Re: [PATCH] pack-objects: only throw away data during memory pressure

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, 12 Feb 2008, Brian Downing wrote:

> On Mon, Feb 11, 2008 at 11:00:33AM -0500, Nicolas Pitre wrote:
> > On Mon, 11 Feb 2008, Martin Koegler wrote:
> > > If pack-objects hit the memory limit, it deletes objects from the
> > > delta window.
> > > 
> > > This patch make it only delete the data, which is recomputed, if
> > > needed again.
> > > 
> > > Signed-off-by: Martin Koegler <mkoegler@xxxxxxxxxxxxxxxxx>
> > 
> > Looks fine.
> > 
> > Acked-by: Nicolas Pitre <nico@xxxxxxx>
> 
> Unfortunately this patch (if I understand what it's doing correctly)
> basically defeats my intended use-case for which I wrote the memory
> limiter.  I have a repository with files of very mixed size.  I want the
> window to be very large for small files, for good archival repacking,
> but I don't want it to be very large for my 20+MB files with hundreds of
> revisions, because I want it to finish someday.
> 
> Also, I've gotten into the habit of just doing:
>     git repack --window=100000 --window-memory=256m
> for archival repacks and just letting the memory limit automatically
> size the window.  Basically, I don't really want to specify a window
> size, I just want it to use 512mb of RAM (and go at the speed that size
> of a window would entail.)  While this is slow, it tends to be a
> relatively constant speed, and it tends to find some very interesting
> deltas in my trees that I wouldn't have otherwise expected.
> 
> If this patch is accepted, I'd really like a way to maintain the old
> behavior as an option.

I think your use case has merits, but the previous behavior had 
semantics problems.  We always had constant window size with dynamic 
memory usage, and now we have constant window size with bounded memory 
usage.

If what you want is really to have a dynamic window size using a 
constant memory usage then it needs a different and coherent way to be 
specified.


Nicolas
-
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux