Subject: Re: Patch: fix lock contention for HASHHDR.mutex
To: pgsql-hackers <pgsql-hackers@postgresql.org>
References: <20151211170001.78ded9d7@fujitsu>
 <20151215122513.GC17724@alap3.anarazel.de>
 <CA+TgmoavsNuBNQf6ROn020F6g9gXJjtMrSkSFZ=RL1t2eJHQ5Q@mail.gmail.com>
 <20151217151418.GB2224@awork2.anarazel.de> <20151217190342.07e4533a@fujitsu>
 <CA+Tgmob1-7=Bh21LyLXk7Yst1nzcA1P=rv0UKq1zH+A9PVehrg@mail.gmail.com>
 <56740E41.4010708@sigaev.ru>
 <CA+Tgmobp1Dsk5M0UDjm91kCtt+LLJ+tTzs0nucw4Vw2ZfQx29w@mail.gmail.com>
 <20151222183953.771cb58b@fujitsu>
 <CA+TgmoZC6RsDt09LsYvSgmZFjy-Jwc6rD4GiySY3d4QTh8c3ow@mail.gmail.com>
 <20151228120112.6a847dfd@fujitsu> <20151229184851.1bb7d1bd@fujitsu>
 <20151230160137.034371e0@fujitsu>
From: Anastasia Lubennikova <a.lubennikova@postgrespro.ru>
Message-ID: <56A0E536.8020504@postgrespro.ru>
Date: Thu, 21 Jan 2016 17:03:34 +0300
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.5.1
MIME-Version: 1.0
In-Reply-To: <20151230160137.034371e0@fujitsu>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

30.12.2015 16:01, Aleksander Alekseev:
> Here is a clean version of the patch. Step 1 is an optimization. Step 2
> refactors this:
>
>   HTAB *
>   ShmemInitHash(const char *name, /* table string name for shmem index */
> -             long init_size,   /* initial table size */
> +             long init_size,   /* initial table size XXX ignored,
>   refactor! */

Hi, I found that the patch is still not reviewed, so I've looked it over.
I haven't dived into this subject before, so my comments will be more 
general than relating to your investigation.
Sorry if some things seem like nitpicks, I just suppose that clear patch 
would be more convenient for reviewers.

First of all, why not merge both patches into one? They aren't too big 
anyway.
Then, this thread became too tangled. I think it's worth to write a new 
message with the patch, the test script, some results and brief overview 
of how does it really works. It will make following review much easier.

+       /* number of entries in hash table */
+       long            nentries[NUM_LOCK_PARTITIONS];
+       /* linked list of free elements */
+       HASHELEMENT *freeList[NUM_LOCK_PARTITIONS];

I think comments should be changed, to be more informative here.

+               if (IS_PARTITIONED(hashp->hctl))
+                       partitions_number = NUM_LOCK_PARTITIONS;
+               else
+                       partitions_number = 1;
+
+               nelem_alloc = ((int) nelem) / partitions_number;
+               if (nelem_alloc == 0)
+                       nelem_alloc = 1;

Add a comment here too.

-/* Number of partitions of the shared buffer mapping hashtable */
-#define NUM_BUFFER_PARTITIONS  128
-
  /* Number of partitions the shared lock tables are divided into */
-#define LOG2_NUM_LOCK_PARTITIONS  4
+#define LOG2_NUM_LOCK_PARTITIONS  7
  #define NUM_LOCK_PARTITIONS  (1 << LOG2_NUM_LOCK_PARTITIONS)

+/* Number of partitions of the shared buffer mapping hashtable */
+#define NUM_BUFFER_PARTITIONS NUM_LOCK_PARTITIONS
+

I'm sure, it was discussed above. but these definitions do not look 
obvious at all.
Maybe you should explain this magic number 7 in the comment above?

-- 
Anastasia Lubennikova
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers