[Info-vax] Updated HPE/VSI OpenVMS V8.4-2L1 Marketing Brochures

[David Froble]

Bob Gezelter wrote:

> David,
> 
> Actually, as I noted in my presentation at last week's Bootcamp, the port to
> x86-64 is simpler than the port to Itanium (I will be uploading my
> presentations to my firm's www site in the next weeks).

Bob, I don't know how it could be much simpler.  The Alpha to itanic port, with 
one exception, which was a programming bug on my part on image rundown that 
Alpha didn't report, and itanic squaked over, was compile, link, and run.

Your stuff is always an interesting read.

> Put simply, there are no representation issues when going from Itanium to
> x86-64. They are both 64-bit environments with IEEE floating point. If one's
> source base is current with regards to language standards, there is little
> issue.
> 
> If one's Alpha-based code is using VAX floating point (Alpha support both VAX
> and IEEE formats), there is the issue of verifying that there are no issues
> raised by the slight differences in computation related to the differences
> between the two formats. As before, my suggestion is to do the change in
> floating point BEFORE changing architectures.

We use D-float, and already know of the issues, which we've overcome.  As long 
as that doesn't change, we're expecting another compile, link, and run.

> Some other small changes should be made to any conditionalization, to reflect
> that the architecture is not the relevant issue, but the representations are
> the issue. There are those who will notice that data structures need reflect
> the fact that the size of a longword is not necessarily the size of a
> pointer.
> 
> - Bob Gezelter, http://www.rlgsc.com