Dr. Philip Endecott
Address: |
1
Petworth Street, |
Email: |
phil05@chezphil.org |
Home Phone: |
01223 529198 |
||
Mobile Phone: |
07906 263138 |
||
Date of Birth: |
17th May 1970 |
Summary of Skills and Interests
My skills cover a wide range of subjects from VHDL to databases, allowing me to contribute to interdisciplinary projects involving hardware / software co-design and embedded systems.
Digital Design
|
Microprocessors
|
Design Tools
|
Software
|
Employment and Education - Summary
March
2004 |
Self-employed / consultant (see below) |
August
2003 |
Senior Electronics Design Engineer - Video compression FPGA for a hardware VNC encoder. |
August
1999 |
Member
of Technical Staff - Design of a low-power embedded RISC processor and its toolset. |
October
1998 |
Principal
Software Engineer |
October
1995 |
Research
Associate - Simulation tools for a hardware description language. |
October
1993 |
Ph.D.
Student |
October
1992 |
M.Sc.
Student |
September
1991 |
Support
Engineer |
September
1988 |
B.Sc.
Student |
Employment and Education - Details
I am currently self employed and divide my time between consultancy work, primarily offering advice about digital FPGA and ASIC design to companies in the Cambridge area that I know personally, and an online venture called TREEFIC. This is a web site for family history enthusiasts that uses algorithms similar to those used in PCB and ASIC layout to draw family tree diagrams; it started as a personal project – I was curious to learn more about database and web technology - and now generates a small income.
My last permanent employment was with ADDER TECHNOLOGY where I designed and implemented the video compression accelerator for their AdderLink IP product. This unit allows remote administration of a computer by connecting to its video, keyboard and mouse ports on one side and Ethernet on the other. A low-cost Xilinx FPGA is used to pre-process the high-speed video data, reducing the load on the main processor and so offering better performance than competing solutions. The main design challenges were the high speeds required for video (up to 165 million pixels/second), interfacing to SDRAM, and minimising cost. The FPGA worked first time.
During my two and a half years at AT&T Laboratories Cambridge I was chief architect and hardware designer for the JCN CPU project. JCN was a conventional 32-bit RISC microprocessor, prototyped in an FPGA but also suitable for ASIC implementation. The aim was to give us more flexibility at less cost than we would get buying in a processor IP block for use in lab projects. We also used it as the basis for experiments in low-power design, and through a number of small changes to the instruction set, hardware and tools reduced the activity on the external buses by more than 40%. Several patents are pending on these power-saving inventions.
The JCN project involved two engineers at AT&T and a lecturer on sabbatical from Cambridge University Computer Laboratory. A substantial part of the project was the development of the simulator and in-circuit debugging system. My responsibilities included JTAG in-circuit debugging, the graphical user interface, and configuration of the tool based on the instruction set specification.
The processor core was nearly all written in high-level VHDL and was remarkably small in terms of lines-of-code. The exception was the control circuit providing clock gating and halt/wake-up; I used an asynchronous state-machine synthesis tool to implement this, allowing the processor’s clock to be removed in halt mode.
AT&T Laboratories Cambridge closed in April 2002 due to AT&T budget cuts. I spent my redundancy payment on a year-long round-the-world trip.
My previous job was for a start-up company called Systolix who had invented a DSP processor consisting of an array of multiply-accumulate units with FPGA-style interconnect. My role, as the fifth employee, was to implement the required software tools, including FPGA-like place and route, HDL-based simulation and filter synthesis.
Prior to this I spent six years with the AMULET group at the University of Manchester, led by professor Steve Furber. The AMULET group is best known for its asynchronous implementations of ARM processors. My postgraduate work looked at how - or rather if - less conventional architectures would be more suited to asynchronous low-power implementation than the ARM. This involved analysis of the power efficiency and code density issues in conventional instruction sets and building a model of a superscalar asynchronous processor.
I then spent three years with the same group as a postdoctoral research associate, developing a hardware description language for high-level modelling of asynchronous systems. The lack of tools had been cited as a reason for the slow adoption of asynchronous techniques in industry. My language, LARD, adds occam-style channel communication to the usual hardware description mechanisms. Use of this language led to a ten-fold improvement in productivity during the initial design phase.
Before starting my Ph.D. I spent a year working for ARM, who at that time were a small bunch of hard-working engineers in a barn out in the Cambridgeshire countryside. This gave me a useful insight into start-up companies, chip design, and the chip business. My work included test engineering and code reviewing for a video controller chip, and contributions to what became the AMBA on-chip bus.
Other Skills and Interests
I speak French and Spanish more than adequately for my holidays, having taken GCSE Spanish evening classes some years ago and AS French this year.
I am the author of a small number of open-source software projects, including Decimail, a mail server based on a PostgreSQL database, and Anyterm, a terminal emulator on a web page.
I enjoy escaping to the hills and have been known to take the Fort William sleeper after work on a Friday when the forecast is for good weather - sun or snow - in the highlands. I am also an occasional caver.
This CV is online at http://chezphil.org/cv/