Muse and Atlas

By the mid 1950's Britain was behind the United States in the production of high performance computers as confirmed in a report dated May 1956 to the DSIR Advisory Committee on High Speed Calculating Machines which remarked "We ought to be a little worried that no serious effort is being made to produce one really large fast machine" By autumn 1956 Tom Kilburn had initiated such an effort known as the MUSE (microsecond) computer.

While in its early stages Kilburn talked to several potential users of large computers, the design specifications from these meetings included the desire for an instruction speed approaching one order per microsecond and the need to attach a large number of peripherals of various types. Also required was an immediate-access storage capacity far in excess of anything then available.

It was clear from the outset that special techniques would be needed if efficient use of equipment and rapid turnaround were to be achieved in a computer system consisting of many interconnected units of widely differing speeds. This applied not only to slow and fast peripherals and backing stores, but also to high speed transistor logic circuits interfacing with medium-speed moderate-size main stores.

Special techniques that were eventually employed included what are now known as

None of these techniques had been invented when project started in 1956.

By 1959 the computer had been renamed the Atlas and was subsequently developed as a joint University/Ferranti venture under Tom Kilburn.

Performance

Atlas was inaugurated on the 7th December '62. It was considered to be the most powerful computer in the world. It was 80 times more powerful than Meg/Mercury and 2400 times more powerful than the Mark 1.

The Atlas had a 48-bit word, single address format, allowing for double B-modification and a user virtual address of 1 million words.

There were 127 half word B-registers, mostly held in a fast 0.7 microsecond cycle time core store. System routines and some frequently-used user library routines were held in an 8k read only store having an access time of 0.3 milliseconds.

The core+drum one level store used fixed-sized 512 word page, 32 page address registers forming an associative store for virtual to real address translation. A `drum learning program' held in read-only memory attempted to optimise page swapping by accumulating statistics on page utilisation. The learning program was the first use of AI in a conventional computer.

The Storage Hierarchy :

Main Store : 16k, 2 microsecond cycle time 4-way interleaved. Backed by 4*drums, each 24k, 12 millisecond revolution time, capable of transferring 1 block of 512 words every 2 milliseconds.
System Working Store: 1k* 2 microsecond cycle time.
System read-only: 8k, 0.3 microsecond access time.
Magnetic tapes: 8 decks (on 8 channels) each 90 kc/s giving one word every 88 microseconds.
file disc (added in 1967): 16M words.

The interrupt structure allowed for up to 512 peripherals. The Atlas had 17 conventional input/output devices.

The instruction speed on the 7th Dec 62 were:

fixed-point addition 1.59 microseconds
floating-point add, no modification 1.61 microseconds
floating-point add, double modification 2.61 microseconds
floating-point multiply, double modification 4.97 microseconds
floating-point division 10.66 microseconds min to 29.8 microseconds max

The full supervisor (operating system) became available from Jan 1964, followed by languages such as Algol and Fortran. The Atlas provided a computing service to many universities and industry until September 1971.

Ferranti sold 2 full versions of the Atlas. The Main impact of the Atlas was the wealth of new ideas the project introduced.