... | ... | @@ -108,7 +108,10 @@ Comparison of the update phase, without considering task overlapping. We will me |
|
|
| Strategy 2 | 4 |210 - 230 ms |
|
|
|
| Strategy 3 | 4 | |
|
|
|
|
|
|
# Ring communications
|
|
|
# Multi-reduce + broadcast communications
|
|
|
|
|
|
Some traces, with 4 ranks, 1 rank/socket, B = 4, T=1024
|
|
|
|
|
|
![RC_legend](uploads/74660146f17bc0d52b76d2d48dd95455/RC_legend.png)
|
|
|
|
|
|
![RC_implementation](uploads/12ae0ed35cb016331ce4da4a3f7d9aac/RC_implementation.png)
|
... | ... | @@ -120,3 +123,12 @@ Comparison of the update phase, without considering task overlapping. We will me |
|
|
![mpi_comms](uploads/cc578008512fec4a6c810fd4936d588f/mpi_comms.png)
|
|
|
|
|
|
![mpi_comms.code_legend](uploads/b2a0c1e1c7f98529751cbe14084768aa/mpi_comms.code_legend.png)
|
|
|
|
|
|
|
|
|
Communication cost with number of ranks :
|
|
|
| Number of ranks | Communication time from last to first | Communication time from first to last |
|
|
|
| ------ | ------ | ------ |
|
|
|
| 2 | 100 ms | 180 ms |
|
|
|
| 4 | 70 ms | 100 ms |
|
|
|
| 8 | 150 ms | 300 ms |
|
|
|
| 16 | 175 ms | 300 ms | |
|
|
\ No newline at end of file |