Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.lang.clipper.visual-objects > #4609 > unrolled thread

Thread-safe garbage collector

Started byserggio2011@gmail.com
First post2013-01-18 11:08 -0800
Last post2013-01-28 10:25 -0500
Articles 12 — 6 participants

Back to article view | Back to comp.lang.clipper.visual-objects


Contents

  Thread-safe garbage collector serggio2011@gmail.com - 2013-01-18 11:08 -0800
    Re: Thread-safe garbage collector "Valery N." <mustangsoftcom@gmail.com> - 2013-01-18 13:54 -0800
    Re: Thread-safe garbage collector "Arne Ortlinghaus" <Arne.Ortlinghaus@acs.it> - 2013-01-21 18:30 +0100
      Re: Thread-safe garbage collector serggio2011@gmail.com - 2013-01-24 07:57 -0800
        Re: Thread-safe garbage collector "Ginny Caughey" <gcaughey@online.hotmail.com> - 2013-01-25 09:48 -0500
        Re: Thread-safe garbage collector "Arne Ortlinghaus" <Arne.Ortlinghaus@acs.it> - 2013-01-25 18:40 +0100
          Re: Thread-safe garbage collector "Ginny Caughey" <gcaughey@online.hotmail.com> - 2013-01-26 07:48 -0500
        Re: Thread-safe garbage collector "Valery N." <mustangsoftcom@gmail.com> - 2013-01-27 07:34 -0800
        Re: Thread-safe garbage collector "Jamal" <vodotnet_nospam@yahoo.com> - 2013-01-25 15:23 -0500
          Re: Thread-safe garbage collector "Dirk (Belgium)" <dirk.dot.herijgers@pbprojects.dot.be> - 2013-01-28 08:50 +0000
            Re: Thread-safe garbage collector serggio2011@gmail.com - 2013-01-28 03:43 -0800
            Re: Thread-safe garbage collector "Jamal" <vodotnet_nospam@yahoo.com> - 2013-01-28 10:25 -0500

#4609 — Thread-safe garbage collector

Fromserggio2011@gmail.com
Date2013-01-18 11:08 -0800
SubjectThread-safe garbage collector
Message-ID<ca4c37d9-767f-4710-98cf-c8c8c9238e73@googlegroups.com>
Well, it had been widely and numerously discussed that GC is not thread-safe and one should avoid using dynamic objects when dealing with multiple threads.
And you know that's very-very uncomfortable :( . Sure, one may use OldSpace, some static "memalloc"-based constructions, etc. But that's not a way out in most cases. There's quite a common practice to involve multithreading as an extension to existing functionality, so preserving the majority of code is unevitable.
Are there any plans to make a THREAD-SAFE GC IN SP5 ?

Please don't say me about Vulcan. It's not so easy to move, when you have project (*.VOA + *.IDX files) sized 1.7Gb.

[toc] | [next] | [standalone]


#4612

From"Valery N." <mustangsoftcom@gmail.com>
Date2013-01-18 13:54 -0800
Message-ID<7c33037e-7fc7-4dfc-8448-8b8282939c9e@g6g2000vbk.googlegroups.com>
In reply to#4609
Your letter with an example I received. The example corrected. Errors
- No. I sent you the letter with a corrected example.

Good Luck!

Valery Novossiolov.

[toc] | [prev] | [next] | [standalone]


#4622

From"Arne Ortlinghaus" <Arne.Ortlinghaus@acs.it>
Date2013-01-21 18:30 +0100
Message-ID<kdju0j$n33$1@dont-email.me>
In reply to#4609
Hello Serggio,

I believe that it depends much on the classes/functions you want to use and 
it does not depend on the GC directly. GUI for example should never be used 
for Threads, only the main thread must have the right to communicate with 
the GUI. There are also limitations accessing databases (DBF, SQL), so it 
may depend more on these circumstances.

I have seen colleagues having trouble also with C# multi threading. So it is 
not only the programming language but understanding threads in general.

Arne
Ortlinghaus

<serggio2011@gmail.com> schrieb im Newsbeitrag 
news:ca4c37d9-767f-4710-98cf-c8c8c9238e73@googlegroups.com...
Well, it had been widely and numerously discussed that GC is not thread-safe 
and one should avoid using dynamic objects when dealing with multiple 
threads.
And you know that's very-very uncomfortable :( . Sure, one may use OldSpace, 
some static "memalloc"-based constructions, etc. But that's not a way out in 
most cases. There's quite a common practice to involve multithreading as an 
extension to existing functionality, so preserving the majority of code is 
unevitable.
Are there any plans to make a THREAD-SAFE GC IN SP5 ?

