WU types


log in

Advanced search

Message boards : Number crunching : WU types

1 · 2 · Next
Author Message
Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1192372 - Posted: 8 Feb 2012, 7:28:42 UTC

I know this is silly and I have been crunching since 1999 but what are the different work units type and what should they be crunched with my 17 920 CPU or my 2 galaxy GTX 460 @768 memory GPU. I read about vlar and shorties and astropulse are there more?
____________

LadyL
Volunteer tester
Avatar
Send message
Joined: 14 Sep 11
Posts: 1679
Credit: 5,230,097
RAC: 0
Message 1192417 - Posted: 8 Feb 2012, 10:51:29 UTC

There are two types of applications: Multibeam (MB) and Astropulse (AP).

So there are basically MB and AP workunits.

MB workunits have a property called the angle range (AR) that vastly influences processing. VLAR are 'very low angle range' - those units are atm not going to NVidia GPUs because they were found to process very badly there, often causing huge (screen) lags, making some systems impossible to use or even crashing them. 'shorties' VHAR are on the other end of the spectrum - very high angle range. They process very fast and are issued with very short deadlines accounting for that fact.

Running stock you could crunch MB on the GPUS and MB+AP on the CPUs.
Running optimised (Lunatics) the same - there is a beta app to run AP on the NVidia GPU via OpenCL, but it has an issue with newer drivers.

As you are already using optimised apps you'd only want to look into why you don't seem to be getting any GPU tasks.

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1192438 - Posted: 8 Feb 2012, 12:14:54 UTC - in response to Message 1192417.

Thank you
Very informative I am now schooled.
In November my power supply burned up and I was running my 2 video cards and doing AP and MB on both video cards as well as my i7 920 CPU and had a rack of 35000. So I had to fall back on an older CPU that at one time had a q6600 but the CPU went bad and I put a celeron in it until I built my i7 machine. I rma the psu until then I started to use the celeron which had a low rac. I replaced the celeron with a q8400 and it had a rac of 3000 it went up a little when I used the 210 video card with a 450 psu which could not run my 2 460 cards. Cooler master replaced my power supply. I just started to run my i7 again and I wanted to see what the rac would be without the video cards. Then I will start with the video cards. So I have read about the NV cards and the vlar wu and shorties and ghost wu and people using rescheduler etc.

So what are ghost wu?

____________

LadyL
Volunteer tester
Avatar
Send message
Joined: 14 Sep 11
Posts: 1679
Credit: 5,230,097
RAC: 0
Message 1192453 - Posted: 8 Feb 2012, 12:51:35 UTC - in response to Message 1192438.

So what are ghost wu?


Ghost WU are something that mostly happens when the bandwidth is maxxed out.

Your host asks for work. The scheduler assigns work (it shows up on your web tasklist) but that reply gets lost. So your host never actually recieves the order to download and crunch those tasks. They are 'ghosts'.

If 'resent lost results' is enabled, each time your host asks for work it sends along a list of what it has and the scheduler checks that against what you should have. If some are missing on your host, it reissues them to you. A side effect is that work may show up as 'timed out' with impossibly short deadlines. Those are in fact cancelled units. Your host asked for work for CPU and got assigned some VLAR. it doesn't receive them. It then asks for GPU work. The scheduler notices that you should have those units. At the same time those units have been deemed unsuitable for the NVidia GPU. Instead of checking whether you are asking for CPU work as well and directing them there, those units are timed out. (and resent to a different host at a later time).

msattlerProject donor
Volunteer tester
Avatar
Send message
Joined: 9 Jul 00
Posts: 38674
Credit: 572,959,641
RAC: 554,815
United States
Message 1192456 - Posted: 8 Feb 2012, 12:57:46 UTC - in response to Message 1192438.
Last modified: 8 Feb 2012, 12:58:56 UTC

Ghost WUs are WUs that are assigned to your computer by the Seti scheduler but not downloaded by your computer.

Your computer makes a work request, the scheduler issues WUs to it. But for some reason, comms fail at that point, and your computer does not successfully download them. So, the Seti servers show that your computer has that work, but your computer does not have it there to crunch.

Right now, there should not be many ghosties, because the 'resend' feature is turned on (thankfully). This feature compares the WU list your computer sends to Seti when requesting work, and if WUs are in the list for that computer on the Seti servers, but your computer does not include them in the work request list, they should be sent to you again. You may see a response to a work request in the Boinc messages saying 'resending lost tasks'. If the resend feature was not enabled, those would have become ghosts, and new work would have been issued instead.

