[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: i386->powerpc cross compiler

Konstantinos Margaritis wrote:

> On Mon, 10 Apr 2000, Brendan J Simon wrote:
> > I've been thinking about this just recently.
> > We are developing an embedded product using the mpc860 and maybe a mpc8260.
> > There is the HardHat Linux distribution which is based on RPMs but I think you
> > have to buy a maintenance contract at US$5000 per developer per project.
> > I was thinking about making a Debian distribution for embedded powerpc
> > processors.  I don't know where to start but I figured that all the sources
> > would be the same as the powerpc workstations, just the compile flags need to be
> > changed (ie. -mcpu=860, -fsoft-float, etc).
> > I was thinking about native powerpc builds and i386 cross-builds.
> > Does this sound feasible ?
> After you have a working cross-compiler, your next step should be to
> install an autobuilder. However, many packages use configure scripts. For
> them to use a different compiler all that is necessary is to set
> CC=cross-gcc and the package will be produced for the correct
> architecture. I know it works because I have used it many times. The
> problem is the packages which are more hardcoded (no standard makefiles,
> etc.). But in the end you end up creating a new distro. Wrt this matter,
> I thought the kernel provided some sort of compatibility between
> PPC60x/7xx <-> MPC860 chips?

I believe that recent kerenls compiled for mpc860 can execute ordinary powerpc
binaries, especially now that floating point emulation exists in the kernel.
I guess I'd prefer to have a very small distrribution that is compiled with the
mpc860 flags, so that there is no unnecessary floating point emulation calls.

It's probably better that a Debian based distribution for embedded processors be
distributed in source form only (using the standard source) and just have specific
build scripts for the embedded targets.  Initially this would only work on powerpc
architectures but cross-compilation would also be very nice.  A binrary distribution
can be organised later if there is a reasonable demand.

Brendan Simon.

Reply to: