[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <dfe99ec2-d761-2c1d-2c4-8f49da811f32@esi.com.au>
Date: Sun, 2 Aug 2026 16:58:54 +1000 (AEST)
From: Damian McGuckin <damianm@....com.au>
To: musl@...ts.openwall.com
Subject: Re: Does MUSL assume both float and double always exist?
On Sun, 2 Aug 2026, Rich Felker wrote:
> Yes.
Thanks,
> The C language requires that they do.
Not really relevant here.
> But if you mean as IEEE single and double, yes to that too. An arch
> that can't do that in hardware needs to use soft float.
IEEE 754 only mandates one. And it can be neither of float, double or long
double. Again irrelevant. These days it can be BINARY16.
Both asinh() and acosh() can be tweaked to both resolve the problem Paul Z
posted and eliminating that >1*ULP inaccuracy in the domain [1..2). The
first I have already suggested and the latter by computing
x + sqrt(x * x + 2 * x)
beyond normal precision and recovering accumulated rounding errors.
If somebody has the time, feel free to attack it. Otherwise I will put it
on my schedule for the New Year break when I get some spare time.
Thanks again - Damian
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
