Check page-cache residency of a file without fincore/vmtouch: 15 lines of Python with ctypes mincore

On hosts without util-linux fincore or vmtouch (and without root), you can still check how much of a file is cached before calling a benchmark run 'cold'. Use mmap plus mincore through ctypes (Linux):

import ctypes, os, sys
libc = ctypes.CDLL(None, use_errno=True)
libc.mmap.restype = ctypes.c_void_p
libc.mmap.argtypes = [ctypes.c_void_p, ctypes.c_size_t, ctypes.c_int, ctypes.c_int, ctypes.c_int, ctypes.c_long]
libc.mincore.argtypes = [ctypes.c_void_p, ctypes.c_size_t, ctypes.c_char_p]
libc.munmap.argtypes = [ctypes.c_void_p, ctypes.c_size_t]
PROT_READ, MAP_SHARED = 1, 1
path = sys.argv[1]; fd = os.open(path, os.O_RDONLY); n = os.fstat(fd).st_size
pg = os.sysconf('SC_PAGE_SIZE'); npg = (n + pg - 1) // pg
addr = libc.mmap(None, n, PROT_READ, MAP_SHARED, fd, 0)
vec = ctypes.create_string_buffer(npg)
if libc.mincore(ctypes.c_void_p(addr), n, vec) != 0: raise OSError(ctypes.get_errno(), 'mincore')
res = sum(b & 1 for b in vec.raw)
print(f'{res*pg/1e6:.1f} MB resident of {n/1e6:.1f} MB ({100*res/npg:.1f}%)')
libc.munmap(ctypes.c_void_p(addr), n); os.close(fd)

Notes: - mmap with PROT_READ doesn't fault pages in, so the check doesn't warm the file itself. - For multi-GB files the vector is one byte per page (~2.5 MB for 10 GB), which is fine. - To make a file cold without root: os.posix_fadvise(fd, 0, 0, os.POSIX_FADV_DONTNEED). It works for clean pages, and rerunning the check confirms it. - Log the resident MB next to every 'cold' timing. Label a run 'partly warm' if more than a few % of the file is resident.