Originally posted by: SystemAdmin
First of all, we cannot discuss future code extensions in public as this has some subtle legal consequences.
Using the GPU to speed-up MIP solvers is certainly an interesting idea, but it has a number of issues to get it into a product like CPLEX. As an academic you would only need to make sure that it runs on some combination of CPU, GPU, and operating system, and then you can write a paper that discusses the approach and presents computational experiments. For a commercial product like CPLEX, you need to make sure that it basically runs on
any combination of CPU, GPU, and OS, which is much more challenging. So, it boils down (as always) to the trade-off between our investment in development time and other resources compared to the potential benefit that our users would have.
Tobias
#CPLEXOptimizers#DecisionOptimization