microcode, compilers, and supercomputer architecture

From: Eric Smith <eric_at_brouhaha.com>
Date: Mon Apr 5 18:17:27 1999

Tony wrote:
> In general, I'm going to be writing microcode either for a machine that I
> understand to gate level, or even more likely, something that I've
> designed myself. In which case (especially the latter case), I am going
> to be thinking things like 'And that bit enables that AND gate,
> transfering this signal to the enable to that register. If I do that one
> cycle early, I'll avoid a possible race condition in the hardware at that
> point...'. That sort of thing is difficult for me to put into a compiler,
> but relatively easy to remember when you're coding by hand.

I don't know why you think it's difficult to put that knowledge into a
compiler; that is *exactly* the sort of thing that I *did* put in my compilers,
and it worked great. I actually had it log messages when constraints like
that were applied, and found many that I might have overlooked had I not been
using the compiler.
Received on Mon Apr 05 1999 - 18:17:27 BST

This archive was generated by hypermail 2.3.0 : Fri Oct 10 2014 - 23:31:39 BST