Dear Keith,
Am Donnerstag, 22. Oktober 2009 01:00 schrieben Sie:
> Mr. Stimming, we are attempting to use lapack++ in our application. We are
> working in a multi-thread environment and are experiencing issues
Please define "experiencing issues".
> with our data results and we believe that thread issues in lapack++ may be
> the cause. To your knowledge, is lapack++ thread safe?
Please define "thread safe". The data container classes (LaGenMatDouble etc)
in lapack++ are just that - data container classes. Those are not protected
against access from multiple threads concurrently for good reasons (way way
too high runtime cost in the normal use case). If you planned to access any
of the data classes from multiple threads concurrently, you must set up your
multi-threading protection on top of the lapack++ classes.
Other than that, all computational functions are thread safe in the sense that
as long as they work on different container class objects, calling the
computational functions from multiple threads concurrently will not access
any common (e.g. static) storage and hence is thread safe. In particular, any
temporary workspace memory is allocated in each function call separately
which means this is indeed thread safe.
Regards,
Christian Stimming
> Any response that you
> can give would be helpful. Thanks.
>
>
>
> Keith Blonquist
>
> Junior Engineer
>
> Lidar Pacific Corporation
|
