[Mesa-dev] killing off the address reg in tgsi

[Rob Clark]

On Thu, Jan 29, 2015 at 4:20 PM, Roland Scheidegger <sroland at vmware.com> wrote:
> Hi,
>
> the address reg in tgsi is quite a nuisance. glsl-to-tgsi code assumes
> that indirections can only be done through the address reg and has quite
> some extra code to deal with this. Even though hardware and apis which
> worked like that are definitely old by now.
> Thus, I'm proposing the address reg be nuked. I am however not quite
> sure what the implications for drivers are, other than I'm certain
> llvmpipe can handle that already.
> For that reason, I suspect at least initially a new cap bit would be
> required so glsl-to-tgsi would skip the extra code. I tend to think
> longer term it would be great if it could be nuked completely, I am
> however not sure if that is easily done with drivers for old hw (such as
> r300) - I guess if necessary we could keep operations such as ARL (or
> even ARR though clearly not UARL!) and just define them to be usable
> with temp regs.
>
> Opinions?

Not sure about other hw (although I would assume radeon is similar)..
but adreno has real actual address register, which need to be used for
relative addressing of the register (or const) file.  Otherwise we'd
need to do something like arrays in memory and load/store
instructions.

Possibly w/ more smarts in the compiler, we could infer address
register usage.  It would require some smarts that we don't have yet
to eliminate redundant writes to the address register generated for
each instruction with indirect reference.

(Although eventually I'd like to use tgsi->nir.. and haven't yet
looked at how nir handles indirects so maybe eventually I need to
implement those smarts in my compiler anyway.)

BR,
-R


> Roland
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev