Scaling
hybrid scaling on jugene
- benchmark version: 1.0.0
- IBM BG/P jugene
- time needed for one time step, whereas it is computed by averaging the last 5 steps
- used time output is provided in the stdout as t_all
data
- number of particle: 100,000
- input file: small.para
mpi ranks | time [s] | speedup | speedup w.r.t. 1
|
1 | 39.95 | 1 | 1
|
2 | 20.86 | 1.92 | 1.92
|
4 | 10.8 | 3.7 | 3.7
|
8 | 5.64 | 7.08 | 7.08
|
16 | 3.07 | 13.02 | 13.02
|
32 | 1.76 | 22.67 | 22.67
|
64 | 1.07 | 37.3 | 37.3
|
128 | 0.7 | 56.74 | 56.74
|
256 | 0.56 | 71.58 | 71.58
|
- number of particle: 1,600,000
- input file: medium.para
mpi ranks | time [s] | speedup | speedup w.r.t. 1
|
16 | 61.9 | 1 | 16
|
32 | 32.1 | 1.93 | 30.85
|
64 | 16.41 | 3.77 | 60.35
|
128 | 8.72 | 7.1 | 113.58
|
256 | 4.72 | 13.11 | 209.83
|
512 | 2.67 | 23.19 | 371.08
|
1024 | 1.76 | 35.23 | 563.69
|
2048 | 1.51 | 40.94 | 655.03
|
- number of particle: 25,600,000
- input file: large.para
mpi ranks | time [s] | speedup | speedup w.r.t. 1
|
256 | 91.41 | 1 | 256
|
512 | 46.97 | 1.95 | 498.21
|
1024 | 24.78 | 3.69 | 944.35
|
2048 | 13.61 | 6.72 | 1719.39
|
4096 | 8.63 | 10.59 | 2711.27
|
8192 | 7.94 | 11.51 | 2945.74
|
- number of particle: 102,400,000
- input file: xlarge.para
mpi ranks | time [s] | speedup | speedup w.r.t. 1
|
2048 | 57.94 | 1 | 2048
|
4096 | 32.4 | 1.79 | 3662.38
|
8192 | 24.14 | 2.4 | 4915.54
|
plot
timing of individual phases
data
cores | trun | t0_domains | t0_allocate | t0_build | t0_branches | t0_fill | t0_properties | t0_walk | t0_walkc | t0_force | t0_restore | t0_deallocate | t0_all | ttot | Phase 1 | Phase 2 | Phase 3 | Phase 4
|
256 | 1.70E-01 | 1.09E+00 | 2.70E-01 | 2.99E-01 | 2.74E-01 | 6.08E-01 | 8.07E-01 | 3.53E+01 | 9.57E+00 | 4.34E+01 | 2.08E-01 | 1.24E-03 | 9.14E+01 | 9.15E+01 | 1.57 | 1.99 | 78.66 | 4.34E+001
|
512 | 1.90E-01 | 7.66E-01 | 2.57E-01 | 2.00E-01 | 1.73E-01 | 4.06E-01 | 5.04E-01 | 1.75E+01 | 5.67E+00 | 2.15E+01 | 1.41E-01 | 5.29E-04 | 4.70E+01 | 4.71E+01 | 1.16 | 1.28 | 38.98 | 2.15E+001
|
1024 | 1.90E-01 | 3.83E-01 | 2.47E-01 | 1.59E-01 | 1.01E-01 | 3.79E-01 | 3.89E-01 | 8.71E+00 | 3.76E+00 | 1.08E+01 | 7.72E-02 | 6.59E-04 | 2.48E+01 | 2.49E+01 | 0.71 | 1.03 | 19.48 | 1.08E+001
|
2048 | 1.80E-01 | 2.57E-01 | 1.88E-01 | 1.36E-01 | 8.43E-02 | 4.78E-01 | 4.41E-01 | 4.42E+00 | 2.25E+00 | 5.38E+00 | 6.62E-02 | 3.17E-04 | 1.36E+01 | 2.05E+01 | 0.51 | 1.14 | 9.8 | 5.38E+000
|
4096 | 1.80E-01 | 3.29E-01 | 1.74E-01 | 1.23E-01 | 1.20E-01 | 8.22E-01 | 7.05E-01 | 2.27E+00 | 1.38E+00 | 2.71E+00 | 5.41E-02 | 5.55E-03 | 8.63E+00 | 1.95E+01 | 0.56 | 1.77 | 4.97 | 2.71E+000
|
8192 | 1.80E-01 | 6.05E-01 | 1.73E-01 | 1.18E-01 | 2.36E-01 | 1.48E+00 | 1.23E+00 | 1.16E+00 | 1.60E+00 | 1.34E+00 | 5.31E-02 | 3.20E-04 | 7.94E+00 | 2.74E+01 | 0.83 | 3.07 | 2.49 | 1.34E+000
|
plot
Attachments
(6)
Download all attachments as:
.zip
Note:
See
TracWiki
for help on using the wiki.