how much time left?

[Alex Morega]

Hi,

Tim, how long do you want to keep the wide finder project happening?
I'd like to try to improve the Python results a bit more, maybe I can
get Stackless to work on the T2000. Problem is, I don't have time to
do it until Wednesday, at the earliest. Does that fit in?

Cheers,
-- Alex

[Erik Engbrecht]

Good question.

Just to give my opinion on the matter, I think the box should be transitioned from a test platform for WF2 to a test platform for people developing opensource libraries meant to take advantage of having lots of cores or doing public research in that area.  It seems like a very small investment on the part of Sun in order to receive public demonstrations of working software optimized for their architecture.

That's free R&D and free buzz.

--
http://erikengbrecht.blogspot.com/

[Tim Bray]

Well, I'm still getting "show me the test data" emails almost every
day, and "I got a program running, can I have an account?" once or
twice per week. So I'm not in a hurry to shut it down. I have the
computer for a year, which means till early 2009. I would expect that
anybody who's going to do anything significant would have done so in
the next month or two.

However, at some point, I probably agree with Erik. Shift the problem
slightly, from "make it go fast" to "build a programming interface so
that programmers can have it go fast easily without having to
understand all this voodoo".

What do others think? -TIm

[Justin Sheehy]

On Jun 21, 2008, at 10:27 AM, Tim Bray wrote:

> Well, I'm still getting "show me the test data" emails almost every
> day, and "I got a program running, can I have an account?" once or
> twice per week. So I'm not in a hurry to shut it down. I have the
> computer for a year, which means till early 2009. I would expect that
> anybody who's going to do anything significant would have done so in
> the next month or two.
>
> However, at some point, I probably agree with Erik. Shift the problem
> slightly, from "make it go fast" to "build a programming interface so
> that programmers can have it go fast easily without having to
> understand all this voodoo".
>
> What do others think?

+1 to both. I've not had time to really do any work on the core problem
(other than poking holes in oversimplifying/space-splitting, and so on)
and I also think that the later goal of finding better general
solutions is
a very worthwhile use of the system.

-Justin

[Philip Jacob]

On Jun 21, 2008, at 10:41 AM, Justin Sheehy wrote:

>> However, at some point, I probably agree with Erik. Shift the
>> problem
>> slightly, from "make it go fast" to "build a programming interface so
>> that programmers can have it go fast easily without having to
>> understand all this voodoo".
>>
>> What do others think?

In some ways, this is opposed to what a lot of us have been trying to
do for WideFinder (which is to decrease runtime). "Build a
programming interface" in fact means "build a generalized programming
interface" whereas specialized knowledge of the dataset in this case
has allowed many of us to make optimizations that would otherwise not
be possible.

That's not to say that a generalized kind of approach is not a good
idea: it is, so +1 from me.

phil.