Nembered Paramenters accessibility.

This is my kernel version.

cnc@cnc-desktop:~$ cat /proc/version
Linux version 2.6.32-47-generic (buildd@roseapple) (gcc version 4.4.3 (Ubuntu 4.4.3-4ubuntu5.1) ) #109-Ubuntu SMP Tue May 7 02:03:05 UTC 2013
cnc@cnc-desktop:~$

bigalex

Hi Andy.

Sorry for my misunderstanding .
You gave me already this indication.
I'm trying ........

bigalex

Hi People.

Finally I've installed LinuxCNC from the Master branch.
I made the HAL latency test and the result is much worst than before with the Hardy 8.04 installation.
I've already implemented the isolcpu=1 as per indication here.
The numbers are very high especially for the servo thread more than 400000ns !!!
I know that on some hardware the Lucid 10.04 is not a good solution.
In any case I will proceed with Andy solution and try to understand how it works.

Thanks .

bigalex

Hi Andy.

With the Hardy 8.04 installation I did not apply any modification/fix for the SMI.
So the situation "should be the same" except for the Ubuntu Linux version.
I will try to understand better how to fix the SMI issue as you mentioned.
Running the latency without any graphic stress the numbers are fine but when I open Firefox immediately the servo thread jitter
rose above 400000ns and sometimes even more than 500000ns.
The isolcpu=1 is done correctly because I can see it at the GRUB boot list.
Anyway I will try to continue this test .
Thank you so much.

bigalex

I will try to understand better how to fix the SMI issue as you mentioned.
Running the latency without any graphic stress the numbers are fine but when I open Firefox immediately the servo thread jitter
rose above 400000ns and sometimes even more than 500000ns.
The isolcpu=1 is done correctly because I can see it at the GRUB boot list.

A lot of older machines run far better on the non SMP kernel in 8.04
Many will not even install 10.04
It is nothing to do with SMI, you would have had spikes ruining your figures with 8.04 if that was the case.

Unless you have one of the early dual core P4s, I doubt you need isolcupu either,
If you do it is isolcpus=1 ie. plural.
If you pass an unknown kernel line for the most part you get no warnings it just does nothing and you would have to delve through kernel log to find any mention of it.

This is why I don't use glade and either use pyVCP or C++ GUIs, my workshop machines are single core P4s and gurgle along perfectly happily with 9K max jitter on 8.04

regards

Hi ArcEye.

In my case the isolcpu=1 on Hardy 8.04 gave me better numbers than no isolcpu fix .
My computer has an integrated GPU so I know that this is a part of the problem .
But because it is a "small server pc" hardware architecture (Fujitsu Siemens Primergy Econel)
it has only one PCI slot that I'm using for the parallel board for steppers .
For sure the latency numbers were much better with the Hardy installation (servo thread mainly).
As you like pyVCP me too , but this have , from my point of view much more limitations on the HMI side.
I will try to fix the SMI issue because I've verified that the chipsets on my mother board are listed as "affected" .
I will feedback you on the next result.

bigalex