Xref: csiph.com comp.arch:44365
Newsgroups: comp.arch
Path: csiph.com!goblin2!goblin3!goblin.stu.neva.ru!news.misty.com!kithrup.com!mrs
From: mrs@kithrup.com (Mike Stump)
Subject: Re: A right alternative to IEEE-754's format
Message-ID: <p7LrJM.1H2v@kithrup.com>
Date: Sun, 22 Apr 2018 19:59:46 GMT
References: <0d4dc7f8-1819-43e5-8082-6ff7aee5f41b@googlegroups.com> <00f57960-d344-4e58-bd88-e197ceb656a5@googlegroups.com> <paqahj$1g3a$1@gioia.aioe.org> <paqegh$n9d$1@dont-email.me>
Organization: Kithrup Enterprises, Ld.
X-Newsreader: trn 4.0-test77 (Sep 1, 2010)

In article <paqegh$n9d$1@dont-email.me>,
David Brown  <david.brown@hesbynett.no> wrote:
>At no point do you do: "compile --variant-optimal-for-everything" and get:
>
>	pick-best-machine-code-for-lsl r1, r0, #2
>
>and then use different assembly lines for different targets, expecting
>the assembler to generated different types of machine code.

Well, to be complete, yes, a port writter for clang/gcc can trivially
do this, some do.  Mill as an example does this to a degree.  Your
point remains however, most do not do this for most of the constructs.