Well you managed to slide right past my point.

It has been said that the clients are not equal.
The Mac client does not use Altivec therefore missing a major part of the processor.
Linux has always been shunned by the small credit for their work. The new credit system shows that they were shortchanged.
Windows gets the most attention by the developers.
None of these clients are using all of the processors power in any platform.
This is what needs to get fixed for my computers to go back online.
If all the clients were written to take advantage of all the features of operating system including hardware (Optimized) there would be a balance between them.
Then the project would get so much more science done than they know what to do with.

You seem to assume that this is a trivial task? It is not. The project staff dedicated a lot of energy in optimizing Rosetta for all platforms. In another post you assume that they focus mainly on Windows, which they do not. Where do you have your information that they do? Rosetta is a very complex app and much harder to optimize as for example Seti or Einstein, which basically search for something which isn't there in noise (pun intended). ;-)

As I said some people already looked at the code and the first they agreed on is, that further optimizations are not trivial and obvious for all platforms. The G4 and G5 processors just are not that efficient for the Rosetta app, as for many other apps. For some apps they are much more efficient. That is basically what makes the difference between different architectures.

Bottom line: The develpers put as much effort in optimizing for Mac as reasonable and possible. The current dissatisfying speed is not due to laziness from them. If you happen to know any person who might be able to suggest on concrete optimizations for Altivec please feel free to contact me: joachim@iwanuschka.de

I don't agree with you GHz relation.

If the 12.3GHz under Linux worked slower than the 12.3GHz under Windows then it does less work. Be it the fault of the Operating System, the Compiler that compiler Rosetta, so what since it's crunches at a slower wait.
Similar thing with the Mac.
It's the through put of actual useful calculations that is the measure. Not just a frequency of cycles on a CPU.

A computer that does a job/task/work unit should get the exact same 'points/credit/work done' as another computer that does that same workunit no matter what type of computer it is.


What I would like to know is if the Rosetta Windows compile is slower or faster than the Rosetta Linux compile.
Has anyone seen (or can anyone run a task manually) to find out ?

Well you managed to slide right past my point.

It has been said that the clients are not equal.
The Mac client does not use Altivec therefore missing a major part of the processor.
Linux has always been shunned by the small credit for their work. The new credit system shows that they were shortchanged.
Windows gets the most attention by the developers.
None of these clients are using all of the processors power in any platform.
This is what needs to get fixed for my computers to go back online.
If all the clients were written to take advantage of all the features of operating system including hardware (Optimized) there would be a balance between them.
Then the project would get so much more science done than they know what to do with.

And you missed my point, Not everything is equal.


As for the linux points lower, nothing to do with Rosetta and not their fault, that is BOINC's doing (and hence Rosetta have now treid to do something about that) It does not make it any faster or slower though).


I said if (or maybe implied ;-) If you put Linux and Windows on a computer (the same computer, dual boot), if the Linux Rosetta takes 1hr longer than the Windows Rosetta then Linux credits/hr should be lower no matter what. It does not matter what is under the hood.

But you seem to miss something,
The Rosetta code needs to remain compatible across platforms, putting optimisations in that cannot be switch of and/or maintained properly would be a nightmare. It is used in more places than just here.
So like most places they use a compiler to do the hard work, if the compilers crap then don't blame Rosetta, blame to people that make the compiler.

They also have limited resources, this is not normally money but time and (wo)man hours. So if it was my time I would aim for Windows platform since by far the majority of users run Windows. You improve the code a little for windows users with less effort and you get a large reward. You mess about adjusting the code for a limited user base with 'Altivec' dodars where it's something the compiler should do for them, you waste a lot of time when it could have ben used for their science part (remember they are scientist not dedicated programmers).

I would much rather do somthing not using all the power than do nothing at all. My way at least something gets done and I can help while they try to improve things.


