1 | /* |
2 | * Copyright (C) 2009-2018 Apple Inc. All rights reserved. |
3 | * |
4 | * Redistribution and use in source and binary forms, with or without |
5 | * modification, are permitted provided that the following conditions |
6 | * are met: |
7 | * 1. Redistributions of source code must retain the above copyright |
8 | * notice, this list of conditions and the following disclaimer. |
9 | * 2. Redistributions in binary form must reproduce the above copyright |
10 | * notice, this list of conditions and the following disclaimer in the |
11 | * documentation and/or other materials provided with the distribution. |
12 | * |
13 | * THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY |
14 | * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
15 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
16 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR |
17 | * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
18 | * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
19 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
20 | * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY |
21 | * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
22 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
23 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
24 | */ |
25 | |
26 | #include "config.h" |
27 | #include "YarrJIT.h" |
28 | |
29 | #include <wtf/ASCIICType.h> |
30 | #include "LinkBuffer.h" |
31 | #include "Options.h" |
32 | #include "VM.h" |
33 | #include "Yarr.h" |
34 | #include "YarrCanonicalize.h" |
35 | #include "YarrDisassembler.h" |
36 | |
37 | #if ENABLE(YARR_JIT) |
38 | |
39 | namespace JSC { namespace Yarr { |
40 | |
41 | template<YarrJITCompileMode compileMode> |
42 | class YarrGenerator : public YarrJITInfo, private MacroAssembler { |
43 | |
44 | #if CPU(ARM_THUMB2) |
45 | static const RegisterID input = ARMRegisters::r0; |
46 | static const RegisterID index = ARMRegisters::r1; |
47 | static const RegisterID length = ARMRegisters::r2; |
48 | static const RegisterID output = ARMRegisters::r3; |
49 | |
50 | static const RegisterID regT0 = ARMRegisters::r4; |
51 | static const RegisterID regT1 = ARMRegisters::r5; |
52 | static const RegisterID initialStart = ARMRegisters::r8; |
53 | |
54 | static const RegisterID returnRegister = ARMRegisters::r0; |
55 | static const RegisterID returnRegister2 = ARMRegisters::r1; |
56 | |
57 | #define HAVE_INITIAL_START_REG |
58 | #elif CPU(ARM64) |
59 | // Argument registers |
60 | static const RegisterID input = ARM64Registers::x0; |
61 | static const RegisterID index = ARM64Registers::x1; |
62 | static const RegisterID length = ARM64Registers::x2; |
63 | static const RegisterID output = ARM64Registers::x3; |
64 | static const RegisterID freelistRegister = ARM64Registers::x4; |
65 | static const RegisterID freelistSizeRegister = ARM64Registers::x5; |
66 | |
67 | // Scratch registers |
68 | static const RegisterID regT0 = ARM64Registers::x6; |
69 | static const RegisterID regT1 = ARM64Registers::x7; |
70 | static const RegisterID regT2 = ARM64Registers::x8; |
71 | static const RegisterID remainingMatchCount = ARM64Registers::x9; |
72 | static const RegisterID regUnicodeInputAndTrail = ARM64Registers::x10; |
73 | static const RegisterID initialStart = ARM64Registers::x11; |
74 | static const RegisterID supplementaryPlanesBase = ARM64Registers::x12; |
75 | static const RegisterID leadingSurrogateTag = ARM64Registers::x13; |
76 | static const RegisterID trailingSurrogateTag = ARM64Registers::x14; |
77 | static const RegisterID endOfStringAddress = ARM64Registers::x15; |
78 | |
79 | static const RegisterID returnRegister = ARM64Registers::x0; |
80 | static const RegisterID returnRegister2 = ARM64Registers::x1; |
81 | |
82 | const TrustedImm32 surrogateTagMask = TrustedImm32(0xfffffc00); |
83 | #define HAVE_INITIAL_START_REG |
84 | #define JIT_UNICODE_EXPRESSIONS |
85 | #elif CPU(MIPS) |
86 | static const RegisterID input = MIPSRegisters::a0; |
87 | static const RegisterID index = MIPSRegisters::a1; |
88 | static const RegisterID length = MIPSRegisters::a2; |
89 | static const RegisterID output = MIPSRegisters::a3; |
90 | |
91 | static const RegisterID regT0 = MIPSRegisters::t4; |
92 | static const RegisterID regT1 = MIPSRegisters::t5; |
93 | static const RegisterID initialStart = MIPSRegisters::t6; |
94 | |
95 | static const RegisterID returnRegister = MIPSRegisters::v0; |
96 | static const RegisterID returnRegister2 = MIPSRegisters::v1; |
97 | |
98 | #define HAVE_INITIAL_START_REG |
99 | #elif CPU(X86) |
100 | static const RegisterID input = X86Registers::eax; |
101 | static const RegisterID index = X86Registers::edx; |
102 | static const RegisterID length = X86Registers::ecx; |
103 | static const RegisterID output = X86Registers::edi; |
104 | |
105 | static const RegisterID regT0 = X86Registers::ebx; |
106 | static const RegisterID regT1 = X86Registers::esi; |
107 | |
108 | static const RegisterID returnRegister = X86Registers::eax; |
109 | static const RegisterID returnRegister2 = X86Registers::edx; |
110 | #elif CPU(X86_64) |
111 | #if !OS(WINDOWS) |
112 | // Argument registers |
113 | static const RegisterID input = X86Registers::edi; |
114 | static const RegisterID index = X86Registers::esi; |
115 | static const RegisterID length = X86Registers::edx; |
116 | static const RegisterID output = X86Registers::ecx; |
117 | static const RegisterID freelistRegister = X86Registers::r8; |
118 | static const RegisterID freelistSizeRegister = X86Registers::r9; // Only used during initialization. |
119 | #else |
120 | // If the return value doesn't fit in 64bits, its destination is pointed by rcx and the parameters are shifted. |
121 | // http://msdn.microsoft.com/en-us/library/7572ztz4.aspx |
122 | COMPILE_ASSERT(sizeof(MatchResult) > sizeof(void*), MatchResult_does_not_fit_in_64bits); |
123 | static const RegisterID input = X86Registers::edx; |
124 | static const RegisterID index = X86Registers::r8; |
125 | static const RegisterID length = X86Registers::r9; |
126 | static const RegisterID output = X86Registers::r10; |
127 | #endif |
128 | |
129 | // Scratch registers |
130 | static const RegisterID regT0 = X86Registers::eax; |
131 | #if !OS(WINDOWS) |
132 | static const RegisterID regT1 = X86Registers::r9; |
133 | static const RegisterID regT2 = X86Registers::r10; |
134 | #else |
135 | static const RegisterID regT1 = X86Registers::ecx; |
136 | static const RegisterID regT2 = X86Registers::edi; |
137 | #endif |
138 | |
139 | static const RegisterID initialStart = X86Registers::ebx; |
140 | #if !OS(WINDOWS) |
141 | static const RegisterID remainingMatchCount = X86Registers::r12; |
142 | #else |
143 | static const RegisterID remainingMatchCount = X86Registers::esi; |
144 | #endif |
145 | static const RegisterID regUnicodeInputAndTrail = X86Registers::r13; |
146 | static const RegisterID leadingSurrogateTag = X86Registers::r14; |
147 | static const RegisterID endOfStringAddress = X86Registers::r15; |
148 | |
149 | static const RegisterID returnRegister = X86Registers::eax; |
150 | static const RegisterID returnRegister2 = X86Registers::edx; |
151 | |
152 | const TrustedImm32 supplementaryPlanesBase = TrustedImm32(0x10000); |
153 | const TrustedImm32 trailingSurrogateTag = TrustedImm32(0xdc00); |
154 | const TrustedImm32 surrogateTagMask = TrustedImm32(0xfffffc00); |
155 | #define HAVE_INITIAL_START_REG |
156 | #define JIT_UNICODE_EXPRESSIONS |
157 | #endif |
158 | |
159 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
160 | struct ParenContextSizes { |
161 | size_t m_numSubpatterns; |
162 | size_t m_frameSlots; |
163 | |
164 | ParenContextSizes(size_t numSubpatterns, size_t frameSlots) |
165 | : m_numSubpatterns(numSubpatterns) |
166 | , m_frameSlots(frameSlots) |
167 | { |
168 | } |
169 | |
170 | size_t numSubpatterns() { return m_numSubpatterns; } |
171 | |
172 | size_t frameSlots() { return m_frameSlots; } |
173 | }; |
174 | |
175 | struct ParenContext { |
176 | struct ParenContext* next; |
177 | uint32_t begin; |
178 | uint32_t matchAmount; |
179 | uintptr_t returnAddress; |
180 | struct Subpatterns { |
181 | unsigned start; |
182 | unsigned end; |
183 | } subpatterns[0]; |
184 | uintptr_t frameSlots[0]; |
185 | |
186 | static size_t sizeFor(ParenContextSizes& parenContextSizes) |
187 | { |
188 | return sizeof(ParenContext) + sizeof(Subpatterns) * parenContextSizes.numSubpatterns() + sizeof(uintptr_t) * parenContextSizes.frameSlots(); |
189 | } |
190 | |
191 | static ptrdiff_t nextOffset() |
192 | { |
193 | return offsetof(ParenContext, next); |
194 | } |
195 | |
196 | static ptrdiff_t beginOffset() |
197 | { |
198 | return offsetof(ParenContext, begin); |
199 | } |
200 | |
201 | static ptrdiff_t matchAmountOffset() |
202 | { |
203 | return offsetof(ParenContext, matchAmount); |
204 | } |
205 | |
206 | static ptrdiff_t returnAddressOffset() |
207 | { |
208 | return offsetof(ParenContext, returnAddress); |
209 | } |
210 | |
211 | static ptrdiff_t subpatternOffset(size_t subpattern) |
212 | { |
213 | return offsetof(ParenContext, subpatterns) + (subpattern - 1) * sizeof(Subpatterns); |
214 | } |
215 | |
216 | static ptrdiff_t savedFrameOffset(ParenContextSizes& parenContextSizes) |
217 | { |
218 | return offsetof(ParenContext, subpatterns) + (parenContextSizes.numSubpatterns()) * sizeof(Subpatterns); |
219 | } |
220 | }; |
221 | |
222 | void initParenContextFreeList() |
223 | { |
224 | RegisterID parenContextPointer = regT0; |
225 | RegisterID nextParenContextPointer = regT2; |
226 | |
227 | size_t parenContextSize = ParenContext::sizeFor(m_parenContextSizes); |
228 | |
229 | parenContextSize = WTF::roundUpToMultipleOf<sizeof(uintptr_t)>(parenContextSize); |
230 | |
231 | // Check that the paren context is a reasonable size. |
232 | if (parenContextSize > INT16_MAX) |
233 | m_abortExecution.append(jump()); |
234 | |
235 | Jump emptyFreeList = branchTestPtr(Zero, freelistRegister); |
236 | move(freelistRegister, parenContextPointer); |
237 | addPtr(TrustedImm32(parenContextSize), freelistRegister, nextParenContextPointer); |
238 | addPtr(freelistRegister, freelistSizeRegister); |
239 | subPtr(TrustedImm32(parenContextSize), freelistSizeRegister); |
240 | |
241 | Label loopTop(this); |
242 | Jump initDone = branchPtr(Above, nextParenContextPointer, freelistSizeRegister); |
243 | storePtr(nextParenContextPointer, Address(parenContextPointer, ParenContext::nextOffset())); |
244 | move(nextParenContextPointer, parenContextPointer); |
245 | addPtr(TrustedImm32(parenContextSize), parenContextPointer, nextParenContextPointer); |
246 | jump(loopTop); |
247 | |
248 | initDone.link(this); |
249 | storePtr(TrustedImmPtr(nullptr), Address(parenContextPointer, ParenContext::nextOffset())); |
250 | emptyFreeList.link(this); |
251 | } |
252 | |
253 | void allocateParenContext(RegisterID result) |
254 | { |
255 | m_abortExecution.append(branchTestPtr(Zero, freelistRegister)); |
256 | sub32(TrustedImm32(1), remainingMatchCount); |
257 | m_hitMatchLimit.append(branchTestPtr(Zero, remainingMatchCount)); |
258 | move(freelistRegister, result); |
259 | loadPtr(Address(freelistRegister, ParenContext::nextOffset()), freelistRegister); |
260 | } |
261 | |
262 | void freeParenContext(RegisterID headPtrRegister, RegisterID newHeadPtrRegister) |
263 | { |
264 | loadPtr(Address(headPtrRegister, ParenContext::nextOffset()), newHeadPtrRegister); |
265 | storePtr(freelistRegister, Address(headPtrRegister, ParenContext::nextOffset())); |
266 | move(headPtrRegister, freelistRegister); |
267 | } |
268 | |
269 | void saveParenContext(RegisterID parenContextReg, RegisterID tempReg, unsigned firstSubpattern, unsigned lastSubpattern, unsigned subpatternBaseFrameLocation) |
270 | { |
271 | store32(index, Address(parenContextReg, ParenContext::beginOffset())); |
272 | loadFromFrame(subpatternBaseFrameLocation + BackTrackInfoParentheses::matchAmountIndex(), tempReg); |
273 | store32(tempReg, Address(parenContextReg, ParenContext::matchAmountOffset())); |
274 | loadFromFrame(subpatternBaseFrameLocation + BackTrackInfoParentheses::returnAddressIndex(), tempReg); |
275 | storePtr(tempReg, Address(parenContextReg, ParenContext::returnAddressOffset())); |
276 | if (compileMode == IncludeSubpatterns) { |
277 | for (unsigned subpattern = firstSubpattern; subpattern <= lastSubpattern; subpattern++) { |
278 | loadPtr(Address(output, (subpattern << 1) * sizeof(unsigned)), tempReg); |
279 | storePtr(tempReg, Address(parenContextReg, ParenContext::subpatternOffset(subpattern))); |
280 | clearSubpatternStart(subpattern); |
281 | } |
282 | } |
283 | subpatternBaseFrameLocation += YarrStackSpaceForBackTrackInfoParentheses; |
284 | for (unsigned frameLocation = subpatternBaseFrameLocation; frameLocation < m_parenContextSizes.frameSlots(); frameLocation++) { |
285 | loadFromFrame(frameLocation, tempReg); |
286 | storePtr(tempReg, Address(parenContextReg, ParenContext::savedFrameOffset(m_parenContextSizes) + frameLocation * sizeof(uintptr_t))); |
287 | } |
288 | } |
289 | |
290 | void restoreParenContext(RegisterID parenContextReg, RegisterID tempReg, unsigned firstSubpattern, unsigned lastSubpattern, unsigned subpatternBaseFrameLocation) |
291 | { |
292 | load32(Address(parenContextReg, ParenContext::beginOffset()), index); |
293 | storeToFrame(index, subpatternBaseFrameLocation + BackTrackInfoParentheses::beginIndex()); |
294 | load32(Address(parenContextReg, ParenContext::matchAmountOffset()), tempReg); |
295 | storeToFrame(tempReg, subpatternBaseFrameLocation + BackTrackInfoParentheses::matchAmountIndex()); |
296 | loadPtr(Address(parenContextReg, ParenContext::returnAddressOffset()), tempReg); |
297 | storeToFrame(tempReg, subpatternBaseFrameLocation + BackTrackInfoParentheses::returnAddressIndex()); |
298 | if (compileMode == IncludeSubpatterns) { |
299 | for (unsigned subpattern = firstSubpattern; subpattern <= lastSubpattern; subpattern++) { |
300 | loadPtr(Address(parenContextReg, ParenContext::subpatternOffset(subpattern)), tempReg); |
301 | storePtr(tempReg, Address(output, (subpattern << 1) * sizeof(unsigned))); |
302 | } |
303 | } |
304 | subpatternBaseFrameLocation += YarrStackSpaceForBackTrackInfoParentheses; |
305 | for (unsigned frameLocation = subpatternBaseFrameLocation; frameLocation < m_parenContextSizes.frameSlots(); frameLocation++) { |
306 | loadPtr(Address(parenContextReg, ParenContext::savedFrameOffset(m_parenContextSizes) + frameLocation * sizeof(uintptr_t)), tempReg); |
307 | storeToFrame(tempReg, frameLocation); |
308 | } |
309 | } |
310 | #endif |
311 | |
312 | void optimizeAlternative(PatternAlternative* alternative) |
313 | { |
314 | if (!alternative->m_terms.size()) |
315 | return; |
316 | |
317 | for (unsigned i = 0; i < alternative->m_terms.size() - 1; ++i) { |
318 | PatternTerm& term = alternative->m_terms[i]; |
319 | PatternTerm& nextTerm = alternative->m_terms[i + 1]; |
320 | |
321 | // We can move BMP only character classes after fixed character terms. |
322 | if ((term.type == PatternTerm::TypeCharacterClass) |
323 | && (term.quantityType == QuantifierFixedCount) |
324 | && (!m_decodeSurrogatePairs || (term.characterClass->hasOneCharacterSize() && !term.m_invert)) |
325 | && (nextTerm.type == PatternTerm::TypePatternCharacter) |
326 | && (nextTerm.quantityType == QuantifierFixedCount)) { |
327 | PatternTerm termCopy = term; |
328 | alternative->m_terms[i] = nextTerm; |
329 | alternative->m_terms[i + 1] = termCopy; |
330 | } |
331 | } |
332 | } |
333 | |
334 | void matchCharacterClassRange(RegisterID character, JumpList& failures, JumpList& matchDest, const CharacterRange* ranges, unsigned count, unsigned* matchIndex, const UChar32* matches, unsigned matchCount) |
335 | { |
336 | do { |
337 | // pick which range we're going to generate |
338 | int which = count >> 1; |
339 | char lo = ranges[which].begin; |
340 | char hi = ranges[which].end; |
341 | |
342 | // check if there are any ranges or matches below lo. If not, just jl to failure - |
343 | // if there is anything else to check, check that first, if it falls through jmp to failure. |
344 | if ((*matchIndex < matchCount) && (matches[*matchIndex] < lo)) { |
345 | Jump loOrAbove = branch32(GreaterThanOrEqual, character, Imm32((unsigned short)lo)); |
346 | |
347 | // generate code for all ranges before this one |
348 | if (which) |
349 | matchCharacterClassRange(character, failures, matchDest, ranges, which, matchIndex, matches, matchCount); |
350 | |
351 | while ((*matchIndex < matchCount) && (matches[*matchIndex] < lo)) { |
352 | matchDest.append(branch32(Equal, character, Imm32((unsigned short)matches[*matchIndex]))); |
353 | ++*matchIndex; |
354 | } |
355 | failures.append(jump()); |
356 | |
357 | loOrAbove.link(this); |
358 | } else if (which) { |
359 | Jump loOrAbove = branch32(GreaterThanOrEqual, character, Imm32((unsigned short)lo)); |
360 | |
361 | matchCharacterClassRange(character, failures, matchDest, ranges, which, matchIndex, matches, matchCount); |
362 | failures.append(jump()); |
363 | |
364 | loOrAbove.link(this); |
365 | } else |
366 | failures.append(branch32(LessThan, character, Imm32((unsigned short)lo))); |
367 | |
368 | while ((*matchIndex < matchCount) && (matches[*matchIndex] <= hi)) |
369 | ++*matchIndex; |
370 | |
371 | matchDest.append(branch32(LessThanOrEqual, character, Imm32((unsigned short)hi))); |
372 | // fall through to here, the value is above hi. |
373 | |
374 | // shuffle along & loop around if there are any more matches to handle. |
375 | unsigned next = which + 1; |
376 | ranges += next; |
377 | count -= next; |
378 | } while (count); |
379 | } |
380 | |
381 | void matchCharacterClass(RegisterID character, JumpList& matchDest, const CharacterClass* charClass) |
382 | { |
383 | if (charClass->m_table && !m_decodeSurrogatePairs) { |
384 | ExtendedAddress tableEntry(character, reinterpret_cast<intptr_t>(charClass->m_table)); |
385 | matchDest.append(branchTest8(charClass->m_tableInverted ? Zero : NonZero, tableEntry)); |
386 | return; |
387 | } |
388 | |
389 | JumpList unicodeFail; |
390 | if (charClass->m_matchesUnicode.size() || charClass->m_rangesUnicode.size()) { |
391 | JumpList isAscii; |
392 | if (charClass->m_matches.size() || charClass->m_ranges.size()) |
393 | isAscii.append(branch32(LessThanOrEqual, character, TrustedImm32(0x7f))); |
394 | |
395 | if (charClass->m_matchesUnicode.size()) { |
396 | for (unsigned i = 0; i < charClass->m_matchesUnicode.size(); ++i) { |
397 | UChar32 ch = charClass->m_matchesUnicode[i]; |
398 | matchDest.append(branch32(Equal, character, Imm32(ch))); |
399 | } |
400 | } |
401 | |
402 | if (charClass->m_rangesUnicode.size()) { |
403 | for (unsigned i = 0; i < charClass->m_rangesUnicode.size(); ++i) { |
404 | UChar32 lo = charClass->m_rangesUnicode[i].begin; |
405 | UChar32 hi = charClass->m_rangesUnicode[i].end; |
406 | |
407 | Jump below = branch32(LessThan, character, Imm32(lo)); |
408 | matchDest.append(branch32(LessThanOrEqual, character, Imm32(hi))); |
409 | below.link(this); |
410 | } |
411 | } |
412 | |
413 | if (charClass->m_matches.size() || charClass->m_ranges.size()) |
414 | unicodeFail = jump(); |
415 | isAscii.link(this); |
416 | } |
417 | |
418 | if (charClass->m_ranges.size()) { |
419 | unsigned matchIndex = 0; |
420 | JumpList failures; |
421 | matchCharacterClassRange(character, failures, matchDest, charClass->m_ranges.begin(), charClass->m_ranges.size(), &matchIndex, charClass->m_matches.begin(), charClass->m_matches.size()); |
422 | while (matchIndex < charClass->m_matches.size()) |
423 | matchDest.append(branch32(Equal, character, Imm32((unsigned short)charClass->m_matches[matchIndex++]))); |
424 | |
425 | failures.link(this); |
426 | } else if (charClass->m_matches.size()) { |
427 | // optimization: gather 'a','A' etc back together, can mask & test once. |
428 | Vector<char> matchesAZaz; |
429 | |
430 | for (unsigned i = 0; i < charClass->m_matches.size(); ++i) { |
431 | char ch = charClass->m_matches[i]; |
432 | if (m_pattern.ignoreCase()) { |
433 | if (isASCIILower(ch)) { |
434 | matchesAZaz.append(ch); |
435 | continue; |
436 | } |
437 | if (isASCIIUpper(ch)) |
438 | continue; |
439 | } |
440 | matchDest.append(branch32(Equal, character, Imm32((unsigned short)ch))); |
441 | } |
442 | |
443 | if (unsigned countAZaz = matchesAZaz.size()) { |
444 | or32(TrustedImm32(32), character); |
445 | for (unsigned i = 0; i < countAZaz; ++i) |
446 | matchDest.append(branch32(Equal, character, TrustedImm32(matchesAZaz[i]))); |
447 | } |
448 | } |
449 | |
450 | if (charClass->m_matchesUnicode.size() || charClass->m_rangesUnicode.size()) |
451 | unicodeFail.link(this); |
452 | } |
453 | |
454 | #ifdef JIT_UNICODE_EXPRESSIONS |
455 | void advanceIndexAfterCharacterClassTermMatch(const PatternTerm* term, JumpList& failures, const RegisterID character) |
456 | { |
457 | ASSERT(term->type == PatternTerm::TypeCharacterClass); |
458 | |
459 | if (term->characterClass->hasOneCharacterSize() && !term->invert()) |
460 | add32(TrustedImm32(term->characterClass->hasNonBMPCharacters() ? 2 : 1), index); |
461 | else { |
462 | add32(TrustedImm32(1), index); |
463 | failures.append(atEndOfInput()); |
464 | Jump isBMPChar = branch32(LessThan, character, supplementaryPlanesBase); |
465 | add32(TrustedImm32(1), index); |
466 | isBMPChar.link(this); |
467 | } |
468 | } |
469 | #endif |
470 | |
471 | // Jumps if input not available; will have (incorrectly) incremented already! |
472 | Jump jumpIfNoAvailableInput(unsigned countToCheck = 0) |
473 | { |
474 | if (countToCheck) |
475 | add32(Imm32(countToCheck), index); |
476 | return branch32(Above, index, length); |
477 | } |
478 | |
479 | Jump jumpIfAvailableInput(unsigned countToCheck) |
480 | { |
481 | add32(Imm32(countToCheck), index); |
482 | return branch32(BelowOrEqual, index, length); |
483 | } |
484 | |
485 | Jump checkNotEnoughInput(RegisterID additionalAmount) |
486 | { |
487 | add32(index, additionalAmount); |
488 | return branch32(Above, additionalAmount, length); |
489 | } |
490 | |
491 | Jump checkInput() |
492 | { |
493 | return branch32(BelowOrEqual, index, length); |
494 | } |
495 | |
496 | Jump atEndOfInput() |
497 | { |
498 | return branch32(Equal, index, length); |
499 | } |
500 | |
501 | Jump notAtEndOfInput() |
502 | { |
503 | return branch32(NotEqual, index, length); |
504 | } |
505 | |
506 | BaseIndex negativeOffsetIndexedAddress(Checked<unsigned> negativeCharacterOffset, RegisterID tempReg, RegisterID indexReg = index) |
507 | { |
508 | RegisterID base = input; |
509 | |
510 | // BaseIndex() addressing can take a int32_t offset. Given that we can have a regular |
511 | // expression that has unsigned character offsets, BaseIndex's signed offset is insufficient |
512 | // for addressing in extreme cases where we might underflow. Therefore we check to see if |
