RE: success building powerpc-405 & powerpc-750 using kernel 2.6.6

[Ankur Sheth <asheth at ixiacom dot com>]

> Ankur Sheth wrote:
> > FYI : We were able to build a version of the toolchain for 
> powerpc-405 &
> > powerpc-750 using crosstool-0.28-rc19.   It worked right 
> out of the box for
> > using gcc-3.3.3-glib-2.3.2.dat.  It also worked right away 
> when we changed
> > the Linux kernel version to 2.6.6 instead of 2.4.24.  So 
> the final versions
> > of the various tools in our toolchain are as follows :-
> > 	- gcc 3.3.3
> > 	- glibc 2.3.2
> > 	- binutils 2.15
> > 	- Linux kernel 2.6.6
> > 
> > We were also able to run the regression tests on both.  
> I've attached the
> > summaries for both with this email.  
> 
> Excellent!   (What, no sh4? :-)

sh4 is on the way, can't get rid of it yet :(  In process of running the
tests right now.  Will send out the results when done.

> But hang on, the failures on ppc750 for the following tests
> 
> gcc.c-torture/execute/920501-8.c
> gcc.c-torture/execute/930513-1.c
> gcc.c-torture/execute/980709-1.c
> gcc.c-torture/execute/990826-0.c
> gcc.c-torture/execute/ieee/920810-1.c
> gcc.c-torture/execute/struct-ret-1.c
> 26_numerics/complex_inserters_extractors.cc
> 26_numerics/complex_value.cc
> 27_io/ostream_inserter_arith.cc
> 
> (essentially anything that failed for ppc750 but not ppc405)
> seem suspicious.  Can you have a look at the detailed test log
> (buried in build-gcc/gcc/testsuite/...) for those tests,
> or maybe run a couple of them by hand, and see what's failing?

Will get back to you on those later in a separate email, haven't really
begun analyzing all the failures.  Just wanted to send out what we had so
far.

> It might be worth trying gcc-3.4.0 or gcc-3.4.1, by the way.
> It's pickier, but you might notice improved performance
> (especially if you can turn on the profile driven optimization!).
> Plus it has a bunch more testcases.

Wanted to be a bit conservative and stick with something that's been out
there for a while.  Nevertheless we'll try and give it a shot (won't be able
to do it right away, deadlines...:(   ).

> (Finally, if you get a chance, do future runs with 
> crosstool-0.28-rc24,
> it's nearly final)

Ok. We'll try it out for the next version of the tool-chain.  I think of the
version that we just built as a beta and not the final thing.

ankur

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com