Re: Performance of records with variant parts

[John Perry <john.perry@usm.edu>]

On Monday, March 22, 2021 at 12:39:54 PM UTC-5, Dmitry A. Kazakov wrote:
> On 2021-03-22 18:02, John Perry wrote: 
> 
> > Is there a reason that the record with variant parts runs so much slower? I can make the complete source available if need be.
> It must check the discriminant if the object is dynamically constrained.

I'm not quite sure I understand. Are you saying that it has to check the discriminant at run-time? How would that differ from checking the value if it were merely a field of the record?

(Sorry if this is is a dumb question, but I guess I don't understand variant records as well as I thought.)

> Checking the discriminant and accessing a scalar field is roughly in the 
> same league, so 15-30% penalty is a quite good outcome. I would expect 
> 50% or even worse with cache misses between independent memory accesses. 

OK, so 15-30% is actually good news. Interesting.

john perry