Squeak-dev October 1999

squeak-dev@lists.squeakfoundation.org

141 participants
396 discussions

Continuous Process creation
by David N. Smith (IBM) 06 Oct '99

06 Oct '99

Process>>#forContext:priority:, the method that creates processes, seems to get called about 250 times per second to create a process which is immediately thrown away. The calls seem to be coming from ProcessScheduler>>#yield which creates a process to signal a new Semaphore. Here is the code: yield "Give other Processes at the current priority a chance to run." | semaphore | semaphore _ Semaphore new. [semaphore signal] fork. semaphore wait It looks like a cute trick, but I'm wondering what the 250 new processes per second (on my old PPC 9500/200) do to garbage collection and processor usage. The #yield messages seem to be constantly sent by controllers and I'm in a MVC world (because that's what runs on tiny processors). Dave _______________________________ David N. Smith IBM T J Watson Research Center Hawthorne, NY _______________________________ Any opinions or recommendations herein are those of the author and not of his employer.

2 1

RE: Add instVar in Process
by Michael Klein 06 Oct '99

06 Oct '99

Rather than naming processes, I have a method Process >># initiatorBlock than answers the block that got forked. I find this is enough to identify most processes, even if they are unnamed (very usefull if you are debugging somebody else's code). I ported from VW to Squeak easily (while composing this response), mostly because both systems have 'thisContext'. I am presently trying to port to VA, but the exposure of the stack is causing all sorts of iritating problems. You may find this easier and more usefull than adding a name to Process. In Squeak 2.5, (Process allInstances collect: [:p| p initiatorBlock ]) reject: [:each | each isNil ] answers: ( [] in Delay class>>DoIt [] in WeakArray class>>DoIt [] in InputSensor>>installInterruptWatcher [] in ProcessorScheduler class>>startUp [] in SystemDictionary>>installLowSpaceWatcher [] in ControlManager>>activeController: ) The first to process come from the bad habit of releasing images with processes forked from a workspace or from changed code. The Delay process should be: [] in Delay class>>startTimerInterruptWatcher This can be remedied by doing a: Delay startTimerInterruptWatcher. I'll figure out the WeakArray one later. If anyone cares, I'll clean the code and send it on up. -- Mike Klein Squeak versus VisualAge Hmmm, tough call.... BlockClosures would tip the scale for me. On Wednesday, October 06, 1999 10:53 AM, David N. Smith (IBM) [SMTP:dnsmith@watson.ibm.com] wrote: > At 9:38 -0700 10/5/99, Xu Wang wrote: > >I noticed Squeak dose not allow me to add extra instVars in Process. Is > >there a special reason for it (e.g. hardcode in VM etc.)? > >I need to add a process context which holds user defined process > >dependent info. > > I was looking at adding one yesterday, but backed off before I tried > it. In another Smalltalk that I use, processes can be named, and I > find it useful when a lot of processes are running around. > > Dave > _______________________________ > David N. Smith > IBM T J Watson Research Center > Hawthorne, NY > _______________________________ > Any opinions or recommendations > herein are those of the author > and not of his employer. >

