Vm-dev April 2008

vm-dev@lists.squeakfoundation.org

16 participants
24 discussions

FFI 64-bit updates - progress report
by David T. Lewis 20 Apr '08

20 Apr '08

I now have FFI working on a 64-bit AMD Linux system for both 32-bit and 64-bit images. All unit tests pass, including some new tests that I added to provide more coverage of the FFI primitives. The X11Display examples also work. This is using the "cdecl" interface in platforms/unix/plugins/SqueakFFIPrims/lib-anyffi.c. The next step will be to move the changes to a 32 bit Linux system and do the updates to platforms/unix/plugins/SqueakFFIPrims/x86-sysv.c. This is work in progress, but I'm attaching a snapshot of the work that I have done so far. There are changes to the VM, the FFI plugin, and to the FFI support classes in the image (note: after applying the changes, classes with cdecl pragmas must be recompiled). More to follow, Dave

1 2

Fwd: [squeak-dev] Image crashes just after start-up
by John M McIntosh 17 Apr '08

17 Apr '08

Begin forwarded message: > From: David Röthlisberger <squeak(a)c3com.ch> > Date: April 17, 2008 2:56:19 AM PDT (CA) > To: johnmci(a)smalltalkconsulting.com > Subject: Re: [squeak-dev] Image crashes just after start-up > > Hi John, > >> If you can setup a test where it halts before giving the bad data >> to FormCanvas >> drawString:from:to:in:font: >> then we can see how to guard against that condition and fail the >> primitive call. > > I tried this, but it's actually pretty difficult... Stepping through > the application somehow "changes" the behavior in the sense that > this error either happens immediately when I inspect the corrupt > data in the debugger, or it works well, the data seems to be just > fine, but nonetheless it crashes when the data gets accessed again > later on, even though it seems to be okay in the debugger. > > I found a workaround, so I changed the application to use another > piece of code providing the same functionality. This would work, but > of course the problem is not really solved. > >> Also you should make a mantis report > > I did so, including a link to the problematic image: > http://bugs.squeak.org/view.php?id=7022 > > Cheers, > David > -- = = = ======================================================================== John M. McIntosh <johnmci(a)smalltalkconsulting.com> Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================

1 0

Fwd: [squeak-dev] Image crashes just after start-up
by John M McIntosh 16 Apr '08

16 Apr '08

fyi Begin forwarded message: > From: David Röthlisberger <squeak(a)c3com.ch> > Date: April 16, 2008 1:15:57 AM PDT (CA) > To: johnmci(a)smalltalkconsulting.com > Subject: Re: [squeak-dev] Image crashes just after start-up > > Hi John, > >> Ok, so does this VM work with a base image, or it only a problem >> with your image? > > The VM works well with a normal image, it's only a problem with this > specific image. > And even for that image it worked well for a couple of weeks. Then I > implemented an > application drawing a graph on a BalloonCanvas. Whenever I execute > this application, > the image crashes. Sometimes directly after starting the > application, but in most > cases I can use the app without problems, but when I then save the > image after having > executed one specific feature, it crashes on image start-up. > I could track the error down to the method FormCanvas >> > drawString:from:to:in:font: > color:. > Whenever this method gets executed with some specific parameters, > then the crash will > happen. It seems to me that the bounds parameter has corrupt data, > eg. numbers not > properly stored. But how this happens I don't know. > > >> usually points to a problem with the VM crashing either because the >> image is corrupt or you >> have corrupted memory via an FFI call, or perhaps passing bad data >> to the sound system. > > I don't use FFI, neither I pass data to the sound system, so > probably this means that > the image is corrupt. I guess that some objects get created that are > corrupt. But why > can this happen? > >> I also note you run Application Enhancer, perhaps turn that off and >> see if it changes anything? > > ok, I turned it off, but this doesn't make a difference. > > > David -- = = = ======================================================================== John M. McIntosh <johnmci(a)smalltalkconsulting.com> Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================

2 2

Better type than void* for primitiveFunctionPointer?
by tim Rowledge 16 Apr '08

16 Apr '08

In Interpreter>declareCVarsIn: we currently have:- declareC: 'void *primitiveFunctionPointer' . "xxxx FIX THIS STUPIDITY xxxx - ikp. What he means is use a better type than void *, apparently - tpr" Anyone care to offer the better type than void* ? Speak now or stop whining later. tim -- tim Rowledge; tim(a)rowledge.org; http://www.rowledge.org/tim Fractured Idiom:- QUE SERA SERF - Life is feudal

