18:43 <nixuser22> hey, I am benchmarking an application on pypy and it takes (at least) 3x to 4x as much memory, is that usual? How can I profile what takes that much memory? (tracemalloc and vmprof do not support memory profiling on pypy afaict)

19:24 <nixuser22> sry got disconnected, did I miss a response?

19:59 <cfbolz> nixuser22: nope, no response yet

19:59 <cfbolz> nixuser22: how much memory are we talking about?

20:17 <mattip> there are some gc hooks you can use to try to figure out what is taking up memory

20:17 <mattip> https://doc.pypy.org/en/latest/gc_info.html#gc-get-stats

20:18 <mattip> as well as a jit hook, just in case that is generating too much code

20:18 <mattip> https://doc.pypy.org/en/latest/jit-hooks.html#get_stats_asmmemmgr

21:33 <nixuser22> cfbolz: like 200M per python process (from 50M)

21:33 <nixuser22> will try the hooks, thx

22:04 <mattip> do you have interactions with C (cffi, cytpes, c-extension modules ...)? Some of the uses of these can result in pinned memory.

22:05 <mattip> the pinning should be short-lived, but if the objects are not quickly freed then suboptimal memory usage could maybe occur