A Bottom-up Approach for Parallelizing CAPE Software

• Main Authors: K. Alloula, J.P. Belaud, L. Sanroma, J.M. Le Lann
• Format: Article
• Language: English
• Published: AIDIC Servizi S.r.l. 2017-03-01
• Series: Chemical Engineering Transactions
• Online Access: https://www.cetjournal.it/index.php/cet/article/view/2234

One way to parallelize a CAPE software is to simulate unit operations simultaneously. This parallelization at the upper level -the flow sheet level- takes place usually in the context of distributed memory parallelism, where distinct computation nodes execute different pieces of code and exchange data with each other by means of a communication network. Taking advantage of today shared memory architectures, we suggest parallelizing the simulation codes from the lowest level -mathematical expressions and control structures- up to the upper level -the process-. Such an approach requires that the pieces of code to be parallelized should be written in an OpenMP compliant language, and that the source codes may be slightly modified. Assuming that these technical constraints are satisfied, from this systematic and progressive way to parallelize a CAPE software one can obtain an interesting speedup.