Welcome to the main channel on the development of MoarVM, a virtual machine for NQP and Rakudo (moarvm.org). This channel is being logged for historical purposes.
Set by lizmat on 24 May 2021.
jdv timo: i couldn't find it 12:41
also, i think i saw similar mem growth in a $work project that had no subproc stuff - but it did use async stuff
MasterDuke huh. perf shows MVM_get_lexical_by_name at the top with 11% for `raku -e 'loop { put 1 }' >/dev/null` 21:01
next are MVM_frame_find_dynamic_using_frame_walker, MVM_spesh_frame_walker_next, MVM_spesh_frame_walker_get_lex 21:03
huh. running that for 1_000_000 iterations results in 10_000_000 cases of the lexical_names hash *not* being built (for $*PROMISE and $*OUT) and 3_000_000 cases of it being built (for the same two vars) 21:15
and when the lexical_names hash is not built, 20_011_570 times the name isn't found in sf->body.lexical_names_list 21:21
hm, and in 3_015_965 of the times MVM_string_equal is called, b doesn't have a cached_hash_code 21:34