On 26 Nov 2013, at 11:41, Serge Stinckwich <[hidden email]> wrote: > >>> Anything preventing Pharo and VisualWorks for using the technology also? >>> >> Short answer: Time and money. >> The way we did it requires significant changes to the virtual machine >> (as we believe this is the only way to get a decent performance). Indeed, if >> somebody going to pay for it, everything's possible ;-) > > You should try to ask ESUG about financial support. > I do not think that he talks about this kind of money. But MONEY. Marcus |
In reply to this post by Jan Vrany
I downloaded ST/X this morning and looked around for a couple hours.
Impressive system. And impressive clients list/activities etc. <rant> And wow, Claus is yet another individual with top notch computer chops... Seems that this community has the highest density of high perf brains. I feel humbled indeed.
</rant> What struck me was the speed. Pharo really needs a huge speedup on the UI front. I am using a top of the line desktop system and still Pharo sometimes feels slow (some is due to algorithms - like the finder tool taking ages for a lot of things, but some is due to the UI system).
This is not linked to the VM, as the MVC projects in Squeak are uber fast. Is the Text rewrite going to have an impact on this? On the front of interoperability, I personally have chosen to go the RabbitMQ route, it allows me to wire all kinds of things together with some room for scale.
On the Java front itself, in the TCL community, we do have tclbend and also a package to do the same thing as STX:LIBJAVA. Truth be told, there hasn't been much traction in there. It works fine but that's it. http://www2.tcl.tk/1313 - usage sample http://www2.tcl.tk/14919
It also has fell behind in terms of keeping up with the latest versions of the environment. In PHP, there is the Quercus and Caucho Resin Server that allows to run PHP on top of the JVM and invoke Java from there. Some people run very fast stuff on that, and benefiting from all JEE abilities (JMS, JTA, Clustering, JAAS...) is really nice. http://www.caucho.com/resin-3.1/doc/quercus.xtp
Phil On Tue, Nov 26, 2013 at 12:26 PM, Jan Vrany <[hidden email]> wrote: On 26/11/13 10:40, Serge Stinckwich wrote: |
In reply to this post by Marcus Denker-4
On Tue, Nov 26, 2013 at 12:40 PM, Marcus Denker <[hidden email]> wrote:
> > On 26 Nov 2013, at 11:41, Serge Stinckwich <[hidden email]> wrote: >> >>>> Anything preventing Pharo and VisualWorks for using the technology also? >>>> >>> Short answer: Time and money. >>> The way we did it requires significant changes to the virtual machine >>> (as we believe this is the only way to get a decent performance). Indeed, if >>> somebody going to pay for it, everything's possible ;-) >> >> You should try to ask ESUG about financial support. >> > > I do not think that he talks about this kind of money. But MONEY. Ok I understand ;-) BIG MONEY -- Serge Stinckwich UCBN & UMI UMMISCO 209 (IRD/UPMC) Every DSL ends up being Smalltalk http://www.doesnotunderstand.org/ |
To have Java running on top of Pharo (which means on top of Cog), you would need to have as they did in STX, a JIT compatible with java bytecode, an interpreter for java bytecode, runtime objects such as JavaContext, ... It is at least 1 year of work for an expert in virtual machines (I think there's mainly 1 expert in the Cog VM but there are several on java VM). Therefore this project requires several hundred thousands dollars (considering you can find a virtual machine expert willing to do it). A bit too much for ESUG in my opinion.
2013/11/26 Serge Stinckwich <[hidden email]>
|
Here is the video of his Smalltalk 2013 talk for the ones interested: 2013/11/26 Clément Bera <[hidden email]>
|
In reply to this post by philippeback
Try to switch back to bitmap DejaVu font and I'm sure you'll notice a difference. 2013/11/26 [hidden email] <[hidden email]>
|
In reply to this post by philippeback
2013/11/26 [hidden email] <[hidden email]>:
> Pharo really needs a huge speedup on the UI front. I am using a top of the > line desktop system and still Pharo sometimes feels slow (some is due to > algorithms - like the finder tool taking ages for a lot of things, but some > is due to the UI system). > > This is not linked to the VM, as the MVC projects in Squeak are uber fast. > > Is the Text rewrite going to have an impact on this? It doesn't feel slow, it is slow. Perfectly usable, though. I even feel it was faster before. But I think the problem comes from the "rendering" side, and not because of "not UI" operations running in the main thread. Regards, |
In reply to this post by philippeback
Yes but need people to try.
