Re: how much work is it to port a backend from 2.95.3 to head: wasRe: define_delay working only when optimization on!

[Hans-Peter Nilsson <hp at bitrange dot com>]

On 20 Mar 2003, Spundun Bhatt wrote:
> On Tue, 2003-03-18 at 18:35, Hans-Peter Nilsson wrote:
> > On 18 Mar 2003, Spundun Bhatt wrote:
> > > I just added a define_delay statement to my machine description file
> > > (derived from rs6000-2.95.3).
> >
> > I suggest you use the CVS trunk, updating once in a while when
> > gcc-regression shows no regressions, carefully saving the tree
> > before you update.
> >
> The developement for this backend started some 1.5 years back. We had
> decided to use the motorola altivec gcc 2.95.2 . When I started working
> on this 6 months back, some porting had already taken place. Now quite a
> lot of effort has already gone in to porting this backend, I was
> wondering how much work it is to port a backend derived from a
> motorola-derived altivec version of backend merged with 2.95.3 changes
> and then worked on for another 4 months :). Are there any major api
> changes

There are some, yes.

> or could I just put my backend directrory in the cvs head and
> put another week and get rolling?

Usually that should fit for a random port, but...

> I know that motorola guy made some
> changes outside the backend config directory so those things I will have
> to fix atleast.

If this "port" is about ppc stuff for the "motorola-derived
altivec" I guess you're up for major clashes.  You'll notice if
you try to use CVS HEAD that there's already syntax-incompatible
altivec support.  Actually, IIRC it's generic vector mode
support, but I haven't paid attention to the details.

brgds, H-P