Vm-dev April 2017

vm-dev@lists.squeakfoundation.org

44 participants
105 discussions

Re: [Vm-dev] [Pharo-dev] patching vmmaker
by Clément Bera 16 Apr '17

16 Apr '17

Hi, The main VMMaker repo is on http:/source.squeak/org/VMMaker, not on smalltalkhub. The repo is common with other programming languages than Pharo. To integrate your changes into the main repo you may send a mail to vm-dev (in CC). If you plan to commit frequently in the VMMaker repository you may ask for commit rights on the repo. Alternatively you may hope that someone with rights will look into your changes. Any change that break one of the builds (including all platforms, programming languages supported, vm versions) that has no massive impact on the VM (performance, stability or feature) is usually ignored. The opensmalltalk-vm github repo contains the generated C sources and most builds (all builds except some oddities like the android version that is not obvious to build with free build frameworks) are generated automatically from this repo using common build frameworks (travis, appveyor). On Sat, Apr 15, 2017 at 11:34 AM, K K Subbu <kksubbu.ml(a)gmail.com> wrote: > Hi, > > While building the VM from generator.image through eval handler, I noticed > spurious blank lines in the log output. I created a changeset that removes > these and creates a compact log (all in VMMaker package). > > My generator.image only throws up local repos for MC. smalltalkhub.com > seems to be down. How do I upload the patch to VMMaker mainline? > > Regards .. Subbu >

3 2

String hash function
by Clément Bera 15 Apr '17

15 Apr '17

Hi all, I am trying to investigate performance overhead in benchmarks to improve the VM performance. In a benchmark strings are used as dictionary keys, which seems to be an OK pattern to me that may be actually used in real code from time to time. Something stroke me: *the hash function of a string is actually iterating over the whole string. * As a result, I get something like this: #( 'String size: 0, Hash function performance: 17,817,776 per second' 'String size: 1, Hash function performance: 15,395,388 per second' 'String size: 3, Hash function performance: 14,853,109 per second' 'String size: 4, Hash function performance: 15,151,954 per second' 'String size: 5, Hash function performance: 14,139,881 per second' 'String size: 10, Hash function performance: 11,545,749 per second' 'String size: 65, Hash function performance: 3,431,067 per second' 'String size: 130, Hash function performance: 1,879,047 per second' 'String size: 520, Hash function performance: 511,934 per second' ) This means that if a hash of a 1MB String is computed it takes a hell of a time. I wonder why we don't compute the string hash from a small number of inner characters taken randomly but deterministically from the String. From example, for all strings longer than 5 characters, one could pick only 5 characters in the String to compute its hash. The hash function is this way decent enough and the performance does not drop for large strings. One implementation could be to replace: 1 to: stringSize do: [:pos | by: 1 to: stringSize by: stringSize // 5 do: [:pos | in: ByteString>>stringHash: aString initialHash: speciesHash Then we need to rehash the whole image. What do you think ? Is there a hash expert around that can confirm this is a good idea ?

9 11

VM Maker: Cog-eem.339.mcz
by commits＠source.squeak.org 14 Apr '17

14 Apr '17

Eliot Miranda uploaded a new version of Cog to project VM Maker: http://source.squeak.org/VMMaker/Cog-eem.339.mcz ==================== Summary ==================== Name: Cog-eem.339 Author: eem Time: 14 April 2017, 11:42:22.888965 am UUID: 66f40e21-3f37-47a6-b722-4fd5031d8c54 Ancestors: Cog-eem.338 Fix ancient slip in GdbARMAlien>>cResultRegister =============== Diff against Cog-eem.338 =============== Item was changed: ----- Method: GdbARMAlien>>cResultRegister (in category 'accessing-abstract') ----- cResultRegister + ^self r0! - self r0!

2 1

Re: [Vm-dev] String hash function
by Ralph Boland 14 Apr '17

14 Apr '17

I had an application where I needed to store arrays into sets and performance depended heavily on this operation. Standard array classes were unacceptable because hashing was either too poor (then) or too expensive (now). So I created my own array class (say HashArray, I forget the name now). HashArray stored a hash value rather than compute one. The hash value of an array was the XOR of its elements (which in my application have good hash values). This worked very well but came at a price: each time an element was added or removed or changed for a HashArray instance the hash value had to be recomputed (O(1) cost). For my application this wasn't a problem so my solution worked well. Notes: 1) In my case two arrays containing the same elements should have the same hash value. Your situation might be different. If so you could also add the the hash value some value associated with the array itself. 2) While contained in a set the array couldn't be changed (didn't happen in my application) but this is true for most objects anyway since its has value cannot change; just something to be aware of. 3) In my application the elements of the array also stored their hash values. The has values were generated using a random number generator. This could be done in my application but obviously can't always be done. This gave very good hashing for my arrays. If you are having performance problems with array classes because of hashing perhaps you should try a scheme similar to mine. Whether or not such a class should be part of Squeak, I don't know. Perhaps there should be a package of HashArray like classes that could be loaded. Ralph Boland

1 0

Fwd: latest vm - 32bit - crash on mac
by Tudor Girba 14 Apr '17

14 Apr '17

I forgot to send this to the VM mailing list. Doru > Begin forwarded message: > > From: Tudor Girba <tudor(a)tudorgirba.com> > Subject: latest vm - 32bit - crash on mac > Date: April 13, 2017 at 10:50:43 PM GMT+2 > To: Pharo Development List <pharo-dev(a)lists.pharo.org> > > Hi, > > I just got a crash on Mac when using the latest VM. I attached the dump. > > This was a short session. I played a bit with a few Roassal visualizations, and then I started some coding in Nautilus, and the image crashed while typing the body of a method. > > I am on OS X, 10.11.6 (Mac OS X El Capitain). > > Here are the details of the VM: > CoInterpreter VMMaker.oscog-eem.2188 uuid: ff4ca601-cd05-4792-ab0d-dcdf19975239 Apr 9 2017 > StackToRegisterMappingCogit VMMaker.oscog-eem.2188 uuid: ff4ca601-cd05-4792-ab0d-dcdf19975239 Apr 9 2017 > VM: 201704090811 https://github.com/OpenSmalltalk/opensmalltalk-vm.git $ Date: Sun Apr 9 10:11:47 2017 +0200 $ Plugins: 201704090811 https://github.com/OpenSmalltalk/opensmalltalk-vm.git $ > > Here is the image build was: 60464. > > Cheers, > Doru > > -- > www.tudorgirba.com > www.feenk.com > > "Problem solving efficiency grows with the abstractness level of problem understanding." > > > > -- www.tudorgirba.com www.feenk.com "Reasonable is what we are accustomed with."

