| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: FightAIDS@Home
Thread: 64 bit version of FAAH application - up to 25% faster... |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 32
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There is one more point which is attainable much quicker - 64 bit computing. A lot of people are moving to 64 bits, some companies like toshiba start selling PCs preinstalled with Windows 7 x64 preinstalled.
Now our Fight aids at home application in World community grid comes only in old 32 bit verrsion. It could be easilly recompiled in 64 bits gaining some speed, possibly up to 25%. As more and more computers with x64 bit are popping up, this would be excelent opportunity to speed up project. I do not have source code of the distributed FAAH application, but people at the institute have it and possibly could recompile it or let me or others to do that. This is realtively simple job that could be done meanwhile we go for CUDA or OpenCL solution. |
||
|
|
nickoli
Senior Cruncher USA Joined: Nov 28, 2005 Post Count: 167 Status: Offline Project Badges: 
|
I agree with you 100%. Just as CUDA takes advantage of the processing power of GPUs, we would be foolish not to fully exploit the functions of our x64 PCs!
---------------------------------------- |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
See reply in other thread to same post on Autodock-CUDA
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Sekerob: Moving application from 32 bits (X86) to 64 bits (X64) is usually trivial task, with guaranteed improvements (more on this later in this post). It means usually recompiling and correcting some places that were wrongly coded and will make problems on X64. Original BOINC is both in 32 and 64 versions, so porting Fight Aids at Home module for running as native 64 or 32 application under original BOINC should be job of some hours, days at most.
World Community Grid has only 32 apllication so far, so running FAAH module in 64 bit might be tricky in WCG clinet version. I supouse it could be constructed from 2 parts, frist being 32 module called from WCG client that would later run new navie 64 bit process. I believe this might have more complications. Why 64 bit application is always faster? Because X64 mode has many more registers to use when executing compiled C++ code, registers are larger (64 bit which may help but not always is a case). Also, OS is optimized for X64 and runs quicker by itself, running more power to applications. And finally calling Win32 OS methods use Register entry, which is more effective than using stack/memory. |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges: 
|
Why 64 bit application is always faster? Because X64 mode has many more registers to use when executing compiled C++ code, registers are larger (64 bit which may help but not always is a case). Also, OS is optimized for X64 and runs quicker by itself, running more power to applications. And finally calling Win32 OS methods use Register entry, which is more effective than using stack/memory. It is for all these valid reasons that 32-bit applications running in a 64-bit OS are already much faster than in a 32-bit OS, without having to manage dual distribution. I would be curious to see numbers comparing a 32-bit app and the same one compiled in 64-bit running in the same 64-bit device. Cheers. Jean. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I would be curious to see numbers comparing a 32-bit app and the same one compiled in 64-bit running in the same 64-bit device.  I would like to see it too. I expect it would be at least 10% faster than a 32-bit application running on a 64-bit OS. But I am not seeing something that the staff are overlooking. The fact that they have not asked the project scientists to validate a 64-bit compiled version of an application makes me think that the labor involved in servicing another job queue outweighs the gain in speed. We are trying to deliver the maximum science for a given number of man-hours and funding. I do not know the figures for time spent managing job queues, but the staff do - - and they seem to have decided not to add a 64-bit job queue.But there are always more 64-bit systems on the WCG. At some point, they may decide to experiment on one project to see if it is profitable or not. Lawrence |
||
|
|
Movieman
Veteran Cruncher Joined: Sep 9, 2006 Post Count: 1042 Status: Offline |
I would be curious to see numbers comparing a 32-bit app and the same one compiled in 64-bit running in the same 64-bit device. I would like to see it too. I expect it would be at least 10% faster than a 32-bit application running on a 64-bit OS. But I am not seeing something that the staff are overlooking. The fact that they have not asked the project scientists to validate a 64-bit compiled version of an application makes me think that the labor involved in servicing another job queue outweighs the gain in speed. We are trying to deliver the maximum science for a given number of man-hours and funding. I do not know the figures for time spent managing job queues, but the staff do - - and they seem to have decided not to add a 64-bit job queue.But there are always more 64-bit systems on the WCG. At some point, they may decide to experiment on one project to see if it is profitable or not. Lawrence Lawrence; Did you ever have a job to do and look at it and say to yourself: "Darn, this will take a long time to do and I'm just not motivated so I'll take the easy way out and do a "bandaid" approach and just get by." This is how I see ignoring taking advantage of 64 bit computing which is here and in such numbers that to ignore it seems criminal. I know the vast numbers of machines are on XP32 bit but just for the sake of arguement say that 15-20% of the machines on WCG are on 64 bit OS's..Is that a fair guess? Is there any way to find out what percentage of all the machines are on BOINC 64 bit from the data thats returned to you guys? Let's also just assume that by having a 64 bit compiler those machines JUST cut 10% off the timeframes. Add it up and it's like adding hundreds of machines to the project. Some of us sit back here in frustration because we see what can be and see it ignored. Yes, it is work and yes it might cost but it's a one time work with a one time cost for a benefit that goes on forever. Please try to make this point to the people in decision making positions. If you can't tell I'm a big beleiver in the old adage of : "Do it once, Do it right, and never have to do it again." (Takes deep breath and steps off of soap box)  |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges:
|
Since I am not in the technical staff I have nothing against multiplying the number of feeding groups by two but, still, I am uncomfortable having not seen a single set of numbers about what seems to be a rather simple comparison. There are so many members also crunching in other grids and there have been so many posters saying "hey! you should have 64-bit apps, it's so much better!" how is it that after almost 3 years at WCG I have never seen anything more precise than "I expect it would be at least 10% faster" or "Let's also just assume that by having a 64 bit compiler those machines JUST cut 10% off the timeframes."
----------------------------------------By quoting Lawrence and Movieman I am not shooting at them, I am simply quoting the most recent statements on the subject. Why do we have to expect or assume if somebody has a true positive experience on this matter. Could it be that the benefit is not even close to these 10 % or that it cannot be clearly demonstrated? Then I would understand that the staff hesitates to double the complexity of the feeding/distribution process. Any real numbers anybody? Jean. |
||
|
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges:
|
Instead of rewriting current projects, why not compile new projects in 64 bit? They are coding anyways so do it in 64 bit.
----------------------------------------
Crunching for humanity since 2007!
 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
With all due respect to the scientists and techs who are, I am sure, all smarter than I am I would like to offer the following viewpoint.
Any real numbers anybody? Jean. Only the people who wrote the science code are truely capable of answering the question. Just because one program runs faster does not mean another will. Likely yes due to 64 bit registers but to see if there will be substantial improvements we will have to wait for the scientists to give it a go. I would like to suggest that a beta run of one recoded 64 bit version that is very specifically targeted to one OS/ processor for each of the projects that are going to last at least another year will help answer the question. I understand resources are limited but looking at the huge amount of time required to complete these projects, even if only the small fraction of WCG users who are already running 64 bit could be improve by a small fraction, it would produce a substantial improvement. Also, the occurence of 64 bit machines is going to increase over time so whatever stats you have now that might say the time is not now will be changing shortly. |
||
|
|
|