This processor is one of my favorite terrible ideas.
It’s a terrible idea because they didn’t talk to the compiler team until after they’d taped out the core so they ended up with a bunch of common operations requiring 200 or more cycles. Several things ended up being indirection tables that meant that even an optimised follow-on improved microarchitecture would still have been slow.
My favourite, because it was fully memory safe and capability-safe hardware, in the early ‘80s. This is basically solving the same problems that I am, just 40 years earlier. A lot of what we did in the CHERI project was explicitly not what the 432 did, because it explored a lot of the design space of things that don’t work.
There is a really fun book about the Rekursiv computer that was built by some random audio company in the UK in the 80s. They were trying to make hardware to run a Smalltalk-like system. The book itself is a really good overview of the architecture, as well as having its fair share of smart-assery. In the end, it looks like one of these computers ended up in the bottom of a canal somewhere.
The Mushroom project at Manchester in the ’80s also tried to build a computer for Smalltalk. They had hardware-assisted GC and a bunch of other interesting features, prototyped across 7 FPGAs. Some of the same folks ended up working at Sun on Project Maxwell, which almost added GC support to SPARC before being cancelled (their approach was really neat: hardware young-generation compacting in the cache, barriers to allow old-generation compacting in software to be efficient). Reading a bit about the Rekursiv computer, I wonder if any of the same folks were involved: some of the concepts resurfaced in Maxwell…
There are a number of capability systems that are basically gone into the mists of time. I was hoping to persuade someone to write an emulator for the Cambridge CAP Computer before all of the folks that worked on it died. I think someone found a printout of the system software a few years ago.
An FPGA 432 would be a fun student project if the data sheets or software existed.
This processor is one of my favorite terrible ideas.
It’s a terrible idea because they didn’t talk to the compiler team until after they’d taped out the core so they ended up with a bunch of common operations requiring 200 or more cycles. Several things ended up being indirection tables that meant that even an optimised follow-on improved microarchitecture would still have been slow.
My favourite, because it was fully memory safe and capability-safe hardware, in the early ‘80s. This is basically solving the same problems that I am, just 40 years earlier. A lot of what we did in the CHERI project was explicitly not what the 432 did, because it explored a lot of the design space of things that don’t work.
There is a really fun book about the Rekursiv computer that was built by some random audio company in the UK in the 80s. They were trying to make hardware to run a Smalltalk-like system. The book itself is a really good overview of the architecture, as well as having its fair share of smart-assery. In the end, it looks like one of these computers ended up in the bottom of a canal somewhere.
Linn. They’re still around.
https://www.linn.co.uk/uk/
The Mushroom project at Manchester in the ’80s also tried to build a computer for Smalltalk. They had hardware-assisted GC and a bunch of other interesting features, prototyped across 7 FPGAs. Some of the same folks ended up working at Sun on Project Maxwell, which almost added GC support to SPARC before being cancelled (their approach was really neat: hardware young-generation compacting in the cache, barriers to allow old-generation compacting in software to be efficient). Reading a bit about the Rekursiv computer, I wonder if any of the same folks were involved: some of the concepts resurfaced in Maxwell…
I’ve always loved this architecture.
I’m surprised that there doesn’t seem to be an emulator anywhere, nor even an effort for one.
There are a number of capability systems that are basically gone into the mists of time. I was hoping to persuade someone to write an emulator for the Cambridge CAP Computer before all of the folks that worked on it died. I think someone found a printout of the system software a few years ago.
An FPGA 432 would be a fun student project if the data sheets or software existed.
I think because the data sheets for this don’t exist, and what’s out there isn’t enough to go by to implement an emulator.
http://www.bitsavers.org/components/intel/iAPX_432/