Date:  4 May 1981 1313-EDT (Monday)
From: David.Smith at CMU-10A (C410DS30)
To: space at mit-mc
Subject: Apollo guidance computer

In Bell & Newell's book "Computer Structures:  Readings and Examples" there
is an article about the Apollo Guidance Computer (AGC), reprinted from
IEEE Trans. EC-12, Dec. 1963.  It doesn't unambiguously say that it is to
be used in the moon flights, but does say, "The AGC is an onboard computer
for one of the forthcoming manned space projects."  It is rather refreshing
in that it doesn't claim to be the best thing since sliced bread, and
admits that some design decisions (such as a 15-bit word) were brought
along from previous (incompatible) designs.

Some of the specifics are as follows.  Ram was core;  rom was rope memory.
Both cycled in 12 microseconds.  Word length was 15 bits, except for
accumulator, which had 16 bits.  (This was for overflow indication.)
Instruction format was 3 bits for opcode, 12 bits for address.  When
designing the predecessor computer, the MOD-3C, they had decided that
4000 words would be plenty, but now they realized that they really
needed 10,000 words.  So instead of making the computer into a 17-bit
machine, they provided bank switching for addresses 6000-7777 (octal).
"The possibility of using two bank registers is worthy of consideration,
but it did not occur to us."

Rather than an index register, they provided an index instruction which
added its operand to the following instruction (for purposes of execution
only).  This addition could affect the following opcode as well.
They even wired up the machine so that overflow from the index operation
was treated as an opcode bit.  This provided room for the instructions
subtract, multiply, and divide.

The Pdp-11 stole its memory-mapped I/O from the AGC.  There were also
a few active memory locations for shifting right and left.

If I recall correctly, the PR about the shuttle's computers rated them
as around 40 times the speed of the AGC, or 400,000 instructions/sec.
So AGC must have been rather slow.  On top of that, "Most of the
programs relevant to navigation were written in a parenthesis-free
pseudocode notation for economy of storage..."  No wonder it was
flashing OVERLOAD as Armstrong and Aldrin were looking for a place to
set down!

		David Smith

Date:  5 May 1981 1225-EDT (Tuesday)
From: David.Smith at CMU-10A (C410DS30)
To: space at mit-mc
Subject:  Gemini onboard computer

If I recall flight and astronaut correctly, it was Stafford on Gemini 6
who used a 15 inch diameter circular slide rule to do orbital rendezvous
calculations.  Ground computers were primary, but he got the same answers
in good enough time.

[I guessed Stafford, because it stuck in my mind that it was the first
rendezvous, but it also sticks in my mind that it was Aldrin, who flew
later.]