513 | // negativeCharacterOffset will underflow directly or after converting for 16 bit characters. |
514 | // If so, we do our own address calculating by adjusting the base, using the result register |
515 | // as a temp address register. |
516 | unsigned maximumNegativeOffsetForCharacterSize = m_charSize == Char8 ? 0x7fffffff : 0x3fffffff; |
517 | unsigned offsetAdjustAmount = 0x40000000; |
518 | if (negativeCharacterOffset.unsafeGet() > maximumNegativeOffsetForCharacterSize) { |
519 | base = tempReg; |
520 | move(input, base); |
521 | while (negativeCharacterOffset.unsafeGet() > maximumNegativeOffsetForCharacterSize) { |
522 | subPtr(TrustedImm32(offsetAdjustAmount), base); |
523 | if (m_charSize != Char8) |
524 | subPtr(TrustedImm32(offsetAdjustAmount), base); |
525 | negativeCharacterOffset -= offsetAdjustAmount; |
526 | } |
527 | } |
528 | |
529 | Checked<int32_t> characterOffset(-static_cast<int32_t>(negativeCharacterOffset.unsafeGet())); |
530 | |
531 | if (m_charSize == Char8) |
532 | return BaseIndex(input, indexReg, TimesOne, (characterOffset * static_cast<int32_t>(sizeof(char))).unsafeGet()); |
533 | |
534 | return BaseIndex(input, indexReg, TimesTwo, (characterOffset * static_cast<int32_t>(sizeof(UChar))).unsafeGet()); |
535 | } |
536 | |
537 | #ifdef JIT_UNICODE_EXPRESSIONS |
538 | void tryReadUnicodeCharImpl(RegisterID resultReg) |
539 | { |
540 | ASSERT(m_charSize == Char16); |
541 | |
542 | JumpList notUnicode; |
543 | |
544 | load16Unaligned(regUnicodeInputAndTrail, resultReg); |
545 | and32(surrogateTagMask, resultReg, regT2); |
546 | notUnicode.append(branch32(NotEqual, regT2, leadingSurrogateTag)); |
547 | addPtr(TrustedImm32(2), regUnicodeInputAndTrail); |
548 | notUnicode.append(branchPtr(AboveOrEqual, regUnicodeInputAndTrail, endOfStringAddress)); |
549 | load16Unaligned(Address(regUnicodeInputAndTrail), regUnicodeInputAndTrail); |
550 | and32(surrogateTagMask, regUnicodeInputAndTrail, regT2); |
551 | notUnicode.append(branch32(NotEqual, regT2, trailingSurrogateTag)); |
552 | sub32(leadingSurrogateTag, resultReg); |
553 | sub32(trailingSurrogateTag, regUnicodeInputAndTrail); |
554 | lshift32(TrustedImm32(10), resultReg); |
555 | or32(regUnicodeInputAndTrail, resultReg); |
556 | add32(supplementaryPlanesBase, resultReg); |
557 | notUnicode.link(this); |
558 | } |
559 | |
560 | void tryReadUnicodeChar(BaseIndex address, RegisterID resultReg) |
561 | { |
562 | ASSERT(m_charSize == Char16); |
563 | |
564 | getEffectiveAddress(address, regUnicodeInputAndTrail); |
565 | |
566 | if (resultReg == regT0) |
567 | m_tryReadUnicodeCharacterCalls.append(nearCall()); |
568 | else |
569 | tryReadUnicodeCharImpl(resultReg); |
570 | } |
571 | #endif |
572 | |
573 | void readCharacterDontDecodeSurrogates(Checked<unsigned> negativeCharacterOffset, RegisterID resultReg, RegisterID indexReg = index) |
574 | { |
575 | BaseIndex address = negativeOffsetIndexedAddress(negativeCharacterOffset, resultReg, indexReg); |
576 | |
577 | if (m_charSize == Char8) |
578 | load8(address, resultReg); |
579 | else |
580 | load16Unaligned(address, resultReg); |
581 | } |
582 | |
583 | void readCharacter(Checked<unsigned> negativeCharacterOffset, RegisterID resultReg, RegisterID indexReg = index) |
584 | { |
585 | BaseIndex address = negativeOffsetIndexedAddress(negativeCharacterOffset, resultReg, indexReg); |
586 | |
587 | if (m_charSize == Char8) |
588 | load8(address, resultReg); |
589 | #ifdef JIT_UNICODE_EXPRESSIONS |
590 | else if (m_decodeSurrogatePairs) |
591 | tryReadUnicodeChar(address, resultReg); |
592 | #endif |
593 | else |
594 | load16Unaligned(address, resultReg); |
595 | } |
596 | |
597 | Jump jumpIfCharNotEquals(UChar32 ch, Checked<unsigned> negativeCharacterOffset, RegisterID character) |
598 | { |
599 | readCharacter(negativeCharacterOffset, character); |
600 | |
601 | // For case-insesitive compares, non-ascii characters that have different |
602 | // upper & lower case representations are converted to a character class. |
603 | ASSERT(!m_pattern.ignoreCase() || isASCIIAlpha(ch) || isCanonicallyUnique(ch, m_canonicalMode)); |
604 | if (m_pattern.ignoreCase() && isASCIIAlpha(ch)) { |
605 | or32(TrustedImm32(0x20), character); |
606 | ch |= 0x20; |
607 | } |
608 | |
609 | return branch32(NotEqual, character, Imm32(ch)); |
610 | } |
611 | |
612 | void storeToFrame(RegisterID reg, unsigned frameLocation) |
613 | { |
614 | poke(reg, frameLocation); |
615 | } |
616 | |
617 | void storeToFrame(TrustedImm32 imm, unsigned frameLocation) |
618 | { |
619 | poke(imm, frameLocation); |
620 | } |
621 | |
622 | #if CPU(ARM64) || CPU(X86_64) |
623 | void storeToFrame(TrustedImmPtr imm, unsigned frameLocation) |
624 | { |
625 | poke(imm, frameLocation); |
626 | } |
627 | #endif |
628 | |
629 | DataLabelPtr storeToFrameWithPatch(unsigned frameLocation) |
630 | { |
631 | return storePtrWithPatch(TrustedImmPtr(nullptr), Address(stackPointerRegister, frameLocation * sizeof(void*))); |
632 | } |
633 | |
634 | void loadFromFrame(unsigned frameLocation, RegisterID reg) |
635 | { |
636 | peek(reg, frameLocation); |
637 | } |
638 | |
639 | void loadFromFrameAndJump(unsigned frameLocation) |
640 | { |
641 | jump(Address(stackPointerRegister, frameLocation * sizeof(void*)), YarrBacktrackPtrTag); |
642 | } |
643 | |
644 | unsigned alignCallFrameSizeInBytes(unsigned callFrameSize) |
645 | { |
646 | if (!callFrameSize) |
647 | return 0; |
648 | |
649 | callFrameSize *= sizeof(void*); |
650 | if (callFrameSize / sizeof(void*) != m_pattern.m_body->m_callFrameSize) |
651 | CRASH(); |
652 | callFrameSize = (callFrameSize + 0x3f) & ~0x3f; |
653 | return callFrameSize; |
654 | } |
655 | void initCallFrame() |
656 | { |
657 | unsigned callFrameSizeInBytes = alignCallFrameSizeInBytes(m_pattern.m_body->m_callFrameSize); |
658 | if (callFrameSizeInBytes) { |
659 | #if CPU(X86_64) || CPU(ARM64) |
660 | if (Options::zeroStackFrame()) { |
661 | // We need to start from the stack pointer, because we could have spilled callee saves |
662 | move(stackPointerRegister, regT0); |
663 | subPtr(Imm32(callFrameSizeInBytes), stackPointerRegister); |
664 | if (callFrameSizeInBytes <= 128) { |
665 | for (unsigned offset = 0; offset < callFrameSizeInBytes; offset += sizeof(intptr_t)) |
666 | storePtr(TrustedImm32(0), Address(regT0, -8 - offset)); |
667 | } else { |
668 | Label zeroLoop = label(); |
669 | subPtr(TrustedImm32(sizeof(intptr_t) * 2), regT0); |
670 | #if CPU(ARM64) |
671 | storePair64(ARM64Registers::zr, ARM64Registers::zr, regT0); |
672 | #else |
673 | storePtr(TrustedImm32(0), Address(regT0)); |
674 | storePtr(TrustedImm32(0), Address(regT0, sizeof(intptr_t))); |
675 | #endif |
676 | branchPtr(NotEqual, regT0, stackPointerRegister).linkTo(zeroLoop, this); |
677 | } |
678 | } else |
679 | #endif |
680 | subPtr(Imm32(callFrameSizeInBytes), stackPointerRegister); |
681 | |
682 | } |
683 | } |
684 | void removeCallFrame() |
685 | { |
686 | unsigned callFrameSizeInBytes = alignCallFrameSizeInBytes(m_pattern.m_body->m_callFrameSize); |
687 | if (callFrameSizeInBytes) |
688 | addPtr(Imm32(callFrameSizeInBytes), stackPointerRegister); |
689 | } |
690 | |
691 | void generateFailReturn() |
692 | { |
693 | move(TrustedImmPtr((void*)WTF::notFound), returnRegister); |
694 | move(TrustedImm32(0), returnRegister2); |
695 | generateReturn(); |
696 | } |
697 | |
698 | void generateJITFailReturn() |
699 | { |
700 | if (m_abortExecution.empty() && m_hitMatchLimit.empty()) |
701 | return; |
702 | |
703 | JumpList finishExiting; |
704 | if (!m_abortExecution.empty()) { |
705 | m_abortExecution.link(this); |
706 | move(TrustedImmPtr((void*)static_cast<size_t>(-2)), returnRegister); |
707 | finishExiting.append(jump()); |
708 | } |
709 | |
710 | if (!m_hitMatchLimit.empty()) { |
711 | m_hitMatchLimit.link(this); |
712 | move(TrustedImmPtr((void*)static_cast<size_t>(-1)), returnRegister); |
713 | } |
714 | |
715 | finishExiting.link(this); |
716 | removeCallFrame(); |
717 | move(TrustedImm32(0), returnRegister2); |
718 | generateReturn(); |
719 | } |
720 | |
721 | // Used to record subpatterns, should only be called if compileMode is IncludeSubpatterns. |
722 | void setSubpatternStart(RegisterID reg, unsigned subpattern) |
723 | { |
724 | ASSERT(subpattern); |
725 | // FIXME: should be able to ASSERT(compileMode == IncludeSubpatterns), but then this function is conditionally NORETURN. :-( |
726 | store32(reg, Address(output, (subpattern << 1) * sizeof(int))); |
727 | } |
728 | void setSubpatternEnd(RegisterID reg, unsigned subpattern) |
729 | { |
730 | ASSERT(subpattern); |
731 | // FIXME: should be able to ASSERT(compileMode == IncludeSubpatterns), but then this function is conditionally NORETURN. :-( |
732 | store32(reg, Address(output, ((subpattern << 1) + 1) * sizeof(int))); |
733 | } |
734 | void clearSubpatternStart(unsigned subpattern) |
735 | { |
736 | ASSERT(subpattern); |
737 | // FIXME: should be able to ASSERT(compileMode == IncludeSubpatterns), but then this function is conditionally NORETURN. :-( |
738 | store32(TrustedImm32(-1), Address(output, (subpattern << 1) * sizeof(int))); |
739 | } |
740 | |
741 | void clearMatches(unsigned subpattern, unsigned lastSubpattern) |
742 | { |
743 | for (; subpattern <= lastSubpattern; subpattern++) |
744 | clearSubpatternStart(subpattern); |
745 | } |
746 | |
747 | // We use one of three different strategies to track the start of the current match, |
748 | // while matching. |
749 | // 1) If the pattern has a fixed size, do nothing! - we calculate the value lazily |
750 | // at the end of matching. This is irrespective of compileMode, and in this case |
751 | // these methods should never be called. |
752 | // 2) If we're compiling IncludeSubpatterns, 'output' contains a pointer to an output |
753 | // vector, store the match start in the output vector. |
754 | // 3) If we're compiling MatchOnly, 'output' is unused, store the match start directly |
755 | // in this register. |
756 | void setMatchStart(RegisterID reg) |
757 | { |
758 | ASSERT(!m_pattern.m_body->m_hasFixedSize); |
759 | if (compileMode == IncludeSubpatterns) |
760 | store32(reg, output); |
761 | else |
762 | move(reg, output); |
763 | } |
764 | void getMatchStart(RegisterID reg) |
765 | { |
766 | ASSERT(!m_pattern.m_body->m_hasFixedSize); |
767 | if (compileMode == IncludeSubpatterns) |
768 | load32(output, reg); |
769 | else |
770 | move(output, reg); |
771 | } |
772 | |
773 | enum YarrOpCode : uint8_t { |
774 | // These nodes wrap body alternatives - those in the main disjunction, |
775 | // rather than subpatterns or assertions. These are chained together in |
776 | // a doubly linked list, with a 'begin' node for the first alternative, |
777 | // a 'next' node for each subsequent alternative, and an 'end' node at |
778 | // the end. In the case of repeating alternatives, the 'end' node also |
779 | // has a reference back to 'begin'. |
780 | OpBodyAlternativeBegin, |
781 | OpBodyAlternativeNext, |
782 | OpBodyAlternativeEnd, |
783 | // Similar to the body alternatives, but used for subpatterns with two |
784 | // or more alternatives. |
785 | OpNestedAlternativeBegin, |
786 | OpNestedAlternativeNext, |
787 | OpNestedAlternativeEnd, |
788 | // Used for alternatives in subpatterns where there is only a single |
789 | // alternative (backtracking is easier in these cases), or for alternatives |
790 | // which never need to be backtracked (those in parenthetical assertions, |
791 | // terminal subpatterns). |
792 | OpSimpleNestedAlternativeBegin, |
793 | OpSimpleNestedAlternativeNext, |
794 | OpSimpleNestedAlternativeEnd, |
795 | // Used to wrap 'Once' subpattern matches (quantityMaxCount == 1). |
796 | OpParenthesesSubpatternOnceBegin, |
797 | OpParenthesesSubpatternOnceEnd, |
798 | // Used to wrap 'Terminal' subpattern matches (at the end of the regexp). |
799 | OpParenthesesSubpatternTerminalBegin, |
800 | OpParenthesesSubpatternTerminalEnd, |
801 | // Used to wrap generic captured matches |
802 | OpParenthesesSubpatternBegin, |
803 | OpParenthesesSubpatternEnd, |
804 | // Used to wrap parenthetical assertions. |
805 | OpParentheticalAssertionBegin, |
806 | OpParentheticalAssertionEnd, |
807 | // Wraps all simple terms (pattern characters, character classes). |
808 | OpTerm, |
809 | // Where an expression contains only 'once through' body alternatives |
810 | // and no repeating ones, this op is used to return match failure. |
811 | OpMatchFailed |
812 | }; |
813 | |
814 | // This structure is used to hold the compiled opcode information, |
815 | // including reference back to the original PatternTerm/PatternAlternatives, |
816 | // and JIT compilation data structures. |
817 | struct YarrOp { |
818 | explicit YarrOp(PatternTerm* term) |
819 | : m_term(term) |
820 | , m_op(OpTerm) |
821 | , m_isDeadCode(false) |
822 | { |
823 | } |
824 | |
825 | explicit YarrOp(YarrOpCode op) |
826 | : m_op(op) |
827 | , m_isDeadCode(false) |
828 | { |
829 | } |
830 | |
831 | // For alternatives, this holds the PatternAlternative and doubly linked |
832 | // references to this alternative's siblings. In the case of the |
833 | // OpBodyAlternativeEnd node at the end of a section of repeating nodes, |
834 | // m_nextOp will reference the OpBodyAlternativeBegin node of the first |
835 | // repeating alternative. |
836 | PatternAlternative* m_alternative; |
837 | size_t m_previousOp; |
838 | size_t m_nextOp; |
839 | |
840 | // The operation, as a YarrOpCode, and also a reference to the PatternTerm. |
841 | PatternTerm* m_term; |
842 | YarrOpCode m_op; |
843 | |
844 | // Used to record a set of Jumps out of the generated code, typically |
845 | // used for jumps out to backtracking code, and a single reentry back |
846 | // into the code for a node (likely where a backtrack will trigger |
847 | // rematching). |
848 | Label m_reentry; |
849 | JumpList m_jumps; |
850 | |
851 | // Used for backtracking when the prior alternative did not consume any |
852 | // characters but matched. |
853 | Jump m_zeroLengthMatch; |
854 | |
855 | // This flag is used to null out the second pattern character, when |
856 | // two are fused to match a pair together. |
857 | bool m_isDeadCode; |
858 | |
859 | // Currently used in the case of some of the more complex management of |
860 | // 'm_checkedOffset', to cache the offset used in this alternative, to avoid |
861 | // recalculating it. |
862 | Checked<unsigned> m_checkAdjust; |
863 | |
864 | // Used by OpNestedAlternativeNext/End to hold the pointer to the |
865 | // value that will be pushed into the pattern's frame to return to, |
866 | // upon backtracking back into the disjunction. |
867 | DataLabelPtr m_returnAddress; |
868 | }; |
869 | |
870 | // BacktrackingState |
871 | // This class encapsulates information about the state of code generation |
872 | // whilst generating the code for backtracking, when a term fails to match. |
873 | // Upon entry to code generation of the backtracking code for a given node, |
874 | // the Backtracking state will hold references to all control flow sources |
875 | // that are outputs in need of further backtracking from the prior node |
876 | // generated (which is the subsequent operation in the regular expression, |
877 | // and in the m_ops Vector, since we generated backtracking backwards). |
878 | // These references to control flow take the form of: |
879 | // - A jump list of jumps, to be linked to code that will backtrack them |
880 | // further. |
881 | // - A set of DataLabelPtr values, to be populated with values to be |
882 | // treated effectively as return addresses backtracking into complex |
883 | // subpatterns. |
884 | // - A flag indicating that the current sequence of generated code up to |
885 | // this point requires backtracking. |
886 | class BacktrackingState { |
887 | public: |
888 | BacktrackingState() |
889 | : m_pendingFallthrough(false) |
890 | { |
891 | } |
892 | |
893 | // Add a jump or jumps, a return address, or set the flag indicating |
894 | // that the current 'fallthrough' control flow requires backtracking. |
895 | void append(const Jump& jump) |
896 | { |
897 | m_laterFailures.append(jump); |
898 | } |
899 | void append(JumpList& jumpList) |
900 | { |
901 | m_laterFailures.append(jumpList); |
902 | } |
903 | void append(const DataLabelPtr& returnAddress) |
904 | { |
905 | m_pendingReturns.append(returnAddress); |
906 | } |
907 | void fallthrough() |
908 | { |
909 | ASSERT(!m_pendingFallthrough); |
910 | m_pendingFallthrough = true; |
911 | } |
912 | |
913 | // These methods clear the backtracking state, either linking to the |
914 | // current location, a provided label, or copying the backtracking out |
915 | // to a JumpList. All actions may require code generation to take place, |
916 | // and as such are passed a pointer to the assembler. |
917 | void link(MacroAssembler* assembler) |
918 | { |
919 | if (m_pendingReturns.size()) { |
920 | Label here(assembler); |
921 | for (unsigned i = 0; i < m_pendingReturns.size(); ++i) |
922 | m_backtrackRecords.append(ReturnAddressRecord(m_pendingReturns[i], here)); |
923 | m_pendingReturns.clear(); |
924 | } |
925 | m_laterFailures.link(assembler); |
926 | m_laterFailures.clear(); |
927 | m_pendingFallthrough = false; |
928 | } |
929 | void linkTo(Label label, MacroAssembler* assembler) |
930 | { |
931 | if (m_pendingReturns.size()) { |
932 | for (unsigned i = 0; i < m_pendingReturns.size(); ++i) |
933 | m_backtrackRecords.append(ReturnAddressRecord(m_pendingReturns[i], label)); |
934 | m_pendingReturns.clear(); |
935 | } |
936 | if (m_pendingFallthrough) |
937 | assembler->jump(label); |
938 | m_laterFailures.linkTo(label, assembler); |
939 | m_laterFailures.clear(); |
940 | m_pendingFallthrough = false; |
941 | } |
942 | void takeBacktracksToJumpList(JumpList& jumpList, MacroAssembler* assembler) |
943 | { |
944 | if (m_pendingReturns.size()) { |
945 | Label here(assembler); |
946 | for (unsigned i = 0; i < m_pendingReturns.size(); ++i) |
947 | m_backtrackRecords.append(ReturnAddressRecord(m_pendingReturns[i], here)); |
948 | m_pendingReturns.clear(); |
949 | m_pendingFallthrough = true; |
950 | } |
951 | if (m_pendingFallthrough) |
952 | jumpList.append(assembler->jump()); |
953 | jumpList.append(m_laterFailures); |
954 | m_laterFailures.clear(); |
955 | m_pendingFallthrough = false; |
956 | } |
957 | |
958 | bool isEmpty() |
959 | { |
960 | return m_laterFailures.empty() && m_pendingReturns.isEmpty() && !m_pendingFallthrough; |
961 | } |
962 | |
963 | // Called at the end of code generation to link all return addresses. |
964 | void linkDataLabels(LinkBuffer& linkBuffer) |
965 | { |
966 | ASSERT(isEmpty()); |
967 | for (unsigned i = 0; i < m_backtrackRecords.size(); ++i) |
968 | linkBuffer.patch(m_backtrackRecords[i].m_dataLabel, linkBuffer.locationOf<YarrBacktrackPtrTag>(m_backtrackRecords[i].m_backtrackLocation)); |
969 | } |
970 | |
971 | private: |
972 | struct ReturnAddressRecord { |
973 | ReturnAddressRecord(DataLabelPtr dataLabel, Label backtrackLocation) |
974 | : m_dataLabel(dataLabel) |
975 | , m_backtrackLocation(backtrackLocation) |
976 | { |
977 | } |
978 | |
979 | DataLabelPtr m_dataLabel; |
980 | Label m_backtrackLocation; |
981 | }; |
982 | |
983 | JumpList m_laterFailures; |
984 | bool m_pendingFallthrough; |
985 | Vector<DataLabelPtr, 4> m_pendingReturns; |
986 | Vector<ReturnAddressRecord, 4> m_backtrackRecords; |
987 | }; |
988 | |
989 | // Generation methods: |
990 | // =================== |
991 | |
992 | // This method provides a default implementation of backtracking common |
993 | // to many terms; terms commonly jump out of the forwards matching path |
994 | // on any failed conditions, and add these jumps to the m_jumps list. If |
995 | // no special handling is required we can often just backtrack to m_jumps. |
996 | void backtrackTermDefault(size_t opIndex) |
997 | { |
998 | YarrOp& op = m_ops[opIndex]; |
999 | m_backtrackingState.append(op.m_jumps); |
1000 | } |
1001 | |
1002 | void generateAssertionBOL(size_t opIndex) |
1003 | { |
1004 | YarrOp& op = m_ops[opIndex]; |
1005 | PatternTerm* term = op.m_term; |
1006 | |
1007 | if (m_pattern.multiline()) { |
1008 | const RegisterID character = regT0; |
1009 | |
1010 | JumpList matchDest; |
1011 | if (!term->inputPosition) |
1012 | matchDest.append(branch32(Equal, index, Imm32(m_checkedOffset.unsafeGet()))); |
1013 | |
1014 | readCharacter(m_checkedOffset - term->inputPosition + 1, character); |
1015 | matchCharacterClass(character, matchDest, m_pattern.newlineCharacterClass()); |
1016 | op.m_jumps.append(jump()); |
1017 | |
1018 | matchDest.link(this); |
1019 | } else { |
1020 | // Erk, really should poison out these alternatives early. :-/ |
1021 | if (term->inputPosition) |
1022 | op.m_jumps.append(jump()); |
1023 | else |
1024 | op.m_jumps.append(branch32(NotEqual, index, Imm32(m_checkedOffset.unsafeGet()))); |
1025 | } |
1026 | } |
1027 | void backtrackAssertionBOL(size_t opIndex) |
1028 | { |
1029 | backtrackTermDefault(opIndex); |
1030 | } |
1031 | |
1032 | void generateAssertionEOL(size_t opIndex) |
1033 | { |
1034 | YarrOp& op = m_ops[opIndex]; |
1035 | PatternTerm* term = op.m_term; |
1036 | |
1037 | if (m_pattern.multiline()) { |
1038 | const RegisterID character = regT0; |
1039 | |
1040 | JumpList matchDest; |
1041 | if (term->inputPosition == m_checkedOffset.unsafeGet()) |
1042 | matchDest.append(atEndOfInput()); |
1043 | |
1044 | readCharacter(m_checkedOffset - term->inputPosition, character); |
1045 | matchCharacterClass(character, matchDest, m_pattern.newlineCharacterClass()); |
1046 | op.m_jumps.append(jump()); |
1047 | |
1048 | matchDest.link(this); |
1049 | } else { |
1050 | if (term->inputPosition == m_checkedOffset.unsafeGet()) |
1051 | op.m_jumps.append(notAtEndOfInput()); |
1052 | // Erk, really should poison out these alternatives early. :-/ |
1053 | else |
1054 | op.m_jumps.append(jump()); |
1055 | } |
1056 | } |
1057 | void backtrackAssertionEOL(size_t opIndex) |
1058 | { |
1059 | backtrackTermDefault(opIndex); |
1060 | } |
1061 | |
1062 | // Also falls though on nextIsNotWordChar. |
1063 | void matchAssertionWordchar(size_t opIndex, JumpList& nextIsWordChar, JumpList& nextIsNotWordChar) |
1064 | { |
1065 | YarrOp& op = m_ops[opIndex]; |
1066 | PatternTerm* term = op.m_term; |
1067 | |
1068 | const RegisterID character = regT0; |
1069 | |
1070 | if (term->inputPosition == m_checkedOffset.unsafeGet()) |
1071 | nextIsNotWordChar.append(atEndOfInput()); |
1072 | |
1073 | readCharacter(m_checkedOffset - term->inputPosition, character); |
1074 | |
1075 | CharacterClass* wordcharCharacterClass; |
1076 | |
1077 | if (m_unicodeIgnoreCase) |
1078 | wordcharCharacterClass = m_pattern.wordUnicodeIgnoreCaseCharCharacterClass(); |
1079 | else |
1080 | wordcharCharacterClass = m_pattern.wordcharCharacterClass(); |
1081 | |
1082 | matchCharacterClass(character, nextIsWordChar, wordcharCharacterClass); |
