Hi.
I am working on an open source hardware/software project, which I think might be interesting to finding Mersenne numbers, or not - I have not come to a conclusion yet, and thought I would get some feedback from the list...
We have a small (5" x 7"), cheap (sub $150) board based on the Blackfin processor (jointly developed with Analog Devices and Intel, 756MHz, multiple parallel computational blocks including a dual multiply accumulate per instruction cycle, and optimized instructions for FFTs), with 128Meg of SDRAM (133MHz), 4 Meg of Flash, 10/100 Ethernet, and power supply.
http://blackfin.uclinux.org/projects/stamp
We have alphas of gcc (3.4.x), gdb and uClinux (2.8.x kernel - including networking) up and running on this platform. Everything is mostly working, and issues are being resolved on a daily basis.
I would be interested in having the discussion on this mailing list or otherwise of what would be the best use of a platform like this (fixed point, 750 MHz core, supported with gcc and uClinux - optimized instructions/assembly for FFTs).
Am I better modifying the algorithms work for arbitrary precision rather than floating point?
Since the processor does not have an FPU (soft-floating point), if I am going to be running the standard algorithms - that Double-checking, would be the best thing (vs factoring, or Lucas-Lehmer testing?)
Or is this useful at all?
Feedback/comments/questions welcome.
-Robin
_______________________________________________ Prime mailing list [EMAIL PROTECTED] http://hogranch.com/mailman/listinfo/prime
