Dear Valgrind Development Team,

I hope this message finds you well. I am a frequent user of Callgrind for performance profiling and greatly appreciate the tool’s flexibility and robustness. However, I have encountered a scenario where I believe a new feature could significantly enhance its usability.

Proposed Feature:
I propose adding an optional capability for Callgrind to automatically save a snapshot of profiling data upon the monitored process’s termination (normal or abnormal). This feature could be controlled via a command-line option or environment variable, such as --auto-dump-on-exit, and applied selectively using existing NPID control mechanisms.

Use Case:
In workflows where applications may terminate unexpectedly (e.g., due to crashes or external signals), the current manual invocation of CALLGRIND_DUMP_STATS can result in missed profiling opportunities. Automating this step would ensure critical profiling data is captured without requiring changes to the application code or extensive wrapper scripts.

Feasibility:
Given that Callgrind already supports NPID-based control, implementing this feature seems feasible by extending its termination handling logic. Users could opt-in to this behavior only for specific processes of interest to avoid unnecessary data generation.

Benefits:
-Ensures profiling data is not lost due to unforeseen process terminations.
-Reduces the need for custom instrumentation or external monitoring scripts.
-Enhances user control while maintaining Callgrind’s efficiency.
I would be happy to discuss this further or provide additional examples and use cases. If this feature aligns with the tool’s development goals.

Thank you for considering this request, and I look forward to your feedback.

Best regards,
Elie
Expert Performance Test Engineer