nv50/ir: fix OP_UNION resolving when used for vector values
authorKarol Herbst <kherbst@redhat.com>
Thu, 1 Sep 2022 19:45:20 +0000 (21:45 +0200)
committerMarge Bot <emma+marge@anholt.net>
Thu, 8 Sep 2022 11:35:35 +0000 (11:35 +0000)
When an OP_UNION def takes part in a vector source e.g. for a tex
instruction we failed to clean up the OP_UNION instruction as rep() points
towards the coalesced value instead.

This fixes a regression on nv50 moving to NIR, but also potentially issues
with nvc0.

The main reason this is common in nv50 is, that we lower OP_SLCT to a set,
predicated movs and a union.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6406
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7117
Cc: mesa-stable
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: M Henning <drawoc@darkrefraction.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18377>

src/nouveau/codegen/nv50_ir_peephole.cpp

index 4e2b3e6..9088e61 100644 (file)
@@ -53,7 +53,7 @@ Instruction::isNop() const
       if (!getDef(0)->equals(getSrc(0)))
          return false;
       if (op == OP_UNION)
-         if (!def(0).rep()->equals(getSrc(1)))
+         if (!getDef(0)->equals(getSrc(1)))
             return false;
       return true;
    }