Date: Fri, 14 Sep 2012 17:18:08 -0700 From: Steve Kargl <sgk@troutmask.apl.washington.edu> To: Brooks Davis <brooks@freebsd.org> Cc: Ian Lepore <freebsd@damnhippie.dyndns.org>, Tijl Coosemans <tijl@coosemans.org>, current@freebsd.org, Dimitry Andric <dim@freebsd.org>, toolchain@freebsd.org Subject: Re: Clang as default compiler November 4th Message-ID: <20120915001808.GA70215@troutmask.apl.washington.edu> In-Reply-To: <20120914202319.GB5244@lor.one-eyed-alien.net> References: <20120911120649.GA52235@freebsd.org> <20120911132410.GA87126@troutmask.apl.washington.edu> <504F4645.4070900@FreeBSD.org> <504F4A6B.4010001@coosemans.org> <504F5101.8090906@FreeBSD.org> <505101C3.70203@freebsd.org> <20120913020833.GA8255@troutmask.apl.washington.edu> <1347550332.1110.108.camel@revolution.hippie.lan> <20120913161024.GA13846@troutmask.apl.washington.edu> <20120914202319.GB5244@lor.one-eyed-alien.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Sep 14, 2012 at 03:23:19PM -0500, Brooks Davis wrote:
> On Thu, Sep 13, 2012 at 09:10:24AM -0700, Steve Kargl wrote:
> > ok 1 - cexp zero
> > Abort trap (core dumped)
> > *** [tests] Error code 134
> >
> > Stop in /usr/src/tools/regression/lib/msun.
>
> Prompted by this post, I did a bit of testing and it looks like we have
> two classes of failures that need to be investigated. First, some tests
> are failing with a gcc compiled world and clang compiled test code.
> They seem to mostly be unexpected fp exception state when testing with
> NaNs. I suspect that someone more knowledgeable in this area could come
> up with a reduced test case and explication of what clang is doing wrong
> pretty quickly.
>
> The second class is tests that fail when libm is compiled using clang.
> I've not investigate those at all. I'd tend to guess that we have a
> wider range of issues there.
>
A third class of failure appears to be that clang emits
i387 fpu instructions for at least sinf and cosf instead
of calls to the library routines. AFAIK, the library
routines are faster and more accurate.
% cat a.c
#include <math.h>
float
foo(float x) {
float a;
a = sinf(x) * cosf(x);
return (a);
}
% clang -S -O a.c
foo: # @foo
# BB#0: # %entry
flds 4(%esp)
fld %st(0)
fcos
fxch
fsin
fmulp
ret
% cc -S -O a.c
foo:
pushl %ebp
movl %esp, %ebp
pushl %ebx
subl $20, %esp
movl 8(%ebp), %ebx
movl %ebx, (%esp)
call sinf
fstps -8(%ebp)
movl %ebx, (%esp)
call cosf
fmuls -8(%ebp)
addl $20, %esp
popl %ebx
popl %ebp
ret
% clang -S -O -fno-builtin a.c
foo: # @foo
# BB#0: # %entry
pushl %ebp
movl %esp, %ebp
subl $24, %esp
flds 8(%ebp)
fsts -16(%ebp) # 4-byte Folded Spill
fstps (%esp)
calll sinf
fstpt -12(%ebp) # 10-byte Folded Spill
flds -16(%ebp) # 4-byte Folded Reload
fstps (%esp)
calll cosf
fldt -12(%ebp) # 10-byte Folded Reload
fmulp
addl $24, %esp
popl %ebp
ret
Using -fno-builtin would seem to be a non-starter in that it
disables all builtin functions.
--
Steve
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120915001808.GA70215>