2 3

VM Maker: VMMaker.oscog-eem.2194.mcz
by commits＠source.squeak.org 14 Apr '17

14 Apr '17

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.2194.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.2194 Author: eem Time: 13 April 2017, 3:06:53.413343 pm UUID: 84683bad-a517-40de-bc34-4c1ac734826c Ancestors: VMMaker.oscog-eem.2193 Fix a regression in VMMaker.oscog-eem.2138 & VMMaker.oscog-eem.2143 where receiverTags work broke cointer trips in special selector comparisons of known SmallIntegers. Have compileAbstractInstructionsFrom:through: only check that extensions are consumed if the last bytewcode generator succeeded. Rename Allocat(e)able to Allocable (OED). Add some missing simulation glue. Comment the existence of the variable new inline primitives. =============== Diff against VMMaker.oscog-eem.2193 =============== Item was added: + ----- Method: CurrentImageCoInterpreterFacade>>maybeSelectorOfMethod: (in category 'accessing') ----- + maybeSelectorOfMethod: methodOop + ^self oopForObject: (self objectForOop: methodOop) selector! Item was changed: ----- Method: RegisterAllocatingCogit>>freeAnyRegNotConflictingWith: (in category 'simulation stack') ----- freeAnyRegNotConflictingWith: regMask "Spill the closest register on stack not conflicting with regMask. + Override to unassign assigned temp regs if necessary." - Override so no assertion failure if no register can be allocated." <var: #desc type: #'CogSimStackEntry *'> + | index desc | - | reg index | self assert: needsFrame. - reg := NoReg. index := simSpillBase max: 0. + [index < simStackPtr] whileTrue: + [desc := self simStackAt: index. - [reg = NoReg and: [index < simStackPtr]] whileTrue: - [ | desc | - desc := self simStackAt: index. desc type = SSRegister ifTrue: [(regMask anyMask: (self registerMaskFor: desc register)) ifFalse: + [self ssAllocateRequiredReg: desc register. + ^desc register]]. - [reg := desc register]]. index := index + 1]. + 0 to: methodOrBlockNumTemps - 1 do: + [:i| + desc := self simStackAt: i. + (desc liveRegister = NoReg + or: [self register: desc liveRegister isInMask: regMask]) ifFalse: + [self ssAllocateRequiredReg: desc liveRegister. + ^desc liveRegister]]. + ^NoReg! - reg ~= NoReg ifTrue: - [self ssAllocateRequiredReg: reg]. - ^reg! Item was added: + ----- Method: SimpleStackBasedCogit>>genExtEnsureAllocableSlots (in category 'bytecode generators') ----- + genExtEnsureAllocableSlots + "SistaV1 * 236 11101100 iiiiiiii Ensure Allocable Slots (+ Extend A * 256)" + | slots skip | + slots := (extA bitShift: 8) + byte1. + extA := 0. + self + MoveAw: objectMemory freeStartAddress R: TempReg; + CmpCq: objectMemory getScavengeThreshold - (objectMemory bytesPerOop * slots) R: TempReg. + skip := self JumpBelow: 0. + objectRepresentation genSetGCNeeded. + self CallRT: ceCheckForInterruptTrampoline. + skip jmpTarget: self Label. + self annotateBytecode: skip getJmpTarget. + ^0! Item was removed: - ----- Method: SimpleStackBasedCogit>>genExtEnsureAllocatableSlots (in category 'bytecode generators') ----- - genExtEnsureAllocatableSlots - "SistaV1 * 236 11101100 iiiiiiii Ensure Allocatable Slots (+ Extend A * 256)" - | slots skip | - slots := (extA bitShift: 8) + byte1. - extA := 0. - self - MoveAw: objectMemory freeStartAddress R: TempReg; - CmpCq: objectMemory getScavengeThreshold - (objectMemory bytesPerOop * slots) R: TempReg. - skip := self JumpBelow: 0. - objectRepresentation genSetGCNeeded. - self CallRT: ceCheckForInterruptTrampoline. - skip jmpTarget: self Label. - self annotateBytecode: skip getJmpTarget. - ^0! Item was changed: ----- Method: SistaCogit>>genSpecialSelectorComparison (in category 'bytecode generators') ----- genSpecialSelectorComparison "Override to count inlined branches if followed by a conditional branch. We borrow the following conditional branch's counter and when about to inline the comparison we decrement the counter (without writing it back) and if it trips simply abort the inlining, falling back to the normal send which will then continue to the conditional branch which will trip and enter the abort." | nextPC postBranchPC targetPC primDescriptor branchDescriptor rcvrIsInt rcvrIsConst argIsIntConst argInt jumpNotSmallInts inlineCAB counterAddress countTripped counterReg index | <var: #countTripped type: #'AbstractInstruction *'> <var: #primDescriptor type: #'BytecodeDescriptor *'> <var: #jumpNotSmallInts type: #'AbstractInstruction *'> <var: #branchDescriptor type: #'BytecodeDescriptor *'> (coInterpreter isOptimizedMethod: methodObj) ifTrue: [ ^ self genSpecialSelectorComparisonWithoutCounters ]. self ssFlushTo: simStackPtr - 2. primDescriptor := self generatorAt: byte0. argIsIntConst := self ssTop type = SSConstant and: [objectMemory isIntegerObject: (argInt := self ssTop constant)]. rcvrIsInt := ((rcvrIsConst := (self ssValue: 1) type = SSConstant) and: [objectMemory isIntegerObject:(self ssValue: 1) constant]) or: [self mclassIsSmallInteger and: [(self ssValue: 1) isSameEntryAs: (self addressOf: simSelf)]]. "short-cut the jump if operands are SmallInteger constants." (argIsIntConst and: [rcvrIsInt and: [rcvrIsConst]]) ifTrue: [^ self genStaticallyResolvedSpecialSelectorComparison]. self