论文部分内容阅读
The scale of some datasets generated by simulations on tens of thousands of cores are gigabyte or larger per output step. It is imperative that efficient coupling of these simulations and parallel visualization. Therefore, a patch-based parallel I/O strategy has been designed and implemented in JASMIN. These programs reconstructed or developed on JASMIN can write plotting data in many files. These datasets in files can be read directly by Visit software with low I/O overheads. Finally, four large-scale datasets generated have been visualized efficiently on 72 cores by using Visit software. These make it possible to analyze and understand these datasets in a timely fashion.