When I capture a performance profile with, say, perf record --call-graph dwarf $(command -v kgpg), should I provide the raw perf.data (albeit generally compressed in an archive), or should I provide the .perfparser file that I can generate with something like hotspot? I ask because IDK whether the perf.data can be parsed solely on the generatee’s OS installation, or another machine entirely (like gdb’s generate-core-file’s coredumps appear to be able to).