[PATCH 1:2] [MTD][NAND]omap: Adding support for nand prefetch-read and post-write, in MPU mode.

[David Brownell]

On Friday 10 July 2009, vimal singh wrote:
> +static void omap_read_buf8(struct mtd_info *mtd, u_char *buf, int len)
> +{
> +       struct nand_chip *nand = mtd->priv;
> +       u_char *p = (u_char *)buf;
> +
> +       while (len--)
> +               *p++ = __raw_readb(nand->IO_ADDR_R);
> +}

Better as __raw_readsb() yes?  Or maybe ioread8_rep().

... speaking of which, maybe the MTD layer can finally
default to using the accelerated block PIO calls, now
that ioreadX_rep() is supposed to work on all arches...

Last time I measured, that change alone was worth
something like a 10% speedup.  Maybe more.