#include "precompiled.h"#include "lib/sysdep/numa.h"#include "lib/bits.h"#include "lib/sysdep/os_cpu.h"
Include dependency graph for unuma.cpp:
Functions | |
| size_t | numa_NumNodes () |
| size_t | numa_NodeFromProcessor (size_t processor) |
| uintptr_t | numa_ProcessorMaskFromNode (size_t node) |
| size_t | numa_AvailableMemory (size_t node) |
| double | numa_Factor () |
| bool | numa_IsMemoryInterleaved () |
Function Documentation
◆ numa_AvailableMemory()
| size_t numa_AvailableMemory | ( | size_t | node | ) |
- Parameters
-
node
- Returns
- bytes of memory available for allocation on <node>.
◆ numa_Factor()
| double numa_Factor | ( | ) |
- Returns
- the ratio between maximum and minimum times that one processor from each node required to fill a globally allocated array. in other words, this is the maximum slowdown for NUMA-oblivious memory accesses. Microsoft guidelines require it to be <= 3.
◆ numa_IsMemoryInterleaved()
| bool numa_IsMemoryInterleaved | ( | ) |
- Returns
- an indication of whether memory pages are node-interleaved.
note: this requires ACPI access, which may not be available on least-permission accounts. the default is to return false so as not to cause callers to panic and trigger performance warnings.
◆ numa_NodeFromProcessor()
| size_t numa_NodeFromProcessor | ( | size_t | processor | ) |
- Parameters
-
processor
- Returns
- node number (zero-based) to which <processor> belongs.
◆ numa_NumNodes()
| size_t numa_NumNodes | ( | ) |
- Returns
- number of NUMA "nodes" (i.e. groups of CPUs with local memory).
◆ numa_ProcessorMaskFromNode()
| uintptr_t numa_ProcessorMaskFromNode | ( | size_t | node | ) |
- Parameters
-
node
- Returns
- bit-mask of all processors constituting <node>.
