1 Introduction
2 Related work
3 A real-time tsunami inundation forecast system
3.1 System configuration
3.2 Tsunami source modeling system
Obs. data | Information |
---|---|
Earthquake early warning | Hypocenter, magnitude |
GRiD-MT | Hypocenter, magnitude, fault mechanism |
RAPiD/REGARD | Finite fault model |
3.3 Simulation system
3.4 Delivery/mapping server
4 Tsunami simulation
4.1 Tsunami simulation model
4.2 Computational domain
Domain | Grid size (m) | Number of grid points | |
---|---|---|---|
\(x\)-direction | \(y\)-direction | ||
1 | 810 | 1536 | 1020 |
2 | 270 | 1680 | 990 |
3 | 90 | 2292 | 1260 |
4 | 30 | 1782 | 1188 |
5 | 10 | 3504 | 2364 |
4.3 Implementation of the TUNAMI model on SX-ACE
5 Performance evaluation
5.1 Experimental setup
5.1.1 SX-ACE
5.1.2 LX 406Re-2
System | Clock freq. (GHz) | Perf./CPU (Gflop/s) | No. of cores | Perf./core (Gflop/s) | Mem. BW (GB/s) |
---|---|---|---|---|---|
SX-ACE | 1 | 256 | 4 | 64 | 256 |
LX 406Re-2 | 2.4 (Base) | 230.4 | 12 | 19.2 (Base) | 59.7 |
(Xeon E5-2695 v2) | 3.2 (Turbo) | 25.6 (Turbo) |
5.1.3 Evaluated model
Kochi | Shizuoka | Ishinomaki | |
---|---|---|---|
Total grid points |
\(16.5~\times ~10^{6}\)
|
\(8.0~\times ~10^{6}\)
|
\(15.7~\times ~10^{6}\)
|
Time interval (s) | 0.1 | 0.05 | 0.1 |
Period (h) | 6 | 6 | 6 |
Flop count |
\(0.63~\times ~10^{15}\)
|
\(0.67~\times ~10^{15}\)
|
\(0.63~\times ~10^{15}\)
|
System | Compiler | Options |
---|---|---|
SX-ACE | FORTRAN90/SX | -Chopt -pi |
LX 406Re-2 | Intel Fortran Composer XE | -O3 -ipo |
5.2 Overall performance
Kochi | Shizuoka | Ishinomaki | |
---|---|---|---|
Execution time (min) | 4.0 | 4.5 | 3.9 |
5.3 Processor performance
5.3.1 Single-core performance
5.3.2 Single-processor performance
5.4 Multi-node performance
System | Perf./CPU (Gflop/s) | No. of cores | Perf./core (Gflop/s) | Mem. BW (GB/s) |
---|---|---|---|---|
K computer | 128 | 8 | 16 | 64 |