Just a quick update. I went to this site

Yes, I've been there too - but haven't tried any of it out.
I gather these are libraries that you link to if you want enhanced DFP coverage. The "decNumber" package appears to be software only, while DFPAL can tap into available hardware capabilities, falling back to decNumber where necessary.

The mpfr home page links to decNumber, along with a link to what is presumably Intel's equivalent.
And there's links there to a couple of other decimal packages - MathCW and mpdecimal.
So there's plenty of additional chasing up to do for anyone so inclined :-)

BTW, the original wiki page that you linked to contains a nice little atodecimal implementation that works for both _Decimal64 and _Decimal128.
Unfortunately it won't evaluate 'inf', 'nan', or scientific notation inputs correctly and, by the time I add the perl code to accommodate that limitation, it probably won't be much faster than what I'm already doing.
But I think I might use it anyway as it's cleaner than what I've presently got.

Cheers
Rob