Message boards :
Number crunching :
Updating for upcoming apps.....
Message board moderation
Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 · Next
Author | Message |
---|---|
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Too bad that the BOINC server had its couple of disk crashes last year, those gobbled up all those old versions and their release dates which were at that time still available from the download site and its mirrors. No more. :-( Looking at one of my oldest boxes I have BOINC 4.25.0, dated 2005/03/08, on it with an app: ap_2.11_windows_intelx86.exe. Perhaps I should fire it up and see if it will run the v6 AP application. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
All of which reminds me - to get back on topic - that I have to downgrade that server from v6.10.58 to v5.something so I can test the service mode restart code for v5 in the new installer. Does your 'remove all traces of BOINC from my computer' script contain something to strip off the sandbox security group memberships from the data folder, so I can get back to a clean v5 service without losing work in progress, either here or for Einstein? That script works on Windows 2000 and XP, not so good on Vista and above. Better to do it manually from this FAQ. By the way, it appears that a full uninstall option is in the works for the installer after direct complaints to David. ;-) |
LadyL Send message Joined: 14 Sep 11 Posts: 1679 Credit: 5,230,097 RAC: 0 |
ETA for the v0.40 Lunatics installer is tomorrow, provided we get the green light from Eric, and I get all the information I need to put into the release notes. If you were unhappy with those from last time for some reason, NOW is the time to tell me. Constructive criticism only please. At my current stress levels anything else will find you on my ignore list. I'm not the Pope. I don't speak Ex Cathedra! |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
Looking forward. I've gone ahead and set NNT and suspended all the "ready to start" APs I've got. One thing I'm wondering about though is over on the tasks page where it shows the stock application, once I go over to anon. platform, will that update, or will tasks originally assigned to the stock app still show and count to the application details as stock even though they were done with optimized? Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Looking forward. I've gone ahead and set NNT and suspended all the "ready to start" APs I've got. The webpage always displays what the tasks were assigned to. So once you switch to the anon platform only new tasks will be marked as such. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Looking forward. I've gone ahead and set NNT and suspended all the "ready to start" APs I've got. The way to update the details on the task page is to reset the project after you're installed the Optimised apps, (If you save your Wu files to somewhere safe before you reset the project, then restore them afterwards, you'll then skip the downloads as the Wu's are resent) (It'll also mean those Optimised Wu's will go towards the anonymous platform APR instead of the Stock APR too) Claggy |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
The webpage always displays what the tasks were assigned to. So once you switch to the anon platform only new tasks will be marked as such. Noted. After installing the new package when it is released, I'll just drop back down to a tiny cache to make sure I don't end up with more than I can handle. ETAs should be huge and slowly come down though. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Looking forward. I've gone ahead and set NNT and suspended all the "ready to start" APs I've got. As Cosmic_Ocean is trying to keep clean statistics, and depending on circumstances, he might consider doing a project reset just before installing Lunatics. That clears all work, and work requests after the Lunatics install will get those tasks resent and assigned to the optimized app. If the download pipe is saturated, that would probably not be a good move. But if the tasks were "resent" just before the Tuesday outage and the download servers remain enabled during the outage as has been usual lately, they would probably download fairly easily. This is not a recommendation, just a possibility to consider. Joe |
LadyL Send message Joined: 14 Sep 11 Posts: 1679 Credit: 5,230,097 RAC: 0 |
