__aeabi_dadd/__aeabi_dsub sizes for nRF51 vs STM32F1

[gfwilliams]

Just noticed this...

• STM32F1 - __aeabi_dadd = ~600 bytes
• nRF51 - __aeabi_dadd = ~1700 bytes

I guess it could be Cortex M0 vs M3 - but that seems like a hell of a difference

[gfwilliams]

__aeabi_dadd is also ~600 bytes on the nRF52832 - so this seems like a Cortex M0 issue :(

[gfwilliams]

Still 630 on nRF528540, 1772 bytes on nRF51.

We did manage to make a new __aeabi_dsub which just flipped the sign and called __aeabi_add so that halves the memory use.

@fanoush any thoughts? If it were fixable it'd be an easy way to gain 1k of space on nRF51. But maybe there's just no nice way to do this efficiently on the M0...