[PATCH] e7501 register pressure

[Stefan Reinauer]

See patch

[ron minnich]

There's a way to test this?

ron

[Stefan Reinauer]

On 2/23/10 12:45 AM, ron minnich wrote:
> There's a way to test this?

Not for me... it's 2 boards out of 20+

But we did the same conversion for i945 and i830 before, so it's
generally a good idea.

The smarter thing would be to convert the board to CAR, but if I had to
choose between those two alternatives, getting rid of a memory
indirection for moving 0x80000000 to a register seemed the safer path.

Does anyone have the hardware?

[ron minnich]

On Mon, Feb 22, 2010 at 4:03 PM, Stefan Reinauer  wrote:
> Does anyone have the hardware?

I doubt anyone does. The question here is, if no one can test it, are
we better off making a change we can not test, or removing the
support. If there is hardware so old that we can not test, i'm more
comfortable pointing people to an old "known good" version, rather
than let them think that new untested code is good.

This change should work, however.

Acked-by: Ronald G. Minnich

ron

[Uwe Hermann]

On Mon, Feb 22, 2010 at 04:30:20PM -0800, ron minnich wrote:
> On Mon, Feb 22, 2010 at 4:03 PM, Stefan Reinauer  wrote:
> > Does anyone have the hardware?
>
> I doubt anyone does. The question here is, if no one can test it, are
> we better off making a change we can not test,

Yes.
> or removing the support.

No.

We have to live with the fact that we will never be able to test _all_
changes on hardware all the time. That doesn't mean we should stop
working on the code. We've been doing tree-wide changes in recent weeks
and months on a regular basis (high tables, cbfs, kconfig,
tinybootblock, and many many more) which have a _lot_ higher potential
for breaking (lots of!) boards than this patch. That doesn't mean we should
not have done those changes. At most we should try to carefully review
more of our patches and encourage more users to test on their boards
more often.

Yes, bugs may slip in, but they will be noticed next time someone tests
on hardware (or a user reports a bug on IRC or the list), and then we can
bisect and fix the bug. No big deal.
> Acked-by: Ronald G. Minnich

Acked-by: Uwe Hermann

Uwe.