1 0

ClearType for Squeak/Sub-pixel antialiasing
by Henrik Gedenryd 06 Oct '99

06 Oct '99

The TrueType font rendering and antialiasing in Squeak is very cool. I realized it would be quite easy to implement sub-pixel rendering, at least in a simple form. If you haven't heard of it, sub-pixel rendering is when you manipulate the individual R, G and B cells of an LCD screen's pixels to achieve a 3x increase in horizontal antialiasing resolution, which makes especially text much sharper. In effect, you use the three cells of each pixel as independent pixels, and balance their colors so you don't notice that they are of different colors (like you don't notice that a white pixel is made of one red, one green, and one blue cell). Now this only works on LCD screens (more or less), but that's what a Dynabook is meant to use after all, isn't it?!! Or you could use that 22-inch Apple $4000 monster LCD monitor, for example... Microsoft is implementing this as ClearType, boasting about it and claiming to have invented it, although it's an old invention by Apple, Xerox, and others (hmm, doesn't that sound familiar?). A good description of how it works and of how to implement it, with good pictures and a simple demo implementation, is at http://www.grc.com/ctwhat.htm (Those three pages are all the information I needed to do my preliminary implementation, in my spare time. It ended up being only 20-30 lines of Squeak.) I attached a small sample GIF as an illustration. In it, the four variants are, top to bottom: 1. Standard Morphic anti-aliasing 2. Simple sub-pixel version without color balancing (notice color artifacts). 3. 3-pixel color balancing. 4. 5-pixel color balancing. Should be sharper than #3 but appears not to be; I don't know why. Note that the vertical resolution is not improved by this technique. However, Andreas Raab tells me, "the reason that none of the TrueType stuff is actually used is that we don't have grid-fitting or hinting in Squeak which means that despite the efforts of anti-aliasing stuff fonts at small resolutions are hardly readable (too blurry)". Now sub-pixel techniques improve this a lot (as the right sample in the image is meant to show; the effect is even greater with italics). The lack of grid-fitting is the cause, I believe, of the gray blur that can be seen in particular inside the lower case 'e', and below the horizontal bar of the 'A', in the small samples. Actually, also the SP-rendered images would become even sharper with improved type rendering, as they use the same anti-aliasing code. Are these improvements hard to do? I'm not sure, does anyone out there know? There is eg. a free-source (C) package called FreeType that implements TrueType rendering with hinting and all. Now I'm just hoping that others also would like better TrueType support in Squeak, and would like to chip in w r t hinting etc. and implementing general use of outline type in Squeak. Note that we have a working cross-platform version of ClearType even before Microsoft has released any version at all! As for now, this is just an experimental implementation. It can be made more general, much faster, and better integrated into Morphic, for example. I am on my part certainly willing to do the fixes needed to make it general enough, and to optimize it and produce a C-compiled version for Balloon. Anyone? Henrik < = > . Attachment converted: Anon:sub-pixel.gif (GIFf/8BIM) (00014B78)

2 1

Add instVar in Process
by Xu Wang 06 Oct '99

06 Oct '99

I noticed Squeak dose not allow me to add extra instVars in Process. Is there a special reason for it (e.g. hardcode in VM etc.)? I need to add a process context which holds user defined process dependent info. Thanks, -xwang -- Xu Wang Distributed Systems Group Cincom Systems Voice: (408) 216 4591 Fax: (408) 216 4500 ----------------------- Opentalk for open minds -----------------------

3 3

[ENH] Speed boost for hierarchical lists
by Bob Arning 06 Oct '99

06 Oct '99

The following changeset can provide a *major* speed boost in cases where you collapse an item in a hierarchical list with many thousands of sub-items. Cheers, Bob ======== 'From Squeak 2.5 of August 6, 1999 on 6 October 1999 at 12:29:24 pm'! "Change Set: MorphRemove Date: 6 October 1999 Author: Bob Arning A significant speed-up in Morph>>removeAllMorphsIn: when dropping large numbers of submorphs (e.g. when collapsing an item in a hierarchical list with thousands of subelements)"! !Morph methodsFor: 'submorphs-add/remove' stamp: 'RAA 10/6/1999 12:23'! removeAllMorphsIn: aCollection "greatly speeds up the removal of *lots* of submorphs" | set | self changed. aCollection do: [:m | m privateOwner: nil]. set _ IdentitySet new: aCollection size * 4 // 3. aCollection do: [:each | set add: each]. submorphs _ submorphs reject: [ :each | set includes: each]. self layoutChanged. ! !

1 0

Re: wearable computers
by Peter Novak 06 Oct '99

06 Oct '99

Try Computer Magazine 1997 issue No. 02 (february). There's good article about Steve Mann, guy who is something like artist of wearable computing. http://computer.org/computer/co1997/r2toc.htm (to download it you need to be a member of E-library, but still you can find it in the paper version). Peter Novak. -----Original Message----- From: Ivan Tomek <ivan.tomek(a)acadiau.ca> To: squeak(a)cs.uiuc.edu <squeak(a)cs.uiuc.edu> Date: 5. október 1999 23:25 Subject: wearable computers Check out http://www.usnews.com/usnews/issue/991011/nycu/wearable.htm Dr. Ivan Tomek, Jodrey School of Computer Science Acadia University Nova Scotia, Canada fax: (902) 585-1067 voice: (902) 585-1467 Attachment converted: Anon:Peter Novak.vcf 10 (TEXT/R*ch) (00014710)

1 0

RE: An image is a lonely place
by Raab, Andreas 06 Oct '99

06 Oct '99

Whoa! Do you have any demos (images) of this stuff?! Sounds really cool, I'd love to see how far you got. Andreas -- +===== Andreas Raab ========= (andreasr(a)wdi.disney.com) ==+ | Walt Disney Imagineering Phone: +1 818 544 5016 I I Glendale, CA Fax: +1 818 544 4544 I +======< http://isgwww.cs.uni-magdeburg.de/~raab >========+ > ---------- > From: Luciano Notarfrancesco > Reply To: squeak(a)cs.uiuc.edu > Sent: Wednesday, September 29, 1999 1:47 PM > To: Tom Morgan > Cc: recipient list not shown > Subject: Re: An image is a lonely place > > > Tom, > I've implemented: > o. speech synthesis (text-to-speech and singing voices) > o. simple gestures synthesis (2D faces, SouthPark-like toons) > o. 3D facial animation (the Waters' muscles model) > I haven't released it yet, but if you want anything of this just e-mail me > and I will send you a changeset. I think this stuff might be useful for > you if you want to implement avatars. > > Cheers, > Luciano.- > > > -----Original Message----- > From: Tom Morgan <tmorgan(a)acm.org> > To: squeak(a)cs.uiuc.edu > Date: Wed, 29 Sep 1999 07:58:02 -0400 > Subject: An image is a lonely place > > > I've been noticing how lonesome it is inside an image. > > > > There are lots of strong place based/community elements > > which surround Squeak. These are currently formed > > by the basic networked apps of email, the discussion Swiki, > > the mail archives, and so on. > > > > There is now lots of machinery which is available which > > would allow basic collaborative support right inside > > a running image (and there are some spacey ones imagineable > > using the 3D stuff that feel like 'Snowcrash') > > > > The kinds of things that I have in mind revolve around > > establishing a sense of 'place', which is shared, 'embodiment' > > of you the user explicitly some how in the interface and > > some amount of 'awareness' of the presence of > > others in the shared place, all done with a light enough > > touch that we don't end up with avatar fashion shows. > > > > In this picture, an Morphic book wouldn't be found by > > reaching out across the network and loading it; instead, there'd be > > a more or less far away part of your image (maybe > > dressed up as an Alice world) that would let you wander > > around and find the book you wanted. Perahaps the > > author would hand it to you, if he happened to be around. > > > > While working, instead of the discussion Swiki, there'd be > > small knots of others, gathered around places of interest > > to them. You could glance around and see if anyone > > was nearby, possibly interested in what you were up to. > > > > Over the course of a few years, I have seen naive users > > comfortably navigate incredibly complex graph structures, when > > they are presented with a place based metaphor. > > > > I am not sure how universally appealing these interface > > metaphors are, but they seem to be captivating for > > at least some portion of the population. > > > > I remember one posting about trying to incorporate the ICQ > > protocols into Squeak. > > > > Are there others padding around this collaborative territory? > > > > ...Tom M > > > > > > >

