v5.22 Experimental optimized app

Message boards : Number crunching : v5.22 Experimental optimized app

To post messages, you must log in.

Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 13 · Next

AuthorMessage
Profile mdoerner
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 30 Jul 08
Posts: 202
Credit: 6,998,388
RAC: 0
Message 901 - Posted: 11 May 2009, 10:43:14 UTC - in response to Message 900.  

When I tested his app with the benchmark, I didn't change the rights on the file, and the benchmark ran fine. Maybe omitting step 6 is all you have to do?

Mike D
ID: 901 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grubix

Send message
Joined: 20 Jul 08
Posts: 44
Credit: 6,270,837
RAC: 0
Message 903 - Posted: 11 May 2009, 11:27:31 UTC

I did not run benchmark, i have only restarted BOINC. Furthermore, i change the rights to "rwxrwxrwx", more is not possible.

I will try only to copy the file, not to change any rights.



PS: I search the forum for "benchmark", but don't found how do i run it.
ID: 903 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mdoerner
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 30 Jul 08
Posts: 202
Credit: 6,998,388
RAC: 0
Message 904 - Posted: 11 May 2009, 12:19:38 UTC - in response to Message 903.  

Here's the link to the benchmark app....eb.tgz

Basically, unpack the file, replace the "enigma" file with "enigma_static" (i.e. rename it) or one of the apps from TJM's test.tgz file, and then type "./start" within the shell. The benchmark should start and complete within 3-6 minutes depending on the processor and app you've selected. You might want to stop BOINC while this runs, but it's not absolutely needed.

Mike Doerner
ID: 904 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grubix

Send message
Joined: 20 Jul 08
Posts: 44
Credit: 6,270,837
RAC: 0
Message 907 - Posted: 11 May 2009, 18:17:14 UTC - in response to Message 904.  

Here's the link to the benchmark app....eb.tgz

Thanks, benchmark works fine with original app and TJM opt app. The client from Buninek does not run.

grubi@Laptop:~/Enigma/benchmark/enigma_benchmark$ ls -l
insgesamt 1288
-rw-r--r-- 1 grubi grubi   6152 2007-08-22 02:35 00bigr.naval
-rw-r--r-- 1 grubi grubi     88 2007-09-05 03:34 00hc_.resume
-rw-r--r-- 1 grubi grubi     88 2009-05-11 19:55 00hc.resume
-rw-r--r-- 1 grubi grubi 134860 2007-08-22 02:35 00trigr.naval
-rw-r--r-- 1 grubi grubi    100 2007-09-02 14:06 bench_cipher
-r-xr-xr-x 1 grubi grubi 640584 2009-05-09 18:32 enigma
-r-xr-xr-x 1 grubi grubi 502068 2007-09-22 23:42 enigma_original
-rwxr-xr-x 1 grubi grubi    132 2007-11-30 15:23 start

grubi@Laptop:~/Enigma/benchmark/enigma_benchmark$ ./start
rm: Entfernen von „results.txt“ nicht möglich: No such file or directory
./start: line 4: ./enigma: cannot execute binary file

real    0m0.095s
user    0m0.000s
sys     0m0.000s

grubi@Laptop:~/Enigma/benchmark/enigma_benchmark$ md5sum enigma
28804d55b81c28e3091dd0655c76d818  enigma


Any idea? Agrees the md5 checksum from Buninek's enigma app?

I use Kubuntu 8.10, may be a missing packet?
ID: 907 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
buninek

Send message
Joined: 30 Sep 08
Posts: 2
Credit: 136,916
RAC: 0
Message 908 - Posted: 11 May 2009, 19:08:08 UTC - in response to Message 907.  
Last modified: 11 May 2009, 19:52:23 UTC

It is static version. Compiled using gcc-3.4.6 (glibc 2.7).
In me it works with linux x86 and x86_64.
I have no idea why it does not work.

Perhaps your version of glibc is too old?

Try with this version - Link
ID: 908 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grubix

Send message
Joined: 20 Jul 08
Posts: 44
Credit: 6,270,837
RAC: 0
Message 909 - Posted: 11 May 2009, 22:22:18 UTC - in response to Message 908.  

Thank you for your help and patience.

Try with this version - Link

Yea, nice, this version works. :-)

Thank you for this version you have created for me, i will report about the speed.

Perhaps your version of glibc is too old?

This information will help you eventually?

