Vm-dev March 2012

vm-dev@lists.squeakfoundation.org

23 participants
41 discussions

Fwd: Strange bug in numerical computation
by Bert Freudenberg 11 Mar '12

11 Mar '12

Hi folks, we're having some problems with floats on ARM (iPad, specifically). Any idea? - Bert - Begin forwarded message: > From: Hilaire Fernandes <hilaire.fernandes(a)gmail.com> > Subject: Strange bug in numerical computation > Date: 10. März 2012 07:28:33 MEZ > To: "drgeo-dev(a)googlegroups.com" <drgeo-dev(a)googlegroups.com> > Reply-To: drgeo-dev(a)googlegroups.com > > Esteban, > > There are some bugs hanging in the VM (I guess) related to numerical > computation. > I found some sketch does not load or bring error like out of memory error. > > For example, when loading the sketch "symetrie axiale.fgeo", we have an > out of memory error, indeed during the construction of the interactive > sketch out of this file, a polygon with extent 2808173@957466 is > constructed. > > Esteban , I took a basic Pharo image, installed the iPhone and > ObjectiveC packages, then run the 'KernelTests-Numbers'. > On Linux I got no error nor failure. > On iPad, I got 23 failures and 8 errors, then the image hang at the > testSqrFloor test. > > Guys, it is the place to look at. > To make your testing easier, here is the image you can use on iPad to test: > http://ubuntuone.com/0IzxMtXtdHk5KjWeAQBZVU (expect a few minute for the > file to be published on ubuntuone) > > In the mean time I will continue polishing the user interface. > > Hilaire > > -- > Dr. Geo -- http://www.drgeo.eu

1 0

Re: [Pharo-project] GSoC: ARM jitter for Squeak VM
by Eliot Miranda 08 Mar '12

08 Mar '12

On Thu, Mar 8, 2012 at 1:20 PM, Nick Ager <nick.ager(a)gmail.com> wrote: > Hi Elliot, > > I'm probably the least qualified of the people who have responded so far - > but if at all possible I'd love to try and help. > You're welcome too. > If you want something to do now, the first thing to do is to produce an >> ARM simulator plugin, analogous to the Bochs simulator. The first thing to >> do is go out and find a suitable software simulator, written in either C or >> C++ (cuz these suit the plugin compilation system) and play with it, >> configuring an ARM and running some simple assembler on it. >> > > In terms of ARM emulators a popular one seems to be QEMU [1] - the Android > simulator uses it. > Right, but this one is unsuitable because it does dynamic translation. Because the Squeak VM has a moving garbage collector and because, when simulating, the entire Cog heap, including generated machine-code, lives in a single ByteArray, the machine code generated by Cog, when simulating, can move. So the interface between the processor simulator and the generated machine code and memory must be based on virtual addresses. I'm pretty certain that a JIT like QEMU will use real addresses and so its own code generation would get confused by any movements of the Cog heap ByteArray. Hence it is I think better to stick with a simple interpretive simulator such as the Armulator <http://en.wikipedia.org/wiki/ARMulator>, and of course Bochs <http://en.wikipedia.org/wiki/Bochs>. BTW, the ARM is so simple that the core instruction set simulator in the ARM is only 1500 lines of C++ (including whitespace). So another approach might be to implement ARM in Smalltalk, e.g. by back-porting the Armulator code, although I expect it'll be significantly slower than optimized C++. > > Cheers > > Nick > > [1] http://en.wikipedia.org/wiki/QEMU > -- best, Eliot

1 0

CompiledMethod header bits index 29 (flag bit)
by Stefan Marr 07 Mar '12

07 Mar '12

Hi: What is the intention of the flag bit in the CompiledMethod header? Just interested what it is good for, or was used for originally. More specifically, I would like to verify that I can just reuse it without having to worry about to many incompatibilities. My understanding so far is that it is only used to highlight methods in a browser. Thanks Stefan PS: the split encoding of the primitive code seems also to be a beautiful artifact of an old legacy ;) -- Stefan Marr Software Languages Lab Vrije Universiteit Brussel Pleinlaan 2 / B-1050 Brussels / Belgium http://soft.vub.ac.be/~smarr Phone: +32 2 629 2974 Fax: +32 2 629 3525

4 7

Re: [squeak-dev] Smalltalk as embedded platform/language?
by Eliot Miranda 07 Mar '12

07 Mar '12

On Tue, Mar 6, 2012 at 4:35 PM, Igor Stasenko <siguctua(a)gmail.com> wrote: > It is very long post, so i decided to put it on my blog. > > http://computeradventures.wordpress.com/ > > i more or less realizing the scale of changes in order to achieve it, > but i also wonder why we do not pay attention to this aspect.. > maybe i miss something, maybe we don't need it at all. or maybe > there's less painful alternatives... > My main task this year at Cadence is to make Newspeak embeddable. So I will have cycles to work on this :) I've done it before with VisualWorks, although it never made production. I'd *like* to use this as cover for working on the new object representation too, but that's proving harder to sell. > > -- > Best regards, > Igor Stasenko. > > -- best, Eliot

2 1

New Research Group would like to participate from MUM in Iowa
by Mohammed Farrag 07 Mar '12

