| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: Is Code optimization for AMD possible? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 29
|
|
| Author |
|
|
[B@H] Kokomiko
Cruncher Joined: Jul 11, 2008 Post Count: 6 Status: Offline |
When I compare the runtime between my old Intel D925 and my Phenom 9850, I have in most projects more than the treble output. On WCG only factor 1.1, that's look for me like the application is not optimized for this modern CPU.
----------------------------------------Is it's possible to give the compiler some flags to generate a optimized code?   |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Herzlich Willkommen bei unseren forums B@H] Kokomiko
----------------------------------------A tech will come and help , have a nice day Best regards [Edit 1 times, last edit by Former Member at Aug 3, 2008 11:20:18 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
When I compare the runtime between my old Intel D925 and my Phenom 9850, I have in most projects more than the treble output. On WCG only factor 1.1, that's look for me like the application is not optimized for this modern CPU. Is it's possible to give the compiler some flags to generate a optimized code? Regrettably no. The WCG sciences are all 32 bit to allow the widest possible participation by all that have a CPU with the Intel x86 code base. At your other projects they probably have x64 or as some keep referring to as AMD64 optimised sciences. WCG keeps track of the CPUs and OSses contacting and contributing. Hopefully one day when there are enough will they compile a version for 64 bit to use that 3x capacity (like that is available from my Q6600 quad). [added: Just looked in BOINCstats and it is pretty staggering that there are over 8,200 of the Q6600 type having contributed work.... that's a multiple capacity available. http://boincstats.com/stats/host_cpu_stats.php?pr=wcg&st=0 I'll vote for it to happen soonest :D ]
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 2 times, last edit by Sekerob at Aug 3, 2008 11:37:12 AM] |
||
|
|
[B@H] Kokomiko
Cruncher Joined: Jul 11, 2008 Post Count: 6 Status: Offline |
I'll vote for it to happen soonest :D I think, if a optimization is made, much more volunteers will come and help. Not much technical versed people let his high end PC willingly crunch old code for a 80386 CPU  |
||
|
|
Dmitrio
Advanced Cruncher Joined: Sep 6, 2007 Post Count: 68 Status: Offline Project Badges: 
|
Hi!
----------------------------------------If I'm not mistaken, AMD processors always had high scores in integer MIPS test, and Intel - in floating point MIPS. So any WCG user can compare different project and choose one that has the biggest number of integer operation in WU. Maybe I'm wrong and CAs would correct me, but that would be a HCC project. Dmitrio  |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
I do not choose a project because it gives high credit or crunches through some faster than another project. I crunch at WCG because it's a life choice and me being convinced that contributing here makes the biggest difference.
----------------------------------------There are a few life science projects that have a 64bit compilation, think SIMAP is one, but they have only work for 2 - 3 days a month. Of course, a 64bit OS is another premiss. AND, now the crux. Most parts of WCG sciences are floating point intensive. Does not matter what compilation, they process in 32 bit, which is why you wont see the Whetstone benchmark vary that much between 32 and 64 bits CPU/OS. The integer part of the benchmark (Dhrystone) is usually good to prop up the claim value for work. But, it's here I see most value to donating most of the idle CPU cycles.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Dmitrio
Advanced Cruncher Joined: Sep 6, 2007 Post Count: 68 Status: Offline Project Badges:
|
I do not choose a project because it gives high credit or crunches through some faster than another project. I crunch at WCG because it's a life choice and me being convinced that contributing here makes the biggest difference. I totally agree with you on this, but if I'm ready to crunch any WCG project, why not choose the one that is more suitable to my CPU and will give the biggest return? Dmitrio [Edit 1 times, last edit by Dmitrio at Aug 4, 2008 10:00:42 AM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Maybe HCC has more Integer, i don't know.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
So any WCG user can compare different project and choose one that has the biggest number of integer operation in WU. Maybe I'm wrong and CAs would correct me, but that would be a HCC project. Dmitrio No, you are not wrong and I have mentioned it several times with figures to support it. Processed through the same Q6600 clocked at the same speed a HCC WU needs an average 3.75 hours under XP32 vs 2.90 under Ubuntu 64. And it's not a matter of claiming more credits it's a matter of being more efficient with the same machine. The claims are in the same range in both cases with a slight advantage to XP32 (65.37) over Ubuntu 64 (62.40). And when talking of granted credits the difference is even worse for Ubuntu 64 since Linux is often underpaid: 70.14 for XP32 vs 62.13 for Ubuntu 64. Overall the machine is still getting more credits per day (but not as much as it should) and that is perfectly normal since it crunches 33 HCC WUs in Ubuntu 64 vs 26 under XP32. Note that I described the case between Windows 32 and Linux 64 because I have no Windows 64 system, but I think the differences should be the same between a Win32 and a Win64, at least as far as runtime is concerned (maybe not for credits). Cheers. Jean. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges: 
|
Hi,
----------------------------------------just to confirm Jean's assessment. I have one powerful system running XP64 and it is chronically (since 1 year) under granted. The contrast is particularly obvious if I compare a Q6600 running XP32 (over 170 results) and the double E5345 running XP64 (over 320 results). The first host receives generally 12.67% more than claimed and the second host receives 9.14% less than claimed. I use boinc 32 with XP32 and boinc 64 with XP64. Considering the theoretical advantage of 64 bits systems, the recognition of the host performance seems to be not really appropriate. Furthermore, even if the Q6600 has a lightly higher frequency (2.4 GHz) then the double E5345 (2.33 GHz), thanks the Xeon design, the E5345 should bring better performance. It is difficult to identify the root cause, but it looks that the 64bits systems experience some credit prejudices. You can take a look on this comparison here . Anyway, have a good crunching ! Cheers, |
||
|
|
|