2021-08-06 22:55
hkaiser changed the topic of #ste||ar to: STE||AR: Systems Technology, Emergent Parallelism, and Algorithm Research | stellar-group.org | HPX: A cure for performance impaired parallel applications | github.com/STEllAR-GROUP/hpx | This channel is logged: irclog.cct.lsu.edu
01:09
hkaiser has quit [Quit: Bye!]
01:28
<
gonidelis[m] >
i know there is ctad
01:28
<
gonidelis[m] >
but what about functions?
01:29
<
gonidelis[m] >
1. is there function tempalte argument deduciton?
01:29
<
gonidelis[m] >
2. how can i force the compiler to show me what the "3" arguments that I am passing are?
01:29
<
gonidelis[m] >
hkaiser:
01:46
Yorlik_ has joined #ste||ar
01:49
Yorlik__ has quit [Ping timeout: 265 seconds]
02:03
<
gonidelis[m] >
what! `std::cout << std::is_same<std::remove_reference<unsigned int &>, unsigned int>::value; \\false`
04:01
HHN93 has joined #ste||ar
04:03
HHN93 has quit [Client Quit]
06:24
Yorlik_ has quit [Read error: Connection reset by peer]
06:24
K-ballo has quit [Ping timeout: 260 seconds]
06:24
K-ballo1 has joined #ste||ar
06:25
Yorlik_ has joined #ste||ar
06:27
K-ballo1 is now known as K-ballo
09:00
karamemp[m] has quit [Quit: You have been kicked for being idle]
09:00
PoornaGunathilak has quit [Quit: You have been kicked for being idle]
09:00
gdaiss[m] has quit [Quit: You have been kicked for being idle]
10:19
<
satacker[m] >
<gonidelis[m]> "what! `std::cout << std::..." <- std::remove_reference_t
10:25
<
satacker[m] >
> <@gonidelis:matrix.org> 1. is there function tempalte argument deduciton?
10:25
<
satacker[m] >
> 2. how can i force the compiler to show me what the "3" arguments that I am passing are?
10:25
<
satacker[m] >
if there's so suitable candidate the compiler lists all the candidates and also lists how the function was instantiated
11:48
HHN93 has joined #ste||ar
11:48
<
HHN93 >
Executable /hpx/build/bin/set_operations_3442_test does not exist <_io.TextIOWrapper name='<stderr>' mode='w' encoding='utf-8'>
11:48
<
HHN93 >
any idea what this test failure means?
11:48
<
HHN93 >
I have tried building the test executable and it does build
11:49
HHN93 has quit [Client Quit]
11:49
<
satacker[m] >
please paste the complete error somewhere including CLI args
11:56
HHN93 has joined #ste||ar
11:56
<
HHN93 >
13/179 Test #504: tests.regressions.modules.algorithms.set_operations_3442 ............................................***Failed 0.04 sec
11:56
<
HHN93 >
Executable /hpx/build/bin/set_operations_3442_test does not exist <_io.TextIOWrapper name='<stderr>' mode='w' encoding='utf-8'>
11:56
<
HHN93 >
encountered the error during CI checks
11:57
<
satacker[m] >
it does not exist, probably it has not been built
11:57
<
satacker[m] >
it would be better if you send the CI log link
11:59
<
HHN93 >
I did change a few things in that file test
11:59
<
HHN93 >
ok got the error
12:06
HHN93 has quit [Quit: Client closed]
12:38
hkaiser has joined #ste||ar
13:53
K-ballo1 has joined #ste||ar
13:54
K-ballo has quit [Ping timeout: 264 seconds]
13:54
K-ballo1 is now known as K-ballo
15:11
parsa[fn] has joined #ste||ar
15:56
scofield_zliu has joined #ste||ar
16:00
KhushiBalia[m] has quit [Quit: You have been kicked for being idle]
16:02
scofield_zliu has quit [Ping timeout: 252 seconds]
19:07
HHN93 has joined #ste||ar
19:08
<
HHN93 >
where exactly does hpx request for a CPU thread pool or GPU threads to run the algorithm?
19:24
<
hkaiser >
HHN93: in the executor
19:30
<
HHN93 >
do we provide an API to request for a pool of HPX threads?
19:31
<
hkaiser >
the index_queue_spawning is doing that, no?
19:31
<
hkaiser >
why do you ask?
19:32
<
HHN93 >
untill now I have been running parallel tasks with HPX by finding number of cores using (processing_units_count) and launching those many async tasks
19:32
<
HHN93 >
is that the ideal way of doing it?
19:32
<
hkaiser >
not always
19:33
<
hkaiser >
this works well if your tasks are large enough to compensate for the overheads involved
19:33
<
HHN93 >
ok, that's how most of the hpx algorithms do it, right?
19:33
<
hkaiser >
no, the algorithms rely on the executor to spawn tasks - those also decide how that's done
19:35
<
hkaiser >
my talk at ccpcon last year has the basics explained
19:38
<
HHN93 >
for most algorithms I have observed they call bulk_async_execute using the policy.executor()
19:38
<
HHN93 >
does the policy.executor() mention where the processing happens?
19:39
<
hkaiser >
policy.executor() returns a reference to the executor used to do the spawning
19:44
HHN93 has quit [Quit: Client closed]
22:23
K-ballo1 has joined #ste||ar
22:24
K-ballo has quit [Ping timeout: 255 seconds]
22:24
K-ballo1 is now known as K-ballo