JimmyZ timotimo: I didn't do that, I was reading spesh code to under thing well. Though I found many code needs to be opitmized :) 01:26
*understand 03:41
05:06 [Coke]_ joined 05:12 harrow joined 07:33 FROGGS joined 08:23 brrt joined 09:06 kjs_ joined
brrt yesterday nights spectest wasn't very pretty (linux, master HEAD) 09:17
nwc10 oh? Still works on "my" machine.
JimmyZ: "well volunteered" :-) 09:18
(it would be wonderful if you could figure this stuff out, but please don't feel that you have to) 09:19
brrt maybe i was doing it wrong :-)
nwc10 maybe I am
but I'm doing a git clean before the build
brrt as did i :-) 09:20
i don't have the machine at me at the moment so i can't really check
09:38 FROGGS joined 10:02 dalek joined 10:09 dalek joined 10:31 FROGGS joined 10:49 rurban joined 10:57 lizmat joined 11:56 nebuchad` joined 12:09 kjs_ joined 12:39 japhb joined 12:45 dalek joined
timotimo i had a fully clean spectest run yesterday 13:41
with all masters and freshly reconfigured
JimmyZ: can you name examples for things that should be optimized by spesh?
JimmyZ: maybe you can try implementing the "boxing/unboxing tracking" thing i've tried to implement once before 13:42
it blew up in my face sadly; couldn't build nqp (or rakudo?) any more with it
but there's multiple occasions where the post-spesh code boxes an int and immediately unboxes it again
probably due to in-lined stuff
brrt timotimo - most likely /me did something wrong like not install fresh nqp before moving to the rakudo spectest 13:55
if everybody else is seeing a clean spectest 13:56
(btw, in the category of 'because you can', i was trying to get rakudo to work on minix :-))
timotimo haha
wow
brrt but so far had limited success due to virtualbox-minix network driver suckiness
BinGOs toy OS is toy. 13:57
brrt i used to have success though
but yeah :-)
as far as toys go, i think it is kind of cool
BinGOs they appear to have shot the pooch. It was once possible to build perl5 on minix, but not with the latest version 13:58
lizmat brrt: what kind of filesystem (semantices) does Minix have ? 13:59
unixlike I would hope?
brrt pretty much, yes
they claim netbsd compatibility 14:00
lizmat ah, ok :-)
14:01 kjs_ joined
BinGOs en.wikipedia.org/wiki/MINIX_file_system 14:01
en.wikipedia.org/wiki/MINIX#MINIX_3 14:02
nwc10 how did they manage to break the ability to build Perl? That seems pretty impressive for an OS that is attempting to be Unix-like, considering *how* portable Perl 5 is 14:03
brrt that is really odd
lizmat
.oO( a system on which rakudo builds, but perl 5 doesn't? what is happening to the world? )
 14:04
BinGOs uname -a 14:05
Minix 192.168.1.85 3.3.0 Minix 3.3.0 (GENERIC) i386
brrt lizmat: i haven't actually built rakudo yet :-) 14:06
i said i was going to
lizmat yeah, I know, I was just dreaming :-)
BinGOs I can't find my perl5 build on this vm, so I must have destroyed the old one and haven't got around to trying on this one yet. 14:07
I think the conclusion I was coming to was that a hints/minix.sh would be needed.
and the build failure was possibly clang-related. I can't remember. 14:08
14:10 rurban joined
brrt hmmm 14:13
ideally we'd be able to build moarvm without perl5
nwc10 that's about step 5 14:14
logical pre-reqs are get Rakudo testing and installing without Perl 5
then configuring and building without Perl 5
(easy enough, NQP is around. Also potentially better, as a Rakudo-based Test::Harness replacement can run in parallel more effectively, and also on Win32) 14:15
and then do the same with NQP
and only then, think about MoarVM
none of this is needed to get Perl 6 to 6.0.0
or to get performance properly juiced.
BinGOs you can get a perl5 from the packages. 14:17
it reports itself as 'i386-netbsd' in perl -v 14:18
JimmyZ timotimo: maybe I'm wrong, but I think we need more base optimization code like CSE, AA, Constant folding etc to do more work :) 14:26
brrt optimization is hard 14:28
JimmyZ yeah
brrt let's go shopping :-P
for my clarity, what is CSE? 14:31
and AA?
JimmyZ brrt: wiki.luajit.org/Optimizations
brrt that is some impressive list there 14:38
JimmyZ yeah 14:39
14:43 FROGGS[mobile] joined
JimmyZ looks like CSE is not that hard, except I don't know how to get the CSE entry in spesh 14:44
brrt well.... 14:48
ehm
doesn't SSE form make give you that? 14:49
in the sense that for any 'final' register value, you can determine the graph of how it was computed, and hence you can find out if there are any subexpressions within that graph 14:50
BinGOs brrt: re: minix, I've just had to change it to a manual network configuration instead of the dhcp/nonamed bullshit. 15:05
and the network is actually usable
vbox+lance 15:06
brrt aha 15:09
i'm using the detected (emulated) intel pro card and got major crashing
JimmyZ but looks like it relies on AA, at least in luajit :( 15:10
ah , let's sleep, it's hard 15:11
BinGOs you should have more success with AMD *mumble* *mumble* and the lance driver
but GNU/Hurd feels like a more polished OS than MINIX 15:12
JimmyZ Good night 15:15
lizmat gnight JimmyZ 15:18
brrt BingOS - that is pretty damning for MINIX 15:31
sleep well JimmyZ
15:31 colomon joined 15:47 kjs_ joined 16:00 zakharyas joined
BinGOs oh my. 16:00
This is perl 5, version 21, subversion 8 (v5.21.8 (v5.21.7-46-g5e75412)) built for i386-minix
brrt perl5 works \o/ 16:01
via pkgsrc?
BinGOs that's blead 16:02
I've made a very basic hints file to get it working.
for your purposes, you could probably get away with the perl5 from using pkgin 16:03
brrt excellent 16:32
16:32 brrt left 16:36 colomon joined 17:23 rurban joined 17:36 kjs_ joined 17:56 kjs_ joined 18:12 FROGGS[mobile] joined 18:45 JimmyZ joined 19:48 FROGGS_ joined 21:22 rurban joined 21:32 kjs_ joined 23:06 kjs_ joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!