[Insight-developers] Re: More on Adaptors vs Filter precision

[Luis Ibanez]

Bill,

It seems that the (double) casts in the math unary Filters
are really needed to avoid the SGI's warnings.

The epsilon of 1e-10 for the test is really not necesary,
it is now set to 1e-6, so probably we can put the (double)
casting back, and have them working on all the platforms.

Should I make this modifications ?


Luis


--------------------

On Wed, 4 Apr 2001, Lorensen, William E (CRD) wrote:

> Luis,
> I believe that I am the culprit on the adaptor/filter precision problems. A couple of weeks ago, I
> added an explict cast to double on all the image filters that used arthmetic. I thought this was an
> sgi'ism. Turns out it was an sgi loader problem that requires -lm after each --lVXLNumerics.
>
> I will remove the cast to double (since I broke them). This will fix, I believe, the precision
> "problem".
>
> Bill
>
>