On Wed, 5 Jul 2023 19:12:45 +0100
Valentin Schneider <vschneid@xxxxxxxxxx> wrote:
> +/* Optimisation of do_filter_cpumask() for scalar values */
> +static inline int
> +do_filter_cpumask_scalar(int op, unsigned int cpu, const struct cpumask *mask)
> +{
> + switch (op) {
> + case OP_EQ:
> + return cpumask_equal(mask, cpumask_of(cpu));
> + case OP_NE:
> + return !cpumask_equal(mask, cpumask_of(cpu));
The above two can be optimized much more if the cpu is a scalar. If mask is
anything but a single CPU, then EQ will always be false, and NE will always
be true. If the mask contains a single CPU, then we should convert the mask
into the scalar CPU and just do:
case OP_EQ:
return mask_cpu == cpu;
case op_NE:
return maks_cpu != cpu;
-- Steve
> + case OP_BAND:
> + return cpumask_test_cpu(cpu, mask);
> + default:
> + return 0;
> + }
> +}
