Git super slow on Windows 7

Hi Johannes,

I am working with Git for Windows right now and it is dramatically slower than on OS X. I executed "time git status" on Windows and OS X with the same repository and the following results:

## Windows git version 2.6.3.windows.1 (with enabled experimental flag on install): real 0m1.327s user 0m0.000s sys 0m0.015s

## OS X git version 2.4.9 (Apple Git-60): git status 0.06s user 0.13s system 102% cpu 0.186 total

Initially it was even slower on Windows (~1.6s). According to [1] I used the following settings to make it faster: $ git config --global core.preloadindex true $ git config --global core.fscache true

Is this behavior normal/expected? If it is not normal, how would you debug the issue? How can I find out why it is so slow?

My user drive is not on a net share and the machine has a SSD.

Thanks, Lars

[1] http://stackoverflow.com/questions/4485059/git-bash-is-extremely-slow-in-windows-7-x64

After some investigation I figured that ~50 Submodules are the culprit. Does anyone have an idea how to speed up Git on Windows while keeping 50 Submodules?

Thanks, Lars

On Wed, Nov 25, 2015 at 10:42 AM, Lars Schneider <larsxschneider@gmail.com> wrote:

Use the latest version of Git ;)

Checkout the series merged at 65e1449 (2015-10-05, Merge branch 'sb/submodule-helper')

    The infrastructure to rewrite "git submodule" in C is being built
    incrementally.  Let's polish these early parts well enough and make
    them graduate to 'next' and 'master', so that the more involved
    follow-up can start cooking on a solid ground.

    * sb/submodule-helper:
      submodule: rewrite `module_clone` shell function in C
      submodule: rewrite `module_name` shell function in C
      submodule: rewrite `module_list` shell function in C

More specifically the commits in there:

submodule: rewrite `module_name` shell function in C

    This implements the helper `name` in C instead of shell,
    yielding a nice performance boost.

    Before this patch, I measured a time (best out of three):

      $ time ./t7400-submodule-basic.sh  >/dev/null
        real 0m11.066s
        user 0m3.348s
        sys 0m8.534s

    With this patch applied I measured (also best out of three)

      $ time ./t7400-submodule-basic.sh  >/dev/null
        real 0m10.063s
        user 0m3.044s
        sys 0m7.487s

On Wed, Nov 25, 2015 at 7:47 PM, Stefan Beller <sbeller@google.com> wrote:

Does it do parallel refresh yet? I think it would help. I only looked at "git log --merges origin/pu" and nothing caught my eyes.

On Wed, Nov 25, 2015 at 12:23 PM, Duy Nguyen <pclouds@gmail.com> wrote:

No. The hinted patch series only does a partial shell->C conversion, which is the best guess for improving git status here.

I punted on parallel local operations inside "git submodule update" for now, too as when things go wrong there, you need a human to resolve the merge conflict, and as a user you only want to deal with one merge conflict at a time instead of being left there with a ton of unresolved issues (according to the git log of older patches in the submodule area).

git status should require not human interaction if things go bad within submodules, so we may want to speed that up by parallelizing the submodule part. The status command gathers all information of submodules by a call to "git submodule summary" and does some slight post processing on the output. "git submodule summary" however is written completely in shell code (200 lines, so I estimate 400 lines of C).

I will benchmark that later today and check if it's worth for us to rewrite that in C for our case (we plan to have lots more submodules, but we're a linux shop)