|
|
![Plot_GadgetSwift](/uploads/5bc62a62664d7378f2b76fa2eb1a5313/Plot_GadgetSwift.png)
|
|
|
![Plot_GadgetSwift_SPH200000](/uploads/42cc721a951b63fe4defa172dad3cbe1/Plot_GadgetSwift_SPH200000.png)![Plot_GadgetSwift_SPH200000_Hybrid](/uploads/e1b17c8b8ad458b3898a494eb4f09826/Plot_GadgetSwift_SPH200000_Hybrid.png)![Plot_GadgetSwift_SPH26000](/uploads/0fc915795a5a2f794d5e763b359fb140/Plot_GadgetSwift_SPH26000.png)
|
|
|
|
|
|
|
|
|
The time taken for a single step in Gadget-2 and Swift. Gadget is run with 1, 2, 4, 8 and 12 cores. Swift-Threads is run with 1, 2, 4, 8 and 12 threads. Swift-MPI is run on 1,2,4,8 and 12 cores with 1 thread per core/rank. I'm clearly doing something wrong with the MPI runs. This is for the isothermal sphere - no gravity with ~ 26000 particles. I'll look at gravity next - that requires a modified Gadget-2 which at the moment is broken :). It is also run out out of the Gravity Particles branch..... |
|
|
\ No newline at end of file |