Squeak-dev January 1998

squeak-dev@lists.squeakfoundation.org

81 participants
196 discussions

Squeak Music: Siren
by Richard Smol 21 Jan '98

21 Jan '98

Hi there, Well I just discovered on Stephen Pope's web-site that there is a successor to Smallmusic's MODE, called Siren. When I followed the link there was nothing there though! Bummer. I am really very curious about it. Does anyone here know any more? TIA! RS

1 0

RE: Squeak 1.3 for Windows
by Kevin Driedger 21 Jan '98

21 Jan '98

Can anyone confirm the existance of a version of Squeak for Windows CE? > -----Original Message----- > From: Andreas Raab [SMTP:raab@isgnw.cs.Uni-Magdeburg.DE] > Sent: Saturday, January 17, 1998 6:29 PM > To: squeak-annc(a)cs.uiuc.edu; squeak(a)cs.uiuc.edu > Cc: @cs.uiuc.edu@nfsbrute > Subject: Squeak 1.3 for Windows > > Fellow Squeakers, > > Squeak 1.3 is available for Win32 at > > http://isgwww.cs.uni-magdeburg.de/~raab/squeak/ > ftp://ftp.cs.uni-magdeburg.de/pub/Smalltalk/free/squeak/win32/ > > in the image/1.30 directory and should show up at the UIUC archive > soon. > Besides the new features of Squeak 1.3 and some bug fixes (such as > time > zones and a net work fix fir PWS) there are couple of changes to the > VM > (quoting the ReleaseNotes): > > Notes for the 1.3 version of Squeak > ----------------------------------- > > The directory contains two executables one simply called Squeak.exe > the > other one called SqueakJ.exe. The latter contains the dynamic > translator > (called Jitter) developed by Ian Piumarta. Note that this VM will > _not_ > run with images prior to 1.3! > > Also, there are now some preferences which can be set in the VM. You > can > access these from the system menu or by pressing the F2 key. The > preferences currently include: > * "Defer display update" > Some people didn't like it therefore you can turn it now on and off > (and > compare the speed difference ;-) > * "Run full screen" > An idea of Reinier van der Loon (thanks!) allowing Squeak to occupy > the > whole screen. You won't notice that you're running Windows -- you > feel > like running SqueakOS ;-) > Remember: Since you can't access the system menu in this mode you > need > to press F2 to make Squeak run in a window again! > * "Show output console" > Extremely helpful for VM developers. Opens a small window at the > bottom > showing the standard output messages. Thus, if you have any > "printf"s in > your C code you can see it right there. > * "Display version information" > Displays the version of the sqWin32Window.c file being used. Good > for > bug reports if the VM crashes. > > All of the above settings are stored in a file named "Squeak.ini" > which > can be found in your VMs directory. Note that the settings are > immediately > updated meaning that if you > - start Squeak > - Make it run full screen > - quit Squeak WITHOUT saving > - start Squeak again > you will be in full screen mode - that's why you have to > > ------> REMEMBER THE F2 KEY! <------ > > If you don't, you can still modify/delete the Squeak.ini file. > > > Andreas > -- > Linear algebra is your friend - Trigonometry is your enemy. > +===== Andreas Raab ============= (raab(a)isg.cs.uni-magdeburg.de) > =====+ > I Department of Simulation and Graphics Phone: +49 391 671 8065 > I > I University of Magdeburg, Germany Fax: +49 391 671 1164 > I > +=============< http://isgwww.cs.uni-magdeburg.de/~raab > >=============+

2 1

Re: Squeak Code Etiquette?
by Andreas Raab 21 Jan '98

21 Jan '98

John, > For example in squeak there is a method r:g:b. In a set of programs that I am > trying to convert, red:green:blue is used. Should I change all the programs in > the set I am working on, or add the method to the Squeak base? I would place such methods in a category called "compatibility" or if it something specific to a certain dialect "dialect-compatibility". In my eyes its far less effort to incorporate such modifications in the Squeak base than changing a number of methods. Andreas -- Linear algebra is your friend - Trigonometry is your enemy. +===== Andreas Raab ============= (raab(a)isg.cs.uni-magdeburg.de) =====+ I Department of Simulation and Graphics Phone: +49 391 671 8065 I I University of Magdeburg, Germany Fax: +49 391 671 1164 I +=============< http://isgwww.cs.uni-magdeburg.de/~raab >=============+

