And maybe it's worth considering if we want to enable this for some run mode of run-javascriptcore-tests to ensure it doesn't break.
Created attachment 260808 [details] patch
Comment on attachment 260808 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260808&action=review > Tools/Scripts/run-jsc-stress-tests:791 > + runDFGFlushEverything Should we have this as part of the default run or no?
Comment on attachment 260808 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260808&action=review > Source/JavaScriptCore/JavaScriptCore.xcodeproj/project.pbxproj:1037 > + 79B531EA1B9CEE180039CC8E /* DFGFlushEverythingInsertionPhase.cpp in Sources */ = {isa = PBXBuildFile; fileRef = 79B531E81B9CEE180039CC8E /* DFGFlushEverythingInsertionPhase.cpp */; }; > + 79B531EB1B9CEE180039CC8E /* DFGFlushEverythingInsertionPhase.h in Headers */ = {isa = PBXBuildFile; fileRef = 79B531E91B9CEE180039CC8E /* DFGFlushEverythingInsertionPhase.h */; settings = {ATTRIBUTES = (Private, ); }; }; I would call this MaximalFlushInsertionPhase. The names of the phases are meant to parse in English when used in the following two forms: phase name: "Maximal Flush Insertion Phase" command: "perform Maximal Flush Insertion" I think that "Maximal Flush Insertion" is more grammatical than "Flush Everything Insertion", particularly in the two uses above. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:48 > + RELEASE_ASSERT(m_graph.m_form == LoadStore); DFG_ASSERT(m_graph, nullptr, m_graph.m_form == LoadStore) > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:52 > + for (BasicBlock* block : m_graph.blocksInPreOrder()) { m_graph.blocksInNaturalOrder() is what you should use if you don't care about the order. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:60 > + if (!treatedRootBlock) { > + treatRootBlock(block, insertionSet, currentBlockAccessData); > + treatedRootBlock = true; > + isRootBlock = true; > + } I think you're doing this in the loop so that treatRegularBlock() doesn't see the SetLocals that this inserted. I sort of buy that. If you want to do that, then I recommend this code: if (block == m_graph.block(0)) treatRootBlock(block, insertionSet, currentBlockAccessData); But this phase doesn't need to be super fast. It just needs to be accurate. So, I would recommend calling treatRootBlock() *after* this loop. It will need to use its own insertion set and it will essentially recompute currentBlockAccessData, but that's fine. It's more clear that way. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:76 > + if (node->op() != SetLocal && node->op() != SetArgument) > + continue; If you change this so that you're searching forward to find the first VariableAccessData* for each local, then you can just use hasVariableAccessData() instead of casing for particular node types. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:88 > + if (hasSetInRootBlock.operand(operand)) > + return; This seems like a strange behavior. It means that if you have a local that is assigned in the root block, then its initial undefined value will not be flushed. I think that it would be good to avoid this kind of inconsistency. I recommend having this phase unconditionally insert SetLocal(Undefined) for any non-argument at the top of the root block, and use the search above to find the right VariableAccessData* to use. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:99 > + for (unsigned i = 0; i < block->variablesAtTail.numberOfArguments(); i++) > + insertSetLocal(virtualRegisterForArgument(i)); I think that this is unnecessary because: 1) It's wrong. You don't want to set any arguments to Undefined. 2) It's dead code. Every argument will have a SetArgument in the root block, so hasSetInRootBlock.operand(virtualRegisterForArgument(i)) is always true. Instead, I think that you should make the intent more deliberate: you want to set every local to Undefined but you don't want to set any arguments to anything, instead relying on the fact that the arguments already have values. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:102 > + void treatRegularBlock(BasicBlock* block, InsertionSet& insertionSet, Operands<VariableAccessData*>& currentBlockAccessData, bool isRootBlock) In general, instead of passing around isRootBlock, you could say "block == m_graph.block(0)". > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:113 > + case SetArgument: > + case SetLocal: { I think that you only want to insert Flush before SetLocal. I don't think it makes sense to insert Flush before SetArgument. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:128 > + if (isRootBlock) { > + // We don't want to flush before the very first SetLocal/SetArgument in the root block. > + bool alreadySkippedSetLocal = skippedSetAlreadyBecauseRoot.operand(operand); > + if (!alreadySkippedSetLocal) { > + skippedSetAlreadyBecauseRoot.operand(operand) = true; > + continue; > + } > + } You could get rid of this if you just unconditionally SetLocal(Undefined) at the top of the root block. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:141 > + case Flush: { > + VariableAccessData* accessData = node->variableAccessData(); > + VirtualRegister operand = accessData->local(); > + currentBlockAccessData.operand(operand) = accessData; Outside of this switch statement, you could have "if (node->hasVariableAccessData()) currentBlockAccessData.operand(node->local()) = node->variableAccessData()". That way, you don't have to case on node types to figure out if it has variable access data and you don't need to have two copies of this code. > Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.h:35 > + > +bool performFlushEverythingInsertionPhase(Graph& graph); You should put a comment here describing what this phase is about.
Comment on attachment 260808 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260808&action=review >> Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:99 >> + insertSetLocal(virtualRegisterForArgument(i)); > > I think that this is unnecessary because: > > 1) It's wrong. You don't want to set any arguments to Undefined. > 2) It's dead code. Every argument will have a SetArgument in the root block, so hasSetInRootBlock.operand(virtualRegisterForArgument(i)) is always true. > > Instead, I think that you should make the intent more deliberate: you want to set every local to Undefined but you don't want to set any arguments to anything, instead relying on the fact that the arguments already have values. Thanks for your review. The code is much cleaner now. I have a question on (2), I don't see why this holds just by looking at the varargs inlining code in byte code parser. BytecodeParser, when inlining a a varargs call, will add SetArgument nodes to the graph. Why is it then that (2) holds?
Created attachment 260834 [details] patch
Comment on attachment 260834 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260834&action=review > Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:117 > + // FIXME: maybe we should assert this is always a Flush. But would this ever be a GetLocal? Fil, do you have any comments on this?
Created attachment 260835 [details] patch switch to DFG_ASSERT
Attachment 260835 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:88: Place brace on its own line for function definitions. [whitespace/braces] [4] ERROR: Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.h:51: The parameter name "graph" adds no information, so it should be removed. [readability/parameter_name] [5] ERROR: Source/JavaScriptCore/dfg/DFGPlan.cpp:45: Alphabetical sorting problem. [build/include_order] [4] Total errors found: 3 in 10 files If any of these errors are false positives, please file a bug against check-webkit-style.
Comment on attachment 260834 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260834&action=review >> Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:117 >> + // FIXME: maybe we should assert this is always a Flush. But would this ever be a GetLocal? > > Fil, do you have any comments on this? What I really mean here, is, when we loop over numberOfLocals below, if we cached operand=>node here, should we assert that all those nodes were Flushes. (Obviously we may encounter SetArguments here).
(In reply to comment #4) > Comment on attachment 260808 [details] > patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=260808&action=review > > >> Source/JavaScriptCore/dfg/DFGFlushEverythingInsertionPhase.cpp:99 > >> + insertSetLocal(virtualRegisterForArgument(i)); > > > > I think that this is unnecessary because: > > > > 1) It's wrong. You don't want to set any arguments to Undefined. > > 2) It's dead code. Every argument will have a SetArgument in the root block, so hasSetInRootBlock.operand(virtualRegisterForArgument(i)) is always true. > > > > Instead, I think that you should make the intent more deliberate: you want to set every local to Undefined but you don't want to set any arguments to anything, instead relying on the fact that the arguments already have values. > > Thanks for your review. The code is much cleaner now. > I have a question on (2), I don't see why this holds just by looking at the > varargs inlining code in byte code parser. > BytecodeParser, when inlining a a varargs call, will add SetArgument nodes > to the graph. Why is it then that (2) holds? Had to think about this a bit, and in the process I think that I found an error in some of my other feedback. But first, your question. First of all, for (2) to be true, there needs to be at least one [sic] SetLocal or SetArgument to that argument in the root block. So, if varargs inserted some additional SetArgument nodes, it would have no impact on the fact that there was already at least one SetArgument in the root block, since we insert a SetArgument for every argument in the root block. But anyway, varargs inserts SetArgument nodes for calls to inlined functions, not for the arguments to the function we're compiling. So, if it did emit a SetArgument node, it would not be for an argument - it would be for a local. And now, the error: I told you to not insert Flushes before SetArguments. That's wrong. Failure to insert a Flush before a SetArgument would mean that we aren't flushing whatever the last value of the local was prior to that SetArgument. So, you had it right by looking for both SetArgument and SetLocal nodes, and prepending Flush nodes to them. It's true that when you do this, your Flush inserter must not insert a Flush before those SetArgument nodes that already happened to be there. You basically need to be able to ask "is this one of the SetArguments that I am using as the initial value of this local?". Luckily there is a way to ask this question: if (node->op() == SetArgument) { if (node->local().isArgument() && node == m_graph.m_arguments[node->local().toArgument()]) // this node is a primordial SetArgument that represents the value of the local from before the start of IR execution. }
Comment on attachment 260835 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260835&action=review LGTM, but I think you also need to Flush before SetArgument, provided that it's not a primordial SetArgument. > Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:70 > + if (node->op() == SetLocal) { I was wrong before. You also need to consider those SetArgument nodes that are not primordial (i.e. they don't belong to m_graph.m_arguments). > Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:118 > + // FIXME: maybe we should assert this is always a Flush. But would this ever be a GetLocal? > + // Do we ever have GetLocals before a SetLocal in the root block? This is interesting. The first access to a local in the root block cannot be a GetLocal. It should be either a SetLocal or SetArgument. Due to your other transformation, it'll be either SetArgument or Flush.