4 6

dispatchOn:in: and inlining bytecode routines
by André Wendt 13 Apr '08

13 Apr '08

Hi, as far as I can see, there's nothing special about BytecodeTable in the Interpreter that instructs VMMaker to generate a switch statement that includes: case 112: /* pushReceiverBytecode */ { /* begin fetchNextBytecode */ currentBytecode = byteAtPointer(++localIP); /* begin internalPush: */ longAtPointerput(localSP += BytesPerWord, foo->receiver); } break; However, when I want to dispatch on a code in a subset of bytecodes using my own table (for integrating multiple bytecodes into one), I get case 0: pushReceiverBytecode(); break; instead of an inlined pushReceiverBytecode routine. Needless to say my code doesn't compile because of the pushReceiverBytecode() call. Why does pushReceiverBytecode get inlined when used with BytecodeTable, but not in my own? Thanks, André

2 2

3.10-1 for Unix
by Ian Piumarta 12 Apr '08

12 Apr '08

Hi, 3.10-1 (sources plus binaries for Darwin and Linux) is now up at http://squeakvm.org/unix Release notes are here: http://squeakvm.org/unix/release/ RELEASE_NOTES_3.10-1 If you want them, both the image and VMM config used to generate the core VM are available (look for the obvious file names) in http:// squeakvm.org/unix/release Enjoy! Ian

2 1

Re: [Vm-dev] Re: VM performance discrepancy on Linux and Windows
by Andreas Raab 11 Apr '08

11 Apr '08

Paolo Bonzini wrote: > GCC 3.x should be in the same ballpark as 2.95. I have not found this to be true. I tried multiple times (gcc 3.1, 3.3 and 3.4) and each time the resulting VM was *significantly* slower than 2.95.2. Cheers, - Andreas

3 2

Checking contributors to VM code base
by tim Rowledge 11 Apr '08

11 Apr '08

I'm going to make a pass at establishing the identities of all the contributors to the VM codebase so we can (hopefully) declare it MIT clean on squeaksource etc. If anyone has code that can help to extract method stamp initials from all the VMMaker package code I'd be very happy to receive a copy; I don't want to duplicate effort. There's several dozen mcz packages to scan and there are many, many changeset files from before the existence of VMMaker as well. I think it unlikely that there is anyone not already signed up for the relicensing but better safe than sued. tim -- tim Rowledge; tim(a)rowledge.org; http://www.rowledge.org/tim Useful random insult:- A room temperature IQ.

4 8

Re: [squeak-dev] Re: [Vm-dev] Re: VM performance discrepancy on Linux and Windows
by Yoshiki Ohshima 11 Apr '08

11 Apr '08

At Fri, 11 Apr 2008 09:47:13 +0200, Paolo Bonzini wrote: > > > >> %esi is almost used for IP but use %eax for fetching the next byte, > >> jmp also seems to use %eax so right before it is spilled and the > >> destination address is brought into %eax. > > > > Sounds more like the static register assignments get ignored. > > It does not really get ignored, but the compiler performs more > aggressive live range splitting (because it uses SSA in 4.x so live > range splitting comes from free -- sometimes even if you don't want > it...). OTOH the optimizer is better, which is why sends are faster. > > GCC 3.x should be in the same ballpark as 2.95. Hmm, so simple increment assigns a new value and the range is splitted there? BTW, I got a report that says the result from gcc 3.3 was comparable with 4.1. Now, I still have a feeling that there is some difference between the compiler in http://squeakvm.org/win32/release/Squeak-Win32-Tools-1.2.zip and 2.95.2 I compiled on my Linux as there seems some unexplained discrepancy. Can anybody with dual boot machine do some more testing? (If the machine is green and white cute one, that would be really interesting.) -- Yoshiki

1 0

VM performance discrepancy on Linux and Windows
by Yoshiki Ohshima 11 Apr '08

11 Apr '08

Hello, I had some suspicous for a while but we did a little test with a computer that dual boot Windows XP and (Ubuntu) Linux to run tinyBenchmarks. (The computer happens to be a 1.8GHz Pentium-M Dell laptop.) On Windows, 3.10.6 VM (pre-compiled one on the site) with etoys-dev.image, the result was: 311 million bytecodes/sec, 8.9 million sends/sec On Linux, 3.9-8 VM (pre-compiled one on the site) with etoys-dev.image the result was: 190 million bytecodes/ec, 5.7 million seonds/sec Has any of you been experiencing similar gap? Have anybody looked at the generated code, or has anybody done some experiment recently? -- Yoshiki

3 4

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev April 2008