From: Johannes Sixt <j6t@kdbg.org>
Date: Thu, 26 Jul 2018 08:12:40 GMT
Subject: Re: [PATCH v2] pack-objects: fix performance issues on packing large deltas
Message-ID: <c872c664-a078-853d-4ce9-631dfd110c94@kdbg.org>
In-Reply-To: <20180722080421.12887-1-pclouds@gmail.com>

Am 22.07.2018 um 10:04 schrieb Nguyễn Thái Ngọc Duy:
> +	if (size < pack->oe_delta_size_limit) {
> +		e->delta_size_ = size;
> +		e->delta_size_valid = 1;
> +	} else {
> +		packing_data_lock(pack);
> +		if (!pack->delta_size)
> +			ALLOC_ARRAY(pack->delta_size, pack->nr_alloc);
> +		packing_data_unlock(pack);
> +
> +		pack->delta_size[e - pack->objects] = size;

My first thought was that this is wrong (falling prey to the same 
mistake as the double-checked locking pattern). But after thinking twice 
over it again, I think that this unprotected access of pack->delta_size 
is thread-safe.

Of course, I'm assuming that different threads never assign to the same 
array index.

> +		e->delta_size_valid = 0;
> +	}
>   }

-- Hannes