Re: [boost] [thread] best practice to lock multiple mutexes

From: Andrey Tcherepanov (moyt63c02_at_[hidden])
Date: 2008-02-12 13:03:55

Hello,

On Tue, 12 Feb 2008 00:52:47 -0700, Anthony Williams  
<anthony_w.geo_at_[hidden]> wrote:

> Andrey Tcherepanov <moyt63c02 <at> sneakemail.com> writes:
>
>>
> How shall I get them to you? Each build is around 1Mb, and zips to  
> around 280k.

as long as zip file is not above couple megs, I think you can email me to  
andrey<dot>tcherepanov<at>gmail<dot>com . Or, even better, put it on  
vault, so if somebody else wants to rerun them on their machine can do it  
too.

> I have a variety of builds I would like to run: 3 mutex implementations,
> yield/no yield on each, with 5, 16 or 32 threads (less than, equal, and  
> greater
> than the number of cores). That's 18 combinations. Each should take a  
> lot less
> than a minute to run (5 threads on my dual-core machine takes around  
> 20-30s).
> Are you willing to run all those builds (18) several times each to get  
> reliable
> figures? If not, I'll try and pick a subset.

I think I will be able to run them all, just create a batch script that  
will run them in any sequence that you feel like. Just don't override  
result files with each run, so if time will permit I will rerun them more  
times.

One thing that I am not sure is if you need completely clean machine (no  
other processes running)? I am asking because these machines are parts of  
the db clusters (standby node in the active-passive cluster), and might  
have some small processes running anyway.

In regards of cores and cpus and hyperthreading -- I will add full specs  
of the machines to the results.

> Your assistance is much appreciated.
>
> Thanks,
>
> Anthony

You are very welcome.

Regards,
   Andrey