Re: [ros-dev] ping Alex regarding log2() for scheduler

[Royce Mitchell III]

Mark Junker wrote:
> Royce Mitchell III schrieb:
> > I managed to shave off another comparison in favor of a shift:
>
> What about using a 64k table?
>
> static
> BYTE HighestBits[65536] = {
> 0x00, 0x01, 0x02, 0x02, ... // and so on
> };
>
> // returns 0xFF when no bits set in dwValue
> BYTE highest_bit(DWORD dwValue)
> {
>  BYTE ret;
>  if (i > 0xffff) {
>    ret = HighestBits[(int) (dwValue >> 16)];
>  } else {
>    ret = HighestBits[(int) dwValue];
>  }
>  return ret - 1;
> }
>
> Regards,
> Mark

There's a couple bugs in your code, but...

The table lookups are a little bit faster when they take full advantage 
of available cpu caching. I'm not sure that they are good enough to 
continue to fare well when put into our target environment ( at least 
once per interrupt/thread schedule ). Besides, it's wasting a lot of 
npool ram for a slight performance improvement.