Is there a difference between the work done by Lin and Win WUS?


Advanced search

Message boards : Application Info : Is there a difference between the work done by Lin and Win WUS?

Sort
Author Message
Profile Saenger
Volunteer tester
Avatar

Joined: Sep 13 06
Posts: 125
ID: 79
Credit: 411,959
RAC: 0
Message 587 - Posted 23 Sep 2006 19:41:06 UTC

In the number crunching forum I stumbled over this question by looking for something else:

I took a look at 2 puter with the same CPU, but different OS:
#217 (linux)
#254 (windows)

The crunching times for their results differ considerably:
Linux : about 4800 seconds
Windows : about 10500 sec

Both imho quite consistanly. So my questions (asked in the other thread as well, but I think here's the more appropriate part of this forum):

Is the application for linux and windows the same or do the windows users more of whatever?
Is the app perhaps more suited for Lin than Win?
How does this reflect on the scientific output (however it will be measured)?
Is there a possibility to measure the scientific output at the servers?

Profile Andre Kerstens
Forum moderator
Project tester
Volunteer tester
Avatar

Joined: Sep 11 06
Posts: 749
ID: 1
Credit: 15,199
RAC: 0
Message 592 - Posted 24 Sep 2006 4:17:59 UTC - in response to Message ID 587 .

I've put this in your other thread, but will put it here again so that people can more easily find it:

1) The source code for windows, linux and mac is exactly the same for all platforms. But of course source code has to be compiled in a binary so that it can be run on those platforms. And that's where the difference might be as optimizations, etc will be different for windows, linux and mac. I'm not a compiler expert myself, but that's where the difference might come from.
2) This app (charmm) was originally written for a Cray supercomputer :-) It has been in development since the early 70's and is written in Fortran77. The windows port was only recently made. Therefore it is probably better suited to Lin/Mac than Win. And what we see is that it runs a whole lot faster on unix than it does on windows.
3) There will actually be differences in the output when run on different platforms. That's why we use homogeneous redundancy in boinc which makes sure that workunit replicas all are send to the same platform (or win, or lin, or mac). That way we do not compromise our scientific results and make sure people actually get credit for their results.
4) I'm not sure what you mean with this. Can you rephrase this question? I suspect your question might have been answered by 3 above...

The whole benchmark problem we are experiencing now is actually because we are using Boinc's standard integer and floating point benchmarks. These benchmarks of course are not Charmm specific, but very general purpose benchmarks. So whatever the boinc client measures is probably way off of what the app actually is using. That's why we will develop our own benchmarks in the future (Boinc has an API to tie your own benchmark in the boinc client) so that we are not dependent on the standard benchmarks anymore. This also means of course that optimized clients will get exactly the same flops and ints numbers as standard clients, which is what we all want in the end: fairness and credit based on realistic numbers.

Hope that helps!
Andre

In the number crunching forum I stumbled over this question by looking for something else:

I took a look at 2 puter with the same CPU, but different OS:
#217 (linux)
#254 (windows)

The crunching times for their results differ considerably:
Linux : about 4800 seconds
Windows : about 10500 sec

Both imho quite consistanly. So my questions (asked in the other thread as well, but I think here's the more appropriate part of this forum):

Is the application for linux and windows the same or do the windows users more of whatever?
Is the app perhaps more suited for Lin than Win?
How does this reflect on the scientific output (however it will be measured)?
Is there a possibility to measure the scientific output at the servers?


____________
D@H the greatest project in the world... a while from now!
Profile Saenger
Volunteer tester
Avatar

Joined: Sep 13 06
Posts: 125
ID: 79
Credit: 411,959
RAC: 0
Message 598 - Posted 24 Sep 2006 11:26:38 UTC

PLease look in the "original" thread for further discussiom.

And thans for the answers:)
____________
Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

Honza
Volunteer tester

Joined: Sep 13 06
Posts: 25
ID: 72
Credit: 5,064
RAC: 0
Message 620 - Posted 26 Sep 2006 17:36:42 UTC - in response to Message ID 587 .
Last modified: 26 Sep 2006 17:42:03 UTC

Is there a possibility to measure the scientific output at the servers?
I think this question suggest possibility of credit based on server-side (a-la CPDN, Einstein) - fixed credit per work done and abandom ill-numbered benchmarks.
Credit award was never fair among project (weak correlation) and differences withing project when difference platforms gives different numbers doesn't help the situation. This can be solved - on some projects - using server-side credit award, which also eliminate those with unrealistic benchmarks.

