I have been running successfully with the version of IT++ packaged with ubuntu
10.04. I then started using ubuntu 11.10 with weird results sometimes. I then
tried building 4.2 from source with same weird results as a result. Finally
after installing 4.0.7 I got my old results back. The weirdness manifested
itself as e.g. different values of random variables being generated despite
first reseting using the same seed. I know it can happen with multiple threads
- but my code should be single thread (unless something happens that I don't
understand).
I am using a Dell E6410 latitude with an "Intel(R) Core(TM) i5" CPU and run
64bit version of ubuntu 11.10.
BR/
Per
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
A second thought. The weird things I have seen have all involved binary
numbers (bvec). Could there possibly be something related to how they are
implemented in IT++ for 64bit processors .... just a wild guess.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am the maintainer of the IT++ packages for Debian (and thus, Ubuntu). You
need to give me a sample program to try out and see why this could be
happening. That said, the first thing I'd look at is whether you are using
vanilla lapack and blas, or the Atlas versions and whether this is causing
some adverse effects.
In any case, I'd suggest you try building your own version of IT++ and running
make check to run the tests.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi All,
I have been running successfully with the version of IT++ packaged with ubuntu
10.04. I then started using ubuntu 11.10 with weird results sometimes. I then
tried building 4.2 from source with same weird results as a result. Finally
after installing 4.0.7 I got my old results back. The weirdness manifested
itself as e.g. different values of random variables being generated despite
first reseting using the same seed. I know it can happen with multiple threads
- but my code should be single thread (unless something happens that I don't
understand).
I am using a Dell E6410 latitude with an "Intel(R) Core(TM) i5" CPU and run
64bit version of ubuntu 11.10.
BR/
Per
A second thought. The weird things I have seen have all involved binary
numbers (bvec). Could there possibly be something related to how they are
implemented in IT++ for 64bit processors .... just a wild guess.
Hi.
I am the maintainer of the IT++ packages for Debian (and thus, Ubuntu). You
need to give me a sample program to try out and see why this could be
happening. That said, the first thing I'd look at is whether you are using
vanilla lapack and blas, or the Atlas versions and whether this is causing
some adverse effects.
In any case, I'd suggest you try building your own version of IT++ and running
make check to run the tests.