Thursday, October 25, 2018

Gurobi 8.1

Quite some improvements in MIQP performance. Of course the smaller improvements in other model types also help: over time these things add up to substantial performance gains.


Announcement


Gurobi is pleased to announce the release of Gurobi Version 8.1. This latest version improves the overall performance of Gurobi Optimizer, and adds enhancements to Gurobi Instant Cloud, including support for Microsoft Azure® and for the latest Amazon Web Services® machines, and more.

Version 8.1 demonstrates our commitment to delivering the new features our users request, and includes:

Performance Improvements

Gurobi Optimizer v8.1 continues to push the envelope of solver speed and performance. The overall v8.1 performance improvements versus v8.0 include:

MIQP 
  • More than a factor of 2.8x faster overall.
  • More than a factor of 6x faster on difficult models that take more than 100 seconds to solve.
MIQCP
  • 38% faster overall.
  • 92% faster on difficult models that take more than 100 seconds to solve.
LP
  • 2.9% faster overall in default settings.
  • 6.5% faster on difficult models that take more than 100 seconds to solve.
LP barrier
  • 4.4% faster overall.
  • 11% faster on difficult models that take more than 100 seconds to solve.
LP dual simplex
  • 4.2% faster overall.
  • 10.5% faster on difficult models that take more than 100 seconds to solve.

Enhancements
  • Gurobi Instant Cloud now supports Microsoft Azure®: Instant Cloud users can now use Microsoft Azure, in several regions.
  • Gurobi Instant Cloud adds faster and more powerful machines on Amazon EC2®: The new version supports c5, r5 and z1 instance types.
  • New Q matrix linearization for MIQP and MIQCP models: We added a new Q matrix linearization approach in presolve. This new option can be chosen by setting parameter PreQLinearize to the new value of 2.
  • Improved Mac Installation Package: Users no longer need to install Xcode to perform the installation.
  • Support for Python 3.7: We have added support for Python 3.7 on Windows, Linux and Mac platforms.
  • A callback function for multi-objective optimization: We now let users terminate optimization for individual objectives for multi-objective MIP models.
To learn more about all of the new features and improvements, visit What's New in Gurobi 8.1.

No comments:

Post a Comment