1083 | } |
1084 | |
1085 | void generateAssertionWordBoundary(size_t opIndex) |
1086 | { |
1087 | YarrOp& op = m_ops[opIndex]; |
1088 | PatternTerm* term = op.m_term; |
1089 | |
1090 | const RegisterID character = regT0; |
1091 | |
1092 | Jump atBegin; |
1093 | JumpList matchDest; |
1094 | if (!term->inputPosition) |
1095 | atBegin = branch32(Equal, index, Imm32(m_checkedOffset.unsafeGet())); |
1096 | readCharacter(m_checkedOffset - term->inputPosition + 1, character); |
1097 | |
1098 | CharacterClass* wordcharCharacterClass; |
1099 | |
1100 | if (m_unicodeIgnoreCase) |
1101 | wordcharCharacterClass = m_pattern.wordUnicodeIgnoreCaseCharCharacterClass(); |
1102 | else |
1103 | wordcharCharacterClass = m_pattern.wordcharCharacterClass(); |
1104 | |
1105 | matchCharacterClass(character, matchDest, wordcharCharacterClass); |
1106 | if (!term->inputPosition) |
1107 | atBegin.link(this); |
1108 | |
1109 | // We fall through to here if the last character was not a wordchar. |
1110 | JumpList nonWordCharThenWordChar; |
1111 | JumpList nonWordCharThenNonWordChar; |
1112 | if (term->invert()) { |
1113 | matchAssertionWordchar(opIndex, nonWordCharThenNonWordChar, nonWordCharThenWordChar); |
1114 | nonWordCharThenWordChar.append(jump()); |
1115 | } else { |
1116 | matchAssertionWordchar(opIndex, nonWordCharThenWordChar, nonWordCharThenNonWordChar); |
1117 | nonWordCharThenNonWordChar.append(jump()); |
1118 | } |
1119 | op.m_jumps.append(nonWordCharThenNonWordChar); |
1120 | |
1121 | // We jump here if the last character was a wordchar. |
1122 | matchDest.link(this); |
1123 | JumpList wordCharThenWordChar; |
1124 | JumpList wordCharThenNonWordChar; |
1125 | if (term->invert()) { |
1126 | matchAssertionWordchar(opIndex, wordCharThenNonWordChar, wordCharThenWordChar); |
1127 | wordCharThenWordChar.append(jump()); |
1128 | } else { |
1129 | matchAssertionWordchar(opIndex, wordCharThenWordChar, wordCharThenNonWordChar); |
1130 | // This can fall-though! |
1131 | } |
1132 | |
1133 | op.m_jumps.append(wordCharThenWordChar); |
1134 | |
1135 | nonWordCharThenWordChar.link(this); |
1136 | wordCharThenNonWordChar.link(this); |
1137 | } |
1138 | void backtrackAssertionWordBoundary(size_t opIndex) |
1139 | { |
1140 | backtrackTermDefault(opIndex); |
1141 | } |
1142 | |
1143 | #if ENABLE(YARR_JIT_BACKREFERENCES) |
1144 | void matchBackreference(size_t opIndex, JumpList& characterMatchFails, RegisterID character, RegisterID patternIndex, RegisterID patternCharacter) |
1145 | { |
1146 | YarrOp& op = m_ops[opIndex]; |
1147 | PatternTerm* term = op.m_term; |
1148 | unsigned subpatternId = term->backReferenceSubpatternId; |
1149 | |
1150 | Label loop(this); |
1151 | |
1152 | readCharacterDontDecodeSurrogates(0, patternCharacter, patternIndex); |
1153 | readCharacterDontDecodeSurrogates(m_checkedOffset - term->inputPosition, character); |
1154 | |
1155 | if (!m_pattern.ignoreCase()) |
1156 | characterMatchFails.append(branch32(NotEqual, character, patternCharacter)); |
1157 | else { |
1158 | Jump charactersMatch = branch32(Equal, character, patternCharacter); |
1159 | ExtendedAddress characterTableEntry(character, reinterpret_cast<intptr_t>(&canonicalTableLChar)); |
1160 | load16(characterTableEntry, character); |
1161 | ExtendedAddress patternTableEntry(patternCharacter, reinterpret_cast<intptr_t>(&canonicalTableLChar)); |
1162 | load16(patternTableEntry, patternCharacter); |
1163 | characterMatchFails.append(branch32(NotEqual, character, patternCharacter)); |
1164 | charactersMatch.link(this); |
1165 | } |
1166 | |
1167 | |
1168 | add32(TrustedImm32(1), index); |
1169 | add32(TrustedImm32(1), patternIndex); |
1170 | |
1171 | branch32(NotEqual, patternIndex, Address(output, ((subpatternId << 1) + 1) * sizeof(int))).linkTo(loop, this); |
1172 | } |
1173 | |
1174 | void generateBackReference(size_t opIndex) |
1175 | { |
1176 | YarrOp& op = m_ops[opIndex]; |
1177 | PatternTerm* term = op.m_term; |
1178 | |
1179 | if (m_pattern.ignoreCase() && m_charSize != Char8) { |
1180 | m_failureReason = JITFailureReason::BackReference; |
1181 | return; |
1182 | } |
1183 | |
1184 | unsigned subpatternId = term->backReferenceSubpatternId; |
1185 | unsigned parenthesesFrameLocation = term->frameLocation; |
1186 | |
1187 | const RegisterID characterOrTemp = regT0; |
1188 | const RegisterID patternIndex = regT1; |
1189 | const RegisterID patternTemp = regT2; |
1190 | |
1191 | storeToFrame(index, parenthesesFrameLocation + BackTrackInfoBackReference::beginIndex()); |
1192 | if (term->quantityType != QuantifierFixedCount || term->quantityMaxCount != 1) |
1193 | storeToFrame(TrustedImm32(0), parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex()); |
1194 | |
1195 | JumpList matches; |
1196 | |
1197 | if (term->quantityType != QuantifierNonGreedy) { |
1198 | load32(Address(output, (subpatternId << 1) * sizeof(int)), patternIndex); |
1199 | load32(Address(output, ((subpatternId << 1) + 1) * sizeof(int)), patternTemp); |
1200 | |
1201 | // An empty match is successful without consuming characters |
1202 | if (term->quantityType != QuantifierFixedCount || term->quantityMaxCount != 1) { |
1203 | matches.append(branch32(Equal, TrustedImm32(-1), patternIndex)); |
1204 | matches.append(branch32(Equal, patternIndex, patternTemp)); |
1205 | } else { |
1206 | Jump zeroLengthMatch = branch32(Equal, TrustedImm32(-1), patternIndex); |
1207 | Jump tryNonZeroMatch = branch32(NotEqual, patternIndex, patternTemp); |
1208 | zeroLengthMatch.link(this); |
1209 | storeToFrame(TrustedImm32(1), parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex()); |
1210 | matches.append(jump()); |
1211 | tryNonZeroMatch.link(this); |
1212 | } |
1213 | } |
1214 | |
1215 | switch (term->quantityType) { |
1216 | case QuantifierFixedCount: { |
1217 | Label outerLoop(this); |
1218 | |
1219 | // PatternTemp should contain pattern end index at this point |
1220 | sub32(patternIndex, patternTemp); |
1221 | op.m_jumps.append(checkNotEnoughInput(patternTemp)); |
1222 | |
1223 | matchBackreference(opIndex, op.m_jumps, characterOrTemp, patternIndex, patternTemp); |
1224 | |
1225 | if (term->quantityMaxCount != 1) { |
1226 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex(), characterOrTemp); |
1227 | add32(TrustedImm32(1), characterOrTemp); |
1228 | storeToFrame(characterOrTemp, parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex()); |
1229 | matches.append(branch32(Equal, Imm32(term->quantityMaxCount.unsafeGet()), characterOrTemp)); |
1230 | load32(Address(output, (subpatternId << 1) * sizeof(int)), patternIndex); |
1231 | load32(Address(output, ((subpatternId << 1) + 1) * sizeof(int)), patternTemp); |
1232 | jump(outerLoop); |
1233 | } |
1234 | matches.link(this); |
1235 | break; |
1236 | } |
1237 | |
1238 | case QuantifierGreedy: { |
1239 | JumpList incompleteMatches; |
1240 | |
1241 | Label outerLoop(this); |
1242 | |
1243 | // PatternTemp should contain pattern end index at this point |
1244 | sub32(patternIndex, patternTemp); |
1245 | matches.append(checkNotEnoughInput(patternTemp)); |
1246 | |
1247 | matchBackreference(opIndex, incompleteMatches, characterOrTemp, patternIndex, patternTemp); |
1248 | |
1249 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex(), characterOrTemp); |
1250 | add32(TrustedImm32(1), characterOrTemp); |
1251 | storeToFrame(characterOrTemp, parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex()); |
1252 | if (term->quantityMaxCount != quantifyInfinite) |
1253 | matches.append(branch32(Equal, Imm32(term->quantityMaxCount.unsafeGet()), characterOrTemp)); |
1254 | load32(Address(output, (subpatternId << 1) * sizeof(int)), patternIndex); |
1255 | load32(Address(output, ((subpatternId << 1) + 1) * sizeof(int)), patternTemp); |
1256 | |
1257 | // Store current index in frame for restoring after a partial match |
1258 | storeToFrame(index, parenthesesFrameLocation + BackTrackInfoBackReference::beginIndex()); |
1259 | jump(outerLoop); |
1260 | |
1261 | incompleteMatches.link(this); |
1262 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoBackReference::beginIndex(), index); |
1263 | |
1264 | matches.link(this); |
1265 | op.m_reentry = label(); |
1266 | break; |
1267 | } |
1268 | |
1269 | case QuantifierNonGreedy: { |
1270 | JumpList incompleteMatches; |
1271 | |
1272 | matches.append(jump()); |
1273 | |
1274 | op.m_reentry = label(); |
1275 | |
1276 | load32(Address(output, (subpatternId << 1) * sizeof(int)), patternIndex); |
1277 | load32(Address(output, ((subpatternId << 1) + 1) * sizeof(int)), patternTemp); |
1278 | |
1279 | // An empty match is successful without consuming characters |
1280 | Jump zeroLengthMatch = branch32(Equal, TrustedImm32(-1), patternIndex); |
1281 | Jump tryNonZeroMatch = branch32(NotEqual, patternIndex, patternTemp); |
1282 | zeroLengthMatch.link(this); |
1283 | storeToFrame(TrustedImm32(1), parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex()); |
1284 | matches.append(jump()); |
1285 | tryNonZeroMatch.link(this); |
1286 | |
1287 | // Check if we have input remaining to match |
1288 | sub32(patternIndex, patternTemp); |
1289 | matches.append(checkNotEnoughInput(patternTemp)); |
1290 | |
1291 | storeToFrame(index, parenthesesFrameLocation + BackTrackInfoBackReference::beginIndex()); |
1292 | |
1293 | matchBackreference(opIndex, incompleteMatches, characterOrTemp, patternIndex, patternTemp); |
1294 | |
1295 | matches.append(jump()); |
1296 | |
1297 | incompleteMatches.link(this); |
1298 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoBackReference::beginIndex(), index); |
1299 | |
1300 | matches.link(this); |
1301 | break; |
1302 | } |
1303 | } |
1304 | } |
1305 | void backtrackBackReference(size_t opIndex) |
1306 | { |
1307 | YarrOp& op = m_ops[opIndex]; |
1308 | PatternTerm* term = op.m_term; |
1309 | |
1310 | unsigned subpatternId = term->backReferenceSubpatternId; |
1311 | |
1312 | m_backtrackingState.link(this); |
1313 | op.m_jumps.link(this); |
1314 | |
1315 | JumpList failures; |
1316 | |
1317 | unsigned parenthesesFrameLocation = term->frameLocation; |
1318 | switch (term->quantityType) { |
1319 | case QuantifierFixedCount: |
1320 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoBackReference::beginIndex(), index); |
1321 | break; |
1322 | |
1323 | case QuantifierGreedy: { |
1324 | const RegisterID matchAmount = regT0; |
1325 | const RegisterID patternStartIndex = regT1; |
1326 | const RegisterID patternEndIndexOrLen = regT2; |
1327 | |
1328 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex(), matchAmount); |
1329 | failures.append(branchTest32(Zero, matchAmount)); |
1330 | |
1331 | load32(Address(output, (subpatternId << 1) * sizeof(int)), patternStartIndex); |
1332 | load32(Address(output, ((subpatternId << 1) + 1) * sizeof(int)), patternEndIndexOrLen); |
1333 | sub32(patternStartIndex, patternEndIndexOrLen); |
1334 | sub32(patternEndIndexOrLen, index); |
1335 | |
1336 | sub32(TrustedImm32(1), matchAmount); |
1337 | storeToFrame(matchAmount, parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex()); |
1338 | jump(op.m_reentry); |
1339 | break; |
1340 | } |
1341 | |
1342 | case QuantifierNonGreedy: { |
1343 | const RegisterID matchAmount = regT0; |
1344 | |
1345 | failures.append(atEndOfInput()); |
1346 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex(), matchAmount); |
1347 | if (term->quantityMaxCount != quantifyInfinite) |
1348 | failures.append(branch32(AboveOrEqual, Imm32(term->quantityMaxCount.unsafeGet()), matchAmount)); |
1349 | add32(TrustedImm32(1), matchAmount); |
1350 | storeToFrame(matchAmount, parenthesesFrameLocation + BackTrackInfoBackReference::matchAmountIndex()); |
1351 | jump(op.m_reentry); |
1352 | break; |
1353 | } |
1354 | } |
1355 | failures.link(this); |
1356 | m_backtrackingState.fallthrough(); |
1357 | } |
1358 | #endif |
1359 | |
1360 | void generatePatternCharacterOnce(size_t opIndex) |
1361 | { |
1362 | YarrOp& op = m_ops[opIndex]; |
1363 | |
1364 | if (op.m_isDeadCode) |
1365 | return; |
1366 | |
1367 | // m_ops always ends with a OpBodyAlternativeEnd or OpMatchFailed |
1368 | // node, so there must always be at least one more node. |
1369 | ASSERT(opIndex + 1 < m_ops.size()); |
1370 | YarrOp* nextOp = &m_ops[opIndex + 1]; |
1371 | |
1372 | PatternTerm* term = op.m_term; |
1373 | UChar32 ch = term->patternCharacter; |
1374 | |
1375 | if ((ch > 0xff) && (m_charSize == Char8)) { |
1376 | // Have a 16 bit pattern character and an 8 bit string - short circuit |
1377 | op.m_jumps.append(jump()); |
1378 | return; |
1379 | } |
1380 | |
1381 | const RegisterID character = regT0; |
1382 | #if CPU(X86_64) || CPU(ARM64) |
1383 | unsigned maxCharactersAtOnce = m_charSize == Char8 ? 8 : 4; |
1384 | #else |
1385 | unsigned maxCharactersAtOnce = m_charSize == Char8 ? 4 : 2; |
1386 | #endif |
1387 | uint64_t ignoreCaseMask = 0; |
1388 | #if CPU(BIG_ENDIAN) |
1389 | uint64_t allCharacters = ch << (m_charSize == Char8 ? 24 : 16); |
1390 | #else |
1391 | uint64_t allCharacters = ch; |
1392 | #endif |
1393 | unsigned numberCharacters; |
1394 | unsigned startTermPosition = term->inputPosition; |
1395 | |
1396 | // For case-insesitive compares, non-ascii characters that have different |
1397 | // upper & lower case representations are converted to a character class. |
1398 | ASSERT(!m_pattern.ignoreCase() || isASCIIAlpha(ch) || isCanonicallyUnique(ch, m_canonicalMode)); |
1399 | |
1400 | if (m_pattern.ignoreCase() && isASCIIAlpha(ch)) { |
1401 | #if CPU(BIG_ENDIAN) |
1402 | ignoreCaseMask |= 32 << (m_charSize == Char8 ? 24 : 16); |
1403 | #else |
1404 | ignoreCaseMask |= 32; |
1405 | #endif |
1406 | } |
1407 | |
1408 | for (numberCharacters = 1; numberCharacters < maxCharactersAtOnce && nextOp->m_op == OpTerm; ++numberCharacters, nextOp = &m_ops[opIndex + numberCharacters]) { |
1409 | PatternTerm* nextTerm = nextOp->m_term; |
1410 | |
1411 | // YarrJIT handles decoded surrogate pair as one character if unicode flag is enabled. |
1412 | // Note that the numberCharacters become 1 while the width of the pattern character becomes 32bit in this case. |
1413 | if (nextTerm->type != PatternTerm::TypePatternCharacter |
1414 | || nextTerm->quantityType != QuantifierFixedCount |
1415 | || nextTerm->quantityMaxCount != 1 |
1416 | || nextTerm->inputPosition != (startTermPosition + numberCharacters) |
1417 | || (U16_LENGTH(nextTerm->patternCharacter) != 1 && m_decodeSurrogatePairs)) |
1418 | break; |
1419 | |
1420 | nextOp->m_isDeadCode = true; |
1421 | |
1422 | #if CPU(BIG_ENDIAN) |
1423 | int shiftAmount = (m_charSize == Char8 ? 24 : 16) - ((m_charSize == Char8 ? 8 : 16) * numberCharacters); |
1424 | #else |
1425 | int shiftAmount = (m_charSize == Char8 ? 8 : 16) * numberCharacters; |
1426 | #endif |
1427 | |
1428 | UChar32 currentCharacter = nextTerm->patternCharacter; |
1429 | |
1430 | if ((currentCharacter > 0xff) && (m_charSize == Char8)) { |
1431 | // Have a 16 bit pattern character and an 8 bit string - short circuit |
1432 | op.m_jumps.append(jump()); |
1433 | return; |
1434 | } |
1435 | |
1436 | // For case-insesitive compares, non-ascii characters that have different |
1437 | // upper & lower case representations are converted to a character class. |
1438 | ASSERT(!m_pattern.ignoreCase() || isASCIIAlpha(currentCharacter) || isCanonicallyUnique(currentCharacter, m_canonicalMode)); |
1439 | |
1440 | allCharacters |= (static_cast<uint64_t>(currentCharacter) << shiftAmount); |
1441 | |
1442 | if ((m_pattern.ignoreCase()) && (isASCIIAlpha(currentCharacter))) |
1443 | ignoreCaseMask |= 32ULL << shiftAmount; |
1444 | } |
1445 | |
1446 | if (m_decodeSurrogatePairs) |
1447 | op.m_jumps.append(jumpIfNoAvailableInput()); |
1448 | |
1449 | if (m_charSize == Char8) { |
1450 | auto check1 = [&] (Checked<unsigned> offset, UChar32 characters) { |
1451 | op.m_jumps.append(jumpIfCharNotEquals(characters, offset, character)); |
1452 | }; |
1453 | |
1454 | auto check2 = [&] (Checked<unsigned> offset, uint16_t characters, uint16_t mask) { |
1455 | load16Unaligned(negativeOffsetIndexedAddress(offset, character), character); |
1456 | if (mask) |
1457 | or32(Imm32(mask), character); |
1458 | op.m_jumps.append(branch32(NotEqual, character, Imm32(characters | mask))); |
1459 | }; |
1460 | |
1461 | auto check4 = [&] (Checked<unsigned> offset, unsigned characters, unsigned mask) { |
1462 | if (mask) { |
1463 | load32WithUnalignedHalfWords(negativeOffsetIndexedAddress(offset, character), character); |
1464 | if (mask) |
1465 | or32(Imm32(mask), character); |
1466 | op.m_jumps.append(branch32(NotEqual, character, Imm32(characters | mask))); |
1467 | return; |
1468 | } |
1469 | op.m_jumps.append(branch32WithUnalignedHalfWords(NotEqual, negativeOffsetIndexedAddress(offset, character), TrustedImm32(characters))); |
1470 | }; |
1471 | |
1472 | #if CPU(X86_64) || CPU(ARM64) |
1473 | auto check8 = [&] (Checked<unsigned> offset, uint64_t characters, uint64_t mask) { |
1474 | load64(negativeOffsetIndexedAddress(offset, character), character); |
1475 | if (mask) |
1476 | or64(TrustedImm64(mask), character); |
1477 | op.m_jumps.append(branch64(NotEqual, character, TrustedImm64(characters | mask))); |
1478 | }; |
1479 | #endif |
1480 | |
1481 | switch (numberCharacters) { |
1482 | case 1: |
1483 | // Use 32bit width of allCharacters since Yarr counts surrogate pairs as one character with unicode flag. |
1484 | check1(m_checkedOffset - startTermPosition, allCharacters & 0xffffffff); |
1485 | return; |
1486 | case 2: { |
1487 | check2(m_checkedOffset - startTermPosition, allCharacters & 0xffff, ignoreCaseMask & 0xffff); |
1488 | return; |
1489 | } |
1490 | case 3: { |
1491 | check2(m_checkedOffset - startTermPosition, allCharacters & 0xffff, ignoreCaseMask & 0xffff); |
1492 | check1(m_checkedOffset - startTermPosition - 2, (allCharacters >> 16) & 0xff); |
1493 | return; |
1494 | } |
1495 | case 4: { |
1496 | check4(m_checkedOffset - startTermPosition, allCharacters & 0xffffffff, ignoreCaseMask & 0xffffffff); |
1497 | return; |
1498 | } |
1499 | #if CPU(X86_64) || CPU(ARM64) |
1500 | case 5: { |
1501 | check4(m_checkedOffset - startTermPosition, allCharacters & 0xffffffff, ignoreCaseMask & 0xffffffff); |
1502 | check1(m_checkedOffset - startTermPosition - 4, (allCharacters >> 32) & 0xff); |
1503 | return; |
1504 | } |
1505 | case 6: { |
1506 | check4(m_checkedOffset - startTermPosition, allCharacters & 0xffffffff, ignoreCaseMask & 0xffffffff); |
1507 | check2(m_checkedOffset - startTermPosition - 4, (allCharacters >> 32) & 0xffff, (ignoreCaseMask >> 32) & 0xffff); |
1508 | return; |
1509 | } |
1510 | case 7: { |
1511 | check4(m_checkedOffset - startTermPosition, allCharacters & 0xffffffff, ignoreCaseMask & 0xffffffff); |
1512 | check2(m_checkedOffset - startTermPosition - 4, (allCharacters >> 32) & 0xffff, (ignoreCaseMask >> 32) & 0xffff); |
1513 | check1(m_checkedOffset - startTermPosition - 6, (allCharacters >> 48) & 0xff); |
1514 | return; |
1515 | } |
1516 | case 8: { |
1517 | check8(m_checkedOffset - startTermPosition, allCharacters, ignoreCaseMask); |
1518 | return; |
1519 | } |
1520 | #endif |
1521 | } |
1522 | } else { |
1523 | auto check1 = [&] (Checked<unsigned> offset, UChar32 characters) { |
1524 | op.m_jumps.append(jumpIfCharNotEquals(characters, offset, character)); |
1525 | }; |
1526 | |
1527 | auto check2 = [&] (Checked<unsigned> offset, unsigned characters, unsigned mask) { |
1528 | if (mask) { |
1529 | load32WithUnalignedHalfWords(negativeOffsetIndexedAddress(offset, character), character); |
1530 | if (mask) |
1531 | or32(Imm32(mask), character); |
1532 | op.m_jumps.append(branch32(NotEqual, character, Imm32(characters | mask))); |
1533 | return; |
1534 | } |
1535 | op.m_jumps.append(branch32WithUnalignedHalfWords(NotEqual, negativeOffsetIndexedAddress(offset, character), TrustedImm32(characters))); |
1536 | }; |
1537 | |
1538 | #if CPU(X86_64) || CPU(ARM64) |
1539 | auto check4 = [&] (Checked<unsigned> offset, uint64_t characters, uint64_t mask) { |
1540 | load64(negativeOffsetIndexedAddress(offset, character), character); |
1541 | if (mask) |
1542 | or64(TrustedImm64(mask), character); |
1543 | op.m_jumps.append(branch64(NotEqual, character, TrustedImm64(characters | mask))); |
1544 | }; |
1545 | #endif |
1546 | |
1547 | switch (numberCharacters) { |
1548 | case 1: |
1549 | // Use 32bit width of allCharacters since Yarr counts surrogate pairs as one character with unicode flag. |
1550 | check1(m_checkedOffset - startTermPosition, allCharacters & 0xffffffff); |
1551 | return; |
1552 | case 2: |
1553 | check2(m_checkedOffset - startTermPosition, allCharacters & 0xffffffff, ignoreCaseMask & 0xffffffff); |
1554 | return; |
1555 | #if CPU(X86_64) || CPU(ARM64) |
1556 | case 3: |
1557 | check2(m_checkedOffset - startTermPosition, allCharacters & 0xffffffff, ignoreCaseMask & 0xffffffff); |
1558 | check1(m_checkedOffset - startTermPosition - 2, (allCharacters >> 32) & 0xffff); |
1559 | return; |
1560 | case 4: |
1561 | check4(m_checkedOffset - startTermPosition, allCharacters, ignoreCaseMask); |
1562 | return; |
1563 | #endif |
1564 | } |
1565 | } |
1566 | } |
1567 | void backtrackPatternCharacterOnce(size_t opIndex) |
1568 | { |
1569 | backtrackTermDefault(opIndex); |
1570 | } |
1571 | |
1572 | void generatePatternCharacterFixed(size_t opIndex) |
1573 | { |
1574 | YarrOp& op = m_ops[opIndex]; |
1575 | PatternTerm* term = op.m_term; |
1576 | UChar32 ch = term->patternCharacter; |
1577 | |
1578 | const RegisterID character = regT0; |
1579 | const RegisterID countRegister = regT1; |
1580 | |
1581 | if (m_decodeSurrogatePairs) |
1582 | op.m_jumps.append(jumpIfNoAvailableInput()); |
1583 | |
1584 | move(index, countRegister); |
1585 | Checked<unsigned> scaledMaxCount = term->quantityMaxCount; |
1586 | scaledMaxCount *= U_IS_BMP(ch) ? 1 : 2; |
1587 | sub32(Imm32(scaledMaxCount.unsafeGet()), countRegister); |
1588 | |
1589 | Label loop(this); |
1590 | readCharacter(m_checkedOffset - term->inputPosition - scaledMaxCount, character, countRegister); |
1591 | // For case-insesitive compares, non-ascii characters that have different |
1592 | // upper & lower case representations are converted to a character class. |
1593 | ASSERT(!m_pattern.ignoreCase() || isASCIIAlpha(ch) || isCanonicallyUnique(ch, m_canonicalMode)); |
1594 | if (m_pattern.ignoreCase() && isASCIIAlpha(ch)) { |
1595 | or32(TrustedImm32(0x20), character); |
1596 | ch |= 0x20; |
1597 | } |
1598 | |
1599 | op.m_jumps.append(branch32(NotEqual, character, Imm32(ch))); |
1600 | #ifdef JIT_UNICODE_EXPRESSIONS |
1601 | if (m_decodeSurrogatePairs && !U_IS_BMP(ch)) |
1602 | add32(TrustedImm32(2), countRegister); |
1603 | else |
1604 | #endif |
1605 | add32(TrustedImm32(1), countRegister); |
1606 | branch32(NotEqual, countRegister, index).linkTo(loop, this); |
1607 | } |
1608 | void backtrackPatternCharacterFixed(size_t opIndex) |
1609 | { |
