New version of sys_basher, a multithreaded system exerciser andbenchmarking tool

  • Thread starter Thread starter General Schvantzkoph
  • Start date Start date
G

General Schvantzkoph

I've just put a new version of sys_basher on the web,

http://www.polybus.com/sys_basher_web/

sys_basher is a multithreaded system exerciser and benchmarking tool. It
run's extensive memory and I/O diagnostics as well as floating point
tests. I wrote it because there didn't seem to be a good way test the
stability of my new Core2 system. Memtest86+ hasn't been updated for over
a year so it doesn't have any support for Core2 systems. Also it's single
threaded so it doesn't stress modern multicore processors.

sys_basher runs under Linux so it can run on any system that Linux runs
on. It uses pthreads to run from 1 to 256 simultaneous threads. In
addition to running memory tests (at the maximum possible speed), it also
runs disk tests to stress the I/O system, and floating point array
processing to stress the CPU.

I've run sys_basher on a Core2 Duo. Athlon 64 X2, dual Xeon, and two single
core Athlon 64 systems using Fedora Core(s) 4,5 and 6 in both 32 and 64
bit modes. It's straight POSIX C so it should be able to compile and run
on any *nix system.

sys_basher is free and open source, it's licensed under the BSD.
 
General said:
I've just put a new version of sys_basher on the web,

http://www.polybus.com/sys_basher_web/
[snip]
I've run sys_basher on a Core2 Duo. Athlon 64 X2, dual Xeon, and two
single core Athlon 64 systems using Fedora Core(s) 4,5 and 6 in both 32
and 64 bit modes. It's straight POSIX C so it should be able to compile
and run on any *nix system.
Click to expand...
[snip]
I'm using Mandriva 2007 on an Athalon 64 x2 with smp disabled.

Do you want any feedback?

Here are some snippets of output...

Total RAM: 514732K
Free RAM: 96500K
Total Swap: 1534168K
Free Swap: 1533952K

could that be reformatted like this?

Total RAM: 514 732K
Free RAM: 96 500K
Total Swap: 1 534 168K
Free Swap: 1 533 952K

Or this?

Total RAM: 514,732K
Free RAM: 96,500K
Total Swap: 1,534,168K
Free Swap: 1,533,952K


Another snippet of output...

Error Status
Array Size Add Mpy Div Sum of Products
1KB PASS PASS PASS PASS
4KB PASS PASS PASS PASS
16KB PASS PASS PASS PASS
64KB PASS PASS PASS PASS
256KB PASS PASS PASS PASS
1MB PASS PASS PASS PASS
4MB PASS PASS PASS PASS
16MB PASS PASS PASS FAIL
fp error

Have you got any suggestions about how to pursue that "FAIL"?

Thank you.
 
General said:
I've just put a new version of sys_basher on the web,

http://www.polybus.com/sys_basher_web/
[snip]
I've run sys_basher on a Core2 Duo. Athlon 64 X2, dual Xeon, and two
single core Athlon 64 systems using Fedora Core(s) 4,5 and 6 in both 32
and 64 bit modes. It's straight POSIX C so it should be able to compile
and run on any *nix system.
Click to expand...
[snip]
I'm using Mandriva 2007 on an Athalon 64 x2 with smp disabled.

Do you want any feedback?

Here are some snippets of output...

Total RAM: 514732K
Free RAM: 96500K
Total Swap: 1534168K
Free Swap: 1533952K

could that be reformatted like this?

Total RAM: 514 732K
Free RAM: 96 500K
Total Swap: 1 534 168K
Free Swap: 1 533 952K

Or this?

Total RAM: 514,732K
Free RAM: 96,500K
Total Swap: 1,534,168K
Free Swap: 1,533,952K
Click to expand...

That's a reasonable suggestion, I'll add it to the next rev.

Another snippet of output...

Error Status
Array Size Add Mpy Div Sum of Products
1KB PASS PASS PASS PASS
4KB PASS PASS PASS PASS
16KB PASS PASS PASS PASS
64KB PASS PASS PASS PASS
256KB PASS PASS PASS PASS
1MB PASS PASS PASS PASS
4MB PASS PASS PASS PASS
16MB PASS PASS PASS FAIL
fp error

Have you got any suggestions about how to pursue that "FAIL"?
Click to expand...

Does this failure happen consistently? Did any of the memory tests
fail?

The way the floating point tests work is by doing the same
calculation twice, once entirely in registers and once where the operands
are fetched from memory. The program is brand new so I'm not 100%
confident that there aren't potential compiler issues, however it seems
unlikely because the test passed for every case except the last. Your CPU
will be getting hotter and hotter as it runs these tests so if you have a
cooling problem this is the kind of result that you would expect. Please
try the following experiment, set the scaling frequency to the lowest
value and rerun the test. I don't know if Mandriva has the CPU Frequency
Scaling Monitor applet, if it does you can use that the frequency down. If
it doesn't it can be done by writing the following files,


echo userspace > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo userspace > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
echo 500000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed
echo 500000 > /sys/devices/system/cpu/cpu1/cpufreq/scaling_setspeed

The number that you put into scaling_setspeed has to be equal to or less
than the lowest number in

/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
 
General said:
I've just put a new version of sys_basher on the web,

http://www.polybus.com/sys_basher_web/
[snip]
I've run sys_basher on a Core2 Duo. Athlon 64 X2, dual Xeon, and two
single core Athlon 64 systems using Fedora Core(s) 4,5 and 6 in both 32
and 64 bit modes. It's straight POSIX C so it should be able to compile
and run on any *nix system.
Click to expand...
[snip]
I'm using Mandriva 2007 on an Athalon 64 x2 with smp disabled.

