float &fy,
float &fz)
{
-#if (GMX_PTX_ARCH >= 300)
if (!(order & (order - 1))) // Only for orders of power of 2
{
const unsigned int activeMask = c_fullWarpMask;
}
}
else
-#endif
{
// We use blockSize shared memory elements to read fx, or fy, or fz, and then reduce them to fit into smemPerDim elements
// which are stored separately (first 2 dimensions only)
const int atomIndexGlobal = atomIndexOffset + atomIndexLocal;
/* Early return for fully empty blocks at the end
- * (should only happen on Fermi or billions of input atoms)
+ * (should only happen for billions of input atoms)
*/
if (atomIndexOffset >= kernelParams.atoms.nAtoms)
{