Comment on attachment 260835 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260835&action=review >> Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:118 >> + // Do we ever have GetLocals before a SetLocal in the root block? > > This is interesting. The first access to a local in the root block cannot be a GetLocal. It should be either a SetLocal or SetArgument. Due to your other transformation, it'll be either SetArgument or Flush. So, if we do see the first node as being a SetArgument here, then it must be a primordial SetArgument node (otherwise we would have inserted a Flush before it). Should we still then insert a SetLocal(Undefined) prior to the primordial node? Or should we skip such SetLocals?
Created attachment 260845 [details] patch
Attachment 260845 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:89: Place brace on its own line for function definitions. [whitespace/braces] [4] Total errors found: 1 in 10 files If any of these errors are false positives, please file a bug against check-webkit-style.
Some performance numbers as an experiment to see the penalty of turning this on: VMs tested: "og" at /Users/saambarati/WK/Clean/WebKitBuild/Release/jsc (r189532) "flush" at /Users/saambarati/WK/ternary/WebKitBuild/Release/jsc (r189532) export JSC_enableMaximalFlushInsertionPhase=1 Collected 10 samples per benchmark/VM, with 10 VM invocations per benchmark. Emitted a call to gc() between sample measurements. Used 1 benchmark iteration per VM invocation for warm-up. Used the jsc-specific preciseTime() function to get microsecond-level timing. Reporting benchmark execution times with 95% confidence intervals in milliseconds. og flush SunSpider: 3d-cube 6.4755+-1.3743 6.1424+-0.3787 might be 1.0542x faster 3d-morph 6.6136+-0.8172 ! 9.4456+-1.0623 ! definitely 1.4282x slower 3d-raytrace 7.8758+-1.1171 ? 9.1533+-1.4357 ? might be 1.1622x slower access-binary-trees 3.3455+-0.6262 ? 3.7128+-0.7768 ? might be 1.1098x slower access-fannkuch 7.6505+-0.7021 ! 13.6345+-1.1741 ! definitely 1.7822x slower access-nbody 2.9725+-0.3136 ! 3.8264+-0.2817 ! definitely 1.2873x slower access-nsieve 3.8612+-0.2273 ? 4.5258+-0.8317 ? might be 1.1721x slower bitops-3bit-bits-in-byte 1.7207+-0.4163 ! 2.8676+-0.3093 ! definitely 1.6666x slower bitops-bits-in-byte 4.0922+-0.2879 ? 4.2477+-0.4752 ? might be 1.0380x slower bitops-bitwise-and 2.2932+-0.0456 ! 6.0569+-0.2203 ! definitely 2.6412x slower bitops-nsieve-bits 4.1909+-0.8229 ! 16.9594+-0.8995 ! definitely 4.0467x slower controlflow-recursive 3.7956+-0.8616 ? 4.3204+-0.6235 ? might be 1.1383x slower crypto-aes 4.6878+-0.4036 ! 8.5556+-1.5396 ! definitely 1.8251x slower crypto-md5 3.0358+-0.3988 ? 3.9798+-0.6452 ? might be 1.3110x slower crypto-sha1 3.6357+-0.7153 3.3074+-0.2253 might be 1.0993x faster date-format-tofte 9.5870+-0.2603 ! 11.4124+-1.0823 ! definitely 1.1904x slower date-format-xparb 5.6403+-0.3881 ! 7.4350+-1.2285 ! definitely 1.3182x slower math-cordic 4.2436+-0.6257 ! 7.9952+-1.6071 ! definitely 1.8840x slower math-partial-sums 5.8466+-0.4142 ! 7.2021+-0.3569 ! definitely 1.2319x slower math-spectral-norm 3.0351+-0.6890 ! 4.5550+-0.5845 ! definitely 1.5008x slower regexp-dna 7.7560+-0.7649 7.1574+-0.2536 might be 1.0836x faster string-base64 5.7308+-0.8192 ? 6.2714+-1.2159 ? might be 1.0943x slower string-fasta 8.7600+-1.7726 ! 13.0052+-2.0591 ! definitely 1.4846x slower string-tagcloud 9.9262+-0.6069 ! 11.9114+-1.0513 ! definitely 1.2000x slower string-unpack-code 26.2628+-2.5058 25.0744+-2.2236 might be 1.0474x faster string-validate-input 6.8533+-0.9303 ? 7.0679+-0.9044 ? might be 1.0313x slower <arithmetic> 6.1495+-0.1439 ! 8.0701+-0.2622 ! definitely 1.3123x slower og flush LongSpider: 3d-cube 986.3528+-10.3882 ! 1031.8449+-12.6571 ! definitely 1.0461x slower 3d-morph 1848.9913+-21.0805 ? 1861.3791+-18.2894 ? 3d-raytrace 776.9968+-41.2581 ? 824.0996+-5.9851 ? might be 1.0606x slower access-binary-trees 1087.5335+-11.9101 ! 1185.0039+-4.9847 ! definitely 1.0896x slower access-fannkuch 350.5218+-3.2144 ! 1683.8007+-16.2968 ! definitely 4.8037x slower access-nbody 632.2262+-8.8302 ! 852.5363+-6.0429 ! definitely 1.3485x slower access-nsieve 572.8992+-4.6685 ! 1106.0133+-6.0152 ! definitely 1.9306x slower bitops-3bit-bits-in-byte 43.1412+-2.3015 ! 138.7816+-2.7007 ! definitely 3.2169x slower bitops-bits-in-byte 99.7509+-6.8719 ! 152.4313+-7.6513 ! definitely 1.5281x slower bitops-nsieve-bits 596.2120+-3.7789 ! 740.5570+-11.6822 ! definitely 1.2421x slower controlflow-recursive 536.2677+-9.3553 535.7968+-6.1016 crypto-aes 712.8685+-6.4402 ! 3360.8066+-30.8215 ! definitely 4.7145x slower crypto-md5 613.6552+-26.6674 ! 829.6543+-37.2248 ! definitely 1.3520x slower crypto-sha1 791.3703+-30.3242 ! 2251.4377+-99.8072 ! definitely 2.8450x slower date-format-tofte 611.1212+-7.2368 ? 612.7934+-6.4772 ? date-format-xparb 796.5434+-4.9867 ! 874.4074+-8.9806 ! definitely 1.0978x slower hash-map 199.7482+-3.5486 ! 285.6361+-12.8789 ! definitely 1.4300x slower math-cordic 605.3922+-9.1437 ! 683.1072+-9.5399 ! definitely 1.1284x slower math-partial-sums 582.6919+-6.6212 ! 611.3866+-21.6257 ! definitely 1.0492x slower math-spectral-norm 688.3118+-5.8348 ! 2341.8443+-20.5792 ! definitely 3.4023x slower string-base64 447.2260+-4.1232 ! 460.8508+-6.3296 ! definitely 1.0305x slower string-fasta 459.6027+-10.9716 ! 719.4856+-8.2178 ! definitely 1.5655x slower string-tagcloud 221.3917+-2.7579 ! 233.3106+-4.2087 ! definitely 1.0538x slower <geometric> 495.9340+-2.4693 ! 759.7270+-3.0344 ! definitely 1.5319x slower og flush V8Spider: crypto 94.8586+-3.5770 ! 276.1156+-5.6876 ! definitely 2.9108x slower deltablue 149.7938+-15.5448 ! 218.4037+-5.3969 ! definitely 1.4580x slower earley-boyer 77.8475+-5.0552 ! 126.6976+-10.2494 ! definitely 1.6275x slower raytrace 63.7298+-7.3514 ! 167.8530+-14.3517 ! definitely 2.6338x slower regexp 79.4428+-2.5620 ? 81.7718+-2.0992 ? might be 1.0293x slower richards 116.7410+-4.8730 ! 165.1984+-8.4731 ! definitely 1.4151x slower splay 68.2214+-5.0650 ? 71.4285+-10.6122 ? might be 1.0470x slower <geometric> 88.7072+-1.1428 ! 