Looking forward. I've gone ahead and set NNT and suspended all the "ready to start" APs I've got. Provided we are out before maintenance... I'm not the Pope. I don't speak Ex Cathedra! |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
...If the download pipe is saturated, that would probably not be a good move. But if the tasks were "resent" just before the Tuesday outage and the download servers remain enabled during the outage as has been usual lately, they would probably download fairly easily. Good point, though of course the installer will be out before some Tuesday maintenance ;^) A more complex alternative is to save all the WU files just before doing the project reset, then put them back after doing the Lunatics install. That way when they are "resent" the core client will find they're already there and not have to download again. I've never run 6.2.19 as Cosmic_Ocean is using, but AFAIK that ought to work. The project reset does lose whatever partial crunching has been done on a task, of course. But it gives a clean transition that way. Joe |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
The stats are easy enough to figure out. Since I have 7 tasks in the cache that have not started and are currently suspended, it is easy enough to cut & paste the entries in the spreadsheet from one worksheet to another. Or I can just type in the taskID and wuIDs again. I did think about just running the cache down, but that will take about 2.5 days at the moment and.. well I don't want to wait that long. Besides, it would be nice to have something to crunch and a tiny bit of cache while the work requests endlessly try to get more tasks to fill the cache even more. Starting from an empty cache is rough and seems to take a lot longer to fill up than if you already have something in it. The application details page will figure itself out. These 7 tasks that I have will end up going toward the stock tally from what I've interpreted, which may bring the APR up a bit for a comparison though. I'll just make a note of what the APR was in the spreadsheet when I switch apps so there's something to compare to, if average runtime isn't good enough. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
DMMD Send message Joined: 14 Feb 00 Posts: 118 Credit: 71,564,960 RAC: 0 |
Hi Joe, and other interested parties; Couldn't post this in the most appropriate venue, and this seemed a ~ reasonable 2nd choice....D'tor if not, of course: i7 3960@4.8 GHz, BOINC running Seti@Home, only Ftst_v7 completed successfully. ========================================================= Ftst_v7_J45 started. Optimal function choices: -------------------------------------------------------- name timing error -------------------------------------------------------- v_BaseLineSmooth (no other) v_GetPowerSpectrum 0.000084 0.00000 test v_vGetPowerSpectrum 0.000042 0.00000 test v_vGetPowerSpectrum2 0.000056 0.00000 test v_vGetPowerSpectrumUnrolled 0.000039 0.00000 test v_vGetPowerSpectrumUnrolled2 0.000060 0.00000 test v_avxGetPowerSpectrum 0.000037 0.00000 test v_avxGetPowerSpectrum 0.000037 0.00000 choice v_ChirpData 0.003177 0.00000 test fpu_ChirpData 0.008388 0.00000 test fpu_opt_ChirpData 0.003173 0.00000 test v_vChirpData_x86_64 0.040736 0.00000 test sse1_ChirpData_ak 0.004972 0.00000 test sse1_ChirpData_ak8e 0.004023 0.00000 test sse1_ChirpData_ak8h 0.004222 0.00000 test sse2_ChirpData_ak 0.004688 0.00000 test sse2_ChirpData_ak8 0.003043 0.00000 test sse3_ChirpData_ak 0.004506 0.00000 test sse3_ChirpData_ak8 0.003041 0.00000 test avx_ChirpData_a 0.001442 0.00000 test avx_ChirpData_b 0.001636 0.00000 test avx_ChirpData_c 0.001453 0.00000 test avx_ChirpData_d 0.001338 0.00000 test avx_ChirpData_d 0.001338 0.00000 choice v_Transpose 0.002160 0.00000 test v_Transpose2 0.002891 0.00000 test v_Transpose4 0.001206 0.00000 test v_Transpose8 0.002194 0.00000 test v_pfTranspose2 0.001133 0.00000 test v_pfTranspose4 0.001106 0.00000 test v_pfTranspose8 0.002250 0.00000 test v_vTranspose4 0.000594 0.00000 test v_vTranspose4np 0.001025 0.00000 test v_vTranspose4ntw 0.016963 0.00000 test v_vTranspose4x8ntw 0.006324 0.00000 test v_vTranspose4x16ntw 0.001870 0.00000 test v_vpfTranspose8x4ntw 0.016348 0.00000 test v_avxTranspose4x8ntw 0.006472 0.00000 test v_avxTranspose4x16ntw 0.001450 0.00000 test v_avxTranspose8x4ntw 0.016965 0.00000 test v_avxTranspose8x8ntw_a 0.005262 0.00000 test v_avxTranspose8x8ntw_b 0.005684 0.00000 test v_vTranspose4 0.000594 0.00000 choice FPU opt folding 0.001603 0.00000 test AK SSE folding 0.000352 0.00000 test BH SSE folding 0.000336 0.00000 test JS AVX_a folding 0.000297 0.00000 test JS AVX_c folding 0.000299 0.00000 test JS AVX_a folding 0.000297 0.00000 choice Test duration 3.30 seconds Ftst_v7 completed successfully. D |