EDIT....Oops, the Lady beat me...
____________
*********************************************
Embrace your inner kitty...ya know ya wanna!

I have met a few friends in my life.
Most were cats.

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1192464 - Posted: 8 Feb 2012, 13:27:34 UTC

Thanks to LadyL and Mark.

From another post "memory timmings" in which Jason talks about using different applications such sse4.1 for certain wu and ssse3x for certain wu so that got me to wondering about the different wu.

Your replies helped me understand what he was talking about.

See below.


Yeah there's a few things going on that make the best app different depending on the setup & dominant work mix. Firstly with respect to latencies, VHAR are most prone to cache thrashing, and there are different strides used to trigger the hardware prefetchers depending on the build. In the case of the SSE4.1 build, those strides are longer & so initiate fewer transactions, effectively reducing bus contention, though can be prone to starve on a lower memory subsystem performance setup, where the 'Xeon half stride' feature pulls ahead, in SSSE3x builds.

Another consideration is with the SSSE3x & SSE4.1 builds is that a major feature of those was Alex's SSSE3 hand pulsefinding optimisations, intended to deal with misaligned loads & other issues. As Intel's transistor budget rises year to year, it appears they are continuing to reduce the sensitivity to misaligned accesses, and so in some cases SSE3 code can be a better choice due to its inherent lower instruction count & complexity.

As it stands, one of my long term goals is to integrate many of the builds to a single build per platform, which will use the best code path available for hot functions on a given chip. The way things are going, this may end up being controlled via a static dispatch mechanism controlled by an external bench utility & advanced user configurable parameters. The exact form this should take isn't completely clear yet, though certainly there are indications that some sub-functions involved at different weights by angle range do perform differently depending on the task angle range content (FFT lengths etc), so it will likely end up more complex than 'use SSE3 for this...' etc.

Jason

____________

Profile Michel448a
Volunteer tester
Avatar
Send message
Joined: 27 Oct 00
Posts: 1201
Credit: 2,891,635
RAC: 0
Canada
Message 1192466 - Posted: 8 Feb 2012, 13:31:30 UTC - in response to Message 1192456.
Last modified: 8 Feb 2012, 13:33:17 UTC

wow sweet awesome

I almost have 1 million of credits and I'm just starting to understand what I've done and I've made so far since 3-4 months :P

Can I ask more explanations ?

about : AR (Angle range) along with VLAR (Very Low Angle range) - VHAR (Very High Angle range).

My misunderstanding comes from when i try to figure our what is an "Angle range", the picture that comes in my mind is our eye with a telescope versus the horizon (Low) or directly over our head (High). => which is totally a wrong picture since we are talking about frequencies.

What is an "Angle range" inside a "frequency" ?
____________

Profile ignorance is no excuse
Avatar
Send message
Joined: 4 Oct 00
Posts: 9529
Credit: 44,433,274
RAC: 0
Korea, North
Message 1192469 - Posted: 8 Feb 2012, 13:39:10 UTC - in response to Message 1192466.

does anyone else here seem amused that the astropulse WU's are designated as ap.XXX...wu why does an astropulse wu need the wu on the end. Is that an artifact from back when it was being instituted?
____________
In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

End terrorism by building a school

ClaggyProject donor
Volunteer tester
Send message
Joined: 5 Jul 99
Posts: 4058
Credit: 32,799,910
RAC: 5,052
United Kingdom
Message 1192477 - Posted: 8 Feb 2012, 13:45:51 UTC - in response to Message 1192466.

wow sweet awesome

I almost have 1 million of credits and I'm just starting to understand what I've done and I've made so far since 3-4 months :P

Can I ask more explanations ?

about : AR (Angle range) along with VLAR (Very Low Angle range) - VHAR (Very High Angle range).

My misunderstanding comes from when i try to figure our what is an "Angle range", the picture that comes in my mind is our eye with a telescope versus the horizon (Low) or directly over our head (High). => which is totally a wrong picture since we are talking about frequencies.

What is an "Angle range" inside a "frequency" ?

