Re: [PATCH/RFC 0/6] commit caching

On Wed, Jan 30, 2013 at 2:18 PM, Jeff King [off-list ref] wrote:

On Wed, Jan 30, 2013 at 10:31:43AM +0700, Nguyen Thai Ngoc Duy wrote:

quoted

On Tue, Jan 29, 2013 at 4:14 PM, Jeff King [off-list ref] wrote:

quoted

The timings from this one are roughly similar to what I posted earlier.
Unlike the earlier version, this one keeps the data for a single commit
together for better cache locality (though I don't think it made a big
difference in my tests, since my cold-cache timing test ends up touching
every commit anyway).  The short of it is that for an extra 31M of disk
space (~4%), I get a warm-cache speedup for "git rev-list --all" of
~4.2s to ~0.66s.

Some data point on caching 1-parent vs 2-parent commits on webkit
repo, 26k commits. With your changes (caching 2-parent commits), the
.commits file takes 2241600 bytes. "rev-list --all --quiet":

Hmm. My webkit repo has zero merges in it (though it is the older
svn-based one). What percentage of the one you have are merges? How does
your 1-parent cache perform on something like git.git, where about 25%
of all commits are merges?

git.git performs worse with 1-parent cache. But the point is it should
be customizable.

quoted

The performance loss in 1-parent case is not significant while disk
saving is (although it'll be less impressive after you do Shawn's
suggestion not storing SHA-1 directly)

Yeah, I think moving to offsets instead of sha1s is going to be a big
enough win that it won't matter anymore.

Yeah, if we use uint32_t instead of sha-1, the cache is just about
400k 2 parents for webkit, 312k for 1 parent. The total size is so
small that reduction does not really matter anymore.
-- 
Duy