2018-06-19 22:05 AEST

View Issue Details Jump to Notes ]
IDProjectCategoryView StatusLast Update
0000333mercuryBugpublic2018-03-13 14:38
Reporterwangp 
Assigned Topbone 
PrioritynormalSeverityminorReproducibilityalways
StatusacknowledgedResolutionopen 
Product Version 
Target VersionFixed in Version 
Summary0000333: MR_num_idle_engines count goes wrong
DescriptionHi Paul, I think you can solve this faster than I can.

I've made a copy of the parallel fib benchmark, except that it repeats. Parallel execution works for the first fib call. Afterwards, fewer engines do any work. Soon, parallel execution never occurs again.

MR_num_idle_engines should be the culprit:

Program received signal SIGINT, Interrupt.
0x0000000000405dac in fib_par_plain_module1 () at fib_par_plain.c:206
206 MR_parent_sv(3) = MR_r1;
(gdb) p MR_num_idle_engines
$2 = -2
(gdb) c
Continuing.
fib(40) = 165580141
fib(40) = 165580141
^C
Program received signal SIGINT, Interrupt.
0x00007ffff6789d00 in sem_wait () from /lib64/libpthread.so.0
(gdb) p MR_num_idle_engines
$3 = 0
TagsNo tags attached.
Attached Files

-Relationships
+Relationships

-Notes

~0000700

pbone (administrator)

It looks very odd where MR_num_idle_engines is -2. I'll take a look but I think 0000334 should be done first and might even help with this bug.

Thanks.

~0000982

pbone (administrator)

Unassigned things from me.
+Notes

-Issue History
Date Modified Username Field Change
2014-06-16 11:41 wangp New Issue
2014-06-16 11:41 wangp Status new => assigned
2014-06-16 11:41 wangp Assigned To => pbone
2014-06-16 11:41 wangp File Added: fib_par_plain.m
2014-06-16 12:14 pbone Note Added: 0000700
2018-03-13 14:38 pbone Status assigned => acknowledged
2018-03-13 14:38 pbone Note Added: 0000982
+Issue History