GPU acceleration for dense/sparse matrix multiplication on finite fields

[Pixels 么]

Dear madam or sir,

I’m a junior in Shandong University in China, majoring in Computer Science and Technology. I would like to take part in the project which concerning GPU acceleration for dense/sparse matrix multiplication on finite fields.

I'm interested in this project because of its fundamental promotion to scientific compution. What's more, I have the skill of c and c++ programming and familiarization with Linux and have learned parallel computing last semester. Through studing the parallel computing, I get the parallel computing knowledge and the basic parallel programming skills about some parallel libraries, such as OpenMP, Pthreads, CUDA and so on. Added the linear algebra knowledge which was learned when I'm in the first year after I was admitted to the university, I think I can achieve this project. And now I'm studing a course --- Scientific Computing, which introducing the floating-point system, methods and priciples to solve linear system and so on. This curriculum can also promote my work in this project. Since I decided to do this project, I have been looking through the materials to prepare for the project.

In the case of implementing the GPU version of matrix multiplication on finite fields, I think I should do several steps. First, I should master algorithms about matrix multiplication, such as Strassen's method. Second, I should learn the fast numerical GPU BLAS libraries(cuBLAS, cuSPARSE). Third, I should implement the algorithms which I use and test the program. While doing these steps, I can study the related work, including the LinBox GPU acceleration implementation and related papers, to help my work.

I hope I can participate in this project under the mentor's help. Because lacking related work experience, I need mentors' help. And I would appreciate you very much if you could give me some suggestions to prepare for this project, such as selecting appropriate algorithm and optimize the implementation about the algorithms and so on.

 

 Sincerely yours,

Guangzhong YAO

[Brice Boyer]

Hi,

well now you need to build up a proposal and share it very soon.

best,

brice.

--
You received this message because you are subscribed to the Google Groups "linbox-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to linbox-devel...@googlegroups.com.
To post to this group, send email to linbox...@googlegroups.com.
Visit this group at http://groups.google.com/group/linbox-devel.
For more options, visit https://groups.google.com/d/optout.

[Semyon Grigorev]

Hi!

What about GPU acceleration of matrix multiplication over finite fields? Any progress?

[Semyon Grigorev]

Hi,

what about GPU-accelerated matrix multiplication over finite fields?