Questions and Answers :
Macintosh :
SaH 5.22 oddness
Message board moderation
Previous · 1 · 2 · 3 · Next
Author | Message |
---|---|
Bob Merrill Send message Joined: 7 Jun 99 Posts: 120 Credit: 8,531,677 RAC: 19 |
Also work units seem to be taking about 3 times longer to run. I don't know if we are being issued bigger work units or if the program is slow. Have you also noticed that the clock in the screen saver graphics doesn't work now and the Seti @ Home graphic is missing? I also have noticed that the work units take a lot longer to run on 522. It looks like this version has a lot of new bug since they fixed the old one. |
Dotsch Send message Joined: 9 Jun 99 Posts: 2422 Credit: 919,393 RAC: 0 |
[quote]Also, as I have asked before, have you verified your results with other PPC hosts which claims the same credit. The runtimes of the SETI task are variable and so credits are variable. You can only compare WUs with the same claimed credits. No, 5.22 claimes more credits as 5.1x, but the more claimed credits are not granted if the resuls was vailidated with a 5.1x application. Please read my previos posts.
Which problem ? Your credit problem ? - It is no problem. Please read my last posts, there are the explanations why it is no problem. Your impression that the application is slow ? - Without any excatly describition how you have measured/compared it with the other systems, I can nothing do for you. Please read my last postings where I have asked you about the compare of WUs with the same clamed credits and variable WU length. I have the impression, that you searching for problems which are not existant and talking something without any facts. |
johnh Send message Joined: 16 Apr 03 Posts: 8 Credit: 6,684,929 RAC: 1 |
2 MHz iMAC G5 PPC Running under OSX 10.4.10 with SETI 5.22. Same problem with WUs. Takes much longer to finish. I have no trouble with Einstein. Additional problem I noted: I run Einstein 50% and SETI 50% and SETI will restart when time comes to run and CPU time may increase, but time to completion wil not decrease and sometimes evem increase. I have suspended SETI until 5.22 many problems get fixed; Also work units seem to be taking about 3 times longer to run. I don't know if we are being issued bigger work units or if the program is slow. |
Dena Wiltsie Send message Joined: 19 Apr 01 Posts: 1628 Credit: 24,230,968 RAC: 26 |
Dotsch, click on my name. Look at the bottom of the screen. I am running 100% on both project and daily credits should be close to the same. THEY ARE NOT. Both task have been running. All other available information is saying the same thing. THERE IS A PROBLEM with 5.22. You must trust me when I say I know something about computers as I have been working with them for 35 years. If I don't know something I will tell you because I have no pride. I only care about getting the problem fixed and I don't care if I am wrong as long as the problem is fixed. |
Dotsch Send message Joined: 9 Jun 99 Posts: 2422 Credit: 919,393 RAC: 0 |
2 MHz iMAC G5 PPC Running under OSX 10.4.10 with SETI 5.22. Same problem with WUs. Takes much longer to finish. I have no trouble with Einstein. This is the point I had tried to explain in my recent posts.... The crunching time of the SETI WUs differs. This resulting of the recording of the different angle rates (AR) at the telescope. So you can only compare crunching times from WUs with excatly the same AR to compare the performance of the application. The only real method to make a benchmark of the SETI application is to take WUs from different angle rates and let them running against the different application versions. Einstein had - I did not know excatly if it is still so - WUs with the same crunching times. Additional problem I noted: I run Einstein 50% and SETI 50% and SETI will restart when time comes to run and CPU time may increase, but time to completion wil not decrease and sometimes evem increase. This is harmless, looks like a problem with the update of the BOINC client. Does the CPU time and progress counter has the same behavior at the screensaver ?
I see no reason, why you not should crunch. The bugs are only minor problems. |
Dotsch Send message Joined: 9 Jun 99 Posts: 2422 Credit: 919,393 RAC: 0 |
Dotsch, click on my name. Look at the bottom of the screen. I am running 100% on both project and daily credits should be close to the same. THEY ARE NOT. No. The credits SHOULD not be the same between differnent projects and this is OK !!! Please read my previous posts, there I have written that the credit assignment is a project specifiy issue. Each project assign the credits like they think it is good. There are activities to make the credits more compareable between different projects, like you wan't it.
Which information ? Which problem ? - Your problem reports are totaly useless. I have asked you more times how you have compared the crunching times and how you mean that 5.22 be slower. Also I gave you the background why the different crunching times are OK and normal. Also I the difference in credit assingments has I described you more times and states that this is OK.
