Four bit adder: Difference between revisions

 

A quick note on the use of random stimulus. You might think that, with an input space of only 2**8 (256) distinct inputs, that exhaustive testing (i.e. just loop through all the possible inputs) would be appropriate. In this case you might be right. But as a HW verification engineer I'm used to dealing with coverage spaces closer to 10**80 (every state element -- bit of memory) increases the space). It's not practical to verify such hardware exhaustively -- indeed, it's hard to know where the interesting cases are -- so we use constrained random verification. If you want to, to can work thought the statistics to figure out the probability that we missed a bug when sampling 20 cases from a space of 2**8 -- it's quite scary when you realize that every complex digital chip that you ever bought (cpu, gpu, networking, etc.) was 0% verified (zero to at least 50 decimal places).

 

=={{header|Tcl}}==