Skip to content

Conversation

@HeatCrab
Copy link
Contributor

This PR resolves memory leaks in the original implementation and improves stability with a warm-up phase. Previously, the initialization function was called multiple times within the testing function, reallocating memory for the context array without releasing earlier allocations, resulting in leaks detected by Valgrind (43,632 bytes across 909 blocks per test). Now, the initialization happens only once, and memory is released when testing completes, preventing any leaks.

Additionally, a warm-up step is added in the measurement function by discarding the first batch of data. This filters out initial anomalies (such as cache misses or memory allocation delays), aligning with the design principles of the dudect framework and enhancing measurement consistency.

These changes ensure reliability, eliminate memory issues, and adhere to the lightweight and stable testing approach outlined in the dudect paper.

Change-Id: I3ab522ff205fab8a4c1ad55f8e552d7d8178b69f

This commit resolves memory leaks in the original implementation and improves stability with a warm-up phase. Previously, the initialization function was called multiple times within the testing function, reallocating memory for the context array without releasing earlier allocations, resulting in leaks detected by Valgrind (43,632 bytes across 909 blocks per test). Now, the initialization happens only once, and memory is released when testing completes, preventing any leaks. Additionally, a warm-up step is added in the measurement function by discarding the first batch of data. This filters out initial anomalies (such as cache misses or memory allocation delays), aligning with the design principles of the dudect framework and enhancing measurement consistency. These changes ensure reliability, eliminate memory issues, and adhere to the lightweight and stable testing approach outlined in the dudect paper. Change-Id: I3ab522ff205fab8a4c1ad55f8e552d7d8178b69f
@HeatCrab HeatCrab closed this Mar 29, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

1 participant