PMEM read media monitored when issuing NTSTORE 256B writes

[Hao Guo]

When I use "pmem_memcpy(p_start, arr, 256, PMEM_F_MEM_NONTEMPORAL)" in PMDK to issue NTSTORE 256B writes to PMEM, and use "ipmctl show -dimm -performance TotalMediaReads" to monitor media read counter at the same time, I discovered that the read counter is increasing at almost the same speed as the write counter, which can be monitored by command "ipmctl show -dimm -performance TotalMediaWrites". But the read request counter ("ipmctl show -dimm -performance TotalReadRequests") shows no increase. Why read media can be monitored at this pure-aligned-write workload? Thanks in advance!

[Wu, Dennis]

PMem internal is a Read-Modify-Write, so you modify the media status, it will have the write. But if you monitor from the DDR-T bus like the pcm, you will not see any read BW.

--
You received this message because you are subscribed to the Google Groups "pmem" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pmem+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pmem/4cfcd50f-1847-4f4f-b0ae-181dff79ec0bn%40googlegroups.com.

[Hao Guo]

Thanks for your consideration and timely reply! I have another question. Does **any** granuality of PMem write cause PMem read-modify-write inside PMem? If so, will PMem write operations influence read operations' bandwidth? That is, if I issue a read op at 6GB/s, then issue a write op at 2GB/s, will the read op's bandwidth drop to less than 4GB/s?