Sunday, April 17, 2016

Digest for comp.programming.threads@googlegroups.com - 25 updates in 10 topics

Ramine <ramine@1.1>: Apr 17 09:12PM -0700

Hello,
 
I think that we have also to prove by contradiction...
 
If the serial part of the Amdahl's law is 1/16 the overall
parallel program, so when you will test with USL with
fewer cores and fewer threads, let say 8, you can
escape contention on the serial part, and this will
make the linear regression of USL to fail to predict
in this particular case, so since it fails on this
particular case , so this makes the USL methodology
to fail, it is like a proof by contradiction in
mathematics.
 
So from this, how can we believe this USL methodology ?
i think that Dr. Gunther the author of USL methodology
must explain to us what is this magical thing that makes
his USL methodology succeed.
 
 
Thank you,
Amine Moulay Ramdane.
bleachbot <bleachbot@httrack.com>: Apr 17 07:36PM +0200

bleachbot <bleachbot@httrack.com>: Apr 17 08:12PM +0200

bleachbot <bleachbot@httrack.com>: Apr 17 08:15PM +0200

bleachbot <bleachbot@httrack.com>: Apr 17 08:33PM +0200

bleachbot <bleachbot@httrack.com>: Apr 17 08:39PM +0200

bleachbot <bleachbot@httrack.com>: Apr 17 08:46PM +0200

bleachbot <bleachbot@httrack.com>: Apr 17 10:02PM +0200

bleachbot <bleachbot@httrack.com>: Apr 18 12:17AM +0200

bleachbot <bleachbot@httrack.com>: Apr 18 01:04AM +0200

bleachbot <bleachbot@httrack.com>: Apr 18 01:26AM +0200

bleachbot <bleachbot@httrack.com>: Apr 18 02:27AM +0200

bleachbot <bleachbot@httrack.com>: Apr 18 02:28AM +0200

bleachbot <bleachbot@httrack.com>: Apr 18 03:11AM +0200

Ramine <ramine@1.1>: Apr 17 08:30PM -0700

Hello,
 
USL methodology is like a playing a Dice game with much greater
number of sides than 6..
 
When you want to test with fewer cores and fewer threads using
the USL methodology to predict scalability, the bigger serial parts
of the Amdahl's law that causes contention and the smallest serial parts
of the Amdahl's law that are much farther from the the smallest big
serial part that causes contention constitutes the greater majorities of
the sides of a Dice with much greater
number of sides than 6.., so there probability are , much higher, so
since the bigger serial parts that causes contention on the serial part
makes the nonlinear regression will succeed in predicting scalability..
and since for the smallest serial parts of the Amdahl's law that are far
from the smallest big parts that causes contention have a higher chance
to happen than the smallest serial parts that are near the smallest
bigger parts that causes contention, this is why when you test with
fewer cores and fewer threads with USL methodology there is a much
higher chance that the forecasting of scalability farther succeeds, that
means it's a better approximation.
 
 
This is all about mathematical probability, and my reasonning makes
USL methodology a successful and great tool that predict scalability.
 
 
 
Thank you,
Amine Moulay Ramdane.
Ramine <ramine@1.1>: Apr 17 08:28PM -0700

Hello....
 
USL methodology is like a playing a Dice game with much greater
sides than 6..
 
When you want to test with fewer cores and fewer threads using
the USL methodology to predict scalability, the bigger serial parts
of the Amdahl's law that causes contention and the smallest serial parts
of the Amdahl's law that are much farther from the the smallest big
serial part that causes contention constitutes the greater majorities of
the sides of a Dice with much greater
sides than 6.., so there probability are , much higher, so since the
bigger serial parts that causes contention on the serial part
makes the nonlinear regression will succeed in predicting scalability..
and since for the smallest serial parts of the Amdahl's law that are far
from the smallest big parts that causes contention have a higher chance
to happen than the smallest serial parts that are near the smallest
bigger parts that causes contention, this is why when you test with
fewer cores and fewer threads with USL methodology there is a much
higher chance that the forecasting of scalability farther succeeds, that
means it's a better approximation.
 
 
This is all about mathematical probability, and my reasonning makes
USL methodology a successful and great tool that predict scalability.
 
 
 
Thank you,
Amine Moulay Ramdane.
Ramine <ramine@1.1>: Apr 17 07:29PM -0700

Hello.....
 
I will make my proof more precise...
 
For bigger serial parts of the Amdahl's law my reasonning is correct and
here it is:
 
If the serial part of the Amdahl's law is bigger, you have more
chance probabilistically to get contention on the serial part
with fewer threads and fewer cores with the USL methodology,
and this contention will enable the nonlinear regression to
approximate more the predicted scalability, because this
mathematical fact will deviate the graph of the nonlinear
regression in a more right direction up to a farther predicted
scalability, so this enable the nonlinear regression to predict
scalability farther, so this reasonning will make the USL methodology to
succeed on a more bigger serial parts of the Amdahl's law.
 
