Posts by Claggy

1) Message boards : Cruncher's Corner : New Driver: AMD Catalyst Omega Driver (Message 136578)
Posted 2 days ago by Claggy

You want the ATI Driver Cheat Sheet, As well as matching CAL to APP OpenCL runtimes, it also has links to different drivers, and to Previous Driver Lists,

Claggy


I don't quite under the list. Is the SDK, OCL and CAL version important? Do they not come with the driver package?

Nowadays the Catalyst package does contain all the drivers, Earlier on ATI/AMD was producing drivers with just CAL drivers, and you had to install the SDK to get OpenCL support,
Then they produced drivers that needed an extra OpenCL runtime download, then an Choice of without and with the OpenCL runtime, Now all drivers included the OpenCL runtime,
just like Nvidia's drivers,

How do I match the matching CAL to APP OpenCL runtimes?

Einstein's OpenCL apps don't produce much diagnostic info, ie what OpenCL runtime is used, This Cheat Sheet was produced for Setiathome users, there the Seti v7 and Astropulse v6/v7 apps report the OpenCL runtime, while the Host details page reports the CAL version,

The problem with ATI/AMD drivers is the CAL and OpenCL drivers can become mismatched, we see this at Setiathome, hosts with recent CAL driver and an ancient OpenCL runtime,
or with just old CAL and OpenCL runtimes, But mismatched from different SDK versions, either the Seti or Astropulse app doesn't work, or produces rubbish,
Maybe windows update updated the driver in those cases, But not the OpenCL runtime, Or they have upgraded the Catalyst package, then windows reverted the driver at some point.

For example here's a host running CAL driver 1.4.1546, and OpenCL 1.2 AMD-APP (923.1), the CAL driver is from Cat 11.9, and uses SDK 2.5,
and the OpenCL runtime is from Cat 12.4 and uses SDK 2.7, GPU unsurprisingly produces inconclusive/invalids.

Claggy
2) Message boards : Cruncher's Corner : New Driver: AMD Catalyst Omega Driver (Message 136532)
Posted 2 days ago by Claggy
I rolled back the driver from 14.12 to version 13.12 for windows 7 64 bit. I had to search for the driver since I could not find any of the older version on the AMD website.

You want the ATI Driver Cheat Sheet, As well as matching CAL to APP OpenCL runtimes, it also has links to different drivers, and to Previous Driver Lists, So you can get Cat 14.4 or 14.9 if needed:

ATI Driver Version Cheat Sheet

Claggy
3) Message boards : Problems and Bug Reports : BRP4 Intel GPU app feedback thread (Message 136512)
Posted 3 days ago by Claggy
The last host isn't completing any intel_gpu Wu's, some end in maximum time exceeded, maybe some a more allowance is needed, But it's a huge difference between the valid result of 904.63secs (on the other host), and the 27,726.57secs some of its results take,
looking at the stderr.txt shows no checkpointing recorded in any of those long running Wu's, some throw an Access Violation error:

All tasks for computer 7536812

Host 7536812:
http://einstein.phys.uwm.edu/host_sched_logs/7536/7536812

While it hasn't asked for work here, it is attached to other projects, a stderr.txt from a Seti v7 result shows:

http://setiathome.berkeley.edu/show_host_detail.php?hostid=7026234
Name: Intel(R) HD Graphics 4000 Vendor: Intel(R) Corporation Driver version: 10.18.10.3316 Version: OpenCL 1.2


Claggy
4) Message boards : Problems and Bug Reports : BRP4 Intel GPU app feedback thread (Message 136510)
Posted 3 days ago by Claggy
Host 11672358 has asked for work:

Host 11672358
http://einstein.phys.uwm.edu/host_sched_logs/11672/11672358

2014-12-17 07:01:19.9178 [PID=23748] [version] Checking plan class 'opencl-intel_gpu'
2014-12-17 07:01:19.9179 [PID=23748] [version] parsed project prefs setting 'gpu_util_brp': 0.000000
2014-12-17 07:01:19.9179 [PID=23748] [version] OpenCL driver version: 10.18.10.3977; platform version: OpenCL 1.2; device version: OpenCL 1.2
2014-12-17 07:01:19.9179 [PID=23748] [version] Peak flops supplied: 6.4e+10
2014-12-17 07:01:19.9179 [PID=23748] [version] plan class ok


Claggy
5) Message boards : Problems and Bug Reports : BRP4 Intel GPU app feedback thread (Message 136490)
Posted 4 days ago by Claggy
The 4th WU is a validation nightmare, but at least contains an example of a Haswell validating against the NEON app. That Haswell might simply use the older driver, though.

