The ARM architecture (originally the Acorn RISC Machine) is a 32-bit RISC processor architecture that is widely used in a number of embedded designs. Due to their power saving features, ARM CPUs are dominant in the mobile electronics market, where low-power is a critical design goal.
Today, the ARM family accounts for over 75% of all 32-bit embedded CPUs, making it one of the most prolific 32-bit architectures in the world. ARM CPUs are found in all corners of consumer electronics, from portable devices (PDAs, mobile phones, media players, handheld gaming units, and calculators) to computer peripherals (hard drives, desktop routers.) The most noticeable branch in this family nowadays is Intel’s XScale.
The ARM design was started in 1983 as a development project at Acorn Computers Ltd.
The team, led by Roger Wilson and Steve Furber, started development of what in some ways resembles an advanced MOS Technology 6502. Acorn had a long line of computers based on the 6502, so a chip that was similar to program could represent a significant advantage for the company.
The team completed development samples called ARM1 by 1985, and the first “real” production systems as ARM2 the following year. The ARM2 featured a 32-bit data bus, a 26-bit address space giving a 64 Mbyte address range and 16 32-bit registers. One of these registers served as the (word aligned) program counter with its top 6 bits and lowest 2 bits holding the processor status flags. The ARM2 was possibly the simplest useful 32-bit microprocessor in the world, with only 30,000 transistors (compare with Motorola’s four-year older 68000 with around 68,000). Much of this simplicity comes from not having microcode (which represents about 1/4 to 1/3rd of the 68000) and, like most CPUs of the day, not including any cache. This simplicity led to its low power usage, while performing better than the 286. A successor, ARM3, was produced with a 4KB cache which further improved performance.
In the late 1980s Apple Computer started working with Acorn on newer versions of the ARM core. The work was so important that Acorn spun off the design team in 1990 into a new company called Advanced RISC Machines. For this reason you often see ARM lengthened to Advanced RISC Machine instead of Acorn RISC Machine. Advanced RISC Machines became ARM Limited when the company floated on the London Stock Exchange and NASDAQ in 1998.
This work would eventually turn into the ARM6. The first models were released in 1991, and Apple used the ARM6-based ARM 610 as the basis for their Apple Newton PDA. In 1994, Acorn used the ARM 610 as the main CPU in their RiscPC computers.
The core has remained largely the same size throughout these changes. ARM2 had 30,000 transistors, while the ARM6 grew to only 35,000. The idea is that the end-user combines the ARM core with a number of optional parts to produce a complete CPU, one that can be built on old semiconductor fabs and still deliver lots of performance at a low cost.
The most successful implementation has been the ARM7TDMI with hundreds of millions sold in mobile phones, handheld video game systems, and Sega Dreamcasts. While ARM’s business has always been to sell IP cores, some of the licensees generated microcontrollers based on this core.
The Dreamcast features a SH4 processor which only borrows concepts from ARM (low power consumption, optional compact instruction set etc.), but is otherwise different from an ARM. The Dreamcast also features a sound chip designed by Yamaha with an ARM7 core. Nintendo’s Gameboy Advance, however, uses the ARM7TDMI at 16.78MHz.
DEC licensed the architecture (which caused some confusion because they also produced the DEC Alpha) and produced the StrongARM. At 233 MHz this CPU drew only 1 watt of power (more recent versions draw far less). This work was later passed to Intel as a part of a lawsuit settlement, and Intel took the opportunity to supplement their aging i960 line with the StrongARM. Intel have since developed its own high performance implementation known by the name XScale.
Perhaps in part because of the conditional execution facility using up four bits of every instruction, newer ARM processors have a 16-bit instruction mode, called Thumb. The smaller opcodes have less functionality; for example, only branches can be conditional, and many opcodes cannot access all of the CPU’s registers. However, the shorter opcodes give improved code density overall, even though some operations will require more opcodes to be executed. Particularly in situations where the memory port or bus width is constrained to less than 32 bits, the shorter Thumb opcodes allows greater performance than with 32-bit code because of the more efficient use of the limited memory bandwidth. Typically in embedded applications a small range of addresses have a 32-bit datapath and the rest are 16 bits wide or narrower (e.g. the Game Boy Advance); in this situation, it usually makes sense to compile Thumb code and hand-optimise a few of the most CPU-intensive sections using the 32-bit instruction set, placing them in the limited 32-bit bus width memory.
The first processor with Thumb technology was the ARM7TDMI. All ARM9 and later families, including XScale have included Thumb technology.
ARM has implemented a technology that allows certain of their architectures to execute Java bytecode natively in hardware, in another execution mode alongside the existing ARM and Thumb modes and accessed in a similar fashion to ARM/Thumb interworking.
The first processor with Jazelle technology was the ARM926EJ-S: Jazelle being denoted by the ‘J’ in the CPU name. It has been used by mobile phone manufacturers to speed up execution of Java ME games and applications, which is probably what drove development of the technology.
Thumb-2 technology made its debut in the ARM1156 core, announced in 2003. Thumb-2 extends the limited 16-bit instruction set of Thumb with additional 32-bit instructions to give the instruction set more breadth. As a result the stated aim for Thumb-2 is to achieve code density that is similar to Thumb with performance similar to the ARM instruction set on 32-bit memory.