1610 | backtrackTermDefault(opIndex); |
1611 | } |
1612 | |
1613 | void generatePatternCharacterGreedy(size_t opIndex) |
1614 | { |
1615 | YarrOp& op = m_ops[opIndex]; |
1616 | PatternTerm* term = op.m_term; |
1617 | UChar32 ch = term->patternCharacter; |
1618 | |
1619 | const RegisterID character = regT0; |
1620 | const RegisterID countRegister = regT1; |
1621 | |
1622 | move(TrustedImm32(0), countRegister); |
1623 | |
1624 | // Unless have a 16 bit pattern character and an 8 bit string - short circuit |
1625 | if (!((ch > 0xff) && (m_charSize == Char8))) { |
1626 | JumpList failures; |
1627 | Label loop(this); |
1628 | failures.append(atEndOfInput()); |
1629 | failures.append(jumpIfCharNotEquals(ch, m_checkedOffset - term->inputPosition, character)); |
1630 | |
1631 | add32(TrustedImm32(1), index); |
1632 | #ifdef JIT_UNICODE_EXPRESSIONS |
1633 | if (m_decodeSurrogatePairs && !U_IS_BMP(ch)) { |
1634 | Jump surrogatePairOk = notAtEndOfInput(); |
1635 | sub32(TrustedImm32(1), index); |
1636 | failures.append(jump()); |
1637 | surrogatePairOk.link(this); |
1638 | add32(TrustedImm32(1), index); |
1639 | } |
1640 | #endif |
1641 | add32(TrustedImm32(1), countRegister); |
1642 | |
1643 | if (term->quantityMaxCount == quantifyInfinite) |
1644 | jump(loop); |
1645 | else |
1646 | branch32(NotEqual, countRegister, Imm32(term->quantityMaxCount.unsafeGet())).linkTo(loop, this); |
1647 | |
1648 | failures.link(this); |
1649 | } |
1650 | op.m_reentry = label(); |
1651 | |
1652 | storeToFrame(countRegister, term->frameLocation + BackTrackInfoPatternCharacter::matchAmountIndex()); |
1653 | } |
1654 | void backtrackPatternCharacterGreedy(size_t opIndex) |
1655 | { |
1656 | YarrOp& op = m_ops[opIndex]; |
1657 | PatternTerm* term = op.m_term; |
1658 | |
1659 | const RegisterID countRegister = regT1; |
1660 | |
1661 | m_backtrackingState.link(this); |
1662 | |
1663 | loadFromFrame(term->frameLocation + BackTrackInfoPatternCharacter::matchAmountIndex(), countRegister); |
1664 | m_backtrackingState.append(branchTest32(Zero, countRegister)); |
1665 | sub32(TrustedImm32(1), countRegister); |
1666 | if (!m_decodeSurrogatePairs || U_IS_BMP(term->patternCharacter)) |
1667 | sub32(TrustedImm32(1), index); |
1668 | else |
1669 | sub32(TrustedImm32(2), index); |
1670 | jump(op.m_reentry); |
1671 | } |
1672 | |
1673 | void generatePatternCharacterNonGreedy(size_t opIndex) |
1674 | { |
1675 | YarrOp& op = m_ops[opIndex]; |
1676 | PatternTerm* term = op.m_term; |
1677 | |
1678 | const RegisterID countRegister = regT1; |
1679 | |
1680 | move(TrustedImm32(0), countRegister); |
1681 | op.m_reentry = label(); |
1682 | storeToFrame(countRegister, term->frameLocation + BackTrackInfoPatternCharacter::matchAmountIndex()); |
1683 | } |
1684 | void backtrackPatternCharacterNonGreedy(size_t opIndex) |
1685 | { |
1686 | YarrOp& op = m_ops[opIndex]; |
1687 | PatternTerm* term = op.m_term; |
1688 | UChar32 ch = term->patternCharacter; |
1689 | |
1690 | const RegisterID character = regT0; |
1691 | const RegisterID countRegister = regT1; |
1692 | |
1693 | m_backtrackingState.link(this); |
1694 | |
1695 | loadFromFrame(term->frameLocation + BackTrackInfoPatternCharacter::matchAmountIndex(), countRegister); |
1696 | |
1697 | // Unless have a 16 bit pattern character and an 8 bit string - short circuit |
1698 | if (!((ch > 0xff) && (m_charSize == Char8))) { |
1699 | JumpList nonGreedyFailures; |
1700 | nonGreedyFailures.append(atEndOfInput()); |
1701 | if (term->quantityMaxCount != quantifyInfinite) |
1702 | nonGreedyFailures.append(branch32(Equal, countRegister, Imm32(term->quantityMaxCount.unsafeGet()))); |
1703 | nonGreedyFailures.append(jumpIfCharNotEquals(ch, m_checkedOffset - term->inputPosition, character)); |
1704 | |
1705 | add32(TrustedImm32(1), index); |
1706 | #ifdef JIT_UNICODE_EXPRESSIONS |
1707 | if (m_decodeSurrogatePairs && !U_IS_BMP(ch)) { |
1708 | Jump surrogatePairOk = notAtEndOfInput(); |
1709 | sub32(TrustedImm32(1), index); |
1710 | nonGreedyFailures.append(jump()); |
1711 | surrogatePairOk.link(this); |
1712 | add32(TrustedImm32(1), index); |
1713 | } |
1714 | #endif |
1715 | add32(TrustedImm32(1), countRegister); |
1716 | |
1717 | jump(op.m_reentry); |
1718 | nonGreedyFailures.link(this); |
1719 | } |
1720 | |
1721 | if (m_decodeSurrogatePairs && !U_IS_BMP(ch)) { |
1722 | // subtract countRegister*2 for non-BMP characters |
1723 | lshift32(TrustedImm32(1), countRegister); |
1724 | } |
1725 | |
1726 | sub32(countRegister, index); |
1727 | m_backtrackingState.fallthrough(); |
1728 | } |
1729 | |
1730 | void generateCharacterClassOnce(size_t opIndex) |
1731 | { |
1732 | YarrOp& op = m_ops[opIndex]; |
1733 | PatternTerm* term = op.m_term; |
1734 | |
1735 | const RegisterID character = regT0; |
1736 | |
1737 | if (m_decodeSurrogatePairs) { |
1738 | op.m_jumps.append(jumpIfNoAvailableInput()); |
1739 | storeToFrame(index, term->frameLocation + BackTrackInfoCharacterClass::beginIndex()); |
1740 | } |
1741 | |
1742 | JumpList matchDest; |
1743 | readCharacter(m_checkedOffset - term->inputPosition, character); |
1744 | // If we are matching the "any character" builtin class we only need to read the |
1745 | // character and don't need to match as it will always succeed. |
1746 | if (term->invert() || !term->characterClass->m_anyCharacter) { |
1747 | matchCharacterClass(character, matchDest, term->characterClass); |
1748 | |
1749 | if (term->invert()) |
1750 | op.m_jumps.append(matchDest); |
1751 | else { |
1752 | op.m_jumps.append(jump()); |
1753 | matchDest.link(this); |
1754 | } |
1755 | } |
1756 | #ifdef JIT_UNICODE_EXPRESSIONS |
1757 | if (m_decodeSurrogatePairs && (!term->characterClass->hasOneCharacterSize() || term->invert())) { |
1758 | Jump isBMPChar = branch32(LessThan, character, supplementaryPlanesBase); |
1759 | add32(TrustedImm32(1), index); |
1760 | isBMPChar.link(this); |
1761 | } |
1762 | #endif |
1763 | } |
1764 | void backtrackCharacterClassOnce(size_t opIndex) |
1765 | { |
1766 | #ifdef JIT_UNICODE_EXPRESSIONS |
1767 | if (m_decodeSurrogatePairs) { |
1768 | YarrOp& op = m_ops[opIndex]; |
1769 | PatternTerm* term = op.m_term; |
1770 | |
1771 | m_backtrackingState.link(this); |
1772 | loadFromFrame(term->frameLocation + BackTrackInfoCharacterClass::beginIndex(), index); |
1773 | m_backtrackingState.fallthrough(); |
1774 | } |
1775 | #endif |
1776 | backtrackTermDefault(opIndex); |
1777 | } |
1778 | |
1779 | void generateCharacterClassFixed(size_t opIndex) |
1780 | { |
1781 | YarrOp& op = m_ops[opIndex]; |
1782 | PatternTerm* term = op.m_term; |
1783 | |
1784 | const RegisterID character = regT0; |
1785 | const RegisterID countRegister = regT1; |
1786 | |
1787 | if (m_decodeSurrogatePairs) |
1788 | op.m_jumps.append(jumpIfNoAvailableInput()); |
1789 | |
1790 | move(index, countRegister); |
1791 | |
1792 | Checked<unsigned> scaledMaxCount = term->quantityMaxCount; |
1793 | |
1794 | #ifdef JIT_UNICODE_EXPRESSIONS |
1795 | if (m_decodeSurrogatePairs && term->characterClass->hasOnlyNonBMPCharacters() && !term->invert()) |
1796 | scaledMaxCount *= 2; |
1797 | #endif |
1798 | sub32(Imm32(scaledMaxCount.unsafeGet()), countRegister); |
1799 | |
1800 | Label loop(this); |
1801 | JumpList matchDest; |
1802 | readCharacter(m_checkedOffset - term->inputPosition - scaledMaxCount, character, countRegister); |
1803 | // If we are matching the "any character" builtin class we only need to read the |
1804 | // character and don't need to match as it will always succeed. |
1805 | if (term->invert() || !term->characterClass->m_anyCharacter) { |
1806 | matchCharacterClass(character, matchDest, term->characterClass); |
1807 | |
1808 | if (term->invert()) |
1809 | op.m_jumps.append(matchDest); |
1810 | else { |
1811 | op.m_jumps.append(jump()); |
1812 | matchDest.link(this); |
1813 | } |
1814 | } |
1815 | |
1816 | #ifdef JIT_UNICODE_EXPRESSIONS |
1817 | if (m_decodeSurrogatePairs) { |
1818 | if (term->characterClass->hasOneCharacterSize() && !term->invert()) |
1819 | add32(TrustedImm32(term->characterClass->hasNonBMPCharacters() ? 2 : 1), countRegister); |
1820 | else { |
1821 | add32(TrustedImm32(1), countRegister); |
1822 | Jump isBMPChar = branch32(LessThan, character, supplementaryPlanesBase); |
1823 | op.m_jumps.append(atEndOfInput()); |
1824 | add32(TrustedImm32(1), countRegister); |
1825 | add32(TrustedImm32(1), index); |
1826 | isBMPChar.link(this); |
1827 | } |
1828 | } else |
1829 | #endif |
1830 | add32(TrustedImm32(1), countRegister); |
1831 | branch32(NotEqual, countRegister, index).linkTo(loop, this); |
1832 | } |
1833 | void backtrackCharacterClassFixed(size_t opIndex) |
1834 | { |
1835 | backtrackTermDefault(opIndex); |
1836 | } |
1837 | |
1838 | void generateCharacterClassGreedy(size_t opIndex) |
1839 | { |
1840 | YarrOp& op = m_ops[opIndex]; |
1841 | PatternTerm* term = op.m_term; |
1842 | |
1843 | const RegisterID character = regT0; |
1844 | const RegisterID countRegister = regT1; |
1845 | |
1846 | if (m_decodeSurrogatePairs && (!term->characterClass->hasOneCharacterSize() || term->invert())) |
1847 | storeToFrame(index, term->frameLocation + BackTrackInfoCharacterClass::beginIndex()); |
1848 | move(TrustedImm32(0), countRegister); |
1849 | |
1850 | JumpList failures; |
1851 | Label loop(this); |
1852 | #ifdef JIT_UNICODE_EXPRESSIONS |
1853 | if (term->characterClass->hasOneCharacterSize() && !term->invert() && term->characterClass->hasNonBMPCharacters()) { |
1854 | move(TrustedImm32(1), character); |
1855 | failures.append(checkNotEnoughInput(character)); |
1856 | } else |
1857 | #endif |
1858 | failures.append(atEndOfInput()); |
1859 | |
1860 | if (term->invert()) { |
1861 | readCharacter(m_checkedOffset - term->inputPosition, character); |
1862 | matchCharacterClass(character, failures, term->characterClass); |
1863 | } else { |
1864 | JumpList matchDest; |
1865 | readCharacter(m_checkedOffset - term->inputPosition, character); |
1866 | // If we are matching the "any character" builtin class for non-unicode patterns, |
1867 | // we only need to read the character and don't need to match as it will always succeed. |
1868 | if (!term->characterClass->m_anyCharacter) { |
1869 | matchCharacterClass(character, matchDest, term->characterClass); |
1870 | failures.append(jump()); |
1871 | } |
1872 | matchDest.link(this); |
1873 | } |
1874 | |
1875 | #ifdef JIT_UNICODE_EXPRESSIONS |
1876 | if (m_decodeSurrogatePairs) |
1877 | advanceIndexAfterCharacterClassTermMatch(term, failures, character); |
1878 | else |
1879 | #endif |
1880 | add32(TrustedImm32(1), index); |
1881 | add32(TrustedImm32(1), countRegister); |
1882 | |
1883 | if (term->quantityMaxCount != quantifyInfinite) { |
1884 | branch32(NotEqual, countRegister, Imm32(term->quantityMaxCount.unsafeGet())).linkTo(loop, this); |
1885 | failures.append(jump()); |
1886 | } else |
1887 | jump(loop); |
1888 | |
1889 | failures.link(this); |
1890 | op.m_reentry = label(); |
1891 | |
1892 | storeToFrame(countRegister, term->frameLocation + BackTrackInfoCharacterClass::matchAmountIndex()); |
1893 | } |
1894 | void backtrackCharacterClassGreedy(size_t opIndex) |
1895 | { |
1896 | YarrOp& op = m_ops[opIndex]; |
1897 | PatternTerm* term = op.m_term; |
1898 | |
1899 | const RegisterID countRegister = regT1; |
1900 | |
1901 | m_backtrackingState.link(this); |
1902 | |
1903 | loadFromFrame(term->frameLocation + BackTrackInfoCharacterClass::matchAmountIndex(), countRegister); |
1904 | m_backtrackingState.append(branchTest32(Zero, countRegister)); |
1905 | sub32(TrustedImm32(1), countRegister); |
1906 | storeToFrame(countRegister, term->frameLocation + BackTrackInfoCharacterClass::matchAmountIndex()); |
1907 | |
1908 | if (!m_decodeSurrogatePairs) |
1909 | sub32(TrustedImm32(1), index); |
1910 | else if (term->characterClass->hasOneCharacterSize() && !term->invert()) |
1911 | sub32(TrustedImm32(term->characterClass->hasNonBMPCharacters() ? 2 : 1), index); |
1912 | else { |
1913 | // Rematch one less |
1914 | const RegisterID character = regT0; |
1915 | |
1916 | loadFromFrame(term->frameLocation + BackTrackInfoCharacterClass::beginIndex(), index); |
1917 | |
1918 | Label rematchLoop(this); |
1919 | readCharacter(m_checkedOffset - term->inputPosition, character); |
1920 | |
1921 | sub32(TrustedImm32(1), countRegister); |
1922 | add32(TrustedImm32(1), index); |
1923 | |
1924 | #ifdef JIT_UNICODE_EXPRESSIONS |
1925 | Jump isBMPChar = branch32(LessThan, character, supplementaryPlanesBase); |
1926 | add32(TrustedImm32(1), index); |
1927 | isBMPChar.link(this); |
1928 | #endif |
1929 | |
1930 | branchTest32(Zero, countRegister).linkTo(rematchLoop, this); |
1931 | |
1932 | loadFromFrame(term->frameLocation + BackTrackInfoCharacterClass::matchAmountIndex(), countRegister); |
1933 | } |
1934 | jump(op.m_reentry); |
1935 | } |
1936 | |
1937 | void generateCharacterClassNonGreedy(size_t opIndex) |
1938 | { |
1939 | YarrOp& op = m_ops[opIndex]; |
1940 | PatternTerm* term = op.m_term; |
1941 | |
1942 | const RegisterID countRegister = regT1; |
1943 | |
1944 | move(TrustedImm32(0), countRegister); |
1945 | op.m_reentry = label(); |
1946 | |
1947 | #ifdef JIT_UNICODE_EXPRESSIONS |
1948 | if (m_decodeSurrogatePairs) { |
1949 | if (!term->characterClass->hasOneCharacterSize() || term->invert()) |
1950 | storeToFrame(index, term->frameLocation + BackTrackInfoCharacterClass::beginIndex()); |
1951 | } |
1952 | #endif |
1953 | |
1954 | storeToFrame(countRegister, term->frameLocation + BackTrackInfoCharacterClass::matchAmountIndex()); |
1955 | } |
1956 | |
1957 | void backtrackCharacterClassNonGreedy(size_t opIndex) |
1958 | { |
1959 | YarrOp& op = m_ops[opIndex]; |
1960 | PatternTerm* term = op.m_term; |
1961 | |
1962 | const RegisterID character = regT0; |
1963 | const RegisterID countRegister = regT1; |
1964 | |
1965 | JumpList nonGreedyFailures; |
1966 | |
1967 | m_backtrackingState.link(this); |
1968 | |
1969 | #ifdef JIT_UNICODE_EXPRESSIONS |
1970 | if (m_decodeSurrogatePairs) { |
1971 | if (!term->characterClass->hasOneCharacterSize() || term->invert()) |
1972 | loadFromFrame(term->frameLocation + BackTrackInfoCharacterClass::beginIndex(), index); |
1973 | } |
1974 | #endif |
1975 | |
1976 | loadFromFrame(term->frameLocation + BackTrackInfoCharacterClass::matchAmountIndex(), countRegister); |
1977 | |
1978 | nonGreedyFailures.append(atEndOfInput()); |
1979 | nonGreedyFailures.append(branch32(Equal, countRegister, Imm32(term->quantityMaxCount.unsafeGet()))); |
1980 | |
1981 | JumpList matchDest; |
1982 | readCharacter(m_checkedOffset - term->inputPosition, character); |
1983 | // If we are matching the "any character" builtin class for non-unicode patterns, |
1984 | // we only need to read the character and don't need to match as it will always succeed. |
1985 | if (term->invert() || !term->characterClass->m_anyCharacter) { |
1986 | matchCharacterClass(character, matchDest, term->characterClass); |
1987 | |
1988 | if (term->invert()) |
1989 | nonGreedyFailures.append(matchDest); |
1990 | else { |
1991 | nonGreedyFailures.append(jump()); |
1992 | matchDest.link(this); |
1993 | } |
1994 | } |
1995 | |
1996 | #ifdef JIT_UNICODE_EXPRESSIONS |
1997 | if (m_decodeSurrogatePairs) |
1998 | advanceIndexAfterCharacterClassTermMatch(term, nonGreedyFailures, character); |
1999 | else |
2000 | #endif |
2001 | add32(TrustedImm32(1), index); |
2002 | add32(TrustedImm32(1), countRegister); |
2003 | |
2004 | jump(op.m_reentry); |
2005 | |
2006 | nonGreedyFailures.link(this); |
2007 | sub32(countRegister, index); |
2008 | m_backtrackingState.fallthrough(); |
2009 | } |
2010 | |
2011 | void generateDotStarEnclosure(size_t opIndex) |
2012 | { |
2013 | YarrOp& op = m_ops[opIndex]; |
2014 | PatternTerm* term = op.m_term; |
2015 | |
2016 | const RegisterID character = regT0; |
2017 | const RegisterID matchPos = regT1; |
2018 | #ifndef HAVE_INITIAL_START_REG |
2019 | const RegisterID initialStart = character; |
2020 | #endif |
2021 | |
2022 | JumpList foundBeginningNewLine; |
2023 | JumpList saveStartIndex; |
2024 | JumpList foundEndingNewLine; |
2025 | |
2026 | if (m_pattern.dotAll()) { |
2027 | move(TrustedImm32(0), matchPos); |
2028 | setMatchStart(matchPos); |
2029 | move(length, index); |
2030 | return; |
2031 | } |
2032 | |
2033 | ASSERT(!m_pattern.m_body->m_hasFixedSize); |
2034 | getMatchStart(matchPos); |
2035 | |
2036 | #ifndef HAVE_INITIAL_START_REG |
2037 | loadFromFrame(m_pattern.m_initialStartValueFrameLocation, initialStart); |
2038 | #endif |
2039 | saveStartIndex.append(branch32(BelowOrEqual, matchPos, initialStart)); |
2040 | Label findBOLLoop(this); |
2041 | sub32(TrustedImm32(1), matchPos); |
2042 | if (m_charSize == Char8) |
2043 | load8(BaseIndex(input, matchPos, TimesOne, 0), character); |
2044 | else |
2045 | load16(BaseIndex(input, matchPos, TimesTwo, 0), character); |
2046 | matchCharacterClass(character, foundBeginningNewLine, m_pattern.newlineCharacterClass()); |
2047 | |
2048 | #ifndef HAVE_INITIAL_START_REG |
2049 | loadFromFrame(m_pattern.m_initialStartValueFrameLocation, initialStart); |
2050 | #endif |
2051 | branch32(Above, matchPos, initialStart).linkTo(findBOLLoop, this); |
2052 | saveStartIndex.append(jump()); |
2053 | |
2054 | foundBeginningNewLine.link(this); |
2055 | add32(TrustedImm32(1), matchPos); // Advance past newline |
2056 | saveStartIndex.link(this); |
2057 | |
2058 | if (!m_pattern.multiline() && term->anchors.bolAnchor) |
2059 | op.m_jumps.append(branchTest32(NonZero, matchPos)); |
2060 | |
2061 | ASSERT(!m_pattern.m_body->m_hasFixedSize); |
2062 | setMatchStart(matchPos); |
2063 | |
2064 | move(index, matchPos); |
2065 | |
2066 | Label findEOLLoop(this); |
2067 | foundEndingNewLine.append(branch32(Equal, matchPos, length)); |
2068 | if (m_charSize == Char8) |
2069 | load8(BaseIndex(input, matchPos, TimesOne, 0), character); |
2070 | else |
2071 | load16(BaseIndex(input, matchPos, TimesTwo, 0), character); |
2072 | matchCharacterClass(character, foundEndingNewLine, m_pattern.newlineCharacterClass()); |
2073 | add32(TrustedImm32(1), matchPos); |
2074 | jump(findEOLLoop); |
2075 | |
2076 | foundEndingNewLine.link(this); |
2077 | |
2078 | if (!m_pattern.multiline() && term->anchors.eolAnchor) |
2079 | op.m_jumps.append(branch32(NotEqual, matchPos, length)); |
2080 | |
2081 | move(matchPos, index); |
2082 | } |
2083 | |
2084 | void backtrackDotStarEnclosure(size_t opIndex) |
2085 | { |
2086 | backtrackTermDefault(opIndex); |
2087 | } |
2088 | |
2089 | // Code generation/backtracking for simple terms |
2090 | // (pattern characters, character classes, and assertions). |
2091 | // These methods farm out work to the set of functions above. |
2092 | void generateTerm(size_t opIndex) |
2093 | { |
2094 | YarrOp& op = m_ops[opIndex]; |
2095 | PatternTerm* term = op.m_term; |
2096 | |
2097 | switch (term->type) { |
2098 | case PatternTerm::TypePatternCharacter: |
2099 | switch (term->quantityType) { |
2100 | case QuantifierFixedCount: |
2101 | if (term->quantityMaxCount == 1) |
2102 | generatePatternCharacterOnce(opIndex); |
2103 | else |
2104 | generatePatternCharacterFixed(opIndex); |
2105 | break; |
2106 | case QuantifierGreedy: |
2107 | generatePatternCharacterGreedy(opIndex); |
2108 | break; |
2109 | case QuantifierNonGreedy: |
2110 | generatePatternCharacterNonGreedy(opIndex); |
2111 | break; |
2112 | } |
2113 | break; |
2114 | |
2115 | case PatternTerm::TypeCharacterClass: |
2116 | switch (term->quantityType) { |
2117 | case QuantifierFixedCount: |
2118 | if (term->quantityMaxCount == 1) |
2119 | generateCharacterClassOnce(opIndex); |
2120 | else |
2121 | generateCharacterClassFixed(opIndex); |
2122 | break; |
2123 | case QuantifierGreedy: |
2124 | generateCharacterClassGreedy(opIndex); |
2125 | break; |
2126 | case QuantifierNonGreedy: |
2127 | generateCharacterClassNonGreedy(opIndex); |
2128 | break; |
2129 | } |
2130 | break; |
2131 | |
2132 | case PatternTerm::TypeAssertionBOL: |
2133 | generateAssertionBOL(opIndex); |
2134 | break; |
2135 | |
2136 | case PatternTerm::TypeAssertionEOL: |
2137 | generateAssertionEOL(opIndex); |
2138 | break; |
2139 | |
2140 | case PatternTerm::TypeAssertionWordBoundary: |
2141 | generateAssertionWordBoundary(opIndex); |
2142 | break; |
2143 | |
2144 | case PatternTerm::TypeForwardReference: |
2145 | m_failureReason = JITFailureReason::ForwardReference; |
2146 | break; |
2147 | |
2148 | case PatternTerm::TypeParenthesesSubpattern: |
2149 | case PatternTerm::TypeParentheticalAssertion: |
2150 | RELEASE_ASSERT_NOT_REACHED(); |
2151 | |
2152 | case PatternTerm::TypeBackReference: |
2153 | #if ENABLE(YARR_JIT_BACKREFERENCES) |
2154 | generateBackReference(opIndex); |
2155 | #else |
2156 | m_failureReason = JITFailureReason::BackReference; |
2157 | #endif |
2158 | break; |
2159 | case PatternTerm::TypeDotStarEnclosure: |
2160 | generateDotStarEnclosure(opIndex); |
2161 | break; |
2162 | } |
2163 | } |
2164 | void backtrackTerm(size_t opIndex) |
2165 | { |
2166 | YarrOp& op = m_ops[opIndex]; |
2167 | PatternTerm* term = op.m_term; |
2168 | |
2169 | switch (term->type) { |
2170 | case PatternTerm::TypePatternCharacter: |
2171 | switch (term->quantityType) { |
2172 | case QuantifierFixedCount: |
2173 | if (term->quantityMaxCount == 1) |
2174 | backtrackPatternCharacterOnce(opIndex); |
2175 | else |
2176 | backtrackPatternCharacterFixed(opIndex); |
2177 | break; |
2178 | case QuantifierGreedy: |
2179 | backtrackPatternCharacterGreedy(opIndex); |
2180 | break; |
2181 | case QuantifierNonGreedy: |
2182 | backtrackPatternCharacterNonGreedy(opIndex); |
2183 | break; |
2184 | } |
2185 | break; |
2186 | |
2187 | case PatternTerm::TypeCharacterClass: |
2188 | switch (term->quantityType) { |
2189 | case QuantifierFixedCount: |
2190 | if (term->quantityMaxCount == 1) |
2191 | backtrackCharacterClassOnce(opIndex); |
2192 | else |
2193 | backtrackCharacterClassFixed(opIndex); |
2194 | break; |
2195 | case QuantifierGreedy: |
2196 | backtrackCharacterClassGreedy(opIndex); |
2197 | break; |
2198 | case QuantifierNonGreedy: |
2199 | backtrackCharacterClassNonGreedy(opIndex); |
2200 | break; |
2201 | } |
2202 | break; |
2203 | |
2204 | case PatternTerm::TypeAssertionBOL: |
2205 | backtrackAssertionBOL(opIndex); |
2206 | break; |
2207 | |
2208 | case PatternTerm::TypeAssertionEOL: |
2209 | backtrackAssertionEOL(opIndex); |
2210 | break; |
2211 | |
2212 | case PatternTerm::TypeAssertionWordBoundary: |
2213 | backtrackAssertionWordBoundary(opIndex); |
2214 | break; |
2215 | |
2216 | case PatternTerm::TypeForwardReference: |
2217 | m_failureReason = JITFailureReason::ForwardReference; |
2218 | break; |
2219 | |
2220 | case PatternTerm::TypeParenthesesSubpattern: |
2221 | case PatternTerm::TypeParentheticalAssertion: |
2222 | RELEASE_ASSERT_NOT_REACHED(); |
2223 | |
2224 | case PatternTerm::TypeBackReference: |
2225 | #if ENABLE(YARR_JIT_BACKREFERENCES) |
2226 | backtrackBackReference(opIndex); |
