Thanks, I suspected of this too.
I will reinstall sage with this command
$ /usr/bin/cpufreq-selector -g performance
$ ./sage
$ /usr/bin/cpufreq-selector -g ondemand
21.12.2010 06:50, Dima Pasechnik пишет:
to install Atlas well, you should switch to "performance" CPU mode.
"On demand" might produce inconsistent timings during the tuning
stage.
http://math-atlas.sourceforge.net/atlas_install/atlas_install.html#SECTION00032000000000000000
On Dec 18, 6:07 am, "Alexey U. Gudchenko"<p...@goodok.ru> wrote:
Sorry for my English,
I carried out some experiments, used the system reinstalation to excuse.
Configuration:
Ubuntu 10.10 with 2.6.35-23 kernel
2 GB of memory
AMD Athlon II X2 245
Processor clock frequency is on the "On demand" mode
$ cd /home/goodok/ta/sage-4.6
$ ./make
On about 0:43 minute I have the error while ATLAS was tuning
====
best cgemvT : ID=106, mu=1, nu=128 at 6748.29 MFLOPS
...
make[5]: *** [res/cMVRES]
...
ERROR 734 DURING MVTUNE!!. CHECK INSTALL_LOG/cMVTUNE.LOG FOR DETAILS.
====
Then, just in case, I went to the sage-4.6 directory so that the path
pointed to sage-4.6 do not contains any soft links.
(it was /home/goodok/ta --> /home/goodok/Tools-arsenal )
$ cd /home/goodok/Tools-arsenal/
$ rm -r sage-4.6
$ tar -xf sage-4.6.tar
$ cd sage-4.6
I run "./make" once more.
Installation ended with success.
./sage shell worked normally too.
Then I again remove "sage-4.6" dir and chage dir as at first case
$ cd /home/goodok/ta/
$ rm -r sage-4.6
$ tar -xf sage-4.6.tar
$ cd sage-4.6
$ ./make
Installation ended with success too.
So I decided it is independent of the file system links.
I suppose that the ATLAS tuning is unstable in random.
Or may be the screen saver GLMatrix has significance for this process.
16.12.2010 20:48, Jan Groenewald:
Hi
I had ATLAS fail to build from sage-3.6 running Ubuntu 10.04.1
on an HP Probook 4515s with and AMD CPU (yesterday).
Due to time pressure I installed a prebuilt binary instead.
Sorry, no more info.
regards,
Jan
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