1 0

Re: Common byte code sequences
by Ian Piumarta 21 Jan '98

21 Jan '98

Andreas, > While looking around in the DynamicTranslators documentation category I > noticed that you've documented a number of common byte code sequences. > Where do these come from? Are this just "good guesses" or did you analyze > the control flow? They came from a combination of three sources. Some are just "obvious". Things like "pushInteger {add,sub,mul}" occur all over the place. Some were spotted just by looking through the bytecodes generated by pieces of typical Smalltalk -- e.g. the "loopStep" opcode and those like "push{True,False} jumpIf{False,True}" (which occur whenever you use #and: or #or:). Some were the result of analysing the control flow (I suspect this is what you wanted to hear ;-). Some unexpected ones popped out of this analysis -- and sure enough, implementing them noticably improved the overall performance. > If so, how did you do this? I modified the VM to print out the index of the opcode that it was dispatching. I generated (from Smalltalk) a C header file containing a map from the opcode indices onto opcode names, and then wrote a C program to extract and print all sequences of N opcodes that did not contain a branch (or send or return) in the middle. The output of this program was fed through an appropriate combination of "sort" and "uniq" to generate tables of the most popular sequences. If you (or anyone else) is interested in this, I can place the raw data file (which contains a trace of something like 730 million opcodes taken while the VM regenerated itself from Smalltalk source), the analysis program, a script for processing the output and some pre-generated lists of common sequences, in alix's FTP area. FWIW, I've appended the five most common sequences for lengths 2 to 5 for the first 4 million opcodes in the trace. (The time to process this stuff goes up exponentially because of the several sorting steps involved: processing just the first 4 million opcodes required about 10 minutes "per length" on the fastest machine to which I have access.) Regards, Ian length 2 191220 PushReceiverVariable PushTemporaryVariable 151178 PushTemporaryVariable Send 104733 PushReceiver Send 70731 PushReceiverVariable Send 61065 PushTemporaryVariable PrimPointX length 3 58745 PushReceiverVariable PushTemporaryVariable PrimPointX 55108 PushReceiverVariable PushTemporaryVariable PrimPointY 35651 StoreTemporaryVariable PushReceiverVariable PushTemporaryVariable 28760 PushConstant PrimEquivalent LongJumpIfTrue 28074 PushTemporaryVariable StoreReceiverVariable ReturnReceiver length 4 35590 StoreTemporaryVariable PushReceiverVariable PushTemporaryVariable PrimPointX 23285 MacroMoveTempRcvr PushTemporaryVariable StoreReceiverVariable ReturnReceiver 20445 PushTemporaryVariable PushConstant PrimEquivalent LongJumpIfTrue 18516 PushTemporaryVariable PrimPointY PrimAdd SpecialPrimitive 18516 PushReceiverVariable PushTemporaryVariable PrimPointY PrimAdd length 5 18516 StoreTemporaryVariable PushReceiverVariable PushTemporaryVariable PrimPointX PrimAdd 18512 PushTemporaryVariable PrimPointX PrimAdd PushReceiverVariable PushTemporaryVariable 18512 PushReceiverVariable PushTemporaryVariable PrimPointY PrimAdd SpecialPrimitive 18512 PushReceiverVariable PushTemporaryVariable PrimPointX PrimAdd PushReceiverVariable 18512 PrimPointX PrimAdd PushReceiverVariable PushTemporaryVariable PrimPointY

1 0

Re: Coordinate order in Array2D
by Mike Klein 21 Jan '98

21 Jan '98

> I guess this is a long winded way of asking if it's a big deal to > bring coordinate ordering in Squeak into conformance with conventional > matrix notation. A search through the 1.23 image tells me that > at least there, there's nothing using it save for the shrinks which > trash it. I don't know if there is heavy use in the new image. Would > anyone out there get hurt by such a change? I'd go for it but then, I'm a rampant change kinda guy. Why, I'd also change Number >># positive. See if that method offends your mathematical sensabilities! -- Mike Klein

1 0

RE: Coordinate order in Array2D
by Leandro Caniglia 21 Jan '98

21 Jan '98

Cristopher Oliver wrote: CO> Hey folk! Why does Array2D at:at: and at:at:put: use a coordinate order reversed from normal mathematical conventions? The first coordinate is traditionally the matrix row. I just got bitten while transliterating some cryptology code using matrices. This sort of difference can cause severe brain hemorrhage. I don't know why Array2D have this coordinate order but please note that the traditional matrix notation, in which the first index is the row, opposes to the also traditional (x, y) order. Smalltalk uses the (x , y) order in the class Point, so using first the column (x) and then the row (y) seems to adhere to a homogeneous convention. Saludos, Leandro

2 1

RE: Coordinate order in Array2D
by Mike Klein 21 Jan '98

21 Jan '98

> Cristopher Oliver wrote: > > CO> Hey folk! Why does Array2D at:at: and at:at:put: use a coordinate > order reversed from normal mathematical conventions? The first > coordinate is traditionally the matrix row. I just got bitten while > transliterating some cryptology code using matrices. This sort of > difference can cause severe brain hemorrhage. > > I don't know why Array2D have this coordinate order but please note that the traditional matrix notation, in which the first index is the row, opposes to the also traditional (x, y) order. Smalltalk uses the (x , y) order in the class Point, so using first the column (x) and then the row (y) seems to adhere to a homogeneous convention. > > Saludos, > Leandro Personaly, In my Matrix classes, I use the selector down:accross: this way I never have to remember, (Unless you are used to reading right to left... this is an international list :-) You can then add to Matrix: down: i accross: j ^self at: j at: i Also, notice that Point reverses the sign of the traditional y-axis orientation, but in math, it is ordinary for matrices to be numbered downwards. The moral of the story is that in math, there are all sorts of natural isomorphisms which we use painlessly, but in CS, these may not seem so natural. The coordinates of a location in a 2-d axis-aligned euclidean plane bears as much resemblance to matrix indeces as an Array of Booleans resembles an Integer. -- Mike Klein

1 0

Little problems in 1.3
by eguevara＠dc.uba.ar 21 Jan '98

21 Jan '98

Hi everybody! I've found 2 litte "errors"in 1.3 version: 1) #formta - in some messages a PopUp error window appears. 2) #findClass - it doesn' show me all possible classes starting with the characters I've typed. I hope I' being useful. -- Eric Maximiliano Rodriguez Guevara eguevara(a)dc.uba.ar

3 2

Morphic
by Matthijs van Kempen 21 Jan '98

21 Jan '98

Hi, The new UI looks a lot better, escpecially the resident help menu, sigh.. The Gui still has some refresh problems: In the Play with me -3 world if I ask the actions of the ellipse morph and then close it, the whole screen gets refreshed and it still takes a long time to do that. I think I will look carefullyu at all refreshes and try to fix them, sound good? (I'll need help I'm just starting.). Thank god I have a slow machine handy.

1 0

Morphic + Distr.Comp. >>> @!&#Beans?
by Matthijs van Kempen 21 Jan '98

21 Jan '98

Hi Everyone, I'm a beginner and very much in love with Smalltalk. I have several observations as a user of Squeak: If we could develop a vast library of useful models, buttons etc. that would be very easy to configure visually and behavorially, this would give us @!&#Beans all over again. But this time, they would be smaller, faster?, very portable and they would provide a means of adding or changing behavior in components. So we could exchange programs in the form of small reusable components. For instance, I could download several buttons, sliders and models with a native lookand feel, use them in my Morphs or easily delete them from my system. Or I could download a complete 'applet' and use it. I'm not sure how to relate this to the class hierarchy. It seems that a construct like an interface would be necessary to handle the interaction between the component and the base classes. (Bases classes would not be BASE classes anymore, they would act as libraries which can be dynamically loaded and unloaded. A core of undeletable classes would still remain though.). The pluggable web server goes a long way in this direction. It is quite easy to setup and use. It's the fastest way to build a CGI enabled web server ever. If it's interface were a little better, you could fathom it being a component that would be usable for any user even without Smalltalk knowledge.... Suppose you had a PWS and distributed clients for it...That would be great for unloading the server. Besides with what Squeak has, it could probably outrun JavaBeans.That could also be a great way of adressing our own problems of managing code. Think about it and if you don't like it, don't think about it.

1 0

← Newer
1
...
7
8
9
10
11
12
13
...
20
Older →

Jump to page: