[Miu]

Originally Posted by DeagLe.Studio Declaring the variable outside the "for" block: ...

You need to do something with the variable so it won't optimize it out

[TheDS1337]

Originally Posted by Miu You need to do something with the variable so it won't optimize it out

Oh, I didnt know that.

But anyways guys, I always declare vars outside the loop. the thing is that I love to choose a good coding style if you know what I mean, and recently I saw a lot of people declaring the vars inside the loop I thought it'll be cool to do it (but only if compiler actually optimize it).

[Miu]

PHP Code:

public void OnPluginStart()
{
    for (int i = 0; i < 100; i++)
    {
        char buf[512];
        PrintToServer("%s", buf);
    }
} 


Lysis reversal:

PHP Code:

public void:OnPluginStart()
{
    new i;
    while (i < 100)
    {
        new String:buf[512];
        PrintToServer("%s", buf);
        i++;
    }
    return void:0;
} 


SPEdit disassembly:

PHP Code:

; OnPluginStart
; 23 instruction(s)
; starts at code address 0xb68
---
0xb6c: break                           
0xb70: break                           
0xb74: push.c 0x0                       ; 0
0xb7c: jump 0xb90                       ; +0x14
0xb84: break                           
0xb88: inc.s 0xfffffffc                 ; i
0xb90: load.s.pri 0xfffffffc            ; i
0xb98: const.alt 0x64                   ; 100
0xba0: jsgeq 0xbf8                      ; +0x58
0xba8: break                           
0xbac: stack 0xfffffe00                 ; -512
0xbb4: zero.pri                        
0xbb8: addr.alt 0xfffffdfc              ; buf
0xbc0: fill 0x200                       ; 512
0xbc8: break                           
0xbcc: push.adr 0xfffffdfc              ; buf
0xbd4: push.c 0x8d4                     ; 2260
0xbdc: sysreq.n 2 0x2                   ; PrintToServer 2
0xbe8: stack 0x200                      ; 512
0xbf0: jump 0xb84                       ; -0x6c
0xbf8: stack 0x4                        ; 4
0xc00: zero.pri                        
0xc04: retn 


So yeah no optimization

[h3bus]

Wow

Nice to know... I wonder why such an implementation.
Is there cases where a variable cannot be statically allocated function-wide (stack pointer pushed at beginnign on function and restored at the end)? Those would explain that choice.

[Miu]

Originally Posted by h3bus Wow Nice to know... I wonder why such an implementation. Is there cases where a variable cannot be statically allocated function-wide (stack pointer pushed at beginnign on function and restored at the end)? Those would explain that choice.

of course not, but the allocation isn't the issue anyway, the implicit zeroing of the entire array is, which can't be moved outside of the loop

[Potato Uno]

Originally Posted by Miu of course not, but the allocation isn't the issue anyway, the implicit zeroing of the entire array is, which can't be moved outside of the loop

You can move it to the global scope if you're using it in a function that is called frequently. (Most notorious example is OnGameFrame, lesser knowns are some sdkhook callbacks.)

Then the zeroing is done on compile time (according to asherkin).

[TheDS1337]

Originally Posted by Miu PHP Code: public void OnPluginStart() {     for (int i = 0; i < 100; i++)     {         char buf[512];         PrintToServer("%s", buf);     } }  Lysis reversal: PHP Code: public void:OnPluginStart() {     new i;     while (i < 100)     {         new String:buf[512];         PrintToServer("%s", buf);         i++;     }     return void:0; }  SPEdit disassembly: PHP Code: ; OnPluginStart ; 23 instruction(s) ; starts at code address 0xb68 --- 0xb6c: break                            0xb70: break                            0xb74: push.c 0x0                       ; 0 0xb7c: jump 0xb90                       ; +0x14 0xb84: break                            0xb88: inc.s 0xfffffffc                 ; i 0xb90: load.s.pri 0xfffffffc            ; i 0xb98: const.alt 0x64                   ; 100 0xba0: jsgeq 0xbf8                      ; +0x58 0xba8: break                            0xbac: stack 0xfffffe00                 ; -512 0xbb4: zero.pri                         0xbb8: addr.alt 0xfffffdfc              ; buf 0xbc0: fill 0x200                       ; 512 0xbc8: break                            0xbcc: push.adr 0xfffffdfc              ; buf 0xbd4: push.c 0x8d4                     ; 2260 0xbdc: sysreq.n 2 0x2                   ; PrintToServer 2 0xbe8: stack 0x200                      ; 512 0xbf0: jump 0xb84                       ; -0x6c 0xbf8: stack 0x4                        ; 4 0xc00: zero.pri                         0xc04: retn  So yeah no optimization

Thank you!

Now I'll just continue declaring it outside the loop block

[h3bus]

Originally Posted by Miu of course not, but the allocation isn't the issue anyway, the implicit zeroing of the entire array is, which can't be moved outside of the loop

I'm really not familiar with the new syntax. Does this mean decl has no equivalent in new syntax?

[Powerlord]

Originally Posted by h3bus I'm really not familiar with the new syntax. Does this mean decl has no equivalent in new syntax?

Exactly. There is no decl in new syntax.

[friagram]

it's not all pre allocated, pretty often I do
decl array [size]
where size is the call count of a dynamic array or some config variable.

you can use static too, which should allocate it on runtime and keep it there...

decl still works too, as reading out a string or variable that is not initialized gives you the appropriate garbage output.