|
github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today Set by moderator on 6 May 2018. |
|||
|
00:17
statisfiable6 joined
|
|||
| MasterDuke | should the gc_seq_number ever be -1 (if there haven't been enough GCs for it to wrap around)? | 01:43 | |
|
01:56
ilbot3 joined
|
|||
| moderator | github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today | ||
|
02:45
Kaypie joined
03:25
nativecallable6 joined
05:38
zakharyas joined
06:21
nwc10 joined
06:33
quotable6 joined
06:52
domidumont joined
06:59
domidumont joined
07:03
robertle joined
07:57
jsimonet joined
08:13
zakharyas joined
09:22
zakharyas joined
|
|||
| Geth | MoarVM: samcv++ created pull request #855: Implement a optimized memmem 32 bit search 4-15x faster |
10:07 | |
|
10:19
domidumont joined
|
|||
| timotimo | MasterDuke: can you give me the script that gives you the 0 vs 1 discrepancy? | 10:35 | |
| Geth | MoarVM: a519028373 | (Timo Paulssen)++ | src/profiler/instrument.c commit a missing break statement |
10:45 | |
| MoarVM: a91b3d465a | (Timo Paulssen)++ | src/profiler/instrument.c insert prof_allocated in the exact right place must be after all PHIs, but since insert_ins puts the instruction after the one given in the arguments, we have to take a step back first. |
|||
| MoarVM: 37d5b9850d | (Samantha McVey)++ | src/strings/ops.c Factor out the memmem wrapper for searching Grapheme32 blobs Also rename some variables in MVM_string_char_at_in_string to make it more clear what the haystack is. |
10:46 | ||
| MoarVM: e71ba73815 | (Samantha McVey)++ | 4 files Implement a optimized memmem 32 bit search 4-15x faster This is a modification of the memmem used in FreeBSD to allow us to quickly search 32 bit strings. This is much more efficient than searching by byte since we can skip every 4 bytes instead of every 1 byte, as well as the fact that in a 32 bit integer, some of the bytes will be empty, making it even less efficient. ... (6 more lines) |
|||
| MoarVM: 6e2b54c7d1 | (Samantha McVey)++ (committed using GitHub Web editor) | 4 files Merge pull request #855 from samcv/changes Implement a optimized memmem 32 bit search 4-15x faster |
|||
|
11:06
travis-ci joined
|
|||
| travis-ci | MoarVM build failed. Timo Paulssen 'insert prof_allocated in the exact right place | 11:06 | |
| travis-ci.org/MoarVM/MoarVM/builds/375831423 github.com/MoarVM/MoarVM/compare/8...1b3d465afb | |||
|
11:06
travis-ci left
|
|||
| Geth | MoarVM: eefd9a4362 | (Timo Paulssen)++ | src/profiler/instrument.c fix off-by-one in profile's GC dumping |
12:02 | |
| samcv | timotimo: wondering if you can take a quick look at this code github.com/samcv/MoarVM/commit/f8f...a7a9bc5R50 | 12:27 | |
| almost certain it's right and all tests pass. but i'm pretty sure i'm ok casting and reading the blob as 64 bit integers moving 32 bits down at a time | 12:28 | ||
| timotimo | oh, you mean with regards to endianness? | ||
| lizmat | no, just in general ( samcv's answer to me here verbally) | 12:30 | |
| samcv | timotimo: well that could be too, but there shouldn't by any endianess issues since i'm comparing like and like | 12:32 | |
| and just treating it as raw data | |||
| though i may want to get ahold of a 32 bit computer and make sure reading 64 bits at a time isn't slower on that system. | 12:33 | ||
|
12:34
travis-ci joined
|
|||
| travis-ci | MoarVM build failed. Samantha McVey 'Implement memmem_two_uint32, up to 45% faster index of short 32bitstr | 12:34 | |
| travis-ci.org/samcv/MoarVM/builds/375867773 github.com/samcv/MoarVM/compare/87...f4b9c1b309 | |||
|
12:34
travis-ci left
|
|||
| samcv | ah nqp wanted a newer version i guess | 12:34 | |
|
12:43
travis-ci joined
|
|||
| travis-ci | MoarVM build failed. Samantha McVey 'Implement memmem_two_uint32, up to 45% faster index of short 32bitstr | 12:43 | |
| travis-ci.org/samcv/MoarVM/builds/375871214 github.com/samcv/MoarVM/compare/f8...00f7892a19 | |||
|
12:43
travis-ci left
13:22
ilbot3 joined
|
|||
| moderator | github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today | ||
|
13:57
Kaypie joined
14:11
committable6 joined,
quotable6 joined,
undersightable6 joined
|
|||
| Geth | MoarVM/master: 10 commits pushed by (Bart Wiegmans)++, (Elizabeth Mattijsen)++ review: github.com/MoarVM/MoarVM/compare/e...92140315de |
