Apologies if this is not the place to ask these kinds of questions.
We are currently trying to analyze the overhead caused by icache misses on benchmarks running on a RISCV machine. To do this we will be using the cache modeling plugin. This works with user space emulation, but we are having some trouble cross compiling some benchmarks. A workaround that we are considering would be to build the benchmarks on a RISCV VM. However, we don't want the cache modeling plugin to be running during the entire process. Would it be possible to run the VM, pause execution or save a snapshot, enable the plugin, and then continue execution or load from the snapshot while the plugin is running? If not, would there be a better solution?
Thanks,
Conan