Optomized S5 SSE3

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 561
Credit: 4527270
RAC: 0

RE: out of stderr.txt:

Message 39078 in response to message 39076

Quote:

out of stderr.txt: E@H S5R1 4.02 SSE-0307 TEST

is that ok, or did you upload the wrong file? ;)

It's not OK!

I can read this in my stderr.txt: E@H S5R1 4.02 SSE3-0709 TEST

Perhaps, the file was wrong, eh...
Try to download again: S5T0709

LiborA
LiborA
Joined: 8 Dec 05
Posts: 74
Credit: 337135
RAC: 0

RE: RE: S5T0308 SSE /

Message 39079 in response to message 39066

Quote:
Quote:

S5T0308 SSE / 2431,00 sec / 40,5min / result
S5T0308 SSE / 2466,48 sec / 41,1min / result

S5T0308 is very fast, but the results seemed to be invalid :-/

Thanks!

So, I'm sure the validator works with zero tolerance... :-(
Bad news... we cannot reach much bigger speed-up.

Version S5T0308 produce output file with diferent size as original app (and other tested optimalization)!!! So I'm not sure if the problem is in validator or in this version of app.

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 561
Credit: 4527270
RAC: 0

Hello

Message 39080 in response to message 39077

Hello S56RKO,

Quote:

S5S0003 on long WUs on Pentium M 755 (2GHz) (CPU time approx 10 hours):

validates OK
speedup approx 1%

Going for S5T0307 (no SSE3 on this CPU :( )

Try to use S5S0007, it will give you about 15% speed-up.

Metod, S56RKO
Metod, S56RKO
Joined: 11 Feb 05
Posts: 135
Credit: 826243590
RAC: 84775

RE: Try to use S5S0007, it

Message 39081 in response to message 39080

Quote:
Try to use S5S0007, it will give you about 15% speed-up.


Right. I first checked thi thread and noticed you mentioned next stable release. Only later I checked the other thread and noticed the latest link.

I'm running the S5S0007 now ... another two long WUs ahead (the first one half done uing S5S0003 and slightly touched by S5T0307). I'll be back ... tomorrow ;)

Metod ...

Crunchers For More Power
Crunchers For M...
Joined: 3 Aug 05
Posts: 69
Credit: 1071273
RAC: 0

@akosf: isn't S5T0307 =

Message 39082 in response to message 39080

@akosf: isn't S5T0307 = S5S0007 ?

Now the stderr.txt (SSE3-0709) is correct :) THX !!!!
The speed is very good, times in 30min...

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 561
Credit: 4527270
RAC: 0

RE: @akosf: isn't S5T0307 =

Message 39083 in response to message 39082

Quote:
@akosf: isn't S5T0307 = S5S0007 ?

The output text is different.

S -> Stable
T -> Test

Pepperammi
Pepperammi
Joined: 20 Feb 05
Posts: 131
Credit: 437943
RAC: 0

Odd question. On a dual core

Message 39084 in response to message 39082

Odd question. On a dual core or hyper threaded machine

Is it possible to enter a patched version into one of the 'slots' in the boinc file directory and a different patched version into another 'slot' and run them. If that is possible i can test two versions at once and only one maybe unstable- avoid ruining two units.

I supposse then (if it is possible) you might ae able to copy a wu into two slot2 and run standard app and patch app at same time for a comparison?
would have to run cache out before even trying this but is it possible?

Crunchers For More Power
Crunchers For M...
Joined: 3 Aug 05
Posts: 69
Credit: 1071273
RAC: 0

i have the same idea, but i

i have the same idea, but i did't test it.
Try it and tell us :)

Pepperammi
Pepperammi
Joined: 20 Feb 05
Posts: 131
Credit: 437943
RAC: 0

RE: i have the same idea,

Message 39086 in response to message 39085

Quote:
i have the same idea, but i did't test it.
Try it and tell us :)


:-)
It'll be tomorrow before i run my cache out and i'll give it a go but i'll make a backupof everything now and give the two patched versions idea a try in a minute.
Anyone knows better please let me know before i waste my time and/or mess things up.

Metod, S56RKO
Metod, S56RKO
Joined: 11 Feb 05
Posts: 135
Credit: 826243590
RAC: 84775

RE: Odd question. On a dual

Message 39087 in response to message 39084

Quote:

Odd question. On a dual core or hyper threaded machine

Is it possible to enter a patched version into one of the 'slots' in the boinc file directory and a different patched version into another 'slot' and run them. If that is possible i can test two versions at once and only one maybe unstable- avoid ruining two units.


If you're willing to try: in slot directory (slots/0 or something) you have file with name einstein_S5R1_4.02_windows_intelx86.exe. This implies executable but it's rather an XML file saying it's a soft link pointing to real executable.

I wonder what happens if you replace this XML file with a real executable? Is BOINC CC smar enough to notice it and run it directly or will it choke?

[edit]
Another idea: what if you simply edit the XML file to point to different executable?

Metod ...

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.