2 @comment node-name, next, previous, up
5 SBCL supports a fairly low-level threading interface that maps onto
6 the host operating system's concept of threads or lightweight
7 processes. This means that threads may take advantage of hardware
8 multiprocessing on machines that have more than one CPU, but it does
9 not allow Lisp control of the scheduler. This is found in the
12 This requires x86/x86-64 and Linux kernel 2.6 or systems with NPTL
19 * Waitqueue/condition variables::
20 * Sessions/Debugging::
21 * Implementation (Linux x86)::
24 @node Threading basics
25 @comment node-name, next, previous, up
26 @section Threading basics
29 (make-thread (lambda () (write-line "Hello, world")))
32 @include struct-sb-thread-thread.texinfo
33 @include var-sb-thread-star-current-thread-star.texinfo
34 @include fun-sb-thread-make-thread.texinfo
35 @include fun-sb-thread-thread-alive-p.texinfo
36 @include fun-sb-thread-list-all-threads.texinfo
37 @include condition-sb-thread-interrupt-thread-error.texinfo
38 @include fun-sb-thread-interrupt-thread-error-thread.texinfo
39 @include fun-sb-thread-interrupt-thread.texinfo
40 @include fun-sb-thread-terminate-thread.texinfo
42 @node Special Variables
43 @comment node-name, next, previous, up
44 @section Special Variables
46 The interaction of special variables with multiple threads is mostly
47 as one would expect, with behaviour very similar to other
52 global special values are visible across all threads;
54 bindings (e.g. using LET) are local to the thread;
56 threads do not inherit dynamic bindings from the parent thread
59 The last point means that
64 (sb-thread:make-thread (lambda () (print *x*))))
67 prints @code{0} and not @code{1} as of 0.9.6.
70 @comment node-name, next, previous, up
71 @section Mutex Support
73 Mutexes are used for controlling access to a shared resource. One
74 thread is allowed to hold the mutex, others which attempt to take it
75 will be made to wait until it's free. Threads are woken in the order
76 that they go to sleep.
78 There isn't a timeout on mutex acquisition, but the usual WITH-TIMEOUT
79 macro (which throws a TIMEOUT condition after n seconds) can be used
80 if you want a bounded wait.
83 (defpackage :demo (:use "CL" "SB-THREAD" "SB-EXT"))
87 (defvar *a-mutex* (make-mutex :name "my lock"))
90 (format t "Thread ~A running ~%" *current-thread*)
91 (with-mutex (*a-mutex*)
92 (format t "Thread ~A got the lock~%" *current-thread*)
94 (format t "Thread ~A dropped lock, dying now~%" *current-thread*)))
96 (make-thread #'thread-fn)
97 (make-thread #'thread-fn)
100 @include struct-sb-thread-mutex.texinfo
101 @include fun-sb-thread-make-mutex.texinfo
102 @include fun-sb-thread-mutex-name.texinfo
103 @include fun-sb-thread-mutex-value.texinfo
104 @include fun-sb-thread-get-mutex.texinfo
105 @include fun-sb-thread-release-mutex.texinfo
106 @include macro-sb-thread-with-mutex.texinfo
107 @include macro-sb-thread-with-recursive-lock.texinfo
109 @node Waitqueue/condition variables
110 @comment node-name, next, previous, up
111 @section Waitqueue/condition variables
113 These are based on the POSIX condition variable design, hence the
114 annoyingly CL-conflicting name. For use when you want to check a
115 condition and sleep until it's true. For example: you have a shared
116 queue, a writer process checking ``queue is empty'' and one or more
117 readers that need to know when ``queue is not empty''. It sounds
118 simple, but is astonishingly easy to deadlock if another process runs
119 when you weren't expecting it to.
