RFC: powerpc-rtems multilibs

[Thomas Doerfler]

Hi Ralf,

> Hi,
> 
> Currently, powerpc-rtems-gcc-3-0.4 uses these multilib-variants:
> 
> # powerpc-rtems-gcc --print-multi-lib
> .;@mrelocatable-lib at mno-eabi@mstrict-align
> nof;@msoft-float at mrelocatable-lib@mno-eabi at mstrict-align
> le;@mlittle at mrelocatable-lib@mno-eabi at mstrict-align
> ca;@mcall-aix at mrelocatable-lib@mno-eabi at mstrict-align
> und;@fleading-underscore at mrelocatable-lib@mno-eabi at mstrict-align
> le/und;@mlittle at fleading-underscore@mrelocatable-lib at mno-eabi@mstrict-align
> nof/le;@msoft-float at mlittle@mrelocatable-lib at mno-eabi@mstrict-align
> nof/ca;@msoft-float at mcall-aix@mrelocatable-lib at mno-eabi@mstrict-align
> nof/und;@msoft-float at fleading-underscore@mrelocatable-lib at mno-eabi@mstrict-align
> nof/le/und;@msoft-float at mlittle@fleading-underscore at mrelocatable-lib@mno-eabi at mstrict-align
> 
> My question is: Do these actually give sense for RTEMS or can some of
> them be eliminated rsp. shall others be added?
> 
> Remember that we can set the conventions/standards to be used by
> RTEMS-gcc, so "sticking to some arbitrary defaults", as
> powerpc-rtems-gcc currently does, might be worth reconsideration.
> 
> AFAIS, currently, only nof (==softfloat) and "default" are used by the
> public BSPs. Little endian does not seem to be used, however would give
> sense if the ppc-ports support it (AFAIS, they at least claim to do so,
> but there doesn't seem to exist a BSP which actually uses it.)
> 
> How about "call-aix" and "fleading-underscore"?
> I don't know what these actually do, but they don't seem to be used by
> RTEMS, AFAICT.

I would assume that "leading_underscore" will add an 
underscore to all symbols and "call-aix" adapts gcc to AIX 
calling conventions. AFAIK both configuration options are 
irrelevant for RTEMS.

I would also tend to remove little-endian support, because the 
idea behind it is not a straight little-endian memory layout, 
but a rather cheap trick to make applications belive they 
would run on a little endian machine. It has mainly been added 
to the PPC arch to support WinNT, OS/2 and WinCE. Using little 
endian with PPC requires many changes to the drivers and the 
OS core, so the way to PPC little endian support in RTEMS 
would be very long.

Any other opinions on that?

Thomas.

> 
> I.e. I am in favor of removing the "ca" and "und" multilib variants,
> unless there is a technical reason to keep them. This would
> significantly reduce the size of powerpc-rtems-gcc and the time required
> to build multilibs.
> 
> Opinions, comments?
> 
> Ralf
> 
> 


--------------------------------------------
IMD Ingenieurbuero fuer Microcomputertechnik
Thomas Doerfler           Herbststrasse 8
D-82178 Puchheim          Germany
email:    Thomas.Doerfler at imd-systems.de
PGP public key available at: http://www.imd-
systems.de/pgp_key.htm