GMP, Windows, Visual Studio
mathimagics at yahoo.co.uk
Sun Jul 8 20:26:51 CEST 2007
> Date: Sat, 7 Jul 2007 09:55:10 -0500
> From: "Jeremy Rann" <xerog1 at gmail.com>
> ... I think for now I'll just use the C library and
> go from there.
> I've been thinking about benchmarking the various
> different ways of using the library - static
> vs. dynamic, using all mingw vs Visual Studio vs
> Linux, etc.
If by benchmarking, you mean performance of a client
application doing gmp arithmetic, then it should make
no difference how you build the client (other than the
varying degrees of resistance from the different
client development environments!).
On Win32 platforms there should be no discernible
difference in the execution speed of code in a static
library or in a dll - other than in very extreme, and
highly improbable, circumstances (eg, running 1000
simulataneous copies of a statically-linked client
MinGW is not only the easiest way to build GMP dll's,
it's gcc compiler/linker are stand-alone, you can run
them from a CMD (command) window, use BAT files to
create build scripts, etc). This does assume you
don't already prefer Unix-style "make"'s, of course.
That reminds me, too, of one specific advantage of a
DLL over a static library - a DLL can called by
virtually any client language.
Have fun, whatever you do ...
More information about the gmp-discuss