VM-beginners April 2012

vm-beginners@lists.squeakfoundation.org

8 participants
15 discussions

Unix VM building - Monticello can't find DBus-Plugin-dtl.33(bf.32).mcd
by Paul DeBruicker 05 May '12

05 May '12

After updating the Squeak One-click to trunk and loading the update-dtl.11 package from the VMMaker repository on source.squeak.org When I run VMMaker updateFromServer the process eventually trips up when MCHttpRepository>>#loadVersionFromFileNamed: attempts to load DBus-Plugin-dtl.33(bf.32).mcd If I download DBus-Plugin-dtl.33.mcz and stick it in my package-cache the error continues to happen. I'll try it without updating to trunk.

1 1

Unix VM building instructions point to Squeaksource VMMaker repository.
by Paul DeBruicker 30 Apr '12

30 Apr '12

The instructions here: http://squeakvm.org/unix/devel.html and http://wiki.squeak.org/squeak/6177 still point to the squeaksource monticello repo here: MCHttpRepository location: 'http://www.squeaksource.com/VMMaker' user: '' password: '' while they should point to: MCHttpRepository location: 'http://source.squeak.org/VMMaker' user: '' password: '' I'm going to attempt following them on Ubuntu amd64 (12.04) and will report other things that come up. Thanks for posting the detailed instructions Paul

2 1

Combining C and Objective-C headers
by Chris Cunnington 30 Apr '12

30 Apr '12

Actually, I think I misconceived this problem. This isn't about headers; it's about compilation and cross compilation. It's cmake that allows C and Objective-C files to be linked. Thanks, Chris

1 0

Combining C and Objective-C headers
by Chris Cunnington 30 Apr '12

30 Apr '12

MicroSqueak produces several source files for a VM. At bottom, you'd need three: sq.h interp.c msqMain.c (msqFilePrims.c and msqMiscPrims.c can be stubbed out) msqMain.c is ~1000 lines long. It would be shorter in Objective-C. But then it would no longer be able to use sq.h as a header file, because you cannot mix languages in a header. So, msqMain.m would require an msqMain.h? You'd have: sq.h msqMain.h interp.c msqMain.m And then the two header files could import each other. Is that how Cocoa VMs combine the two languages? Chris

1 0

[ANN] ImageWriter, the SystemTracer's son
by Guillermo Polito 26 Apr '12

26 Apr '12

First public push of the first step of the hazelnut GSoC project :) *In a nutshell* For those who didn't know SystemTracer (the father): It is a tool that traces a graph of objects and writes an image from it. ImageWriter is a refactored, renamed and repackaged version of SystemTracer2 with the following changes: - The image format can be provided as a parameter. - You have to provide an SmalltalkImage object as root to write the image (so in the future we can write bootstrap with this). - The object format was reified in objects. So the code is easier to understand + change, and people can learn how are our objects in memory easily :D. Right now, only 32 bit image format is supported. *Structure* - ImageWriter-Core has the image formats and the image writer in itself - ImageWriter-ObjectFormats contains objects that represent the memory layout of the objects and it's headers. - ImageWriter-Tests has some tests. *Tests done* - Only tested in Pharo-1.4 - There are specific tests written for the size and headers of the objects - Also, I've written a Pharo-1.4 image and then run the tests in the copied image. Only one failure compared to the original :) *Known Bugs :)* - WeakRegistryTest>>#testFinalizationWithOnFork is failing, and when I run it manually it works :/. In the image I use to write, it does not happen, only in the clone. - When opening the new image, the window is too small. It is not respecting the original image size. Don't know yet what is the problem. *How To Download* Gofer it squeaksource3: 'ImageWriter'; package: 'ImageWriter-Core'; package: 'ImageWriter-ObjectFormats'; package: 'ImageWriter-Tests'; load *How to Use it* HzImageWriter new format: HzCogImageFormat new; "or HzInterpreterImageFormat new" imageFileName: 'someFile.image'; onInitializeDo: [ "Some initialization?" ]; "this is optional, probably the same as an startup script, probably nonsense" writeImage: Smalltalk "Yes, we provide the our lovely global Smalltak, but any SmalltalkImage instance could be" *Next steps on this direction* - support to write an isolated image different from the host one (I'll know what I need later, hehe) - class comments! - metacello configuration - better api :) *Links* http://ss3.gemstone.com/ss/ImageWriter.html Cheers, Guille

