The OPTMILP Option Tuner

Example 16.3 Tuning a Defined Set of Options for Multiple Problems in Distributed Mode

This example demonstrates how to run the tuner in distributed mode. The example is similar to Example 16.2. The only difference between single-machine and distributed mode is that the PERFORMANCE statement specifies the number of threads and nodes to be used. The following statement changes the operating mode to distributed:

   /* set the numbers of nodes and threads and get performance details */
   performance nodes=5 nthreads=4 details;

The performance information and procedure task timing tables are displayed in Output 16.3.1. The NODES=5 and NTHREADS=4 options in the PERFORMANCE statement cause the tuner to run in distributed mode, where each computer node processes up to four threads simultaneously.

Output 16.3.1: Performance Information in Distributed Mode: Output

Tuner Output

Performance Information
Host Node << your grid host >>
Execution Mode Distributed
Number of Compute Nodes 5
Number of Threads per Node 4

Tuner Output

The OPTMILP Procedure

Procedure Task Timing
Task Time
(sec.)
% Time
Data Loading 1.42 0.05%
Data Transfer 0.18 0.01%
Tuner 0.74 0.02%
Solver 3065.59 99.84%
Idle 2.46 0.08%