This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] patches for fixed-point support
On 1/10/09, Denis Chertykov <chertykov@gmail.com> wrote:
> 2009/1/10 Sean D'Epagnier <geckosenator@gmail.com>:
>> Hi,
>>
>> I have 3 patches.
>>
>> gcc-fixedpoint-avr.patch -- support for fixed-point in avr backend
>> gcc-fixedpoint-varasm.patch -- fixes what I believe was a bug but only
>> for fixedpoint on 8bit cpus but is not specific to avr
>> gcc-fixedpoint-dwarf2support.patch -- support for fixed point in dwarf
>> output
>>
>> Please let me know how I'm doing with these patches.
>
> Are you have a signed FSF copyright assignment for GCC ?
>
I am trying to do this, I got this request also for my fixed point
patches for GDB. I sent am email to the fsf, and I never got a reply
or anything. Maybe I should email them again. They have my mailing
address, so maybe they have to mail me actual paperwork?
>>
>> gcc configured with:
>> ./configure --target=avr --enable-languages=c --disable-libssp
>> --enable-fixed-point
>>
>> Denis,
>>
>> I tried both of your suggestions for define_expand and define_split,
>> and neither worked.
>
> What kind of error you have with define_expand-only version ?
>
Usually something like:
error: unrecognizable insn:
(insn 6 5 7 3 avrtest.c:14 (set (reg:SF 42 [ z.1 ])
(fract_convert:SF (reg:QQ 43 [ x.0 ]))) -1 (nil))
I suspect gcc is too smart for its own good, and folds expressions like:
(fract_convert:X (fract_convert:Y Z))
into (fract_convert:X Z)
>
>> Maybe you can try it?
>
> Ohh, sorry.
> May be Anatoly Sokolov <aesok@post.ru> can try.
>
I hope so too. I'm not even sure this is what I want, maybe this
stuff should remain in a library which would make things slightly
smaller sometimes.. and bigger other times. It's hard to make gcc do
the best thing.
Sean
> Denis.
>