its too bad cpu makers dont build in some functionality to let us use the vector units as native float128 / float256 / etc
When FPGA caches become standard on new CPUs it will give software devs a new tool to play with much how GPUs are now used for work rather than play. Suppose someone writes an FPGA core with 1024 bit or higher calcs. Even if the core only runs at 400Mhz, it could do arbitrary far faster than a 4Ghz CPU would allow. And only programming instructions you need like add and multiply, you reduce the transistor count so you could duplicate the bignum core until it fills the FPGA cache.
And not just fractals. Cryptography uses some pretty big numbers that would benefit from unlimited precision. SETI. Protein Bending. Prime search. Bitcoin farming. Look at 4k BluRay players using tiny ASICs to decode video. It would require an immensely powerful PC to play back said content at 60fps.
November 23, 2009, 01:13:21 AM
Dilber MC Theme by HarzeM