strange stats

strange stats

[Giacomo Tesio]

Hi, I've noted a strange behaviour in stats during the system compilation.

When I run `mk install` at the rio terminal, stats shows that load,
context and syscalls grow to 100% untill the window buffer is full and
then go down to the previous values. If I scroll down the window, the
stats grow up again until the the window is full, than drop again.

Instead if I send stdout to a file, that stats remain 100% till the
compilation completes.

Such behaviour makes me wonder if writing logs to the console is
blocking the compiler.
Is this the expected behaviour?


Moreover, does the compiler benefit from multiple cores?
`xl top` at xen dom0 shows the cpu usage of the 9front vm at 112% but
I did expect it around 400%, since I assigned 4 cores to the system.


Giacomo

Re: [9front] strange stats

[cinap_lenrek]

yes, unless the window is in "scroll" mode (middle mouse button menu),
the writer will block when the window filled up. read rio(1)

and yes, it benefits from multiple processors (you can set NPROC
environment variable before mk to get more stuff run in parallel).

but nobody knows if cpu is the bottleneck on your xen box. remember,
you'r probably the first running 9front on xen-hvm :)

--
cinap