Re: Population count in SSE2, again

"James Van Buskirk"  wrote in message
 (code)

Someone tell me why 21 instructions and a branch, per word, is somehow
better/faster than

Initializtion all methods need.

  zero summing index   (eg ax or eax)
  load address of string to esi
  load string count to ecx
  load table address to ebx
back:                              ; 3 instructions per word follow
  add to current sum, from count table indexed by the 8 byte or 16 bit
word (one instruction)
  increment string pointer
(esi)                                            (one instruction)
  branch to repeat if not finished string byte/word string length
(dec-branch instruction)

?
I want to know, if I haven't understood what you all seem to be trying
to do!
Is it that all those instuctions are all somehow loaded to the cpu
chip inststruction store and processed on chip? Because there is
always the memory access to be considered.