From: Jonathan Nieder <jrnieder@gmail.com>
To: David Barr <davidbarr@google.com>
Cc: Git List <git@vger.kernel.org>,
Thomas Rast <trast@student.ethz.ch>,
Dmitry Ivankov <divanorama@gmail.com>,
Sverre Rabbelier <srabbelier@gmail.com>
Subject: [PATCH 1/4] fast-import: allow object_table to grow dynamically
Date: Wed, 11 Apr 2012 07:13:42 -0500 [thread overview]
Message-ID: <20120411121342.GC19568@burratino> (raw)
In-Reply-To: <20120411121259.GB19568@burratino>
From: David Barr <david.barr@cordelta.com>
Date: Thu, 31 Mar 2011 22:59:58 +1100
The current custom hash tables in fast-import.c do not grow. This
causes poor performance for very large imports.
Shawn O. Pearce writes:
> I can tell you why its fixed size... when I wrote fast-import to
> support the Mozilla repository import, we had an estimate on the
> number of objects that we needed fast-import to handle in a given run.
> From that estimate we concluded that a table of 2^16 was sufficiently
> large enough, as the hash chains would only be some small N long given
> the total number of objects we needed to handle for Mozilla. Doubling
> that into 2^17 or larger wasn't useful, and using a smaller table like
> 2^14 produced too long of a chain.
>
> Once this code was working, we moved on to other features, and never
> reconsidered the table size. This table should be growing if the
> initial 2^16 isn't big enough. Its just that nobody has ever noticed
> that I hardcoded the size. :-)
Fortunately, we have struct hash_table and friends so there's no need
to write cumbersome hash table growth code.
[jn: using native endianness for hash, new commit message; with
init_hash call, even though it's not currently needed]
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
---
fast-import.c | 27 ++++++++++++++++++++-------
1 file changed, 20 insertions(+), 7 deletions(-)
diff --git a/fast-import.c b/fast-import.c
index 78d97868..a79a1260 100644
--- a/fast-import.c
+++ b/fast-import.c
@@ -313,7 +313,7 @@ static off_t pack_size;
/* Table of objects we've written. */
static unsigned int object_entry_alloc = 5000;
static struct object_entry_pool *blocks;
-static struct object_entry *object_table[1 << 16];
+static struct hash_table object_table;
static struct mark_set *marks;
static const char *export_marks_file;
static const char *import_marks_file;
@@ -553,11 +553,18 @@ static struct object_entry *new_object(unsigned char *sha1)
return e;
}
+static unsigned int hash_sha1(const unsigned char *sha1)
+{
+ unsigned int h;
+ memcpy(&h, sha1, sizeof(unsigned int));
+ return h;
+}
+
static struct object_entry *find_object(unsigned char *sha1)
{
- unsigned int h = sha1[0] << 8 | sha1[1];
+ unsigned int h = hash_sha1(sha1);
struct object_entry *e;
- for (e = object_table[h]; e; e = e->next)
+ for (e = lookup_hash(h, &object_table); e; e = e->next)
if (!hashcmp(sha1, e->idx.sha1))
return e;
return NULL;
@@ -565,8 +572,9 @@ static struct object_entry *find_object(unsigned char *sha1)
static struct object_entry *insert_object(unsigned char *sha1)
{
- unsigned int h = sha1[0] << 8 | sha1[1];
- struct object_entry *e = object_table[h];
+ unsigned int h = hash_sha1(sha1);
+ struct object_entry *e = lookup_hash(h, &object_table);
+ void **pos;
while (e) {
if (!hashcmp(sha1, e->idx.sha1))
@@ -575,9 +583,13 @@ static struct object_entry *insert_object(unsigned char *sha1)
}
e = new_object(sha1);
- e->next = object_table[h];
+ e->next = NULL;
e->idx.offset = 0;
- object_table[h] = e;
+ pos = insert_hash(h, e, &object_table);
+ if (pos) {
+ e->next = *pos;
+ *pos = e;
+ }
return e;
}
@@ -3261,6 +3273,7 @@ int main(int argc, const char **argv)
atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
+ init_hash(&object_table);
marks = pool_calloc(1, sizeof(struct mark_set));
global_argc = argc;
--
1.7.10
next prev parent reply other threads:[~2012-04-11 12:13 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-03-31 11:59 fast-import: use struct hash_table David Barr
2011-03-31 11:59 ` [PATCH 1/2] fast-import: use struct hash_table for atom strings David Barr
2011-04-02 2:42 ` Jonathan Nieder
2011-04-02 3:33 ` Jonathan Nieder
2011-03-31 11:59 ` [PATCH 2/2] fast-import: use struct hash_table for objects David Barr
2011-04-02 2:46 ` Jonathan Nieder
2011-04-02 2:48 ` fast-import: use struct hash_table Jonathan Nieder
2012-04-11 12:11 ` [PATCH/RFC v2 0/4] " Jonathan Nieder
2012-04-11 12:12 ` [PATCH/RFC v2 0/4 resend] " Jonathan Nieder
2012-04-11 12:13 ` Jonathan Nieder [this message]
2012-04-11 12:14 ` [PATCH 2/4] fast-import: allow atom_table to grow dynamically Jonathan Nieder
2012-04-11 12:15 ` [PATCH 3/4] fast-import: allow branch_table " Jonathan Nieder
2012-04-11 12:15 ` [PATCH 4/4] fast-import: use DIV_ROUND_UP Jonathan Nieder
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
List information: http://vger.kernel.org/majordomo-info.html
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20120411121342.GC19568@burratino \
--to=jrnieder@gmail.com \
--cc=davidbarr@google.com \
--cc=divanorama@gmail.com \
--cc=git@vger.kernel.org \
--cc=srabbelier@gmail.com \
--cc=trast@student.ethz.ch \
/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.
Code repositories for project(s) associated with this public inbox
https://80x24.org/mirrors/git.git
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 read-only IMAP folder(s) and NNTP newsgroup(s).