News for hackers. Stuff that matters.
Re:The Scheme Machine (1994) [pdf] by rjsw on Tuesday August 07, 2018 @11:44AM (#17707651)
One road not taken would have been to port the Lisp Machine software to a commercial CPU, not on top of a multitasking Operating System. Then add a native compiler. A fast 68020 plus a custom MMU would have been competitive with the microcoded machines.

A unikernel like MirageOS is basically an OCaml Machine, except that you don't develop software within it. Someone could create a Lisp on Xen equivalent.

Re:The Scheme Machine (1994) [pdf] by lispm on Tuesday August 07, 2018 @01:06PM (#17708410)
The view at Symbolics was mostly (AFAIK) that 32bit were not enough for their software - their microprocessor was a 40bit + 8bit ECC architecture. They also ported their software to the DEC Alpha, because that was a 64bit machine.

But generally I would agree that the 68020 + MMU would be mostly okay - given that MCL then ran fine under 68030 + 68040. But the MCL runtime was quite a bit weaker: cooperative multitasking, less CLOS and a less capable memory management.

Re:The Scheme Machine (1994) [pdf] by jecel on Tuesday August 07, 2018 @01:15PM (#17708496)
The very last version of the Symbolics machine wasn't hardware but instead was a virtual machine running on the Alpha processor. So I would say it was a road that was taken, though it was not a success (probably bad timing since I agree it was a good idea).

http://pt.withy.org/publications/VLM.html

Re:The Scheme Machine (1994) [pdf] by rjsw on Tuesday August 07, 2018 @02:00PM (#17708934)
The VLM doesn't have a native (to Alpha) compiler.