Re: PATCH: Fix broken PIO with libata

[Tejun Heo]

Jeff Garzik wrote:
> Kevin Radloff wrote:
> > On 5/16/06, Alan Cox <alan@xxxxxxxxxxxxxxxxxxx> wrote:
> > > On Maw, 2006-05-16 at 11:33 -0400, Kevin Radloff wrote:
> > > > However, I still have a problem with pata_pcmcia (that I actually
> > > > experienced also with the ide-cs driver) where sustained reading or
> > > > writing to the CF card spikes the CPU with nearly 100% system time.
> > >
> > > That is normal. The PCMCIA devices don't support DMA. As a result of
> > > this the processor has to fetch each byte itself over the ISA speed
> > > PCMCIA bus link.
> >
> > Hrm, as I recall that only started happening with ide-cs sometime in
> > the single digits of 2.6.x.. And note that it's only maxing out at
> > about 1.5MB/s. Should that saturate my laptop's 1.1GHz Pentium M
> > processor?
>
> Doing data xfer using PIO rather than DMA definitely eats tons of CPU 
> cycles.

Yeap, in addition, if doing real PIO (unbuffered by the HBA), the time 
it takes is soley determined by what PIO mode is in use.  It doesn't 
matter how fast the CPU is.  Faster CPUs only end up wasting more 
cycles.  :-(

--
tejun
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/