On 15/05/07, Adrian McMenamin <adrianmcmenamin@xxxxxxxxx> wrote:
> On 15/05/07, Heikki Orsila <shdl@xxxxxxxxxx> wrote:
> > I still noticed some type cleanups, and possibly a good debug message..
> >
> > On Mon, May 14, 2007 at 11:14:55PM +0100, Adrian McMenamin wrote:
> > > +/* spu_memload - write to SPU address space */
> > > +static void spu_memload(u32 toi, void __iomem * from, int length)
> > > +{
> > > + u32 __iomem *froml = from;
> > > + u32 __iomem *to = (u32 __iomem *) (SPU_MEMORY_BASE + toi);
> > > + int i, val;
> >
> > You should use "void *from" because it's not IO memory. Also, use "u32 val"
> > since that is the native type of "u32 *from".
> >
> > > + length = DIV_ROUND_UP(length, 4);
> > > + spu_write_wait();
> > > + for (i = 0; i < length; i++) {
> > > + val = *froml;
> > > + writel(val, to);
> > > + froml++;
> > > + to++;
> > > + if (i && !(i % 8))
> > > + spu_write_wait();
> > > + }
> > > +}
> >
> > It seems a small simplification is possible:
> >
> > /* remove first spu_write_wait(); */
> > for (i = 0; i < length; i++) {
> > if (!(i % 8))
> > spu_write_wait();
> > val = *froml;
> > ...
> > }
> >
>
> No, that would be dangerous because we don't know what state the FIFO
> is in when this is called.
>
Sorry, should think before I post!
_______________________________________________
Alsa-devel mailing list
Alsa-devel@xxxxxxxxxxxxxxxx
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
