Hello Everyone,
In this and coming next posts, I am going to discuss the Gsoc 2018 project I am applying for. To have a glimpse of the idea of the project go through the link.
I have already started working on this project and this is the work I have done till now on my Github repo. Go to this link and take an overview of the work and the details about the project, it's limitations or the current part being implemented.
I will be updating this blog from now on for this project and the issues related to this project. Any kind of suggestions and feedbacks are most welcome.
Update 1:
The current issue is with the speed of multiplication of two matrix of size 100 by 100, which takes around 317.748 milliseconds in SymMat class as compared to 24.72 milliseconds in case of Eigen:: Matrix class.
The Eigen Matrix class uses a method similar to BLASS GEMM. So the current plan is to implement something similar to this.
Update 2:
I have updated the symmetric.cpp file with using multithreading in case of multiplication of one SYmMat and One Eigen Matrix, both of size 100 by 100.
By doing so I was able to reduce the initial time complexity by 27% i.e now the time taken is 232.278 milliseconds.
In this and coming next posts, I am going to discuss the Gsoc 2018 project I am applying for. To have a glimpse of the idea of the project go through the link.
I have already started working on this project and this is the work I have done till now on my Github repo. Go to this link and take an overview of the work and the details about the project, it's limitations or the current part being implemented.
I will be updating this blog from now on for this project and the issues related to this project. Any kind of suggestions and feedbacks are most welcome.
Update 1:
The current issue is with the speed of multiplication of two matrix of size 100 by 100, which takes around 317.748 milliseconds in SymMat class as compared to 24.72 milliseconds in case of Eigen:: Matrix class.
The Eigen Matrix class uses a method similar to BLASS GEMM. So the current plan is to implement something similar to this.
Update 2:
I have updated the symmetric.cpp file with using multithreading in case of multiplication of one SYmMat and One Eigen Matrix, both of size 100 by 100.
By doing so I was able to reduce the initial time complexity by 27% i.e now the time taken is 232.278 milliseconds.
No comments:
Post a Comment