|
Parrot 6.10.0 "New Caledonian lorikeet" | parrot.org/ | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Fridays 13:30 UTC Set by moderator on 21 November 2014. |
|||
|
01:03
particle1 joined
02:43
kid51 joined
03:14
kid51 joined
03:17
kid51_ joined
03:24
kid51_ joined
03:32
kid51_ joined
04:25
sorear joined
07:21
FROGGS joined
|
|||
| dalek | rrot/smoke-me/rpa-offset-gh1152: 63ed0d4 | rurban++ | src/pmc/ (2 files): [pmc] rpa: fix visit we need an offset for visit, we cannot use the fixedpmcarray method. |
11:27 | |
| rrot/smoke-me/rpa-offset-gh1152: 6b4db01 | rurban++ | / (2 files): [pmc] rpa: fix resize with offset optimization we moved the wrong items, and had wrong slack overwrites. Add testcases. |
|||
| rrot/smoke-me/rpa-offset-gh1152: c8a7211 | rurban++ | / (4 files): [pmc]: various minor refactors * Access attr struct more directly. * Fix one syntax problem with UNLIKELY with non-std compilers. * Optimize *1.5 by using integer math. * Change rpa overallocation to use 1.5, not 2. This is better to handle free blocks. * Always fill slack space, as we extend rpa via set. In a better version we should fill the slack space only when needed: only on set. * Change push_{non-pmc} to use push_pmc and not set_keyed_int. push is offset optimized, set not. Still WIP. object was probably modified after being freed |
|||
|
11:38
travis-ci joined
|
|||
| travis-ci | parrot/parrot#1361 (smoke-me/rpa-offset-gh1152 - c8a7211 : Reini Urban): The build is still failing. | 11:38 | |
| Change view : github.com/parrot/parrot/compare/5...a7211520f9 | |||
| Build details : travis-ci.org/parrot/parrot/builds/42726894 | |||
|
11:38
travis-ci left
13:09
kid51 joined
|
|||
| dalek | rrot/smoke-me/rpa-offset-gh1152: b4e7b00 | rurban++ | / (4 files): [pmc]: rpa: optimize slack filling with PMCNULL Only fill the rest with PMCNULL when necessary, i.e. when we set an index > size, the intermediate indices between size and index-1. This saves a lot of slack filling code cycles. Error on delete oob. |
14:23 | |
| rrot/smoke-me/rpa-offset-gh1152: 6a1917f | rurban++ | src/pmc/resizablepmcarray.pmc: [pmc]: rpa: optimize delete_keyed_int on index=0 just bump the offset as with shift, else memmove en-block, not item by item in a loop. This passes now most rpa related tests, just reverse has a conflict with fpa. with reverse renamed to nreverse make test and benchmarking is possible. |
|||
|
14:41
travis-ci joined
|
|||
| travis-ci | parrot/parrot#1362 (smoke-me/rpa-offset-gh1152 - 6a1917f : Reini Urban): The build is still failing. | 14:41 | |
| Change view : github.com/parrot/parrot/compare/c...1917f00185 | |||
| Build details : travis-ci.org/parrot/parrot/builds/42744107 | |||
|
14:41
travis-ci left
|
|||
| dalek | rrot/smoke-me/rpa-offset-gh1152: c3742a5 | rurban++ | / (4 files): [ext] fix the remaining rpa reverse problem nqp ResizablePMCArray provided its own reverse method, which is now not needed anymore. We use the better pir method instead. Also found a syntax problem in parrot_old Remaining todos: t/op/lexicals.t (Wstat: 256 Tests: 54 Failed: 1) Failed test: 27 t/compilers/pge/p5regex/p5rx.t (Wstat: 0 Tests: 960 Failed: 7) Failed tests: 172, 420-421, 442-443, 869-870 t/src/extend_vtable.t (Wstat: 512 Tests: 134 Failed: 2) Failed tests: 34, 37 |
14:44 | |
|
15:05
travis-ci joined
|
|||
| travis-ci | parrot/parrot#1363 (smoke-me/rpa-offset-gh1152 - c3742a5 : Reini Urban): The build has errored. | 15:05 | |
| Change view : github.com/parrot/parrot/compare/6...742a5c0879 | |||
| Build details : travis-ci.org/parrot/parrot/builds/42746413 | |||
|
15:05
travis-ci left
15:55
Reini joined
|
|||
| Reini | rpa's almost ready. let's now see how much faster they are. should be faster than qrpa's. | 15:56 | |
| dalek | rrot/smoke-me/rpa-offset-gh1152: ec0b57a | rurban++ | / (4 files): [core] apifify -t20 as PARROT_TRACE_ARRAY_STATE_FLAG Only available without --optimize, i.e. #ifndef NDEBUG |
15:59 | |
