While CxxProfiler is pretty functional already now, there are few improvements I've been working on.
Main change would be using ETW API for sampling call-stacks. It would allow much higher frequency and lower overhead than current mechanism. On Windows 10 it allows to sample stack every 0.122 ms (1/8192). I still want to keep current sampling mechanism (basically a debugger) because it can be useful in various situations - for example, if you need to run profiler it as non-admin user. So to do all this I'm refactoring non-UI code, so I can have two backends in code. I'm cleaning up backend code (original one was written ~9 years ago) and making it more handmade-like, that means controlling memory allocations, and not relying on Qt. At the same time I'm separating common functionality for two backends - symbol lookup, string caching, communication with UI process. Now debugger backend looks pretty nice and clean (check out backend_debugger.c file). Profiler will now launch backend in separate process and communicate data back to UI process over pipe. I decided running backend into separate process because ETW will require admin rights, but UI process doesn't need to be running with admin privileges. Also it provides a bit of safety, in case backend will crash, the UI will be still alive and will be able to display partially received data.
You can follow latest changes in branch "next" on github: https://github.com/mmozeiko/CxxProfiler/commits/next