From: "U.Nakamura" <usa@...>
Date: 2012-02-28T00:51:10+09:00
Subject: [ruby-core:42974] Re: 50$ for answering my stackoverflow question about Ruby GIL 500 reputation stackoverflow question about Ruby GIL 50$ for answering my stackoverflow question.

Hello,

In message "[ruby-core:42972] 50$ for answering my stackoverflow question about Ruby GIL 500 reputation stackoverflow question about Ruby GIL 50$ for answering my stackoverflow question."
    on Feb.27,2012 23:00:14, <grigory.v.p@gmail.com> wrote:
> Ok, seems no one needs stackoverflow reputation :(. So i mirrored my
> question as oDesk job posting here:
> https://www.odesk.com/jobs/200703042
> If someone can spare a few minutes and give a comprehensive answer to
> this question i will gladly pay via oDesk or PayPal.

Here is a patch for 1.9.3-p0 (because you said that you are
using the version):

--- thread.c.orig	2012-02-28 00:37:55.894924200 +0900
+++ thread.c	2012-02-28 00:45:11.171820700 +0900
@@ -1312,7 +1312,7 @@ rb_threadptr_execute_interrupts_common(r
 	}
 
 	if (timer_interrupt) {
-	    unsigned long limits_us = 250 * 1000;
+	    unsigned long limits_us = 10 * 1000;
 
 	    if (th->priority > 0)
 		limits_us <<= th->priority;
--- thread.c.orig	2012-02-28 00:37:55.894924200 +0900
+++ thread.c	2012-02-28 00:45:11.171820700 +0900
@@ -1312,7 +1312,7 @@ rb_threadptr_execute_interrupts_common(r
 	}
 
 	if (timer_interrupt) {
-	    unsigned long limits_us = 250 * 1000;
+	    unsigned long limits_us = 10 * 1000;
 
 	    if (th->priority > 0)
 		limits_us <<= th->priority;
--- thread_pthread.c.orig	2012-02-28 00:40:35.453050400 +0900
+++ thread_pthread.c	2012-02-28 00:45:25.791656900 +0900
@@ -1114,7 +1114,7 @@ close_communication_pipe(void)
 /* 100ms.  10ms is too small for user level thread scheduling
  * on recent Linux (tested on 2.6.35)
  */
-#define TIME_QUANTUM_USEC (100 * 1000)
+#define TIME_QUANTUM_USEC (10 * 1000)
 
 static void *
 thread_timer(void *p)


But I don't recommend to use it because this is only a workaround,
so you don't have to pay for me :)


Regards,
-- 
U.Nakamura <usa@garbagecollect.jp>