LadyL Send message Joined: 14 Sep 11 Posts: 1679 Credit: 5,230,097 RAC: 0 |
Which BOINC version you suggest to use ? still the 6.10.58 ? or go for the 6.12.34 ? up to you. 6.10.58 is old but reliable. 6.12.34 has a few useful extra features and combined with SIV to stop it from going into excessive backoffs works fine. what s about the cc_fonfig and/or the app_info ? what about them? the installer generates a new app_info.xml according to the choices made on the installer screens. You have to redo flops, counts and other manual alterations. cc_config.xml is completely independant. EDIT: You have NVidia GPUs. The installer doesn't have an NVidia AP app. The app is in beta and will be made availble separately (i.e. you'll have to do a manual install) The installer will offer a choice of MB and AP for CPU, MB for NVidia and MB and AP for ATI. I'm not the Pope. I don't speak Ex Cathedra! |
red-ray Send message Joined: 24 Jun 99 Posts: 308 Credit: 9,029,848 RAC: 0 |
The installer generates a new app_info.xml according to the choices made on the installer screens. You have to redo <flops>, <counts> and other manual alterations. Why does the installer not insert <flops> lines? All it needs to know is the speed of the CPU and the GPU Peak GFLOPS of the slowest GPU both of which are almost trivial to get. It could easily use the code from BOINC.exe. Having worked out what the values should be it should then popup a panel with these as recommended values and the user should have options to use them, use different ones or not include <flops> at all. Doing things this way would be far easier and less error prone that manually editing the app_info.xml. I have not written code to get AMD/ATI flops, but I can't imagine it's any harder. Ideally the CPU benchmark would run threads on ALL the CPUs at the same time. The current BOINC code only seems to do one CPU at a time. [GPU CUDA Information] <- SIV64X - System Information Viewer V4.28 Beta-01 RED::ray nVidia CUDA V4.10 Peak Compute CPUs and Memory Shader Level 2 Current Total Threads Registers CUDA Cache Stack FIFO Heap |Bus-Numb-Fun| Device Name GFLOPS Capability Threads Clock Clock Cache Memory Memory per Block per Block API Config Size Size Size [3 - 00 - 0] GeForce GTX 460 683.6 V2.01 7 1,536 1.90GHz 1.53GHz 512KB 439MB 1.00GB 1,024 32,768 V3.20 None 1KB 1MB 8MB [4 - 00 - 0] GeForce GTX 460 683.6 V2.01 7 1,536 1.90GHz 1.53GHz 512KB 424MB 1.00GB 1,024 32,768 V3.20 None 1KB 1MB 8MB [5 - 00 - 0] GeForce GT 430 179.2 V2.01 2 1,536 700MHz 1.40GHz 128KB 359MB 512MB 1,024 32,768 V3.20 None 1KB 1MB 8MB [7 - 00 - 0] GeForce GT 520 103.7 V2.01 1 1,536 667MHz 1.62GHz 64KB 371MB 512MB 1,024 32,768 V3.20 None 1KB 1MB 8MB [CPU Benchmark] <- SIV64X - System Information Viewer V4.28 Beta-01 RED::ray SIV64X - CPU Dhrystone and Whetstone Benchmarks on \\RED - Windows 7 x64 Ultimate V6.01 Build 7601 Service Pack 1 Intel Core i7 Extreme 980X (Gulftown) 3.33GHz [B1] Include Integer Time Floating Time Include Integer Time Floating Time [Y] CPU-0 5,524 2.07 3,529 2.01 [Y] CPU-1 5,546 2.07 3,557 2.00 [Y] CPU-2 5,520 2.06 3,457 2.00 [Y] CPU-3 5,590 2.04 3,529 2.01 [Y] CPU-4 5,636 1.98 3,507 2.00 [Y] CPU-5 5,554 2.07 3,529 2.01 [Y] CPU-6 5,441 1.92 3,507 2.00 [Y] CPU-7 5,546 2.06 3,529 2.01 [Y] CPU-8 5,537 2.07 3,557 2.00 [Y] CPU-9 5,471 2.07 3,429 2.01 [Y] CPU-10 5,537 2.07 3,483 1.98 [Y] CPU-11 5,558 2.01 3,529 2.01 [#############################################################################################################] Elapsed 4.2 Total 66,460 + 42,142 Average 5,538 + 3,511 Minimum 5,441 + 3,429 Maximum 5,636 + 3,557 CPU Time 48.55 Include Integer Time Floating Time Include Integer Time Floating Time [Y] CPU-0 8,689 2.07 3,678 2.01 [_] CPU-1 [_] CPU-2 [_] CPU-3 [_] CPU-4 [_] CPU-5 [_] CPU-6 [_] CPU-7 [_] CPU-8 [_] CPU-9 [_] CPU-10 [_] CPU-11 [#############################################################################################################] Elapsed 4.2 Total 8,689 + 3,678 Average 8,689 + 3,678 Minimum 8,689 + 3,678 Maximum 8,689 + 3,678 CPU Time 4.09 |
LadyL Send message Joined: 14 Sep 11 Posts: 1679 Credit: 5,230,097 RAC: 0 |
