Parallel Performance of the IMSL C Library
Click here to download now
(1)Overview: The use of OpenMP directives within the IMSL C Numerical Library is an effective means of achieving scalable parallelism in a portable fashion across a wide variety of computing platforms. While other technologies exist for SMP parallelization, OpenMP is a stable, well‐known and mature choice with broad platform support, which is a key for the IMSL C Numerical Library. Some functions show very good, nearly linear, scaling up to 8 threads, which is quite good considering the algorithms were not re‐written, but rather just had OpenMP directives added in strategic places. This paper discusses a wide range of algorithms commonly used to develop financial models parallelized in IMSL C Numerical Library version 7.0, allowing quantitative analysts to improve finance computation by tackling bigger models , make data-driven decisions faster and increase responsiveness to volatile financial markets.
