Memory allocated to variables of a function on different function calls

Hello World !!
Today I was reading a topic on 'Local Variables' from 'Thinking in C++ by Bruce Eckel', there Bruce has mentioned that "variables defined local to a function disappear at the end of the function scope. When you call a function again, storage for the variables is created anew and the values are reinitialized."

Now from this I inferred:

Inference 1. variables local to a scope get destroyed as soon as their scope ends.
Say, when we call a function again and again, all its variables are allocated new memory locations and as soon as the function execution gets over all the memory locations are set free.

And I wrote the following code to seal my inference

Code:

---

#include<iostream>using namespace std;


void func()
{
    int i;
    cout<<"Address of i in func(): "<<(long)&i<<endl;
}


int main()
{
    int i;
    cout<<"The Address of i in main(): "<<(long)&i<<endl;
    for(int i=0;i<3;i++)
    {
        func();
    }
    cout<<"The Address of i in main() again: "<<(long)&i<<endl;
    func();


    return 0;
}

---

The results were as follows:

Code:

---

The Address of i in main(): 2293512
Address of i in func(): 2293468
Address of i in func(): 2293468
Address of i in func(): 2293468
The Address of i in main() again: 2293512
Address of i in func(): 2293468

---

According to Inference 1, on every func() call, 'int i' inside it should have been allocated different memory location, whereas here I see that on every func() call the memory location allocated to 'int i' is same

from the results of the above code I am forced to infer this:

Inference 2: for every func() call, the variables disappear at the end of the func() execution and the memory is set free but this freed memory still resides in the big memory chunk allocated to the program at the beginning of its execution, so at the beginning of every func() call the compiler does allocate anew memory location to 'int i' but every time it allocates the memory it lays its hands on the readily available memory locations which turn out to be the same in all the func() calls.
this may explain the same memory address for 'int i' on every call

so my question is: which of the two inferences is true or are they both a big fat fallacy ?
friends, please help me out, I am just a humble beginner of C++ :)

Quote:

---

When you call a function again, storage for the variables is created anew and the values are reinitialized

---

Variables created anew may not mean they 'must' have a different memory location.
Here they have, but if you, for example put a call to func() inside another function, you may see a different address.

This is about memory segmentation. Inference 2 is sort of correct, as is manasij7479.

When a function gets called, its data is placed in the (call) stack segment. When it exits, it is removed. This means if you call a function that calls a function that calls a function you will have a number of stack frames, one for each active function call.

So if you call a function sequentially from somewhere, its data will end up in the same place each time, as you surmise. But if you call it (eg) recursively, each call has its own frame in a new location. Try this:

Code:

---

#include <stdio.h>

void eg (int r) {
        if (!r) return;
        fprintf(stderr,"%d %p\n", r, &r);
        eg(r-1);
}

int main() {
        eg(3);
        eg(1);
        eg(1);
        eg(1);
        return 0;
}

---

Notice in the first three the address of r is different whereas in the last three it is always the same.

Worth noting that this is not a product of any C or C++ standard, and so is not guaranteed. It's determined by the operating system and compiler, which can organize memory in any way what so ever (but mostly as described).

You may find some value in:
Understanding the Stack
Call Stack wiki

As a side note, there is no need to "infer" in computer science. This isn't cosmology or quantum mechanics where you are observing unknown behavior and then having to figure out why you observed it. Computer language implementation is a well documented field as well as OS design.

Quote:

---

there is no need to "infer" in computer science

---

Reverse Engineering FTW .

Quote:

---

Originally Posted by manasij7479

Reverse Engineering FTW .

---

This board always makes me chuckle. :rolleyes:

Quote:

---

Originally Posted by AndrewHunter

As a side note, there is no need to "infer" in computer science.

---

Very strongly disagree. Ideally speaking, you might have a point about how it might be nice if it were really so. But in reality, this is totally ridiculous.

Also, inferring is a useful skill worth developing. Sort of like poker might be, YMMV ;)

Quote:

---

This isn't cosmology or quantum mechanics where you are observing unknown behavior

---

I did not invent atomic physics, the universe, microprocessors, assembly instructions, the C standard, or my operating system, so I'm afraid these all provide me with ample opportunity to observe "unknown behavior".

