Re: [PATCH 2/2] Implement a simple delta_base cache

From: Julian Phillips <julian@quantumfyre.co.uk>
Date: Mar 18, 2007, 05:30 UTC
Message-ID: <Pine.LNX.4.64.0703180517360.24626@beast.quantumfyre.co.uk>
In-Reply-To: <Pine.LNX.4.64.0703172013340.6730@woody.linux-foundation.org>

On Sat, 17 Mar 2007, Linus Torvalds wrote:

Show 14 quoted lines

> On Sat, 17 Mar 2007, Nicolas Pitre wrote:
>>
>> Sure.  But at this point the reference to compare GIT performance
>> against might be GIT itself.  And while 1 second is really nice in this
>> case, there are some repos where it could be (and has already been
>> reported to be) much more.
>
> I'd still like to see the KDE repo, that thing went quiet after it was
> supposed to hit sneaker-net..
>
> If it was 30 seconds before to do a "git log" for some individual file,
> after the recent optimizations it should hopefully be down to 10. And I
> agree that I might be more motivated to try to get it down further if I
> could just find a repository where it's that much.

In my test repository (which emulates a real repository in terms of approximate size in terms of commits, branches and tags) "git log f12000" takes about 15m (using 1.5.0.4). After applying patches 1/2 and 2/2 on top of master I get ~3m50s. With 3/2 as well it goes down a bit more to ~3m20s.

I've attached the script that generated the repository in case you feel the urge to try some move time shaving exercises ... ;)

(This is a rather unrealistic repository consisting of a long series of commits of new binary files, but I don't have access to the repository that is being approximated until I get back to work on Monday ...)

-- 
Julian

  ---
That must be wonderful: I don't understand it at all.
 		-- Moliere

#!/bin/bash

# no. of commits branches and tags to make
commits=25000;
branches=900;
tags=8000;

# create a new file of this size (kb) for each commit
commit_size=102;
bs=1024;

large=$1;

((bg = $commits / $branches));
((tg = $commits / $tags));

echo "creating $large";
mkdir $large;
cd $large;

git init-db;

i=0
while [ $i -lt $commits ]; do
  dd if=/dev/urandom of=f$i bs=${bs} count=${commit_size} > /dev/null 2>&1

  git add f$i;
  git commit -m "add t$i";

  ((ig = $i % $tg));
  if [ $ig -eq 0 ]; then
    git tag t$i;
    echo -n "t";
  fi

  ((ig = $i % $bg));
  if [ $ig -eq 0 ]; then
    git branch b$i;
    echo -n "b";
  fi

  echo -n "$i ";
  ((i = $i + 1))
done

echo;
echo "complete.";