Runtimes on a HD5770, 2 tasks at a time.
[pre]
App Average of Count of
Run Time(s) results
Binary Radio Pulsar Search (Arecibo) v1.24 (opencl-ati) 14490.90 38
Binary Radio Pulsar Search (Arecibo) v1.28 (opencl-ati) 6105.84 7
[/pre]
Quite in improvement, I would say. Next step would be collecting a few more results on the 1.28 app, after that reverting to running one task at a time.
App v1.28 halved both my GPU and CPU run times. From ~4,600 seconds GPU and ~750 seconds CPU to ~2,300 seconds GPU and ~370 seconds CPU. Thumbs up. AMD HD6850, 2GB.
Saw the same problem with 12.8 -> increased run times. I run Win7 64bit. I don't know why but I changed from "Balanced" performance to "High" performance under "Power Options" and saw run times return to the 12.6 values. This was on Milky Way at Home. I haven't timed Einstein yet.
After update E@H OpenCL application from 1.24 to version 1.28 error with the lack of free memory at once disappeared (which confirms that the problem was not in amount of free RAM, but a bug - in the application or in the AMD drivers or probably both).
But 1.28 app still not working - just another error in all Wus: http://einsteinathome.org/host/2896418/tasks
[15:51:43][3012][ERROR] Error during OpenCL kernel setup: HSFFB (error: -5)
[15:51:43][3012][ERROR] Demodulation failed (error: 2019)!
(0x7e3) - exit code 2019 (0x7e3)
or
[16:51:34][6032][ERROR] Error during OpenCL kernel setup: TSMP_T (error: -5)
[16:51:34][6032][ERROR] Demodulation failed (error: 2019)!
(0x7e3) - exit code 2019 (0x7e3)
Hmm, you right.
More exactly - at 1.24 app was 2 errors:
-Demodulation failed (error: 2013)! - this related to bug with lack of free RAM (when lot of free RAM actually)
-Demodulation failed (error: 2019)!
At 1.28 app first error (2013) gone, but 2019 still persist.
I run einstein in the background & pause it when I need to do processor intensive work. I recently started getting WU's for the gpu but these won't pause when I need to do graphics intensive work. The Boinc manager tasks status is listed as suspended, but the task keeps on using the gpu. I had to abort the task to get on with work. Do you have any suggestions so that I won't have to continue to abort opencl-ati-lion tasks?
BRP4 1.28 OpenCL and CUDA app : feedback thread
)
This is a report not a problem. :)
Runtimes on a HD5770, 2 tasks at a time.
[pre]
App Average of Count of
Run Time(s) results
Binary Radio Pulsar Search (Arecibo) v1.24 (opencl-ati) 14490.90 38
Binary Radio Pulsar Search (Arecibo) v1.28 (opencl-ati) 6105.84 7
[/pre]
Quite in improvement, I would say. Next step would be collecting a few more results on the 1.28 app, after that reverting to running one task at a time.
Today I installed latest
)
Today I installed latest AMD-drivers 12.8. Runtimes on my HD6950 increased from ~3200 sec's (2wu's) to ~3770 sec's (2 wu's).
Will switch back to 12.6
App v1.28 halved both my GPU
)
App v1.28 halved both my GPU and CPU run times. From ~4,600 seconds GPU and ~750 seconds CPU to ~2,300 seconds GPU and ~370 seconds CPU. Thumbs up. AMD HD6850, 2GB.
Alex, Saw the same problem
)
Alex,
Saw the same problem with 12.8 -> increased run times. I run Win7 64bit. I don't know why but I changed from "Balanced" performance to "High" performance under "Power Options" and saw run times return to the 12.6 values. This was on Milky Way at Home. I haven't timed Einstein yet.
Joe
After update E@H OpenCL
)
After update E@H OpenCL application from 1.24 to version 1.28 error with the lack of free memory at once disappeared (which confirms that the problem was not in amount of free RAM, but a bug - in the application or in the AMD drivers or probably both).
But 1.28 app still not working - just another error in all Wus:
http://einsteinathome.org/host/2896418/tasks
[15:51:43][3012][ERROR] Error during OpenCL kernel setup: HSFFB (error: -5)
[15:51:43][3012][ERROR] Demodulation failed (error: 2019)!
