[Mesa-dev] [PATCH 5/9] i965/nir: Emit MUL + ADD for MAD on GEN <= 5

[Kenneth Graunke]

On Wednesday, March 18, 2015 11:27:19 PM Connor Abbott wrote:
> On Wed, Mar 18, 2015 at 11:24 PM, Kenneth Graunke <kenneth at whitecape.org> wrote:
> > On Tuesday, March 17, 2015 07:17:19 PM Jason Ekstrand wrote:
> >> ---
> >>  src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 8 +++++++-
> >>  1 file changed, 7 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> >> index 5da8423..41f9ae2 100644
> >> --- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> >> +++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> >> @@ -1234,7 +1234,13 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr)
> >>        break;
> >>
> >>     case nir_op_ffma:
> >> -      inst = emit(MAD(result, op[2], op[1], op[0]));
> >> +      if (brw->gen >= 6) {
> >> +         inst = emit(MAD(result, op[2], op[1], op[0]));
> >> +      } else {
> >> +         fs_reg temp = vgrf(glsl_type::float_type);
> >> +         emit(MUL(temp, op[0], op[1]));
> >> +         inst = emit(ADD(result, temp, op[2]));
> >> +      }
> >>        inst->saturate = instr->dest.saturate;
> >>        break;
> >>
> >>
> >
> > I think this is a reasonable plan for now - it's simple and works.
> > Connor is probably right that we can do better with nir_opt_algebraic,
> > but I think doing that as a follow-on patch justified by statistics
> > would be wise.
> 
> Actually, it seems like we already have a lower_ffma option, so just
> setting that to true on gen < 6 should be even less code than this...

You make a good point.  Let's do that.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20150318/20ca99d7/attachment.sig>