All Your Base Are Belong To Us - All Comments

re: HTML5 Web Workers: Classic Message Passing Concurrency

Sasha Goldshtein — Tue, 07 Feb 2012 08:14:54 GMT

@David: "Better" is relative :-) For experienced developers with a strong background in operating systems, cache coherence protocols, and synchronization mechanisms -- I agree that the "traditional" way is more flexible and more powerful. Most developers I know, however, abuse that power due to lack of that strong background...

@peter: Because the "isPrime" implementation is very naive -- it checks all possible factors from 2 to n-1.

re: HTML5 Web Workers: Classic Message Passing Concurrency

peter — Mon, 06 Feb 2012 20:00:50 GMT

Why is the distribution uneven?

re: HTML5 Web Workers: Classic Message Passing Concurrency

David Nelson — Mon, 06 Feb 2012 17:16:02 GMT

I don't envy JavaScript developers at all. It is trivial to write multi-threaded code in C# using a message passing style if that is what you want to do. But C# ALSO has the ability to use shared state and synchronization where it is appropriate (and there are many such cases). Being less powerful and less functionally complete does not make JavaScript better.

re: Windows To Go

Matt — Sat, 04 Feb 2012 08:36:27 GMT

And to think I was gonna stuff around with Virtualisation to make sure it was easy to upgrade to new hardware later at home, THIS IS AWESOME, no more annoying hardware upgrades...

re: Windows To Go

Anant Kamath — Tue, 31 Jan 2012 09:26:29 GMT

This has benn possible with several Linux distros since ages

re: Garbage Collection in The Age of Smart Pointers

Fulvio Esposito — Mon, 30 Jan 2012 20:22:27 GMT

Well, C++11 offers a minimal support for garbage collection, but the point is that C++ memory model is always deterministic.

When you use automatic storage, dtors are granted to be excuted upon block leaving (whatever is the reason, so don't need to try/catch/finally).

With unique_ptr<T> we have the same behaviour of automatic objects for objects allocated on the heap

With shared_prt<T>/weak_ptr<T> dtors are granted to be executed when last owner is destoyed.

All of this means that we manage all the resources, not just the memory, the same way and there's always a guarantee about resource deallocation time (that's a very valuable and desiderable property).

If fragmentation is a problem, C++ let you override standard allocatino mechanism. But fragmentation is a problem just in very limited contexts. If allocation time is a problem, no-one forbid to provide a memory allocation scheme similar to what a garbage collection system do, preallocate a large block and just returns pointer to internal offset of the block on memory demand).

In my opinion, with GC we trade all resource management for just automatic memory management.

in c++

...

{

fstream f( "a.out" );

f << "hello";

}

{

fstream f( "a.out" );

f << "hello";

}

...

the equivalent correct java version is

...

{

FileStream fs = new FileStream( ... );

try

{

fs.write( "hello" );

}

finally

{

if( fs != null )

fs.close();

}

try

{

fs.write( "hello" );

}

finally

{

if( fs != null )

fs.close();

}

it's really an improvement in productivity?

Just my 2 cents

re: Garbage Collection in The Age of Smart Pointers

Karan Kadam — Sun, 29 Jan 2012 17:35:27 GMT

Excellent article. Good insight into GC vs smart pointers. Thanks!

re: Garbage Collection in The Age of Smart Pointers

Catalin — Sat, 28 Jan 2012 22:23:41 GMT

You missed the core aspect of the things: what about DETERMINISTIC vs NON-DETERMINISTIC facet of the problem. This is not a minor issue, fast system programming can't rely on non-deterministic destruction, so I don't agree with the conclusion.

re: Managed-Unmanaged Interoperability in the 2010s: Part 2, P/Invoke – Unmanaged to Managed

Jeroen Frijters — Sat, 28 Jan 2012 10:16:17 GMT

LPARAM should be IntPtr

re: .NET to C++ Bridge

Kiran — Fri, 27 Jan 2012 19:34:03 GMT

Hi there!

I tried using this solution, but does not work for my environment. One small difference is that the application that loads my native library fails to load, if there are mixed dll dependencies -- or so, it seems. It loads perfectly, if i comment out all references to the mixed model dll, in my native library.

Has anyone run into this issue? Any luck?

Thanks,

re: CLR Stack Explorer – Preview

Andrew — Fri, 20 Jan 2012 14:03:07 GMT

The tool is useful to me too.

Having source code would be great.

Several bugs: do not display services even if Explorer running as admin

If it can't attach to process (already debugged) it shows error message, but after several tries, crashes.

Also all "Service" application crashes when I start them. I don't know if I need to start them at all.

I have Win 7 x64

re: Garbage Collection in The Age of Smart Pointers

Sasha Goldshtein — Fri, 20 Jan 2012 13:21:44 GMT

@POKE53280,0: You can always manage non-memory resources yourself, the same way you do with destructors in C++ (after all, you're responsible for calling the destructor). And there's the "using" statement. But I agree, there's room for improvement here.

@Jon: No, most garbage collection implementations do not use reference counting.

re: Garbage Collection in The Age of Smart Pointers

POKE53280,0 — Sun, 15 Jan 2012 00:00:11 GMT

The problem is that .NET GC can be OK for memory resources, but doesn't have a clue on non-memory resources (like textures, sockets, files...). Instead, C++ destructors (and smart pointers) can be used to handle every kind of resource uniformly, releasing it as soon as possible.

re: Windows 7 on 512 MB of RAM

Maisum — Sat, 14 Jan 2012 04:32:31 GMT

I am having pentium 4 with cpu 2.80 ghz and 2.79 ghz processor an 512 of ram so can i install windows 7 ....????

re: Garbage Collection in The Age of Smart Pointers

Guest — Sat, 14 Jan 2012 03:42:26 GMT

agreed, plus, reference counting doesn't do compacting, so it leads to fragmentation.