Clearly, a superior strategy is to compute possible scenarios well ahead of time using a cloud or at least an informal network of processors, and store the results on an easily distributed medium. After an event one might not have connectivity to massive processing resources or the electricity to use them. It is reasonable to expect that regardless of whether CPUs are hard at work responding to queries from all over the world about the status of individual people or collating damage and casualty reports, execution cycles will be at a premium. Ultra-parallel programmers would take advantage of the mostly idle GPUs. To save time during compute-intensive mega-moments
we ruthlessly exploit the marvelous parallel computing potential of GPUs. Of general interest to any of our exanomic modelling colleagues,
but especially those in Mozambique, is the consideration that once AIDS obtains a significant presence (perhaps as low as 1% of adults; certainly 2.5%) traditional rho statistics like product moment correlations are no longer accurate or even very useful. AIDS introduces some very nasty non-linear effects into an environment already contaminated with manipulated variables (currency exchange rates are not random when there is pegging, for example) and weakened by violations of joint normalcy. One must use eta statistics (correlation ratio), which take a bit more computing than rho. As there aren't any hemorrhagic plagues like the Black Death sweeping across Planet Earth, as far as we know,
no other factor besides AIDS can singlehandedly wreck an economy.
This first step is to have the application determine what GPU resources are available. While one might complain about how the intrinsic C function _cpuid has been implemented and extended over the years GPUs weren't even a gleam in the eyes of Kernighan and Ritchie long ago. As far as we can determine, GPU information is NOT available through WMI using Win32_Processors nor in the Registry at SYSTEM\CurrentControlSet\Control\Video.