
1735 Reviewed by Geoff Garen.
1737 Port more of the JIT to use the MacroAssembler interface.
1739 Everything in the main pass, bar a few corner cases (operations with required
1740 registers, or calling convention code). Slightly refactors array creation,
1741 moving the offset calculation into the callFrame into C code (reducing code
1744 Overall this appears to be a 1% win on v8-tests, due to the smaller immediates
1745 being planted (in jfalse in particular).
1747 * interpreter/Interpreter.cpp:
1748 (JSC::Interpreter::cti_op_new_array):
1750 (JSC::JIT::privateCompileMainPass):
1751 (JSC::JIT::privateCompileSlowCases):
1753 * wrec/WRECGenerator.cpp:
1754 (JSC::WREC::Generator::generateEnter):
1756 2008-12-10 Sam Weinig <sam@webkit.org>
1760 * bytecode/CodeBlock.h:
1762 2008-12-10 Sam Weinig <sam@webkit.org>
1764 Reviewed by Geoffrey Garen.
1766 <rdar://problem/6428332> Remove the CTI return address table from CodeBlock
1770 Convert the return address table from a HashMap to a sorted Vector. This
1771 reduces the size of the data structure by ~4.5MB on Membuster head.
1773 SunSpider reports a 0.5% progression.
1775 * bytecode/CodeBlock.cpp:
1776 (JSC::sizeInBytes): Generic method to get the cost of a Vector.
1777 (JSC::CodeBlock::dumpStatistics): Add dumping of member sizes.
1778 * bytecode/CodeBlock.h:
1779 (JSC::PC::PC): Struct representing NativePC -> VirtualPC mappings.
1780 (JSC::getNativePC): Helper for binary chop.
1781 (JSC::CodeBlock::getBytecodeIndex): Used to get the VirtualPC from a
1782 NativePC using a binary chop of the pcVector.
1783 (JSC::CodeBlock::pcVector): Accessor.
1785 * interpreter/Interpreter.cpp:
1786 (JSC::vPCForPC): Use getBytecodeIndex instead of jitReturnAddressVPCMap().get().
1787 (JSC::Interpreter::cti_op_instanceof): Ditto.
1788 (JSC::Interpreter::cti_op_resolve): Ditto.
1789 (JSC::Interpreter::cti_op_resolve_func): Ditto.
1790 (JSC::Interpreter::cti_op_resolve_skip): Ditto.
1791 (JSC::Interpreter::cti_op_resolve_with_base): Ditto.
1792 (JSC::Interpreter::cti_op_throw): Ditto.
1793 (JSC::Interpreter::cti_op_in): Ditto.
1794 (JSC::Interpreter::cti_vm_throw): Ditto.
1797 (JSC::JIT::privateCompile): Reserve exact capacity and fill the pcVector.
1799 2008-12-09 Geoffrey Garen <ggaren@apple.com>
1801 Reviewed by Oliver Hunt.
1803 Added WREC support for an assertion followed by a quantifier. Fixed
1806 * wrec/WRECParser.cpp:
1807 (JSC::WREC::Parser::parseParentheses): Throw away the quantifier, since
1808 it's meaningless. (Firefox does the same.)
1810 * pcre/pcre_compile.cpp:
1811 (compileBranch): ditto.
1813 2008-12-09 Geoffrey Garen <ggaren@apple.com>
1815 Reviewed by Cameron Zwarich.
1817 In preparation for compiling WREC without PCRE:
1819 Further relaxed WREC's parsing to be more web-compatible. Fixed PCRE to
1820 match in cases where it didn't already.
1822 Changed JavaScriptCore to report syntax errors detected by WREC, rather
1823 than falling back on PCRE any time WREC sees an error.
1825 * pcre/pcre_compile.cpp:
1826 (checkEscape): Relaxed parsing of \c and \N escapes to be more
1829 * runtime/RegExp.cpp:
1830 (JSC::RegExp::RegExp): Only fall back on PCRE if WREC has not reported
1834 (JSC::WREC::Generator::compileRegExp): Fixed some error reporting to
1837 * wrec/WRECParser.cpp: Added error messages that match PCRE.
1839 (JSC::WREC::Parser::consumeGreedyQuantifier):
1840 (JSC::WREC::Parser::parseParentheses):
1841 (JSC::WREC::Parser::parseCharacterClass):
1842 (JSC::WREC::Parser::parseNonCharacterEscape): Updated the above functions to
1843 use the new setError API.
1845 (JSC::WREC::Parser::consumeEscape): Relaxed parsing of \c \N \u \x \B
1846 to be more web-compatible.
1848 (JSC::WREC::Parser::parseAlternative): Distinguish between a malformed
1849 quantifier and a quantifier with no prefix, like PCRE does.
1851 (JSC::WREC::Parser::consumeParenthesesType): Updated to use the new setError API.
1853 * wrec/WRECParser.h:
1854 (JSC::WREC::Parser::error):
1855 (JSC::WREC::Parser::syntaxError):
1856 (JSC::WREC::Parser::parsePattern):
1857 (JSC::WREC::Parser::reset):
1858 (JSC::WREC::Parser::setError): Store error messages instead of error codes,
1859 to provide for exception messages. Use a setter for reporting errors, so
1860 errors detected early are not overwritten by errors detected later.
1862 2008-12-09 Gavin Barraclough <barraclough@apple.com>
1864 Reviewed by Oliver Hunt.
1866 Use va_args to access cti function arguments.
1867 https://bugs.webkit.org/show_bug.cgi?id=22774
1869 This may be a minor regression, but we'll take the hit if so to reduce fragility.
1871 * interpreter/Interpreter.cpp:
1872 * interpreter/Interpreter.h:
1874 2008-12-09 Sam Weinig <sam@webkit.org>
1876 Reviewed twice by Cameron Zwarich.
1878 Fix for https://bugs.webkit.org/show_bug.cgi?id=22752
1879 Clear SymbolTable after codegen for Function codeblocks that
1880 don't require an activation
1882 This is a ~1.5MB improvement on Membuster-head.
1884 * bytecode/CodeBlock.cpp:
1885 (JSC::CodeBlock::dumpStatistics): Add logging of non-empty symbol tables
1886 and total size used by symbol tables.
1887 * bytecompiler/BytecodeGenerator.cpp:
1888 (JSC::BytecodeGenerator::generate): Clear the symbol table here.
1890 2008-12-09 Sam Weinig <sam@webkit.org>
1892 Reviewed by Geoffrey Garen.
1894 Remove unnecessary extra lookup when throwing an exception.
1895 We used to first lookup the target offset using getHandlerForVPC
1896 and then we would lookup the native code stub using
1897 nativeExceptionCodeForHandlerVPC. Instead, we can just pass around
1900 * bytecode/CodeBlock.cpp:
1901 (JSC::CodeBlock::handlerForVPC): Return the HandlerInfo.
1902 * bytecode/CodeBlock.h: Remove nativeExceptionCodeForHandlerVPC.
1904 * interpreter/Interpreter.cpp:
1905 (JSC::Interpreter::throwException): Return a HandlerInfo instead of
1906 and Instruction offset.
1907 (JSC::Interpreter::privateExecute): Get the offset from HandlerInfo.
1908 (JSC::Interpreter::cti_op_throw): Get the native code from the HandleInfo.
1909 (JSC::Interpreter::cti_vm_throw): Ditto.
1910 * interpreter/Interpreter.h:
1912 2008-12-09 Eric Seidel <eric@webkit.org>
1914 Build fix only, no review.
1916 Speculative fix for the Chromium-Windows bot.
1917 Add JavaScriptCore/os-win32 to the include path (for stdint.h)
1918 Strangely it builds fine on my local windows box (or at least doesn't hit this error)
1920 * JavaScriptCore.scons:
1922 2008-12-09 Eric Seidel <eric@webkit.org>
1924 No review, build fix only.
1926 Add ExecutableAllocator files missing from Scons build.
1928 * JavaScriptCore.scons:
1930 2008-12-09 Dimitri Glazkov <dglazkov@chromium.org>
1932 Reviewed by Timothy Hatcher.
1934 https://bugs.webkit.org/show_bug.cgi?id=22631
1935 Allow ScriptCallFrame query names of functions in the call stack.
1937 * JavaScriptCore.exp: added InternalFunction::name and
1938 UString operator==() as exported symbol
1940 2008-12-08 Judit Jasz <jasy@inf.u-szeged.hu>
1942 Reviewed and tweaked by Cameron Zwarich.
1944 Bug 22352: Annotate opcodes with their length
1945 <https://bugs.webkit.org/show_bug.cgi?id=22352>
1947 * bytecode/Opcode.cpp:
1948 * bytecode/Opcode.h:
1949 * interpreter/Interpreter.cpp:
1950 (JSC::Interpreter::privateExecute):
1952 (JSC::JIT::privateCompileMainPass):
1953 (JSC::JIT::privateCompileSlowCases):
1955 2008-12-08 Geoffrey Garen <ggaren@apple.com>
1957 Reviewed by Oliver Hunt.
1959 Implemented more of the relaxed and somewhat weird rules for deciding
1960 how to interpret a non-pattern-character.
1963 (JSC::WREC::Escape::):
1964 (JSC::WREC::Escape::Escape): Eliminated Escape::None because it was
1965 unused. If you see an '\\', it's either a valid escape or an error.
1967 * wrec/Quantifier.h:
1968 (JSC::WREC::Quantifier::Quantifier):
1969 * wrec/WRECGenerator.cpp:
1970 (JSC::WREC::Generator::generateNonGreedyQuantifier):
1971 (JSC::WREC::Generator::generateGreedyQuantifier): Renamed "noMaxSpecified"
1972 to "Infinity", since that's what it means.
