I also wrote a copy version that orders file sequence on disk
efficiently, using write through, and posted it. This speeds
up any
subsequent file system operations done in the directory order
as if you
have done a defrag. Great for hard drives, but not needed for
ssd.
https://github.com/jnorwood
Any change you could turn this into a pull request and submit
to Phobos?
Looks like I never promoted the write through code on github. I
just posted it on the forum.dlang.org, along with some timings. I
was only experimenting on ntfs, and so the extension I posted
doesn't force write-through on linux.
http://forum.dlang.org/thread/gmkocaqzmlmfbuozh...@forum.dlang.org
After re-reading my post, I don't think I experimented with
write through in copydir. I show the timings in the post for
ntfs operations on a 2GB layout after unzipping with a write
through version of unzip. I suspect you'd see similar improvement
following a copydir that uses write through. The problem I saw
with ntfs was that the target directory was very fragmented when
allowing ntfs to do its lazy flushes. I tried several
experiments, and using the write through was the only solution
that resulted in a reasonably defragged target immediately as the
result of the unzip.
