This started as a presentation for US EPA's Office of Research and Development in 2002 when the author was at MCNC—North Carolina Supercomputing Center, and has been updated from time to time since. It is amazing just how well the lessons of that time have held up—the only thing is that the tendencies noted in 2002 have become even stronger over time—main memory (and even caches) is even slower relative to processor-cores, for example
is the direct English translation of a typical I/O API
Read layer 1 of variable 'OZONE' from 'CONCFILE' for 5:00 PM GMT on July 19, 2018 and put the result into array A.
I/O API Home Page
I/O API Programmers Manual
PAVE Users Guide
PAVE Frequently Asked Questions