1974 * wrec/WRECParser.cpp:
1975 (JSC::WREC::Parser::consumeGreedyQuantifier): Re-wrote {n,m} parsing rules
1976 because they were too strict before. Added support for backtracking
1977 in the case where the {n,m} fails to parse as a quantifier, and yet is
1980 (JSC::WREC::Parser::parseCharacterClass):
1981 (JSC::WREC::Parser::parseNonCharacterEscape): Eliminated Escape::None,
1984 (JSC::WREC::Parser::consumeEscape): Don't treat ASCII and _ escapes
1985 as syntax errors. See fast/regex/non-pattern-characters.html.
1987 * wrec/WRECParser.h:
1988 (JSC::WREC::Parser::SavedState::SavedState):
1989 (JSC::WREC::Parser::SavedState::restore): Added a state backtracker,
1990 since parsing {n,m} forms requires backtracking if the form turns out
1991 not to be a quantifier.
1993 2008-12-08 Geoffrey Garen <ggaren@apple.com>
1995 Reviewed by Oliver Hunt.
1997 Refactored WREC parsing so that only one piece of code needs to know
1998 the relaxed and somewhat weird rules for deciding how to interpret a
1999 non-pattern-character, in preparation for implementing those rules.
2001 Also, implemented the relaxed and somewhat weird rules for '}' and ']'.
2003 * wrec/WREC.cpp: Reduced the regular expression size limit. Now that
2004 WREC handles ']' properly, it compiles fast/js/regexp-charclass-crash.html,
2005 which makes it hang at the old limit. (The old limit was based on the
2006 misimpression that the same value in PCRE limited the regular expression
2007 pattern size; in reality, it limited the expected compiled regular
2008 expression size. WREC doesn't have a way to calculate an expected
2009 compiled regular expression size, but this should be good enough.)
2011 * wrec/WRECParser.cpp:
2012 (JSC::WREC::parsePatternCharacterSequence): Nixed this function because
2013 it contained a second copy of the logic for handling non-pattern-characters,
2014 which is about to get a lot more complicated.
2016 (JSC::WREC::PatternCharacterSequence::PatternCharacterSequence):
2017 (JSC::WREC::PatternCharacterSequence::size):
2018 (JSC::WREC::PatternCharacterSequence::append):
2019 (JSC::WREC::PatternCharacterSequence::flush): Helper object for generating
2020 an optimized sequence of pattern characters.
2022 (JSC::WREC::Parser::parseNonCharacterEscape): Renamed to reflect the fact
2023 that the main parseAlternative loop handles character escapes.
2025 (JSC::WREC::Parser::parseAlternative): Moved pattern character sequence
2026 logic from parsePatternCharacterSequence to here, using
2027 PatternCharacterSequence to help with the details.
2029 * wrec/WRECParser.h: Updated for renames.
2031 2008-12-08 Alexey Proskuryakov <ap@webkit.org>
2033 Reviewed by Geoff Garen.
2035 <rdar://problem/6166088> Give JSGlobalContextCreate a behavior that is concurrency aware,
2038 * API/JSContextRef.cpp: (JSGlobalContextCreate):
2039 * API/JSContextRef.h:
2040 Use a unique context group for the context, unless the application was linked against old
2043 2008-12-08 Sam Weinig <sam@webkit.org>
2045 Reviewed by Cameron Zwarich.
2047 Fix for <rdar://problem/6428332> Remove the CTI return address table from CodeBlock
2051 Remove use of jitReturnAddressVPCMap when looking for vPC to store Structures
2052 in for cached lookup. Instead, use the offset in the StructureStubInfo that is
2055 * bytecode/CodeBlock.cpp:
2056 (JSC::CodeBlock::dumpStatistics): Fix extraneous semicolon.
2057 * interpreter/Interpreter.cpp:
2058 (JSC::Interpreter::tryCTICachePutByID):
2059 (JSC::Interpreter::tryCTICacheGetByID):
2060 (JSC::Interpreter::cti_op_get_by_id_self_fail):
2061 (JSC::Interpreter::cti_op_get_by_id_proto_list):
2063 (JSC::JIT::compileGetByIdSelf):
2064 (JSC::JIT::compileGetByIdProto):
2065 (JSC::JIT::compileGetByIdChain):
2066 (JSC::JIT::compilePutByIdReplace):
2067 (JSC::JIT::compilePutByIdTransition):
2068 * jit/JITPropertyAccess.cpp:
2069 (JSC::JIT::privateCompilePutByIdTransition):
2070 (JSC::JIT::patchGetByIdSelf):
2071 (JSC::JIT::patchPutByIdReplace):
2072 (JSC::JIT::privateCompilePatchGetArrayLength): Remove extra call to getStubInfo.
2073 (JSC::JIT::privateCompileGetByIdSelf):
2074 (JSC::JIT::privateCompileGetByIdProto):
2075 (JSC::JIT::privateCompileGetByIdChain):
2076 (JSC::JIT::privateCompilePutByIdReplace):
2078 2008-12-08 Gavin Barraclough <barraclough@apple.com>
2080 Reviewed by Oliver Hunt.
2082 Port the op_j?n?eq_null JIT code generation to use the MacroAssembler,
2083 and clean up slightly at the same time. The 'j' forms currently compare,
2084 then set a register, then compare again, then branch. Branch directly on
2085 the result of the first compare.
2087 Around a 1% progression on deltablue, crypto & early boyer, for about 1/2%
2088 overall on v8-tests.
2091 (JSC::JIT::privateCompileMainPass):
2092 * jit/JITPropertyAccess.cpp:
2093 (JSC::JIT::compileGetByIdSlowCase):
2095 2008-12-08 Gavin Barraclough <barraclough@apple.com>
2097 Reviewed by Geoff Garen.
2099 Expand MacroAssembler to support more operations, required by the JIT.
2101 Generally adds more operations and permutations of operands to the existing
2102 interface. Rename 'jset' to 'jnz' and 'jnset' to 'jz', which seem clearer,
2103 and require that immediate pointer operands (though not pointer addresses to
2104 load and store instructions) are wrapped in a ImmPtr() type, akin to Imm32().
2106 No performance impact.
2108 * assembler/MacroAssembler.h:
2109 (JSC::MacroAssembler::):
2110 (JSC::MacroAssembler::ImmPtr::ImmPtr):
2111 (JSC::MacroAssembler::add32):
2112 (JSC::MacroAssembler::and32):
2113 (JSC::MacroAssembler::or32):
2114 (JSC::MacroAssembler::sub32):
2115 (JSC::MacroAssembler::xor32):
2116 (JSC::MacroAssembler::loadPtr):
2117 (JSC::MacroAssembler::load32):
2118 (JSC::MacroAssembler::storePtr):
2119 (JSC::MacroAssembler::store32):
2120 (JSC::MacroAssembler::poke):
2121 (JSC::MacroAssembler::move):
2122 (JSC::MacroAssembler::testImm32):
2123 (JSC::MacroAssembler::jae32):
2124 (JSC::MacroAssembler::jb32):
2125 (JSC::MacroAssembler::jePtr):
2126 (JSC::MacroAssembler::je32):
2127 (JSC::MacroAssembler::jnePtr):
2128 (JSC::MacroAssembler::jne32):
2129 (JSC::MacroAssembler::jnzPtr):
2130 (JSC::MacroAssembler::jnz32):
2131 (JSC::MacroAssembler::jzPtr):
2132 (JSC::MacroAssembler::jz32):
2133 (JSC::MacroAssembler::joSub32):
2134 (JSC::MacroAssembler::jump):
2135 (JSC::MacroAssembler::sete32):
2136 (JSC::MacroAssembler::setne32):
2137 (JSC::MacroAssembler::setnz32):
2138 (JSC::MacroAssembler::setz32):
2139 * assembler/X86Assembler.h:
2140 (JSC::X86Assembler::addl_mr):
2141 (JSC::X86Assembler::andl_i8r):
2142 (JSC::X86Assembler::cmpl_rm):
2143 (JSC::X86Assembler::cmpl_mr):
2144 (JSC::X86Assembler::cmpl_i8m):
2145 (JSC::X86Assembler::subl_mr):
2146 (JSC::X86Assembler::testl_i32m):
2147 (JSC::X86Assembler::xorl_i32r):
2148 (JSC::X86Assembler::movl_rm):
2149 (JSC::X86Assembler::modRm_opmsib):
2151 (JSC::JIT::privateCompileMainPass):
2152 * jit/JITInlineMethods.h:
2153 (JSC::JIT::emitGetVirtualRegister):
2154 (JSC::JIT::emitPutCTIArgConstant):
2155 (JSC::JIT::emitPutCTIParam):
2156 (JSC::JIT::emitPutImmediateToCallFrameHeader):
2157 (JSC::JIT::emitInitRegister):
2158 (JSC::JIT::checkStructure):
2159 (JSC::JIT::emitJumpIfJSCell):
2160 (JSC::JIT::emitJumpIfNotJSCell):
2161 (JSC::JIT::emitJumpSlowCaseIfNotImmNum):
2163 2008-12-08 Geoffrey Garen <ggaren@apple.com>
2165 Reviewed by Sam Weinig.
2167 Fixed a bug where WREC would allow a quantifier whose minimum was
2168 greater than its maximum.
2170 * wrec/Quantifier.h:
2171 (JSC::WREC::Quantifier::Quantifier): ASSERT that the quantifier is not
2174 * wrec/WRECParser.cpp:
2175 (JSC::WREC::Parser::consumeGreedyQuantifier): Verify that the minimum
2176 is not greater than the maximum.
