broadcom/compiler: ldvary doesn't implicitly write to r3 since V3D 4.1

total instructions in shared programs: 13805979 -> 13786037 (-0.14%)
instructions in affected programs: 2263244 -> 2243302 (-0.88%)
helped: 10646
HURT: 1508
Instructions are helped.

total threads in shared programs: 412220 -> 412242 (<.01%)
threads in affected programs: 58 -> 80 (37.93%)
helped: 17
HURT: 6
Threads are helped.

total uniforms in shared programs: 3793200 -> 3790401 (-0.07%)
uniforms in affected programs: 131281 -> 128482 (-2.13%)
helped: 1547
HURT: 281
Uniforms are helped.

total max-temps in shared programs: 2326309 -> 2324834 (-0.06%)
max-temps in affected programs: 31836 -> 30361 (-4.63%)
helped: 1139
HURT: 153
Max-temps are helped.

total spills in shared programs: 5932 -> 5940 (0.13%)
spills in affected programs: 80 -> 88 (10.00%)
helped: 2
HURT: 3

total fills in shared programs: 13370 -> 13372 (0.01%)
fills in affected programs: 480 -> 482 (0.42%)
helped: 2
HURT: 3

total sfu-stalls in shared programs: 30829 -> 30685 (-0.47%)
sfu-stalls in affected programs: 2190 -> 2046 (-6.58%)
helped: 570
HURT: 533
Sfu-stalls are helped.

total inst-and-stalls in shared programs: 13836808 -> 13816722 (-0.15%)
inst-and-stalls in affected programs: 2276152 -> 2256066 (-0.88%)
helped: 10643
HURT: 1525
Inst-and-stalls are helped.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9430>

src/broadcom/qpu/qpu_instr.c

@@ -815,7 +815,7 @@ v3d_qpu_writes_r3(const struct v3d_device_info *devinfo,
                 return true;
         }
 
-        return inst->sig.ldvary || inst->sig.ldvpm;
+        return (devinfo->ver < 41 && inst->sig.ldvary) || inst->sig.ldvpm;
 }
 
 bool