In the lecture, we have a question about whether compilers prefer SIMD or multithreading. In the case of using supercomputer which has countless SIMD ALUs and cores, what are some principles that supercomputer compilers follow to best utilize the supercomputer resource? Is bandwidth also a significant limit to how fast supercomputer can run?
tcm
Bandwidth is often a major concern for programs running on supercomputers. They are usually very data-intensive programs.
In the lecture, we have a question about whether compilers prefer SIMD or multithreading. In the case of using supercomputer which has countless SIMD ALUs and cores, what are some principles that supercomputer compilers follow to best utilize the supercomputer resource? Is bandwidth also a significant limit to how fast supercomputer can run?
Bandwidth is often a major concern for programs running on supercomputers. They are usually very data-intensive programs.