Re: pessimal storage allocation, Continuations

On 15/07/2024 21:07, John Levine wrote:

... In the 1960s the IBM 360 Algol F compiler,
which I always assumed was written because some European institution
had it on a checklist rather than because they expected anyone to use
it, did the same thing. Every time it entered a block it did a GETMAIN
system call, and when it left the block it did a FREEMAIN, with all of
the performance you'd expect.

The /360 F level ALGOL compiler was notoriously bad at procedure calls,
and now I know why.  On the Ackermann function benchmark it is thought
to have required about 820 instructions per call.  Not surprising if it
was using an OS store allocation routine to set up each stack frame.
For comparison with near contemporary machines, and compilers:
IBM, F Level,     /360 ALGOL 60: 820 ins/call*
NU                1108 ALGOL 60: 175 ins/call*
Delft U,          /360 ALGOL 60: 142 ins/call
Wirth@Stanford U, /360 ALGOL W:   74 ins/call
Edinburgh U, EMAS /360 ALGOL 60:  21 ins/call
York U,            VAX Ada:        8 ins/call
* Estimated
The VAX Ada compiler was 12 years later.
Data taken from Brian Wichmann's study.
See <http://www.findlayw.plus.com/KDF9/Documents/Benchmarking/>
and
<http://www.findlayw.plus.com/KDF9/The%20KDF9%20and%20Benchmarking.pdf>)
--
Bill F.