Issaku Kanamori

Hi Peter, thank you very much for taking your time, the force and smearing are slow. The solvers are fine. Here is an example of comparison with 6 KNL (...

Thanks so much, I will wait.

Hi Peter, thank you very much, I will try it. Issaku

It runs much faster than before (great!), but the result is wrong: the Hamailtonian is no longer coservred (dH=0.00583017617464066 --> 2125889.73222768). I will pin down and report which part causes...

Hi, it seems the forces for domainwall/moebius fermion have problem. I tried tests/core, tests/hmc, tests/forces, and found some tests fail. (see below). Since I have not configured fft, so the...

with an earlier version https://github.com/paboyle/Grid/commit/07c0c02f8c1d58605150c4729f7d1b3b32416045 (the latest version takes too long to compile, by the way, some object file take almost two hours and the result is not yet coming)...

Hi Peter, thank you so much, nice to hear that you have reproduced the problem.

Hi Peter, I confirmed that the forces are all fine, with Unified virtual memory : yes I will continue with checking HMC. Thank you very much!

I confirmed that the hmc tests are fine, and my original HMC (2+1 flavor Mobius with stout smearing) also gives the correct result. Thanks so much! The performance (with 16...