From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.5-pre1 (2020-06-20) on
	ip-172-31-74-118.ec2.internal
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 required=3.0 tests=BAYES_00 autolearn=ham
	autolearn_force=no version=3.4.5-pre1
X-Received: by 2002:aed:2143:: with SMTP id 61mr1943890qtc.136.1616450054085;
        Mon, 22 Mar 2021 14:54:14 -0700 (PDT)
X-Received: by 2002:a25:2484:: with SMTP id k126mr972523ybk.165.1616450053951;
 Mon, 22 Mar 2021 14:54:13 -0700 (PDT)
Path: eternal-september.org!reader02.eternal-september.org!weretis.net!feeder8.news.weretis.net!proxad.net!feeder1-2.proxad.net!209.85.160.216.MISMATCH!news-out.google.com!nntp.google.com!postnews.google.com!google-groups.googlegroups.com!not-for-mail
Newsgroups: comp.lang.ada
Date: Mon, 22 Mar 2021 14:54:13 -0700 (PDT)
In-Reply-To: <ibs7okF5i52U1@mid.individual.net>
Injection-Info: google-groups.googlegroups.com; posting-host=2601:3c3:401:f550:f6d1:8c0b:6974:417;
 posting-account=JSxOkAoAAADa00TJoz2WZ_46XrZCdXeS
NNTP-Posting-Host: 2601:3c3:401:f550:f6d1:8c0b:6974:417
References: <07a56dcc-9e17-49b2-a980-3a5a2d265cedn@googlegroups.com> <ibs7okF5i52U1@mid.individual.net>
User-Agent: G2/1.0
MIME-Version: 1.0
Message-ID: <b8037068-5da7-4c2a-ac74-7cc7790e11a3n@googlegroups.com>
Subject: Re: Performance of records with variant parts
From: John Perry <john.perry@usm.edu>
Injection-Date: Mon, 22 Mar 2021 21:54:14 +0000
Content-Type: text/plain; charset="UTF-8"
Xref: reader02.eternal-september.org comp.lang.ada:61645
List-Id: <comp.lang.ada>

On Monday, March 22, 2021 at 1:59:34 PM UTC-5, Niklas Holsti wrote:
> What is your optimization level? -O2? 

I've tried several. The one with the best performance is -O3 -funroll-loops -gnatn.

> Moreover, in your case the compiler should 
> statically know the offset of each discriminant-dependent component, so 
> accessing such a component should not require any overhead compared to a 
> record without discriminants. 

Precisely this perplexes me. Nothing in this section of the code would be unknown after the discriminant check. -- Nothing that seems obvious to me, anyway. I'll copy some code in a reply to someone else in a moment.

> However, in the past I've seen unexpectedly slow code for copying 
> records with discriminants. ...
> [very interesting material excised for brevity]
> If the compiler is stupidly checking the discriminant of obj for each 
> obj-component access, you may be able to speed up the code by creating 
> separate sub-procedures

I liked this idea and thought it might be the problem, but that also didn't help out. I may have to look at the assembly, which I am loath to do. :-/

john perry