Take a look at the ARM crypto lib (GPL license) or implement it directly from the RFC here. It is quite a simple procedure, more about re-grouping bits than anything else.
When encoding to base64, the output size will increase by a factor of about 4/3. This is because every 3 bytes are encoded into 4 bytes, and finally padded to always be a multiple of 4 bytes. So the destination array needs to be larger. A more precise size than 4/3 is (src_length + 2 - ((src_length + 2) % 3)) / 3 * 4. You should then get that your string encodes into "ADUAAAAAAAAAADU=", which is reversed properly back to your start array.
When encoding to base64, the output size will increase by a factor of about 4/3. This is because every 3 bytes are encoded into 4 bytes, and finally padded to always be a multiple of 4 bytes. So the destination array needs to be larger. A more precise size than 4/3 is (src_length + 2 - ((src_length + 2) % 3)) / 3 * 4. You should then get that your string encodes into "ADUAAAAAAAAAADU=", which is reversed properly back to your start array.
