Fig. 13(b) labels show the level of parallelism or speedup achieved (limited to a theoretical maximum of 32 in our current setup), and is computed by dividing the CPU time and wall time. However, unlike the computations performed on a volume per-voxel basis, the stream tracing workload ...