| … | |
… | |
| 3 | TODO: should explore PerlIO::coroaio (perl leaks like hell). |
3 | TODO: should explore PerlIO::coroaio (perl leaks like hell). |
| 4 | TODO: maybe implement a default message channel, very much like Erlang's |
4 | TODO: maybe implement a default message channel, very much like Erlang's |
| 5 | actor model (which is cool in a lot of important aspects (failures!), |
5 | actor model (which is cool in a lot of important aspects (failures!), |
| 6 | but very lacking in others (higher level ipc)). |
6 | but very lacking in others (higher level ipc)). |
| 7 | |
7 | |
|
|
8 | TODO: Coro/Intro |
|
|
9 | - wrote a small introduction - Coro::Intro. |
|
|
10 | - convert Coro::Timer, Coro::Select and Coro::Util to rouse API. |
|
|
11 | - Coro::Select did errornously dup the file descriptors |
|
|
12 | and didn't work with all AnyEvent backends. |
|
|
13 | - disassociate c contexts from coro objects - this is agruably more |
|
|
14 | correct, but mostly allows sharing of cctxs between coro and state |
|
|
15 | objects, for added memory savings and speed increases. |
|
|
16 | - bumped $Coro::POOL_RSS up to 32kb by default. |
|
|
17 | - no longer set the optype to OP_CUSTOM, as B::* understandably |
|
|
18 | doesn't like this very much (and we *are* a type of entersub). |
|
|
19 | - implement state cloning, just to prove that call/cc can be done. |
|
|
20 | - automatically load Coro::AnyEvent in Coro::Handle. |
|
|
21 | - wrap ->cancel calls in eval inside Coro::Handle as EV watchers |
|
|
22 | do not have this method (and don't need it either). |
|
|
23 | - speed up generic anyevent methods in Coro::Handle by using rouse |
|
|
24 | callbacks. |
|
|
25 | - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and |
|
|
26 | others. It also makes the debugger happier, as you can trace |
|
|
27 | through the idle threads now. |
|
|
28 | - eg/event was pretty broken. |
|
|
29 | - better 5.8.6 compatibility. |
|
|
30 | |
|
|
31 | 5.0 Thu Nov 20 10:35:05 CET 2008 |
|
|
32 | - NEW ARCHITECTURE: use the latest 4.x version if you experience |
|
|
33 | stability issues. |
|
|
34 | - bump API version to 7 - all dependents must be recompiled. |
|
|
35 | - removed timed_* functions - they were not being used anyways |
|
|
36 | and should be replaced by a more generic mechanism - |
|
|
37 | and were annoying to support anyways :) |
|
|
38 | - removed SemaphoreSet's waiter method - use sem method instead. |
|
|
39 | - Coro::Semaphore->adjust didn't correctly wake up enough waiters. |
|
|
40 | - async_pool did free a scalar value twice |
|
|
41 | ("Attempt to unreference..."). |
|
|
42 | - fix a longstanding bug where calling terminate on a coro that |
|
|
43 | was waiting for a semaphore that was just becoming available |
|
|
44 | would cause a deadlock (semaphore would get into a state where |
|
|
45 | it was available but waiters were still blocked). |
|
|
46 | - calling throw on a coroutine that is waiting for a semaphore will |
|
|
47 | no longer make it acquire the semaphore (and thus leak a count). |
|
|
48 | - perl's process emulation is now not even theoretically supported |
|
|
49 | anymore. |
|
|
50 | - new functions Coro::rouse_cb and Coro::rouse_wait for easier |
|
|
51 | conversion of callback-style to blocking-style. |
|
|
52 | - new methods $coro->schedule_to and ->cede_to, to specifically |
|
|
53 | schedule or cede to a specific coroutine. |
|
|
54 | - new function Coro::Semaphore::wait. |
|
|
55 | - use named constants in Coro::Channel (Richard Hundt). |
|
|
56 | - directly patch the entersub opcode calling SLF functions (cede, |
|
|
57 | transfer and so on). this does speed up context switching, but |
|
|
58 | more importanly, it frees us from the hardcoded behaviour of |
|
|
59 | entersub, so we might actually be able to return something from |
|
|
60 | those functions and atcually create new ones. |
|
|
61 | - take advantage of __builtin_frame_address on gcc. |
|
|
62 | - expose THX in coroapi (not sure whether this was a wise decision, |
|
|
63 | as "threaded" perls are running at half speed anyways). |
|
|
64 | - implement execute_slf (schedule-like-function) interface that makes |
|
|
65 | it possible to implement schedule-like-functions in XS. |
|
|
66 | - use new SLF interface to massively speed up Coro::EV by roughly a |
|
|
67 | factor of two. |
|
|
68 | - used new SLF interface to massively speed up Coro::Semaphore by a |
|
|
69 | factor of three. |
|
|
70 | - used new SLF interface to speed up Coro::AIO by roughly a factor of |
|
|
71 | four and reduce its memory usage considerably. |
|
|
72 | - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore, |
|
|
73 | for a nice speedup and vastly more correct behaviour. Also implement |
|
|
74 | a new method "sem" to get at the underlying semaphore object. |