/lib/libc.so.6

GNU C Library development release version 2.8.90, by Roland McGrath et al.
Copyright (C) 2008 Free Software Foundation, Inc.                         
This is free software; see the source for copying conditions.             
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A       
PARTICULAR PURPOSE.                                                       
Compiled by GNU CC version 4.3.2.                                         
Compiled on a Linux >>2.6.24-23-server<< system on 2009-01-29.            
Available extensions:                                                     
        crypt add-on version 2.1 by Michael Glad and others               
        GNU Libidn by Simon Josefsson                                     
        Native POSIX Threads Library by Ulrich Drepper et al              
        BIND-8.2.3-T5B                                                    
For bug reporting instructions, please see:                               
<http://www.gnu.org/software/libc/bugs.html>.


By, Grubix.
ID: 909 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
MJ

Send message
Joined: 17 Nov 07
Posts: 16
Credit: 95,844
RAC: 0
Message 911 - Posted: 12 May 2009, 5:35:15 UTC - in response to Message 909.  

I get no work sent w/ the upgraded app. Your app_info.xml files doesn't have version Enigma .76b. I think you need to update the files and have really easy step by step instructions. I have windows vista.
ID: 911 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grubix

Send message
Joined: 20 Jul 08
Posts: 44
Credit: 6,270,837
RAC: 0
Message 913 - Posted: 12 May 2009, 12:09:24 UTC


Speed test

CPU: Intel Core2 Quad Q6700 @ 2.66GHz
OS: Linux, 2.6.28-11-generic (Kubuntu 9.04)
Workunits: hceyz72_2_xxxxxxx_r0_0

Standard App
------------
3257,94
3259,26
3254,34
3278,12
-------
3262,42 = 54:22


athlon_tb
---------
2596,79
2604,58
2627,45
2638,82
-------
2616,91 = 43:36 -> -19,79%


athlon_xp
---------
2609,53
2627,78
2631,20
2640,16
-------
2627,17 = 43:47 -> -19,47%


default
-------
2764,89
2773,14
2766,39
2788,74
-------
2773,29 = 46:13 -> -14,99%


k6
-------
2606,18
2597,32
2615,24
2613,94
-------
2608,17 = 43:28 -> -20,05%


k6-2: app binary identical with k6
k6-3: app binary identical with k6


pentium3
--------
2629,99
2618,91
2595,76
2613,12
2608,36
2617,53
2619,44
2614,23
--------
2614,67 = 43:35 -> -19,86%


pentium3_sse2
-------------
2572,67
2546,68
2543,49
2534,80
2548,17
2559,23
2558,73
2558,97
-------
2552,84 = 42:33 -> -21,75%


pentium4
--------
2621,04
2627,81
2624,36
2735,95
-------
2652,29 = 44:12 -> -18,70%


pentiumpro
----------
2613,37
2594,89
2593,02
2591,95
2613,10
2600,22
2615,04
2619,90
-------
2605,19 = 43:25 -> -20,15%


buninek (with Intel-CPU out of competition.)
-------
3647,88
3640,56
3643,89
3659,42
-------
3647,94 = 60:48 -> +11,81%


Summary:
--------
3262,42 = 54:22 (standard app)
3647,94 = 60:48 -> +11,81% (buninek)
2773,29 = 46:13 -> -14,99% (default)
2652,29 = 44:12 -> -18,70% (pentium4)
2627,17 = 43:47 -> -19,47% (athlon_xp)
2616,91 = 43:36 -> -19,79% (athlon_tb)
2614,67 = 43:35 -> -19,86% (pentium3)
2608,17 = 43:28 -> -20,05% (k6)
2605,19 = 43:25 -> -20,15% (pentiumpro)
2552,84 = 42:33 -> -21,75% (pentium3_sse2)


Test with benchmark:
--------------------
3:36,11 (standard app)
4:06,77 -> +14,19% (buninek)
2:51,37 -> -20,70% (pentium3)
2:50,25 -> -21,22% (pentiumpro)
2:44,99 -> -23,66% (pentium3_sse2)


I hope the test is interesting for someone. The AMD X2 will follow.

If someone has a Core2 optimized app, I would test it. :-)
ID: 913 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mdoerner
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 30 Jul 08
Posts: 202
Credit: 6,998,388
RAC: 0
Message 916 - Posted: 12 May 2009, 19:33:58 UTC - in response to Message 913.  
Last modified: 12 May 2009, 19:36:17 UTC

I can compile a 32-bit Open64 CORE optimized app if you like (there's already a 64-bit one in the file). It will be compiled on an AMD Phenom, but it should still work.

