On 3/14/21 11:36 AM, Guy Sotomayor via cctalk wrote: > > I can say with a fair amount of certainty, that we at IBM knew of the > existence of the LOADALL instructions including all of it's warts (and > its inability to switch back from protected mode) from the earliest days. >
ca. 1980, we were in search of a decent 16-bit processor for upward migration. Code base compatibility wasn't much of a concern--we were resigned to the prospect that everything had to be recoded, so code migration wasn't a concern. We played a bit with the 8086, but couldn't see a path forward in the immediate future. So we in engineering decided that the best CPU was the Moto 68K and we wrapped up a test board and started cutting some code for it. Bill Davidow was on our BOD and when he got wind of our efforts, he nearly went through the roof. When we defended our decision by saying that the 8086 was limited in possibilities, with the rumored iAPX432 nowhere near reality, so the 68K was the logical next step. Zilog had their Z800, but after looking at the cost of the MMU and the fact that it, too was a segmented memory CPU and slower than the 68K, we never got any farther than talking about it. We did have an Onyx box for software development in the lab, however. Bill put us on the OEM pre-release steppings for both the 186 and the 286. We got the 186 going long before the 286 (Intel, IIRC, had taken on the job of writing the Xenix kernel for Microsoft). Davidow's stubbornness cost us months of product delay in getting a multi-user system out the door. We certainly were not advised about the LOADALL instruction at that time. Nor, I suspect would it have made much of a difference. In our system, the 186 did the I/O heavy lifting for the 286. My impression was that the design for the 286 never envisioned the need to switch back and forth between real and protected mode. We never did. --Chuck
