Open vikramsubramanian opened 3 months ago
To implement the feature of using whiteboard tags for finding more performance bugs, you can follow these steps:
fxperf
.whiteboard_keywords
function in bugbug/bug_features.py
to filter bugs that contain performance-related tags.whiteboard_keywords
function to create a new feature or enhance an existing feature that identifies performance bugs based on whiteboard tags.Here's a pseudo-code example of how you might modify the whiteboard_keywords
function to include a check for performance-related tags:
def whiteboard_keywords(bug):
# ... existing code ...
# Define a list of performance-related tags
perf_tags = ['fxperf', 'perf', 'performance']
# Filter splits to include only those with performance-related tags
perf_related_splits = [split for split in splits if any(tag in split for tag in perf_tags)]
return perf_related_splits
Then, you can use this function within the existing feature extraction process to identify performance bugs. For example, you might create a new feature class that uses the whiteboard_keywords
function to classify bugs:
class PerformanceBugs(SingleBugFeature):
name = "Performance related bugs"
def __call__(self, bug, **kwargs):
keywords = whiteboard_keywords(bug)
return any(keyword in keywords for keyword in ['fxperf', 'perf', 'performance'])
This new feature can then be used in the machine learning model or any other analysis to identify and prioritize performance bugs.
Remember to test the new feature thoroughly to ensure it correctly identifies performance bugs and integrates well with the rest of the system.
There are some whiteboard tags that are used for performance bugs (e.g. fxperf if I'm not mistaken). We can find more examples of perf bugs by using those tags. )