By Noah | April 7, 2014
It’s been way too long since I posted here, but the 50th anniversary of the announcement of the IBM 360 seems like a good excuse. Of course, the 360 was to its era what the Intel architecture is today, and in fact even more. Not only was its instruction set the code of choice for business and much other computing, the 360 also defined the I/O interconnection architecture for a generation of IBM computers. If your disk or tape drive couldn’t talk to the IBM 360 channel, its market was very limited.
What I think too often gets missed is that the 360 still stands up as a remarkable achievement in system design. This is an instruction set and machine architecture that was created in the early 1960s, at a time when fully integrated circuits were too new for IBM to adopt. Cycle times were still measured in microseconds for all but the fastest machines. Today, programs written 50 years ago continue to run on CMOS mainframes with cycle times of under 1 nanosecond. Indeed, assembler code that I wrote for 370 machines in the 1970s is still being used, unmodified, by IBM’s customers today.
The designers of the 360 included Gene Amdahl, Gerrit Blaauw and Fred Brooks, and they achieved something truly remarkable. When I got to Stanford in the late 1970s everyone was excited about the DEC VAX, a machine with sophisticated instruction formats and a built in stack. The 360 was seen as old fashioned. Time went by and as manufacturers tried to scale their architectures DEC came to realize that the VAX could not be properly pipelined. The same instructions that were easy targets for compilers were hard to decode in an efficient way. There were probably other problems as well. The Vax was in many ways a lovely machine, especially for programmers, but it didn’t scale for even 20 years. The same was true for many other architectures that came and went. Indeed, one can argue that the only reason Intel has managed to scale the 8086 architecture is that chips now have so many circuits that tremendously complex techniques can be applied.
For anyone who’s interested in computer architecture this is a good time to step back and celebrate the achievements of the team that built the first truly scalable, compatible computer architecture.
BTW: IBM will apparently be live streaming a celebration event tomorrow, April 8, 2014.