[Home] [Groups] - Message: [Prev in Group] [Next in Group]

nu.kanga.list.mud-dev

10080: RE: [MUD-Dev] Sockets

[Full Header] [Plain Text]
From: Petri Virkkula <pvirkkul@iki.fi>
Newsgroups: nu.kanga.list.mud-dev
Date: Sat, 8 May 1999 10:57:43 +0300 (EEST)
References: [1]
Organization: Kanga.Nu
>>>>> "Chris" == Chris Gray <cg@ami-cg.GraySage.Edmonton.AB.CA> writes:

Chris> [Jon A. Lambert:]
>> I think the performance boost 
>> from threading has been understated or misunderstood on single-processor 
>> machines probably due to early thread implementations.

Chris> Is this deliberate flamebait, or just accidental?

Chris> Please explain to me how adding the overhead of thread switches, locks
Chris> and extra memory usage can increase performance. Or, perhaps you meant

	How are you measuring your performance figures? As average
	per command response time or average commands per a second?
	Which one of the two choices below has better performance:

	Singlethreaded:
		Average per command response time: 10ms
		Average commands per a second: 100 commands/s

	Multithreaded:
		Average per command response time: 11ms
		Average commands per a second: 120 commands/s

	Above numbers are just examples, but I wouldn't be too
	suprised if they were near the truth in a real system.

	I prefer the commands per a second performance measurement,
	because that number tells more directly how many players can
	be supported by a system.

Chris> to say "multi-processor machines"? One could perhaps imagine how a good
Chris> threaded implementation of something could perform better a poor
Chris> non-threaded implementation, but I don't think that's what you mean.

	Actually there aren't many "single-procesor" machines in the
	market today (neither so many SMP machines however). Many
	(most?) disks for example have their own processors, thus a
	PC could be said to be a "multi-processor" machine. However
	the term "multi-processor" usually refers to SMP.


	Petri



_______________________________________________
MUD-Dev maillist  -  MUD-Dev@kanga.nu
http://www.kanga.nu/lists/listinfo/mud-dev