Question About New Stock Apps (v8, that is)

Message boards : Number crunching : Question About New Stock Apps (v8, that is)
Message board moderation

To post messages, you must log in.

AuthorMessage
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1760530 - Posted: 30 Jan 2016, 1:40:03 UTC

I am a bit confused over just what the stock apps are, especially when it is true (I think) that, at least for now, they are the same as the "new" Lunatics apps. How does that work? Are the apps smart enough to figure out whether I am SSE4.2 or AVX or whatever?
IIRC, the Lunatics apps have several versions depending on CPU specs.
So what actually will get installed by stock?
And if not the "best" (AVX I assume), what are the relative performance of the apps for an i7-4790k, say?

Thanks for your help!
ID: 1760530 · Report as offensive
bluestar

Send message
Joined: 5 Sep 12
Posts: 7031
Credit: 2,084,789
RAC: 3
Message 1760543 - Posted: 30 Jan 2016, 2:42:10 UTC
Last modified: 30 Jan 2016, 2:45:53 UTC

Perhaps by adding autocorrelation to the search, making things possibly more complicated, but also giving a chance at being able to detect an intelligent signal coming from space.

If you happen to be familiar with the .avi files for the WOW signal, it was not detected by the numbers that were being visible.

I think that the GPU application for v8 should be much anticipated, at least when having a capable graphics card.
ID: 1760543 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13736
Credit: 208,696,464
RAC: 304
Australia
Message 1760545 - Posted: 30 Jan 2016, 2:43:01 UTC - in response to Message 1760530.  

So what actually will get installed by stock?

The application that is the most compatible with older hardware, so basically MMX, 3DNow!, SSE etc support.

And if not the "best" (AVX I assume), what are the relative performance of the apps for an i7-4790k, say?

If your CPU & OS can make use of AVX then it's worth using.

When there was no v8 GPU application I ran the stock CPU application and run times for longer running WUs were around 4-4.5hrs. Now I'm running the GPU application (which uses a good chunk of CPU time) the AVX CPU application run time for longer running WUs is around 3.5-3.45hrs (with the occasional 4.5hr extra long running WU).
A big improvement.
Grant
Darwin NT
ID: 1760545 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34258
Credit: 79,922,639
RAC: 80
Germany
Message 1760648 - Posted: 30 Jan 2016, 8:55:33 UTC

Lunatics apps for CPU are not the same as stock.
Only GPU apps.


With each crime and every kindness we birth our future.
ID: 1760648 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1760703 - Posted: 30 Jan 2016, 12:17:56 UTC

1) Stock CPU app has embedded benchmark that selects best available processing codelets for particular host. Among those codelets SSE, SSE2, SSE3 and AVX ones.
2) Lunatics CPU build pre-tuned for particular available SSE level. So SSE3 for example has no AVX codelets.
3) For tested hardware Lunatics build of corresponding SIMD level was faster than stock binary with one exclusion - SSE.
ID: 1760703 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1760715 - Posted: 30 Jan 2016, 14:18:35 UTC - in response to Message 1760703.  

Raistmer - thanks for the info

3) For tested hardware Lunatics build of corresponding SIMD level was faster than stock binary with one exclusion - SSE.


Any rough idea how much faster for AVX CPUs like i7-4790k, i7-4820?
(i.e., 10% 25%)?

T
ID: 1760715 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1760890 - Posted: 30 Jan 2016, 22:26:38 UTC - in response to Message 1760715.  

Win10 x64 IP, Q9450 idle:

WU : PG0009_v7.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 687.071 secs
CPU 683.690 secs
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 460.338 secs, speedup: 33.00% ratio: 1.49x
CPU 457.906 secs, speedup: 33.02% ratio: 1.49x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 518.305 secs, speedup: 24.56% ratio: 1.33x
CPU 515.813 secs, speedup: 24.55% ratio: 1.33x

WU : PG0395_v7.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 808.595 secs
CPU 805.168 secs
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 447.799 secs, speedup: 44.62% ratio: 1.81x
CPU 445.250 secs, speedup: 44.70% ratio: 1.81x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 537.061 secs, speedup: 33.58% ratio: 1.51x
CPU 534.438 secs, speedup: 33.62% ratio: 1.51x

WU : PG0444_v7.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 738.052 secs
CPU 734.640 secs
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 419.239 secs, speedup: 43.20% ratio: 1.76x
CPU 416.656 secs, speedup: 43.28% ratio: 1.76x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 499.487 secs, speedup: 32.32% ratio: 1.48x
CPU 497.000 secs, speedup: 32.35% ratio: 1.48x