Quote:

---

Computer language implementation is a well documented field as well as OS design.

---

And the people who produce much of that "documentation", whatever their skills at programming and engineering, are often totally pathetic at the task, and produce material that no doubt makes sense to the person who wrote it because they understand said material already, but fails to explain anything to anyone else.

In fact, I believe it is the norm in programming to code first, answer questions later. Meaning the docs easily lag behind the implementation -- if they even exist.

I think the OP is completely right to read some and theorize some if what you are recommending is that instead s/he should "just read EVERYTHING first", which no single individual could possible have time to do in a normal lifetime, and would probably not get them far anyway without some good old fashioned conjecture and experiment (and BTW, without that you do not have a "science").

Trust no one 100%, and instead prove things for yourself.

Ok, before we get way off track here, let's revisit my side comment. No offense was intended to the OP, and after providing useful links to answering his question, I was simply attempting to state that these resources are out there to learn from. I do strongly agree with experimenting and learning from those experiences in programming.

My comment was more or less just a statement about the word he used and the connotation that comes with it. As you have stated you did not write the C++ language or the OS you are using, however there is a C++ standard that defines what the language will and will not do, and an associated OS documentation that clearly lays out what it will and will not do and how it will do it.

I do not have a problem with the OP's question, I was simply pointing out that once you observe some pattern from your programs that you should strive to figure out why, by using the documentation available of which there is plenty. This is what we tell people all the time here.

Quote:

---

"variables defined local to a function disappear at the end of the function scope. When you call a function again, storage for the variables is created anew and the values are reinitialized."

---

What you haven't realised is that this is not a statement about defining the exact actual inner workings of a compiled program. This is a statement about the concepts and guarantees (or lack thereof) in what you should rely on to be true with regards to the programming language itself.

Nothing in the quoted text actually amounts to testable behaviour, so writing any program to test the statement, is wrong before you've started. It is like testing a "No Trespassing" sign by attempting to trespass, succeeding in doing so, and then claiming that the sign is wrong. Of course you might have been instantly electrocuted the moment you tried to trespass, but as fate would have it, you weren't. In reality it simply means that you missed the point.

You shouldn't be trying to infer specific behaviour from the statement made. That's not to say inferring in general is wrong, but it's pretty much wrong in this case.

I could easily write a basic program for you that does happen to show a variable inside a function having a different address from a different call. A recursive function would be one such easy example. But I wont, because it isn't a point that needs proving.

Quote:

---

Originally Posted by MK27

This is about memory segmentation. Inference 2 is sort of correct, as is manasij7479. When a function gets called, its data is placed in the (call) stack segment.....

---

yeah you're right, I think the "inferred" conclusion can be as follows:

if a function is called sequentially from somewhere in the program then all its data will end up in the same memory locations. But if we call a function from inside of another function(eg. recursively), then each call will have its own frame in a new memory location

Quote:

---

Originally Posted by AndrewHunter

My comment was more or less just a statement about the word he used and the connotation that comes with it.

---

"Inference" is essential to, and part of, "being human". we engage in inference every day. we infer it is raining/sunny when we see someone with an open umbrella. we infer people are thirsty if they ask for a glass of water. we infer that evidence in a text is authoritative when it is attributed to a scholar in the field. Similarly we "infer" the functioning of the computer/compiler by observing the behaviour of the program and thats what I did and I'm good at doing ;)

Quote:

---

I do not have a problem with the OP's question, I was simply pointing out that once you observe some pattern from your programs that you should strive to figure out why, by using the documentation available of which there is plenty. This is what we tell people all the time here.

---

I strongly disagree with your "authoritative advice" and why is it that before or after experimentation, instead of going for a knowledgeable discussion, one should force himself to drown in the vast ocean of documentations just to find a small piece of useful information
if this was the way computer scientists wanted us learn C++ then Bjarne, in his books, after every topic would have forced us to go through some link to documentation before proceeding further

Quote:

---

My comment was more or less just a statement about the word he used and the connotation that comes with it. As you have stated you did not write the C++ language or the OS you are using, however there is a C++ standard that defines what the language will and will not do, and an associated OS documentation that clearly lays out what it will and will not do and how it will do it.

