Re: [PATCH] remove ARM and Mozilla SHA1 implementations

From: Nicolas Pitre <nico@cam.org>
Date: Aug 24, 2009, 17:27 UTC
Message-ID: <alpine.LFD.2.00.0908241318000.6044@xanadu.home>
In-Reply-To: <alpine.DEB.1.00.0908241849160.8306@pacific.mpi-cbg.de>

On Mon, 24 Aug 2009, Johannes Schindelin wrote:

Show 40 quoted lines

> Hi,
> 
> On Mon, 24 Aug 2009, Johannes Schindelin wrote:
> 
> > On Mon, 24 Aug 2009, Nicolas Pitre wrote:
> > 
> > > On Tue, 18 Aug 2009, Johannes Schindelin wrote:
> > > 
> > > > On Mon, 17 Aug 2009, Nicolas Pitre wrote:
> > > > 
> > > > > They are both slower than the new BLK_SHA1 implementation, so it is 
> > > > > pointless to keep them around.
> > > > > 
> > > > > Signed-off-by: Nicolas Pitre <nico@cam.org> ---
> > > > > 
> > > > > Someone else would need to make the call for the PPC version.
> > > > 
> > > > If I don't forget, I can test tomorrow on 2 different 32-bit PPCs and 
> > > > possibly one 64-bit PPC.
> > > 
> > > Did you forget?  ;-)
> > 
> > Yes.
> 
> At long last (XTools took a real long time to install here; it is an old 
> PowerPC running MacOSX):
> 
> Best of 10 "git rev-list --all" runs on a full Git repository (including 
> my own tree):
> 
> before 30ae47b4
> 
> 	1.85 real, 1.52 user, 0.28 sys
> 
> after 30ae47b4
> 
> 	1.86 real, 1.52 user, 0.28 sys
> 
> To be frank, the 1.85 looks like an outlier, so I think there is exactly 0 
> difference between the two.

Maybe there wasn't any _code_ difference after all. According to the Makefile, only mingw defines NO_OPENSSL (although there might certainly be others).

TRy a build with PPC_SHA1=1, and then compare with BLK_SHA1=1. And best is to time a fsck --full.

Nicolas