mmap-cxlmem.fio

#
# run command for this workload:
# you can configure tmpfs to emulate cxl.mem if you do not have device as below. or configure cxl.mem into fsdax mode to test
# mount -t tmpfs -o size=20G tmpfs /mnt/mytmpfs
#
[global]
thread
group_reporting
norandommap
overwrite=1
thinktime=0
sync=0
direct=0

ioengine=mmap
#
# Important note: this workload is run on a tmpfs as emulation of cxl.mem, file system, mounting
filename=/mnt/mytmpfs/test1
size=1GB
#offset_increment=50G  # each of 'numjobs' will get its own space within the device

time_based
runtime=300
ramp_time=30

[SeqW]
bs=4k
numjobs=1  # depending on CPU core count more can be better, but also can be a bottleneck
iodepth=1

rw=write
#rwmixread=100
#percentage_random=0