For a medium-sized protein pair, the GPU calculation achieved a 17.3-fold speed up compared with the one CPU core, which was much greater than that for a small-sized protein.
Strengths of this study include the randomised design, blinding of the primary outcome assessors, sample size determined from power calculation achieved, appropriate statistical methods including intention-to-treat analysis and linear mixed models applied, moderate withdrawal and reporting according to CONSORT guidelines.
For large-sized protein pairs, the GPU calculations achieved 17.8-fold speed up compared with the one CPU core and the acceleration rate was almost same as that with a medium-sized protein pair.
An improvement by more than two orders of magnitude in the execution speed of the calculation is achieved by the Newton iteration compared to time integration to a periodic state.
Unlike distance vector, it eliminates the need to calculate the distance for each class, therefore it can speed up the calculation and achieve better classification performance by ignoring the disturbance from the distant classes.
The calculations were achieved by Langevin dynamics and Monte Carlo simulation methods.
All the calculations are achieved by using steady-state theoretical model, which is based on the time-dependent Hamiltonian model.
Additional accuracy for the calculations is achieved through grid refinement by re-generating the FEM grid manually.
In the calculation examples, to achieve the integrated optimization of multiple thermal systems design and performance prediction, by describing the influence of system thermomechanical parameters and operating duty to underhood heat transfer process, performance evaluation of the engine cooling circuit and the air conditioning circuit are realized.
Additionally, some shortcomings about experimental process and results need to be pointed out and explained: 1) The power flow calculation has achieved better performance in stand-alone model when the number of buses is relatively small.
