From: compudj@krystal.dyndns.org (Mathieu Desnoyers)
Subject: [ltt-dev] [PATCH 05/10 round10] add max_size arguemnt
Date: Wed, 23 Nov 2011 02:15:46 -0500 [thread overview]
Message-ID: <20111123071546.GA1969@Krystal> (raw)
In-Reply-To: <4ECC9D0D.8080300@cn.fujitsu.com>
* Lai Jiangshan (laijs at cn.fujitsu.com) wrote:
> On 11/22/2011 06:11 PM, Mathieu Desnoyers wrote:
> > arguemnt -> argument
> >
> > I think we need to discuss the semantic of max_size: here, you propose
> > that max_size is just the number of buckets, but it does not limit the
> > ability of the table to contain more elements than max_size.
> >
> > IMHO, it would be nice to have max_size limit the number of elements
> > that the table can accept, e.g. ensuring that insertion fails if the
> > number of elements gets over the max_size threshold. It might not need
> > to be an exact thing though, just giving an approximated upper limit
> > might be fine (using the split-counters), as long as it is documented.
>
> documented: "max_size: the max size of buckets for future growing"
>
> Is max_bucket_size better?
> and min_alloc_size need to be changed to min_alloc_bucket_size?
>
> I'm not planing to add "max_size" for the number of elements in this
> round or near future.
maybe max_nr_buckets then ?
nr_ seems to be more descriptive than "size" in this context, because
"size" could lead to some confusion, leading to think that we would be
specifying the size of a bucket rather than actually specifying the
number of buckets. Seems reasonable ?
Thanks,
Mathieu
>
> Thanks,
> Lai
>
> >
> > Thanks,
> >
> > Mathieu
> >
> > * Lai Jiangshan (laijs at cn.fujitsu.com) wrote:
> >> Signed-off-by: Lai Jiangshan <laijs at cn.fujitsu.com>
> >> ---
> >> rculfhash.c | 17 +++++++++++++++++
> >> tests/test_urcu_hash.c | 2 +-
> >> urcu/rculfhash.h | 9 ++++++---
> >> 3 files changed, 24 insertions(+), 4 deletions(-)
> >>
> >> diff --git a/rculfhash.c b/rculfhash.c
> >> index 2df133c..b7be893 100644
> >> --- a/rculfhash.c
> >> +++ b/rculfhash.c
> >> @@ -266,6 +266,7 @@ struct cds_lfht {
> >> struct rcu_table t;
> >> unsigned long min_alloc_order;
> >> unsigned long min_alloc_size;
> >> + unsigned long max_size;
> >> int flags;
> >> /*
> >> * We need to put the work threads offline (QSBR) when taking this
> >> @@ -1351,6 +1352,7 @@ void cds_lfht_create_bucket(struct cds_lfht *ht, unsigned long size)
> >>
> >> struct cds_lfht *_cds_lfht_new(unsigned long init_size,
> >> unsigned long min_alloc_size,
> >> + unsigned long max_size,
> >> int flags,
> >> void (*cds_lfht_call_rcu)(struct rcu_head *head,
> >> void (*func)(struct rcu_head *head)),
> >> @@ -1369,11 +1371,22 @@ struct cds_lfht *_cds_lfht_new(unsigned long init_size,
> >> /* min_alloc_size must be power of two */
> >> if (!min_alloc_size || (min_alloc_size & (min_alloc_size - 1)))
> >> return NULL;
> >> +
> >> /* init_size must be power of two */
> >> if (!init_size || (init_size & (init_size - 1)))
> >> return NULL;
> >> +
> >> + if (!max_size)
> >> + max_size = 1UL << (MAX_TABLE_ORDER - 1);
> >> +
> >> + /* max_size must be power of two */
> >> + if (!max_size || (max_size & (max_size - 1)))
> >> + return NULL;
> >> +
> >> min_alloc_size = max(min_alloc_size, MIN_TABLE_SIZE);
