How Computers Work
NUMA
memory closer to some cores than others (many-socket)
Source board
This definition comes from the board below. Open it to see the term in context.
The Memory HierarchyHow Computers WorkMore terms from this board
- cachesmall fast memory near the CPU
- cache linethe ~64B unit moved between levels
- DRAMmain memory (RAM)
- hit ratefraction of hits
- hit/missfound in cache / had to go slower
- latency vs bandwidthper-access delay vs throughput
- LRU/associativitythe cache's eviction policy
- memory wallCPU-vs-RAM speed gap
- MMU/page tablethe address translator
- prefetchCPU guesses & preloads the next line
- spatial localityuse nearby data
- swap/pagingspilling RAM to disk when full
- temporal localityreuse soon
- TLBa cache of translations
- virtual memoryper-process address space
- working setthe memory a program touches often