142.6580+-3.6462 ! definitely 1.6082x slower og flush Octane: encrypt 0.20833+-0.00219 ! 1.64099+-0.01495 ! definitely 7.8768x slower decrypt 4.11218+-0.04958 ! 29.59152+-0.30892 ! definitely 7.1961x slower deltablue x2 0.19316+-0.00501 ! 0.39175+-0.00924 ! definitely 2.0280x slower earley 0.36940+-0.00461 ! 0.69386+-0.00701 ! definitely 1.8783x slower boyer 5.61721+-0.03895 ! 5.76294+-0.09224 ! definitely 1.0259x slower navier-stokes x2 5.96279+-0.03181 ! 19.67241+-0.17705 ! definitely 3.2992x slower raytrace x2 1.27117+-0.03471 ! 8.14602+-0.25205 ! definitely 6.4083x slower richards x2 0.13635+-0.00234 ! 0.15132+-0.00210 ! definitely 1.1098x slower splay x2 0.57314+-0.02375 ? 0.64384+-0.05069 ? might be 1.1233x slower regexp x2 32.20959+-0.55590 ! 34.95541+-0.29008 ! definitely 1.0852x slower pdfjs x2 48.61470+-1.44142 ! 74.64875+-1.19304 ! definitely 1.5355x slower mandreel x2 74.27947+-2.84910 ! 145.90357+-9.56777 ! definitely 1.9643x slower gbemu x2 63.13872+-2.44900 ! 103.79944+-6.19983 ! definitely 1.6440x slower closure 0.71367+-0.03379 0.70335+-0.00465 might be 1.0147x faster jquery 9.16253+-0.13578 9.10879+-0.07827 box2d x2 17.63766+-0.27134 ! 39.86000+-0.87973 ! definitely 2.2599x slower zlib x2 488.99212+-8.53742 ! 2216.60297+-28.75483 ! definitely 4.5330x slower typescript x2 1009.89226+-41.82195 ! 2111.40337+-52.23943 ! definitely 2.0907x slower <geometric> 7.81414+-0.04852 ! 16.33091+-0.08677 ! definitely 2.0899x slower og flush Kraken: ai-astar 278.866+-4.920 ^ 240.490+-12.914 ^ definitely 1.1596x faster audio-beat-detection 62.770+-1.912 ! 443.506+-17.910 ! definitely 7.0656x slower audio-dft 138.224+-1.634 ! 619.470+-4.812 ! definitely 4.4817x slower audio-fft 41.938+-1.392 ! 378.318+-3.210 ! definitely 9.0208x slower audio-oscillator 73.863+-2.448 ! 84.509+-1.902 ! definitely 1.1441x slower imaging-darkroom 78.734+-5.339 ! 128.095+-12.593 ! definitely 1.6269x slower imaging-desaturate 79.888+-2.877 ! 98.080+-12.020 ! definitely 1.2277x slower imaging-gaussian-blur 105.801+-1.225 ! 2670.923+-14.659 ! definitely 25.2448x slower json-parse-financial 48.941+-1.630 ? 49.052+-1.831 ? json-stringify-tinderbox 28.361+-1.511 ? 29.028+-1.806 ? might be 1.0235x slower stanford-crypto-aes 52.205+-3.029 ! 69.305+-4.135 ! definitely 1.3275x slower stanford-crypto-ccm 48.476+-5.088 ? 51.123+-3.648 ? might be 1.0546x slower stanford-crypto-pbkdf2 115.529+-2.310 ! 131.612+-9.622 ! definitely 1.1392x slower stanford-crypto-sha256-iterative 46.548+-3.903 ? 49.198+-2.559 ? might be 1.0569x slower <arithmetic> 85.725+-1.103 ! 360.193+-2.649 ! definitely 4.2017x slower og flush JSRegress: abc-forward-loop-equal 36.2369+-1.5910 ? 36.8451+-1.5136 ? might be 1.0168x slower abc-postfix-backward-loop 38.4581+-3.6916 36.1085+-1.6206 might be 1.0651x faster abc-simple-backward-loop 35.6084+-0.6875 ? 36.1783+-1.4550 ? might be 1.0160x slower abc-simple-forward-loop 35.9196+-1.2207 ? 36.9598+-1.3882 ? might be 1.0290x slower abc-skippy-loop 26.1297+-1.1851 ? 30.0142+-4.7898 ? might be 1.1487x slower abs-boolean 4.2531+-1.2975 ! 19.3236+-5.4636 ! definitely 4.5434x slower adapt-to-double-divide 19.0427+-0.6309 ? 19.8230+-1.0605 ? might be 1.0410x slower aliased-arguments-getbyval 1.5630+-0.1784 1.5410+-0.1467 might be 1.0143x faster allocate-big-object 2.8970+-0.2622 ! 5.2094+-1.1949 ! definitely 1.7982x slower arguments-named-and-reflective 13.0180+-1.0919 ! 21.5797+-1.2849 ! definitely 1.6577x slower arguments-out-of-bounds 10.8698+-0.4764 ? 11.8577+-0.8397 ? might be 1.0909x slower arguments-strict-mode 11.5181+-0.5805 ! 21.2183+-1.4619 ! definitely 1.8422x slower arguments 9.7277+-0.4208 ! 19.4979+-1.4178 ! definitely 2.0044x slower arity-mismatch-inlining 1.1466+-0.1064 ! 2.4599+-0.3752 ! definitely 2.1454x slower array-access-polymorphic-structure 6.6236+-0.2953 ! 11.9403+-0.4399 ! definitely 1.8027x slower array-nonarray-polymorhpic-access 32.1410+-3.9112 ! 121.5956+-5.8815 ! definitely 3.7832x slower array-prototype-every 94.1156+-3.0431 ? 99.0219+-2.8021 ? might be 1.0521x slower array-prototype-forEach 97.7763+-9.7905 ? 99.4583+-5.4616 ? might be 1.0172x slower array-prototype-map 106.4638+-13.3644 105.2023+-3.8074 might be 1.0120x faster array-prototype-reduce 91.6383+-3.4412 ! 98.0850+-2.7725 ! definitely 1.0703x slower array-prototype-reduceRight 95.0788+-6.2936 ! 330.7264+-133.7211 ! definitely 3.4784x slower array-prototype-some 94.1731+-3.1109 ? 100.7670+-3.9325 ? might be 1.0700x slower array-splice-contiguous 24.5892+-1.0344 ? 25.0117+-1.2646 ? might be 1.0172x slower array-with-double-add 4.6191+-0.5037 ! 17.8230+-1.1763 ! definitely 3.8585x slower array-with-double-increment 4.8849+-1.0672 ! 21.2274+-0.8173 ! definitely 4.3455x slower array-with-double-mul-add 5.8231+-0.8361 ! 19.6176+-0.8416 ! definitely 3.3689x slower array-with-double-sum 3.9224+-0.4721 ! 10.5637+-0.6528 ! definitely 2.6931x slower array-with-int32-add-sub 7.2990+-1.0520 ! 27.0044+-1.2790 ! definitely 3.6997x slower array-with-int32-or-double-sum 4.6833+-0.6236 ! 10.2377+-1.1663 ! definitely 2.1860x slower ArrayBuffer-DataView-alloc-large-long-lived 40.1973+-4.4443 ? 41.9108+-3.0707 ? might be 1.0426x slower ArrayBuffer-DataView-alloc-long-lived 17.4900+-4.9675 15.9530+-0.8451 might be 1.0963x faster ArrayBuffer-Int32Array-byteOffset 4.2590+-0.1849 ! 56.0901+-1.7925 ! definitely 13.1699x slower ArrayBuffer-Int8Array-alloc-large-long-lived 38.5888+-2.1469 ? 43.2180+-3.7880 ? might be 1.1200x slower ArrayBuffer-Int8Array-alloc-long-lived-buffer 24.4612+-1.2645 ! 29.4157+-1.7662 ! definitely 1.2025x slower ArrayBuffer-Int8Array-alloc-long-lived 14.2633+-0.8176 ! 15.8942+-0.5954 ! definitely 1.1143x slower ArrayBuffer-Int8Array-alloc 12.3252+-0.6221 ! 14.3099+-0.8270 ! definitely 1.1610x slower arrowfunction-call 14.4289+-1.1569 ! 23.4082+-1.8670 ! definitely 1.6223x slower asmjs_bool_bug 9.6811+-0.7120 ! 25.5718+-4.6831 ! definitely 2.6414x slower assign-custom-setter-polymorphic 2.8131+-0.1252 ? 2.9217+-0.1885 ? might be 1.0386x slower assign-custom-setter 4.3260+-0.6512 4.2207+-0.7175 might be 1.0249x faster basic-set 13.3983+-1.7246 13.1190+-1.2141 might be 1.0213x faster big-int-mul 4.8923+-0.8293 ! 11.1212+-0.7961 ! definitely 2.2732x slower boolean-test 3.6345+-0.2439 ! 9.3721+-0.8178 ! definitely 2.5786x slower branch-fold 5.1593+-1.1317 ! 15.2857+-1.5919 ! definitely 2.9628x slower branch-on-string-as-boolean 20.5817+-1.0587 ! 188.2343+-9.5655 ! definitely 9.1457x slower by-val-generic 3.4852+-0.9387 ! 7.0102+-0.6465 ! definitely 2.0114x slower call-spread-apply 38.1132+-2.3438 ? 41.3398+-2.8373 ? might be 1.0847x slower call-spread-call 32.0654+-2.7988 ? 32.9398+-1.8339 ? might be 1.0273x slower captured-assignments 0.5537+-0.1370 ? 0.5609+-0.0606 ? might be 1.0131x slower cast-int-to-double 8.0809+-1.9312 ! 24.0669+-1.5250 ! definitely 2.9782x slower cell-argument 7.8215+-0.7967 ? 8.6214+-1.0893 ? might be 1.1023x slower cfg-simplify 3.6302+-0.3321 ! 16.5681+-1.6980 ! definitely 4.5640x slower chain-getter-access 10.6837+-0.5138 ! 32.9755+-1.9395 ! definitely 3.0865x slower cmpeq-obj-to-obj-other 16.3178+-1.8925 ! 21.4969+-1.1624 ! definitely 1.3174x slower constant-test 7.2209+-1.9719 ! 30.4463+-1.0282 ! definitely 4.2164x slower create-lots-of-functions 13.0061+-1.0946 12.6707+-0.8475 might be 1.0265x faster cse-new-array-buffer 2.7004+-0.2896 ? 3.7300+-0.7904 ? might be 1.3813x slower cse-new-array 2.8946+-0.3602 ! 4.9702+-1.5382 ! definitely 1.7170x slower DataView-custom-properties 43.1387+-2.4015 ? 46.6504+-1.6149 ? might be 1.0814x slower delay-tear-off-arguments-strictmode 15.6605+-0.9905 ? 17.3408+-1.2989 ? might be 1.1073x slower deltablue-varargs 256.7865+-26.2379 ! 494.3535+-7.3611 ! definitely 1.9252x slower destructuring-arguments 205.1812+-4.1031 ! 239.7262+-22.7441 ! definitely 1.1684x slower destructuring-parameters-overridden-by-function 0.7164+-0.1527 0.6647+-0.2112 might be 1.0778x faster destructuring-swap 6.5506+-0.7190 ! 16.4976+-0.8554 ! definitely 2.5185x slower direct-arguments-getbyval 1.4086+-0.1575 ? 1.7538+-0.3392 ? might be 1.2451x slower div-boolean-double 5.9781+-0.1262 ! 17.1858+-2.3253 ! definitely 2.8748x slower div-boolean 9.2585+-0.2872 ! 16.8307+-1.9740 ! definitely 1.8179x slower double-get-by-val-out-of-bounds 6.2255+-0.7216 ! 9.2747+-2.0831 ! definitely 1.4898x slower double-pollution-getbyval 10.0823+-0.7109 ! 13.4969+-0.4461 ! definitely 1.3387x slower double-pollution-putbyoffset 5.6852+-1.9779 ! 11.6257+-2.3126 ! definitely 2.0449x slower double-real-use 28.8953+-1.3447 ? 31.9679+-2.0721 ? might be 1.1063x slower double-to-int32-typed-array-no-inline 3.0676+-0.3657 ! 4.2497+-0.1576 ! definitely 1.3854x slower double-to-int32-typed-array 2.8201+-0.2846 ? 3.4303+-0.5793 ? might be 1.2164x slower double-to-uint32-typed-array-no-inline 3.3543+-0.7309 ? 5.1137+-1.4893 ? might be 1.5245x slower double-to-uint32-typed-array 2.8142+-0.1931 ! 3.6139+-0.4440 ! definitely 1.2842x slower elidable-new-object-dag 46.3427+-1.6536 ! 383.1435+-4.5447 ! definitely 8.2676x slower elidable-new-object-roflcopter 44.2102+-2.5504 ! 84.0348+-14.6290 ! definitely 1.9008x slower elidable-new-object-then-call 39.3565+-1.0407 ! 328.4919+-4.2910 ! definitely 8.3466x slower elidable-new-object-tree 51.5827+-2.1613 ! 372.6899+-3.5277 ! definitely 7.2251x slower empty-string-plus-int 6.5535+-1.4651 ? 8.3255+-1.1904 ? might be 1.2704x slower emscripten-cube2hash 38.0299+-1.7908 ! 43.8591+-2.9333 ! definitely 1.1533x slower exit-length-on-plain-object 15.1953+-1.0233 ? 16.2955+-1.7607 ? might be 1.0724x slower external-arguments-getbyval 1.5006+-0.1500 ? 1.6693+-0.3213 ? might be 1.1124x slower external-arguments-putbyval 3.0386+-0.7179 2.7756+-0.3185 might be 1.0948x faster fixed-typed-array-storage-var-index 1.6211+-0.3203 ! 2.6029+-0.1639 ! definitely 1.6056x slower fixed-typed-array-storage 1.3412+-0.4183 ! 2.1509+-0.3166 ! definitely 1.6037x slower Float32Array-matrix-mult 4.6122+-0.4043 ! 14.1384+-1.6629 ! definitely 3.0654x slower Float32Array-to-Float64Array-set 58.9069+-8.1720 ! 140.2476+-5.1763 ! definitely 2.3808x slower Float64Array-alloc-long-lived 87.0905+-9.1038 ! 209.1907+-3.5556 ! definitely 2.4020x slower Float64Array-to-Int16Array-set 73.1120+-6.5942 ! 155.3192+-5.1075 ! definitely 2.1244x slower fold-double-to-int 14.4433+-0.6667 ! 28.4853+-3.4887 ! definitely 1.9722x slower fold-get-by-id-to-multi-get-by-offset-rare-int 16.2968+-1.8056 ! 29.2257+-0.8944 ! definitely 1.7933x slower fold-get-by-id-to-multi-get-by-offset 15.7767+-2.7428 ! 28.6864+-3.0272 ! definitely 1.8183x slower fold-multi-get-by-offset-to-get-by-offset 14.5865+-1.9321 ! 23.5783+-1.1925 ! definitely 1.6164x slower fold-multi-get-by-offset-to-poly-get-by-offset 13.9940+-1.1554 ! 22.9715+-1.7125 ! definitely 1.6415x slower fold-multi-put-by-offset-to-poly-put-by-offset 14.1350+-1.8678 ! 24.2919+-3.2653 ! definitely 1.7186x slower fold-multi-put-by-offset-to-put-by-offset 8.9274+-1.6666 ! 20.9736+-2.3411 ! definitely 2.3493x slower fold-multi-put-by-offset-to-replace-or-transition-put-by-offset 15.0437+-2.2198 ! 25.0214+-1.9026 ! definitely 1.6633x slower fold-put-by-id-to-multi-put-by-offset 16.7672+-2.5838 ! 24.1920+-1.3224 ! definitely 1.4428x slower fold-put-by-val-with-string-to-multi-put-by-offset 15.8056+-1.7919 ! 25.0901+-1.5448 ! definitely 1.5874x slower fold-put-by-val-with-symbol-to-multi-put-by-offset 16.6414+-2.4914 ! 25.2231+-1.4244 ! definitely 1.5157x slower fold-put-structure 9.6407+-2.7432 ! 18.9726+-0.9975 ! definitely 1.9680x slower for-of-iterate-array-entries 14.1694+-2.0111 ! 18.0928+-1.3780 ! definitely 1.2769x slower for-of-iterate-array-keys 4.7365+-0.5921 ? 5.4310+-0.5810 ? might be 1.1466x slower for-of-iterate-array-values 5.2173+-0.8548 5.1474+-0.5871 might be 1.0136x faster fround 20.9188+-0.6834 ! 117.1662+-1.9544 ! definitely 5.6010x slower ftl-library-inlining-dataview 73.5906+-2.0658 ! 86.8898+-2.4924 ! definitely 1.1807x slower ftl-library-inlining 124.2193+-2.4074 ! 162.3463+-10.4739 ! definitely 1.3069x slower function-call 13.5845+-0.5068 ! 22.2193+-0.6902 ! definitely 1.6356x slower function-dot-apply 3.5512+-1.1468 ! 13.4200+-0.7761 ! definitely 3.7790x slower function-test 3.4983+-0.4685 ! 