I've had BASE64 masm code on my homepage for some time, but it was just a transcription form C code, and thus just couldn't be really good. There were also some errors.
Finally I rewrote it from scratch. I post here both fasm and masm versions in attachment. There are 3 functions: base64_encode, base64_decode and base64_decode_fast. Two decoding functions, because the first one shares data table with encode func. and the other needs new table of 122 bytes - imho it's worth it, as the second function is significantly faster.
Enjoy!

EDIT:
Newer version in the attachment:
- fixed error when 0 was passed as size
- option to put newline after each n-th character (for example, for MIME encoding)
- only one decoding function (the faster one)
- ignores characters from outside the decoding table

base64 following which standard (there are more of them)?
supports padding?
what does it do if someone passes length=0?
supports breaking base64 data into multiple lines?
ignores all characters on base64_decode, as defined in RFC 2045?

Asnwers:
- supports padding (ie. pads with '=')
- length = 0
a) encode will work ok
b) decode will fail - I'll have to change it
- no breaking into multiple lines (may add it as an option - TRUE/FALSE flag will change behaviour)
- decode assumes all characters are ok - will have to change it

Thanks for that post vid.

glad to see you take it seriously. Think about how to report error from procedures (bad base64 data, zero length). If you make these procedures seriously working, i would like to include them to FASMLIB


rather add option for number of chars per line, no breaking if it's zero.

New version is in the first, edited post.
As for the error reporting:
- eax contains size of encoded/decoded data - so eax = 0 iff size = 0
- characters outside decoding table are not reported, they are just ignored.

I'd be happy if it shows in FASMLIB

not all characters should be ignored, only some of them...

"Additionally it specifies that any extra-alphabetic characters must be ignored by a compliant decoder" (quote from http://en.wikipedia.org/wiki/Base64)

It was about MIME encoding and generally that was what I wrote BASE64 for.

you are right. But i find it pretty stupid

But do not take wikipedia as reliable source, you must use the RFC next time

Vintage base64 encoding thread
http://board.flatassembler.net/topic.php?t=3063

The algorithm seems to run very fast on pentiums but it doesn't follow a standard implementation (input is null terminated string not sized buffer). But that can be easily changed.

I agree 100% with vid on this. WikiPedia is a great resource and can be used for some initial research, but you shouldn't use use it as an exclusive source of information, and you should always practice source criticism.

vid, f0dder, I agree with both of you. And I do not take Wikipedia as a omniscient source

r22, thanks for reminding me of this thread. I even joined discussion then and forgot of it