Re: GNAT vs Matlab - operation on multidimensional complex matrices

["J-P. Rosen" <rosen@adalog.fr>]

Le 24/03/2020 à 00:16, darek a écrit :
> Hi Everyone, 
>  I am working on radar signal processing algorithms that use multidimensional complex arrays. 
> 
> To my surprise, the performance of some Matlab functions is much better than compiled Ada code. 
> 
[...]

Two remarks:
1) why do you need pointers? Ada is very good at using pointer only when
absolutely necessary, which is good for safety and efficiency.

2) you are comparing a hand-written, naive[1] algorithm with a
specialized function that may well be written in assembly language,
making use of special SIMD instructions. I would say that Ada does not
behave that bad...

[1] Naive in that you simply use the "obvious" loops. If you want
maximum efficiency, a lot of tricky factors come into play. For example,
changing the order of the nesting of the loops changes the spread of
memory accesses, which can (or not) have a dramatic effect on
efficiency, given the various caches, whether successive memory accesses
belong to the same memory bank, etc. High efficiency mathematical
libraries are designed to consider these effects.

-- 
J-P. Rosen
Adalog
2 rue du Docteur Lombard, 92441 Issy-les-Moulineaux CEDEX
Tel: +33 1 45 29 21 52, Fax: +33 1 45 29 25 00
http://www.adalog.fr