I am trying to currently focus on tie-breaking (of winners) explained in section 5 of the original Schulze paper and this seems more complicated than the basics of the method.

Regarding your post, I am having two questions:

1) Have you thought of implementing “margins” as explained in http://m-schulze.9mail.de/schulze1.pdf

2) Have you thought of possible caveats of sequential dropping (your while loop, as I understand it). Basically, given N candidates, we calculate (using Condorcet and/or Schulze scheme) a winner. Then we drop the winner and compute the next winner from N-1 candidates. We proceed till there is one member left.

I have tried to find clear indications of this procedure in the 1st Schulze paper but couldn’t. Would you have any comment on this?

Thanks.

Alfred

]]>gcc -std=gnu99 -I../include -I. -I../extra -DHAVE_CONFIG_H -DR_DLL_BUILD -c -o e_pow.o e_pow.S

e_pow.S: Assembler messages:

e_pow.S:110: Error: invalid instruction suffix for `pop'`

pop’

e_pow.S:111: Error: invalid instruction suffix for

e_pow.S:227: Error: invalid instruction suffix for `pop'`

pop’

e_pow.S:228: Error: invalid instruction suffix for

e_pow.S:278: Error: invalid instruction suffix for `pop'`

pop’

e_pow.S:279: Error: invalid instruction suffix for

e_pow.S:313: Error: invalid instruction suffix for `pop'`

pop’

e_pow.S:314: Error: invalid instruction suffix for

make[3]: *** [e_pow.o] Error 1

make[2]: *** [../../bin/i386/R.dll] Error 2

make[1]: *** [rbuild] Error 2

make: *** [all] Error 2

Any suggestion?

Thanks.

compared with original R, my Blas is about 10 times faster. NotBlas functions differs, A/B is like 3 times faster, simple calc is around 5% to 10% or no significant improvement. Overall, well done!

have you compiled for 32bit as well? thanks.

]]>