On Fri, 9 Jun 2000, Hannu Savolainen wrote:
> On Thu, 8 Jun 2000, Britton wrote:
>
> >
> > The guide of 4front's web page says total_buffer_size/2. I know the
> > kernel buffer size used to be defined at kernel compile time. How does
> > one go about determining it now? DSP_GETBLKSZ?
> Wy would you need this information?
I'm letting my users specify the fragment size, and I try to check things
like whether what they specify is out of range before actually making the
syscall.
> If you like to get as large fragments as possible just set the fragment
> size to very big. The driver will then select the largest one it could
> support.
Ok, I didn't know it would do that. I guess there is no real reason the
user needs to know if their large fragment size was not actually set (they
already get a warning if the fragment size they request is big enough that
it will likely cause latency). That behavior might be something to add to
the 4front page.
> However changing the fragment size bigger than the default doesn't give
> any benefit. You can always write/read multiple fragments at time to get
> the same effect.
>
> > Does DSP_GETBLKSZ still
> > fix the block size so that the device must be closed and opened again in
> > order to set the fragment size?
> This is the current and future behaviour.
Thanks,
Britton
