Re: Fix "git commit directory/" performance anomaly

From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Aug 10, 2007, 17:14 UTC
Message-ID: <alpine.LFD.0.999.0708101008230.30176@woody.linux-foundation.org>
In-Reply-To: <alpine.LFD.0.999.0708100924570.30176@woody.linux-foundation.org>

On Fri, 10 Aug 2007, Linus Torvalds wrote:

>
> This improves the cold-cache case immensely, since we don't need to bring 
> in all the file contents, just the index and any files dirty in the index.

The extreme - but not necessarily unusual - case of this is when the index and stat information is hot-cache (which is quite normal, in case you've done a "git diff" or something like that), but the whole source tree is otherwise not. Then the numbers look like this:

Before:
	[torvalds@woody linux]$ time git commit .
	real    1m33.751s
	user    0m3.864s
	sys     0m2.160s

After:
	[torvalds@woody linux]$ time git commit .
	real    0m1.415s
	user    0m1.176s
	sys     0m0.260s

The all-hot-cache numbers are better too, of course, just not nearly as noticeable:

Before:
	[torvalds@woody linux]$ time git commit .
	real    0m3.960s
	user    0m3.304s
	sys     0m0.672s

(and the after case is that 1.415s above, of course, so it's still more than twice as fast - it's just no longer a 60x performance difference!).

(Honesty in advertising: the "after" numbers also contain the "runstatus" speedup, which is the 0.5s -> 0.3s improvement)

			Linus