Parallel Efficiency Tests
Tests in this page were performed on a server with the following hardware and software configuration:
- 2 * Intel Xeon Gold 6226 CPU @ 2.7GHz (2 * 12 cores, 2 * 12 * 2 threads, hyperthreading disabled)
- 6 * 32 GB DDR4 memory
- Red Hat Enterprise Linux 7.6 (kernel 3.10.0-957.12.1.el7)
- Intel Parallel Studio Cluster version 2019.5
- ICC optimization flags: -O3 -xHost
- OpenMP environment variables
- OMP_NUM_THREADS=1, 2, 4, 8, 12, 24
- OMP_PLACES=cores
- OMP_PROC_BIND=close
Test point sets: 800,000 uniformly and randomly distributed points in a 3D unit ball
Prescribed QR relative error tolerance: 1e-6
Report H2-build and H2-matvec timings
Test kernel: Coulomb
| Components and Running modes | Parallel Efficiency Using 12 Cores | Parallel Efficiency Using 24 Cores |
|---|---|---|
| H2-construction, AOT | 65.0% | 56.5% |
| H2-matvec, AOT | 68.8% | 58.3% |
| H2-construction, JIT | 71.2% | 56.9% |
| H2-matvec, JIT | 66.7% | 54.8% |
| H2-matvec, JIT, projected by CPU frequency | 89.7% | 73.7% |
The parallel efficiency figure above looks different from the one in our paper. The reason for this is that here we used a much larger point set and the computing node has fewer processors. Nevertheless, the conclusions in the paper still hold.