07 Mar '12

Dear developers, Greetings, We are a new parallel programming research group in Maharishi University of Management and we would like to take a step in ly programming language development. From our perspective, we believe that Non-deterministic programming languages have a great future in the next few years so we would like to participate and share our knowledge with yours to come up with long steps in that field (possibly a paper will be published for the work achieved in this research and technically reviewed by Bruce P. Lester, author of "The Art of Parallel Programming" book). We have a few questions; some of them to emphasize our thoughts and the rest we are asking because we didn't find appropriate answers for them. Please accept our apologies for some trivial questions because we are at the "very first beginning steps" in Ly. First, here's our system specification I7 Quad core, 16 GB 3 units of 300 GB – Raid 5 for Data 1 Unit of 300 GB – Primary HDD for OS Two NIC cards Notes: 1. RAM – DDR3 1333 2. Processor – i7 2.67 GHz Are these specification will be appropriate for our research? Note that the university won't be able to buy 16core system now for financial reasons. In this case, What's the most appropriate Operating System which provides good tools to write, compile and debug our code?. Second, Our group already read "Harnessing emergence for manycore programming: early experience integrating ensembles, adverbs, and object-based inheritance" and it provides nice background about ly so our next step is to pick a book which describes the ly syntax. We will be glad if you could provide us good tutorials about Ly. Finally, We would like to know about the current progress to avoid the duplicates and conflicts so we will be glad if you mention the status of your development so we can catch up a new branch to extend the capabilities of this language. Regards, Mohamed

1 0

Re: [Pharo-project] NativeBoost VM's
by Eliot Miranda 06 Mar '12

06 Mar '12

Hi Igor, On Sun, Mar 4, 2012 at 1:04 PM, Igor Stasenko <siguctua(a)gmail.com> wrote: > On 4 March 2012 22:46, Jan van de Sandt <jvdsandt(a)gmail.com> wrote: > > Hello, > > > > On the Pharo Jenkins server [1] there is a recent build of a VM with > > NativeBoost support for the Mac. But the Unix version is six months old > and > > there is no Windows version. > > The Unix one was not built because i forgot to set the upstream project, :) > so build doesn't triggers when everything else builds. > The Windows one, i was built manually on my machine, because we were > lacking > windows slave for jenkins. > You can download windows VM from nativeboost project downloads page: > http://code.google.com/p/nativeboost/downloads/list > > I'm not happy with VM i built (Cog + NativeBoost Win32 Binaries (7 Feb, > 2012) ). > I built the threaded VM, and besides the warning which pops up every > time you launch it, > i had couple crashes when simply loading MC packages. > I am still not sure what happens, and need to investigate, because the > other VM i built recently > on another windows machine (by installing everything from scratch) > behaves quite stable. > I would definitely treat the threaded VM as experimental for now. It is in development (stalled due to lack of time), bleeding edge and very much not ready for prime time. I used to use it regularly on Mac but then it started locking up. I plan to get back to it this year sometime, but right now it is "use at your own risk" and not something to inflict on the unwitting. > > > > > Is this just a Jenkins issue and is NativeBoost working on Unix and > Windows? > > Or are these platforms not working yet? > > > Everything is working. It is just jenkins. > I hope in following few days, after we deal with weird free space > issues on jenkins, > we can finally set up windows builds for vanilla vms as well as for > nativeboost. > > > Jan. > > > > [1] https://ci.lille.inria.fr/pharo/view/NativeBoost/ > > > > -- > Best regards, > Igor Stasenko. > > -- best, Eliot

2 2

Re: [Pharo-project] Cog VM binaries as per VMMaker.oscog-eem.152/r2538.
by Eliot Miranda 05 Mar '12

05 Mar '12

On Sun, Mar 4, 2012 at 12:36 PM, Frank Shearar <frank.shearar(a)gmail.com>wrote: > On 4 March 2012 19:44, Philippe Marschall <kustos(a)gmx.net> wrote: > > On 03.03.2012 23:21, Eliot Miranda wrote: > >> > >> > >> > >> > >> > >> Cog VM binaries as per VMMaker.oscog-eem.152/r2538 > >> <http://www.mirandabanda.org/files/Cog/VM/VM.r2538/>. > >> > >> Fix tricky context state bug that can cause crashes in the GC. > >> Contextpart>runUntilErrorOrReturnFrom: may pop the last argument. > >> Hence if a frame is married in this state its spouse context won't have > >> its last argument slot initialized. If later the frame is divorced and > >> the > >> arguments are not updated the last argument slot can be left with this > >> uninitialized slot and ... bang. > >> The solution is to update arguments as well as stack contents when > >> divorcing. This is good also because it removes the VM's assumption > >> that method arguments are read-only, and that's only enforced by the > >> Smalltalk bytecode compiler, not by the bytecode. > > > > > > I just wanted to report that my LD_LIBRARY_PATH problems are gone and > the VM > > starts again on 64bit. Also TCP_CORK seems to work. > > On my Ubuntu Lucid install the magic statement in the case returns > > /lib/tls/i686/cmov/libc.so.6 > > So with the addition of > > /lib/tls/i686/cmov/libc*) > SVMLLP="/lib/tls/i686/cmov:/usr/lib";; > > to the script, things seem OK: image booted, it updated, it passed the > UUID tests. I'm just running the full suite of tests now. Right, and > it didn't crash on playing that beep, which scared the __ out of me. > OK, the entire suite ran without crashing, which is great! Thanks, > Thanks Frank! People, please keep these corrections to LD_LIBRARY_PATH coming. And if any ubuntu guru can figure out how to discard the sceme on ubuntu and what the script shoudl do on ubuntu I'll happily integrate that too. > Eliot! > > I did have 18 failures and 38 errors, but that's the topic of a separate > mail. > > frank > > > Thanks > > Philippe > > > > > > -- best, Eliot

1 0

Cog VM binaries as per VMMaker.oscog-eem.152/r2538.
by Eliot Miranda 03 Mar '12

03 Mar '12

Cog VM binaries as per VMMaker.oscog-eem.152/r2538<http://www.mirandabanda.org/files/Cog/VM/VM.r2538/> . Fix tricky context state bug that can cause crashes in the GC. Contextpart>runUntilErrorOrReturnFrom: may pop the last argument. Hence if a frame is married in this state its spouse context won't have its last argument slot initialized. If later the frame is divorced and the arguments are not updated the last argument slot can be left with this uninitialized slot and ... bang. The solution is to update arguments as well as stack contents when divorcing. This is good also because it removes the VM's assumption that method arguments are read-only, and that's only enforced by the Smalltalk bytecode compiler, not by the bytecode. -- best, Eliot

1 0

VM Maker: VMMaker.oscog-eem.152.mcz
by commits＠source.squeak.org 03 Mar '12

03 Mar '12

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.152.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.152 Author: eem Time: 3 March 2012, 12:46:18 pm UUID: d39d62b0-45ca-4621-8f67-e3981bbe1875 Ancestors: VMMaker.oscog-eem.151 Fix tricky context state bug that can cause crashes in the GC. Contextpart>runUntilErrorOrReturnFrom: may pop the last argument. Hence if a frame is married in this state its spouse context won't have its last argument slot initialized. If later the frame is divorced and the arguments are not updated the last argument slot can be left with this uninitialized slot and ... bang. The sloution is to update arguments as well as stack contents when divorcing. This is good also becausde it removes the VM's assumption that method arguments are read-only, and that's only enforced by the Smalltalk bytecode compiler, not by the bytecode. =============== Diff against VMMaker.oscog-eem.151 =============== Item was changed: ----- Method: CoInterpreter>>updateStateOfSpouseContextForFrame:WithSP: (in category 'frame access') ----- updateStateOfSpouseContextForFrame: theFP WithSP: theSP "Update the frame's spouse context with the frame's current state except for the sender and instruction pointer, which are used to mark the context as married." + | theContext tempIndex pointer argsPointer | - | theContext tempIndex pointer | <inline: false> <var: #theFP type: #'char *'> <var: #theSP type: #'char *'> <var: #pointer type: #'char *'> + <var: #argsPointer type: #'char *'> self assert: (self frameHasContext: theFP). theContext := self frameContext: theFP. self assert: (self isContext: theContext). self assert: (self frameReceiver: theFP) = (objectMemory fetchPointer: ReceiverIndex ofObject: theContext). (self isMachineCodeFrame: theFP) ifTrue: [tempIndex := self mframeNumArgs: theFP. pointer := theFP + FoxMFReceiver - BytesPerWord] ifFalse: [tempIndex := self iframeNumArgs: theFP. pointer := theFP + FoxIFReceiver - BytesPerWord]. + "update the arguments. this would appear not to be strictly necessary, but is for two reasons. + First, the fact that arguments are read-only is only as convention in the Smalltalk compiler; + other languages may choose to modify arguments. + Second, the Squeak runUntilErrorOrReturnFrom: nightmare pops the stack top, which may, in + certain circumstances, be the last argument, and hence the last argument may not have been + stored into the context." + argsPointer := theFP + (self frameStackedReceiverOffsetNumArgs: tempIndex). + 1 to: tempIndex do: + [:i| + argsPointer := argsPointer - BytesPerWord. + self assert: (objectMemory addressCouldBeOop: (stackPages longAt: argsPointer)). + objectMemory storePointer: ReceiverIndex + i + ofObject: theContext + withValue: (stackPages longAt: argsPointer)]. + "now update the non-argument stack contents." [pointer >= theSP] whileTrue: [self assert: (objectMemory addressCouldBeOop: (stackPages longAt: pointer)). tempIndex := tempIndex + 1. objectMemory storePointer: ReceiverIndex + tempIndex ofObject: theContext withValue: (stackPages longAt: pointer). pointer := pointer - BytesPerWord]. self assert: ReceiverIndex + tempIndex < (objectMemory lengthOf: theContext). objectMemory storePointerUnchecked: StackPointerIndex ofObject: theContext withValue: (objectMemory integerObjectOf: tempIndex)! Item was changed: ----- Method: StackInterpreter>>updateStateOfSpouseContextForFrame:WithSP: (in category 'frame access') ----- updateStateOfSpouseContextForFrame: theFP WithSP: theSP "Update the frame's spouse context with the frame's current state except for the sender and instruction pointer, which are used to mark the context as married." | theContext tempIndex pointer | <inline: false> <var: #theFP type: #'char *'> <var: #theSP type: #'char *'> <var: #pointer type: #'char *'> + <var: #argsPointer type: #'char *'> self assert: (self frameHasContext: theFP). theContext := self frameContext: theFP. self assert: (self frameReceiver: theFP) = (objectMemory fetchPointer: ReceiverIndex ofObject: theContext). tempIndex := self frameNumArgs: theFP. + "update the arguments. this would appear not to be strictly necessary, but is for two reasons. + First, the fact that arguments are read-only is only as convention in the Smalltalk compiler; + other languages may choose to modify arguments. + Second, the Squeak runUntilErrorOrReturnFrom: nightmare pops the stack top, which may, in + certain circumstances, be the last argument, and hence the last argument may not have been + stored into the context." + pointer := theFP + (self frameStackedReceiverOffsetNumArgs: tempIndex). + 1 to: tempIndex do: + [:i| + pointer := pointer - BytesPerWord. + self assert: (objectMemory addressCouldBeOop: (stackPages longAt: pointer)). + objectMemory storePointer: ReceiverIndex + i + ofObject: theContext + withValue: (stackPages longAt: pointer)]. + "now update the non-argument stack contents." pointer := theFP + FoxReceiver - BytesPerWord. [pointer >= theSP] whileTrue: [self assert: (objectMemory addressCouldBeOop: (stackPages longAt: pointer)). tempIndex := tempIndex + 1. objectMemory storePointer: ReceiverIndex + tempIndex ofObject: theContext withValue: (stackPages longAt: pointer). pointer := pointer - BytesPerWord]. self assert: ReceiverIndex + tempIndex < (objectMemory lengthOf: theContext). objectMemory storePointerUnchecked: StackPointerIndex ofObject: theContext withValue: (objectMemory integerObjectOf: tempIndex)!

