The AMULET group was established in late 1990 and is part of the Computer Science Department at the University of Manchester. The group started with a primary interest in the application of asynchronous computing techniques, with special regard to a reduction in electrical power consumption. This has now expanded across several related projects, working on a number of asynchronous logic and low power problems.
In April 1994 the group took delivery of the AMULET1 microprocessor, the world's first implementation of a commercial microprocessor architecture (ARM) in asynchronous logic.
Work on the asynchronous ARM architecture did not stop there. The development team continued with AMULET2 within the OMI-DE project. This is a major rework of AMULET1 in the light of past experience and the instruction throughput has been increased whilst the electrical power consumption was reduced. AMULET2e (an embedded system chip with an AMULET2 core and a self-timed cache) was received in October 1996. This appears (we are not sure that we are comparing quite the same numbers) to give AMULET2 more MIPs/W than a contemporary synchronous ARM on the same geometry - a very difficult target!
The power/performance qualities of AMULET are sufficiently encouraging to continue the line of investigation. As the performance of AMULET2 still lags that of contemporary synchronous devices the next task is to improve the MIPS rate, without sacrificing excess power. This problem is being addressed by AMULET3, which is now under development.
The AMULET microprocessors are not the only project AMULET projects however. Other members of the group have implemented an AMULET microprocessor in a differential CML process. This promised to be significantly faster than a CMOS implementation, although it is certainly not lower power. The size of the design challenged the implementation technology and eventually only a subset part was produced; the result was interesting, but not encouraging in view of the rapid development of CMOS technology.
Another AMULET sub-group was involved in the ExACT project, an ESPRIT consortium led by Philips Research Labs. in Eindhoven. This project drew together many asynchronous logic design strategies and produced commercially viable technology demonstrators. Much of this project was concerned with the low-power aspects and synthesisability of asynchronous circuits to allow the easy and quick production of devices for domestic applications.
A further AMULET interest - again in asynchronous design - is in the OMI-HORN project. This has the intention of producing an asynchronous implementation of a different microprocessor architecture.
Current AMULET projects include the production of asynchronous-logic development and analysis tools. The first of these is called LARD, for reasons which will be obscure to anyone outside the group. Another recently started project is PowerPack, where we are a member of a consortium working on low-power logic design; we'll probably do some asynchronous things out of habit!
There are also a number of individual research students with their own projects which will be outlined (eventually!) in their own pages.
In the meantime the some earlier members of AMULET, including the designers of the AMULET1 & AMULET2 chips, have left us. In 1996 they formed their own start-up company - Cogency Technology - which is doing asynchronous things for money!
The AMULET group is also forming links with several formal methods groups, including the group here at Manchester University.