This was on my todo for a topics for students. On Nov 25, 2013, at 11:14 PM, [hidden email] wrote:
|
In reply to this post by philippeback
On Nov 26, 2013, at 1:21 PM, [hidden email] wrote:
This is why the work of Eliot on Spur are also important. What is also important is that more people can work on the VM for cleaning it so that we get a documented and good vehicule. And this will take time. We are working on a prototype to get forced to think about VM and learn. But this prototype will not replace Cog in the near future. So we should be happy that cog is there and improving. Stef
|
FWIW, I'd love to have a working Pharo bytecode interpreter that works. VMMaker currently doesn't have one it seems (earlier experiments didn't worked for me). I am very interested with the VM, read the blue book, understand the primitives, can somewhat read bytecode but what is needed now is the ability to run/debug a VM inside Pharo itself. GDB'ing is okay but a pain in the ass to understand what's going on.
Also read the Tour of the OE of Tim Rowledge and Porting the VM etc. Also looked at the VMMaker package (Interpreter and Object Memory) + Slang. Now, getting an working interpreter would help me reach the next step. I am not talking about the Stack interpreter, but the plain Interpreter.
Any plans? Phil --- Philippe Back Dramatic Performance Improvements
Mob: +32(0) 478 650 140 | Fax: +32 (0) 70 408 027
Blog: http://philippeback.be | Twitter: @philippeback
High Octane SPRL rue cour Boisacq 101 | 1301 Bierges | Belgium
Pharo Consortium Member - http://consortium.pharo.org/ Featured on the Software Process and Measurement Cast - http://spamcast.libsyn.com Sparx Systems Enterprise Architect and Ability Engineering EADocX Value Added Reseller On Tue, Nov 26, 2013 at 9:33 PM, Stéphane Ducasse <[hidden email]> wrote:
|
In reply to this post by Esteban A. Maringolo
I also noticed that Sean does a replacement of the small size fonts with large size fonts in a specific way to speed up rendering of larger font (In his Configuration that does Preferences). I've got eyesight issues, so I am also using larger fonts. That doesn't help with speed indeed. But the Paragraph rendering code is also a culprit I think, as it endlessly recreates objects all the time. Morphs with step methods and a short stepping time are pretty fast. I've a educational game on the iPad that uses this and it is pretty nice in terms of rendering speed (dragging and dropping formula pieces around).
--- Philippe Back Dramatic Performance Improvements Mob: +32(0) 478 650 140 | Fax: +32 (0) 70 408 027
Blog: http://philippeback.be | Twitter: @philippeback
High Octane SPRL rue cour Boisacq 101 | 1301 Bierges | Belgium
Pharo Consortium Member - http://consortium.pharo.org/ Featured on the Software Process and Measurement Cast - http://spamcast.libsyn.com Sparx Systems Enterprise Architect and Ability Engineering EADocX Value Added Reseller On Tue, Nov 26, 2013 at 8:27 PM, Esteban A. Maringolo <[hidden email]> wrote: 2013/11/26 [hidden email] <[hidden email]>: |
On Nov 26, 2013, at 9:53 PM, [hidden email] wrote:
Sadly he was working well with fernando. We will see if we can help him getting back to energy. Stef
|
In reply to this post by philippeback
On Tue, Nov 26, 2013 at 12:49 PM, [hidden email] <[hidden email]> wrote:
David Lewis and I want to see the Cog branch and the VMMaker proper merged and I definitely want the standard Interpreter to be married to Spur. But I have no cycles to do this, and I don't think David has many either. Volunteers welcome.
best, Eliot
|
In reply to this post by Stéphane Ducasse
That's the issue with such an immense amount of work to do and brainpower to use.
Lots of rest, getting away from everything for while definitely helps. Just an advice from the burned-out anonymous...
Phil On Tue, Nov 26, 2013 at 10:04 PM, Stéphane Ducasse <[hidden email]> wrote:
|
Administrator
|
In reply to this post by Jan Vrany
Jan:
Can you give us a bit of history of STX:LIBJAVA? Even if it was just for fun what was the rational? What killer app might there be? What business models are there to pursue? How much effort was used and who funded your group? Can you debug Java like all the power of Smalltalk debugger? Congratulations again. How about an easier name like StJ? or SXJ? All the best, Aik-Siong Koh |
In reply to this post by philippeback
On Tue, Nov 26, 2013 at 5:49 PM, [hidden email] <[hidden email]> wrote:
I been writing a lot about that, at least as much as I could.
Yes, the simulator does not run these days.
