Pbench: A benchmark suite for characterizing 3D printing prefabrication

3D printing is revolutionizing the next-generation manufacturing industry. With increasing design complexity, computation in the prefabrication process is becoming the bottleneck of 3D printing. For example, a multi-scale, multi-material 3D design (e.g., a bionic bone) takes a few hours or even days to complete the prefabrication computation. Therefore, it is prudent to improve the performance of 3D printing prefabrication. In this paper, we investigate the computational challenges in 3D printing. First, we develop the PBench, an open-source standard benchmark suite for 3D printing prefabrication. To the best of our knowledge, this is the first benchmark suite for 3D printing prefabrication. Second, we study the properties of PBench using TotalChar, a proposed benchmark characterization framework analyzing PBench from three complementary dimensions, i.e., (1) microarchitecture independent analysis, (2) architecture bottleneck analysis and (3) functional performance analysis. Experiments show that 3D printing prefabrication can be potentially optimized through specified function accelerator design or parallelism exploration. Overall, this work establishes the potential for accelerating 3D printing prefabrication.