Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.programming.threads > #2554 > unrolled thread

Parallel Conjugate gradient linear system solver

Back to article view | Back to comp.programming.threads

Contents

  Parallel Conjugate gradient linear system solver Ramine <ramine@1.1> - 2014-06-29 18:41 -0700
    Re: Parallel Conjugate gradient linear system solver Ramine <ramine@1.1> - 2014-06-29 18:43 -0700

#2554 — Parallel Conjugate gradient linear system solver

Hello,

As you have noticed i have implemented my Parallel Conjugate gradient 
linear system solver, here it is:


https://sites.google.com/site/aminer68/parallel-implementation-of-conjugate-gradient-linear-system-solver


But as you will notice i am using arrays to implement it ,
but arrays do take too much memory even if the matrix is more
sparse, but i have decided to implement it like that with arrays,
so a matrix of 100000 by 1000000 elements will take around
64 Gbytes of memory, so you have to have a server with 32 Gbytes
of memory or more to be able to solve large industrial problems
with my parallel Parallel Conjugate gradient linear system solver.

To use much less memory i have to use linklists , but linklists
are not parallel freindly with my parallel algorithm, they will
be too CPU expensive , hence too slow, so this why i have decided
to keep using arrays in my parallel algorithm


Thank you,
Amine Moulay Ramdane.

[toc] | [next] | [standalone]

#2555

On 6/29/2014 6:41 PM, Ramine wrote:
>
> Hello,
>
> As you have noticed i have implemented my Parallel Conjugate gradient
> linear system solver, here it is:
>
>
> https://sites.google.com/site/aminer68/parallel-implementation-of-conjugate-gradient-linear-system-solver
>
>
>
> But as you will notice i am using arrays to implement it ,
> but arrays do take too much memory even if the matrix is more
> sparse, but i have decided to implement it like that with arrays,
> so a matrix of 100000 by 1000000 elements will take around

I mean a matrix of 100000 by 100000 elements

> 64 Gbytes of memory, so you have to have a server with 32 Gbytes
> of memory or more to be able to solve large industrial problems
> with my parallel Parallel Conjugate gradient linear system solver.
>
> To use much less memory i have to use linklists , but linklists
> are not parallel freindly with my parallel algorithm, they will
> be too CPU expensive , hence too slow, so this why i have decided
> to keep using arrays in my parallel algorithm
>
>
> Thank you,
> Amine Moulay Ramdane.
>
>
>
>
>
>

[toc] | [prev] | [standalone]

Back to top | Article view | comp.programming.threads

csiph-web