But for smaller serial parts of the Amdahl's law, i will make my
proof more precise by example..
 
On smaller serial parts of the Amdahl's law, let's imagine
the serial part is 1/8 the overall parallel program, so this
1/8 has a probability of happening empirically on the overall
number of parallel applications, and this 1/8 can cause contention
on fewer cores and fewer threads using the USL methodology, but this 1/8
is not the only one in the distribution of probability, because we have
other that are 1/9 and 1/10 and 1/11 that has a
probability to happen empirically... and my reasonning
by mathematical probability and by making a better approximation
will make us affirm that there is a higher chance to get a more smaller
serial parts than 1/8 , so this is all about mathematical probability,
and my reasonning with mathematical probability that i am making makes
the nonlinear regression of the USL methodology to succeed in predicting
much farther.
 
Hence from the above proof, the bigger parts of the Amdahl's law
that causes more contention with fewer threads an fewer cores using
the USL methodology and the much smaller parts of the Amdahl's law have
a much higher chance to happen than the rest, so since they have a
higher chance to happen, that means mathematically that USL methodology
is successful in predicting scalability and that means that it is a
really a great tool.
 
 
I have included the 32 bit and 64 bit windows executables of my
programs inside the zip file to easy the job for you.
 
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.
Ramine <ramine@1.1>: Apr 17 07:07PM -0700

Hello.....
 
I will make my proof more precise...
 
For bigger serial parts of the Amdahl's law my reasonning is correct and
here it is:
 
If the serial part of the Amdahl's law is bigger, you have more
chance probabilistically to get contention on the serial part
with fewer threads and fewer cores with the USL methodology,
and this contention will enable the nonlinear regression to
approximate more the predicted scalability, because this
mathematical fact will deviate the graph of the nonlinear
regression in a more right direction up to a farther predicted
scalability, so this enable the nonlinear regression to predict
scalability farther, so this reasonning will make the USL methodology to
succeed on a more bigger serial parts of the Amdahl's law.
 
But for smaller serial parts of the Amdahl's law, i will make my
proof more precise by example..
 
On smaller serial parts of the Amdahl's law, let's imagine
the serial part is 1/8 the overall parallel program, so this
1/8 has a probability of happening empirically on the overall
number of parallel applications, and this 1/8 can cause contention
on fewer cores and fewer threads using the USL methodology, but this 1/8
is not the only one in the distribution of probability, because we have
other that are 1/9 and 1/10 and 1/11 that has a
probability to happen empirically... and my reasonning
by mathematical probability and by making a better approximation
will make us affirm that there is a higher chance to get a more smaller
serial parts than 1/8 , so this is all about mathematical probability,
and my reasonning with mathematical probability that i am making makes
the nonlinear regression of the USL methodology to succeed in predicting
much farther.
 
Hence from the above proof, the bigger parts of the Amdahl's law
that causes more contention with fewer threads an fewer cores using
the USL methodology and the much smaller parts of the Amdahl's law has a
much higher chance to happen than the rest, so since they have a higher
chance to happen that means mathematically that USL methodology is
successful in predicting scalability and that means that it is a really
a great tool.
 
 
I have included the 32 bit and 64 bit windows executables of my
programs inside the zip file to easy the job for you.
 
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.
Ramine <ramine@1.1>: Apr 17 06:19PM -0700

Hello....
 
Because Dr. Gunther he author of USL has not explain why
regression analyses works in his methodology, so i will now give my
final mathematical proof...
 
Here is my final mathematical proof that i will explain...
 
First you know that Amdahl's law predict scalability using
the serial part S and the parallel part P of a parallel program..
 
Now i will continu my proof with the Amdahl's law, making some
good approximation by simplifying a little bit the model..
 
Now here is my proof:
 
If the serial part of the Amdahl's law is bigger, you have more
chance probabilistically to get contention on the serial part,
and this contention will enable the nonlinear regression to
approximate more the predicted scalability, because this
mathematical fact will deviate the graph of the nonlinear
regression in a more right direction up to a farther predicted
scalability, so this enable the nonlinear regression to predict
scalability farther, so this reasonning will make the USL methodology to
succeed on a more bigger serial parts of the Amdahl's law.
 
For smaller serial parts, if the serial part is smaller ,
you have less chance probabilistically to get contention on
the serial part, and this mathematical fact will enable
the nonlinear regression to predict farther scalability with
fewer threads and fewer cores.
 
So this two mathematical facts makes the mathematical probability
distribution of the success of the forecasting of scalability farther
higher, so this is all about mathematical probability, and this
mathematical probability of my proof makes the USL methodology
successful and enable the USL methodology to forecast farther.
 
