mirror of
https://github.com/torvalds/linux.git
synced 2024-12-13 22:53:20 +00:00
fdb6649ab7
If x is not 0, __ffs(x) is equivalent to:
(unsigned long)__builtin_ctzl(x)
And if x is not ~0UL, ffz(x) is equivalent to:
(unsigned long)__builtin_ctzl(~x)
Because __builting_ctzl() returns an int, a cast to (unsigned long) is
necessary to avoid potential warnings on implicit casts.
Concerning the edge cases, __builtin_ctzl(0) is always undefined,
whereas __ffs(0) and ffz(~0UL) may or may not be defined, depending on
the processor. Regardless, for both functions, developers are asked to
check against 0 or ~0UL so replacing __ffs() or ffz() by
__builting_ctzl() is safe.
For x86_64, the current __ffs() and ffz() implementations do not
produce optimized code when called with a constant expression. On the
contrary, the __builtin_ctzl() folds into a single instruction.
However, for non constant expressions, the __ffs() and ffz() asm
versions of the kernel remains slightly better than the code produced
by GCC (it produces a useless instruction to clear eax).
Use __builtin_constant_p() to select between the kernel's
__ffs()/ffz() and the __builtin_ctzl() depending on whether the
argument is constant or not.
** Statistics **
On a allyesconfig, before...:
$ objdump -d vmlinux.o | grep tzcnt | wc -l
3607
...and after:
$ objdump -d vmlinux.o | grep tzcnt | wc -l
2600
So, roughly 27.9% of the calls to either __ffs() or ffz() were using
constant expressions and could be optimized out.
(tests done on linux v5.18-rc5 x86_64 using GCC 11.2.1)
Note: on x86_64, the BSF instruction produces TZCNT when used with the
REP prefix (which explain the use of `grep tzcnt' instead of `grep bsf'
in above benchmark). c.f. [1]
[1]
|
||
|---|---|---|
| .. | ||
| asm | ||
| uapi/asm | ||