Seti@Home optimized science apps and information

Optimized Seti@Home apps => Windows => Topic started by: hiamps on 27 Nov 2006, 08:37:07 pm

Title: Weird results with 2.0
Post by: hiamps on 27 Nov 2006, 08:37:07 pm

My credit claimed was way out there...
<core_client_version>5.4.11</core_client_version>
<stderr_txt>
ChirpData--[ak's_sse3_chirp]:   9857556 (chosen)
 GetPeak--[hand_opt]:      3229 (chosen)
   f_sum--[original]:     32352 (chosen)
GetChiSq--[hoisted+abs(]:     36890 (chosen)
IPP FFT SSE2(64K)[original]:   4214609 (chosen)
Bench Time: 1.95 seconds
work_len=1048576
Optimized Windows SETI@Home Enhanced application

Optimizers: Ben Herndon, Josef Segur, Alex Kan, Simon Zadra
  Version: Windows SSE2 32-bit based on seti V5.15  'Chicken Good!'
      Rev: (R-2.0|QxB|FFT:IPP_SSE2|Ben-Joe|)
    CPUID: 'AMD K8 Opteron DC 1 (Denmark)'
     cpus: 1 cores: 2 threads: 1   cache: L1=64K  L2=1024K L3=0K
 features: mmx 3Dnow 3Dnow+ sse sse2 sse3 
    speed: 2562 MHz  -- read megs/sec: L1=14050, L2=7023, RAM=3346

Work Unit Info
True angle range:  0.426464
ChirpData--[ak's_sse3_chirp]:   9872393 (chosen)
 GetPeak--[hand_opt]:      2462 (chosen)
   f_sum--[original]:     32503 (chosen)
GetChiSq--[hoisted+abs(]:     35852 (chosen)
IPP FFT SSE2(64K)[original]:   4207780 (chosen)
Bench Time: 2.00 seconds
work_len=1048576

Spikes Pulses Triplets Gaussians Flops
   1      2       5        4     13426431975160353000
</stderr_txt>

 
Validate state Valid
Claimed credit 52058503.6073926
Granted credit 62.3993140302431
application version 5.17

Title: Re: Weird results with 2.0
Post by: Pepo on 29 Nov 2006, 07:06:40 pm

Quote from: hiamps on 27 Nov 2006, 08:37:07 pm

My credit claimed was way out there...
[...]
Validate state Valid
Claimed credit 52058503.6073926

If we could have the -10 Million club (http://setiweb.ssl.berkeley.edu/beta/forum_thread.php?id=229), back when the app was not that optimized, why not having a +52 Million club now?

Peter

Title: Re: Weird results with 2.0
Post by: Simon on 29 Nov 2006, 08:29:21 pm

Something similar happened to msattler's Athlon FX or Opteron rig, as well.

That one was overclocked pretty much to the limit and sometimes claimed wildly outrageous numbers. When he dialed it back a bit, those went away, if I remember correctly.

All in all, I wouldn't freak out about it as it's very likely those claims will be ignored, or rather, you'll get the average quorum (which doesn't include your overclaim in this case).

I had something like that happen recently as well when I was testing a pre-2.0 beta app because I had defined a variable as a double instead of an int, but that's what happened for those results. I did post about them in the "computation errors" thread at the s@h forums to make sure though.

HTH,
Simon.

Title: Re: Weird results with 2.0
Post by: hiamps on 30 Nov 2006, 03:29:53 pm

Thanks Simon, I am overclocked a bit......

Title: Re: Weird results with 2.0
Post by: Vyper on 01 Dec 2006, 10:58:31 am

This is offtopic but has to do with Rev2.0 application as well.

I tried to modify your source to force a core2 specific optimized rev2.0 app but it kinda fails..

The compilation goes well and in the progress window i have looked up the USE_SSSE3 and it executes as it should ... but..

when i run the application the output in the stderr.txt has printed that it uses MMX mode.

Rev: (R-2.0|QxW|FFT:IPP_MMX|Ben-Joe)

Instead of for instance SSE2, SSSE3 and so on.. I've managed to compile a version that is 1 sec faster than Simons Rev 1.3 in the testprogram for fast mid and slow test, for Core2 duo but when i try on live WU's its about 600 sec slower.. on a regular WU but validates perfectly..

And it says that it uses IPP_MMX..

Check out:

core_client_version>5.3.12.tx36</core_client_version>
<stderr_txt>
Optimized SETI@Home Enhanced application

Optimizers: Ben Herndon, Josef Segur, Alex Kan, Simon Zadra
  Version: Windows SSSE3 32-bit 'Ni!' based on seti V5.15  'Chicken Good!'
      Rev: (R-2.0|MMX|FFT:IPP_MMX|Ben-Joe)
    CPUID: 'Intel C2D Core 2 Duo (Conroe)'
     cpus: 1 cores: 2 threads: 1   cache: L1=32K  L2=4096K L3=0K
 features: mmx sse sse2 sse3 ssse3 
    speed: 3294 MHz  -- read megs/sec: L1=12297, L2=10649, RAM=6562

Work Unit Info
True angle range:  0.670041

Spikes Pulses Triplets Gaussians Flops
   0      0       0        0     8917670335985
</stderr_txt>

Does anyone know why the code jumps to a MMX optimized code when a Core2 is using the app??

Kind Regards Vyper

Title: Re: Weird results with 2.0
Post by: BenHer on 01 Dec 2006, 03:36:32 pm

Vyper,

Some parts of the "version" line are constant, and some others are variable-depending on DEFINE'd values.

All of them are created inside the 'Optimizer/chicken.h' header file.

Title: Re: Weird results with 2.0
Post by: Vyper on 01 Dec 2006, 04:35:07 pm

Ok thx.. Will look into it..