Sparse Matrix extension and graduate project

[OSMAN YASAL]

Hello I'm Osman from Istanbul/Turkey.

Here's a little about me:

I am a new graduate student at one of the best universities of my country and my research is about parallel programming in general. I am currently associated with a project called Sparcity (europhian funded) which is about sparse matrices in general and I am looking a topic to conduct my research.

About Research and My Request:

My advisor would like me to start a research asap! and when I was looking online I came across this library and read comments about how it is efficient. Although I have not examined in greater detail the library wondering maybe I can apply whatever inside the library for sparse matrices as well as an extension.

How are you doing in this project? 

is this idea feasible to you? 

Any answer will be appreciated

~Osman.

[Jed Brown]

Sparse matrices generally don't have sufficient locality for the BLIS tiling approach.

Sparse matrix-matrix products (such as appear in algebraic multigrid setup) are a good topic, but there are soooo many ways in which people set up benchmarks that are not representative of applications and thus arrive at spurious causal claims. Creating a good performance model for existing implementations would be insightful and valuable. The relative performance of sparse matrix-matrix products are not well correlated with either sparse matrix-vector products or dense GEMM (comparing Intel, AMD, POWER9, and ARM).

"'OSMAN YASAL' via blis-discuss" <blis-d...@googlegroups.com> writes:

> Hello I'm Osman from Istanbul/Turkey.
>

> *Here's a little about me:*

> I am a new graduate student at one of the best universities of my country
> and my research is about parallel programming in general. I am currently
> associated with a project called Sparcity (europhian funded) which is about
> sparse matrices in general and I am looking a topic to conduct my research.
>

> *About Research and My Request:*

> My advisor would like me to start a research asap! and when I was looking
> online I came across this library and read comments about how it is
> efficient. Although I have not examined in greater detail the library
> wondering maybe I can apply whatever inside the library for sparse matrices
> as well as an extension.
> How are you doing in this project?
> is this idea feasible to you?
>
> Any answer will be appreciated
> ~Osman.
>
>

> --
> You received this message because you are subscribed to the Google Groups "blis-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to blis-discuss...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/blis-discuss/c27a7292-413b-44ea-9449-b0f01c7bd0bbn%40googlegroups.com.