This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Use mantissa_t in __mpexp
- From: Siddhesh Poyarekar <siddhesh at redhat dot com>
- To: Andreas Jaeger <aj at suse dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Tue, 2 Apr 2013 17:43:41 +0530
- Subject: Re: [PATCH] Use mantissa_t in __mpexp
- References: <20130402115802 dot GB1410 at spoyarek dot pnq dot redhat dot com> <3665968 dot Zdk1WMNLRa at byrd>
On Tue, Apr 02, 2013 at 02:05:19PM +0200, Andreas Jaeger wrote:
> On Tuesday, April 02, 2013 17:28:02 Siddhesh Poyarekar wrote:
> > Hi,
> >
> > Here's a minor cleanup in mpexp to use mantissa_t instead of double.
> > I have also adjusted HALFRAD accordingly so that it is integral on
> > x86_64. Verified that this works fine on x86_64 and does not regress
> > the testsuite. OK to commit?
>
> I see usage of HALFRAD also in sincos32.c and mpsqrt.c. Will you change
> have any effect on these?
>
In both cases they're used for mantissa digits, so they will not have
any negative effect. In mpsqrt.c, the constant will be written in by
the link editor according to the type of mantissa_t. In sincos32.c,
there should be a minor performance advantage since the compiler will
not do the unnecessary conversion to double for the comparison.
Siddhesh