Mike D's Linux Apps

Mike D
ID: 916 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grubix

Send message
Joined: 20 Jul 08
Posts: 44
Credit: 6,270,837
RAC: 0
Message 917 - Posted: 12 May 2009, 21:32:16 UTC

Yes, that would be glad. This is really kind from you.

@buninek: your app seems to be 30% faster on AMD X2.


Thanks, Grubix.
ID: 917 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mdoerner
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 30 Jul 08
Posts: 202
Credit: 6,998,388
RAC: 0
Message 918 - Posted: 12 May 2009, 21:34:13 UTC - in response to Message 917.  

OK, I've added a 32-bit Core optimized app to the above link. I got 3m 38s on my Phenom (compared to 3m 14s with the Athlon 32-bit app). Try it. See if it works better for you.

Mike D
ID: 918 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grubix

Send message
Joined: 20 Jul 08
Posts: 44
Credit: 6,270,837
RAC: 0
Message 920 - Posted: 13 May 2009, 11:52:04 UTC - in response to Message 918.  

OK, I've added a 32-bit Core optimized app...

Thank you, but the app is not faster than P3 version. I have tested all the 32bit versions from your archive. Here are the benchmark values.

CPU: Intel Core2 Quad Q6700 @ 2.66GHz
OS: Linux, 2.6.28-11-generic (Kubuntu 9.04)

3:36,11 (standard app)
3:33,38 -> -1,16% (Opteron32)
3:29,02 -> -3,28% (Barcelona32)
3:28,52 -> -3,51% (Athlon64fx32)
3:28,29 -> -3,62% (Athlon64_32)
3:26,39 -> -5,49% (Intel-Core_32)
3:06,59 -> -13,66% (Athlon_32)

By, Grubix.
ID: 920 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mdoerner
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 30 Jul 08
Posts: 202
Credit: 6,998,388
RAC: 0
Message 921 - Posted: 13 May 2009, 19:50:49 UTC - in response to Message 920.  
Last modified: 13 May 2009, 19:51:11 UTC

Right, the only one that SLIGHTLY beat the TJM P3 app was my Athlon 32-Bit version, but it was within a second of TJM's app so I call it even money between the two.

Mike Doerner
ID: 921 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grubix

Send message
Joined: 20 Jul 08
Posts: 44
Credit: 6,270,837
RAC: 0
Message 922 - Posted: 14 May 2009, 20:19:10 UTC

CPU: AMD Turion X2 Dual Core Mobile RM-70 @ 2,0 GHz
OS: Linux, 2.6.27-11-generic (Kubuntu 8.10)
WorkUnits: hceyz72_0_xxxxxxx_r0_0

Standard App
------------
3737,49
3696,81
3813,37
3787,63
3768,74
3758,75
---------
3760,47 = 1:02:40


athlon_tb
---------
3170,84
3154,62
3118,37
3156,68
-------
3150,13 = 0:52:30 -> -16,23%


athlon_xp
---------
3172,84
3145,07
3162,78
3151,61
-------
3158,08 = 0:52:38 -> -16,02%


default
-------
3538,51
3492,93
3531,42
3510,04
-------
3518,23 = 0:58:38 -> -6,44%


k6
-------
3204,84
3219,68
3187,92
3250,36
-------
3215,70 = 0:53:35 -> -14,49%


k6-2: app binary identical with k6
k6-3: app binary identical with k6


pentium3
--------
3163,00
3167,30
3150,47
3159,18
---------
3159,98 = 0:52:40 -> -15,97%


pentium3_sse2
-------------
4758,62
4793,38
4736,76
4738,28
-------
4756,76 = 1:19:17 -> +26,49%


pentium4
--------
3172,47
3212,23
3159,40
3212,52
-------
3189,16 = 0:53:09 -> -15,19%


pentiumpro
----------
läuft
3164,34
3171,39
3148,74
3187,17
-------
3167,91 = 0:52:48 -> -15,76%


bunirek
-------
2562,35
2552,33
2559,85
2607,16
2546,66
2574,52
-------
2567,15 = 0:42:47 -> -31,73%


Summary:
--------
3760,47 = 1:02:40 (standard app)
4756,76 = 1:19:17 -> +26,49% (pentium3_sse2)
3518,23 = 0:58:38 -> -6,44% (default)
3215,70 = 0:53:35 -> -14,49% (k6)
3189,16 = 0:53:09 -> -15,19% (pentium4)
3159,98 = 0:52:40 -> -15,97% (pentium3)
3158,08 = 0:52:38 -> -16,02% (athlon_xp)
3150,13 = 0:52:30 -> -16,23% (athlon_tb)
2567,15 = 0:42:47 -> -31,73% (bunirek)