Ray, believe me, if it was trivial, we'd be doing it. The flops you put in are a measure of the app performance on the GPU not a static GPU value. They even fluctuate with AR. The way the installer is written atm does not allow for simple insertion of values. We'd be looking at a complete revamp. While that was something that 'would be nice to do' we simply don't have the resources. I'm not the Pope. I don't speak Ex Cathedra! |
red-ray Send message Joined: 24 Jun 99 Posts: 308 Credit: 9,029,848 RAC: 0 |
Ray, believe me, if it was trivial, we'd be doing it. I should have made it clear, it's only getting the raw flops values I consider trivial. As I have not looked at the installer code I would not comment on how hard it would be to change it. I would expect the code to convert from GPU Peak GFLOPs to <flops> would not be too dificult given the data from a selection of GPUs. To get it started this is what I get for my GTX 460 and GT 520. Currently <flops>6.0e09</flops> is being used. 20/03/2012 00:00:54 | | NVIDIA GPU 0: GeForce GTX 460 (driver version 28562, CUDA version 4010, compute capability 2.1, 1024MB, 684 GFLOPS peak) 20/03/2012 12:36:37 | SETI@home | [dcf] DCF: 1.133698->1.052642, raw_ratio 0.323136, adj_ratio 0.285029. 20/03/2012 00:00:54 | | NVIDIA GPU 3: GeForce GT 520 (driver version 28562, CUDA version 4010, compute capability 2.1, 512MB, 104 GFLOPS peak) 20/03/2012 12:40:11 | SETI@home | [dcf] DCF: 0.973314->1.841232, raw_ratio 1.841232, adj_ratio 1.891714 If you are wondering why my DCF is daft look at http://setiathome.berkeley.edu/forum_thread.php?id=67273 |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14649 Credit: 200,643,578 RAC: 874 |
The updated installer is now ready, but I still haven't had a reply from Eric. We could see that they were having a bit of a struggle with the AP splitters yesterday, so maybe he was busy with that. Also, we're getting close to Maintenance - earlier than normal for us Europeans, becuse DST changes aren't in step - and I wouldn't like to release when these message boards are down and we can't give technical support or receive reports of any problems. So it's looking like it'll be tomorrow before we do the full release. Sorry about that, folks. In the meantime, 'Release Candidate' builds are available to Lunatics Beta testers - same place as before. Last chance to break it before we go live... ;P Edit - I should have mentioned. I did get a reply from David Anderson - he still doesn't understand why we got zero credit in the first place (that shouldn't have happened). There may be some further breakages to fix before they're happy to see a mass influx of optimised apps. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14649 Credit: 200,643,578 RAC: 874 |
LadyL is offline just at the moment, but I'll try to live up to her high standards. You have NVidia GPUs. The installer doesn't have an NVidia AP app. The app is in beta and will be made availble separately (i.e. you'll have to do a manual install) Actually, neither of the applications uses the Open Graphics Library (OpenGL). They both use the Open Computing Language (OpenCL): and you are right, OpenCL is implemented on both the ATI and NVidia platforms. The actual scientific (crunching) programs are the same on both sets of hardware, but the management and control part of the program (running on the host CPU) has to be different. The current situation is that the OpenCL developer (Raistmer) believes that the NVidia version of the AP program is complete, but testers are reporting problems with the implementation. Because we've been concentrating on incorporating the late v6.01 modification into the tried-and-tested ATI implementation, the NVidia version remains in Beta testing while we catch our breath. 2nd questions :) Correct. Carry on using that until the dust settles on this rollout, and you are ready to choose which applications you wish to run. The updated application to run AP (the new version) on CPU is the only benefit you will find in the new installer. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
LOL, just around the time that the new science applications are being released, do I get one v5.05 AP. I haven't had one in over 6 months. Serious coincidence. ;-) |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14649 Credit: 200,643,578 RAC: 874 |
LOL, just around the time that the new science applications are being released, do I get one v5.05 AP. I haven't had one in over 6 months. Serious coincidence. ;-) Shhhh. Don't shout, and wave it about, or the rest will be wanting one too. (with apologies to the anonymous limerick writer) |
©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.