etmc / tmLQCD

tmLQCD is a freely available software suite providing a set of tools to be used in lattice QCD simulations. This is mainly a HMC implementation (including PHMC and RHMC) for Wilson, Wilson Clover and Wilson twisted mass fermions and inverter for different versions of the Dirac operator. The code is fully parallelised and ships with optimisations for various modern architectures, such as commodity PC clusters and the Blue Gene family.
http://www.itkp.uni-bonn.de/~urbach/software.html
GNU General Public License v3.0
32 stars 47 forks source link

LapH #161

Open ggscorzato opened 12 years ago

ggscorzato commented 12 years ago

I plan to remove the option of LapH from the master branch and start a LapH-only branch. From LapH-only I will slowly remove what is not needed. The idea is that LapH-only should one day be merged into other analysis codes like Ahmidas or nissa.

Any comment?

Luigi

urbach commented 12 years ago

currently I'd really like to keep it in the main branch, as I'd like to use it, if that is possible.

ggscorzato commented 12 years ago

ok. I have used it and it works. You are welcome to use it.

urbach commented 12 years ago

and how?

ggscorzato commented 12 years ago

the code was broken, after—I believe—standardizing the complex in the rest of the code, which was not ported to all lapH routines. I have fixed it. Tomorrow I check how to commit (I haven't committed with git yet...) I will also commit a typical LapH.input file with more instructions.

urbach commented 12 years ago

okay, would be great...

our procedure is that you prepare a branch with your changes in your own tmLQCD fork. Then you open a pull request which we discuss and check, before it will be included in the master branch...

kostrzewa commented 12 years ago

I have also omitted OpenMP parallelization for many (or all) of the LapH routines because I didn't know if they were still in use and I already had hundreds of functions to worry about...

ggscorzato commented 12 years ago

I don't think that openMP parallelization is necessary for LapH. It is not as performance-critical as the rest. MPI parallelization was necessary, but it will be enough. It would be nice to add a check in the configure that openmp and laph are not enabled together.

I am going to make a pull request for some bug fixes in laph (I have to find out where to click in github ... :-) )

kostrzewa commented 12 years ago

It would be nice to add a check in the configure that openmp and laph are not enabled together.

that's not strictly necessary, the parallelization doesn't affect correctness anyway, it would just make things slow when running on BG/Q say, because one thread will do the work of 64 threads...

kostrzewa commented 12 years ago

I am going to make a pull request for some bug fixes in laph (I have to find out where to click in github ... :-) )

you have to navigate to your branch which contains the changes you wish to submit and click "create pull-request". There is also a convenience button on the main view of your fork which suggests creating a pull-request for your most recently pushed-to branch.