🚀 go-pugleaf

Dave Gladstein <daveg@com.icad> writes
      From: mikef@dcs.ed.ac.uk (Mike Fourman)
      Poly/ML is a mature system; it has evolved over a period of more than
      six years. The compiler and run-time system are both reliable products
      that have passed the test of time; Poly/ML is being used to produce
      industrial-scale software products. Large source files will not break
      the Poly/ML system (modules containing thousands of lines of code are
      common) nor will large data-structures (each workspace may grow to
      32Mb). The speed and code quality of the Poly/ML compiler are both up
   
   As a matter of curiousity, why the restriction?  We've had complaints
   from customers when our lisp-based product hit a hardware vendor's
   limit of 256Mb per process.
   

Mike Crawley of AHL writes:

We also have a hardware vendor's limit. The Sparc workstations
we use cannot address more than 256MB per process either.
We are waiting for new workstations and operating systems from Sun
that make these limits much bigger - like 4GB per process.

Poly/ML allows 256MB of dynamic data and 36MB per level in each persistent
hierarchic database (each database is roughly 32MB immutables and 4MB 
mutables), with up to 4 levels of hierarchy giving a total of 144MB of 
static data. (A further 36MB is reserved for garbage collection of the 
persistent store.)

There is an address range 0x01000000 to 0x0D000000 available
for the static data in databases. This is actually 201,326,592 bytes.
Hopefully newer machines won't have this limit.

-------------------------------------------------------------------------------