[PATCH]: SH 2a - Part 4: Begin using RETURN_VALUE
Tue Oct 5 20:09:00 GMT 2004
Corinna Vinschen writes:
> this is SH 2a patch 4. It does not introduce new functionality, but
> instead it converts sh-tdep.c to use RETURN_VALUE instead of
> STORE_RETURN_VALUE, EXTRACT_RETURN_VALUE, RETURN_VALUE_ON_STACK and
> The old functions are still used and just called from the RETURN_VALUE
> implementations. We have two of them, one for fpu and one for nofpu
> What this patch does *not* do is this: I would love to rename the
> functions sh_default_store/extract_value and sh3e_sh4_store/extract_value
> to sh_store/extract_value_nofpu and sh_store/extract_value_fpu.
> This would decouple the sense of these functions from the cpu names,
> which is rather irritating and wrong anyway. If that's ok, I'd
> submit another patch which just performs the renaming.
It wasn't wrong and irritating at the time it was introduced.
> Ok, for the sake of completeness, here are the testresults of various
> SH cpu types, tested on GNU/Linux host and sh-sim, before and after
> applying the whole set of SH 2a patches. I hope the choice of CPUs is
> representative enough:
> BEFORE AFTER
> ----- -----
> sh-sim/-m2a -- 88
> sh-sim/-m2a-nofpu -- 79
> sh-sim/-m2e 100 100
> sh-sim/-m3 79 79
> sh-sim/-m4 88 88
> sh-sim/-m4/-ml 86 86
> sh-sim/-m4a-single 88 88
Thanks for providing the test results. Are these with
sources.redhat.com sources? Or some internal only tree? What version
of gcc? So we have a good reference point.
More information about the Gdb-patches