Please don't say me about Vulcan. It's not so easy to move, when you have 
project (*.VOA + *.IDX files) sized 1.7Gb. 

[toc] | [prev] | [next] | [standalone]


#4654

Fromserggio2011@gmail.com
Date2013-01-24 07:57 -0800
Message-ID<79dfed67-ce5b-43d7-8ffb-3307526790ce@googlegroups.com>
In reply to#4622
Arne, experience in thread-safe programming is a basic axiom. I do not even argue about it. I'm just saying that GC is not designed well enough to be thread-safe.
Let me cite some paragraphs from CAVO's help:

------------------------------
When multiple threads are running, each dynamic memory pointer becomes a shared resource. One thread might load a pointer to a string into a register. (The GC can only update memory locations, not registers.) Before actually performing an operation with that pointer, execution is transferred to another thread that triggers the GC. Now the string referred to by the register in the first thread is moved, and when execution returns to the first thread, the pointer in the register is not valid anymore.

The problem described above WILL BE RESOLVED IN A FUTURE RELEASE of Visual Objects, but right now the programmer has to be aware of it.
------------------------------

So I'm just asking, when that "future release" is going to be released ? :)

[toc] | [prev] | [next] | [standalone]


#4655

From"Ginny Caughey" <gcaughey@online.hotmail.com>
Date2013-01-25 09:48 -0500
Message-ID<51029b3b$0$63311$c3e8da3$3a1a2348@news.astraweb.com>
In reply to#4654
That sounds like a documentation error. In any case, and on any platform, 
you need to design your app for what is available today.

Ginny

wrote in message 
news:79dfed67-ce5b-43d7-8ffb-3307526790ce@googlegroups.com...

Arne, experience in thread-safe programming is a basic axiom. I do not even 
argue about it. I'm just saying that GC is not designed well enough to be 
thread-safe.
Let me cite some paragraphs from CAVO's help:

------------------------------
When multiple threads are running, each dynamic memory pointer becomes a 
shared resource. One thread might load a pointer to a string into a 
register. (The GC can only update memory locations, not registers.) Before 
actually performing an operation with that pointer, execution is transferred 
to another thread that triggers the GC. Now the string referred to by the 
register in the first thread is moved, and when execution returns to the 
first thread, the pointer in the register is not valid anymore.

The problem described above WILL BE RESOLVED IN A FUTURE RELEASE of Visual 
Objects, but right now the programmer has to be aware of it.
------------------------------

So I'm just asking, when that "future release" is going to be released ? :)

[toc] | [prev] | [next] | [standalone]


#4656

From"Arne Ortlinghaus" <Arne.Ortlinghaus@acs.it>
Date2013-01-25 18:40 +0100
Message-ID<kdug1k$sm$1@dont-email.me>
In reply to#4654
Hello Serggio,

I would say that if you use entercriticalsecion and leavecriticalsection (or 
similar) in your program parts then it should be also possible to make 
thread safe programs.  In any case it depends much on the ressorces used. 
The ODBC drivers I am using have an explicit setting "Multithreaded 
application" that must be set and where the manufactorer says that there is 
a "small" performance difference when setting the switch.

There is another possibility of "multithreading" I am using: Multiple exe 
files partly hidden that communicate via posting messages and database 
changes. There is the nice Windows message WM_COPYDATA with which also long 
commands and answers can be transferred between two applications. It is 
important to use postmessage and not sendmessage.

Arne Ortlinghaus

<serggio2011@gmail.com> schrieb im Newsbeitrag 
news:79dfed67-ce5b-43d7-8ffb-3307526790ce@googlegroups.com...
Arne, experience in thread-safe programming is a basic axiom. I do not even 
argue about it. I'm just saying that GC is not designed well enough to be 
thread-safe.
Let me cite some paragraphs from CAVO's help:

------------------------------
When multiple threads are running, each dynamic memory pointer becomes a 
shared resource. One thread might load a pointer to a string into a 
register. (The GC can only update memory locations, not registers.) Before 
actually performing an operation with that pointer, execution is transferred 
to another thread that triggers the GC. Now the string referred to by the 
register in the first thread is moved, and when execution returns to the 
first thread, the pointer in the register is not valid anymore.