1 0

VM Maker: VMMaker.oscog-eem.151.mcz
by commits＠source.squeak.org 03 Mar '12

03 Mar '12

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.151.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.151 Author: eem Time: 3 March 2012, 12:31:42.725 pm UUID: 04f1d165-afc5-4f65-8b65-3be15e84b11d Ancestors: VMMaker.oscog-eem.150 Various clean-ups in debugging VM crash (fixes to follow). Provide checkedOkayOop: et al for checking (no fatal error). Correct some receivers in simulation (objectMemory vs coInterpreter). Provide shortPrintFramesInPage: =============== Diff against VMMaker.oscog-eem.150 =============== Item was changed: ----- Method: CoInterpreter>>ceDynamicSuperSend:to:numArgs: (in category 'trampolines') ----- ceDynamicSuperSend: selector to: rcvr numArgs: numArgs "Entry-point for an unlinked dynamic super send in a CogMethod. Smalltalk stack looks like receiver args head sp -> sender return pc If an MNU then defer to handleMNUInMachineCodeTo:... which will dispatch the MNU and may choose to allocate a closed PIC with a fast MNU dispatch for this send. Otherwise attempt to link the send site as efficiently as possible. All link attempts may fail; e.g. because we're out of code memory. Continue execution via either executeMethod or interpretMethodFromMachineCode: depending on whether the target method is cogged or not." <api> <option: #NewspeakVM> | class canLinkCacheTag errSelIdx cogMethod newCogMethod mClassMixin mixinApplication | <inline: false> <var: #cogMethod type: #'CogMethod *'> <var: #newCogMethod type: #'CogMethod *'> "self printExternalHeadFrame" "self printStringOf: selector" cogit assertCStackWellAligned. self assert: ((objectMemory isIntegerObject: rcvr) or: [objectMemory addressCouldBeObj: rcvr]). self sendBreak: selector + BaseHeaderSize point: (objectMemory lengthOf: selector) receiver: rcvr. mClassMixin := self mMethodClass. mixinApplication := self findApplicationOfTargetMixin: mClassMixin startingAtBehavior: (objectMemory fetchClassOf: rcvr). self assert: (objectMemory lengthOf: mixinApplication) > (InstanceSpecificationIndex + 1). lkupClass := self superclassOf: mixinApplication. + class := objectMemory fetchClassOf: rcvr. - class := self fetchClassOf: rcvr. canLinkCacheTag := (objectMemory isYoungObject: class) not or: [cogit canLinkToYoungClasses]. "We set the messageSelector and lkupClass for executeMethod below since things like the at cache read messageSelector and lkupClass and so they cannot be left stale." messageSelector := selector. lkupClass := self superclassOf: mixinApplication. argumentCount := numArgs. (self lookupInMethodCacheSel: selector class: lkupClass) ifTrue:"check for coggability because method is in the cache" [self ifAppropriateCompileToNativeCode: newMethod selector: selector] ifFalse: [(errSelIdx := self lookupMethodNoMNUEtcInClass: lkupClass) ~= 0 ifTrue: [self handleMNU: errSelIdx InMachineCodeTo: rcvr classForMessage: lkupClass. "NOTREACHED" self assert: false]]. "Method found and has a cog method. Attempt to link to it." (self maybeMethodHasCogMethod: newMethod) ifTrue: [cogMethod := self cogMethodOf: newMethod. cogMethod selector = objectMemory nilObject ifTrue: [cogit setSelectorOf: cogMethod to: selector]. "Deal with anonymous accessors, e.g. in Newspeak. The cogMethod may not have the correct selector. If not, try and compile a new method with the correct selector." cogMethod selector ~= selector ifTrue: [self assert: (self methodClassAssociationOf: newMethod) = objectMemory nilObject. newCogMethod := cogit cog: newMethod selector: selector. newCogMethod ifNotNil: [cogMethod := newCogMethod]]. (cogMethod selector = selector and: [canLinkCacheTag]) ifTrue: [cogit linkSendAt: (stackPages longAt: stackPointer) in: (self mframeHomeMethod: framePointer) to: cogMethod offset: cogit dynSuperEntryOffset receiver: rcvr]. instructionPointer := self popStack. self executeNewMethod. self assert: false "NOTREACHED"]. instructionPointer := self popStack. ^self interpretMethodFromMachineCode "NOTREACHED"! Item was changed: ----- Method: CogVMSimulator>>ceMNUFromPICMNUMethod:receiver: (in category 'trampolines') ----- ceMNUFromPICMNUMethod: aMethodObj receiver: rcvr + "self halt." - self halt. ^super ceMNUFromPICMNUMethod: aMethodObj receiver: rcvr! Item was removed: - ----- Method: CogVMSimulator>>printCurrPageFrames (in category 'debug printing') ----- - printCurrPageFrames - self printFrameAndCallers: localFP SP: localSP! Item was changed: ----- Method: Cogit>>closedPICRefersToUnmarkedObject: (in category 'garbage collection') ----- closedPICRefersToUnmarkedObject: cPIC "Answer if the ClosedPIC refers to any unmarked objects or freed/freeable target methods, applying markAndTraceOrFreeCogMethod:firstVisit: to those targets to determine if freed/freeable." <var: #cPIC type: #'CogMethod *'> | pc offsetToLiteral object entryPoint targetMethod | <var: #targetMethod type: #'CogMethod *'> (objectMemory isMarked: cPIC selector) ifFalse: [^true]. pc := cPIC asInteger + firstCPICCaseOffset. "First jump is unconditional; subsequent ones are conditional" offsetToLiteral := backEnd jumpLongByteSize. 