[SchlumPF*]

"When there a 32 slots on the server.. MAX_ENTs = 1365"

fail,
maxents = 900 + maxplayers * 15
maxents_32 = 900 + 32 * 15 = 1380

anyway, thats the hl calculations which is used if the -num_edicts param is not used, else maxents = -num_edicts

[Hawk552]

Originally Posted by Empowers why? g_isAlive is a good optimization & it's pretty save: PHP Code: //code by MeRcyLeZZ public fw_PlayerSpawn( id ) {     if ( !is_user_alive( id ) || !cs_get_user_team( id ) )         return;          g_isAlive[id] = true } public fw_PlayerKilled( id ) {     g_isAlive[id] = false }

Running two forwards like that is probably slower than just using is_user_alive() wherever you need it (unless you're checking it in client_PreThink or some other forward that's called very frequently). This is even more likely since you have two native calls in the fw_PlayerSpawn() function. On top of that, you lose precision as the player could have been marked as killed some other way (although unlikely).

Either way, it's a trivial optimization, even if it was faster. People need to learn how to balance their time. Your plugin doesn't have to be 100% optimized to the point that not a single CPU cycle is wasted. Just apply optimizations that don't make you go extremely out of your way to do them. This, for example, is an insanely big waste of time.

[joaquimandrade]

Hawk is alive!

[AntiBots]

This:

PHP Code:

static cell AMX_NATIVE_CALL strip_user_weapons(AMX *amx, cell *params) // index
{
    CHECK_PLAYER(params[1]);

    edict_t* pPlayer = MF_GetPlayerEdict(params[1]);

    string_t item = MAKE_STRING("player_weaponstrip");
    edict_t *pent = CREATE_NAMED_ENTITY(item);
    
    if (FNullEnt(pent)) 
    {
        return 0;
    }

    MDLL_Spawn(pent);
    MDLL_Use(pent, pPlayer);
    REMOVE_ENTITY(pent);

    *reinterpret_cast<int *>(MF_PlayerPropAddr(params[1], Player_CurrentWeapon)) = 0;

    return 1;
} 


Is the same like this.

PHP Code:

stock fm_strip_user_weapons(index) {
    static ent; ent = fm_create_entity("player_weaponstrip");
    if (!pev_valid(ent))
        return 0;

    dllfunc(DLLFunc_Spawn, ent);
    dllfunc(DLLFunc_Use, ent, index);
    engfunc(EngFunc_RemoveEntity, ent);

    return 1;
} 


And have less things.

Module using module is more faster than a call of the plugin?

MDLL_Spawn Is a call in the module.

[joaquimandrade]

Originally Posted by AntiBots This: PHP Code: static cell AMX_NATIVE_CALL strip_user_weapons(AMX *amx, cell *params) // index {     CHECK_PLAYER(params[1]);     edict_t* pPlayer = MF_GetPlayerEdict(params[1]);     string_t item = MAKE_STRING("player_weaponstrip");     edict_t *pent = CREATE_NAMED_ENTITY(item);          if (FNullEnt(pent))      {         return 0;     }     MDLL_Spawn(pent);     MDLL_Use(pent, pPlayer);     REMOVE_ENTITY(pent);     *reinterpret_cast<int *>(MF_PlayerPropAddr(params[1], Player_CurrentWeapon)) = 0;     return 1; }  Is the same like this. PHP Code: stock fm_strip_user_weapons(index) {     static ent; ent = fm_create_entity("player_weaponstrip");     if (!pev_valid(ent))         return 0;     dllfunc(DLLFunc_Spawn, ent);     dllfunc(DLLFunc_Use, ent, index);     engfunc(EngFunc_RemoveEntity, ent);     return 1; }  And have less things. Module using module is more faster than a call of the plugin? MDLL_Spawn Is a call in the module.

It doesn't have more code. You are forgetting that fm_create_entity has more calls inside it.

Calling functions inside modules is less expensive than plugins calling module provided functions.

[xPaw]

Just tested strip_user_guns and give_item vs fakemeta, and i was surprised about results..

                 name |      calls | time / min / max
--------------------------------------------------
                equal |       1001 | 0.001641 / 0.000001 / 0.000229
              engfunc |       6006 | 0.037257 / 0.000001 / 0.002556
            pev_valid |       2002 | 0.003879 / 0.000001 / 0.000697
                  pev |       4004 | 0.007338 / 0.000001 / 0.000470
              set_pev |       2002 | 0.003648 / 0.000001 / 0.000293
              dllfunc |       4004 | 0.126556 / 0.000002 / 0.003295
         fm_give_item |       1001 | 0.018707 / 0.000016 / 0.000473
fm_strip_user_weapons |       1001 | 0.012963 / 0.000008 / 0.000943

            give_item |       1001 | 0.307619 / 0.000152 / 0.016896
   strip_user_weapons |       1001 | 0.068136 / 0.000033 / 0.001293

0.001641 + 0.037257 + 0.003879 + 0.007338 + 0.003648 + 0.126556 + 0.018707 + 0.012963 = 0.211989
0.307619 + 0.068136 = 0.375755

Fun is slower with less calls?

