I recently learned of a new C-based language specifically designed to
use the massively-parallel multi-core graphics chips such as nVidia
and others produce for the game industry, among other uses.
Since I once worked with Transputers and the Occam parallel Fortran
(again "way back"), I wonder, are there any moves to do the same for
Fortran?.
Given the teraflop processing capability of the better graphics chips,
why do we still use compilers that really only work well wiith a
single core CPU?
DNA, RNA, decrypting and other pattern searching and list processing
problems are simply made for these chips.
Of course there's no I/O to match, but many "intractable" problems may
now be attacked without waiting for quantum computers to be
practicable tools.
|
