2/ We implemented fast, memory efficient sim mimicking a variety of tactile hardware physics. With per-probe parallelization and some good old fashioned engineering, we achieve much better FPS throughput, easily scaling past 16k parallel environments and 1k+ taxels on one GPU.