-forcesse/-advemthods no longer necessary.

csimoncsimon Acadiana Icrontian
edited May 2004 in Folding@Home
Today after looking over a thread someone I made the observation that with core_78 v1.64 -forcesse was no longer necessary to get the sse boost for athlons. I tested the observation myself with core_78 v1.62 and guess what?
sse boost ok!

Arguments: -service

[14:41:50] - Ask before connecting: No
[14:41:50] - User name: csimon (Team 93)
[14:41:50] - User ID = 6C6F721974B9C5E6
[14:41:50] - Machine ID: 1
[14:41:50]
[14:41:50] Loaded queue successfully.
[14:41:50] + Benchmarking ...
[14:41:52]
[14:41:52] + Processing work unit
[14:41:52] Core required: FahCore_78.exe
[14:41:52] Core found.
[14:41:52] Working on Unit 00 [May 8 14:41:52]
[14:41:52] + Working ...
[14:41:53]
[14:41:53] *
*
[14:41:53] Folding@home Gromacs Core
[14:41:53] Version 1.62 (April 2, 2004)
[14:41:53]
[14:41:53] Preparing to commence simulation
[14:41:53] - Looking at optimizations...
[14:41:53] - Files status OK
[14:41:53] - Expanded 510733 -> 2524933 (decompressed 494.3 percent)
[14:41:53]
[14:41:53] Project: 910 (Run 65, Clone 90, Gen 27)
[14:41:53]
[14:41:53] Assembly optimizations on if available.
[14:41:53] Entering M.D.
[14:42:13] (Starting from checkpoint)
[14:42:13] Protein: p910_vill_str0_rf
[14:42:13]
[14:42:13] Writing local files
[14:42:16] Completed 240000 out of 250000 steps (96)
[14:42:16] Extra SSE boost OK.
[14:46:40] Writing local files
[14:46:40] Completed 242500 out of 250000 steps (97)
[14:51:00] Writing local files

this is on the sig system.

since the new point scheme -advmethods is also no longer necessary.
«1

Comments

  • KingFishKingFish
    edited May 2004
    did you mean to publish that second flag?
    I have vers 1.64. What is the difference between that and the vers you have here (1.62)?

    KingFish
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    KingFish wrote:
    did you mean to publish that second flag?
    I have vers 1.64. What is the difference between that and the vers you have here (1.62)?

    KingFish
    whew ...that'll be our little secret!!! :wink:

    Science-wise, 1.64 includes support for a new implicit solvent model (see projects 923-4). Besides this, it supports 735's results, which 1.62 didn't.
    It's now the standard download, not just beta. Guha
  • KingFishKingFish
    snickers to himself and ponders what to do with the newfound knowledge
    edited May 2004
    snickers to himself and ponders what to do with the newfound knowledge
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    KingFish wrote:
    * KingFish snickers to himself and ponders what to do with the newfound knowledge
    thanks for catching that and pointing it out KF ...I could get in trouble if I leaked it ...right marc? ;D j/k

    Anyway ...if you want to do beta all you need to do is drop a PM to one of the admins at the community and read and follow the rules/guidelines and you're in!!! Rule number one ...don't reveal teh knowlege! LOL

    Anyway ...folding beta wu's means reporting any issues you may have with the wu's at -verbosity 9 I think and then they mostly do the troubleshooting from there!

    Some risk involved but hey ...we live in louisiana so that's on par!
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited May 2004
    I use

    -forcesse -service -forceasm

    Is this ok?
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    Thrax wrote:
    I use

    -forcesse -service -forceasm

    Is this ok?
    sure it's ok ...are you using FAH4Console.exe?

    I may have been a little premature about -forcesse ...I just noticed that with v1.64 I started a new wu and it said neither sse boost or 3dnow so I'm thinking maybe -forcesse is still necessary to fold using sse with an athlon.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited May 2004
    Yeah, FAH4Console.

    Will it be faster by adding or removing any flags?
  • MrBillMrBill Missouri Member
    edited May 2004
    Thrax wrote:
    Yeah, FAH4Console.

    Will it be faster by adding or removing any flags?
    That depends on who you ask. You should not need forcesse and forceasm.

    I use -advmethods -forcesse
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited May 2004
    OK, this is like all the premature health "research" reports that the idiotic press loves for soundbites: oat bran - great, no - just good, ah nevermind; cholesteral - LDL, XYZ, ABC - good, bad, not so bad; coffee - good, bad, good, who knows....and so on.

    This is my new flag for Athlons: -turnonwhateverthehellworksbestdamnit
  • MrBillMrBill Missouri Member
    edited May 2004
    Leonardo wrote:
    This is my new flag for Athlons: -turnonwhateverthehellworksbestdamnit
    Doesn't work on mine. Said invalid argument or something like that. ;):D;D



    p.s. i didn't really try it
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited May 2004
    -turnonwhateverthehellworksbestdamnit
    -screwthe'invalidargument'crap
    -pretendthiscpuis6000mhz
    -deathtoseti
    -ignoremyoverthetopoverclocking
    -thisuserisalwaysright
    -givememorefriggingdoublegromacsNOW
  • DragstkDragstk Syracuse, N.Y.
    edited May 2004
    Leonardo wrote:
    -turnonwhateverthehellworksbestdamnit
    -screwthe'invalidargument'crap
    -pretendthiscpuis6000mhz
    -deathtoseti
    -ignoremyoverthetopoverclocking
    -thisuserisalwaysright
    -givememorefriggingdoublegromacsNOW