1 0

Trying to make systemTracer write a CogImage
by Guillermo Polito 24 Apr '12

24 Apr '12

Ok, I'm writing the image header as with the extra words. But my image is kind of not living. It seems the Display is not waking up. When I open my new image with a cog vm, I get an image that does not break, but comes up with a window in black. I sent a SIGUSR1 to the vm Process and get the following. All Smalltalk process stacks (active first): Process 0x776c9498 priority 10 0xbfc2fbf0 M ProcessorScheduler class>idleProcess 2003034484: a(n) ProcessorScheduler class 2006384288 s [] in ProcessorScheduler class>startUp 2008868524 s [] in BlockClosure>newProcess Process 0x77636ce8 priority 60 2003063132 s SmalltalkImage>lowSpaceWatcher 2003240984 s [] in SmalltalkImage>installLowSpaceWatcher 2003695368 s [] in BlockClosure>newProcess Process 0x77638668 priority 50 2003074716 s WeakArray class>finalizationProcess 2003263900 s [] in WeakArray class>restartFinalizationProcess 2003718900 s [] in BlockClosure>newProcess Process 0x779860e4 priority 80 2009271856 s Delay class>handleTimerEvent 2010697732 s Delay class>runTimerEventLoop 2011549844 s [] in Delay class>startTimerEventLoop 2012016284 s [] in BlockClosure>newProcess Process 0x78809050 priority 40 0xbfc2eac4 M [] in Semaphore>critical: 2006475044: a(n) Semaphore 0xbfc2eae4 M BlockClosure>ensure: 2021697252: a(n) BlockClosure 0xbfc2eb04 M Semaphore>critical: 2006475044: a(n) Semaphore 0xbfc2eb28 I Delay>schedule 2021697000: a(n) Delay 0xbfc2eb48 I Delay>wait 2021697000: a(n) Delay 0xbfc2eb70 I WorldState>interCyclePause: 2004530692: a(n) WorldState 0xbfc2eb94 I WorldState>doOneCycleFor: 2004530692: a(n) WorldState 0xbfc2ebb8 I PasteUpMorph>doOneCycle 2003571264: a(n) PasteUpMorph 0xbfc2ebd0 M [] in MorphicUIManager>spawnNewProcess 2006644196: a(n) MorphicUIManager 0xbfc2ebf0 I [] in BlockClosure>newProcess 2021691252: a(n) BlockClosure Most recent primitives relinquishProcessorForMicroseconds: relinquishProcessorForMicroseconds: ... (more relinquishProcessorForMicroseconds:) Now, looking at the primitives called, It seems the idle process is living but not giving any place to the other processes... any ideas? Does cog do something different with processes at system startup? Thanks! Guille

1 0

Re: [Vm-dev] Class checks in the vm
by Guillermo Polito 22 Apr '12

22 Apr '12

On Sat, Apr 21, 2012 at 9:50 PM, stephane ducasse < stephane.ducasse(a)gmail.com> wrote: > > > > > I'm playing to bootstrap a Pharo image, trying to get a new environment > running into my 'host' image, and after that, dump those new objects into a > new image file. > > > > Now, to do that, I'm creating new classes for String, Character... > Which are well known by the vm specialObjectsArray. And trying to > initialize my new image classes, which creates new character objects, and > new string objects which is causing me some troubles :P. > > > > Now, my question is about this check in vmmaker (and some others that > look similar): > > > > Interpreter>>#asciiOfCharacter: characterObj "Returns an integer object" > > <inline: false> > > self assertClassOf: characterObj is: (self splObj: ClassCharacter). > > successFlag > > ifTrue: [^ self fetchPointer: CharacterValueIndex ofObject: > characterObj] > > ifFalse: [^ ConstZero] "in case some code needs an int" > > > Guillermo when you change the specialObjectsArray it does not solve your > problem? > I imagine that check a class pointer is way faster than looking in the > method dict. > Now why when you change the specialObjectArray it does not solve your > problem? > Do you need two? > > yes, or more :). (or none in the ideal case :( )

1 0

[JOB] Young Engineer: Smalltalk VM related
by Marcus Denker 18 Apr '12

18 Apr '12

2 year Young Engineer Position =============================== INRIA RMoD is looking for a full-time young engineer for helping with developing our research infrastructure. Young engineer means someone who finished with a degree (Master or Engineer) not longer than 2 years ago. The engineer position is focussed on virtual machine related work. As researchers, we need a system to explore new ideas with. The job of the engineer is to help us by improving the infrastructure we use for our daily work. What this of course means is that it's kind of a cool job: work in a research lab, but no paper writing, no teaching, no administration. In addition: - live in France (one hour from Paris, 1h20 from London, 35 min from Brussels) - try a lot of good beers (our office is just 10km from the border to Belgium) - french food - Smalltalk! (in the form of Pharo) Lille is an active city but not expensive (lot of students...), centrally located with good connections by train to everywhere (including CDG and Brussels airports). The idea of the job is to strengthen RMoD with regard to virtual machine development and improve our research infrastructure. In research we are often faced with doing experiments that need VM changes. We want to be able to make these changes as easy as possible so researchers can focus on trying out new ideas. In particular we are working on an easy to understand and modify research virtual machine that is implemented in Smalltalk. The goal here is not a fast production VM but instead infrastructure that is easy to learn and extend for research experiments. The job will include working on the research VM as well as helping with Pharo related VM level development: - extend and develop the research VM. - implement a simple GC in Smalltalk for this VM. - help with Pharo VM related development (build system, fixes, improvements) - improvement of the NativeBoost native code backend. Therefore we are looking for a candidate with some prior experience with Smalltalk (Pharo/Squeak) and possibly some background in implementation of dynamic languages. - Knowledge in Dynamic Languages, preferably Smalltalk - Compiler construction - Implementation of dynamic langauges (Virtual Machines) - Good knowledge in low-level programming (C, Asm) - English Duration: 24 months Starting date: 2012 October or later Sallary: 2 527,74 euros brut / month Location: Lille (no remote job possible) Needed: - degree (Master or Engineering, 5 years) - need to have finished the degree max 2 years ago. More about INRIA RMOD: http://rmod.lille.inria.fr INRIA Lille: http://www.inria.fr/lille/ INRIA in General: http://www.inria.fr Engineers at INRIA: http://www.inria.fr/travailler/metiers/accompagnement/dev_applications.en.h… Lille: http://en.wikipedia.org/wiki/Lille http://www.mairie-lille.fr/en http://wikitravel.org/en/Lille Do not hesitate to contact us for more information. Link to this offer: http://rmod.lille.inria.fr/web/pier/blog/2012-04-18 More RMoD Jobs: http://rmod.lille.inria.fr/web/pier/Jobs -- Marcus Denker -- http://marcusdenker.de

