Posts by Raistmer

1) Message boards : Number crunching : Multi-core AMD stories (aka: Ryzen 7, Threadripper and Threadripper2) (Message 1961525)
Posted 1 day ago by Profile Raistmer
Post:

I just benchmarked it on my WIN10 1950x machine that runs at 3.6GHz. It has 128GB of 2400 speed ECC udimm.

WU : blc04_2bit_blc04_guppi_57898_17662_DIAG_KIC8462852_OFF_0020.12914.818.18.27.159.vlar.WU
MB8_win_x64_AVX_VS2010_r3330.exe -verb -nog :
  Elapsed 2638.336 secs
      CPU 2633.938 secs
MB8_win_x64_AVX_VS2017_r3714_Vlock.exe   :
  Elapsed 2472.117 secs, speedup: 6.30%  ratio: 1.07x
      CPU 2467.734 secs, speedup: 6.31%  ratio: 1.07x

It looks good. I will release it on my WIN10 Machine. Do you know if Raistmer updated the optimized branch of the source code? I would like to try to compile on Linux again. I have been successful in compiling it on 18.04, but the executable finds no signal. If the optimized code is updated, I would like to try it again.

It's Windows-targeted update.
That's, to be useful for Linux, Linux-oriented API calls should be made instead.
App uses locked memory to circumvent possible suboptimal Windows choices for paging. Same as opt AstroPulse does.

Also I attempted to use LargePages (in another build) to save TLB cache entries and improve memory cache hits, but it seems too much efforts from user are needed to run it. It requires special user right and default UAC admin can't use such right. That is, one need to create non-administrator account and directly assing corresponding right to it.
Another way is to hack UAC management. In any case, user's action required.
Vlock can be used under usual user rights.
2) Message boards : Number crunching : Anyone have any SuperMicro Boards? China hardware hack (Message 1959714)
Posted 12 days ago by Profile Raistmer
Post:
ROFL, fake news - very modern approach to advertising :)
3) Message boards : Number crunching : Anyone have any SuperMicro Boards? China hardware hack (Message 1959308)
Posted 15 days ago by Profile Raistmer
Post:
I'd say what you give is what you get in return. Nothing new. Absolutely same measures required when one deals with "good old" non-outsourced Americans firms and their hardware/software .
Yes, and you're a Russian, anything you say has to be not believed or let's call out the Russian Ambassador on that. :-)

Yep, didn't you know your all PCs, especially ones with GPUs pwned already? ;) :D
4) Message boards : Number crunching : Anyone have any SuperMicro Boards? China hardware hack (Message 1959307)
Posted 15 days ago by Profile Raistmer
Post:
So, still only driver/ runtime could be responsible....
Hard to believe you keep saying that when the problem doesn't exist when a US marketed card is used. Obviously the Fake Chinese 970 is responsible. This is a Fake 970, to fool the BIOS the card uses a Boot ROM prior to entering BIOS. If you watch the screen closely you can see the card info appear on screen just before the machine enters BIOS. This is how they can fool the machine/OS into reporting a 550Ti is really a 970. Remove the Chinese Fakery and the problem goes away.

So you suppose GPU BIOS is able to redirect text output to file? Hm... I would like to look at such card, really. Sounds like this masterpiece would be much more costly that usual 1080Ti perhaps :)
5) Message boards : Number crunching : Anyone have any SuperMicro Boards? China hardware hack (Message 1959278)
Posted 15 days ago by Profile Raistmer
Post:
The only question is how the Chinese card trigger the App to write stderrs in Chinese.

Yep, very hard question :)
And if app is Lunatics based I'm sure it has no any localization strings beside English at all. So, still only driver/ runtime could be responsible.
Trigger factor could be ID string in ROM. Hardly believable though driver/runtime selects language based on card ID.
6) Message boards : Number crunching : Anyone have any SuperMicro Boards? China hardware hack (Message 1959233)
Posted 15 days ago by Profile Raistmer
Post:

So watch out what you order on the internet. Or have it vetted. Or don't put it on your network without some air gap or sandboxing.


