🚀 go-pugleaf

I have been following this discussion only loosely, but Peter Grandi's
comment on APL permits me an opening to climb on my soapbox.

I have been for several years convinced that APL should be a wonderful
language for parallel processors, particularly SIMD machines but maybe also
MIMD machines (see my article in TOPLAS in 1984).  I am convinced that 
the cost of parallel hardware is coming down, and the accessibility of
such hardware to ``normal'' programmers will increase dramatically in the
next few years, and yet acceptance of such hardware will be slow as long as
programmers need to think about parallelism.  That is, programmers have
enough to think about as it is, and if they need to provide explicit
parallel directives in their language then only the committed will do it.
Ergo we need languages that have lots of parallelism implicit, such as APL.

The catch - making a compiler for a high level language such as APL you can 
never get all the efficiency of a low level language, such as C or Assembly.
Currently the people using parallel hardware are very committed to getting
the last iota of speed out of their machines, and if you say something like
``I can write my program in APL and spend a tenth the time in writing, but
it executes only half as fast as your C version'' you're going to get
laughed out of the room.  I feel like I'm in the 1950's again with the
assembly/HLL debate.  I'm waiting for ten years to go by, so I can say ``I
told you so''.  In the meantime, I'll keep plugging away at my algorithms
for APL and parallel hardware.
--tim budd, budd@cs.orst.edu