|
Thanks for the link, Pete. That one looks really good 
"Real programmers just throw a bunch of 1s and 0s at the computer to see what sticks" - Pete O'Hanlon
|
|
|
|
|
Hi, I read some paper about how clr works on managed modules, managed heap and jit the managed assembly. That works on heap (reference type). It seems me that only the reference to those oblects are on the stack but the real objects are on the heap. Value type instead are allocate on the stack. I'm wondering what changes if also objects were allocate on the stack; what'll happen? If also objects will be on the stack where do the jitting go? Will I have need of assembly; What changes are necessary, please?
thank you.
|
|
|
|
|
MarKus0 wrote: I'm wondering what changes if also objects were allocate on the stack; what'll happen?
Nothing happens because object instances are NEVER allocated on the stack.
|
|
|
|
|
yes, I know that.
But the question was different: what changes should I do inside CLR if I want allocate object on the stack, what changes on the CLR Should I do? (I want change CLR for this, it's a theoretical question, my curiousity)
thanks
|
|
|
|
|
You can't allocate objects on the stack because each position in the stack will only hold, on a 32-bit proc, a 32-bit VALUE. You can't allocate on object on the stack since an object can be any type, any size, and composed of one or more types, plus code.
It's possible to convert an object to a serialized array of values that can be pushed onto the stack, but the method that you call would have to deserialize the data, recreate the object, and could then work it. But, that would defeat the entire purpose of putting it on the stack in the first place AND introduce a huge performance penality with the serialization/deserialzation process.
I'm starting to question if you really know what the execution stack is and how it works.
|
|
|
|
|
The very nature of a reference type excludes the possibility of it going on the stack.
For example take this bit of code
public MyMethod(SomeObject obj)
{
MyReferenceType rt = new MyReferenceType();
obj.PropertyToKeep = rt;
}
Now, rt is destroyed at this point becuase the stack has shrunk. But, obj still maintains a reference to it. What happens?
In C++ you would end up with a bad pointer that points to a piece of memory that no longer contains the object. The CLR simply doesn't allow this situation to occur in the first place.
Upcoming events:
* Glasgow: Mock Objects, SQL Server CLR Integration, Reporting Services, db4o, Dependency Injection with Spring ...
"I wouldn't say boo to a goose. I'm not a coward, I just realise that it would be largely pointless."
My website
|
|
|
|
|
OK; I have to explain better. Focus on .NET C# (but it isn't the point of my question). I have to design a new CLR changing the actual CLR. My new CLR must allocate objects on the stack. So, how do I have to change the current CLR to allocate oblìjects on the stack? (probabily this new CLR doens't make sense (probabily because everythings goes on the stack)), but I need to think what I have to change to reach this aim...
thanks
|
|
|
|
|
Make everything a value type?
|
|
|
|
|
I don't know how....I'm thinking the way ( and after I'll can say if the thing is good or not). But what I want find is a solution that the CLR can manages objects on the stack instead on the heap. I know this hard/strange question....but will it be possible? I'm finding an rough idea (what'll do the loader...., the pointer, the Jit)...any idea?
|
|
|
|
|
A stack is a data structure where things get added on one end, in a particular
order, and get removed, at the same end, in reverse order. That is the only thing
a stack is capable of (and it is very good at that).
you can't allocate reference types on a stack: the stack is used for
keeping track of program flow (calling and returning from methods) and for
local value types (which live as long as the method is executing).
If you return from a method, the stack is reset to the state it had when
you entered the method, that is the only thing acceptable to the calling method.
Ref types must live from construction (with new) until death (last reference
got lost); they don't die in reverse order of construction, and they get
constructed and die independent of program flow, hence there is no room
whatsoever to allocate ref types on the stack.
In summary: a system can not allocate ref types on the stack.
It needs separate memory to hold them, such memory is called a heap.
Actually ref types come from a couple of heaps, depending on the size of
the objects (small objects get shuffled around to avoid heap fragmentation,
large objects come from the large object heap, which is not reshuffled,
and may get exhausted due to fragmentation, something Micro$oft tries not
to focus your attention on).
Luc Pattyn [Forum Guidelines] [My Articles]
this weeks tips:
- make Visual display line numbers: Tools/Options/TextEditor/...
- show exceptions with ToString() to see all information
- before you ask a question here, search CodeProject, then Google
|
|
|
|
|
That was an excellent answer right up until
Luc Pattyn wrote: Micro$oft
I hate that. Microsoft is a business. It is their job to make money and they happen to be very good at it.
|
|
|
|
|
Hi Colin,
Colin Angus Mackay wrote: Microsoft is a business. It is their job to make money and they happen to be very good at it.
I agree.
But I would like them to be a little better at offering things that make life easy
for programmers/developers, even if that would (seem to) take away one percent
of their profits.
Why are line numbers not displayed by default in Visual source files ?
Why does Windows show a silly error number when trying to run a .NET app on
a system that does not have (the right version of) .NET installed ?
Why don't they provide a full set of P/Invoke prototypes for all Win32 structures
and functions ?
Why does it take a network guru to set up a wireless network ? (are there any
non-technical people who get any level of security?)
Why do we have to view or enter stuff in 200*200 pixel windows and dropdowns,
when the screen has over 1M pixel ?
Why does most MSDN documentation on API stuff come without examples ?
Why don't they provide a list of probable causes, in non-academic terms, for
the most frequently occuring error situations ?
...
the list is infinite.
