Squeak-dev April 2005

squeak-dev@lists.squeakfoundation.org

158 participants
287 discussions

Re: [FIX] WebpageMorphFix
by karl 28 Apr '05

28 Apr '05

Larry Trutter wrote: > Hi, > > I am one of the Janitors. I am attempt to load the fix into the image > - I used 3.9a #6646. I was able to load the fix into the image > successfully but I am having trouble trying to verify if the fix > actually works after loading. > > I create a WebpageMorph from the Squeak menu that lets me create the > object. Everytime I tried to interact with the WebpageMorph, it keeps > giving me all kinds of walkback errors. > > Is there some other way that I can verify that it works in the image? > > thanks, > Larry Trutter > > > karl.ramberg(a)chello.se wrote: > >> from preamble: >> >> "Change Set: WebpageMorphFix >> Date: 25 April 2005 >> Author: Karl Ramberg >> >> Adds a method to WebPageMorph so horizontal scrollbars pop up"! >> >> >> ------------------------------------------------------------------------ >> >> >> >> > > > I think the WebPageMorph really should be part of the Scamper package, I don't think you can create an intance of it apart from Scamper. karl

1 0

Dorado bytecodes per second
by Jecel Assumpcao Jr 28 Apr '05

28 Apr '05

I am trying to get a feel for how different hardware options perform running Squeak and how that translates in practice into the user's experience. We currently use mainly the bytecodes/second and sends/second numbers for this which aren't available for older machines. I was particularly interested in the Dorado since that was considered the minimum performance for "acceptable" Smalltalking and other machines were evaluated in terms of how many Dorados they got (normally a small percentage in the early days). By running the benchmarks for the "green book" and doing a lot of rough extrapolations, my guess is that the Dorado would get between 200K and 400K bytecodes/sec. That is better than what I got running Squeak 1.16 on a 33MHz 486 machine which was some 13 years newer, but far below what I expected. My impression was that the old 20MHz ECL computer was able to reach a peak of one bytecode per clock which would indicate a number around three or four times better. After looking at a bunch of numbers my conclusion is that Squeak is usable on machines capable of at least 20M bps. If my guess above is corrent, that would be around 60 Dorados. If both estimates are true, then I wonder if our definition of "acceptable" has changed or if Squeak has become less efficient. Certainly Morphic is always being blamed for slowing everything down, so the latter is probably the case. If anyone has actual numbers, I would like to see them. The Dorado was really, really impressive. If re-implemented on a modern FPGA (not the low cost ones) it would be at most about six times faster than the original. That would be too slow for Squeak. As a custom chip in a slightly older process (0.18um for example) it would be around ten times faster than that, which would be just "usable". This doesn't make sense to me. The fastest machines are getting about 200M bps out of the interpreter and Bryce's compiler improves that to nearly 700M bps. I think I have lost an order of magnitude somewhere along the way. -- Jecel

1 0

Sun Fortress language spec
by David Faught 27 Apr '05

27 Apr '05

>Fortress is a new language from Sun. Thanks Marcus for having pointed >out it. A spec is available on >http://research.sun.com/projects/plrg/fortress0618.pdf There was an article "A Conversation with Guy Steele, Jr." in the April 2005 issue of DDJ that mentioned this briefly. __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

2 1

Contacting the Box-Admins for Support
by Ken Causey 27 Apr '05

27 Apr '05

If you are having any trouble with any of the services hosted on the Box1 server (We will attempt to maintain a list of such services at http://box1.squeakfoundation.org/ ) then you can request support by sending email to box-admins(a)discuss.squeakfoundation.org . If for some reason mail cannot be delivered to that address then feel free to contact me directly or send email to the Squeak list itself. Ken

1 0

Re: Bookmark icon
by karl 27 Apr '05

27 Apr '05

karl wrote: > It would be nice if the new site had a bookmark icon. > The plain white document icon I get in Firefox is a little boring. > Any thoughts? > > There is a howto here: > http://www.jamesshuggins.com/h/web1/favicon.htm > > Karl I created a so called favicon, attached, but that should be linked in the page head: "<LINK REL="SHORTCUT ICON" HREF="favicon.ico">" PS: Seems the menubox is misbehaving again :-( Karl