2227 | #else |
2228 | m_failureReason = JITFailureReason::BackReference; |
2229 | #endif |
2230 | break; |
2231 | |
2232 | case PatternTerm::TypeDotStarEnclosure: |
2233 | backtrackDotStarEnclosure(opIndex); |
2234 | break; |
2235 | } |
2236 | } |
2237 | |
2238 | void generate() |
2239 | { |
2240 | // Forwards generate the matching code. |
2241 | ASSERT(m_ops.size()); |
2242 | size_t opIndex = 0; |
2243 | |
2244 | do { |
2245 | if (m_disassembler) |
2246 | m_disassembler->setForGenerate(opIndex, label()); |
2247 | |
2248 | YarrOp& op = m_ops[opIndex]; |
2249 | switch (op.m_op) { |
2250 | |
2251 | case OpTerm: |
2252 | generateTerm(opIndex); |
2253 | break; |
2254 | |
2255 | // OpBodyAlternativeBegin/Next/End |
2256 | // |
2257 | // These nodes wrap the set of alternatives in the body of the regular expression. |
2258 | // There may be either one or two chains of OpBodyAlternative nodes, one representing |
2259 | // the 'once through' sequence of alternatives (if any exist), and one representing |
2260 | // the repeating alternatives (again, if any exist). |
2261 | // |
2262 | // Upon normal entry to the Begin alternative, we will check that input is available. |
2263 | // Reentry to the Begin alternative will take place after the check has taken place, |
2264 | // and will assume that the input position has already been progressed as appropriate. |
2265 | // |
2266 | // Entry to subsequent Next/End alternatives occurs when the prior alternative has |
2267 | // successfully completed a match - return a success state from JIT code. |
2268 | // |
2269 | // Next alternatives allow for reentry optimized to suit backtracking from its |
2270 | // preceding alternative. It expects the input position to still be set to a position |
2271 | // appropriate to its predecessor, and it will only perform an input check if the |
2272 | // predecessor had a minimum size less than its own. |
2273 | // |
2274 | // In the case 'once through' expressions, the End node will also have a reentry |
2275 | // point to jump to when the last alternative fails. Again, this expects the input |
2276 | // position to still reflect that expected by the prior alternative. |
2277 | case OpBodyAlternativeBegin: { |
2278 | PatternAlternative* alternative = op.m_alternative; |
2279 | |
2280 | // Upon entry at the head of the set of alternatives, check if input is available |
2281 | // to run the first alternative. (This progresses the input position). |
2282 | op.m_jumps.append(jumpIfNoAvailableInput(alternative->m_minimumSize)); |
2283 | // We will reenter after the check, and assume the input position to have been |
2284 | // set as appropriate to this alternative. |
2285 | op.m_reentry = label(); |
2286 | |
2287 | m_checkedOffset += alternative->m_minimumSize; |
2288 | break; |
2289 | } |
2290 | case OpBodyAlternativeNext: |
2291 | case OpBodyAlternativeEnd: { |
2292 | PatternAlternative* priorAlternative = m_ops[op.m_previousOp].m_alternative; |
2293 | PatternAlternative* alternative = op.m_alternative; |
2294 | |
2295 | // If we get here, the prior alternative matched - return success. |
2296 | |
2297 | // Adjust the stack pointer to remove the pattern's frame. |
2298 | removeCallFrame(); |
2299 | |
2300 | // Load appropriate values into the return register and the first output |
2301 | // slot, and return. In the case of pattern with a fixed size, we will |
2302 | // not have yet set the value in the first |
2303 | ASSERT(index != returnRegister); |
2304 | if (m_pattern.m_body->m_hasFixedSize) { |
2305 | move(index, returnRegister); |
2306 | if (priorAlternative->m_minimumSize) |
2307 | sub32(Imm32(priorAlternative->m_minimumSize), returnRegister); |
2308 | if (compileMode == IncludeSubpatterns) |
2309 | store32(returnRegister, output); |
2310 | } else |
2311 | getMatchStart(returnRegister); |
2312 | if (compileMode == IncludeSubpatterns) |
2313 | store32(index, Address(output, 4)); |
2314 | move(index, returnRegister2); |
2315 | |
2316 | generateReturn(); |
2317 | |
2318 | // This is the divide between the tail of the prior alternative, above, and |
2319 | // the head of the subsequent alternative, below. |
2320 | |
2321 | if (op.m_op == OpBodyAlternativeNext) { |
2322 | // This is the reentry point for the Next alternative. We expect any code |
2323 | // that jumps here to do so with the input position matching that of the |
2324 | // PRIOR alteranative, and we will only check input availability if we |
2325 | // need to progress it forwards. |
2326 | op.m_reentry = label(); |
2327 | if (alternative->m_minimumSize > priorAlternative->m_minimumSize) { |
2328 | add32(Imm32(alternative->m_minimumSize - priorAlternative->m_minimumSize), index); |
2329 | op.m_jumps.append(jumpIfNoAvailableInput()); |
2330 | } else if (priorAlternative->m_minimumSize > alternative->m_minimumSize) |
2331 | sub32(Imm32(priorAlternative->m_minimumSize - alternative->m_minimumSize), index); |
2332 | } else if (op.m_nextOp == notFound) { |
2333 | // This is the reentry point for the End of 'once through' alternatives, |
2334 | // jumped to when the last alternative fails to match. |
2335 | op.m_reentry = label(); |
2336 | sub32(Imm32(priorAlternative->m_minimumSize), index); |
2337 | } |
2338 | |
2339 | if (op.m_op == OpBodyAlternativeNext) |
2340 | m_checkedOffset += alternative->m_minimumSize; |
2341 | m_checkedOffset -= priorAlternative->m_minimumSize; |
2342 | break; |
2343 | } |
2344 | |
2345 | // OpSimpleNestedAlternativeBegin/Next/End |
2346 | // OpNestedAlternativeBegin/Next/End |
2347 | // |
2348 | // These nodes are used to handle sets of alternatives that are nested within |
2349 | // subpatterns and parenthetical assertions. The 'simple' forms are used where |
2350 | // we do not need to be able to backtrack back into any alternative other than |
2351 | // the last, the normal forms allow backtracking into any alternative. |
2352 | // |
2353 | // Each Begin/Next node is responsible for planting an input check to ensure |
2354 | // sufficient input is available on entry. Next nodes additionally need to |
2355 | // jump to the end - Next nodes use the End node's m_jumps list to hold this |
2356 | // set of jumps. |
2357 | // |
2358 | // In the non-simple forms, successful alternative matches must store a |
2359 | // 'return address' using a DataLabelPtr, used to store the address to jump |
2360 | // to when backtracking, to get to the code for the appropriate alternative. |
2361 | case OpSimpleNestedAlternativeBegin: |
2362 | case OpNestedAlternativeBegin: { |
2363 | PatternTerm* term = op.m_term; |
2364 | PatternAlternative* alternative = op.m_alternative; |
2365 | PatternDisjunction* disjunction = term->parentheses.disjunction; |
2366 | |
2367 | // Calculate how much input we need to check for, and if non-zero check. |
2368 | op.m_checkAdjust = Checked<unsigned>(alternative->m_minimumSize); |
2369 | if ((term->quantityType == QuantifierFixedCount) && (term->type != PatternTerm::TypeParentheticalAssertion)) |
2370 | op.m_checkAdjust -= disjunction->m_minimumSize; |
2371 | if (op.m_checkAdjust) |
2372 | op.m_jumps.append(jumpIfNoAvailableInput(op.m_checkAdjust.unsafeGet())); |
2373 | |
2374 | m_checkedOffset += op.m_checkAdjust; |
2375 | break; |
2376 | } |
2377 | case OpSimpleNestedAlternativeNext: |
2378 | case OpNestedAlternativeNext: { |
2379 | PatternTerm* term = op.m_term; |
2380 | PatternAlternative* alternative = op.m_alternative; |
2381 | PatternDisjunction* disjunction = term->parentheses.disjunction; |
2382 | |
2383 | // In the non-simple case, store a 'return address' so we can backtrack correctly. |
2384 | if (op.m_op == OpNestedAlternativeNext) { |
2385 | unsigned parenthesesFrameLocation = term->frameLocation; |
2386 | op.m_returnAddress = storeToFrameWithPatch(parenthesesFrameLocation + BackTrackInfoParentheses::returnAddressIndex()); |
2387 | } |
2388 | |
2389 | if (term->quantityType != QuantifierFixedCount && !m_ops[op.m_previousOp].m_alternative->m_minimumSize) { |
2390 | // If the previous alternative matched without consuming characters then |
2391 | // backtrack to try to match while consumming some input. |
2392 | op.m_zeroLengthMatch = branch32(Equal, index, Address(stackPointerRegister, term->frameLocation * sizeof(void*))); |
2393 | } |
2394 | |
2395 | // If we reach here then the last alternative has matched - jump to the |
2396 | // End node, to skip over any further alternatives. |
2397 | // |
2398 | // FIXME: this is logically O(N^2) (though N can be expected to be very |
2399 | // small). We could avoid this either by adding an extra jump to the JIT |
2400 | // data structures, or by making backtracking code that jumps to Next |
2401 | // alternatives are responsible for checking that input is available (if |
2402 | // we didn't need to plant the input checks, then m_jumps would be free). |
2403 | YarrOp* endOp = &m_ops[op.m_nextOp]; |
2404 | while (endOp->m_nextOp != notFound) { |
2405 | ASSERT(endOp->m_op == OpSimpleNestedAlternativeNext || endOp->m_op == OpNestedAlternativeNext); |
2406 | endOp = &m_ops[endOp->m_nextOp]; |
2407 | } |
2408 | ASSERT(endOp->m_op == OpSimpleNestedAlternativeEnd || endOp->m_op == OpNestedAlternativeEnd); |
2409 | endOp->m_jumps.append(jump()); |
2410 | |
2411 | // This is the entry point for the next alternative. |
2412 | op.m_reentry = label(); |
2413 | |
2414 | // Calculate how much input we need to check for, and if non-zero check. |
2415 | op.m_checkAdjust = alternative->m_minimumSize; |
2416 | if ((term->quantityType == QuantifierFixedCount) && (term->type != PatternTerm::TypeParentheticalAssertion)) |
2417 | op.m_checkAdjust -= disjunction->m_minimumSize; |
2418 | if (op.m_checkAdjust) |
2419 | op.m_jumps.append(jumpIfNoAvailableInput(op.m_checkAdjust.unsafeGet())); |
2420 | |
2421 | YarrOp& lastOp = m_ops[op.m_previousOp]; |
2422 | m_checkedOffset -= lastOp.m_checkAdjust; |
2423 | m_checkedOffset += op.m_checkAdjust; |
2424 | break; |
2425 | } |
2426 | case OpSimpleNestedAlternativeEnd: |
2427 | case OpNestedAlternativeEnd: { |
2428 | PatternTerm* term = op.m_term; |
2429 | |
2430 | // In the non-simple case, store a 'return address' so we can backtrack correctly. |
2431 | if (op.m_op == OpNestedAlternativeEnd) { |
2432 | unsigned parenthesesFrameLocation = term->frameLocation; |
2433 | op.m_returnAddress = storeToFrameWithPatch(parenthesesFrameLocation + BackTrackInfoParentheses::returnAddressIndex()); |
2434 | } |
2435 | |
2436 | if (term->quantityType != QuantifierFixedCount && !m_ops[op.m_previousOp].m_alternative->m_minimumSize) { |
2437 | // If the previous alternative matched without consuming characters then |
2438 | // backtrack to try to match while consumming some input. |
2439 | op.m_zeroLengthMatch = branch32(Equal, index, Address(stackPointerRegister, term->frameLocation * sizeof(void*))); |
2440 | } |
2441 | |
2442 | // If this set of alternatives contains more than one alternative, |
2443 | // then the Next nodes will have planted jumps to the End, and added |
2444 | // them to this node's m_jumps list. |
2445 | op.m_jumps.link(this); |
2446 | op.m_jumps.clear(); |
2447 | |
2448 | YarrOp& lastOp = m_ops[op.m_previousOp]; |
2449 | m_checkedOffset -= lastOp.m_checkAdjust; |
2450 | break; |
2451 | } |
2452 | |
2453 | // OpParenthesesSubpatternOnceBegin/End |
2454 | // |
2455 | // These nodes support (optionally) capturing subpatterns, that have a |
2456 | // quantity count of 1 (this covers fixed once, and ?/?? quantifiers). |
2457 | case OpParenthesesSubpatternOnceBegin: { |
2458 | PatternTerm* term = op.m_term; |
2459 | unsigned parenthesesFrameLocation = term->frameLocation; |
2460 | const RegisterID indexTemporary = regT0; |
2461 | ASSERT(term->quantityMaxCount == 1); |
2462 | |
2463 | // Upon entry to a Greedy quantified set of parenthese store the index. |
2464 | // We'll use this for two purposes: |
2465 | // - To indicate which iteration we are on of mathing the remainder of |
2466 | // the expression after the parentheses - the first, including the |
2467 | // match within the parentheses, or the second having skipped over them. |
2468 | // - To check for empty matches, which must be rejected. |
2469 | // |
2470 | // At the head of a NonGreedy set of parentheses we'll immediately set the |
2471 | // value on the stack to -1 (indicating a match skipping the subpattern), |
2472 | // and plant a jump to the end. We'll also plant a label to backtrack to |
2473 | // to reenter the subpattern later, with a store to set up index on the |
2474 | // second iteration. |
2475 | // |
2476 | // FIXME: for capturing parens, could use the index in the capture array? |
2477 | if (term->quantityType == QuantifierGreedy) |
2478 | storeToFrame(index, parenthesesFrameLocation + BackTrackInfoParenthesesOnce::beginIndex()); |
2479 | else if (term->quantityType == QuantifierNonGreedy) { |
2480 | storeToFrame(TrustedImm32(-1), parenthesesFrameLocation + BackTrackInfoParenthesesOnce::beginIndex()); |
2481 | op.m_jumps.append(jump()); |
2482 | op.m_reentry = label(); |
2483 | storeToFrame(index, parenthesesFrameLocation + BackTrackInfoParenthesesOnce::beginIndex()); |
2484 | } |
2485 | |
2486 | // If the parenthese are capturing, store the starting index value to the |
2487 | // captures array, offsetting as necessary. |
2488 | // |
2489 | // FIXME: could avoid offsetting this value in JIT code, apply |
2490 | // offsets only afterwards, at the point the results array is |
2491 | // being accessed. |
2492 | if (term->capture() && compileMode == IncludeSubpatterns) { |
2493 | unsigned inputOffset = (m_checkedOffset - term->inputPosition).unsafeGet(); |
2494 | if (term->quantityType == QuantifierFixedCount) |
2495 | inputOffset += term->parentheses.disjunction->m_minimumSize; |
2496 | if (inputOffset) { |
2497 | move(index, indexTemporary); |
2498 | sub32(Imm32(inputOffset), indexTemporary); |
2499 | setSubpatternStart(indexTemporary, term->parentheses.subpatternId); |
2500 | } else |
2501 | setSubpatternStart(index, term->parentheses.subpatternId); |
2502 | } |
2503 | break; |
2504 | } |
2505 | case OpParenthesesSubpatternOnceEnd: { |
2506 | PatternTerm* term = op.m_term; |
2507 | const RegisterID indexTemporary = regT0; |
2508 | ASSERT(term->quantityMaxCount == 1); |
2509 | |
2510 | // Runtime ASSERT to make sure that the nested alternative handled the |
2511 | // "no input consumed" check. |
2512 | if (!ASSERT_DISABLED && term->quantityType != QuantifierFixedCount && !term->parentheses.disjunction->m_minimumSize) { |
2513 | Jump pastBreakpoint; |
2514 | pastBreakpoint = branch32(NotEqual, index, Address(stackPointerRegister, term->frameLocation * sizeof(void*))); |
2515 | abortWithReason(YARRNoInputConsumed); |
2516 | pastBreakpoint.link(this); |
2517 | } |
2518 | |
2519 | // If the parenthese are capturing, store the ending index value to the |
2520 | // captures array, offsetting as necessary. |
2521 | // |
2522 | // FIXME: could avoid offsetting this value in JIT code, apply |
2523 | // offsets only afterwards, at the point the results array is |
2524 | // being accessed. |
2525 | if (term->capture() && compileMode == IncludeSubpatterns) { |
2526 | unsigned inputOffset = (m_checkedOffset - term->inputPosition).unsafeGet(); |
2527 | if (inputOffset) { |
2528 | move(index, indexTemporary); |
2529 | sub32(Imm32(inputOffset), indexTemporary); |
2530 | setSubpatternEnd(indexTemporary, term->parentheses.subpatternId); |
2531 | } else |
2532 | setSubpatternEnd(index, term->parentheses.subpatternId); |
2533 | } |
2534 | |
2535 | // If the parentheses are quantified Greedy then add a label to jump back |
2536 | // to if we get a failed match from after the parentheses. For NonGreedy |
2537 | // parentheses, link the jump from before the subpattern to here. |
2538 | if (term->quantityType == QuantifierGreedy) |
2539 | op.m_reentry = label(); |
2540 | else if (term->quantityType == QuantifierNonGreedy) { |
2541 | YarrOp& beginOp = m_ops[op.m_previousOp]; |
2542 | beginOp.m_jumps.link(this); |
2543 | } |
2544 | break; |
2545 | } |
2546 | |
2547 | // OpParenthesesSubpatternTerminalBegin/End |
2548 | case OpParenthesesSubpatternTerminalBegin: { |
2549 | PatternTerm* term = op.m_term; |
2550 | ASSERT(term->quantityType == QuantifierGreedy); |
2551 | ASSERT(term->quantityMaxCount == quantifyInfinite); |
2552 | ASSERT(!term->capture()); |
2553 | |
2554 | // Upon entry set a label to loop back to. |
2555 | op.m_reentry = label(); |
2556 | |
2557 | // Store the start index of the current match; we need to reject zero |
2558 | // length matches. |
2559 | storeToFrame(index, term->frameLocation + BackTrackInfoParenthesesTerminal::beginIndex()); |
2560 | break; |
2561 | } |
2562 | case OpParenthesesSubpatternTerminalEnd: { |
2563 | YarrOp& beginOp = m_ops[op.m_previousOp]; |
2564 | if (!ASSERT_DISABLED) { |
2565 | PatternTerm* term = op.m_term; |
2566 | |
2567 | // Runtime ASSERT to make sure that the nested alternative handled the |
2568 | // "no input consumed" check. |
2569 | Jump pastBreakpoint; |
2570 | pastBreakpoint = branch32(NotEqual, index, Address(stackPointerRegister, term->frameLocation * sizeof(void*))); |
2571 | abortWithReason(YARRNoInputConsumed); |
2572 | pastBreakpoint.link(this); |
2573 | } |
2574 | |
2575 | // We know that the match is non-zero, we can accept it and |
2576 | // loop back up to the head of the subpattern. |
2577 | jump(beginOp.m_reentry); |
2578 | |
2579 | // This is the entry point to jump to when we stop matching - we will |
2580 | // do so once the subpattern cannot match any more. |
2581 | op.m_reentry = label(); |
2582 | break; |
2583 | } |
2584 | |
2585 | // OpParenthesesSubpatternBegin/End |
2586 | // |
2587 | // These nodes support generic subpatterns. |
2588 | case OpParenthesesSubpatternBegin: { |
2589 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
2590 | PatternTerm* term = op.m_term; |
2591 | unsigned parenthesesFrameLocation = term->frameLocation; |
2592 | |
2593 | // Upon entry to a Greedy quantified set of parenthese store the index. |
2594 | // We'll use this for two purposes: |
2595 | // - To indicate which iteration we are on of mathing the remainder of |
2596 | // the expression after the parentheses - the first, including the |
2597 | // match within the parentheses, or the second having skipped over them. |
2598 | // - To check for empty matches, which must be rejected. |
2599 | // |
2600 | // At the head of a NonGreedy set of parentheses we'll immediately set 'begin' |
2601 | // in the backtrack info to -1 (indicating a match skipping the subpattern), |
2602 | // and plant a jump to the end. We'll also plant a label to backtrack to |
2603 | // to reenter the subpattern later, with a store to set 'begin' to current index |
2604 | // on the second iteration. |
2605 | // |
2606 | // FIXME: for capturing parens, could use the index in the capture array? |
2607 | if (term->quantityType == QuantifierGreedy || term->quantityType == QuantifierNonGreedy) { |
2608 | storeToFrame(TrustedImm32(0), parenthesesFrameLocation + BackTrackInfoParentheses::matchAmountIndex()); |
2609 | storeToFrame(TrustedImmPtr(nullptr), parenthesesFrameLocation + BackTrackInfoParentheses::parenContextHeadIndex()); |
2610 | |
2611 | if (term->quantityType == QuantifierNonGreedy) { |
2612 | storeToFrame(TrustedImm32(-1), parenthesesFrameLocation + BackTrackInfoParentheses::beginIndex()); |
2613 | op.m_jumps.append(jump()); |
2614 | } |
2615 | |
2616 | op.m_reentry = label(); |
2617 | RegisterID currParenContextReg = regT0; |
2618 | RegisterID newParenContextReg = regT1; |
2619 | |
2620 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoParentheses::parenContextHeadIndex(), currParenContextReg); |
2621 | allocateParenContext(newParenContextReg); |
2622 | storePtr(currParenContextReg, newParenContextReg); |
2623 | storeToFrame(newParenContextReg, parenthesesFrameLocation + BackTrackInfoParentheses::parenContextHeadIndex()); |
2624 | saveParenContext(newParenContextReg, regT2, term->parentheses.subpatternId, term->parentheses.lastSubpatternId, parenthesesFrameLocation); |
2625 | storeToFrame(index, parenthesesFrameLocation + BackTrackInfoParentheses::beginIndex()); |
2626 | } |
2627 | |
2628 | // If the parenthese are capturing, store the starting index value to the |
2629 | // captures array, offsetting as necessary. |
2630 | // |
2631 | // FIXME: could avoid offsetting this value in JIT code, apply |
2632 | // offsets only afterwards, at the point the results array is |
2633 | // being accessed. |
2634 | if (term->capture() && compileMode == IncludeSubpatterns) { |
2635 | const RegisterID indexTemporary = regT0; |
2636 | unsigned inputOffset = (m_checkedOffset - term->inputPosition).unsafeGet(); |
2637 | if (term->quantityType == QuantifierFixedCount) |
2638 | inputOffset += term->parentheses.disjunction->m_minimumSize; |
2639 | if (inputOffset) { |
2640 | move(index, indexTemporary); |
2641 | sub32(Imm32(inputOffset), indexTemporary); |
2642 | setSubpatternStart(indexTemporary, term->parentheses.subpatternId); |
2643 | } else |
2644 | setSubpatternStart(index, term->parentheses.subpatternId); |
2645 | } |
2646 | #else // !YARR_JIT_ALL_PARENS_EXPRESSIONS |
2647 | RELEASE_ASSERT_NOT_REACHED(); |
2648 | #endif |
2649 | break; |
2650 | } |
2651 | case OpParenthesesSubpatternEnd: { |
2652 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
2653 | PatternTerm* term = op.m_term; |
2654 | unsigned parenthesesFrameLocation = term->frameLocation; |
2655 | |
2656 | // Runtime ASSERT to make sure that the nested alternative handled the |
2657 | // "no input consumed" check. |
2658 | if (!ASSERT_DISABLED && term->quantityType != QuantifierFixedCount && !term->parentheses.disjunction->m_minimumSize) { |
2659 | Jump pastBreakpoint; |
2660 | pastBreakpoint = branch32(NotEqual, index, Address(stackPointerRegister, parenthesesFrameLocation * sizeof(void*))); |
2661 | abortWithReason(YARRNoInputConsumed); |
2662 | pastBreakpoint.link(this); |
2663 | } |
2664 | |
2665 | const RegisterID countTemporary = regT1; |
2666 | |
2667 | YarrOp& beginOp = m_ops[op.m_previousOp]; |
2668 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoParentheses::matchAmountIndex(), countTemporary); |
2669 | add32(TrustedImm32(1), countTemporary); |
2670 | storeToFrame(countTemporary, parenthesesFrameLocation + BackTrackInfoParentheses::matchAmountIndex()); |
2671 | |
2672 | // If the parenthese are capturing, store the ending index value to the |
2673 | // captures array, offsetting as necessary. |
2674 | // |
2675 | // FIXME: could avoid offsetting this value in JIT code, apply |
2676 | // offsets only afterwards, at the point the results array is |
2677 | // being accessed. |
2678 | if (term->capture() && compileMode == IncludeSubpatterns) { |
2679 | const RegisterID indexTemporary = regT0; |
2680 | |
