From: Brian P. <br...@tu...> - 2002-01-16 14:49:23
|
Jacob =Jouk Jansen wrote: > > br...@tu... wrote on 15-JAN-2002 19:17:53.17 > >jo...@hr...m.tudelft wrote: > >> What I see from the code is that division by zero is allowed by the > >> programmer. This means that on OpenVMS the division operator should be over > >> loaded by something like > >> > >> if ( g == 0.0 ) > >> if ( h > 0.0 ) > >> r = __MAXFLOAT; > >> else > >> r = -__MAXFLOAT; > >> else > >> r = h / g; > >> > >> > >> What is the reason to put all this infinite stuff into the routines? > > > >That's what the NV_vertex_program spec calls for. Part of the agreement > >to allow NV_vertex_program into Mesa was that it would be conformant > >with the specification. > > > >It looks like I'll need to use a few #ifdef VMS / #else / #endif > >conditionals to approximate the NV_vertex_program behaviour on VMS. > > The trunk CVS compiles again and all my own applications seem to run fine. > (no special vertex either). I'll try to run some of the vertex demo programs > later and do some of the VMS tweaking where needed (if I find some time) OK, good. Though none of the vertex program tests that I've checked in use the EXP or LOG functions. -Brian |