mirror of https://gitee.com/openkylin/numactl.git
43ff68b8ba
|
||
|---|---|---|
| .. | ||
| README | ||
| bind_range | ||
| checkaffinity | ||
| checktopology | ||
| distance.c | ||
| ftok.c | ||
| mbind_mig_pages.c | ||
| migrate_pages.c | ||
| move_pages.c | ||
| mynode.c | ||
| node-parse.c | ||
| nodemap.c | ||
| numademo | ||
| pagesize.c | ||
| prefered.c | ||
| printcpu | ||
| randmap.c | ||
| realloc_test.c | ||
| regress | ||
| regress-io | ||
| regress2 | ||
| runltp | ||
| shmtest | ||
| tbitmap.c | ||
| tshared.c | ||
README
Various simple test scripts to verify some parts of the NUMA API.
To do a full regression test run make test
You should have at least two nodes on a NUMA system for the test suite.
The tests in regress assume that there is enough memory free on nodes 0/1.
They consider PREFERRED/INTERLEAVE not hitting the first choice node an
error.
They also require a relatively idle machine to avoid too much
noise from memory allocation from other processes. Without
that regress1 might fail.
You can run the tests under valgrind with VALGRIND=valgrind make test
Older valgrind versions incorrectly report a uninitialized byte error
on set_mempolicy. That is a false positive.
TBD: more detailed unit tests for mbind / shm / {get,set}_mempolicy
Currently everything is tested using numactl only.