large(ish) BCD to binary

[kripacharya]

This is the opposite prob from what most folks want.

I need to convert a BCD number -- 6 digit -- into its binary/ hex equivalent. The BCD can be packed/ not packed. I prefer not-packed for now.

Can someone provide pointers to an 8-bit assembly implementation for an efficient algorithm ? Its for an AVR project, using Atmega48 as of now. Fair amount of RAM is available if required.

Once again -- this is BCD to Binary.

 

[FvM]

Essentially multiply digits by ten and add next digit. Repeat from left to right. Don't need other RAM than 16-Bit result variable.

 

[kripacharya]

Essentially multiply digits by ten and add next digit. Repeat from left to right. Don't need other RAM than 16-Bit result variable.

yes of course !! i was clearly too sleepy to see the obvious.
But i will need at least 20-bits for the result.

thanks & cheers!

 

[FvM]

But i will need at least 20-bits for the result.

Yes, of course. Unfortunately the least significant digit can overflow to any result bit, so you have to implement a full 24 Bit (or 32, when using standard number formats) adder.

Any reasons against utilizing a standard C compiler to implement the calculations?

 

[kripacharya]

Oh ? How would it overflow ? Max for 6-digit is 999,999 which fits in 20 bits. Anyhow I have to use 3x 8-bit regs so no troubles there.
This project is very tight on time/ real-time, and its somehow just easier to use assembly, where i can SEE everything. But other than that just force of habit.
Though i admit I sometimes copy bits of code output by gcc with optimisation set ;-)

 

[FvM]

Sorry for misunderstanding, I didn't mean overflow, just regular carry. In any case, it's normal integer math, nothing special.