extractMaybeBranchDescriptorInto: [ :descr :next :postBranch :target | branchDescriptor := descr. nextPC := next. postBranchPC := postBranch. targetPC := target ]. "Only interested in inlining if followed by a conditional branch." inlineCAB := branchDescriptor isBranchTrue or: [branchDescriptor isBranchFalse]. "Further, only interested in inlining = and ~= if there's a SmallInteger constant involved. The relational operators successfully statically predict SmallIntegers; the equality operators do not." (inlineCAB and: [primDescriptor opcode = JumpZero or: [primDescriptor opcode = JumpNonZero]]) ifTrue: [inlineCAB := argIsIntConst or: [rcvrIsInt]]. inlineCAB ifFalse: [^self genSpecialSelectorSend]. argIsIntConst ifTrue: [(self ssValue: 1) popToReg: ReceiverResultReg. self ssPop: 2] ifFalse: [self marshallSendArguments: 1]. jumpNotSmallInts := (rcvrIsInt and: [argIsIntConst]) ifFalse: [argIsIntConst ifTrue: [objectRepresentation genJumpNotSmallInteger: ReceiverResultReg] ifFalse: [rcvrIsInt ifTrue: [objectRepresentation genJumpNotSmallInteger: Arg0Reg] ifFalse: [objectRepresentation genJumpNotSmallIntegersIn: ReceiverResultReg and: Arg0Reg scratch: TempReg]]]. counterReg := self allocateRegNotConflictingWith: (self registerMaskFor: ReceiverResultReg and: Arg0Reg). self genExecutionCountLogicInto: [ :cAddress :countTripBranch | counterAddress := cAddress. countTripped := countTripBranch ] counterReg: counterReg. argIsIntConst ifTrue: [self CmpCq: argInt R: ReceiverResultReg] ifFalse: [self CmpR: Arg0Reg R: ReceiverResultReg]. "Cmp is weird/backwards so invert the comparison. Further since there is a following conditional jump bytecode define non-merge fixups and leave the cond bytecode to set the mergeness." self genConditionalBranch: (branchDescriptor isBranchTrue ifTrue: [primDescriptor opcode] ifFalse: [self inverseBranchFor: primDescriptor opcode]) operand: (self ensureNonMergeFixupAt: targetPC) asUnsignedInteger. self genFallsThroughCountLogicCounterReg: counterReg counterAddress: counterAddress. self Jump: (self ensureNonMergeFixupAt: postBranchPC). countTripped jmpTarget: self Label. + jumpNotSmallInts + ifNil: [(self fixupAt: nextPC) notAFixup ifTrue: + [branchReachedOnlyForCounterTrip := true]] + ifNotNil: [jumpNotSmallInts jmpTarget: countTripped getJmpTarget]. - jumpNotSmallInts ifNil: - [self annotateInstructionForBytecode. - self ensureFixupAt: postBranchPC. - self ensureFixupAt: targetPC. - deadCode := true. - ^0]. - jumpNotSmallInts jmpTarget: countTripped getJmpTarget. argIsIntConst ifTrue: [self MoveCq: argInt R: Arg0Reg]. index := byte0 - self firstSpecialSelectorBytecodeOffset. ^self genMarshalledSend: index negated - 1 numArgs: 1 sendTable: ordinarySendTrampolines! Item was changed: ----- Method: SistaCogitClone>>genBinaryConstOpVarInlinePrimitive: (in category 'inline primitive generators') ----- genBinaryConstOpVarInlinePrimitive: prim "Const op var version of binary inline primitives." "SistaV1: 248 11111000 iiiiiiii mjjjjjjj Call Primitive #iiiiiiii + (jjjjjjj * 256) m=1 means inlined primitive, no hard return after execution. See EncoderForSistaV1's class comment and StackInterpreter>>#binaryInlinePrimitive:" <option: #SistaVM> | ra val untaggedVal adjust | ra := self allocateRegForStackEntryAt: 0. self ssTop popToReg: ra. self ssPop: 1. val := self ssTop constant. self ssPop: 1. untaggedVal := val - objectMemory smallIntegerTag. prim caseOf: { "0 through 6, +, -, *, /, //, \\, quo:, SmallInteger op SmallInteger => SmallInteger, no overflow" [0] -> [self AddCq: untaggedVal R: ra]. [1] -> [self MoveCq: val R: TempReg. self SubR: ra R: TempReg. objectRepresentation genAddSmallIntegerTagsTo: TempReg. self MoveR: TempReg R: ra]. [2] -> [objectRepresentation genShiftAwaySmallIntegerTagsInScratchReg: ra. self MoveCq: untaggedVal R: TempReg. self MulR: TempReg R: ra. objectRepresentation genSetSmallIntegerTagsIn: ra]. + "2011 through 2015 Variable-sized pointers new, byte new, 16-bit new, 32-bit new, 64-bit new" + "2016 through 2020, bitAnd:, bitOr:, bitXor, bitShiftLeft:, bitShiftRight:, SmallInteger op SmallInteger => SmallInteger, no overflow" [16] -> [ self AndCq: val R: ra ]. [17] -> [ self OrCq: val R: ra ]. [18] -> [ self XorCw: untaggedVal R: ra. ]. [19] -> [ objectRepresentation genConvertSmallIntegerToIntegerInReg: ra. self MoveCq: untaggedVal R: TempReg. self LogicalShiftLeftR: ra R: TempReg. objectRepresentation genAddSmallIntegerTagsTo: TempReg. self MoveR: TempReg R: ra]. [20] -> [objectRepresentation genConvertSmallIntegerToIntegerInReg: ra. self MoveCq: untaggedVal R: TempReg. self ArithmeticShiftRightR: ra R: TempReg. objectRepresentation genClearAndSetSmallIntegerTagsIn: TempReg. self MoveR: TempReg R: ra]. "2032 through 2037, >, <, >=, <=. =, ~=, SmallInteger op SmallInteger => Boolean (flags?? then in jump bytecodes if ssTop is a flags value, just generate the instruction!!!!)" "CmpCqR is SubRCq so everything is reversed, but because no CmpRCq things are reversed again and we invert the sense of the jumps." [32] -> [ self CmpCq: val R: ra. self genBinaryInlineComparison: JumpLess opFalse: JumpGreaterOrEqual destReg: ra ]. [33] -> [ self