9.7888+-1.0762 ! definitely 2.7982x slower function-with-eval 124.6163+-2.5951 ^ 118.4492+-2.2798 ^ definitely 1.0521x faster gcse-poly-get-less-obvious 16.8511+-0.9114 ! 24.5656+-0.6818 ! definitely 1.4578x slower gcse-poly-get 19.4942+-0.8727 ! 25.2003+-0.9371 ! definitely 1.2927x slower gcse 5.9769+-0.9823 ! 20.5536+-0.5872 ! definitely 3.4388x slower get-by-id-bimorphic-check-structure-elimination-simple 3.4257+-0.3817 ! 6.7776+-0.4321 ! definitely 1.9784x slower get-by-id-bimorphic-check-structure-elimination 7.3161+-1.0568 ! 27.5379+-1.4041 ! definitely 3.7640x slower get-by-id-chain-from-try-block 6.4805+-0.2792 6.3278+-0.2074 might be 1.0241x faster get-by-id-check-structure-elimination 7.0885+-2.5671 ! 22.4256+-0.7793 ! definitely 3.1637x slower get-by-id-proto-or-self 20.1523+-1.8790 ! 60.8088+-1.6905 ! definitely 3.0175x slower get-by-id-quadmorphic-check-structure-elimination-simple 3.8960+-0.4494 ! 7.3301+-0.7466 ! definitely 1.8814x slower get-by-id-self-or-proto 22.1856+-3.0034 ! 61.4364+-2.0132 ! definitely 2.7692x slower get-by-val-out-of-bounds 6.9009+-1.2803 ? 8.9898+-2.6326 ? might be 1.3027x slower get-by-val-with-string-bimorphic-check-structure-elimination-simple 3.6171+-0.3908 ! 6.5019+-0.1865 ! definitely 1.7975x slower get-by-val-with-string-bimorphic-check-structure-elimination 8.1378+-1.3899 ! 44.8433+-1.2782 ! definitely 5.5105x slower get-by-val-with-string-chain-from-try-block 6.7893+-0.6502 6.5742+-0.5373 might be 1.0327x faster get-by-val-with-string-check-structure-elimination 6.4010+-0.7396 ! 34.9073+-1.4368 ! definitely 5.4534x slower get-by-val-with-string-proto-or-self 19.1684+-1.0331 ! 62.1516+-1.6500 ! definitely 3.2424x slower get-by-val-with-string-quadmorphic-check-structure-elimination-simple 4.7652+-1.0811 ! 7.1636+-0.5301 ! definitely 1.5033x slower get-by-val-with-string-self-or-proto 21.0238+-2.1847 ! 62.8205+-1.7583 ! definitely 2.9881x slower get-by-val-with-symbol-bimorphic-check-structure-elimination-simple 4.2852+-0.6608 ! 7.1046+-0.6481 ! definitely 1.6579x slower get-by-val-with-symbol-bimorphic-check-structure-elimination 20.1847+-2.4568 ! 40.9891+-1.5891 ! definitely 2.0307x slower get-by-val-with-symbol-chain-from-try-block 6.4933+-0.2982 6.4924+-0.2668 get-by-val-with-symbol-check-structure-elimination 16.2686+-2.0628 ! 38.7757+-1.8159 ! definitely 2.3835x slower get-by-val-with-symbol-proto-or-self 20.2952+-1.3293 ! 64.6987+-3.4841 ! definitely 3.1879x slower get-by-val-with-symbol-quadmorphic-check-structure-elimination-simple 4.9643+-0.7995 ! 7.6127+-0.7014 ! definitely 1.5335x slower get-by-val-with-symbol-self-or-proto 21.6627+-2.6810 ! 61.9943+-1.8829 ! definitely 2.8618x slower get_callee_monomorphic 4.6729+-1.3657 ! 7.9187+-1.6666 ! definitely 1.6946x slower get_callee_polymorphic 3.8980+-0.5189 ! 11.4289+-1.4680 ! definitely 2.9320x slower getter-no-activation 7.0242+-1.5534 ! 23.5575+-1.0730 ! definitely 3.3537x slower getter-prototype 11.7370+-0.6768 ! 15.0888+-1.9689 ! definitely 1.2856x slower getter-richards 208.7221+-17.9131 ! 354.8555+-11.3760 ! definitely 1.7001x slower getter 8.8970+-0.9809 ! 21.3742+-3.7127 ! definitely 2.4024x slower global-object-access-with-mutating-structure 7.9831+-1.0914 7.9659+-0.7604 global-var-const-infer-fire-from-opt 1.0016+-0.2676 0.8986+-0.1004 might be 1.1146x faster global-var-const-infer 0.9620+-0.2065 0.7925+-0.0696 might be 1.2139x faster hard-overflow-check-equal 40.5881+-1.5243 ! 44.9469+-1.7957 ! definitely 1.1074x slower hard-overflow-check 40.1301+-1.4012 ! 44.6665+-1.7326 ! definitely 1.1130x slower HashMap-put-get-iterate-keys 52.5608+-4.2018 ! 75.6922+-5.2835 ! definitely 1.4401x slower HashMap-put-get-iterate 52.9183+-7.5641 ! 74.2751+-4.7750 ! definitely 1.4036x slower HashMap-string-put-get-iterate 42.3989+-5.1755 ! 83.6054+-12.1558 ! definitely 1.9719x slower hoist-make-rope 11.8137+-0.6652 ? 12.9692+-1.1106 ? might be 1.0978x slower hoist-poly-check-structure-effectful-loop 6.3422+-1.3643 ? 9.1316+-1.5300 ? might be 1.4398x slower hoist-poly-check-structure 4.6680+-0.6013 ! 6.1999+-0.6861 ! definitely 1.3282x slower imul-double-only 9.2800+-1.1242 ? 10.3639+-1.0776 ? might be 1.1168x slower imul-int-only 10.5328+-1.3136 ! 13.8904+-0.6364 ! definitely 1.3188x slower imul-mixed 8.9117+-1.0609 ! 12.9668+-1.3950 ! definitely 1.4550x slower in-four-cases 22.8276+-2.0746 ! 27.1950+-1.0831 ! definitely 1.1913x slower in-one-case-false 19.0893+-5.0241 18.1751+-0.6301 might be 1.0503x faster in-one-case-true 15.5210+-1.4276 ? 17.2963+-0.5694 ? might be 1.1144x slower in-two-cases 15.5184+-1.3943 ! 18.3199+-1.1636 ! definitely 1.1805x slower indexed-properties-in-objects 3.8268+-0.5142 ? 4.5459+-0.4213 ? might be 1.1879x slower infer-closure-const-then-mov-no-inline 5.1927+-1.5539 ? 6.8443+-1.3474 ? might be 1.3181x slower infer-closure-const-then-mov 23.5951+-1.0511 ! 77.3867+-9.7216 ! definitely 3.2798x slower infer-closure-const-then-put-to-scope-no-inline 11.9400+-0.4520 ! 28.7938+-1.2700 ! definitely 2.4115x slower infer-closure-const-then-put-to-scope 26.9369+-0.7376 ! 135.4386+-2.3720 ! definitely 5.0280x slower infer-closure-const-then-reenter-no-inline 55.4021+-2.4284 ! 135.6624+-3.5846 ! definitely 2.4487x slower infer-closure-const-then-reenter 27.9515+-1.1787 ! 136.1804+-2.8693 ! definitely 4.8720x slower infer-constant-global-property 4.3422+-0.5824 ! 34.6919+-1.3402 ! definitely 7.9895x slower infer-constant-property 3.4117+-0.5334 ! 20.1033+-2.1260 ! definitely 5.8924x slower infer-one-time-closure-ten-vars 9.0290+-0.5946 ? 10.7989+-1.9634 ? might be 1.1960x slower infer-one-time-closure-two-vars 8.9368+-0.7504 ? 9.0528+-0.7868 ? might be 1.0130x slower infer-one-time-closure 8.7344+-0.5005 ? 9.7772+-1.3689 ? might be 1.1194x slower infer-one-time-deep-closure 12.9463+-1.0072 ? 14.5968+-2.6068 ? might be 1.1275x slower inline-arguments-access 5.3688+-0.7133 ! 7.9195+-0.8497 ! definitely 1.4751x slower inline-arguments-aliased-access 5.0132+-0.3607 ! 8.2745+-0.7581 ! definitely 1.6505x slower inline-arguments-local-escape 5.1780+-0.6714 ! 8.2256+-0.8362 ! definitely 1.5886x slower inline-get-scoped-var 5.9087+-0.6100 ! 