And then there's:

$ time grep karl very_huge.file
nose cuke karl

real:  2.127s == user: 1.845s + sys: 0.283s   ## CLOSE ENOUGH!

$ time ./MCE-1.608/bin/mce_grep karl very_huge.file
nose cuke karl

real: 1.061s != user: 2.176s + sys: 1.616s    ## NOT EVEN CLOSE to: 3.
+792s

$ time ./MCE-1.608/examples/egrep.pl karl very_huge.file
nose cuke karl

real: 0.690s != user: 2.165s + sys: 0.362s    ## NOR IS THIS CLOSE to:
+ 2.527s
[download]

Looks dodgy to me.

And then the claim that:

The following code snippet parses the 2 GiB file in 1 second.

Let's examine that. The code: slurps the entire 2GiB file into memory and then starts 4 workers that each get a reference to (1/4 of???) the slurped data and then:

1. Run a regex against their chunk of the file to see if it contains the sought string;
2. And if it does, open their referenced chunk of the file as a memory file, and then re-run the regex on a line-by-line basis in order to get the whole lines that contain the sought string;
3. And then they return those lines to the parent process (via a pipe??) for it to print out.

So the regex is run against 1.25 times as much data as is contained in the file, and takes "less than one second"; which is less than the 2.127 seconds the real grep takes, despite that it only processes the file's data once?

Have you heard the song: Funny numbers?