2681 | unsigned inputOffset = (m_checkedOffset - term->inputPosition).unsafeGet(); |
2682 | if (inputOffset) { |
2683 | move(index, indexTemporary); |
2684 | sub32(Imm32(inputOffset), indexTemporary); |
2685 | setSubpatternEnd(indexTemporary, term->parentheses.subpatternId); |
2686 | } else |
2687 | setSubpatternEnd(index, term->parentheses.subpatternId); |
2688 | } |
2689 | |
2690 | // If the parentheses are quantified Greedy then add a label to jump back |
2691 | // to if we get a failed match from after the parentheses. For NonGreedy |
2692 | // parentheses, link the jump from before the subpattern to here. |
2693 | if (term->quantityType == QuantifierGreedy) { |
2694 | if (term->quantityMaxCount != quantifyInfinite) |
2695 | branch32(Below, countTemporary, Imm32(term->quantityMaxCount.unsafeGet())).linkTo(beginOp.m_reentry, this); |
2696 | else |
2697 | jump(beginOp.m_reentry); |
2698 | |
2699 | op.m_reentry = label(); |
2700 | } else if (term->quantityType == QuantifierNonGreedy) { |
2701 | YarrOp& beginOp = m_ops[op.m_previousOp]; |
2702 | beginOp.m_jumps.link(this); |
2703 | op.m_reentry = label(); |
2704 | } |
2705 | #else // !YARR_JIT_ALL_PARENS_EXPRESSIONS |
2706 | RELEASE_ASSERT_NOT_REACHED(); |
2707 | #endif |
2708 | break; |
2709 | } |
2710 | |
2711 | // OpParentheticalAssertionBegin/End |
2712 | case OpParentheticalAssertionBegin: { |
2713 | PatternTerm* term = op.m_term; |
2714 | |
2715 | // Store the current index - assertions should not update index, so |
2716 | // we will need to restore it upon a successful match. |
2717 | unsigned parenthesesFrameLocation = term->frameLocation; |
2718 | storeToFrame(index, parenthesesFrameLocation + BackTrackInfoParentheticalAssertion::beginIndex()); |
2719 | |
2720 | // Check |
2721 | op.m_checkAdjust = m_checkedOffset - term->inputPosition; |
2722 | if (op.m_checkAdjust) |
2723 | sub32(Imm32(op.m_checkAdjust.unsafeGet()), index); |
2724 | |
2725 | m_checkedOffset -= op.m_checkAdjust; |
2726 | break; |
2727 | } |
2728 | case OpParentheticalAssertionEnd: { |
2729 | PatternTerm* term = op.m_term; |
2730 | |
2731 | // Restore the input index value. |
2732 | unsigned parenthesesFrameLocation = term->frameLocation; |
2733 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoParentheticalAssertion::beginIndex(), index); |
2734 | |
2735 | // If inverted, a successful match of the assertion must be treated |
2736 | // as a failure, so jump to backtracking. |
2737 | if (term->invert()) { |
2738 | op.m_jumps.append(jump()); |
2739 | op.m_reentry = label(); |
2740 | } |
2741 | |
2742 | YarrOp& lastOp = m_ops[op.m_previousOp]; |
2743 | m_checkedOffset += lastOp.m_checkAdjust; |
2744 | break; |
2745 | } |
2746 | |
2747 | case OpMatchFailed: |
2748 | removeCallFrame(); |
2749 | generateFailReturn(); |
2750 | break; |
2751 | } |
2752 | |
2753 | ++opIndex; |
2754 | } while (opIndex < m_ops.size()); |
2755 | } |
2756 | |
2757 | void backtrack() |
2758 | { |
2759 | // Backwards generate the backtracking code. |
2760 | size_t opIndex = m_ops.size(); |
2761 | ASSERT(opIndex); |
2762 | |
2763 | do { |
2764 | --opIndex; |
2765 | |
2766 | if (m_disassembler) |
2767 | m_disassembler->setForBacktrack(opIndex, label()); |
2768 | |
2769 | YarrOp& op = m_ops[opIndex]; |
2770 | switch (op.m_op) { |
2771 | |
2772 | case OpTerm: |
2773 | backtrackTerm(opIndex); |
2774 | break; |
2775 | |
2776 | // OpBodyAlternativeBegin/Next/End |
2777 | // |
2778 | // For each Begin/Next node representing an alternative, we need to decide what to do |
2779 | // in two circumstances: |
2780 | // - If we backtrack back into this node, from within the alternative. |
2781 | // - If the input check at the head of the alternative fails (if this exists). |
2782 | // |
2783 | // We treat these two cases differently since in the former case we have slightly |
2784 | // more information - since we are backtracking out of a prior alternative we know |
2785 | // that at least enough input was available to run it. For example, given the regular |
2786 | // expression /a|b/, if we backtrack out of the first alternative (a failed pattern |
2787 | // character match of 'a'), then we need not perform an additional input availability |
2788 | // check before running the second alternative. |
2789 | // |
2790 | // Backtracking required differs for the last alternative, which in the case of the |
2791 | // repeating set of alternatives must loop. The code generated for the last alternative |
2792 | // will also be used to handle all input check failures from any prior alternatives - |
2793 | // these require similar functionality, in seeking the next available alternative for |
2794 | // which there is sufficient input. |
2795 | // |
2796 | // Since backtracking of all other alternatives simply requires us to link backtracks |
2797 | // to the reentry point for the subsequent alternative, we will only be generating any |
2798 | // code when backtracking the last alternative. |
2799 | case OpBodyAlternativeBegin: |
2800 | case OpBodyAlternativeNext: { |
2801 | PatternAlternative* alternative = op.m_alternative; |
2802 | |
2803 | if (op.m_op == OpBodyAlternativeNext) { |
2804 | PatternAlternative* priorAlternative = m_ops[op.m_previousOp].m_alternative; |
2805 | m_checkedOffset += priorAlternative->m_minimumSize; |
2806 | } |
2807 | m_checkedOffset -= alternative->m_minimumSize; |
2808 | |
2809 | // Is this the last alternative? If not, then if we backtrack to this point we just |
2810 | // need to jump to try to match the next alternative. |
2811 | if (m_ops[op.m_nextOp].m_op != OpBodyAlternativeEnd) { |
2812 | m_backtrackingState.linkTo(m_ops[op.m_nextOp].m_reentry, this); |
2813 | break; |
2814 | } |
2815 | YarrOp& endOp = m_ops[op.m_nextOp]; |
2816 | |
2817 | YarrOp* beginOp = &op; |
2818 | while (beginOp->m_op != OpBodyAlternativeBegin) { |
2819 | ASSERT(beginOp->m_op == OpBodyAlternativeNext); |
2820 | beginOp = &m_ops[beginOp->m_previousOp]; |
2821 | } |
2822 | |
2823 | bool onceThrough = endOp.m_nextOp == notFound; |
2824 | |
2825 | JumpList lastStickyAlternativeFailures; |
2826 | |
2827 | // First, generate code to handle cases where we backtrack out of an attempted match |
2828 | // of the last alternative. If this is a 'once through' set of alternatives then we |
2829 | // have nothing to do - link this straight through to the End. |
2830 | if (onceThrough) |
2831 | m_backtrackingState.linkTo(endOp.m_reentry, this); |
2832 | else { |
2833 | // If we don't need to move the input poistion, and the pattern has a fixed size |
2834 | // (in which case we omit the store of the start index until the pattern has matched) |
2835 | // then we can just link the backtrack out of the last alternative straight to the |
2836 | // head of the first alternative. |
2837 | if (m_pattern.m_body->m_hasFixedSize |
2838 | && (alternative->m_minimumSize > beginOp->m_alternative->m_minimumSize) |
2839 | && (alternative->m_minimumSize - beginOp->m_alternative->m_minimumSize == 1)) |
2840 | m_backtrackingState.linkTo(beginOp->m_reentry, this); |
2841 | else if (m_pattern.sticky() && m_ops[op.m_nextOp].m_op == OpBodyAlternativeEnd) { |
2842 | // It is a sticky pattern and the last alternative failed, jump to the end. |
2843 | m_backtrackingState.takeBacktracksToJumpList(lastStickyAlternativeFailures, this); |
2844 | } else { |
2845 | // We need to generate a trampoline of code to execute before looping back |
2846 | // around to the first alternative. |
2847 | m_backtrackingState.link(this); |
2848 | |
2849 | // No need to advance and retry for a sticky pattern. |
2850 | if (!m_pattern.sticky()) { |
2851 | // If the pattern size is not fixed, then store the start index for use if we match. |
2852 | if (!m_pattern.m_body->m_hasFixedSize) { |
2853 | if (alternative->m_minimumSize == 1) |
2854 | setMatchStart(index); |
2855 | else { |
2856 | move(index, regT0); |
2857 | if (alternative->m_minimumSize) |
2858 | sub32(Imm32(alternative->m_minimumSize - 1), regT0); |
2859 | else |
2860 | add32(TrustedImm32(1), regT0); |
2861 | setMatchStart(regT0); |
2862 | } |
2863 | } |
2864 | |
2865 | // Generate code to loop. Check whether the last alternative is longer than the |
2866 | // first (e.g. /a|xy/ or /a|xyz/). |
2867 | if (alternative->m_minimumSize > beginOp->m_alternative->m_minimumSize) { |
2868 | // We want to loop, and increment input position. If the delta is 1, it is |
2869 | // already correctly incremented, if more than one then decrement as appropriate. |
2870 | unsigned delta = alternative->m_minimumSize - beginOp->m_alternative->m_minimumSize; |
2871 | ASSERT(delta); |
2872 | if (delta != 1) |
2873 | sub32(Imm32(delta - 1), index); |
2874 | jump(beginOp->m_reentry); |
2875 | } else { |
2876 | // If the first alternative has minimum size 0xFFFFFFFFu, then there cannot |
2877 | // be sufficent input available to handle this, so just fall through. |
2878 | unsigned delta = beginOp->m_alternative->m_minimumSize - alternative->m_minimumSize; |
2879 | if (delta != 0xFFFFFFFFu) { |
2880 | // We need to check input because we are incrementing the input. |
2881 | add32(Imm32(delta + 1), index); |
2882 | checkInput().linkTo(beginOp->m_reentry, this); |
2883 | } |
2884 | } |
2885 | } |
2886 | } |
2887 | } |
2888 | |
2889 | // We can reach this point in the code in two ways: |
2890 | // - Fallthrough from the code above (a repeating alternative backtracked out of its |
2891 | // last alternative, and did not have sufficent input to run the first). |
2892 | // - We will loop back up to the following label when a repeating alternative loops, |
2893 | // following a failed input check. |
2894 | // |
2895 | // Either way, we have just failed the input check for the first alternative. |
2896 | Label firstInputCheckFailed(this); |
2897 | |
2898 | // Generate code to handle input check failures from alternatives except the last. |
2899 | // prevOp is the alternative we're handling a bail out from (initially Begin), and |
2900 | // nextOp is the alternative we will be attempting to reenter into. |
2901 | // |
2902 | // We will link input check failures from the forwards matching path back to the code |
2903 | // that can handle them. |
2904 | YarrOp* prevOp = beginOp; |
2905 | YarrOp* nextOp = &m_ops[beginOp->m_nextOp]; |
2906 | while (nextOp->m_op != OpBodyAlternativeEnd) { |
2907 | prevOp->m_jumps.link(this); |
2908 | |
2909 | // We only get here if an input check fails, it is only worth checking again |
2910 | // if the next alternative has a minimum size less than the last. |
2911 | if (prevOp->m_alternative->m_minimumSize > nextOp->m_alternative->m_minimumSize) { |
2912 | // FIXME: if we added an extra label to YarrOp, we could avoid needing to |
2913 | // subtract delta back out, and reduce this code. Should performance test |
2914 | // the benefit of this. |
2915 | unsigned delta = prevOp->m_alternative->m_minimumSize - nextOp->m_alternative->m_minimumSize; |
2916 | sub32(Imm32(delta), index); |
2917 | Jump fail = jumpIfNoAvailableInput(); |
2918 | add32(Imm32(delta), index); |
2919 | jump(nextOp->m_reentry); |
2920 | fail.link(this); |
2921 | } else if (prevOp->m_alternative->m_minimumSize < nextOp->m_alternative->m_minimumSize) |
2922 | add32(Imm32(nextOp->m_alternative->m_minimumSize - prevOp->m_alternative->m_minimumSize), index); |
2923 | prevOp = nextOp; |
2924 | nextOp = &m_ops[nextOp->m_nextOp]; |
2925 | } |
2926 | |
2927 | // We fall through to here if there is insufficient input to run the last alternative. |
2928 | |
2929 | // If there is insufficient input to run the last alternative, then for 'once through' |
2930 | // alternatives we are done - just jump back up into the forwards matching path at the End. |
2931 | if (onceThrough) { |
2932 | op.m_jumps.linkTo(endOp.m_reentry, this); |
2933 | jump(endOp.m_reentry); |
2934 | break; |
2935 | } |
2936 | |
2937 | // For repeating alternatives, link any input check failure from the last alternative to |
2938 | // this point. |
2939 | op.m_jumps.link(this); |
2940 | |
2941 | bool needsToUpdateMatchStart = !m_pattern.m_body->m_hasFixedSize; |
2942 | |
2943 | // Check for cases where input position is already incremented by 1 for the last |
2944 | // alternative (this is particularly useful where the minimum size of the body |
2945 | // disjunction is 0, e.g. /a*|b/). |
2946 | if (needsToUpdateMatchStart && alternative->m_minimumSize == 1) { |
2947 | // index is already incremented by 1, so just store it now! |
2948 | setMatchStart(index); |
2949 | needsToUpdateMatchStart = false; |
2950 | } |
2951 | |
2952 | if (!m_pattern.sticky()) { |
2953 | // Check whether there is sufficient input to loop. Increment the input position by |
2954 | // one, and check. Also add in the minimum disjunction size before checking - there |
2955 | // is no point in looping if we're just going to fail all the input checks around |
2956 | // the next iteration. |
2957 | ASSERT(alternative->m_minimumSize >= m_pattern.m_body->m_minimumSize); |
2958 | if (alternative->m_minimumSize == m_pattern.m_body->m_minimumSize) { |
2959 | // If the last alternative had the same minimum size as the disjunction, |
2960 | // just simply increment input pos by 1, no adjustment based on minimum size. |
2961 | add32(TrustedImm32(1), index); |
2962 | } else { |
2963 | // If the minumum for the last alternative was one greater than than that |
2964 | // for the disjunction, we're already progressed by 1, nothing to do! |
2965 | unsigned delta = (alternative->m_minimumSize - m_pattern.m_body->m_minimumSize) - 1; |
2966 | if (delta) |
2967 | sub32(Imm32(delta), index); |
2968 | } |
2969 | Jump matchFailed = jumpIfNoAvailableInput(); |
2970 | |
2971 | if (needsToUpdateMatchStart) { |
2972 | if (!m_pattern.m_body->m_minimumSize) |
2973 | setMatchStart(index); |
2974 | else { |
2975 | move(index, regT0); |
2976 | sub32(Imm32(m_pattern.m_body->m_minimumSize), regT0); |
2977 | setMatchStart(regT0); |
2978 | } |
2979 | } |
2980 | |
2981 | // Calculate how much more input the first alternative requires than the minimum |
2982 | // for the body as a whole. If no more is needed then we dont need an additional |
2983 | // input check here - jump straight back up to the start of the first alternative. |
2984 | if (beginOp->m_alternative->m_minimumSize == m_pattern.m_body->m_minimumSize) |
2985 | jump(beginOp->m_reentry); |
2986 | else { |
2987 | if (beginOp->m_alternative->m_minimumSize > m_pattern.m_body->m_minimumSize) |
2988 | add32(Imm32(beginOp->m_alternative->m_minimumSize - m_pattern.m_body->m_minimumSize), index); |
2989 | else |
2990 | sub32(Imm32(m_pattern.m_body->m_minimumSize - beginOp->m_alternative->m_minimumSize), index); |
2991 | checkInput().linkTo(beginOp->m_reentry, this); |
2992 | jump(firstInputCheckFailed); |
2993 | } |
2994 | |
2995 | // We jump to here if we iterate to the point that there is insufficient input to |
2996 | // run any matches, and need to return a failure state from JIT code. |
2997 | matchFailed.link(this); |
2998 | } |
2999 | |
3000 | lastStickyAlternativeFailures.link(this); |
3001 | removeCallFrame(); |
3002 | generateFailReturn(); |
3003 | break; |
3004 | } |
3005 | case OpBodyAlternativeEnd: { |
3006 | // We should never backtrack back into a body disjunction. |
3007 | ASSERT(m_backtrackingState.isEmpty()); |
3008 | |
3009 | PatternAlternative* priorAlternative = m_ops[op.m_previousOp].m_alternative; |
3010 | m_checkedOffset += priorAlternative->m_minimumSize; |
3011 | break; |
3012 | } |
3013 | |
3014 | // OpSimpleNestedAlternativeBegin/Next/End |
3015 | // OpNestedAlternativeBegin/Next/End |
3016 | // |
3017 | // Generate code for when we backtrack back out of an alternative into |
3018 | // a Begin or Next node, or when the entry input count check fails. If |
3019 | // there are more alternatives we need to jump to the next alternative, |
3020 | // if not we backtrack back out of the current set of parentheses. |
3021 | // |
3022 | // In the case of non-simple nested assertions we need to also link the |
3023 | // 'return address' appropriately to backtrack back out into the correct |
3024 | // alternative. |
3025 | case OpSimpleNestedAlternativeBegin: |
3026 | case OpSimpleNestedAlternativeNext: |
3027 | case OpNestedAlternativeBegin: |
3028 | case OpNestedAlternativeNext: { |
3029 | YarrOp& nextOp = m_ops[op.m_nextOp]; |
3030 | bool isBegin = op.m_previousOp == notFound; |
3031 | bool isLastAlternative = nextOp.m_nextOp == notFound; |
3032 | ASSERT(isBegin == (op.m_op == OpSimpleNestedAlternativeBegin || op.m_op == OpNestedAlternativeBegin)); |
3033 | ASSERT(isLastAlternative == (nextOp.m_op == OpSimpleNestedAlternativeEnd || nextOp.m_op == OpNestedAlternativeEnd)); |
3034 | |
3035 | // Treat an input check failure the same as a failed match. |
3036 | m_backtrackingState.append(op.m_jumps); |
3037 | |
3038 | // Set the backtracks to jump to the appropriate place. We may need |
3039 | // to link the backtracks in one of three different way depending on |
3040 | // the type of alternative we are dealing with: |
3041 | // - A single alternative, with no simplings. |
3042 | // - The last alternative of a set of two or more. |
3043 | // - An alternative other than the last of a set of two or more. |
3044 | // |
3045 | // In the case of a single alternative on its own, we don't need to |
3046 | // jump anywhere - if the alternative fails to match we can just |
3047 | // continue to backtrack out of the parentheses without jumping. |
3048 | // |
3049 | // In the case of the last alternative in a set of more than one, we |
3050 | // need to jump to return back out to the beginning. We'll do so by |
3051 | // adding a jump to the End node's m_jumps list, and linking this |
3052 | // when we come to generate the Begin node. For alternatives other |
3053 | // than the last, we need to jump to the next alternative. |
3054 | // |
3055 | // If the alternative had adjusted the input position we must link |
3056 | // backtracking to here, correct, and then jump on. If not we can |
3057 | // link the backtracks directly to their destination. |
3058 | if (op.m_checkAdjust) { |
3059 | // Handle the cases where we need to link the backtracks here. |
3060 | m_backtrackingState.link(this); |
3061 | sub32(Imm32(op.m_checkAdjust.unsafeGet()), index); |
3062 | if (!isLastAlternative) { |
3063 | // An alternative that is not the last should jump to its successor. |
3064 | jump(nextOp.m_reentry); |
3065 | } else if (!isBegin) { |
3066 | // The last of more than one alternatives must jump back to the beginning. |
3067 | nextOp.m_jumps.append(jump()); |
3068 | } else { |
3069 | // A single alternative on its own can fall through. |
3070 | m_backtrackingState.fallthrough(); |
3071 | } |
3072 | } else { |
3073 | // Handle the cases where we can link the backtracks directly to their destinations. |
3074 | if (!isLastAlternative) { |
3075 | // An alternative that is not the last should jump to its successor. |
3076 | m_backtrackingState.linkTo(nextOp.m_reentry, this); |
3077 | } else if (!isBegin) { |
3078 | // The last of more than one alternatives must jump back to the beginning. |
3079 | m_backtrackingState.takeBacktracksToJumpList(nextOp.m_jumps, this); |
3080 | } |
3081 | // In the case of a single alternative on its own do nothing - it can fall through. |
3082 | } |
3083 | |
3084 | // If there is a backtrack jump from a zero length match link it here. |
3085 | if (op.m_zeroLengthMatch.isSet()) |
3086 | m_backtrackingState.append(op.m_zeroLengthMatch); |
3087 | |
3088 | // At this point we've handled the backtracking back into this node. |
3089 | // Now link any backtracks that need to jump to here. |
3090 | |
3091 | // For non-simple alternatives, link the alternative's 'return address' |
3092 | // so that we backtrack back out into the previous alternative. |
3093 | if (op.m_op == OpNestedAlternativeNext) |
3094 | m_backtrackingState.append(op.m_returnAddress); |
3095 | |
3096 | // If there is more than one alternative, then the last alternative will |
3097 | // have planted a jump to be linked to the end. This jump was added to the |
3098 | // End node's m_jumps list. If we are back at the beginning, link it here. |
3099 | if (isBegin) { |
3100 | YarrOp* endOp = &m_ops[op.m_nextOp]; |
3101 | while (endOp->m_nextOp != notFound) { |
3102 | ASSERT(endOp->m_op == OpSimpleNestedAlternativeNext || endOp->m_op == OpNestedAlternativeNext); |
3103 | endOp = &m_ops[endOp->m_nextOp]; |
3104 | } |
3105 | ASSERT(endOp->m_op == OpSimpleNestedAlternativeEnd || endOp->m_op == OpNestedAlternativeEnd); |
3106 | m_backtrackingState.append(endOp->m_jumps); |
3107 | } |
3108 | |
3109 | if (!isBegin) { |
3110 | YarrOp& lastOp = m_ops[op.m_previousOp]; |
3111 | m_checkedOffset += lastOp.m_checkAdjust; |
3112 | } |
3113 | m_checkedOffset -= op.m_checkAdjust; |
3114 | break; |
3115 | } |
3116 | case OpSimpleNestedAlternativeEnd: |
3117 | case OpNestedAlternativeEnd: { |
3118 | PatternTerm* term = op.m_term; |
3119 | |
3120 | // If there is a backtrack jump from a zero length match link it here. |
3121 | if (op.m_zeroLengthMatch.isSet()) |
3122 | m_backtrackingState.append(op.m_zeroLengthMatch); |
3123 | |
3124 | // If we backtrack into the end of a simple subpattern do nothing; |
3125 | // just continue through into the last alternative. If we backtrack |
3126 | // into the end of a non-simple set of alterntives we need to jump |
3127 | // to the backtracking return address set up during generation. |
3128 | if (op.m_op == OpNestedAlternativeEnd) { |
3129 | m_backtrackingState.link(this); |
3130 | |
3131 | // Plant a jump to the return address. |
3132 | unsigned parenthesesFrameLocation = term->frameLocation; |
3133 | loadFromFrameAndJump(parenthesesFrameLocation + BackTrackInfoParentheses::returnAddressIndex()); |
3134 | |
3135 | // Link the DataLabelPtr associated with the end of the last |
3136 | // alternative to this point. |
3137 | m_backtrackingState.append(op.m_returnAddress); |
3138 | } |
3139 | |
3140 | YarrOp& lastOp = m_ops[op.m_previousOp]; |
3141 | m_checkedOffset += lastOp.m_checkAdjust; |
3142 | break; |
3143 | } |
3144 | |
3145 | // OpParenthesesSubpatternOnceBegin/End |
3146 | // |
3147 | // When we are backtracking back out of a capturing subpattern we need |
3148 | // to clear the start index in the matches output array, to record that |
3149 | // this subpattern has not been captured. |
3150 | // |
3151 | // When backtracking back out of a Greedy quantified subpattern we need |
3152 | // to catch this, and try running the remainder of the alternative after |
3153 | // the subpattern again, skipping the parentheses. |
3154 | // |
3155 | // Upon backtracking back into a quantified set of parentheses we need to |
3156 | // check whether we were currently skipping the subpattern. If not, we |
3157 | // can backtrack into them, if we were we need to either backtrack back |
