Posts by Bob Guy

1) Message boards : Number crunching : How to deal with a moderator that (Message 27856) Posted 20 Sep 2006 by Bob Guy Post: Well, I must thank you for your last post, it was a very reasonable response. I'd like more of that kind, please. I want you to know that I read ALL your posts, and if I have a complaint it is that repeating the same thing over and over does not make me more likely to believe you.
2) Message boards : Number crunching : How to deal with a moderator that (Message 27849) Posted 20 Sep 2006 by Bob Guy Post: I've been around pretty much since Rosetta started, just so you know. Call me names if you like, I'll only think the worse of you. My ego doesn't depend on the acclamation of others, although it's nice to get it. Regarding my recent post - I did not bring up the subject, I merely responded.
3) Message boards : Number crunching : How to deal with a moderator that (Message 27842) Posted 20 Sep 2006 by Bob Guy Post: I assume from that this was created as a secondary account for you to flame with. You are making an unsubstantiated speculation, of course. This account is for a single P4 attached to multiple projects to which I donate my spare computing cycles. I think that complies with the designed intention of Boinc.
4) Message boards : Number crunching : How to deal with a moderator that (Message 27837) Posted 20 Sep 2006 by Bob Guy Post: Yes: I am trying to force the moderators to be consistent or to accept that what they did in the past was wrong. I don't thing the moderators want to eliminate me from the database. That move can come back to haunt them. :) Your threats (yes, that's what they are) are starting to be tedious. Continue if you must, I'm sure the MODs will give you enough rope to hang yourself.
5) Message boards : Number crunching : Another discussion on the New Credit System (Message 27832) Posted 20 Sep 2006 by Bob Guy Post: It is unethical to change the rules ex post facto. In fact, it is illegal to makes such laws in the US. Wrong! The Court and the Legislature can do just about as they please as long as it's constitutional. I'd like you to quote the constitutional requirement that you appear to believe applies. In California (a couple of years ago) a law was passed that required smog tests in perpetuity for all vehicles manufactured in 1976 and later. Previously the law had been that any vehicle 30 years old was classified as 'antique' and was exempted from many of the ordinary requirements including smog testing and maintaining an absolutely stock engine. I owned (still own) one of those vehicles, a 1977 Trans-Am 6.6L Limited Edition, which I had kept in the hope that it's value would increase as it passed it's 30th year. This law made my investment greatly decrease in value since the car is not so desirable to collectors owing mainly to the smog and stock engine requirement. If this is not 'ex post facto' perhaps you could explain the difference. The law has not been challenged because it is considered constitutional by all legal experts. I will allow that California is known to have a slightly deranged political climate and legal system. The majority of stock users had no chance to compete... Sure they did. Just upgrade the client like everyone else with the "competition" mindset. This is the weakest argument that has been made since the beginning of all the credit fuss. It is transparently simple to create a Boinc client that 'evens the playing field' with respect to AMD-Intel-P4-Linux WITHOUT the 3X credit inflation. Yet, no one did this. I think we all know why.
6) Message boards : Number crunching : HAL ACPI Multiprocessor back to Uniprocessor (Message 26565) Posted 11 Sep 2006 by Bob Guy Post: Then how come I can select different HAL's if I do it on my Windows XP computer ? and in the link I gave some posts up http://boinc.bakerlab.org/forum_thread.php?id=2264#26457 it says WindowsXP can (a limited choice in WinXP as opposed to any in Win2K) You cannot switch between ACPI and non-ACPI in device manager in WinXP. Extracted from the MS post: 4. On Windows XP and later versions, the ACPI Uniprocessor HAL and the MPS Uniprocessor HAL recognize the existence of more than one processor and report the MP ID. Plug and Play detects that the computer devnode's hardware ID list has changed and moves the devnode back through the "found new hardware" detection process. Therefore, when you add a second processor, the MP files (HAL and kernels) are automatically installed, and you do not have to manually update the driver in Device Manager. ___________________________________ This has just not been my experience - XP SP2 will not do what item #4 says that it will. I believe that the article applies to (original release) vanilla XP and that XP SP2 changes this behavior - such that SP2 disables that feature more or less completely. I can testify that when I changed my processor from single to dual, XP SP2 did NOT change the HAL automatically nor was there an option to do so in Device Manager. References to 'forcing' a particular HAL are only relevant when installing a fresh copy of Windows and possibly relevant when doing a 'repair' install. There may still exist the ability to change the HAL if you started with a current 'Standard PC' HAL, although I can't verify that.