I'd say what you give is what you get in return. Nothing new. Absolutely same measures required when one deals with "good old" non-outsourced Americans firms and their hardware/software .
History of different hardware bookmarks in Intel's production lasts decades. "Ooops, just bug-just bug" every time when disclosed :)
Not to say about last Google/M$ trends that got in user's underwear already. Re-phrase Twain, rumors of NK danger are greatly exaggerated.
7) Message boards : Number crunching : Petition: Metal API support in GPU Mac (Apple) Tasks (Message 1959221)
Posted 15 days ago by Profile Raistmer
Post:
From inception Apple chosed closed-source way. There are consequencies of such decision. Users of Apple should be very aware of that. Why this company still has such big support eludes me.
From one side to make special app for Apple is good short-term decision for SETI, from another, it's just one more step on the road of coercion this company goes from beginning...
8) Message boards : Number crunching : GTX1080 - Low usage memory controller load? (Message 1959219)
Posted 15 days ago by Profile Raistmer
Post:
Biggest possible GPU usage load with smallest possible GPU memory controller load is the key to good processing performance. So nothing wrong in low memory controller % as long as GPU load stays ~100%.
(of course it's the rule with own exceptions but to get into those exceptions app code changes are required).
9) Message boards : Number crunching : Errors on android, all tasks (Message 1959217)
Posted 15 days ago by Profile Raistmer
Post:
just as it now doesn't allow for using all foreground and background CPUs at the same time, without especially disabling the code that sets that.

So, finally, number of CPUs in device became same advertizing hype as number of megapixels on shit-optics cameras or number of gigabytes of device memoty already filled with preinstalled crap.
Free market, nothing honest, only money.

Modern Android versions seems to actively resist to computaional load. On both my Digma phones (Android 6) I experience silent BOINC shut downs. They happen w/o any other phone activity.
Device just cools down so I understand that computations stopped again. Sometime it can continue for hours or even days, sometimes it stops very shortly after relaunching.
On relaunch BOINC says that "another instance detected" and refuses to start. Second attempt usually results in operational (for some time) BOINC client with working scientific (SETI) apps.
10) Message boards : Number crunching : Anyone have any SuperMicro Boards? China hardware hack (Message 1959216)
Posted 15 days ago by Profile Raistmer
Post:
I've got a Chinese story of my own. It involves those Chinese Video cards selling on eBay. I ended up with one and decided it might be useful to just run the Display on My development machine. As long as you don't try to use it for compute, it seems to run the monitor fine in Linux. Most of My cards are in the Mining machine so I was short a card or two. All seemed well, the Apps passed the Benchmark App so I placed it on the Miner. Then I noticed the Inconclusives. Then I checked the results...most of the Stderr output was missing. Then I checked the Stderr output in the slots as it was being written. The Stderr output in the Slots were all in Chinese. You can't make this up. The client_state apparently doesn't do Chinese so the results were just missing in the Client_state file. The files in the Slots were full of Chinese though, even though the App wasn't being run on a Chinese card. Weird. So, I replaced the Chinese card, compiled another App, and all is well, No more Chinese in the Stderr output.
Be careful with your Chinese video cards...

Did you try just to install generic driver instead one provided by supplier?
AFAIK GPU can't write into stderr on its own. So, it's driver API messages from localized driver. The single place that could give chinese symbols directly from hardware is device model name stored in it's ROM.
Such weakness of BOINC XML parser is sad thing. What if app's stderr will contain let say cyrillic, will it break too?...
11) Message boards : Number crunching : Can the Movidius Neural Compute Stick be used as a GPU for Processing Seti etc. data (Message 1954802)
Posted 11 Sep 2018 by Profile Raistmer
Post:
Then its a case of not possible at the moment, but possibly a useful tool in the future?

Michael B.

Well, if neuronal net would be able to learn how to produce valid results for our SETI algorithm it would be remarkable result itself, IMHO. Another step to AI :)
12) Message boards : Number crunching : Arecibo Overflow Tasks Run With CUDA Special App Sometimes Filled With Triplets Only (Message 1954801)
Posted 11 Sep 2018 by Profile Raistmer
Post:
If you want me to do offline task modifications please save task on some storage, to retrieve it later.
13) Message boards : Number crunching : Arecibo Overflow Tasks Run With CUDA Special App Sometimes Filled With Triplets Only (Message 1954010)
Posted 6 Sep 2018 by Profile Raistmer
Post:

For some reason the Only App reporting Arecibo Overflows filled with Triplets is the CUDA App, Why?
There Should be at least Some Arecibo Overflows filled with Triplets, Why aren't the Other Apps reporting any?
[/pre]

