 |
|
|
What is ARM? An informal history lesson
In the late 1980s a company called Acorn Computers, based in Cambridge, England, developed a RISC CPU and various support chips. Acorn had been using the 8-bit 6502 for its range of personal computers. It wanted to moved to 32 bits but without sacrificing the efficiency of their existing design. This first CPU shipped as a result was called the ARM2 (which stood for Acorn Risc Machine 2). The ARM2 found its way into Acorn's range of computers and offered excellent performance at a low cost. Acorn later produced the ARM3, which had a cache attached to speed up operation. Apple Computer, a US-based computer company, wanted a CPU for their battery-powered Newton. They knew it had to be small and power-efficient and yet very fast. The ARM caught their eye and as a result Advanced RISC Machines Ltd was formed in 1990 as a joint venture between Acorn Computer and Apple Computer. The design team moved from Acorn and a new CEO, Robin Saxby, was hired in February 1991. The company set about developing the 'ARM6' core and the ARM610, which had a cache, write buffer and a full MMU. The business model for the company was one of licensing the technology, rather than making chips itself and this proved very successful from the start. ARM also invested in software tools, producing a compiler, assembler and linker for the ARM. These were very reliable but initially command-line only. With the release of the SDT 2.0 in 1995, ARM produced a GUI Project Manager and Debugger. ARM has since invested heavily and the software tools are now some of the bestin the embedded world. After the ARM6 came the ARM7, which supported better operation at 3V and added embedded debug capability. The ARM7TDMI ('Thumb') was a development of the ARM7, which included an additional 16-bit instruction set to improve the already-excellent code density. The ARM7TDMI was introduced in 1995 and is still ARM's best selling product. It was far superior to anything else on the market in terms of cost, performance, power consumption and code density. It is only now that competitors have caught up with that product. After ARM7, which was designed around von Neuman architecture, ARM experimented with Havard architecture, which has separate data and address buses. In ARM's case these buses went to separate caches, which were connected to unified memory. Thus the extra cost was fairly small. The ARM8 was a step along the way, but the ARM9 followed soon afterwards and this has become the true successor to ARM7. ARM had grown rapidly from its initial 12 people and by the end of 1996 around 160 people were employed. ARM organised itself into six separate business units which saw the light of day early in 1997. The company continued to expand in many directions and today there are over 400 employees. However, by semiconductor standards ARM is still very small. It is dwarfed by most of its major customers. ARM opened offices in Tokyo and Silicon Valley in 1994, and has since expanded to many other locations in Asia, the USA and Europe. A design centre was opened in Austin, Texas in 1997 and work is well advanced there on the next generation: ARM10.
So what is so different about ARM technology? The main rationale between ARM technology is efficiency. In an ARM core, every transistor is there for a reason and offers a benefit over and above its cost. Whereas other designers might choose a slightly faster and much larger implementation method, ARM designers always look for the best compromise between the various requirements. The original ARM2 was a 'low-power' CPU almost by accident. It was much faster than the Intel '286 and '386 chips of its day, and technically superior to embedded CPUs available at the time (e.g. the Z80). Yet it had very few transistors and a very simple internal logic. The ARM instruction set has about 11 different instruction types, all of which are 32-bits in length (16 for Thumb).
|
Send mail to webmaster@bluewaternz.com
with questions or comments about this web site. |
