[gmx-users] Installing GROMACS on Windows

[Konrad Piwowarczyk]

> On Mon, Apr 26, 2004 at 08:39:24PM +0200, Konrad Piwowarczyk wrote:
> > > ----- Original Message ----- 
> > > From: "Nuno R. L. Ferreira" <nunolf at ci.uc.pt>
> > > To: <gmx-users at gromacs.org>
> > > Sent: Saturday, April 24, 2004 8:18 PM
> > > Subject: Re: [gmx-users] Installing GROMACS on Windows
> > > 
> > > 
> > > > Hi Michael.
> > > >
> > > > In all my history of program compilations, I never knew what is
> > > compiling
> > > > under dozze environ. ;)
> > > > Couldn't you contribute the gmx3.2.1 compiled to the contributions
> > > page?
> > > I
> > > > really do not know how to compile under win.
> > > 
> > > No work, no pay!
> > > Now I know what cgwin is. Gmx 3.2.1 ready and runing on a WXP box.
> > > Sorry about that.
> > > 
> > > Nuno
> > > 
> > 
> > I am thinking about contributing windows binaries since windows package
> with gromacs 3.1.2 at gromacs site is outdated. If I have some time...
> 
> compiled by which compiller? cause I compared gromacs 3.1.4 under win32
> compiled by MS Visual C++ 6.0 compiler and Intel C++ compiler 8.01
> without
> asm loops (optimized with no special options) and on AMD Athlon processor
> was intel 19% better). I wasn't compare gcc yet :(
> 
> Milan

I suppose so! Once I compared a very simple md code compiled by bcc32 5.5, gcc 2.95 and VC++ 7.0 (they are ordered by the efficiency of generated code). VC++ was really pure especially compared to Borland which is excellent product IMHO - its binaries needed 2/3 time of VC++ binaries time (if I correctly remember) and gcc was always a bit better than VC++. 
What more if I forced VC++ to make optimizations it produced even slower binaries - I am still not sure why.
So if I manage to compile gromacs with Borland or even better Watcom compiler I will post results :)

I suppose that for code with assembly loops enabled compilers will affect less the performance of code - am I right? 

have a nice evening
Konrad