GBT ('guppi') .vlar tasks will be send to GPUs, what you think about this?

Message boards : Number crunching : GBT ('guppi') .vlar tasks will be send to GPUs, what you think about this?
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 10 · Next

AuthorMessage
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1785964 - Posted: 8 May 2016, 22:31:50 UTC - in response to Message 1785962.  

It's still a way to go.

Before the current app can be considered for release to run as stock.

Before VLAR could be enabled on main, perhaps? Don't mix things up.

Yes indeed, and that keeps us on topic for this thread. I don't think that GBT VLARs should be sent to (NVidia) GPUs until we have an application ready to handle them, and I don't think we have such an application - either CUDA or OpenCL - yet.

So far, there has been enough non-VLAR work to keep my GPUs busy. My cache dives a bit now and again, but seems to recover well enough.
It should improve when Eric has the time to start splitting AP work from the Green Bank datasets, which he has indicated he will do.
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1785964 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1785994 - Posted: 9 May 2016, 0:22:53 UTC - in response to Message 1785964.  
Last modified: 9 May 2016, 0:24:12 UTC

It's still a way to go.

Before the current app can be considered for release to run as stock.

Before VLAR could be enabled on main, perhaps? Don't mix things up.

Yes indeed, and that keeps us on topic for this thread. I don't think that GBT VLARs should be sent to (NVidia) GPUs until we have an application ready to handle them, and I don't think we have such an application - either CUDA or OpenCL - yet.

So far, there has been enough non-VLAR work to keep my GPUs busy. My cache dives a bit now and again, but seems to recover well enough.
It should improve when Eric has the time to start splitting AP work from the Green Bank datasets, which he has indicated he will do.

As I recall one of the reason for MB v8 was to be able to accommodate data from sources other than Arecibo. So we will probably need a new AP app as well? I believe the only thing I have read so far was Eric mentioning there are plans to do AP for the GBT data as well.

Maybe they can get some summer inters to bang out the new splitter and base apps.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1785994 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1785996 - Posted: 9 May 2016, 0:25:21 UTC - in response to Message 1785994.  

It's still a way to go.

Before the current app can be considered for release to run as stock.

Before VLAR could be enabled on main, perhaps? Don't mix things up.

Yes indeed, and that keeps us on topic for this thread. I don't think that GBT VLARs should be sent to (NVidia) GPUs until we have an application ready to handle them, and I don't think we have such an application - either CUDA or OpenCL - yet.

So far, there has been enough non-VLAR work to keep my GPUs busy. My cache dives a bit now and again, but seems to recover well enough.
It should improve when Eric has the time to start splitting AP work from the Green Bank datasets, which he has indicated he will do.

As I recall one of the reason for MB v8 was to be able to accommodate data from sources other than Arecibo. So we will probably need a new AP app as well? I believe the only thing I have read so far was Eric mentioning there are plans to do AP for the GBT data as well.

I am not sure a new app will be required. AP is a different animal than MB. And indications I have are that the current app will work properly.
Of course, subject to change as the winds blow...LOL.
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1785996 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 1786010 - Posted: 9 May 2016, 2:45:21 UTC

I've started SETI Beta to use my AMD HD 7770. The first task ended in about a minute and is waiting for validation. But what is SoG? It's the first time i see this in a task on my SuSE Linux 42.1 host.
Tullio
ID: 1786010 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1786088 - Posted: 9 May 2016, 8:39:45 UTC

A 0.005 guppi vlar under 10 minutes. No screen lag. System fully usable.
How many credits do you get from a guppi vlar?


