Closed dtabell closed 4 years ago
One Zgoubi test parallelized. We examined strong scaling, with results consistent with the 10% non-parallel overhead. This means only very long parallel runs will benefit from large numbers of processors.
A bug—quite possibly a compiler bug—has prevented us limiting the I/O to image 1. This is a serious stumbling block. Am closing this.
Perform a scaling test with the JLEIC lattice v. 15.1 on 1, 5, 10, 20, 40, and 60 cores. The test will use a minimum of 120 particles and run for approximately 100 turns. The goal is to show roughly linear speed-up.