SMP2 enters next phase of testing at Stanford!

By , the in Folding@Home Project - No comment
It has been announced on the official Folding@Home project blog that SMP2 is currently undergoing beta testing. But what is it?

SMP1 was a clever method of simulating true multithreading on multicore systems by using the MPI API in conjunction with a modified monocore (the Gromacs 3.99 core, to be exact). This method has been used since 2007, with some success, but has also shown some limits, especially on the stability side of things.

SMP2 is revolutionary, because it is a major overhaul of the Gromacs method, and uses the native capabilities of multicore CPUs. In terms of reliability, we should once again find ourselves on less treacherous ground, as the intermediary layer that is MPI - which could occasionally crash the core and/or screw up the calculations - has been removed. This core has provided a great deal of grief to Peter Kasson and the Gromacs development team.

Though... are we speaking of a single core? No; we must mention SMP2 cores. For the moment, only SMP2 A3 is being tested. This one is the heir to the SMP1 A2 MacOSX/Linux core. But other methods of calculations are also being considered and directly implemented in SMP2 in the form of the core known as Desmond.

In other important news, SMP2 also makes widespread use of the bonus system, which until recently was reserved for BigAdv units. This means the faster a unit is completed and returned to Stanford, the greater the points reward will be. We do not yet know what the basis for calculating points will be, or how performance will be positioned relative to the A1 and A2 cores today.

We will keep you informed as soon as we have more information.

Source: The Official Blog