Such statement could imply that reported triplets are invalid ones, not just another subset of all task signals.
What I would propose to check this possibility is to change thresholds and limits in task under question.
This would allow:
1) increasing pulse threshold will effectively remove pulse signals from output so if CPU (for example) app overflows on Pulses while CUDA - on Tripet such modification would allow CPU app to continue search and maybe to find those Triplets too.
2) increasing limits for reporting signals (better to test with CPU apps cause GPU apps could be hardwired to particular number due to GPU memory restrictions and peculierities of design) will allow CPU apps to prolong processign and report both subsets of signals original CPU and GPU runs did. But if there are no same Triplets in whole output of modded task and CPU app as were in original task and GPU app - then severity of issue increases indeed - that would mean not just another subset of valid signals but invalid signals instead.
So the check is possible, just testing task modification required.
14) Message boards : Number crunching : Arecibo Overflow Tasks Run With CUDA Special App Sometimes Filled With Triplets Only (Message 1954008)
Posted 6 Sep 2018 by Profile Raistmer
Post:
One of possible ways to fix this issue is to store more signals of each type than limit says separately.
And do CPU-based reordering and selection from these signal pools on final report.
This would allow fully parallell processing on GPU side still improving validation rate.
That's how OpenCL SoG operates currently. Of course same issue there too - on overflow too many signals exist to be reported as whole so some selection required.
15) Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users (Message 1951838)
Posted 24 Aug 2018 by Profile Raistmer
Post:
I found newly prepared flash with Ubuntu booting on one PC but not on another. No bootable OS found.
Doing some testing I reproduced issue on original PC too. Flash boots when being chosen via EFI but fails then chosen just as USB device (BIOS).
On installation time I said ~ "install grub on "drive" (not in "partition") and make it EFI/BIOS aware".
So expected to be able to boot via both routes.
How could I reinstall grub (Linux bootloader, right?) w/o reinstalling whole OS and how to make sure it will boot in both EFI and BIOS cases?

Not sure what you are trying to do but I hope I phrased my search correctly. I think you want the grub bootloader on the drive and not in the EFI partition?
How can I reinstall GRUB to the EFI partition?

Thanks,I try it.
Not sure it will allow to boot in both modes though.
The idea is to be able to boot both EFI and non-EFI ways.
16) Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users (Message 1951669)
Posted 23 Aug 2018 by Profile Raistmer
Post:
I found newly prepared flash with Ubuntu booting on one PC but not on another. No bootable OS found.
Doing some testing I reproduced issue on original PC too. Flash boots when being chosen via EFI but fails then chosen just as USB device (BIOS).
On installation time I said ~ "install grub on "drive" (not in "partition") and make it EFI/BIOS aware".
So expected to be able to boot via both routes.
How could I reinstall grub (Linux bootloader, right?) w/o reinstalling whole OS and how to make sure it will boot in both EFI and BIOS cases?
17) Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users (Message 1950872)
Posted 20 Aug 2018 by Profile Raistmer
Post:
Even didn't finish logon process yet but already got " Failed to update" message.
Looks like even didn't changed OS at all, same nasty things as in Windows here.

Anyway, I don't need to like Ubuntu, just use it on purpose... And purpose was to create Portable Linux bootable flash to test Linux-based software.
Will check if it can boot on PC with 1050Ti being created on netbook...
18) Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users (Message 1950801)
Posted 19 Aug 2018 by Profile Raistmer
Post:
Live booted Ubuntu doesn't have MidnightCommander. What happened with the world %)
EDIT: I observe very strange installation procedure.
Minimal install was chosen. So it installed... seems ALL. And now it REMOVES stuff. For example, it removed GNOME docs and now preparing to remove Majong game. Is it not common logic to NOT install everything on MINIMAL install? Even Windows at least pretends not to install smth that didn't selected for install...
EDIT2: Remmina (RDP client) asks to send usage statistics.. Is it really *nix OS or M$ won everywhere ???
19) Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users (Message 1950261)
Posted 17 Aug 2018 by Profile Raistmer
Post:
Will be possible to install on the same drive from that booted? Just in another partition?

Yes, if there is enough room. I gather you are trying to install to a USB stick? The Live USB takes about 1.9GB. So if you have say a 32GB stick then you could partition a 2GB partition and a 30GB partition and boot from the first partition say and install to the second partition. Should work. USB running at 3.0 or 3.1 speeds is still slower than normal hard drives or SSD speeds. Doable but slow.

I used (as recommended in Ubuntu tutorial) Rufus to flash ISO into USB stick.
Unfortunately it recreates partition and allocates full drive to that partition. So, before OS installation I need to shrink bootable partition with ISO itself.
Currently I decided to go another route - to boot ISO into VirtualBox having USB stick attached to it.
I hope this will make possible to install OS on USB stick w/o using same stick for booting ISO image.
Work in progress cause on my netbook booting Ubuntu under VM isn't speedy process :)
EDIT: it sits here https://clip2net.com/s/3W7nyTe more than half hour already :/
CD icon flashing so I hope it didn''t hang, just veeeery slow....
20) Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users (Message 1949967)
Posted 15 Aug 2018 by Profile Raistmer
Post:
Will be possible to install on the same drive from that booted? Just in another partition?


Next 20


 
©2018 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.