[Info-vax] What Will Drive More OpenVMS Adoption?

[Arne Vajhøj]

On 12/5/2021 11:40 AM, Bill Gunshannon wrote:
> On 12/5/21 5:52 AM, John Dallman wrote:
>> In article <61ac02f5$0$704$14726298 at news.sunsite.dk>, arne at vajhoej.dk
>> (Arne Vajhøj) wrote:
>>
>> When you need the best performance you can reasonably get, using
>> manufacturer extensions becomes necessary.
> 
> And yet, for at least the past 15 years, anytime I have brought
> up a need for efficient programming I am told (by Professors
> and practitioners of the trade) that it is irrelevant, "Just
> throw more hardware at it!"  Wouldn't the same apply to using
> non-standard extensions to get "the best performance you can
> reasonably get"?  We're not running VAXen any more and we have
> more than 12MB of memory.

Performance is still very important today.

But what was relevant for good performance 30 years ago and
what is relevant for good performance today is very different.

Writing complex code to save a little CPU or memory is
typical a total waste of time today. Programming
languages and compiler settings are also often without
significance today. The change in HW prices has done that.

(typical is a normal business application - there are
exceptions in HPC simulations, cryptography brute force
attacks and other special cases)

But there are plenty of performance problems in the real
world. Data being stored the wrong way, data being accessed
the wrong way, too many network hops, something being interpreted
that should have been compiled, bad algorithms etc..

In general I will say that bad performance today is typical
due to error in architecture or high level design and rarely
due to error in coding.

Arne