From: Jeff King <peff@peff.net>
Date: Wed, 22 Aug 2018 15:24:54 GMT
Subject: Re: [ANNOUNCE] Git v2.19.0-rc0
Message-ID: <20180822152453.GD32630@sigill.intra.peff.net>
In-Reply-To: <459f138f-78ba-6465-0f2b-b2419bbc36ff@gmail.com>

On Wed, Aug 22, 2018 at 10:28:56AM -0400, Derrick Stolee wrote:

> In my testing, I've had the best luck with this change:
> 
> diff --git a/cache.h b/cache.h
> index b1fd3d58ab..6c8b51c390 100644
> --- a/cache.h
> +++ b/cache.h
> @@ -1023,7 +1023,14 @@ extern const struct object_id null_oid;
> 
>  static inline int hashcmp(const unsigned char *sha1, const unsigned char
> *sha2)
>  {
> -       return memcmp(sha1, sha2, the_hash_algo->rawsz);
> +       switch (the_hash_algo->rawsz) {
> +               case 20:
> +                       return memcmp(sha1, sha2, 20);
> +               case 32:
> +                       return memcmp(sha1, sha2, 32);
> +               default:
> +                       assert(0);
> +       }
>  }
> 
> The fact that '20' and '32' are constants here may be helpful to the
> compiler. Can someone else test the perf?

I tested that one last night (and just re-tested it now to be sure). It
seems to just generate two separate calls to memcmp, with no speed
improvement.

-Peff