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.
Nicholas good *, #moarvm 06:48
lizmat and there's another day almost past :-) 12:55
target has a :noinline instruction - ins: param_rp_o
is that an easily fixable inline issue ?
jnthnwrthngtn It's more an argument optimization issue; the argument handling optimization usually rewrites all of those 13:12
So by the time we reach inlining, they ideally are gone (since there's not a good way to inline them)
So the question is "why weren't they written to unchecked parameter instructions" 13:13
Time for me to POETS :) o/ 13:28
lizmat full disclosure: 13:51
2x write_uint32_at BB(7, 68 bytes) -> add_uint32 BB(330):
target has a :noinline instruction - ins: param_rp_o
3x write_uint16_at BB(9, 68 bytes) -> add_uint16 BB(331):
target has a :noinline instruction - ins: param_rp_o
japhb POETS? 17:53
MasterDuke lizmat: what's the raku code where you see that? 18:37
lizmat it's from running test-t with -MSIL 18:50
raku -MSIL test-t.pl <hello.csv
Nicholas japhb: en.wikipedia.org/wiki/POETS_day 18:51
"It is tradition to begin the POETS day at 3:30 p.m" is definately [Citation needed] 18:52
MasterDuke huh, i don't see that 18:53
ok, i do on master (i was on my branch that converts a lot of `find_symbol([...], :setting-only)` to `find_symbol_in_setting([...])`) 19:02
japhb Nicholas: Ah, got it! 19:04
MasterDuke hm. none of the obvious things that would cause param_rp_o to not get optimized in src/spesh/args.c seem to be the cause 19:33
MasterDuke i seem to get different results with -MSIL and just running with MVM_SPESH_LOG=1. there's no mention of param_rp_o or write_uint* when i do MVM_SPESH_LOG=1 20:33

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!