15.0081+-2.0512 ! definitely 2.5400x slower inlined-put-by-id-transition 12.8620+-0.9233 ! 20.5158+-2.4995 ! definitely 1.5951x slower inlined-put-by-val-with-string-transition 92.0535+-10.4707 ? 94.1393+-3.5350 ? might be 1.0227x slower inlined-put-by-val-with-symbol-transition 87.0357+-3.8408 ? 99.2666+-11.1930 ? might be 1.1405x slower int-or-other-abs-then-get-by-val 5.7901+-0.7903 ! 13.0004+-0.5193 ! definitely 2.2453x slower int-or-other-abs-zero-then-get-by-val 18.9187+-1.8591 ! 35.1314+-3.6222 ! definitely 1.8570x slower int-or-other-add-then-get-by-val 6.8373+-2.6033 ! 21.8582+-3.8979 ! definitely 3.1969x slower int-or-other-add 7.7194+-2.4695 ! 18.3746+-0.8007 ! definitely 2.3803x slower int-or-other-div-then-get-by-val 4.5402+-0.4645 ! 10.6128+-0.7782 ! definitely 2.3375x slower int-or-other-max-then-get-by-val 4.4627+-0.2050 ! 13.2897+-1.8469 ! definitely 2.9780x slower int-or-other-min-then-get-by-val 4.5274+-0.4220 ! 13.1027+-1.5816 ! definitely 2.8941x slower int-or-other-mod-then-get-by-val 4.1331+-0.2138 ! 10.3755+-0.6169 ! definitely 2.5103x slower int-or-other-mul-then-get-by-val 4.6000+-0.7620 ! 13.0182+-1.1320 ! definitely 2.8300x slower int-or-other-neg-then-get-by-val 5.5203+-1.2761 ! 12.8440+-0.7784 ! definitely 2.3267x slower int-or-other-neg-zero-then-get-by-val 19.4502+-2.6198 ! 33.2040+-1.1882 ! definitely 1.7071x slower int-or-other-sub-then-get-by-val 5.2088+-0.9778 ! 21.4134+-1.3163 ! definitely 4.1110x slower int-or-other-sub 4.2792+-0.4854 ! 17.6304+-1.1386 ! definitely 4.1200x slower int-overflow-local 6.3099+-2.1352 ! 20.0244+-1.0933 ! definitely 3.1735x slower Int16Array-alloc-long-lived 58.7143+-2.3246 ! 176.2055+-2.5309 ! definitely 3.0011x slower Int16Array-bubble-sort-with-byteLength 22.9224+-2.5537 ! 167.6531+-3.2031 ! definitely 7.3140x slower Int16Array-bubble-sort 22.5332+-1.0715 ! 165.7354+-3.7791 ! definitely 7.3552x slower Int16Array-load-int-mul 2.7540+-1.9854 ? 4.3138+-0.6768 ? might be 1.5664x slower Int16Array-to-Int32Array-set 57.9792+-5.5828 ! 135.9252+-2.4811 ! definitely 2.3444x slower Int32Array-alloc-large 13.4997+-0.7216 13.4947+-0.6153 Int32Array-alloc-long-lived 67.2095+-2.8275 ! 185.3483+-3.2665 ! definitely 2.7578x slower Int32Array-alloc 3.3480+-0.1441 ! 12.1094+-0.6615 ! definitely 3.6168x slower Int32Array-Int8Array-view-alloc 8.5204+-1.0707 ? 9.9158+-0.5412 ? might be 1.1638x slower int52-spill 5.6178+-0.4976 ! 7.8764+-0.9944 ! definitely 1.4020x slower Int8Array-alloc-long-lived 49.2457+-1.8184 ! 168.9125+-2.6052 ! definitely 3.4300x slower Int8Array-load-with-byteLength 4.4459+-0.7002 ? 5.0690+-1.1252 ? might be 1.1402x slower Int8Array-load 4.3267+-0.3802 4.1923+-0.2135 might be 1.0321x faster integer-divide 13.4728+-1.4025 ! 38.9909+-2.6587 ! definitely 2.8940x slower integer-modulo 2.2507+-0.3318 ? 2.3192+-0.6046 ? might be 1.0304x slower is-boolean-fold-tricky 4.6336+-0.3807 ! 40.0188+-1.7900 ! definitely 8.6367x slower is-boolean-fold 3.5681+-0.4345 ! 40.3766+-4.4992 ! definitely 11.3159x slower is-function-fold-tricky-internal-function 12.6308+-1.3900 ! 64.7312+-2.2639 ! definitely 5.1249x slower is-function-fold-tricky 4.7707+-0.6351 ! 65.9397+-5.2015 ! definitely 13.8219x slower is-function-fold 3.4196+-0.1994 ! 57.6359+-1.5674 ! definitely 16.8544x slower is-number-fold-tricky 4.9429+-0.7854 ! 39.8758+-1.8367 ! definitely 8.0673x slower is-number-fold 3.6564+-0.6610 ! 39.2987+-0.9036 ! definitely 10.7480x slower is-object-or-null-fold-functions 3.4802+-0.3256 ! 63.2132+-2.4791 ! definitely 18.1639x slower is-object-or-null-fold-less-tricky 5.5565+-0.9301 ! 57.9133+-1.4846 ! definitely 10.4226x slower is-object-or-null-fold-tricky 6.3273+-1.9640 ! 58.6008+-4.9430 ! definitely 9.2616x slower is-object-or-null-fold 3.5485+-0.5388 ! 57.4842+-1.4712 ! definitely 16.1996x slower is-object-or-null-trickier-function 4.6817+-0.4450 ! 70.7637+-11.4575 ! definitely 15.1149x slower is-object-or-null-trickier-internal-function 12.6362+-1.3150 ! 84.7006+-2.3037 ! definitely 6.7030x slower is-object-or-null-tricky-function 5.2824+-0.9259 ! 61.6878+-1.7306 ! definitely 11.6781x slower is-object-or-null-tricky-internal-function 11.3121+-1.9161 ! 61.3288+-1.5657 ! definitely 5.4215x slower is-string-fold-tricky 4.3635+-0.2015 ! 42.9304+-6.8784 ! definitely 9.8385x slower is-string-fold 3.3063+-0.1938 ! 39.1142+-1.3810 ! definitely 11.8301x slower is-undefined-fold-tricky 4.0644+-0.5880 ! 37.9107+-1.5697 ! definitely 9.3274x slower is-undefined-fold 3.5675+-0.6150 ! 39.5000+-2.5921 ! definitely 11.0723x slower JSONP-negative-0 0.3187+-0.0398 ? 0.3352+-0.0398 ? might be 1.0519x slower large-int-captured 6.6284+-0.9842 ? 6.7652+-1.3071 ? might be 1.0206x slower large-int-neg 17.7499+-1.3269 ? 17.8587+-0.8557 ? large-int 15.6849+-0.8655 ? 16.7169+-0.8138 ? might be 1.0658x slower load-varargs-elimination 24.5343+-0.5830 ! 101.0584+-4.0822 ! definitely 4.1191x slower logical-not-weird-types 4.2276+-0.9032 ! 8.0032+-1.2538 ! definitely 1.8931x slower logical-not 5.6143+-0.9678 ! 10.0175+-0.9738 ! definitely 1.7843x slower lots-of-fields 11.2842+-1.0352 ? 13.2509+-1.8930 ? might be 1.1743x slower make-indexed-storage 4.1772+-0.7153 3.6693+-0.2748 might be 1.1384x faster make-rope-cse 4.9659+-0.4463 ! 17.1632+-1.2979 ! definitely 3.4562x slower marsaglia-larger-ints 40.9223+-3.9692 ! 337.9191+-4.4497 ! definitely 8.2576x slower marsaglia-osr-entry 26.1932+-1.3750 ! 235.0953+-4.6408 ! definitely 8.9754x slower math-with-out-of-bounds-array-values 28.7523+-2.7885 ? 29.1496+-2.7996 ? might be 1.0138x slower max-boolean 3.9564+-0.6864 ! 23.0802+-0.8798 ! definitely 5.8336x slower method-on-number 18.2689+-0.3956 ! 22.7109+-1.1523 ! definitely 1.2431x slower min-boolean 3.3705+-0.2560 ! 23.4064+-3.4204 ! definitely 6.9445x slower minus-boolean-double 3.9931+-0.4473 ! 17.8640+-4.0415 ! definitely 4.4737x slower minus-boolean 2.9460+-0.2471 ! 15.6392+-0.6275 ! definitely 5.3085x slower misc-strict-eq 38.1254+-3.3074 ! 277.4525+-4.2441 ! definitely 7.2774x slower mod-boolean-double 13.2034+-0.5772 ! 24.7171+-1.0452 ! definitely 1.8720x slower mod-boolean 9.2789+-0.2263 ! 47.4875+-1.5158 ! definitely 5.1178x slower mul-boolean-double 4.3112+-0.2003 ! 15.8345+-0.6221 ! definitely 3.6729x slower mul-boolean 4.2944+-1.5226 ! 17.5113+-3.0244 ! definitely 4.0777x slower neg-boolean 3.7730+-0.3176 ! 14.3991+-0.4316 ! definitely 3.8163x slower negative-zero-divide 0.4387+-0.0229 0.4359+-0.0185 negative-zero-modulo 0.4500+-0.0423 ? 0.4740+-0.1191 ? might be 1.0532x slower negative-zero-negate 0.4143+-0.0491 ? 0.4621+-0.0537 ? might be 1.1156x slower nested-function-parsing 53.7693+-2.6419 ? 55.9494+-0.8708 ? might be 1.0405x slower new-array-buffer-dead 117.9383+-4.4844 ! 128.8156+-3.2865 ! definitely 1.0922x slower new-array-buffer-push 7.6528+-0.3258 ! 13.0722+-1.9744 ! definitely 1.7082x slower new-array-dead 19.9125+-2.1405 ! 175.3632+-2.2423 ! definitely 8.8067x slower new-array-push 5.3215+-0.6580 ! 8.2686+-0.9185 ! definitely 1.5538x slower no-inline-constructor 40.9579+-1.3909 ! 69.3798+-2.1550 ! definitely 1.6939x slower number-test 3.5924+-0.3216 ! 8.5958+-0.2227 ! definitely 2.3927x slower object-closure-call 7.8067+-1.0353 ! 16.6759+-0.6527 ! definitely 2.1361x slower object-get-own-property-symbols-on-large-array 5.0529+-0.4812 4.5991+-0.2405 might be 1.0987x faster object-test 3.7323+-0.4689 ! 9.9879+-0.9859 ! definitely 2.6760x slower obvious-sink-pathology-taken 132.9863+-1.7820 ! 266.4994+-4.0600 ! definitely 2.0040x slower obvious-sink-pathology 128.3428+-3.3598 ! 222.2878+-3.6189 ! definitely 1.7320x slower obviously-elidable-new-object 39.5181+-2.7647 ! 217.5721+-2.3406 ! definitely 5.5056x slower plus-boolean-arith 3.1805+-0.3958 ! 13.4702+-0.5692 ! definitely 4.2353x slower plus-boolean-double 3.6560+-0.1119 ! 20.7110+-0.5764 ! definitely 5.6649x slower plus-boolean 3.5553+-0.6016 ! 20.3255+-0.7949 ! definitely 5.7170x slower poly-chain-access-different-prototypes-simple 4.3825+-0.6384 ! 6.3326+-0.2091 ! definitely 1.4450x slower poly-chain-access-different-prototypes 4.1479+-0.4492 ! 6.7883+-0.4267 ! definitely 1.6366x slower poly-chain-access-simpler 4.2150+-0.4309 ! 6.7003+-0.5677 ! definitely 1.5896x slower poly-chain-access 4.3766+-0.7424 ! 6.3009+-0.1473 ! definitely 1.4397x slower poly-stricteq 63.3683+-1.4464 ! 173.7628+-4.7418 ! definitely 2.7421x slower polymorphic-array-call 1.7815+-0.4437 ? 2.1294+-0.2368 ? might be 1.1952x slower polymorphic-get-by-id 3.3923+-0.2653 ! 4.9760+-1.1175 ! definitely 1.4669x slower polymorphic-put-by-id 29.9508+-0.8734 ! 32.3879+-1.5561 ! definitely 1.0814x slower polymorphic-put-by-val-with-string 32.0327+-1.7706 ? 33.8009+-1.4157 ? might be 1.0552x slower polymorphic-put-by-val-with-symbol 30.0580+-1.2529 ! 35.0311+-2.0604 ! definitely 1.1654x slower polymorphic-structure 15.6949+-0.7022 ! 72.8844+-3.1071 ! definitely 4.6438x slower polyvariant-monomorphic-get-by-id 11.0788+-1.0677 ! 36.0304+-2.1938 ! definitely 3.2522x slower proto-getter-access 10.9920+-0.8198 ! 33.5039+-1.2537 ! definitely 3.0480x slower prototype-access-with-mutating-prototype 7.4747+-0.7040 ? 8.4605+-1.1711 ? might be 1.1319x slower put-by-id-replace-and-transition 9.6918+-1.1376 ! 14.0371+-1.0388 ! definitely 1.4483x slower put-by-id-slightly-polymorphic 3.5475+-0.4101 ! 4.8743+-0.8292 ! definitely 1.3740x slower put-by-id 15.2245+-1.5722 ! 39.0799+-1.3854 ! definitely 2.5669x slower put-by-val-direct 0.4409+-0.0412 ? 0.5633+-0.2072 ? might be 1.2776x slower put-by-val-large-index-blank-indexing-type 6.5604+-0.8096 ? 7.4476+-0.6552 ? might be 1.1352x slower put-by-val-machine-int 3.1236+-0.4260 ! 8.4699+-0.6124 ! definitely 2.7116x slower put-by-val-with-string-replace-and-transition 15.4233+-1.4809 ! 21.9935+-1.1713 ! definitely 1.4260x slower put-by-val-with-string-slightly-polymorphic 3.7784+-0.3497 ! 5.9395+-0.8641 ! definitely 1.5720x slower put-by-val-with-string 15.0273+-0.8015 ! 44.5506+-2.6395 ! definitely 2.9646x slower put-by-val-with-symbol-replace-and-transition 16.0079+-1.4694 ! 21.2311+-1.5060 ! definitely 1.3263x slower put-by-val-with-symbol-slightly-polymorphic 4.3782+-0.8310 ! 5.8390+-0.5212 ! definitely 1.3337x slower put-by-val-with-symbol 16.1651+-1.9525 ! 43.6369+-2.1282 ! definitely 2.6995x slower rare-osr-exit-on-local 17.5793+-2.1222 ! 22.1360+-1.7243 ! definitely 1.2592x slower register-pressure-from-osr 19.2285+-0.4855 ! 30.5737+-1.6897 ! definitely 1.5900x slower repeat-multi-get-by-offset 25.9802+-1.1731 ! 40.4344+-2.0533 ! definitely 1.5564x slower setter-prototype 10.2299+-0.9010 ? 11.9988+-1.2441 ? might be 1.1729x slower setter 8.3947+-0.5623 ! 17.2484+-0.4705 ! definitely 2.0547x slower simple-activation-demo 30.4850+-1.2812 ^ 21.2868+-1.4074 ^ definitely 1.4321x faster simple-getter-access 13.8254+-1.6532 ! 49.4520+-1.7668 ! definitely 3.5769x slower simple-poly-call-nested 12.3222+-1.3590 ? 14.3152+-0.6636 ? might be 1.1617x slower simple-poly-call 1.7571+-0.3615 ? 1.9461+-0.2613 ? might be 1.1076x slower sin-boolean 23.9516+-1.8362 ! 27.2089+-0.7587 ! definitely 1.1360x slower singleton-scope 71.8764+-5.5603 ! 120.2474+-12.2424 ! definitely 1.6730x slower sink-function 12.9145+-0.8941 ! 16.4322+-1.4109 ! definitely 1.2724x slower sink-huge-activation 19.0061+-1.3264 ! 33.9748+-2.2591 ! definitely 1.7876x slower sinkable-new-object-dag 74.4398+-3.0207 ! 411.7730+-4.0284 ! definitely 5.5316x slower sinkable-new-object-taken 57.5688+-2.0080 ! 263.7635+-4.7913 ! definitely 4.5817x slower sinkable-new-object 40.1321+-2.4958 ! 226.5844+-2.8291 ! definitely 5.6460x slower slow-array-profile-convergence 4.4752+-1.7772 ? 6.1125+-0.6224 ? might be 1.3659x slower slow-convergence 2.8415+-0.2345 ! 4.7941+-0.8181 ! definitely 1.6872x slower slow-ternaries 20.1699+-0.6580 ! 67.8537+-2.3657 ! definitely 3.3641x slower sorting-benchmark 21.3766+-1.2767 ! 32.8064+-5.0816 ! definitely 1.5347x slower sparse-conditional 1.5777+-0.1668 ! 4.4415+-0.6127 ! definitely 2.8152x slower splice-to-remove 14.5874+-1.1946 ! 21.8753+-5.7244 ! definitely 1.4996x slower string-char-code-at 16.7470+-1.3709 ! 26.7406+-1.0017 ! definitely 1.5967x slower string-concat-object 2.7770+-0.2810 ! 