Your permanent statings about your computer knowledge did not impress me. If you have a such knowledge, why did you not awnser my questions or gave me a detailed feedback ? - Till you not gave any feedback, I can nothing do for you and did not belive your "there is a problem" crusade. The only issues I see in 5.22 are the sometimes higher memory requirements, pixy's graphic issues and the CPU time/progress counter issues. - In my opinion only minor issues. So I not see any cause why not to crunch 5.22. |
Dena Wiltsie Send message Joined: 19 Apr 01 Posts: 1628 Credit: 24,230,968 RAC: 26 |
Dotsch, I have been crunching these two projects for close to a year. I know the credits are not always the same but there is a difference between 5% and 50% and we are seeing the latter. I did not see this with 5.18. Plain and simple, I don't think you belive anything anyone who is having this problem because you are not seeing it on your own system. We are hurting with 5.22 and I don't blame anyone who is having these problem for quitting if this is the type of problem resolution they have to look forward to. 5.22 is not a clean release but it will run after a fact if you don't require to much from it. We all hope 5.23 will resolve these problems so we can get back to work crunching for SETI. The problem we have is that there has been no information about efforts to resolve these problems. If we knew a fix was on the way it would help keep us interested in the project. |
Dotsch Send message Joined: 9 Jun 99 Posts: 2422 Credit: 919,393 RAC: 0 |
The problem we have is that there has been no information about efforts to resolve these problems. Please RTFM, and try to understand how credits are assigned over the projects and how the credits are assigned at SETI (FLOPS counting). The inoffical BOINC wiki is a good stating point. The credits are variable here, depending on the AR. The FLOP counting depends in a range on the CPU types. In the end, some batches with some ARs can gave more credits as some other batches, because some ARs calculate more FLOPS and are so weighted a little bit more. So a not constant RAC at SETI is OK and normal, and I see still no "problem" in the credit assignments. Also, with such imprencipe describtions you gave, I can not clasify your "problem" and so I can/wan't not talk with the developers about this. For me is here EOD of your "credits problem", till you give some detailed information. |
Dena Wiltsie Send message Joined: 19 Apr 01 Posts: 1628 Credit: 24,230,968 RAC: 26 |
Dotsch, you have not worked with software or bugs very much because this thread is loaded with information and unless ask for something that we would not normaly be able to provide, you will not get it. Just saying you want numbers will not do it. You need to tell us what numbers you want and where to get them. Also, when you have worked with software as long as I have, you develope a 6th sense about bugs. I have had customers call me and just by the fact they called me, I know what the problems was. Other times, without looking at the code, other programmers started telling me what they wanted to do and I knew where the problem was. All information is important even if it only gives you a starting point to ask questions. You want us to hand you a solution and if you are a gate keeper, it is your job to ask the question, not our to figure out what will make you happy so you will pass it on. If the develpers are not aware of the things in this thread, you are not doing your job. Also, no problem is minor to the person who has to deal with it. I currently work on a device that has to come as close to 100% uptime as possible. With over 100,000 lines assembler, this is not easy, but we have units that stay up for over a year without problems. I don't expect SETI to meet this standard but I would hope when problems like these are found, they are addressed. |
Dotsch Send message Joined: 9 Jun 99 Posts: 2422 Credit: 919,393 RAC: 0 |
Dena, I have written you a lot question what I would to know, which you newer awnsered. If you where a such a good software guru as you write, where are the facts ? - Without any facts I can not do anything for you. |
Dena Wiltsie Send message Joined: 19 Apr 01 Posts: 1628 Credit: 24,230,968 RAC: 26 |
Dena, I have written you a lot question what I would to know, which you newer awnsered. If you where a such a good software guru as you write, where are the facts ? - Without any facts I can not do anything for you. Dotsch, you must be from Missouri. Click on my name, You will see my credits are down. I am not the only one. If you can't beleive SETI's own numbers, we have a problem. Jobs that ran under 6 hours are now taking 12 hours but only grant 6 hours of credit. My results for SETI would overflow into two screens, but now only take about 1/2 screen. I can send you screen shots if you tell me what you want and where you want them. If you need something else, explain how to get the numbers you want, because the ones I know about already prove the problem. Show Graphic only sometimes comes up when you click on it. Show Graphics screen has a big gray square in it. Picture is elsewhere in this form. Theses are problems I have seen and can reproduce. With SETI, I am an end user. I can spot flaws in the software BUT HAVE NOT LOOKED AT THE SOURCE. I currently program in 68000 assembler on