[Mesa-dev] [PATCH] i965/fs: Fix saturate on MAD and LRP with the NIR backend.

[Eero Tamminen]

Hi,

On 02/03/2015 05:06 PM, Jason Ekstrand wrote:
> Ooh! I bet this fixes our rendering problems on some of those benchmarks
> too!  I was wondering why generating made was causing problems.

NIR in today's Mesa renders fine all the benchmarks in which I've had 
noticed rendering artifacts last week.  Performance gap (in uniform 
heavy) things is still there though. :-)


	- Eero

> Reviewed-by: Jason Ekstrand <jason.ekstrand at intel.com
> <mailto:jason.ekstrand at intel.com>>
>
> On Feb 3, 2015 1:18 AM, "Kenneth Graunke" <kenneth at whitecape.org
> <mailto:kenneth at whitecape.org>> wrote:
>
>     Fixes misrendering in "Witcher 2" with INTEL_USE_NIR=1, and probably
>     many other programs.
>
>     Signed-off-by: Kenneth Graunke <kenneth at whitecape.org
>     <mailto:kenneth at whitecape.org>>
>     ---
>       src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 6 ++++--
>       1 file changed, 4 insertions(+), 2 deletions(-)
>
>     This depends on Jason's 3 patch series that removes emit_percomp.
>     It's available in the 'nir-madfix' branch of my tree.
>
>     This was caught by tests/spec/arb_fragment_program/lrp_sat.shader_test
>     with my in-progress Mesa IR -> NIR converter code, so I don't think we
>     need to write more Piglit tests.  We just don't have a GLSL based one.
>
>     diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
>     b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
>     index fbb1622..153a1be 100644
>     --- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
>     +++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
>     @@ -1070,12 +1070,14 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr)
>             break;
>
>          case nir_op_ffma:
>     -      emit(MAD(result, op[2], op[1], op[0]));
>     +      inst = emit(MAD(result, op[2], op[1], op[0]));
>     +      inst->saturate = instr->dest.saturate;
>             break;
>
>          case nir_op_flrp:
>             /* TODO emulate for gen < 6 */
>     -      emit(LRP(result, op[2], op[1], op[0]));
>     +      inst = emit(LRP(result, op[2], op[1], op[0]));
>     +      inst->saturate = instr->dest.saturate;
>             break;
>
>          case nir_op_bcsel:
>     --
>     2.2.2
>
>     _______________________________________________
>     mesa-dev mailing list
>     mesa-dev at lists.freedesktop.org <mailto:mesa-dev at lists.freedesktop.org>
>     http://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
>
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev
>