This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH][ARM] Thumb2 constant loading optimization


On Tue, 2011-04-12 at 11:14 +0100, Andrew Stubbs wrote:
> [Ignoring the other issues for now ...]
> 
> On 12/04/11 11:02, Richard Earnshaw wrote:
> > Also, your change to use a double-letter sequence beginning with 'j'
> > means any hand-written inline assembly code using a single 'j' will
> > break (that's a backwards compatibility issue for users); is there
> > really no other letter that can be used to prefix the operation?  Also,
> > think very carefully about whether your new constraints should be
> > internal only or really public.  I think it would be better to leave 'j'
> > as is, and add some more P-prefixed constraints that are internal only
> > for the additional cases.
> 
> The constraints were part of the patch I expected to have to fix up 
> following review. :)
> 
> I chose to modify the 'j' constraint because it was not documented, last 
> I looked, and the new use was related to the meaning of 'j'. I thought 
> this might make the patterns a little easier to read. Do we have to 
> assume that all constraints might be used by third parties, documented 
> or otherwise?
> 

It may not be in invoke.texi, but it isn't marked @internal in
constraints.md.

> I can switch to using P-prefixed constraints, of course.
> 

I think that would be better.  I think there's a reasonable argument for
keeping 'j' /and/ having it 'public'.

R.

> Andrew
> 



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]