On Mon, 23 Sep 2019 at 13:27, Clem Cole <clemc@ccc.com> wrote:

On Fri, Sep 20, 2019 at 7:08 PM Robert Clausecker <fuz@fuz.su> wrote:
  It's a bit rough around
the edges and the runtime code needs some work (division and
remainder are missing), but it does compile B code correctly,
generating acceptable code (for my taste, though the function call
sequence could be better).
A suggestion,   Load TSS/8 on to your simh system with its Algol compiler and look at how it generated code.  I would suspect you can use Algol's calling conventions and probably some of its runtime.   Google is your friend.  I had it running a while back, but do not have it active at the moment - the key is all the pieces should be findable in the wild,.

I hope some of you enjoy this compiler for an important historical
language for an important historical computer (makes me wonder why
the two weren't married before).
Might have been, although when Ken created B for the PDP-7, BCPL was his model and there were already implementations of BCPL around for a number of processors.  I would not be surprised if there was a BCPL/8.  I would check in the DECUS library, much of which I think can be found online these days ??bit savers??.

I checked here: http://so-much-stuff.com/pdp8/software/decus.php but did not immediately see anything BCPL related.  8-330 is TSS/8 ALGOL.