From: | Simon Riggs <simon(at)2ndQuadrant(dot)com> |
---|---|
To: | Takashi Horikawa <t-horikawa(at)aj(dot)jp(dot)nec(dot)com> |
Cc: | PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Typo in bufmgr.c that result in waste of memory |
Date: | 2016-02-19 09:24:29 |
Message-ID: | CANP8+jKYgenEEKu+HP=X=+crk5Gat6K=y3w0G8+u5QPtX8++SA@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On 19 February 2016 at 02:58, Takashi Horikawa <t-horikawa(at)aj(dot)jp(dot)nec(dot)com>
wrote:
> I have just found a typo in the source code (not in a comment) of bufmgr.c
> that result in waste of memory. It might be a 'bug' but it does not result
> in any incorrect operation but just results in waste of a few memory
> resource.
>
> As sizeof(PrivateRefCountArray) found in InitBufferPoolAccess() is 64 and
> sizeof(PrivateRefCountEntry) which should be used here is 8, this typo
> produces 56 byte of unused memory area per one PrivateRefCount entry in the
> hash table. I think this result in not only the waste of memory but also
> reduces the cache hit ratio.
>
> X hash_ctl.entrysize = sizeof(PrivateRefCountArray);
> O hash_ctl.entrysize = sizeof(PrivateRefCountEntry);
>
I see the problem, but I don't buy the argument that it wastes large
amounts of memory. Or do you have some evidence that it does?
I think we should fix it, but not backpatch.
--
Simon Riggs http://www.2ndQuadrant.com/
<http://www.2ndquadrant.com/>
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
From | Date | Subject | |
---|---|---|---|
Next Message | Simon Riggs | 2016-02-19 09:26:08 | Re: Bug in StartupSUBTRANS |
Previous Message | Fabien COELHO | 2016-02-19 09:16:41 | Re: checkpointer continuous flushing - V16 |