Pooly wrote:
2005/9/27, Lefteris Tsintjelis <[EMAIL PROTECTED]>:Hi, What makes me wonder is that the same test, with the code stripped down, to my surprise, is significantly faster that the multi threaded one, no matter how many times I run the tests. I am including the code for both tests I run. Since I couldn't find a good example of mutex locking the following one is something that worked for me. However, I am not sure if its as optimized as it should be, so I would appreciate an expert's opinion about this. Is this a good example of mutex locking? Are there any other better ways for this? Is this an OS or MySQL issue? I am currently running 4.1.14 on a FreeBSD5 box.You ran several queries with multiple thread, fine, but they are all serialised over one connection, so you get all the overhead of locking and thread-creation, for no advantage... So that's the result expected. (So, yes forthe troll, it's an OS issue, threads creation are somewhat slow on FreeBSD :)
My intention was to avoid the overhead of multiple network connections and I didn't expect it to have that much difference. It is probably an OS issue a bit here as well, I have to agree with that. I will test and see what happens with a few network connections but I have a bad feeling about this one also. I don't think it will get much better and not even close to a non multi thread implementation, but further tests will show. I am just curious if anyone could run the same tests in some other OS and maybe compare some notes. I looked around but the few things I found are doubtful. There are no good performance tests between threads and no threads with random access reads and writes, or maybe even better MyISAM and InnoDB as well as threads/no threads. This should also be interesting due to the locking differences of those two databases among other things. Lefteris -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
