[cpu minus startup time]     [sort] [sort]   Source Code CPU (sec) Mem (KB) Lines Code Log gcc 0.15 296 31 log ocaml 0.26 472 22 log bash 0.39 1348 1 log mlton 0.41 384 56 log cmucl 0.56 4168 32 log mawk 0.59 528 6 log gawk 0.63 796 6 log bigforth 0.74 832 34 log g++ 0.77 676 31 log smlnj 1.20 3480 60 log perl 1.26 1224 9 log java 1.94 7532 33 log python 1.98 1300 13 log bigloo 2.00 676 22 log se 2.19 388 40 log ruby 2.36 6528 9 log stalin 2.47 55064 13 log pike 3.63 2396 27 log icon 4.13 1996 14 log lua 4.37 584 13 log ocamlb 4.74 628 22 log gforth 5.42 688 34 log ghc 8.44 1640 26 log php 9.14 1964 13 log tcl 9.21 1108 18 log erlang 10.04 135252 19 log njs 15.59 2800 27 log guile 50.66 1448 18 log Languages that compile to native code are in Bold Italics.

[Note: Values have been normalized to fall in the range of 0-10 for aesthetic reasons. Original value ranges are included on the X-axis. Click here for more detailed data and graphs. [Results last updated: Tue Oct 9 18:40:32 2001 CDT]

About this test

For this test, each program should be implemented to do the same thing, following the guidelines below:

Each program reads the input from standard input, and counts the lines, words (whitespace delimited tokens), and characters, and outputs each count. The programs should not read the input by more than 4K at a time. To give a baseline of expected performance I allow bash to use an external process (wc). All other solutions should be implemented natively.

This test is essentially the same as the wordcount test from Timing Trials, or, the Trials of Timing: Experiments with Scripting and User-Interface Languages by Brian W. Kernighan and Christopher J. Van Wyk.

Note that as in the original version of this test, whitespace is defined as space, newline and tab characters. This is a little different from the Unix wc command, which defines a few more characters to be whitespace.

The programs can assume that the file ends in a newline, and they should be able to handle arbitrarily long lines.

Input file (it is repeated N times).

The correct output (for N = 500, i.e. a 500 copies of the input) looks like this:

  12500 68500 3048000

Observations

The original C program is significantly slower than the version here which bypasses stdio.

Alternates

This section is for displaying alternate solutions that are either slower than ones above or perhaps don't quite meet my criteria for the competition, but are otherwise worthy of comment.