> >> init_size = max(init_size, MIN_TABLE_SIZE);
> >> + max_size = max(max_size, min_alloc_size);
> >> + init_size = min(init_size, max_size);
> >> ht = calloc(1, sizeof(struct cds_lfht));
> >> assert(ht);
> >> ht->flags = flags;
> >> @@ -1393,6 +1406,7 @@ struct cds_lfht *_cds_lfht_new(unsigned long init_size,
> >> ht->t.resize_target = 1UL << order;
> >> ht->min_alloc_size = min_alloc_size;
> >> ht->min_alloc_order = get_count_order_ulong(min_alloc_size);
> >> + ht->max_size = max_size;
> >> cds_lfht_create_bucket(ht, 1UL << order);
> >> ht->t.size = 1UL << order;
> >> return ht;
> >> @@ -1752,6 +1766,7 @@ void resize_target_update_count(struct cds_lfht *ht,
> >> unsigned long count)
> >> {
> >> count = max(count, MIN_TABLE_SIZE);
> >> + count = min(count, ht->max_size);
> >> uatomic_set(&ht->t.resize_target, count);
> >> }
> >>
> >> @@ -1809,6 +1824,7 @@ void cds_lfht_resize_lazy_grow(struct cds_lfht *ht, unsigned long size, int grow
> >> {
> >> unsigned long target_size = size << growth;
> >>
> >> + target_size = min(target_size, ht->max_size);
> >> if (resize_target_grow(ht, target_size) >= target_size)
> >> return;
> >>
> >> @@ -1827,6 +1843,7 @@ void cds_lfht_resize_lazy_count(struct cds_lfht *ht, unsigned long size,
> >> if (!(ht->flags & CDS_LFHT_AUTO_RESIZE))
> >> return;
> >> count = max(count, MIN_TABLE_SIZE);
> >> + count = min(count, ht->max_size);
> >> if (count == size)
> >> return; /* Already the right size, no resize needed */
> >> if (count > size) { /* lazy grow */
> >> diff --git a/tests/test_urcu_hash.c b/tests/test_urcu_hash.c
> >> index b9e3e81..2961177 100644
> >> --- a/tests/test_urcu_hash.c
> >> +++ b/tests/test_urcu_hash.c
> >> @@ -967,7 +967,7 @@ int main(int argc, char **argv)
> >> * thread from the point of view of resize.
> >> */
> >> rcu_register_thread();
> >> - test_ht = cds_lfht_new(init_hash_size, min_hash_alloc_size,
> >> + test_ht = cds_lfht_new(init_hash_size, min_hash_alloc_size, (1UL << 18),
> >> (opt_auto_resize ? CDS_LFHT_AUTO_RESIZE : 0) |
> >> CDS_LFHT_ACCOUNTING, NULL);
> >> ret = populate_hash();
> >> diff --git a/urcu/rculfhash.h b/urcu/rculfhash.h
> >> index c13d3df..2eefd2c 100644
> >> --- a/urcu/rculfhash.h
> >> +++ b/urcu/rculfhash.h
> >> @@ -98,6 +98,7 @@ enum {
> >> */
> >> struct cds_lfht *_cds_lfht_new(unsigned long init_size,
> >> unsigned long min_alloc_size,
> >> + unsigned long max_size,
> >> int flags,
> >> void (*cds_lfht_call_rcu)(struct rcu_head *head,
> >> void (*func)(struct rcu_head *head)),
> >> @@ -112,8 +113,9 @@ struct cds_lfht *_cds_lfht_new(unsigned long init_size,
> >>
> >> /*
> >> * cds_lfht_new - allocate a hash table.
> >> - * @init_size: number of nodes to allocate initially. Must be power of two.
> >> - * @min_alloc_size: the smallest allocation size to use. Must be power of two.
> >> + * @init_size: number of buckets to allocate initially. Must be power of two.
> >> + * @min_alloc_size: the smallest allocation buckets size to use. Must be power of two.
> >> + * @max_size: the max size of buckets for future growing. Must be power of two.
> >> * @flags: hash table creation flags (can be combined with bitwise or: '|').
> >> * 0: no flags.
