Moved ForceEnergy into useful.h, added atomicAdd for ForceEnergy

Commented unwanted conditional in ComputeKernel that made simulations using the softcore code include self-interactions

Made rigid body grid--grid kernels use atomicAdd to communicate force/torque for entire kernel instead of each block writing data

(1) Protect dynamic memory allocations with conditionals to prevent zero-byte allocations

(2) Correctly initialize device data for particles_d using sizeof(int) instead of sizeof(int*)

Conflicts:
	src/ComputeForce.cu
	src/ComputeGridGrid.cu
	src/ComputeGridGrid.cuh
	src/GrandBrownTown.cu
	src/Makefile
	src/RigidBody.cu
	src/RigidBody.h
	src/RigidBodyController.cu
	src/RigidBodyController.h
	tests/argon-small/BrownDyn.bd
	tests/argon/BrownDyn.bd

hchou10 authored 7 years ago and cmaffeo2 committed 6 years ago

implement openMP option for rigid body simulation and fix bug in rigid body replicas, need to be reviewed and tested

Conflicts:
	src/GrandBrownTown.cu
	src/Makefile

hchou10 authored 7 years ago and cmaffeo2 committed 6 years ago

Conflicts:
	src/Configuration.cpp
	src/GrandBrownTown.cu

hchou10 authored 7 years ago and cmaffeo2 committed 6 years ago

Conflicts:
	src/RandomCUDA.cu

hchou10 authored 7 years ago and cmaffeo2 committed 6 years ago

Conflicts:
	src/Configuration.cpp
	src/GrandBrownTown.cu

Removed some cudaDeviceSync commands, but got unexpectedly worse performance; in profiler it looks like all but 1 devsyncs are still there, and a slow cudamemcpy appeared

Consolidated grid-particle memcpy operations, moving them to RigidBodyController using pinned memory