I was doing some calculations for my research and I needed more precision than afforded by the long float type
Using 'long double' on a x86 machine should give you access to 80 bit FP math (as supported by the x87 FPU instructions). Unfortunately x87 and SSE instructions are mutually exclusive (SSE instructions only support 64bit doubles) so there may be a performance hit (although 80bit x87math is still a heck of a lot faster than extended precision software implementations).