EDIT: ha, found it on the other thread ...at least my assumption was correct.

Mentioning early 70's and Fortran77...
I'm not sure how complex the code is but probably way more simpler than climate models...which were developed in Fortran over the decades as well.
A cooperation with CPDN on compilation of Fortran stuff and cross-platform consistency may be usefull...CPDN is quite mature among BOINC projects and porting from 64-bit Cray over to Linux and 32-bit Windows was done even prior BOINC era.

Message boards : Application Info : Is there a difference between the work done by Lin and Win WUS?

Database Error
: The MySQL server is running with the --read-only option so it cannot execute this statement
array(3) {
  [0]=>
  array(7) {
    ["file"]=>
    string(47) "/boinc/projects/docking/html_v2/inc/db_conn.inc"
    ["line"]=>
    int(97)
    ["function"]=>
    string(8) "do_query"
    ["class"]=>
    string(6) "DbConn"
    ["object"]=>
    object(DbConn)#9 (2) {
      ["db_conn"]=>
      resource(66) of type (mysql link persistent)
      ["db_name"]=>
      string(7) "docking"
    }
    ["type"]=>
    string(2) "->"
    ["args"]=>
    array(1) {
      [0]=>
      &string(50) "update DBNAME.thread set views=views+1 where id=54"
    }
  }
  [1]=>
  array(7) {
    ["file"]=>
    string(48) "/boinc/projects/docking/html_v2/inc/forum_db.inc"
    ["line"]=>
    int(60)
    ["function"]=>
    string(6) "update"
    ["class"]=>
    string(6) "DbConn"
    ["object"]=>
    object(DbConn)#9 (2) {
      ["db_conn"]=>
      resource(66) of type (mysql link persistent)
      ["db_name"]=>
      string(7) "docking"
    }
    ["type"]=>
    string(2) "->"
    ["args"]=>
    array(3) {
      [0]=>
      object(BoincThread)#3 (16) {
        ["id"]=>
        string(2) "54"
        ["forum"]=>
        string(2) "11"
        ["owner"]=>
        string(2) "79"
        ["status"]=>
        string(1) "0"
        ["title"]=>
        string(63) "Is there a difference between the work done by Lin and Win WUS?"
        ["timestamp"]=>
        string(10) "1159292202"
        ["views"]=>
        string(4) "1241"
        ["replies"]=>
        string(1) "3"
        ["activity"]=>
        string(20) "1.4953310766294e-130"
        ["sufferers"]=>
        string(1) "0"
        ["score"]=>
        string(1) "0"
        ["votes"]=>
        string(1) "0"
        ["create_time"]=>
        string(10) "1159040466"
        ["hidden"]=>
        string(1) "0"
        ["sticky"]=>
        string(1) "0"
        ["locked"]=>
        string(1) "0"
      }
      [1]=>
      &string(6) "thread"
      [2]=>
      &string(13) "views=views+1"
    }
  }
  [2]=>
  array(7) {
    ["file"]=>
    string(63) "/boinc/projects/docking/html_v2/user/community/forum/thread.php"
    ["line"]=>
    int(184)
    ["function"]=>
    string(6) "update"
    ["class"]=>
    string(11) "BoincThread"
    ["object"]=>
    object(BoincThread)#3 (16) {
      ["id"]=>
      string(2) "54"
      ["forum"]=>
      string(2) "11"
      ["owner"]=>
      string(2) "79"
      ["status"]=>
      string(1) "0"
      ["title"]=>
      string(63) "Is there a difference between the work done by Lin and Win WUS?"
      ["timestamp"]=>
      string(10) "1159292202"
      ["views"]=>
      string(4) "1241"
      ["replies"]=>
      string(1) "3"
      ["activity"]=>
      string(20) "1.4953310766294e-130"
      ["sufferers"]=>
      string(1) "0"
      ["score"]=>
      string(1) "0"
      ["votes"]=>
      string(1) "0"
      ["create_time"]=>
      string(10) "1159040466"
      ["hidden"]=>
      string(1) "0"
      ["sticky"]=>
      string(1) "0"
      ["locked"]=>
      string(1) "0"
    }
    ["type"]=>
    string(2) "->"
    ["args"]=>
    array(1) {
      [0]=>
      &string(13) "views=views+1"
    }
  }
}
query: update docking.thread set views=views+1 where id=54