Takuma Kawamura, Tomoyuki Noda and Yasuhiro Idomura
We examine the performance portability of the in-situ visualization system based on the Particle Based Volume Rendering (in-situ PBVR). In this system, parallelized in-situ processing converts extreme scale volume data into small rendering primitive data given by particles without costly visibility ordering, and the small particles can be rendered as view-independent volume rendering image on client user PC. These features enable us to avoid severe bottlenecks on latest many core platforms such as limited memory size per node and significant performance gap between computation and inter-node communication. The system shows excellent strong scaling up to about 50k threads on the Oakforest-PACS, which consists of 8,208 Intel Xeon Phi7250 (Knights Landing) processors. This performance is compatible with the capability of remote in-situ data exploration based on asynchronous file-based control sequences.