English / 日本語
← Back to the Fundamentals home Chapter 5: Part 2 · How you write the contract changes the machine

Chapter 5　MIPS — the RISC textbook, ancestor of RISC-V

The aim of this chapter. In Chapter 4 we saw the RISC way of thinking that keeps instructions simple and uniform. In this chapter we visit MIPS, which first put that RISC into “textbook form” and became the ancestor of today’s RISC-V. Why was MIPS studied in classrooms around the world? And why, even so, do we choose RISC-V rather than MIPS? Tracing the history, the preface’s “the contract is shared, the machine is chosen” and Chapter 4’s “simplicity pays” connect into a single story.

5.1 Two sources of RISC — Berkeley and Stanford

Around 1980, the idea that “if you keep instructions simple and uniform, the machine becomes fast and light” (RISC) sprouted at almost the same time at two universities. “RISC-I / RISC-II” by Professor David Patterson and colleagues at the University of California, Berkeley, and MIPS by Professor John Hennessy and colleagues at Stanford University. Both were pioneers that brought Chapter 4’s “simple instructions, many, fast” down into real designs.

These two later received, for this RISC achievement, the 2017 Turing Award — regarded as the highest honor in computer science, also called “the Nobel Prize of the computing field” — jointly. The co-authored textbook bearing their names is still a standard read in classrooms around the world. The RISC road we’re about to use as a matter of course was one these forerunners carefully opened.

This is an important distinction that connects to later. MIPS is from Stanford. And RISC-V, which appears later, is from the other side, Berkeley. From the same RISC aspiration, two streams were born — keep that in mind.

5.2 Why did MIPS become “the textbook”?

What made MIPS special was its clean simplicity. The kinds and shapes of instructions were well uniform, and it rides very docilely on Chapter 3’s assembly line (pipeline). In kitchen terms, the way of writing the order slips is tightly unified, and the five workers’ steps flow smoothly — such a “model” contract.

This “ease of teaching, ease of understanding” was prized, and it is said to have been adopted as the subject for learning CPUs in the standard textbook Computer Organization and Design by Professors Patterson and Hennessy. Ever since, students around the world have learned CPU design with MIPS. MIPS became, so to speak, “the RISC textbook.” The “workbench, stove, five steps” you’re learning now are also on the road that textbook paved.

5.3 But MIPS too had a “tollgate”

However, there was one wall. MIPS’s instruction set (contract) was a proprietary one whose rights are held by a particular company, and using it commercially required a license (permission). This is the “tollgate” of the reader and Chapter 4. Like x86 and ARM, to pass through the MIPS contract, you had to pay a toll at the gate.

Here, let’s also look at other instruction sets. x86, familiar from PCs, is the mainstream of the CISC seen in Chapter 4, and it too is a contract with a tollgate, handled only by limited companies such as Intel and AMD. ARM, on the other hand, is actually a member of RISC — a stream said to have been created in the 1980s by the UK’s Acorn, inspired by Berkeley’s RISC research. Now a greatly successful RISC used in many smartphones and embedded devices, but using its contract also requires a license (permission). That is, ARM too has the same tollgate as MIPS. This contrast makes the “absence of a tollgate” in the RISC-V we see next stand out all the more.

In the preface we said “a contract can be shared.” But if the contract itself has an owner, not everyone can use it freely. You can learn it in a textbook, but the moment you try to use it in your own chip, a tollgate stands in the way — this was a long-standing worry.

5.4 From MIPS to RISC-V — an opened shared language

That’s where RISC-V appeared. Around 2010, at Berkeley, the other source of RISC, it was born from the idea, “a tollgate-free instruction set that anyone can use freely.” Keeping the textbook-like docility as is, it opened the contract itself, for free and openly.

And symbolically, in 2021, the company that had handled MIPS itself is reported to have switched its own path from MIPS to RISC-V. The company behind MIPS, long taken as “the textbook,” moved to the RISC-V side of its own accord — that is what happened.

The lineage of RISC — from MIPS to RISC-V Around 1980 RISC was born at Berkeley and Stanford. MIPS was commercialized and became a textbook standard but had a license tollgate. Around 2010 the free and open RISC-V was born at Berkeley. In 2021 MIPS itself moved to RISC-V. around 1980 RISC is born Berkeley / Stanford 1985– MIPS, the standard commercial · textbook around 2010 RISC-V is born from Berkeley from research license tollgate free · open And in 2021 — MIPS itself, too, to RISC-V.
The lineage of RISC. It sprouted at two universities, and MIPS became the textbook standard. But the contract had a tollgate. Around 2010, the free and open RISC-V was born at Berkeley, and in 2021 the company behind MIPS itself, too, headed toward RISC-V — it is reported.

5.5 The designer’s view — why do we choose RISC-V

The history up to here becomes, as is, the reason for the design decision.

The designer decides here. What: which instruction set (contract) to choose for your own chip’s brain (core). How to decide: choose one that combines textbook-like docility (easy to learn, rides docilely on the assembly line) and free/open (no tollgate, anyone can implement and verify). Why: because it inherits MIPS’s virtues (simplicity, ease of teaching) while having no tollgate. So now, for us who “choose and assemble” our own chip, RISC-V becomes a docile option.

MIPS is the source of that docility, so to speak the ancestor of RISC-V. The “workbench, stove, five steps” view we’ve used throughout this Fundamentals section, traced back, is also on this textbook lineage.

Which decision is this knowledge for. The choice of instruction set (contract) is the root of the “Brain” in the preface’s “decisions map.” Knowing MIPS’s history, you can explain the reason for choosing RISC-V not as “just because it’s popular,” but with two contents — “textbook-like docility + tollgate-free freedom” — in your own words. In the next Chapter 6, we look at how the way you write that contract bears on the machine’s “low power.”