[Scilab-Dev] Scilab5.3 SSE3

[Eduardo Tarasiuk]

*This message was transferred with a trial version of CommuniGate(r) Pro*
We take the functions or programs in the link using but tested only MKL,
we did not attached the results today in a mail to this list because
some crazy work in my office :) but for sure Ronit , my colleague, will
do it tomorrow, there are a lot of graphs in the document.
We take your work as a reference and as I wrote yesterday we did not saw
any difference between SSE3 and SSE, sorry it was my mistake (a wrong
installation without the MKL libraries)

"The CONCLUSION: 
Benchmarks on operation based on MKL will have the same result between
5.3.0 and
with 5.3.0-SSE3 because MKL uses all features of the CPU by default."


You are completely right !!.

     Thanks
               Eduardo



-----Original Message-----
From: Netanel [mailto:offirbs at gmail.com] 
Sent: ב 31 ינואר 2011 11:33
To: dev at lists.scilab.org
Subject: Re: [Scilab-Dev] Scilab5.3 SSE3

*This message was transferred with a trial version of CommuniGate(r)
Pro*
Michaël Baudin <michael.baudin at ...> writes:

> 
> Hi,
> 
> Can you be more specific about the particular improvements that you 
> experienced ? What functions ?
> 
> Best regards,
> 
> Michaël Baudin
> 
> Le 07/01/2011 06:52, Eduardo Tarasiuk a écrit :
> > According to my very preliminary performance tests for the commands
I
> > checked, I can see that the improvement is amazing, even better than
in
> > Matlab.
> 


Hi, 

I did some performance tests: Scilab 5.3.0 Vs Scilab 5.3.0-SSE3

The benchmark is set of Scilab Math core functions, It is performed on
Scilab
v5.3.0/5.3.0-SSE3 under Windows XP 64 bits. The processor is an Intel
Core 2 Duo
E7500 at 2.93 GHz.
I used Intel Math Kernel Library(MKL). 

I used the tic and toc functions to measure the performances of an
algorithm. 
I repeat the timing 10 times to get a more reliable estimate of the
performance.

I compare the performance of Scilab math core operations (such as:
identity,random,zeros,ones,matrix-matrix multiply,Trigonometric
functions,logarithms, exponentials etc.)
I compare the performance with matrix order increasing.
I also compare the performance of Scilab Demos.

The CONCLUSION: 
Benchmarks on operation based on MKL will have the same result between
5.3.0 and
with 5.3.0-SSE3 because MKL uses all features of the CPU by default.

References
[1] "Programming in Scilab", Michael Baudin, 2010
[2] http://wiki.scilab.org/Linalg_performances 


I would appreciate your comments on this. ‏