Re: [PATCH] fast-import: use hashcmp() for SHA1 hash comparison

From: René Scharfe <l.s.r@web.de>
Date: Jul 18, 2014, 19:14 UTC
Message-ID: <53C971FD.6040500@web.de>
In-Reply-To: <20140718184246.GS12427@google.com>

Am 18.07.2014 20:42, schrieb Jonathan Nieder:

Show 25 quoted lines

> René Scharfe wrote:
>
>> Signed-off-by: Rene Scharfe <l.s.r@web.de>
>> ---
>>   fast-import.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> Before:
>
>    $ size git-fast-import
>       text    data     bss     dec     hex filename
>     804138    6768  754160 1565066  17e18a git-fast-import
>
> After:
>
>    $ size git-fast-import
>       text    data     bss     dec     hex filename
>     804154    6768  754160 1565082  17e19a git-fast-import
>
> So this makes the text size worse on this machine (amd64, building
> with gcc 4.8.2 -O2).  That's probably because the old code does 'call
> memcmp', while the new code inlines it.  Inlining is presumably the
> better choice.
>
> More importantly, the new code is more readable.

Yes, the latter point is the important one.

If inlining is really better is another matter; I don't understand how 1a812f3a (hashcmp(): inline memcmp() by hand to optimize) could have made git gc 18% faster, as it claimed. I would expect memcmp(), which can compare more than a byte at a time, to be significantly faster -- or at least just as fast as whatever the compiler does with the inlined version.

René