[stupok]

I noticed something curious, but please keep in mind that I don't know how this profiler works. Maybe this behavior is expected.

The data below shows what happens when a function is called 1 time, 10 times, etc. up to 100 million times. The function is called in a for() loop.

The curiosity is that the "max" time required for one call increases at a rate proportional to the number of total calls.

date: Sat Apr 18 14:27:33 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |          1 | 0.000002 / 0.000002 / 0.000002
   f |                    UseGetPlayers |          1 | 0.000001 / 0.000001 / 0.000001

date: Sat Apr 18 14:27:57 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |         10 | 0.000006 / 0.000000 / 0.000002
   f |                    UseGetPlayers |         10 | 0.000007 / 0.000001 / 0.000001

date: Sat Apr 18 14:28:04 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |        100 | 0.000045 / 0.000000 / 0.000001
   f |                    UseGetPlayers |        100 | 0.000063 / 0.000001 / 0.000001

date: Sat Apr 18 14:28:13 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |       1000 | 0.000425 / 0.000000 / 0.000001
   f |                    UseGetPlayers |       1000 | 0.000636 / 0.000001 / 0.000001

date: Sat Apr 18 14:28:20 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |      10000 | 0.004230 / 0.000000 / 0.000002
   f |                    UseGetPlayers |      10000 | 0.006582 / 0.000001 / 0.000051

date: Sat Apr 18 14:28:29 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |     100000 | 0.042843 / 0.000000 / 0.000090
   f |                    UseGetPlayers |     100000 | 0.066326 / 0.000001 / 0.000171

date: Sat Apr 18 14:28:38 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |    1000000 | 0.425573 / 0.000000 / 0.000282
   f |                    UseGetPlayers |    1000000 | 0.646957 / 0.000001 / 0.000462

date: Sat Apr 18 14:39:49 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |   10000000 | 4.284324 / 0.000000 / 0.000629
   f |                    UseGetPlayers |   10000000 | 6.440324 / 0.000001 / 0.000622

date: Sat Apr 18 14:48:39 2009 map: de_dust
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                      get_players |  100000000 | 43.063107 / 0.000000 / 0.031469
   f |                    UseGetPlayers |  100000000 | 64.614610 / 0.000001 / 0.001876

[joaquimandrade]

Stupok, good work with the profiling. Let me just say this:

While caching variables is faster no matter what, the comparison is not fair. That because you just have 32 to client_authorized/client_disconnect. On a "real" situation, that number would grow bigger during the lifetime of the plugin (even if you don't use the cached values). Another point is that forwards are called module side. So you can't profile with amxmodx profiler the time that a forward takes before happening plugin side.

__

By the way, does Sawce still visits the forums?

[Arkshine]

Last Activity: Yesterday 12:03

It seems.

[joaquimandrade]

Originally Posted by arkshine It seems.

He should be tired of working for free here.

[danielkza]

Originally Posted by sawce's profile statistics Last Activity: Yesterday 07:03 AM Join Date: 08-10-04 Referrals: 5

[joaquimandrade]

By the way, arkshine:

[IMG]http://img4.**************/img4/567/arkd.png[/IMG]

[stupok]

Originally Posted by joaquimandrade While caching variables is faster no matter what, the comparison is not fair.

You're right, but at least I made Emp` happy.

Any ideas about the max time per call increasing proportionally with the number of total calls?

I'm too much of a C++ noob to inspect the source, and I'm not very knowledgeable on the topic.

[joaquimandrade]

Originally Posted by stupok You're right, but at least I made Emp` happy. Any ideas about the max time per call increasing proportionally with the number of total calls? I'm too much of a C++ noob to inspect the source, and I'm not very knowledgeable on the topic.

You got that profile results with windows or linux?

[stupok]

windows

[Emp`]

Could someone do a profile for:

random(var);

vs

random_num(0,var-1);

and

new var = 10;

vs

static var;
var = 10;