1 0

Re: [Vm-dev] Class checks in the vm
by Guillermo Polito 18 Apr '12

18 Apr '12

On Tue, Apr 17, 2012 at 6:13 PM, Eliot Miranda <eliot.miranda(a)gmail.com>wrote: > > Hi Guillermo, > > On Tue, Apr 17, 2012 at 8:33 AM, Guillermo Polito < > guillermopolito(a)gmail.com> wrote: > >> >> Hi! >> >> I'm playing to bootstrap a Pharo image, trying to get a new environment >> running into my 'host' image, and after that, dump those new objects into a >> new image file. >> >> Now, to do that, I'm creating new classes for String, Character... Which >> are well known by the vm specialObjectsArray. And trying to initialize my >> new image classes, which creates new character objects, and new string >> objects which is causing me some troubles :P. >> >> Now, my question is about this check in vmmaker (and some others that >> look similar): >> >> Interpreter>>#asciiOfCharacter: characterObj "Returns an integer object" >> <inline: false> >> *self assertClassOf: characterObj is: (self splObj: ClassCharacter).* >> successFlag >> ifTrue: [^ self fetchPointer: CharacterValueIndex ofObject: >> characterObj] >> ifFalse: [^ ConstZero] "in case some code needs an int" >> >> >> - Can't we put a more intelligent assersion like checking if the object >> responds true to #isCharacter? >> > > The issue is performance. One wants as quick a test here as possible, so > testing the class of the receiver is faster than doing a probe on the > method lookup cache and much faster than possibly having to do a full class > lookup. So realistically the answer is "no, we shouldn't". > > >> - And removing the static check from the vm and adding the dynamic and >> nice #isCharacter check into the image? >> > > This doesn't make a lot of sense. asciiOfCharacter: is being used in the > VM by various primitives (e.g. at:put: on strings) and its use is to fail > if one attempts to store other than a character in a string. So moving the > test to the image means the primitive is not safe. Keep on moving in this > direction and you end up with a system that will crash when one makes an > error instead of raising an error message. Smalltalk is a safe language; > it is extremely important that the system not crash and instead provide > walkbacks that afford diagnosis of errors. > > > The real issue here is that you want more flexibility in the VM while > bootstrapping than one wants or needs in normal use. Instead of making the > VM slow and more general than it need be the right approach is to create a > variant of the VM for your bootstrapping use. You can do this by creating > a subclass of Interpreter which reimplements asciiOfCharacter: (and others > you may need) and testing it in the simulator. You can then use your > flexible VM for the bootstrap and once complete use the normal VM. > I was not totally aware of the performance issues (I knew there were going to be some tradeoffs, but not so critical :) ). Now, for the bootstrap, I'll talk with my mentors to see which should be a nice approach to reach where we want to be. BTW, you were talking about better object representation and inmediate characters, can you give me a pointer of what you mean? (just curious) Thank you!! Guille > > >> Thanks! >> Guille >> >> >> > > > -- > best, > Eliot > > >

1 0

Class checks in the vm
by Guillermo Polito 17 Apr '12

17 Apr '12

Hi! I'm playing to bootstrap a Pharo image, trying to get a new environment running into my 'host' image, and after that, dump those new objects into a new image file. Now, to do that, I'm creating new classes for String, Character... Which are well known by the vm specialObjectsArray. And trying to initialize my new image classes, which creates new character objects, and new string objects which is causing me some troubles :P. Now, my question is about this check in vmmaker (and some others that look similar): Interpreter>>#asciiOfCharacter: characterObj "Returns an integer object" <inline: false> *self assertClassOf: characterObj is: (self splObj: ClassCharacter).* successFlag ifTrue: [^ self fetchPointer: CharacterValueIndex ofObject: characterObj] ifFalse: [^ ConstZero] "in case some code needs an int" - Can't we put a more intelligent assersion like checking if the object responds true to #isCharacter? - And removing the static check from the vm and adding the dynamic and nice #isCharacter check into the image? Thanks! Guille

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

VM-beginners April 2012