Posts by jason_gee


log in
1) Message boards : Number crunching : IntelĀ® iGPU AP bench test run (Message 1648489)
Posted 22 hours ago by Profile jason_gee
Hi Dirk,

I would recommend to take settings from your best 5 runs, then run each 30 times. After that you can calculate a variance for each setting. Once you know a variance for each, then it makes it easier to choose the best setting by probability.

Example:
Somesetting has best time of 900 seconds, average of 1000 seconds, variance of +/- 100 seconds.

Othersetting has best time of 850 second (better), average of 950 seconds, variance of +/- 150 seconds (worse).

Which is better ? Well you can calculate that with 1000's of samples, but a model with 30 runs would be enough to draw a curve of times. When you layer them all on the same scale, the one with highest probability density has the most area to the left, and would stand out (if your scale is fine enough).

5 second 'bins' is probably reasonable resolution, and you count the population of runs in each time bin, ending up with a graph of time bin (x axis) by numer of runs in that bin (y-axis)
2) Message boards : Number crunching : Problems with 26ja13ah tasks (Message 1648122)
Posted 2 days ago by Profile jason_gee
Hmm, I'm not aware of any issues running Kepler Class cards with Cuda 4.2 & recent drivers, though certainly something to keep an eye on with all the changes going on. Certainly Cuda 3.2 stock distribution has been halted to Maxwell and Kepler as a preemptive measure ( 4.2 & 5 running better on those anyway).

That particular wingman looks to be generating some invalids, and some valid work. Could be a simple matter of dust bunnies, unstable overclock, or power issue. Still pays to look for behavioural patterns across different hosts.
3) Message boards : Number crunching : Panic Mode On (95) Server Problems? (Message 1647610)
Posted 3 days ago by Profile jason_gee
...It would be wonderful if there was crosstalk between the two projects...
There's been a fair bit of that in bursts... Though mostly about Boinc design flaws and bugs.
4) Message boards : Number crunching : Linux static compiled x86_64 Boinc CLIENT? (Message 1647295)
Posted 3 days ago by Profile jason_gee
sweet, did you manage static ? or just lucky enough to have it work with existing shared libs ?

so far for x86_64 I have openssl, zlib and c-ares building static libs , still working on curl.

If you don't need a static client right now, I'll make do with the dynamic one myself ;)

[Edit:] ... and static curl library built, so that would just leave the boinc client itself , well maybe later
5) Message boards : Number crunching : Linux static compiled x86_64 Boinc CLIENT? (Message 1647193)
Posted 4 days ago by Profile jason_gee
Well the procedure to make a portable static client package, using the old debian virtualbox compatibility image, works for i686. ( no ssl, crypto or curl shared libraries turn up in ldd boinc ). It's an old version though 6.12.something, and I don't see an x86_64 vm image, so building with the configure option --build=x86_64-pc-linux-gnu fails, the image having no 64 bit libs (though i imagine the cross compile is possible if the libs were there)

Short of building a similar old debian VM from scratch, is there a way to put in the 64 bit libs (command line debian), or is the only option for a static build to replicate the old debian image in 64 bit ?

The compatibility vm image also has no Git installed ::S
6) Message boards : Number crunching : Windows new WDDM drivers (Message 1647142)
Posted 4 days ago by Profile jason_gee
Use nVidia ones please Jim, not Windows update ones.
7) Message boards : Number crunching : Linux static compiled x86_64 Boinc CLIENT? (Message 1647141)
Posted 4 days ago by Profile jason_gee
[Edit:] Along with Samsungs server drive manager, found some cool filesystem/SSD guide at https://wiki.archlinux.org/index.php/Solid_State_Drives , so plenty to read, lol.



Haha, lucky I wasn't too quick on the update. Turns out I must have arrived at the site sometime after some firmware updates were pulled due to bricking of drives. Mine's working fine, so I'll leave it until I read they fix that.
8) Message boards : Number crunching : Linux static compiled x86_64 Boinc CLIENT? (Message 1646938)
Posted 4 days ago by Profile jason_gee
Ugh, somehow they made the build system put in a squillion shared libraries even when setting things to be supposedly static, lol

I'll try their VM image in a day or so, Though I suspect multiple breakages there/

@Urs: yeah its a new Samsung 850 Pro. so it seems to not have any new firmware I can see yet, but will keep an eye out before I pile on too much

[Edit:] Along with Samsungs server drive manager, found some cool filesystem/SSD guide at https://wiki.archlinux.org/index.php/Solid_State_Drives , so plenty to read, lol.
9) Message boards : Number crunching : The Highest Ranked SETI AMD Host is a MAC: Time for a STOCK MAC APP? (Message 1646848)
Posted 5 days ago by Profile jason_gee
fftw location, IIRC is picked up during autosetup/configure, therefore you'd want to delete the old one, setup/make/install your dependancies boinc and fftw, and then autosetup/configure/make your application last. different order, or old file present, would probably pick up the repository one first. Probably we should stop putting binaries into repositories, but well it happens.
10) Message boards : Number crunching : Linux static compiled x86_64 Boinc CLIENT? (Message 1646747)
Posted 5 days ago by Profile jason_gee
Drive arrived safe and sound.

alright, path of least resistance for both Cuda and Boinc, seems to be Centos 7 (which I'll be happy with), and if compatibility issues arise there, there's the virtualbox VM option within that.

