The System/36 developers in 1993 recognized that the RISC processor to be used for the AS/400 was sufficiently fast to emulate the MSP instruction set without the need for a separate hardware processor. If such an emulator was built into SLIC, along with all the CSP code, then the SSP operating system would run directly on the new RISC processor. The effort the developers undertook was to create the emulator and to rewrite the CSP code in C++ as a part of SLIC.

The interface between the original MSP and CSP was a Supervisor Call (SVC) interface. An SVC is an instruction that executes in the MSP and requests an action to be taken in the CSP. This is conceptually the same as executing an instruction at the MI level to request SLIC to perform some action. The developers reasoned that if they extended the MI to include this SVC interface, they could run the SSP on top of the MI. This would also make the SSP technology independent. They called this extension to the MI the Technology Independent Emulation Interface.⁷

---

⁷This emulation interface also formed the basis for interpreting Java on the AS/400. We examine this further in Chapter 11. Until this emulation interface was added, all MI code had to be translated to the hardware level before being executed. Now there is the possibility for some of the MI to be interpreted.

---

With the decision to use an emulator rather than a separate hardware processor for the MSP instruction set, the developers had decided to create a design that internally looked more like a System/32 than either a System/34 or a System/36. As happens so often, history repeats itself. The bionic desk (aka the System/32) lives again in a new black box!

Advanced 36

We suddenly had a brand new System/36 that ran on the new 64-bit RISC hardware using the SLIC kernel. Furthermore, it was a pure System/36. No modifications were made to the SSP that changed any application interfaces. The new system was totally binary-compatible with the System/36. Recompiling was not even necessary to move an application to the new system.

Clearly, this new System/36 personality could run on the AS/400 when the RISC processors were announced. There was, however, another possibility. We could build an early version of the Cobra processor with everything (the processor, the cache, and the I/O interface) on a single chip. We put a small Rochester team in place and soon they had a single-chip processor based on the Endicott design. This processor ran at 50 MHz, which was fast enough for any System/36 application. We called the processor Cobra-Lite, because it was missing about 17 instructions from the required 64-bit PowerPC set. These instructions, which were primarily floating-point instructions, were implemented in the software, but that didn’t matter. The missing instructions, including a combination floating-point multiply and add instruction used for matrix calculations, were not used by the SLIC and had no effect on the new System/36.

IBM confirmed that its semiconductor facility in Burlington, Vermont, could produce these special chips in sufficient quantity to ship RISC-based System/36s late in 1994. We knew we could package an early version of SLIC along with a new release of SSP for this new product. We also knew we had a major dilemma on our hands.

Here we were, able to ship IBM’s first 64-bit RISC computer — but it was a System/36! We had been telling our customers for years to get off the System/36, and now we were ready to ship a new one with the latest technology available. This was going to be a hard sell.

IBM organizations all over the world turned thumbs down on the idea of a new System/36. Finally, we convinced management to let a few of us talk to customers and business partners about the new product and let the market decide whether we should announce the new system in 1994. I had the opportunity to disclose the existence of this new System/36 at our largest business partner conference in early 1994 and to ask the business partners whether they were interested. The feedback was overwhelmingly in favor of announcing the system. Many business partners offered to write us a check on the spot for the demo machine we showed them. IBM and business partner marketing people quickly got behind the idea, realizing the potential of this system. In October 1994, the Advanced 36 was announced, and it became an immediate success.

The initial model of the Advanced 36 ran only the SSP operating system. Later models run OS/400 and SSP side by side on the same processor. The Advanced 36 showed the power of the AS/400 architecture and its ability to integrate new function, and even whole operating systems, in a totally seamless manner. Every RISC model of the AS/400 now has a System/36 inside the SLIC. Perhaps a small label saying “System/36 Inside” would be appropriate for every new AS/400.

Conclusions

Integration creates a unique value proposition for the AS/400. The components are designed to work together in a complementary fashion. Integration also makes it difficult to study the individual components in the same way you can study some other system. For example, we saw earlier that the implementation of a component such as security is split between OS/400 and SLIC. Studying just OS/400 security would not give a complete picture. Contrast this to some other system where security is a separate, self-contained package on top of the operating system.

Examining the AS/400 as a series of horizontal layers does not give a very complete picture of the system. A better way to understand the AS/400 is to look at vertical slices — in other words, to look at the whole function, parts of which are in OS/400, in SLIC, and in the hardware.

But first, in Chapter 4 we are going to look at the MI layer in the AS/400. This will give you a better understanding of the types of functions that are implemented in OS/400 and in the SLIC. We also will see how an application program is compiled down to the hardware level before it is executed.

Following Chapter 4, we look at the major AS/400 components as a series of vertical slices. Only then can we see that the old adage “The whole is greater than the sum of its parts” is very true for the AS/400.