Do you want any feedback?

Here are some snippets of output...

Total RAM: 514732K
Free RAM: 96500K
Total Swap: 1534168K
Free Swap: 1533952K

could that be reformatted like this?

Total RAM: 514 732K
Free RAM: 96 500K
Total Swap: 1 534 168K
Free Swap: 1 533 952K

Or this?

Total RAM: 514,732K
Free RAM: 96,500K
Total Swap: 1,534,168K
Free Swap: 1,533,952K


Another snippet of output...

Error Status
Array Size Add Mpy Div Sum of Products
1KB PASS PASS PASS PASS
4KB PASS PASS PASS PASS
16KB PASS PASS PASS PASS
64KB PASS PASS PASS PASS
256KB PASS PASS PASS PASS
1MB PASS PASS PASS PASS
4MB PASS PASS PASS PASS
16MB PASS PASS PASS FAIL
fp error

Have you got any suggestions about how to pursue that "FAIL"?

Thank you.
Click to expand...

I was able to get the same error, I'll post a new rev tomorrow.
 
General said:
General Schvantzkoph wrote:
Click to expand...

[snip]
Total RAM: 514,732K
Free RAM: 96,500K
Total Swap: 1,534,168K
Free Swap: 1,533,952K
Click to expand...

That's a reasonable suggestion, I'll add it to the next rev.
Click to expand...

Ahh good. It is substantially a mater of personal preference, but it
is my preference. ;-) Here is a snippet from my .bashrc file which
might be relevant.

,----[ /home/psd/.bashrc ]
|
| # Sat May 1 16:37:06 EST 2004
| # "'" adds thousands separator to ls, du, df
| # from LC_NUMERIC
| export BLOCK_SIZE=\'1
|
`----
Does this failure happen consistently? Did any of the memory tests
fail?
Click to expand...

I was assuming that your code was good and that I have hardware/firmware
problems. I do have stability problems here.

I've only just started playing with /sys_basher/, memtest does not
report any problems. Your other post indicates there might be a problem
in the code, but that does not rule out an ASUS problem as well. :-(
The way the floating point tests work is by doing the same
calculation twice, once entirely in registers and once where the operands
are fetched from memory. The program is brand new so I'm not 100%
confident that there aren't potential compiler issues, however it seems
unlikely because the test passed for every case except the last. Your CPU
will be getting hotter and hotter as it runs these tests so if you have a
cooling problem this is the kind of result that you would expect.
Click to expand...

I'm already under clocking the memory and the processor and only using
one of the cores.
Please
try the following experiment, set the scaling frequency to the lowest
value and rerun the test. I don't know if Mandriva has the CPU Frequency
Scaling Monitor applet, if it does you can use that the frequency down.
Click to expand...

Haven't seen it, but I haven't looked for it. What is it called?
If
it doesn't it can be done by writing the following files,


echo userspace > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo userspace > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
echo 500000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed
echo 500000 > /sys/devices/system/cpu/cpu1/cpufreq/scaling_setspeed

The number that you put into scaling_setspeed has to be equal to or less
than the lowest number in

/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
Click to expand...

No such files on my system.

->pwd; ls -al
/sys/devices/system/cpu/cpu0
total 0
drwxr-xr-x 2 root root 0 Jan 2 18:07 ./
drwxr-xr-x 3 root root 0 Jan 2 18:07 ../
-r-------- 1 root root 4,096 Jan 2 18:07 crash_notes

Should I recompile the kernel?
 
General said:
I've just put a new version of sys_basher on the web,

http://www.polybus.com/sys_basher_web/
[snip]
I've run sys_basher on a Core2 Duo. Athlon 64 X2, dual Xeon, and two
single core Athlon 64 systems using Fedora Core(s) 4,5 and 6 in both 32
and 64 bit modes. It's straight POSIX C so it should be able to compile
and run on any *nix system.
Click to expand...
[snip]
I'm using Mandriva 2007 on an Athalon 64 x2 with smp disabled.

Do you want any feedback?

Here are some snippets of output...

Total RAM: 514732K
Free RAM: 96500K
Total Swap: 1534168K
Free Swap: 1533952K

could that be reformatted like this?

Total RAM: 514 732K
Free RAM: 96 500K
Total Swap: 1 534 168K
Free Swap: 1 533 952K

Or this?

Total RAM: 514,732K
Free RAM: 96,500K
Total Swap: 1,534,168K
Free Swap: 1,533,952K


Another snippet of output...

Error Status
Array Size Add Mpy Div Sum of Products
1KB PASS PASS PASS PASS
4KB PASS PASS PASS PASS
16KB PASS PASS PASS PASS
64KB PASS PASS PASS PASS
256KB PASS PASS PASS PASS
1MB PASS PASS PASS PASS
4MB PASS PASS PASS PASS
16MB PASS PASS PASS FAIL
fp error

Have you got any suggestions about how to pursue that "FAIL"?

Thank you.
Click to expand...

I've fixed the problem with the sum of products test. The new rev is
1.0.4, please give it a try. I'm seeing some failures in the Disk tests
which I think are code problems, for now you should ignore any failures in
the disk tests.
 
Back
Top