a proprietary platform that looks nothing like SETI. If you are going to work the problem, give me SPECIFIC request an I will go my best to fill them. If you are going to call the above information lies or useless, don't bother responding because you will not believe any other response I generate. And, don't forget, we are customers of your service. Your job is to do your best to make us happy by resolving our problems. If you are unable to solve a problem, there is no shame in bumping it up to the next level, or admitting to the customer the problem can not be solved. Ths is a rule I live by every day. And by the way, what i am really good at is finding the bug no one else could find, coming up with a solution to a problem that evaded everyone else and taking a large, slow application and making it a clean fast application. If you want someone who can turn out volumes of code fast, I am not the person you are after. Most programmers out there are not like me so take that into consideration when dealing with me. |
hanyou23 Send message Joined: 14 May 00 Posts: 46 Credit: 2,357,323 RAC: 0 |
Common guys, calm down here. All of this arguing about who has better credentials, etc. is counter-productive. Let's try to pinpoint the problem and try to find a solution without trying to knock each other over. I'm not trying to be a forum moderator here, but this is ridiculous. It's about time one starts to ignore the other like adults. Geesh! Why do Tech Specs always have to be (territorial) like this >_> ?... Ok, anyways, one report I have to give is every time I quit the BOINC app and reenter it, the message log always clears, like this: Thu Jul 26 15:06:59 2007||Starting BOINC client version 5.10.7 for powerpc-apple-darwin Thu Jul 26 15:06:59 2007||log flags: task, file_xfer, sched_ops Thu Jul 26 15:06:59 2007||Libraries: libcurl/7.16.1 OpenSSL/0.9.7l zlib/1.2.3 Thu Jul 26 15:06:59 2007||Data directory: /Library/Application Support/BOINC Data Thu Jul 26 15:06:59 2007||Processor: 1 Power Macintosh Power Macintosh [Power Macintosh Model PowerMac3,5] [AltiVec] Thu Jul 26 15:06:59 2007||Processor features: Thu Jul 26 15:06:59 2007||Memory: 1.00 GB physical, 32.68 GB virtual Thu Jul 26 15:06:59 2007||Disk: 111.65 GB total, 32.44 GB free Thu Jul 26 15:06:59 2007|Einstein@Home|URL: http://einstein.phys.uwm.edu/; Computer ID: 932951; location: home; project prefs: default Thu Jul 26 15:06:59 2007|SETI@home|URL: http://setiathome.berkeley.edu/; Computer ID: 3051447; location: home; project prefs: default Thu Jul 26 15:06:59 2007||General prefs: from SETI@home (last modified 2007-06-19 18:05:28) Thu Jul 26 15:06:59 2007||Host location: home Thu Jul 26 15:06:59 2007||General prefs: no separate prefs for home; using your defaults Thu Jul 26 15:06:59 2007||Preferences limit memory usage when active to 512.00MB Thu Jul 26 15:06:59 2007||Preferences limit memory usage when idle to 921.60MB Thu Jul 26 15:06:59 2007||Preferences limit disk usage to 31.04GB I never really have a clue what's been going on. Before the log was always complete whether I quitted it & reopened or not (holding every message since start up of the computer), but now it always clears itself when opening the app. Any clue why it's doing this now??? I think it began at 5.22. Sorry for my straightforwardness ~_~ ..... |
Dena Wiltsie Send message Joined: 19 Apr 01 Posts: 1628 Credit: 24,230,968 RAC: 26 |
I reserve the right for someone to provide a better response. I thought the console always cleared when you restarted Bionc. Clearing the log would be a function of Bionc as all task log there. I took advantage of the fact it cleared while tracking down a problem in World Community Grid. I would run till I had the failure, capture the log and restart fresh. |
C Send message Joined: 3 Apr 99 Posts: 240 Credit: 7,716,977 RAC: 0 |
Dena: Your computer is definitely operating very slowly. My G5 at work takes around 12.3K sec to complete a work unit and receive 65 credits for it. Here's one: http://setiathome.berkeley.edu/result.php?resultid=576133634 I'm running Alex's optimized app on it. C Join Team MacNN |
Bob Merrill Send message Joined: 7 Jun 99 Posts: 120 Credit: 8,531,677 RAC: 19 |
Dena: Your running 5.13, the problems are with 5.22... |
Bob Foertsch Send message Joined: 19 Jun 99 Posts: 25 Credit: 11,969,667 RAC: 25 |
See above. I had network problems while making this post. |
Bob Foertsch Send message Joined: 19 Jun 99 Posts: 25 Credit: 11,969,667 RAC: 25 |
It has been over a week since my initial post started this thread and my (and obviously many others) experience with SaH 5.22 on PPC Macs is basically unchanged. - My machine is still running 5 degrees F hotter than it did with SaH 5.18. (It typically ran with one processor at 157 and the other 146, now it runs 163 and 151) - The client routinely crashes 4 or 5 times a day. - The client does wierd stuff with system memory. (The application typically uses 65MB but on occasion has gobbled up over 900MB and reeked havoc on my machine) - The client has problems managing tasks. This might be related to the last point. (Often the task status says 'waiting for memory' when there is over 1GB of memory free. Sometimes no tasks are running - both active tasks say 'waiting for memory') - The screensaver still has issues. Sometimes it does not start. When it does start, there is this gray box on the left towards the bottom. - I had been getting 300 or so credits per day for the WUs processed. That number over the last 3 or 4 days is now abround 100. I, and I'm sure many PPC users are frustrated by this development. The fact that there has not been any kind of 'official' statement or status regarding all of these reports doesn't help the issue. Neither do some of the statements made by volunteer testers. Basically the SaH 5.22 is broken/crippled and that is what needs to get addressed. If the code developers need additional information there are several folks here who are experiencing the problems first-hand and would be willing (at least I believe they are) to provide any further information in an effort to identify how the application is behaving in an attempt to get it fixed. |
