[Info-vax] Dave Cutler, Prism, DEC, Microsoft, etc.

[Stanley F. Quayle]

On Wed, 2009-12-09 at 11:07 -0500, Richard B. Gilbert wrote:
> ISTR that a VAX to Alpha port was usually just a matter of compile,
> link and run.

There are lots of other "oops".  For instance, shared memory (global
sections) required a memory barrier instruction before any other process
could "see" the changes.  And there were lots of pain with default
longword-padding in C/C++.

A 200,000+ line C/BASIC/MACRO conversion I did "came up" in a few weeks.
It was about a year before it worked the same as the VAX version.  And
there's a 20,000 line C program that I've given up on -- no
documentation, all AST-driven, fast, efficient, and ACCVIO's at random
intervals.

I did discover that the MACRO compiler doesn't "like" programs past a
certain size.  I had to develop tools that would "subdivide" those
programs into separate modules.  And that limits those programs, since
they can't just jump willy-nilly into the other modules. which was a
"feature".

> People who had data files with binary floating point had to convert 
> them. I don't recall it as being a problem although I'm sure some 
> people did have problems.

On Alpha, no problem, since the processor supports VAX float values.  On
Itanium, you have to convert everything explicitly.

-- 
Stan Quayle
Quayle Consulting Inc.

----------
Stanley F. Quayle, P.E. N8SQ  +1 614-868-1363
8572 North Spring Ct., Pickerington, OH  43147  USA
stan-at-stanq-dot-com       http://www.stanq.com
"OpenVMS, when downtime is not an option"