That being said, if you want to look over the code OR pay for somebady to look over the code for them and dedicate their time for you Altivec code then I think they would be very happy :-)

For Rosetta

---

Team mauisun.org

With regard to "optimization" of the Rosetta application(s), the info from the projet team, along with many constructive discussion threads, suggests that, to Rosetta, the "optimal" application is one that can be modified cheaply and quickly, since the algorithms and heuristics that the application uses are under constant revision in order to further the science. While speed and memory footprint are important--and have been addressed by the developers--they are secondary to having an application that is "agile" and able to be redirected with data files, along with code changes. "Optimal" just doesn't always mean "fast."
I am fond of the old saying about system development that has served me and my colleagues well over the years:

----> "Make it work first, then make it work fast." <----

Edit: BBCode-challenged typos...

This to David Baker
It seems you have a No Win situation here
#1 you can not take away the points given in the past for the members that used opt. clients because is was said it was said by your staff that it was not against Rosetta's rules to do so
#2 If you do not reduce the past points the other 1/2 will feel they have been cheated. And will be mad at Rosetta for saying it was OK to use the opt. clients
So you (Rosetta) are in a No win situation here
Here is one thing you might consider to end this mater
#1 End Rosetta close the points

#2 Start Rosetta 2 with a new secure client that is run on the new point system and lets all start all over again

This post was from the call of carl.h for "O.k. guys...Positive input only. That means no tearing the other guys post.

---

If You Want The Best You Must forget The Rest
---------------And Join Free-DC----------------

#2 There is no other half. Another group of few people might leave when they finally realise that the credits will not be backdated.

---

list of my results

This to David Baker
It seems you have a No Win situation here
#1 you can not take away the points given in the past for the members that used opt. clients because is was said it was said by your staff that it was not against Rosetta's rules to do so
#2 If you do not reduce the past points the other 1/2 will feel they have been cheated. And will be mad at Rosetta for saying it was OK to use the opt. clients
So you (Rosetta) are in a No win situation here
Here is one thing you might consider to end this mater
#1 End Rosetta close the points

#2 Start Rosetta 2 with a new secure client that is run on the new point system and lets all start all over again

This post was from the call of carl.h for "O.k. guys...Positive input only. That means no tearing the other guys post.

The situation is indeed unpleasant. Backdating was mainly discarded because of the tumultuous protest from XtremeSystem. In the meantime they left anyway and now we have exactly the situation as described here. I doubt this was intentionally or planned by XS but the result is the same.

That said I don't think starting over with zero credit will be a solution. Others will protest and leave because of that and in the end more people will be unhappy. Or could you guarantee lots of new hosts if the credit system was reset? ;-)

XS_Vietnam_Soldiers,

I could present it at the boinc workshop (unless he/she plans to, I don't know who is going to attend) as I am sure all would be interested.

I saw this, looks interesting, and I think it would be good for some crunchers to go -
If there is still room. It is suppose to be open for all interested in BOINC, not just developers but crunchers as well.. Might be a good opportunity to chat about the crunchers frusterations and idea's.

---

SETI.USA

..and now I am gone.
Movieman

Does this mean you are quiting the project ? Just as our SIMAP challenge has ended and I am going to be coming back online, at least partly -
And I wanted to give you a good chase

Where are you going ? Maybe I should go to your team web site and say hi,

Stop in and visit me at www.setiusa.net

---

SETI.USA

This to David Baker
#1 you can not take away the points given in the past for the members that used opt. clients because is was said it was said by your staff that it was not against Rosetta's rules to do so
#2 If you do not reduce the past points the other 1/2 will feel they have been cheated. And will be mad at Rosetta for saying it was OK to use the opt. clients
So you (Rosetta) are in a No win situation here
Here is one thing you might consider to end this mater

The crunchers that feel they have been cheated by not using the optimized BOINC app.. well - all I can say to that is hmmm.. if they feel this way, maybe they should have upgraded.

If they do not care about credits and only the science then it should not matter one way or another. But don't complain.

Credits will not be back dated, so everyone just get over it and get on with it.

---

SETI.USA

From Tony DeBari

The crunchers that feel they have been cheated by not using the optimized BOINC app.. well - all I can say to that is hmmm.. if they feel this way, maybe they should have upgraded.

Third-party BOINC core clients are not an upgrade. They are a niche product created for a specific purpose. Using them to increase credit output - but not science output - by exploiting Rosetta's previous quorum-of-one credit system (and having the project staff regard this behavior as acceptable) irks people who believe in equal credit for equal work.

