This is the mail archive of the
mailing list for the GCC project.
Re: RFA: DSE replace_read revisited
- From: Kenneth Zadeck <zadeck at naturalbridge dot com>
- To: Eric Botcazou <ebotcazou at libertysurf dot fr>, gcc-patches at gcc dot gnu dot org, zadeck at naturalbridge dot com, rsandifo at nildram dot co dot uk
- Date: Wed, 07 Nov 2007 07:59:52 -0500
- Subject: Re: RFA: DSE replace_read revisited
- References: <email@example.com> <firstname.lastname@example.org> <email@example.com>
Richard Sandiford wrote:
> Eric Botcazou <firstname.lastname@example.org> writes:
>> I'm a little reluctant to introduce this machinery at this point. So
>> I'd try to fix the wrong-code code as long as extract_low_bits is not
>> required; if it is, I'd punt for now.
>> What do you think?
> The reason I didn't do that is that it would mean adding even more
> special cases. GCC has a history of bolting on "except for this
> special case" conditions, which makes the code even harder to work
> with when you next need to change it. (And we might we need to change
> this code during the 4.3 maintanence period.)
> IMO, adding extract_low_bits makes the DSE code conceptually simpler,
> and I imagine is what Eric and Kenny would have used from the outset
> had extract_low_bits existed at that point.
> Kenny, what do you think?
The truth is that eric christopher and i completely underestimated how
hard removing read after writes really was. Had we understood what we
were getting into, i would have gone to either you, iant, or bonzini
from the start.
in fact my plan for reviewing this patch is to ask bonzini or one of the
middle end maintainers to review it when this spu issue is resolved (or
else to just trust you). Even after writing the interference graph
scanner, this is still way over my skill level.