This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: powerpc patch 1 of 9
- From: Geoff Keating <geoffk at geoffk dot org>
- To: Alan Modra <amodra at bigpond dot net dot au>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: 22 Feb 2002 17:56:05 -0800
- Subject: Re: powerpc patch 1 of 9
- References: <20020220110029.GH1054@bubble.sa.bigpond.net.au>
Alan Modra <amodra@bigpond.net.au> writes:
> It's time I contributed back some of the fixes I've been making to
> gcc for IBM people working on PowerPC64 linux.
>
> This is part 1 of 9. The whole group has been bootstrapped and
> regression tested on powerpc-linux.
>
> After back-porting 3.1 handling of const_ints to 3.0, I noticed a
> number of places where ppc64 linux kernel code got worse. Tracked
> these down to sign extension of negative SImode constants.
>
> * config/rs6000/rs6000.c (u_short_cint_operand): Mask off
> irrelevant bits before testing with 'K' constraint.
> (includes_lshift_p): Mask irrelevant bits of SImode const_int.
> (includes_rshift_p): Likewise.
These seem to be optimising for cases that shouldn't be happening.
Can you find out where these constants are being generated and fix it
there instead?
--
- Geoffrey Keating <geoffk@geoffk.org> <geoffk@redhat.com>