There is a way of finding out (the present driver version), while you can't see it by looking at the host details page, you can by looking at the last contact as long as they ask for work:

Host 7454946:
http://einstein.phys.uwm.edu/host_sched_logs/7454/7454946
2014-12-16 06:00:44.7290 [PID=19867] [version] Checking plan class 'opencl-intel_gpu'
2014-12-16 06:00:44.7291 [PID=19867] [version] parsed project prefs setting 'gpu_util_brp': 1.000000
2014-12-16 06:00:44.7291 [PID=19867] [version] OpenCL driver version: 10.18.10.3960; platform version: OpenCL 1.2; device version: OpenCL 1.2
2014-12-16 06:00:44.7291 [PID=19867] [version] Peak flops supplied: 6.4e+10
2014-12-16 06:00:44.7291 [PID=19867] [version] plan class ok


Host 7536812:
http://einstein.phys.uwm.edu/host_sched_logs/7536/7536812

Host 8362996
http://einstein.phys.uwm.edu/host_sched_logs/8362/8362996
2014-12-16 12:03:08.4868 [PID=22374] [version] Checking plan class 'opencl-intel_gpu'
2014-12-16 12:03:08.4868 [PID=22374] [version] parsed project prefs setting 'gpu_util_brp': 0.000000
2014-12-16 12:03:08.4868 [PID=22374] [version] OpenCL driver version: 10.18.10.3960; platform version: OpenCL 1.2; device version: OpenCL 1.2
2014-12-16 12:03:08.4868 [PID=22374] [version] Peak flops supplied: 6.4e+10
2014-12-16 12:03:08.4868 [PID=22374] [version] plan class ok


Host 11672358
http://einstein.phys.uwm.edu/host_sched_logs/11672/11672358

Host 10823173:
http://einstein.phys.uwm.edu/host_sched_logs/10823/10823173
2014-12-16 13:49:54.5165 [PID=4684 ] [version] Checking plan class 'opencl-intel_gpu'
2014-12-16 13:49:54.5165 [PID=4684 ] [version] parsed project prefs setting 'gpu_util_brp': 0.000000
2014-12-16 13:49:54.5165 [PID=4684 ] [version] OpenCL driver version: 9.18.10.3186; platform version: OpenCL 1.2; device version: OpenCL 1.2
2014-12-16 13:49:54.5165 [PID=4684 ] [version] Peak flops supplied: 6.4e+10
2014-12-16 13:49:54.5165 [PID=4684 ] [version] plan class ok


Host 11674682:
http://einstein.phys.uwm.edu/host_sched_logs/11674/11674682
2014-12-16 13:33:33.6684 [PID=29873] [version] Checking plan class 'opencl-intel_gpu'
2014-12-16 13:33:33.6684 [PID=29873] [version] parsed project prefs setting 'gpu_util_brp': 0.000000
2014-12-16 13:33:33.6684 [PID=29873] [version] OpenCL driver version: 10.18.10.3960; platform version: OpenCL 1.2; device version: OpenCL 1.2
2014-12-16 13:33:33.6684 [PID=29873] [version] Peak flops supplied: 6.4e+10
2014-12-16 13:33:33.6684 [PID=29873] [version] plan class ok


Claggy
6) Message boards : Problems and Bug Reports : BRP4 Intel GPU app feedback thread (Message 136484)
Posted 4 days ago by Claggy
All we can say for certain is that there was a change in behaviour after the 10.18.10.3621 driver, which produces a high validation rate. Quite what goes wrong - what changes in the result file - I'm afraid I can't say: that might be easier for you or Bikeman to extract from the returned results.

For anyone needing the last driver that works, I've looked through my downloads (I checked the readme), and my download history (I use Getright),
the original (direct) download url for the 10.18.10.3621 driver for 64bit Windows is:

http://downloadmirror.intel.com/23885/a08/win64_153322.zip

README FILE


Release Version: Production Version

Driver Version: 15.33.22.64.3621

Operating System(s):

Microsoft Windows* 7 64
Microsoft Windows* 8 64
Microsoft Windows* 8.1 64

Release Date: May 21, 2014


CONTENTS

I. Product Support
II. Installation Information
III. Disclaimer
IV. Important Note


I. Product Support


Supports Intel(R) Iris(TM) graphics, Intel(R) Iris(TM) Pro graphics and Intel(R) HD graphics on:

4th Generation Intel(R) Core(TM) Processor Platform
4th Generation Intel(R) Core(TM) Processor U Series-based Platform
4th Generation Intel(R) Core(TM) Processor Y Series-based Platform
3rd Generation Intel(R) Core(TM) Processor Platform
3rd Generation Intel(R) Core(TM) Processor U Series-based Platform
3rd Generation Intel(R) Core(TM) Processor Y Series-based Platform
Intel(R) Pentium(TM) Processor 2000/A1000/G2000/G2100 Series
Intel(R) Celeron(TM) Processor 900/1000/G1000 Series
Intel(R) Pentium(R) N3510/ 3520 Processor
Intel(R) Celeron(R) N2810/ 2820 Processor
Intel(R) Celeron(R) N2910/ 2920 Processor
Intel(R) Celeron(R) N2805/ N2806 Processor
Intel(R) Celeron(R) N2815 Processor
Intel(R) Pentium(R) J2850/ J2900 Processor
Intel(R) Celeron(R) J1850/ J1900 Processor
Intel(R) Celeron(R) J1750/ J1800 Processor


