This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: SH: use register names instead of numbers
- To: Joern Rennecke <amylaar at redhat dot com>
- Subject: Re: SH: use register names instead of numbers
- From: Alexandre Oliva <aoliva at redhat dot com>
- Date: 23 Nov 2000 01:09:55 -0200
- Cc: gcc-patches at gcc dot gnu dot org
- Organization: GCC Team, Red Hat
- References: <200011230258.eAN2wc018539@tantra.cygnus.co.uk>
On Nov 23, 2000, Joern Rennecke <amylaar@redhat.com> wrote:
> Well, the REG names are usually pretty short, most have only one or
> two chars before the _REG. Morever, it's pretty close to plaintext:
> e.g. T_REG is the T register.
> OTOH ALIGN_UNSPEC is not a directive to make unspecs aligned. Rather,
> it's an unspec that is used for alignment.
Just like ALIGN_UNSPEC is the align unspec.
Of course, you could read both as the register named T or the unspec
used for alignment, but I fail to see a difference.
I'm doing the change, anyway.
>> I actually prefer them all as suffixes, but it's your call.
> It not only differentiates them, it also makes it easier to read
> the name of the symbol to get the meaning.
Ok, I'll got for UNSPECV, then.
>> >> + (BREDIR_UNSPEC 4)
>>
>> > I'm having a hard time to associate this with block_branch_redirect.
>> > It would be easier if it was called UNSPEC_BBR or UNSPEC_BL_BR_REDIR
>>
>> Please pick one of these, and I'll do the change.
> UNSPEC_BBR.
Fixed.
>> >> + (CALLER_UNSPEC 10)
>>
>> > I don't know what this is, and I couldn't find it anywhere in the code,
>> > either.
>>
>> Oops. It's part of a patch I haven't posted yet. I may keep it out,
>> or book it in advance, whatever you prefer.
> Keep it out for now, please.
Aye :-)
>> Is changing FPSCR_REG to the mode in which it's used in all other
>> places a semantic change? I take it as fixing a typo.
> It is a semantic change. It can cause tests not to match, expanders to
> get out of sync with insns, or new combinations being attempted.
Ok. Since I suppose you're probably going to bed in a quite near
future, may I at least consider pre-approved a patch that only changes
register and unspec numbers to names, and nothing else?
--
Alexandre Oliva Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist *Please* write to mailing lists, not to me