Following on from a discussion on IRC, then in email with khr, maps could do a better
job when hashing small keys. 

What steps will reproduce the problem?
1. go get github.com/davecheney/bfk
2. % go test -bench=. 

What is the expected output? What do you see instead?

On my linux/amd64 SNB core i5 I get something like this

BenchmarkArray  20000000                89.5 ns/op
BenchmarkSwitch 10000000               189 ns/op
BenchmarkMap      500000              4518 ns/op

Clearly for a this sort of problem, using an array if func pointers is the best
solution, but it does raise the question if performance for the map variant could be
improved. Looking at the profile

lucky(~/src/github.com/davecheney/bfk) % go tool pprof bfk.test cpu.out 
Welcome to pprof!  For help, type 'help'.
(pprof) top
Total: 232 samples
      84  36.2%  36.2%      145  62.5% hash_lookup
      58  25.0%  61.2%       58  25.0% runtime.aeshashbody
      38  16.4%  77.6%      232 100.0% github.com/davecheney/bfk.BenchmarkMap
      27  11.6%  89.2%      194  83.6% runtime.mapaccess1
      16   6.9%  96.1%       16   6.9% runtime.memmove
       6   2.6%  98.7%       22   9.5% runtime.memcopy
       2   0.9%  99.6%        2   0.9% runtime.memequal8
       1   0.4% 100.0%        1   0.4% runtime.aeshash
       0   0.0% 100.0%      232 100.0% runtime.gosched0
       0   0.0% 100.0%      232 100.0% testing.(*B).launch

aeshash{,body} is called when the key is a byte and the body is two words. khr thought
that this could be improved.

Please use labels and text to provide additional information.