On Thu, Sep 27, 2012 at 7:47 AM, Shawn Pearce <spea...@spearce.org> wrote:
> Google has published a series of patches (see links below) to JGit to

Should discussions about this series happen in here, jgit mailing or
gerrit? I just want to make sure I'll discuss it at the right place.

> improve fetch and clone performance by adding compressed bitmaps to
> the pack-*.idx structure.
> Operation                   Index V2               Index VE003
> Clone                       37530ms (524.06 MiB)     82ms (524.06 MiB)
> Fetch (1 commit back)          75ms                 107ms
> Fetch (10 commits back)       456ms (269.51 KiB)    341ms (265.19 KiB)
> Fetch (100 commits back)      449ms (269.91 KiB)    337ms (267.28 KiB)
> Fetch (1000 commits back)    2229ms ( 14.75 MiB)    189ms ( 14.42 MiB)
> Fetch (10000 commits back)   2177ms ( 16.30 MiB)    254ms ( 15.88 MiB)
> Fetch (100000 commits back) 14340ms (185.83 MiB)   1655ms (189.39 MiB)

Beautiful. And curious, why do 100->1000 and 10000->10000 have such
big leaps in time (V2)?

> The basic gist of the implementation is a bitmap has a 1 bit set for
> each object that is reachable from the commit the bitmap is associated
> with. An index file may have a unique bitmap for hundreds of commits
> in the corresponding pack file. The set of objects to send is
> performed by doing a simple computation:
>   OR (all want lines) AND NOT OR (all have lines)
> There are two key patches in the series that implement the file format
> change and logic involved:
> * https://git.eclipse.org/r/7939
>   Defines the new E003 index format and the bit set
>   implementation logic.

I suppose the index format is not set in stone yet? My java-foo is
rusty and I'm not familiar with jgit, so I more likely read things

It seems the bitmap data follows directly after regular index content.
I'd like to see some sort of extension mechanism like in
$GIT_DIR/index, so that we don't have to increase pack index version
often. What I have in mind is optional commit cache to speed up
rev-list and merge, which could be stored in pack index too.

In PackIndexVE003 class

+               // Read the bitmaps for the Git types
+               SimpleDataInput dataInput = new SimpleDataInput(fd);
+               this.commits = readBitmap(dataInput);
+               this.trees = readBitmap(dataInput);
+               this.blobs = readBitmap(dataInput);
+               this.tags = readBitmap(dataInput);

Am I correct in saying that you have four different on-disk bitmaps,
one for each object type? If so, for compression efficient reasons?

> :-)

Definitely :-). I have shown my interest in this topic before. So I
should probably say that I'm going to work on this on C Git, but
sllloooowwwly. As this benefits the server side greatly, perhaps a
GitHubber ;-) might want to work on this on C Git, for GitHub itself
of course, and, as a side effect, make the rest of us happy?
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to