Hello crunchers!

I guess most of you already recognized the current apps of Milkyway@home are grossly inefficient. There was quite some discussion about the possible improvements and when a new app would be released to the public. Just some hours ago Dave posted, there will be a further delay to the release
Dave wrote:

Well Travis hasn't compiled the code changes I pushed through so we cant release until that happens.

I offered our help and supplied the project with compiled versions of an improved (i.e. vastly faster and delivering correct results) version implementing the changes described in the code discussion forum. They were tried by the project and I got the answer:

Your applications are indeed working great!

So I guess it will be okay when I link these versions here in the forum. To make that process fair, the links will be posted tomorrow (31st of October) at 18:00 UTC as some kind of a Halloween gift ;)

That way there is enough time for everyone to prepare for the release and for the project to adjust their credit scheme (or distribute the apps as default). I would suggest 4 to 5 credits for the long WUs (now 260 credits) and something around 2.5 credits for the medium ones (now 139.29 credits). That would still be a quite generous level (comparable to the current state).

If the project decides not to make some adjustments to the credits, nobody should expect to make as much credits as Milksop is doing now. You will run into the credit limit implemented as a reaction to the improvements of Crunch3rs app. But hopefully MW raises the limit of 300 WUs/day, otherwise fast machines won't be able to tap their full potential.

Just as some preliminary information, the apps are fully compatible and interchangeable with the official applications and seamlessly continue from the checkpoints of the official one. The Linux version is statically linked (as the semi-official 1.24 for Linux64 compiled by Crunch3r), and should therefore also work on 64Bit systems (tested on Ubuntu64 8.10). And as a bonus, the progress bar is finally linear ;)

Both versions base on the exact same source code and are compiled for quite ancient machines as compiler target. It should work on virtually every x86 CPU capable running Windows, they do not require any SSE capabilities. As a consequence the speed is quite similar for both Linux and Windows. More information about the apps will come tomorrow.

PS:
When the improved app is released, I will stop to crunch for this account.

Good news ! :-)

Is a Mac OS client version also foreseen and if yes, for x86 only or also PPC ?

Thx !

Good news ! :-)

Is a Mac OS client version also foreseen and if yes, for x86 only or also PPC ?

Thx !

In the moment I have only versions for Win and Linux on x86er CPUs. But I told already the project staff that they could again ask Crunch3r, as he has some experience in compiling things for OSX.
I can't make a promise, but maybe I could get access to a PPC and create a version for it over the weekend.

Btw., I planned to post a link to this topic in the fora/forums of some of the big teams. But frankly, my French isn't sufficient for this. It would be great if you could bring this to the attention of your team members.

G'day Milksop,
BOINC@Australia got your message no problems. Hope to see the new applic as soon as it's available.

Thanks for the heads up.

---

[http://milkyway.cs.rpi.edu/milkyway/team_display.php?teamid=15]Click here to join the #1 Aussie Alliance on Milky Way.[/url]

Hi Milksop, we at Cruncher Junkies also are awaiting the improved application.

good work mate !

good work mate !

Thanks!

I want to mention I just got some kind of official blessing to release the app. As an additional information Dave told me the project plans to roll out their new application (with some functional changes I suppose) sometimes next week, so real soon now (tm) ;) But I guess we are really getting closer to that event.

OK -- does that mean the improved application release will wait until next week and be handled on the Milkyway site?

I assume we'd need to clear out the existing work units first before moving to the new application (that's what goes on with the optimized SETI application) -- is that correct?

I want to mention I just got some kind of official blessing to release the app. As an additional information Dave told me the project plans to roll out their new application (with some functional changes I suppose) sometimes next week, so real soon now (tm) ;) But I guess we are really getting closer to that event.

---

I want to mention I just got some kind of official blessing to release the app. As an additional information Dave told me the project plans to roll out their new application (with some functional changes I suppose) sometimes next week, so real soon now (tm) ;) But I guess we are really getting closer to that event.

Now I just kind of wondered...
Milksop, is your app an improvement of the old app or did you already improved the code from the new, next week to be released app?
If first one is the case, don't you think the new release will negate all your efforts? I'm pretty sure there will be some changes made to scheduler and WU format when the new app is out.
So why did you wait so long until the new app comes out instead of releasing it way earlier...

---

Member of BOINC@Heidelberg and ATA!

My BOINCstats

Thanks for the work on the new application Milksop, Dave, Crunch3r et al.

I look forwards to the link to Milksop's code release tonight, and will install ASAP. After that, when the slightly modified MW version comes along, next week/early the week after, then I will download and install that version in it's place.

I assume a straight download and install is all that is needed to replace the exisitng client? For various projects using enhanced clients (SETI, Einstein, etc) the usual, and recommended, process was -

1. Run down the existing cache.
2. Close BOINC Manager.
3. In the BOINC projects folder (wherever it lay) lable the old client files/dlls/app_info files as .old (to deactivate them but leave them ready) in case).
4. Install the new client.
5. Restart BOINC and download WUs.

Is the process similar to this description?