2178 2008-12-08 Eric Seidel <eric@webkit.org>
2180 Build fix only, no review.
2182 * JavaScriptCore.scons: add bytecode/JumpTable.cpp
2184 2008-12-08 Sam Weinig <sam@webkit.org>
2186 Reviewed by Geoffrey Garen.
2188 Patch for https://bugs.webkit.org/show_bug.cgi?id=22716
2189 <rdar://problem/6428315>
2190 Add RareData structure to CodeBlock for infrequently used auxiliary data
2193 Reduces memory on Membuster-head by ~.5MB
2195 * bytecode/CodeBlock.cpp:
2196 (JSC::CodeBlock::dump):
2197 (JSC::CodeBlock::dumpStatistics):
2198 (JSC::CodeBlock::mark):
2199 (JSC::CodeBlock::getHandlerForVPC):
2200 (JSC::CodeBlock::nativeExceptionCodeForHandlerVPC):
2201 (JSC::CodeBlock::shrinkToFit):
2202 * bytecode/CodeBlock.h:
2203 (JSC::CodeBlock::numberOfExceptionHandlers):
2204 (JSC::CodeBlock::addExceptionHandler):
2205 (JSC::CodeBlock::exceptionHandler):
2206 (JSC::CodeBlock::addFunction):
2207 (JSC::CodeBlock::function):
2208 (JSC::CodeBlock::addUnexpectedConstant):
2209 (JSC::CodeBlock::unexpectedConstant):
2210 (JSC::CodeBlock::addRegExp):
2211 (JSC::CodeBlock::regexp):
2212 (JSC::CodeBlock::numberOfImmediateSwitchJumpTables):
2213 (JSC::CodeBlock::addImmediateSwitchJumpTable):
2214 (JSC::CodeBlock::immediateSwitchJumpTable):
2215 (JSC::CodeBlock::numberOfCharacterSwitchJumpTables):
2216 (JSC::CodeBlock::addCharacterSwitchJumpTable):
2217 (JSC::CodeBlock::characterSwitchJumpTable):
2218 (JSC::CodeBlock::numberOfStringSwitchJumpTables):
2219 (JSC::CodeBlock::addStringSwitchJumpTable):
2220 (JSC::CodeBlock::stringSwitchJumpTable):
2221 (JSC::CodeBlock::evalCodeCache):
2222 (JSC::CodeBlock::createRareDataIfNecessary):
2224 2008-11-26 Peter Kasting <pkasting@google.com>
2226 Reviewed by Anders Carlsson.
2228 https://bugs.webkit.org/show_bug.cgi?id=16814
2229 Allow ports to disable ActiveX->NPAPI conversion for Media Player.
2230 Improve handling of miscellaneous ActiveX objects.
2232 * wtf/Platform.h: Add another ENABLE(...).
2234 2008-12-08 Sam Weinig <sam@webkit.org>
2236 Reviewed by Mark Rowe.
2238 Add dumping of CodeBlock member structure usage.
2240 * bytecode/CodeBlock.cpp:
2241 (JSC::CodeBlock::dumpStatistics):
2242 * bytecode/EvalCodeCache.h:
2243 (JSC::EvalCodeCache::isEmpty):
2245 2008-12-08 David Kilzer <ddkilzer@apple.com>
2247 Bug 22555: Sort "children" sections in Xcode project files
2249 <https://bugs.webkit.org/show_bug.cgi?id=22555>
2251 Reviewed by Eric Seidel.
2253 * JavaScriptCore.xcodeproj/project.pbxproj: Sorted.
2255 2008-12-08 Tony Chang <tony@chromium.org>
2257 Reviewed by Eric Seidel.
2259 Enable Pan scrolling only when building on PLATFORM(WIN_OS)
2260 Previously platforms like Apple Windows WebKit, Cairo Windows WebKit,
2261 Wx and Chromium were enabling it explicitly, now we just turn it on
2262 for all WIN_OS, later platforms can turn it off as needed on Windows
2263 (or turn it on under Linux, etc.)
2264 https://bugs.webkit.org/show_bug.cgi?id=22698
2268 2008-12-08 Sam Weinig <sam@webkit.org>
2270 Reviewed by Cameron Zwarich.
2272 Add basic memory statistics dumping for CodeBlock.
2274 * bytecode/CodeBlock.cpp:
2275 (JSC::CodeBlock::dumpStatistics):
2276 (JSC::CodeBlock::CodeBlock):
2277 (JSC::CodeBlock::~CodeBlock):
2278 * bytecode/CodeBlock.h:
2280 2008-12-08 Simon Hausmann <simon.hausmann@nokia.com>
2282 Fix the Linux build with newer gcc/glibc.
2284 * jit/ExecutableAllocatorPosix.cpp: Include unistd.h for
2285 getpagesize(), according to
2286 http://opengroup.org/onlinepubs/007908775/xsh/getpagesize.html
2288 2008-12-08 Simon Hausmann <simon.hausmann@nokia.com>
2290 Fix the build with Qt on Windows.
2292 * JavaScriptCore.pri: Compile ExecutableAllocatorWin.cpp on Windows.
2294 2008-12-07 Oliver Hunt <oliver@apple.com>
2296 Reviewed by NOBODY (Buildfix).
2300 * runtime/RegExp.cpp:
2301 (JSC::RegExp::RegExp):
2303 2008-12-07 Oliver Hunt <oliver@apple.com>
2305 Reviewed by NOBODY (Build fix).
2307 Put ENABLE(ASSEMBLER) guards around use of ExecutableAllocator in global data
2309 Correct Qt and Gtk project files
2312 * JavaScriptCore.pri:
2313 * runtime/JSGlobalData.h:
2315 2008-12-07 Oliver Hunt <oliver@apple.com>
2317 Reviewed by NOBODY (Build fix).
2319 Add new files to other projects.
2322 * JavaScriptCore.pri:
2323 * JavaScriptCore.pro:
2325 2008-12-07 Oliver Hunt <oliver@apple.com>
2327 Rubber stamped by Mark Rowe.
2329 Rename ExecutableAllocatorMMAP to the more sensible ExecutableAllocatorPosix
2331 * JavaScriptCore.xcodeproj/project.pbxproj:
2332 * jit/ExecutableAllocator.h:
2333 * jit/ExecutableAllocatorPosix.cpp: Renamed from JavaScriptCore/jit/ExecutableAllocatorMMAP.cpp.
2334 (JSC::ExecutableAllocator::intializePageSize):
2335 (JSC::ExecutablePool::systemAlloc):
2336 (JSC::ExecutablePool::systemRelease):
2338 2008-12-07 Oliver Hunt <oliver@apple.com>
2340 Reviewed by Cameron Zwarich and Sam Weinig
2342 <rdar://problem/6309878> Need more granular control over allocation of executable memory (21783)
2343 <https://bugs.webkit.org/show_bug.cgi?id=21783>
2345 Add a new allocator for use by the JIT that provides executable pages, so
2346 we can get rid of the current hack that makes the entire heap executable.
2348 1-2% progression on SunSpider-v8, 1% on SunSpider. Reduces memory usage as well!
2350 * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
2351 * JavaScriptCore.vcproj/jsc/jsc.vcproj:
2352 * JavaScriptCore.xcodeproj/project.pbxproj:
2353 * assembler/AssemblerBuffer.h:
2354 (JSC::AssemblerBuffer::size):
2355 (JSC::AssemblerBuffer::executableCopy):
2356 * assembler/MacroAssembler.h:
2357 (JSC::MacroAssembler::size):
2358 (JSC::MacroAssembler::copyCode):
2359 * assembler/X86Assembler.h:
2360 (JSC::X86Assembler::size):
2361 (JSC::X86Assembler::executableCopy):
2362 * bytecode/CodeBlock.cpp:
2363 (JSC::CodeBlock::~CodeBlock):
2364 * bytecode/CodeBlock.h:
2365 (JSC::CodeBlock::executablePool):
2366 (JSC::CodeBlock::setExecutablePool):
2367 * bytecode/Instruction.h:
2368 (JSC::PolymorphicAccessStructureList::derefStructures):
2369 * interpreter/Interpreter.cpp:
2370 (JSC::Interpreter::~Interpreter):
2371 * interpreter/Interpreter.h:
2372 * jit/ExecutableAllocator.cpp: Added.
2373 * jit/ExecutableAllocator.h: Added.
2374 (JSC::ExecutablePool::create):
2375 (JSC::ExecutablePool::alloc):
2376 (JSC::ExecutablePool::~ExecutablePool):
2377 (JSC::ExecutablePool::available):
2378 (JSC::ExecutablePool::ExecutablePool):
2379 (JSC::ExecutablePool::poolAllocate):
2380 (JSC::ExecutableAllocator::ExecutableAllocator):
2381 (JSC::ExecutableAllocator::poolForSize):
2382 (JSC::ExecutablePool::sizeForAllocation):
2383 * jit/ExecutableAllocatorMMAP.cpp: Added.
2384 (JSC::ExecutableAllocator::intializePageSize):
2385 (JSC::ExecutablePool::systemAlloc):
2386 (JSC::ExecutablePool::systemRelease):
2387 * jit/ExecutableAllocatorWin.cpp: Added.
2388 (JSC::ExecutableAllocator::intializePageSize):
2389 (JSC::ExecutablePool::systemAlloc):
2390 (JSC::ExecutablePool::systemRelease):
2392 (JSC::JIT::privateCompile):
2393 (JSC::JIT::privateCompileCTIMachineTrampolines):
2395 (JSC::JIT::compileCTIMachineTrampolines):
2396 * jit/JITPropertyAccess.cpp:
2397 (JSC::JIT::privateCompilePutByIdTransition):
2398 (JSC::JIT::privateCompilePatchGetArrayLength):
2399 (JSC::JIT::privateCompileGetByIdSelf):
2400 (JSC::JIT::privateCompileGetByIdProto):
2401 (JSC::JIT::privateCompileGetByIdSelfList):
2402 (JSC::JIT::privateCompileGetByIdProtoList):
2403 (JSC::JIT::privateCompileGetByIdChainList):
2404 (JSC::JIT::privateCompileGetByIdChain):
2405 (JSC::JIT::privateCompilePutByIdReplace):
2407 (JSC::RegExpNode::emitBytecode):
2408 * runtime/JSGlobalData.h:
2409 (JSC::JSGlobalData::poolForSize):
2410 * runtime/RegExp.cpp:
2411 (JSC::RegExp::RegExp):
2412 (JSC::RegExp::create):
2413 (JSC::RegExp::~RegExp):
2415 * runtime/RegExpConstructor.cpp:
2416 (JSC::constructRegExp):
2417 * runtime/RegExpPrototype.cpp:
2418 (JSC::regExpProtoFuncCompile):
2419 * runtime/StringPrototype.cpp:
2420 (JSC::stringProtoFuncMatch):
2421 (JSC::stringProtoFuncSearch):
2423 (JSC::WREC::Generator::compileRegExp):
2424 * wrec/WRECGenerator.h:
2425 * wtf/FastMalloc.cpp:
2427 * wtf/TCSystemAlloc.cpp:
2431 (TCMalloc_SystemRelease):
2433 2008-12-06 Sam Weinig <sam@webkit.org>
2437 * jit/JITPropertyAccess.cpp:
2438 (JSC::JIT::compileGetByIdHotPath):
2439 (JSC::JIT::compilePutByIdHotPath):
2441 2008-12-06 Sam Weinig <sam@webkit.org>
2443 Reviewed by Cameron Zwarich,
2445 Move CodeBlock constructor into the .cpp file.
2447 Sunspider reports a .7% progression, but I can only assume this
2450 * bytecode/CodeBlock.cpp:
2451 (JSC::CodeBlock::CodeBlock):
2452 * bytecode/CodeBlock.h:
2454 2008-12-06 Sam Weinig <sam@webkit.org>
2456 Reviewed by Cameron Zwarich.
2458 Split JumpTable code into its own file.
2461 * JavaScriptCore.pri:
2462 * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
2463 * JavaScriptCore.xcodeproj/project.pbxproj:
2464 * JavaScriptCoreSources.bkl:
2465 * bytecode/CodeBlock.cpp:
2466 * bytecode/CodeBlock.h:
2467 * bytecode/JumpTable.cpp: Copied from bytecode/CodeBlock.cpp.
2468 * bytecode/JumpTable.h: Copied from bytecode/CodeBlock.h.
2470 2008-12-05 Sam Weinig <sam@webkit.org>
2472 Reviewed by Cameron Zwarich.
2474 Fix for https://bugs.webkit.org/show_bug.cgi?id=22715
2475 Encapsulate more CodeBlock members in preparation
2476 of moving some of them to a rare data structure.
2478 * bytecode/CodeBlock.cpp:
2479 (JSC::locationForOffset):
2480 (JSC::printConditionalJump):
2481 (JSC::printGetByIdOp):
2482 (JSC::printPutByIdOp):
2483 (JSC::CodeBlock::printStructure):
2484 (JSC::CodeBlock::printStructures):
2485 (JSC::CodeBlock::dump):
2486 (JSC::CodeBlock::~CodeBlock):
2487 (JSC::CodeBlock::unlinkCallers):
2488 (JSC::CodeBlock::derefStructures):
2489 (JSC::CodeBlock::refStructures):
2490 (JSC::CodeBlock::mark):
2491 (JSC::CodeBlock::getHandlerForVPC):
2492 (JSC::CodeBlock::nativeExceptionCodeForHandlerVPC):
2493 (JSC::CodeBlock::lineNumberForVPC):
2494 (JSC::CodeBlock::expressionRangeForVPC):
2495 (JSC::CodeBlock::shrinkToFit):
2496 * bytecode/CodeBlock.h:
2497 (JSC::CodeBlock::CodeBlock):
2498 (JSC::CodeBlock::addCaller):
2499 (JSC::CodeBlock::removeCaller):
2500 (JSC::CodeBlock::isKnownNotImmediate):
2501 (JSC::CodeBlock::isConstantRegisterIndex):
2502 (JSC::CodeBlock::getConstant):
2503 (JSC::CodeBlock::isTemporaryRegisterIndex):
2504 (JSC::CodeBlock::getStubInfo):
2505 (JSC::CodeBlock::getCallLinkInfo):
2506 (JSC::CodeBlock::instructions):
2507 (JSC::CodeBlock::setJITCode):
2508 (JSC::CodeBlock::jitCode):
2509 (JSC::CodeBlock::ownerNode):
2510 (JSC::CodeBlock::setGlobalData):
2511 (JSC::CodeBlock::setThisRegister):
2512 (JSC::CodeBlock::thisRegister):
2513 (JSC::CodeBlock::setNeedsFullScopeChain):
2514 (JSC::CodeBlock::needsFullScopeChain):
2515 (JSC::CodeBlock::setUsesEval):
2516 (JSC::CodeBlock::usesEval):
2517 (JSC::CodeBlock::setUsesArguments):
2518 (JSC::CodeBlock::usesArguments):
2519 (JSC::CodeBlock::codeType):
2520 (JSC::CodeBlock::source):
2521 (JSC::CodeBlock::sourceOffset):
2522 (JSC::CodeBlock::addGlobalResolveInstruction):
2523 (JSC::CodeBlock::numberOfPropertyAccessInstructions):
2524 (JSC::CodeBlock::addPropertyAccessInstruction):
2525 (JSC::CodeBlock::propertyAccessInstruction):
2526 (JSC::CodeBlock::numberOfCallLinkInfos):
2527 (JSC::CodeBlock::addCallLinkInfo):
2528 (JSC::CodeBlock::callLinkInfo):
2529 (JSC::CodeBlock::numberOfJumpTargets):
2530 (JSC::CodeBlock::addJumpTarget):
2531 (JSC::CodeBlock::jumpTarget):
2532 (JSC::CodeBlock::lastJumpTarget):
2533 (JSC::CodeBlock::numberOfExceptionHandlers):
2534 (JSC::CodeBlock::addExceptionHandler):
2535 (JSC::CodeBlock::exceptionHandler):
2536 (JSC::CodeBlock::addExpressionInfo):
2537 (JSC::CodeBlock::numberOfLineInfos):
2538 (JSC::CodeBlock::addLineInfo):
2539 (JSC::CodeBlock::lastLineInfo):
2540 (JSC::CodeBlock::jitReturnAddressVPCMap):
2541 (JSC::CodeBlock::numberOfIdentifiers):
2542 (JSC::CodeBlock::addIdentifier):
2543 (JSC::CodeBlock::identifier):
2544 (JSC::CodeBlock::numberOfConstantRegisters):
2545 (JSC::CodeBlock::addConstantRegister):
2546 (JSC::CodeBlock::constantRegister):
2547 (JSC::CodeBlock::addFunction):
2548 (JSC::CodeBlock::function):
2549 (JSC::CodeBlock::addFunctionExpression):
2550 (JSC::CodeBlock::functionExpression):
2551 (JSC::CodeBlock::addUnexpectedConstant):
2552 (JSC::CodeBlock::unexpectedConstant):
2553 (JSC::CodeBlock::addRegExp):
2554 (JSC::CodeBlock::regexp):
2555 (JSC::CodeBlock::symbolTable):
2556 (JSC::CodeBlock::evalCodeCache):
2557 New inline setters/getters.
2559 (JSC::ProgramCodeBlock::ProgramCodeBlock):
2560 (JSC::ProgramCodeBlock::~ProgramCodeBlock):
2561 (JSC::ProgramCodeBlock::clearGlobalObject):
2562 * bytecode/SamplingTool.cpp:
2563 (JSC::ScopeSampleRecord::sample):
2564 (JSC::SamplingTool::dump):
2565 * bytecompiler/BytecodeGenerator.cpp:
2566 * bytecompiler/BytecodeGenerator.h:
2567 * bytecompiler/Label.h:
2568 * interpreter/CallFrame.cpp:
2569 * interpreter/Interpreter.cpp:
2572 * jit/JITInlineMethods.h:
2573 * jit/JITPropertyAccess.cpp:
2575 * runtime/Arguments.h:
2576 * runtime/ExceptionHelpers.cpp:
2577 * runtime/JSActivation.cpp:
2578 * runtime/JSActivation.h:
2579 * runtime/JSGlobalObject.cpp:
2580 Change direct access to use new getter/setters.
2582 2008-12-05 Gavin Barraclough <barraclough@apple.com>
2584 Reviewed by Oliver Hunt.
2586 Prevent GCC4.2 from hanging when trying to compile Interpreter.cpp.
2587 Added "-fno-var-tracking" compiler flag.
2589 https://bugs.webkit.org/show_bug.cgi?id=22704
2591 * JavaScriptCore.xcodeproj/project.pbxproj:
2593 2008-12-05 Gavin Barraclough <barraclough@apple.com>
2595 Reviewed by Oliver Hunt.
2597 Ordering of branch operands in MacroAssembler in unnecessarily inconsistent.
2599 je, jg etc take an immediate operand as the second argument, but for the
2600 equality branches (je, jne) the immediate operand was the first argument. This
2601 was unnecessarily inconsistent. Change je, jne methods to take the immediate
2602 as the second argument.
