2 Jan
2007
2 Jan
'07
9:13 p.m.
On 2 Jan 2007 at 20:24, Brent Hilpert wrote:
Now that was a fun read. I'm still trying to
understand his description of
the core rope ROM though. I'm familiar with core-rope ROM (or at least one
version of it) from attempting to make a reader to dump the contents of a Wang
calc microcode ROM, but the AGC version sounds like the
address-decoding/word-selection is done differently (..need a diagram).
Somewhere, on one of the AGC sites linked to by the guy who's
building his own, is a diagram of both the ROM and RAM setup.
In the paper, the author says: "I don?t know of any other computer
that uses 1?s complements ." He obviously never met Seymour Cray.
One's complement arithmetic isn't hard to deal with; and it helped
that Cray designed his ALU as a subtracter rather than an adder, so
that the only way (using addition and subtraction) to get a negative
zero was to subtract negative zero from negative zero. In practice,
this came up very rarely. The zero and nonzero conditional branches
worked on both species of zero at any rate. There are a few very
clever bit-twiddling tricks possible on a 1's complement machine that
don't carry over to 2's complement.
Cheers,
Chuck