I'm using git 1.6.1.3 and it seems to be limited to a single core. Given that I've seen cases where the cpu has been basically pinned for minutes on end (initial clone of a repository, for instance) has there been any discussion of taking advantage of multiple cores?
Chris
Re: git and multiple cores
On Tue, Jun 2, 2009 at 6:40 PM, Chris Friesen wrote:
> > I'm using git 1.6.1.3 and it seems to be limited to a single core. > Given that I've seen cases where the cpu has been basically pinned for > minutes on end (initial clone of a repository, for instance) has there > been any discussion of taking advantage of multiple cores?
Sounds like you're mostly concerned about packing.
The good news is, your version of git already has a threaded packer. You just need to enable it. See "pack.threads" in "git help config".
1.6.2 and newer use multiple threads by default.
Peter Harris
Shawn O. PearceRe: git and multiple cores
Peter Harris <git@peter.is-a-geek.org> wrote:
> On Tue, Jun 2, 2009 at 6:40 PM, Chris Friesen wrote: > > > > I'm using git 1.6.1.3 and it seems to be limited to a single core. > > Given that I've seen cases where the cpu has been basically pinned for > > minutes on end (initial clone of a repository, for instance) has there > > been any discussion of taking advantage of multiple cores? > > Sounds like you're mostly concerned about packing. > > The good news is, your version of git already has a threaded packer. > You just need to enable it. See "pack.threads" in "git help config". > > 1.6.2 and newer use multiple threads by default.
True, but he was talking about initial clone, which on the client side is git-index-pack. Which is not threaded.
-- Shawn.
Re: git and multiple cores
On Tue, Jun 2, 2009 at 7:02 PM, Shawn O. Pearce wrote:
> Peter Harris <git@peter.is-a-geek.org> wrote: >> On Tue, Jun 2, 2009 at 6:40 PM, Chris Friesen wrote: >> > >> > I'm using git 1.6.1.3 and it seems to be limited to a single core. >> > Given that I've seen cases where the cpu has been basically pinned for >> > minutes on end (initial clone of a repository, for instance) has there >> > been any discussion of taking advantage of multiple cores? >> >> Sounds like you're mostly concerned about packing. >> >> The good news is, your version of git already has a threaded packer. >> You just need to enable it. See "pack.threads" in "git help config". >> >> 1.6.2 and newer use multiple threads by default. > > True, but he was talking about initial clone, which on the client > side is git-index-pack. Which is not threaded.
Ah. I thought he was talking about the server side. My mistake.
Peter Harris
Re: git and multiple cores
Peter Harris wrote:
> On Tue, Jun 2, 2009 at 7:02 PM, Shawn O. Pearce wrote: > >>Peter Harris <git@peter.is-a-geek.org> wrote: >> >>>On Tue, Jun 2, 2009 at 6:40 PM, Chris Friesen wrote: >>> >>>>I'm using git 1.6.1.3 and it seems to be limited to a single core. >>>>Given that I've seen cases where the cpu has been basically pinned for >>>>minutes on end (initial clone of a repository, for instance) has there >>>>been any discussion of taking advantage of multiple cores? >>> >>>Sounds like you're mostly concerned about packing. >>> >>>The good news is, your version of git already has a threaded packer. >>>You just need to enable it. See "pack.threads" in "git help config". >>> >>>1.6.2 and newer use multiple threads by default. >> >>True, but he was talking about initial clone, which on the client >>side is git-index-pack. Which is not threaded. > > > Ah. I thought he was talking about the server side. My mistake.
Sorry, I wasn't clear. I was talking about the client side, although the server side information is useful to have.
I have a 4-way machine as a client, and it just seemed odd that git could only use one core.
Chris
Nicolas PitreRe: git and multiple cores
On Tue, 2 Jun 2009, Chris Friesen wrote:
> Peter Harris wrote: > > On Tue, Jun 2, 2009 at 7:02 PM, Shawn O. Pearce wrote: > > > >>True, but he was talking about initial clone, which on the client > >>side is git-index-pack. Which is not threaded. > > > > > > Ah. I thought he was talking about the server side. My mistake. > > Sorry, I wasn't clear. I was talking about the client side, although > the server side information is useful to have. > > I have a 4-way machine as a client, and it just seemed odd that git > could only use one core.
It can use them all when repacking or pushing. But on a clone or fetch your client machine deals with incoming data using index-pack which is much more trickier to thread.
Nicolas