memhog - Allocates memory with policy for testing
memhog [ -r<NUM> ] [ size kmg ] [
policy nodeset ] [ -f<filename> ]
memhog mmaps a memory region for a given size and sets the
numa policy (if specified). It then updates the memory region for the given
number of iterations using memset.
| -r<num> |
Repeat memset NUM times |
| -f<file> |
Open file for mmap backing |
| -H |
Disable transparent hugepages |
| -size |
Allocation size in bytes, may have case-insensitive order |
|
suffix (G=gigabyte, M=megabyte, K=kilobyte) |
Supported numa-policies:
- interleave
- Memory will be allocated using round robin on nodes. When memory cannot be
allocated on the current interleave, target fall back to other nodes.
Multiple nodes may be specified.
- membind
- Only allocate memory from nodes. Allocation will fail when there is not
enough memory available on these nodes. Multiple nodes may be
specified.
- preferred
- Preferably allocate memory on node, but if memory cannot be allocated
there fall back to other nodes. This option takes only a single node
number.
- default
- Memory will be allocated on the local node (the node the thread is running
on)
- # Allocate a 1G region, mmap backed by memhog.mmap file, membind to node
0, repeat test 6 times
- memhog -r6 1G --membind 0 -fmemhog.mmap
- # Allocate a 1G region, iterleave across nodes 0,1,2,3, repeat test 4
times
- memhog -r4 1G --interleave 0-3
- # Allocate a 1G region, (implicit) default policy, repeat test 8
times
- memhog -r8 1G