This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Status of SSE builtins
> >>>>> "Richard" == Richard Henderson <rth@redhat.com> writes:
>
> Oh yeah, I'll probably tackle this next week. I'm just a bit busy
> with my try/finally and CLASS_CANNOT_CHANGE_MODE patches which are
> both 98% finished ;-).
Great. I am little bit more busy right now so I will hapilly leave the
generic stuff for you :)
>
> > On Thu, Oct 31, 2002 at 05:54:39PM +0100, Jan Hubicka wrote:
> >> What names would you propose?
> >> Perhaps extract_vec_fieldM/set_vec_fieldM?
>
> > No idea. I was hoping that Aldy would invent something. ;-)
>
> I was thinking more along the lines of aldyvec_fieldM :).
>
> I think a combination of the vec_select and vec_merge patterns which
> we already have (but don't emit), could be used for most
> architectures. I'll look into it.
We do have VEC_MERGE/VEC_SELECT RTL expressions. How this is related to
named patterns?
To access specific field of V16QI register in SSE, you need serious
amount of work, as almost nothing in SSE supports byte operations.
This is extreme case of course, but we should consider it when doing
API. If you have more specific ideas, please let me know.
Having vec_setM to set specific field of register and vec_getM to read
it. Both accepting an integer operand for the field, vector input and
vector/scalar intput/output should be generic enought for SSE, but
producing lousy code sequences. WIth some luck, combine will clean
these up.
Honza
>
> AldyVec