Posts by Raistmer


log in
1) Message boards : Number crunching : OpenCl 2.2 Provisional Specification (Message 1784692)
Posted 5 hours ago by Profile Raistmer
Immediately I see how this could help (if it is what its name implies): Program scope local memory
EDIT: all other is more for new code simplification and generalization. If app already written putting it from C to C++ adds not too much for performance.
2) Message boards : Number crunching : Panic Mode On (102) Server Problems? (Message 1784581)
Posted 17 hours ago by Profile Raistmer
GPU still dry. Only VLARs from the Russian billionaires project available. No non VLARs from Arecibo. I suspected it would go this way, when the Russian billionaire got involved (or should I say "bought" the project.)

Maybe Putin will own any eventual find of ET....

Yup, and all my builds have direct IP-connect to Kremlin to notify about EVERYTHING EVERYWHERE. You all pwned... ;)
3) Message boards : Number crunching : Help! No GPU work for Days (Message 1783498)
Posted 5 days ago by Profile Raistmer
Example of such modification based on NV SoG build:

<app_info> <app> <name>setiathome_v8</name> </app> <file_info> <name>MB8_win_x86_SSE3_OpenCL_NV_r3430_SoG.exe</name> <executable/> </file_info> <file_info> <name>libfftw3f-3-3-4_x86.dll</name> <executable/> </file_info> <file_info> <name>mb_cmdline_win_x86_SSE3_OpenCL_NV.txt</name> </file_info> <app_version> <app_name>setiathome_v8</app_name> <version_num>800</version_num> <platform>windows_intelx86</platform> <cmdline>-gpu_lock -total_GPU_instances_num 6 -instances_per_device 2</cmdline> <file_ref> <file_name>MB8_win_x86_SSE3_OpenCL_NV_r3430_SoG.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>libfftw3f-3-3-4_x86.dll</file_name> </file_ref> <file_ref> <file_name>mb_cmdline_win_x86_SSE3_OpenCL_NV.txt</file_name> <open_name>mb_cmdline.txt</open_name> </file_ref> </app_version> </app_info>


Cause BOINC will think it's CPU app app's own scheduling mechanism should be used instead.
so enable GPUlock and set how many tasks should be distributed to each GPU.

This particular example implies 3 GPU devices each of those should run 2 tasks at a time.
If different nums required edit -total_GPU_instances_num 6 -instances_per_device 2 string.
Also, tuning string can be added to mb_cmdline*.txt file (or inside <cmdline> tag).

This example implies there are enough CPUs to allow running of 6 tasks. If not either edit number of CPUs in cc_config.xml or set fraction usage of CPU in app_info.
4) Message boards : Number crunching : Help! No GPU work for Days (Message 1783274)
Posted 5 days ago by Profile Raistmer
If you completely out of work with powerful GPUs you could sacrifice CPU performance but let GPU run. Just install anonymous platform apps (via Lunatics installer for example) and then remove GPU coprocessor tags from CUDA app section. This will make BOINC to think that it's CPU app and schedule CPU-eligible work to it. Such way you can get VLARs on GPU.
But "real" CPU app will get nothing (cause there is no way to feed 2 different apps under same plan class AFAIK) and its section should be removed to not to intercept tasks from GPU app.
So, CPU will sit idle (can be used in another project) and GPU will work on VLARs.
5) Message boards : Number crunching : HSA, Boltzman, HIP and other AMD stuff... (Message 1782876)
Posted 7 days ago by Profile Raistmer
It seems there is possibility to port CUDA-based NV apps to AMD hardware w/o re-writing them.

http://gpuopen.com/tag/hip/

Also it seems it's mostly Linux's part of compute world.
Anyone had some experience with all this stuff?
6) Message boards : Number crunching : Panic Mode On (102) Server Problems? (Message 1782709)
Posted 7 days ago by Profile Raistmer
@Tbar, To my knowledge (unless something changed), the AP OpenCL App does not use the NVAPI detection as does Seti Cuda MB, GPUz, and Precision-X, but instead standard figures reported before the application even initialises the device, so it's not a measurement.

It reports what runtime returns so it's measurement in the same sense as quering freq via CUDA runtime.
On NV it's inevitable to go through CUDA driver API with OpenCL runtime frontend cause NV would't invent same thing twice.
7) Message boards : Number crunching : GBT MESSIER031 work on GPU NVIDIA (Message 1782003)
Posted 10 days ago by Profile Raistmer
Except for the possible overflows, are triplets and their respective scores perhaps a better indicator than a similar gaussian score?

