Stripped down reference C source size less headers, debug/intermediate print-outs and some large superficial comments. Reference source size is of course approximate and some of the submissions are not portable, but it reflects more or less the complexity of the reference implementation of each algorithm that people will have to deal with while trying to understand and/or implement the algorithm. Some of the submissions have extensive comments, some don't. Nevertheless, if the reference source is that complicated that it requires tons of additional comments, the weight of all those comments must count as well.
cubehash ¦ 3480
streamhash ¦ 6866
enrupt ¦ 7246
mcssha3 ¦ 7623
dynamicsha ¦ 8441
jh ¦ 9530
dynamicsha_2 ¦ 10854
lux ¦ 10968
shabal ¦ 11306
keccak ¦ 11498
meshhash ¦ 11964
waterfall ¦ 12752
fugue ¦ 13872
cheetah ¦ 14891
dch ¦ 16029
echo ¦ 16940
tangle ¦ 17122
luffa ¦ 17196
groestl ¦ 17308
boole ¦ 17343
sha2* ¦ 17784
shamata ¦ 17855
abacus ¦ 17868
khichidi-1 ¦ 21447
sarmal ¦ 22659
ponic* ¦ 22849
blake ¦ 24389
simd ¦ 25442
arirang ¦ 25833
spectral_hash ¦ 26142
twister ¦ 26439
edon-r ¦ 26855
chi ¦ 27263
aurora ¦ 27266
bmw ¦ 27446
lesamnta ¦ 29538
tib3 ¦ 33594
nasha ¦ 34315
lane ¦ 41907
sandstorm ¦ 45245
shavite-3 ¦ 54282
swifftx ¦ 63355
md6 ¦ 69665
vortex ¦ 71575
blender ¦ 81146
skein ¦ 84089
essence ¦ 90513
sgail ¦ 424579
wamm ¦ 462864
ecoh ¦ 479490
hamsi ¦ 943575
fsb ¦ 1467 K
crunch ¦ 3119 K
* - not a part of the SHA-3 competition, included for comparison.