3158 | // out of the start of the parentheses, or jump back to the forwards |
3159 | // matching start, depending of whether the match is Greedy or NonGreedy. |
3160 | case OpParenthesesSubpatternOnceBegin: { |
3161 | PatternTerm* term = op.m_term; |
3162 | ASSERT(term->quantityMaxCount == 1); |
3163 | |
3164 | // We only need to backtrack to this point if capturing or greedy. |
3165 | if ((term->capture() && compileMode == IncludeSubpatterns) || term->quantityType == QuantifierGreedy) { |
3166 | m_backtrackingState.link(this); |
3167 | |
3168 | // If capturing, clear the capture (we only need to reset start). |
3169 | if (term->capture() && compileMode == IncludeSubpatterns) |
3170 | clearSubpatternStart(term->parentheses.subpatternId); |
3171 | |
3172 | // If Greedy, jump to the end. |
3173 | if (term->quantityType == QuantifierGreedy) { |
3174 | // Clear the flag in the stackframe indicating we ran through the subpattern. |
3175 | unsigned parenthesesFrameLocation = term->frameLocation; |
3176 | storeToFrame(TrustedImm32(-1), parenthesesFrameLocation + BackTrackInfoParenthesesOnce::beginIndex()); |
3177 | // Jump to after the parentheses, skipping the subpattern. |
3178 | jump(m_ops[op.m_nextOp].m_reentry); |
3179 | // A backtrack from after the parentheses, when skipping the subpattern, |
3180 | // will jump back to here. |
3181 | op.m_jumps.link(this); |
3182 | } |
3183 | |
3184 | m_backtrackingState.fallthrough(); |
3185 | } |
3186 | break; |
3187 | } |
3188 | case OpParenthesesSubpatternOnceEnd: { |
3189 | PatternTerm* term = op.m_term; |
3190 | |
3191 | if (term->quantityType != QuantifierFixedCount) { |
3192 | m_backtrackingState.link(this); |
3193 | |
3194 | // Check whether we should backtrack back into the parentheses, or if we |
3195 | // are currently in a state where we had skipped over the subpattern |
3196 | // (in which case the flag value on the stack will be -1). |
3197 | unsigned parenthesesFrameLocation = term->frameLocation; |
3198 | Jump hadSkipped = branch32(Equal, Address(stackPointerRegister, (parenthesesFrameLocation + BackTrackInfoParenthesesOnce::beginIndex()) * sizeof(void*)), TrustedImm32(-1)); |
3199 | |
3200 | if (term->quantityType == QuantifierGreedy) { |
3201 | // For Greedy parentheses, we skip after having already tried going |
3202 | // through the subpattern, so if we get here we're done. |
3203 | YarrOp& beginOp = m_ops[op.m_previousOp]; |
3204 | beginOp.m_jumps.append(hadSkipped); |
3205 | } else { |
3206 | // For NonGreedy parentheses, we try skipping the subpattern first, |
3207 | // so if we get here we need to try running through the subpattern |
3208 | // next. Jump back to the start of the parentheses in the forwards |
3209 | // matching path. |
3210 | ASSERT(term->quantityType == QuantifierNonGreedy); |
3211 | YarrOp& beginOp = m_ops[op.m_previousOp]; |
3212 | hadSkipped.linkTo(beginOp.m_reentry, this); |
3213 | } |
3214 | |
3215 | m_backtrackingState.fallthrough(); |
3216 | } |
3217 | |
3218 | m_backtrackingState.append(op.m_jumps); |
3219 | break; |
3220 | } |
3221 | |
3222 | // OpParenthesesSubpatternTerminalBegin/End |
3223 | // |
3224 | // Terminal subpatterns will always match - there is nothing after them to |
3225 | // force a backtrack, and they have a minimum count of 0, and as such will |
3226 | // always produce an acceptable result. |
3227 | case OpParenthesesSubpatternTerminalBegin: { |
3228 | // We will backtrack to this point once the subpattern cannot match any |
3229 | // more. Since no match is accepted as a successful match (we are Greedy |
3230 | // quantified with a minimum of zero) jump back to the forwards matching |
3231 | // path at the end. |
3232 | YarrOp& endOp = m_ops[op.m_nextOp]; |
3233 | m_backtrackingState.linkTo(endOp.m_reentry, this); |
3234 | break; |
3235 | } |
3236 | case OpParenthesesSubpatternTerminalEnd: |
3237 | // We should never be backtracking to here (hence the 'terminal' in the name). |
3238 | ASSERT(m_backtrackingState.isEmpty()); |
3239 | m_backtrackingState.append(op.m_jumps); |
3240 | break; |
3241 | |
3242 | // OpParenthesesSubpatternBegin/End |
3243 | // |
3244 | // When we are backtracking back out of a capturing subpattern we need |
3245 | // to clear the start index in the matches output array, to record that |
3246 | // this subpattern has not been captured. |
3247 | // |
3248 | // When backtracking back out of a Greedy quantified subpattern we need |
3249 | // to catch this, and try running the remainder of the alternative after |
3250 | // the subpattern again, skipping the parentheses. |
3251 | // |
3252 | // Upon backtracking back into a quantified set of parentheses we need to |
3253 | // check whether we were currently skipping the subpattern. If not, we |
3254 | // can backtrack into them, if we were we need to either backtrack back |
3255 | // out of the start of the parentheses, or jump back to the forwards |
3256 | // matching start, depending of whether the match is Greedy or NonGreedy. |
3257 | case OpParenthesesSubpatternBegin: { |
3258 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3259 | PatternTerm* term = op.m_term; |
3260 | unsigned parenthesesFrameLocation = term->frameLocation; |
3261 | |
3262 | if (term->quantityType != QuantifierFixedCount) { |
3263 | m_backtrackingState.link(this); |
3264 | |
3265 | RegisterID currParenContextReg = regT0; |
3266 | RegisterID newParenContextReg = regT1; |
3267 | |
3268 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoParentheses::parenContextHeadIndex(), currParenContextReg); |
3269 | |
3270 | restoreParenContext(currParenContextReg, regT2, term->parentheses.subpatternId, term->parentheses.lastSubpatternId, parenthesesFrameLocation); |
3271 | |
3272 | freeParenContext(currParenContextReg, newParenContextReg); |
3273 | storeToFrame(newParenContextReg, parenthesesFrameLocation + BackTrackInfoParentheses::parenContextHeadIndex()); |
3274 | |
3275 | const RegisterID countTemporary = regT0; |
3276 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoParentheses::matchAmountIndex(), countTemporary); |
3277 | Jump zeroLengthMatch = branchTest32(Zero, countTemporary); |
3278 | |
3279 | sub32(TrustedImm32(1), countTemporary); |
3280 | storeToFrame(countTemporary, parenthesesFrameLocation + BackTrackInfoParentheses::matchAmountIndex()); |
3281 | |
3282 | jump(m_ops[op.m_nextOp].m_reentry); |
3283 | |
3284 | zeroLengthMatch.link(this); |
3285 | |
3286 | // Clear the flag in the stackframe indicating we didn't run through the subpattern. |
3287 | storeToFrame(TrustedImm32(-1), parenthesesFrameLocation + BackTrackInfoParentheses::beginIndex()); |
3288 | |
3289 | if (term->quantityType == QuantifierGreedy) |
3290 | jump(m_ops[op.m_nextOp].m_reentry); |
3291 | |
3292 | // If Greedy, jump to the end. |
3293 | if (term->quantityType == QuantifierGreedy) { |
3294 | // A backtrack from after the parentheses, when skipping the subpattern, |
3295 | // will jump back to here. |
3296 | op.m_jumps.link(this); |
3297 | } |
3298 | |
3299 | m_backtrackingState.fallthrough(); |
3300 | } |
3301 | #else // !YARR_JIT_ALL_PARENS_EXPRESSIONS |
3302 | RELEASE_ASSERT_NOT_REACHED(); |
3303 | #endif |
3304 | break; |
3305 | } |
3306 | case OpParenthesesSubpatternEnd: { |
3307 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3308 | PatternTerm* term = op.m_term; |
3309 | |
3310 | if (term->quantityType != QuantifierFixedCount) { |
3311 | m_backtrackingState.link(this); |
3312 | |
3313 | unsigned parenthesesFrameLocation = term->frameLocation; |
3314 | |
3315 | if (term->quantityType == QuantifierGreedy) { |
3316 | // Check whether we should backtrack back into the parentheses, or if we |
3317 | // are currently in a state where we had skipped over the subpattern |
3318 | // (in which case the flag value on the stack will be -1). |
3319 | Jump hadSkipped = branch32(Equal, Address(stackPointerRegister, (parenthesesFrameLocation + BackTrackInfoParentheses::beginIndex()) * sizeof(void*)), TrustedImm32(-1)); |
3320 | |
3321 | // For Greedy parentheses, we skip after having already tried going |
3322 | // through the subpattern, so if we get here we're done. |
3323 | YarrOp& beginOp = m_ops[op.m_previousOp]; |
3324 | beginOp.m_jumps.append(hadSkipped); |
3325 | } else { |
3326 | // For NonGreedy parentheses, we try skipping the subpattern first, |
3327 | // so if we get here we need to try running through the subpattern |
3328 | // next. Jump back to the start of the parentheses in the forwards |
3329 | // matching path. |
3330 | ASSERT(term->quantityType == QuantifierNonGreedy); |
3331 | |
3332 | const RegisterID beginTemporary = regT0; |
3333 | const RegisterID countTemporary = regT1; |
3334 | |
3335 | YarrOp& beginOp = m_ops[op.m_previousOp]; |
3336 | |
3337 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoParentheses::beginIndex(), beginTemporary); |
3338 | branch32(Equal, beginTemporary, TrustedImm32(-1)).linkTo(beginOp.m_reentry, this); |
3339 | |
3340 | JumpList exceededMatchLimit; |
3341 | |
3342 | if (term->quantityMaxCount != quantifyInfinite) { |
3343 | loadFromFrame(parenthesesFrameLocation + BackTrackInfoParentheses::matchAmountIndex(), countTemporary); |
3344 | exceededMatchLimit.append(branch32(AboveOrEqual, countTemporary, Imm32(term->quantityMaxCount.unsafeGet()))); |
3345 | } |
3346 | |
3347 | branch32(Above, index, beginTemporary).linkTo(beginOp.m_reentry, this); |
3348 | |
3349 | exceededMatchLimit.link(this); |
3350 | } |
3351 | |
3352 | m_backtrackingState.fallthrough(); |
3353 | } |
3354 | |
3355 | m_backtrackingState.append(op.m_jumps); |
3356 | #else // !YARR_JIT_ALL_PARENS_EXPRESSIONS |
3357 | RELEASE_ASSERT_NOT_REACHED(); |
3358 | #endif |
3359 | break; |
3360 | } |
3361 | |
3362 | // OpParentheticalAssertionBegin/End |
3363 | case OpParentheticalAssertionBegin: { |
3364 | PatternTerm* term = op.m_term; |
3365 | YarrOp& endOp = m_ops[op.m_nextOp]; |
3366 | |
3367 | // We need to handle the backtracks upon backtracking back out |
3368 | // of a parenthetical assertion if either we need to correct |
3369 | // the input index, or the assertion was inverted. |
3370 | if (op.m_checkAdjust || term->invert()) { |
3371 | m_backtrackingState.link(this); |
3372 | |
3373 | if (op.m_checkAdjust) |
3374 | add32(Imm32(op.m_checkAdjust.unsafeGet()), index); |
3375 | |
3376 | // In an inverted assertion failure to match the subpattern |
3377 | // is treated as a successful match - jump to the end of the |
3378 | // subpattern. We already have adjusted the input position |
3379 | // back to that before the assertion, which is correct. |
3380 | if (term->invert()) |
3381 | jump(endOp.m_reentry); |
3382 | |
3383 | m_backtrackingState.fallthrough(); |
3384 | } |
3385 | |
3386 | // The End node's jump list will contain any backtracks into |
3387 | // the end of the assertion. Also, if inverted, we will have |
3388 | // added the failure caused by a successful match to this. |
3389 | m_backtrackingState.append(endOp.m_jumps); |
3390 | |
3391 | m_checkedOffset += op.m_checkAdjust; |
3392 | break; |
3393 | } |
3394 | case OpParentheticalAssertionEnd: { |
3395 | // FIXME: We should really be clearing any nested subpattern |
3396 | // matches on bailing out from after the pattern. Firefox has |
3397 | // this bug too (presumably because they use YARR!) |
3398 | |
3399 | // Never backtrack into an assertion; later failures bail to before the begin. |
3400 | m_backtrackingState.takeBacktracksToJumpList(op.m_jumps, this); |
3401 | |
3402 | YarrOp& lastOp = m_ops[op.m_previousOp]; |
3403 | m_checkedOffset -= lastOp.m_checkAdjust; |
3404 | break; |
3405 | } |
3406 | |
3407 | case OpMatchFailed: |
3408 | break; |
3409 | } |
3410 | |
3411 | } while (opIndex); |
3412 | } |
3413 | |
3414 | // Compilation methods: |
3415 | // ==================== |
3416 | |
3417 | // opCompileParenthesesSubpattern |
3418 | // Emits ops for a subpattern (set of parentheses). These consist |
3419 | // of a set of alternatives wrapped in an outer set of nodes for |
3420 | // the parentheses. |
3421 | // Supported types of parentheses are 'Once' (quantityMaxCount == 1), |
3422 | // 'Terminal' (non-capturing parentheses quantified as greedy |
3423 | // and infinite), and 0 based greedy / non-greedy quantified parentheses. |
3424 | // Alternatives will use the 'Simple' set of ops if either the |
3425 | // subpattern is terminal (in which case we will never need to |
3426 | // backtrack), or if the subpattern only contains one alternative. |
3427 | void opCompileParenthesesSubpattern(PatternTerm* term) |
3428 | { |
3429 | YarrOpCode parenthesesBeginOpCode; |
3430 | YarrOpCode parenthesesEndOpCode; |
3431 | YarrOpCode alternativeBeginOpCode = OpSimpleNestedAlternativeBegin; |
3432 | YarrOpCode alternativeNextOpCode = OpSimpleNestedAlternativeNext; |
3433 | YarrOpCode alternativeEndOpCode = OpSimpleNestedAlternativeEnd; |
3434 | |
3435 | if (UNLIKELY(!m_vm->isSafeToRecurse())) { |
3436 | m_failureReason = JITFailureReason::ParenthesisNestedTooDeep; |
3437 | return; |
3438 | } |
3439 | |
3440 | // We can currently only compile quantity 1 subpatterns that are |
3441 | // not copies. We generate a copy in the case of a range quantifier, |
3442 | // e.g. /(?:x){3,9}/, or /(?:x)+/ (These are effectively expanded to |
3443 | // /(?:x){3,3}(?:x){0,6}/ and /(?:x)(?:x)*/ repectively). The problem |
3444 | // comes where the subpattern is capturing, in which case we would |
3445 | // need to restore the capture from the first subpattern upon a |
3446 | // failure in the second. |
3447 | if (term->quantityMinCount && term->quantityMinCount != term->quantityMaxCount) { |
3448 | m_failureReason = JITFailureReason::VariableCountedParenthesisWithNonZeroMinimum; |
3449 | return; |
3450 | } |
3451 | |
3452 | if (term->quantityMaxCount == 1 && !term->parentheses.isCopy) { |
3453 | // Select the 'Once' nodes. |
3454 | parenthesesBeginOpCode = OpParenthesesSubpatternOnceBegin; |
3455 | parenthesesEndOpCode = OpParenthesesSubpatternOnceEnd; |
3456 | |
3457 | // If there is more than one alternative we cannot use the 'simple' nodes. |
3458 | if (term->parentheses.disjunction->m_alternatives.size() != 1) { |
3459 | alternativeBeginOpCode = OpNestedAlternativeBegin; |
3460 | alternativeNextOpCode = OpNestedAlternativeNext; |
3461 | alternativeEndOpCode = OpNestedAlternativeEnd; |
3462 | } |
3463 | } else if (term->parentheses.isTerminal) { |
3464 | // Select the 'Terminal' nodes. |
3465 | parenthesesBeginOpCode = OpParenthesesSubpatternTerminalBegin; |
3466 | parenthesesEndOpCode = OpParenthesesSubpatternTerminalEnd; |
3467 | } else { |
3468 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3469 | // We only handle generic parenthesis with non-fixed counts. |
3470 | if (term->quantityType == QuantifierFixedCount) { |
3471 | // This subpattern is not supported by the JIT. |
3472 | m_failureReason = JITFailureReason::FixedCountParenthesizedSubpattern; |
3473 | return; |
3474 | } |
3475 | |
3476 | m_containsNestedSubpatterns = true; |
3477 | |
3478 | // Select the 'Generic' nodes. |
3479 | parenthesesBeginOpCode = OpParenthesesSubpatternBegin; |
3480 | parenthesesEndOpCode = OpParenthesesSubpatternEnd; |
3481 | |
3482 | // If there is more than one alternative we cannot use the 'simple' nodes. |
3483 | if (term->parentheses.disjunction->m_alternatives.size() != 1) { |
3484 | alternativeBeginOpCode = OpNestedAlternativeBegin; |
3485 | alternativeNextOpCode = OpNestedAlternativeNext; |
3486 | alternativeEndOpCode = OpNestedAlternativeEnd; |
3487 | } |
3488 | #else |
3489 | // This subpattern is not supported by the JIT. |
3490 | m_failureReason = JITFailureReason::ParenthesizedSubpattern; |
3491 | return; |
3492 | #endif |
3493 | } |
3494 | |
3495 | size_t parenBegin = m_ops.size(); |
3496 | m_ops.append(parenthesesBeginOpCode); |
3497 | |
3498 | m_ops.append(alternativeBeginOpCode); |
3499 | m_ops.last().m_previousOp = notFound; |
3500 | m_ops.last().m_term = term; |
3501 | Vector<std::unique_ptr<PatternAlternative>>& alternatives = term->parentheses.disjunction->m_alternatives; |
3502 | for (unsigned i = 0; i < alternatives.size(); ++i) { |
3503 | size_t lastOpIndex = m_ops.size() - 1; |
3504 | |
3505 | PatternAlternative* nestedAlternative = alternatives[i].get(); |
3506 | opCompileAlternative(nestedAlternative); |
3507 | |
3508 | size_t thisOpIndex = m_ops.size(); |
3509 | m_ops.append(YarrOp(alternativeNextOpCode)); |
3510 | |
3511 | YarrOp& lastOp = m_ops[lastOpIndex]; |
3512 | YarrOp& thisOp = m_ops[thisOpIndex]; |
3513 | |
3514 | lastOp.m_alternative = nestedAlternative; |
3515 | lastOp.m_nextOp = thisOpIndex; |
3516 | thisOp.m_previousOp = lastOpIndex; |
3517 | thisOp.m_term = term; |
3518 | } |
3519 | YarrOp& lastOp = m_ops.last(); |
3520 | ASSERT(lastOp.m_op == alternativeNextOpCode); |
3521 | lastOp.m_op = alternativeEndOpCode; |
3522 | lastOp.m_alternative = 0; |
3523 | lastOp.m_nextOp = notFound; |
3524 | |
3525 | size_t parenEnd = m_ops.size(); |
3526 | m_ops.append(parenthesesEndOpCode); |
3527 | |
3528 | m_ops[parenBegin].m_term = term; |
3529 | m_ops[parenBegin].m_previousOp = notFound; |
3530 | m_ops[parenBegin].m_nextOp = parenEnd; |
3531 | m_ops[parenEnd].m_term = term; |
3532 | m_ops[parenEnd].m_previousOp = parenBegin; |
3533 | m_ops[parenEnd].m_nextOp = notFound; |
3534 | } |
3535 | |
3536 | // opCompileParentheticalAssertion |
3537 | // Emits ops for a parenthetical assertion. These consist of an |
3538 | // OpSimpleNestedAlternativeBegin/Next/End set of nodes wrapping |
3539 | // the alternatives, with these wrapped by an outer pair of |
3540 | // OpParentheticalAssertionBegin/End nodes. |
3541 | // We can always use the OpSimpleNestedAlternative nodes in the |
3542 | // case of parenthetical assertions since these only ever match |
3543 | // once, and will never backtrack back into the assertion. |
3544 | void opCompileParentheticalAssertion(PatternTerm* term) |
3545 | { |
3546 | if (UNLIKELY(!m_vm->isSafeToRecurse())) { |
3547 | m_failureReason = JITFailureReason::ParenthesisNestedTooDeep; |
3548 | return; |
3549 | } |
3550 | |
3551 | size_t parenBegin = m_ops.size(); |
3552 | m_ops.append(OpParentheticalAssertionBegin); |
3553 | |
3554 | m_ops.append(OpSimpleNestedAlternativeBegin); |
3555 | m_ops.last().m_previousOp = notFound; |
3556 | m_ops.last().m_term = term; |
3557 | Vector<std::unique_ptr<PatternAlternative>>& alternatives = term->parentheses.disjunction->m_alternatives; |
3558 | for (unsigned i = 0; i < alternatives.size(); ++i) { |
3559 | size_t lastOpIndex = m_ops.size() - 1; |
3560 | |
3561 | PatternAlternative* nestedAlternative = alternatives[i].get(); |
3562 | opCompileAlternative(nestedAlternative); |
3563 | |
3564 | size_t thisOpIndex = m_ops.size(); |
3565 | m_ops.append(YarrOp(OpSimpleNestedAlternativeNext)); |
3566 | |
3567 | YarrOp& lastOp = m_ops[lastOpIndex]; |
3568 | YarrOp& thisOp = m_ops[thisOpIndex]; |
3569 | |
3570 | lastOp.m_alternative = nestedAlternative; |
3571 | lastOp.m_nextOp = thisOpIndex; |
3572 | thisOp.m_previousOp = lastOpIndex; |
3573 | thisOp.m_term = term; |
3574 | } |
3575 | YarrOp& lastOp = m_ops.last(); |
3576 | ASSERT(lastOp.m_op == OpSimpleNestedAlternativeNext); |
3577 | lastOp.m_op = OpSimpleNestedAlternativeEnd; |
3578 | lastOp.m_alternative = 0; |
3579 | lastOp.m_nextOp = notFound; |
3580 | |
3581 | size_t parenEnd = m_ops.size(); |
3582 | m_ops.append(OpParentheticalAssertionEnd); |
3583 | |
3584 | m_ops[parenBegin].m_term = term; |
3585 | m_ops[parenBegin].m_previousOp = notFound; |
3586 | m_ops[parenBegin].m_nextOp = parenEnd; |
3587 | m_ops[parenEnd].m_term = term; |
3588 | m_ops[parenEnd].m_previousOp = parenBegin; |
3589 | m_ops[parenEnd].m_nextOp = notFound; |
3590 | } |
3591 | |
3592 | // opCompileAlternative |
3593 | // Called to emit nodes for all terms in an alternative. |
3594 | void opCompileAlternative(PatternAlternative* alternative) |
3595 | { |
3596 | optimizeAlternative(alternative); |
3597 | |
3598 | for (unsigned i = 0; i < alternative->m_terms.size(); ++i) { |
3599 | PatternTerm* term = &alternative->m_terms[i]; |
3600 | |
3601 | switch (term->type) { |
3602 | case PatternTerm::TypeParenthesesSubpattern: |
3603 | opCompileParenthesesSubpattern(term); |
3604 | break; |
3605 | |
3606 | case PatternTerm::TypeParentheticalAssertion: |
3607 | opCompileParentheticalAssertion(term); |
3608 | break; |
3609 | |
3610 | default: |
3611 | m_ops.append(term); |
3612 | } |
3613 | } |
3614 | } |
3615 | |
3616 | // opCompileBody |
3617 | // This method compiles the body disjunction of the regular expression. |
3618 | // The body consists of two sets of alternatives - zero or more 'once |
3619 | // through' (BOL anchored) alternatives, followed by zero or more |
3620 | // repeated alternatives. |
3621 | // For each of these two sets of alteratives, if not empty they will be |
3622 | // wrapped in a set of OpBodyAlternativeBegin/Next/End nodes (with the |
3623 | // 'begin' node referencing the first alternative, and 'next' nodes |
3624 | // referencing any further alternatives. The begin/next/end nodes are |
3625 | // linked together in a doubly linked list. In the case of repeating |
3626 | // alternatives, the end node is also linked back to the beginning. |
3627 | // If no repeating alternatives exist, then a OpMatchFailed node exists |
3628 | // to return the failing result. |
3629 | void opCompileBody(PatternDisjunction* disjunction) |
3630 | { |
3631 | if (UNLIKELY(!m_vm->isSafeToRecurse())) { |
3632 | m_failureReason = JITFailureReason::ParenthesisNestedTooDeep; |
3633 | return; |
3634 | } |
3635 | |
3636 | Vector<std::unique_ptr<PatternAlternative>>& alternatives = disjunction->m_alternatives; |
3637 | size_t currentAlternativeIndex = 0; |
3638 | |
3639 | // Emit the 'once through' alternatives. |
3640 | if (alternatives.size() && alternatives[0]->onceThrough()) { |
3641 | m_ops.append(YarrOp(OpBodyAlternativeBegin)); |
3642 | m_ops.last().m_previousOp = notFound; |
3643 | |
3644 | do { |
3645 | size_t lastOpIndex = m_ops.size() - 1; |
3646 | PatternAlternative* alternative = alternatives[currentAlternativeIndex].get(); |
3647 | opCompileAlternative(alternative); |
3648 | |
3649 | size_t thisOpIndex = m_ops.size(); |
3650 | m_ops.append(YarrOp(OpBodyAlternativeNext)); |
3651 | |
3652 | YarrOp& lastOp = m_ops[lastOpIndex]; |
3653 | YarrOp& thisOp = m_ops[thisOpIndex]; |
3654 | |
3655 | lastOp.m_alternative = alternative; |
3656 | lastOp.m_nextOp = thisOpIndex; |
3657 | thisOp.m_previousOp = lastOpIndex; |
3658 | |
3659 | ++currentAlternativeIndex; |
3660 | } while (currentAlternativeIndex < alternatives.size() && alternatives[currentAlternativeIndex]->onceThrough()); |
3661 | |
3662 | YarrOp& lastOp = m_ops.last(); |
3663 | |
3664 | ASSERT(lastOp.m_op == OpBodyAlternativeNext); |
3665 | lastOp.m_op = OpBodyAlternativeEnd; |
