Monday, August 04, 2008

My elevator pitch for BitGrid

I had an idea back when I was in college (1981-82) about using an array of look up tables to do programmable logic. I've never really had a chance to get a chip made, as my work is nowhere near that field these days. I'm wondering if you think my idea has any merit.

I've got a blog up at where I write about this subject, trying to get a chip made some day.

The idea is simple, really... a grid of cells with 4 inputs, a look up table, and 4 outputs. The 64 bits determine the outputs for any possible input combination.

Routing logic is even simpler... there is none. If you want to route through a cell, you have to program the cell to do it.

Thus any cell can be routing or computation, or both.

An unsigned n bit adder takes n cells

An unsigned n bit multiply takes n*(n-1) cells

A divider takes (n+1)*n cells, unless you want to divide by zero...then it's (n+1)^2 cells

Sound interesting? Waste of time?

I'd like to know what you think.

No comments: