Patch fixing 3.3 bug PR 9745 and PR 10021

[Marcus Meissner]

> I am concerned about my ability to work on this problem, since I don't
> have access to the hardware involved, and I am finding it very time
> consuming.  Time I spend on this problem is time I can't spend on
> reviewing patches sent to gcc-patches.
> 
> I think the next step here is testing of my latest patch.  We need
> performance testing which I can't do.  We also need some execution
> testing to see if it fixes the bugs 9745 and 10021.  I could do 9745 by
> building an uberbaum cross compiler, but that will take a lot of time. 
> It would be faster if someone else could do this for me.  If this patch
> works, then I think this is safe for the 3.3.1 branch and then we can
> close the two PRs.  We still have long term issues with
> loop_regs_update/record_base_value which will effect performance and
> possibly also correctness.

I tested it against my testcase for 10021 on powerpc-linux with current
gcc 3.3 branch + your loop patch.

The loop patch attached to your mail compiled with -O2 does not change
the assembler output and does not fix the testcase in 10021.

Ciao, Marcus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 481 bytes
Desc: not available
URL: <https://gcc.gnu.org/pipermail/gcc/attachments/20030709/6b9d884e/attachment.sig>