Joseph Sarkes writes:
|> 
|> System is a 233 MHz multia.
|> 
|> I found much to my surprise, that by monitoring the 
|> make build results using tail -f buildlog to the console
|> was eating up around 50% of my cpu time! If I rlogin from
|> another machine it is miniscule. Did I miss something in
|> configuring my multia? Am I using some deprecated console
|> system or something? Needless to say, I was VERY surprised.
|> 

If the console works like I think it does, when doing
console serial IO you are calling a purely polling
based code path.  It literally waits for each character
to leave the UART before it'll send another one.

So, wasn't someone mentioning kernel profiling before?

Anyone up for porting IPROBE?


-- 
Dave Cherkus ------- UniMaster, Inc. ------ Contract Software Development
Specialties: UNIX Internals/Kernel TCP/IP Alpha Clusters Performance ISDN
Email: cherkus@UniMaster.COM  When the music's over, turn out the lights!