Figure 6
(a) Comparison of real and estimated execution times (in s) considering different phases for the Hornby dataset with varying number of compute nodes at Stampede. The reconstruction algorithm is APMLR. The dimensions of the dataset are 180 × 2048 × 2048. Workflow of the reconstruction is configured as W: Chameleon–Stampede–Chameleon. (b) Distribution (%) of phases with respect to total execution time grouped in real and estimated configurations. |