| … | |
… | |
| 2 | |
2 | |
| 3 | TODO: should explore PerlIO::coroaio (perl leaks like hell). |
3 | TODO: should explore PerlIO::coroaio (perl leaks like hell). |
| 4 | TODO: channel->maxsize(newsize)? |
4 | TODO: channel->maxsize(newsize)? |
| 5 | TODO: http://www.microsoft.com/msj/archive/s2ce.aspx |
5 | TODO: http://www.microsoft.com/msj/archive/s2ce.aspx |
| 6 | |
6 | |
|
|
7 | 6.06 Mon Aug 8 23:59:48 CEST 2011 |
|
|
8 | - cygwin unfortunately patches the stack at runtime, so we use the pthreads |
|
|
9 | backend, which is an order of magnitutde slower. unfortunately, cygwins |
|
|
10 | pthread implementation isn't very complete either, so allocate the stack |
|
|
11 | twice just to be sure. |
|
|
12 | (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS, |
|
|
13 | in its ithreaded perl - best recompile cygwin and use the 'w'indows |
|
|
14 | backend for much better performance. also disable ithreads for |
|
|
15 | even better performance...). |
|
|
16 | |
|
|
17 | 6.05 Thu Aug 4 21:36:36 CEST 2011 |
|
|
18 | - blush, condvar values would not be propagated from send to recv anymore |
|
|
19 | (reported by Chip Salzenberg). |
|
|
20 | - use exponential increase for the readline buffer length in |
|
|
21 | Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes. |
|
|
22 | |
|
|
23 | 6.04 Wed Aug 3 17:15:45 CEST 2011 |
|
|
24 | - use even more efficient and more compatible condvars for |
|
|
25 | compatibility to AnyEvent 6.x :) |
|
|
26 | - more inconsequential ecb.h updates. |
|
|
27 | |
|
|
28 | 6.03 Wed Aug 3 11:41:30 CEST 2011 |
|
|
29 | - change how Coro patches AnyEvent condvars for compatibility to |
|
|
30 | AnyEvent 6.x. |
|
|
31 | - update ecb.h, to no longer include <pthread.h> in case WinNT.h |
|
|
32 | hasn't been included. |
|
|
33 | |
|
|
34 | 6.02 Wed Jul 13 04:35:19 CEST 2011 |
|
|
35 | - "improve portability to Gentoo" - gentoo manages to put perl variables |
|
|
36 | in memory areas that are farther than 2gb apart, which the jit couldn't |
|
|
37 | handle and barfed. now it's just a bit slower on gentoo and similar |
|
|
38 | systems. |
|
|
39 | |
|
|
40 | 6.01 Sun Jul 3 12:31:14 CEST 2011 |
|
|
41 | - workarounds are good, but the test for whether pthreads are used |
|
|
42 | was not good. thisone should be better. |
|
|
43 | - check differently whether gcc generates cfi instructions itself. |
|
|
44 | |
|
|
45 | 6.0 Wed Jun 29 19:43:35 CEST 2011 |
|
|
46 | - INCOMPATIBLE CHANGE: unreferenced coro objects will now be |
|
|
47 | destroyed and cleaned up automatically (e.g. async { schedule }). |
|
|
48 | - implement a JIT compiler for part of the thread switch code, |
|
|
49 | which gives a 50% speed improvement on threaded perls, and |
|
|
50 | about 4% on non-thraeded perls (so threaded perls now finally |
|
|
51 | reach about half the speed of non-threaded perls). |
|
|
52 | - slightly modernise Coro::Intro, add section about rouse functions. |
|
|
53 | - avoid DEFSV and ERRSV, giving another 10% improvement |
|
|
54 | in thread switching. |
|
|
55 | - Coro::State->is_destroyed is now called is_zombie. |
|
|
56 | - implement a Coro->safe_cancel method that might fail, but |
|
|
57 | cancels in a "safer" way if it succeeds. |
|
|
58 | - add preliminary support for DEBUGGING perls. |
|
|
59 | - get rid of two hash-accesses when initialising a new Coro - this |
|
|
60 | speeds up coro creation by almost a factor of two. |
|
|
61 | - croak when a coro that is being cancelled tries to block |
|
|
62 | (e.g. while executing a guard block), instead of crashing or |
|
|
63 | deadlocking. |
|
|
64 | - use a more robust and also faster method to identify Coro::State |
|
|
65 | objects - speeds up everything a bit. |
|
|
66 | - implement Coro->cancel in XS for a 20% speed improvement, and to |
|
|
67 | be able to implement mutual cancellation. |
|
|
68 | - speed up context switches by a percent or two by more efficiently |
|
|
69 | allocating context stack entries. |
|
|
70 | - implement Coro->join and Coro->on_destroy in XS for a speedup and |
|
|
71 | a reduction in memory use. |
|
|
72 | - cancelling a coro while it itself is cancelling another coro is |
|
|
73 | now supported and working, instead of triggering an assertion. |
| 7 | - be a bit more crash-resistant when calling (buggy) on_destroy |
74 | - be a bit more crash-resistant when calling (buggy) on_destroy |
| 8 | callbacks (bets effort). |
75 | callbacks (best effort). |
|
|
76 | - move on_destroy into the slf_frame, to allow extension slf |
|
|
77 | functions to have destructors. |
|
|
78 | - get rid if coro refcounting - simply crash in other interpreter |
|
|
79 | threads by nulling the pointers on clone. |
|
|
80 | - simplify warn/die hook handling when loading Coro - the convoluted |
|
|
81 | logic seems to be no longer neccessary. |
|
|
82 | - use libecb instead of our own home-grown gcc hacks. |
|
|
83 | - document alternatives to Coro::LWP. Please use them :) |
|
|
84 | - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds |
|
|
85 | sigaltstack. Really. wine suffers from it, erlang suffers from it, |
|
|
86 | and it's known since at least 2006. |
| 9 | |
87 | |
| 10 | 5.372 Wed Feb 23 06:14:30 CET 2011 |
88 | 5.372 Wed Feb 23 06:14:30 CET 2011 |
| 11 | - apparently mingw doesn't provide a working gettimeofday, try to |
89 | - apparently mingw doesn't provide a working gettimeofday, try to |
| 12 | work around that by relying on Time::HiRes (indirectly brought to |
90 | work around that by relying on Time::HiRes (indirectly brought to |
| 13 | my attention by Max Maischein). |
91 | my attention by Max Maischein). |