2604 https://bugs.webkit.org/show_bug.cgi?id=22703
2606 * assembler/MacroAssembler.h:
2607 (JSC::MacroAssembler::je32):
2608 (JSC::MacroAssembler::jne32):
2610 (JSC::JIT::compileOpStrictEq):
2611 * wrec/WRECGenerator.cpp:
2612 (JSC::WREC::Generator::generateEnter):
2613 (JSC::WREC::Generator::generateNonGreedyQuantifier):
2614 (JSC::WREC::Generator::generateGreedyQuantifier):
2615 (JSC::WREC::Generator::generatePatternCharacterPair):
2616 (JSC::WREC::Generator::generatePatternCharacter):
2617 (JSC::WREC::Generator::generateCharacterClassInvertedRange):
2618 (JSC::WREC::Generator::generateCharacterClassInverted):
2619 (JSC::WREC::Generator::generateAssertionBOL):
2620 (JSC::WREC::Generator::generateAssertionWordBoundary):
2622 2008-12-05 Gavin Barraclough <barraclough@apple.com>
2624 Reviewed by Geoff Garen.
2626 Second tranche of porting JIT.cpp to MacroAssembler interface.
2628 * assembler/MacroAssembler.h:
2629 (JSC::MacroAssembler::mul32):
2630 (JSC::MacroAssembler::jl32):
2631 (JSC::MacroAssembler::jnzSub32):
2632 (JSC::MacroAssembler::joAdd32):
2633 (JSC::MacroAssembler::joMul32):
2634 (JSC::MacroAssembler::jzSub32):
2636 (JSC::JIT::emitSlowScriptCheck):
2637 (JSC::JIT::privateCompileMainPass):
2638 (JSC::JIT::privateCompileSlowCases):
2639 (JSC::JIT::privateCompileCTIMachineTrampolines):
2641 * jit/JITInlineMethods.h:
2642 (JSC::JIT::emitJumpIfNotJSCell):
2643 (JSC::JIT::emitJumpSlowCaseIfNotJSCell):
2645 2008-12-05 David Kilzer <ddkilzer@apple.com>
2647 Bug 22609: Provide a build-time choice when generating hash tables for properties of built-in DOM objects
2649 <https://bugs.webkit.org/show_bug.cgi?id=22609>
2650 <rdar://problem/6331749>
2652 Reviewed by Darin Adler.
2654 Initial patch by Yosen Lin. Adapted for ToT WebKit by David Kilzer.
2656 Added back the code that generates a "compact" hash (instead of a
2657 perfect hash) as a build-time option using the
2658 ENABLE(PERFECT_HASH_SIZE) macro as defined in Lookup.h.
2660 * create_hash_table: Rename variables to differentiate perfect hash
2661 values from compact hash values. Added back code to compute compact
2662 hash tables. Generate both hash table sizes and emit
2663 conditionalized code based on ENABLE(PERFECT_HASH_SIZE).
2664 * runtime/Lookup.cpp:
2665 (JSC::HashTable::createTable): Added version of createTable() for
2666 use with compact hash tables.
2667 (JSC::HashTable::deleteTable): Updated to work with compact hash
2669 * runtime/Lookup.h: Defined ENABLE(PERFECT_HASH_SIZE) macro here.
2670 (JSC::HashEntry::initialize): Set m_next to zero when using compact
2672 (JSC::HashEntry::setNext): Added for compact hash tables.
2673 (JSC::HashEntry::next): Added for compact hash tables.
2674 (JSC::HashTable::entry): Added version of entry() for use with
2675 compact hash tables.
2676 * runtime/Structure.cpp:
2677 (JSC::Structure::getEnumerablePropertyNames): Updated to work with
2678 compact hash tables.
2680 2008-12-05 Gavin Barraclough <barraclough@apple.com>
2682 Reviewed by Geoff Garen.
2684 Remove redundant calls to JIT::emitSlowScriptCheck.
2685 This is checked in the hot path, so is not needed on the slow path - and the code
2686 was being planted before the start of the slow case, so was completely unreachable!
2689 (JSC::JIT::privateCompileSlowCases):
2691 2008-12-05 Gavin Barraclough <barraclough@apple.com>
2693 Reviewed by Geoff Garen.
2695 Move JIT::compileOpStrictEq to MacroAssembler interface.
2697 The rewrite also looks like a small (<1%) performance progression.
2699 https://bugs.webkit.org/show_bug.cgi?id=22697
2702 (JSC::JIT::compileOpStrictEq):
2703 (JSC::JIT::privateCompileSlowCases):
2705 * jit/JITInlineMethods.h:
2706 (JSC::JIT::emitJumpIfJSCell):
2707 (JSC::JIT::emitJumpSlowCaseIfJSCell):
2709 2008-12-05 Gavin Barraclough <barraclough@apple.com>
2711 Reviewed by Geoff Garen.
2713 Remove m_assembler from MacroAssembler::Jump.
2714 Keeping a pointer allowed for some syntactic sugar - "link()" looks nicer
2715 than "link(this)". But maintaining this doubles the size of Jump, which
2716 is even more unfortunate for the JIT, since there are many large structures
2717 holding JmpSrcs. Probably best to remove it.
2719 https://bugs.webkit.org/show_bug.cgi?id=22693
2721 * assembler/MacroAssembler.h:
2722 (JSC::MacroAssembler::Jump::Jump):
2723 (JSC::MacroAssembler::Jump::link):
2724 (JSC::MacroAssembler::Jump::linkTo):
2725 (JSC::MacroAssembler::JumpList::link):
2726 (JSC::MacroAssembler::JumpList::linkTo):
2727 (JSC::MacroAssembler::jae32):
2728 (JSC::MacroAssembler::je32):
2729 (JSC::MacroAssembler::je16):
2730 (JSC::MacroAssembler::jg32):
2731 (JSC::MacroAssembler::jge32):
2732 (JSC::MacroAssembler::jl32):
2733 (JSC::MacroAssembler::jle32):
2734 (JSC::MacroAssembler::jnePtr):
2735 (JSC::MacroAssembler::jne32):
2736 (JSC::MacroAssembler::jnset32):
2737 (JSC::MacroAssembler::jset32):
2738 (JSC::MacroAssembler::jump):
2739 (JSC::MacroAssembler::jzSub32):
2740 (JSC::MacroAssembler::joAdd32):
2741 (JSC::MacroAssembler::call):
2743 (JSC::WREC::Generator::compileRegExp):
2744 * wrec/WRECGenerator.cpp:
2745 (JSC::WREC::Generator::generateEnter):
2746 (JSC::WREC::Generator::generateBackreferenceQuantifier):
2747 (JSC::WREC::Generator::generateNonGreedyQuantifier):
2748 (JSC::WREC::Generator::generateGreedyQuantifier):
2749 (JSC::WREC::Generator::generatePatternCharacter):
2750 (JSC::WREC::Generator::generateCharacterClassInvertedRange):
2751 (JSC::WREC::Generator::generateCharacterClassInverted):
2752 (JSC::WREC::Generator::generateCharacterClass):
2753 (JSC::WREC::Generator::generateParenthesesAssertion):
2754 (JSC::WREC::Generator::generateParenthesesInvertedAssertion):
2755 (JSC::WREC::Generator::generateParenthesesNonGreedy):
2756 (JSC::WREC::Generator::generateParenthesesResetTrampoline):
2757 (JSC::WREC::Generator::generateAssertionBOL):
2758 (JSC::WREC::Generator::generateAssertionEOL):
2759 (JSC::WREC::Generator::generateAssertionWordBoundary):
2760 (JSC::WREC::Generator::generateBackreference):
2761 (JSC::WREC::Generator::terminateAlternative):
2762 (JSC::WREC::Generator::terminateDisjunction):
2763 * wrec/WRECParser.h:
2765 2008-12-05 Gavin Barraclough <barraclough@apple.com>
2767 Reviewed by Geoffrey Garen.
2769 Simplify JIT generated checks for timeout code, by moving more work into the C function.
2770 https://bugs.webkit.org/show_bug.cgi?id=22688
2772 * interpreter/Interpreter.cpp:
2773 (JSC::Interpreter::cti_timeout_check):
2774 * interpreter/Interpreter.h:
2776 (JSC::JIT::emitSlowScriptCheck):
2778 2008-12-05 Sam Weinig <sam@webkit.org>
2780 Reviewed by Geoffrey Garen.
2782 Encapsulate access to jump tables in the CodeBlock in preparation
2783 of moving them to a rare data structure.
2785 * bytecode/CodeBlock.cpp:
2786 (JSC::CodeBlock::dump):
2787 (JSC::CodeBlock::shrinkToFit):
2788 * bytecode/CodeBlock.h:
2789 (JSC::CodeBlock::numberOfImmediateSwitchJumpTables):
2790 (JSC::CodeBlock::addImmediateSwitchJumpTable):
2791 (JSC::CodeBlock::immediateSwitchJumpTable):
2792 (JSC::CodeBlock::numberOfCharacterSwitchJumpTables):
2793 (JSC::CodeBlock::addCharacterSwitchJumpTable):
2794 (JSC::CodeBlock::characterSwitchJumpTable):
2795 (JSC::CodeBlock::numberOfStringSwitchJumpTables):
2796 (JSC::CodeBlock::addStringSwitchJumpTable):
2797 (JSC::CodeBlock::stringSwitchJumpTable):
2798 * bytecompiler/BytecodeGenerator.cpp:
2799 (JSC::BytecodeGenerator::generate):
2800 (JSC::BytecodeGenerator::endSwitch):
2801 * interpreter/Interpreter.cpp:
2802 (JSC::Interpreter::privateExecute):
2803 (JSC::Interpreter::cti_op_switch_imm):
2804 (JSC::Interpreter::cti_op_switch_char):
2805 (JSC::Interpreter::cti_op_switch_string):
2807 (JSC::JIT::privateCompileMainPass):
2809 2008-12-05 Adam Roben <aroben@apple.com>
2811 Windows build fix after r39020
2813 * jit/JITInlineMethods.h:
2814 (JSC::JIT::restoreArgumentReference):
2815 (JSC::JIT::restoreArgumentReferenceForTrampoline):
2816 Add some apparently-missing __.
2818 2008-12-04 Geoffrey Garen <ggaren@apple.com>
2820 Reviewed by Darin Adler.
2822 https://bugs.webkit.org/show_bug.cgi?id=22673
2824 Added support for the assertion (?=) and inverted assertion (?!) atoms
2827 * wrec/WRECGenerator.cpp:
2828 (JSC::WREC::Generator::generateParenthesesAssertion):
2829 (JSC::WREC::Generator::generateParenthesesInvertedAssertion): Split the
2830 old (unused) generateParentheses into these two functions, with more
2831 limited capabilities.
2833 * wrec/WRECGenerator.h:
2834 (JSC::WREC::Generator::): Moved an enum to the top of the class definition,
2835 to match the WebKit style, and removed a defunct comment.
2837 * wrec/WRECParser.cpp:
2838 (JSC::WREC::Parser::parseParentheses):
2839 (JSC::WREC::Parser::consumeParenthesesType):
2840 * wrec/WRECParser.h:
2841 (JSC::WREC::Parser::): Added support for parsing (?=) and (?!).