CmpCq: val R: ra. self genBinaryInlineComparison: JumpGreater opFalse: JumpLessOrEqual destReg: ra ]. [34] -> [ self CmpCq: val R: ra. self genBinaryInlineComparison: JumpLessOrEqual opFalse: JumpGreater destReg: ra ]. [35] -> [ self CmpCq: val R: ra. self genBinaryInlineComparison: JumpGreaterOrEqual opFalse: JumpLess destReg: ra ]. [36] -> [ self CmpCq: val R: ra. self genBinaryInlineComparison: JumpZero opFalse: JumpNonZero destReg: ra ]. [37] -> [ self CmpCq: val R: ra. self genBinaryInlineComparison: JumpNonZero opFalse: JumpZero destReg: ra ]. "2064 through 2068, Pointer Object>>at:, Byte Object>>at:, Short16 Word Object>>at: LongWord32 Object>>at: Quad64Word Object>>at:. obj op 0-rel SmallInteger => oop" [64] -> [objectRepresentation genConvertSmallIntegerToIntegerInReg: ra. adjust := (objectMemory baseHeaderSize >> objectMemory shiftForWord) - 1. "shift by baseHeaderSize and then move from 1 relative to zero relative" adjust ~= 0 ifTrue: [ self AddCq: adjust R: ra. ]. self genMoveConstant: val R: TempReg. self MoveXwr: ra R: TempReg R: ra]. [65] -> [objectRepresentation genConvertSmallIntegerToIntegerInReg: ra. adjust := objectMemory baseHeaderSize - 1. "shift by baseHeaderSize and then move from 1 relative to zero relative" self AddCq: adjust R: ra. self genMoveConstant: val R: TempReg. self MoveXbr: ra R: TempReg R: ra. objectRepresentation genConvertIntegerToSmallIntegerInReg: ra] } otherwise: [^EncounteredUnknownBytecode]. self ssPushRegister: ra. ^0! Item was added: + ----- Method: SistaCogitClone>>getCogCodeZoneThreshold (in category 'accessing') ----- + getCogCodeZoneThreshold + <doNotGenerate> + ^methodZone getCogCodeZoneThreshold! Item was added: + ----- Method: StackDepthFinder>>ensureAllocableSlots: (in category 'instruction decoding') ----- + ensureAllocableSlots: numSlots + "nothing to do here..."! Item was removed: - ----- Method: StackDepthFinder>>ensureAllocateableSlots: (in category 'instruction decoding') ----- - ensureAllocateableSlots: numSlots - "nothing to do here..."! Item was changed: ----- Method: StackInterpreter class>>initializeBytecodeTableForSistaV1 (in category 'initialization') ----- initializeBytecodeTableForSistaV1 "See e.g. the cass comment for EncoderForSistaV1" "StackInterpreter initializeBytecodeTableForSistaV1" "Note: This table will be used to generate a C switch statement." initializationOptions at: #SistaV1BytecodeSet put: (SistaV1BytecodeSet := true). BytecodeTable := Array new: 256. BytecodeEncoderClassName := #EncoderForSistaV1. BytecodeSetHasDirectedSuperSend := true. BytecodeSetHasExtensions := true. LongStoreBytecode := 245. self table: BytecodeTable from: #( "1 byte bytecodes" ( 0 15 pushReceiverVariableBytecode) ( 16 31 pushLiteralVariable16CasesBytecode) ( 32 63 pushLiteralConstantBytecode) ( 64 75 pushTemporaryVariableBytecode) ( 76 pushReceiverBytecode) ( 77 pushConstantTrueBytecode) ( 78 pushConstantFalseBytecode) ( 79 pushConstantNilBytecode) ( 80 pushConstantZeroBytecode) ( 81 pushConstantOneBytecode) ( 82 extPushPseudoVariable) ( 83 duplicateTopBytecode) ( 84 87 unknownBytecode) ( 88 returnReceiver) ( 89 returnTrue) ( 90 returnFalse) ( 91 returnNil) ( 92 returnTopFromMethod) ( 93 returnNilFromBlock) ( 94 returnTopFromBlock) ( 95 extNopBytecode) ( 96 bytecodePrimAdd) ( 97 bytecodePrimSubtract) ( 98 bytecodePrimLessThanSistaV1) "for booleanCheatSistaV1:" ( 99 bytecodePrimGreaterThanSistaV1) "for booleanCheatSistaV1:" (100 bytecodePrimLessOrEqualSistaV1) "for booleanCheatSistaV1:" (101 bytecodePrimGreaterOrEqualSistaV1) "for booleanCheatSistaV1:" (102 bytecodePrimEqualSistaV1) "for booleanCheatSistaV1:" (103 bytecodePrimNotEqualSistaV1) "for booleanCheatSistaV1:" (104 bytecodePrimMultiply) (105 bytecodePrimDivide) (106 bytecodePrimMod) (107 bytecodePrimMakePoint) (108 bytecodePrimBitShift) (109 bytecodePrimDiv) (110 bytecodePrimBitAnd) (111 bytecodePrimBitOr) (112 bytecodePrimAt) (113 bytecodePrimAtPut) (114 bytecodePrimSize) (115 bytecodePrimNext) "i.e. a 0 arg special selector" (116 bytecodePrimNextPut) "i.e. a 1 arg special selector" (117 bytecodePrimAtEnd) (118 bytecodePrimIdenticalSistaV1) "for booleanCheatSistaV1:" (119 bytecodePrimClass) (120 bytecodePrimNotIdenticalSistaV1) "was blockCopy:" (121 bytecodePrimValue) (122 bytecodePrimValueWithArg) (123 bytecodePrimDo) "i.e. a 1 arg special selector" (124 bytecodePrimNew) "i.e. a 0 arg special selector" (125 bytecodePrimNewWithArg) "i.e. a 1 arg special selector" (126 bytecodePrimPointX) "i.e. a 0 arg special selector" (127 bytecodePrimPointY) "i.e. a 0 arg special selector" (128 143 sendLiteralSelector0ArgsBytecode) (144 159 sendLiteralSelector1ArgBytecode) (160 175 sendLiteralSelector2ArgsBytecode) (176 183 shortUnconditionalJump) (184 191 shortConditionalJumpTrue) (192 199 shortConditionalJumpFalse) (200 207 storeAndPopReceiverVariableBytecode) (208 215 storeAndPopTemporaryVariableBytecode) (216 popStackBytecode) (217 unconditionnalTrapBytecode) (218 223 unknownBytecode) "2 byte bytecodes" (224 extABytecode) (225 extBBytecode) (226 extPushReceiverVariableBytecode) (227 extPushLiteralVariableBytecode) (228 extPushLiteralBytecode) (229 longPushTemporaryVariableBytecode) (230 