powerpc/xive: Use cpumask_intersects()
authorCosta Shulyupin <costa.shul@redhat.com>
Thu, 26 Sep 2024 09:26:22 +0000 (12:26 +0300)
committerMichael Ellerman <mpe@ellerman.id.au>
Thu, 14 Nov 2024 11:43:52 +0000 (22:43 +1100)
commit6da1cab4f5f8eb778fd61f0eb6ca5b0a011dd44d
treefad97c75f7f3b2e53f6a363b4b6eefd13aef38f4
parent7ca93aa9204b706e4afcd4fae0dc8798500598d5
powerpc/xive: Use cpumask_intersects()

Replace `cpumask_any_and(a, b) >= nr_cpu_ids`
with the more readable `!cpumask_intersects(a, b)`.

Comparison between cpumask_any_and() and cpumask_intersects()

The cpumask_any_and() function expands using FIND_FIRST_BIT(),
resulting in a loop that iterates through each bit of the bitmask:

for (idx = 0; idx * BITS_PER_LONG < sz; idx++) {
val = (FETCH);
if (val) {
sz = min(idx * BITS_PER_LONG + __ffs(MUNGE(val)), sz);
break;
}
}

The cpumask_intersects() function expands using __bitmap_intersects(),
resulting in that the first loop iterates through each long word of the bitmask,
and the second through each bit within a long word:

unsigned int k, lim = bits/BITS_PER_LONG;
for (k = 0; k < lim; ++k)
if (bitmap1[k] & bitmap2[k])
return true;

if (bits % BITS_PER_LONG)
if ((bitmap1[k] & bitmap2[k]) & BITMAP_LAST_WORD_MASK(bits))
return true;

Conclusion: cpumask_intersects() is at least as efficient as cpumask_any_and(),
if not more so, as it typically performs fewer loops and comparisons.

Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://patch.msgid.link/20240926092623.399577-2-costa.shul@redhat.com
arch/powerpc/sysdev/xive/common.c