 
Thank you,
Amine Moulay Ramdane.
Ramine <ramine@1.1>: Apr 17 04:04PM -0700

Hello.....
 
 
How to validate my USL programs to be sure that they work correctly ?
 
I have first tested my USL programs with polynomial regression
against the R package of USL with the default solver with the raytracer
performance data of the R package and they are giving the same results
that is the peak number of processors at 449 and the same predicted
scalability, but my nonlinear solver that uses the simplex method as
a function minimization is giving a very good approximation
of the predicted scalability, i have also tested with other performance
data from my parallel LZMA algorithm and parallel LZ4 algorithm
of my parallel compression library and the R package is giving
the same results as my USL program solvers.
 
So you can be confident with my USL programs because they are working
great and are great tools for predicting scalability.
 
I have included the 32 bit and 64 bit windows executables of my
programs inside the zip file to easy the job for you.
 
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.
Ramine <ramine@1.1>: Apr 17 02:13PM -0700

Hello....
 
Here is my contributions of my USL programs..
 
I have first implemented a solver for my USL program that
is polynomial regression, this solver must make
the a0 coefficient of the mathematical serie to 0, but this solver
is not so efficient as my other solver that i have implemented
that is nonlinear regression using the simplex method of
of Nelder and Mead as a function minimization, this nonlinear
solver that i have implemented works perfectly and is more
efficient than the solver that uses polynomial regression,
also my contribution is my USL programs that is called usl_graph
that provides you with a more interractive graphical chart that
permit you to optimize more the criterion of the cost, i think
that the other R package is less powerful on this option.
 
So i think my USL programs are great tools that can predict
scalability.
 
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.
Ramine <ramine@1.1>: Apr 17 02:17PM -0700

Hello,
 
I have included the 32 bit and 64 bit windows executables of my
programs inside the zip file to easy the job for you.
 
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.
Ramine <ramine@1.1>: Apr 17 02:47PM -0700

Hello.....
 
Here is my contributions of my USL programs..
 
I have first implemented a solver for my USL program that
is polynomial regression, this solver must make
the a0 coefficient of the mathematical serie to 0, but this solver
is not so efficient as my other solver that i have implemented
that is nonlinear regression using the simplex method of
of Nelder and Mead as a function minimization, this nonlinear
solver that i have implemented works perfectly and is more
efficient than the solver that uses polynomial regression,
also my contribution is my USL programs that is called usl_graph
that provides you with a more interractive graphical chart that
permit you to optimize more the criterion of the cost, i think
that the other R package is less powerful on this option.
 
Also in my USL programs i have calculated and feed my nonlinear solver
with partial derivatives of the USL equation:
 
C(N) = N/(1 + α (N − 1) + β N (N − 1))
 
I have calculated the partial derivative with respect to
α of the above USL equation, and i have calculated the partial
derivative with respect to β of the above USL equation, and the
two partial derivatives must be given to my nonlinear solver
that uses the simplex method of of Nelder and Mead as a function
minimization.
 
Please try my USL programs because they are working great and
they predict scalability !
 
I have included the 32 bit and 64 bit windows executables of my
programs inside the zip file to easy the job for you.
 
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.
Ramine <ramine@1.1>: Apr 17 02:34PM -0700

Hello......
 
I have wrote before that:
 
"I have first implemented a solver for my USL program that
is polynomial regression, this solver must make
the a0 coefficient of the mathematical serie to 0, but this solver
is not so efficient as my other solver that i have implemented
that is nonlinear regression using the simplex method of
of Nelder and Mead as a function minimization, this nonlinear
solver that i have implemented works perfectly and is more
efficient than the solver that uses polynomial regression"
 
Also in my USL programs i have calculated and feed my nonlinear solver
with partial derivates of the USL equation:
 
C(N) = N/(1 + α (N − 1) + β N (N − 1))
 
I have calculated the partial derivative with respect to
α of the above USL equation, and i have calculated the partial
derivative with respect to β of the above USL equation, and the
two partial derivatives must be given to my nonlinear solver
that uses the simplex method of of Nelder and Mead as a function
minimization.
 
Please try my USL programs because they are working great and
they predict scalability !
 
I have included the 32 bit and 64 bit windows executables of my
programs inside the zip file to easy the job for you.
 
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.
Ramine <ramine@1.1>: Apr 17 02:41PM -0700

On 4/17/2016 2:34 PM, Ramine wrote:
> efficient than the solver that uses polynomial regression"
 
> Also in my USL programs i have calculated and feed my nonlinear solver
> with partial derivates of the USL equation:
 
 
I mean: partial derivatives, not partial derivates.
 
You received this digest because you're subscribed to updates for this group. You can change your settings on the group membership page.
To unsubscribe from this group and stop receiving emails from it send an email to comp.programming.threads+unsubscribe@googlegroups.com.
Posted by at

No comments:

Post a Comment

Subscribe to: Post Comments (Atom)