[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 6 Jan 2020 11:17:44 +0300 (MSK)
From: Alexander Monakov <amonakov@...ras.ru>
To: musl@...ts.openwall.com
Subject: Re: [PATCH] math: move x86_64 fabs, fabsf to C with inline
asm
On Sun, 5 Jan 2020, Rich Felker wrote:
> I'm not sure if "this logic" carries over. fabs is a common operation
> (ideally compiler would inline it anyway in the caller, though).
> copysign not so much.
Indeed, in practice we want compilers to recognize and inline simple math
functions, especially on x86 where all float registers are call-clobbered.
This makes it moot, except maybe for musl-internal uses like in floatscan.c,
but there the right solution is probably to use __builtin_ forms if available.
(to be clear: I think musl may not implement math functions in terms of
math builtins because it might result in a circular dependency, but using
builtins for complex math and higher-level functions like in floatscan.c
should be fair game)
For functions like sqrt GCC can be changed to default to -fno-math-errno
on *-musl targets, as progress on flipping it globally seems slow.
> Really I'm not even sure it makes sense to have the asm here at all
> for fabs either, but perhaps with the gratuitous stack access in the
> older-GCC version it does...?
That's why I'm asking, the situation just looks ambiguous to me, I can't
deduce the intent and I could imagine arguments both ways. I think I'd
prefer a more inclusive playground. Please come up with some policy or
guidelines if possible.
Alexander
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.