Thumb-2 also extends both the ARM and Thumb instruction set with yet more instructions, including bit-field manipulation, table branches, and conditional execution.
Thumb-2EE, marketed as Jazelle RCT, was announced in 2005, first appearing in the Cortex-A8 processor. Thumb-2EE provides a small extension to Thumb-2, making the instruction set particularly suited to code generated at runtime (e.g. by JIT compilation) in managed Execution Environments. Thumb-2EE is a target for languages such as Limbo, Java, C#, Perl and Python, and allows JIT compilers to output smaller compiled code without impacting performance.
New features provided by Thumb-2EE include automatic null pointer checks on every load and store instruction, an instruction to perform an array bounds check, and the ability to branch to handlers, which are small sections of frequently called code, commonly used to implement a feature of a high level language, such as allocating memory for a new object.
NEON technology is a combined 64 and 128bit SIMD (Single Instruction Multiple Data) instruction set that provides standardized acceleration for media and signal processing applications. NEON can execute MP3 audio decoder on CPU running at 10 MHz and can run the GSM AMR (Adaptive Multi-Rate) speech codec using CPU running at no more than 13 MHz. It features a comprehensive instruction set, separate register files and independent execution hardware. NEON supports 8-, 16-, 32- and 64-bit integer and single precision floating-point data and operates in SIMD operations for handling audio/video processing as well as graphics and gaming processing. SIMD is a crucial element in vector supercomputers which feature simultaneous multiple operations. In NEON, the SIMD supports up to 16 operations at the same time.
VFP technology is a coprocessor extension to the ARM architecture. It provides low-cost single-precision and double-precision floating-point computation that is fully compliant with the ANSI/IEEE Std 754-1985 Standard for Binary Floating-Point Arithmetic. VFP provides floating-point computation suitable for a wide spectrum of applications such as PDA, smartphones, voice compression and decompression, three-dimensional graphics and digital audio, printers, set-top boxes, and automotive applications. The VFP architecture also supports execution of short vector instructions allowing SIMD (Single Instruction Multiple Data) parallelism. This is useful in graphics and signal-processing applications by reducing code size and increasing throughput.
ARM Ltd does not manufacture and sell CPU devices based on their own designs, but rather, licenses the processor architecture to interested parties. ARM offers a variety of licensing terms, broken down by cost and deliverables. To all licensees, ARM provides an integratable hardware-description of the ARM core, as well as complete set of software development toolset (compiler, debugger, SDK), and the right to sell manufactured-silicon containing the ARM CPU. Fabless licensees, who wish to integrate an ARM core into their own chip design, are usually only interested in acquiring a ready-to-manufacture, pre-verified IP-core. For these customers, ARM delivers a gate-netlist description of the chosen ARM core, along with an abstracted simulation-model and test programs to aid design integration and verification. More ambitiuous customers, including integrated device manufacturers (IDM) and foundry operators, chose to acquire the processor IP in synthesizable RTL (Verilog) form. With the synthesizable RTL, the customer has the ability to perform architectural level optimizations and extensions. These allow the designer to achieve exotic design goals not otherwise possible with an unmodified netlist (high clock speed, very low power-consumption, instruction-set extensions, etc.) While ARM does not grant the licensee the right to re-sell the ARM-architecture itself, licensees may freely sell manufactured product (chip devices, evaluation boards, complete systems, etc.) Merchant foundries can be a special case; not only are they allowed to sell finished silicon containing ARM-cores, they generally hold the right to re-manufacture ARM-cores for other customers.
Like most IP vendors, ARM prices its IP based on perceived value. In architectural terms, the lower-performance ARM cores command a lower license cost than the higher-performance cores. In terms of silicon implementation, a synthesizable core is more expensive than a hard-macro (black-box) core. Complicating price matters, merchant foundries who hold an ARM license (such as Samsung and Fujitsu) can offer reduced licensing costs to its fab customers. In exchange for acquiring the ARM core through the foundry’s in-house design-services, the customer can reduce or eliminate payment of ARM’s upfront license fee. Compared to dedicated semicounductor foundries (such as TSMC and UMC) without in-house design-services, Fujitsu/Samsung charge 2-3x more per manufactured wafer. For low-mid volume applications, a design-service foundry offers lower overall pricing (through subsidization of the license-fee.) For high volume mass-produced parts, the long-term cost-reduction achievable through lower wafer-pricing reduces the impact of ARM’s NRE cost, making the dedicated foundry a better choice.
Many hightech semiconductor firms hold ARM licenses: Broadcom, Cirrus Logic, Freescale (spun off from Motorola in 2004), Fujitsu, Intel (through its settlement with DEC), IBM, Infineon Technologies, Texas Instruments, Nintendo, Philips, VLSI, Atmel, Sharp, Samsung, and STMicroelectronics are some of the many companies who have licensed the ARM in one form or another. Although ARM’s license terms are covered by NDA, within the IP industry, ARM is widely known to be among the most expensive CPU cores. A single customer product containing a basic ARM-core can incur a one-time license fee in excess of (USD) $200,000. Where significant quantity and architectural modification are involved, the license fee can exceed $10M.