This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Higher level RTL issues
- To: law at redhat dot com
- Subject: Re: Higher level RTL issues
- From: Joern Rennecke <amylaar at onetel dot net dot uk>
- Date: Tue, 23 Oct 2001 21:12:22 +0100 (BST)
- Cc: jh at suse dot cz (Jan Hubicka), dan at cgsoftware dot com (Daniel Berlin), gcc at gcc dot gnu dot org
> I don't think you'd need to expose both generic and target RTL patterns
> to genattrtab since attributes are generally related to low level target
> specific stuff (scheduling, insn lengths, delay slots, etc).
Well, you might want to provide default definitions for high level rtl insn
attributes that are different from the defaults of the machine description.
For example, you might want to default most md patterns to have a 'normal'
type, for standard length and one-cycle execution. Obviously, high-level
rtl patterns should default to something else.
So we could start with having an automatically provided attribute that
says if a pattern is high-level rtl, i.e. not described in the md file,
and the default attribute definitions can then test this attribute.
Or - that should make the build machinery simpler - we could have a new
(optional) slot in attribute definitions, which provides the value for
high-level rtl patterns.
--
Joern Rennecke | gcc expert for hire
amylaar@onetel.net.uk | send enquiries to: jwr_jobs@onetel.net.uk