git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: Filters should be parallelized if needed

From
Lars Schneider <larsxschneider@gmail.com>
Date
Jun 21, 2016, 09:36 UTC
Message-ID
<A61012AD-BC22-4640-AE99-000D99D7375F@gmail.com>
In-Reply-To
<ABCC694798EF214E91DB2ACCA61703690BBADA33@MMXDC2642.hermes.si.socgen>
Show 18 quoted lines
> On 21 Jun 2016, at 09:50, LUCAS Thomas (EXT) <thomas.lucas-ext@sgcib.com> wrote:
> 
> Setup:
> - 2.9.0.windows.1 x64 on a windows 7 16g ram, ssd
> - No particular installation settings, default.
> 
> Details:
> - Running git from either a cmd or GitExtensions,  I encountered an issue while staging or checkout of lfs files. I am processing 3000 files for a total of 500MB.
> Files are proccessed sequentially which is too slow for this amount of files, while lfs offer in some cases a way to run it from the command line in parallel, I think the git filters should be run in parallel. I think there is no need (or I am forgetting something) to run it sequentially.
> 
> If I were to use a filter processing files in a parallel way, I would have my time to checkout these files be reduced from 1h30min to 1min30s.
> 
> What actually happened instead?
> - This is a no go for our enterprise to use git-lfs because this is too slow.
> 
> I can't give you my repository because it is confidential.
> 
> Anyway, I hope you can do something about it.
Hi Lucas,

this is a known issue and there are a few ideas how to improve the situation: https://github.com/github/git-lfs/issues/1059 https://github.com/github/git-lfs/issues/931#issuecomment-172939381

I am not sure if anyone is working on the batch/parallel smudge filter processing already (I might have some time for that later this year). AFAIK this is not a trivial change.

That being said, I think the Git mailing list is the wrong place to discuss GitLFS issues as these two projects are separate. I recommend to raise an issue here: https://github.com/github/git-lfs/issues

Cheers, Lars

Previous: LUCAS Thomas (EXT)
Message 2 of 2 in “Filters should be parallelized if needed”
  1. LUCAS Thomas (EXT)Jun 21, 2016
  2. Lars SchneiderJun 21, 2016

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.