Re: git and larger trees, not so fast?

On Sat, Aug 11, 2007 at 10:06:30PM +0200, moe wrote:

On Sat, Aug 11, 2007 at 11:47:42AM -0700, Linus Torvalds wrote:

quoted

On Thu, 9 Aug 2007, moe wrote:

quoted

here's a test-case (should be safe to
copy/paste on linux, bash):

moe: with current git (and thus the 1.5.3 release), the "git status" 
commands now take half a second for me, and the git commit takes just 
under a second.

The *initial* commit that adds everything still takes almost 5 seconds, 
but that was due to generating the diffstat summary - with a "-q" on the 
commit line that too drops down to just under a second.

In fact, the only thing that took more than a second for me with the 
current git is that initial "git add .", which took 1.791s for me. 
Considering that it had to hash all the 100,000 objects, I'm not 
surprised.

Anyway, it would be good if you re-did your real work tree with current 
commit, just to verify. You have slower hardware than I do, but hopefully 
it is now just about as fast as it can be.

 
hi linus,

thx for your efforts, the figures look very promising.
i'm out of town right now but will test when i get
stationary internet again (sometime tomorrow evening
i think).

sorry for late followup, i was blocked on paid work
for longer than i expected.

i can happily confirm what others have already reported;
with the patches applied git works well even for my
bigger repo.

git status              : 0m1.036s
git commit (single file): 0m1.846s

http://www.gosimpsons.com/ProdImages/krustysealkeychain.jpg


regards, moe