NDEBUG, affects floating point math?

From: Stephen Crowley (stephenc_at_[hidden])
Date: 2003-02-28 00:55:35

I'm getting some seriously strange problems with regard to numerical
accuracy when computing matrix products that depends on whether or not
NDEBUG is defined when I compile.

I wrote a small piece of C code to calculate a symmetric-matrix by
row-matrix product by directly accessing the memory (which btw is about
1000% faster than ublas's generated code with gcc 3.2.3 and -O3, why is
ublas so slow?) And when I compile with -DNDEBUG the results differ 
by around +/- 1e-16, but when its not defined the results match almost
perfectly. Is there something strange going on with regard to FPU control
words or types or anything? I'm using type double btw. If more info is
needed I can post a test case, but I thought it'd make a quick inquery to
see if there is a fast answer.

-- 
Stephen
--------
This message was scanned for viruses by Digital Passage.