timotimo heyo brrt, have you accidentally rebased x64-dynamic-registers onto nfa_to_statelist at some point? 00:24
timotimo is that branch even still interesting? 00:30
timotimo goes through moar's branches to see if anything interesting fell between the cracks
i'd still like to get early-death-percentages up and running again :\ 00:31
timotimo i put a debug statement in that's supposed to help me find out where this one spesh opt (iscont and iscont_[ins] to constant values) triggers 01:15
but the output doesn't get interleaved properly
anyway, there was a lonely little branch that implemented iscont in spesh and i added the _i, _n, and _s parts and will spectest now 01:17
timotimo turns out that optimization isn't good enough yet. 02:11
as it breaks many a spectest
i should try to get some rest now 02:12
o/
nwc10 good *, #moarvm 07:34
FROGGS o/ 09:06
nwc10 \o 09:09
dalek arVM/tiler-handle-args: 37f4764 | brrt++ | src/jit/ (4 files):
Move branch labelling to analysis time

As part of my effort to simplify the compilation of tiles I move the branch labelling to the tree analysis stage. Labels are now stored relative to the tree, meaning they won't mind if the tree is rearranged.
09:23
MoarVM/tiler-handle-args: 0b227eb | brrt++ | docs/jit/tiles.md:
MoarVM/tiler-handle-args: Add tile documentation
MoarVM/tiler-handle-args:
MoarVM/tiler-handle-args: During the writing of this documentation it occured to me that
MoarVM/tiler-handle-args: 'constant' (i.e. non-symbolic arguments) can be loaded in the
arVM/even-moar-jit: 37f4764 | brrt++ | src/jit/ (4 files):
Move branch labelling to analysis time

As part of my effort to simplify the compilation of tiles I move the branch labelling to the tree analysis stage. Labels are now stored relative to the tree, meaning they won't mind if the tree is rearranged.
MoarVM/even-moar-jit: 0b227eb | brrt++ | docs/jit/tiles.md:
MoarVM/even-moar-jit: Add tile documentation
MoarVM/even-moar-jit:
MoarVM/even-moar-jit: During the writing of this documentation it occured to me that
MoarVM/even-moar-jit: 'constant' (i.e. non-symbolic arguments) can be loaded in the
brrt good * 09:26
timotimo: x64-dynamic-registers is the dynasm branch, not the moar branch
but I have been merging in master quite a few times over the last few weeks, so, possibly it is in there 09:27
jnthn o/ brrt 09:30
brrt \o jnthn
retupmoca jnthn, *: backtrace from my VS2015 stack overflow when building CORE.setting: gist.github.com/retupmoca/e5f9da34105352d7d947 17:26
hoelzro retupmoca++ # debugging VS2015 issues 17:28
retupmoca interesting - MVM_JIT_DISABLE=1 makes it build fine 17:38
so...VS2015 jitbug?
[Coke] oh, wait, moar jit is working, moar-nonjit is dead. 17:40