You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There are couple of problems that should be fixed:
The support_numa.c with alloc_1MB_shared opens and mmaps a file and while the test is executed in unique temporary directory the helper is called a few times for the same file, we should ensure that the pages are not present in the page cache on subsequent calls. I guess that unlinking the file before we open it should suffice.
The measurements extracted from numastat output show only overall status for the whole process, we should parse /proc/$pid/numa_maps instead where we can get precise number of pages per node for the test file.
The test should be split into separate testcases in the numa runtest file, so instead of "numa tests failed" we will get more granular results.
The text was updated successfully, but these errors were encountered:
Also the scenarios where the memory is backed up by a file should be executed against all supported filesystems to make sure the allocation is done correctly.
There are couple of problems that should be fixed:
The support_numa.c with alloc_1MB_shared opens and mmaps a file and while the test is executed in unique temporary directory the helper is called a few times for the same file, we should ensure that the pages are not present in the page cache on subsequent calls. I guess that unlinking the file before we open it should suffice.
The measurements extracted from numastat output show only overall status for the whole process, we should parse /proc/$pid/numa_maps instead where we can get precise number of pages per node for the test file.
The test should be split into separate testcases in the numa runtest file, so instead of "numa tests failed" we will get more granular results.
The text was updated successfully, but these errors were encountered: