This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] Fix one-sized FP arrays returning on MIPS
We have such constructs in our Ada code, mainly in generic optimizer
functions, if the input or output dimension is one or two we'll have
such small arrays.
But anyway our 60 MIPS R12000 processors Origin 2000 died last week,
and the replacement (inflation & moore law adjusted :) will be around
200 AMD Opteron 248 so we won't use the GCC MIPS backend any more...
On Fri, 2004-06-18 at 09:42, Eric Botcazou wrote:
> > Not really, sorry. I really think the new behaviour follows the spirit
> > of the ABI. I agree that we generate poor code, but to be honest, the
> > output you quote isn't any worse than I expected it to be, given what
> > GCC currently does.
> As you wish. I can only say that I'm probably not as used as you to seeing
> this... interesting code. :-)
> > Is this sort of construct (one-dimensional arrays) really likely
> > to occur in performance critical code? Probably a naive question. ;)
> No idea. However, since Ada95 has generics, I'd think it is not uncommon:
> Max : Positive;
> package my_array
> a : array (1..Max) of Float;
> FYI, this construct (hence the ICE) appears in 4 testcases of the ACT
> internal testsuite.