RE: [PATCH 7/8] lib: add fast path for find_next_*_bit()

From: David Laight
Date: Mon Feb 01 2021 - 11:05:29 EST


From: Andy Shevchenko
> Sent: 01 February 2021 13:49
>
> On Sat, Jan 30, 2021 at 11:17:18AM -0800, Yury Norov wrote:
> > Similarly to bitmap functions, find_next_*_bit() users will benefit
> > if we'll handle a case of bitmaps that fit into a single word. In the
> > very best case, the compiler may replace a function call with a
> > single ffs or ffz instruction.
>
> Would be nice to have the examples how it reduces the actual code size (based
> on the existing code in kernel, especially in widely used frameworks /
> subsystems, like PCI).

I bet it makes the kernel bigger but very slightly faster.
But the fact that the wrappers end up in the i-cache may
mean that inlining actually makes it slower for some calling
sequences.

If a bitmap fits in a single word (as a compile-time constant)
then you should (probably) be using different functions if
you care about performance.

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)