If they do not care about credits and only the science then it should not matter one way or another. But don't complain.

I care about the science first and foremost. And even though some on this board would label me a ZeroRACer, I care about the credits in the sense that the only reason someone should have a higher RAC than me is if they actually produce more science output than I do. I have the utmost respect for people who have the dedication and the resources to crunch more work in one hour than I can in a week.

[flamebait comment deleted]

Credits will not be back dated, so everyone just get over it and get on with it.

This I agree with. There is so much to pick apart with the new credit system that there's no need to keep rehashing the old one.

Regards,

-- Tony

---

Rosetta Moderator: Mod.Tymbrimi
ROSETTA@home FAQ
Moderator Contact

The crunchers that feel they have been cheated by not using the optimized BOINC app.. well - all I can say to that is hmmm.. if they feel this way, maybe they should have upgraded.

It has been stated b4 that it was impossible for us to do that because then we would have been causing the same problem at other projects that happened here with overclaiming.

---

98SE XP2500+ @ 2.1GHz Boinc v5.8.8

And God said"Let there be light."But then the program crashed because he was trying to access the 'light' property of a NULL universe pointer.

Steve, why do other projects matter ? It`s been stated that other projects did/do not have this problem due to quorums or other ?

Do you think in those other projects that work on quorums you have at anytime gained by getting extra from other crunchers in the quorum who have used 5.5 or such ? Did you give the points back ?

---

Not all Czech`s bounce but I`d like to try with Barbar ;-)

Make no mistake This IS the TEDDIES TEAM.

Steve, why do other projects matter ? It`s been stated that other projects did/do not have this problem due to quorums or other ?

Do you think in those other projects that work on quorums you have at anytime gained by getting extra from other crunchers in the quorum who have used 5.5 or such ? Did you give the points back ?

Other projects matter. Perhaps not for you, but for a lot of people, especially from the big teams, like BoincSynergy, SETI.Germany, SETI.USA and such.

There is a standard in BOINC, and though it's not a really good and perfect one, the "optimised" 5.5.0 went off far too much on non-optimised applications like Rosetta, LHC, QMC and others. LHC has a quorum, so it didn't really matter, as 5.5.0 is here as everywhere just a client for a minority, and it's sorted out by the quorum. Rosetta and QMC don't have a quorum, so those who use it get an unfair advantage in credits without doing more in science. Why should the majority (stock client) follow a vocal minority ("opt." client) to get even? If you just keep everything in the standard BOINC, it's just fine.

It isn`t sorted by a quorum though.

If say, there is a quorum of three and two are claiming using 5.5, the higher points are awarded even to the std.

I`m not arguing the point for or against 5.5 just stating what I believe are facts.

---

Not all Czech`s bounce but I`d like to try with Barbar ;-)

Make no mistake This IS the TEDDIES TEAM.

Steve, why do other projects matter ? It`s been stated that other projects did/do not have this problem due to quorums or other ?

Do you think in those other projects that work on quorums you have at anytime gained by getting extra from other crunchers in the quorum who have used 5.5 or such ? Did you give the points back ?

Other projects matter. Perhaps not for you, but for a lot of people, especially from the big teams, like BoincSynergy, SETI.Germany, SETI.USA and such.

There is a standard in BOINC, and though it's not a really good and perfect one, the "optimised" 5.5.0 went off far too much on non-optimised applications like Rosetta, LHC, QMC and others. LHC has a quorum, so it didn't really matter, as 5.5.0 is here as everywhere just a client for a minority, and it's sorted out by the quorum. Rosetta and QMC don't have a quorum, so those who use it get an unfair advantage in credits without doing more in science. Why should the majority (stock client) follow a vocal minority ("opt." client) to get even? If you just keep everything in the standard BOINC, it's just fine.

Hows about this then ? Picked at random. I give no names for obvious reasons.

A Seti.Germany member.

<core_client_version>5.5.0</core_client_version>
<stderr_txt>
# random seed: 3432459
# cpu_run_time_pref: 28800
# DONE :: 1 starting structures built 79 (nstruct) times
# This process generated 80 decoys from 80 attempts


A BoincSynergy member.

