Hello...


Universal scalability Law (USL) methodology is really amazing !

Here is my own and more intuitive mathematical proof using probability
that explains the USL methodology...

As state the Amdahl's law, a parallel program can be constituted of
serial part and a parallel part, that we call S and P.

Now since in the overall application the S (serial part) is variable ,
it can be smaller or bigger, when you will do the benchmark
of your parallel program with fewer number of cores and threads, let say
up to 8 cores and 8 threads, so if the serial part is bigger so you will
get more chance to get contention and this will allow
the USL methodology to get a good approximation, but the chance on the
overall parallel application to get a bigger serial part that get
contention with 8 cores and 8 threads is small, so since you will have
less chance to get contention on 8 cores, so the USL methodology with
fewer cores and fewer threads will constitute a better approximation of
the predicted scalability and will allow the USL methodology to
gerecat scalability farther, so this is amazing and successful !
and i say USL methodology is an amazing and succefull tool.


Hope you have understood my proof.

So i think that USL methodology is a great tool.

You can download my USL programs version 3.0 with the source code from:

https://sites.google.com/site/aminer68/universal-scalability-law-for-delphi-and-freepascal


Thank you,
Amine Moulay Ramdane.

Thank you,
Amine Moulay Ramdane.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)