---

if the world of science(in particular comp sc.) was to follow your advice then there would have been no scope for inferences(or learning), because every discussion in this world would have been ended before even a start by someone like you with a link to some boring-indecipherable piece of text called as "documentation"
after all "questioning, experimenting and inferring" is the process we call learning :)

Quote:

---

Originally Posted by MK27

Trust no one 100%, and instead prove things for yourself.

---

I second you on that :)
and thanx a lot for clearly explaining and somewhat backing the "Inference 2" :)

Quote:

---

Originally Posted by iMalc

Nothing in the quoted text actually amounts to testable behaviour, so writing any program to test the statement, is wrong before you've started. It is like testing a "No Trespassing" sign by attempting to trespass, succeeding in doing so, and then claiming that the sign is wrong. Of course you might have been instantly electrocuted the moment you tried to trespass, but as fate would have it, you weren't. In reality it simply means that you missed the point.....

---

I strongly disagree with your "No Trespassing" theory where one should force himself into categorizing down any statement on programming behaviour into either some theoretical "one-should-not-care-to-test" statement or a statement (according to you) having some worthy "testable behavior"; because if I would'nt have tested that statement and initiated a discussion here then I might not have(if not later) learnt the "inferred conclusion" in the post above.
OR can you please go ahead and kindly define the parameters you use to judge a statement's worth to be tested or not

Quote:

---

I could easily write a basic program for you that does happen to show a variable inside a function having a different address from a different call. A recursive function would be one such easy example. But I wont, because it isn't a point that needs proving.

---

Revered Sir,
proving that might be an utter foolishness and worthless for you but for a humble beginner like me, the learning from the proof is worth its salt :)

\\Peace :)//

The point iMalc et al are making is that compiler writers can do as they damn well please, because unworthies such as us (who are writing the code) are not allowed (by the language) to actually inspect what's going on behind the scenes from the behavior of the code itself. [This doesn't mean people don't cheat, sometimes, by writing "illegal" code.] The machine has to behave "as if" the memory was returned and then reallocated, but it doesn't have to actually do that; if you've got the optimizer turned on high enough, and the compiler writer was sufficiently clever, then your code at the beginning could elide the stack building and un-building that happens during a function call and use the same memory for the whole program. You could tell by examining the generated machine code, but not from the behavior of the program (you would just see it as "luck" that the machine chose the same address each time).

If you make the code more complex, you make it more likely that the program has to behave to specs (i.e. you make optimizations harder to find), and calling more than one function in a recursive manner will almost certainly guarantee different values since you will have to have more than one i running around.

Quote:

---

Originally Posted by tabstop

The point iMalc et al are making is that compiler writers can do as they damn well please, because unworthies such as us (who are writing the code) are not allowed (by the language) to actually inspect what's going on behind the scenes from the behavior of the code itself.....

---

You nailed your point :)
I think the explanation MK27 gave if viewed from your "optimization" point of view yields a clearer picture.
Thanx a lot for contributing this invaluable view point

Quote:

---

You could tell by examining the generated machine code, but not from the behavior of the program.

---

can you please tell me how can we examine the generated machine code and deduce the findings? or is it asking for more(which I'm daringly fond of) ;)

"Generated machine code" might be a little bit of a stretch, but presumably if you knew all the opcodes for the chip, and the binary format, etc, then you could "read" an exe file. I don't know anyone who's ever even tried this, and wouldn't expect anyone to actually do so.

But you can get some idea by having your compiler stop at generating the assembly code, which has got pretty much all the gory details while still at least pretending to be readable. How you do that depends on your compiler. (You can find some information about assembly language on the net; the last time I looked I didn't find anything extremely up-to-date (because people just aren't writing assembly that much any more), but for programs such as the one you've got it will be good enough.)

If you are looking into assembly language then check out: Art of Assembly Language - Hyde. It provides a pretty good tutorial on the subject matter, specifically the HLA or High Level Assembly section.

Quote:

---

Originally Posted by tabstop

"Generated machine code" might be a little bit of a stretch, but presumably if you knew all the opcodes for the chip, and the binary format, etc, then you could "read" an exe file. I don't know anyone who's ever even tried this, and wouldn't expect anyone to actually do so.

