Re: Multithreading

Michael Haggerty <[email protected]> Fri, 26 Jul 2013 05:24:44 -0700
Newsgroups gmane.comp.version-control.subversion.cvs2svn.user
Message-ID <[email protected]>
On 07/25/2013 03:40 AM, Fabian Schäfer wrote:
> I am searching for performance optimizations concerning an cvs2git
> migration. Therefore my question: Does cvs2git support multiple processors?
> 
> If not, is there any chance to distribute the load?

For cvs2git, some parallelization is possible:

* If you use the external blob generator, then the revision contents are
extracted from CVS in a separate process.

* Once FilterSymbolsPass is done, you can already start loading the
blobs into git using git-fast-import before the rest of the cvs2git
passes are done.  This can even be done on a different computer by
moving the blobfile.

Some other work could be parallelized relatively easily.  For example,
if the dumpfile were written a bit differently, then it could be piped
directly into "git fast-import" rather than written to a file that needs
to be loaded after the conversion is done.  But so far nobody has
volunteered to work on this.

To get the fastest conversion, I recommend running it on a computer with
a big ramdisk, and store all of the temporary files on the ramdisk.
(It's probably not necessary to hold the blobfile or dumpfile on the
ramdisk.)

Michael

-- 
Michael Haggerty
[email protected]
http://softwareswirl.blogspot.com/

------------------------------------------------------
http://cvs2svn.tigris.org/ds/viewMessage.do?dsForumId=1670&dsMessageId=3061385

To unsubscribe from this discussion, e-mail: [[email protected]].