Many modern multi-socket systems are based on non-uniform memory access (NUMA), where access latency and bandwidth depend on the location of the physical memory relative to its use. The art of memory object placement in a NUMA system is in finding patterns to drive heuristics. Introducing new memory hierarchy elements (e.g., high-speed, on-package MCDRAM in Knights Landing) introduces an additional NUMA factor for which a performance-minded engineer needs to account. Understanding the effect of memory object placement on the memory subsystem is key to extracting the best performance out of your platform. We will demonstrate how to use Intel® VTune™ Amplifier to analyze memory objects (dynamic, global, and stack), understand the effects of your choices in data placement on an object basis, and extract the best possible performance out of your system.
Download Slides [PDF 1.47MB]
Benchmark results were obtained prior to the implementation of recent software patches and firmware updates intended to address exploits referred to as "Spectre" and "Meltdown". Implementation of these updates may make these results inapplicable to your device or system.
Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information, see Performance Benchmark Test Disclosure.