Specifications

April 2012 v1 AMD Opteron™ 6200 Linux Tuning Guide
23
• 16 jobs - Use all L2 caches and shared floating point units: 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28,
30.
• 32 jobs - Use all L2 caches and shared floating point units: 0,1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18,
19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31.
- The best resource hogging allocation would maximize the area under the graph if plotting total node
performance for each number of cores active going from 0 to 32 cores.
- Sharing is only an issue when a particular program or workload mix on a node hits a limit.
- Jobs can be bound to cores with the numactl utility. For example, the following runs a job on core 8
allocating memory local to that core.
- numactl --physcpubind 8 –localalloc job_name