|
|
Threads-Devel : |
From: Anthony Williams (anthony_at_[hidden])
Date: 2006-09-25 05:07:05
> @Anthony:
>
> The barrier and read_write_mutex obviously can be implemented on top of
> the core boost::thread primitives. So no platform split is needed. True?
In theory, yes. However, if you want to make use of the pthread native
read write mutex, then we need a platform split (pthreads vs everything
else). I'm also not sure whether or not we can make use of native Windows
stuff to make it work better.
> I did not yet try to understand the code, but your read_write_mutex
> is far less code lines than the original. Does this mean your code
> is not finished yet, it is semantically different or it is that cute? ;-)
My read_write_mutex code is not yet finished. Peter and Cory have provided
sample implementations, but I haven't yet looked at them. I also have an
older implementation that I wrote, which I will revisit, to check it
actually sorts all the issues.
> I am also wondering why you have put read_write_mutex.hpp into a
> platform dispatcher?
See above.
> Another question: Did you already try to run any of the regression tests
> on thread_rewrite? If so, could you please help me to get a head start?
Everything except TSS runs fine on MSVC 7.1 & 8.0. They also ran fine on
gcc 4 on Windows last time I tried, but it was a while ago.
Or did you mean something else?
Anthony
-- Anthony Williams Just Software Solutions Ltd http://www.justsoftwaresolutions.co.uk