Mailing-List: contact users-help@flex.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@flex.apache.org
Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 9.3 \(3124\))
Subject: Re: Multithreading
From: Justin Mclean <justin@classsoftware.com>
In-Reply-To: <1471140026513-13274.post@n4.nabble.com>
Date: Sun, 14 Aug 2016 12:41:39 +1000
Content-Transfer-Encoding: quoted-printable
Message-Id: <C35153B4-9356-4A71-8899-AE42850BF2CB@classsoftware.com>
References: <1471140026513-13274.post@n4.nabble.com>
To: users@flex.apache.org
archived-at: Sun, 14 Aug 2016 02:41:58 -0000

Hi,

> I am likely to write a simple command line C++ program that gets data =
out of
> a database, processes it, and then writes the data to a database table =
to be
> consumed by mobile devices.  I am very interested in seeing the speed
> difference in processing the same data in C++ versus my AIR =
application.

Given it a fairly straight forward calculation I=E2=80=99s probably go =
with C rather than C++ to avoid some of it=E2=80=99s complexity. It =
likely to run faster than in AIR.

> I currently know nothing about C++, so I don't know if it is like AIR =
and
> natively runs in a single thread, or if it somehow takes advantage of
> multiple threads on your computer.

By default no, in C you need to write code to take advantages of threads =
and threads are not part of the language. While reasonably =
straightforward [1] it can be a little tricky at times to get =
working/debug.

>  it would be more efficient for me if I was able to run multiple
> instances of the C++ application side by side so that I could process
> several cities of data at the same time.

Not more efficient but it would be simpler. A thread has less overhead =
than a process. It is however easier to communicate between threads than =
communicate between processes if you need to do that.

>  I was wondering what happens if you run several instances of a single =
threaded application at the same time?=20

The OS will give each of them some time on the number or cores it has. =
Remember there also other programs likely to be running at the same time =
so you are not going to get 100% of the CPU cores all the time just for =
just your code.

> Does the OS put each instance in its own thread

The OS works at a process level not a thread level. [2] The code works =
at a thread level.

For a problem like this something like Spark [3] is going to be a more =
scalable solution as it can scale over multiple cores and machines. The =
problem, basically taking a set of data and filtering it fits the map / =
reduce [4] concept quite well. There is however (at a guess) going to be =
a significant learning curve here.

Thanks,
Justin

1. http://www.geeksforgeeks.org/multithreading-c-2/
2. =
http://www.programmerinterview.com/index.php/operating-systems/thread-vs-p=
rocess/
3. http://spark.apache.org
4. https://en.wikipedia.org/wiki/MapReduce