II. Installation Information

This document makes references to products developed by Intel. There are some restrictions on how these products may be used, and what information may be disclosed to others. Please read the Disclaimer section and contact your Intel field representative if you would like more information.

III. Disclaimer

Intel is making no claims of usability, efficacy or warranty. The INTEL SOFTWARE LICENSE AGREEMENT contained herein completely defines the license and use of this software.

This document contains information on products in the design phase of development. The information here is subject to change without notice. Do not finalize a design with this information.

IV. Important Note

Your computer is configured to run the WinSAT* test, which determines whether the computer can support the Windows Aero* desktop theme. The test will run after this driver is installed and the computer is restarted. If your computer can support it, the Windows Aero desktop theme will automatically be enabled.

This test will take approximately 10 seconds.

*Other names and brands may be claimed as the property of others.


Copyright (C) 2014 Intel Corporation. All rights reserved.


Claggy
7) Message boards : Problems and Bug Reports : BRP4 Intel GPU app feedback thread (Message 136429)
Posted 5 days ago by Claggy
Up to you to judge whether the current error return rate from the Intel GPU app is acceptable to the project (especially considering whether validations arising from two 'new driver' hosts matching are scientifically useful).

My fleet of Android and Linux Arm hosts have five invalid tasks, and a number of inconclusives between them where they have been matched with a pair of intel_gpu hosts:

http://einstein.phys.uwm.edu/hosts_user.php?userid=238511

they also run at Albert, they have no invalids, nor inconclusives there:

http://albertathome.org/account/65462/computers

Claggy
8) Message boards : Cruncher's Corner : Boinc no longer sees Nvidia GPU after update, Mint 17. (Message 136368)
Posted 6 days ago by Claggy
Does Mint 17 have a nvidia_uvm module?

Message 136336

Claggy
9) Message boards : Cruncher's Corner : Work Unit odditys (Message 135930)
Posted 16 days ago by Claggy
And do you also have Seti as part of the mix? If you do, and Seti can't supply, it might explain why you got a bunch of Einstein tasks. BOINC tends to do things like that. Until you get properly established, keep your cache settings low - say 0.5 to 1.0 days until things settle.

Of Ryan's four hosts, two have the minimum buffer set to 0.1 days, one to 10 days, and the last to 0 days.

Claggy
10) Message boards : Problems and Bug Reports : Android app release: feedback thread (Message 135928)
Posted 16 days ago by Claggy
A heads up if no one is watching the boinc_projects mailing list:

http://lists.ssl.berkeley.edu/mailman/private/boinc_projects/2014-November/011225.html

Android apps (if your project has them, read this!)‏

At some point, Google introduced the notion of
"position-independent executable" (PIE) native-mode programs:
http://stackoverflow.com/questions/24818902/

Support for PIE programs was added in Android 4.1.
Starting with Android 5.0 (which is now shipping on some devices)
native-mode applications *must* be PIE.
This means that your current Android apps will fail on Android 5.0.

(BTW, the BOINC client is a native-mode app; we've built a PIE version
of it, which seems to work on Android 5.0).

Building PIE versions of apps just requires adding a couple of compile flags; see
http://boinc.berkeley.edu/trac/wiki/AndroidBuildApp#Position-independentexecutablesPIE

So, I recommend that you do the following:

- build PIE versions of your applications
- deploy the new versions, with a plan class that sends them
only to Android 4.1+ clients
- if you like, deploy your old non-PIE app versions with a plan class
that sends them only to pre-4.0 clients.
Note: pre-4.0 clients account for less than 10% of the pool.

Currently, Android clients report their Linux kernel version,
not their Android version.
I believe that Android 4.1 uses Linux 3.0.31,
while earlier Androids use earlier Linux.
So, if you use XML plan class specs, use
<min_os_version>30031</min_os_version>
for your PIE version.

Sorry for the late notice; we found out about this ourselves only a few days ago.

-- David


Claggy


Next 10

Home · Your account · Message boards

This material is based upon work supported by the National Science Foundation (NSF) under Grants PHY-1104902, PHY-1104617 and PHY-1105572 and by the Max Planck Gesellschaft (MPG). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the investigators and do not necessarily reflect the views of the NSF or the MPG.

Copyright © 2014 Bruce Allen