unknownBytecode) (231 pushNewArrayBytecode) (232 extPushIntegerBytecode) (233 extPushCharacterBytecode) (234 extSendBytecode) (235 extSendSuperBytecode) + (236 extEnsureAllocableSlots) - (236 extEnsureAllocatableSlots) (237 extUnconditionalJump) (238 extJumpIfTrue) (239 extJumpIfFalse) (240 extSistaStoreAndPopReceiverVariableBytecode) (241 extSistaStoreAndPopLiteralVariableBytecode) (242 longStoreAndPopTemporaryVariableBytecode) (243 extSistaStoreReceiverVariableBytecode) (244 extSistaStoreLiteralVariableBytecode) (245 longStoreTemporaryVariableBytecode) (246 247 unknownBytecode) "3 byte bytecodes" (248 callPrimitiveBytecode) (249 extPushFullClosureBytecode) (250 extPushClosureBytecode) (251 extPushRemoteTempOrInstVarLongBytecode) (252 extStoreRemoteTempOrInstVarLongBytecode) (253 extStoreAndPopRemoteTempOrInstVarLongBytecode) (254 extJumpIfNotInstanceOfBehaviorsBytecode) (255 unknownBytecode) )! Item was added: + ----- Method: StackInterpreter>>extEnsureAllocableSlots (in category 'miscellaneous bytecodes') ----- + extEnsureAllocableSlots + "SistaV1 * 236 11101100 iiiiiiii Ensure Allocable Slots (+ Extend A * 256)" + | slots ok | + slots := (extA bitShift: 8) + self fetchByte. + self fetchNextBytecode. + extA := 0. + ok := objectMemory checkForAvailableSlots: slots. + ok ifFalse: + [self externalizeIPandSP. + self checkForEventsMayContextSwitch: true. + self browserPluginReturnIfNeeded. + self internalizeIPandSP]! Item was removed: - ----- Method: StackInterpreter>>extEnsureAllocatableSlots (in category 'miscellaneous bytecodes') ----- - extEnsureAllocatableSlots - "SistaV1 * 236 11101100 iiiiiiii Ensure Allocatable Slots (+ Extend A * 256)" - | slots ok | - slots := (extA bitShift: 8) + self fetchByte. - self fetchNextBytecode. - extA := 0. - ok := objectMemory checkForAvailableSlots: slots. - ok ifFalse: - [self externalizeIPandSP. - self checkForEventsMayContextSwitch: true. - self browserPluginReturnIfNeeded. - self internalizeIPandSP]! Item was changed: ----- Method: StackToRegisterMappingCogit class>>initializeBytecodeTableForSistaV1 (in category 'class initialization') ----- initializeBytecodeTableForSistaV1 "StackToRegisterMappingCogit initializeBytecodeTableForSistaV1" numPushNilsFunction := #sistaV1:Num:Push:Nils:. pushNilSizeFunction := #sistaV1PushNilSize:numInitialNils:. BytecodeSetHasDirectedSuperSend := true. FirstSpecialSelector := 96. NumSpecialSelectors := 32. self flag: 'Special selector send class must be inlined to agree with the interpreter, which inlines class. If class is sent to e.g. a general instance of ProtoObject then unless class is inlined there will be an MNU. It must be that the Cointerpreter and Cogit have identical semantics. We get away with not hardwiring the other special selectors either because in the Cointerpreter they are not inlined or because they are inlined only to instances of classes for which there will always be a method.'. self generatorTableFrom: #( "1 byte bytecodes" "pushes" (1 0 15 genPushReceiverVariableBytecode isInstVarRef needsFrameNever: 1) (1 16 31 genPushLitVarDirSup16CasesBytecode needsFrameNever: 1) (1 32 63 genPushLiteralConstantBytecode needsFrameNever: 1) (1 64 75 genPushTemporaryVariableBytecode needsFrameIfMod16GENumArgs: 1) (1 76 76 genPushReceiverBytecode needsFrameNever: 1) (1 77 77 genPushConstantTrueBytecode needsFrameNever: 1) (1 78 78 genPushConstantFalseBytecode needsFrameNever: 1) (1 79 79 genPushConstantNilBytecode needsFrameNever: 1) (1 80 80 genPushConstantZeroBytecode needsFrameNever: 1) (1 81 81 genPushConstantOneBytecode needsFrameNever: 1) (1 82 82 genExtPushPseudoVariable) (1 83 83 duplicateTopBytecode needsFrameNever: 1) (1 84 87 unknownBytecode) "returns" (1 88 88 genReturnReceiver return needsFrameIfInBlock: isMappedInBlock 0) (1 89 89 genReturnTrue return needsFrameIfInBlock: isMappedInBlock 0) (1 90 90 genReturnFalse return needsFrameIfInBlock: isMappedInBlock 0) (1 91 91 genReturnNil return needsFrameIfInBlock: isMappedInBlock 0) (1 92 92 genReturnTopFromMethod return needsFrameIfInBlock: isMappedInBlock -1) (1 93 93 genReturnNilFromBlock return needsFrameNever: -1) (1 94 94 genReturnTopFromBlock return needsFrameNever: -1) (1 95 95 genExtNopBytecode needsFrameNever: 0) "sends" (1 96 96 genSpecialSelectorArithmetic isMapped AddRR) (1 97 97 genSpecialSelectorArithmetic isMapped SubRR) (1 98 98 genSpecialSelectorComparison isMapped JumpLess) (1 99 99 genSpecialSelectorComparison isMapped JumpGreater) (1 100 100 genSpecialSelectorComparison isMapped JumpLessOrEqual) (1 101 101 genSpecialSelectorComparison isMapped JumpGreaterOrEqual) (1 102 102 genSpecialSelectorComparison isMapped JumpZero) (1 103 103 genSpecialSelectorComparison isMapped JumpNonZero) (1 104 109 genSpecialSelectorSend isMapped) " #* #/ #\\ #@ #bitShift: //" (1 110 110 genSpecialSelectorArithmetic isMapped AndRR) (1 111 111 genSpecialSelectorArithmetic isMapped OrRR) (1 112 117 genSpecialSelectorSend isMapped) "#at: #at:put: #size #next #nextPut: #atEnd" (1 118 118 genSpecialSelectorEqualsEquals needsFrameNever: notMapped -1) "not mapped because it is directly inlined (for now)" (1 119 119 genSpecialSelectorClass