This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: regclass.c: scan_one_insn special casing
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- To: echristo at redhat dot com
- Cc: gcc at gcc dot gnu dot org
- Date: Sat, 27 Mar 04 18:55:35 EST
- Subject: Re: regclass.c: scan_one_insn special casing
> For those very few cases that are machine-specific.
There are tons of things that go through define_split now. You should
look at a few ports as examples:
I know. It has been used a lot, some would say abused. But that's no
reason to require each port to add yet more of them.
I know, you put it in there in 92. I won't ask why - I don't even want
to try to think of things I did 11 years ago :)
Actually, I *thought* I wrote that, but wasn't sure.
So that you aren't penalizing every call to scan_one_insn with this
code that is almost never called.
But that's a small cost compared to the maintenance issue of having
that define_split in every port.
Now, you could certainly claim that there's some reason that this
transformation can't occur (experimental isn't that interesting here)
and that the code is therefore obsolete. But I haven't seen that claim.