PHP Code:

#include <amxmodx>
#include <fun>
#include <fakemeta_util>

new omg2;

public plugin_init() {
    register_clcmd("say /profile","profile");
}

public profile( id ) {
    // set_Task because loop crushes server.
    set_task( 0.1, "omg", id,_,_,"b" );
}

public omg( id ) {
    omg2++;
    strip_user_weapons( id );
    fm_strip_user_weapons( id );
    
    give_item( id, "weapon_knife" );
    fm_give_item( id, "weapon_knife" );
    
    if( omg2 > 1000 )
        server_cmd("quit");
} 


Yeah my PC is slow.

[Hawk552]

Run them separately. It's not fair the way you have it set up since fm_strip_user_weapons() and fm_give_item() have no work to do.

[xPaw]

U'r right. But still strip_user_weapons is slower...

                             name |      calls | time / min / max
-------------------------------------------------------------------
                          engfunc |       3003 | 0.018500 / 0.000002 / 0.001438
                        pev_valid |       1001 | 0.004189 / 0.000001 / 0.001656
                          dllfunc |       2002 | 0.015975 / 0.000002 / 0.000834
            fm_strip_user_weapons |       1001 | 0.012942 / 0.000008 / 0.000651

               strip_user_weapons |       1001 | 0.086086 / 0.000033 / 0.004615
-------------------------------------------------------------------
                            equal |       1001 | 0.003534 / 0.000001 / 0.001859
                          engfunc |       3003 | 0.016452 / 0.000002 / 0.001325
                        pev_valid |       1001 | 0.001539 / 0.000001 / 0.000400
                              pev |       4004 | 0.009467 / 0.000001 / 0.000779
                          set_pev |       2002 | 0.004025 / 0.000001 / 0.000494
                          dllfunc |       2002 | 0.157066 / 0.000002 / 0.004978
                     fm_give_item |       1001 | 0.032702 / 0.000016 / 0.002458

                        give_item |       1001 | 0.173836 / 0.000115 / 0.002795
-------------------------------------------------------------------

strip_user_weapons = 0.086086
fm_strip_user_weapons = 0.051606

give_item = 0.173836
fm_give_item = 0.224785

Someone make, strip_user_weapons & give_item || fm_strip_user_weapons & give_item

[joaquimandrade]

xPaw, tasks are not suitable for profiling it seems.

Edit: Of course they aren't. Profiling tasks is like profiling forwards. They are called from the module so you just can't do it.

PHP Code:


date: Tue Apr 21 18:38:46 2009 map: de_dust2
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                            equal |        100 | 0.000242 / 0.000002 / 0.000005
   n |                          engfunc |        300 | 0.001387 / 0.000002 / 0.000010
   n |                        pev_valid |        100 | 0.000269 / 0.000002 / 0.000004
   n |                              pev |        400 | 0.001077 / 0.000002 / 0.000004
   n |                          set_pev |        200 | 0.000563 / 0.000002 / 0.000027
   n |                          dllfunc |        200 | 0.002190 / 0.000003 / 0.000021
   n |                   register_clcmd |          1 | 0.000017 / 0.000017 / 0.000017
   n |                        give_item |        100 | 0.002403 / 0.000020 / 0.000084
   p |                      plugin_init |          1 | 0.000007 / 0.000007 / 0.000007
   p |                          profile |          1 | 0.000425 / 0.000425 / 0.000425
   f |                     fm_give_item |        100 | 0.003121 / 0.000028 / 0.000055
0 natives, 0 public callbacks, 2 function calls were not executed. 


PHP Code:

date: Tue Apr 21 18:40:26 2009 map: de_dust2
type |                             name |      calls | time / min / max
-------------------------------------------------------------------
   n |                          engfunc |        300 | 0.001517 / 0.000002 / 0.000031
   n |                        pev_valid |        100 | 0.000281 / 0.000002 / 0.000004
   n |                          dllfunc |        200 | 0.001160 / 0.000002 / 0.000023
   n |                   register_clcmd |          1 | 0.000018 / 0.000018 / 0.000018
   n |               strip_user_weapons |        100 | 0.001220 / 0.000010 / 0.000022
   p |                      plugin_init |          1 | 0.000007 / 0.000007 / 0.000007
   p |                          profile |          1 | 0.000427 / 0.000427 / 0.000427
   f |            fm_strip_user_weapons |        100 | 0.001602 / 0.000014 / 0.000031
0 natives, 0 public callbacks, 2 function calls were not executed. 


[Bad_Bud]

I thought the concept behind the fakemeta high was that the engine module eats up CPU time when it's doing nothing, whereas fakemeta does not. I haven't seen anyone do a profile with fakemeta running for long periods of time with the engine module disabled, versus engine running for long periods of time. In most profiles, I see the fakemeta being compared with the engine in the same plugin, which guarantees that the engine module is still running -- defeating the purpose of a comparison between engine and "pure fakemeta"; we've taken away the most preached aspect of the comparison: engine module uses CPU when it is not being used.

I'm not taking sides, because I'm not advanced enough of a scripter to really care, but I think these tests are less intuitive than they appear, and that if we are going to state things as solid facts with profiles, fair tests should be run.