Making, drinking tea and reading an opus magnum from Alexei G. Malinin:
[Charset ISO-8859-1 unsupported, filtering to ASCII...]
> Michael Shalayeff wrote:
>
> >Making, drinking tea and reading an opus magnum from Alexei G. Malinin:
> >
> >
> >>Hello.
> >>
> >>Can anybody tell me what is the idea of setting
> >>TX trigger to maximum? In this case we have
> >>no performance gain during servicing TX
> >>interrupts because, for example, when TX fifo
> >>is 64 bytes and TX trigger is 56 bytes and TX fifo
> >>is full - TX interrupt is activated after TX fifo
> >>transmits every 8 bytes if traffic is high.
> >>I think that setting TX trigger to minimum can
> >>give us performance gain because TX interrupt
> >>will be activated less frequently then in the
> >>above example and the kernel will have more time
> >>to service other tasks.
> >>
> >>
> >
> >8 chars at 115200 be about 1000 ints/sec -- not a problem.
> >the problem at hands is to make sure there is always smth
> >in the tx buffer while interrupt is being serviced thus
> >avoiding stalls due to delys for the intr handler to
> >push new data into the tx buffer.
> >
> >cu
> >
> >
> >
> ok but setting TX trigger to reasonably low value, e. x. 4 or 8 bytes,
> solves this problem but with a performance gain?!

what performance gain?
it's still gonna push out 115200 worth of data...
8 chars thus about 1ms may not be enough
to replenish the tx buffer (and also do many
other things elsewhere)

cu

--
paranoic mickey (my employers have changed but, the name has remained)