From: Jeff King <peff@peff.net>
Date: Fri, 20 Jul 2018 05:30:32 GMT
Subject: Re: 2.18.0 Regression: packing performance and effectiveness
Message-ID: <20180720053031.GA16126@sigill.intra.peff.net>
In-Reply-To: <20180720052829.GA3852@sigill.intra.peff.net>

On Fri, Jul 20, 2018 at 01:28:29AM -0400, Jeff King wrote:

> @@ -162,8 +166,10 @@ struct object_entry *packlist_alloc(struct packing_data *pdata,
>  
>  		if (!pdata->in_pack_by_idx)
>  			REALLOC_ARRAY(pdata->in_pack, pdata->nr_alloc);
> +		pack_delta_lock(pdata);
>  		if (pdata->delta_size)
>  			REALLOC_ARRAY(pdata->delta_size, pdata->nr_alloc);
> +		pack_delta_unlock(pdata);
>  	}

This made me wonder if in_pack_by_idx needs a similar lock. But
actually, I don't think either is necessary. We only allocate new
entries during the counting phase, not during the threaded delta search.

So this hunk could be dropped (and the system CPU benefit I saw is just
from reduced lock contention).

-Peff