RISC Processor Background

RISC, or Reduced Instruction Set Computer, is a term that is conventionally used to describe a type of microprocessor architecture that employs a small but highly-optimized set of instructions, rather than the large set of more specialized instructions often found in other types of architectures. This other type of processor is traditionally referred to as CISC, or Complex Instruction Set Computer.

History

The early RISC processors came from research projects at Stanford and Berkeley universities in the late 1970s and early 1980s. These processors were designed with a similar philosophy, which has become known as RISC. The basic design architecture of all RISC processors has generally followed the characteristics that came from these early research projects and which can be summarized as follows:

• One instruction per clock cycle execution time: RISC processors have a CPI (clock per instruction) of one cycle. This is due to the optimization of each instruction on the CPU and a technique called pipelining. This technique allows each instruction to be processed in a set number of stages. This in turn allows for the simultaneous execution of a number of different instructions, each instruction being at a different stage in the pipeline.

• Load/Store machine with a large number of internal registers: The RISC design philosophy typically uses a large number (commonly 32) of registers. Most instructions operate on these registers, with access to memory made using a very limited set of Load and Store instructions. This limits the need for continuous access to slow memory for loading and storing data.

• Separate Data Memory and Instruction Memory access paths: Different stages of the pipeline perform simultaneous accesses to memory. This Harvard style of architecture can either be used with two completely different memory spaces, a single dual-port memory space or, more commonly, a single memory space with separate data and instruction caches for the two pipeline stages.

Over the last 20-25 years, RISC processors have been steadily improved and optimized. In one sense, the original simplicity of the RISC architecture has been lost – replaced by super-scalar, multiple-pipelined hardware, often running in the gigahertz range.

"Soft" FPGA Processors

With the advent of low-cost, high-capacity programmable logic devices, there has been something of a resurgence in the use of processors with simple RISC architectures. Register-rich FPGAs, with their synchronous design requirements, have found the ideal match when paired with these simple pipelined processors.

As a result, most 32-bit FPGA soft processors are adopting this approach. They could even be considered as "Retro-processors".

Why use "Soft" Processors?

There are a number of benefits to be gained from using soft processors on reconfigurable hardware. The following sections explore some of the more significant of these benefits in more detail.

Field Reconfigurable Hardware

For certain specific applications, the ability to change the design once it is in the field can be a significant competitive advantage. Applications in general can benefit from this ability also. It allows commitment to shipping early in the development cycle. It also allows field testing to be used to help drive the latter part of the design cycle without requiring new "board-spins" based on the outcome. This is very similar to the way in which alpha, beta, pre-release and release cycles currently drive the closure of software products.

The ability to update embedded software in a device in the field has long been an advantage enjoyed by designers of embedded systems. With FPGAs, this has now become a reality for the hardware side of the design. For end-users, this translates as "Field Upgradeable Hardware".

Faster Time to Market

FPGAs offer the fastest time to market due to their programmable nature. Design problems, or feature changes, can be made quickly and simply by changing the FPGA design - with no changes in the board-level design.

Improving and Extending Product Life-Cycles

Fast time to market is usually synonymous with a weaker feature set – a traditional trade-off. With FPGA-based system designs you can have the best of both worlds. You can get your product to market quickly with a limited feature set, then follow-up with more extensive features over time, upgrading the product while it is already in the field.

This not only extends product life-cycles but also lowers the risk of entry, allowing new protocols to be added dynamically and hardware bugs to be fixed without product RMA.

Creating Application-Specific Coprocessors

Algorithms can easily be moved between hardware and software implementations. This allows the design to be initially implemented in software, later off-loading intensive tasks into dedicated hardware, in order to meet performance objectives. Again, this can happen even after commitment to the board-level design.

Implementing Multiple Processors within a Single Device

Extra processors can be added within a single FPGA device, simply by modifying the design with which the device is programmed. Once again, this can be achieved after the board-level design has been finalized and a commitment to production made.

Lowering System Cost

Processors, peripherals, memory and I/O interfaces can be integrated into a single FPGA device, greatly reducing system complexity and cost. Once the FPGA-based embedded application moves to 32-bit, cost becomes an even more powerful driver.

As large FPGAs become cheaper, both Hybrids and soft cores move into the same general cost area as dedicated processors. At the heart of this argument is also the idea that once you have paid for the FPGA, any extra IP that you place in the device is free functionality.

Avoiding Processor Obsolescence

As products mature, processor supply may become an increasing problem, particularly where the processor is one of many variants supplied by the semiconductor vendor. Switching to a new processor usually requires design software changes or logical hardware changes.

With FPGA implementations, the design can be easily moved to a different device with little or no change to the hardware logic and probably no change to the application software. Peripherals are created dynamically in the hardware, so lack of availability of specific processor variants is never a problem.