2843 2008-12-05 Simon Hausmann <simon.hausmann@nokia.com>
2845 Rubber-stamped by Tor Arne Vestbø.
2847 Disable the JIT for the Qt build alltogether again, after observing
2848 more miscompilations in a wider range of newer gcc versions.
2850 * JavaScriptCore.pri:
2852 2008-12-05 Simon Hausmann <simon.hausmann@nokia.com>
2854 Reviewed by Tor Arne Vestbø.
2856 Disable the JIT for the Qt build on Linux unless gcc is >= 4.2,
2857 due to miscompilations.
2859 * JavaScriptCore.pri:
2861 2008-12-04 Gavin Barraclough <barraclough@apple.com>
2863 Reviewed by Geoff Garen.
2865 Start porting the JIT to use the MacroAssembler.
2867 https://bugs.webkit.org/show_bug.cgi?id=22671
2868 No change in performance.
2870 * assembler/MacroAssembler.h:
2871 (JSC::MacroAssembler::Jump::operator X86Assembler::JmpSrc):
2872 (JSC::MacroAssembler::add32):
2873 (JSC::MacroAssembler::and32):
2874 (JSC::MacroAssembler::lshift32):
2875 (JSC::MacroAssembler::rshift32):
2876 (JSC::MacroAssembler::storePtr):
2877 (JSC::MacroAssembler::store32):
2878 (JSC::MacroAssembler::poke):
2879 (JSC::MacroAssembler::move):
2880 (JSC::MacroAssembler::compareImm32ForBranchEquality):
2881 (JSC::MacroAssembler::jnePtr):
2882 (JSC::MacroAssembler::jnset32):
2883 (JSC::MacroAssembler::jset32):
2884 (JSC::MacroAssembler::jzeroSub32):
2885 (JSC::MacroAssembler::joverAdd32):
2886 (JSC::MacroAssembler::call):
2887 * assembler/X86Assembler.h:
2888 (JSC::X86Assembler::shll_i8r):
2890 (JSC::JIT::privateCompileMainPass):
2891 (JSC::JIT::privateCompile):
2892 (JSC::JIT::privateCompileCTIMachineTrampolines):
2894 * jit/JITArithmetic.cpp:
2895 (JSC::JIT::compileBinaryArithOp):
2896 * jit/JITInlineMethods.h:
2897 (JSC::JIT::emitGetVirtualRegister):
2898 (JSC::JIT::emitPutCTIArg):
2899 (JSC::JIT::emitPutCTIArgConstant):
2900 (JSC::JIT::emitGetCTIArg):
2901 (JSC::JIT::emitPutCTIArgFromVirtualRegister):
2902 (JSC::JIT::emitPutCTIParam):
2903 (JSC::JIT::emitGetCTIParam):
2904 (JSC::JIT::emitPutToCallFrameHeader):
2905 (JSC::JIT::emitPutImmediateToCallFrameHeader):
2906 (JSC::JIT::emitGetFromCallFrameHeader):
2907 (JSC::JIT::emitPutVirtualRegister):
2908 (JSC::JIT::emitInitRegister):
2909 (JSC::JIT::emitNakedCall):
2910 (JSC::JIT::restoreArgumentReference):
2911 (JSC::JIT::restoreArgumentReferenceForTrampoline):
2912 (JSC::JIT::emitCTICall):
2913 (JSC::JIT::checkStructure):
2914 (JSC::JIT::emitJumpSlowCaseIfNotJSCell):
2915 (JSC::JIT::emitJumpSlowCaseIfNotImmNum):
2916 (JSC::JIT::emitJumpSlowCaseIfNotImmNums):
2917 (JSC::JIT::emitFastArithDeTagImmediate):
2918 (JSC::JIT::emitFastArithDeTagImmediateJumpIfZero):
2919 (JSC::JIT::emitFastArithReTagImmediate):
2920 (JSC::JIT::emitFastArithPotentiallyReTagImmediate):
2921 (JSC::JIT::emitFastArithImmToInt):
2922 (JSC::JIT::emitFastArithIntToImmOrSlowCase):
2923 (JSC::JIT::emitFastArithIntToImmNoCheck):
2924 (JSC::JIT::emitTagAsBoolImmediate):
2925 * jit/JITPropertyAccess.cpp:
2926 (JSC::JIT::privateCompilePutByIdTransition):
2928 2008-12-04 Geoffrey Garen <ggaren@apple.com>
2930 Reviewed by Oliver Hunt.
2932 Some refactoring for generateGreedyQuantifier.
2934 SunSpider reports no change (possibly a 0.3% speedup).
2936 * wrec/WRECGenerator.cpp:
2937 (JSC::WREC::Generator::generateGreedyQuantifier): Clarified label
2938 meanings and unified some logic to simplify things.
2940 * wrec/WRECParser.h:
2941 (JSC::WREC::Parser::parseAlternative): Added a version of parseAlternative
2942 that can jump to a Label, instead of a JumpList, upon failure. (Eventually,
2943 when we have a true Label class, this will be redundant.) This makes
2944 things easier for generateGreedyQuantifier, because it can avoid
2945 explicitly linking things.
2947 2008-12-04 Simon Hausmann <simon.hausmann@nokia.com>
2949 Reviewed by Holger Freyther.
2951 Fix crashes in the Qt build on Linux/i386 with non-executable memory
2952 by enabling TCSystemAlloc and the PROT_EXEC flag for mmap.
2954 * JavaScriptCore.pri: Enable the use of TCSystemAlloc if the JIT is
2956 * wtf/TCSystemAlloc.cpp: Extend the PROT_EXEC permissions to
2959 2008-12-04 Simon Hausmann <simon.hausmann@nokia.com>
2961 Reviewed by Tor Arne Vestbø.
2963 Enable ENABLE_JIT_OPTIMIZE_CALL, ENABLE_JIT_OPTIMIZE_PROPERTY_ACCESS
2964 and ENABLE_JIT_OPTIMIZE_ARITHMETIC, as suggested by Niko.
2966 * JavaScriptCore.pri:
2968 2008-12-04 Kent Hansen <khansen@trolltech.com>
2970 Reviewed by Simon Hausmann.
2972 Enable the JSC jit for the Qt build by default for release builds on
2973 linux-g++ and win32-msvc.
2975 * JavaScriptCore.pri:
2977 2008-12-04 Gavin Barraclough <barraclough@apple.com>
2979 Reviewed by Oliver Hunt.
2981 Allow JIT to function without property access repatching and arithmetic optimizations.
2982 Controlled by ENABLE_JIT_OPTIMIZE_PROPERTY_ACCESS and ENABLE_JIT_OPTIMIZE_ARITHMETIC switches.
2984 https://bugs.webkit.org/show_bug.cgi?id=22643
2986 * JavaScriptCore.xcodeproj/project.pbxproj:
2988 (JSC::JIT::privateCompileMainPass):
2989 (JSC::JIT::privateCompileSlowCases):
2991 * jit/JITArithmetic.cpp: Copied from jit/JIT.cpp.
2992 (JSC::JIT::compileBinaryArithOp):
2993 (JSC::JIT::compileBinaryArithOpSlowCase):
2994 * jit/JITPropertyAccess.cpp: Copied from jit/JIT.cpp.
2995 (JSC::JIT::compileGetByIdHotPath):
2996 (JSC::JIT::compileGetByIdSlowCase):
2997 (JSC::JIT::compilePutByIdHotPath):
2998 (JSC::JIT::compilePutByIdSlowCase):
2999 (JSC::resizePropertyStorage):
3000 (JSC::transitionWillNeedStorageRealloc):
3001 (JSC::JIT::privateCompilePutByIdTransition):
3002 (JSC::JIT::patchGetByIdSelf):
3003 (JSC::JIT::patchPutByIdReplace):
3004 (JSC::JIT::privateCompilePatchGetArrayLength):
3007 2008-12-03 Geoffrey Garen <ggaren@apple.com>
3009 Reviewed by Oliver Hunt.
3011 Optimized sequences of characters in regular expressions by comparing
3012 two characters at a time.
3014 1-2% speedup on SunSpider, 19-25% speedup on regexp-dna.
3016 * assembler/MacroAssembler.h:
3017 (JSC::MacroAssembler::load32):
3018 (JSC::MacroAssembler::jge32): Filled out a few more macro methods.
3020 * assembler/X86Assembler.h:
3021 (JSC::X86Assembler::movl_mr): Added a verion of movl_mr that operates
3022 without an offset, to allow the macro assembler to optmize for that case.
3025 (JSC::WREC::Generator::compileRegExp): Test the saved value of index
3026 instead of the index register when checking for "end of input." The
3027 index register doesn't increment by 1 in an orderly fashion, so testing
3028 it for == "end of input" is not valid.
3030 Also, jump all the way to "return failure" upon reaching "end of input,"
3031 instead of executing the next alternative. This is more logical, and
3032 it's a slight optimization in the case of an expression with many alternatives.
3034 * wrec/WRECGenerator.cpp:
3035 (JSC::WREC::Generator::generateIncrementIndex): Added support for
3036 jumping to a failure label in the case where the index has reached "end
3039 (JSC::WREC::Generator::generatePatternCharacterSequence):
3040 (JSC::WREC::Generator::generatePatternCharacterPair): This is the
3041 optmization. It's basically like generatePatternCharacter, but it runs two
3042 characters at a time.
3044 (JSC::WREC::Generator::generatePatternCharacter): Changed to use isASCII,
3045 since it's clearer than comparing to a magic hex value.
3047 * wrec/WRECGenerator.h:
3049 2008-12-03 Gavin Barraclough <barraclough@apple.com>
3051 Reviewed by Cameron Zwarich.
3053 Allow JIT to operate without the call-repatching optimization.
3054 Controlled by ENABLE(JIT_OPTIMIZE_CALL), defaults on, disabling
3055 this leads to significant performance regression.
