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

Groups > comp.arch > #21387 > unrolled thread

More about waitfree and lockfree

Back to article view | Back to comp.arch

Contents

  More about waitfree and lockfree aminer <aminer@toto.net> - 2014-04-22 20:21 -0700
    Re: More about waitfree and lockfree MitchAlsup <MitchAlsup@aol.com> - 2014-04-22 18:07 -0700
      Re: More about waitfree and lockfree "Chris M. Thomasson" <no@spam.invalid> - 2014-04-23 12:52 -0700
        Re: More about waitfree and lockfree aminer <aminer@toto.net> - 2014-04-23 15:58 -0700
        Re: More about waitfree and lockfree aminer <aminer@toto.net> - 2014-04-23 16:08 -0700
        Re: More about waitfree and lockfree aminer <aminer@toto.net> - 2014-04-23 16:19 -0700
          Re: More about waitfree and lockfree aminer <aminer@toto.net> - 2014-04-23 16:32 -0700
            Re: More about waitfree and lockfree aminer <aminer@toto.net> - 2014-04-23 16:41 -0700
      Re: More about waitfree and lockfree Kaz Kylheku <kaz@kylheku.com> - 2014-04-23 23:13 +0000

#21387 — More about waitfree and lockfree

Hello...


Question:

Amine , we are kind of lost, what is this waitfree
concurrent FIFO queue , and lockfree concurrent FIFO queue ?


Answer:

We have to be smart and understand that waitfree and lockfree
algorithms maximize the parallelism, but the lockfree
suffers from starvation and it's less efficient than
other algorithms whith a scenario of high contention and the lockfree
algorithm can not be used in realtime systems when you need
the process to complete any operation in a finite number of steps of 
finite time, the lockfree algorithm can not do that, and what about
waitfree ? as i have said the waitfree algorithm do maximize the 
parallelism and it is also starvation-free and it finishes
an operatoin such us the push() and a pop() of a concurrent FIFO queue
in a finite number of steps , and the waitfree is efficient in a 
scenario with high contention , so i have presented to you a
concurrent FIFO queue that is waitfree, but the number of threads
has to be fixed to calculate the number of steps that the push()
and pop() takes in a realtime time system.


I have also presented to you a faster concurrent FIFO queue
that is not waitfree but as i said before it has more parallelism than 
the two lock algorithm and the items will be pushed and popped in
a FIFO order and it scales well even if the number of threads are 
greater than the number of cores.

You can download all those concurrent FIFO queues from my website:

http://pages.videotron.com/aminer/




Thank you,
Amine Moulay Ramdane.

[toc] | [next] | [standalone]

#21388

On Tuesday, April 22, 2014 7:20:54 PM UTC-5, aminer wrote:
> Hello...

I have a suggestion for you.

Continue to post, but never change the subject line again. Just add and add and
add posts to the current open post. Even if you come out with a new version
(like in 3 hours or so), just leave it all in one thread. All of it. 

Did you even notice that everyone else in this newsgroup happens to maintain
threads and adds to them as they feel fit. But you seem to spew new posts
over a wide berth with virtually no changes.

I understand you feel justifiably proud of your achievements, but cut the SPAM,
please.

By the way, have you noticed how few reponses you are getting?

Mitch

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

#21397

> "MitchAlsup"  wrote in message 
> news:cd1edc26-da1f-40e1-ad24-7979e621e935@googlegroups.com...

> On Tuesday, April 22, 2014 7:20:54 PM UTC-5, aminer wrote:
> > Hello...

> I have a suggestion for you.
> Continue to post, but never change the subject line again.

nice and succinct Mitch!

:^D      Thanks.



Well, IMVHO, animer seems to be fond of attributing my
name to a point were it can be confused with me actually
working with him. This makes me want to   ASK  him to
_refrain_ wrt creating  content that seems to suggest that
we have ever worked together.

:^O

Never happened.



My algorithms are my own work. I have never worked with "animer".

Blah!

:^( 

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

#21398

On 4/23/2014 12:52 PM, Chris M. Thomasson wrote:
>> "MitchAlsup"  wrote in message
>> news:cd1edc26-da1f-40e1-ad24-7979e621e935@googlegroups.com...
>
>> On Tuesday, April 22, 2014 7:20:54 PM UTC-5, aminer wrote:
>> > Hello...
>
>> I have a suggestion for you.
>> Continue to post, but never change the subject line again.
>
> nice and succinct Mitch!
>
> :^D      Thanks.
>
>
>
> Well, IMVHO, animer seems to be fond of attributing my
> name to a point were it can be confused with me actually
> working with him. This makes me want to   ASK  him to
> _refrain_ wrt creating  content that seems to suggest that
> we have ever worked together.
>
> :^O
>
> Never happened.
>
>
>
> My algorithms are my own work. I have never worked with "animer".
>
> Blah!
>
> :^(



My dear Chriss Thomasson, your algorithm is your own work that's a true, 
but what i have tried to do is make your algorithm available to
the Object pascal community, you have given me this algorithm
a long time ago, and if you have noticed i have put your name
as the author of the algorithm and i have put my name as the author
of the Object pascal source code and as the author of SemaMonitor,
so i don't think there is a problem. What do you think ?