Indicator of what?
8) Message boards : Number crunching : Cuda 50 V8 Weirdness (Message 1781988)
Posted 10 days ago by Profile Raistmer
So what magic number do you put on how long the valid result must take to shut down ?


Well, I've had some thoughts about that every now and then, not really worked it through though. Instead of a fixed magic number the client could see if the app has been given a chance to exit. Take a snapshot of the app's CPU time when finish file is first seen and wait until the app's CPU time is snapshot+10.

I had one Rosetta CPU tasks trashed by finish file too present long stuff. The CPU may have been busy with something else at that time, or the disk may have been busy, or the machine may have been swapping heavily. Or maybe even all three at the same time. A simple CPU time snapshot wouldn't catch too busy disk. To solve that it would need... umm, something.

At which point the word over-engineering comes to mind. There's already the maximum time check. On one hand it makes me feel like it would be a waste of resources to wait for maximum time instead of waiting just ten seconds for the app to exit. On the other hand, just how much resources have been wasted for the "finish file present too long"?

So I don't know. Maybe getting rid of the check is the right way.


Good points rised indeed... but IMHO in wrong place. Until someone (like Richard) will deliberately place lot of efforts to bring this ideas to BOINC devs they will be "lost in the noise" here. BOINC devs don't read these boards it seems at all. The place with better (only better) chances to get heard are the BOINC's dev/project mail-lists.

More on topic - I would agree with definitions of what it's all about and priority of science result preservation over "programmatically correct exit". Seems BOINC's part of code changing is required though.

Regarding additional timer of exit - what if to restart from last checkpoint in case of BOINC-perceived failure instead of bring computation error? That will attempt to save most of time spent for computation still not awaiting too long to realize there are some difficulties on exit (as would be in case of awaiting full time reserved for task processing).

And awaiting for CPU progress instead of elapsed time can have some negative effect with GPU app. Different runtimes react differently on failure (like driver restart). For OpenCL apps I saw at least 3 reactions:
1) return failure code to API call - most nice one cause allow app to know about issue in some way.
2) doesn't return from API call, 100% core consumption - here CPU time check can see progress.
3) doesn't return from API call, zero CPU consumption - here app will never progress on CPU time until external process termination.
9) Message boards : Number crunching : GBT MESSIER031 work on GPU NVIDIA (Message 1781985)
Posted 10 days ago by Profile Raistmer
The interesting thing of the overflows is that the result are not simple spikes but triplets.

Yeah, high triplets count, but also spikes in the same WU's.


I have a relatively high number of inconclusives on these tasks. All my inconclusive GPU tasks overflow with 30 triplets, but the wingmen for these tasks (CPU) report a high number of triplets (typically 25 or 26) and also a number of spikes.
Anyone else notice this ?

Tom


Most probably it's the result of current validator weakness in overflowed tasks comparison. It relates to only CPU-like sequential style of task processing. But overflows report only part of all signals inside data, so, depend from processing order. Seems CUDA build affected too (and I expect even greater effect on SoG builds).
With Arecibo data where overflows were early ones and mostly on spikes there were attempts to simulate serial order of execution on early processing stages, where overflow most probably. This apparently doesn't help a lot with new types of overflows we seeing with GBT data.

Until validator will account for different processing order possibility such inconclusives (and some of them will even turn into invalids later though relatively small part) are unavoidable.
10) Message boards : Number crunching : amdkmdap stopped responding for tasks blc0_2bit_guppi_*_MESSIER031_* (Message 1781981)
Posted 10 days ago by Profile Raistmer
Hello.
At some point of processing these blc0 tasks my display driver crushes with message like 'display driver amd stopped responding and has recovered'. This happens only for long (90 mins or so) tasks at about 15.16% of progress. Counter of progress then stops, CPUID Monitor shows that video card is loaded at 99% (~90% is usual load).
It did not happen for non-blc0 tasks.
Could you please help with that? As I understand, I can't disable getting these blc0 tasks.

