|
|
# Memory use of the SWIFT FoF implementation
|
|
|
|
|
|
This page shows results of running FoF on the EAGLE DMONLY 50Mpc box on cosma7. The box is replicated 1-5 times in each dimension and we run on 1, 8, 27, 64 and 125 nodes. There are 2 MPI ranks per node and 14 threads per rank.
|
|
|
This page shows peak memory use when running FoF on the EAGLE DMONLY 50Mpc box on cosma7. The box is replicated 1-5 times in each dimension and we run on 1, 8, 27, 64 and 125 nodes. There are 2 MPI ranks per node and 14 threads per rank.
|
|
|
|
|
|
## On one node
|
|
|
The plots show total memory use on one node.
|
|
|
|
|
|
## 1 node / 1 replication
|
|
|
|
|
|
![memuse_EAGLE50_x1](/uploads/566752c6f23279f7bde1099ba12928b9/memuse_EAGLE50_x1.png)
|
|
|
|
|
|
## 125 nodes, 5*5*5 replications
|
|
|
|
|
|
![memuse_EAGLE50_x5](/uploads/cc964ba6d3f317c075a84469b16f72c0/memuse_EAGLE50_x5.png) |