Dotsch Send message Joined: 9 Jun 99 Posts: 2422 Credit: 919,393 RAC: 0 |
It has been over a week since my initial post started this thread and my (and obviously many others) experience with SaH 5.22 on PPC Macs is basically unchanged. Sorry, please be patient. SETI is underfunded, has less money and personal. A lot of thing are made by volunteers, which have go to work every day and has otherthings todoy too, as me for example. Also, I think your tone is a little bit rude. We are no software company where you can put in a call an get a awnser some minutes later back. There are many of other posts here in this thread, so it could also happen that some messages would be oversean and not readen. Also, I think it is better to make for each problem an own thread an post there the problems. It is better to handle for us volunteers.
We have too look at this issue... - The client routinely crashes 4 or 5 times a day. Your clients are hidden. I can not look into the logs. With which message does the client crash ? - Could you please look at the WUs at the website and post the error messages. - The client does wierd stuff with system memory. (The application typically uses 65MB but on occasion has gobbled up over 900MB and reeked havoc on my machine)151) We have too look at this issue, too... - The client has problems managing tasks. This might be related to the last point. (Often the task status says 'waiting for memory' when there is over 1GB of memory free. Sometimes no tasks are running - both active tasks say 'waiting for memory') The client stops the application, because the high memory allocation you have reported in the previos error report.
Eventualy a problem in the screensaver code... - I had been getting 300 or so credits per day for the WUs processed. That number over the last 3 or 4 days is now abround 100. See my discusion with Dena, I have explained that the WUs, crunching times, credits are variable and the RAC can change. I, and I'm sure many PPC users are frustrated by this development. The fact that there has not been any kind of 'official' statement or status regarding all of these reports doesn't help the issue. Neither do some of the statements made by volunteer testers. Please read from the begining, and you will see, that I try to catch up this problems, clasify them and report them to the devs. Also, as I have stated before, it is easier to open a new thread for each problem, instead posting everything in one single thread. - And I mean the other posters here, too. Also, I think, you can be happy that there IS a application where the most bugs are fixed (as the G3 Bug), and running one the most systems stable and you can crunch some WUs. |
Dena Wiltsie Send message Joined: 19 Apr 01 Posts: 1628 Credit: 24,230,968 RAC: 26 |
If the developers are aware of our problems and scratching their heads, we are happy. If we know a fix is on the way at some latter date, we should be quite because only a few people on the SETI project receive a pay check for their work. The rest of the SETI effort is paid by that good feeling they have in their heart. To others about chicken apps. I know they exist and can provide good results. The down side is you need to do a manual upgrade and if you miss an important upgrade, you interfere with the project by introducing problem data. Also I am unable to get chicken apps for World Community Grid which means I am not out to score points or I would only run SETI. Chicken apps are not wrong, they are just not for me because I am after a set and forget environment. |
Dotsch Send message Joined: 9 Jun 99 Posts: 2422 Credit: 919,393 RAC: 0 |
I will now try to summarize and comment the different problem reports.... After analysing the whole thread I see now following problems : - System is 5°F hotter than with 5.18 -> Has other users this problem, too ? - Sometimes screensaver did not start - Sometimes screensaver shows strage fields - Memory usage sometimes high -> Has other users this problem, too ? - Bob, are you shure that you use 5.22? - Application crashs -> Mostly happens with 5.21. Bob, are you shure, that you use 5.22 ? Has other users the same problem with 5.22 ? - Please verify, if you have 5.22, and if not, do a reset project - WUs are slower compared with 5.18. -> Hard to reproduced, I need some comparement values from 5.18 results on the same PPC system. Eventualy a application benchmark is needed. Also, please verify, that you compare WUs with the AR (shown in the WU field in the report of the WU) |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.