 |
|
||
|
|
Camp Smalltalk (OOPSLA 2000) Wednesday |
Sunday Oct 15th 2000
Wednesday Oct 18th 2000
I must admit it's late morning. Since I was up til 4AM, and didn't have a session until the afternoon to attend I'm afraid the alarm clock didn't wake me. Hey that's only 2:00AM PDT and I don't think I've changed time zones. But because of this and my busy schedule today the report is somewhat short.
One of the things about doing the trip report during the show is that people were much more aggressive in catching up to me and asking for correction. I think in future I'll post them to a Swiki. However don't be afraid to ask me to change/correct things. The objective of course is to get correct information out there and sometimes in the effort to type, listen, learn etc I might get a few things wrong. Of course I can't help with the translation to Japanese or be responsible for the poor quality of the translation, but isn't the Internet amazing?
PS It's Camp Smalltalk at ESUG, not ECOOP. And a few changes were done to the BOF report.
In the Camp Smalltalk area about 28 people and 8+ computers were busying humming, various people were playing with the neatest toy, a clear rubble ball from ThoughtWorks which encloses some LEDs. I'm sure it got the attention of some of the campers who have kids at home, awaiting some trinket from Dad's visit to OOPLSA. For most of the day the numbers in the Camp Smalltalk area floated around 30.
Tim was busy setting up an environment on the ARM to compile GCC. Perhaps he's the first to do this with a newer version GCC because it doesn't work out of the box. Yes we need that for an ARM version of JIT 3. I was spending time with a Disney fellow getting Squeak JIT 3 to compile on the Macintosh for OS-9. Yes JIT 3 is coming.
Chris dropped by and admitted he had bought the Ruby book. Why? It's kinda like actor, it's a replacement for Python, but won't a Smalltalk script language make more sense?
Someone had brought a photocopy page from the Squeak book by. So it is real…
Chris asked did in fact the non-jitter version on my machine run faster than the jitter version on Ian's machine last night? "Yes that's why Ian is locked up in his hotel room this morning trying to understand why" I replied, in fact he was there all day.
Ok I'll skip ahead about 11 hours and explain. This evening Ian and I spent a few hours in the hotel lobby tinkering and trying some things to understand why. My machine is at least 25% faster than his machine but the numbers came out somewhat identical. That does of course mean JIT3 is at least 25% faster, but it should be more.
So what we did along other things was replace his instruction cache flushing logic with some logic I had found a few months back taken from the LinuxPPC Kernel, then we confirmed the rate of instruction flushes and the amount of bytes flushed seemed reasonable. Other things like sleeping, GC events and general handwaving was visited. After much exploration we found that his image had twice as many objects in it, and had 1/2 as much memory as my image. So I had more memory to play with, and had fewer objects to manage. Although going to somewhat identical images and memory sizes improved his numbers they weren't as good as he felt they could be. So more work will be done and I believe he was going to run a profile to understand where the bottlenecks are. Therefore I'm not going to publish any benchmark numbers since JIT3 hasn't been released yet and something Ian has done lately has degraded performance.
Now back to Camp Smalltalk. Conversation around the table ranged from how JIT 3 did the code translation to Ruby to ARM power consumption and Intel CPU step down technology.
Later in the morning I spent some time with a onlooking showing him Squeak, MPEG3 of course and explaining why Squeak was important and what you could do with Squeak. The fellow was from Tektronics, for the older Smalltalkers among you it's a sad reflection from where we came from to have to explain Smalltalk to one of the original licensers of the technology.
Later speaking of 'dead' technology Andy Hodges was telling me that they in fact have lots of OS/2 machines happily running Smalltalk. Discussion then moved into an issue with ignoring user input in VA which Enoch was happy to address.
It seems this morning very little work was being done on Camp Smalltalk projects, rather questions were being asked, then answered. But this is a great thing, where else can you come and get expert answers without having to pay for support.
However Campers should pay attention: numerous people suggested to me that at OOPSLA we need to have "Some Greeters". Someone to catch and assist the curious, and to ensure they don’t get away without some interaction with the Campers. Although it's great to come work on projects, the intent of the venue is to get more than us interested in the technology. I know Don spent time doing this (thanks Don) but I don't think we were as successful at it as we could have been because of the input I received.
I spent the afternoon attending the Garbage Collection tutorial. It was good to confirm that I do have some sort of understanding of the area, but it was also a chance to talk to Richard Jones about some of the finer points of the train algorithm, and tenuring. He had in fact a question back to me, which one the readers might know.
In Java you can only resurrect a finalized object ONCE. True or false? Anyone know? Better yet why?
Strange as it may seem today I talked to no less than four senior computer science types about a common theme, the poor showing of computer science university student with regard to knowing anything about theory, history, and the application of theory. This wasn't my idea rather it's what was expressed to me.
In the evening we had a somewhat disappointing OOPLSA event, but the conversion was good and I missed most of a spectacular event when you mix oh say a dozen senior Smalltalkers around a table and talk about history. Do you know where the name Envy comes from? Well I do, but you'll need to ask… Did you know a couple of us have worked on Cray computers, well perhaps not in Smalltalk. Did you know … Well all good things must come to an end and it seem Ian wanted to get back to his Jitter 3 performance concerns, and I need to do some pair programming and it's difficult for me to get Ian to visit my place so I grabbed the chance to spend a few hours with him, which does cut into my time to gather Smalltalk tales from OOPSA
Oh what if you had Squeak as a Admin console/control environment for Unix support?
So as you see neat exciting things are happening. So tomorrow OOPLSA ends and you'll need to wait for the last trip report.
PS I did have some poeple commit to working on the windows verson of the MPEG plugin, but I need someone with some MMX background so if you've a few minutes/hours free?