Date: Thu, 9 Oct 2014 15:34:25 -0700 From: Steve Kargl <sgk@troutmask.apl.washington.edu> To: Walter <dev.for@saveouraquifer.org> Cc: freebsd-fortran@freebsd.org Subject: Re: 128-bit floats Message-ID: <20141009223425.GA67845@troutmask.apl.washington.edu> In-Reply-To: <5437010E.7080200@saveouraquifer.org> References: <201410060832.s968WCLm095448@mech-as221.men.bris.ac.uk> <54331392.9040808@saveouraquifer.org> <20141006233011.GA14956@troutmask.apl.washington.edu> <5437010E.7080200@saveouraquifer.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Oct 09, 2014 at 05:41:34PM -0400, Walter wrote: > On 10/6/14, 7:30 PM, Steve Kargl wrote: > > > > 128-bit floating point should work without a hitch (knock on wood). > > Just install lang/gcc. Look for /usr/local/bin/gfortranXX where XX > > is 47 or 48 and off you go. > > Thanks, Steve. After the installs, and resolving the "ld: library not > found for -lc" error, I got the code to compile. After I get a data > set together, I'll know if it runs. Odd. gfortran should be able to find libc. I would need to see how you are building things. > Maybe this is a bit premature, but the code might be more robust (I'm > not the developer/mathematician) using 256-bit floats. Are such things > available? Not as a implicit type. Asumming a i386 or x86_64 class cpu, gfortran will give you real(4), real(8), real(10), and real(16). These are 32, 64, 80, and 128 bit floating point types. You you need more precision, then netlib is your friend. http://www.netlib.org/mpfun/index.html -- Steve
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20141009223425.GA67845>