On Thu, 2017-03-23 at 13:50 -0700, Matt Turner wrote: > SEL can only convert between a few integer types, which we basically > never do. > > Fixes fs/vs-double-uniform-array-direct-indirect-non-uniform-control- > flow > Cc: mesa-sta...@lists.freedesktop.org
I sent a similar but wrong patch (taking only into account the type size) some time ago, but after discussing it with Curro, the solution was to fix it inside d2x pass. This is what this patch "i965/fs: generalize the legalization d2x pass" does. I am still working on improving that patch but I expect to have something soon. If you prefer to land this now, please add my R-b but you probably want to discuss it with Curro before: Reviewed-by: Samuel Iglesias Gonsálvez <sigles...@igalia.com> Sam > --- > src/intel/compiler/brw_fs_sel_peephole.cpp | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/src/intel/compiler/brw_fs_sel_peephole.cpp > b/src/intel/compiler/brw_fs_sel_peephole.cpp > index 8cd897f..4c1c160 100644 > --- a/src/intel/compiler/brw_fs_sel_peephole.cpp > +++ b/src/intel/compiler/brw_fs_sel_peephole.cpp > @@ -165,6 +165,8 @@ fs_visitor::opt_peephole_sel() > then_mov[i]->exec_size != else_mov[i]->exec_size || > then_mov[i]->group != else_mov[i]->group || > then_mov[i]->force_writemask_all != else_mov[i]- > >force_writemask_all || > + then_mov[i]->dst.type != then_mov[i]->src[0].type || > + else_mov[i]->dst.type != else_mov[i]->src[0].type || > then_mov[i]->is_partial_write() || > else_mov[i]->is_partial_write() || > then_mov[i]->conditional_mod != BRW_CONDITIONAL_NONE ||
signature.asc
Description: This is a digitally signed message part
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
