When a target process calls dlopen(), a dlopen event is generated and must be handled by TotalView. Because dlopen event handling can affect debugger performance for a variety of reasons, especially if the application loads many shared libraries or the debugger is controlling many processes, TotalView provides ways to configure dlopen for better performance and scalability in HPC computing environments:
• Filtering dlopen events to avoid stopping a process for each event.
• Handling dlopen events in parallel, reducing client/server communication overhead to fetch library information. Note: Both this option and MRNet must be enabled for TotalView to fetch libraries in parallel.