commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Herbert <alex.d.herb...@gmail.com>
Subject Re: [RNG] Utility for displaying progress (Was: [...] estimate for stress test tasks)
Date Mon, 14 Oct 2019 23:09:40 GMT


> On 14 Oct 2019, at 23:16, Gilles Sadowski <gilleseran@gmail.com> wrote:
> 
> Hello Alex.
> 
>> [...]
>> 
>> commit 7177cdc82fd5020442f833ea8947151366dec725
>> Author: Alex Herbert <aherbert@apache.org>
>> AuthorDate: Mon Oct 14 21:26:48 2019 +0100
>> 
>>    Better time estimate for stress test tasks.
>> ---
>> .../rng/examples/stress/StressTestCommand.java     | 35 ++++++++++++++++++----
>> 1 file changed, 30 insertions(+), 5 deletions(-)
>> 
>> [...]
> 
> Did you investigate whether some coding project already provides
> such a utility, that could be reused?

No. Reusing a library would be nice if it works for the timing data generated by PractRand.

I had a simple version for estimating time remaining that worked well for Dieharder and BigCrush
since they always take about the same length of time. Using the mean execution time, the number
of concurrent tasks and the number of tasks left is pretty accurate.

When I recently ran the repeat of BigCrush it counted down from 96 hours remaining quite nicely.

I have been testing with PractRand and that can fail very quickly (1000x faster than a full
length run) on certain generators. This makes it difficult to estimate the time for a test
run using the mean or even the median.

It doesn’t really matter for this application that the estimate is correct. The stress test
application currently prints out how far it is through running all the tasks for each of the
generators in the list. The estimate is a nice to have but is not even recorded in the test
results. It prints to console as the tests are run.

I may look into a library if I have time. For now I tweak the code each time I run PractRand.
What is more important is to be able to run PractRand on my test machine with many parallel
jobs. The first attempt had out-of-memory errors as it approached 2GB per PractRand process
(on the C++ side) which exceeded the RAM on the machine with multiple jobs. I’ve changed
the settings to use less tests and will see what happens overnight.

Alex

> 
> Best,
> Gilles
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org


Mime
View raw message