[Patch, fortran] PR64952 - Missing temporary in assignment from elemental function

Dominique d'Humières dominiq@lps.ens.fr
Wed Feb 11 18:02:00 GMT 2015


Dear Paul,

> Le 11 févr. 2015 à 17:57, Paul Richard Thomas <paul.richard.thomas@gmail.com> a écrit :
> 
> Dear Dominique,
> 
>> The patch works as advertised! I have two remarks:
> 
> Of course it does :-)

;-)

>> 
>> (1) AFAIU there is no need for a temporary for
> 
> Indeed not. I cannot see how that can be avoided without a much more
> elaborate patch. Frankly, I do not see that it would be warranted.
> Much better some extra temporaries in corner cases, than wrong code.

Agreed.

>> (2) You wrote: << However, this works and has no perceivable effect on Polyhedron timings. >>. This is hardly a surprise since the Polyhedron tests don't use any elemental procedure.
> 
> :-) You might have gathered that I didn't check! Within errors, it
> didn't affect compile times either.

AFAICT, I have only one test in my archives that time elemental functions: pr40581 for which I don’t see any difference with/without your patch.

Looking at bugzilla it seems that Fran Martinez Fadrique is a heavy consumer of elemental procs (see e.g. pr52332).
May be you can ask him to do some timing of his code with/without your patch.

> Thanks for the feedback

You’re welcome!

Dominique



More information about the Gcc-patches mailing list