Hello, I using cross compiled GNU toolchain for ARM7TDMI-S target. Could anyone clarify why float words are always in big-endian format for ARM? Operations on double are fine until it is treated as whole value. But there are some bit field operations on double variables in my application which expects double words in little-endian format. This application is used on different platforms, So I didn't want to modify the structure of double for each platform. Just making #define FLOAT_WORDS_BIG_ENDIAN 0 in the file gcc-4.4.7/gcc/config/arm/arm.h & rebuilding the toolchain is giving me wrong double precision results. How to modify gcc to support multiple float words in little-endian format? Regards, Parimala