On the relationship between functional size and software code size

Abstract

SLOC (Source Lines-Of-Code) has been used extensively as a Code Size Measure, and as input to parametric software cost and effort estimation tools. SLOC is obtained by measuring FP (Function Points) on the requirements and multiplying by the SLOC/FP ratio from similar projects. This is done even though several studies show large variations in this ratio, due to weak correlation between FP and SLOC. However, in our previous experiments we have obtained strong correlation between CFP (COSMIC Function Points) and Bytes compiled code as Code Size Measure. The experiments were conducted in the automotive industry using software components developed by GM (General Motors). In this paper we explain the reasons behind the strong correlation. The main reasons are that we apply the COSMIC method on software components of similar type, with a 1-to-1 mapping to COSMIC. A strong correlation between the Functional Size Measure and the Code Size Measure is required to obtain accurate Code Size estimation results. To estimate the Code Size before the software is available, is important both for Cost/Effort estimation and design of electronic hardware.