[lnkForumImage]
TotalShareware - Download Free Software

Confronta i prezzi di migliaia di prodotti.
Asp Forum
 Home | Login | Register | Search 


 

Forums >

comp.programming

Let's talk more computer science...

Ramine

1/3/2015 1:53:00 AM


Hello,

Let's talk more computer science...

Finally i have arrived to an important subject that is "optimization",
and you will notice that computer science learns you one
of the most important thing, is how to amortize and optimize by
optimizing more your time complexity that we express mathematically with
a O(), this is how i have done it with my new parallel sort algorithm
that has a "very" good time complexity, and what learns you also
computer science is how to optimize more in the source code's part or
parts that takes the greater percentage of running time, and what learns
you also computer science is to do the greater percentage of all the
optimizations and to not do the smaller optimizations, this is also a
good way to follow to optimize your code, so if you look at my
parallel conjugate gradient solver library , i am not using SIMD SSE2
instructions , but i am just parallelizing the memory transfers cause my
parallel algorithm is NUMA-aware and i am also parallelizing also the
multiplication and addition of the double's type etc. and also my
algorithm is cache-aware so i am also parallelizing more the L2 cache
memory transfers , so this optimization will make it a very greater
percentage of all the optimization that we can do on my parallel
algorithm, so in large scale industrial problems i don't think that my
parallel conjugate gradient algorithm needs to do SIMD SSE2 or AVX
optimization , cause SIMD SSE2 or AVX will give just a small improvement
to my parallel algorithm.


You can download my Parallel conjugate gradient linear system solver
library that is NUMA-aware and cache-aware from:

https://sites.google.com/site/aminer68/scalable-parallel-implementation-of-conjugate-gradient-linear-system-solver-library-that-is-numa-aware-and-c...



Thank you for your time.


Amine Moulay Ramdane.