Firstly, my production code is CPU bound; the example above is not. My single threaded production code on my 4 core box runs at 100% of one core.

Really? Cos when I run a single process reading 10 x 4MB files each consisting of 16,000 lines on my 4-core system, it reads and counts all 40MB/160,000 lines in ~1/3 of a second:

c:\test>dir junk?.dat
 Volume in drive C has no label.
 Volume Serial Number is 8C78-4B42

 Directory of c:\test

09/07/2011  01:39         4,032,000 junk0.dat
09/07/2011  01:39         4,032,000 junk1.dat
09/07/2011  01:39         4,032,000 junk2.dat
09/07/2011  01:39         4,032,000 junk3.dat
09/07/2011  01:39         4,032,000 junk4.dat
09/07/2011  01:39         4,032,000 junk5.dat
09/07/2011  01:39         4,032,000 junk6.dat
09/07/2011  01:39         4,032,000 junk7.dat
09/07/2011  01:39         4,032,000 junk8.dat
09/07/2011  01:39         4,032,000 junk9.dat
              10 File(s)     40,320,000 bytes
               0 Dir(s)  205,414,293,504 bytes free

c:\test>head junk0.dat
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890
1234567890123456789012345678901234567890123456789012345678901234567890
+123456789012345678901234567890123456789012345678901234567890123456789
+012345678901234567890123456789012345678901234567890123456789012345678
+901234567890123456789012345678901234567890

c:\test>wc -l junk0.dat
  16000 junk0.dat

c:\test>perl -MTime::HiRes=time 
-e"BEGIN{ @ARGV = map glob, @ARGV; print time}" 
-nle"$n++ }{ print $n, ' ', time" junk?.dat

1310172599.189
160000 1310172599.43053
[download]

And the cpu usage is 2.7% for less than 1/2 a second. It requires more than that to start a single thread.

So, if you are going to make a case for threading this insignificant piece of code, you are going to have to show something far more convincing than you have so far. Seeing is believing.