* bug#21887: 'monitor' form broken
@ 2015-11-12 15:29 Taylan Ulrich Bayırlı/Kammer
2016-06-24 16:04 ` Andy Wingo
0 siblings, 1 reply; 5+ messages in thread
From: Taylan Ulrich Bayırlı/Kammer @ 2015-11-12 15:29 UTC (permalink / raw)
To: 21887
It seems that the 'monitor' form is currently a no-op. The form
(par-for-each (lambda (x)
(monitor
(foo)))
xs)
should be functionally equivalent to
(let ((mutex (make-mutex)))
(par-for-each (lambda (x)
(with-mutex mutex
(foo)))
xs))
but currently becomes
(par-for-each (lambda (x)
(let ((mutex (make-mutex)))
(with-mutex mutex
(foo))))
xs)
which is ineffective.
I don't know what's the best way to fix this. The simplest thing that
comes to my mind is something along the lines of:
(define-syntax monitor
(lambda (stx)
(syntax-case stx ()
((_ body body* ...)
(let ((uuid (generate-uuid)))
#`(with-mutex (mutex-with-uuid #,uuid)
body body* ...))))))
where mutex-with-uuid looks it up from a hash table at run-time and
instantiates it when it doesn't exist, this operation also being
synchronized across threads, like:
(define mutex-table (make-hash-table))
(define mutex-table-mutex (make-mutex))
(define (mutex-with-uuid uuid)
(with-mutex mutex-table-mutex
(or (hash-ref mutex-table uuid)
(let ((mutex (make-mutex)))
(hash-set! mutex-table uuid mutex)
mutex))))
If that looks OK, I can try to make a proper patch from it. I'm not
sure what I'd use in place of `generate-uuid' though. Would `gensym' be
good enough?
Shameless advertisement: with SRFI-126, the (or (hash-ref ...) ...) bit
would have been just:
(hashtable-intern! mutex-table uuid make-mutex)
It's borrowed from MIT/GNU Scheme. Seems pretty useful.
Taylan
^ permalink raw reply [flat|nested] 5+ messages in thread
* bug#21887: 'monitor' form broken
2015-11-12 15:29 bug#21887: 'monitor' form broken Taylan Ulrich Bayırlı/Kammer
@ 2016-06-24 16:04 ` Andy Wingo
2016-06-25 14:51 ` Taylan Ulrich Bayırlı/Kammer
0 siblings, 1 reply; 5+ messages in thread
From: Andy Wingo @ 2016-06-24 16:04 UTC (permalink / raw)
To: Taylan Ulrich "Bayırlı/Kammer"; +Cc: 21887
Hi Taylan,
On Thu 12 Nov 2015 16:29, taylanbayirli@gmail.com (Taylan Ulrich "Bayırlı/Kammer") writes:
> It seems that the 'monitor' form is currently a no-op. The form
>
> (par-for-each (lambda (x)
> (monitor
> (foo)))
> xs)
>
> should be functionally equivalent to
>
> (let ((mutex (make-mutex)))
> (par-for-each (lambda (x)
> (with-mutex mutex
> (foo)))
> xs))
>
> but currently becomes
>
> (par-for-each (lambda (x)
> (let ((mutex (make-mutex)))
> (with-mutex mutex
> (foo))))
> xs)
>
> which is ineffective.
>
> I don't know what's the best way to fix this. The simplest thing that
> comes to my mind is something along the lines of:
>
> (define-syntax monitor
> (lambda (stx)
> (syntax-case stx ()
> ((_ body body* ...)
> (let ((uuid (generate-uuid)))
> #`(with-mutex (mutex-with-uuid #,uuid)
> body body* ...))))))
>
> where mutex-with-uuid looks it up from a hash table at run-time and
> instantiates it when it doesn't exist, this operation also being
> synchronized across threads, like:
>
> (define mutex-table (make-hash-table))
>
> (define mutex-table-mutex (make-mutex))
>
> (define (mutex-with-uuid uuid)
> (with-mutex mutex-table-mutex
> (or (hash-ref mutex-table uuid)
> (let ((mutex (make-mutex)))
> (hash-set! mutex-table uuid mutex)
> mutex))))
>
> If that looks OK, I can try to make a proper patch from it. I'm not
> sure what I'd use in place of `generate-uuid' though. Would `gensym' be
> good enough?
You're totally right on all points. Please do prepare a patch :) I
wish we could do something faster for the "embedded" mutex but
correctness should come first.
