Re: [PATCH] lib/sha1: use the git implementation of SHA-1

[Nicolas Pitre]

On Sun, 7 Aug 2011, Linus Torvalds wrote:

> On Sun, Aug 7, 2011 at 1:48 PM, Joachim  Eastwood <manabian@xxxxxxxxx> wrote:
> >
> > yes, this works. At least my board boots as normal.
> 
> Ok, I'll remove it for -rc1, just to have a working ARM setup. Maybe
> we can re-introduce it later (either together with some arm-specific
> hack for SHA_WORKSPACE_WORDS or by having an arm-optimized version of
> the *good* sha1 routine).
> 
> But I doubt it: there used to be an ARM-optimized thing in git too. It
> was removed two years ago with the commit message:
> 
>     remove ARM and Mozilla SHA1 implementations
> 
>     They are both slower than the new BLK_SHA1 implementation, so it is
>     pointless to keep them around.
> 
> and quite frankly, that removed code seems to be the same as the
> in-kernel one. So I bet the ARM "optimized" SHA1 is simply not worth
> keeping around.

Indeed.  The ARM code in the kernel is mine:

|commit c09f98271f685af349d3f0199360f1c0e85550e0
|Author: Nicolas Pitre <nico@xxxxxxx>
|Date:   Fri Oct 28 15:26:40 2005 +0100
|
|    [ARM] 2930/1: optimized sha1 implementation for ARM
|
|    Patch from Nicolas Pitre
|
|    Here's an ARM assembly SHA1 implementation to replace the default C
|    version. It is approximately 50% faster than the generic C version. On
|    an XScale processor running at 400MHz:
|        generic C version:      9.8 MB/s
|        my version:             14.5 MB/s
[...]

I eventually added it to Git as well.  After a while I worked on the C 
version in Git:

|$ git log --reverse --oneline --author="Nicolas Pitre" git/block-sha1/
|30ba0de block-sha1: move code around
|dc52fd2 block-sha1: split the different "hacks" to be individually selected
|660231a block-sha1: support for architectures with memory alignment restrictions
|ee7dc31 block-sha1: more good unaligned memory access candidates
|d5f6a96 block-sha1: make the size member first in the context struct
|51ea551 make sure byte swapping is optimal for git
|e9c5dcd block-sha1: guard gcc extensions with __GNUC__
|30ae47b remove ARM and Mozilla SHA1 implementations

I therefore think that the ARM version in the kernel should also be 
removed if the Git C version goes in.


Nicolas
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/