Performance difference??

Shy Agam

Hello experts, There's something that has been bothering me for quite some time now... Is there a performance difference between the following two implementations?

for (...)
{
    int x = ...;
    // Do some stuff
}

int x;

for (...)
{
    // Do same stuff
}

In the first case, is x getting allocated for every step of the loop? Or does .NET knows to allocate it once, and use it again and again? Thanks in advance, Shy.

BoneSoft

I haven't looked at the IL, but I'm pretty sure you're creating and dumping a new integer every interation in your first example. And therefore, yes, it performs worse.

Try code model generation tools at BoneSoft.com.

Shy Agam

I see... Is it the same for reference types?

Luc Pattyn

Hi, you should not believe everything they tell you. int is a value type, it probably gets "allocated" on stack (meaning the stack pointer gets lowered a bit more when you enter this method to reserve a spot for it), this has zero cost. (the probably means, maybe the int variable does not need storage, the JIT might decide to keep it in one of the CPU registers, but I am not relying on this) Anyway, there is no "new", no object, and no garbage collection involved. both code snippets could and should result in exactly the same MSIL code; and I would be very surprised if you can measure the slightest difference in execution time (which would hint a compiler inefficiency, not an inherent difference between both code snippets). :) :)

Luc Pattyn

try { [Search CP Articles] [Search CP Forums] [Forum Guidelines] [My Articles] } catch { [Google] }

Paul Conrad

Time it through the PerformanceCounter class for large iterations in the for loop and compare results...

"Real programmers just throw a bunch of 1s and 0s at the computer to see what sticks" - Pete O'Hanlon

Luc Pattyn

Hi, please read my other reply first. if it is:

string x="aha";

for (...) {
string y=x;
}

then again it wont make a difference, since y=x simply copies a reference, it does not create an object; it has the same cost as int y=x; And again, if the compiler notices x can not change during the loop, it should and would move the statement outside the loop. :)

Luc Pattyn

try { [Search CP Articles] [Search CP Forums] [Forum Guidelines] [My Articles] } catch { [Google] }

Shy Agam

Exactly the answer I was looking for! ;) Thanks!

BoneSoft

I stand corrected, both of these...

public void Inside() {
for (int i = 0; i < 1000; i++) {
int x = i;
}
}

public void Outside() {
int x;
for (int i = 0; i < 1000; i++) {
x = i;
}
}

decompile to something similar to this...

public void Outside() {
int num1;
int num2;
for (num2 = 0; (num2 < 1000); num2 = (num2 + 1)) {
num1 = num2;
}
}

The compiler does you the favor of moving your variable outside the loop where it belongs.

Try code model generation tools at BoneSoft.com.

Luc Pattyn

:rose:

Luc Pattyn

try { [Search CP Articles] [Search CP Forums] [Forum Guidelines] [My Articles] } catch { [Google] }

snorkie

For what its worth, I just ran a test and got the following results. It looks like it doesn't matter for something this small. Inside the loop took 552645639 ticks. Outside the loop took 549520699 ticks. The test code: long timeIn = 0; long timeOut = 0; long l = System.DateTime.Now.Ticks; for (long x = 0; x < 10000000000; x++) { int i = 1; i++; } timeIn = System.DateTime.Now.Ticks - l; l = System.DateTime.Now.Ticks; int i2; for (long x = 0; x < 10000000000; x++) { i2 = 1; i2++; } timeOut = System.DateTime.Now.Ticks - l; MessageBox.Show("Time In: " + timeIn.ToString() + Environment.NewLine + "Time Out: " + timeOut.ToString()); Hogan

PIEBALDconsult

Indeed, I found the same thing last week.

Luc Pattyn

So the measured difference is less than 1% which is I guess well below your measurement accuracy: you are measuring elapsed time (good) over a long period (50 sec, good), but circumstances vary: there are hundreds of threads in an idle system, watching each and every part of your machine, there are disk transfers with DMA, there is a varying amount of Ethernet traffic including broadcast messages you are not interested in, etc etc. So my first impression is the execution times are the same. What you could do is reduce the loop count by a factor of 100, run it 100 times, and observe the distribution of the 100 timing results. The disappointment is neither the C# compiler nor the JIT compiler was smart enough to recognize the entire loop body could and should be moved outside, and then the loop is without subject; so it should have been hundred ticks at most ! On a non-CLR language (such as good old C) it takes quite some tricks to prevent the compiler from throwing it all out, and still get an accurate measurement ! :)

Luc Pattyn

try { [Search CP Articles] [Search CP Forums] [Forum Guidelines] [My Articles] } catch { [Google] }

Luc Pattyn

Hi, This is related to the formatting, not the topic, of your message. I am preparing an article on copying code snippets from a CodeProject message board to Visual Studio; seems there are no problems when the browser used is FireFox, Safari, ... but often problems arise when using Internet Explorer. One of the factors is the tags that are present inside the < PRE> </PRE> block. In your message the lines are separated by <br> tags; most other messages have just regular newlines. Could you please tell me how you get that, i.e. what tools you use, and how you go about it. Thanks in advance.

Luc Pattyn

try { [Search CP Articles] [Search CP Forums] [Forum Guidelines] [My Articles] } catch { [Google] }

Shy Agam

I'm not really sure if I understood what you mean, hence I don't know if this is the answer your after... I use Maxthon browser, and when inserting code snippets to my posts I wrap the <code></code> tags inside of the <pre></pre> tags, and not the other way around. If you could explain again, maybe I would be able to give you a better answer. :) Regards, Shy.

Luc Pattyn

Thanks, that is exactly what I wanted to know. I am not familiar with Maxthon browser, I never heard of it. The background is: if I select (part of) your code snippet using Internet Explorer, then copy it in Visual Studio, all formatting is lost, there are no line breaks any more. I created a utility that improves the situation in most cases; with it I manage to get newlines, and most of the time the indentation remains intact too; on your code snippet, with the tool, I do get newlines (so that is an improvement) but now I loose indentation. My article will suggest people use the PRE tags without adding CODE tags to it; could you please give that a try, so just reply to this message, and insert a code snippet as you usually do (that is using PRE tags) but this time without using CODE tags. Thanks in advance.

Luc Pattyn

try { [Search CP Articles] [Search CP Forums] [Forum Guidelines] [My Articles] } catch { [Google] }