Bug #5106 noticeable lower speed of 4.1.3 compared to 4.0 (PowerPC)
Submitted: 19 Aug 2004 13:15 Modified: 7 Feb 2005 10:27
Reporter: Gunnar von Boehn Email Updates:
Status: Can't repeat Impact on me:
Category:MySQL Server Severity:S2 (Serious)
Version:4.1.3 OS:Linux (Debian Linux PowerPC)
Assigned to: Igor Babaev CPU Architecture:Any

[19 Aug 2004 13:15] Gunnar von Boehn
I use our MySQL binaries for PowerPC.

I noticed that 4.1.3 runs around 15-20% slower than 4.0.18/20 on my system.
I could verify that a number of different queries are slower on 4.1.3 than on 4.0.20.

Used table type is myisam, the mysql.config is the same.

I wonder if this behavior is expected
and 4.1 is supposed to be around 20% slower than 4.0,
or did I spot a problem here?

I have included an explain of one typical query but I doubt that it will help.

Execution time of test query (times verified by running test many times)
MySQL 4.0.18  =  4.11 sec
MySQL 4.1.3   =  4.80 sec

I can upload real testdata is needed


How to repeat:
see internal
[23 Aug 2004 13:15] Michael Widenius
The above two explains doesn't help to solve the problem as they are identical (4.1 just uses slightly different format).

If you can attach test data to this bug that shows the speed difference, we will look into what could be the cause of the slowdown.
[23 Aug 2004 14:01] Gunnar von Boehn
SQL-Bench 4.0.20

Attachment: peg_4.0.20_RUN (application/octet-stream, text), 7.77 KiB.

[23 Aug 2004 14:02] Gunnar von Boehn
SQL-Bench 4.1.3

Attachment: peg_4.1.3_RUN (application/octet-stream, text), 7.77 KiB.

[23 Aug 2004 14:07] Gunnar von Boehn
I've attached SQL_Bench results which show some speed differences.

The two results are from the SQL-BENCH test which came
bundled with the source of 4.1.4(daily snapshot 2004-08-18)
One result show the MySQL PPC binary 4.0.20 and the other one MySQL PPC binary 4.1.3
Both test ran on the same machine with my-big config and no other load.

I'll prepare a testcase with my data which propably show more clearly the slowdown of joins.

Kind regards
[15 Sep 2004 16:34] Gunnar von Boehn
verified with 4.1.4 and 4.0.12 and new testcase uploaded
(sorry took me some time without broadband)

Maybe I found somehting else as well something to optimise in optimizer.
Please see below
[7 Feb 2005 10:27] Gunnar von Boehn
Hi Igor,

I'm very surprised that you can not repeate the bug.
Would you mind telling me which MySQL versions on what computer/system you used for testing?

Thanks in advance