head: MIPS: A workaround for the R4000 divide/shift errata

Eric Christopher echristo@redhat.com
Sat Feb 21 13:45:00 GMT 2004


> FWIW, I'd happy to see the patch go in with or without a separate option.
> Let's see what Eric thinks.
> 

Hrm. Of two minds...

> I think the -mfix-* options come into their own if it's likely (or at least
> possible) that the bugs will be fixed in a later revision of the processor.
> That's certainly not the case for the r4000.  So I don't see a problem with
> using:
> 
> #define TARGET_FIX_R4000 (TARGET_MIPS4000 && !TARGET_MIPS16)
> 
> for the time being and maybe adding an option later if it causes trouble.

Basically right now we've got -mips3 defaulting to TARGET_MIPS4000 and
so we'd be adding this to the default mips3 (i.e. mips64-*) target where
we haven't before.

Now if most people are actually running on r4000 processors that's fine,
but I somehow doubt it. Then again, we do specify that it's going to be
an r4k target so it should be there.

How about adding the option so that users can manually turn it off for
faster performance (or other mips3 processors can turn it off by default
in their configurations) and leaving it on as Richard suggests above?

-eric

-- 
Eric Christopher <echristo@redhat.com>



More information about the Gcc-patches mailing list