OK -- does that mean the improved application release will wait until next week and be handled on the Milkyway site?

I assume we'd need to clear out the existing work units first before moving to the new application (that's what goes on with the optimized SETI application) -- is that correct?

No. I will post the links today as said.

You don't need to clear the caches. As mentioned, the app is completely compatible and able to continue the computation of WUs started with the official app from a checkpoint. You only need to put the app and the supplied app_info.xml into the MW folder and restart BOINC, that's all. I will detail a special case when the links are posted.

Now I just kind of wondered...
Milksop, is your app an improvement of the old app or did you already improved the code from the new, next week to be released app?
If first one is the case, don't you think the new release will negate all your efforts? I'm pretty sure there will be some changes made to scheduler and WU format when the new app is out.
So why did you wait so long until the new app comes out instead of releasing it way earlier...

What gets released today is the improved old app. For the reasons you mentioned, most probably it would not be possible to run the "real" new app without accompanying changes at the project side. But as said by the project, the changes implemented here are carried over to the new one as the the overall algorithm does not change that much. I don't see how this could negate my efforts.

Concerning your question why it gets released now, I have to say that we initially had not planned to release it for the reasons I wrote maybe 10 times here in the discussions. Actually the goal was to push for the release of the new and improved app by the project. It was announced by MW as early as March and you will find messages about two months old in the forum that it is basicaly ready and will be released within 2 weeks. So maybe the remaining week will be a long one ;) But to me it appears Dave is really pushing it this time.

So when looking at the time scales, the 4 weeks we gave the project isn't exactly that long. If you followed the discussions closely, you will remember that I made a statement that a deviation from the non-distribution policy would be possible only two weeks ago. We had to think about a rollout scheme as undisruptive as possible that does not necessarily ends up in chaos and the possible death of the project. The decision and also the date was determined before the announcement of the possible release of the new app by the project (that's why I was so sure about the release here and hinted to halloween before Dave stated a date). It would have happened the way it happens in almost any case. Well, if the new app would have been distributed by the project this week... But the delay fits the picture ;)

...as the the overall algorithm does not change that much. I don't see how this could negate my efforts.

Okay, that kinda explains it. I erroneously thought, the new app would be completely different from the one now used. I'm not very familiar with the source code and the stuff "behind the scenes". Sorry for that. ;-)

Actually the goal was to push for the release of the new and improved app by the project. It was announced by MW as early as March and you will find messages about two months old in the forum that it is basicaly ready and will be released within 2 weeks.

I know. I followed the whole story, but watched it from the background. ;-)
There were times I would liked to say something to it, but decided otherwise. Maybe it was a mistake to not discuss with, but whatever...

If you followed the discussions closely, you will remember that I made a statement that a deviation from the non-distribution policy would be possible only two weeks ago.

Was much stuff to read, I don't remember that anymore unfortunately...

We had to think about a rollout scheme as undisruptive as possible that does not necessarily ends up in chaos and the possible death of the project.

I said that somewhere else recently, but I don't think MW will "die" that fast only because of a credit inflation...
Even this one project (I think it began with P...) which obviously mistreated his own users by banishment (which I think is way worser than credit abuse) and was boycotted by some users still existed and continued the work after the happenings - although it's not working much around this time now...

As for me, I will give your app a try, too of course, but I won't use it 'til the end, only for finally reaching my 100k milestone probably... ;-)

---

Member of BOINC@Heidelberg and ATA!

My BOINCstats

Watching the time and counting down to 18.00 GMT

Watching the time and counting down to 18.00 GMT

Most probably I will post it a bit earlier, as I won't have time at 18:00 UTC.

Good news ! :-)

Is a Mac OS client version also foreseen and if yes, for x86 only or also PPC ?

Thx !

In the moment I have only versions for Win and Linux on x86er CPUs. But I told already the project staff that they could again ask Crunch3r, as he has some experience in compiling things for OSX.
I can't make a promise, but maybe I could get access to a PPC and create a version for it over the weekend.

Btw., I planned to post a link to this topic in the fora/forums of some of the big teams. But frankly, my French isn't sufficient for this. It would be great if you could bring this to the attention of your team members.

No problem, it was actually already done on the Alliance Francophone newsgroup a couple of minutes after my post. :-)

That way there is enough time for everyone to prepare for the release and for the project to adjust their credit scheme (or distribute the apps as default). I would suggest 4 to 5 credits for the long WUs (now 260 credits) and something around 2.5 credits for the medium ones (now 139.29 credits). That would still be a quite generous level (comparable to the current state).

What is the expected runtime for this new app?

What I mean is, for the suggested 4 - 5 credits how long will a wu be running?
I realise this is dependent on CPU speed!

---

What is the expected runtime for this new app?

What I mean is, for the suggested 4 - 5 credits how long will a wu be running?
I realise this is dependent on CPU speed!

Look here, some people give examples from their testings.
It is a runtime around Crunch3rs first coded app, for me that took 6 minutes.

---

Member of BOINC@Heidelberg and ATA!

My BOINCstats