[Pawn] Opinion on PVars

[Pinch]

Hi everyone, I've gotten bored so I started writing a gamemode for fun, I want it to be a RolePlay gamemode with all of the stuff (like the NG-RP) but I want it to be smooth on bigger player base too..



I ran some tests and those were the results

Code:

Timing "PawnPlus list"...

      Mean = 480.00ns

      Mode = 475.00ns

    Median = 477.00ns

     Range = 5.00ns

Timing "Normal var"...

      Mean = 200.00ns

      Mode = 199.00ns

    Median = 200.00ns

     Range = 3.00ns

Timing "Pvar timings"...

      Mean = 483.00ns

      Mode = 481.00ns

    Median = 483.00ns

     Range = 4.00ns

The code I used:

Code:

new List:pp_list[1000];



enum E_PLAYER_DATA {

    E_PLAYER_SOME_VAR,

    E_PLAYER_SOME_VAR_2,

    E_PLAYER_SOME_VAR_3

};

new PlayerData[1000][E_PLAYER_DATA];



hook OnScriptInit()

{

    for(new i = 0; i < 1000; ) {

        pp_list[i] = list_new();

        list_resize_var(pp_list[i], 3, VAR_NULL);    

    }

    return Y_HOOKS_CONTINUE_RETURN_1;

}



public OnPlayerConnect(playerid)

{

    wait_ms(5000);

    print("A");

    RUN_TIMING("PawnPlus list")

    {

        list_set(pp_list[playerid], 0, 10000000);

        list_set(pp_list[playerid], 1, 10000000);

        list_set(pp_list[playerid], 2, 10000000);

    }

    RUN_TIMING("Normal var")

    {

        PlayerData[playerid][E_PLAYER_SOME_VAR] = 10000000;

        PlayerData[playerid][E_PLAYER_SOME_VAR_2] = 10000000;

        PlayerData[playerid][E_PLAYER_SOME_VAR_3] = 10000000;

    }

    RUN_TIMING("Pvar timings")

    {

        SetPVarInt(playerid, "key1", 10000000);

        SetPVarInt(playerid, "key2", 10000000);

        SetPVarInt(playerid, "key3", 10000000);

    }

    return 1;

}

So, is it worth it?

Should I use PVars/PP lists? Does it even matter that much when the diff is 200ns?

[IllidanS4]

The question is incomplete. Use X for what? All 3 features have other differences, and are suited for other things. PP lists are resizable; PVars are maps (so PP maps would be better in a comparison), plus their lookup is linear so the difference will only become apparent for a larger collection.



And the difference should be interpreted in another way. It doesn't matter at all what the absolute difference is for a single call, you should be looking at how it scales. So for 1000 calls it will be 200 ?s and so on. But I don't think it matters (for this specific use case) that much since twice slower is still more or less within the same order of magnitude.

[Pinch]

Hmm, I want to store the user data so that's why I tested it this way..????

[Y_Less]

He is just saying that your test isn't big enough. You're only testing 3 writes to three variables, you need more like 100.