    I'm not sure I can fit that all in the little Target box :scratch:;D;D
  • EyesOnlyEyesOnly Sweden New
    edited May 2004
    Would it work on P2:s? ;D:eek2::D
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited May 2004
    -pretendthiscpuis6000mhz
  • MedlockMedlock Miramar, Florida Member
    edited May 2004
    Leonardo wrote:
    -turnonwhateverthehellworksbestdamnit
    -screwthe'invalidargument'crap
    -pretendthiscpuis6000mhz
    -deathtoseti
    -ignoremyoverthetopoverclocking
    -thisuserisalwaysright
    -givememorefriggingdoublegromacsNOW
    It'd be great if all that worked! :rant: Especially the Double Gromacs one :thumbsup:

    Which arguments help a P4's production? (Besides -advmethods and -forcesse) Anything I'm missing? I don't use -service 'cause the only user acct on this machine is my own, so there's no need right?
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    TheGr81 wrote:
    It'd be great if all that worked! :rant: Especially the Double Gromacs one :thumbsup:

    Which arguments help a P4's production? (Besides -advmethods and -forcesse) Anything I'm missing? I don't use -service 'cause the only user acct on this machine is my own, so there's no need right?
    don't use any at all.
  • MedlockMedlock Miramar, Florida Member
    edited May 2004
    :eek2: :confused2 :scratch:

    Ok, now I'm confused. A P4's point production is greater with gromacs than with tinkers right? That's why I use -advmethods, it tends to give out more gromacs and dgromacs WU's. And while I'm still not sure about -forcesse, some people have experienced greater production with the tag than without it. (But I haven't really given it enough attention to notice it myself.) I know I need -local, but that's different.

    And one more thing...

    Where the heck did you get all those folding computers!? :rant:
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    TheGr81 wrote:
    :eek2: :confused2 :scratch:

    Ok, now I'm confused. A P4's point production is greater with gromacs than with tinkers right? That's why I use -advmethods, it tends to give out more gromacs and dgromacs WU's. And while I'm still not sure about -forcesse, some people have experienced greater production with the tag than without it. (But I haven't really given it enough attention to notice it myself.) I know I need -local, but that's different.

    And one more thing...

    Where the heck did you get all those folding computers!? :rant:

    It's all relative on a P4 ...points are assigned from a P4 so whatever you get it's relative. So at this point that flag should give you no advantage or disadvantage.
  • KingFishKingFish
    edited May 2004
    Hey csimon
    I won't leak the info any further. I don't wish to be on the beta team at this point because I don't have the time necessary to report my findings and such. I have 1.64 gromacs core and it is turning my optimizations on as necessary with my athlons cpu's. The only flag I have with my athlons is -verbosity 9 and they are all chugging along with sse (as verified in the log files) just fine. I have no idea why others wouldn't have the optimizations working without the -forcesse flag.

    KingFish
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    KingFish wrote:
    Hey csimon
    I won't leak the info any further. I don't wish to be on the beta team at this point because I don't have the time necessary to report my findings and such. I have 1.64 gromacs core and it is turning my optimizations on as necessary with my athlons cpu's. The only flag I have with my athlons is -verbosity 9 and they are all chugging along with sse (as verified in the log files) just fine. I have no idea why others wouldn't have the optimizations working without the -forcesse flag.

    KingFish

    Ok ...The next time you get gromacs make sure that it uses SSE automatically. :thumbsup:
  • KingFishKingFish
    edited May 2004
    I've already watched a few cycles and it is using them automatically. All appears well.

    KingFish
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    KingFish wrote:
    I've already watched a few cycles and it is using them automatically. All appears well.

    KingFish
    I'll give this one another shot and see what happens also ..thanks!
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited May 2004
    Which arguments help a P4's production?
    It may not make any difference, but I would defintely add -advmethods for a P4. You really want to snag Gromacs as often as possible.
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    KingFish wrote:
    I've already watched a few cycles and it is using them automatically. All appears well.

    KingFish
    Even if you stop the client and restart you still get SSE boost ok!?
    I'm just not seeing it ...maybe I'm nervous for nothing idunno.
  • fatcatfatcat Mizzou Icrontian
    edited May 2004
    do ya get the new cores when u get a WU that needs them?...I'm still using 1.56.
  • csimoncsimon Acadiana Icrontian
    edited May 2004
    fatcat wrote:
    do ya get the new cores when u get a WU that needs them?...I'm still using 1.56.
    yes
  • mmonninmmonnin Centreville, VA
    edited May 2004
    You can delete the core if you want to try this. Best idea is to wait til a WU is finished as always.

    This is interesting news. I hadnt heard of this yet.

    Hmm yes that secret better stay a secret.
  • MedlockMedlock Miramar, Florida Member
    edited May 2004
    Aw, you guys suck! I dunno what you're talkin' about! :rant:

    lol
  • mmonninmmonnin Centreville, VA
    edited May 2004
    The beta team is open to anyone who wants to help test the clients and work units.
  • KingFishKingFish
    edited May 2004
    csimon wrote:
    Even if you stop the client and restart you still get SSE boost ok!?
    I'm just not seeing it ...maybe I'm nervous for nothing idunno.

    Yessir, when I stop for a reboot all my nodes using athlon xp's are automatically detecting sse optimizations and utilizing them all on their own without any -forcesse flags. Hmmm, I wonder what this other flag does? hehe

    KingFish
Sign In or Register to comment.