7) Message boards : Number crunching : HAL ACPI Multiprocessor back to Uniprocessor (Message 26519) Posted 10 Sep 2006 by Bob Guy Post: I'm wondering how the system seems to remember that you had a dual-core in it. Did you clear the CMOS after you reinstalled the Sempron? I think you are suggesting that Windows XP should detect that the CPU is now a single processor and then install and use the proper HAL. I'll repeat: that function (reinstalling a different HAL) is deliberately disabled in XP. The only way to change the HAL is by manually installing the proper fileset or by wiping the disk and doing a full install. WinXP will never change the currently installed HAL no matter what CPU is detected and no matter what the BIOS says. The Compaq BIOS may well be dumbed down and may be partly what has led to the current problem. While the motherboard may have been originally intended to support a multiprocessor, Compaq may have deliberately disabled that feature on the board supplied with that particular computer package in order to save costs. Any XP install CD can be used to install XP. All that changes is the XP key and the key is not recorded on the disk. You can borrow any XP install CD and use it with your genuine XP key. You can use a 'burned' CD copy as well as an original CD. You do not violate your license by using a different XP install CD no matter the source. The Compaq support people may gripe that you're not using the recovery method supplied with your computer - let them gripe!
8) Message boards : Number crunching : HAL ACPI Multiprocessor back to Uniprocessor (Message 26460) Posted 9 Sep 2006 by Bob Guy Post: Windows won't start (I think) if you just delete those files. The feature to change the HAL in Device Manager in XP is deliberately disabled by Microsoft - it works for previous versions of Windows but not for XP. See troubleshoot HAL from Microsoft. Below you'll find the names of the relevant files for easier access. You need to replace three files in the system32 folder then restart Windows. The best set of files can be found in the cab files from the XP redistributable SP2 (yes I know it's big) found at Microsoft. You can also use the files (already expanded) found in Windows/ServicePackFiles/i386 - this would be best if you have that folder and it contains the proper files. It is still problematic to change these 3 files because they can have been updated by regular Windows updates. If you get Windows running after you change these 3 files it is best to manually run the Windows update process to see if MS will detect if these 3 files need further updating. The article refers to Win2000 but it should apply equally to XP. Folder winnt/system32 is changed to Windows/system32 for XP. You can change these filenames from the Recovery Console - at the command prompt (at least I think you can). You need to be able to run Recovery Console from a Win install CD or have installed Recovery Console into Windows. Be sure you can successfully run the Recovery Console before you begin. Windows (online) help file has more info on how to run it or install it. To make this easier you need to find the files with your operating copy of Windows first and put the files in place in the system32 folder with a changed name such as hal.new.dll, you'll be able to find them quicker. Be sure you know how to get the files extracted/expanded out of the .cab first, either with a built-in extractor such as that in the 7-zip app or by using the Windows provided extract/expand utility. I mention this because it is possible to end up with an un-expanded copy of the file. Check the extracted files with File\|Properties in Explorer - good files will show the version info. Just to be safe, copy your existing 3 files and keep them in case this process fails. An alternative to hunting down the files from the cab files is just to find another running version of Windows with the uniprocessor files and then just copy them (like on a floppy). ___________________________________ The following is extracted from the url above. 2. After you log on to your computer, from the %SystemRoot%System32 folder of your original Windows 2000 installation, use the command console to copy and rename the following files as noted in the following lists. First determine which HAL and kernel files your computer requires by using the following list of supported computer types: NOTE: Those marked with an asterisk character are Standard computer types (non-ACPI). (You need only one of the following!) i386 source file: i386driver.cabhalmacpi.dll Computer type: ACPI Multiprocessor PC Copy to this folder: winntSystem32 Rename to this file name: hal.dll i386 source file: i386driver.cabhalaacpi.dll Computer type: ACPI Uniprocessor PC Copy to this folder: winntSystem32 Rename to this file name: hal.dll i386 source file: i386driver.cabhalacpi.dll Computer type: Advanced Configuration and Power Interface (ACPI) PC Copy to this folder: winntSystem32 Rename to this file name: hal.dll i386 source file: i386driver.cabhalsp.dll Computer type: Compaq SystemPro Multiprocessor or 100% Compatible Copy to this folder: winntSystem32 Rename to this file name: hal.dll i386 source file: i386driver.cabhalapic.dll Computer type: MPS Uniprocessor PC Copy to this folder: winntSystem32 Rename to this file name: hal.dll i386 source file: i386driver.cabhalmps.dll Computer type: MPS Multiprocessor PC Copy to this folder: winntSystem32 Rename to this file name: hal.dll i386 source file: i386driver.cabhal.dll Computer type: Standard PC Copy to this folder: winntSystem32 Rename to this file name: hal.dll i386 source file: *i386driver.cabhalborg.dll Computer type: SGI mp Copy to this folder: winntSystem32 Rename to this file name: hal.dll 3. As indicated in the following table, copy all the appropriate kernel files common for your system (Uniprocessor vs. Multiprocessor) and the appropriate HAL file based on your computer supported computer type from the Windows 2000 CD-ROM (i386driver.cab) or installed service pack to the corresponding folder listed in the table under your original Windows 2000 installation, and then rename them (if required) to Ntoskrnl.exe and Hal.dll, respectively. Common Multiprocessor kernel files Copy to the following directory Rename to the following filename --------------------------------------------------------- i386driver.cabntkrnlmp.exe winntSystem32 ntoskrnl.exe i386driver.cabntkrpamp.exe winntSystem32 ntkrnlpa.exe (following 4 files are the same - don't need to copy ) (i386driver.cabkernel32.dll winntSystem32 kernel32.dll ) (i386driver.cabntdll.dll winntSystem32 ntdll.dll ) (i386driver.cabwin32k.sys winntSystem32drivers win32k.sys ) (i386driver.cabwinsrv.dll winntSystem32 winsrv.dll ) Common Uniprocessor kernel files Copy to the following directory Rename to the following filename --------------------------------------------------------- i386driver.cabntoskrnl.exe winntSystem32 ntoskrnl.exe i386driver.cabntkrnlpa.exe winntSystem32 ntkrnlpa.exe (following 4 files are the same - don't need to copy ) (i386driver.cabkernel32.dll winntSystem32 kernel32.dll ) (i386driver.cabntdll.dll winntSystem32 ntdll.dll ) (i386driver.cabwin32k.sys winntSystem32 win32k.sys ) (i386driver.cabwinsrv.dll ) ___________________________________________________________ So, for computer type: ACPI Uniprocessor PC i386driver.cabhalaacpi.dll -> windowsSystem32hal.dll i386driver.cabntoskrnl.exe -> windowsSystem32 ntoskrnl.exe i386driver.cabntkrnlpa.exe -> windowsSystem32 ntkrnlpa.exe The additional 4 files may or may not need to match, you may need to extract them also from the SP2 cab file. This is usually unnecessary. Good luck! I suppose you've thought about just wiping the disk and doing a full 'real' install. I don't know whether the Compaq files will permit this - I'm guessing not.
9) Message boards : Number crunching : Who do you like for PSU's ? (Message 25858) Posted 1 Sep 2006 by Bob Guy Post: Antec TruePower 2.0 or SmartPower 2.0, both have a dedicated CPU rail and a second rail for the rest. I have a SmartPower 500W and it runs fairly cool and very quiet. There is enough power to run 4 HDs, two DVDs and a power hungry NVidia video card while keeping very stable voltages on the CPU.
10) Message boards : Number crunching : Upgrade complete, sort of (Message 25552) Posted 30 Aug 2006 by Bob Guy Post: This report and hotfix from Microsoft may be relevant to bad multiprocessor performance. Ethan@: I'll remember you said that, and I'm glad it works for you, I just haven't had any experience with the new dual CPUs. It would seem reasonable that the license is 'per socket' but I long ago stopped trusting Microsoft to be reasonable. I just didn't know what the software actually did although I expected XP Home to work properly on the new dual CPUs.
11) Message boards : Number crunching : Upgrade complete, sort of (Message 25537) Posted 30 Aug 2006 by Bob Guy Post: Well, I'm thoroughly confused now! Microsoft says, in various places, that XP Home supports only one physical processor. I can't find anywhere that SP2 actually changes that support. However, the question remains whether an X2 or a Core Duo or a Core 2 Duo is considered as a single physical CPU or as two physical CPUs. It would seem logical that they are two physical CPUs even though they are in the same physical package. It remains, though, whether the XP installer detects the dual-processors as single or multiple CPUs for the purpose of determining processor support. Clearly it is a multiple CPU in terms of hardware but is it multiple in terms of the software license. All of this is irrelevant if SP2 changes what it considers a multiple CPU system. But I've not been able to confirm what SP2 does consider as single/multiple in terms of the software license. My processor is HT so all this is academic for me. I suppose we need someone to report in who has XP Home properly installed on a dual processor, and ask them if it seems to be working as advertised. I've never heard of this problem with XP Home so I'm assuming SP2 should run properly with the dual processor chips. It's possible that XP Home doesn't actually run properly in this case and nobody has noticed it yet. In case you haven't noticed, I've written all this with CYA as the objective and with the intent of a disclaimer. And now that I read it it's mostly useless info. Well, I'll post it anyway in the hope that someone else won't have to go through the same exercise.