Ok, for debugging the VM with gdb I have: http://marianopeck.wordpress.com/tag/gdb/
If you want to learn, besides all those mentioned books, you can start with http://marianopeck.wordpress.com/2011/03/31/journey-through-the-vm/
that is a sequence of a lots of posts related to the VM. Cheers,
Mariano http://marianopeck.wordpress.com |
On Tue, Nov 26, 2013 at 5:20 PM, Mariano Martinez Peck <[hidden email]> wrote:
That's simply not true. I use the simulator *every day*. It is vital for the development of Spur. Starting it up these days is as simple as (| vm | vm := StackInterpreterSimulator newWithOptions: #().
vm openOn: '/Users/eliot/Cog/startreader.image'. vm openAsMorph; run) something slightly more complex would be (| vm | vm := CogVMSimulator newWithOptions: #(ObjectMemory Spur32BitCoMemoryManager
Cogit StackToRegisterMappingCogit). vm desiredNumStackPages: 8. vm objectMemory setCheckForLeaks: 1. vm openOn: '/Users/eliot/Cog/spurreader.image'.
vm openAsMorph; toggleTranscript; halt; run) See the class comments in StackInterpreterSimulator and CogVMSimulator.
best, Eliot
|
In reply to this post by Eliot Miranda-2
On Tue, Nov 26, 2013 at 01:30:40PM -0800, Eliot Miranda wrote:
> On Tue, Nov 26, 2013 at 12:49 PM, [hidden email] <[hidden email]>wrote: > > > FWIW, I'd love to have a working Pharo bytecode interpreter that works. > > VMMaker currently doesn't have one it seems (earlier experiments didn't > > worked for me). > > > > I am very interested with the VM, read the blue book, understand the > > primitives, can somewhat read bytecode but what is needed now is the > > ability to run/debug a VM inside Pharo itself. GDB'ing is okay but a pain > > in the ass to understand what's going on. > > > > Also read the Tour of the OE of Tim Rowledge and Porting the VM etc. > > > > Also looked at the VMMaker package (Interpreter and Object Memory) + Slang. > > > > Now, getting an working interpreter would help me reach the next step. I > > am not talking about the Stack interpreter, but the plain Interpreter. > > > > Any plans? > > > > David Lewis and I want to see the Cog branch and the VMMaker proper merged > and I definitely want the standard Interpreter to be married to Spur. But > I have no cycles to do this, and I don't think David has many either. > Volunteers welcome. Fully agree :-) With respect to the interpreter simulator, the simulators tend to get bit rotted when not used, but I think that overall they are in reasonable shape. Granted that we currently have to fumble around with multiple code bases, but it's fair to say that if you want to run a Cog/StackInterpreter/Spur simulator, you can use the appropriate classes in the oscog branch (after all, that is what Eliot is using for his active development, and it's quite unlikely that he could do this without a working simulator). And if you want to run an image using the classic interpreter, you should use the interpreter simulator in the "trunk" VMMaker branch. I realize this may be a bit confusing, but as Eliot says there are only so may free cycles available, so if someone wants to help ... I just tried loading an image into the ("trunk") InterpreterSimulator and found a problem in loading an image that had been saved from Cog. This would be a problem if you wanted to load a Pharo image into the InterpreterSimulator to try running bytecodes using a simple interpreter. The fix is in VMMaker-dtl.330 in the source.squeak.org/VMMaker repository. Hopefully it works for you now, please give it a try. Dave |
In reply to this post by Mariano Martinez Peck
Thanks. I've read the all.
Especially when I wanted to get the iPad version running and customized some things in there and ran into CMake bugs at the time. Thanks for writing them, they helped!
Phil On Wed, Nov 27, 2013 at 2:20 AM, Mariano Martinez Peck <[hidden email]> wrote:
|
In reply to this post by David T. Lewis
Hi Dave, Eliot,
Dave, Yes, that is the problem I was running into. I'll check that out. I could get a bit further by hacking some version numbers if I remember well but then got stuck again. I do not remember why exactly. It felt a bit like crossing the desert with a single bottle of water :-)
Eliot, Thanks for the pointers. Maybe would it be great to have a Google Hangout or two with the people doing the VM work (you both, Clement, Esteban, ...) and go through what matters in the process of dealing with the simulators.
Phil On Wed, Nov 27, 2013 at 4:01 AM, David T. Lewis <[hidden email]> wrote: On Tue, Nov 26, 2013 at 01:30:40PM -0800, Eliot Miranda wrote: |
Free forum by Nabble | Edit this page |