3057 https://bugs.webkit.org/show_bug.cgi?id=22639
3059 * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
3060 * JavaScriptCore.xcodeproj/project.pbxproj:
3062 (JSC::JIT::privateCompileSlowCases):
3064 * jit/JITCall.cpp: Copied from jit/JIT.cpp.
3065 (JSC::JIT::compileOpCallInitializeCallFrame):
3066 (JSC::JIT::compileOpCallSetupArgs):
3067 (JSC::JIT::compileOpCallEvalSetupArgs):
3068 (JSC::JIT::compileOpConstructSetupArgs):
3069 (JSC::JIT::compileOpCall):
3070 (JSC::JIT::compileOpCallSlowCase):
3072 * jit/JITInlineMethods.h: Copied from jit/JIT.cpp.
3073 (JSC::JIT::checkStructure):
3074 (JSC::JIT::emitFastArithPotentiallyReTagImmediate):
3075 (JSC::JIT::emitTagAsBoolImmediate):
3078 2008-12-03 Eric Seidel <eric@webkit.org>
3080 Rubber-stamped by David Hyatt.
3082 Make HAVE_ACCESSIBILITY only define if !defined
3086 2008-12-03 Sam Weinig <sam@webkit.org>
3090 * assembler/X86Assembler.h:
3091 (JSC::X86Assembler::orl_i32r):
3093 2008-12-03 Sam Weinig <sam@webkit.org>
3095 Reviewed by Geoffrey Garen.
3097 Remove shared AssemblerBuffer 1MB buffer and instead give AssemblerBuffer
3098 an 256 byte inline capacity.
3100 1% progression on Sunspider.
3102 * assembler/AssemblerBuffer.h:
3103 (JSC::AssemblerBuffer::AssemblerBuffer):
3104 (JSC::AssemblerBuffer::~AssemblerBuffer):
3105 (JSC::AssemblerBuffer::grow):
3106 * assembler/MacroAssembler.h:
3107 (JSC::MacroAssembler::MacroAssembler):
3108 * assembler/X86Assembler.h:
3109 (JSC::X86Assembler::X86Assembler):
3110 * interpreter/Interpreter.cpp:
3111 (JSC::Interpreter::Interpreter):
3112 * interpreter/Interpreter.h:
3116 (JSC::RegExpNode::emitBytecode):
3117 * runtime/RegExp.cpp:
3118 (JSC::RegExp::RegExp):
3119 (JSC::RegExp::create):
3121 * runtime/RegExpConstructor.cpp:
3122 (JSC::constructRegExp):
3123 * runtime/RegExpPrototype.cpp:
3124 (JSC::regExpProtoFuncCompile):
3125 * runtime/StringPrototype.cpp:
3126 (JSC::stringProtoFuncMatch):
3127 (JSC::stringProtoFuncSearch):
3129 (JSC::WREC::Generator::compileRegExp):
3130 * wrec/WRECGenerator.h:
3131 (JSC::WREC::Generator::Generator):
3132 * wrec/WRECParser.h:
3133 (JSC::WREC::Parser::Parser):
3135 2008-12-03 Geoffrey Garen <ggaren@apple.com>
3137 Reviewed by Oliver Hunt, with help from Gavin Barraclough.
3139 orl_i32r was actually coded as an 8bit OR. So, I renamed orl_i32r to
3140 orl_i8r, changed all orl_i32r clients to use orl_i8r, and then added
3141 a new orl_i32r that actually does a 32bit OR.
3143 (32bit OR is currently unused, but a patch I'm working on uses it.)
3145 * assembler/MacroAssembler.h:
3146 (JSC::MacroAssembler::or32): Updated to choose between 8bit and 32bit OR.
3148 * assembler/X86Assembler.h:
3149 (JSC::X86Assembler::orl_i8r): The old orl_i32r.
3150 (JSC::X86Assembler::orl_i32r): The new orl_i32r.
3153 (JSC::JIT::emitFastArithPotentiallyReTagImmediate):
3154 (JSC::JIT::emitTagAsBoolImmediate): Use orl_i8r, since we're ORing 8bit
3157 2008-12-03 Dean Jackson <dino@apple.com>
3159 Reviewed by Dan Bernstein.
3161 Helper functions for turn -> degrees.
3162 https://bugs.webkit.org/show_bug.cgi?id=22497
3168 2008-12-02 Cameron Zwarich <zwarich@apple.com>
3170 Reviewed by Geoff Garen.
3172 Bug 22504: Crashes during code generation occur due to refing of ignoredResult()
3173 <https://bugs.webkit.org/show_bug.cgi?id=22504>
3175 Since ignoredResult() was implemented by casting 1 to a RegisterID*, any
3176 attempt to ref ignoredResult() results in a crash. This will occur in
3177 code generation of a function body where a node emits another node with
3178 the dst that was passed to it, and then refs the returned RegisterID*.
3180 To fix this problem, make ignoredResult() a member function of
3181 BytecodeGenerator that simply returns a pointe to a fixed RegisterID
3182 member of BytecodeGenerator.
3184 * bytecompiler/BytecodeGenerator.h:
3185 (JSC::BytecodeGenerator::ignoredResult):
3186 * bytecompiler/RegisterID.h:
3188 (JSC::NullNode::emitBytecode):
3189 (JSC::BooleanNode::emitBytecode):
3190 (JSC::NumberNode::emitBytecode):
3191 (JSC::StringNode::emitBytecode):
3192 (JSC::RegExpNode::emitBytecode):
3193 (JSC::ThisNode::emitBytecode):
3194 (JSC::ResolveNode::emitBytecode):
3195 (JSC::ObjectLiteralNode::emitBytecode):
3196 (JSC::PostfixResolveNode::emitBytecode):
3197 (JSC::PostfixBracketNode::emitBytecode):
3198 (JSC::PostfixDotNode::emitBytecode):
3199 (JSC::DeleteValueNode::emitBytecode):
3200 (JSC::VoidNode::emitBytecode):
3201 (JSC::TypeOfResolveNode::emitBytecode):
3202 (JSC::TypeOfValueNode::emitBytecode):
3203 (JSC::PrefixResolveNode::emitBytecode):
3204 (JSC::AssignResolveNode::emitBytecode):
3205 (JSC::CommaNode::emitBytecode):
3206 (JSC::ForNode::emitBytecode):
3207 (JSC::ForInNode::emitBytecode):
3208 (JSC::ReturnNode::emitBytecode):
3209 (JSC::ThrowNode::emitBytecode):
3210 (JSC::FunctionBodyNode::emitBytecode):
3211 (JSC::FuncDeclNode::emitBytecode):
3213 2008-12-02 Geoffrey Garen <ggaren@apple.com>
3215 Reviewed by Cameron Zwarich.
3217 Fixed https://bugs.webkit.org/show_bug.cgi?id=22537
3218 REGRESSION (r38745): Assertion failure in jsSubstring() at ge.com
3220 The bug was that index would become greater than length, so our
3221 "end of input" checks, which all check "index == length", would fail.
3223 The solution is to check for end of input before incrementing index,
3224 to ensure that index is always <= length.
3226 As a side benefit, generateJumpIfEndOfInput can now use je instead of
3227 jg, which should be slightly faster.
3230 (JSC::WREC::Generator::compileRegExp):
3231 * wrec/WRECGenerator.cpp:
3232 (JSC::WREC::Generator::generateJumpIfEndOfInput):
3234 2008-12-02 Gavin Barraclough <barraclough@apple.com>
3236 Reviewed by Geoffrey Garen.
3238 Plant shift right immediate instructions, which are awesome.
3239 https://bugs.webkit.org/show_bug.cgi?id=22610
3240 ~5% on the v8-crypto test.
3243 (JSC::JIT::privateCompileMainPass):
3244 (JSC::JIT::privateCompileSlowCases):
3246 2008-12-02 Geoffrey Garen <ggaren@apple.com>
3248 Reviewed by Sam Weinig.
3250 Cleaned up SegmentedVector by abstracting segment access into helper
3253 SunSpider reports no change.
3255 * bytecompiler/SegmentedVector.h:
3256 (JSC::SegmentedVector::SegmentedVector):
3257 (JSC::SegmentedVector::~SegmentedVector):
3258 (JSC::SegmentedVector::size):
3259 (JSC::SegmentedVector::at):
3260 (JSC::SegmentedVector::operator[]):
3261 (JSC::SegmentedVector::last):
3262 (JSC::SegmentedVector::append):
3263 (JSC::SegmentedVector::removeLast):
3264 (JSC::SegmentedVector::grow):
3265 (JSC::SegmentedVector::clear):
3266 (JSC::SegmentedVector::deleteAllSegments):
3267 (JSC::SegmentedVector::segmentFor):
3268 (JSC::SegmentedVector::subscriptFor):
3269 (JSC::SegmentedVector::ensureSegmentsFor):
3270 (JSC::SegmentedVector::ensureSegment):
3272 2008-12-02 Geoffrey Garen <ggaren@apple.com>
3274 Reviewed by Geoffrey Garen. (Patch by Cameron Zwarich <zwarich@apple.com>.)
3276 Fixed https://bugs.webkit.org/show_bug.cgi?id=22482
3277 REGRESSION (r37991): Occasionally see "Scene rendered incorrectly"
3278 message when running the V8 Raytrace benchmark
3280 Rolled out r37991. It didn't properly save xmm0, which is caller-save,
3281 before calling helper functions.
