On 02/19/2015 03:38 PM, Matt Turner wrote: > On Thu, Feb 19, 2015 at 3:31 PM, Ian Romanick <[email protected]> wrote: >> On 02/10/2015 11:09 AM, Matt Turner wrote: >>> We were special casing OPCODE_END but no other instructions that have no >>> destination, like OPCODE_KIL, leading us to emitting MOVs with null >>> destinations. >> >> Should we (or do we somehow already) handle instructions with null >> destinations? So... >> >> ADD.sat.nz null, g38, g22 >> >> Or is that just impossible? > > I don't think that's possible. The cases I saw were because we were > emitting MOV.SAT to copy results of instructions that don't have > destinations. :) > > After I made this change, I grepped all of the output of shader-db and > found no more MOVs (without conditional mod) with null destinations.
That's about what I figured. This patch is Reviewed-by: Ian Romanick <[email protected]> _______________________________________________ mesa-dev mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/mesa-dev
