This patch introduce M:N thread scheduler for Ractor system.
In general, M:N thread scheduler employs N native threads (OS threads)
to manage M user-level threads (Ruby threads in this case).
On the Ruby interpreter, 1 native thread is provided for 1 Ractor
and all Ruby threads are managed by the native thread.
From Ruby 1.9, the interpreter uses 1:1 thread scheduler which means
1 Ruby thread has 1 native thread. M:N scheduler change this strategy.
Because of compatibility issue (and stableness issue of the implementation)
main Ractor doesn't use M:N scheduler on default. On the other words,
threads on the main Ractor will be managed with 1:1 thread scheduler.
There are additional settings by environment variables:
`RUBY_MN_THREADS=1` enables M:N thread scheduler on the main ractor.
Note that non-main ractors use the M:N scheduler without this
configuration. With this configuration, single ractor applications
run threads on M:1 thread scheduler (green threads, user-level threads).
`RUBY_MAX_CPU=n` specifies maximum number of native threads for
M:N scheduler (default: 8).
This patch will be reverted soon if non-easy issues are found.
[Bug #19842]
Checking by `__STDC_VERSION__` is unreliable because old gcc 4.8
supports `-std=gnu11` option but does not implement `_Thread_local`.
Check the implementation directly instead.
Show native thread's serial on `RUBY_DEBUG_LOG`.
`nt->serial` is also stored into `ruby_nt_serial` if the compiler
supports `RB_THREAD_LOCAL_SPECIFIER`.
`NON_SCALAR_THREAD_ID` shows `pthread_t` is non-scalar (non-pointer)
and only s390x is known platform. However, the supporting code is
very complex and it is only used for deubg print information.
So this patch removes the support of `NON_SCALAR_THREAD_ID`
and make the code simple.
`rb_thread_t` contained `native_thread_data_t` to represent
thread implementation dependent data. This patch separates
them and rename it `rb_native_thread` and point it from
`rb_thraed_t`.
Now, 1 Ruby thread (`rb_thread_t`) has 1 native thread (`rb_native_thread`).
Now GVL is not process *Global* so this patch try to use
another words.
* `rb_global_vm_lock_t` -> `struct rb_thread_sched`
* `gvl->owner` -> `sched->running`
* `gvl->waitq` -> `sched->readyq`
* `rb_gvl_init` -> `rb_thread_sched_init`
* `gvl_destroy` -> `rb_thread_sched_destroy`
* `gvl_acquire` -> `thread_sched_to_running` # waiting -> ready -> running
* `gvl_release` -> `thread_sched_to_waiting` # running -> waiting
* `gvl_yield` -> `thread_sched_yield`
* `GVL_UNLOCK_BEGIN` -> `THREAD_BLOCKING_BEGIN`
* `GVL_UNLOCK_END` -> `THREAD_BLOCKING_END`
* removed
* `rb_ractor_gvl`
* `rb_vm_gvl_destroy` (not used)
There are GVL functions such as `rb_thread_call_without_gvl()` yet
but I don't have good name to replace them. Maybe GVL stands for
"Greate Valuable Lock" or something like that.
To make some kind of Ractor related extensions, some functions
should be exposed.
* include/ruby/thread_native.h
* rb_native_mutex_*
* rb_native_cond_*
* include/ruby/ractor.h
* RB_OBJ_SHAREABLE_P(obj)
* rb_ractor_shareable_p(obj)
* rb_ractor_std*()
* rb_cRactor
and rm ractor_pub.h
and rename srcdir/ractor.h to srcdir/ractor_core.h
(to avoid conflict with include/ruby/ractor.h)
To access TLS, it is faster to use language TLS specifier instead
of using pthread_get/setspecific functions.
Original proposal is: Use native thread locals. #3665
This commit introduces Ractor mechanism to run Ruby program in
parallel. See doc/ractor.md for more details about Ractor.
See ticket [Feature #17100] to see the implementation details
and discussions.
[Feature #17100]
This commit does not complete the implementation. You can find
many bugs on using Ractor. Also the specification will be changed
so that this feature is experimental. You will see a warning when
you make the first Ractor with `Ractor.new`.
I hope this feature can help programmers from thread-safety issues.
According to MSVC manual (*1), cl.exe can skip including a header file
when that:
- contains #pragma once, or
- starts with #ifndef, or
- starts with #if ! defined.
GCC has a similar trick (*2), but it acts more stricter (e. g. there
must be _no tokens_ outside of #ifndef...#endif).
Sun C lacked #pragma once for a looong time. Oracle Developer Studio
12.5 finally implemented it, but we cannot assume such recent version.
