Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2014-10-18

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
03:28 lue joined #moarvm
04:45 ggoebel11111113 joined #moarvm
06:42 FROGGS joined #moarvm
06:44 flaviusb joined #moarvm
06:45 camelia joined #moarvm
07:26 camelia joined #moarvm
08:08 vendethiel- joined #moarvm
09:35 kjs_ joined #moarvm
10:07 cognome joined #moarvm
12:02 cognome joined #moarvm
13:01 timotimo the decont_all that's put into the nativecall MAST Op ... it can't get jitted because param_sp_o
13:01 timotimo can we get that to work better?
13:01 timotimo i don't even know exactly why param_sp_o gets emitted sometimes
13:02 * timotimo tries $args instead of @args
13:05 colomon joined #moarvm
13:07 timotimo seems like that gives the same op
13:10 timotimo er
13:10 timotimo i meant to write param_rp_o
13:10 timotimo "required positional object", right? how come we can't jit that :S
13:11 timotimo we don't jit named parameters? or something like that?
13:12 timotimo i differentiate between relative and absolute move_to and line_to with a :relative at the end
13:12 timotimo seems like the jit doesn't like that so much
13:16 timotimo oh interesting
13:16 timotimo 10 bails on "multicachefind"
13:16 timotimo that should be easy
13:19 moritz does removing reasons to bail actually improve performance measurably?
13:23 timotimo depends entirely!
13:23 timotimo it's just more stuff i can do
13:23 timotimo i'd think moving between jit and interpeter has a tiny amount of cost
13:23 timotimo otherwise we can inline and stuff
13:24 timotimo hm. invokewithcapture follows all instances of multicachefind
13:24 timotimo invokey ops are a bit harder
13:25 dalek MoarVM/multicachefind: 7fd90e7 | (Timo Paulssen)++ | src/jit/graph.c:
13:25 dalek MoarVM/multicachefind: multicachefind for the jit
13:25 dalek MoarVM/multicachefind:
13:25 dalek MoarVM/multicachefind: needs invokewithcapture in order to actually work.
13:25 dalek MoarVM/multicachefind: review: https://github.com/MoarVM/MoarVM/commit/7fd90e7569
13:25 timotimo brrt will have to do something about this :
13:25 timotimo :P
13:26 timotimo the three ops that bail the most in my game are: param_sn and param_rp_o with 45 frames each
13:26 timotimo 26 frames with param_sp
13:26 timotimo 10 with multicachefind (and after that invokewithcapture)
13:26 timotimo 9 argconst_s, 7 scgetobjidx
13:28 timotimo exception, setobjsc, setcodeobj, captureposelems, capturehasnameds, reprname, param_on2_o
13:28 timotimo newexception, isinvokable, getcomp, ctx, cmp_n,
13:28 timotimo all after that have 2 or 1 bails
13:34 dalek MoarVM: 41216a0 | (Timo Paulssen)++ | src/jit/graph.c:
13:34 dalek MoarVM: jit pow_n and sqrt_n
13:34 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/41216a009d
13:38 cognome joined #moarvm
13:55 timotimo there's really not terribly much to add ...
14:10 _sri joined #moarvm
14:13 timotimo but it's still too slow! oh no!
14:16 colomon joined #moarvm
14:40 kjs_ joined #moarvm
14:51 zakharyas joined #moarvm
16:35 cognome joined #moarvm
16:38 cognome_ joined #moarvm
17:04 zakharyas joined #moarvm
17:50 cognome joined #moarvm
19:19 Ven joined #moarvm
22:27 kjs_ joined #moarvm
22:49 flussence joined #moarvm
22:58 cognome joined #moarvm
23:49 kjs_ joined #moarvm
23:55 colomon joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary