Novice Programmers’ Coping with Multi-Threaded Software Design
Multi-threaded software design is considered to be difficult, especially to novice programmers. In this study, we explored how students cope with a task that its solution requires a multi-threaded architecture to achieve optimal runtime.
An efficient exploit of multicore processors architecture requires computer programs that use parallel programming techniques. However, parallel programming is difficult to understand and apply by novice programmers.
The students had to address a two-stage problem: (1) design an optimal runtime solution to a given problem with no additional instructions; and (2) provide an optimal runtime multi-threaded design to the same problem. Interviews were conducted with a representative group of students to understand the underlying causes of their provided designs. We used qualitative research methods to gain refined insights regarding the students’ decision making during the design process. To analyze the gained data, we used content analysis tools.
The case study presented in this paper will help the teacher to stress the merits and limitations of various parallel architectures and confront students with the consequences of their solutions via performances’ benchmark.
Analysis of the student’s solutions to the first stage revealed that the majority of them did not provide a multi-threaded solution ignoring the optimal runtime requirement. At the second stage, seven various architectures were provided differing in the number of involved threads, the data structures used, and the synchronization mechanism employed. The majority of the solutions were sub-optimal and only a few students provided an optimal one.
We recommend conducting class discussions that will follow a task similar to the one used in this study.
To be able to generalize the received results this research should be repeated with larger study participant groups from various academic institutions.
Understanding the difficulties of novice programmers may lead to quality software systems.
To be able to generalize the received results this research should be repeated with larger study participant groups from various academic institutions.
Back