powerpc switches
Aldy Hernandez
aldyh@redhat.com
Fri Jan 17 17:21:00 GMT 2003
> I've just had a confusing time investigating why a ppc compiler
> accepted -mcpu=8540 and -mabi=spe, yet refused to recognise the builtins.
> Turned out it was configured wrongly and TARGET_SPE and TARGET_SPE_ABI
> were hardwired to false.
He he. That was Zack and David's doing for reducing the size of the
compiler on configurations that wouldn't need spe/altivec/etc.
I'm ok with this patch, but I don't agree with this:
> if (rs6000_cpu == PROCESSOR_PPC8540)
> ! rs6000_isel = 1;
>
> /* If we are optimizing big endian systems for space, use the load/store
> multiple and string instructions. */
> --- 575,585 ----
> }
>
> if (rs6000_cpu == PROCESSOR_PPC8540)
> ! {
> ! rs6000_isel = 1;
> ! if (!TARGET_SPE)
> ! warning ("not configured for spe extensions");
> ! }
Motorola has said future (non SPE) chips may have the isel
instruction, so this patch wouldn't permit us to do -misel=yes on such
a target.
Aldy
More information about the Gcc-patches
mailing list