Message boards :
Number crunching :
Optimized OS X Applications
Message board moderation
Author | Message |
---|---|
Send message Joined: 8 Nov 08 Posts: 178 Credit: 6,140,854 RAC: 0 |
Despite the title, I don't have any optimized applications yet, but I'm still working on it. I have however lowered the crunch time from 1h04m to 48m on my dual-core MacBook using the x86_64 application. I have yet to benchmark either the i686 or PPC applications yet, but I have compiled versions. So far, this has just been achieved by changing the GCC compilation flags. I'll keep updates in here. |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
So how do we get a copy to test with? |
Send message Joined: 27 Feb 09 Posts: 45 Credit: 305,963 RAC: 0 |
I've already PM'd jedirock. Hopefully he wants testers. Would be nice to get a little closer to the windows apps. Mars rules this confectionery war! |
Send message Joined: 8 Nov 08 Posts: 178 Credit: 6,140,854 RAC: 0 |
So how do we get a copy to test with? I'm going to attempt to speed them up a bit more before releasing them. Glad to know people are interested though. ;) |
Send message Joined: 12 Apr 08 Posts: 621 Credit: 161,934,067 RAC: 0 |
Of course we are interested ... |
Send message Joined: 29 Jan 09 Posts: 32 Credit: 1,962,668 RAC: 0 |
Of course we are interested ... Me too :) |
Send message Joined: 12 Oct 07 Posts: 77 Credit: 404,471,187 RAC: 0 |
Of course we are interested ... and me :) |
Send message Joined: 6 Apr 08 Posts: 2018 Credit: 100,142,856 RAC: 0 |
|
Send message Joined: 12 Dec 08 Posts: 56 Credit: 269,889,439 RAC: 0 |
I've got a G4 that is currently using the stock app...if you need a tester for the PPC, I can help... Click to help Seti City. |
Send message Joined: 22 Nov 07 Posts: 36 Credit: 1,224,316 RAC: 0 |
|
Send message Joined: 2 Sep 07 Posts: 18 Credit: 180,611 RAC: 0 |
jup, opti apps for PPC G4 and G3 too would be very nice, now with the double workload WUs around. Thanks for your work. |
Send message Joined: 8 Nov 08 Posts: 178 Credit: 6,140,854 RAC: 0 |
jup, opti apps for PPC G4 and G3 too would be very nice, now with the double workload WUs around. Thanks for your work. Just a quick note that I can't test the PPC version natively. I can test it using Rosetta, but I wouldn't trust it for final numbers. So while I'm compiling PPC apps, I'm basically doing it blind. |
Send message Joined: 8 Nov 08 Posts: 178 Credit: 6,140,854 RAC: 0 |
So how do we get a copy to test with? Well, I don't think I'm going to get any more huge speed increases anytime soon. It's too late here to compile the PowerPC and Intel 32-bit variants right now, but I'll do so tomorrow and release them. I'll also benchmark the Intel 32- and 64-bit versions, see if there's any difference, and give you my findings. As always, YMMV, especially with the PPC apps, but hopefully they'll help. As for why it's taken so long, I've been compiling some apps for yoyo@home in the past few days too. Seem to be going pretty speedily along now, but that could also be because they have hand-tweaked ASM code in there. ;) |
Send message Joined: 27 Feb 09 Posts: 45 Credit: 305,963 RAC: 0 |
Jedi, no need to apologise for taking your time. Your doing this off of your own back. The many of us mac users are very grateful for your efforts. Mars rules this confectionery war! |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
I don't have a Mac. If you give me one I'll try the apps out. ;p And good job. Atleast someone is willing to take the time to do those apps and share their work too. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 8 Nov 08 Posts: 178 Credit: 6,140,854 RAC: 0 |
All right, my latest compiles. I actually tried something I haven't done with BOINC apps before, and stuck all three architectures (Intel 32- and 64-bit, and PowerPC) in one fat binary for convenience. So it's the same file for any Mac. I don't think there's really any performance difference between Intel 32- and 64-bit versions, other than possibly different versions of SSE. For PowerPC, I've tried to do as much as possible, and that means GCC optimizations, along with Altivec. This is the zip with everything you should need. Instructions for installing: 1) Unzip file to produce an application and an app_info 2) Shut down BOINC 3) Copy files to /Library/Application Support/BOINC Data/projects/milkyway.cs.rpi.edu_milkyway 4) Open a Terminal.app window, and copy and paste "sudo chown boinc_master:boinc_project /Library/Application Support/BOINC Data/projects/milkyway.cs.rpi.edu_milkyway/app_info.xml /Library/Application Support/BOINC Data/projects/milkyway.cs.rpi.edu_milkyway/milkyway_0.18_apple_darwin" in and hit Enter. Enter your password, and that fixes permissions. 5) Start BOINC I haven't done a full crunch of a WU with these yet as they're so fresh, but they pass Travis's test WUs and my first crunch is over 20% done at this point. Enjoy! |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
Installed on my iMac, will have to wait and see though. I always just reinstall the Manager to fix the permissions problem. |
Send message Joined: 26 Dec 07 Posts: 41 Credit: 2,582,082 RAC: 0 |
3 to 4 per cent faster on a C2D. |
Send message Joined: 8 Nov 08 Posts: 178 Credit: 6,140,854 RAC: 0 |
3 to 4 per cent faster on a C2D. I took off 25% of my crunching time with these on my 2GHz C2D. From 1h4m to 48-49m. |
Send message Joined: 26 Dec 07 Posts: 41 Credit: 2,582,082 RAC: 0 |
That's odd. My 2.8GHz C2D (E8235) has reduced the longer tasks from 35.49 (2149 sec) to 34.20 (2060 sec), a reduction of about 4%. The shorter tasks have gone from 22.05 (1325 sec) to 21.23 (1283 sec), a reduction of about 3%. Maybe my C2D is newer than yours and the stock app was already making use of instructions not available on an older chip? |
©2024 Astroinformatics Group