dmtcp VS timemory

I would love to make more use of CRIU. E.g. I considered to use CRIU for my Python preloaded logic (https://github.com/albertz/python-preloaded). Unfortunately, at that point in time, CRIU must be used with root access, which was not an option. However, I see that the PR was merged now, so maybe it works now? (https://github.com/checkpoint-restore/criu/pull/1930)

There is also DMTCP (https://github.com/dmtcp/dmtcp/) but this might have other problems for my use case.

My solution was to use a fork server instead, which works almost equally well. There are not really much downsides with this approach. And this is actually quite simple, and also quite cross-platform (except Windows).

Or maybe DMTCP is a better alternative to CRIU?

Anybody that's done it before on Windows could probably get a prototype ready in an hour or two bc it already supports not using fork -- there's a timem-mpi exe built from the same source that uses MPI_Comm_spawn_multiple instead of fork bc OpenMPI will seg-fault when you fork inside a rank.

example enum #1 example usage #1.1 example usage #1.2

This can as simple as using a command-line tool like time) or timem

W.r.t. to performing all this analysis on the code at a high-level and a low-level, checkout the performance analysis toolkit I've been working for the last year and half or so: timemory -- it has a high-level interface and some pre-built tools (such as the aforementioned timem) but it also exposes C++ template interface that's designed so that you can trivially bundle together different measurements and/or create custom measurement types and then use them with as little overhead as you desire, e.g. lightweight_tuple foo{}; foo.start(); ...; foo.stop() will create wall-clock and cpu-clock timers on the stack and never touch any global memory, component_tuple, vtune_frame> bar{ "mylabel"}; will report a wall-clock time, total-cycles, and the L1 cache misses in JSON and text files at the end, and create a "mylabel" frame in VTune when it is connected.