3283 SunSpider and v8 benchmarks show little change -- possibly a .2%
3284 SunSpider regression, possibly a .2% v8 benchmark speedup.
3286 * assembler/X86Assembler.h:
3287 (JSC::X86Assembler::):
3288 * bytecode/CodeBlock.cpp:
3289 (JSC::CodeBlock::dump):
3290 * bytecode/Instruction.h:
3291 (JSC::Instruction::):
3292 * bytecompiler/BytecodeGenerator.cpp:
3293 (JSC::BytecodeGenerator::emitUnaryOp):
3294 * bytecompiler/BytecodeGenerator.h:
3295 (JSC::BytecodeGenerator::emitToJSNumber):
3296 (JSC::BytecodeGenerator::emitTypeOf):
3297 (JSC::BytecodeGenerator::emitGetPropertyNames):
3298 * interpreter/Interpreter.cpp:
3299 (JSC::Interpreter::privateExecute):
3300 * interpreter/Interpreter.h:
3302 (JSC::JIT::privateCompileMainPass):
3303 (JSC::JIT::privateCompileSlowCases):
3306 (JSC::UnaryOpNode::emitBytecode):
3307 (JSC::BinaryOpNode::emitBytecode):
3308 (JSC::EqualNode::emitBytecode):
3309 * parser/ResultType.h:
3310 (JSC::ResultType::isReusable):
3311 (JSC::ResultType::mightBeNumber):
3312 * runtime/JSNumberCell.h:
3314 2008-12-01 Gavin Barraclough <barraclough@apple.com>
3316 Reviewed by Geoffrey Garen.
3318 Remove unused (sampling only, and derivable) argument to JIT::emitCTICall.
3319 https://bugs.webkit.org/show_bug.cgi?id=22587
3322 (JSC::JIT::emitCTICall):
3323 (JSC::JIT::compileOpCall):
3324 (JSC::JIT::emitSlowScriptCheck):
3325 (JSC::JIT::compileBinaryArithOpSlowCase):
3326 (JSC::JIT::privateCompileMainPass):
3327 (JSC::JIT::privateCompileSlowCases):
3328 (JSC::JIT::privateCompile):
3331 2008-12-02 Dimitri Glazkov <dglazkov@chromium.org>
3333 Reviewed by Eric Seidel.
3335 Fix the inheritance chain for JSFunction.
3337 * runtime/JSFunction.cpp:
3338 (JSC::JSFunction::info): Add InternalFunction::info as parent class
3340 2008-12-02 Simon Hausmann <hausmann@webkit.org>
3342 Reviewed by Tor Arne Vestbø.
3344 Fix ability to include JavaScriptCore.pri from other .pro files.
3346 * JavaScriptCore.pri: Moved -O3 setting into the .pro files.
3347 * JavaScriptCore.pro:
3350 2008-12-01 Geoffrey Garen <ggaren@apple.com>
3352 Reviewed by Cameron Zwarich, with help from Gavin Barraclough.
3354 Fixed https://bugs.webkit.org/show_bug.cgi?id=22583.
3356 Refactored regular expression parsing to parse sequences of characters
3357 as a single unit, in preparation for optimizing sequences of characters.
3359 SunSpider reports no change.
3361 * JavaScriptCore.xcodeproj/project.pbxproj:
3362 * wrec/Escapes.h: Added. Set of classes for representing an escaped
3365 * wrec/Quantifier.h:
3366 (JSC::WREC::Quantifier::Quantifier): Simplified this constructor slightly,
3367 to match the new Escape constructor.
3369 * wrec/WRECGenerator.cpp:
3370 (JSC::WREC::Generator::generatePatternCharacterSequence):
3371 * wrec/WRECGenerator.h: Added an interface for generating a sequence
3372 of pattern characters at a time. It doesn't do anything special yet.
3374 * wrec/WRECParser.cpp:
3375 (JSC::WREC::Parser::consumeGreedyQuantifier):
3376 (JSC::WREC::Parser::consumeQuantifier): Renamed "parse" to "consume" in
3377 these functions, to match "consumeEscape."
3379 (JSC::WREC::Parser::parsePatternCharacterSequence): New function for
3380 iteratively aggregating a sequence of characters in a pattern.
3382 (JSC::WREC::Parser::parseCharacterClassQuantifier):
3383 (JSC::WREC::Parser::parseBackreferenceQuantifier): Renamed "parse" to
3384 "consume" in these functions, to match "consumeEscape."
3386 (JSC::WREC::Parser::parseCharacterClass): Refactored to use the common
3387 escape processing code in consumeEscape.
3389 (JSC::WREC::Parser::parseEscape): Refactored to use the common
3390 escape processing code in consumeEscape.
3392 (JSC::WREC::Parser::consumeEscape): Factored escaped token processing
3393 into a common function, since we were doing this in a few places.
3395 (JSC::WREC::Parser::parseTerm): Refactored to use the common
3396 escape processing code in consumeEscape.
3398 * wrec/WRECParser.h:
3399 (JSC::WREC::Parser::consumeOctal): Refactored to use a helper function
3400 for reading a digit.
3402 2008-12-01 Cameron Zwarich <zwarich@apple.com>
3404 Reviewed by Oliver Hunt.
3406 Bug 20340: SegmentedVector segment allocations can lead to unsafe use of temporary registers
3407 <https://bugs.webkit.org/show_bug.cgi?id=20340>
3409 SegmentedVector currently frees segments and reallocates them when used
3410 as a stack. This can lead to unsafe use of pointers into freed segments.
3412 In order to fix this problem, SegmentedVector will be changed to only
3413 grow and never shrink. Also, rename the reserveCapacity() member
3414 function to grow() to match the actual usage in BytecodeGenerator, where
3415 this function is used to allocate a group of registers at once, rather
3416 than merely saving space for them.
3418 * bytecompiler/BytecodeGenerator.cpp:
3419 (JSC::BytecodeGenerator::BytecodeGenerator): Use grow() instead of
3421 * bytecompiler/SegmentedVector.h:
3422 (JSC::SegmentedVector::SegmentedVector):
3423 (JSC::SegmentedVector::last):
3424 (JSC::SegmentedVector::append):
3425 (JSC::SegmentedVector::removeLast):
3426 (JSC::SegmentedVector::grow): Renamed from reserveCapacity().
3427 (JSC::SegmentedVector::clear):
3429 2008-12-01 Mark Rowe <mrowe@apple.com>
3431 Rubber-stamped by Anders Carlsson.
3433 Disable WREC for x86_64 since memory allocated by the system allocator is not marked executable,
3434 which causes 64-bit debug builds to crash. Once we have a dedicated allocator for executable
3435 memory we can turn this back on.
3439 2008-12-01 Antti Koivisto <antti@apple.com>
3441 Reviewed by Maciej Stachowiak.
3443 Restore inline buffer after vector is shrunk back below its inline capacity.
3447 (WTF::VectorBuffer::restoreInlineBufferIfNeeded):
3448 (WTF::::shrinkCapacity):
3450 2008-11-30 Antti Koivisto <antti@apple.com>
3452 Reviewed by Mark Rowe.
3454 Try to return free pages in the current thread cache too.
3456 * wtf/FastMalloc.cpp:
3457 (WTF::TCMallocStats::releaseFastMallocFreeMemory):
3459 2008-12-01 David Levin <levin@chromium.org>
3461 Reviewed by Alexey Proskuryakov.
3463 https://bugs.webkit.org/show_bug.cgi?id=22567
3464 Make HashTable work as expected with respect to threads. Specifically, it has class-level
3465 thread safety and constant methods work on constant objects without synchronization.
3467 No observable change in behavior, so no test. This only affects debug builds.
3469 * wtf/HashTable.cpp:
3470 (WTF::hashTableStatsMutex):
3471 (WTF::HashTableStats::~HashTableStats):
3472 (WTF::HashTableStats::recordCollisionAtCount):
3473 Guarded variable access with a mutex.
3477 (WTF::::lookupForWriting):
3478 (WTF::::fullLookupForWriting):
3483 Changed increments of static variables to use atomicIncrement.
3485 (WTF::::invalidateIterators):
3487 (WTF::removeIterator):
3488 Guarded mutable access with a mutex.
3490 2008-11-29 Gavin Barraclough <barraclough@apple.com>
3492 Reviewed by Cameron Zwarich.
3494 Enable WREC on PLATFORM(X86_64). This change predominantly requires changes to the
3495 WREC::Generator::generateEnter method to support the x86-64 ABI, and addition of
3496 support for a limited number of quadword operations in the X86Assembler.
3498 This patch will cause the JS heap to be allocated with RWX permissions on 64-bit Mac
3499 platforms. This is a regression with respect to previous 64-bit behaviour, but is no
3500 more permissive than on 32-bit builds. This issue should be addressed at some point.
3501 (This is tracked by bug #21783.)
3503 https://bugs.webkit.org/show_bug.cgi?id=22554
3504 Greater than 4x speedup on regexp-dna, on x86-64.
3506 * assembler/MacroAssembler.h:
3507 (JSC::MacroAssembler::addPtr):
3508 (JSC::MacroAssembler::loadPtr):
3509 (JSC::MacroAssembler::storePtr):
3510 (JSC::MacroAssembler::pop):
3511 (JSC::MacroAssembler::push):
3512 (JSC::MacroAssembler::move):
3513 * assembler/X86Assembler.h:
3514 (JSC::X86Assembler::):
3515 (JSC::X86Assembler::movq_rr):
3516 (JSC::X86Assembler::addl_i8m):
3517 (JSC::X86Assembler::addl_i32r):
3518 (JSC::X86Assembler::addq_i8r):
3519 (JSC::X86Assembler::addq_i32r):
3520 (JSC::X86Assembler::movq_mr):
3521 (JSC::X86Assembler::movq_rm):
3523 * wrec/WRECGenerator.cpp:
3524 (JSC::WREC::Generator::generateEnter):
3525 (JSC::WREC::Generator::generateReturnSuccess):
3526 (JSC::WREC::Generator::generateReturnFailure):
3528 * wtf/TCSystemAlloc.cpp:
3530 2008-12-01 Cameron Zwarich <zwarich@apple.com>
3532 Reviewed by Sam Weinig.
3534 Preliminary work for bug 20340: SegmentedVector segment allocations can lead to unsafe use of temporary registers
3535 <https://bugs.webkit.org/show_bug.cgi?id=20340>
3537 SegmentedVector currently frees segments and reallocates them when used
3538 as a stack. This can lead to unsafe use of pointers into freed segments.
git://git.webkit.org / WebKit-https.git / blob