| IRC logs at
Set by AlexDaniel on 12 June 2018.
03:20 evalable6 left, linkable6 left 03:22 linkable6 joined 03:23 evalable6 joined 06:32 frost-lab joined 07:53 evalable6 left, linkable6 left 07:55 linkable6 joined 07:56 evalable6 joined 08:08 Altai-man joined 09:09 sena_kun joined 09:10 Altai-man left 09:18 dumarchie joined
dumarchie Hi MasterDuke, I got the following error in gmake: 09:19
tellable6 2020-11-05T18:40:28Z #moarvm <timotimo> dumarchie please check out if fixes build on MinGW
dumarchie Found C:\raku\rakudo\install\bin\moar.exe version 2020.11-47-g1c7358004, which is too old. Wanted at least 2020.11-85-g82a34e25a at C:/raku/rakudo/nqp/3rdparty/nqp-configure/lib/NQP/ line 192.
Should I just remove the whole install directory?
MasterDuke dumarchie: i think if you have a new enough rakudo there's a `--force-rebuild` option 09:30
dumarchie Ah, I will try to remember that for next time. I removed the install directory already and am now rebuilding everything from scratch. 09:32
MasterDuke that'll work too. looks like --force-rebuild was added end of last month 09:33
dumarchie (y)
nine Usually the message is simply correct though 09:37
dumarchie Hmm, `gmake test` failed because there was no plan found in the TAP output of t\04-nativecall\17-libnames.t 09:39
nine So it crashed before even putting out a test plan 09:40
dumarchie I'll `gmake install` anyway
Offline life is calling me. I'll check in again later today. 09:44
nine Offline life? I seem to vaguely remember that... 09:45
dumarchie =D
09:45 dumarchie left 13:08 Altai-man joined 13:10 sena_kun left 13:43 brrt joined
nwc10 good *, brrt 13:44
brrt good * nwc10 13:46
13:51 MasterDuke left, lucasb joined 14:20 MasterDuke joined
MasterDuke brrt: picked up an apple m1 yet? 14:21
brrt didn't, couldn't justify the expense 14:24
and $work needs far more than the 8gb that apple ships 14:25
I might get a raspberry pi 400 though
MasterDuke ah, too bad
brrt that stuff is awesome
MasterDuke yeah, those are pretty cool though
how generic would an aarch64 port of the jit be? i assume it would work on both the rpi4 and the m1? 14:28
brrt it depends on the call convention I guess
MasterDuke we don't do any specialization on the x86-64 version for cpu capabilities, right? e.g., sse3 vs avx 14:29
brrt I think in the basic opcode set there's very little specialization you can do 14:30
all the special stuff, as far as I know anyway, is vector/simd processing
and you need a bunch of analysis to prove that's necessary or even useful 14:31
I imagine that a hypothetical port to aarch64 will mean a major refactor as well
MasterDuke runtime analysis? 14:32
brrt not sure I parse correctly, but you'd need to analyze the code-under-optimization, at runtime, to prove that the code benefits from vectorization, I'm actually not sure the JIT is the right place for it 14:33
the spesh IR is much more powerful
14:40 sxmx left
MasterDuke interesting 14:41
14:44 sxmx joined
nwc10 the raspberry pi 400 has less RAM than the bare-bones 8GB Pi (well, d'oh!) 14:46
and last night's discovery is that I can build Rakudo 32bit on a 2GB Mips machine, but not 64bit on the same 2GB Mips machine 14:47
so 4GB likely is enough for aarch64 64 bit
(I have the hardware - I could test this if I rewrite an SD card)
but I'm not sure how much of a squeeze it is
on the other hand, the raspberry pi 400 is far more cute
I believe that we're doing better than PyPy 14:48
16:15 dumarchie joined
dumarchie MasterDuke, I added a nice backtrace to #1237. Can you let me know if you think if it's useful to try again with `MVM_JIT_DISABLE=1` and/or `MVM_SPESH_DISABLE=1`? 16:17
brrt dumarchie: definitely always helps :-) 16:34
dumarchie Okay, I'll try and trigger the panic again with both disabled. 16:40
Still panic. I added the `gdb` output to 16:48
It looks like the panic is triggered by `prefix:<⚛>`, but what do I know?=# 16:51
Let me know it there's anything else I can do. 16:56
16:58 zakharyas joined 17:09 sena_kun joined 17:10 Altai-man left 17:13 brrt left 17:35 leont joined
lizmat Looking at 17:35
wouldn't it make sense to change the < 0 condition to < 160 ? 17:36
so it can bypass the large switch statement because in the end it would end up in the default anyway ?
hmmm.. I guess not: /* If it's within ASCII just pass it through */ 17:40
if (0 <= codepoint && codepoint <= 127) {
19:00 vesper11 left 21:08 Altai-man joined
Geth MoarVM/modern-raku-script-extensions: 16 commits pushed by (Elizabeth Mattijsen)++
MoarVM: lizmat++ created pull request #1401:
Modern raku script extensions
21:11 sena_kun left 21:51 patrickb joined 22:15 zakharyas left 22:41 Altai-man left 22:42 patrickb left 22:53 dumarchie left