How to compile a double-precision version?

[jiandingzhe]

In dgTypes.h, there has following codes:

Code: Select all

#ifdef __USE_DOUBLE_PRECISION__
   typedef double dgFloat32;
#else
   typedef float dgFloat32;
#endif


However, if I define __USE_DOUBLE_PRECISION__, it will not compile, because there are polymorphisms related with dgFloat32 and dgFloat64 in other parts of Newton.

How can I get a double-precision version of Newton?

[Julio Jerez]

Update to SVN I just added the nessesary #ifndef __USE_DOUBLE_PRECISION__

[jiandingzhe]

Update to SVN I just added the nessesary #ifndef __USE_DOUBLE_PRECISION__

Not solved all niches.

I added -D__USE_DOUBLE_PRECISION__ to FLAGS of linux64 makefile. It still have something wrong with SIMD:

Code: Select all

../../source/physics/dgCollisionConvex.cpp: In member function ‘virtual dgVector dgCollisionConvex::SupportVertexSimd(const dgVector&) const’:
../../source/physics/dgCollisionConvex.cpp:919:2: error: ‘simd_type’ was not declared in this scope


After modified that, it compiles correctly.

Code: Select all

dgVector dgCollisionConvex::SupportVertexSimd (const dgVector& direction) const
{
#ifdef DG_BUILD_SIMD_CODE
    blah blah blah
#else
    return 0;
#endif


[jiandingzhe]

Update to SVN I just added the nessesary #ifndef __USE_DOUBLE_PRECISION__

Moreover, dMath, dScene, dCustomJoints, dContainers still requires a patch...

And why does dMath exists? There are dgVector, dgMatrix in the coreLibrary_200 which looks similiar.

[Julio Jerez]

but the dMath and teh rest librarys uses if __USE_DOUBLE_PRECISION__
if you just use that it soudl compiel fine.

I fixe the convexhull sind error.
thanks