The problem described above WILL BE RESOLVED IN A FUTURE RELEASE of Visual 
Objects, but right now the programmer has to be aware of it.
------------------------------

So I'm just asking, when that "future release" is going to be released ? :)

[toc] | [prev] | [next] | [standalone]


#4659

From"Ginny Caughey" <gcaughey@online.hotmail.com>
Date2013-01-26 07:48 -0500
Message-ID<5103d07d$0$63268$c3e8da3$3a1a2348@news.astraweb.com>
In reply to#4656
Great tips, Arne. I especially like the pseudo-multithreading using 
PostMessage.

Ginny

"Arne Ortlinghaus"  wrote in message news:kdug1k$sm$1@dont-email.me...

Hello Serggio,

I would say that if you use entercriticalsecion and leavecriticalsection (or
similar) in your program parts then it should be also possible to make
thread safe programs.  In any case it depends much on the ressorces used.
The ODBC drivers I am using have an explicit setting "Multithreaded
application" that must be set and where the manufactorer says that there is
a "small" performance difference when setting the switch.

There is another possibility of "multithreading" I am using: Multiple exe
files partly hidden that communicate via posting messages and database
changes. There is the nice Windows message WM_COPYDATA with which also long
commands and answers can be transferred between two applications. It is
important to use postmessage and not sendmessage.

Arne Ortlinghaus

<serggio2011@gmail.com> schrieb im Newsbeitrag
news:79dfed67-ce5b-43d7-8ffb-3307526790ce@googlegroups.com...
Arne, experience in thread-safe programming is a basic axiom. I do not even
argue about it. I'm just saying that GC is not designed well enough to be
thread-safe.
Let me cite some paragraphs from CAVO's help:

------------------------------
When multiple threads are running, each dynamic memory pointer becomes a
shared resource. One thread might load a pointer to a string into a
register. (The GC can only update memory locations, not registers.) Before
actually performing an operation with that pointer, execution is transferred
to another thread that triggers the GC. Now the string referred to by the
register in the first thread is moved, and when execution returns to the
first thread, the pointer in the register is not valid anymore.

The problem described above WILL BE RESOLVED IN A FUTURE RELEASE of Visual
Objects, but right now the programmer has to be aware of it.
------------------------------

So I'm just asking, when that "future release" is going to be released ? :)

[toc] | [prev] | [next] | [standalone]


#4663

From"Valery N." <mustangsoftcom@gmail.com>
Date2013-01-27 07:34 -0800
Message-ID<c5424a79-1341-4551-aee2-959d83cf6a53@v7g2000yqv.googlegroups.com>
In reply to#4654
I sent the letter with one more way of the solution of a task.

[toc] | [prev] | [next] | [standalone]


#4665

From"Jamal" <vodotnet_nospam@yahoo.com>
Date2013-01-25 15:23 -0500
Message-ID<ke4ul6$k4g$2@dont-email.me>
In reply to#4654
Yet in another part of the VO help it says:

"Note: The kernel runtime support (garbage collector) has been made 
thread-safe in this version of Visual Objects, easing the development of 
multithreaded applications.  Thread support is a basic requisite for 
building server components, such as ISAPI extensions or components running 
in an application server framework.  For an example of multithreading, see 
the Multithreaded Standard MDI Application, which is accessible from the 
Samples tab page of the Application Gallery."

Jamal

wrote in message 
news:79dfed67-ce5b-43d7-8ffb-3307526790ce@googlegroups.com...

Arne, experience in thread-safe programming is a basic axiom. I do not even 
argue about it. I'm just saying that GC is not designed well enough to be 
thread-safe.
Let me cite some paragraphs from CAVO's help:

------------------------------
When multiple threads are running, each dynamic memory pointer becomes a 
shared resource. One thread might load a pointer to a string into a 
register. (The GC can only update memory locations, not registers.) Before 
actually performing an operation with that pointer, execution is transferred 
to another thread that triggers the GC. Now the string referred to by the 
register in the first thread is moved, and when execution returns to the 
first thread, the pointer in the register is not valid anymore.

The problem described above WILL BE RESOLVED IN A FUTURE RELEASE of Visual 
Objects, but right now the programmer has to be aware of it.
------------------------------

So I'm just asking, when that "future release" is going to be released ? :)

[toc] | [prev] | [next] | [standalone]


#4670

From"Dirk (Belgium)" <dirk.dot.herijgers@pbprojects.dot.be>
Date2013-01-28 08:50 +0000
Message-ID<ke5e4k$toq$1@dont-email.me>
In reply to#4665
Jamal,

In which version of the help file is this written?
VO 2.5 ?
VO 2.6 ?
VO 2.7 ?
VO 2.8 ?