needsFrameIfStackGreaterThanOne: notMapped 0) "not mapped because it is directly inlined (for now)" (1 120 120 genSpecialSelectorNotEqualsEquals needsFrameNever: notMapped -1) "not mapped because it is directly inlined (for now)" (1 121 127 genSpecialSelectorSend isMapped) "#value #value: #do: #new #new: #x #y" (1 128 143 genSendLiteralSelector0ArgsBytecode isMapped) (1 144 159 genSendLiteralSelector1ArgBytecode isMapped) (1 160 175 genSendLiteralSelector2ArgsBytecode isMapped) "jumps" (1 176 183 genShortUnconditionalJump branch v3:ShortForward:Branch:Distance:) (1 184 191 genShortJumpIfTrue branch isBranchTrue isMapped "because of mustBeBoolean" v3:ShortForward:Branch:Distance:) (1 192 199 genShortJumpIfFalse branch isBranchFalse isMapped "because of mustBeBoolean" v3:ShortForward:Branch:Distance:) (1 200 207 genStoreAndPopReceiverVariableBytecode isInstVarRef is1ByteInstVarStore isMappedIfImmutability needsFrameIfImmutability: -1) (1 208 215 genStoreAndPopTemporaryVariableBytecode) (1 216 216 genPopStackBytecode needsFrameNever: -1) (1 217 217 genUnconditionalTrapBytecode isMapped) (1 218 223 unknownBytecode) "2 byte bytecodes" (2 224 224 extABytecode extension) (2 225 225 extBBytecode extension) "pushes" (2 226 226 genExtPushReceiverVariableBytecode isInstVarRef) "Needs a frame for context inst var access" (2 227 227 genExtPushLitVarDirSupBytecode needsFrameNever: 1) (2 228 228 genExtPushLiteralBytecode needsFrameNever: 1) (2 229 229 genLongPushTemporaryVariableBytecode) (2 230 230 unknownBytecode) (2 231 231 genPushNewArrayBytecode) (2 232 232 genExtPushIntegerBytecode needsFrameNever: 1) (2 233 233 genExtPushCharacterBytecode needsFrameNever: 1) "returns" "sends" (2 234 234 genExtSendBytecode isMapped) (2 235 235 genExtSendSuperBytecode isMapped) "sista bytecodes" + (2 236 236 genExtEnsureAllocableSlots isMapped) - (2 236 236 genExtEnsureAllocatableSlots isMapped) "jumps" (2 237 237 genExtUnconditionalJump branch isMapped "because of interrupt check" v4:Long:Branch:Distance:) (2 238 238 genExtJumpIfTrue branch isBranchTrue isMapped "because of mustBeBoolean" v4:Long:Branch:Distance:) (2 239 239 genExtJumpIfFalse branch isBranchFalse isMapped "because of mustBeBoolean" v4:Long:Branch:Distance:) "stores" (2 240 240 genSistaExtStoreAndPopReceiverVariableBytecode isInstVarRef isMappedIfImmutability) (2 241 241 genSistaExtStoreAndPopLiteralVariableBytecode isMappedIfImmutability) (2 242 242 genLongStoreAndPopTemporaryVariableBytecode) (2 243 243 genSistaExtStoreReceiverVariableBytecode isInstVarRef isMappedIfImmutability) (2 244 244 genSistaExtStoreLiteralVariableBytecode isMappedIfImmutability) (2 245 245 genLongStoreTemporaryVariableBytecode) (2 246 247 unknownBytecode) "3 byte bytecodes" (3 248 248 genCallPrimitiveBytecode) (3 249 249 genExtPushFullClosureBytecode) (3 250 250 genExtPushClosureBytecode block v4:Block:Code:Size:) (3 251 251 genExtPushRemoteTempOrInstVarLongBytecode) (3 252 252 genExtStoreRemoteTempOrInstVarLongBytecode isMappedIfImmutability) (3 253 253 genExtStoreAndPopRemoteTempOrInstVarLongBytecode isMappedIfImmutability) (3 254 254 genExtJumpIfNotInstanceOfBehaviorsBytecode branch v4:Long:BranchIfNotInstanceOf:Distance:) (3 255 255 unknownBytecode))! Item was changed: ----- Method: StackToRegisterMappingCogit>>compileAbstractInstructionsFrom:through: (in category 'compile abstract instructions') ----- compileAbstractInstructionsFrom: start through: end "Loop over bytecodes, dispatching to the generator for each bytecode, handling fixups in due course." | nextOpcodeIndex descriptor nExts fixup result | <var: #descriptor type: #'BytecodeDescriptor *'> <var: #fixup type: #'BytecodeFixup *'> self traceSimStack. bytecodePC := start. nExts := result := 0. descriptor := nil. deadCode := false. [self maybeHaltIfDebugPC. fixup := self fixupAt: bytecodePC. self mergeWithFixupIfRequired: fixup. self assertCorrectSimStackPtr. descriptor := self loadBytesAndGetDescriptor. nextOpcodeIndex := opcodeIndex. result := deadCode ifTrue: [self mapDeadDescriptorIfNeeded: descriptor] ifFalse: [self perform: descriptor generator]. + result = 0 ifTrue: [self assertExtsAreConsumed: descriptor]. - self assertExtsAreConsumed: descriptor. self traceDescriptor: descriptor; traceSimStack. self patchFixupTargetIfNeeded: fixup nextOpcodeIndex: nextOpcodeIndex. self maybeDumpLiterals: descriptor. bytecodePC := self nextBytecodePCFor: descriptor exts: nExts. result = 0 and: [bytecodePC <= end]] whileTrue: [nExts := descriptor isExtension ifTrue: [nExts + 1] ifFalse: [0]]. self checkEnoughOpcodes. ^result! Item was added: + ----- Method: StackToRegisterMappingCogit>>genExtEnsureAllocableSlots (in category 'bytecode generators') ----- + genExtEnsureAllocableSlots + "SistaV1 * 236 11101100 iiiiiiii Ensure Allocable Slots (+ Extend A * 256)" + self ssFlushTo: simStackPtr. + ^super genExtEnsureAllocableSlots! Item was removed: - ----- Method: StackToRegisterMappingCogit>>genExtEnsureAllocatableSlots (in category 'bytecode generators') ----- - genExtEnsureAllocatableSlots - "SistaV1 * 236 11101100 iiiiiiii Ensure Allocatable Slots (+ Extend A * 256)" - self ssFlushTo: simStackPtr. - ^super genExtEnsureAllocatableSlots!

