Announcement

Collapse
No announcement yet.

How to make Genome@home go faster!

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • How to make Genome@home go faster!

    Getting your Go Faster Stripes on Genome@Home

    Ofcourse, a fast CPU and lots of memory help a lot, but I found these little things help quite a lot as well

    After downloading a GU, let the FILTER program run for a minimum of one sequence. First though click on the properties of the FILTER program, go to the Misc tab and set the Idle time for the CPU to LOW. Uncheck Allow screensaver. Click Apply.

    When FILTER has run at least one sequence, quit it when it's at the lower % (before 50% !!), then start up GHCLIENT. In the now opened up CLI window also choose properties, go to the Misc tab, set the slider on the Idle time for CPU to LOW and disable Allow Screensaver. Click Apply.

    If you have Genome Voyeur, start it up now and hide the Genome CLI.

    My times have started to plummet doing this. Whereas a sequence would take 60 minutes or more on my P3-450, it now takes no more than 30 minutes per sequence.

    Add your own solutions for tuning, and I'll add this whole thread to the FAQs at some point.

    Jord.
    Jordâ„¢

  • #2
    Thanks Jorden, interesting post!
    Fred

    Comment


    • #3
      Every time I restart the client, it is back to the original settings.
      Any ideas on what I'm missing?

      Thanks,
      Mark F.

      ------------------
      OH NO, my retractable cup holder swallowed a DVD...
      and burped out a movie
      Mark F. (A+, Network+, & CCNA)
      --------------------------------------------------
      OH NO, my retractable cup holder swallowed a DVD...
      and burped out a movie

      Comment


      • #4
        You're not missing much, Mark.
        I have to do that as well, but it's quickly done. Just have to put the slider to low again

        Btw, by putting the slider to LOW, you give more CPU time to the program, in case you were wondering

        Jord.
        Jordâ„¢

        Comment


        • #5
          That is a bad idea if you're going to do something else on the computer at the same time.
          Someday, we'll look back on this, laugh nervously and change the subject.

          Comment


          • #6
            I assume this is a Win9x thing...?

            Assuming you are just increasing the priority, you can do it in Win2K with the Task Manager

            Comment


            • #7
              Jorden, that OS are you running on???

              Genome@home on NT contains two exe-files, ghclient.exe and filter.exe

              ghclient.exe is the main program used to download/upload and do 99%+ of the work.
              filter.exe is called by ghclient.exe once for a downloaded wu, doing some preliminary stuff. If you re-loop the same wu, filter.exe isn't called again.

              So please tell me why you're letting filter run one sequence, and afterwards stopping it before 50%. Filter doesn't show any percents, it just counts up to the aa-length and quits.


              To change the priority should have negligible effect if you're not running something like seti@home as backup for the re-loop pause or if genome crashes. Hiding should probably cut some minutes, but running as service should probably cut a little bit more.

              I've heard locking the affinity leads to faster times.

              As for the large drop in times, that is the aa-length before/after?

              Comment


              • #8
                I just remember the slider-thing, it's used on NT also in the shortcut for 16-bit-dos-programs. Windows-programs don't use it, not even 16-bits.

                Comment


                • #9
                  I use WinME. Filter runs through after filtering the unit it just got. It doesn't switch over to GHClient, and if I quit Filter before at least one sequence has run, start up GHClient to do the work, it tries to reconnect to the net and get a new Unit.

                  I tested the slider for CPU Idle Time, because under WinME the ghclient and filter run in a DOS box. It works on my machine, especially when I go to bed and just let the machine rip through the night

                  Usually when I get up in the morning, it's already rerunning the last unit again.

                  Jord.
                  Jordâ„¢

                  Comment


                  • #10
                    Hmm, I have to look at those settings, Jorden you've caught up to me in no time ...

                    Hopefully I'll get my P3-500 running today and then ... (me laying the smack down)




                    [This message has been edited by CHHAS (edited 15 March 2001).]
                    "That's right fool! Now I'm a flying talking donkey!"

