Customer Experiences with the New ADINA Sparse Solver
On the launch of ADINA version 9.4 (ADINA 9.4) one year ago, it was announced that the new ADINA sparse solver released in ADINA 9.4 achieves considerable speedups with no compromise in accuracy. This speed increase is of great value to industry — in particular, the automotive, aerospace, defense, industrial equipment, heavy machinery, civil, and consumer product industries — that frequently need to run large, high-fidelity simulations.
Over the past year, we have heard from our ADINA customers that the new ADINA sparse solver has significantly increased their productivity, while maintaining a high level of accuracy, giving them a significant advantage.
In this Tech Brief, we will show some examples of the performance of the new ADINA sparse solver released in version 9.4. We show representative test models and actual customer models. The models are solved using ADINA 9.3 and ADINA 9.5, and the solution times and memory usage are compared.
All models except no. 4 have been solved using an Intel Xeon E5-2667 v4 workstation with 16-cores and SMP (no DMP). Model no. 4 was solved using an Intel Xeon CPU E5-2687W workstation with 16-cores and SMP (no DMP).
Models
This section describes the models considered.
Model 1 – Linear static analysis
Figure 1 Model 1 – Finite element mesh for linear static analysis of a 3D block
Model 2 – Nonlinear contact analysis
Figure 2 Model 2 – Finite element mesh for nonlinear contact analysis of 3D blocks
Model 3 – Nonlinear contact analysis
Figure 3a Model 3 – Finite element mesh for nonlinear contact analysis of a ring bearing
Figure 3b Model 3 – Animation of effective stress results
Model 4 – Frequency analysis of a six cylinder engine block
Figure 4a Model 4 – Finite element mesh for frequency analysis with contact of a six cylinder engine block
Figure 4b Model 4 – Animation of mode shape 1
Model 5 – FSI analysis
Figure 5a Model 5 – Fluid finite element mesh for FSI analysis of a hydraulic engine mount
Figure 5b Model 5 – Animation of deformation with effective stresses and fluid velocity vectors
Solution time and memory usage results
The solution speedup is defined as:
The memory usage reduction is defined as:
Table 1 compares the total solution times of ADINA 9.3 and ADINA 9.5.
Table 1 Total solution times for ADINA 9.3 and ADINA 9.5
Models |
Total Solution Time (s)
|
Speedup |
|
ADINA 9.3 |
ADINA 9.5 |
||
Model 1 |
1,624 |
238 |
6.8 |
Model 2 |
10,459 |
1,795 |
5.8 |
Model 3 |
23,619 |
7,896 |
3.0 |
Model 4 |
56,300 |
7,088 |
7.9 |
Model 5 |
38,368 |
11,130 |
3.5 |
Table 2 compares the memory usage of ADINA 9.3 and ADINA 9.5.
Table 2 Memory usage of ADINA 9.3 and ADINA 9.5
Models |
Total Memory Usage (GB)
|
Memory Usage Reduction |
|
ADINA 9.3 |
ADINA 9.5 |
||
Model 1 |
84.4 |
47.3 |
1.8 |
Model 2 |
122.6 |
76.9 |
1.6 |
Model 3 |
89.8 |
68.7 |
1.3 |
Model 4 |
207.4 |
162.3 |
1.2 |
Model 5 |
50.6 |
38.2 |
1.3 |
Tables 1 and 2 show that the new ADINA sparse solver released in version 9.4 achieves considerable speedups and memory usage reductions for a wide range of applications, including linear, nonlinear, contact, frequency, and FSI analyses.
For the models considered in this Tech Brief, the new ADINA sparse solver is 3 to 7 times faster than the sparse solver in ADINA 9.3.
SMP parallel-processing performance results
The SMP performance rate is defined as:
Table 3 summarizes the SMP performance rates of ADINA 9.5 for Model 1.
Table 3 SMP performance rates of ADINA 9.5 for Model 1
Number of Cores
|
Total Solution Time (s)
|
SMP Performance Rate
|
1 |
2,641 |
1.0 |
2 |
1,297 |
2.0 |
4 |
680 |
3.9 |
8 |
382 |
6.9 |
16 |
238 |
11.1 |
The table shows that the ADINA sparse solver released in version 9.4 achieves good SMP scalability as the number of cores is increased.
Conclusions
The new ADINA sparse solver released in version 9.4 clearly greatly strengthens the ADINA System allowing customers to increase their productivity, while maintaining a high level of accuracy, for a wide range of analysis problems.
Keywords:
Sparse solver, frequency analysis, contact analysis, efficiency, improvement, productivity