Why PGaudi?

PGaudi is a complementary package for the GaudiMM suite to solve the problem of the speed of execution. GaudiMM can generate very precise solutions for systems with very little structural data. However, in the execution, the software only uses one core and the execution time skyrocket into a very large amount of time.

From this performance issue it came the idea of PGaudi, a complementary package able to use all the cores available in the machine to accelerate the GaudiMM’s execution run without detriment the output quality.

The parallelization

The idea of the performance of PGaudi is to take the initial input and generates new subprocesses of Gaudi runs. These new subprocesses are cheaper computationally to speed up the global execution run and independently, they not share information between them. To simplify the new subprocesses, PGaudi decreases the values of the variables of the Genetical Algorithm of the number of generations and the number of individuals per population. The decreasing is proportional to the number of subprocesses generated.

GaudiMM’s, at the beginning of the run, generates random solutions, so the creation of new independent subprocesses allows the generation of a very diverse pool of solutions ensuring an output of quality.

Once all the subprocesses are finished, PGaudi collects all the solutions obtained and checks if there are some repeated solution between the subprocess, in which case, PGaudi removes one of the double solution.

Finally, PGaudi stores all the solutions in an unique output file with the extension .gaudi-output.

Tip

You can visualize all the solutions of the output file in a interactively way with the UCSF Chimera’s extension interface GaudiView. Checkout this tutorial of the GaudiMM’s documentation for visualizing the output.

Diagram of the main idea of PGaudi’s performance