Re: [PATCH] get_order() centralization

[Jamie Lokier]

Jakub Jelinek wrote:
> I removed the bsrl implementation for Intel since the last patch because it
> really seems it is slower than a simple loop.

Besides, the timing is more important for small regions than large ones,
and that's how the loop is biased.  If you're allocating a large region
you're likely to be doing a lot with it, so a little overhead in a loop
doesn't matter.

But seeing as we're here, did you try this code?

  static const unsigned char table [32] =
    { 0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 
      5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5 };

  size = (size - 1) >> PAGE_SHIFT; /* Let's assume it's 12. */
  if (size >= (1 << 10)) { size >>= 10; order += 10; }
  if (size >= (1 <<  5)) { size >>= 5 ; order += 5; }
  order += table [size];

enjoy,
-- Jamie