Read again, i correct a last typo

[Horizon68]

Hello...


Read again, i correct a last typo:

More explanation about scalability..

As you have noticed i said below the following:

"The scalability of my Parallel Sort Library in the mergesort mode will
go to around 12X"

You have to distinguish between the scalability that is the combination
of scalability of hardware and software, and the scalability that is
only software or only hardware.

So you have to know that i am speaking below about the "total"
scalability of both the hardware(from one channel memory to 8 channels
memory) and software on actual Intel and AMD processors, but 3D stacking
of memory will give much more scalability.


Read again:

About Multi-channel memory architecture and my Parallel Sort Library..

On actual Intel and AMD processors data is moved from and to main memory
only in 64 bytes chunks, so in a 8 channels of a Multi-channel memory
architecture in a ganged mode, i think that the memory speed of
my Parallel Sort library in the mergesort mode will go to 8X speed,
and the actual scalability of my Parallel Sort Library in the mergesort
mode will go to around 12X. So i will advice you to use the ganged mode
of the Multi-channel memory architecture because it works great.

My Parallel Sort Library was updated to version 3.64

I have enhanced it more, and i think it is stable and fast
and it scales more.

You can download it from:

https://sites.google.com/site/scalable68/parallel-sort-library


Thank you,
Amine Moulay Ramdane.

[Horizon68]

Hello..

Read again, i correct a last typo

Here is my new invention..

I have just invented a new scalable lock that is "better" than
my following scalable lock:

https://sites.google.com/site/scalable68/scalable-mlock


I think i will sell my new scalable lock to Embarcadero software company
or such software companies.

My new scalable lock is a node based Lock that is scalable, FIFO fair
and starvation-free.

- Discovered by Amine Moulay Ramdane

- This lock is scalable

- It has the same space requirement as the scalable MCS lock

- Doesn't require a local "queue node" to be passed in as a parameter as
is doing the MCS and CLH locks.

- Spins only on local locations on a cache-coherent machine

- And it's fast.


So you have to know that my new scalable lock doesn't require a local
"queue node" to be passed in as a parameter as is doing the MCS and CLH
locks, my scalable MLock doesn't require any parameter to be passed,
just call the Enter() and Leave() methods and that's all.