From: Muchun Song <muchun.song@linux.dev>
To: Gang Li <gang.li@linux.dev>
Cc: David Hildenbrand <david@redhat.com>,
David Rientjes <rientjes@google.com>,
Mike Kravetz <mike.kravetz@oracle.com>,
Andrew Morton <akpm@linux-foundation.org>,
Tim Chen <tim.c.chen@linux.intel.com>,
Linux-MM <linux-mm@kvack.org>,
LKML <linux-kernel@vger.kernel.org>,
ligang.bdlg@bytedance.com
Subject: Re: [PATCH v4 6/7] hugetlb: parallelize 2M hugetlb allocation and initialization
Date: Mon, 22 Jan 2024 19:30:39 +0800 [thread overview]
Message-ID: <849D7EA4-BCF4-4587-8A78-F3B35B63EAE9@linux.dev> (raw)
In-Reply-To: <14e38e95-2bc6-4571-b502-4e3954b4bcc4@linux.dev>
> On Jan 22, 2024, at 18:12, Gang Li <gang.li@linux.dev> wrote:
>
> On 2024/1/22 15:10, Muchun Song wrote:> On 2024/1/18 20:39, Gang Li wrote:
>>> +static void __init hugetlb_alloc_node(unsigned long start, unsigned long end, void *arg)
>>> {
>>> - unsigned long i;
>>> + struct hstate *h = (struct hstate *)arg;
>>> + int i, num = end - start;
>>> + nodemask_t node_alloc_noretry;
>>> + unsigned long flags;
>>> + int next_node = 0;
>> This should be first_online_node which may be not zero.
>
> That's right. Thanks!
>
>>> - for (i = 0; i < h->max_huge_pages; ++i) {
>>> - if (!alloc_bootmem_huge_page(h, NUMA_NO_NODE))
>>> + /* Bit mask controlling how hard we retry per-node allocations.*/
>>> + nodes_clear(node_alloc_noretry);
>>> +
>>> + for (i = 0; i < num; ++i) {
>>> + struct folio *folio = alloc_pool_huge_folio(h, &node_states[N_MEMORY],
>>> + &node_alloc_noretry, &next_node);
>>> + if (!folio)
>>> break;
>>> + spin_lock_irqsave(&hugetlb_lock, flags);
>> > I suspect there will more contention on this lock when parallelizing.
>
> In the worst case, there are only 'numa node number' of threads in
> contention. And in my testing, it doesn't degrade performance, but
> rather improves performance due to the reduced granularity.
So, the performance does not change if you move the lock out of
loop?
>
>> I want to know why you chose to drop prep_and_add_allocated_folios()
>> call in the original hugetlb_pages_alloc_boot()?
>
> Splitting him to parallelize hugetlb_vmemmap_optimize_folios.
Unfortunately, HVO should be enabled before pages go to the pool list.
>
>>> +static unsigned long __init hugetlb_pages_alloc_boot(struct hstate *h)
>>> +{
>>> + struct padata_mt_job job = {
>>> + .fn_arg = h,
>>> + .align = 1,
>>> + .numa_aware = true
>>> + };
>>> +
>>> + job.thread_fn = hugetlb_alloc_node;
>>> + job.start = 0;
>>> + job.size = h->max_huge_pages;
>>> + job.min_chunk = h->max_huge_pages / num_node_state(N_MEMORY) / 2;
>>> + job.max_threads = num_node_state(N_MEMORY) * 2;
>> I am curious the magic number of 2 used in assignments of ->min_chunk
>> and ->max_threads, does it from your experiment? I thinke it should
>> be a comment here.
>
> This is tested and I can perform more detailed tests and provide data.
>
>> And I am also sceptical about the optimization for a small amount of
>> allocation of hugepages. Given 4 hugepags needed to be allocated on UMA
>> system, job.min_chunk will be 2, job.max_threads will be 2. Then, 2
>> workers will be scheduled, however each worker will just allocate 2 pages,
>> how much the cost of scheduling? What if allocate 4 pages in single
>> worker? Do you have any numbers on parallelism vs non-parallelism in
>> a small allocation case? If we cannot gain from this case, I think we shold
>> assign a reasonable value to ->min_chunk based on experiment.
>> Thanks.
>>
>
> That's a good suggestion, I'll run some tests and choose the best
> values.
>
>
next prev parent reply other threads:[~2024-01-22 11:31 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-01-18 12:39 [RESEND PATCH v4 0/7] hugetlb: parallelize hugetlb page init on boot Gang Li
2024-01-18 12:39 ` [PATCH v4 1/7] hugetlb: code clean for hugetlb_hstate_alloc_pages Gang Li
2024-01-18 12:39 ` [PATCH v4 2/7] hugetlb: split hugetlb_hstate_alloc_pages Gang Li
2024-01-22 3:43 ` Muchun Song
2024-01-18 12:39 ` [PATCH v4 3/7] padata: dispatch works on different nodes Gang Li
2024-01-18 23:04 ` Tim Chen
2024-01-19 15:05 ` Gang Li
2024-01-19 2:59 ` Muchun Song
2024-01-19 15:04 ` Gang Li
2024-01-18 12:39 ` [PATCH v4 4/7] hugetlb: pass *next_nid_to_alloc directly to for_each_node_mask_to_alloc Gang Li
2024-01-18 23:01 ` Tim Chen
2024-01-19 2:54 ` Muchun Song
2024-01-22 6:16 ` Muchun Song
2024-01-22 9:14 ` Gang Li
2024-01-22 9:50 ` Muchun Song
2024-01-18 12:39 ` [PATCH v4 5/7] hugetlb: have CONFIG_HUGETLBFS select CONFIG_PADATA Gang Li
2024-01-18 12:39 ` [PATCH v4 6/7] hugetlb: parallelize 2M hugetlb allocation and initialization Gang Li
2024-01-22 7:10 ` Muchun Song
2024-01-22 10:12 ` Gang Li
2024-01-22 11:30 ` Muchun Song [this message]
2024-01-23 2:12 ` Gang Li
2024-01-23 3:32 ` Muchun Song
2024-01-18 12:39 ` [PATCH v4 7/7] hugetlb: parallelize 1G hugetlb initialization Gang Li
2024-01-18 14:22 ` Kefeng Wang
2024-01-19 14:45 ` Gang Li
2024-01-24 9:23 ` Muchun Song
2024-01-24 10:52 ` Gang Li
2024-01-25 2:48 ` Muchun Song
2024-01-25 3:47 ` Gang Li
2024-01-25 3:56 ` Gang Li
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=849D7EA4-BCF4-4587-8A78-F3B35B63EAE9@linux.dev \
--to=muchun.song@linux.dev \
--cc=akpm@linux-foundation.org \
--cc=david@redhat.com \
--cc=gang.li@linux.dev \
--cc=ligang.bdlg@bytedance.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mike.kravetz@oracle.com \
--cc=rientjes@google.com \
--cc=tim.c.chen@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).