Benchmark:
----------
08:22,23 (standard app)
10:29,78 -> +25,40% (pentium3_sse2)
07:41,00 -> -8,21% (default)
07:04,58 -> -15,46% (k6)
06:56,03 -> -17,16% (pentium4)
06:56,00 -> -17,17% (athlon_tb)
06:54,41 -> -17,49% (athlon_xp)
06:54,41 -> -17,49% (pentium3)
05:36,42 -> -33,02% (bunirek)


By, Grubix.
ID: 922 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mdoerner
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 30 Jul 08
Posts: 202
Credit: 6,998,388
RAC: 0
Message 923 - Posted: 14 May 2009, 21:36:44 UTC - in response to Message 922.  
Last modified: 14 May 2009, 21:39:09 UTC

Bunirek's 32-bit app was only about 8-10 seconds slower on the benchmark than my 64-bit optimized app. Definitely beat TJM's P3 app. Should be posted somewhere for people to d/load. FWIW.

Mike Doerner
ID: 923 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile (_KoDAk_)

Send message
Joined: 13 May 09
Posts: 2
Credit: 10,021
RAC: 0
Message 960 - Posted: 26 May 2009, 16:41:57 UTC
Last modified: 26 May 2009, 16:42:25 UTC

Is there a fast kernel for Intel Q9550 win-x64 ?
ID: 960 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mdoerner
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 30 Jul 08
Posts: 202
Credit: 6,998,388
RAC: 0
Message 962 - Posted: 26 May 2009, 19:35:57 UTC - in response to Message 960.  

Is there a fast kernel for Intel Q9550 win-x64 ?



If you mean an optimized app for Win 64-bit, no not yet. I think you can download a copy of Microsoft Visual C++ for your own personal use, and you can compile a copy that way. Or gcc (MinGW?) for windows may work better since this source code is written to NOT be platform specific.

Mike Doerner
ID: 962 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile TJM
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 25 Aug 07
Posts: 843
Credit: 267,994,998
RAC: 0
Message 963 - Posted: 26 May 2009, 22:03:10 UTC

My Windows apps were compiled with Mingw, I also tried different versions of MS Visual Studio, but executables build with it were significantly slower.

I was going to try Visual Studio + Intel C Compiler for Windows, but I couldn't get it to work (and the Intel C Compiler for Windows isn't free, just a time-limited trial).

M4 Project homepage
M4 Project wiki
ID: 963 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [AF>Libristes] Dudumomo

Send message
Joined: 15 Feb 09
Posts: 20
Credit: 196,260
RAC: 0
Message 964 - Posted: 27 May 2009, 7:33:27 UTC - in response to Message 963.  
Last modified: 27 May 2009, 7:57:38 UTC

Wow A great job has been done here ! (And is doing)
I'm trying to put all together the differents opti app for every boinc projects.

I have decided to start by Enigma. (But I didn't check this part of the forum for a while) So, I am a bit lost...

It would be very nice to know exactly which app is the best according to the proc.

I can help to make some test as well if you need so.
I'm running on Gentoo 64b
ID: 964 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mdoerner
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 30 Jul 08
Posts: 202
Credit: 6,998,388
RAC: 0
Message 967 - Posted: 27 May 2009, 12:51:31 UTC - in response to Message 964.  
Last modified: 27 May 2009, 12:52:20 UTC

Wow A great job has been done here ! (And is doing)
I'm trying to put all together the differents opti app for every boinc projects.

I have decided to start by Enigma. (But I didn't check this part of the forum for a while) So, I am a bit lost...

It would be very nice to know exactly which app is the best according to the proc.

I can help to make some test as well if you need so.
I'm running on Gentoo 64b


Without knowing which processor you're running, it's hard to tell....;-)

Since you're 64-bit linux, I would try one of the apps in the Open64 64-bit linux thread. I have compiled every available processor optimization within the Open64 suite with the exception of Intel Wolfdale (which AMD admits they forgot to put wolfdale.so in the release, so it will be available in early-June).

Mike Doerner
ID: 967 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 13 · Next

Message boards : Number crunching : v5.22 Experimental optimized app




Copyright © 2024 TJM