Angle Range is what the Telescope is doing, A VLAR is when the telescope is fixed to a single point in Space, Note the Earth is Rotating, so the telescope is moving to track that single point in space,
a Normal AR task is where the telescope is stationary, as the Earth rotates the telescope tracks across the sky, a Very High AR task is where the telecope is being motored across the sky,

Claggy

Profile Michel448a
Volunteer tester
Avatar
Send message
Joined: 27 Oct 00
Posts: 1201
Credit: 2,891,635
RAC: 0
Canada
Message 1192492 - Posted: 8 Feb 2012, 15:02:47 UTC - in response to Message 1192477.
Last modified: 8 Feb 2012, 15:12:44 UTC


Angle Range is what the Telescope is doing, A VLAR is when the telescope is fixed to a single point in Space, Note the Earth is Rotating, so the telescope is moving to track that single point in space,
a Normal AR task is where the telescope is stationary, as the Earth rotates the telescope tracks across the sky, a Very High AR task is where the telecope is being motored across the sky,

Claggy



SO,
Normal AR = The telescope which at the radar is linked doesnt move but the earth is rotating.

Low AR = The telescope which at the radar is linked move to follow the rotation of Earth , trying to always point the same spot in the sky.

High AR = The telescope which at the radar is linked move toward whatever the direction they want/need , which is very different of the earth rotation.

right ?

now i need to figure out why it has an impact on the frequencies :) :) :)


PS EDIT : if i say '' the telescope which at (or at which - sorry my english /cry) the radar is linked'' it is not a radar that record the frequencies ? and not a telescope ? because a telescope = it's for pictures , and what we are recording are sounds - frequencies ? so : a radar. no ?
____________

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1192493 - Posted: 8 Feb 2012, 15:15:04 UTC
Last modified: 8 Feb 2012, 15:16:09 UTC

I know when I have a vlar wu just by looking at my task but how can I tell if I get a vhar wu or a ghost wu.
____________

Profile SciManStevProject donor
Volunteer tester
Avatar
Send message
Joined: 20 Jun 99
Posts: 4817
Credit: 80,547,203
RAC: 34,224
United States
Message 1192495 - Posted: 8 Feb 2012, 15:23:22 UTC - in response to Message 1192492.



PS EDIT : if i say '' the telescope which at (or at which - sorry my english /cry) the radar is linked'' it is not a radar that record the frequencies ? and not a telescope ? because a telescope = it's for pictures , and what we are recording are sounds - frequencies ? so : a radar. no ?

We are using radio waves. A radio telescope (Arecibo in Puerto Rico) only collects radio waves. That is what we are analyzing with our computers.

Steve
____________
Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 3993
Credit: 109,970,575
RAC: 136,483
United States
Message 1192496 - Posted: 8 Feb 2012, 15:26:08 UTC - in response to Message 1192493.
Last modified: 8 Feb 2012, 15:30:06 UTC

I know when I have a vlar wu just by looking at my task but how can I tell if I get a vhar wu or a ghost wu.

One way that you can find the angle range is by open the work unit in something such as notepad. Where you will find <true_angle_range> followed by the value such as <true_angle_range>0.37454123643124</true_angle_range>.

EDIT: As I use Fred's reschedule program I can see the number of each kind of tasks that my machine has. You could use it just for that without using any of the reschedule functions if you wanted.

To tell if you have "ghost" tasks. You can check the count of the tasks on your machine vs. the number the on your machines tasks pages on the website. If the in progress number is larger than the count on your machine then you have "ghost" tasks. When the servers are enabled to send out lost tasks you don't have to worry about ghosts. At the moment this is enabled.
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

Profile Michel448a
Volunteer tester
Avatar
Send message
Joined: 27 Oct 00
Posts: 1201
Credit: 2,891,635
RAC: 0
Canada
Message 1192498 - Posted: 8 Feb 2012, 15:34:43 UTC - in response to Message 1192495.
Last modified: 8 Feb 2012, 15:35:59 UTC



PS EDIT : if i say '' the telescope which at (or at which - sorry my english /cry) the radar is linked'' it is not a radar that record the frequencies ? and not a telescope ? because a telescope = it's for pictures , and what we are recording are sounds - frequencies ? so : a radar. no ?

We are using radio waves. A radio telescope (Arecibo in Puerto Rico) only collects radio waves. That is what we are analyzing with our computers.

Steve


ya that s what i was presently figuring out hehe

by "telescope"" he meant "radio-telescope" in fact ^^

thx for the precision, Steve :))
____________

Profile Michel448a
Volunteer tester
Avatar
Send message
Joined: 27 Oct 00
Posts: 1201
Credit: 2,891,635
RAC: 0
Canada
Message 1192500 - Posted: 8 Feb 2012, 15:41:46 UTC - in response to Message 1192496.

Where you will find <true_angle_range> followed by the value such as <true_angle_range>0.37454123643124</true_angle_range>


0.37454123643124 which kind of AR it is ? VLAR ? VHAR ? or Normal ?
____________

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1192501 - Posted: 8 Feb 2012, 15:51:15 UTC - in response to Message 1192496.

