«

Hi

I would like some ideas on way to solve the concurrency issue. The
problem is , I have an object say X which multiple readers need to
access and also to update. They do so by say two functions :
// simplistic view of the problem
Read()
{
return X;
}

Update(Y)
{
X= Y;
}

Now I would like to block any reader while calling update. With the
general solutions provided of having say a mutex and using it in Read
and Update would solve the problem. But then this would block multiple
readers also which is not desired. i.e a block on read should be
allowed only when update is called. Having a boolean and checkin is
not ideal as the atomicity of the operation is not guaranteed. Any
ideas on this?

On 21 Feb, 11:08, "Hunk" <santosh.udyav...@gmail.com> wrote:
> Hi
>
> I would like some ideas on way to solve the concurrency issue. The
> problem is , I have an object say X which multiple readers need to
> access and also to update. They do so by say two functions :
> // simplistic view of the problem
> Read()
> {
> return X;
>
> }
>
> Update(Y)
> {
> X= Y;
>
> }
>
> Now I would like to block any reader while calling update. With the
> general solutions provided of having say a mutex and using it in Read
> and Update would solve the problem. But then this would block multiple
> readers also which is not desired. i.e a block on read should be
> allowed only when update is called. Having a boolean and checkin is
> not ideal as the atomicity of the operation is not guaranteed. Any
> ideas on this?

you nead reader-writer lock. the best implementation I've
seen is Valery Pryamikov's mrsw_guard

On Feb 21, 3:08 am, "Hunk" <santosh.udyav...@gmail.com> wrote:
> Hi
>
> I would like some ideas on way to solve the concurrency issue. The
> problem is , I have an object say X which multiple readers need to
> access and also to update. They do so by say two functions :
> // simplistic view of the problem
> Read()
> {
> return X;
>
> }
>
> Update(Y)
> {
> X= Y;
>
> }
>
> Now I would like to block any reader while calling update. With the
> general solutions provided of having say a mutex and using it in Read
> and Update would solve the problem. But then this would block multiple
> readers also which is not desired. i.e a block on read should be
> allowed only when update is called. Having a boolean and checkin is
> not ideal as the atomicity of the operation is not guaranteed. Any
> ideas on this?

Is this a multi-threaded program? If it is, the best solution is to
use read-write locks in pthread library. See the manpage of
pthread_rwlock_wrlock and pthread_rwlock_rdlock. Note that you need to
have -DUSE_UNIX98 in your compiler flags.

On Feb 21, 7:55 pm, "Bharath" <cbku...@gmail.com> wrote:
> On Feb 21, 3:08 am, "Hunk" <santosh.udyav...@gmail.com> wrote:
>
>
>
>
>
> > Hi
>
> > I would like some ideas on way to solve the concurrency issue. The
> > problem is , I have an object say X which multiple readers need to
> > access and also to update. They do so by say two functions :
> > // simplistic view of the problem
> > Read()
> > {
> > return X;
>
> > }
>
> > Update(Y)
> > {
> > X= Y;
>
> > }
>
> > Now I would like to block any reader while calling update. With the
> > general solutions provided of having say a mutex and using it in Read
> > and Update would solve the problem. But then this would block multiple
> > readers also which is not desired. i.e a block on read should be
> > allowed only when update is called. Having a boolean and checkin is
> > not ideal as the atomicity of the operation is not guaranteed. Any
> > ideas on this?
>
> Is this a multi-threaded program? If it is, the best solution is to
> use read-write locks in pthread library. See the manpage of
> pthread_rwlock_wrlock and pthread_rwlock_rdlock. Note that you need to
> have -DUSE_UNIX98 in your compiler flags.- Hide quoted text -
>
> - Show quoted text -

Thanks... yes it is a multithreaded program. i'll take a look at the
reader writer locks mentioned. Any idea on how this is implemented in
zthreads?

On Feb 21, 9:59 pm, "Hunk" <santosh.udyav...@gmail.com> wrote:
> On Feb 21, 7:55 pm, "Bharath" <cbku...@gmail.com> wrote:
>
>
>
>
>
> > On Feb 21, 3:08 am, "Hunk" <santosh.udyav...@gmail.com> wrote:
>
> > > Hi
>
> > > I would like some ideas on way to solve the concurrency issue. The
> > > problem is , I have an object say X which multiple readers need to
> > > access and also to update. They do so by say two functions :
> > > // simplistic view of the problem
> > > Read()
> > > {
> > > return X;
>
> > > }
>
> > > Update(Y)
> > > {
> > > X= Y;
>
> > > }
>
> > > Now I would like to block any reader while calling update. With the
> > > general solutions provided of having say a mutex and using it in Read
> > > and Update would solve the problem. But then this would block multiple
> > > readers also which is not desired. i.e a block on read should be
> > > allowed only when update is called. Having a boolean and checkin is
> > > not ideal as the atomicity of the operation is not guaranteed. Any
> > > ideas on this?
>
> > Is this a multi-threaded program? If it is, the best solution is to
> > use read-write locks in pthread library. See the manpage of
> > pthread_rwlock_wrlock and pthread_rwlock_rdlock. Note that you need to
> > have -DUSE_UNIX98 in your compiler flags.- Hide quoted text -
>
> > - Show quoted text -
>
> Thanks... yes it is a multithreaded program. i'll take a look at the
> reader writer locks mentioned. Any idea on how this is implemented in
> zthreads?- Hide quoted text -
>
> - Show quoted text -

I never used zthreads but this link might help.
http://zthread.sourceforge.net/html/...WriteLock.html

"dasjotre" <> wrote in message
news: ups.com...
> On 21 Feb, 11:08, "Hunk" <santosh.udyav...@gmail.com> wrote:
[...]
> you nead reader-writer lock. the best implementation I've
> seen is Valery Pryamikov's mrsw_guard

how does it compare to the following algorithm:

DOH! sorry for the other msg... I left out the link!

"dasjotre" <> wrote in message
news: ups.com...
> On 21 Feb, 11:08, "Hunk" <santosh.udyav...@gmail.com> wrote:
[...]
> you nead reader-writer lock. the best implementation I've
> seen is Valery Pryamikov's mrsw_guard

how does it compare to the following algorithm:

http://groups.google.com/group/comp....fb0ed46a4bcb4b

"Hunk" <> wrote in message
news: ups.com...
> Hi
>
> I would like some ideas on way to solve the concurrency issue.

Sure:

http://groups.google.com/group/comp....6ad37c34dd673a

http://groups.google.com/group/comp....fb0ed46a4bcb4b

http://appcore.home.comcast.net/


How about we transfer this thread over to comp.programming.threads before we
go any further?

A while back I was thinking about fully integrating a working solution to
the reader-writer problem directly into the C++ STL:

http://groups.google.com/group/comp....c9c2673682d4cf

"Chris Thomasson" <> wrote in message
news. ..
> DOH! sorry for the other msg... I left out the link!
>
> "dasjotre" <> wrote in message
> news: ups.com...
>> On 21 Feb, 11:08, "Hunk" <santosh.udyav...@gmail.com> wrote:
> [...]
>> you nead reader-writer lock. the best implementation I've
>> seen is Valery Pryamikov's mrsw_guard
>
> how does it compare to the following algorithm:
>
> http://groups.google.com/group/comp....fb0ed46a4bcb4b

Or this rw-spinlock implementation:

http://appcore.home.comcast.net/appc...k_algo1_c.html

http://appcore.home.comcast.net/appc...k_algo1_h.html