How did the Apollo Guidance Computer actually work?

The Apollo Guidance Computer (AGC) was a real-time digital control computer with many features of a general-purpose machine. It did two kinds of work: it solved the guidance and navigation equations for the lunar mission, and it acted as a control computer that aligned the inertial measurement unit, processed radar data, managed the astronaut displays and controls, and generated commands for the spacecraft engines. Astronauts talked to it through a keyboard and display unit called the DSKY. [1] [2]

MIT's final report describes a parallel, whole-number binary machine with three principal sections: a memory, a central processor, and a control section. The memory was split into a fixed, read-only portion of 36,864 words and an erasable portion of 2,048 words. The central processor held an adder, an instruction decoder, interrupt control, a memory address decoder and several addressable registers, linked by a set of 16 write buses. The control section supplied timing and sequencing, including the microprograms that carry out each instruction. [1]

Those figures are for the final design. The earlier Block 1 version had 1,024 words of erasable ferrite-core memory, 15 data bits per word plus a parity bit, one's-complement arithmetic, and a memory cycle time of 11.7 microseconds. [3]

Each spacecraft carried its own computer: one in the command module and one in the lunar module. They were electrically identical but ran different software and had different interfaces. The command module had two DSKYs and the lunar module had one. [1] [2]

A simple real-time operating system designed by J. Halcombe Laning managed the work. It had two parts. The Exec was a job scheduler using cooperative multitasking. The Waitlist was an interrupt-driven pre-emptive scheduler for timer-driven tasks. Tasks were short threads of execution that could reschedule themselves, or start a longer job under the Exec. [4]

On the trip between Earth and Moon, the command module crew used a sextant to measure the angle between a known star and an Earth or Moon feature. After calibrating the instrument, the navigator lined up the images and pressed the MARK button. The computer recorded the angle and the time, then used a Kalman recursive filter to work out how the measurement would change the spacecraft's position and velocity. The navigator could accept or reject that update. The Atlantic describes the computer's two main jobs as computing the course to the Moon from such measurements and controlling the spacecraft's physical components, with connections to about 150 devices. [2] [5]

For the rendezvous after the lunar module left the Moon's surface, the lunar module's computer combined rendezvous radar ranges, radar antenna angles and inertial measurements, again using a Kalman filter, to produce targeting for the maneuvers. The flight plan was designed so astronauts or ground tracking could solve it with simple methods if the primary system failed. [2]