1 0

VM Maker: ImageFormat-dtl.23.mcz
by commits＠source.squeak.org 13 Apr '17

13 Apr '17

David T. Lewis uploaded a new version of ImageFormat to project VM Maker: http://source.squeak.org/VMMaker/ImageFormat-dtl.23.mcz ==================== Summary ==================== Name: ImageFormat-dtl.23 Author: dtl Time: 13 April 2017, 10:15:41.267 am UUID: 0eacae94-ff21-4207-8f57-e92e3952977b Ancestors: ImageFormat-dtl.22 Refactoring and cleaner magic file output by Subbu. =============== Diff against ImageFormat-dtl.22 =============== Item was changed: ----- Method: ImageFormat class>>unixMagicFileEntries (in category 'unix magic file entries') ----- unixMagicFileEntries "Answer a string that can be appended to /etc/magic on a Unix system to support the file(1) utility" "ImageFormat unixMagicFileEntries" " | fs | fs := FileStream newFileNamed: 'magic'. [ fs nextPutAll: ImageFormat unixMagicFileEntries ] ensure: [ fs close ] " ^String streamContents: [:s | s nextPutAll: '# Smalltalk image file formats'; lf. KnownVersionNumbers do: [ :num | | fmt | #( 'le' 'be' ) do: [ :endian | fmt := self fromInteger: num. (fmt is64Bit and: [ endian = 'be' ]) ifTrue: [ s nextPut: $4 ] ifFalse: [ s nextPut: $0 ]. s tab; nextPutAll: endian; nextPutAll: 'long'; tab; nextPutAll: num asString; tab; nextPutAll: 'Smalltalk '. fmt printTerseDescriptionOn: s. s lf. + s nextPutAll: '!!:mime application/'; + nextPutAll: fmt simpleName; + nextPutAll: '-image'; + lf - s nextPutAll: '!!:mime application/'. - fmt requiresSpurSupport - ifTrue: [ s nextPutAll: 'spur'] - ifFalse: [fmt requiresClosureSupport ifTrue: [ s nextPutAll: 'cog'] - ifFalse: [s nextPutAll: 'squeak']]. - fmt is64Bit ifTrue: [ s nextPutAll: '64']. - s nextPutAll: '-image'. - s lf ] ]. s lf. ]! Item was changed: ----- Method: ImageFormat>>printTerseDescriptionOn: (in category 'printing') ----- printTerseDescriptionOn: stream "Shortened description as may be required for unix magic file entries" + stream + nextPutAll: self simpleName; + nextPutAll: ' image '. - stream - nextPutAll: 'image '; - nextPutAll: (self requiresSpurSupport - ifTrue: [ 'Spur ' ] - ifFalse: [ 'V3 ' ]); - nextPutAll: (wordSize * 8) asString; - nextPutAll: 'b '. self requiresClosureSupport ifTrue: [stream nextPutAll: '+C']. self requiresNativeFloatWordOrder ifTrue: [stream nextPutAll: '+NF']. self requiresNewSpur64TagAssignment ifTrue: [stream nextPutAll: '+Tag' ]. stream nextPutAll: ' (%d)'. ^ stream ! Item was added: + ----- Method: ImageFormat>>simpleName (in category 'printing') ----- + simpleName + + "Return a simple name for the format, suitable for use as filename or mimetype. + (ImageFormat fromInteger: 6500) simpleName." + + ^String streamContents: [:s | + self requiresSpurSupport + ifTrue: [ s nextPutAll: 'spur'] + ifFalse: [self requiresClosureSupport + ifTrue: [ s nextPutAll: 'cog'] + ifFalse: [s nextPutAll: 'squeak']]. + self is64Bit ifTrue: [ s nextPutAll: '64']]. + !

1 0

[VM-dev] What the difference in method format between 32 and 64 bits images?
by Denis Kudriashov 13 Apr '17

13 Apr '17

Hi I am testing serialization library and now I have problem with compiled methods. I try to materialise 64bits method in 32bits image and otherwise. Both is not working (but in same image everything is fine). I investigated following difference with "Margitude>>between:and:" as example: (Margitude>>between:and:) size. "=> 64bits->46; 32bits->30" (Margitude>>between:and:) header. "same 34078723 for 64 and 32 bits" (Margitude>>between:and:) initialPC. "=> 64bits-> 33; 32bits->17" Is there any formula to convert these values between 64 and 32 bits? Best regards, Denis

2 2

VM Maker: ImageFormat-dtl.22.mcz
by commits＠source.squeak.org 13 Apr '17

13 Apr '17

David T. Lewis uploaded a new version of ImageFormat to project VM Maker: http://source.squeak.org/VMMaker/ImageFormat-dtl.22.mcz ==================== Summary ==================== Name: ImageFormat-dtl.22 Author: dtl Time: 12 April 2017, 7:52:22.953 pm UUID: 4abaa676-0204-4ac8-94a9-4ebe43a73949 Ancestors: ImageFormat-dtl.21 K K Subbu updates for magic file generation (vm-dev list) =============== Diff against ImageFormat-dtl.21 =============== Item was changed: ----- Method: ImageFormat class>>unixMagicFileEntries (in category 'unix magic file entries') ----- unixMagicFileEntries "Answer a string that can be appended to /etc/magic on a Unix system to support the file(1) utility" "ImageFormat unixMagicFileEntries" " | fs | fs := FileStream newFileNamed: 'magic'. [ fs nextPutAll: ImageFormat unixMagicFileEntries ] ensure: [ fs close ] " ^String streamContents: [:s | s nextPutAll: '# Smalltalk image file formats'; lf. KnownVersionNumbers do: [ :num | | fmt | #( 'le' 'be' ) do: [ :endian | fmt := self fromInteger: num. (fmt is64Bit and: [ endian = 'be' ]) ifTrue: [ s nextPut: $4 ] ifFalse: [ s nextPut: $0 ]. s tab; nextPutAll: endian; nextPutAll: 'long'; tab; nextPutAll: num asString; tab; nextPutAll: 'Smalltalk '. fmt printTerseDescriptionOn: s. + s lf. + s nextPutAll: '!!:mime application/'. + fmt requiresSpurSupport + ifTrue: [ s nextPutAll: 'spur'] + ifFalse: [fmt requiresClosureSupport ifTrue: [ s nextPutAll: 'cog'] + ifFalse: [s nextPutAll: 'squeak']]. + fmt is64Bit ifTrue: [ s nextPutAll: '64']. + s nextPutAll: '-image'. s lf ] ]. s lf. ]! Item was changed: ----- Method: ImageFormat>>printTerseDescriptionOn: (in category 'printing') ----- printTerseDescriptionOn: stream "Shortened description as may be required for unix magic file entries" stream nextPutAll: 'image '; nextPutAll: (self requiresSpurSupport ifTrue: [ 'Spur ' ] ifFalse: [ 'V3 ' ]); nextPutAll: (wordSize * 8) asString; + nextPutAll: 'b '. + self requiresClosureSupport ifTrue: [stream nextPutAll: '+C']. + self requiresNativeFloatWordOrder ifTrue: [stream nextPutAll: '+NF']. + self requiresNewSpur64TagAssignment ifTrue: [stream nextPutAll: '+Tag' ]. + stream nextPutAll: ' (%d)'. - nextPutAll: '-bit '. - self requiresClosureSupport ifFalse: [stream nextPutAll: 'no ']. - stream nextPutAll: 'closure '. - self requiresNativeFloatWordOrder - ifTrue: [stream nextPutAll: 'native float'] - ifFalse: [stream nextPutAll: 'no native float']. - stream nextPutAll: ' ('; - nextPutAll: self asInteger asString; - nextPut: $). ^ stream !

