This is the mail archive of the
libstdc++@sourceware.cygnus.com
mailing list for the libstdc++ project.
Re: fixed check for sinl in acinclude.m4
- To: "Benjamin Kosnik" <bkoz at redhat dot com>
- Subject: Re: fixed check for sinl in acinclude.m4
- From: Steven King <sxking at uswest dot net>
- Date: Thu, 22 Jun 2000 20:06:27 -0700
- Cc: libstdc++ at sourceware dot cygnus dot com
- Organization: is the root of all evil
- References: <Pine.SOL.3.91.1000622193433.13146B-100000@cse.cygnus.com>
- Reply-To: sxking at uswest dot net
On Thu, 22 Jun 2000, Benjamin Kosnik wrote:
>
> No. That's the whole point. Try just doing a simple case for
> __builtin_sinf: the x86 compiler converts this to a call to sinf.
>
> Thus, you can't link against libm (which has a def for sinf) if you want
> to detect what calls actually are __builtin....
Did you try generating a .s file? The x86 compiler generates a fsqrt
instruction, then a check of the fpu status, and then a call to the
sqrt function. The sqrt function is only called if the fsqrt instruction sets
the error bit in the fpu status register. This looks pretty builtin to me, but
it fails the tests as they are currently written.
>
> Anyway.
>
> This code is now working correctly..... it might help if you can compile
> on different platforms than x86. It becomes amazingly clear what's up
> when you do this....
Its not working correctly on x86. I'd love to be able to test this stuff on
different archs. Know anybody willing to loan me a G4?
--
Steven King
sxking@uswest.net