WU : PG1327_v7.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 499.871 secs
CPU 496.567 secs
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 375.160 secs, speedup: 24.95% ratio: 1.33x
CPU 372.578 secs, speedup: 24.97% ratio: 1.33x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 432.332 secs, speedup: 13.51% ratio: 1.16x
CPU 429.875 secs, speedup: 13.43% ratio: 1.16x

WU : reference_work_unit_v8_r3215.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 3892.964 secs
CPU 3889.292 secs
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 2241.100 secs, speedup: 42.43% ratio: 1.74x
CPU 2236.563 secs, speedup: 42.49% ratio: 1.74x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 2612.256 secs, speedup: 32.90% ratio: 1.49x
CPU 2608.219 secs, speedup: 32.94% ratio: 1.49x

IvyBridge, idle, Win7x64

WU : PG0009_v7.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 325.118 secs
CPU 321.580 secs
MB8_win_x64_AVX_VS2010_r3308.exe :
Elapsed 235.782 secs, speedup: 27.48% ratio: 1.38x
CPU 233.502 secs, speedup: 27.39% ratio: 1.38x
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 250.492 secs, speedup: 22.95% ratio: 1.30x
CPU 248.135 secs, speedup: 22.84% ratio: 1.30x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 276.847 secs, speedup: 14.85% ratio: 1.17x
CPU 274.499 secs, speedup: 14.64% ratio: 1.17x

WU : PG0395_v7.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 481.977 secs
CPU 478.767 secs
MB8_win_x64_AVX_VS2010_r3308.exe :
Elapsed 246.780 secs, speedup: 48.80% ratio: 1.95x
CPU 244.469 secs, speedup: 48.94% ratio: 1.96x
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 264.227 secs, speedup: 45.18% ratio: 1.82x
CPU 261.660 secs, speedup: 45.35% ratio: 1.83x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 288.556 secs, speedup: 40.13% ratio: 1.67x
CPU 286.293 secs, speedup: 40.20% ratio: 1.67x

WU : PG0444_v7.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 438.119 secs
CPU 435.071 secs
MB8_win_x64_AVX_VS2010_r3308.exe :
Elapsed 232.818 secs, speedup: 46.86% ratio: 1.88x
CPU 230.382 secs, speedup: 47.05% ratio: 1.89x
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 249.832 secs, speedup: 42.98% ratio: 1.75x
CPU 247.464 secs, speedup: 43.12% ratio: 1.76x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 270.848 secs, speedup: 38.18% ratio: 1.62x
CPU 268.353 secs, speedup: 38.32% ratio: 1.62x

WU : PG1327_v7.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 285.986 secs
CPU 282.908 secs
MB8_win_x64_AVX_VS2010_r3308.exe :
Elapsed 210.985 secs, speedup: 26.23% ratio: 1.36x
CPU 208.573 secs, speedup: 26.28% ratio: 1.36x
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 238.697 secs, speedup: 16.54% ratio: 1.20x
CPU 236.326 secs, speedup: 16.47% ratio: 1.20x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 258.525 secs, speedup: 9.60% ratio: 1.11x
CPU 256.169 secs, speedup: 9.45% ratio: 1.10x

WU : reference_work_unit_v8_r3215.wu
setiathome_8.00_windows_intelx86.exe -verb -nog :
Elapsed 2274.983 secs
CPU 2271.250 secs
MB8_win_x64_AVX_VS2010_r3308.exe :
Elapsed 1245.220 secs, speedup: 45.26% ratio: 1.83x
CPU 1241.721 secs, speedup: 45.33% ratio: 1.83x
MB8_win_x64_SSE3_VS2008_r3308.exe :
Elapsed 1318.780 secs, speedup: 42.03% ratio: 1.73x
CPU 1315.120 secs, speedup: 42.10% ratio: 1.73x
MB8_win_x86_SSE3_VS2008_r3299.exe :
Elapsed 1467.830 secs, speedup: 35.48% ratio: 1.55x
CPU 1464.257 secs, speedup: 35.53% ratio: 1.55x
ID: 1760890 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1760929 - Posted: 31 Jan 2016, 0:50:41 UTC - in response to Message 1760890.  

Thanks, @Raistmer. That info is very helpful for me. I will definitely go the new Lunatics after I figure out whether to HT or not and experiment a bit with the no_gpus in app_config.xml.

Probably in a couple of weeks or so.
ID: 1760929 · Report as offensive

Message boards : Number crunching : Question About New Stock Apps (v8, that is)


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.