Re: complex instruction set compiler

Eric Hildum <Eric.Hildum@Japan.NCR.COM>
18 Mar 1998 23:03:56 -0500

          From comp.compilers

Related articles
complex instruction set compiler (Vipin Jain) (1998-03-15)
Re: complex instruction set compiler (1998-03-15)
Re: complex instruction set compiler (David Chase) (1998-03-15)
Re: complex instruction set compiler Eric.Hildum@Japan.NCR.COM (Eric Hildum) (1998-03-18)
Re: complex instruction set compiler (1998-03-18)
Re: complex instruction set compiler (Tom Moog) (1998-03-20)
Re: complex instruction set compiler (Jan Vorbrueggen) (1998-03-20)
| List of all articles for this month |

From: Eric Hildum <Eric.Hildum@Japan.NCR.COM>
Newsgroups: comp.compilers,comp.arch,
Date: 18 Mar 1998 23:03:56 -0500
Organization: NCR Japan
References: 98-03-144 98-03-149
Keywords: architecture says...
> > I m looking for a rich complex instruction set which has a code
> >generator (or code generator generator ) to go along with it. It would
> > Vipin
> >[There's always the Vax back end for GCC. -John]

Ian Kemmish wrote:
> The compiler for Franz Lisp in BSD 4.2 (also for VAX, of course, and free on
> the BSD source tapes) actually *used* the complex instructions. I *think* it
> used the procedure call stuff, and I definitely remember it used the POLY
> instruction, and lots of lovely addressing modes....
> [I was under the impression that the very complex Vax instructions usually
> ran slower than a sequence of simpler instructions. -John]

I have heard, but do not know for certain, that the COBOL compiler
used the Edit string instruction. As I recall, the code generator for
the Fortran compiler became the base for the VAX standard code
generator used in the later compilers - that code generator created
the best code, and also knew the various instruction sequences to
avoid and ideal alignments for the best performance on all the
different VAX implementations.

By the way, any program that used trig functions on under VMS used the
poly instructions - they were used in the libraries. The backup
program used the CRC instruction - in fact, removing CRC from the HW
in the uVAX I and II was considered a mistake at the time because of
the dramatic effect on backup time requirements when the backup
default was /CRC.
[I can believe that EDIT helped; it was one of the few CISC instructions
in the IBM 360 architecture. But depending on a CRC instruction was pretty
silly since you can do CRC so easily with a small precomputed table and
a few shifts and masks per byte. -John]


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.