[Rd] floating point control on windows
Chris Paulse
cpaulse at aecom.yu.edu
Mon Aug 29 18:23:29 CEST 2005
Thanks for the guidance.
Using the following code fixed my problem:
#ifdef _WIN32
BOOL APIENTRY DllMain( HANDLE hModule,
DWORD ul_reason_for_call,
LPVOID lpReserved
)
{
if (ul_reason_for_call == DLL_PROCESS_ATTACH)
{
Rwin_fpset();
}
return TRUE;
}
#endif
-----Original Message-----
From: Prof Brian Ripley [mailto:ripley at stats.ox.ac.uk]
Sent: Monday, August 29, 2005 10:53 AM
To: Chris Paulse
Cc: r-devel at r-project.org
Subject: Re: [Rd] floating point control on windows
We recommend calling Rwin_fpset() which is exported by R.dll and is known
to do the trick. As it contains
_controlfp(_PC_64, _MCW_PC);
it may be that the args you used are incorrect (and I am not on Windows
to check the docs, but note that is the form used in the URL you cite).
On Mon, 29 Aug 2005, Chris Paulse wrote:
> Hi,
>
> I'm sure that this question has come up many times before. When I load an
R
> extension dll I've built with the Microsoft compiler, I get the warning:
> Warning message:
>
> DLL attempted to change FPU control word from 8001f to 9001f
>
> So, what I did to try to fix the problem (to my knowledge, Microsoft will
> not support compile time switching of floating point control until version
> 8.0 of their compiler - see: http://blogs.msdn.com/ericflee) was to place
> the following statement inside R_init_mylib:
> _controlfp(_MCW_PC, _PC_64); // use extended precision for operations in
x87
> FPU
>
> This didn't fix my problem. I ended up sprinkling these liberally
> throughout my code, but that didn't help. One thing I do like about the
> Microsoft compiler is the debugging facility. But, I'll definitely switch
> to gcc if this is necessary to get rid of the problem.
--
Brian D. Ripley, ripley at stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
More information about the R-devel
mailing list