02:28:42 (1018): Can't open init data file - running in standalone mode
02:28:42 (1018): Can't open init data file - running in standalone mode
setiathome_CUDA: Found 4 CUDA device(s):
  Device 1: GeForce GTX 980, 4095 MiB, regsPerBlock 65536
     computeCap 5.2, multiProcs 16 
     pciBusID = 1, pciSlotID = 0
  Device 2: GeForce GTX 780, 3071 MiB, regsPerBlock 65536
     computeCap 3.5, multiProcs 12 
     pciBusID = 2, pciSlotID = 0
  Device 3: GeForce GTX 980, 4095 MiB, regsPerBlock 65536
     computeCap 5.2, multiProcs 16 
     pciBusID = 3, pciSlotID = 0
  Device 4: GeForce GTX 780, 3071 MiB, regsPerBlock 65536
     computeCap 3.5, multiProcs 12 
     pciBusID = 4, pciSlotID = 0
In cudaAcc_initializeDevice(): Boinc passed DevPref 1
setiathome_CUDA: CUDA Device 1 specified, checking...
   Device 1: GeForce GTX 980 is okay
SETI@home using CUDA accelerated device GeForce GTX 980
Using pfb = 8 from command line args
Using pfp = 120 from command line args

setiathome v8 enhanced x41p_zn, Cuda 7.50 special
Compiled with NVCC 7.5, using 6.5 libraries. Modifications done by petri33.



Detected setiathome_enhanced_v8 task. Autocorrelations enabled, size 128k elements.
Work Unit Info:
...............
WU true angle range is :  0.005109
Sigma 639
Sigma > GaussTOffsetStop: 639 > -575
Thread call stack limit is: 1k
cudaAcc_free() called...
cudaAcc_free() running...
cudaAcc_free() PulseFind freed...
cudaAcc_free() Gaussfit freed...
cudaAcc_free() AutoCorrelation freed...
1,2,3,4,5,6,7,8,9,10,10,11,12,cudaAcc_free() DONE.
13
Flopcounter: 30954930943609.285156

Spike count:    0
Autocorr count: 1
Pulse count:    4
Triplet count:  0
Gaussian count: 0
02:38:15 (1018): called boinc_finish(0)

To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1786088 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1786089 - Posted: 9 May 2016, 8:47:38 UTC - in response to Message 1786088.  

How many credits do you get from a guppi vlar?

Looking at my current test at Beta, the answer seems to be "it depends on the runtime". But Beta averages and workload distributions will be different from what we (currently) have at Main.
ID: 1786089 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786091 - Posted: 9 May 2016, 10:00:07 UTC - in response to Message 1786089.  

How many credits do you get from a guppi vlar?

Looking at my current test at Beta, the answer seems to be "it depends on the runtime". But Beta averages and workload distributions will be different from what we (currently) have at Main.


Yeah, sadly CreditNew penalises optimisation (i.e. computational efficiency) no matter what the designer states.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786091 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 1786542 - Posted: 11 May 2016, 2:39:58 UTC
Last modified: 11 May 2016, 2:40:25 UTC

I am getting guppi vlars on my Windows 10 PC with its Nvidia GTX 750 TI OC board.
Tullio
ID: 1786542 · Report as offensive
Lionel

Send message
Joined: 25 Mar 00
Posts: 680
Credit: 563,640,304
RAC: 597
Australia
Message 1786548 - Posted: 11 May 2016, 3:13:57 UTC - in response to Message 1786542.  

noticed them before ... run times just under an hour ... they seem to take longer than they do on my cpus
ID: 1786548 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1786549 - Posted: 11 May 2016, 3:14:14 UTC

Hmm, 750Ti running 1 vlar Plus 1 non vlar (2 tasks).

vlar time increased by 70% vs non vlar
non vlar time increased by 95%

GPU vs CPU(i5), GPU is approx half the time.

Interesting note, my CPU temps jumped about 4C, and my GPU went down 4C, on both computers.
ID: 1786549 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 1786554 - Posted: 11 May 2016, 3:40:22 UTC

I was expecting guppi .vlars on my Linux box with its AMD HD 7770 but I got none in SETI Beta. Now I am getting them on the Windows PC with its Nvidia GTX 750 Ti OC. They take about 55 minutes.
Tullio
ID: 1786554 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1786557 - Posted: 11 May 2016, 3:49:15 UTC - in response to Message 1786554.  

