> Hi all,
> 
> I am running RTLinux 3.1 on Linux 2.4.18 (+Gearheart patch) on a Pentium
> 3 933 MHz machine.  Running the sample program "measurements" gives me
> numbers I cannot explain:
>

The simple answer is that there are hardware combinations that are not suited
for RT.
 
> ... start tasks, good latency ...
> 
>     min:        0, max:     1664
>     min:        0, max:     2080
>     min:        0, max:     1856
>     min:        0, max:     1472
>     min:        0, max:     1472
> 
> ... start X/GNOME, bad startup latency, levels off ...
> 
>     min:        0, max:   879616

This may indicate that a video card is provoking a PCI-bus stall - there
are video-cards that don't check if there is video-ram available to write to
befor doing so as this is generally the case - so to get a few more X-stones
out of the card they save the check and risk that the PCI-bus blocks whyle 
waiting for the write to complete. If you have the posibility to plug in
some other VGA card to see if that eliminates it that would be helpfull.
it is rather hard to get reliable infos on which cards do something like
this and which don't . 

>     min:        0, max:    24096
>     min:        0, max:     1312
>     min:        0, max:     1824
>     min:        0, max:    27872
>     min:        0, max:    36928
>     min:        0, max:    41440
>     min:        0, max:    49984
>     min:        0, max:    23584
>     min:        0, max:     8064
>     min:        0, max:    15296
>     min:        0, max:     2208
>     min:        0, max:    25024
>     min:        0, max:     6912
>     min:        0, max:    12032
>     min:        0, max:     6848
>     min:       32, max:     2144
>     min:        0, max:     4640
>     min:        0, max:     2240
>     min:        0, max:     3936
>     min:        0, max:     2976
>     min:        0, max:     6048
> 
> ... start "while :; do cat /dev/null; done", periodically bad latency ...
> 
>     min:        0, max:    35104
>     min:       32, max:    34272
>     min:        0, max:    41152
>     min:        0, max:   652608
>     min:       32, max:    35360
>     min:       32, max:    32000
>     min:        0, max:    36544
>     min:        0, max:   801472
>     min:        0, max:    33856
>     min:       64, max:    30560
>     min:        0, max:    33984
>     min:        0, max:   758240

Are these values with X-running or without X ?

> 
> ... kill X, horrible latency spike ...
> 
>     min:       64, max:    41440
>     min:        0, max:    34176
>     min:        0, max:   656128
>     min:        0, max: 997700192
>     min:   992320, max:   997472
>     min:       32, max:   997504
>     min:        0, max:    30496
>     min:        0, max:    32288
>     min:        0, max:    29248
> 
> My hope was to see consistent numbers <= 50,000 ns.  I assume that there
> is either an error in my configuration or in my understanding of the
> numbers that "measurements" is giving me.  Can someone shed some light
> on this?  I have played with the flags in rt_proccess.c but have similar
> results.  Also, what kind of latency should I expect on a machine like
> this?  The RTLinux papers I have read use older computers.

On a K7 800 I get reliable < 24us
   PIII 800 < 34us

load average of something like 30 while ping flooding the box...

But on some Pentium processors you get very bad results

  P III 1GHz Copermine > 10 milli seconds
  P III Laptop > 1milli second

The AMD-Duron (only checked 4 different systems though) show the lowest
jitter values of all PIII class systems I have found - the point is the CPU
is one criteria RAM,Montherboard and some flaky peripherals have large 
influence on the RT-performance as well.

hofrat
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
--
For more information on Real-Time Linux see:
http://www.rtlinux.org/

Reply via email to