I'm a complete novice when it comes to these things -
but is it possible that somewhere, something is doing an optimization? Compiler? CPU? Perhaps some subsystem along the way sees that fair scheduling would take much longer. -mike On 6/10/07, Muli Ben-Yehuda <[EMAIL PROTECTED]> wrote:
On Sun, Jun 10, 2007 at 10:06:40AM +0300, Gilad Ben-Yossef wrote: > AFAIK sched_yeild() precious meaning is and has always been: "put me > at the end of the run queue for my priority. If there is no one else > with my priority, I'll run again". Quote akpm: "Changed sched_yield() semantics. sched_yield() has changed dramatically, and it can seriously impact existing applications. A testcase (this is on 2.5.46, UP, no preempt)". http://lkml.org/lkml/2002/12/2/155. > By calling sched_yielf(), the scheduler we re-arrange tun run queue > so that (assuming all worker threads are of the same priority), the > thread releasing the lock will be put at the back of the run queue > and some other threads, which ahs been marked ready to run after the > lock has been released will get the chance to take the lock, that's > why I expect sched_yield() to do the job. Note that for an SMP > machine things are rather more complicated and also if there are > numeroud tasks in different priroities... :-( Yes, you are correct. sched_yield() will give an appearance of fairness. But consider the cost (see below) - IMO it changes the semantics of the test program in such a way as to make it meaningless. > Shahar, let us know what happens :-) sched_yield() makes it run 20 times longer for TEST_SIZE=10000000, and the difference grows with the test size. moren:/home/muli # unset DO_SCHED_YIELD; time taskset 01 ./mf 5 Thread 40800940 started as thread #0 Thread 41001940 started as thread #1 Thread 41802940 started as thread #2 Thread 42003940 started as thread #3 Thread 42804940 started as thread #4 Results: 234626 1177504 1155320 2187538 5245012 Joined thread #0 Joined thread #1 Joined thread #2 Joined thread #3 Joined thread #4 real 0m1.187s user 0m1.000s sys 0m0.116s moren:/home/muli # export DO_SCHED_YIELD=1; time taskset 01 ./mf 5 using sched_yield() Thread 40800940 started as thread #0 Thread 41001940 started as thread #1 Thread 41802940 started as thread #2 Thread 42003940 started as thread #3 Thread 42804940 started as thread #4 Results: 106710 106711 106712 106711 106709 Results: 213147 213147 213148 213147 213147 Results: 319311 319312 319313 319312 319311 Results: 425650 425652 425653 425652 425651 Results: 531269 531270 531269 531269 531269 Results: 637252 637254 637253 637252 637253 Results: 743678 743681 743680 743679 743679 Results: 849960 849963 849962 849961 849961 Results: 957552 957554 957554 957552 957553 Results: 1065060 1065063 1065064 1065061 1065061 Results: 1172691 1172694 1172695 1172692 1172691 Results: 1280342 1280345 1280345 1280342 1280342 Results: 1388211 1388214 1388214 1388211 1388211 Results: 1495994 1495998 1495999 1495995 1495994 Results: 1603876 1603880 1603881 1603878 1603877 Results: 1711619 1711623 1711624 1711622 1711619 Results: 1819511 1819513 1819515 1819513 1819510 Results: 1927503 1927506 1927507 1927505 1927503 Results: 1999999 2000001 2000003 2000000 1999997 Joined thread #0 Joined thread #1 Joined thread #2 Joined thread #3 Joined thread #4 real 0m19.249s user 0m3.008s sys 0m15.909s #include <stdio.h> #include <stdlib.h> #include <pthread.h> #include <unistd.h> #define TEST_SIZE 10000000 volatile int results[TEST_SIZE], position=0; static int do_sched_yield = 0; int worker( int threadnum ) { static pthread_mutex_t mutex=PTHREAD_MUTEX_INITIALIZER; printf("Thread %lx started as thread #%d\n", pthread_self(), threadnum); pthread_mutex_lock(&mutex); while( position<TEST_SIZE ) { // printf("results[%d]=%d\n", position, threadnum); results[position++]=threadnum; pthread_mutex_unlock(&mutex); if (do_sched_yield) sched_yield(); pthread_mutex_lock(&mutex); } pthread_mutex_unlock(&mutex); return 0; } int *histo; void printres(int numthreads) { int i; printf("Results: "); for( i=0; i<numthreads; ++i ) printf("%d ", histo[i]); printf("\n"); } int main(int argc, char * argv[] ) { int numthreads=atoi(argv[1]); if( argc<2 || numthreads==0 ) exit(1); pthread_t *threads=calloc(sizeof(pthread_t), numthreads); histo=calloc(sizeof(int), numthreads); int i; int oldpos=position; if (getenv("DO_SCHED_YIELD")) { do_sched_yield = 1; printf("using sched_yield()\n"); } for( i=0; i<numthreads; ++i ) { pthread_create(threads+i, NULL, (void *(*)(void *))worker, (void *)(unsigned long)i); } /* Reap the results every second */ while( oldpos<TEST_SIZE ) { sleep(1); for( ; oldpos<position; ++oldpos ) { histo[results[oldpos]]++; } printres(numthreads); } for( i=0; i<numthreads; ++i ) { pthread_join( threads[i], NULL ); printf("Joined thread #%d\n", i); } return 0; } ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