My 750Ti is a little faster;
http://setiathome.berkeley.edu/result.php?resultid=4923071938
This one is about the same;
http://setiathome.berkeley.edu/result.php?resultid=4923086322
Those are running 1 task at a time.
ID: 1786557 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1786564 - Posted: 11 May 2016, 4:03:44 UTC
Last modified: 11 May 2016, 4:06:13 UTC

Running Raismter's SoG on my Titan X's

Running 3 at a time on each GPU with commandlines from Mike,

Each are taking just under 1 hour (56-58 minutes)

http://setiathome.berkeley.edu/result.php?resultid=4923517925

http://setiathome.berkeley.edu/result.php?resultid=4923517967

http://setiathome.berkeley.edu/result.php?resultid=4923517729
ID: 1786564 · Report as offensive
Lionel

Send message
Joined: 25 Mar 00
Posts: 680
Credit: 563,640,304
RAC: 597
Australia
Message 1786571 - Posted: 11 May 2016, 4:31:55 UTC - in response to Message 1786564.  

anyone want to have a guess at what will happen RAC ...
ID: 1786571 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1786572 - Posted: 11 May 2016, 4:33:51 UTC - in response to Message 1786571.  

RAC = Really Aren't Crediting anyone.
ID: 1786572 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 1786573 - Posted: 11 May 2016, 4:38:34 UTC

In Einstein@home I am getting 2000 credits for a CPU task that lasts 90k s and 4400 credits for a GPU task that takes 9k s.
Tullio
ID: 1786573 · Report as offensive
Lionel

Send message
Joined: 25 Mar 00
Posts: 680
Credit: 563,640,304
RAC: 597
Australia
Message 1786587 - Posted: 11 May 2016, 5:32:05 UTC - in response to Message 1786572.  

RAC = Really Aren't Crediting anyone.


love it
ID: 1786587 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1786629 - Posted: 11 May 2016, 8:49:06 UTC
Last modified: 11 May 2016, 9:27:18 UTC

Guppi vlars are coming in and they are done in 9-12 minutes one at a time.

I suspect they are not doing gaussian searching: Sigma > GaussTOffsetStop: 498 > -434

The difference between singma and GaussTOffsetStop would be 64 if the TOffsetStop would be positive ...

analyzePoT.cpp:      PoTInfo.GaussTOffsetStart   = static_cast<int>(floor(swi.analysis_cfg.pot_t_offset * PoTInfo.GaussSigma+0.5));
analyzePoT.cpp:      PoTInfo.GaussTOffsetStop    = swi.analysis_cfg.gauss_pot_length - PoTInfo.GaussTOffsetStart;
gaussfit.cpp:  

should probably be

analyzePoT.cpp:      PoTInfo.GaussTOffsetStart   = static_cast<int>(floor(swi.analysis_cfg.pot_t_offset * PoTInfo.GaussSigma+0.5));
analyzePoT.cpp:      PoTInfo.GaussTOffsetStop    = PoTInfo.GaussTOffsetStart + swi.analysis_cfg.gauss_pot_length;
gaussfit.cpp:  


Just my thoughts.
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1786629 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786634 - Posted: 11 May 2016, 9:01:53 UTC - in response to Message 1786629.  

Sounds about right Petri. Gaussian Search requires a drift over a spot on the sky. i.e. pass the telescope beam over a specific direction and get a bell shaped signal from a constant source. VLAR targeted searches (unblinking eye at one spot) don't pass over a spot, but travel with it.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786634 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1786635 - Posted: 11 May 2016, 9:04:51 UTC - in response to Message 1786634.  
Last modified: 11 May 2016, 9:05:08 UTC

Sounds about right Petri. Gaussian Search requires a drift over a spot on the sky. i.e. pass the telescope beam over a specific direction and get a bell shaped signal from a constant source. VLAR targeted searches (unblinking eye at one spot) don't pass over a spot, but travel with it.


Yes, I know what a gaussian is and how it is formed in to the signal.

I have noticed the same negative value in test wu PG009.

I wonder if ths is a bug in cuda code base or in all sw versions. See edit in my other message.
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1786635 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 10 · Next

Message boards : Number crunching : GBT ('guppi') .vlar tasks will be send to GPUs, what you think about this?


 
©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.