This changeset modifies header files so that each of them include
strictly one #ifndef...#endif. I believe this is the most portable way
to trigger compiler optimizations. [Bug #16770]
*1: https://docs.microsoft.com/en-us/cpp/preprocessor/once
*2: https://gcc.gnu.org/onlinedocs/cppinternals/Guard-Macros.html
`acquired' was an old boolean variable, but nowadays it is a
rb_thread_t pointer; "gvl.owner" seems like a more appropriate
name. And document the contended path including waitq, timer,
and timer_err.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64581 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This matches the behavior of old timer thread more closely
and seems to fix [Bug #14999] when limited to a single CPU.
I cannot reproduce the error on a multi-core system unless
I use schedtool to force affinity to a single CPU:
schedtool -a 0x01 -e make test-spec \
MSPECOPT='-R1000 spec/ruby/library/conditionvariable/wait_spec.rb'
While it may be good enough to pass the spec, I don't have
huge degree of confidence in the interrupt handling robustness
under extremely heavy load (these may be ancient bugs, though).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64467 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Do not waste extra memory for each thread, but make
thread_pthread.c easier-to-follow as a result.
[ruby-core:88475] [Misc #14937]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64375 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This reverts commit 194a6a2c68 (r64203).
Race conditions which caused the original reversion will be fixed
in the subsequent commit.
[ruby-core:88360] [Misc #14937]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64352 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Maybe some platforms have strange condition variable implementations
which have a "memory" of which mutexes they're associated with.
In any case, it makes documentation easier even on GNU/Linux and
FreeBSD.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64124 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
To reduce resource use and reduce CI failure; remove
timer-thread. Single-threaded Ruby processes (including forked
children) will never see extra thread overhead. This prevents
glibc and jemalloc from going into multi-threaded mode and
initializing locks or causing fragmentation via arena explosion.
The GVL is implements its own wait-queue as a ccan/list to
permit controlling wakeup order. Timeslice under contention is
handled by a designated timer thread (similar to choosing a
"patrol_thread" for current deadlock checking).
There is only one self-pipe, now, as wakeups for timeslice are
done independently using condition variables. This reduces FD
pressure slightly.
Signal handling is handled directly by a Ruby Thread (instead
of timer-thread) by exposing signal self-pipe to callers of
rb_thread_fd_select, native_sleep, rb_wait_for_single_fd, etc...
Acquiring, using, and releasing the self-pipe is exposed via 4
new internal functions:
1) rb_sigwait_fd_get - exclusively acquire timer_thread_pipe.normal[0]
2) rb_sigwait_fd_sleep - sleep and wait for signal (and no other FDs)
3) rb_sigwait_fd_put - release acquired result from rb_sigwait_fd_get
4) rb_sigwait_fd_migrate - migrate signal handling to another thread
after calling rb_sigwait_fd_put.
rb_sigwait_fd_migrate is necessary for waitpid callers because
only one thread can wait on self-pipe at a time, otherwise a
deadlock will occur if threads fight over the self-pipe.
TRAP_INTERRUPT_MASK is now set for the main thread directly in
signal handler via rb_thread_wakeup_timer_thread.
Originally, I wanted to use POSIX timers
(timer_create/timer_settime) for this. Unfortunately, this
proved unfeasible as Mutex#sleep resumes on spurious wakeups and
test/thread/test_cv.rb::test_condvar_timed_wait failed. Using
pthread_sigmask to mask out SIGVTALRM fixed that test, but
test/fiddle/test_function.rb::test_nogvl_poll proved there'd be
some unavoidable (and frequent) incompatibilities from that
approach.
Finally, this allows us to drop thread_destruct_lock and
interrupt current ec directly.
We don't need to rely on vm->thread_destruct_lock or a coherent
vm->running_thread on any platform. Separate timer-thread for
time slice and signal handling is relegated to thread_win32.c,
now.
[ruby-core:88088] [Misc #14937]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64107 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
There's no reason to use CLOCK_REALTIME for any condvars in Ruby.
Indeed, we initialized all condvars with RB_CONDATTR_CLOCK_MONOTONIC
anyway; so simplify our code and reduce ifdefs.
[ruby-core:85639] [Misc #14497]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@63238 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This will make it easier for us to try alternative mutex/condvar
implementations while still using pthreads for thread management.
[Feature #10134]
* thread_pthread.h: define RB_NATIVETHREAD_LOCK_INIT and
RB_NATIVETHREAD_COND_INIT macros
* thread_pthread.c (native_mutex_lock, native_mutex_unlock,
native_mutex_trylock, native_mutex_initialize,
native_mutex_destroy, native_cond_wait):
use rb_nativethread_lock_t instead of pthread_mutex_t
* thread_pthread.c (native_mutex_debug): make argument type-agnostic
to avoid later cast.
* thread_pthread.c (register_cached_thread_and_wait):
replace PTHREAD_COND_INITIALIZER with RB_NATIVETHREAD_COND_INIT,
use native_mutex_{lock,unlock}
* thread_pthread.c (use_cached_thread):
use native_mutex_{lock,unlock}
* thread_pthread.c (native_sleep):
use rb_nativethread_lock_t to match th->interrupt_lock,
use native_mutex_{lock,unlock}
* thread_pthread.c (timer_thread_lock): use rb_nativethread_lock_t type
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47185 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.h (struct rb_global_vm_lock_struct):
do not expose pthread type for lock
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46761 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.h: remove unneeded semaphore.h include
We currently do not use POSIX semaphores.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46760 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Move native thread related lines from vm_core.h.
And declare several functions "rb_nativethread_lock_*",
manipulate locking.
* common.mk: add thread_native.h.
* thread.c: add functions "rb_nativethread_lock_*".
* thraed.c, thread_[pthread,win32].[ch]: rename rb_thread_lock_t
to rb_nativethread_lock_t to make it clear that this lock is for
native thraeds, not for ruby threads.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@42133 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
no longer an alias of pthread_cond_t.
* thread_pthread.c: adapt new rb_thread_cond_t type.
* thread.c (mutex_alloc): ditto.
* thread_win32.c (native_cond_initialize): ditto.
* configure.in: add check for pthread_cond_attr_setclock() and
clockid_t type.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@31451 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
APIs to modularize GVL implementation.
* thread_pthread.c, thread_pthread.h: Two GVL implementations.
(1) Simple locking GVL which is same as existing GVL.
(2) Wake-up queued threads. The wake-up order is simple FIFO.
(We can make several queues to support exact priorities, however
this causes some issues such as priority inversion and so on.)
This impl. prevents spin-loop (*1) caused on SMP environemnts.
*1: Only one Ruby thread acqures GVL again and again.
Bug #2359 [ruby-core:26694]
* thread_win32.c, thread_win32.h: Using simple lock
not by CRITICAL_SECTION but by Mutex.
Bug #3890 [ruby-dev:42315]
* vm.c (ruby_vm_destruct): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@29956 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
patched by KOSAKI Motohiro [ruby-dev:40309]
* thread_pthread.c (ruby_init_stack): use get_stack
on platforms which have pthread_attr_get_np.
(FreeBSD, DragonFlyBSD and NetBSD)
This is because FreeBSD and DragonFly BSD must use
pthread_attr_get_np to get stack size of main thread,
but Mac OS X and Linux with LinuxThreads must use getrlimit.
<http://www.nminoru.jp/~nminoru/programming/stackoverflow_handling.html>
<http://d.hatena.ne.jp/nurse/20100204>
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@26572 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
prefixed include guards with RUBY.
* id.h: added include guard.
* regenc.h, regint.h, regparse.h: prefixed include guards with
ONIGURUMA.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@15112 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
thread_win32.c: make some functions static functions.
a patch from Tadashi Saito <shiba AT mail2.accsnet.ne.jp>
in [ruby-core:14407]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@14657 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_win32.ci, thread_win32.h, thread_pthread.ci, thread_pthread.h:
prepare native_cond_*() which are based on pthread_cond_*() spec.
* prelude.rb: fix Mutex#synchronize method.
* vm_core.h, include/ruby/intern.h: change unblock function interface
(to pass some user data).
* file.c, process.c: ditto.
* benchmark/bm_vm2_mutex.rb: add a benchmark for mutex.
* benchmark/bm_vm3_thread_mutex.rb: add a benchmark for mutex
with contension.
* benchmark/run.rb: fix to remove ENV['RUBYLIB'] for matzruby.
* test/ruby/test_thread.rb: add a test.
* common.mk: fix benchmark options.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@13290 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
move cleanup function to thread_start_func_2().
* thread.c, thread_pthread.ci, thread_win32.ci:
add more destruct functions.
(native_thread_destroy() and native_mutex_destroy())
* thread_pthread.ci, thread_pthread.h: make native_mutex_* functions
(check error, etc), it's not macro any more.
* thread_win32.ci (thread_start_func_1): store some values before
running thread (to release these after running thread).
* thread_win32.ci (native_thread_create): fix spaces.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@11671 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Koichi Sasada
Nobuyoshi Nakada
Benoit Daloze
卜部昌平
normal
kosaki
ko1
nobu
naruse
akr