(We are going to blog more often, catch us at http://imagecraft.com/blog https://www.facebook.com/imagecraft.inc and Twitter @imagecraftinc)
One common concession to C Standard on typical 8-bit embedded compilers is that double data type is 32 bits and not 64 bits as required by the standard. Since 8-bit CPU microcontrollers are usually limited by both memory size and clock speed, this is an acceptable tradeoff. However, some of these 8-bit micros are growing in size and speed, driven by the complexity of the firmware.
We added 64-bit double support a couple years back, and now we are adding full printf (e.g. %f, %g, %e”) support on the AVR compiler as well. As usual, this is optional and we even have different versions of the underlying conversion to string function depending on whether you want full range/precision support or faster execution.