This code ...
double d = 1e300;
double f = 1e300;
double r = d+f;
vcl_cout << "d " << d << vcl_endl;
vcl_cout << "f " << f << vcl_endl;
vcl_cout << "r " << r << vcl_endl;
vcl_cout << vcl_flush;
gives this output ...
1e300 is too big for a float right? So I guess my little theory was not quite right.
> -----Original Message-----
> From: Amitha Perera [mailto:perera@...]
> Sent: Thursday, February 20, 2003 9:53 AM
> To: Wheeler, Fred (Research)
> Cc: VXL Maintainers List (E-mail)
> Subject: Re: [Vxl-maintainers] RE: [Vxl-commit] vxl/core/vil2/tests
> test_sample_profile_bilin.cx x,1.9,1.10
> On Thu 20 Feb 2003, Wheeler, Fred (Research) wrote:
> > So it looks like it has nice big doubles. However, I see on some
> > random computational chemistry web site that Google sent me to that
> > the PS2 Emotion Engine (its CPU) has hardware support for 32 bit
> > floating point values only. Looks like the OS and compiler support
> > 8 byte doubles, but must down convert for any operations on them.
> Interesting. Does that mean
> double d = 1e300;
> double f = 1e300;
> double r = d+f;
> results in r==Inf?