Re: [avr-libc-dev] Considerable Compiler improvement needs testing

[E. Weddington]

Andy Hutchinson wrote:

> I am unaware of any negative feedback on the change.
>
> I ran MODES_TIEABLE_P together with Roger's latest "rtx_cost" patch. 
> Together I was seeing reduction from 0x3746 to 0x3462 on my testcode. 
> Of equal importance is that the code is becoming much cleaner and 
> makes it much easier to spot and implement other areas of improvement. 
> I dont like to count the number of times I have tried an "obvious" fix 
> and the whole code when pear shaped!

This is fantastic!

BTW, do you have a link to Roger's "rtx_cost" patch? Has that been 
committed yet to HEAD?

> I have posted bug report against documentation for MODES_TIEABLE_P.
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19815
>
> Although this is DOCUMENTATION BUG  I think we should be able to use 
> as a marker.(Advise - If somebody knows that there should also be a 
> target bug report then I can add).
>
> Can some kind sole submit a patch for MODES_TIEABLE_P
>
> we have
>
> #define MODES_TIEABLE_P(MODE1, MODE2) 0
>
> we need
>
> #define MODES_TIEABLE_P(MODE1, MODE2) 1
>
> Pretty good code metrics on this one!


Marek, perhaps you do this?

Thanks
Eric