Could anyone elaborate on the differences between an implementation of Squeak directly to the hardware ( referred to somewhere as a "language kernel" ) and an implementation of Squeak on top of the Linux kernel accessing the framebuffer? Is anyone working on either of these implementations?
Thanks. Sven
Alan Kay wrote:
Good work!
This is a good platform to start experimenting with.
An interesting question is whether it is really worthwhile to do the port via WinCE, as opposed to Linux (which I've heard is also available on the iPAQ), or even to try a bare port. Tim Rowledge has done quite a bit with ARMs, so he might have good advice here. Also, Jamie Robertson at Disney has done a bare port to the Intel StrongArm 1110a kit, that seems to work pretty well.
Cheers,
Alan
At 6:08 PM +0100 10/29/00, Jörn Eyrich wrote:
Okay, I got a bit sidetracked here when I discovered that there are already ports for a C64 emulator and MAME available for the iPAQ ;-) (playing with this 4-way-pad doesn't work all that well, by the way, in my opinion)
However, thanks to the work Yasuhiro Kurahara and Thomas Nelson did for the Jornadas, I was able to get an ARM Debug version of the 2.7 WinCE VM running on the iPAQ 36xx. It is about 480K and does about 3,8 Mbytecodes/s, 191 Ksends/s on "0 tinyBenchmarks". I use a 760K image obtained by "Smalltalk majorShrink; abandonSources; lastRemoval"
The problems I encounter are:
- part of the squeak window is hidden under the WinCE menu bar (also
noted by Yasuhiro Kurahara)
- sometimes the squeak window is not redrawn
- I can't get to the right button menu
- I don't know how to emulate the "Alt" modifier key (these last two
items conspire to make it impossible for me to do a simple doIt)
- I can't see any new browsers or workspaces that I open (though
they appear in the list after i do "find windows...")
- collapsed windows don't have a title (only close and maximize
icons seperated by free space)
- moving/collapsing windows seems to break sometimes, then while
trying to open the debugger, there seems to be another problem so I get a walkback that ranges from a "ZeroDivide" at the bottom of the screen through a "Fraction>>dNU" to a "MNU: bitAnd:" at the top, but unfortunately ...
- ... I can't find a SqueakDebug.log containing the whole story
(- after that you have 2MB RAM less available, but I guess that's WinCE for you)
Next thing is trying to do the same for Andreas Raab's latest (pre-2.8)sources.
Joern
squeak-dev@lists.squeakfoundation.org