Base16 Speed

Base16 conversion from printable characters to binary data, and back again, is important for one project I work on. Doing conversion as fast as possible is a good thing.

I had some code that did the binary to printable characters conversion well enough, using sprintf(), but I thought it wasn’t optimal, so I decided to find out.

I wrote some quick and dirty code, in my github, to test the performance difference between using sprintf() and using a lookup table. Turns out, the LUT is about 10 times faster on my Core-i7 and about 14 times faster on a BeagleBone. That’s quite an improvement!

I’m sure there’s an even faster way to do other things my code does less than optimally, now I’m on the hunt!

BeagleBone run:

root@localhost:~# ./test test.file 
Read 14659584 bytes
sprintf bin to hex took 20850000 clocks
bintohlut bin to hex took 1470000 clocks
htob hex to bin took 10860000 clocks

Intel Core-i7 run:

andrew@brick:~/git/base16test$ ./test /dev/urandom 
Read 33554432 bytes
sprintf bin to hex took 2750000 clocks
bintohlut bin to hex took 260000 clocks
htob hex to bin took 1080000 clocks

Published