How to wait for thread?

[Brett Romero]

I'd like the main thread to wait until these three threads have
completed their task:

LoadOjbects lo = new LoadOjbects();

lo.thrd1.Start();
lo.thrd2.Start();
lo.thrd3.Start();

while (lo.thrd1.IsAlive || lo.thrd2.IsAlive || lo.thrd3.IsAlive)
{
Thread.Sleep(500);
}

The above works but adds a lot of time to the overall.

Thanks,
Brett

 

[Carl Daniel [VC++ MVP]]

I'd like the main thread to wait until these three threads have
completed their task:

LoadOjbects lo = new LoadOjbects();

lo.thrd1.Start();
lo.thrd2.Start();
lo.thrd3.Start();

while (lo.thrd1.IsAlive || lo.thrd2.IsAlive || lo.thrd3.IsAlive)
{
Thread.Sleep(500);
}

If you just want to wait for all three to end, without regard for the order
in which they end, you can just do:

lo.thrd1.Join();
lo.thrd2.Join();
lo.thrd2.Join();
// You won't get here until all 3 threads have completed
// ...but you can't tell which one completed first
// (nor does it matter).

-cd

 

[Brett Romero]

Thanks. That seems to be working fine. I'm running different task on
each thread. It is taking the same amount of time as if I were doing
everything on the same thread. Here's what I'm doing in the
LoadObjects class:

thrd1 = new Thread( delegate( )
{
//do something here
} );

....and so on for the others

Why would it take the same amount of time? This is a single cpu (Intel
Pentium D 2.8Ghz) WinXP box.

Thanks,
Brett

 

[Nicholas Paldino [.NET/C# MVP]]

Brett,

Well, you mentioned it yourself. You are on a single cpu. Because of
this, the tasks are all sharing the CPU. Each thread will get a slice of
CPU time.

Using threads doesn't guarantee that your program will run faster. Even
if you had a multi-cpu system, it isn't guaranteed. You could have slower
performance due to context switches, for example.

Hope this helps.

 

[Brett Romero]

Hope this helps.

Yes. I was hoping hyperthread or something would come into play but I
guess it more just moving things around to different slices of
execution.

Thanks,
Brett

 

[Brett Romero]

Are there any single core cpus that multi threading can take advantage
of?

Thanks,
Brett

 

[Carl Daniel [VC++ MVP]]

Are there any single core cpus that multi threading can take advantage
of?

You can get some performance gains from a single core hyperthread CPU - if
your program actually has real opportunities for parallelism in it. If your
program is I/O intensive (e.g. reading and writing a lot of files), then
you're unlikely to see significant improvement with multi-threading.

-cd

 

[Brett Romero]

Lots of loading data from the data base (read/writes) for each thread.
So I guess that won't help much.

Thanks,
Brett

 

[Carl Daniel [VC++ MVP]]

Lots of loading data from the data base (read/writes) for each thread.
So I guess that won't help much.

It still might - only testing will tell. If your database is remote
(accessed over a network), then there may be significant opportunities for
improvement parallelism due to network latency (as opposed to actual I/O).
For example, I recently made a change to a web/database app to issue a
series of queries to the database in parallel rather than one after another.
The same total amount of I/I occurred, but the parallelism effectively
removed the network latency from the equation, so the overall process sped
up almost linearly up to ~25 parallel requests. If your database is local,
or you're interacting with local files, then latency is likely not a major
factor and parallel execution will probably have little impact.

-cd

 

[Brett Romero]

It still might - only testing will tell. If your database is remote

(accessed over a network), then there may be significant opportunities for
improvement parallelism due to network latency (as opposed to actual I/O).
For example, I recently made a change to a web/database app to issue a
series of queries to the database in parallel rather than one after another.
The same total amount of I/I occurred, but the parallelism effectively
removed the network latency from the equation, so the overall process sped
up almost linearly up to ~25 parallel requests. If your database is local,
or you're interacting with local files, then latency is likely not a major
factor and parallel execution will probably have little impact.

The database is remote (local network). I don't see any improvement on
my machine. A co worker gets about a 5 second gain. So from 29
seconds down to 24.3. He has a a single Centrino Mobile Pentium M 2 G
processor (laptop).

On my machine, it is 31 and 32.7 (a small increase in time with multi
threading).

On another co workers (slow) machine, it is 2 minutes 48 seconds down
to 58 seconds.

Why do you think there are such differences between our machines?

Thanks,
Brett