Idempotent partition around median of 5?

[tn via Digitalmars-d]

On Saturday, 6 February 2016 at 01:46:58 UTC, Andrei Alexandrescu 
wrote:
> On 02/04/2016 03:30 PM, Timon Gehr wrote:
>> At most 6 comparisons, <=3 swaps, idempotent (optimal number 
>> of swaps):
>
> Oh, also: could you let that bad boy run and let it find 
> anything that does idempotent partition in 6 compares and 0-7 
> swaps? Then slowly tighten the number of swaps until you find 
> the best balance between number of swaps and code size. -- 
> Andrei

That is kind of what I tried to do by hand. My function 
partition5e seems to be practically identical to Ivans solution 
and partition5a is just a copy of Timons solution. Function 
partition5b, partition5c and partition5d are in between of those 
with various tradeoffs between the number of swaps and code size.