Out for planning, locating screwdrivers etc.
11) Message boards : SETI@home Staff Blog : What does loss of net neutrality mean for volunteer computing? (Message 1646620)
Posted 5 days ago by Profile jason_gee
Now I wonder if that means US taxpayers will also get the (never rolled out) internet infrastructure upgrades they already paid for since 1996.

Unfortunately I think that money went to pay bonuses for the CEO's for stealing it from the taxpayer.


Ah yes, a Bonus-worthy endeavour.
12) Message boards : Number crunching : The Highest Ranked SETI AMD Host is a MAC: Time for a STOCK MAC APP? (Message 1646616)
Posted 5 days ago by Profile jason_gee
Um, casting my mind quite far back now, and not checking the current status quo. As far as single precision floating point goes, SSE3 only consists of about 4-5 useful instructions (namely horizontal math ones), which the fftw library being mostly vertical math, would optimise out to SSE (SSE2 is double precision & some caching extensions).

So I'm pretty sure in fftw there are SSE routines, and AVX ones for single precision. [looking for SSE2 or 3 codelets in wisdom files would be a red herring]

For a self built library intended for your own use (as opposed to general), a useful key step would be to set the [compiler] architecture options. While this won't inject fftw codelets with SSE3 in the name, or many SSE3 instruction if any, it probably would inject appropriate prefetch and alignment (which can make a big difference), and do some minor chip family targeted optimisations.
13) Message boards : SETI@home Staff Blog : What does loss of net neutrality mean for volunteer computing? (Message 1646457)
Posted 5 days ago by Profile jason_gee
Now I wonder if that means US taxpayers will also get the (never rolled out) internet infrastructure upgrades they already paid for since 1996.
14) Message boards : Number crunching : TCP Settings (Message 1646388)
Posted 6 days ago by Profile jason_gee
I'll throw in a curly one for anyone with Atheros Chipset Wifi on their host (in my case a cheap PCI express card, under some generic brand name). If you use modified drivers for the Atheros chip available from http://forums.laptopvideo2go.com/forum/128-atheros-chipset/ , then it exposes some extra (network adapter) settings that solve particular network problems with those. (afaik, these should work with vista onwards etc, though not tried other than XP and win7 OS variants for mine)
15) Message boards : Number crunching : The Highest Ranked SETI AMD Host is a MAC: Time for a STOCK MAC APP? (Message 1646247)
Posted 6 days ago by Profile jason_gee
Well, it says SSE3 in the features, but I don't see SSE in the Wisdom file. CPU time is a little long.



haha!, well the fftw issue is probably just the particular libraries you built against (or the options that built them if built locally).

For CPU time, well welcome to the world of massive driver latencies and applications built to work on the smallest cards. The first of any of us to master that particular hurdle should get a medal, or a cigar or something.
16) Message boards : Number crunching : Linux static compiled x86_64 Boinc CLIENT? (Message 1646241)
Posted 6 days ago by Profile jason_gee
Alright, SSD for the Linux box arrives tomorrow. Which distro comes first is the question. I'll need to check the Cuda 7.0 support list and target that first most likely.

[Edit:]
Ahh, I see in yours they're removing the compromised OpenSSL versions (?) good move IMO, if that's the case.
17) Message boards : Number crunching : The Highest Ranked SETI AMD Host is a MAC: Time for a STOCK MAC APP? (Message 1646234)
Posted 6 days ago by Profile jason_gee
Ah yes, the old 'move key system stuff around with an OS revision' so as to 'enhance security' by making it hard for developers ;)
18) Message boards : Number crunching : @Pre-FERMI nVidia GPU users: Important warning (Message 1646232)
Posted 6 days ago by Profile jason_gee
Congrats on persisting through tough times, and coming through :D

It's no secret there are a lot of changes with long term support, coming from internal nVidia reasons, to global MS and Google decisions/changes. Keeping the widest range of backward support open is IMO a win for users, and the companies involved.

Now can you get Google to support OpenCL &/or Cuda on suitably equipped [stock] nexus devices ? ( :P, porting to renderscript is going to be a pain I could have done without )
19) Message boards : Number crunching : Linux static compiled x86_64 Boinc CLIENT? (Message 1645667)
Posted 7 days ago by Profile jason_gee
not on hand at the moment, though I am in process of rebuilding my Linux Box, and will need one in due course as well. If no-one's able to supply one before I get to it (1-2 weeks), then I'll probably be spitting out a slightly customised one.

Did you need a manager as well ? I suppose only the client and command line tool ?
20) Message boards : Number crunching : Suggestions On Recent Increase In Invalids & Validatrion Inconclusives? (Message 1645172)
Posted 9 days ago by Profile jason_gee
I've seen fan starting troubles, like those, happen with larger Noctua case fans when the connection to power isn't great, for example through a Molex adapter that was a little grungy or some such. That's in a situation when the Fan is fine, runs quietly once started etc.

Assuming the fans draw their power from the PCIe via some connector inside the card, I'd definitely be reseating the PSU connectors, and checking out under the shroud for a little connector to wiggle.


Next 20

Copyright © 2015 University of California