VLIW: Very Long Instruction Word

History

Alan Turing

VLIW processing ideas have roots in Alan Turing's 1946 parallel computing studies and Maurice Wilkes's 1951 microprogramming work. Microprogrammed CPUs have a macroinstruction that corresponds to each program instruction. Each of these macroinstructions has a corresponding sequence of microinstructions, kept in ROM on the CPU. These microinstructions can be ordered into wide sets of control signals. This is called horizontal microprogramming.

When Joseph Fisher was working on writing horizontal microcode for a CDC-6600 emulator in 1979, he began to work on the problem of generating long instruction words from short sequential instructions. The techniques he developed, called "trace scheduling" (which will be discussed in detail later), were essential for generating VLIW-compatible code.

In 1984, Fisher and colleagues from Yale University formed Multiflow, with the goal of creating VLIW supercomputers. Cydrome also was formed that year by Bob Rau. Although they both released finished products, the marketplace was not ready, nor was the technology complete. As such, neither company managed much commercial success, and both have since folded. But they have continued to refine VLIW technology, and both founders are now bringing their expertise to Hewlett-Packard Labs, which has produced the Playdoh research machine [PDF].

Once these pioneering companies disappeared, little development appeared to happen on the VLIW front, until engineers found the architecture was ideal for quick computation of complex, repetitive algorithms. Thus the re-emergence of VLIW came with a new wave of media processing chips, led by Chromatic's MPact and Philips's TriMedia, which led to the development of the first really publicly successful VLIW implementation: the Texas Instruments C6X series, which is discussed in greater detail in the implementation section of this site.

The huge success of TI's DSP chips paved the way for a new generation of large-scale VLIW microprocessor designs. In 2000, startup Transmeta released their family of VLIW-based Crusoe processors, designed especially for low-power processing in embedded devices. Intel followed with their 64-bit architecture specification, IA-64, and has continued with the Itanium line of chips.

VLIW's emergence is a matter of the right technology available at the right time. In the 1970s, VLIW was nearly un-implementable due to the prohibitively expensive nature of memory at the time. CISC processors could be made to use as little as eight bits for a simple instruction, making it the technology of choice. These days, however, memory is inexpensive, and VLIW is again a viable solution.

« Previous Next »