I am observing what seams to be rather low utilization of resources when running genBSDF on a system with many cores, and with a model containing only specular surfaces (e.g. a structrure on glass). I am running genBSDF -n 18 -t4 7 -c 262656 (the latter to keep noise low) on a 20 core system with a fast SSD. CPU load per process is at about 8%, the parent rcontrib is at 30%, rfluxmtx at 2.5%. There is plenty of memory. I was first guessing that there would be a io-bottleneck, but iotop shows processes just above 0%. Renicing to -15 did not have any effect. Is this a side-effect of locking the shared files holding the rcontrib output?