git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: David Kastrup <dak@gnu.org>
To: git@vger.kernel.org
Subject: [PATCH] diff-delta.c: pack the index structure
Date: Sat, 08 Sep 2007 11:31:31 +0200	[thread overview]
Message-ID: <0cd39105dcd57a60eca290db598613aafcc8c577.1189243702.git.dak@gnu.org> (raw)

In normal use cases, the performance wins are not overly impressive:
we get something like 5-10% due to the slightly better locality of
memory accesses using the packed structure.

However, since the data structure for index entries saves 33% of
memory on 32-bit platforms and 40% on 64-bit platforms, the behavior
when memory gets limited should be nicer.

This is a rather well-contained change.  One obvious improvement would
be sorting the elements in one bucket according to their hash, then
using binary probing to find the elements with the right hash value.

As it stands, the output should be strictly the same as previously
unless one uses the option for limiting the amount of used memory, in
which case the created packs might be better.

Signed-off-by: David Kastrup <dak@gnu.org>
---
 diff-delta.c |   74 +++++++++++++++++++++++++++++++++++++++++++++------------
 1 files changed, 58 insertions(+), 16 deletions(-)

diff --git a/diff-delta.c b/diff-delta.c
index 0dde2f2..1b4b1c1 100644
--- a/diff-delta.c
+++ b/diff-delta.c
@@ -115,9 +115,13 @@ static const unsigned int U[256] = {
 struct index_entry {
 	const unsigned char *ptr;
 	unsigned int val;
-	struct index_entry *next;
 };
 
+struct unpacked_index_entry {
+	struct index_entry entry;
+	struct unpacked_index_entry *next;
+};	
+
 struct delta_index {
 	unsigned long memsize;
 	const void *src_buf;
@@ -131,7 +135,8 @@ struct delta_index * create_delta_index(const void *buf, unsigned long bufsize)
 	unsigned int i, hsize, hmask, entries, prev_val, *hash_count;
 	const unsigned char *data, *buffer = buf;
 	struct delta_index *index;
-	struct index_entry *entry, **hash;
+	struct unpacked_index_entry *entry, **hash;
+	struct index_entry *aentry, **ahash;
 	void *mem;
 	unsigned long memsize;
 
@@ -148,28 +153,21 @@ struct delta_index * create_delta_index(const void *buf, unsigned long bufsize)
 	hmask = hsize - 1;
 
 	/* allocate lookup index */
-	memsize = sizeof(*index) +
-		  sizeof(*hash) * hsize +
+	memsize = sizeof(*hash) * hsize +
 		  sizeof(*entry) * entries;
 	mem = malloc(memsize);
 	if (!mem)
 		return NULL;
-	index = mem;
-	mem = index + 1;
 	hash = mem;
 	mem = hash + hsize;
 	entry = mem;
 
-	index->memsize = memsize;
-	index->src_buf = buf;
-	index->src_size = bufsize;
-	index->hash_mask = hmask;
 	memset(hash, 0, hsize * sizeof(*hash));
 
 	/* allocate an array to count hash entries */
 	hash_count = calloc(hsize, sizeof(*hash_count));
 	if (!hash_count) {
-		free(index);
+		free(hash);
 		return NULL;
 	}
 
@@ -183,12 +181,13 @@ struct delta_index * create_delta_index(const void *buf, unsigned long bufsize)
 			val = ((val << 8) | data[i]) ^ T[val >> RABIN_SHIFT];
 		if (val == prev_val) {
 			/* keep the lowest of consecutive identical blocks */
-			entry[-1].ptr = data + RABIN_WINDOW;
+			entry[-1].entry.ptr = data + RABIN_WINDOW;
+			--entries;
 		} else {
 			prev_val = val;
 			i = val & hmask;
-			entry->ptr = data + RABIN_WINDOW;
-			entry->val = val;
+			entry->entry.ptr = data + RABIN_WINDOW;
+			entry->entry.val = val;
 			entry->next = hash[i];
 			hash[i] = entry++;
 			hash_count[i]++;
@@ -212,16 +211,59 @@ struct delta_index * create_delta_index(const void *buf, unsigned long bufsize)
 			continue;
 		entry = hash[i];
 		do {
-			struct index_entry *keep = entry;
+			struct unpacked_index_entry *keep = entry;
 			int skip = hash_count[i] / HASH_LIMIT;
 			do {
+				--entries;
 				entry = entry->next;
 			} while(--skip && entry);
+			++entries;
 			keep->next = entry;
 		} while(entry);
 	}
 	free(hash_count);
 
+	/* Now create the packed index in array form rather than
+	 * linked lists */
+
+	memsize = sizeof(*index)
+		+ sizeof(*ahash) * (hsize+1)
+		+ sizeof(*aentry) * entries;
+
+	mem = malloc(memsize);
+
+	if (!mem) {
+		free(hash);
+		return NULL;
+	}
+
+	index = mem;
+	index->memsize = memsize;
+	index->src_buf = buf;
+	index->src_size = bufsize;
+	index->hash_mask = hmask;
+
+	mem = index + 1;
+	ahash = mem;
+	mem = ahash + (hsize+1);
+	aentry = mem;
+
+	/* Coalesce all entries belonging to one linked list into
+	 * consecutive array entries */
+
+	for (i = 0; i < hsize; i++) {
+		ahash[i] = aentry;
+		for (entry = hash[i]; entry; entry = entry->next)
+			*aentry++ = entry->entry;
+	}
+
+	/* Sentinel value to indicate the length of the last hash
+	 * bucket */
+
+	ahash[hsize] = aentry;
+	assert(aentry - (struct index_entry *)mem == entries);
+	free(hash);
+
 	return index;
 }
 
@@ -302,7 +344,7 @@ create_delta(const struct delta_index *index,
 			val ^= U[data[-RABIN_WINDOW]];
 			val = ((val << 8) | *data) ^ T[val >> RABIN_SHIFT];
 			i = val & index->hash_mask;
-			for (entry = index->hash[i]; entry; entry = entry->next) {
+			for (entry = index->hash[i]; entry < index->hash[i+1]; entry++) {
 				const unsigned char *ref = entry->ptr;
 				const unsigned char *src = data;
 				unsigned int ref_size = ref_top - ref;
-- 
1.5.3.GIT

             reply	other threads:[~2007-09-08  9:31 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-09-08  9:31 David Kastrup [this message]
2007-09-08  8:54 ` [PATCH] diff-delta.c: rename {a,}{entry,hash} to {,u}{entry,hash} David Kastrup
2007-09-08 19:41   ` [PATCH] diff-delta.c: Rationalize culling of hash buckets David Kastrup
2007-09-08 20:52   ` [PATCH] diff-delta.c: rename {a,}{entry,hash} to {,u}{entry,hash} Nicolas Pitre
2007-09-08 21:06     ` David Kastrup
2007-09-09  0:38       ` Nicolas Pitre
  -- strict thread matches above, loose matches on Subject: below --
2007-09-08 21:17 [PATCH] diff-delta.c: pack the index structure David Kastrup
2007-09-07 23:38 David Kastrup
2007-09-08  2:34 ` Nicolas Pitre
2007-09-08  6:48   ` David Kastrup
2007-09-08  8:36   ` David Kastrup
2007-09-08 20:50     ` David Kastrup

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=0cd39105dcd57a60eca290db598613aafcc8c577.1189243702.git.dak@gnu.org \
    --to=dak@gnu.org \
    --cc=git@vger.kernel.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.
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).