(0x7e3) - exit code 2019 (0x7e3)
or
[16:51:34][6032][ERROR] Error during OpenCL kernel setup: TSMP_T (error: -5)
[16:51:34][6032][ERROR] Demodulation failed (error: 2019)!
(0x7e3) - exit code 2019 (0x7e3)
And POEM@Home OpenCL app still work fine on same hard&soft:
http://boinc.fzk.de/poem/results.php?hostid=109617
P.S.
Errors with free RAM (in 1.24 version) described in another topic:
http://einsteinathome.org/node/196325&nowrap=true#118714
http://einsteinathome.org/node/196325&nowrap=true#118752
RE: But 1.28 app still not
)
Actually, it's the same error. Before you also had exit code 2019 on all your tasks.
For the developers, so far I've found that this exit code 2019 is the same as There was an error while deleting the color transform. (0x7e3) - exit code 2019 (0x7e3) that I had on Albert. Ask Oliver how he fixed that.
still no dice for me on
)
still no dice for me on HD5670/winXP32.
I get the same "color transform" error #2019 from the earlier 1.24 app.
fyi
7.0.28
There was an error while deleting the color transform. (0x7e3) - exit code 2019 (0x7e3)
Activated exception handling...
[11:53:20][3960][INFO ] Starting data processing...
[11:53:22][3960][INFO ] Using OpenCL platform provided by: Advanced Micro Devices, Inc.
[11:53:22][3960][INFO ] Using OpenCL device "Redwood" by: Advanced Micro Devices, Inc.
[11:53:22][3960][INFO ] Checkpoint file unavailable: status.cpt (No such file or directory).
------> Starting from scratch...
[11:53:22][3960][INFO ] Header contents:
------> Original WAPP file: ./p2030.20101226.G193.96-00.20.C.b4s0g0.00000_DM151.20
------> Sample time in microseconds: 65.4762
------> Observation time in seconds: 274.62705
------> Time stamp (MJD): 55556.179870079133
------> Number of samples/record: 0
------> Center freq in MHz: 1214.289551
------> Channel band in MHz: 0.33605957
------> Number of channels/record: 960
------> Nifs: 1
------> RA (J2000): 61508.1564999
------> DEC (J2000): 163727.051001
------> Galactic l: 0
------> Galactic b: 0
------> Name: G193.96-00.20.C
------> Lagformat: 0
------> Sum: 1
------> Level: 3
------> AZ at start: 0
------> ZA at start: 0
------> AST at start: 0
------> LST at start: 0
------> Project ID: --
------> Observers: --
------> File size (bytes): 0
------> Data size (bytes): 0
------> Number of samples: 4194304
------> Trial dispersion measure: 151.2 cm^-3 pc
------> Scale factor: 0.00137665
[11:53:25][3960][INFO ] Seed for random number generator is 1168041574.
[11:53:34][3960][INFO ] Derived global search parameters:
------> f_A probability = 0.08
------> single bin prob(P_noise > P_thr) = 1.32531e-008
------> thr1 = 18.139
------> thr2 = 21.241
------> thr4 = 26.2686
------> thr8 = 34.6478
------> thr16 = 48.9581
[11:53:41][3960][ERROR] Error during OpenCL kernel setup: PS_R3 (error: -5)
[11:53:41][3960][ERROR] Demodulation failed (error: 2019)!
11:53:41 (3960): called boinc_finish
]]>
You're using XP, so pls check
)
You're using XP, so pls check your driver-version
http://einsteinathome.org/node/196325&nowrap=true#118752
RE: RE: But 1.28 app
)
Hmm, you right.
More exactly - at 1.24 app was 2 errors:
-Demodulation failed (error: 2013)! - this related to bug with lack of free RAM (when lot of free RAM actually)
-Demodulation failed (error: 2019)!
At 1.28 app first error (2013) gone, but 2019 still persist.
I run einstein in the
)
I run einstein in the background & pause it when I need to do processor intensive work. I recently started getting WU's for the gpu but these won't pause when I need to do graphics intensive work. The Boinc manager tasks status is listed as suspended, but the task keeps on using the gpu. I had to abort the task to get on with work. Do you have any suggestions so that I won't have to continue to abort opencl-ati-lion tasks?