Попробуйте добавить -period_iterations_num 500 в файлы (во все) с именем mb_cmdline*.txt
11) Message boards : Number crunching : V8 CUDA for Linux? (Message 1780841)
Posted 13 days ago by Profile Raistmer
Wow :)
That's Linux...
12) Message boards : Number crunching : Can Seti@home keep up with Breakthrough Listen data? (Message 1780839)
Posted 13 days ago by Profile Raistmer
Exactly
13) Message boards : Number crunching : Can Seti@home keep up with Breakthrough Listen data? (Message 1780779)
Posted 13 days ago by Profile Raistmer
Original recordings should be kept even after processing.
For few different reasons.
14) Message boards : Number crunching : Root access to router's OS (Message 1780653)
Posted 14 days ago by Profile Raistmer
It seems many nowadays home routers use some sort of Linux as their OS.
What if there would be possible to get root access to it(at least by re-flashing with specially crafted image) and add BOINC to its applications.
Usually there is 1 such router per flat.
What CPUs there? How they compare with ARMv7 ones? Will such router compute better or worse than cheap smartphone?...

Lets discuss.
15) Message boards : Number crunching : Lunatics Help (Message 1780651)
Posted 14 days ago by Profile Raistmer

that looks like you found a place that prints it out. where's that?
much easier than staring at client_state.xml :D

Yep, it's in host properties web page.
Example:
http://setiathome.berkeley.edu/show_host_detail.php?hostid=4517643
16) Message boards : Number crunching : Lunatics Help (Message 1780644)
Posted 14 days ago by Profile Raistmer

sure. but boinc has mechnisms for that. it's does keep track of uptime and of active time. and that in turn is used to calculated remaining cache.
it _should_ also be taken in account when fetching work - it certainly is when the work arrives.

if not carrying a large cache but processing it at snail's pace could certinly lead to grabbing large amounts of work that don't get done in time.


Like
Fraction of time BOINC is running 76.80%
While BOINC is running, fraction of time computer has an Internet connection 99.08%
While BOINC is running, fraction of time computing is allowed 99.70%
While is BOINC running, fraction of time GPU computing is allowed 99.65%
17) Message boards : Number crunching : Cuda 50 V8 Weirdness (Message 1780640)
Posted 14 days ago by Profile Raistmer
Perhaps, it's DataStore.edb issue: http://superuser.com/questions/539411/pc-boots-then-writes-giant-datastore-edb-file-slowing-the-computer-down

EDIT: in particular:
RAMMap showed that C:\Windows\SoftwareDistribution\DataStore\DataStore.edb mapped file was wasting 400 Mb of my precious RAM. Moreover Windows Update was constantly slowing down the system in process of updating that huge file.
that partially explains huge amount of going to nowhere RAM on modern systems. I have Vista x86 with 3GB RAM but it's in constantly swapping mode almost always these days... Seems all abandoned any care about how much RAM their tools take...

And how to get rid of it at least partially:

In the C:\Windows\SoftwareDistribution\Download directory, you can find 40-char length files name that correspond to the 40 final chars of IUpdateDownloadContent::DownloadUrl property.
For a complete cleaning (clearing also the whole update history):
net stop wuauserv
net stop bits
delete all files inside the C:\Windows\SoftwareDistribution\Download directory
delete the DataStore.edb into C:\Windows\SoftwareDistribution\DataStore
net start bits
net start wuauserv
and, take a look to my free utility to manage your windows updates: www.portableupdate.com
Regards,
gigirex
18) Message boards : Number crunching : New app? (Message 1780061)
Posted 16 days ago by Profile Raistmer

and also that there are templates to work from where can these be found??

What templates, for what task? If you mean template for app_info it's included into binary pack with app binary itself.


also if I start a thread which forum should I start it in I do want to learn and if I can help other newcomers to lunatics as well that will be even better. Thanks for any help you can give.

Number crunching.

And thread about Lunatics installer:
http://setiathome.berkeley.edu/forum_thread.php?id=78878
19) Message boards : Number crunching : New Greenbank Files (Message 1779930)
Posted 16 days ago by Profile Raistmer
Maybe GBT has some different source of artifical signal noise than Arecibo. To have overflowed Triplets and Pulses w/o Spikes is very unusual indeed.
Or, maybe we catch some really strong ET signal here.... fingers crossed :)
20) Message boards : Number crunching : OpenCL NV MultiBeam v8 SoG edition for Windows (Message 1779872)
Posted 17 days ago by Profile Raistmer
Raistmer,

I've created a new thread on the beta site in the Seti@home Enhanced section so that I don't congest this thread.

Here is the link and there are images and links to stderrs for the work in those images.

I probably explained it wrong but look at these and let me know

https://setiweb.ssl.berkeley.edu/beta//forum_thread.php?id=2306


Thanks.

I gave detailed answer in that thread.


Next 20

Copyright © 2016 University of California