[Info-vax] Kednos PL/I

[gah4]

On Wednesday, September 13, 2023 at 6:37:49 AM UTC-7, cao... at pitbulluk.org wrote:

(snip)

> Not that long ago, I was playing with the Kednos PL/I compiler 3.8-003 for VAX.
> It does some rather excellent code optimisations but gets a bit weird with procedures
> that have multiple entry points - I've used them as single-level closures or just
> encapsulating some locally accessible global state.
> I don't recall other PL/I compilers being so sensitive during syntax analysis though.

I don't know anything about the actual compiler at all.  Recursive descent compilers
are well known for failing in a strange way with syntax errors.

> The Kednos compiler trips up really badly on the simplest of syntax errors
> (missing semicolons most obviously) and the fallout is quite spectacular.

The PL/I compiler I used most over the years, is the IBM PL/I(F) compiler
that dates back to the 1960's.  I do remember an error message like
"SEMICOLON MISSING, ONE INSERTED" where it inserts one into the
internal representation and continues on.  Often with strange results.

If I can get a hobbyist license for the VAX compiler, I will try running it.

> Although floating types will map appropriately to F, G/D or H representation,
> I was also a little disappointed that integer handling appears limited to a
> maximum of 32 bits (31 + sign bit). 
> It was a nice trip down memory lane. 

That would agree with VAX.  Though following IBM, and the VAX hardware,
it should have FIXED DEC(15,0). 

Does Alpha have 64 bit multiply and divide?

I even forget now, does x86-64 hardware have multiply and divide with 128 bit
product and dividend?