which code runs faster? (SOLVED)

[EnterTheStory (aka tolworthy)]

I'm trying to optimize the following code:

Code: ags

if(player.variable ==1) dothis();
if(player.variable ==2) dothat();
if(player.variable ==3) dothis();
if(player.variable ==4) dothat();
if(player.variable ==5) dothis();
// repeated a hundred times every frame

Would there be a tiny speed increase to do this:

Code: ags

int n = player.variable;
if(n ==1) dothis();
if(n ==2) dothat();
if(n ==3) dothis();
if(n ==4) dothat();
if(n ==5) dothis();
// etc.

In other words, is finding player.variable a single step, or does it take a couple of steps (find player, find variable)?

Sorry if this is an idiot question!

[Akatosh]

I don't think the difference between the two is measurable, if one even exists. Also... a few hundred times every frame sounds a little bit exaggerated... I may be mistaken, but I've always been under the impression that, say, five times a frame would already be overkill.

[EnterTheStory (aka tolworthy)]

I don't think the difference between the two is measurable, if one even exists. Also... a few hundred times every frame sounds a little bit exaggerated... I may be mistaken, but I've always been under the impression that, say, five times a frame would already be overkill.

Thanks. I need this when moving the mouse over a hotspot that changes according to the story. This is a global function (as most rooms use it) so it involves checking a list of "if player.room ==" choices. Plus checking and changing hotspot text, all in a single frame while other demanding animations may be playing. All those  "if...then" decisions quickly add up!

[Akatosh]

Hah, yeah, I can imagine that. But don't worry, repeatedly_execute and the like aren't run "a few hundred times per frame" anyway - it's executed exactly once per frame (which means ~40 times a second on default GameSpeed settings), which should still be more than enough.

[Pumaman]

Interesting question.

In the first example, each line of code equates to the following AGS byte-code:

Line  59, IP:     324 (SP:98243628) mov mar, 8221696
Line  59, IP:     327 (SP:98243628) push mar
Line  59, IP:     329 (SP:98243632) pop mar
Line  59, IP:     331 (SP:98243628) memread.ptr ax
Line  59, IP:     333 (SP:98243628) push ax
Line  59, IP:     335 (SP:98243632) pop mar
Line  59, IP:     337 (SP:98243628) checknull.ptr
Line  59, IP:     338 (SP:98243628) add mar, 52
Line  59, IP:     341 (SP:98243628) memread ax
Line  59, IP:     343 (SP:98243628) push ax
Line  59, IP:     345 (SP:98243632) mov ax, 1
Line  59, IP:     348 (SP:98243632) pop bx
Line  59, IP:     350 (SP:98243628) cmp bx, ax
Line  59, IP:     353 (SP:98243628) mov bx, ax
Line  59, IP:     356 (SP:98243628) jz 5

(ie. 15 instructions)
In the second example, each of the five "if" statements equates to:

Line  69, IP:     568 (SP:98243632) load.sp.offs 4
Line  69, IP:     570 (SP:98243632) memread ax
Line  69, IP:     572 (SP:98243632) push ax
Line  69, IP:     574 (SP:98243636) mov ax, 1
Line  69, IP:     577 (SP:98243636) pop bx
Line  69, IP:     579 (SP:98243632) cmp bx, ax
Line  69, IP:     582 (SP:98243632) mov bx, ax
Line  69, IP:     585 (SP:98243632) jz 5

(ie 8 instructions).
HOWEVER, in the second case there are 12 additional instructions involved in creating the local variable, which adds an initial extra overhead.

In practice though, the difference is so small that you really shouldn't worry about it; a typical rep_exec in a large game results in thousands of instructions being executed each game loop, which doesn't cause a performance problem. The use of single, slow commands such as raw drawing will have a far larger impact on performance than which method of checks you use in an "if" statement.

Note that using "else if" instead of just "if" here will speed it up a bit, since it can skip the rest of the checks after one of them matches. But again, the difference is marginal.

[EnterTheStory (aka tolworthy)]

Many thanks! That's very, very interesting.