<core_client_version>5.5.0</core_client_version>
<stderr_txt>
# random seed: 1079336
# cpu_run_time_pref: 3600
# DONE :: 1 starting structures built 6 (nstruct) times
# This process generated 7 decoys from 7 attempts


A SETI.USA member.

<core_client_version>5.5.0</core_client_version>
<stderr_txt>
# random seed: 1165285
# cpu_run_time_pref: 10800
# DONE :: 1 starting structures built 17 (nstruct) times
# This process generated 17 decoys from 17 attempts

---

You will find Crunch3r core clients in all bigger teams, especially in those with a strong share for SETI@Home.

None of those more or less inhomogenous teams has any control over the client that the people use. Compare the total number of crunchers for SETI.Germany with the number of people using the SETI.Germany forum frequently and you will understand why it is impossible to influence anything there.

Those teams have quite an average mix of client types, from 4.19 up to the latest one, including all flavours of optimized clients - keyword is "average" here, which is quite a difference to a team which has nearly nothing but the highest optimized clients.


p.s.: Imo. the member movement shows a very clear picture : Those who do use 5.5.0 and still stay here are at Rosetta are really interested to help the Rosetta science. Those who use 5.5.0 and are going to QAH now, are interested only in the credits. I wonder where they will go when QAH stops using the BOINC benchmark system (it has been announced).

Ananas whilst your post stays analytical, I find it relevant and maybe useful, howether your ps tainted it and was an assumption.

Since all Boinc scores are skewed, and the people who called the most for a fairer credit system to be implemented are mostly cross project have they asked at all projects for such ?

Have these people asked Boinc to implement a fair system and reset the scores ?

---

Not all Czech`s bounce but I`d like to try with Barbar ;-)

Make no mistake This IS the TEDDIES TEAM.

The BOINC benchmark has been a matter of many complaints, especially in the beginning when SETI migrated. We often pointed out the flaws. With 4.13 and 4.19, it was possible to double the benchmark results by making a program with high priority eat all CPU power of one CPU on a dual CPU host, 5.x isn't much better, a background program using no CPU time is able to reduce the integer benchmark to 10%, Linux still has totally crappy results - the BOINC benchmark is the most unreliable instrument to measure work done.

Many of those who complained gave up sooner or later and got used to claims varying by factor 3, but there the quorum reduced the impact of the flaws. With projects like Rosetta and QAH those flaws got a new dimension.

The reason, why Akos F. is not interested in helping Rosetta has been the quorum together with the benchmark. Maybe with the new system it would be possible to make him tweak the project clients and support the Rosetta developers team. (p.s.: I bet he will say : "75% speedup is possible" ;-) )

The first thing some people ask in new projects is : when will you switch to fixed credits?

Many multiproject BOINC crunchers are definitely not happy with this situation.

There is a standard in BOINC, and though it's not a really good and perfect one, the "optimised" 5.5.0 went off far too much on non-optimised applications like Rosetta, LHC, QMC and others. LHC has a quorum, so it didn't really matter, as 5.5.0 is here as everywhere just a client for a minority, and it's sorted out by the quorum. Rosetta and QMC don't have a quorum, so those who use it get an unfair advantage in credits without doing more in science. Why should the majority (stock client) follow a vocal minority ("opt." client) to get even? If you just keep everything in the standard BOINC, it's just fine.

Saenger, I counted 4 different flavours of overclaiming clients on the first page alone of SETI.Germany members (your team). I could'nt check the hidden ones on the first page BTW. I stopped after the first page. Hmmmmm 450 members ? That is a lot of members.

---

The is going to come a time when the corss project credit people will have to give up, isnce not all porjects will be CPU based. How do you start scoring projects that do not do work based on CPU strength (like XtremLabs, they tried fixed, one or two people complained so they changed it to Boinc standard, but Xtrem where right slow CPU's do jsut as much work as a super duper fast CPU, so all should get the same). When GPU or projects like DIMES start using boinc, what then ?

Cross project credits is also broken like with seti 'optimised' apps since the optimised apps are not really doing more work (they are just more efficient, they do less work to get teh same result so why should they get more than me here for the same of flip-flops ;-)

---

Team mauisun.org