Compilation directive, time performance optimisation

Hello

We have made a program with the IDE QNX Momentics 6.2.1. We use default
x86 compiler. But when you profile our codes, we saw that every maths
functions (like sqrtf() ) are very slowy (Our target is an pentuim
celeron 1.2GHz, sqrtf() execution time is estimate to 120ns ).
Is default x86 compilation use coprocessor mode ?
How to select gcc cpu type option ?
Have wee need a faster cpu to display 25 images (600400)and calculate
1600
25 square root by second ?

Any sugestions ?