| rrot/smoke-me/rpa-offset-gh1152: 8005345 | rurban++ | docs/ (2 files): [docs] Update stability and memory_internals stability was written in 2007. Just the deprecation cycle needed to be updated. memory_internals talks as if mark&sweep is the only GC. Improve that. |
|||
| rrot: f413a4b | rurban++ | t/compilers/pct/complete_workflow.t: [test] fix GLOB* temp file creation regression from commit caba76f589a9aa. mixed up handle and filename, as run_command actually requires a name and not a handle for STDERR. |
|||
| rrot: 6e5ea61 | rurban++ | docs/ (2 files): [docs] Update stability and memory_internals stability was written in 2007. Just the deprecation cycle needed to be updated. memory_internals talks as if mark&sweep is the only GC. Improve that. |
|||
| Heuristic branch merge: pushed 20 commits to parrot/smoke-me/rpa-offset-gh1152 by rurban | 16:06 | ||
| Reini | 10% slower, huh? | 16:08 | |
| nope, too high load. on my dedicated bench machine it's -22.96% | 16:16 | ||
| 20% was expected. good | |||
| maybe another −20% with ms2 and I call it done | 16:19 | ||
| yep, and perl6 is also faster. good | 16:28 | ||
| lot of shift + push in the parser engines | 16:29 | ||
|
16:31
travis-ci joined
|
|||
| travis-ci | parrot/parrot#1364 (master - 6e5ea61 : Reini Urban): The build passed. | 16:31 | |
| Change view : github.com/parrot/parrot/compare/a...5ea6188a89 | |||
| Build details : travis-ci.org/parrot/parrot/builds/42755786 | |||
|
16:31
travis-ci left
16:42
travis-ci joined
|
|||
| travis-ci | parrot/parrot#1364 (smoke-me/rpa-offset-gh1152 - 8005345 : Reini Urban): The build has errored. | 16:42 | |
| Change view : github.com/parrot/parrot/compare/c...05345686fa | |||
| Build details : travis-ci.org/parrot/parrot/builds/42755797 | |||
|
16:42
travis-ci left
|
|||
| Reini | And now I can repro the performance win on my mac also: 13.59% faster | 16:54 | |
| this is big | |||
| but there's still a double free or corruption | 16:55 | ||
| in splice | 16:56 | ||
|
17:06
travis-ci joined
|
|||
| travis-ci | parrot/parrot#1365 (smoke-me/rpa-offset-gh1152 - 3e1e5df : Reini Urban): The build has errored. | 17:06 | |
| Change view : github.com/parrot/parrot/compare/8...1e5df0c9c7 | |||
| Build details : travis-ci.org/parrot/parrot/builds/42756804 | |||
|
17:06
travis-ci left
|
|||
| dalek | rrot/smoke-me/rpa-offset-gh1152: bd63ea7 | rurban++ | src/pmc/resizablepmcarray.pmc: [pmc] rpa: fix a resize edgecase with size=0 avoid negative offsets. repro: valgrind ./parrot_old -t20 parrot-nqp.pbc --target=pir compilers/data_json/JSON.nqp |
18:28 | |
|
18:58
travis-ci joined
|
|||
| travis-ci | parrot/parrot#1366 (smoke-me/rpa-offset-gh1152 - bd63ea7 : Reini Urban): The build failed. | 18:58 | |
| Change view : github.com/parrot/parrot/compare/3...63ea782e0d | |||
| Build details : travis-ci.org/parrot/parrot/builds/42772372 | |||
|
18:58
travis-ci left
19:07
FROGGS joined
|
|||
| Reini | another 2.5% faster, better shift | 21:55 | |
|
22:08
lauleon joined
|
|||
| dalek | rrot/smoke-me/rpa-offset-gh1152: 2c49485 | rurban++ | src/pmc/resizablepmcarray.pmc: [pmc] rpa: fix invalid read in delete_keyed_int, off by one with (1,9,10) and key=1 we read one too far. repro: valgrind ./parrot_old -t20 t/compilers/pge/p5regex/p5rx.t |
22:08 | |
| rrot/smoke-me/rpa-offset-gh1152: e4fe79c | rurban++ | src/pmc/resizablepmcarray.pmc: [pmc] rpa: shift optim, use data struct directly do not use seperate SELF and temp values for the attrs. 2.40% faster on my mac, %ebp addressing for data. |
|||
| rrot/smoke-me/rpa-offset-gh1152: 73aacbb | rurban++ | src/pmc/resizablepmcarray.pmc: [pmc] rpa: internal rename trace macros add TRACE_RPAdata |
|||
|
22:24
travis-ci joined
|
|||
| travis-ci | parrot/parrot#1367 (smoke-me/rpa-offset-gh1152 - 73aacbb : Reini Urban): The build is still failing. | 22:24 | |
| Change view : github.com/parrot/parrot/compare/b...aacbb414a9 | |||
| Build details : travis-ci.org/parrot/parrot/builds/42796246 | |||
|
22:24
travis-ci left
22:52
Reini left
22:53
Reini joined
22:54
Reini joined
23:06
freudsfeud joined
23:30
kid51 joined
|
|||