Issue #14937 has been updated by MSP-Greg (Greg L).


@normalperson

Eric,

In addition to the build mentioned above, there have been two more ruby-loco builds, both with all three patches, and both passed all tests

>  Thanks for testing; I guess 47 minutes is not unusual for a busy CI machine?

Re Appveyor:
* 6 minutes is updating the build system
* I think each CI is run on 2 cores of a 4 core system, the system may be competing with another CI for disk IO, etc. The build for r64024 completed in 37 minutes. 
* I run both make test-rubyspec and mspec.
* I think MJIT is one of the longer running suites, and with retry, it is always run twice.

Thanks again, Greg

----------------------------------------
Misc #14937: [PATCH] thread_pthread: lazy-spawn timer-thread only on contention
https://bugs.ruby-lang.org/issues/14937#change-73128

* Author: normalperson (Eric Wong)
* Status: Open
* Priority: Normal
* Assignee: 
----------------------------------------
[ruby-core:87773]
```
thread_pthread: lazy-spawn timer-thread only on contention

To reduce resource use and reduce CI failure; lazy spawn
timer-thread only in processes which use Ruby Threads AND those
Ruby Threads hit contention.  Single-threaded Ruby processes
(including forked children) will never have timer-thread
overhead.

To simplify the thread_pthread.c code, I eliminated busy timer
thread [Misc #14851].  Maybe the thread_win32.c code can use
self-pipe, too; and they won't need busy wakeups.

There is only one self-pipe, now, as wakeups for timeslice are
handled via condition variables.  This reduces FD pressure
slightly.

Signal handling is handled directly by one 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) and avoid timer-thread completely.
Unfortunately, this proved unfeasible for one reason:
Mutex#sleep resumes on spurious wakeups and
test/thread/test_cv.rb::test_condvar_timed_wait failed.

In the future, I hope [Feature #14717] is accepted so Threads
may be made non-preemptible.  This will allow users to prevent
timer-thread creation completely.
```

git repository also available at:
  https://80x24.org/ruby.git tt-lazy
  (commit a2990cefccba55300ad44275ee4adf18e6f95ece)


---Files--------------------------------
0001-thread_pthread-lazy-spawn-timer-thread-only-on-conte.patch (42.3 KB)


-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>