1 0

RE: [bug] [fix (? partial)] B3Dmorph no longer works
by Raab, Andreas 06 Oct '99

06 Oct '99

Mark, Thanks for noting this. The original intent of this morph was just to provide something people could play with in the *very* early stages of 3D support. This guy is pretty much obsolete but I'll fix it so that it does use the right (e.g., latest interfaces). Andreas -- +===== Andreas Raab ========= (andreasr(a)wdi.disney.com) ==+ | Walt Disney Imagineering Phone: +1 818 544 5016 I I Glendale, CA Fax: +1 818 544 4544 I +======< http://isgwww.cs.uni-magdeburg.de/~raab >========+ > ---------- > From: Mark Guzdial > Reply To: squeak(a)cs.uiuc.edu > Sent: Tuesday, September 28, 1999 7:23 AM > To: squeak(a)cs.uiuc.edu > Cc: recipient list not shown > Subject: [bug] [fix (? partial)] B3Dmorph no longer works > > Playing around with Squeak 2.6a, I found that B3Dmorph (from the New Morph > menu) no longer worked. Thinking that it might be an exceptions thingie, > I > then tried 2.5. Didn't work there either! > > The problem seems to be that the B3Dmorph no longer has a texture when > it's > first told to draw itself, so it bombs when nil is asked to be asTexture. > > The below fix just generates a form and uses that as the texture. I do it > inside of initialize. Don't know if it's the right fix, but it works now. > > However, the set texture menu item on B3Dmorph still doesn't work. > > 'From Squeak 2.5 of August 6, 1999 on 28 September 1999 at 10:19:52 am'! > > !B3DMorph methodsFor: 'initialize' stamp: 'mjg 9/28/1999 10:19'! > initialize > super initialize. > geometry _ B3DBox from: (-0.7@-0.7@-0.7) to: (0.7@0.7@0.7). > camera _ B3DCamera new. > (self confirm:'Put me into a clipping frame?') > ifTrue:[camera position: 0@0@1.5] > ifFalse:[camera position: 0@0@2. color _ nil]. > camera nearDistance: 0.1. > camera farDistance: 5.0. > self extent: 100@100. > texture _ (Form extent: 100@100) asTexture. > angle _ 0.! ! > > -------------------------- > Mark Guzdial : Georgia Tech : College of Computing : Atlanta, GA > 30332-0280 > (404) 894-5618 : Fax (404) 894-0673 : guzdial(a)cc.gatech.edu > http://www.cc.gatech.edu/gvu/people/Faculty/Mark.Guzdial.html > >

1 0

[BUG?] (Symbol readFromString: '#Smalltalk') class ~~ Symbol
by Mark A. Schwenk 06 Oct '99

06 Oct '99

As part of a high school student's internship at WellThot Inc., we are producing a regression test for Squeak. We are working our way through Goldberg & Robson's Smalltalk-80: The Language (Purple Book), beginning with chapter 6. As part of testing #readFromString:, we noticed that (Symbol readFromString: '#Smalltalk') class == String. when we expected (Symbol readFromString: '#Smalltalk') class == Symbol. What do others think? Is this a bug? Should we fix it? I notice that in VW 3.0, (Symbol readFromString: '#Smalltalk') class == ByteSymbol. -Mark Schwenk WellThot Inc.

1 0

RE: [BUG?] (Symbol readFromString: '#Smalltalk') class ~~ Symbol
by agree＠carltonfields.com 06 Oct '99

06 Oct '99

> As part of testing #readFromString:, we noticed that > > (Symbol readFromString: '#Smalltalk') class == String. > > when we expected > > (Symbol readFromString: '#Smalltalk') class == Symbol. > > What do others think? Is this a bug? Should we fix it? The far more onerous problem is not so much that the class of the expression is String, but that the VALUE of the expression is ''! This derives from the fact that Symbol is a subclass of String (adding the interning functionality), and that String overrides the definition in Object which, I believe would have returned the correct value. For example, (Object readFromString: '#abc') class yields the expected result. Since String expects a string, and an expression beginning without a "'" is not a string, well, you know. So the entire thing devolves to asking the question, what are the expected semantics of the "readFrom:" operation? While it is true that LargePositiveInteger readFromString: '123' returns a SmallInteger and SmallInteger readFromString: '123123123123123' returns a LargePositiveInteger, would this be considered aberrant behavior, or the correct result? Clearly, we expect the Symbol example above to behave more like the Object example, not because of the CLASS returned, so much as because of the actual result given.

1 0

← Newer
1
...
33
34
35
36
37
38
39
40
Older →

Jump to page: