Powerpc IRQ handling breaks strict EABI compliance

Sergei Organov osv at javad.ru
Thu Feb 13 15:25:54 UTC 2003


Joel Sherrill <joel.sherrill at oarcorp.com> writes:
[...]
> My reading of the output of powerpc-rtems-gcc --print-multi-lib shows that
> there are no libraries for -meabi which means there is no valid way to link
> an EABI program.

I believe code compiled with -mmo-eabi couldn't be linked together with code
compiled -meabi only due to the differences in stack alignment requirements
(16 vs 8). However, it could be linked with code compiled with '-mno-eabi
-msdata=eabi' that gives all the EABI advantages but leaves stack alignment to
be 16.

> 
> I am beginning to believe (1) this will not be addressed in 4.6.0

Don't think anybody expected it will.

> and (2) that a 2nd toolset for EABI may be needed.

If EABI has no disadvantages, why don't just turn it on finally (i.e., don't
build anything without EABI)? Maybe on per-BSP basis.

> The C/C++ powerpc-rtems RPM is already 55 MB for C and 70 MB for C++ ! We
> simply can't make it any larger to have EABI and no EABI libraries.

I've already noticed that when attempted to build version 3.3 of ppc-rtems --
it took 83 minutes on 2GHz Xeon :-( Even without EABI I think something is
wrong about it. I can't imagine any users (those who have to build all BSPs
for testing purposes aside as they are experienced enough to get whatever they
wish) will use even 20% of all the multilib variants. Personally I don't care
much as I've always used my own t-ppcgas (t-rtems now), but anyway.

Isn't it enough to have single multilib variant for every BSP?

-- 
Sergei.




More information about the users mailing list