On 7/8/20 11:15 PM, John M. Harris Jr wrote: >> * disk access is literally O(10000) slower than RAM access > This is just false, and you can prove that on your own system using only `dd`. > In fact, if your system is newer than my Lenovo ThinkPad X200 Tablet, you'll > probably have even faster reads/writes from/to disk. > Can you explain how you do this? How do you ensure dd is not using some cache? And how do you measure ram i/o with dd? As far as I understand dd is not a memory benchmarking tool, not even parallised to fully saturate the memory throughput. Just copying to/from ramdisk with dd gave me speeds of below 2 GB/s, but a streaming test [1] gives me 15 GB/s for copy using a single thread and 18 GB/s if I use both cores. Using dd to copy some file, calling time sync before and after, and adding the time after to the estimate by dd gives me an estimate of 30 MB/s for disk speed. So the ratio is only ~1000 but I am not sure how to do it with just dd ... But back to the point - 1000 is still way to slow and I love earlyOOM. I wrote a script myself before I learned about earlyOOM, and I think it is great to have earlyOOM enabled by default. [1] https://www.cs.virginia.edu/stream/FTP/Code/Versions/stream_mpi.c _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx
