Locating Performance Changes in Source Code
In regression benchmarking, it is often complicated to locate the source code modification which caused a performance change detected by a particular benchmark. Manually examining all modifications and resolving if some specific modification has caused the performance change in large projects can be...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Dissertation |
| Language: | English |
| Published: |
2008
|
| Online Access: | http://www.nusl.cz/ntk/nusl-287611 |
| Summary: | In regression benchmarking, it is often complicated to locate the source code modification which caused a performance change detected by a particular benchmark. Manually examining all modifications and resolving if some specific modification has caused the performance change in large projects can be timeconsuming. The thesis analyses possible methods for making that examination faster; focusing in more detail on two of them. The first is to have software that can understand the source language and can resolve which modification can potentially cause performance changes. For example, modifications in comments can not affect performance. The second is selecting source changes that were executed during the test. As a proof of concept, the methods are implemented in Mono Regression Benchmarking Suite. In this master thesis, more solutions how to increase time efficiency of the search are discussed and the two noticed above are implemented. |
|---|