But you can get some idea by having your compiler stop at generating the assembly code, which has got pretty much all the gory details while still at least pretending to be readable. How you do that depends on your compiler. (You can find some information about assembly language on the net; the last time I looked I didn't find anything extremely up-to-date (because people just aren't writing assembly that much any more), but for programs such as the one you've got it will be good enough.)

---

Well, to be honest, I've found myself looking at the assembly of generated C++ code from time to time to find out what the heck is happening.
But to disassemble an entire executable file is probably not very many people who do.

Quote:

---

Originally Posted by tabstop

"Generated machine code" might be a little bit of a stretch, but presumably if you knew all the opcodes for the chip, and the binary format, etc, then you could "read" an exe file......

---

That seems way too complicated for the moment, I think I'll keep this experimentation reserved for future
anyway thanx for the tip :)

Quote:

---

Originally Posted by capt.jack

I strongly disagree with your "No Trespassing" theory where one should force himself into categorizing down any statement on programming behaviour into either some theoretical "one-should-not-care-to-test" statement or a statement (according to you) having some worthy "testable behavior"; because if I would'nt have tested that statement and initiated a discussion here then I might not have(if not later) learnt the "inferred conclusion" in the post above.
OR can you please go ahead and kindly define the parameters you use to judge a statement's worth to be tested or not

---

It's not about worth. Is a no trespassing sign worth testing? It simply doesn't make sense. You made assumptions about the statement such as thinking that variables could not occupy the same space in memory they did last time. Whether they do or not has no bearing on the assumptions the author is asking you to make or not make.

That wasn't a theory I mentioned, it was a metaphor for the particular statement in question. By no means do I intend the same logic to apply to things willy-nilly. Really the original statement being discussed comes down to the author saying "assume nothing". Now you can make assumptions and then test those assumptions, but as I think you've now come to realise, that doesn't and in fact cant invalidate what the original statement says. Just assume no knowledge of the location of your variables is the intent the author is trying to get across here.

Quote:

---

Revered Sir,
proving that might be an utter foolishness and worthless for you but for a humble beginner like me, the learning from the proof is worth its salt :)

---

You simply misunderstood. It's not about not being worthy of proving, it makes no sense to try and prove, just like the no trespassing sign example, because it is not a statement about the behaviour that your compiled program will actually give.

Quote:

---

Originally Posted by iMalc

Really the original statement being discussed comes down to the author saying "assume nothing". Now you can make assumptions and then test those assumptions, but as I think you've now come to realise, that doesn't and in fact cant invalidate what the original statement says. Just assume no knowledge of the location of your variables is the intent the author is trying to get across here.......

---

the question is not whether I misunderstood you or not, whether you mentioned "No Trespassing" as a metaphor or as a theory but rather the question is "How in the world would I know whether the assumptions made by me are true or a big fat fallacy" ? or whether I'm "not allowed" to make any assumptions at all ?
Unless I experiment and post my doubts here, how can I learn that the statement at hand is the "world's-most-obvious-fact" ?

Quote:

---

It's not about not being worthy of proving, it makes no sense to try and prove, just like the no trespassing sign example, because it is not a statement about the behaviour that your compiled program will actually give

---

you being an experienced C++ expert, is in a position to make that call but for beginners like me experimenting and then asking the experts is the right way to proceed on the learning curve, in fact even to make the "obvious" distinction of whether the statement is about the behaviour that your compiled program will give or not, I have to experiment, thats the way I am because for me experimentation is the road to salvation

Quote:

---

Just assume no knowledge of the location of your variables is the intent the author is trying to get across here.

---

I respectfully refuse to take things "granted" and I am determined to experiment and post as many foolish questions as I can, even though it may make me the laughing stock of the forum or bag me many funny jibes(or even "NoTrespassing" metaphors) or whatever, but I don't care until I am learning


Stay Hungry, Stay Foolish :)

Experimentation is good. But there are simply some things that you cannot get a proper understanding from through experimentation alone. Memory management like this is one of them, because it will be different from platform to platform and various circumstances.
If you want to know more, then you shouldn't experiment, but instead study operating system principles and see how they would deal with it.