Thank you,
Amine Moulay Ramdane.

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

#21399

Hello,

And here is the algorithm that you have given me in pascal
source code a long time ago:

Please read this:

https://groups.google.com/forum/#!searchin/comp.programming.threads/Amine$20Moulay$20Ramdane$20and$20bakery/comp.programming.threads/U4-K-a4iWKs/jqZH1plUQMMJ


So what i have done is wrote it in Object pascal source code
and i have added to it my SemaMonitor, and i have wrote
your name as the author of the algorithm and my name as
the author of the Object pascal source code and the author of 
SemaMonitor. Is there any problem ?

My dear Chriss Thomasson, your algorithm is your own work that's  true, 
but what i have tried to do is make your algorithm available to
the Object pascal community, you have given me this algorithm
a long time ago.


Thank you,
Amine Moulay Ramdane.

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

#21401

Please read all the following:

https://groups.google.com/forum/#!searchin/comp.programming.threads/Amine$20Moulay$20Ramdane$20and$20bakery/comp.programming.threads/U4-K-a4iWKs/jqZH1plUQMMJ 


My dear Chriss, as you have noticed , you have given me this algorithm 
in a pascal source code a long time ago , and you have told
me in replyin to my post that every element in the array must be of a 
size of a cache-line and the array must be 64 bytes aligned
i have done all that and i have rewrote it in Object pascal
not pascal and i have added my SemaMonitor.. and i have wrote
your name as the author of the algorithm and my name as
the author of the Object pascal source code and the author
of SemaMonitor, after that i have made the source code available
to the Object pascal community. Is there any problem Chriss ?



Thank you,
Amine Moulay Ramdane.

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

#21403

Chriss, after you have read all my previous posts...
if there is any problem, and you told me to delete it
from my website, i will delete it.

If you told me that's ok , i will keep it.


What do you think Chriss ?



Thank you,
Amine Moulay Ramdane.


On 4/23/2014 4:19 PM, aminer wrote:
>
>
>
> Please read all the following:
>
> https://groups.google.com/forum/#!searchin/comp.programming.threads/Amine$20Moulay$20Ramdane$20and$20bakery/comp.programming.threads/U4-K-a4iWKs/jqZH1plUQMMJ
>
>
> My dear Chriss, as you have noticed , you have given me this algorithm
> in a pascal source code a long time ago , and you have told
> me in replyin to my post that every element in the array must be of a
> size of a cache-line and the array must be 64 bytes aligned
> i have done all that and i have rewrote it in Object pascal
> not pascal and i have added my SemaMonitor.. and i have wrote
> your name as the author of the algorithm and my name as
> the author of the Object pascal source code and the author
> of SemaMonitor, after that i have made the source code available
> to the Object pascal community. Is there any problem Chriss ?
>
>
>
> Thank you,
> Amine Moulay Ramdane.
>
>
>

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

#21404

Hello Chriss Thomasson

After i have read your posts, i think that you don't
want me to make your algorithm available to the Object
pascal community. So i will delete it right now from my website.

And i will stay with my other concurrent FIFO queue
that is lockfree on the pop() side... it has scored a decent
throughput and that's ok for me.


Thank you,
Amine Moulay Ramdane.

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

#21410

On 2014-04-23, MitchAlsup <MitchAlsup@aol.com> wrote:
> On Tuesday, April 22, 2014 7:20:54 PM UTC-5, aminer wrote:
>> Hello...
>
> I have a suggestion for you.
>
> Continue to post, but never change the subject line again. Just add and add and
> add posts to the current open post. Even if you come out with a new version
> (like in 3 hours or so), just leave it all in one thread. All of it. 
>
> Did you even notice that everyone else in this newsgroup happens to maintain
> threads and adds to them as they feel fit. But you seem to spew new posts
> over a wide berth with virtually no changes.
>
> I understand you feel justifiably proud of your achievements, but cut the SPAM,
> please.
>
> By the way, have you noticed how few reponses you are getting?

aminer's ideas are not coherently presented in a structure which takes a low
amount of effort to understand them.

The incremental "status reports" make little sense to anyone who has not
been eagerly been following the progress in detail from the beginning.

The work appears to consists of empirical experimentation with a changing
reference implementation of some locks, in combination with some tests, on
particular hardware configuration.  There is no compelling reason to believe
that the results have any general value that translates to the real world.

Distributing .zip files is an outdated method.

The experimentation could easily be summarized in the commit comments in a git
repository. Anyone who cares about the progress could dig through the history,
whose checking comments explain what was changed, what was the intent of those
changes, and how they actually impacted the benchmarks.

Over the course of several months, the author could dig through this history,
and write a concise paper summarizing the empirical findings.

[toc] | [prev] | [standalone]

Back to top | Article view | comp.arch

csiph-web