Re: [patch 2.6.20-rc1 4/6] PXA GPIO wrappers

[David Brownell <david-b@pacbell.net>]

On Friday 29 December 2006 6:15 pm, Nicolas Pitre wrote:
> On Thu, 28 Dec 2006, David Brownell wrote:
> 
> > Phillip:  is this the final version, then?  It's missing
> > a signed-off-by line, so I can't do anything appropriate.
> > 
> > Nico, your signoff here would be a Good Thing too if it
> > meets your technical review.  (My only comment, ISTR, was
> > that gpio_set_value macro should probably test for whether
> > the value is a constant too, not just the gpio pin.)
> 
> I don't think so.  Expansion of GPIO_bit(x) is pretty simple even if x 
> is not constant.  That probably makes it still less costly than a 
> function call.

I was more concerned with the "value" ... that expands to a conditional,
which is likely to cost a couple more instructions regardless, which in
space terms competes with a function call.

But I concluded much the same thing when I did that experimental
conversion ... not because I was comparing the space for conditional
vs funcall, but because the existing code already had the conditional.
If more code savings can be had later, so be it ... no rush for now.

- Dave