I wonder if there have been any updates on this problem.

Dear Peter, I just wanted to add to this discussion that the single precision version of the code get the same results as the CPU and Reference platforms. With double precision we get other errors in other computers. (Quadro K600) $ python simplified_code.py TotalEnergy 120.0 kJ/mol $ python simplif...

Yes. I'm implementing the Cross-Stacking term of the 3SPN.2C potential. The Cross-Stacking potential is a function of 5 particles: B0, S1, B1, S2 and B2, where B0, S1, and B1 are part of one chain and S2 and B2 are part of another chain (B stands for base and S for Sugar). The potential is a functio...

Hi Peter, Sorry for the late response. I tried to use the compute Angle function, but I don't seem to be able to reach it from the python API or the XML setup. I want to compute the angle between 2 vectors defined by 4 particles, so the angle(...) function from customHBondForce doesn't meet my needs...

Thank you.
Is there a way to incorporate that computeAngle function on a CustomHbond force expresion using the python API?

Hi, I was working with the CustomHbond force using an acos function to calculate the angle between two vectors. In order to compute the angle between two vectors I used the function: t3 = acos(cost3lim); cost3 = sin(t1)*sin(t2)*cos(phi)-cos(t1)*cos(t2); This function shouldn't be more than 1 or less...

Yes, I just noticed that it may be easier to see the error with the sum of force on the system.
On this simulation the error becomes noticeable around frames 75-76.
The particle with indices 1235, 1236, 1237 starts drifting away.
I attached the states.

When I traced back the particles that move fast during the simulation, they accelerate when they are still far from the Hbond aceptor. They don't seem to accelerate as a result of a collision, and they don't seem to be attracted to any other particle. The cutoff for the CustomHBondFoce is 12 nm, whi...

The temperature should raise faster on OpenCl than on the CPU platform. Temperature_increase.png For example on the CPU platform at 3 000 000 steps: #"Step","Time (ps)","Potential Energy (kJ/mole)","Kinetic Energy (kJ/mole)","Total Energy (kJ/mole)","Temperature (K)" 3000000,300000.0000019568,-355.8...

I was using the OpenCL platform. I tried to use the CPU platform and the Reference platform but I could not reproduce the unexpected results.

I attach an example of the problem. Some CAM particles outside the non-bonded cutoff seem to experience a high force.