Re: [RFC PATCH 2/2] sched: Improve cache locality of RSEQ concurrency IDs for intermittent workloads

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

On 2024-09-04 14:28, Mathieu Desnoyers wrote:
> On 2024-09-04 11:50, Mathieu Desnoyers wrote:
>> On 2024-09-04 11:24, Yury Norov wrote:
> [...]
>>>
>>> This all doesn't look like a hot path. And anyways, speculating around
>>> performance without numbers on hands sounds cheap.
>>
>> This is done whenever userspace invokes sched_setaffinity, or changes
>> its cgroup cpuset. It may not be the most important fast-path in the
>> world, but I expect some workloads to issue sched_setaffinity whenever
>> they create a thread, so it's not a purely slow-path either.
>>
>>> In my experience, iterators with a very lightweight payload are ~100
>>> times slower comparing to dedicated bitmap ops. Check this for example:
>>> 3cea8d4753277.
>>>
>>> If you're really cared about performance here, I'd suggest you to
>>> compare your iterators approach with something like this:
>>>
>>>    cpumask_or(mm_allowed, mm_allowed, cpumask);
>>>    atomic_set(&mm->nr_cpus_allowed, cpumask_weight(mm_allowed);
> 
> Here are the benchmark results. Each test use two entirely filled
> bitmaps as input to mimic the common scenario for cpus allowed
> being updated with a subset of the original process CPUs allowed,
> and also the common case where the initial cpumask is filled.
> 
> #define BITMAP_LEN      (4096UL * 8 * 10)
> (len = BITMAP_LEN)
> 
> * Approach 1:
> 
>         int nr_set = 0;
>         for_each_andnot_bit(bit, bitmap, bitmap2, len)
>                 nr_set += !test_and_set_bit(bit, bitmap2);
>         if (nr_set)
>                 atomic_add(nr_set, &total);
> 
> Time: 4680 ns
> 
> * Approach 2:
> 
>         int nr_set = 0;
>         for_each_set_bit(bit, bitmap, len)
>                 nr_set += !test_and_set_bit(bit, bitmap2);
>         if (nr_set)
>                 atomic_add(nr_set, &total);
> 
> Time: 1791537 ns
> 
> * Approach 3:
> 
>         mutex_lock(&lock);
>         bitmap_or(bitmap2, bitmap, bitmap2, len);
>         atomic_set(&total, bitmap_weight(bitmap2, len));
>         mutex_unlock(&lock);
> 
> Time: 79591 ns

The benchmark result is wrong for approach 3, as it was taken with
CONFIG_PROVE_LOCKING=y and lockdep.

Corrected result:

Time: 4500 ns.

So let's go with your approach. I'm wondering whether I should
re-use an existing mutex/spinlock from mm_struct or add a new one.

Thanks,

Mathieu

> 
> The test hardware is a AMD EPYC 9654 96-Core Processor.
> 
> Thanks,
> 
> Mathieu
> 

-- 
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com