00:26 vendethiel joined 00:28 colomon joined 00:58 vendethiel joined
diakopter skipped the laggard May 2016 android update and jumped straight to N beta 01:32
re-precompiling apps is the longest step 01:39
01:48 ilbot3 joined 01:52 vendethiel joined 01:57 zakharyas joined 02:03 tokuhirom joined 03:09 vendethiel joined 04:19 tokuhirom joined 04:38 vendethiel joined 05:07 vendethiel joined 05:42 vendethiel joined 05:51 domidumont joined 05:55 domidumont joined 06:19 domidumont joined 06:20 domidumont joined 06:24 vendethiel joined 06:29 tokuhirom joined 06:59 vendethiel joined 07:17 BinGOs joined 07:50 vendethiel joined 08:07 domidumont joined 08:55 domidumont joined 09:08 domidumont joined 09:30 vendethiel joined 10:18 vendethiel joined 11:21 brrt joined
brrt good * 11:22
jnthn o/ brrt 11:26
brrt \o jnthn
did you know.. that the webkit git repo is over 5gb? 11:27
11:27 vendethiel joined
jnthn No, but...wow :) 11:29
nwc10 what makes it so big? Source code? Or binaries? 11:30
brrt not sure...
code it seems 11:32
JimmyZ so they wrote 5gb code by hande? 11:33
hand
brrt JavascriptCore is 64MB checked-out source 11:34
WebCore is 163mb
it seems that way
moritz several developers working full-time in a verbose-ish language 11:35
yes, seems legit.
brrt 850mb on layouttests
307 mb on websites, too 11:36
anyway, it's interesting the b3 backend, but i think the expr jit is and will be much, much smaller 11:38
JimmyZ Is it luajit 2.0 using another jit api instead of dynasm? 11:39
brrt yes 11:40
not sure if i understand the question 100%, but luajit2 does its own machine code generation, it only uses dynasm to compile an interpreter 11:41
JimmyZ I try to find the code , but most is dynasm
brrt i.e. at boot, dynasm is used to compile the interpreter, and the real JIT then compiles the subroutines without dynasm 11:42
JimmyZ I didn't see where jit api code is
brrt lemessee... 11:43
JimmyZ repo.or.cz/luajit-2.0.git/blob/HEAD...c/lj_asm.c ? 11:44
brrt yeah, that's part of it 11:45
JimmyZ so it is lj_asm prefix
brrt but that's just the assembly 11:46
JimmyZ i see repo.or.cz/luajit-2.0.git/blob/HEAD..._asm_x86.h for X86 11:48
12:06 vendethiel joined 12:24 tokuhirom joined 12:45 brrt joined 13:01 domidumont joined 13:07 vendethiel joined
brrt fwiw, i wouldn't dare to lift that code out of luajit 13:20
13:25 tokuhirom joined
timotimo heh. 13:26
brrt and i have no particular impetus to remove our dependency on dynasm 13:27
13:38 vendethiel joined
brrt timotimo: the bytecode in the MVMJitCode structure is indeed a static bit 13:38
13:42 lizmat joined
dalek Heuristic branch merge: pushed 218 commits to MoarVM/even-moar-jit by bdw 13:50
brrt that merged even-moar-jit, but i've noticed there are some problems with rakudo still...
eh, reframe, into even-moar-jit 13:51
13:51 mst joined
brrt ah, a segv in building CORE.setting 13:52
great
i'm jumping out somewhere to the void.. 13:56
foundit
dalek arVM/even-moar-jit: 5b96b18 | brrt++ | src/jit/x64/emit.dasc:
Don't jump to nonexisting labels

The difference between ->out and ->exit don't exist anymore because we don't use a return value from MVM_jit_enter_code.
14:01
brrt and with that, even-moar-jit is up to date with reframe 14:02
14:09 vendethiel joined
brrt recently realized that pre-coloring would have to happen prior to the other register allocator steps 14:12
what's in a name
14:50 vendethiel joined 15:01 lizmat joined 15:15 domidumont joined, domidumont left 16:03 vendethiel joined 16:58 domidumont joined 17:22 vendethiel joined 17:27 tokuhirom joined 17:54 lizmat joined 18:05 vendethiel joined 18:26 avar joined 18:41 tokuhirom joined 18:54 vendethiel joined 20:22 lizmat joined 20:40 brrt joined 20:43 Ven_ joined 20:49 tokuhirom joined