The main objective of this paper is to propose a software implementation model for concurrency error detection tools that allows reducing process overhead without decreasing its detection capacity. The general model proposed uses software dynamic instrumentation in such a way that an analysis routine can be activated from a signal generated by a hardware event that indicates the possibility of an error occurring. The results obtained showed that, for the case study (an atomicity violation detection algorithm called AVIO), the version that uses the model proposed can detect the same bugs as the original version, but in only 25% of the time (in average) required by it.