Yes, that was a pretty nice system. Certainly not the first ALGOL system, but a decent one even though they did put a bunch of Fortran-like ugliness into the I/O. PDP11 DECUS ALGOL was clearly inspired by that, it?s a subset of Burroughs ALGOL and the generated code looks like a 16-bit variant of B5500 machine code. Note though that some of the discussion was about Algol 68, which is a rather different language. I don?t know that Burroughs ever did anything with it, but some other companies did (CDC for one). ?native? in what sense? There are plenty of machines, from many companies, that support block structured languages well. The PDP11 and VAX are among those, as are the Burroughs mainframes, the Electrologica EL-X8, and many others. If so, they will do well at Algol, Pascal, C, Modula, Ada, and so on. Some other machines don?t make stack operations and recursive functions easy, and as such could be called ?Fortran machines? ? CDC 6000 series, IBM 360, and PDP-8 are examples. Still, with a bit of extra work those too can handle Algol and other block-structured languages. After all, the first Pascal was implemented on the CDC 6600 (by ETH Z?rich). And CDC shipped both Algol 60 and Algol 68 compilers. IBM had an Algol 60 for OS/360, as I recall, though I?ve never used it. If you mean ?native? in the sense of an instruction set tailored for running Algol programs, no ? in that sense, Burroughs was rather unusual, though you might point at the Electrologica EL-X8 as another example. paul

It was not defined at all. I wouldn?t conclude that. The natural way to do I/O in Algol is exactly what it is in C: using a set of libraries. In fact, C is arguably no different than Algol in that I/O is not part of the base language. The difference is that in C there came to be a fairly consistent ?standard library?. The Dutch Algol implementations use that model, and there were several different compilers that used roughly the same set of functions. With that approach, all that is needed by way of extension to Algol is the addition of a ?string? type (for literal strings, not string variables, minimally). Burroughs however took a very different route, which is much more invasive into the underlying language, and feels a lot like a grafting of Fortran FORMAT statements onto Algol. Not in the PDP-11 machine architecture. DECUS Algol is a P-code implementation: the generated binary is pretty much Burroughs machine code apart from the word length, and the runtime library interprets that code. So you see descriptors there, and B5500 style stack operations, and all that, but it isn?t native PDP-11 machine code. No, it was a CDC product, but developed by CDC Holland (at their Rijswijk office). Apparently it was created at the insistence of a number of CDC?s academic customers in Europe. Which makes sense; it demonstrates what nearly everyone now knows, which is that RISC architecture is a very good way to design a computer. I suspect part of the reason is that Algol wasn?t all that popular in the USA even if its heyday. Add to that the fact that most computer designers weren?t all that skilled in software. And finally, as the RISC experience has shown, it isn?t really worth it ? a well designed RISC architecture supports any language well, with a simpler, faster, and more economical hardware architecture. paul

Every machine needs a fast memory system. CISC machines just as much, after all the number of memory references per operation of a given kind doesn?t depend on the sort of CPU architecture you use. All that changes is whether the cycles are issued by regular machine code, or micro-engine actions. A full-up 6600 is 32 way interleaved; half size you get 16 way interleaving. Once nice benefit is that context switching takes only a few microseconds, because the exchange jump would swap current and new context at memory speed: 16 words issued at 100 ns intervals once the operation gets moving. paul