I know when I have a vlar wu just by looking at my task but how can I tell if I get a vhar wu or a ghost wu.

One way that you can find the angle range is by open the work unit in something such as notepad. Where you will find <true_angle_range> followed by the value such as <true_angle_range>0.37454123643124</true_angle_range>.

EDIT: As I use Fred's reschedule program I can see the number of each kind of tasks that my machine has. You could use it just for that without using any of the reschedule functions if you wanted.

To tell if you have "ghost" tasks. You can check the count of the tasks on your machine vs. the number the on your machines tasks pages on the website. If the in progress number is larger than the count on your machine then you have "ghost" tasks. When the servers are enabled to send out lost tasks you don't have to worry about ghosts. At the moment this is enabled.


So what angle ranges are for vlar and intermediate angle range and vhar?
____________

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1192503 - Posted: 8 Feb 2012, 16:02:06 UTC
Last modified: 8 Feb 2012, 16:04:23 UTC

I found this

Notice that the angle range of this wu is 0.65xxx so it would be considered a "high" angle range when compared to a "normal" AR of 0.41/0.42 Also note that this wu would take about 1/3 less time to crunch than a "normal" AR unit.

http://www.boinc-wiki.info/SETI@Home_Enhanced_FAQ#How_do_I_find_what_the_Angle_Range_of_a_work_unit_is.3F

So a normal is 0.41-0.42
So a vhar is above 0.42
So a vlar is below 0.41
____________

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1192508 - Posted: 8 Feb 2012, 16:13:44 UTC - in response to Message 1192503.

I found this

Notice that the angle range of this wu is 0.65xxx so it would be considered a "high" angle range when compared to a "normal" AR of 0.41/0.42 Also note that this wu would take about 1/3 less time to crunch than a "normal" AR unit.

http://www.boinc-wiki.info/SETI@Home_Enhanced_FAQ#How_do_I_find_what_the_Angle_Range_of_a_work_unit_is.3F

So a normal is 0.41-0.42
So a vhar is above 0.42
So a vlar is below 0.41


But this wu is <true_angle_range>0.17057585438786</true_angle_range> would be a vlar but it is not marked as vlar so I guess that is not a true and fast rule observed so if had got sent to an nv card it would have problems.
____________

Profile Michel448a
Volunteer tester
Avatar
Send message
Joined: 27 Oct 00
Posts: 1201
Credit: 2,891,635
RAC: 0
Canada
Message 1192518 - Posted: 8 Feb 2012, 16:39:25 UTC - in response to Message 1192508.

here in your page it s written:

What is an "AR" or What is an Angle Range?

The "AR" is the "Angle Range" of the work unit. This basically describes the angle at which the Arecibo Radio Telescope was "pointed" at the sky to record the information that is contained in that particular work unit.

____________

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 3993
Credit: 109,970,575
RAC: 136,483
United States
Message 1192529 - Posted: 8 Feb 2012, 17:00:56 UTC - in response to Message 1192508.
Last modified: 8 Feb 2012, 17:11:37 UTC

I found this

Notice that the angle range of this wu is 0.65xxx so it would be considered a "high" angle range when compared to a "normal" AR of 0.41/0.42 Also note that this wu would take about 1/3 less time to crunch than a "normal" AR unit.

http://www.boinc-wiki.info/SETI@Home_Enhanced_FAQ#How_do_I_find_what_the_Angle_Range_of_a_work_unit_is.3F

So a normal is 0.41-0.42
So a vhar is above 0.42
So a vlar is below 0.41


But this wu is <true_angle_range>0.17057585438786</true_angle_range> would be a vlar but it is not marked as vlar so I guess that is not a true and fast rule observed so if had got sent to an nv card it would have problems.

IIRC the servers use 0.13 at the threshold the flag tasks as VLAR. The range for normal is much broader than .41 or .42. I think 0.42 was the desired AR for the classic S@H and kind of the center road. If you have seen the charts that have been made by various people where they plot time and AR you can get a good sense of where the VHAR and VLAR ranges are. Often the graph looks like a sideways S that is flattened out in the middle.

Edit: Here is one of the graphs Raistmer did. Not quite a flattened S like I was thinking, but shows where many AR tasks fall.
http://img196.imageshack.us/img196/4914/mbtasks.png
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

1 · 2 · Next

Message boards : Number crunching : WU types

Copyright © 2014 University of California