> >> * CDS_LFHT_AUTO_RESIZE: automatically resize hash table.
> >> @@ -134,10 +136,11 @@ struct cds_lfht *_cds_lfht_new(unsigned long init_size,
> >> static inline
> >> struct cds_lfht *cds_lfht_new(unsigned long init_size,
> >> unsigned long min_alloc_size,
> >> + unsigned long max_size,
> >> int flags,
> >> pthread_attr_t *attr)
> >> {
> >> - return _cds_lfht_new(init_size, min_alloc_size, flags,
> >> + return _cds_lfht_new(init_size, min_alloc_size, max_size, flags,
> >> call_rcu, synchronize_rcu, rcu_read_lock,
> >> rcu_read_unlock, rcu_thread_offline,
> >> rcu_thread_online, rcu_register_thread,
> >> --
> >> 1.7.4.4
> >>
> >
>
>
> _______________________________________________
> ltt-dev mailing list
> ltt-dev at lists.casi.polymtl.ca
> http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
>
--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
next prev parent reply other threads:[~2011-11-23 7:15 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-11-16 6:48 [ltt-dev] [PATCH 00/10 round10] rculfhash memory managements Lai Jiangshan
2011-11-16 6:48 ` [ltt-dev] [PATCH 01/10 round10] introduce bucket_at() and improve readability Lai Jiangshan
2011-11-22 9:58 ` Mathieu Desnoyers
2011-11-23 6:33 ` Lai Jiangshan
2011-11-23 6:50 ` Mathieu Desnoyers
2011-11-16 6:48 ` [ltt-dev] [PATCH 02/10 round10] proper wrapper for bucket table alloc and free Lai Jiangshan
2011-11-22 10:01 ` Mathieu Desnoyers
2011-11-23 6:36 ` Lai Jiangshan
2011-11-23 6:49 ` Mathieu Desnoyers
2011-11-16 6:48 ` [ltt-dev] [PATCH 03/10 round10] ht->t.size is no required larger than ht->min_table_size now Lai Jiangshan
2011-11-22 10:06 ` Mathieu Desnoyers
2011-11-23 7:05 ` Lai Jiangshan
2011-11-23 7:17 ` Mathieu Desnoyers
2011-11-16 6:48 ` [ltt-dev] [PATCH 04/10 round10] remove struct rcu_level Lai Jiangshan
2011-11-22 10:08 ` Mathieu Desnoyers
2011-11-16 6:48 ` [ltt-dev] [PATCH 05/10 round10] add max_size arguemnt Lai Jiangshan
2011-11-22 10:11 ` Mathieu Desnoyers
2011-11-23 7:13 ` Lai Jiangshan
2011-11-23 7:15 ` Mathieu Desnoyers [this message]
2011-11-16 6:48 ` [ltt-dev] [PATCH 06/10 round10] add LFHT_MEMORY_CHUNK low level memory management Lai Jiangshan
2011-11-22 10:20 ` Mathieu Desnoyers
2011-11-23 6:05 ` Mathieu Desnoyers
2011-11-23 8:11 ` Lai Jiangshan
2011-11-23 9:21 ` Mathieu Desnoyers
2011-11-16 6:48 ` [ltt-dev] [PATCH 07/10 round10] add LFHT_MEMORY_RESERVED " Lai Jiangshan
2011-11-17 2:15 ` Lai Jiangshan
2011-11-22 10:25 ` Mathieu Desnoyers
2011-11-16 6:48 ` [ltt-dev] [PATCH 08/10 round10] comments for " Lai Jiangshan
2011-11-16 6:48 ` [ltt-dev] [PATCH 09/10 round10] fix comments of cds_lfht_new() Lai Jiangshan
2011-11-23 5:46 ` Mathieu Desnoyers
2011-11-16 6:48 ` [ltt-dev] [PATCH 10/10 round10] add my name Lai Jiangshan
2011-11-23 5:48 ` Mathieu Desnoyers
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=20111123071546.GA1969@Krystal \
--to=compudj@krystal.dyndns.org \
/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