When performing these measurements, you may want to do your recordings to
linear memory and then print them afterwards.   I think that stdout
(printf) willould affect your measurements as gets proxied back through the
main thread.

On Mon, May 10, 2021 at 12:48 PM Sarang Karandikar <
[email protected]> wrote:

> For easier reading, this is what is happening:
>
>
> --------------------------------------------------------------------------------
> Thread T1(UI Thread)                    | Thread T2 (Non UI Thread)
>       |
>
> ----------------------------------------|---------------------------------------
> [A] PrintTime(); pthread_cond_wait()    |
>       |
>                                         | [X]
> PrintTime(); pthread_cond_signal()|
>                                         |
>       |
> [B] PrintTime();                        |
>       |
>
> --------------------------------------------------------------------------------
>
> The current times printed at [A] [X] and [B] are:
> [A]: t0
> [X]: t0 + k
> [B]: t0 + k + [1 second - 2 seconds],
>
> The extra delay of 1-2 seconds at [B] is the issue, ideally it should be
> no more
> than the scheduler delay (10-15 ms).
>
> Thanks,
> Sarang.
>
>
>
> On Fri, May 7, 2021 at 5:13 PM SK <[email protected]> wrote:
>
>> Hello,
>>
>> With latest EMSDK 2.0.20 I'm trying to use wait and signal using
>> pthread_cond_signal() and pthread_cond_wait() functions.
>>
>> I'm consistently seeing this issue between two non-UI threads that after
>> a pthread_cond_signal() , a previously waiting pthread_cond_timedwait()
>> only comes out after 1-2 seconds delay, expectation is that it should be
>> ready at max within the scheduling interval.
>>
>> The first row shows time just before entering wait(), later are pairs of
>> times of when actual event is generated and point just before signal() is
>> called, there's barely any difference. Many such events pass until we come
>> out of wait and see the last 'Post wait' is printed, it takes 1.56 seconds
>> to return even though event is already signaled. All times are epoch times
>> retrieved via clock_gettime(CLOCK_REALTIME, ...) option.
>>
>> Wait start time 1620415241.405
>>
>> Event generation time 1620415241.859  0.454
>> Signaling time 1620415241.859 0.454
>>
>> Event generation time 1620415242.300 0.895
>> Signaling time 1620415242.300 0.895
>>
>> Event generation time 1620415242.329 0.924
>> Signaling time 1620415242.329 0.924
>>
>> Event generation time 1620415242.465 1.060
>> Signaling time 1620415242.465 1.060
>>
>> Event generation time 1620415242.738 1.333
>> Signaling time 1620415242.738 1.333
>>
>> Event generation time 1620415242.833 1.428
>> Signaling time 1620415242.833 1.428
>>
>> Event generation time 1620415243.163 1.758
>> Signaling time 1620415243.163 1.758
>>
>> Event generation time 1620415243.248 1.843
>> Signaling time 1620415243.249 1.844
>>
>> Post wait time 1620415243.418 2.013 1.559
>>
>> Any pointers would be greatly appreciated.
>>
>> Thanks,
>> Sarang.
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "emscripten-discuss" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/emscripten-discuss/0e73db37-aa12-43fc-948e-ef9622f2dc03n%40googlegroups.com
>> <https://groups.google.com/d/msgid/emscripten-discuss/0e73db37-aa12-43fc-948e-ef9622f2dc03n%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
> --
> You received this message because you are subscribed to the Google Groups
> "emscripten-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/emscripten-discuss/CAHqdETwuTPDxPjBVmCy%2BjU5p3UKxsiwvM9R1ERFURJmrH85p0g%40mail.gmail.com
> <https://groups.google.com/d/msgid/emscripten-discuss/CAHqdETwuTPDxPjBVmCy%2BjU5p3UKxsiwvM9R1ERFURJmrH85p0g%40mail.gmail.com?utm_medium=email&utm_source=footer>
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"emscripten-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/emscripten-discuss/CAL_va2_-LCNmyabXDfkM7R0ULXWBWGkSSLL3PGcfVff0GSEKMQ%40mail.gmail.com.

Reply via email to