3666 | lastOp.m_alternative = 0; |
3667 | lastOp.m_nextOp = notFound; |
3668 | } |
3669 | |
3670 | if (currentAlternativeIndex == alternatives.size()) { |
3671 | m_ops.append(YarrOp(OpMatchFailed)); |
3672 | return; |
3673 | } |
3674 | |
3675 | // Emit the repeated alternatives. |
3676 | size_t repeatLoop = m_ops.size(); |
3677 | m_ops.append(YarrOp(OpBodyAlternativeBegin)); |
3678 | m_ops.last().m_previousOp = notFound; |
3679 | do { |
3680 | size_t lastOpIndex = m_ops.size() - 1; |
3681 | PatternAlternative* alternative = alternatives[currentAlternativeIndex].get(); |
3682 | ASSERT(!alternative->onceThrough()); |
3683 | opCompileAlternative(alternative); |
3684 | |
3685 | size_t thisOpIndex = m_ops.size(); |
3686 | m_ops.append(YarrOp(OpBodyAlternativeNext)); |
3687 | |
3688 | YarrOp& lastOp = m_ops[lastOpIndex]; |
3689 | YarrOp& thisOp = m_ops[thisOpIndex]; |
3690 | |
3691 | lastOp.m_alternative = alternative; |
3692 | lastOp.m_nextOp = thisOpIndex; |
3693 | thisOp.m_previousOp = lastOpIndex; |
3694 | |
3695 | ++currentAlternativeIndex; |
3696 | } while (currentAlternativeIndex < alternatives.size()); |
3697 | YarrOp& lastOp = m_ops.last(); |
3698 | ASSERT(lastOp.m_op == OpBodyAlternativeNext); |
3699 | lastOp.m_op = OpBodyAlternativeEnd; |
3700 | lastOp.m_alternative = 0; |
3701 | lastOp.m_nextOp = repeatLoop; |
3702 | } |
3703 | |
3704 | void generateTryReadUnicodeCharacterHelper() |
3705 | { |
3706 | #ifdef JIT_UNICODE_EXPRESSIONS |
3707 | if (m_tryReadUnicodeCharacterCalls.isEmpty()) |
3708 | return; |
3709 | |
3710 | ASSERT(m_decodeSurrogatePairs); |
3711 | |
3712 | m_tryReadUnicodeCharacterEntry = label(); |
3713 | |
3714 | tagReturnAddress(); |
3715 | |
3716 | tryReadUnicodeCharImpl(regT0); |
3717 | |
3718 | ret(); |
3719 | #endif |
3720 | } |
3721 | |
3722 | void generateEnter() |
3723 | { |
3724 | #if CPU(X86_64) |
3725 | push(X86Registers::ebp); |
3726 | move(stackPointerRegister, X86Registers::ebp); |
3727 | |
3728 | if (m_pattern.m_saveInitialStartValue) |
3729 | push(X86Registers::ebx); |
3730 | |
3731 | #if OS(WINDOWS) |
3732 | push(X86Registers::edi); |
3733 | #endif |
3734 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3735 | if (m_containsNestedSubpatterns) { |
3736 | #if OS(WINDOWS) |
3737 | push(X86Registers::esi); |
3738 | #endif |
3739 | push(X86Registers::r12); |
3740 | } |
3741 | #endif |
3742 | |
3743 | if (m_decodeSurrogatePairs) { |
3744 | push(X86Registers::r13); |
3745 | push(X86Registers::r14); |
3746 | push(X86Registers::r15); |
3747 | |
3748 | move(TrustedImm32(0xd800), leadingSurrogateTag); |
3749 | } |
3750 | // The ABI doesn't guarantee the upper bits are zero on unsigned arguments, so clear them ourselves. |
3751 | zeroExtend32ToPtr(index, index); |
3752 | zeroExtend32ToPtr(length, length); |
3753 | #if OS(WINDOWS) |
3754 | if (compileMode == IncludeSubpatterns) |
3755 | loadPtr(Address(X86Registers::ebp, 6 * sizeof(void*)), output); |
3756 | // rcx is the pointer to the allocated space for result in x64 Windows. |
3757 | push(X86Registers::ecx); |
3758 | #endif |
3759 | #elif CPU(X86) |
3760 | push(X86Registers::ebp); |
3761 | move(stackPointerRegister, X86Registers::ebp); |
3762 | // TODO: do we need spill registers to fill the output pointer if there are no sub captures? |
3763 | push(X86Registers::ebx); |
3764 | push(X86Registers::edi); |
3765 | push(X86Registers::esi); |
3766 | // load output into edi (2 = saved ebp + return address). |
3767 | #if COMPILER(MSVC) |
3768 | loadPtr(Address(X86Registers::ebp, 2 * sizeof(void*)), input); |
3769 | loadPtr(Address(X86Registers::ebp, 3 * sizeof(void*)), index); |
3770 | loadPtr(Address(X86Registers::ebp, 4 * sizeof(void*)), length); |
3771 | if (compileMode == IncludeSubpatterns) |
3772 | loadPtr(Address(X86Registers::ebp, 5 * sizeof(void*)), output); |
3773 | #else |
3774 | if (compileMode == IncludeSubpatterns) |
3775 | loadPtr(Address(X86Registers::ebp, 2 * sizeof(void*)), output); |
3776 | #endif |
3777 | #elif CPU(ARM64) |
3778 | tagReturnAddress(); |
3779 | if (m_decodeSurrogatePairs) { |
3780 | pushPair(framePointerRegister, linkRegister); |
3781 | move(TrustedImm32(0x10000), supplementaryPlanesBase); |
3782 | move(TrustedImm32(0xd800), leadingSurrogateTag); |
3783 | move(TrustedImm32(0xdc00), trailingSurrogateTag); |
3784 | } |
3785 | |
3786 | // The ABI doesn't guarantee the upper bits are zero on unsigned arguments, so clear them ourselves. |
3787 | zeroExtend32ToPtr(index, index); |
3788 | zeroExtend32ToPtr(length, length); |
3789 | #elif CPU(ARM_THUMB2) |
3790 | push(ARMRegisters::r4); |
3791 | push(ARMRegisters::r5); |
3792 | push(ARMRegisters::r6); |
3793 | push(ARMRegisters::r8); |
3794 | #elif CPU(MIPS) |
3795 | // Do nothing. |
3796 | #endif |
3797 | |
3798 | store8(TrustedImm32(1), &m_vm->isExecutingInRegExpJIT); |
3799 | } |
3800 | |
3801 | void generateReturn() |
3802 | { |
3803 | store8(TrustedImm32(0), &m_vm->isExecutingInRegExpJIT); |
3804 | |
3805 | #if CPU(X86_64) |
3806 | #if OS(WINDOWS) |
3807 | // Store the return value in the allocated space pointed by rcx. |
3808 | pop(X86Registers::ecx); |
3809 | store64(returnRegister, Address(X86Registers::ecx)); |
3810 | store64(returnRegister2, Address(X86Registers::ecx, sizeof(void*))); |
3811 | move(X86Registers::ecx, returnRegister); |
3812 | #endif |
3813 | if (m_decodeSurrogatePairs) { |
3814 | pop(X86Registers::r15); |
3815 | pop(X86Registers::r14); |
3816 | pop(X86Registers::r13); |
3817 | } |
3818 | |
3819 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3820 | if (m_containsNestedSubpatterns) { |
3821 | pop(X86Registers::r12); |
3822 | #if OS(WINDOWS) |
3823 | pop(X86Registers::esi); |
3824 | #endif |
3825 | } |
3826 | #endif |
3827 | #if OS(WINDOWS) |
3828 | pop(X86Registers::edi); |
3829 | #endif |
3830 | |
3831 | if (m_pattern.m_saveInitialStartValue) |
3832 | pop(X86Registers::ebx); |
3833 | pop(X86Registers::ebp); |
3834 | #elif CPU(X86) |
3835 | pop(X86Registers::esi); |
3836 | pop(X86Registers::edi); |
3837 | pop(X86Registers::ebx); |
3838 | pop(X86Registers::ebp); |
3839 | #elif CPU(ARM64) |
3840 | if (m_decodeSurrogatePairs) |
3841 | popPair(framePointerRegister, linkRegister); |
3842 | #elif CPU(ARM_THUMB2) |
3843 | pop(ARMRegisters::r8); |
3844 | pop(ARMRegisters::r6); |
3845 | pop(ARMRegisters::r5); |
3846 | pop(ARMRegisters::r4); |
3847 | #elif CPU(MIPS) |
3848 | // Do nothing |
3849 | #endif |
3850 | ret(); |
3851 | } |
3852 | |
3853 | public: |
3854 | YarrGenerator(VM* vm, YarrPattern& pattern, String& patternString, YarrCodeBlock& codeBlock, YarrCharSize charSize) |
3855 | : m_vm(vm) |
3856 | , m_pattern(pattern) |
3857 | , m_patternString(patternString) |
3858 | , m_codeBlock(codeBlock) |
3859 | , m_charSize(charSize) |
3860 | , m_decodeSurrogatePairs(m_charSize == Char16 && m_pattern.unicode()) |
3861 | , m_unicodeIgnoreCase(m_pattern.unicode() && m_pattern.ignoreCase()) |
3862 | , m_fixedSizedAlternative(false) |
3863 | , m_canonicalMode(m_pattern.unicode() ? CanonicalMode::Unicode : CanonicalMode::UCS2) |
3864 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3865 | , m_containsNestedSubpatterns(false) |
3866 | , m_parenContextSizes(compileMode == IncludeSubpatterns ? m_pattern.m_numSubpatterns : 0, m_pattern.m_body->m_callFrameSize) |
3867 | #endif |
3868 | { |
3869 | } |
3870 | |
3871 | void compile() |
3872 | { |
3873 | YarrCodeBlock& codeBlock = m_codeBlock; |
3874 | |
3875 | #ifndef JIT_UNICODE_EXPRESSIONS |
3876 | if (m_decodeSurrogatePairs) { |
3877 | codeBlock.setFallBackWithFailureReason(JITFailureReason::DecodeSurrogatePair); |
3878 | return; |
3879 | } |
3880 | #endif |
3881 | |
3882 | if (m_pattern.m_containsBackreferences |
3883 | #if ENABLE(YARR_JIT_BACKREFERENCES) |
3884 | && (compileMode == MatchOnly || (m_pattern.ignoreCase() && m_charSize != Char8)) |
3885 | #endif |
3886 | ) { |
3887 | codeBlock.setFallBackWithFailureReason(JITFailureReason::BackReference); |
3888 | return; |
3889 | } |
3890 | |
3891 | // We need to compile before generating code since we set flags based on compilation that |
3892 | // are used during generation. |
3893 | opCompileBody(m_pattern.m_body); |
3894 | |
3895 | if (m_failureReason) { |
3896 | codeBlock.setFallBackWithFailureReason(*m_failureReason); |
3897 | return; |
3898 | } |
3899 | |
3900 | if (UNLIKELY(Options::dumpDisassembly() || Options::dumpRegExpDisassembly())) |
3901 | m_disassembler = std::make_unique<YarrDisassembler>(this); |
3902 | |
3903 | if (m_disassembler) |
3904 | m_disassembler->setStartOfCode(label()); |
3905 | |
3906 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3907 | if (m_containsNestedSubpatterns) |
3908 | codeBlock.setUsesPatternContextBuffer(); |
3909 | #endif |
3910 | |
3911 | generateEnter(); |
3912 | |
3913 | Jump hasInput = checkInput(); |
3914 | generateFailReturn(); |
3915 | hasInput.link(this); |
3916 | |
3917 | #ifdef JIT_UNICODE_EXPRESSIONS |
3918 | if (m_decodeSurrogatePairs) |
3919 | getEffectiveAddress(BaseIndex(input, length, TimesTwo), endOfStringAddress); |
3920 | #endif |
3921 | |
3922 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3923 | if (m_containsNestedSubpatterns) |
3924 | move(TrustedImm32(matchLimit), remainingMatchCount); |
3925 | #endif |
3926 | |
3927 | if (compileMode == IncludeSubpatterns) { |
3928 | for (unsigned i = 0; i < m_pattern.m_numSubpatterns + 1; ++i) |
3929 | store32(TrustedImm32(-1), Address(output, (i << 1) * sizeof(int))); |
3930 | } |
3931 | |
3932 | if (!m_pattern.m_body->m_hasFixedSize) |
3933 | setMatchStart(index); |
3934 | |
3935 | initCallFrame(); |
3936 | |
3937 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
3938 | if (m_containsNestedSubpatterns) |
3939 | initParenContextFreeList(); |
3940 | #endif |
3941 | |
3942 | if (m_pattern.m_saveInitialStartValue) { |
3943 | #ifdef HAVE_INITIAL_START_REG |
3944 | move(index, initialStart); |
3945 | #else |
3946 | storeToFrame(index, m_pattern.m_initialStartValueFrameLocation); |
3947 | #endif |
3948 | } |
3949 | |
3950 | generate(); |
3951 | if (m_disassembler) |
3952 | m_disassembler->setEndOfGenerate(label()); |
3953 | backtrack(); |
3954 | if (m_disassembler) |
3955 | m_disassembler->setEndOfBacktrack(label()); |
3956 | |
3957 | generateTryReadUnicodeCharacterHelper(); |
3958 | |
3959 | generateJITFailReturn(); |
3960 | |
3961 | if (m_disassembler) |
3962 | m_disassembler->setEndOfCode(label()); |
3963 | |
3964 | LinkBuffer linkBuffer(*this, REGEXP_CODE_ID, JITCompilationCanFail); |
3965 | if (linkBuffer.didFailToAllocate()) { |
3966 | codeBlock.setFallBackWithFailureReason(JITFailureReason::ExecutableMemoryAllocationFailure); |
3967 | return; |
3968 | } |
3969 | |
3970 | if (!m_tryReadUnicodeCharacterCalls.isEmpty()) { |
3971 | CodeLocationLabel<NoPtrTag> tryReadUnicodeCharacterHelper = linkBuffer.locationOf<NoPtrTag>(m_tryReadUnicodeCharacterEntry); |
3972 | |
3973 | for (auto call : m_tryReadUnicodeCharacterCalls) |
3974 | linkBuffer.link(call, tryReadUnicodeCharacterHelper); |
3975 | } |
3976 | |
3977 | m_backtrackingState.linkDataLabels(linkBuffer); |
3978 | |
3979 | if (m_disassembler) |
3980 | m_disassembler->dump(linkBuffer); |
3981 | |
3982 | if (compileMode == MatchOnly) { |
3983 | if (m_charSize == Char8) |
3984 | codeBlock.set8BitCodeMatchOnly(FINALIZE_REGEXP_CODE(linkBuffer, YarrMatchOnly8BitPtrTag, "Match-only 8-bit regular expression" )); |
3985 | else |
3986 | codeBlock.set16BitCodeMatchOnly(FINALIZE_REGEXP_CODE(linkBuffer, YarrMatchOnly16BitPtrTag, "Match-only 16-bit regular expression" )); |
3987 | } else { |
3988 | if (m_charSize == Char8) |
3989 | codeBlock.set8BitCode(FINALIZE_REGEXP_CODE(linkBuffer, Yarr8BitPtrTag, "8-bit regular expression" )); |
3990 | else |
3991 | codeBlock.set16BitCode(FINALIZE_REGEXP_CODE(linkBuffer, Yarr16BitPtrTag, "16-bit regular expression" )); |
3992 | } |
3993 | if (m_failureReason) |
3994 | codeBlock.setFallBackWithFailureReason(*m_failureReason); |
3995 | } |
3996 | |
3997 | const char* variant() override |
3998 | { |
3999 | if (compileMode == MatchOnly) { |
4000 | if (m_charSize == Char8) |
4001 | return "Match-only 8-bit regular expression" ; |
4002 | |
4003 | return "Match-only 16-bit regular expression" ; |
4004 | } |
4005 | |
4006 | if (m_charSize == Char8) |
4007 | return "8-bit regular expression" ; |
4008 | |
4009 | return "16-bit regular expression" ; |
4010 | } |
4011 | |
4012 | unsigned opCount() override |
4013 | { |
4014 | return m_ops.size(); |
4015 | } |
4016 | |
4017 | void dumpPatternString(PrintStream& out) override |
4018 | { |
4019 | m_pattern.dumpPatternString(out, m_patternString); |
4020 | } |
4021 | |
4022 | int dumpFor(PrintStream& out, unsigned opIndex) override |
4023 | { |
4024 | if (opIndex >= opCount()) |
4025 | return 0; |
4026 | |
4027 | out.printf("%4d:" , opIndex); |
4028 | |
4029 | YarrOp& op = m_ops[opIndex]; |
4030 | PatternTerm* term = op.m_term; |
4031 | switch (op.m_op) { |
4032 | case OpTerm: { |
4033 | out.print("OpTerm " ); |
4034 | switch (term->type) { |
4035 | case PatternTerm::TypeAssertionBOL: |
4036 | out.print("Assert BOL" ); |
4037 | break; |
4038 | |
4039 | case PatternTerm::TypeAssertionEOL: |
4040 | out.print("Assert EOL" ); |
4041 | break; |
4042 | |
4043 | case PatternTerm::TypeBackReference: |
4044 | out.printf("BackReference pattern #%u" , term->backReferenceSubpatternId); |
4045 | term->dumpQuantifier(out); |
4046 | break; |
4047 | |
4048 | case PatternTerm::TypePatternCharacter: |
4049 | out.print("TypePatternCharacter " ); |
4050 | dumpUChar32(out, term->patternCharacter); |
4051 | if (m_pattern.ignoreCase()) |
4052 | out.print(" ignore case" ); |
4053 | |
4054 | term->dumpQuantifier(out); |
4055 | break; |
4056 | |
4057 | case PatternTerm::TypeCharacterClass: |
4058 | out.print("TypePatternCharacterClass " ); |
4059 | if (term->invert()) |
4060 | out.print("not " ); |
4061 | dumpCharacterClass(out, &m_pattern, term->characterClass); |
4062 | term->dumpQuantifier(out); |
4063 | break; |
4064 | |
4065 | case PatternTerm::TypeAssertionWordBoundary: |
4066 | out.printf("%sword boundary" , term->invert() ? "non-" : "" ); |
4067 | break; |
4068 | |
4069 | case PatternTerm::TypeDotStarEnclosure: |
4070 | out.print(".* enclosure" ); |
4071 | break; |
4072 | |
4073 | case PatternTerm::TypeForwardReference: |
4074 | out.print("TypeForwardReference <not handled>" ); |
4075 | break; |
4076 | |
4077 | case PatternTerm::TypeParenthesesSubpattern: |
4078 | case PatternTerm::TypeParentheticalAssertion: |
4079 | RELEASE_ASSERT_NOT_REACHED(); |
4080 | break; |
4081 | } |
4082 | |
4083 | if (op.m_isDeadCode) |
4084 | out.print(" already handled" ); |
4085 | out.print("\n" ); |
4086 | return(0); |
4087 | } |
4088 | |
4089 | case OpBodyAlternativeBegin: |
4090 | out.printf("OpBodyAlternativeBegin minimum size %u\n" , op.m_alternative->m_minimumSize); |
4091 | return(0); |
4092 | |
4093 | case OpBodyAlternativeNext: |
4094 | out.printf("OpBodyAlternativeNext minimum size %u\n" , op.m_alternative->m_minimumSize); |
4095 | return(0); |
4096 | |
4097 | case OpBodyAlternativeEnd: |
4098 | out.print("OpBodyAlternativeEnd\n" ); |
4099 | return(0); |
4100 | |
4101 | case OpSimpleNestedAlternativeBegin: |
4102 | out.printf("OpSimpleNestedAlternativeBegin minimum size %u\n" , op.m_alternative->m_minimumSize); |
4103 | return(1); |
4104 | |
4105 | case OpNestedAlternativeBegin: |
4106 | out.printf("OpNestedAlternativeBegin minimum size %u\n" , op.m_alternative->m_minimumSize); |
4107 | return(1); |
4108 | |
4109 | case OpSimpleNestedAlternativeNext: |
4110 | out.printf("OpSimpleNestedAlternativeNext minimum size %u\n" , op.m_alternative->m_minimumSize); |
4111 | return(0); |
4112 | |
4113 | case OpNestedAlternativeNext: |
4114 | out.printf("OpNestedAlternativeNext minimum size %u\n" , op.m_alternative->m_minimumSize); |
4115 | return(0); |
4116 | |
4117 | case OpSimpleNestedAlternativeEnd: |
4118 | out.print("OpSimpleNestedAlternativeEnd" ); |
4119 | term->dumpQuantifier(out); |
4120 | out.print("\n" ); |
4121 | return(-1); |
4122 | |
4123 | case OpNestedAlternativeEnd: |
4124 | out.print("OpNestedAlternativeEnd" ); |
4125 | term->dumpQuantifier(out); |
4126 | out.print("\n" ); |
4127 | return(-1); |
4128 | |
4129 | case OpParenthesesSubpatternOnceBegin: |
4130 | out.print("OpParenthesesSubpatternOnceBegin " ); |
4131 | if (term->capture()) |
4132 | out.printf("capturing pattern #%u" , term->parentheses.subpatternId); |
4133 | else |
4134 | out.print("non-capturing" ); |
4135 | term->dumpQuantifier(out); |
4136 | out.print("\n" ); |
4137 | return(0); |
4138 | |
4139 | case OpParenthesesSubpatternOnceEnd: |
4140 | out.print("OpParenthesesSubpatternOnceEnd " ); |
4141 | if (term->capture()) |
4142 | out.printf("capturing pattern #%u" , term->parentheses.subpatternId); |
4143 | else |
4144 | out.print("non-capturing" ); |
4145 | term->dumpQuantifier(out); |
4146 | out.print("\n" ); |
4147 | return(0); |
4148 | |
4149 | case OpParenthesesSubpatternTerminalBegin: |
4150 | out.print("OpParenthesesSubpatternTerminalBegin " ); |
4151 | if (term->capture()) |
4152 | out.printf("capturing pattern #%u\n" , term->parentheses.subpatternId); |
4153 | else |
4154 | out.print("non-capturing\n" ); |
4155 | return(0); |
4156 | |
4157 | case OpParenthesesSubpatternTerminalEnd: |
4158 | out.print("OpParenthesesSubpatternTerminalEnd " ); |
4159 | if (term->capture()) |
4160 | out.printf("capturing pattern #%u\n" , term->parentheses.subpatternId); |
4161 | else |
4162 | out.print("non-capturing\n" ); |
4163 | return(0); |
4164 | |
4165 | case OpParenthesesSubpatternBegin: |
4166 | out.print("OpParenthesesSubpatternBegin " ); |
4167 | if (term->capture()) |
4168 | out.printf("capturing pattern #%u" , term->parentheses.subpatternId); |
4169 | else |
4170 | out.print("non-capturing" ); |
4171 | term->dumpQuantifier(out); |
4172 | out.print("\n" ); |
4173 | return(0); |
4174 | |
4175 | case OpParenthesesSubpatternEnd: |
4176 | out.print("OpParenthesesSubpatternEnd " ); |
4177 | if (term->capture()) |
4178 | out.printf("capturing pattern #%u" , term->parentheses.subpatternId); |
4179 | else |
4180 | out.print("non-capturing" ); |
4181 | term->dumpQuantifier(out); |
4182 | out.print("\n" ); |
4183 | return(0); |
4184 | |
4185 | case OpParentheticalAssertionBegin: |
4186 | out.printf("OpParentheticalAssertionBegin%s\n" , term->invert() ? " inverted" : "" ); |
4187 | return(0); |
4188 | |
4189 | case OpParentheticalAssertionEnd: |
4190 | out.printf("OpParentheticalAssertionEnd%s\n" , term->invert() ? " inverted" : "" ); |
4191 | return(0); |
4192 | |
4193 | case OpMatchFailed: |
4194 | out.print("OpMatchFailed\n" ); |
4195 | return(0); |
4196 | } |
4197 | |
4198 | return(0); |
4199 | } |
4200 | |
4201 | private: |
4202 | VM* m_vm; |
4203 | |
4204 | YarrPattern& m_pattern; |
4205 | String& m_patternString; |
4206 | |
4207 | YarrCodeBlock& m_codeBlock; |
4208 | YarrCharSize m_charSize; |
4209 | |
4210 | // Used to detect regular expression constructs that are not currently |
4211 | // supported in the JIT; fall back to the interpreter when this is detected. |
4212 | Optional<JITFailureReason> m_failureReason; |
4213 | |
4214 | bool m_decodeSurrogatePairs; |
4215 | bool m_unicodeIgnoreCase; |
4216 | bool m_fixedSizedAlternative; |
4217 | CanonicalMode m_canonicalMode; |
4218 | #if ENABLE(YARR_JIT_ALL_PARENS_EXPRESSIONS) |
4219 | bool m_containsNestedSubpatterns; |
4220 | ParenContextSizes m_parenContextSizes; |
4221 | #endif |
4222 | JumpList m_abortExecution; |
4223 | JumpList m_hitMatchLimit; |
4224 | Vector<Call> m_tryReadUnicodeCharacterCalls; |
4225 | Label m_tryReadUnicodeCharacterEntry; |
4226 | |
4227 | // The regular expression expressed as a linear sequence of operations. |
4228 | Vector<YarrOp, 128> m_ops; |
4229 | |
4230 | // This records the current input offset being applied due to the current |
4231 | // set of alternatives we are nested within. E.g. when matching the |
4232 | // character 'b' within the regular expression /abc/, we will know that |
4233 | // the minimum size for the alternative is 3, checked upon entry to the |
4234 | // alternative, and that 'b' is at offset 1 from the start, and as such |
4235 | // when matching 'b' we need to apply an offset of -2 to the load. |
4236 | // |
4237 | // FIXME: This should go away. Rather than tracking this value throughout |
4238 | // code generation, we should gather this information up front & store it |
4239 | // on the YarrOp structure. |
4240 | Checked<unsigned> m_checkedOffset; |
4241 | |
4242 | // This class records state whilst generating the backtracking path of code. |
4243 | BacktrackingState m_backtrackingState; |
4244 | |
4245 | std::unique_ptr<YarrDisassembler> m_disassembler; |
4246 | }; |
4247 | |
4248 | static void dumpCompileFailure(JITFailureReason failure) |
4249 | { |
4250 | switch (failure) { |
4251 | case JITFailureReason::DecodeSurrogatePair: |
4252 | dataLog("Can't JIT a pattern decoding surrogate pairs\n" ); |
4253 | break; |
4254 | case JITFailureReason::BackReference: |
4255 | dataLog("Can't JIT some patterns containing back references\n" ); |
4256 | break; |
4257 | case JITFailureReason::ForwardReference: |
4258 | dataLog("Can't JIT a pattern containing forward references\n" ); |
4259 | break; |
4260 | case JITFailureReason::VariableCountedParenthesisWithNonZeroMinimum: |
4261 | dataLog("Can't JIT a pattern containing a variable counted parenthesis with a non-zero minimum\n" ); |
4262 | break; |
4263 | case JITFailureReason::ParenthesizedSubpattern: |
4264 | dataLog("Can't JIT a pattern containing parenthesized subpatterns\n" ); |
4265 | break; |
4266 | case JITFailureReason::FixedCountParenthesizedSubpattern: |
4267 | dataLog("Can't JIT a pattern containing fixed count parenthesized subpatterns\n" ); |
4268 | break; |
4269 | case JITFailureReason::ParenthesisNestedTooDeep: |
4270 | dataLog("Can't JIT pattern due to parentheses nested too deeply\n" ); |
4271 | break; |
4272 | case JITFailureReason::ExecutableMemoryAllocationFailure: |
4273 | dataLog("Can't JIT because of failure of allocation of executable memory\n" ); |
4274 | break; |
4275 | } |
4276 | } |
4277 | |
4278 | void jitCompile(YarrPattern& pattern, String& patternString, YarrCharSize charSize, VM* vm, YarrCodeBlock& codeBlock, YarrJITCompileMode mode) |
4279 | { |
4280 | if (mode == MatchOnly) |
4281 | YarrGenerator<MatchOnly>(vm, pattern, patternString, codeBlock, charSize).compile(); |
4282 | else |
4283 | YarrGenerator<IncludeSubpatterns>(vm, pattern, patternString, codeBlock, charSize).compile(); |
4284 | |
4285 | if (auto failureReason = codeBlock.failureReason()) { |
4286 | if (Options::dumpCompiledRegExpPatterns()) { |
4287 | pattern.dumpPatternString(WTF::dataFile(), patternString); |
4288 | dataLog(" : " ); |
4289 | dumpCompileFailure(*failureReason); |
4290 | } |
4291 | } |
4292 | } |
4293 | |
4294 | }} |
4295 | |
4296 | #endif |
4297 | |