In general, is it so hard to listen to users and to read forums (like this one)
and learn from it, as to how the user's experience could be improved (rather than
just adding features, complexity and sometimes mystery). In my opinion most of
what goes on on these forums should be made redundant by step-by-step improvements
to the MS products and their documentation.
Conclusion: as long as dollars/euros/yens seem to be more important than customers
and their little problems, I tend to occasionally replace an 's' by a '$'
Regards
Luc Pattyn [Forum Guidelines] [My Articles]
this weeks tips:
- make Visual display line numbers: Tools/Options/TextEditor/...
- show exceptions with ToString() to see all information
- before you ask a question here, search CodeProject, then Google
|
|
|
|
|
Gosh! That's not what I see at all. Microsoft are very responsive for such a large company and they do try their best for the community. They are improving constantly, it just takes time. For instance, there are a lot more examples in MSDN than their used to be.
There are a lot of new and innovative things on the horizon which I am sure will address some of your concerns.
|
|
|
|
|
So, it's the OS that allocate space on the stack for the process?! If it's so, I think that can't exists a (new) CLR that manage stack..!? I understand that for my question, CLR would have to do the work of OS; so i would have to change OS too; is it that?
Then arise one question on C#:
<br />
namespace Simple {<br />
class Program {<br />
static void Main(string[] args)<br />
{<br />
Console.WriteLine("that's the program");<br />
int a = 0;<br />
myClass mc = new myClass(); <br />
}<br />
}<br />
}<br />
I read that every value type (int32, enumeration, and other) are on the stack; but if they're inside a class declaration they take part of object and then go on the heap. In c# even main is inside a class (in my example " Program"); with this knowledge, it seems to me that evey things goes on the heap. Could you explain to me what go on the stack ? (use my example code, please)
Thanks...
|
|
|
|
|
I already told you what goes on the stack, and why the other things don't fit
on a stack. Things are the way they must be. For me the subject is closed.
Luc Pattyn [Forum Guidelines] [My Articles]
this weeks tips:
- make Visual display line numbers: Tools/Options/TextEditor/...
- show exceptions with ToString() to see all information
- before you ask a question here, search CodeProject, then Google
|
|
|
|
|
No - let him implement the OS as well. That should be worth seeing.
Deja View - the feeling that you've seen this post before.
|
|
|
|
|
OK, on one condition, Chris should start a separate forum for it, since I expect
a couple more discussion threads before the new OS is up and running...
Luc Pattyn [Forum Guidelines] [My Articles]
this weeks tips:
- make Visual display line numbers: Tools/Options/TextEditor/...
- show exceptions with ToString() to see all information
- before you ask a question here, search CodeProject, then Google
|
|
|
|
|
Seems fair. He'll need a cool name for the project though. How about calling it "Pillock" or "Idiot"? That way we can ask people "Are you running Idiot?"
Deja View - the feeling that you've seen this post before.
|
|
|
|
|
Hmm, learning a least one new word every day now. I like that.
Luc Pattyn [Forum Guidelines] [My Articles]
this weeks tips:
- make Visual display line numbers: Tools/Options/TextEditor/...
- show exceptions with ToString() to see all information
- before you ask a question here, search CodeProject, then Google
|
|
|
|
|
MarKus0 wrote: So, it's the OS that allocate space on the stack for the process?!
No, not entirely. The processor has a stack the is allocated on a per-thread basis. Ever here of the Stack Pointer? Read[^]
It's the processor that's stopping you from doing what you want, not the .NET Framework or the O/S.
|
|
|
|
|
Dave - don't let facts get in the way of our mocking him.
Deja View - the feeling that you've seen this post before.
|
|
|
|
|
My bad! :-> Please, proceed. I love a good show! 
|
|
|
|
|
Dave Kreskowiak wrote: It's the processor that's stopping you from doing what you want, not the .NET Framework or the O/S.
IMO that is not entirely true; this is how I see it:
- most processors don't support heaps at all, yet lots of OS need heaps, so these
get implemented by software;
- stack structures can be implemented by software (e.g. the Stack class in .NET);
- there (still) are processors that don't provide hardware support for a stack,
yet a stack-based language (and OS) can made to run on them; when they have say
a shadow register for PC (into which the PC gets copied upon CALL or INT), each
function must start saving the shadow PC on a software stack;
- alternatively, if the CPU offers stack support (i.e. pushes the PC to a memory
location thru a pointing register) and you don't like the way it works, you can
undo it by software, and keep track of program flow in some other way.
- IIRC Intel's IA432 architecture did not have real stack support, instead it
allocated nodes (on the heap!) that got linked back and forth, resulting in
a distributed structure with stack behavior.
So I would say most OS really want to have hardware supporting a stack, and
hence most chip vendors provide exactly that, but either one can choose to do it
differently (which they seldom do).
Luc Pattyn [Forum Guidelines] [My Articles]
this weeks tips:
- make Visual display line numbers: Tools/Options/TextEditor/...
- show exceptions with ToString() to see all information
- before you ask a question here, search CodeProject, then Google
|
|
|
|
|
I was speaking from the Intel/AMD perspective, which does support execution stacks.
Luc Pattyn wrote: - alternatively, if the CPU offers stack support (i.e. pushes the PC to a memory
location thru a pointing register) and you don't like the way it works, you can
undo it by software, and keep track of program flow in some other way.
I don't think he's up to the task of re-writing the O/S just yet!
|
|
|
|
|
Guys please answer this..Why value type instances are int stack?
M.Sendilkumar
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