12) Message boards : Number crunching : Upgrade complete, sort of (Message 25531) Posted 30 Aug 2006 by Bob Guy Post: For what its worth, the format, re-install took about 20 mins. Sound about the right amount of time? Yes, I did a recent re-install to a 100Mb NTFS partition on a blank harddisk from an original release XP CD and it took about that much time including the SP2 update. I think you can tell if the files and folders are really new ones by looking at the dates. New ones should have a 'date created' (in properties) that is the date you did the install even when the 'date modified' is something different (an earlier or later date).
13) Message boards : Number crunching : Upgrade complete, sort of (Message 25525) Posted 29 Aug 2006 by Bob Guy Post: Sounds to me like you have a defective CPU #1 although I don't know how you could actually determine that that was true.
14) Message boards : Number crunching : Upgrade complete, sort of (Message 25508) Posted 29 Aug 2006 by Bob Guy Post: I don't know whether (all of) the following is true, but I believe it is. First, assume XP is properly set up for multiprocessor/multithreading. If you see 'ACPI Multiprocessor PC' in Device Manager and can see two CPUs in Task Manager I'll assume it is properly set up. For any single instance of Rosetta (or Seti, Einstein or whatever) you CAN see CPU utilization greater than 50% because the application makes calls to the Windows kernel and I/O system which is (also) multithreaded and can use either or both processors. Those calls can and do use the other available CPU resources and will add to the total CPU utilization recorded for that single instance of that application (the Rosetta app). This is of course more likely to happen when there is only one instance of Rosetta (or whatever) running. You can (obviously) never have a total CPU utilization (in Task Manager) for all applications that exceeds 100%, or at least you shouldn't. So, when there are two Rosetta apps running you should see 45%..48% for each app allowing for a couple of percent used for background processes. At least that has been my experience.
15) Message boards : Number crunching : A Challenge (Message 24971) Posted 26 Aug 2006 by Bob Guy Post: (a) Why didn't the ones complaining work within the system and use the same optimized clients? (b) Why would academia waste resources by placating to those with higher pitched voices who have refused to help themselves? For the delivery of treatment, you require the competition of the marketplace. Hence, the pharmaceutical industry with its diligent group of hard-charging "cheaters" is required to make the cure a reality. In this real world people endeavor to be the first to introduce something substantial and new to the marketplace. Therfore, they dig for any advantage they can to be able to come out on top. This attitude ensures the least possible time to market. (a) and (b) Get out your dictionary and look up the word ethics. Big Pharma has only one goal: to make the largest profit. This is not entirely a bad thing for the general public and certainly not a bad thing in a capitalist economy. But, don't expect the pharmaceutical industry to be thinking about how they can lower the cost of their product through the application of technology. Still, that's the price you and I pay, and if it saves your life, I guess it was worth it.
16) Message boards : Number crunching : Discussion of the new credit system (Message 24967) Posted 26 Aug 2006 by Bob Guy Post: Another point with the big, competitive teams is that they compete with each other. If one team suddenly decides to move resources to another project, the others follow suit in order not to be left behind. Our team does not bother with such matters. We have seen an influx of members from other teams, the new members being fed up with serving the teams whims, and ignoring the scientific benefit of what the projects are doing. I'm so glad to hear from someone who has some common sense after hearing so much from the vocal minority. I think most people feel the same as you've said and don't constantly barrage the forum with repeated posts about how bad the new credit system is. I think you know now who the vocal minority is (this is a deliberate poke at certain teams). My take on the new credit system: I'm not entirely satisfied because it does not accurately represent my contribution to the project. I was (for a short time) using the 5.5.0 BOINC client and inflated my credits by 3 times over the 5.4.9 client. I gladly switched back to the 5.4.11 client before the new credit system came on line. Now I'm getting 40..50% more credit awarded than claimed. Don't get me wrong - I'll gladly accept the extra. But, what this means is that the average claim must be much larger (or at least somewhat larger) than my stock claim. This also means that there are still too many overclaiming clients being used which is artificially pushing up the average. Or, my poor P4 has been consistantly underclaiming and continues to do so. It's just a minor annoyance to me and really isn't a complaint, just an observation. My interpretation of the new credit system is that it intends to give to all users the benefit of any advantage that any user's overclaiming client gives. I think the project developers think this would tend to make it less productive and therefore less desirable for any user to implement an overclaiming client. However, it will still increase the averages to use an overclaiming client (as long as it's still not prohibited) so that if enough users (read: teams) use an overclaiming client those users will artificially inflate the credit system to the point that it no longer represents reality. I thought that there once was discussion of using a fixed credit based on a 'golden' standard WU (or a 'golden' standard machine), adjusted of course to each different type of WU. I personally think that's the only reality based credit system, unless you were to implement a FLOP counting method which I understand the programmers have discounted because of the difficulty in coding.