|
|
75 | - implement Coro::Channel in terms of Coro::Semaphore, for a moderate |
|
|
76 | (in comparison) 20-40% speedup. |
|
|
77 | - used new SLF interface to reimplement Coro::Signal gaining |
|
|
78 | some unknown (because I was too lazy), but certain, speedup, and also |
|
|
79 | making signals reliable for the first time. |
|
|
80 | - used new SLF interface and other optimisations to speed up async_pool |
|
|
81 | by a factor of two. It also doesn't rely on perl's exception mechanism |
|
|
82 | to exit anymore. The overhead for terminating an async_pool, coro over |
|
|
83 | a normal async is now very small. |
|
|
84 | - sped up coroutine creation/destruction by 40%. |
|
|
85 | - forgot to include Coro/libcoro/README in the dist for all these years. |
|
|
86 | - work around a freebsd pthreads bug (manual testcancel is required as |
|
|
87 | pthread_cond_wait isn't a cancellation point on freebsd). |
|
|
88 | - use new rouse functions to speed up and simplify Coro::BDB. |
|
|
89 | - make "prefer perl native functions" work with threaded perls. |
|
|
90 | - condense Coro::Debug ps output, hint at v and w flags. |
|
|
91 | - (libcoro) lots of minor cleanups and portability improvements. |
|
|
92 | |
|
|
93 | 4.914 Wed Nov 19 12:54:18 CET 2008 |
|
|
94 | - fix a disastrous bug in the readline optimisation |
|
|
95 | introduced in 4.801. |
|
|
96 | |
|
|
97 | 4.913 Sat Nov 15 07:58:28 CET 2008 |
|
|
98 | - async_pool did free a scalar value twice |
|
|
99 | ("Attempt to unreference..."). |
|
|
100 | |
|
|
101 | 4.912 Thu Nov 13 18:31:23 CET 2008 |
|
|
102 | - minor cleanups. |
|
|
103 | - use much larger stacks on linux and perl < 5.8.8. |
|
|
104 | - Coro::Debug::new_unix_server did not unlink the socket |
|
|
105 | when destroyed. |
|
|
106 | |
|
|
107 | 4.911 Tue Nov 11 04:26:17 CET 2008 |
|
|
108 | - "port" to threaded perls. |
|
|
109 | |
|
|
110 | 4.91 Mon Nov 10 05:36:38 CET 2008 |
|
|
111 | - the ->throw exception object no longer leaks. |
|
|
112 | - creating a new cctx leaked a scopestack entry (memleak). |
|
|
113 | - new coroutines didn't get created with a zero flags field |
|
|
114 | (unknown impact). |
|
|
115 | - calling ->throw on a not-yet-started coroutine should now work |
|
|
116 | instead of being ignored. |
|
|
117 | - ->throw is now supported on Coro::State objects. |
|
|
118 | - clean up cctx creation code a bit. |
|
|
119 | - entersub is actually a UNOP, not a LOGOP (not a bugfix). |
|
|
120 | |
|
|
121 | 4.9 Sat Nov 8 17:45:27 CET 2008 |
|
|
122 | - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). |
|
|
123 | - (libcoro) no longer leak threads in the experimental pthread backend, |
|
|
124 | also speed it up considerably. |
|
|
125 | - (libcoro) do not rely on makecontext passing void *'s unscathed. |
|
|
126 | - fix compiletime dependencies on libcoro in the Makefile. |
|
|
127 | - cctx_count wasn't always updated properly. |
|
|
128 | - Coro::State::cctx_stacksize wasn't applied correctly. |
|
|
129 | - new function Coro::State::cctx_max_idle. |
|
|
130 | - the default max number of idle C contexts is now 4. |
|
|
131 | - (libcoro) try harder to get _setjmp/_longjmp. |
|
|
132 | - (libcoro) cleanup and extend the libcoro API to officially |
|
|
133 | allow the creation of empty source contexts. |
|
|
134 | - very experimental workaround for the totally broken netbsd platform. |
|
|
135 | - tried to hack around openbsd bugs. |
|
|
136 | |
|
|
137 | 4.804 Wed Nov 5 16:36:00 CET 2008 |
|
|
138 | - Coro::Debug::new_unix_server would not create a non-blocking listening |
|
|
139 | socket, sometimes causing freezes. |
|
|
140 | - (libcoro) fix misaligned stack points for setjmp and assembly |
|
|
141 | methods, which can cause crashes on x86/x86_64 with a sufficiently |
|
|
142 | aggressive compiler. |
|
|
143 | - new function: Coro::Debug::new_tcp_server. |
|
|
144 | - move ->throw into the Coro class because it only works on coro objects. |
|
|
145 | |
| 8 | 4.803 Mon Nov 3 17:16:12 CET 2008 |
146 | 4.803 Mon Nov 3 17:16:12 CET 2008 |
| 9 | - use a global asm statement to become independent of gcc otpimisations |
147 | - (libcoro) use a global asm statement to become independent of gcc |
| 10 | for CORO_ASM (thanks to pippijn for the idea). |
148 | otpimisations for CORO_ASM (thanks to pippijn for the idea). |
| 11 | - try to workaround yet another broken bsd, this time dragonfly. |
149 | - try to workaround yet another broken bsd, this time dragonfly. |
| 12 | |
150 | |
| 13 | 4.802 Thu Oct 30 10:56:12 CET 2008 |
151 | 4.802 Thu Oct 30 10:56:12 CET 2008 |
| 14 | - support -fno-omit-frame-pointer on x86 with the assembly method. |
152 | - support -fno-omit-frame-pointer on x86 with the assembly method. |
| 15 | - tune 01_unblock.t tests a bit. |
153 | - tune 01_unblock.t tests a bit. |