16:10 | |
| MoarVM: b38df02a86 | (Samantha McVey)++ | 2 files Ensure `make realclean` cleans 3rdparty/cmp The `cmp` lib had not been added to Makefile.in and setup.pm yet. Make sure make realclean deletes the object file and runs make clean in the directory. |
16:29 | ||
| timotimo | oops :S | 16:36 | |
| build systems, eh? who understands 'em! | |||
|
16:36
lizmat joined
|
|||
| samcv | also. i'm having this issue unrelated to that | 16:38 | |
| timotimo: heh | |||
| fatal: failed to update remote for submodule '3rdparty/dyncall' | |||
| and for some reason it has no remote. so i just delete the folder and then git checkout 3rdparty/dyncall and run ./Configure.pl again and it works. but then the next run i get that error again | |||
| i can ignore it, and things still build but | |||
| timotimo | does "git submodule sync" do anything for that? | 16:42 | |
|
16:47
dogbert17 joined
16:50
statisfiable6 joined
|
|||
| lizmat | .tell brrt merged your jit opcodes, test-t ~2.5% faster :-) | 16:56 | |
| yoleaux | lizmat: I'll pass your message to brrt. | ||
| samcv | timotimo: no | 16:57 | |
| error: could not lock config file /home/samantha/git/MoarVM/.git/modules/3rdparty/dyncall/config: File exists | |||
| i get this message though | |||
| ah i removed /home/samantha/git/MoarVM/.git/modules/3rdparty/dyncall/config.lock and it did work i think | |||
| \\o/ yay | |||
| lizmat | .tell brrt looks like these are new opcodes to target: hllizefor, ctxlexpad, threadlockcount, capturenamedshash. the first 3 of these seem pretty generic for any type of async code running | 16:58 | |
| yoleaux | lizmat: I'll pass your message to brrt. | ||
| timotimo | weird, is any git process still running? | ||
| samcv | timotimo: nope. i think it got left behind | ||
|
17:36
lizmat joined
18:27
Geth joined
18:47
unicodable6 joined
|
|||
| lizmat | perl6 t/spec/S01-perl-5-integration/basic.rakudo.moar | 18:57 | |
|
19:20
brrt joined
|
|||
| brrt | \\o | 19:20 | |
| yoleaux | 16:56Z <lizmat> brrt: merged your jit opcodes, test-t ~2.5% faster :-) | ||
| 16:58Z <lizmat> brrt: looks like these are new opcodes to target: hllizefor, ctxlexpad, threadlockcount, capturenamedshash. the first 3 of these seem pretty generic for any type of async code running | |||
| brrt | thanks! | 19:30 | |
| .ask jnthn given that MVMHLLConfig does *not* start with an MVMCollectable header, can I safely assume that therefore, it doesn't ever move? | |||
| yoleaux | brrt: I'll pass your message to jnthn. | ||
| brrt | I suppose I can... | ||
| given that i already do for the original hllize | 19:39 | ||
| jnthn | brrt: Yes, it's just malluc | 19:40 | |
| yoleaux | 19:30Z <brrt> jnthn: given that MVMHLLConfig does *not* start with an MVMCollectable header, can I safely assume that therefore, it doesn't ever move? | ||
| brrt | hmm, but in this case, the hll name comes from a register, and i don't know if it is a proven constant | 19:42 | |
| timotimo | at least in the template jit we do have facts available | 19:43 | |
| like "known value" | |||
| jnthn | If it is constant then yeah, the fact will be set :) | 19:46 | |
| brrt | hmm | ||
| timotimo | we use that for reprop devirt, remember? | 19:47 | |
| brrt | i do | ||
| i'm wondering if i can get away with only compling hllizefor if we can get the hll name as a constant | 19:48 | ||
| jnthn | Probably it'd hit nearly all the cases | 19:49 | |
| brrt | I think so too | 19:51 | |
| anyway, i'm off for tonight | |||
|
19:54
Kaypie joined
20:29
robertle_ joined
20:48
jnthn1 joined
20:57
ilmari[m] joined
|
|||
| lizmat | and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/05/07/...-solution/ | 21:35 | |
|
21:38
ilmari[m] joined
|
|||
| MasterDuke | timotimo: nice quick fix on the profile issue | 21:48 | |
| timotimo | well, your pinpointing made it easy | ||
| jnthn | lizmat++ # weekly | 21:50 | |
|
21:53
nativecallable6 joined,
unicodable6 joined
|
|||
| MasterDuke | timotimo: i'm sort of annoyed i didn't see the fix myself. i looked for gc_seq_number, but then didn't follow what was done with gc_seq_num. but my excuse is it was late and i was tired... | 21:55 | |
| timotimo | that's just fine | 21:56 | |