标签
This technical blog post explains the performance impact of memory alignment in SIMD vectorization, covering architectures with strict alignment requirements, cacheline crossing, and the behavior of modern CPUs.
本文解释了非统一内存访问（NUMA）的概念、历史背景以及它在多插槽服务器上对性能的影响，同时介绍了Edera在使基于Xen的虚拟化实现端到端NUMA感知方面所做的工作。