                    P4 2.66, 512 mb PC2700, ATI Radeon 9000, Seagate Barracude IV 80 gb, Acer Al 732 17" TFT

                    Comment


                    • #11
                      Better make it quick, CHHAS... already crunched half of the new unit, in less than 4,5 hours

                      Edit:
                      Strangely enough I'm running two ghclient's at the same moment, both of them turning in sequnces every half hour. I didn't see this, as there was only one ghclient window open in my taskbar, and I didn't have Genome Voyeur running. I think they were super-imposed on eachother.

                      But upon starting GV, and hiding the client, I still had a client in my taskbar. I timed both of them for the past hour, and what do you think? Running both on 30 minute per sequence. Cool

                      Jord.

                      [This message has been edited by Jorden (edited 15 March 2001).]
                      Jordâ„¢

                      Comment


                      • #12
                        Extremely cool You must have the Katmai from Hell

                        I've got the Celeron running and I'm installing Windows now, unfortunately the BIOS upgrade was a dead link I've mailed Lucky Star for a BIOS and hope they'll reply (the board is a 6ABX2V if any of you should happen to stumble upon a BIOS for it I'd be very happy).
                        "That's right fool! Now I'm a flying talking donkey!"

                        P4 2.66, 512 mb PC2700, ATI Radeon 9000, Seagate Barracude IV 80 gb, Acer Al 732 17" TFT

                        Comment


                        • #13
                          Well, I'd better warn against running two clients from the same directory, as I was doing earlier. It cost me dearly. For after a reboot, both units that were being crunched were found to be damaged, and no matter what I tried (even deleting everything in the folder), when I started up the ghclient, it wouldn't connect to home and get me a new unit.

                          I'm crunching again, but from a second folder. Every time I try to start up the client in the first genome folder, it still can't get a unit. I think it's time to dump the whole folder in the trashcan.

                          Jord.
                          Jordâ„¢

                          Comment


                          • #14
                            Does the window looks something like this then downloaded a new wu?

                            Creating filtered rotamer library
                            aa positions to filter
                            1
                            2
                            3
                            ...
                            aa
                            Initializing protein design algorithm
                            Designing protein sequence 1 of 30
                            10.0% ....

                            If so, filter.exe is done with it's job then "Initializing protein design algorithm" shows up, and you're running ghclient.exe as you should.
                            The client thinks the wu is bad if it's stopped before done with the first 1/30.

                            If you're running two instances in the same directory, you can get funny results. Every instance of genome@home must have its own directory and a unique id.dat

                            Comment


                            • #15
                              If you use Jorden's advice, along with my batch file and merge the two.

                              ie - create the batch file, and set the properties for the batch file according to Jorden's info, and voila... each process that runs from the batch file inherits the process level from the batch file.

                              No resetting every time you run the process, you don't have to mess with setting it for the filter and ghclient separately...

                              ghclient.bat
                              Code:
                              @echo off
                              call ghclient.exe
                              ghclient.bat
                              Very nice...

                              Thanks Jorden...

                              Guyver

                              [This message has been edited by Guyver (edited 15 March 2001).]
                              Gaming Rig.

                              - Gigabyte GA-7N400-Pro
                              - AMD Athlon 3200+ XP
                              - 1.5GB Dual Channel DDR 433Mhz SDRAM
                              - 6.1 Digital Audio
                              - Gigabit Lan (Linksys 1032)
                              - 4 x 120GB SATA Drives, RAID 0+1 (Striped/Mirrored)
                              - Sony DRU-500A DVD/+/-/R/RW
                              - Creative 8x DVD-ROM
                              - LS120 IDE Floppy
                              - Zip 100 IDE
                              - PNY Ultra 5900 (256MB)
                              - NEC FE950
                              - DTT2500 Cambridge Soundworks

                              Comment

                              Working...
                              X