Project Description

Proposal

The initial GSoC19 proposal from Boost.uBLAS is to add Boost.YAP expression templates to the library. It also included some work on the domain such as Einstein Notation and Policy for Execution device. For more information about the Project have a look at my proposal here.

The Second Project is to add Design Policy to Tensor, for example, choosing storage type at compile time such as Sparse or Band Tensor and improving the design by add compile-time extents and strides which could speed things up. For more information about this project click here. I have explained things in my proposal in great detail.

Mentor

We have a wonderful and helpful mentor named Cem Bassoy

Milestones and tasks

We have 3 milestones each for one of the phases of GSoC. Detailed description for the work of each phase for both of us is tabulated below.

Ashar's Task

Phase	Work	Deadline
1	During this period I will work on to replace the current ublas expression with Boost.YAP expression.	25 June, 2019
2	During this period I work on to write YAP transforms that performs arithematic optimizations on the expression.	22 July 2019
3	During this period I work on to make the expression template compatible for device-based execution and bring execution device chooser API's.	19 August 2019

Each Phase involves 3 Subworks, coding, documenting and testing. The work of each phase is furthur divided into those sub-sections.

Amit's Task

Phase	Work	Deadline
1	For this period I will be designing and implementing static strides and static extents	25 June, 2019
2	For this period I will be designing standard interface for Sparse, Band and Dense Tensor with the implementation of Map based storage	22 July 2019
3	Design and implement slice and subtensor	19 August 2019

Each Phase involves 3 Subworks, coding, documenting and testing. The work of each phase is furthur divided into those sub-sections.

Future Work

We plan to work in the tensor extension even after GSoC19 as the work we are doing is heavy and we have many ideas that could improve uBLAS Tensors. These ideas were not covered in the Proposal we made because we had to make sure that we write an achievable proposal. More information about this could be found at our Future Work Project