The highly anticipated Divide and Conquer Challenge, organized by the Freelancer, on behalf of Bureau of Reclamation Sedimentation and River Hydraulics Group, has reached its exciting conclusion.
Over the course of two phases, talented solvers from around the world showcased their skills and ingenuity in tackling the complex problem of improving flow hydraulics and sediment transport modeling in rivers and reservoirs. After months of intense competition and rigorous evaluation, it is our pleasure to announce the winners who have truly pushed the boundaries of innovation.
A little bit about the winners and solutions:
First Place – Christophe Choquet from Vanves, Ile de France, France.
Christope’s solution impressed the judges with its implementation of a GPU-accelerated Parallel Smoothed Aggregation algorithm. By leveraging previous simulation time steps as input for the finest multi-grid level, Christophe achieved remarkable speedups in the range of 30x to 60x. The parallel solver, combined with optimized CPU to GPU transfers, demonstrated the power of harnessing parallel processing for flow hydraulics and sediment transport modeling.
Second Place – Xiaofeng Liu from State College, PA, USA.
Xiaofeng’s solution provided a unified code that can run in one of three modes: serial, MPI, and OpenMP. This flexibility ensured a seamless transition for users, allowing them to experience the parallel version of the SRH-2D model without any significant changes to their workflow. The solution leveraged the high-performance linear equation system solvers in Hypre, designed for massively parallel computers. Xiaofeng’s approach showcased both performance and usability, making it an exceptional contribution to the field.
Third Place – Mahdi Esmaily from Ithaca, NY, USA.
Mahdi’s solution utilized MPI technology and a graph partitioning algorithm that was developed from scratch. This approach ensured that all components of the SRH-2D model were effectively parallelized. To simplify the parallelization process while maintaining backward compatibility, Mahdi developed a dedicated library for specialized calls to the MPI library. The judges were impressed by the meticulous design and the deep understanding of parallel computing demonstrated in Mahdi’s solution.
Fourth Place – HiCFD team, represented by Zhi Jian Wang from Lawrence, KS, USA.
The HiCFD team adopted domain decomposition and MPI to parallelize the tool across multiple CPU cores. Their solution utilized a preconditioned CGS approach with an incomplete LU decomposition as the linear solver. The combination of domain decomposition and MPI demonstrated their expertise in effectively utilizing distributed computing resources to achieve notable performance improvements.
We extend our sincere congratulations to all the winners for their outstanding achievements throughout the Divide and Conquer Challenge. Their innovative solutions and dedication to advancing the field of flow hydraulics and sediment transport modeling have not only impressed the judges but also brought us closer to more efficient and accurate river and reservoir simulations.
We would also like to express our gratitude to all the participants who contributed their time, skills, and creativity to make this competition a resounding success. The talent and passion displayed by the freelancer community have truly raised the bar for future challenges.
Congratulations once again to the winners and all participants for their remarkable contributions, dedication, and commitment to advancing the field of flow hydraulics and sediment transport modeling.
Stay tuned for more updates on future challenges and opportunities to showcase your skills within the freelancer community!
Leave a Reply