1 0

Re: Traits and Send Performance (was Re: Fwd: Fortress Language Specification)
by Dean_Swan＠Mitel.COM 27 Apr '05

27 Apr '05

Oops. That should have said 'Send Performance', not 'end Performance'. -Dean

1 0

Traits and end Performance (was Re: Fwd: Fortress Language Specification)
by Dean_Swan＠Mitel.COM 27 Apr '05

27 Apr '05

If I understand things correctly, Traits (at least as implemented for Squeak) cannot include references to instance variables, and so the performance will largely be constrained by method send performance. Is this correct? If this is the case (that Traits relies heavily on send performance and therefore method lookups, etc.) I would be concerned about migrating any of Squeak's class library to a Traits based implementation. This may be incentive enough though to seriously consider upgrading the VM to use the 'stack frame based contexts converted to full objects as necessary', or any other techniques that improve send performance. On a slightly related note, using a 3.7 VM, I see around 4 million sends/sec on my 3 GHz Pentium 4 HT, but about 6 million sends/sec on my 1.2 GHz iBook G4. The bytcodes/sec performance is higher on the P4 than the iBook though. Can anyone explain this? -Dean stéphane ducasse <ducasse(a)iam.unibe.ch> Sent by: squeak-dev-bounces(a)lists.squeakfoundation.org 04/27/2005 10:25 AM Please respond to The general-purpose Squeak developers list To: The general-purpose Squeak developers list <squeak-dev(a)lists.squeakfoundation.org> cc: Subject: Fwd: Fortress Language Specification > http://research.sun.com/projects/plrg/fortress0618.pdf Interesting... and worth to read (not only for traits). Stef

1 0

Re: OmniBase and Commits
by marcelo Cortez 27 Apr '05

27 Apr '05

Hi all Daniel in your example base what class is? If base is OrderedCollection or some thing like this should by try: txn := db newTransaction. 1 to: 1000 do: [:i| base := txn root at: 'linkedlist'. base add: anObject. txn markDirty: base. (i \\ 100) = 0 ifTrue: [ txn commit. txn := db newTransaction]] best regards mdc --- Daniel Salama <dsalama(a)user.net> wrote: > Hi, > > I'm trying to test uploading several thousand > objects into an OmniBase > db and I'm trying to figure out how to commit every > 100th object. Can > anyone shed some light into this? Every time I try > something, it must > be surrounded by a transaction, so I can't figure > out how to do it. > > I have stored a LinkedList to a root element. I have > something similar > to this: > > txn := db newTransaction. > 1 to: 1000 do: [:i| > base := txn root at: 'linkedlist'. > base add: anObject. > (i \\ 100) = 0 ifTrue: [ txn commit. txn := db > newTransaction]] > > But this doesn't seem to work. Please help. > > Thanks, > Daniel > > > ___________________________________________________________ 250MB gratis, Antivirus y Antispam Correo Yahoo!, el mejor correo web del mundo http://correo.yahoo.com.ar

1 0

Fwd: Fortress Language Specification
by stéphane ducasse 27 Apr '05

27 Apr '05

> http://research.sun.com/projects/plrg/fortress0618.pdf Interesting... and worth to read (not only for traits). Stef

1 0

Making a server.
by François THIMON 27 Apr '05

27 Apr '05

Hi. To implement a server with the Socket class, I first use the litenOn: method. Its comment says I need to use waitForConnectionUntil:. Since this method doesn't seem to exist anywhere I supposed I could use any other function with a similar name. But all of them work with timeouts. Does someone know whether I can specify infinite timeout or if have to do something else. I mean : how could I make a server that just keeps waiting for any connection without stopping, as any server anywhere? I naturally thought about writing something like : waitForConnectionFor:1000000 ifTimedOut:[^ self waitForConnectionFor:1000000000 ifTimedOut:self makeMeWaitAgain]. But I'd just want an infinite wait not an infinite recursivity!! Can you help me please?

1 0

← Newer
1
2
3
4
5
6
7
8
...
29
Older →

Jump to page: