Created attachment 132618 [details]
IO using cp and kcp

Copying files over KIO is significantly slower than native filesystem actions when performing on many small files.
I did a test with 40000 files of 50K for
a) On a tmpfs cp took about a second, kcp took about 6 seconds. I guess this is a good measure of the cpu overhead of KIO copy.
b) On a hard drive cp took about 53 seconds, kcp took about 3m41s. Now this is a very significant difference.

I monitored the process in KSysGuard, and I found that copying over KIO leads to much less reading operations per second and a much lower reading speed, as you can see from the attached screenshots. (kcp reads data at about 1/10 of the rate of cp!)

I guess one possible reason for this could be an artificial limit to read ops to prevent system freezes on copy operations?