-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hi Ian and fellow GCC'ers
thanks for the info. Compiling "libgcc2.c" for MIPS (-mips1) works OK,
I mean "divdi3" produces correct results.
I'll get back to this in a couple of hours (at my home PC), but i
think i had overseen something important here.
I was coding the DLX backend based on one for nios-2 (very nice
single/double float emulation routines there; they
have written their own). However, both the nios2 backend and an old
DLX backend that i used for reference
(for GCC-2.7.2.3) didn't include patterns for 64-bit moves (like the
MIPS "movdi"). Due to this and some other
observations, I tend to think that neither backend (of these two) was
ever tested for "double" integer arithmetic.
Thus, 64-bit integer arithmetic does not work on my processor and this
is due to the fact that i have not included
patterns (or code emission routines in "dlx.c") for handling 64-bit
moves. Am I missing something else here
(not in detail but any coarse grained info)?
thanks in advance
Nikolaos Kavvadias
PS: libgcc2.{c,h} looks OK, both gcc-3.4.1 and current CVS versions
(actually 16-Oct-06 version)
Ian Lance Taylor wrote:
> Right, libgcc2.c is quite stable.
>
> In a normal build it is the first complicated code to be compiled
> with the new compiler, so it is a normal place for reload problems
> to arise.
>
> Ian
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (MingW32)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFFNSgpMPiy0tCWlz4RAs0uAJ0Y5rGfP27Gbbo2wtXq5u8TKF31WwCeLbZA
55TgLRUnr4yWd+KLE7A2Yyc=
=MR8j
-----END PGP SIGNATURE-----
