2429. Solving LPs and MIPs with Gurobi on the GPU: What's in for me?
Invited abstract in session WD-9: Gurobi Workshop, stream PC Stream.
Wednesday, 15:15-16:00Room: H15
Authors (first author is the speaker)
| 1. | Robert Luce
|
| Gurobi |
Abstract
Recent GPU hardware generations offer high compute throughput for many
applications in machine learning, scientific computing, and beyond.
How does this new compute power translate into improvements in solving
real-world linear and mixed-integer optimization problems in practice?
In this workshop, we review Gurobi's efforts and research findings on
solving optimization problems on the GPU. We touch interior point
methods, first-order methods, and relaxation-based MIP heuristics, and
we frame these in terms of their applicability and usability on the
GPU. Also, we discuss a few implementation aspects that only arise
when implementing such traditional methods on the GPU.
Finally, we'll showcase how Gurobi's preliminary GPU-based components
compare, performance-wise, to our standard CPU-based counterparts
across a wide range of real-world models.
Keywords
- Computational Experiments
- Large Scale Optimization
- Linear Programming
Status: accepted
Back to the list of papers