Re^5: The most precise second (timer)

[misc]

..Again, in pseudocode, this would work:

sub threadwork(){
    lock(MUTEX);
    do{ 
         the work;
    }
}

$t = new threadpool(42 threads);
$t->do( &threadwork() );

lock(MUTEX);
loop{
    ct = timestamp();
    sleep(0.9 seconds); # do a blocking select with timeout in perl fo
+r microsecond timings
    loop {nop;} while (ct==timestamp()); # the spinloop 
         #or, for a interval of 5 seconds: while ( ct+5>timestamp() );
    unlock(MUTEX);
}

Hope that helps

[marioroy]

Hello! I have been following this thread. MCE has yield.

Today, released MCE 1.864 and MCE::Shared 1.864 to use CLOCK_MONOTONIC, if available. Yield gives other workers a chance to run, but helpful to throttle workers as well. Due to the serial nature, yield can also be used to do something periodically.

use strict;
use warnings;

use MCE::Child 1.864;

sub task {
    while () {
        MCE::Child->yield(1.0);
        my $time = MCE::Util::_time();

        printf "pid $$, time %12.06f\n", $time;
    }
}

MCE::Child->create(\&task) for 1..5;
MCE::Child->wait_all;

Output from a Linux VM.

pid 3245, time   151.518361
pid 3246, time   152.518277
pid 3248, time   153.518325
pid 3247, time   154.518268
pid 3249, time   155.518331
pid 3245, time   156.518396
pid 3246, time   157.518181
pid 3248, time   158.518237
pid 3247, time   159.518258
pid 3249, time   160.518230
pid 3245, time   161.518247
pid 3246, time   162.518074
pid 3248, time   163.518094
pid 3247, time   164.518285
pid 3249, time   165.518338
pid 3245, time   166.518081
pid 3246, time   167.518232
pid 3248, time   168.518047
pid 3247, time   169.518250
pid 3249, time   170.518334
pid 3245, time   171.518336
pid 3246, time   172.518232
pid 3248, time   173.518244
pid 3247, time   174.518247
pid 3249, time   175.518338
pid 3245, time   176.518231
pid 3246, time   177.518231
pid 3248, time   178.518238
pid 3247, time   179.518238
pid 3249, time   180.518334
pid 3245, time   181.518237
pid 3246, time   182.518384
pid 3248, time   183.518225
pid 3247, time   184.518203
pid 3249, time   185.518220
pid 3245, time   186.518086
pid 3246, time   187.518264
pid 3248, time   188.518229
pid 3247, time   189.518248
pid 3249, time   190.518250
...

Yield now works similarly for the manager process starting with 1.864. This allows the manager to send work periodically to a pool of workers.

use strict;
use warnings;

use MCE::Child 1.864;
use MCE::Channel;

my $chnl = MCE::Channel->new( impl => 'Mutex' );

sub task {
    while ( my $work = $chnl->recv() ) {
        my $time = MCE::Util::_time();
        printf "pid $$, time %12.06f\n", $time;
    }
}

MCE::Child->create(\&task) for 1..5;

for ( 1..40 ) {
    MCE::Child->yield(1.0);
    $chnl->send("work $_");
}

$chnl->end();

MCE::Child->wait_all();

Output from a Linux VM.

pid 9336, time  2111.790429
pid 9340, time  2112.790575
pid 9337, time  2113.790593
pid 9340, time  2114.790489
pid 9338, time  2115.790415
pid 9339, time  2116.790520
pid 9338, time  2117.790471
pid 9336, time  2118.790468
pid 9337, time  2119.790466
pid 9340, time  2120.790467
pid 9338, time  2121.790340
pid 9339, time  2122.790540
pid 9338, time  2123.790289
pid 9340, time  2124.790484
pid 9337, time  2125.790351
pid 9339, time  2126.790493
pid 9338, time  2127.790459
pid 9340, time  2128.790489
pid 9336, time  2129.790488
pid 9339, time  2130.790487
pid 9338, time  2131.790488
pid 9340, time  2132.790271
pid 9336, time  2133.790449
pid 9337, time  2134.790461
pid 9339, time  2135.790462
pid 9338, time  2136.790445
pid 9340, time  2137.790494
pid 9337, time  2138.790389
pid 9339, time  2139.790528
pid 9336, time  2140.790488
pid 9340, time  2141.790337
pid 9338, time  2142.790339
pid 9340, time  2143.790253
pid 9336, time  2144.790396
pid 9337, time  2145.790544
pid 9340, time  2146.790390
pid 9336, time  2147.790411
pid 9338, time  2148.790463
pid 9339, time  2149.790362
pid 9337, time  2150.790411

Ditto for MCE::Hobo.

Regards, Mario