The Story of Mel...

[Larry sreyoB]

Can't remember this ever being posted here (or here II, or here I), but I wouldn't be surprised if it was...

The Story of Mel.

Here's some information on the computer from the story, the LGP-30.

And there's even a Wikipedia page that tells the story of the story!

[Ward Miller]

Larry, the best programmer I ever knew wrote the FORTRAN compiler for the 4K IBM 1401. With so little memory, he loaded the user's FORTRAN program into the machine, then passed the compiler, in punched cards, against it. It took just two passes. I later sent the programmer to MIT, on IBM's dime, for his PhD.

[Robert Dubner]

The first computer my father programmed was the LGP-30, around 1957 or so. 4096 31-bit words on a drum memory. There was no RAM. It had something over 100 vacuum tubes, and the design was tweaked to minimize that number; tubes were expensive and failed often. Weighed about 750 pounds, but was built on casters, so it was marketed as "portable", meaning that it could be moved near an engineer's desk. Sixteen opcodes. Fifteen milliseconds to do an addition. He used that machine to do the ray-tracing that led to the infrared seeker head on the original Sidewinder missile.

Writing efficient code on that machine wasn't easy. Each and every instruction had an implicit GOTO built into it, and just putting code on the drum sequentially guaranteed terrible performance. (This is the main theme of A Programmer Named Mel.) You had to know how long an instruction would take, and then put the next instruction in a location that would just be coming under the heads of the drum as the current instruction ended. What he learned on it informed his entire career, and, by induction, mine; his influence on me was, well, huge. I never programmed the LGP-30, but sometimes I feel like I did.

A buddy of mine, having learned about the LGP-30 from me, was ready to be contemptuous of the primitive technology. But then he got hold of the "programmers guide", which started off with a lengthy description of the machine's architecture -- you couldn't program it well without knowing how it was designed. He came away very impressed by what the machine's designers pulled off given the inherent limitations of the technology at the time.

[Paul Millner]

Enjoyed reading that! In the early ‘80s, I met the developer of COSMIC: Chevron Optimization System for Modern Industrial Control. The architecture was a thing of beauty... in 4K of core memory, and with a 1k word rotating drum, he achieved advanced process control that industry took 20 years to catch up to. The system was hyper-optimized for running refineries and chemical plants. The last one shutdown ~2005, having been ported to VAX in the mid 90’s, as part of outsourcing support. Operators and engineers complained about the loss of functionality adopting the 25 year newer systems. RIP Roger Humphrey!

oh, and Roger was a pilot, built an RV...