17) Message boards : Number crunching : Converting a Yugo into a Porsche? (Message 24518) Posted 24 Aug 2006 by Bob Guy Post: For the multiprocessor setting: If Windows does not set it on its own you can just enter the control panel, hardware, device manager and then go to "computer" and open the panel for it. You then click "update driver" on the "driver" tab and select HAL.dll to be the multiprocessor driver instead of the single processor driver. After that the computer needs to be restarted. That's all there is to it. I did it in Win2000 two weeks ago and it went without a glitch. This ability is disabled in WinXP (any version) but it is possible to change from single to multiprocessor by manually extracting two files from the appropriate .cab file and replacing them while outside of Windows (i.e. from DOS). I can tell you that this may blow up Windows but I've done it very recently with XP SP2 and it did work. Basically you need to replace the hardware layer and the kernel, both at the same time, with the proper matching set. Post here if you need more specific details, I don't want to post all the details for fear that someone might mess up their computer.
18) Message boards : Number crunching : Page Faults and Rosetta (Message 21871) Posted 5 Aug 2006 by Bob Guy Post: See this TechNet article regarding Memory Usage and System Cache. While it may not be entirely applicable to WinXP see the section 'Confirming Hard Page Faults' for some useful diagnostic techniques. The page faults reported by WinXP (i.e. TaskManager) include both hard and soft page faults so you shouldn't rely on that information alone to make a performance analysis. Soft page faults often occur as a result of the OS reserving memory (real or virtual) for possible future use - it does not indicate that the memory is actually being used or that an actual memory read/write occurred or that an actual disk IO occurred. The page file (virtual memory) may also contain these reserved blocks of memory that in many cases are never written/read - so the size of the paged memory is not a good indicator of its actual use.
19) Message boards : Number crunching : 5.4.9 vs Trux for Claimed Credit (Message 19745) Posted 3 Jul 2006 by Bob Guy Post: I've heard in this forum that the project management frowns on the use of calibrating or optimized clients in Rosetta. It would be nice if the project management would make a statement of policy on this issue. I use a plain vanilla 5.4.9 client and I too feel I'm being shorted on credit but I personally think optimized and calibrating clients in this project are improper. Having said this, I don't think it's possible for the project management to do anything with regard to requiring the use of a plain vanilla client. After all, anybody can make their own client that does whatever it is they want it to do. If and when the project implements an 'internal' credit award system I expect we'll see alot of complaints from the people now using optimized and calibrating clients, much like what has happened at SETI and Einstein. Even so, I think it will be fairer than it is now. I regularly see results claiming 4 to 6 times the credit that I claim (adjusted for processor efficiency and WU run time). I use the credits as a measure of my contribution to this project and when I see results with inflated credit claims it offends me. Is the computer that claims 4 times the credit contributing 4 times the work? I don't think so. Yes, I'm just another disgruntled user. Thanks for listening to my rant.
20) Message boards : Cafe Rosetta : "Is Distributed Computing being Distributed Badly?" (Message 19520) Posted 30 Jun 2006 by Bob Guy Post: I have read here in this thread, in the SlashDot article and in many other places that SETI is doomed to failure because it's not possible to detect "I Love Lucy" type broadcasts. This is probably true. However, SETI is not searching for TV broadcasts. It is very likely that SETI can detect high-energy high-power military radar which, for most technologically advanced civilizations, there is no good substitute. Therefore most scientists agree that those signals should be present, at least for some period of time, and should be detectable. Please stop quoting the "I Love Lucy" argument, it just proves you are either a fool or ignorant (and I mean uneducated) or both, or that you have a political agenda. I always know to disregard anything said by a person who uses this argument. Having said all this, I don't mean this as a personal attack on any one person or group, it's just my observation and opinion.

Next 20