Dirk

================


Jamal wrote:

> Yet in another part of the VO help it says:
> 
> "Note: The kernel runtime support (garbage collector) has been made
> thread-safe in this version of Visual Objects, easing the development
> of multithreaded applications.  Thread support is a basic requisite
> for building server components, such as ISAPI extensions or
> components running in an application server framework.  For an
> example of multithreading, see the Multithreaded Standard MDI
> Application, which is accessible from the Samples tab page of the
> Application Gallery."
> 
> Jamal
> 
> wrote in message
> news:79dfed67-ce5b-43d7-8ffb-3307526790ce@googlegroups.com...
> 
> Arne, experience in thread-safe programming is a basic axiom. I do
> not even argue about it. I'm just saying that GC is not designed well
> enough to be thread-safe.  Let me cite some paragraphs from CAVO's
> help:
> 
> ------------------------------
> When multiple threads are running, each dynamic memory pointer
> becomes a shared resource. One thread might load a pointer to a
> string into a register. (The GC can only update memory locations, not
> registers.) Before actually performing an operation with that
> pointer, execution is transferred to another thread that triggers the
> GC. Now the string referred to by the register in the first thread is
> moved, and when execution returns to the first thread, the pointer in
> the register is not valid anymore.
> 
> The problem described above WILL BE RESOLVED IN A FUTURE RELEASE of
> Visual Objects, but right now the programmer has to be aware of it.
> ------------------------------
> 
> So I'm just asking, when that "future release" is going to be
> released ? :)



-- 

[toc] | [prev] | [next] | [standalone]


#4672

Fromserggio2011@gmail.com
Date2013-01-28 03:43 -0800
Message-ID<29ab9835-156b-4494-b0b0-714dec2929b6@googlegroups.com>
In reply to#4670
Dirk, that's written in the help for VO 2.8. And that's a rather optimistic statement and in fact it is just untrue.
There's a sample multithreaded MDI application in the "samples" directory "\Samples\WinSDK\MThread\Mtstdapp.aef" intended to justify that statement. And it is really quite stable when there are several threads but...  hangs or falls when there are 10-15 threads. Perhaps it's just the time to call gc :)

[toc] | [prev] | [next] | [standalone]


#4673

From"Jamal" <vodotnet_nospam@yahoo.com>
Date2013-01-28 10:25 -0500
Message-ID<ke6592$5ik$1@dont-email.me>
In reply to#4670
Dirk,

It am using 2.8 (sp4) but I don't know what version exactly that statement 
was added in.

Jamal



"Dirk (Belgium)"  wrote in message news:ke5e4k$toq$1@dont-email.me...

Jamal,

In which version of the help file is this written?
VO 2.5 ?
VO 2.6 ?
VO 2.7 ?
VO 2.8 ?

Dirk

================


Jamal wrote:

> Yet in another part of the VO help it says:
>
> "Note: The kernel runtime support (garbage collector) has been made
> thread-safe in this version of Visual Objects, easing the development
> of multithreaded applications.  Thread support is a basic requisite
> for building server components, such as ISAPI extensions or
> components running in an application server framework.  For an
> example of multithreading, see the Multithreaded Standard MDI
> Application, which is accessible from the Samples tab page of the
> Application Gallery."
>
> Jamal
>
> wrote in message
> news:79dfed67-ce5b-43d7-8ffb-3307526790ce@googlegroups.com...
>
> Arne, experience in thread-safe programming is a basic axiom. I do
> not even argue about it. I'm just saying that GC is not designed well
> enough to be thread-safe.  Let me cite some paragraphs from CAVO's
> help:
>
> ------------------------------
> When multiple threads are running, each dynamic memory pointer
> becomes a shared resource. One thread might load a pointer to a
> string into a register. (The GC can only update memory locations, not
> registers.) Before actually performing an operation with that
> pointer, execution is transferred to another thread that triggers the
> GC. Now the string referred to by the register in the first thread is
> moved, and when execution returns to the first thread, the pointer in
> the register is not valid anymore.
>
> The problem described above WILL BE RESOLVED IN A FUTURE RELEASE of
> Visual Objects, but right now the programmer has to be aware of it.
> ------------------------------
>
> So I'm just asking, when that "future release" is going to be
> released ? :)



--  

[toc] | [prev] | [standalone]


Back to top | Article view | comp.lang.clipper.visual-objects


csiph-web