1 to: cPIC cPICNumCases do: [:i| object := backEnd literalBeforeFollowingAddress: pc - offsetToLiteral - backEnd loadLiteralByteSize. ((objectRepresentation couldBeObject: object) and: [(objectMemory isMarked: object) not]) ifTrue: [^true]. object := backEnd literalBeforeFollowingAddress: pc - offsetToLiteral. ((objectRepresentation couldBeObject: object) + and: [(objectMemory isMarked: object) not]) ifTrue: - and: [(coInterpreter isMarked: object) not]) ifTrue: [^true]. entryPoint := backEnd jumpLongTargetBeforeFollowingAddress: pc. "Find target from jump. Ignore jumps to the interpret and MNU calls within this PIC" (entryPoint asUnsignedInteger < cPIC asUnsignedInteger or: [entryPoint asUnsignedInteger > (cPIC asUnsignedInteger + cPIC blockSize) asUnsignedInteger]) ifTrue: [targetMethod := self cCoerceSimple: entryPoint - cmNoCheckEntryOffset to: #'CogMethod *'. self assert: (targetMethod cmType = CMMethod or: [targetMethod cmType = CMFree]). (self markAndTraceOrFreeCogMethod: targetMethod firstVisit: targetMethod asUnsignedInteger > pc asUnsignedInteger) ifTrue: [^true]]. offsetToLiteral := backEnd jumpLongConditionalByteSize. pc := pc + cPICCaseSize]. ^false! Item was added: + ----- Method: NewObjectMemory>>checkOkayOop: (in category 'debug support') ----- + checkOkayOop: oop + "Verify that the given oop is legitimate. Check address, header, and size but not class. + Answer true if OK. Otherwise print reason and answer false." + + <api> + <var: #oop type: #usqInt> + | sz type fmt unusedBit | + + "address and size checks" + (self isIntegerObject: oop) ifTrue: [ ^true ]. + (oop >= self startOfMemory and: [oop < freeStart]) + ifFalse: [ self print: 'oop is not a valid address'; cr. ^false ]. + ((oop \\ BytesPerWord) = 0) + ifFalse: [ self print: 'oop is not a word-aligned address'; cr. ^false ]. + sz := self sizeBitsOf: oop. + (oop + sz) < freeStart + ifFalse: [ self print: 'oop size would make it extend beyond the end of memory'; cr. ^false ]. + + "header type checks" + type := self headerType: oop. + type = HeaderTypeFree + ifTrue: [ self print: 'oop is a free chunk, not an object'; cr. ^false ]. + type = HeaderTypeShort ifTrue: [ + (self compactClassIndexOf: oop) = 0 + ifTrue: [ self print: 'cannot have zero compact class field in a short header'; cr. ^false ]. + ]. + type = HeaderTypeClass ifTrue: [ + ((oop >= BytesPerWord) and: [(self headerType: oop - BytesPerWord) = type]) + ifFalse: [ self print: 'class header word has wrong type'; cr. ^false ]. + ]. + type = HeaderTypeSizeAndClass ifTrue: [ + ((oop >= (BytesPerWord*2)) and: + [(self headerType: oop - (BytesPerWord*2)) = type and: + [(self headerType: oop - BytesPerWord) = type]]) + ifFalse: [ self print: 'class header word has wrong type'; cr. ^false ]. + ]. + + "format check" + fmt := self formatOf: oop. + ((fmt = 5) | (fmt = 7)) + ifTrue: [ self print: 'oop has an unknown format type'; cr. ^false ]. + + "mark and root bit checks" + unusedBit := 16r20000000. + BytesPerWord = 8 + ifTrue: + [unusedBit := unusedBit << 16. + unusedBit := unusedBit << 16]. + ((self longAt: oop) bitAnd: unusedBit) = 0 + ifFalse: [ self print: 'unused header bit 30 is set; should be zero'; cr. ^false ]. + "xxx + ((self longAt: oop) bitAnd: MarkBit) = 0 + ifFalse: [ self print: 'mark bit should not be set except during GC' ]. + xxx" + (((self longAt: oop) bitAnd: RootBit) = 1 and: + [oop >= youngStart]) + ifTrue: [ self print: 'root bit is set in a young object'; cr. ^false ]. + ^true + ! Item was changed: ----- Method: NewObjectMemory>>okayOop: (in category 'debug support') ----- okayOop: signedOop "Verify that the given oop is legitimate. Check address, header, and size but not class." | sz type fmt unusedBit oop | + <var: #oop type: #usqInt> + oop := self cCoerce: signedOop to: #usqInt. - <var: #oop type: 'usqInt'> - oop := self cCoerce: signedOop to: 'usqInt'. "address and size checks" (self isIntegerObject: oop) ifTrue: [ ^true ]. + (oop >= self startOfMemory and: [oop < freeStart]) - (oop < freeStart) ifFalse: [ self error: 'oop is not a valid address'. ^false ]. ((oop \\ BytesPerWord) = 0) ifFalse: [ self error: 'oop is not a word-aligned address'. ^false ]. sz := self sizeBitsOf: oop. + (oop + sz) < freeStart - (oop + sz) < endOfMemory ifFalse: [ self error: 'oop size would make it extend beyond the end of memory'. ^false ]. "header type checks" type := self headerType: oop. type = HeaderTypeFree ifTrue: [ self error: 'oop is a free chunk, not an object'. ^false ]. type = HeaderTypeShort ifTrue: [ (self compactClassIndexOf: oop) = 0 ifTrue: [ self error: 'cannot have zero compact class field in a short header'. ^false ]. ]. type = HeaderTypeClass ifTrue: [ ((oop >= BytesPerWord) and: [(self headerType: oop - BytesPerWord) = type]) ifFalse: [ self error: 'class header word has wrong type'. ^false ]. ]. type = HeaderTypeSizeAndClass ifTrue: [ ((oop >= (BytesPerWord*2)) and: [(self headerType: oop - (BytesPerWord*2)) = type and: [(self headerType: oop - BytesPerWord) = type]]) ifFalse: [ self error: 'class header word has wrong type'. ^false ]. ]. "format check" fmt := self formatOf: oop. ((fmt = 5) | (fmt = 7)) ifTrue: [ self error: 'oop has an unknown format type'. ^false ]. "mark and root bit checks" unusedBit := 16r20000000. BytesPerWord = 8 ifTrue: [unusedBit := unusedBit << 16. unusedBit := unusedBit << 16]. ((self longAt: oop) bitAnd: unusedBit) = 0 ifFalse: [ self error: 'unused header bit 30 is set; should be zero'. ^false ]. "xxx ((self longAt: oop) bitAnd: MarkBit) = 0 ifFalse: [ self error: 'mark bit should not be set except during GC' ]. xxx" (((self longAt: oop) bitAnd: RootBit) = 1 and: [oop >= youngStart]) ifTrue: [ self error: 'root bit is set in a young object'. ^false ]. ^true ! Item was added: + ----- Method: ObjectMemory>>checkOkayOop: (in category 'debug support') ----- + checkOkayOop: oop + "Verify that the given oop is legitimate. Check address, header, and size but not class. + Answer true if OK. Otherwise print reason and answer false." + + <api> + <var: #oop type: #usqInt> + | sz type fmt unusedBit | + + "address and size checks" + (self isIntegerObject: oop) ifTrue: [ ^true ]. + (oop >= self startOfMemory and: [oop < endOfMemory]) + ifFalse: [ self print: 'oop is not a valid address'; cr. ^false ]. + ((oop \\ BytesPerWord) = 0) + ifFalse: [ self print: 'oop is not a word-aligned address'; cr. ^false ]. + sz := self sizeBitsOf: oop. + (oop + sz) < endOfMemory + ifFalse: [ self print: 'oop size would make it extend beyond the end of memory'; cr. ^false ]. + + "header type checks" + type := self headerType: oop. + type = HeaderTypeFree + ifTrue: [ self print: 'oop is a free chunk, not an object'; cr. ^false ]. + type = HeaderTypeShort ifTrue: [ + (self compactClassIndexOf: oop) = 0 + ifTrue: [ self print: 'cannot have zero compact class field in a short header'; cr. ^false ]. + ]. + type = HeaderTypeClass ifTrue: [ + ((oop >= BytesPerWord) and: [(self headerType: oop - BytesPerWord) = type]) + ifFalse: [ self print: 'class header word has wrong type'; cr. ^false ]. + ]. + type = HeaderTypeSizeAndClass ifTrue: [ + ((oop >= (BytesPerWord*2)) and: + [(self headerType: oop - (BytesPerWord*2)) = type and: + [(self headerType: oop - BytesPerWord) = type]]) + ifFalse: [ self print: 'class header word has wrong type'; cr. ^false ]. + ]. + + "format check" + fmt := self formatOf: oop. + ((fmt = 5) | (fmt = 7)) + ifTrue: [ self print: 'oop has an unknown format type'; cr. ^false ]. + + "mark and root bit checks" + unusedBit := 16r20000000. + BytesPerWord = 8 + ifTrue: + [unusedBit := unusedBit << 16. + unusedBit := unusedBit << 16]. + ((self longAt: oop) bitAnd: unusedBit) = 0 + ifFalse: [ self print: 'unused header bit 30 is set; should be zero'; cr. ^false ]. + "xxx + ((self longAt: oop) bitAnd: MarkBit) = 0 + ifFalse: [ self print: 'mark bit should not be set except during GC' ]. + xxx" + (((self longAt: oop) bitAnd: RootBit) = 1 and: + [oop >= youngStart]) + ifTrue: [ self print: 'root bit is set in a young object'; cr. ^false ]. + ^true + ! Item was changed: ----- Method: ObjectMemory>>okayOop: (in category 'debug support') ----- okayOop: signedOop "Verify that the given oop is legitimate. Check address, header, and size but not class." | sz type fmt unusedBit oop | + <var: #oop type: #usqInt> + oop := self cCoerce: signedOop to: #usqInt. - <var: #oop type: 'usqInt'> - oop := self cCoerce: signedOop to: 'usqInt'. "address and size checks" (self isIntegerObject: oop) ifTrue: [ ^true ]. + (oop >= self startOfMemory and: [oop < endOfMemory]) - (oop < endOfMemory) ifFalse: [ self error: 'oop is not a valid address'. ^false ]. ((oop \\ BytesPerWord) = 0) ifFalse: [ self error: 'oop is not a word-aligned address'. ^false ]. sz := self sizeBitsOf: oop. (oop + sz) < endOfMemory ifFalse: [ self error: 'oop size would make it extend beyond the end of memory'. ^false ]. "header type checks" type := self headerType: oop. type = HeaderTypeFree ifTrue: [ self error: 'oop is a free chunk, not an object'. ^false ]. type = HeaderTypeShort ifTrue: [ (self compactClassIndexOf: oop) = 0 ifTrue: [ self error: 'cannot have zero compact class field in a short header'. ^false ]. ]. type = HeaderTypeClass ifTrue: [ ((oop >= BytesPerWord) and: [(self headerType: oop - BytesPerWord) = type]) ifFalse: [ self error: 'class header word has wrong type'. ^false ]. ]. type = HeaderTypeSizeAndClass ifTrue: [ ((oop >= (BytesPerWord*2)) and: [(self headerType: oop - (BytesPerWord*2)) = type and: [(self headerType: oop - BytesPerWord) = type]]) ifFalse: [ self error: 'class header word has wrong type'. ^false ]. ]. "format check" fmt := self formatOf: oop. ((fmt = 5) | (fmt = 7)) ifTrue: [ self error: 'oop has an unknown format type'. ^false ]. "mark and root bit checks" unusedBit := 16r20000000. BytesPerWord = 8 ifTrue: [unusedBit := unusedBit << 16. unusedBit := unusedBit << 16]. ((self longAt: oop) bitAnd: unusedBit) = 0 ifFalse: [ self error: 'unused header bit 30 is set; should be zero'. ^false ]. "xxx ((self longAt: oop) bitAnd: MarkBit) = 0 ifFalse: [ self error: 'mark bit should not be set except during GC' ]. xxx" (((self longAt: oop) bitAnd: RootBit) = 1 and: [oop >= youngStart]) ifTrue: [ self error: 'root bit is set in a young object'. ^false ]. ^true ! Item was added: + ----- Method: StackInterpreter>>checkOopHasOkayClass: (in category 'debug support') ----- + checkOopHasOkayClass: oop + "Attempt to verify that the given oop has a reasonable behavior. The class must be a valid, non-integer oop and must not be nilObj. It must be a pointers object with three or more fields. Finally, the instance specification field of the behavior must match that of the instance. If OK answer true. If not, print reason and answer false." + + <api> + <var: #oop type: #usqInt> + | oopClass formatMask behaviorFormatBits oopFormatBits | + <var: #oopClass type: #usqInt> + + (objectMemory checkOkayOop: oop) ifFalse: + [^false]. + oopClass := self cCoerce: (objectMemory fetchClassOf: oop) to: #usqInt. + + (objectMemory isIntegerObject: oopClass) ifTrue: + [self print: 'a SmallInteger is not a valid class or behavior'; cr. ^false]. + (objectMemory okayOop: oopClass) ifFalse: + [self print: 'class oop is not ok'; cr. ^false]. + ((objectMemory isPointers: oopClass) and: [(objectMemory lengthOf: oopClass) >= 3]) ifFalse: + [self print: 'a class (behavior) must be a pointers object of size >= 3'; cr. ^false]. + formatMask := (objectMemory isBytes: oop) + ifTrue: [16rC00] "ignore extra bytes size bits" + ifFalse: [16rF00]. + + behaviorFormatBits := (objectMemory formatOfClass: oopClass) bitAnd: formatMask. + oopFormatBits := (objectMemory baseHeader: oop) bitAnd: formatMask. + behaviorFormatBits = oopFormatBits ifFalse: + [self print: 'object and its class (behavior) formats differ'; cr. ^false]. + ^true! Item was changed: ----- Method: StackInterpreter>>findHomeForContext: (in category 'debug printing') ----- findHomeForContext: aContext | closureOrNil | <inline: false> + (self isContext: aContext) ifFalse: + [^nil]. closureOrNil := objectMemory fetchPointer: ClosureIndex ofObject: aContext. closureOrNil = objectMemory nilObject ifTrue: [^aContext]. + (objectMemory fetchClassOf: closureOrNil) ~= (objectMemory splObj: ClassBlockClosure) ifTrue: + [^nil]. ^self findHomeForContext: (objectMemory fetchPointer: ClosureOuterContextIndex ofObject: closureOrNil)! Item was changed: ----- Method: StackInterpreter>>frameStackedReceiver:numArgs: (in category 'frame access') ----- frameStackedReceiver: theFP numArgs: numArgs "Answer the stacked receiver given the frame's argument count. The receiver of a message send or the closure of a block activation is always on the stack above any arguments and the frame itself. See the diagram in StackInterpreter class>>initializeFrameIndices." <inline: true> <var: #theFP type: #'char *'> + ^stackPages longAt: theFP + (self frameStackedReceiverOffsetNumArgs: numArgs)! - ^stackPages longAt: theFP + FoxCallerSavedIP + BytesPerWord + (numArgs << ShiftForWord)! Item was changed: ----- Method: StackInterpreter>>frameStackedReceiverOffset: (in category 'frame access') ----- frameStackedReceiverOffset: theFP + "Answer the offset in bytes from the the frame pointer to its stacked receiver. - "Answer the offset in bytes from the a frame pointer to its stacked receiver. The receiver of a message send or the closure of a block activation is always on the stack above any arguments and the frame itself. See the diagram in StackInterpreter class>>initializeFrameIndices." <inline: true> <var: #theFP type: #'char *'> + ^self frameStackedReceiverOffsetNumArgs: (self frameNumArgs: theFP)! - ^FoxCallerSavedIP + BytesPerWord + ((self frameNumArgs: theFP) << ShiftForWord)! Item was removed: - ----- Method: StackInterpreter>>frameStackedReceiverOffset:numArgs: (in category 'frame access') ----- - frameStackedReceiverOffset: theFP numArgs: numArgs - "Answer the offset in bytes from the a frame pointer to its stacked receiver. - The receiver of a message send or the closure of a block activation is - always on the stack above any arguments and the frame itself. See the - diagram in StackInterpreter class>>initializeFrameIndices." - <inline: true> - <var: #theFP type: #'char *'> - ^FoxCallerSavedIP + BytesPerWord + (numArgs << ShiftForWord)! Item was added: + ----- Method: StackInterpreter>>frameStackedReceiverOffsetNumArgs: (in category 'frame access') ----- + frameStackedReceiverOffsetNumArgs: numArgs + "Answer the offset in bytes from the a frame pointer to its stacked receiver, + given the argument count. The receiver of a message send or the closure of + a block activation is always on the stack above any arguments and the frame + itself. See the diagram in StackInterpreter class>>initializeFrameIndices." + <inline: true> + ^FoxCallerSavedIP + BytesPerWord + (numArgs << ShiftForWord)! Item was changed: ----- Method: StackInterpreter>>printFrameAndCallers:SP: (in category 'debug printing') ----- printFrameAndCallers: theFP SP: theSP + self printFrameAndCallers: theFP SP: theSP short: false! - <inline: false> - <var: #theFP type: #'char *'> - <var: #theSP type: #'char *'> - (stackPages couldBeFramePointer: theFP) ifFalse: [^nil]. - (self isBaseFrame: theFP) ifFalse: - [self printFrameAndCallers: (self frameCallerFP: theFP) - SP: (self frameCallerSP: theFP)]. - self cr. - self printFrame: theFP WithSP: theSP! Item was added: + ----- Method: StackInterpreter>>printFrameAndCallers:SP:short: (in category 'debug printing') ----- + printFrameAndCallers: theFP SP: theSP short: printShort + <inline: false> + <var: #theFP type: #'char *'> + <var: #theSP type: #'char *'> + (stackPages couldBeFramePointer: theFP) ifFalse: [^nil]. + (self isBaseFrame: theFP) ifFalse: + [self printFrameAndCallers: (self frameCallerFP: theFP) + SP: (self frameCallerSP: theFP) + short: printShort]. + self cr. + printShort + ifTrue: [self shortPrintFrame: theFP] + ifFalse: [self printFrame: theFP WithSP: theSP]! Item was changed: ----- Method: StackInterpreter>>printFramesInPage: (in category 'debug printing') ----- printFramesInPage: thePage + <export: true> "use export: not api, so it won't be written to cointerp.h. cogit.c is unaware of StackPage" <inline: false> <var: #thePage type: #'StackPage *'> + self printFrameAndCallers: thePage headFP SP: thePage headSP short: false! - self printFrameAndCallers: thePage headFP SP: thePage headSP! Item was changed: ----- Method: StackInterpreter>>shortPrintContext: (in category 'debug printing') ----- shortPrintContext: aContext + | theFP | - | home theFP | <inline: false> <var: #theFP type: #'char *'> (self isContext: aContext) ifFalse: [self printHex: aContext; print: ' is not a context'; cr. ^nil]. - home := self findHomeForContext: aContext. self printHex: aContext. (self isMarriedOrWidowedContext: aContext) ifTrue: [(self checkIsStillMarriedContext: aContext currentFP: framePointer) ifTrue: [(self isMachineCodeFrame: (theFP := self frameOfMarriedContext: aContext)) ifTrue: [self print: ' M ('] ifFalse: [self print: ' I (']. self printHex: theFP; print: ') '] ifFalse: [self print: ' w ']] ifFalse: [self print: ' s ']. + (self findHomeForContext: aContext) + ifNil: [self print: ' BOGUS CONTEXT (can''t determine home)'] + ifNotNil: + [:home| + self printActivationNameFor: (objectMemory fetchPointer: MethodIndex ofObject: aContext) + receiver: (home isNil + ifTrue: [objectMemory nilObject] + ifFalse: [objectMemory fetchPointer: ReceiverIndex ofObject: home]) - self printActivationNameFor: (objectMemory fetchPointer: MethodIndex ofObject: home) - receiver: (objectMemory fetchPointer: ReceiverIndex ofObject: home) isBlock: home ~= aContext + firstTemporary: (objectMemory fetchPointer: 0 + CtxtTempFrameStart ofObject: home)]. - firstTemporary: (objectMemory fetchPointer: 0 + CtxtTempFrameStart ofObject: home). self cr! Item was changed: ----- Method: StackInterpreter>>shortPrintFramesInPage: (in category 'debug printing') ----- shortPrintFramesInPage: thePage + <export: true> "use export: not api, so it won't be written to cointerp.h. cogit.c is unaware of StackPage" <inline: false> <var: #thePage type: #'StackPage *'> + self printFrameAndCallers: thePage headFP SP: thePage headSP short: true! - self shortPrintFrameAndCallers: thePage headFP! Item was removed: - ----- Method: StackInterpreterSimulator>>printCurrPageFrames (in category 'debug printing') ----- - printCurrPageFrames - self printFrameAndCallers: localFP SP: localSP!

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev March 2012