Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2015-01-15

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

All times shown according to UTC.

Time Nick Message
00:05 ggoebel111111117 joined #moarvm
01:54 colomon joined #moarvm
03:58 colomon joined #moarvm
04:27 vendethiel joined #moarvm
05:09 vendethiel joined #moarvm
05:43 vendethiel joined #moarvm
06:10 vendethiel joined #moarvm
06:56 vendethiel joined #moarvm
07:24 dalek MoarVM: e81b97d | (Jimmy Zhuo)++ | src/spesh/graph.h:
07:24 dalek MoarVM: reorder struct in MVMSpeshOperand
07:24 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/e81b97dc74
07:24 dalek MoarVM: ed63097 | (Jimmy Zhuo)++ | src/ (9 files):
07:24 dalek MoarVM: avoid call MVM_callsite_get_common twice
07:24 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/ed6309714f
07:29 dalek MoarVM: 2bce4e5 | (Jimmy Zhuo)++ | src/core/coerce.c:
07:29 dalek MoarVM: removed unused code
07:29 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/2bce4e5d42
07:52 FROGGS joined #moarvm
07:56 zakharyas joined #moarvm
08:03 jnthn JimmyZ: e81b97d is senseless, not to mention weird
08:03 jnthn The notation in the literature is always orig with i as a subscript.
08:05 nwc10 jnthn: not knowing that, it looks initially reasonable, if one assumes that a reasonable approach to avoiding alignment holes on structures is to order them with biggest things first (unless some other reason pre-empts that)
08:06 nwc10 also, why is SSA-computed version signed?
08:30 brrt joined #moarvm
08:32 vendethiel joined #moarvm
08:51 rurban joined #moarvm
09:00 jnthn nwc10: Yeah but there are 64-bit things in the union.
09:00 jnthn So the hole is not making things any worse.
09:01 jnthn Signed? Good question, quite possibly because -1 is used as a sentinel...or was going to be.
09:04 nwc10 not sure whether one ends up with more bugs as a result of negative values accidentally becoming array subscripts, or unsigned "all bits set" not being transfered "correclty" to larger integer types
09:05 jnthn I should check if it ended up using the -1 in such a way
09:05 jnthn If not it can become unsigned
09:09 JimmyZ jnthn: Do I still need to revert it?
09:09 jnthn It just feels like a really pointless patch :/
09:09 jnthn It achieves nothing.
09:09 jnthn Except wasting my time reviewing it.
09:10 JimmyZ sorry :(
09:18 jnthn ed6309714f is sensible overall, though I'd have preferred to see the various extracting of tc->cur_frame to a local in a separate patch.
09:19 jnthn In most cases it's "obviously correct", but in some places ti's in code that is messing with which frame is the current frame. I think it's correct where it's been done, however.
09:29 JimmyZ jnthn: Thanks  for review. I did take a care of right time about extracting of tc->cur_frame.
09:30 jnthn Yeah. It's one of those times where it's borderline whether it's better to extract it, or to do the "obviously correct" thing and leave the compiler's CSE to try and take care of it.
09:31 jnthn (In the unwind/exception run cases where such a refactor could be delicate, I mean)
09:34 JimmyZ I objdumped libmoar.so and found gcc doesn't do CSE to tc->cur_frame->xxxx =  yyyy; the tc->cur_frame part. at least O1 doesn't
09:35 JimmyZ -O1
09:35 JimmyZ so I did some extracting
09:39 kjs_ joined #moarvm
10:05 zakharyas joined #moarvm
11:38 colomon joined #moarvm
14:24 woolfy left #moarvm
14:27 woolfy joined #moarvm
17:10 rurban joined #moarvm
18:12 tgt joined #moarvm
18:51 kjs_ joined #moarvm
19:16 TimToady m: for <aa aba> xx 50 { if not m/b/ { print "."; next }; print "!"; }; say "";
19:16 camelia rakudo-moar 9c74ab: OUTPUT«.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!............................................␤»
19:17 TimToady spesh bug
19:17 TimToady but not jit, inline, or osr
19:19 TimToady removing the not fixes it
19:19 TimToady changing the control flow to if/else fixes it
19:19 TimToady using unless instead of 'if not' fixes it
19:20 TimToady it's quite picky :)
19:21 TimToady well, you can change the 'not' to '!' and it still breaks, anyway
19:22 TimToady m: for <aa aba> xx 50 { if m/b/ { print "."; next }; print "!"; }; say "";
19:22 camelia rakudo-moar 9c74ab: OUTPUT«!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.␤»
19:22 TimToady m: for <aa aba> xx 50 { if not not m/b/ { print "."; next }; print "!"; }; say "";
19:22 camelia rakudo-moar 9c74ab: OUTPUT«!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!␤»
19:23 TimToady hmm
19:23 TimToady m: for <aa aba> xx 50 { if not not not m/b/ { print "."; next }; print "!"; }; say "";
19:23 camelia rakudo-moar 9c74ab: OUTPUT«.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!............................................␤»
19:23 TimToady m: for <aa aba> xx 50 { if not not not !m/b/ { print "."; next }; print "!"; }; say "";
19:23 camelia rakudo-moar 9c74ab: OUTPUT«!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!.!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!␤»
19:25 kjs_ joined #moarvm
19:46 FROGGS joined #moarvm
19:58 rurban joined #moarvm
20:03 kjs_ joined #moarvm
20:33 * timotimo ducks
21:01 * FROGGS quacks
21:02 * lizmat doctor
21:14 vendethiel joined #moarvm
21:36 tgt TimToady: Reverting https://github.com/rakudo/rakudo/commit/661439a595e6b65a306ac01d445cb38966382ef2 fixes that issue. No idea why, I just let git bisect lose on it.
21:37 timotimo i KNEW it'd be my fault :)
21:51 lizmat FWIW, I would think only the Mu candidates need reverting
21:52 lizmat timotimo: something should have ticked you off, because you could optimize that further by removing the Bool candidates :-)
21:52 timotimo oh
21:53 * lizmat is a great fan of lining out similar code
21:53 lizmat you would have seen it then
22:02 jnthn I'd probably not revert the Rakudo patch if it's clearly a VM-level issue
22:03 jnthn Should fix the spesh bug instead.
22:04 lizmat fg
22:05 lizmat oops
22:36 kjs_ joined #moarvm
22:46 nebuchad` joined #moarvm
22:47 moritz_ joined #moarvm
22:47 Util_ joined #moarvm
22:49 bonsaikitten joined #moarvm
22:49 btyler joined #moarvm
22:51 BinGOs_ joined #moarvm
22:51 sivoais_ joined #moarvm
22:53 ggoebel111111117 joined #moarvm
22:53 tadzik joined #moarvm
22:53 ingy joined #moarvm
22:53 daxim joined #moarvm
22:53 jnthn joined #moarvm
22:53 leedo_ joined #moarvm
22:54 BinGOs joined #moarvm
22:54 harrow` joined #moarvm
22:54 pyrimidine joined #moarvm
22:54 vendethiel- joined #moarvm
23:03 ilbot3 joined #moarvm
23:03 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
23:05 leedo joined #moarvm
23:05 camelia joined #moarvm
23:05 flussence joined #moarvm
23:05 dalek joined #moarvm
23:05 JimmyZ joined #moarvm
23:06 btyler joined #moarvm
23:07 arnsholt joined #moarvm
23:09 tadzik joined #moarvm
23:10 ingy joined #moarvm
23:12 dalek joined #moarvm
23:12 moritz joined #moarvm
23:12 jnthn joined #moarvm
23:12 nwc10 joined #moarvm
23:12 pyrimidine joined #moarvm
23:12 harrow` joined #moarvm
23:12 ggoebel111111118 joined #moarvm
23:12 sivoais joined #moarvm
23:13 18VABY4HM joined #moarvm
23:13 daxim joined #moarvm
23:13 TimToady joined #moarvm
23:13 FROGGS joined #moarvm
23:13 synopsebot joined #moarvm
23:23 77CAABZU1 joined #moarvm
23:23 1JTAABMSD joined #moarvm
23:35 64MABRKDD joined #moarvm
23:35 7YUAAFL3J joined #moarvm
23:46 sivoais joined #moarvm
23:49 sivoais_ joined #moarvm

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