Table of Contents:
00:08 - Optimization of Multithreaded application
00:13 - Optimizing thread placement
00:43 - Threads are not bound to any specific cores
00:49 - OS migrating the threads
01:16 - Why this effects the performance?
01:43 - KMP_AFFINITY environment variable
02:08 - Types of thread placement
02:13 - KMP_AFFINITY=compact
02:28 - KMP_AFFINITY=scatter
02:38 - KMP_AFFINITY=calanced
02:51 - Offset and Permute
03:21 - Additional KMP_AFFINITY parameters
03:50 - Affinity for performance
04:24 - Bandwidth-bound application example
05:44 - KMP_AFFINITY for compute-bound applications
05:56 - DGEMM example
06:36 - KMP_AFFINITY=verbose
06:46 - KMP_PLACE_THREADS