1 0

VM Maker: VMMaker.oscog-eem.2193.mcz
by commits＠source.squeak.org 12 Apr '17

12 Apr '17

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.2193.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.2193 Author: eem Time: 12 April 2017, 2:46:53.675491 pm UUID: cf454521-165e-4268-ae52-4e47159af4c4 Ancestors: VMMaker.oscog-eem.2192 Fix ancient bug with non-local return in the StackToRegisterMappingCogit; the stack must be flushed before calling the non-local return trampoline. Solves some, if not all, of the stack depth mismatches on non-local return. Make sure some Siata bytecodes are recognised by the StackDepthFinder. =============== Diff against VMMaker.oscog-eem.2192 =============== Item was changed: ----- Method: CogVMSimulator>>maybeCheckStackDepth:sp:pc: (in category 'debug support') ----- maybeCheckStackDepth: delta sp: sp pc: mcpc | asp bcpc startbcpc cogHomeMethod cogBlockMethod csp debugStackPointers | debugStackDepthDictionary ifNil: [^self]. (self isMachineCodeFrame: framePointer) ifFalse: [^self]. cogBlockMethod := self mframeCogMethod: framePointer. cogHomeMethod := self asCogHomeMethod: cogBlockMethod. debugStackPointers := debugStackDepthDictionary at: cogHomeMethod methodObject ifAbsentPut: [self debugStackPointersFor: cogHomeMethod methodObject]. startbcpc := cogHomeMethod = cogBlockMethod ifTrue: [self startPCOfMethod: cogHomeMethod methodObject] ifFalse: [self startPCOfClosure: (self pushedReceiverOrClosureOfFrame: framePointer)]. bcpc := cogit bytecodePCFor: mcpc startBcpc: startbcpc in: cogBlockMethod. self assert: bcpc ~= 0. ((cogBlockMethod ~= cogHomeMethod or: [cogBlockMethod cmIsFullBlock]) and: [cogit isNonLocalReturnPC: mcpc]) ifTrue: [| lastbcpc | "Method returns within a block (within an unwind-protect) must check the stack depth at the return, not the bytecode following, but the pc mapping maps to the bytecode following the return. lastBytecodePCForBlockAt:in: catches method returns at the end of a block, modifying the bcpc to that of the return. isNonLocalReturnPC: catches method returns not at the end. Assumes method return bytecodes are 1 bytecode long;a dodgy assumption, but good enough." lastbcpc := cogBlockMethod cmIsFullBlock ifTrue: [cogit endPCOf: cogHomeMethod methodObject] ifFalse: [cogit lastBytecodePCForBlockAt: startbcpc in: cogHomeMethod methodObject]. + bcpc > lastbcpc ifTrue: [bcpc := lastbcpc]]. - bcpc := bcpc > lastbcpc ifTrue: [lastbcpc] ifFalse: [bcpc - 1]]. asp := self stackPointerIndexForFrame: framePointer WithSP: sp + objectMemory wordSize. csp := debugStackPointers at: bcpc ifAbsent: [-1]. "Compensate for some edge cases" asp - delta = csp ifTrue: ["Compensate for the implicit context receiver push in a trap bytecode with the absence of a contnuation. Assumes trap bytecodes are 1 byte bytecodes." (SistaVM and: [cogit isTrapAt: mcpc]) ifTrue: [csp := csp + 1]. "Compensate lazily for absent receiver sends (cuz mapping is slow, even though incrememting csp is a dodgy idea)." (NewspeakVM and: [cogit isAbsentReceiverSendAt: mcpc in: cogHomeMethod]) ifTrue: [csp := debugStackPointers at: bcpc put: csp + 1]]. self assert: asp - delta + 1 = csp! Item was added: + ----- Method: StackDepthFinder>>branchIfNotInstanceOf:distance: (in category 'instruction decoding') ----- + branchIfNotInstanceOf: behaviorOrArrayOfBehavior distance: delta + self drop. + self doJump: delta! Item was changed: + ----- Method: StackDepthFinder>>ensureAllocateableSlots: (in category 'instruction decoding') ----- - ----- Method: StackDepthFinder>>ensureAllocateableSlots: (in category 'as yet unclassified') ----- ensureAllocateableSlots: numSlots "nothing to do here..."! Item was changed: ----- Method: StackToRegisterMappingCogit>>genUpArrowReturn (in category 'bytecode generators') ----- genUpArrowReturn "Generate a method return from within a method or a block. Frameless method activation looks like CISCs (x86): receiver args sp-> ret pc. RISCs (ARM): receiver args ret pc in LR. A fully framed activation is described in CoInterpreter class>initializeFrameIndices. Return pops receiver and arguments off the stack. Callee pushes the result." - | framelessReturn | deadCode := true. "can't fall through" inBlock > 0 ifTrue: + [self assert: needsFrame. + self ssFlushTo: simStackPtr. - [self assert: needsFrame. self CallRT: ceNonLocalReturnTrampoline. self annotateBytecode: self Label. ^0]. + (self cppIf: IMMUTABILITY ifTrue: [needsFrame and: [useTwoPaths not]] ifFalse: [needsFrame]) - self - cppIf: IMMUTABILITY - ifTrue: [framelessReturn := needsFrame and: [useTwoPaths not]] - ifFalse: [framelessReturn := needsFrame]. - framelessReturn ifTrue: + [LowcodeVM ifTrue: [ hasNativeFrame ifTrue: [ self leaveNativeFrame ] ]. - [ - LowcodeVM ifTrue: [ hasNativeFrame ifTrue: [ self leaveNativeFrame ] ]. self MoveR: FPReg R: SPReg. self PopR: FPReg. backEnd hasLinkRegister ifTrue: [self PopR: LinkReg]. self RetN: methodOrBlockNumArgs + 1 * objectMemory wordSize] ifFalse: [self RetN: ((methodOrBlockNumArgs > self numRegArgs "A method with an interpreter prim will push its register args for the prim. If the failure body is frameless the args must still be popped, see e.g. Behavior>>nextInstance." or: [regArgsHaveBeenPushed]) ifTrue: [methodOrBlockNumArgs + 1 * objectMemory wordSize] ifFalse: [0])]. ^0! Item was added: + ----- Method: VMObjectProxy>>selector (in category 'accessing') ----- + selector + | selector | + selector := coInterpreter maybeSelectorOfMethod: oop. + ^coInterpreter stringOf: selector!

1 0

← Newer
1
...
4
5
6
7
8
9
10
11
Older →

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev April 2017