[SickAttack]

I posted this thread as most people create their loops that go through strings as follows:

for(new j = 0, k = strlen(string); j < k; j ++)

Which ended up in 3rd.

----------------------------------------------------

***** 1,000,000 ITERATIONS *****

// ** Lowest: 2,666 ms | Highest: 2,724 ms

Код:

new j = (strlen(string) - 1);
while(j != -1)
{
	j --;
}

// ** Lowest: 2,648 ms | Highest: 2,700 ms

Код:

new j, k = strlen(string);
while(j != k)
{
	j ++;
}

// ** Lowest: 2,619 ms | Highest: 2,686 ms

Код:

new j = -1;
while(string[++ j])

// ** Lowest: 2,536 ms | Highest: 2,599 ms

Код:

new j = strlen(string);
do
{
	j --;
}
while(j != 0);

// ** Lowest: 1,428 ms | Highest: 1,450 ms

Код:

for(j = 0; string[j]; j ++)

// ** Lowest: 1,422 ms | Highest: 1,440 ms

Код:

for(new j = 0; string[j] != EOS; j ++)

// ** Lowest: 1,320 ms | Highest: 1,351 ms

Код:

for(new j = (strlen(string) - 1); j > -1; j --)

// ** Lowest: 1,314 ms | Highest: 1,349 ms

Код:

for(new j = strlen(string); --j != -1;)

// ** Lowest: 1,309 ms | Highest: 1,319 ms

Код:

for(new j = 0, k = strlen(string); j < k; j ++)

// ** Lowest: 1,299 ms | Highest: 1,324 ms

Код:

for(new j = 0, k = strlen(string); j != k; j ++)

// ** Lowest: 1,193 ms | Highest: 1,225 ms

Код:

for(new j = strlen(string); j != 0; j --)
{
	// printf("%d", j - 1);
}

[SyS]

Quote:

Originally Posted by SickAttack for(new j = (strlen(string) - 1); j != 0; j --)

It wont reach @ first index.

[SickAttack]

Quote:

Originally Posted by Sreyas It wont reach @ first index.

Yeah, thanks for pointing that out. Anyhow, it still remains the fastest with changes:

// ** Lowest: 1,193 ms | Highest: 1,225 ms

Код:

for(new j = strlen(string); j != 0; j --)
{
	// printf("%d", j - 1);
}

[Gammix]

Sequence matters, i find the regular sequence starting from 0 is most used. So i consider the second last to be best and good thing most people use that already!

[Yashas]

If amx files were compiled executables (exe, dll), then your argument would be wrong as the speed of the logical operations depends on the CPU.

I am not sure how things work for PAWN. You are probably right because the AMX machine has a lot of work (decoding the instruction) to do just before doing the actual CPU > or != instruction so the CPU implementation shouldn't really contribute. PAWN will always be very slow.

[SyS]

hmm can you test its speed? im not home these days so i cant test myself.

Код:

new j=strlen(string);

do
{
  j--;
 //-----
}while(j!=0);

[SickAttack]

Quote:

Originally Posted by Sreyas hmm can you test its speed? im not home these days so i cant test myself. Код: new j=strlen(string); do { j--; //----- }while(j!=0);

It's really slow, I added it to the main post.

-----------------

Well, I'm not sure which one is the best option now.

for(new j = (strlen(cmd) - 1); j > -1; j --)
for(new j = (strlen(cmd) - 1); j != 0; j --)

Both seem to be the fastest in a production scenario.

-----------------

for(new j = strlen(string); j != 0; j --)

It becomes very slow in a production use, but seems to be the fastest in a plain loop.

-----------------

for(new j = 0, k = strlen(string); j != k; j ++)

Slice's fastest loop to his experience seems slower than the first loop in a production use, but faster in a plain loop.

[Luicy.]

Why are everyone upset with the speed? Like,
new something[3]; or new something0, something1, something2;
It just a few ms different, why do it fucking matter?

[SyS]

Quote:

Originally Posted by SickAttack It's really slow, I added it to the main post.

my assumptions sucks
@Luicy. it can be fun in other words and also very useful.How optimized is our code that much is good no? its not a must though.

[SickAttack]

Quote:

Originally Posted by Luicy. Why are everyone upset with the speed? Like, new something[3]; or new something0, something1, something2; It just a few ms different, why do it fucking matter?

"best coding practices"

If you're going to make a game, then you're going to do what it takes to make it cause the less stutter possible.

[Jayse]

I could've tried to find a faster method, but the topic contains what I had on my mind. Also it's always nice to see more optimized ways to do something such as looping. That's the main reason why most of scripters use y_iterate instead of plain player loops.

[Stinged]

How do you do your benchmarks?

I tried out the third slowest one:

Код:

new j = -1;
while(string[++ j])

And this one (which is supposedly faster):

Код:

for (new j = strlen(str)-1; j > -1; j--)

(All according to your list)

It turned out that the while loop is actually faster.
(I tested it out on a 34 characters long string)

Either I'm doing something wrong, or you are.
Can you please post the code?

[Konstantinos]

I ran few tests and if the length is long, these two made a difference:

This is the fastest:

pawn Код:

for(j = strlen(string); --j != -1;)

and second comes:

pawn Код:

for(j = strlen(string); j != 0; j --)

http://pastebin.com/raw/AdJuRncr

Results:

pawn Код:

Bench for 1st: executes, by average, 168.35 times/ms.
Bench for 2nd: executes, by average, 156.76 times/ms.
Bench for 3rd: executes, by average, 173.58 times/ms.
Bench for 4th: executes, by average, 168.80 times/ms.
Bench for 5th: executes, by average, 171.69 times/ms.
Bench for 6th: executes, by average, 149.65 times/ms.
Bench for 7th: executes, by average, 174.24 times/ms.
Bench for 8th: executes, by average, 223.66 times/ms.
Bench for 9th: executes, by average, 174.86 times/ms.
Bench for 10th: executes, by average, 174.98 times/ms.
Bench for 11th: executes, by average, 217.67 times/ms.

EDIT: I removed the part I was setting the text in the loops and did few more. The result for longer strings are the same but for 1 or 2 characters only, this becomes the fastest:

pawn Код:

for(j = 0; string[j]; j ++)