121 There are three components:
125 the condition itself (not represented in code)
128 the condition variable (a.k.a waitqueue) which proxies for it
131 a lock to hold while testing the condition
134 Important stuff to be aware of:
138 when calling condition-wait, you must hold the mutex. condition-wait
139 will drop the mutex while it waits, and obtain it again before
140 returning for whatever reason;
143 likewise, you must be holding the mutex around calls to
147 a process may return from condition-wait in several circumstances: it
148 is not guaranteed that the underlying condition has become true. You
149 must check that the resource is ready for whatever you want to do to
155 (defvar *buffer-queue* (make-waitqueue))
156 (defvar *buffer-lock* (make-mutex :name "buffer lock"))
158 (defvar *buffer* (list nil))
161 (with-mutex (*buffer-lock*)
163 (condition-wait *buffer-queue* *buffer-lock*)
165 (unless *buffer* (return))
166 (let ((head (car *buffer*)))
167 (setf *buffer* (cdr *buffer*))
168 (format t "reader ~A woke, read ~A~%"
169 *current-thread* head))))))
174 (with-mutex (*buffer-lock*)
177 (+ (char-code #\A) (random 26)))))))
178 (setf *buffer* (cons el *buffer*)))
179 (condition-notify *buffer-queue*))))
181 (make-thread #'writer)
182 (make-thread #'reader)
183 (make-thread #'reader)
186 @include struct-sb-thread-waitqueue.texinfo
187 @include fun-sb-thread-make-waitqueue.texinfo
188 @include fun-sb-thread-waitqueue-name.texinfo
189 @include fun-sb-thread-condition-wait.texinfo
190 @include fun-sb-thread-condition-notify.texinfo
191 @include fun-sb-thread-condition-broadcast.texinfo
193 @node Sessions/Debugging
194 @comment node-name, next, previous, up
195 @section Sessions/Debugging
197 If the user has multiple views onto the same Lisp image (for example,
198 using multiple terminals, or a windowing system, or network access)
199 they are typically set up as multiple @dfn{sessions} such that each
200 view has its own collection of foreground/background/stopped threads.
201 A thread which wishes to create a new session can use
202 @code{sb-thread:with-new-session} to remove itself from the current
203 session (which it shares with its parent and siblings) and create a
205 # See also @code{sb-thread:make-listener-thread}.
207 Within a single session, threads arbitrate between themselves for the
208 user's attention. A thread may be in one of three notional states:
209 foreground, background, or stopped. When a background process
210 attempts to print a repl prompt or to enter the debugger, it will stop
211 and print a message saying that it has stopped. The user at his
212 leisure may switch to that thread to find out what it needs. If a
213 background thread enters the debugger, selecting any restart will put
214 it back into the background before it resumes. Arbitration for the
215 input stream is managed by calls to @code{sb-thread:get-foreground}
216 (which may block) and @code{sb-thread:release-foreground}.
218 @code{sb-ext:quit} terminates all threads in the current session, but
219 leaves other sessions running.
221 @node Implementation (Linux x86)
222 @comment node-name, next, previous, up
223 @section Implementation (Linux x86/x86-64)
225 Threading is implemented using pthreads and some Linux specific bits
228 On x86 the per-thread local bindings for special variables is achieved
229 using the %fs segment register to point to a per-thread storage area.
230 This may cause interesting results if you link to foreign code that
231 expects threading or creates new threads, and the thread library in
232 question uses %fs in an incompatible way. On x86-64 the r12 register
235 Queues require the @code{sys_futex()} system call to be available:
236 this is the reason for the NPTL requirement. We test at runtime that
237 this system call exists.
239 Garbage collection is done with the existing Conservative Generational
240 GC. Allocation is done in small (typically 8k) regions: each thread
241 has its own region so this involves no stopping. However, when a
242 region fills, a lock must be obtained while another is allocated, and
243 when a collection is required, all processes are stopped. This is
244 achieved by sending them signals, which may make for interesting
245 behaviour if they are interrupted in system calls. The streams
246 interface is believed to handle the required system call restarting
247 correctly, but this may be a consideration when making other blocking
248 calls e.g. from foreign library code.
250 Large amounts of the SBCL library have not been inspected for
251 thread-safety. Some of the obviously unsafe areas have large locks
252 around them, so compilation and fasl loading, for example, cannot be
253 parallelized. Work is ongoing in this area.
255 A new thread by default is created in the same POSIX process group and
256 session as the thread it was created by. This has an impact on
257 keyboard interrupt handling: pressing your terminal's intr key
258 (typically @kbd{Control-C}) will interrupt all processes in the
259 foreground process group, including Lisp threads that SBCL considers
260 to be notionally `background'. This is undesirable, so background
261 threads are set to ignore the SIGINT signal.
263 @code{sb-thread:make-listener-thread} in addition to creating a new
264 Lisp session makes a new POSIX session, so that pressing
265 @kbd{Control-C} in one window will not interrupt another listener -
266 this has been found to be embarrassing.