This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Inline assembly syntax
- From: <tm_gccmail at kloo dot net>
- To: Paul Koning <pkoning at equallogic dot com>
- Cc: shebs at apple dot com, ian at wasabisystems dot com, gcc at gcc dot gnu dot org
- Date: Mon, 10 May 2004 17:43:14 -0700 (PDT)
- Subject: Re: Inline assembly syntax
On Mon, 10 May 2004, Paul Koning wrote:
> >>>>> "Stan" == Stan Shebs <shebs@apple.com> writes:
>
> Stan> Incidentally, as I was thinking about the assembly grokking
> Stan> problem, it occurred that we already have a rich mine of info
> Stan> about operand constraints - the machine descriptions
> Stan> themselves. Not always expressed as per-instruction, but for at
> Stan> least the one-instruction output strings, one could make up a
> Stan> table of instructions/constraint letters/types that would be
> Stan> pretty accurate, and not need special per-target hackery.
>
> But that only covers the subset of instructions used by the compiler.
> A big reason for assembly code is to do things the compiler doesn't
> do. So that's likely to exercise code patterns not found in the .md
> file.
>
> paul
This is especially true on the SH architecture, where we have vector
instructions which are not describable in the .md file. (They use
non-compiler visible registers).
Toshi