Fig. 5
From: Trace: a high-throughput tomographic reconstruction engine for large-scale datasets
Breakdown of iterative reconstruction times (secs) with respect to varying parallelization configurations. Here ppn stands for processes per node, and t is the number of initialized threads per process. For example, with configuration ppn2-t6, Trace-OC initiates two processes on each compute node, each with six worker threads (i.e., a total of 12 threads per compute node). We used 32 compute nodes for the reconstruction. a Mouse brain, b shale sample