Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2015-09-18

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

All times shown according to UTC.

Time Nick Message
00:55 tokuhiro_ joined #moarvm
03:50 vendethiel joined #moarvm
07:18 FROGGS joined #moarvm
07:38 brrt joined #moarvm
07:39 brrt good UGT
07:41 nwc10 good UGT, brrt
07:42 brrt hey
07:43 brrt you guys fixed it? :-o
07:43 brrt jnthn++
07:44 brrt oh, the 6th arg explains why it didn't go wrong for jnthn
07:45 brrt 6th win32 arg is stack-passed, on unix it's register-passed
07:47 brrt FROGGS: MVM_JIT_REG_VAL and MVM_JIT_REG_VAL_F are really the same op. MVM_JIT_LITERAL is meant for smallish integers, MVM_JIT_LITERAL_64 and friends not
07:48 brrt jnthn++
07:51 brrt we can actually replace that, i think, with a fits_in_32_bit check
08:10 FROGGS :o)
08:20 vendethiel joined #moarvm
08:27 brrt anyway, thanks for fixing that :-)
09:06 * brrt can, somewhat uselessly, confirm that it works for him too
09:07 FROGGS hehe, that's still nice to know :o)
09:07 FROGGS brrt: btw, what are your next steps? fix more bugs?
09:09 brrt ehm, yes, i guess so
09:09 brrt there are still vast maps of NYI in the JIT compiler that should be implemented
09:09 brrt but first and foremost i want to remove bugs
09:10 brrt there is a bug somewhere in the new JIT that seems to return the wrong value in a routine
09:10 brrt i'm going to see if i can golf that
09:10 brrt and why it's broken
09:11 FROGGS brrt: I can't help you in any area, right?
09:11 brrt hmmm
09:11 brrt well, how good are you with GDB? :-P
09:11 brrt i can tell you what i know is going wrong, and maybe you have an idea
09:12 brrt actually, the truth is, i know very little of what is going wrong yet
09:12 FROGGS well, fell free to paste a test case
09:13 brrt hmm
09:13 FROGGS feel*
09:13 brrt my 'testcase' is basically 'run moar in gdb. see where it blows up'
09:14 FROGGS how do you run moar? nqp-m -e'' ?
09:15 brrt https://gist.github.com/bdw/56dd8b724f55e79183ce
09:15 brrt is what i do
09:15 brrt MVM_JIT_EXPR_ENABLE is what is necessary to enable the expression compiler
09:23 FROGGS looks like it blows up here in the elems op
09:24 brrt aye
09:24 brrt but why?
09:24 brrt i'd have to look at the code that is being compiled
09:24 FROGGS valgrind is unhappy too
09:24 brrt oh, that helps
09:24 brrt about what?
09:25 FROGGS https://gist.githubusercontent.com/FROGGS/9dc924b7a89239b18adb/raw/f679483a43cc9a7fa80bbe46442458285dffd602/gistfile1.txt
09:27 brrt aw, that's awful
09:29 FROGGS it is
09:29 brrt hmm
09:29 brrt i wonder why
09:42 brrt FROGGS++ that's a huge help though
09:42 brrt (my suspicion is that valgrind makes it crash because it zeroes out memory?)
09:50 brrt oh, hmm
09:50 brrt you may want to do a make reconfig before you run even-moar-jit?
09:50 brrt that's basically the generic advice in any case :-)
09:51 FROGGS I recompiled moar, yes
09:51 brrt i'm not getting a crash, though
09:51 brrt but still
09:51 FROGGS not nqp though, but I think that's no needed
09:51 FROGGS hmmm
10:01 FROGGS one of the Invalid read of size 8 seems to be about info->tile->right_sym at tile.c:197
10:03 brrt aye
10:03 brrt i see why that should be so
10:05 FROGGS do tell
10:07 brrt it's realloced in assign_tile when we need more space due to tile resolution
10:07 brrt however
10:08 brrt when i remove that
10:08 brrt it still hapens
10:08 brrt happens
10:12 dalek MoarVM/even-moar-jit: c5733aa | brrt++ | src/jit/ (2 files):
10:12 dalek MoarVM/even-moar-jit: Don't cache from realloced array
10:12 dalek MoarVM/even-moar-jit:
10:12 dalek MoarVM/even-moar-jit: This silences valgrind, but does not solve the wrong-value-return
10:12 dalek MoarVM/even-moar-jit: bug
10:12 dalek MoarVM/even-moar-jit: review: https://github.com/MoarVM/MoarVM/commit/c5733aaf46
10:15 FROGGS yeah, the only oddness thath stays is this:
10:15 FROGGS ==10392== Jump to the invalid address stated on the next line
10:15 FROGGS ==10392==    at 0x0: ???
10:15 FROGGS ==10392==    by 0x4F9B2EA: MVM_interp_run (interp.c:2603)
10:17 brrt hmmmm
10:20 FROGGS what happens when you do that? memcpy(88582584, 88582584, 40)
10:20 FROGGS ohh wait
10:21 FROGGS nvm, that was my stupidity
10:23 brrt where should i have done it? :-)
10:23 FROGGS no havn't :o)
10:24 FROGGS lunch &
10:25 * brrt also lunch &
10:50 vendethiel joined #moarvm
10:55 colomon joined #moarvm
11:17 brrt joined #moarvm
11:52 FROGGS joined #moarvm
12:01 brrt joined #moarvm
12:24 brrt ok, i'm now diffing the code pieces
12:28 FROGGS code pieces of what?
12:28 FROGGS generated code?
12:28 FROGGS old jit vs new jit?
12:32 brrt aye
13:05 brrt something seems to be pretty wrong...
13:10 [Coke] new jit, same as the old jit.
13:10 rarara joined #moarvm
13:10 brrt new jit, worse than the old jit :-(
13:11 [Coke] sorry, that was a reference to the band the Who, not your coding skills. :)
13:13 Peter__R joined #moarvm
13:16 brrt :-)
13:21 brrt i wonder what the JIT thought it was doing
13:21 FROGGS O.o
13:29 colomon joined #moarvm
13:31 brrt there is a call in the middle of that code
13:32 brrt and it's weird
13:38 brrt hmmm
13:38 brrt maybe it's the write barier i'm looking at
14:12 FROGGS joined #moarvm
14:15 FROGGS o/
15:08 synbot6 joined #moarvm
15:43 timotimo o/
16:00 Peter_R joined #moarvm
16:24 vendethiel joined #moarvm
16:28 Peter_R joined #moarvm
17:12 Peter_R joined #moarvm
17:31 Peter_R joined #moarvm
17:38 Peter_R joined #moarvm
17:54 Peter_R joined #moarvm
18:00 Peter_R joined #moarvm
19:27 zakharyas joined #moarvm
19:41 FROGGS joined #moarvm
22:57 Peter_R joined #moarvm

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