Multithreading processing?

[Benoit Boissinot]

On Fri, Apr 23, 2010 at 12:10 AM, Greg Ward <greg-hg at gerg.ca> wrote:
> On Thu, Apr 22, 2010 at 5:22 PM, Matt Mackall <mpm at selenic.com> wrote:
>>> Also, conversion is painfully slow, whether with hg convert or an
>>> external tool like cvs2hg.  When I was playing around with various
>>> ways to convert from CVS, git-fast-import was able to process several
>>> gigabytes of output from cvs2git in around 1 minute, whereas cvs2hg
>>> took 16 or 18 hours to generate an equivalent hg repo.  It could be
>>> that I did something horribly wrong in cvs2hg, but "hg convert" and
>>> hg-fastimport have similar performance problems -- so at least I'm not
>>> the only one.  ;-)
>>
>> Conversion is, for most people, a one-time thing. I could sink a lot of
>> effort into optimizing it, but I think that effort is better spent
>> elsewhere.
>
> It's also the first thing people see.  If git-fast-import takes 60 sec
> where hg-fast-import (or hg convert or cvs2hg) takes 15 hours, that
> gets your attention.

And how long is the repack after the git-fast-import (I guess it
doesn't repack during the import) ?
While hg shouldn't need repack (if we had parent-delta, that would be
mostly true).

Benoit