Cheers,
Andy
^ permalink raw reply [flat|nested] 5+ messages in thread
* bug#21887: 'monitor' form broken
2016-06-24 16:04 ` Andy Wingo
@ 2016-06-25 14:51 ` Taylan Ulrich Bayırlı/Kammer
2016-06-27 7:33 ` Andy Wingo
2016-06-27 7:33 ` Andy Wingo
0 siblings, 2 replies; 5+ messages in thread
From: Taylan Ulrich Bayırlı/Kammer @ 2016-06-25 14:51 UTC (permalink / raw)
To: Andy Wingo; +Cc: 21887
[-- Attachment #1: Type: text/plain, Size: 387 bytes --]
Here's a patch, tested minimally by running
(par-for-each (lambda (x)
(monitor
(sleep 1)
(display "foo\n")))
(iota 10))
on a quad-core. Previously it would print the "foo"s in groups of four
with a second between each group; now it prints them one by one with a
second between each, as should be.
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-Fix-monitor-macro.patch --]
[-- Type: text/x-diff, Size: 1471 bytes --]
From 08c7f4cd98c86fbb6551c7c0b6f17262c67e7b23 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Taylan=20Ulrich=20Bay=C4=B1rl=C4=B1/Kammer?=
<taylanbayirli@gmail.com>
Date: Sat, 25 Jun 2016 16:43:36 +0200
Subject: [PATCH] Fix 'monitor' macro.
* module/ice-9/threads.scm (monitor-mutex-table)
(monitor-mutex-table-mutex, monitor-mutex-with-id): New variables.
(monitor): Fix it.
---
module/ice-9/threads.scm | 21 ++++++++++++++++++---
1 file changed, 18 insertions(+), 3 deletions(-)
diff --git a/module/ice-9/threads.scm b/module/ice-9/threads.scm
index 9f9e1bf..14da113 100644
--- a/module/ice-9/threads.scm
+++ b/module/ice-9/threads.scm
@@ -85,9 +85,24 @@
(lambda () (begin e0 e1 ...))
(lambda () (unlock-mutex x)))))
-(define-syntax-rule (monitor first rest ...)
- (with-mutex (make-mutex)
- first rest ...))
+(define monitor-mutex-table (make-hash-table))
+
+(define monitor-mutex-table-mutex (make-mutex))
+
+(define (monitor-mutex-with-id id)
+ (with-mutex monitor-mutex-table-mutex
+ (or (hashq-ref monitor-mutex-table id)
+ (let ((mutex (make-mutex)))
+ (hashq-set! monitor-mutex-table id mutex)
+ mutex))))
+
+(define-syntax monitor
+ (lambda (stx)
+ (syntax-case stx ()
+ ((_ body body* ...)
+ (let ((id (datum->syntax #'body (gensym))))
+ #`(with-mutex (monitor-mutex-with-id '#,id)
+ body body* ...))))))
(define (par-mapper mapper cons)
(lambda (proc . lists)
--
2.8.4
^ permalink raw reply related [flat|nested] 5+ messages in thread
* bug#21887: 'monitor' form broken
2016-06-25 14:51 ` Taylan Ulrich Bayırlı/Kammer
@ 2016-06-27 7:33 ` Andy Wingo
2016-06-27 7:33 ` Andy Wingo
1 sibling, 0 replies; 5+ messages in thread
From: Andy Wingo @ 2016-06-27 7:33 UTC (permalink / raw)
To: Taylan Ulrich "Bayırlı/Kammer"; +Cc: 21887
On Sat 25 Jun 2016 16:51, taylanbayirli@gmail.com (Taylan Ulrich "Bayırlı/Kammer") writes:
> Here's a patch, tested minimally by running
>
> (par-for-each (lambda (x)
> (monitor
> (sleep 1)
> (display "foo\n")))
> (iota 10))
>
> on a quad-core. Previously it would print the "foo"s in groups of four
> with a second between each group; now it prints them one by one with a
> second between each, as should be.
Applied, thanks!
Andy
^ permalink raw reply [flat|nested] 5+ messages in thread
* bug#21887: 'monitor' form broken
2016-06-25 14:51 ` Taylan Ulrich Bayırlı/Kammer
2016-06-27 7:33 ` Andy Wingo
@ 2016-06-27 7:33 ` Andy Wingo
1 sibling, 0 replies; 5+ messages in thread
From: Andy Wingo @ 2016-06-27 7:33 UTC (permalink / raw)
To: 21887-done
thanks
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2016-06-27 7:33 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-11-12 15:29 bug#21887: 'monitor' form broken Taylan Ulrich Bayırlı/Kammer
2016-06-24 16:04 ` Andy Wingo
2016-06-25 14:51 ` Taylan Ulrich Bayırlı/Kammer
2016-06-27 7:33 ` Andy Wingo
2016-06-27 7:33 ` Andy Wingo
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).