This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] ppc ccX splitters
- To: Horst von Brand <vonbrand at sleipnir dot valparaiso dot cl>
- Subject: Re: [patch] ppc ccX splitters
- From: Clinton Popetz <cpopetz at cygnus dot com>
- Date: Tue, 22 Feb 2000 08:33:28 -0600
- Cc: Michael Meissner <meissner at cygnus dot com>, gcc-patches at gcc dot gnu dot org
- References: <meissner@cygnus.com> <200002220039.e1M0dU207019@sleipnir.valparaiso.cl>
On Mon, Feb 21, 2000 at 09:39:30PM -0300, Horst von Brand wrote:
> Michael Meissner <meissner@cygnus.com> said:
> > On Mon, Feb 21, 2000 at 09:05:25AM -0600, Clinton Popetz wrote:
>
> > > This is all done with a perl script (which I can provide to anyone that
> > > is interested.) It matched 170 patterns, which means we have 170 new
> > > splits. There were three insns for which I had to modify the output
> > > template by hand, as they were C code. Also, I had to fix the
> > > anddi3_internal{2,3} cases by hand, since there is no andi (only andi.)
>
> > I can attest to the fact that if you do the 170 patterns by hand as I
> > originally did, somewhere around pattern 150 you accidently put in %4
> > when you meant %5. Good work with the perl script.
>
> In that case, I'd vote for generating (this part of) the .md file from
> templates by this script (or a variant). As things stand, if one of both
> alternatives gets fixed, the other one surely won't be...
I agree that this really needs templatizing, but after talking to Jeff Law, I'm
not convinced there's a way to automate this within the current system. I
think it's a bad idea to maintain the md file and the script in it's current
state, as the latter is too dependent on long term stability of the regexen I
hacked together.
What would be nice is to have a preprocessing step that expanded the md file
before the gen* saw it. In this particular case, we could automate the
addition of the ?y alternatives, and the generation of the splits.
If this is considered a useful long term direction, I can try to come up with a
syntax for it.
-Clint