nak: BMov is always variable-latency

The barrier half is HW scoreboarded by the GPR isn't.  When moving from
a GPR to a barrier, we still need a token for WaR hazards.

Fixes: 7cd9680554 ("nak: Add back OpBMov with better semantics")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29591>
(cherry picked from commit 0a089b1b13)

.pick_status.json

@@ -2434,7 +2434,7 @@
         "description": "nak: BMov is always variable-latency",
         "nominated": true,
         "nomination_type": 1,
-        "resolution": 0,
+        "resolution": 1,
         "main_sha": null,
         "because_sha": "7cd9680554b16e7de07873cd3be428953be9ad07",
         "notes": null

src/nouveau/compiler/nak/ir.rs

@@ -5700,13 +5700,10 @@ impl Instr {
             Op::Bra(_) | Op::Exit(_) => true,
             Op::WarpSync(_) => false,
 
-            // BMOV: barriers only when using gprs (and only valid for the gpr),
-            // no barriers for the others.
-            Op::BMov(op) => match &op.dst {
-                Dst::None => true,
-                Dst::SSA(vec) => vec.file() == RegFile::Bar,
-                Dst::Reg(reg) => reg.file() == RegFile::Bar,
-            },
+            // The barrier half is HW scoreboarded by the GPR isn't.  When
+            // moving from a GPR to a barrier, we still need a token for WaR
+            // hazards.
+            Op::BMov(_) => false,
 
             // Geometry ops
             Op::Out(_) | Op::OutFinal(_) => false,