What does Cudder think of

This? http://dec64.com/

Here's the worst part:

https://github.com/douglascrockford/DEC64/blob/master/dec64.asm

When we decrease the exponent by 1, we must also multiply the coefficient by 10.

...adding requires binary multiplication in a fucking loop! Regular binary floating point normalisation can be done with a comparator and a barrel shifter. Ditto for BCD (the "real" decimal floating point format), where the shift is in 4-bit digit increments.

With this format integer addition is slower than regular integer addition, and floating-point addition is a lot slower than regular floating-point addition. WTF?