3.8235+-0.4012 ! definitely 1.3769x slower string-concat-pair-object 2.6507+-0.1567 ! 3.7127+-0.3542 ! definitely 1.4007x slower string-concat-pair-simple 12.7577+-1.2290 ! 23.8633+-1.4613 ! definitely 1.8705x slower string-concat-simple 12.5810+-1.1017 ! 23.3063+-1.4723 ! definitely 1.8525x slower string-cons-repeat 8.8282+-0.9787 ! 18.4953+-1.1553 ! definitely 2.0950x slower string-cons-tower 9.4249+-1.0383 8.9007+-0.2492 might be 1.0589x faster string-equality 19.5441+-0.8409 ! 45.9753+-2.0289 ! definitely 2.3524x slower string-get-by-val-big-char 8.7064+-0.8391 ! 29.2584+-1.2275 ! definitely 3.3606x slower string-get-by-val-out-of-bounds-insane 3.6912+-0.1633 ! 4.5787+-0.6864 ! definitely 1.2404x slower string-get-by-val-out-of-bounds 5.3067+-0.6636 ! 40.5687+-7.3024 ! definitely 7.6448x slower string-get-by-val 3.8554+-0.5275 ! 6.5041+-0.4795 ! definitely 1.6870x slower string-hash 3.1771+-0.6988 ! 5.9874+-0.7403 ! definitely 1.8845x slower string-long-ident-equality 15.6375+-0.6197 ! 41.1622+-1.6110 ! definitely 2.6323x slower string-out-of-bounds 12.2673+-0.4123 ? 12.4015+-0.9158 ? might be 1.0109x slower string-repeat-arith 32.8656+-1.0940 ? 33.0811+-1.0553 ? string-sub 62.5581+-1.6278 ! 67.0464+-1.7746 ! definitely 1.0717x slower string-test 3.6645+-0.3983 ! 8.6608+-0.5897 ! definitely 2.3635x slower string-var-equality 33.2771+-3.6004 ! 66.8456+-2.0205 ! definitely 2.0088x slower structure-hoist-over-transitions 3.1062+-0.2626 ! 4.8012+-0.5452 ! definitely 1.5457x slower substring-concat-weird 46.7654+-1.5306 ! 106.1436+-2.7419 ! definitely 2.2697x slower substring-concat 52.6562+-2.4365 ! 98.6371+-2.4882 ! definitely 1.8732x slower substring 54.8445+-1.6401 ! 95.1347+-2.9813 ! definitely 1.7346x slower switch-char-constant 3.9276+-0.6907 ! 12.2836+-0.5980 ! definitely 3.1275x slower switch-char 9.2670+-0.4650 ! 14.7153+-0.6990 ! definitely 1.5879x slower switch-constant 10.4661+-1.3055 ! 21.0040+-2.2980 ! definitely 2.0069x slower switch-string-basic-big-var 17.3224+-1.7040 ! 36.3158+-2.2610 ! definitely 2.0965x slower switch-string-basic-big 17.4259+-1.0087 ! 30.2600+-1.8698 ! definitely 1.7365x slower switch-string-basic-var 15.7900+-1.0064 ! 35.3478+-7.2804 ! definitely 2.2386x slower switch-string-basic 13.9976+-0.1540 ! 27.7222+-1.1819 ! definitely 1.9805x slower switch-string-big-length-tower-var 21.7060+-1.3038 ! 37.6562+-1.9627 ! definitely 1.7348x slower switch-string-length-tower-var 15.4286+-0.7221 ! 29.5419+-1.3146 ! definitely 1.9148x slower switch-string-length-tower 13.8021+-0.5783 ! 26.2857+-1.0053 ! definitely 1.9045x slower switch-string-short 14.8100+-0.8622 ! 26.2789+-1.3194 ! definitely 1.7744x slower switch 14.5742+-1.3046 ! 24.8220+-1.8528 ! definitely 1.7031x slower tear-off-arguments-simple 4.6098+-1.2138 ! 7.5572+-0.7363 ! definitely 1.6394x slower tear-off-arguments 6.1999+-1.3919 ! 10.6957+-1.6928 ! definitely 1.7251x slower temporal-structure 13.9481+-1.1058 ! 75.8085+-9.4283 ! definitely 5.4350x slower to-int32-boolean 15.0986+-0.5964 ! 59.5669+-2.0275 ! definitely 3.9452x slower try-catch-get-by-val-cloned-arguments 17.3955+-1.3385 ? 18.2166+-0.6089 ? might be 1.0472x slower try-catch-get-by-val-direct-arguments 7.3842+-0.3806 ! 9.1880+-0.4452 ! definitely 1.2443x slower try-catch-get-by-val-scoped-arguments 9.2681+-1.0040 ? 11.3986+-1.3181 ? might be 1.2299x slower typed-array-get-set-by-val-profiling 35.0747+-1.0969 ! 39.3967+-2.6600 ! definitely 1.1232x slower undefined-property-access 289.5199+-6.0065 ! 992.0490+-7.8074 ! definitely 3.4265x slower undefined-test 3.9520+-0.8702 ! 9.3533+-0.8714 ! definitely 2.3667x slower unprofiled-licm 17.5572+-1.2493 ? 18.8270+-1.1763 ? might be 1.0723x slower varargs-call 17.0667+-0.5611 ! 73.9109+-7.9150 ! definitely 4.3307x slower varargs-construct-inline 28.5676+-2.0357 ! 78.0553+-3.6133 ! definitely 2.7323x slower varargs-construct 25.8002+-1.6392 ! 86.0874+-3.7792 ! definitely 3.3367x slower varargs-inline 9.8420+-1.3100 ! 51.5915+-2.1032 ! definitely 5.2420x slower varargs-strict-mode 11.5740+-1.0298 ! 182.3822+-6.5264 ! definitely 15.7579x slower varargs 11.8531+-2.2834 ! 63.4104+-2.1146 ! definitely 5.3497x slower weird-inlining-const-prop 2.7620+-0.3978 ! 6.7798+-1.4681 ! definitely 2.4546x slower <geometric> 10.2636+-0.0517 ! 21.7292+-0.0663 ! definitely 2.1171x slower og flush Geomean of preferred means: <scaled-result> 35.0641+-0.1263 ! 69.3932+-0.5994 ! definitely 1.9790x slower
Comment on attachment 260845 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260845&action=review > Tools/Scripts/run-jsc-stress-tests:806 > + runDFGMaximalFlushPhase I like that we are running this by default.
> crypto-aes 712.8685+-6.4402 ! > 3360.8066+-30.8215 ! definitely 4.7145x slower Can you file a bug for this? > raytrace x2 1.27117+-0.03471 ! > 8.14602+-0.25205 ! definitely 6.4083x slower LOL this is a bit of a red herring. Flushing everything means that all of our escape analyses think that the values escape, which disables object allocation elimination and varargs simplification. Octane/raytrace is basically just a test of those optimizations. So, I'm curious what the performance of Octane is if you exclude raytrace from the comparison. That might give some insight into the true slow-down. But, anyway, you should probably file a bug for this. It would be great if we could make raytrace faster even when escape analysis fails! > zlib x2 488.99212+-8.53742 ! > 2216.60297+-28.75483 ! definitely 4.5330x slower LOL this needs a bug, I don't know why this failed so hard. > imaging-gaussian-blur 105.801+-1.225 ! > 2670.923+-14.659 ! definitely 25.2448x slower OMG file a bug.
Comment on attachment 260845 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=260845&action=review > Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:133 > + if (initialAccessNodes.operand(operand)->op() == Flush) Remove. > Source/JavaScriptCore/dfg/DFGMaximalFlushInsertionPhase.cpp:138 > + else { > + DFG_ASSERT(m_graph, nullptr, initialAccessNodes.operand(operand)->op() == SetArgument); > + continue; > + } Remove.
landed in: http://trac.webkit.org/changeset/189544