https://bugs.kde.org/show_bug.cgi?id=524939
Marek Rudawski <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Ever confirmed|0 |1 CC| |[email protected] Status|REPORTED |CONFIRMED --- Comment #1 from Marek Rudawski <[email protected]> --- I can confirm this on NFS 4.2 (two exports from the same server, different physical disks), where the impact is much larger than on local disks: ~10 MB/s in Dolphin vs ~215 MB/s with cp. strace -T on dolphin shows each 512 KiB copy_file_range() call takes a nearly constant 42-50 ms: copy_file_range(34, NULL, 36, NULL, 524288, 0) = 524288 <0.041703> copy_file_range(34, NULL, 36, NULL, 524288, 0) = 524288 <0.050174> cp, same mounts: a single call, 6.34 GB in 29.5 s: copy_file_range(4, NULL, 5, NULL, 9223372035781033984, 0) = 6342111173 <29.534756> So with a fixed 512 KiB chunk, throughput is capped at roughly 512 KiB / 0.045 s ≈ 11 MB/s regardless of disk or network speed. This supports using a larger (or adaptive) chunk size for copy_file_range(). Mount options: rw,relatime,vers=4.2,rsize=1048576,wsize=1048576,hard,proto=tcp,timeo=600,retrans=2,sec=sys Dolphin 26.08.2, KIO 6.31.0, client kernel 7.2.11-zen1-1-zen (Arch Linux), server Ubuntu with kernel 7.0.0-38-generic. -- You are receiving this mail because: You are watching all bug changes.
